AU2006222742A1 - Computational linguistic analysis - Google Patents

Description

AUSTRALIA

00

(N

Patents Act 1990 COMPLETE SPECIFICATION FOR A STANDARD PATENT

ORIGINAL

Name of Applicant: Actual Inventor: Address for Service: Invention Title: Details of Associated Provisional Application(s): Marie Louise Fellbaum KORPI Marie Louise Fellbaum KORPI FRASER OLD SOHN Patent Attorneys Level 6, 118 Alfred Street MILSONS POINT NSW 2061 Computational Linguistic Analysis Australian Patent Application No. 2005 905 346 Filed 28 September 2005 The following statement is a full description of this invention, including the best method of performing it known to us: 5161A-AU IDThe present invention relates to computational linguistic analysis. A method of linguistic analysis is disclosed which is intrinsically valuable in its own right as an intellectual advance and which finds commercial utility in the design and conduct of Slanguage courses and in computer programs dealing with speech and word processing.

00 It is known in voice recognition software, for example, to provide various rules which are inherent in the grammar of the language in question to assist the computer in its recognition of the spoken word. However, such rules are restricted to C a single sentence and the analysis of one sentence cannot assist or in any way IND contribute to the analysis of an adjacent sentence.

The genesis of the present invention is a device to advance the state of the computational linguistic analysis art.

In accordance with a first aspect of the present invention there is disclosed a method of computational linguistic analysis, said method comprising the steps of: selecting a passage of text to be analysed, said text containing at least one sentence; (ii) for each word allocating the word to one of four categories comprising: community membership visual linguistic co-presence auditory linguistic co-presence, and indirect co-presence; (iii) for each allocated word allocating a further sub-category comprising: "plus" if the word is already presented in the text or can be inferred, and "minus" if the word is not already presented in the text and is not able to be inferred; and (iv) generating a relationship expressed in the relations of members in sets between all the allocated words.

An embodiment of the present invention will now be described with reference to the following Annexures.

5161A-AU IDAnnexure 1 is a synopsis of a doctorial thesis presented by the inventor, and Annexure 2 is a copy of that thesis.

The inventor has English as a first language and also has a working knowledge 00 of the following languages: Latin, German, French, Spanish, Korean and Japanese.

The invention arises out of a study of interlanguage expression, for example persons N whose first language is Japanese speaking in a second language, namely English.

However, the invention is not limited to interlanguage expression and is thought to be N applicable to all expression and to most, if not all, languages.

IND

Furthermore, the use of sets has been influenced by the hypothesis that the sets represent an actual allocation of resources within the human brain. That is the human brain is thought to have a separate region or centre which relates to nouns, another region or centre which relates to adjectives, a third which relates to indefinites (such as "some", "many", for example) and so on.

In addition, the analysis and generation of relations between members of sets is reduced to three basic or core structures. These relations are not confined to a single sentence but are able to be expanded from one sentence, to the adjacent sentence, and ultimately to the entire text or discourse the subject of the analysis. In addition, the core semantic relations between a speaker and a hearer, or even a reader and a text, may not be the same.

An important consequence of this analysis is that there is a large body of mathematics relating to sets which is now able to be applied to linguistic analysis.

This is thought to open a new vista in computational linguistics and the writing of computer software in fields such as speech or voice recognition, machine translation from one language to another, and the like. The construction of the semantic relations characterizing an entire discourse are thought to differ between languages.

Example 1 The two sentences "a cat has four legs" and "the cat sat on a mat" are spoken by a speaker to a hearer under two conditions. In the first condition (corresponding to Fig. 1) there is no picture. This is the first row of Table 1. In the second condition 5161A-AU

IO

00

N

oo t(N (corresponding to Fig. 2) the speaker is looking at a picture but the hearer is not able to see the picture visible to the speaker. The second row relates to the speaker in the second condition and the third row relates to the hearer in the second condition. Each word of the two sentences is then able to be characterised as follows where C indicates an auditory linguistic co-presence, BI indicates a visible visual linguistic copresence and B2 indicates a visualized visual linguistic co-presence. D indicates an indirect co-presence. The symbols or plus and minus have the meanings referred to above.

Table 1 A CAT HAS FOUR LEGS THE CAT SAT ON THE MAT -C -C -C -C -C +C +C -C -C -C -C -C -C -C -C -C +C +C -C -C -C -C +B1 +B1 +B 1 +B1 +B1 +B1 +B1 +B1 +B1 +B 1 +B1 -C -C -C -C -C +C +C -C -C -C -C +B2 +B2 +B2 +B2 +B2 +B2 +B2 +B2 +B2 +B2 +B2 Table 2 This table illustrates the inferences drawn by the hearer who does not have the benefit of the picture but knows that the speaker has a picture.

THE CAT SAT ON THE MAT +C +C -C -C -C -C +C +C -C -C -C -C +B1 +B1 +B1 +B1 +B1 +B1 +C +C -C -C -C -C +B2 +B2 +B2 +B2 +B2 +B2 +D +D +D +D +D +D "Cat" is a set with two sub-sets of the properties, one being the number four and the other set being "legs". This set then expands to include a set of the action of "sat" and "mat".

The first at time TI can be represented as TI legs four, cat 5161A-AU \0 T2 cat sat, mat In addition the members of the set have a semantic relation of existential, identificational or characterizational as outlined in Appendix 1. Rather than using 00 variables, actual lexemes are used to construct the sets and the relations between existing objects and newly added objects, as indicated in the drawings. The coding of Sdefiniteness properties of the objects is outlined in Tables 1 and 2. This coding of each lexeme constructs the context for each newly added sentence. The core semantic C relations function as a template for referring expressions in the developing discourse.

5161A-AU

NO

0 ANNEXURE1 Semantic Definiteness Structures and Referentiality Introduction common way to approach understanding a complex phenomenon is to decompose it into simpler amponents and more manageable subdomains. After the subdomains are more clearly articulated, the iteractions between and among separate subdomains can be studied. This proposal presents a lethodology for the study of the complex system of semantic meaning of definiteness and :ferentiality in natural and interlanguage, the language spoken by second language learners of any inguage.

First, the concept of definiteness is decomposed into the properties of its most minimal parts, icluding visual, visible, auditory and inferential definiteness, proving that these are independent of ategorical terms such as noun, adjective, verb, and adverb, organized according to four principal ases for establishing mutual knowledge and definiteness: Fellbaum (1999) Basis for Mutual Knowledge 1. Community membership 2. Visible co-presence 3. Acoustic co-presence 4. Indirect Co-presence Visible Acoustic Definiteness Types -Universal, Particular (Proper nouns) -Deixis: Visible, visuo-spatial; -Deixis:Spatio-, Temporal; Anaphora -Inferencing Properties evoked from Visible or visual objects Properties evoked from Acoustic objects Mental schemata; specificity For the purpose of this analysis four tasks were used as the instruments to elicit data for the study of definiteness. These tasks, the four types of definiteness and the psycholinguistic correlates necessary for establishing the type of definiteness are summarized in the chart below.

2006222742 28 Sep 2006 Table 3.2 Four Types of Mutual Knowledge, Cognitive Resources, Linguistic Expression, and Associated Tasks for Assigning Definiteness Basis for Mutual Knowledge Cognitive Resources for Assigning Definiteness Linguistic Expression Tasks I, LI, III, IV 1. Community membership 2. Visual linguistic co-presence a. Immediate b. Potential c. Prior 3. Auditory linguistic co-presenci a. Potential b. Prior 4. Indirect co-presence a. Visual b. Auditory c. Mixture

L

S

S

ommunity comembership, Jniversality of knowledge Simultaneity, attention, rationality imultaneity, attention, rationality, locatability ;imultaneity, attention, rationality, recallability Generics, Proper Names Generics Deixis Il, IV person, spatial (demonstratives) locative, temporal e Anaphora I, II Simultaneity, attention, rationality, locatability, understandability pronouns, definite descriptions Simultaneity, attention, rationality, recallability, understandability demonstratives, proper names Demonstratives Simultaneity, attention, rationality, (locatability or recallability), person, spatial, temporal 11 associativity community membership Simultaneity, attention, rationality, (locatability or recallability) person, spatial, temporal I, II understandability, associativity anaphora, community membership I, IV DEFINITIONS of COGNITIVE RESOURCES for ASSIGNING DEFINITENESS in a CONTEXT I. Community comembership particularly prominent people, places or events which members of a community can safely assume are known by all 2. Universality of knowledge generic and particular knowledge that both [all] interlocutors can safely assume each other knows 3. Simultaneity interlocutor[s:both, all] looking at[/listening to] each other and target in situation simultaneously 4. Attention both [all] interlocutor[s] attending to [each other and] target copresent in the situation Rationality both [all] interlocutors capable of inferring or drawing the same conclusions from the available mutual knowledge, including meaning of visual and auditory signal 6. Locatability-hearer is capable of finding target [incl. abstract auditory in semantic context] and bringing into view or focus [consciousness] simultaneously with speaker 7. Recallability hearer is capable of recalling earlier interaction and targets established between both interlocutors, either from earlier visual focusing or previous mention 8. Understandability hearer is capable of"penetrating indefinite reference" and understanding that its existence is being posited, i.e. its definiteness or specificity 9. Associativity known by the individual[s] in the community that the proposition, entity, or property is certainly, probably, or possibly a particular part of, or in a particular role with the proposition or entity; that the set-subset relationship is an appropriate relationship Task Description: I Introductions; II Description of Hometown Festival; III Information Gap; IV Picture Description of Fest

\D

O Secondly, these are developed into a system of core semantic structures 04 composed of a set of three distinctive definiteness relations, Existential, SCharacterizational, and Identificational.

Semantic Definiteness Structures 00 1. Existential Type where the semantic relation of the proposition is to introduce an entity or express a proposition which is linked neither to an existing discourse entity, or proposition, i.e. a discourse referent. Therefore, the semantic structure contains at least one entity, visible or acoustic, which is newly introduced and nonspecific.

CK 2. Identificational Type where the semantic relation in the proposition relates C, two entities, visible, visual, acoustic, or inferred to an already existing referent.

SThis creates an identity relation between both referents in the proposition, with either San auditory referent in the discourse or visible antecedent external to the proposition.

Thus, the identificational relation is between a definite specific entity and another Cl specific, entity in the proposition and discourse.

3. Characterizational Type where the semantic relation further describes, or denotes a property or characteristic, visible or acoustic, of an already established discourse referent.

The definiteness objects in the semantic relations are coded plus or minus and assigned to a respective set.

An example of the first semantic structure, an Existential, is shown in the following discourse fragment, a small set of sentences from a larger discourse. Two Existential-there semantic propositions are produced at the beginning of s3's discourse fragment i2s3:4 during Week 1, where s the learner and i the interviewer in the discourse, and T the time of utterance.: Example 1 s3: There are... lots of people on the street' (T1) i2: uh huh s3: There are many lights (T2) i2: many light? s3: many alot of lights... on the street. (T2') SProposition Time 1-2' OnSt TI Peo le T 2-2' Lights In this diagram we see a street and a set ('lots of) of people at Time 1 (T1).

We can see that on street and people have a set-subset relation at TI; things are on an entity, the street. The [THINGS] are people. Usually the spatial relation on tells us that a smaller entity (people) is on (top of) another larger entity, (streets). At T2-2' a set of lights is added to the larger set on street. Now we have two smaller sets which IsN are on one larger set, street. In this representation of sets, both people and lights are in O an inclusion relation to the street; they are on the street. This labels the sets, and Sspecifies the inclusion relation.

0^ Jackendoffs theory of semantic structures (1983, 1990) composed of semantic Sroles for arguments and verbs are the basis for this analysis. However where Jackendoff uses [THING], I have replaced [THING] with an object coded either 00 Def or Spec. My approach offers a principled and systematic representation of specificity directly into the Lexical Conceptual Structure (LCS); the arguments of the Existential-there proposition are non-specifics. In my system of conceptual C definiteness, the Existential Be is thus not a stipulated but a derived proposition.

Jackendoff's system allows us to view the same entity from different c-i perspectives and with spatial parameters independently of the entity; the spatial parameters are connected to entities or [THINGS] by prepositions. In Jackendoff's \4 system of LCSs, people on the street is "a two-place relation between a [THING] and 0a [PLACE] which are 'mediated' by the functional category of the verb BE. The preposition "on" is "a one-place function mapping a reference object, a [THING], into a [PLACE]" (1983:69). The overall function is a state (ontological category for Jackendoff).

Jackendoffs conceptual formalization for showing the binding of the variable 'thing', or lights, to the Existential BE (cf. Jackendoff, 1983:180) describes a state: Thing State rX [state F(i, Y, This 'well-formedness rule' says that in the State function F, with two arguments, i and Yj, the Thing, Xi is a variable bound to the function F. This binding relationship prohibits an 'action' and any semantic relation other than 'theme'. In this proposition, a theme bears the semantic relation of a "participant..., being in a state or position", as stated in the second conjunct in Definition 1: Theme, provided above.

(The dotted lines are an early attempt by Jackendoff to express recursion.) Jackendoffs functional (lexical) decomposition for this proposition is: [STATE] [state BE ([THING THINGXi, [PLACE The function is STATE with the verb BE. The [THING] is the argument and is bound to the verb BE and is in the position of PLACE. This transliterates from Jackendoffs LCS to the data in Example 1 above as [STATE] [state BE ([lights], [street])] where the Function is Stative BE and THING X, lights and PLACE y= street. We know from the definition of 'state' and the semantic role of 'theme', that lights (Thing Xi) must be a 'theme,' since state cannot have an agent or actor. Only actions can have agents or actors.

We can specify directly what kind of feature of specificity is mediated by the existential state BE, by adding these features to the semantic description of the existential-there proposition and its 'state' relation of the [THING] and [PLACE]: S-spec +spec [State BEE lightsx, streetY e An example of the second semantic relation, Identificational, and its set D membership, is from Task II, a spoken description of a festival.

00 Auditory Processing -Task II: Description of Hometown Festival Example2) Discourse i2s6:2 i2: Uhm..is there any Jap..uh your favorite festival? Ci s6: Uh huh...oh...my favorite festival is Omagari hama(xx) C Set of Festivals My favorite festival CI O D Omagari Festival The learner's favorite festival is chosen from a set of conceptual festivals. We can use Jackendoffs LCSs for the Identificational Proposition Type to include two [THINGS], yet maintain the state relation with the Identificational BE. In addition, the two [THINGS] must refer to the same object, X=Y.

[State BE, ([THING [THING This particular Identificational Proposition also has a plurality like the Existential Proposition above. Notice that a plural noun phrase does not exist in the discourse, yet we automatically presuppose the existence of a plurality from which a 'favorite' festival is chosen; the adjective favorite implies a relation of one from a plurality.

Conceptually, we can represent this type of Identificational from a collection of objects, with many individual members, from which one is selected as: E -Aspec +Aspec +Aspec [statcBEE your favorite festivali)] ->[stateBEI 2 my favorite festivall) X. OmagariFestival Jackendoff's [THING] for the referents of the Identificational are directly instantiated with the lexemes from the discourse fragment, as well as with their respective definiteness properties. The two references to the antecedent, your favorite festival, are labelled +Aspec for the 'auditorially specific' answer by learner s6 to the question by interviewer i2. The lambda extraction symbol is meant to show that all three NPs are bound together for the LCS of the Identificational Proposition. Notice that the first NP is in an identity relation to the last two NPs.

This binding relationship prohibits an 'action' and any semantic relation other than 'theme,' where a theme bears the semantic relation of a "participant..., being in a state or position," as stated in Definition 1 for Theme and Diagnostic Test 1 for State given in §5.1.1 and 5.1.2 of the submitted thesis, respectively.

A final example of the Identificational Proposition Type is found as the last utterance in the description of Task IV, as the interviewer guesses the identity (name) of the festival. The referent of the first entity is the set of a collection of distinguishing characteristics which help the interviewer visually identify the festival.

N0 This requires visuo-acoustic processing based on a visual mental imagecreated by the O learner's auditory description. A discourse fragment from Task IV is repeated below: Visuo-Auditory Processing SExample 5.6) Discourse i2-s3:4 00 i2: I have some picture of festival Sso you can pick up one and don't show me s3: uh huh...

i2: And describe.it and I will try to guess which festival...

(PICTURE DESCRIPTION by s3...

(N

N s3 That's all. Can you guess? i2 Hhm. I think it is Kanto festival

(N

A diagram of the sets of this identificational relation is given below: Auditory Linguistic Objects Visual (Mental) Image Festival Name people, lights, people, ligh cars, taeko dru cars, taeko drums {<Kanto Festival>} street, stick street, stick MANY MANY One Unique Festival The learner's description is presented as a set of auditory linguistic objects which create a collection of visual images as members of a set for the interviewer.

The visual image, mentally representing a festival in the Tohoku area of Japan for the interviewer is then produced as an auditory object in the form of an identificational proposition in the discourse. This Identificational relation is presented in the final proposition I think it is Kanto.

The proposition I think it is Kanto, I think X and X= it is Kanto can be interpreted in several ways. First, the mental image in the discourse is created from the learner's description, (auditory objects), as an Identificational referring back to which festival, in the introduction to the task. Or secondly, which festival is referring back to one from the set of pictures of festivals. The one then is a unit set of one festival from the folder of 'some' pictures offestivals. The it in describe it, from the introduction to the task, becomes the referent for the it in the final proposition. In this interpretation, the entire Identificational proposition in the final line of the discourse are linked auditorally to the picture, one, and it, and which festival.

A third interpetation is that the auditory objects ranging from TI until T9 result in a final image, and the interviewer produces a proposition with It created mentally as an image of the contents of the picture producing the referents it and Kanto. In this interpretation, It refers simultaneously to all of the above referents as well as the mental image of the picture in the interviewer's mind, and Kanto refers exclusively to the picture.

A final interpretation, closer to Jackendoff's, is that It and Kanto are nouns, resulting in an equational proposition It is Kanto, syntactically. If we only look at the \0 syntactic characteristics of the relations of the nouns in these three examples of 0 Identificational Propositions, they all match one of Jackendoffs examples of an Sidentificational relation: 'Clark Kent is Superman' (Jackendoff, 1983:95).

Q Jackendoff just asserts that this is an identificational, following the traditional Ssyntactician's notion of an equational sentence, or identificational.

SHowever, if this proposition is in a list of characteristics in a discourse 00 fragment about either Clark Kent or Superman, then it is not an identificational. For example, we might be describing Clark Kent by listing that he is an alien, a reporter at the Smallville newspaper, in love with Lois Lane and is Superman in disguise. Then, CN the proposition is an example of 'exhaustive listing' of properties or characteristics and functions as a Characterizational Proposition. The noun Superman is simply a c property among several others; it does not identify him.

r, The Characterizational Semantic relation differs from the Existential- \thererelation, which asserts the existence of an indefinite nonspecific entity, and from Sthe Identificational which links one specific entity it to another specific entity in the discourse network. How can we use Jackendoffs theory of LCSs to distinguish this proposition from the preceding Proposition Types composed of states and themes? States in Jackendoffs diagnostic test can be distinguished from event propositions because they both assign characteristics or properties to entities in existing sets. In Diagnostic Test 1, mentioned above, in his example e, '*the fire truck was red', is a Characterizational Proposition.

Although Jackendoff's LCSs account for this type, he classifies it differently from my semantic core structures, in the Existential and Identificational Types. To distinguish the three proposition types, we can modify his LCSs. Jackendoffs first example of a state can be graphically displayed as below: Truck yell red firetrucks firetruck red THINGS PROPERTIES (Colour Characteristics) This diagram of the set relations for a 'red firetruck' shows several sets.

Semantically, these constructions are composed of individuals from existing sets of THINGS, a subset of THINGS a set of PROPERTIES, a subset of THINGS with properties. Thus, red firetrucks are Characterizational Propositions formed of an existing set, or definite entity 'The firetruck', with a set composed of indefinite properties, the color 'red'. Thus, we can classify the semantic relation in this propositions as a Characterizational Semantic Type.

The set of properties is not in an inclusion relationship to the firetruck. The set with 'a firetruck' and the property red is linked; the properties are characteristics of the firetruck. A learner can learn the word firetruck independently of the words to describe it, especially if a picture is shown to the learner. Likewise, the characteristics of a firetruck are numerous and more properties can be attached to its description, such as 'truck with ladders, 'shape', 'motor vehicle, and colors other than 'red', such as yellow. So one set describes the set of properties linked to the entity of a firetruck.

We can use his system, but we must modify it to distinguish it as an ontological type, distinct from the other two. In this example we have a [THING] and a \0 [PROPERTY] in a two-place relation mediated by the verb 'be' (Jackendoff, 1985: 0 69).

N We need to distinguish this Characterizational state from the Existential BE Sand Identificational BE, including the semantic structure in the description, S+def kr-idef 00 [tate BEc firetruck red where thing x firetruck and property y red. We know from Jackendoff s working C definition of 'state' and the semantic role of'theme', that this must be a 'theme,' since state cannot have an agent or actor. Only actions can have agents or actors 180).

We also need to show, or spell out, the types of definiteness in the structures, (C visual, spatial, temporal, or auditory. We can add these features of definiteness/specificity to help complete the description of the Characterizational 0Structure and its relation of the [THING] and [THING]: r+AVisiSpec r-AVisiSpec [STATE BEc [pcoplex, JL drum.Y 4 s3 there are a lot of lantern in the street (1) 4 i3 mm, mm 4 s3 and there are a lot of people (2) 4 i3 yes 4 s3 on the street 4 i3: there are any people? 4 s3: Yes... (looking at picture) 4 s3 some people has got light... (6) In the proposition, Some people has got light, the learner tells us that a portion of the group of people have a light. This can be illustrated by the diagram below: Proportional Inclusion: Group Characteristics People T6 p+i jli (1) Some people has got light In this diagram, two sets of people are shown. The smaller set of people is included in the first larger set. 'People with a light' are a subset of all the people just introduced as existing in the picture. The set of people with a light are a portion included in the larger group of people.

IDWe can also say that the smaller set of people with a light are a specific group O inside the large group of people. The learner indicates this relation with the pronoun, some, which is specific to the group of people who have/possess a light. This notion Sof specificity is an inclusion relationship; the group 'some people have a light' is a Ssubset of the larger group of people introduced between propositions four and six.

SThe use of some in this description is the proportional meaning; it specifies 00 that a subset, or portion, of the larger group of people (on the street) have lights.

Because it refers to a set of people which is already in the discussion, it is a 'specific' use of the indefinite quantifier.

i Importantly, the use of sets in this analysis is designed around parts of the brain, not linguistic analyses of constituent structures of sentences. One set contains Ci objects such as nouns, other sets contain types of adjectives, members of other sets are proper nouns, such as the names of the festivals. The sets with prepositions are nouns Slinked by instrumentals or places.

0Thirdly, the individuals from these three semantic relations function as referring expressions for establishing three primitives of referentiality within the domain of discourse. These same three semantic relations function locally within the sentence boundary and at the level of the entire sentence, establishing meaning at the level of propositions, and most importantly globally over the entire discourse. These are represented as functions between sets with the discourse as one Power Set consisting of the intersection of sets and their members. This differs from other models which either 'glue' sentences to a discourse, or study only discourse fragments.

One type of representation of the interaction and the semantic relations of the entire discourse between the two participants is given below: X y- LEARNER s3 {<Kanto Festival>} INTERVIEWER-i2 picture of Tl= 8 MR y2 it(u), x=u Kanto Festitval (<Kanto Festiv =:(Piop (entities in Prop 8) 8 objects visi le only-t--3 KA f. Common Ground L y .{<KantoEesti.v.al>_.(y.) K KA I thinkuis y

KA

The largest rectangle represents the context of the entire discourse, the universe. The universe of referents in this discourse structure is given at the top.

Kanto Festival is placed in the highest structure to represent the relation (function) it has over the source of all referents in the discourse: the task and visual world of the learner, the auditory world of the discourse and common ground for both interlocutors, the mental image for the interviewer. This position makes it clear that it Sis irrelevant if the anaphor is visual or auditory in the semantic representation of O relationships.

It also represents its position as the variable for the intentions of both members in the communicative tasks of this discourse: the characterizational set for the learner Sand the identity relation for the hearer. Both sets of semantic relations are embedded inside the matrix DRS.

00 Two small universes are embedded inside the larger, or main, DRS. The left DRS is in boldface with a small Ky in the lower left corner. The boldface and Ky both represent visual knowledge, which includes the learner's source of visible N, referents. In this DRS, the picture in the learner's hands is represented by the visual referent This X represents a single referent for the complete set of objects C composing the elements of the picture of the festival, represented in Ky. This referent N, is at the top of the small visible world knowledge inside the auditory world in the Nmatrix

KA.

0A single variable y is at the top of the right DRS. It represents the referent for Sthe proper name Kanto Festival. The pronoun it represented by u, is a stipulated identity; both interlocutors, the learner and the interviewer, must interpret this variable u as referring to the picture Thus, the visual referent x u, is the stipulated entity or pronoun it. The overall proposition is an assertion because the interviewer is claiming that she thinks her mental representation of the picture created by the objects described orally by the learner is the Kanto Festival. This universe of discourse referents created her mental image and she is asserting these are objects in the real world of the picture. These have created a mental representation of the Kanto Festival. Therefore, she asserts the proposition that she thinks the identity of the picture is the Kanto festival, x is y.' The relation of uniqueness in a set a of festivals is indicated in this DRS. In the proposition, I think it is Kanto festival, the interviewer correctly identifies the festival as the Kanto festival.

Example 4.18) Discourse i2-s3:4 4 s3 That's all. Can you guess? i2: Hhm. I think it is Kanto festival i2: Okay. We gonna finished.

Thank you.

s3: Thank you so much.

The Identity relation is represented after the two entities and in the relation have been introduced. The semantic relation between x and z is an identity relation. Both entities are definite descriptions, and refer to the same referent although they each have their own independently identified individuals. The first referent establishes the set of properties describing/referring to the picture of a festival in the learner's hands, which is already a familiar discourse referent from the interviewers earlier question. The second set identifies the member of that set which the interviewer considers unique among the individual members in this set.

At the beginning of this section I said that we can not fully understand the meaning of each individual proposition without considering the role of each proposition in the overall discourse.

The propositional semantic relations are embedded in the overall discourse context. This is a significant departure from Schank and Abelson's concept of a O frame for the starting point of the discourse and the conclusion to this discourse are O very tightly intertwined. Each proposition of the learner is an extension of visible referents to auditory objects. But these properties do not describe the semantic relation of each proposition for the Hearer, or interviewer. These are creating a Sunique set of properties which culminate in an identificational semantic relation for the interviewer.

00 The it in this proposition refers to the first propositions of the discourse produced by the interviewer which contains the antecedents picture offestival, one, it, and whichfestival: ('i Example 4.19) Discourse i2-s3:4 C- i2: I have some picture of festival C so you can pick up one and don't show me INO s3: uh huh...

i2: And describe.it (change tape) Cq (cont.)and I will try to guess which festival) All of these refer to the same object, the picture in the hands of the learner.

The interviewer's goal is to guess the identity of the picture in the learner's hands. She explicitly tells the learner not to show her the picture. She does not know the identity. Hence, the type of information available between the two interlocutors differs. The information for the interviewer available in the context for interacting in the process of guessing creates a special kind of mutual knowledge. The entire discourse is based on mutual knowledge which is fundamentally different: the learner knows the identity of the picture, the interviewer must guess, or infer, its identity.

In the thesis, Chapter 6 shows the interaction of grammatical relations with the three semantic structures. Chapter 7 outlines the interaction of the three referring types with topic, a binary distinction of given and new in the level of information structure. Like the subject-object distinction, this means that within a clause more than just one entity must be examined to determine the information structure properties. Again the three semantic definiteness structures are shown to exist as a template underlying a level of information structure. A secondary goal is to show that the three semantic structures based on definiteness and specificity, cannot distinguish the notion of topic and are therefore not a useful diagnostic test. Instead, topic and comment is a binary relation superimposed on the semantic definiteness structures.

Chapter 8 investigates the interaction of the referring expressions at the level of the discourse. Discourse Representation Theory (DRT) is used as a representational technique for the study of topics in discourse. Lexical Functional Grammar (LFG), a theory of parallel constraint-based syntax, is used for the organization of the grammar and representation of sentence level properties of subjects and topics. These two theories, DRT and LFG, are used to represent the formal properties for the interaction of the sentence level subject and topic to the discourse topic and definiteness.

However, both of these theories were used only for the purpose of examination of the thesis. Chapters 3-7 all use sets composed of lexical notions as members of sets. It is this representation that I intend to develop more fully in the patent application. Using the concept of object of definiteness and specificity the sets are automatically integrated into the discourse, whether these are composed of \0 interrogative indefinite objects, such as who, which, where, or pronouns, such as 0 one, some, he, or nouns or events.

SThe relevant components of each level in the grammar and the section Saddressing their respective systems for decomposing the sentence "He is a syntactician" is given below (adapted from Fellbaum 2001).

00 LEVELS OF GRAMMAR 00 oo Syntactic Structure: Pro He Subject: Grammatical Relation (Word Order) Semantic (Conceptual) Structure: Semantic Aspectual Types Semantic Roles (Nouns) Det N is a syntactician Predicate (V-Obj) V (0) Subj

S

Theme Stative Theme Semantic Predicate Types Semantic Specificity Information structure Characterizational (+spcc) (-spec) Topic (Given) Focus(New) Pragmatic structure: Sentence topic (Information structure) intersects the discourse Pragmatic Definiteness Def/ Indef N Annexure 2

O

Chapter 1 00 Introduction Introduction SA common way to approach understanding a complex phenomenon is to decompose it into simpler components and more manageable subdomains. After the subdomains are more clearly articulated, the interactions between and among separate subdomains can N be studied. This thesis presents a methodology for the study of the complex system of ND semantic meaning of definiteness and referentiality in interlanguage. The concept of Sdefiniteness is decomposed into its most minimal parts, and then gradually developed ri into a system of patterns which function as referring expressions for establishing meaning within a domain of referentiality. Once the system's parts are established as an autonomous level, their interactions and correspondences with other levels in a grammar, in particular with respect to the behavior of predicate argument structure, grammatical relations of subjects, and finally information structure of topics, are demonstrated.

The levels are organized according to the systems of a parallel-constraint based grammar, such as LFG: constituent structure and linear order of categories and predicate argument functional structure. In addition, autonomous levels of information structure and semantic definiteness and specificity are proposed which operate in parallel.

Chapter 2 describes the tasks and experimental design used for the empirical study of definiteness. The study uses four tasks for the empirical context of aural discourse and visible referents. Since this thesis is not a traditional morphosyntactic analysis, the tasks are described according to characteristics of mutual knowledge and their types of semantic definiteness structures.

Chapter 3 argues for the autonomous nature of definiteness, exemplifying their presence in all parts of speech and constituent structure. Adapting Clark and Marshall's 1981 analysis of definiteness, which distinguishes shared from mutual knowledge, definiteness is decomposed into visible, visual, auditory and inferred semantic structures. The sources of mutual knowledge in the interlanguage discourse are directly connected to the context established by the four tasks. Throughout the chapter, these four types of definiteness in the context of the common ground of the utterance are described according to nine mental resources. This further establishes the unique processing computations of definiteness, independent of constraints at the level of linearity.

Chapter 4 establishes clear patterns of semantic definiteness constructed from three basic 'referring expressions'. These three types of knowledge expressions, are shown to operate locally at the level of the sentence and also more globally over the discourse. Most significantly, the empirical description of one task clearly elucidates the differences in mental representations of speakers and hearers in a discourse.

These are not composed of the same set of referring expression types between the speaker and hearer in the dynamic interactions of discourse. This chapter also shows D that the referring types form core semantic structures of knowledge representation composed of three patterns of definiteness/specificity relations.

The purpose of Chapter 5 is to demonstrate that the three specificity patterns of meaning are another level of representation in conceptual structure together with Jackendoff's lexical-semantic representations. The semantic definiteness structures from Chapter 4 are based on the decomposition of properties of semantic specificity 0and underlie Jackendoff's lexical semantic representations, based on the lexical decomposition of verbs. It is first demonstrated that thematic roles are restrictively neutralized in the patterns, so the three specificity patterns are argued to be rNl autonomous units at the level of Conceptual Structure. A-structure constructions are organized according to Jackendoff's conceptual structures, action tiers and thematic tiers, including all of the a-structure constructions proposed by Manning (1996).

Chapter 6 is divided into two parts. First, I use tests for establishing the Sexistence of subject in the grammatical relations of the developing interlanguage.

OThen I focus on the specificity properties of sentences and the patterns of meaning and their interaction with subject found in the corpus. For evidence of subject in the interlanguage constructions, I look for evidence of subjecthood, such as more than one semantic role, sentence-initially and preverbally, subject ellipsis in complement clauses, case-marking, and subject-verb agreement.

Chapter 7 first reviews the SLA literature on topic, divided into formalist and functionalist studies, concluding with a discussion of the interaction of the three referring types with topic, a binary distinction of given and new in the level of information structure. Like the subject-object distinction, this means that within a clause more than just one entity must be examined to determine the information structure properties. The main goal of this chapter is to show that the three patterns of meaning based on definiteness and specificity at the sentence level, confuse the notion of topic and are therefore not a useful diagnostic test.

Chapter 8 investigates the interaction of the referring expressions at the level of the discourse. Discourse Representation Theory is used as a representational technique for the study of topics in discourse. Lexical Functional Grammar, a theory of parallel constraint-based syntax, is used for the organization of the grammar and representation of sentence level properties of subjects and topics. The thesis culminates in the connection of the sentence to the discourse where the interaction of the two theories, DRT and LFG, is most useful. These two theories are used to represent the formal properties for the interaction of the sentence level subject and topic to the discourse topic and definiteness.

To summarize, the relevant components of each level in the grammar and the chapter addressing their respective systems for decomposing the sentence "He is a syntactician" is given below (adapted from Fellbaum 2001).

IO

0

N

tq c^ oo LEVELS OF GRAMMAR Pragmatic structure: Chapter 8 Sentence topic (Information structure) intersects the discourse (Topic Topic Given Focus/New) Syntactic Structure:

VP

V NP Det N is a syntactician Pro He Chapter 7 Topic (Information structure) Chapter 6 Subject: Grammatical Relation (Word Order) Topic Given Focus/New Subj Predicate (V-Obj) S V (0) Semantic (Conceptual) Structure: Chapter Semantic Aspectual Types Semantic Roles (Nouns) Chapter 4 Semantic Predicate Types Stative Theme Theme Characterizational Chapter 3 Semantic Definiteness Def/(+spcc) Indef(-spec) Chapter 9 is the conclusion to the thesis, presenting a critique of the experiment and corpus, significant findings in the thesis, and questions and directions for future research.

NChapter 2

O

The Experiment 02.0 Introduction This study is an in-depth, empirical description of the conceptual semantics of definiteness and its interaction with the linearity of subject and topic as found in the Ninterlanguage of a task-based experiment. The semantic structures associated with subject, and the pragmatic structures of topic in a discourse intersect at the level of the cK1 sentence in the level of syntax. The data base is an experimentally designed and controlled 24 hour corpus of language structures produced by Japanese learners of NEnglish as a second language (ESL). The full experiment (see Appendix I) includes Onative American speakers as a control group. However, only 12 hours of the nonnative speaker conversational dyads are used for the study, i.e. Japanese NNS-NNS.

A primary focus of the task design is the isolating of types of definiteness and specificity in the context of the interlanguage tasks. It was hypothesized that the shared cultural knowledge of festivals would maximize the production of patterns of mutual knowledge, or definiteness, between speakers. Therefore, Japanese festivals were selected as the discourse topic for the instruments of elicitation. This in turn augmented their use as instruments to study definiteness and topic at the sentence level. A second set of mutual knowledge was the shared academic and residential daily life of both partners in the dyads where the experiment was conducted.

2.1 Background Information In order to have a clear understanding of the 'real world' of the participants and their discussions in the study, it is important to understand the environment in which the experiment took place with respect to its setting in Japan and the academic community.

The experiment was conducted at Minnesota State University-Akita campus, referred to as MSU-A. MSU-A was a small two-year university on the northwest coast of Japan. It was conceived by Prime Minister Nakasone and President Reagan in the mid-1980s to be a joint venture between the two countries. This was an initial attempt to open up trade barriers by importing American education to Japan by founding an American university in Japan. Thus, this distinguishes the university from other American universities which were established in Japan, primarily for recruiting students for home campuses in the United States or as business ventures for the purpose of raising revenue. This difference in goals is evident in such areas as staff recruitment and quality of equipment available in classrooms and on campus.

The university was divided into two departments: the ESL Department and the Department of General Education All classes at the university other than the Japan Area Studies (JAS) Department were conducted in English. The ESL staff were permanently located at the site in Japan. All students were required to complete the equivalent of a six level ESL program, which normally took eighteen months of instruction.

The GE staff, on the other hand, were on one-year contracts from American universities, with priority given to applicants from Minnesota State University academic staff. Thus, although the lecturers came from all over the United States, many came from branch campuses in the Minnesota State University system.

ISome students chose to attend MSU-A only to study English and/or receive an C associate of arts degree and have no intention of further study. Other students completed the ESL program and the equivalent of two years in the General Education Department at MSU-A in Japan, and then transferred to an American university of their choice in the United States. After finishing another two years of tertiary education in the United States, students graduated with a four year degree. Choice of 00 major and possible universities in Minnesota, which have Native American Indian names, such as Mankato, Bemidji, Winona, and St. Cloud, as well as their intention to only study English, are all mentioned in the discussion activities.

2.1.1 Academic Environment N The experiment for this study was conducted as part of a course option for students in the Department of ESL Production Skills Module. The student interviewers were in I the final ESL level, 6, and the teachers were from the General Education Department.

2.1.1.1 English as a Second Language The content-based English as a Second Language (ESL) program for the first three levels for first year students was called Focal Skills, designed by Ashley Hastings, and included five-week modules of Listening, Reading, and Production. Students moved to the next skill, or module, usually requiring ten weeks per module, as soon as they passed a test. The new Focal Skills curriculum had only limited beginning conversation activities at the earlier modules, since they were not speaking modules, but rather listening and reading. If dialogues were present in the classes, they were included as listening practice and grammar review, e.g. Side by Side (Molinsky, 1989) rather than conversation speaking practice. Student interviewers in the conversational dyads make several references to this new "Focal Skills" program, because they were curious to find out how it differed from the ESL program they had just completed.

The Production Skills module, designed for public speaking and practicing speech acts, such as invitations and apologizing, also functioned as a pre-academic course for the required Speech Communications course in the General Education Department.

2.1.1.2 General Education This department was equivalent to the branch campus in Minnesota meeting the curriculum and degree requirements for the Minnesota State system, and included Minnesota and other American university students for a year abroad. All students completed the equivalent of two years of academic course work in the General Education Department with courses taught by academic staff in areas, such as psychology, history, anthropology, mathematics, or speech communications, typical of a liberal arts degree curriculum in the United States. Japanese students began to acquire the language skills necessary for these courses in the content-based preacademic ESL curriculum. When this experiment was conducted, the Production Module functioned similarly to a sheltered ESL class for the required Speech Communication class in the General Education"' Department. Since this Speech Communication class was a required course for the degree, Level 6 students also took a Speech Communication as an ESL adjunct course, once they entered the General Education Department.

The General Education Department also had a Japan Area Studies program.

Students in this program came from America, Australia and even Korea to study the IJapanese language and culture. The interviewers in the experiment, four JSL teachers, two male and two female, were from this program.

(Ni 2.1.2 Non-Academic Environment (Residential Life) Residency at the university site was considered part of the educational plan of the university. All students were required to live on campus in residence halls attached to 0 the school cafeteria during their first year of attendance at MSU-A. Since the university actively recruited American students to come and study Japanese as a second language (JSL), Japanese and American students interacted in each others' (N language while exchanging information about each others' culture. Books and video tapes in both English and Japanese were available from the university library. A (N Student Activities Director and Residential Life Director were constantly providing activities, such as seminars, sports competitions, or sightseeing trips, to encourage Imixing between the ethnic groups on campus. The televisions in the lounges of the residence halls had both Japanese and American broadcasts. Japanese and American students were given the opportunity to live together in an arrangement of two suites of two students per room with a shared bathroom. This created a very small and close community among the students; students knew each other either from sitting next to each other on a bus to a school activity or seeing each other in the cafeteria or in the residence halls.

As explained above, the environment at MSU-A was not a typical foreign language learning environment for an ESL second language learner. ESL students were in a foreign language environment, i.e. rural Japan, where English is of limited use, with Japanese the primary language spoken by non-academic staff at the university. Outside the university, English was non-existent as a means of conducting business with persons in the local area, a rural community near a major fishing and shipping port in the northwestern part of the country on the Sea of Japan.

Nonetheless, even though it was situated in northwestern rural Japan, the close academic and recreational interaction provided ample access for ESL students to native speakers of the target language, English. The university created a community of total immersion in English, if a learner chose to do so. All students, Japanese and non-Japanese, intermingled regularly, discussing current campus activities, each others future plans and possible university locations and academic fields for further study. It is this shared academic and residential daily life of both partners in the dyads of the experiment that created one set of "real world knowledge" in the first couple of minutes of introductions and getting to know each other.

2.2 Method 2.2.1 Experimental Design The study is based on oral production data elicited during 15 minute structured conversations. All participants in the experiment are Japanese, non-native speakers of English, the Target Language. Each dyad consists of an interviewer and a learner, or student, in the "discussion activities" of the experiment.

There are eight interviewers in this report of the experiment, four female and four male. The four Japanese teachers, two of each sex, are from the Japan Area Studies the program in the General Education Department and teach Japanese as a Second Language (JSL). The JSL teachers are the most advanced in their level of English. Two had lived in the United States: one taught Japanese for 25 years; the other was pursuing a doctorate in Japanese linguistics. The other two JSL teachers had O never lived outside of Japan. All of these interviewers were eager to meet the 0 students from the ESL Department to ascertain their level of English skills prior to their entrance in the GE Department. All of these interviewers are able to take credit a for service to the university community in their annual reports to the Dean and tenure Scommittee, so volunteered for the experiment and were not paid.

V)The remaining four interviewers are two students of each sex, who had earlier 00 been my ESL students, and were currently enrolled in Level 6 of the ESL program.

They were now also taking General Education classes, or the ESL adjunct class for Speech Communications. All student interviewers were paid for their participation in CK, the experiment.

The interviewers' status and sex is summarized below: SInterviewer Sex Interviewer Status SJapanese Female 4 Japanese (JSL) Teachers 4 SJapanese Male 4 Japanese students (ESL) 4

NO

The experiment is divided into two parts, Experiment I and Experiment II.

Experiment I included nine students and began in January and ended in March.

Experiment II included another group of six students initially, but reduced to four and began in mid-February and ended in March. Each experiment had a different set of interviewers and the order of appearance differs between the two experiments.

Experiment I has males alternating with females. Experiment II has all male interviewers first, followed by the female interviewers. A summary of the order of interviewers in the experiment is given below.

Experiment I Experiment II (duration 8 weeks) (duration 4' weeks) 1. Japanese Male Teacher 1. Japanese Male Teacher 2. Japanese Female Student 2. Japanese Male Student 3. Japanese Male Student 3. Japanese Female Teacher 4. Japanese.Female Teacher 4. Japanese Female Student The Japanese interviewers are coded for interviewer according to their time of appearance in the experiment. Their identity in the experiment is ii, i2, i3, i4 in Experiment I and i 10, i12, i14, i16 for Experiment IIPV.

The participants, students or learners, in the experiment are twelve native-speaking Japanese learners, eight male Ss and four females Ss. Although initially there were 6 females, two dropped out of the program. However, as sex is not seen as a variable in this analysis of language skills, the fact that there were unequal numbers of male and female participants is not seen as problematic unlike SLA studies of negotiation strategies (Holliday and Fellbaum 1996). No differences in production due to sex were expected, although this can be tested at a later date. All of these participants volunteered for the experiment and were given credit in their class work.

The learners were randomized and assigned a number according to their first appearance in the experiment. Learners (students) are coded with and their rank according to their first interview, e.g. sl s13v. Below is a summary of the learners i.e. 8 students in the first experiment and 4 in the second, for a total of 12 students in both experiments: Experiment I Experiment II 8 x 4 32 interviews 4 x 4 16 interviews ID32 16 48 interviews SWith a total from both experiments of 12 students and 4 interviews each, the a total number of interviews is 48, with 32 in experiment I and 16 in Experiment II.

SEach interview, or dyad, was 15 minutes in length. Below is a summary of V)speech production time in both experiments: 00 8 x 15 minutes 2 hours/student 48 dyads x 15 minutes 12 hours of speech Each dyad was coded for both interviewer and student: ils2 refers to interviewer 1 r and student 2.

In summary, the data for this study is from a corpus of native Japanese N speakers using English. The dyads consist of speakers at several levels of proficiency.

r The experiment controls the type and quantity of language. Although this can be IDlimiting, it also creates a wealth of data to study patterns of shared, mutual knowledge in language within a homogenous community of speakers.

2.2.2 Participants 2.2.2.1 Proficiency Level All of the participants in this study had enrolled ten months earlier, in April, 1993, although they had only had twenty weeks of ESL. The PreTOEFL test scores of the students at the time of the experiment, ranged from 403 to 500, with 500 the top score, for an institutional TOEFL test, administered immediately prior to the start of the experiment.

It should be kept in mind that even though mostv of these Japanese students had not been exposed to native speakers of English prior to their entrance at MSU-A, they had been exposed to English education. All had completed six years of a quasigrammar translation English education required by the Mombosho, or Japanese Ministry of Education, by the end of year twelve in secondary school. The proficiency tests for speaking, listening, and writing administered for the Focal Skills modules, ranked all the students in the experiment as low intermediate. However, since their PreTOEFL scores did not determine their placement, the knowledge and proficiency of grammar and vocabulary for some students who scored 490 or 500 on the PreTOEFL would be considered advanced and sufficient to enter an American university".

At the start of Experiment I the students were enrolled in the first five weeks of the third module, Production Skills, with one class in speaking and discussion, respectively. The second five weeks they matriculated to two new classes, a discussion and writing class, respectively, in the Production Module. The four students in Experiment II, which lasted only five weeks, were enrolled in the first five weeks of the speaking and discussion classes in the Production Module.

Considering their achievement in the ESL program, we could say that the motivation and desire to learn English by all of the students in this study is quite high.

Their motivation, study habits, and attitude (Ellis, 1985) are all positive toward learning English and completing a university degree. Their reasons for choosing an American education instead of attending a Japanese university vary from fascination with America, and/or desire to major in a field which was not available at universities in Japan, such as psychology, environmental science, or automobile design engineeringv"'. Thus, as a group, they can be considered to be both integratively and ND instrumentally motivated. They are integratively motivated because of their desire to identify with American culture and reject the education system of their own country c, for that of an American university. They are instrumentally motivated since their goal was to achieve a university degree from an American university (Gardner Lambert, 1959).

00 2.2.2.2 Age Other than two older policemen, all participants are between the ages of eighteen and twenty-three. The two policemen are twenty-seven and thirty-one, respectively. One other student is twenty-three and is completing her degree at a Japanese university at the same time she is enrolled full-time in the ESL program at MSU-A. Thus, the Ss (N are at an age where we can expect difficulties in acquiring native-like proficiency in (,i pronunciation (Oyama, 1973, 1976) but since they had started English training in the ND sixth grade, around puberty, SLA studies, although controversial and methodologically problematic, suggest that reasonable progress could still be made in N acquiring English grammar and vocabulary (Long, 1988; Long Crookes, 1989).

2.2.2.3 Input from Target Language Contact or exposure to native speakers of English, the target language, was extremely limited prior to the students enrolment at the university. The oldest S, a policeman, had lived abroad for one year, as he studied at a police academy in Los Angeles, California. One other S had visited the U.S. for two weeks, four months prior to the start of the study. Other than these two students, none of the students had prior experience with living among native speakers before entering the university.

However, once the students arrived at the university, all were exposed to Americans but the level of input from the target language is difficult to assess accurately. All students began in the first level of the ESL program, and all but two of the teachers in the ESL program were native speakers of American Englishx. The residential life program at the university provided another means for interacting with Americans. Although none of the students in this class had American roommates, they were able to interact with Americans in the dormitories and some had friends who did share rooms with American students. Even though all of the students had the opportunity to eat and interact with Americans on a daily basis, most of the students, as first year students, were still quite reserved and stayed in groups with other Japanese students at lunch and on school activities. Nor did they interact with American students in the classroom, since they were still enrolled solely in ESL classes.

An additional problem with accurately assessing their input from the American students is created by the desire of the American students to practice their Japanese with the Japanese students, since that was one of their purposes in enrolling in the university. In other words, even if they did have an American friend, the students said they usually spoke in Japanese, not Englishx. The exact level of input from target language speakers is thus difficult to measure. Their only regular, quantitatively assessable interaction with Americans at the time of the study was with their ESL teachers in the Production Module, who were both American.

The nature of the input is relevant when language data is imbalanced between the tasks. For example, Task I asks personal questions regarding hometown, future plans, etc. A student may have experience answering these questions on a daily basis.

Nonetheless, even though the students rehearse these tasks, the data can still be O searched for language patterns consistent with other tasks. Since I am not doing a 0 developmental analysis or classification according to stage of language skills, this is not an inherent problem. It can even be useful if the learner exhibits the same pattern a across all four tasks. We can then hypothesize that the language patterns in the data Sare representative of the individual's language system at that point in time.

00 2.2.3 Materials for Elicitation The four tasks used in this experiment are common in communicative classrooms for language teaching. The fifteen minute discussion activities were divided into two 1 parts: Part I included two tasks with no visual aids: Task I, an introduction (informal greeting) giving personal information with short discussion about the subject; Task II, ,I a description of a festival near the subject's hometown. Part II included two tasks with ,I visual aids: Task III, a chart with a one-way information gap activity; Task IV, a O picture description"'.

S2.2.3.1 Choice of Discussion Activities (-i The topic of Japanese festivals was selected as a theme throughout the fifteen minute activities for several reasons. The first reason is that it is a familiar topic to all the participants, and, therefore, provided an opportunity for students to discuss shared cultural knowledge with each other. The importance of a familiar topic for conversation and the related concept of shared knowledge has been discussed in the second language literature by (Kolb, 1984)who says: immediate personal experience is seen as the focal point for learning giving life, texture, and subjective personal meaning to abstract concepts, and at the same time providing a concrete, publicly shared reference point for testing the implications and validity of ideas created during the learning process....(Kolb 1984:36) The importance of shared cultural knowledge has also been studied in the psycholinguistic literature by Tulving (1985) who says: semantic memory represents our "shared" memories of general knowledge about the world...Semantic retrieval on the other hand, may not be affected by a lack of precise communication to the same extent, as semantic retrieval is assumed to represent a substantial degree of shared knowledge about the world (Andersson 1996: The shared knowledge in this experiment arises from the general context of Japan for Japanese students, the small local community within and surrounding MSU- A, and its ongoing activities, e.g. festivals, a familiar topic with students during the course of the study. Since the Student Activities Director at the university organized the students to attend the festivals, providing special buses for direct transportation to the local festivals, it was assumed that the students had heard of and possibly visited the festivals on the information gap chart (see Announcements were constantly circulating about the festivals and the school hallways were lined with posters furnished by the prefectural tourism office. In addition, the students in this study are all Japanese, and the festivals, famous all over Japan, were covered by the news media on a regular basis.

O Equally important in the choice of festivals as a topic for discussion in the O experimental design was that there were a sufficient number of festivals so that no Sinterview contained the same set of information, even though the format of each a discussion activity was similar. Thus, the content of each information gap activity as Swell as the picture description did not overlap and could be controlled throughout the Sfifteen minute discussion activity. This also reduced the cognitive load of processing 00 the communicative strategies required for performing the tasks. At the same time, repetition of the theme allowed some familiarity with the tasks, thus assisting the students to focus on the communication of meaning alone, allowing each S to perform r at an optimum in terms of language production. In other words, the dependent variable of task complexity was minimized and/or removed so that language ,i production itself could be more optimally tested.

(N A final motivation for the choice of festivals for the discussion and Sinformation gap activity was that it provided an opportunity to add to our knowledge of the acquisition of interrogative structures which can be used to study a learner's stage of acquisition. Previous studies of interrogatives were not as controlled as information gap activities allow. "Spot the Difference" tasks elicited Yes/No or "What" questions with nominatives as answers from the partner (Mackey, 1994, 1999). For example, questions participants would ask were "Do you have or "Is there a The information gap activity designed around the theme of festivals in this study, challenged the participants with seven interrogative (WH questions) forms: Who, What, Where, Why, When, How, What kind of Thus, this provided a complete, controlled set of interrogative structures in English useful for testing and establishing the stages of a learner in the Multi-Dimensional Model (Meisel, Clahsen Pienemann, 1981) or Processability Theory; (Pienemann, 1998). The learner's formation of interrogatives with respect to word order has been used as a criterion for defining the exact stage of acquisition of the target language (Pienemann Johnston, 1987; Pienemann, Johnston, Brindley, 1988; Lightbown Spada 1993; Mackey Philp, 1998). Although this report of the experiment does not address either the processability of questions or stage of development, the word order patterns are briefly noted in the discussion of the Information Gap activity and question formation in syntax discussion in Chapter 7 and discourse features in Chapter 9.

2.2.3.2 Descriptions and Classification of Tasks The materials used for elicitation of the data in this study are all tasks as defined in (Richards, Platt, Weber, 1985): any activity or action which is carried out as the result of processing or understanding language as a response). For example, drawing a map while listening to a tape, listening to an instruction and performing a command, may be referred to as tasks.

The specific tasks chosen for this experiment are: I. An introduction giving personal information; II. A description of a festival near the subject's hometown; III. An information gap activity about Tohoku festivals; IV. A picture description of Tohoku festival.

Key characteristics of task-based language teaching in each task are that they emphasize interaction in the target language, use authentic texts, and enhance the I learner's own personal experiences (Nunan, 1991). The tasks in my experiment, 0 however, are not genuine pedagogic tasks as described by Nunan, because mine do not contain the steps of carefully planning a task in the classroom. These classroom tasks require identifying a target task, providing a model, identifying an enabling skill, Sand culminating in actually simulating the targeted task. The experimental tasks are not related to the classroom activities.

00 Instead, the tasks in this experiment can be described in the sense defined above by (Long, 1989), and are more loosely characterized by (Breen, 1987) as any structured language learning endeavour which has a particular objective, appropriate content, a specified working procedure, and a range of outcomes for those who undertake the task. 'Task' is therefore assumed to refer to a range of work plans which have the overall purpose of facilitating language learning-from the simple and brief Iexercise type, to more complex and lengthy activities such as group problem-solving or simulations and decision making (Breen 1987:23).

The tasks, described more extensively in are simple, designed purely for the purpose of obtaining data on the language processing abilities of the participants. These structures are studied in detail in separate chapters for their patterns of definiteness and specificity and their characteristics with respect to their function in a sentence (Chapters 6, 7, and The semantic structures are tightly correlated to their communicative purpose in a discourse (Chapter Indeed, according to (Brown Yule, 1983) descriptions are easier than instructions which are easier than stories. Two out of the four tasks were descriptions.

In addition, the participants were expected to have some degree of familiarity with the topics in the brief discussions in the first two tasks, which should have facilitated the student's language production in the discussion (Anderson and Lynch, 1988). Moreover, since the interviewers in this report of the experiment are other learners rather than native English speakers and the students are paired with interviewers of higher proficiency level, the tasks also should be optimal for maximizing quantity of speech (Porter, 1983).

2.2.3.2.1 Task I: Greeting and Introduction Activity The first task was an introduction, or greeting, between the two conversationalists and included discussion regarding reasons for attending the university and the subject's future plans (See This task was between a structured and semi-structured interview. The interviewer had a list of questions to guide the interview, but was guided by the responses of the interviewee rather than restricted to a list of questions.

The interests of the interviewer also directed the flow of the conversation (Nunan, 2004). This is a closed task according to Nunan (1991). It is also a "well-defined cognitive process" according to Kitchener's model of cognitive processing, since the personal information offered by the subject in this task, such as the subject's name and hometown had "absolutely correct and knowable solutions" (Kitchener, 1986:223).

O 2.2.3.2.2 Task II: Description of a Festival without a Visual Picture In the second task, participants were asked to describe a festival near their hometown, since small village festivals were common and most cities and towns still have them Stoday. These festivals originated as rituals and rites to a special god/goddess to protect the villagers' livelihood, by warding off danger or praying for a successful harvest. Thus, each festival had evolved with special characteristics unique to the 00 villagers' livelihood, legend or location. Historically, Japanese villages each had their own god, with village festivals being common and it was expected that all students would be able to describe either their own hometown festival or one very near their home. Since festivals were common, it would be unusual, if not impossible, to grow up in an area which did not have one.

(N

2.2.3.2.3 Task III: Information Gap Activity IO The third task is a one-way information gap chart, requiring "required" rather than "optional" information exchange (Long, 1981; Doughty 1988; Pica Doughty, (N 1985). For each discussion activity students have one chart with blanks and the interviewer has a different chart with the missing information in English. Each chart includes four local festivals of the northern Tohoku area (on Honshu, the main island of Japan) with ten different types of information about each of the four festivals. The first line on the chart includes the 'festival name' of each of the four festivals. The second row has information regarding the 'meaning' of the respective Japanese festival names. For example, the festival Namahage means 'Foreign Demon' in English. The third row provides information regarding the 'location' of the festival.

The fourth row gives the 'date' the festival is celebrated each year. The fifth row contains information about the 'origin or purpose' of the festival. For example, the origin of the festival Namahage arose when the ancient Chinese Khan and his men attacked the villages on Oga peninsula, located on the northwest coast of Japan. The purpose of the Kamakura festival is to celebrate the water god and ask him for water.

The sixth row gives the 'age' of each festival. The Dainichi-do Bugaku festival, for example, has been celebrated for over 1200 years, while for others, such as Kamakura, the exact date of origin is unknown. The chart in these cases either has or "very long time". The seventh row contains 'special traits' of the festivals.

For example, the Tsuzureko Odaiko festival has a 2.71 meter taiko drum, the largest in Japan. The eighth row contains 'actions or activities' unique to the respective festivals. In the Dainichi-Do Bugaku festival people dance to ancient court music.

Information on festival 'participants' in the Namahage festival, such as "Men of town", is included in the ninth row. The final piece of information given in the chart is additional 'special facts', such as a 'permanent taiko display' for the Tsuzureko Odaiko festival. A frequent fact in this row is that the festival was a 'national holiday'.

In total, each information gap activity has four festivals with ten items of information about each festival. The students were asked to find information for eight to twelve blanks, the number being randomized over the period of the experiment.

The Information Gap charts are given in Appendix 2x.

2.2.3.2.4. Task IV Picture Description with Visual Picture The final task required students to describe a picture of a Japanese festival. The pictures depict various scenes or activities of festivals in the northern Tohuku area of Japan, taken from a tourism brochure published by the Akita prefectural office of Ntourism. All of the pictures include people but the activities and scenes differ 0 according to the unique characteristics of that festival. For example, two of the pictures are of snow scenes; one has igloo-like structures with people sitting inside the kamakura sharing sake and rice cakes and an old Japanese farmhouse with snow on Sthe rooftop in the background (Kamakura). The other picture with snow has masked men winding down the side of a snow-covered mountain in straw costumes carrying a 00 torch on a long pole (Namahage). A third picture includes men in white costumes carrying a shrine (mikoshi) and running into water under a waterfall. A fourth picture is of a large drum with drummers seated on top of as well as standing on the ground, N dressed in traditional dress for performing taiko drum music and dance. Most are using a drumstick to hit the taiko drum. A fifth picture is of a group of women in traditional summer dress (yukata) and with special hats dancing in a line with their Sleft palm faced down and fingertips turned up (Bon Dance). The final two pictures (,i are night scenes with bright lights. The first has large bursts of fire in the sky, caused Oby fireworks in the night sky. The ground below has little tents in a circle, also brightly lit with lanterns. This is the famous international Omagari festival that attracts famous fireworks' specialists from around the world. The last picture has hundreds of candle-lit lanterns on bamboo poles with people dressed in black shirts and white shorts and headbands, balancing the poles on different parts of their body (hips and arms) while marching in the streets of Akita. This is the local Akita festival known as Kanto and is well-known throughout Japan. Pictures of all the tasks are given in Appendix 2. Learners described these pictures with varying degrees of accuracy and detail during Task IV.

In order to accurately describe the patterns of definiteness, I needed to include the patterns created by visible and visual definiteness. Since the constraints on their use seemed similar, the tasks were described according to their visible as well as acoustic characteristics. Acoustic refers to their speech production in the discourse, and visible refers to their use as antecedents based on visible objects in the discourse.

The four tasks used in this experiment have two with mutual knowledge created by acoustic input only. The last two are a mixture of visible/visual knowledge and acoustic input. They are: Task I: Introduction (informal greeting) giving personal information with short discussion about the subject (acoustic only); Task II: Description of a festival near the subject's hometown (acoustic only); Task III: Chart with a one-way information gap activity (visible and acoustic); Task IV: Picture description with the interviewer guessing its identity (visible/visual/acoustic).

2.2.4 Spontaneous Versus Elicited Speech The tasks in this study are spontaneous to the extent that the students were able to answer freely in the discussion activity and some of the questions were open-ended.

They were controlled in that the topics, tasks and visual cues controlled their options for discussion. The tasks were partially natural and partially elicited. During the report of this study, I use 'interview' instead of discussion activity, for convenience, hence the interviewer label. The recording equipment was in complete view of the students at all times (see D From the point of view of the students, these fifteen minute interactions are termed "discussion activities" to fulfil a homework assignment for the Speaking module. Students were required to speak in English outside the classroom each week.

They were not graded, but a student could not complete the module and progress to Level 4 in the ESL program if they did not do this. Each student determined the number of minutes they would engage in English each week and recorded this in a 0"Homework Log" which was submitted to the teacher at the end of the week. The only requirement for credit was that they complete the amount of time they had contracted at the beginning of the week. Thus, the discussion activities in this N experiment assisted the shy students in seeking out American "homework" partners, since they were allowed to count the time spent during this experiment toward the SHomework Log necessary for fulfilling class requirements and passing the module.

SThis in part was due to ethical concerns of the rights of students in the study. (Note Sthat not all students in the module participated.) O2.3 Procedure 2.3.1 General Description The students are divided into two groups. In the first group, eight students met with an interviewer once a week for fifteen minutes over a twelve week period; in the second group, four students met twice a weekx over a five week period. Thus, the experiment is designed to be an intensive longitudinal study of language change over a short period of time, rather than a longitudinal study with sessions once or twice a month for a year or even three to five times over an eighteen-month period, which is common for longitudinal studies. This was done in order to map more minute changes in more detail in a short period of time sufficient to affect a change in language proficiency. According to psychometric studies of language proficiency tests, such as the TOEFL, six weeks is the normal time for a statistically significant change in proficiency in language development (Porter, 1983).

All learners were required to come to the room across from the experimentation room (also classroom, See fifteen minutes before their scheduled time. Again, the point of the study is not to test their strategies for negotiation of meaning, but rather their language processing abilities. So, every attempt was made to give them opportunity to use English immediately prior to the interviews. During the waiting period prior to entering the experimentation room, they spoke in English with the experimenter and among themselves reviewing the Information Gap chart as a means to both foster discussion and familiarize themselves with its contents. They were not permitted to speak in Japanese. This acted as a warm-up period for the students, so that they could begin thinking in English for the discussion. However, it also ensured that the experiment would run smoothly and all participants were in place at the designated time.

2.3.2 Instructions for Tasks The first task begins with the interviewers introducing themselves and asking for the students name and other questions usual for an introduction or greeting (see Appendix Although these questions are printed on a sheet of paper to assist the interviewer, they were told they could deviate from the guide sheet and introduce other questions if it seemed appropriate or they were interested. However, they were to be sure to ask all the questions on the sheet. The other three tasks all had specific lead-ins to help guide the discussion. For example, the second task began with "I'm interested in the N festivals of Japan, so The third task began with "You have a chart and I have 0 information. So you need to ask me questions to fill in your chart." These directions varied according to each individual interviewer, but remained close to the model. The final item in the fifteen minute interaction was a picture description task. The Sstudents were instructed to select a picture which was face-down in a folder so the interviewer could not see the picture beforehand. students were then asked to describe OO the picture and not show it to the interviewer. These instructions were not printed on the guide sheet since the interviewers had to hold the folder while students selected a picture. Nonetheless, all but one of the interviewers were able to follow the N instructions given by the experimenter. The interviewers often assisted the students in their description by asking for more information, if the S was hesitating, in order to keep the S talking.

ri The experimenter either waited until she heard "Thank you" from the (,i Sinterviewer, or when the 15 minutes had elapsed, she knocked on the door, and the 0 interviewer ended the discussion with "Thank you" and the experimenter opened the door. The S left the room, the tapes were checked, and then the next S entered the (Ni room.

The experimenter controlled the time but instructed the interviewer to keep talking until fifteen minutes had elapsed and a knock would be heard at the door to end the discussion. The importance of this instruction was determined to be necessary from results of the pilot. Some students behaved as though the point of the discussion activity was to finish all four tasks as quickly as possible, and finished in nine minutes. However, research at LARC (Language Acquisition Research Center, Sydney University) had shown a minimum of fifteen minutes of speech production was needed to obtain sufficient speech to determine syntactic stages of acquisition.

The interviewer was instructed to use the final task as a means to ensure fifteen minutes of speech. Usually this meant to keep the S talking by asking questions and not guessing the identity of the festival in the final picture until fifteen minutes had elapsed and they heard the knock at the door.

The last two tasks were conducted as quasi-barrier tasks, with neither task was mounted on a structure which visibly shielded the task from each participant. Instead, both parties held their information gap activity in their hand while they completed Task III, so that neither partner could see the paper held by the other participant. The last task had only one object, the picture of a Japanese festival. The students were instructed to choose one picture and describe it so that the interviewer, who had not seen the picture, could guess its identity. Students were told explicitly not to show it to their partner.

2.3.3 Video and Audio-Taped Interaction All fifteen minute discussion activities were video-taped, audio-taped, and filmed on an 8 mm video cassette. Although video-taping and filming students can be very intimidating, this particular setting was especially relaxed. The room in which the experiment was conducted was the daily classroom with a large, red Persian style rug and a table with tea, coffee, and cocoa. Biscuits and mandarin oranges were usually available for students during the class.

In addition, the textbook for the Speaking module was a personal videotape.

All students were required to purchase a videotape and bring it to each class. During the class, each student was videotaped practicing a speech act, such as 'inviting' or ID'declining an invitation'. Homework assignments were also completed on the O videotape in the classroom after class periods had finished.

Thus, the students were familiar with operating the video equipment and 8 mm a camera in the classroom, since only students and not the teacher(s) operated the d equipment. The discussion activities were added as a homework option ten days after Sthe first term in Experiment 1 and five days for Experiment II. So, the students had 00 been using the room and its equipment and were familiar with the procedures which were exactly the same as the classroom activities and homework assignments. The students were thus relaxed and comfortable during the sessions and frequently N laughed. Conversation analysts have shown that a comfortable, familiar environment can overcome the influence of taped interaction (Have, 1999).

C 2.3.4 Transcription of Data IDBecause I ultimately do an analysis with a psycholinguistic emphasis on data to be Sacquired by the learner, 1 use a transcript format which makes the future analyses Smost transparent. In particular, I have attempted to maintain the integrity of the actual speech produced by the learner, rather than analysing only the clauses, etc. as usually done in both theoretical and empirical analyses of the acquisition of morphosyntax.

This is a step further along the lines proposed by (Sato, 1990)for looking at propositions and paratactic speech described by (Givon, 1979).

I have also attempted to include all of the bits and pieces of the speech which can not be considered propositions, including syllables, such as hmm and other nonverbal cues to check for understanding, suggested by conversation analysts (Gardner, 1987). In order to study mutual knowledge between the interlocutors it is important to determine if the hearer understood the proposition. The use of video as well as audio-taping has aided in checking Japanese features of headnodding, for example. In certain instances, such as the use of so and and, I have analysed the speech with spectrographic analyses for the wave patterns of 'declination reset', to determine if these are hesitation markers or clause boundaries (Couper-Kuhlen Seltig, 1996). These analyses were made on a Sun computer using NWAV, a unixbased software program.

Finally, all of the interactions were transcribed by at least two transcribers. If differences of opinion occurred, a third transcriber also assisted. Native speakers of Japanese also helped with certain transcriptions. All of the transcriptions were then entered into an SGI computer again using a Unix software program. This enabled accessing all 12 hours of data for the presence of certain structures using grep, a tool whose acronym means "search globally for lines matching the regular expression re, and print them. This was especially useful in constructing the examples for Chapters 3, 4, and 5. Later, a grep for windows PC was also used to access certain structures on a smaller set of data.

2.4 Miscellaneous Issues 2.4.1 Effect of Practice on Students Speech Production and Interaction The discussion activities reported in this thesis are actually part of a larger study, which included four additional American interviewers. Therefore, students were given an opportunity to have eight fifteen minute discussion activities. One must therefore address the question of the students' speech production falling victim to the "practice effect". This is a serious issue for negotiation studies, which research the ND learner strategies and their patterns of interaction for acquiring information. These O studies have shown that the strategies of negotiation differ in production styles and amount of speech per partner (Gass Varonis, 1986). Researchers do not want the students to have prior familiarity with a task, as they study learners' adjustments to Stheir communicative strategies according to the sociolinguistic variables of age, sex and ethnicity of their partner. Familiarity with the task as a determinant of syntactic 00 development has not yet been directly studied and shown to be a serious problem for psycholinguistic research questions focusing on language development. In fact, there may be evidence that during planned tasks learners may produce more complex N language and a wider variety of linguistic constructions more relative clauses, noun modifiers) (Ochs, 1979). (Preliminary results from this study suggest that with c, greater familiarity with a task, the same result may occur.) Nonetheless, the practice effect which occurs when the same test is given repeatedly in a study (Odlin, 1989:35) could influence the results in a study such as Othis. Therefore, efforts to control and reduce the practice effect were made. First, the tasks themselves were designed to be challenging. No previous courses in the curriculum at the university, including the class in which the students were currently enrolled, had given the students an opportunity to interact in conversation using dialogues and simple questions and answers; the linguistic structures of question formation had not been taught or practised in the classroom. In addition, although the structural organization for each fifteen minute discussion was the same, and included the same tasks, the discussions varied internally with respect to content and elicitation techniques. The first task, which included questions to guide the interviewer in the greeting and introduction, changed from closed questions, such as "Where are you from?" to more open-ended questions such as, "Tell me about yourself'. As the discussion activities progressed, the number of personal questions in the introduction increased from eight to twelve. The content of the introductory discussions in both Task I and Task II changed with each interviewer, who had different interests, motivations, and styles for agreeing to take part in the experiment. Interviewers were given guidelines to assist in conducting the tasks and discussion, but were told they were free to interact with each S according to their interests and the needs of the conversation. Some interviewers, for example, were more interested in the reasons students were attending an American university; others were more interested in the family background of the students. Some interviewers actively engaged in the discussions by asking questions to prompt and guide the students, while others left the discussions and descriptions entirely to the students forcing them to talk with little or no interactive guidance. These interviewers would use non-verbal gestures, such as nodding, or particles such as "Uh huh" to continue the conversation.

The third task also changed with each discussion activity. No information gap activity was identical. More blanks or required questions were added to the information gap activity as in Task I. Neither did the blanks appear in the same position in each chart. The position of the blanks was randomized so that information was needed in the first blank in the leftmost column for one chart, but in the last blank in the first column in the next chart. In addition, since all of the festivals on each chart were different, each included a different set of facts. The vocabulary required in the content of the information charts and discussion was not part of the curriculum, lesson plans, or text, so students were not necessarily familiar with the items on the charts. The respective labels referring to each blank also contained both familiar and unfamiliar vocabulary items, and these varied with each chart. For example, students N were familiar with the labels "age" and "meaning", but not necessarily "participants".

O The set of eight pictures in the fourth task, picture description, also created a diversity Sof lexical content required to adequately complete the description Moreover, since students chose a picture from a file folder, Task IV was randomized Sand differed across discussions.

Finally, the diversity in ethnicity and status of the interviewers created an 0unexpected diversion. Students began to focus on the identity of the new discussion partner. This interest on the part of the students created an unanticipated distracter from the purpose and content of the elicitation procedures and materials. The students ,1 were very excited and curious as to the identity of their next discussion partner. The experimenter made every attempt to occupy the students with other unrelated conversation as they waited for their turn. The experimenter also actively prevented rstudents from talking to other students who had not yet completed the discussion activity. Although the students thought this was due to the importance of keeping the 0 interviewer's identity secret, the real intent was to keep the level of pre-planning for each discussion activity at an equal level in order to maintain comparability across students.

2.4.2 Tasks with Respect to Definiteness and Shared Cultural Knowledge I have referred to shared cultural knowledge in several places already in this chapter.

This notion is especially important in the study of definiteness and topic. In Chapter 3, I discuss in more detail the notion of shared versus mutual knowledge with respect to definiteness. Shared knowledge refers to the background information described in §2.1 above; mutual knowledge refers to the discourse knowledge of each interviewer at the exact moment of interaction between the dyad partners. This includes the conversational interactions with and without the visual tasks. For this interpretation I appeal to Clark Marshall (1981). For a definition of 'real world knowledge' and its connection to language with specific cultural modifications, I elaborate on descriptions of Chafe (1976) and Anderson Keenan (1985). In Chapter 8, I refer to "situations", similar to 'situation semantics' described by Barwise Perry (1983) and the connection to real world experiences to linguistic constructions.

2.4.3 Tasks with Respect to Topic and Subject This study differs from other studies of interlanguage topic Huebner 1983; Fuller Gundel, 1987) because narrative was not selected as the means for elicitation (see the literature review of topic in Chapter The structure of the four tasks described above contains several key characteristics for the study of topic in interlanguage conversation. The first task is an informal greeting and "get to know you" conversation. The interviewers, with a set of guidelines for initiating and directing the conversation flow, tended to control the conversation in Task 1. With few exceptions, the learners only answered questions and refrained from asking for information from their partner. Hence, most of their utterances contained "new information" as answers to the questions. The next task was designed to elicit descriptions (of their hometown festivals); the learners, therefore, were expected to contribute both old and new information in the conversations, as they gradually developed the description of the festival. The third task was designed to test the ability of the learners to ask questions, testing the students' use of word order in the formation of questions. However, it also tested the students' ability to introduce new D information into the discussion. In this task, students performed the same role as the interviewer in Task 1, as they asked questions for information. The final task was again a description of a festival, forcing students back into a position of description, such as in Task 2, but this time a visual prompt was included. Students were expected to offer new topics in the information structure of their descriptions, thereby altering the cognitive load of Task 2, since they could refer to a visual object 0 Task 4 should, therefore, have added to the data from Task 2 for syntactic structures expected from a description activity.

The use of festivals also added another dimension to the shared knowledge of N the discussion activities which has not hitherto been studied in second language, i.e., use of shared community knowledge and the description of Proper Nouns. A quick review of the information required, especially in Task III, makes it clear that proper ri nouns are a necessary and integral part of the linguistic categories in this information Sgap activity Moreover, proper nouns only have meaning when shared Obetween members of the same community, which Task IV tested.

Summary of the Experiment and Preliminaries to the Study The four tasks were initially conceived as a standard use of instruments for eliciting morphosyntactic interlanguage data described in §2.2.3 above. However, this changed as I began to see the use of the pictures in Task IV being used as antecedents for the learners' descriptions. As I studied their use, it became clear that the constraints on definiteness were very similar to anaphoric utterances in normal spoken discourse. I initially labelled this "visual definiteness". However, as I continued to study the use of the pictures in the experiment, it became obvious that the interviewers were also forming pictures in their minds. In order to distinguish this visual object in their minds from the learners' definiteness when describing the visible object of a picture, I added a new type of definiteness. I relabelled the learners' definiteness "visible definiteness" reserving the label "visual definiteness" for the mental representation developing in the interviewers' minds.

The fourth task was critical in shaping the direction of the thesis for yet another reason. One learner was describing objects that didn't exist in one of the pictures (fruit). In the next week, the same learner started describing people using the definite pronoun they before introducing the group the plural pronoun referred to.

This led to a study of the use of the tasks as objects of inferencing. In addition, it added a new dimension to the study of anaphora, since it became possible to study the use of anaphors linked to real world objects, i.e. the pictures. Both of these issues are addressed in Chapter 8.

Finally, as I described the patterns of specificity in the discourse generated by Task IV for one learner, I suddenly saw patterns developing in his descriptions. Since they changed rapidly in a three week period, I was able to study the components of these patterns in his discourse. Simultaneously, I saw different patterns being created for the interviewer, who was trying to guess the name, or identity, of the festival.

When these patterns were tested in Task III, it became apparent that the type of patterns differed (Chapter It was at this point, the study became one of "patterns of meaning" based on "patterns of specificity" and the pervasiveness of these three patterns throughout language.

Chapter 3 Definiteness 03.0. Introduction Particularly in conversation, speakers have responsibilities other than speaking semantically and syntactically appropriate sentences in order to communicate.

N Among other things, they must also keep track of the knowledge which the hearers should have in mind during the conversation. This includes information Ni, identifying the speakers, the spatial and temporal location of the speech situation, ri and keeping track of the topics under discussion. The judgments by the speaker about the hearers' current mental states and the resulting manifestations in the O propositional structure are described in a component of grammar called information structure.

(Generally, information structure is concerned with knowledge relating something new to something that can already be taken for granted, or presupposed, in the speech situation. The intention in this chapter is to examine the pragmatic behavior of definiteness and referentiality in Keenan's subject list.

Such an approach allows clarification of how subjects differ from, as well as interact with, topics in interlanguage (and natural language as well). Chapter 8 examines information in a situation and context for understanding meaning.

Many studies of the second language acquisition of information structure have proceeded according to referential characteristics, such as topic continuity (Givon, 1979a, 1979b, 1983). In these studies, the number of clauses in which a given pronoun holds for a given referent is counted. This is thought to lead to a psycholinguistic understanding of mental representations and memory load. My primary concern, however, is the organization and behavior of the relations and characteristics of topics in information structure within current and prior utterances and their relationship to grammatical relations, especially subjects.

Like the subject-object distinction, this means that within a clause more than just one entity must be examined to determine the information structure properties. In order to study this behavior, I follow the approach of Clark and Marshall (1981) (henceforth which builds on the psycholinguistic notions of mutual knowledge and behavioral constraints, such as attention, rather than Li and Thompson's (1976) notion of topic based on syntactic processes (discussed in Ch. In addition, I further refine and explicate Chafe's (1976) vague notion 'world knowledge' to achieve an operational notion of the pragmatic behavior which relates the world situation to language behavior. Critical to my analysis is the behavior of entities in the context of mutual knowledge at the exact moment an utterance is spoken, henceforth referred to as 'coding time'.

For the analysis of this pragmatic context, a clear notion of shared versus mutual knowledge is needed. These two notions have been used interchangeably in the literature and are critical for explaining the behavior of definiteness. I adapt C&M's classification of types of 'mutual knowledge', as opposed to 'shared knowledge', and their correlated linguistic manifestations. C&M's ISO categorization is more useful for my study because of its definitions of definiteness correlated with other psycholinguistic behaviors, such as attention, understandability, and associativity (inferencing). These in turn can be a operationalized in the analysis of linguistic utterances found in my data. These Spsycholinguistic notions can be translated easily into already established linguistic behaviors for deixis and anaphora, as well as memory requirements for 00 language learning. Furthermore, C&M's incorporation and treatment of proper nouns, such as those found in my corpus, e.g. the name of Japanese festivals, and presence of visible targets, a core design feature of the tasks in the experiment, Sclarifies the linguistic behaviors required for my study.

Since the classification requires explanation of the behavior of targets and N tasks in the respective contexts, this expands our understanding of the interaction N with objects and tasks used in eliciting language in interlanguage research. These linguistic behaviors are suggested as reflecting concomitant developing processes in memory and language learning.

SThe chapter is organized according to the four principal bases for establishing mutual knowledge: 1. Community Membership 2. Visual Linguistic Co-Presence 3. Auditory Linguistic Co-Presence 4. Indirect (Inferred) Co-Presence The chapter concludes with a summary of the mental resources listed by C&M in their discussion about definiteness. Chapter 4 extends this with a tentative proposal for the study of meaning independent of morphosyntactic stages of other models of second language acquisition such as the Multi- Dimensional Model (Meisel, 1981) or Processability Theory (Pienemann, 1998).

3.1 Sources and Types of Mutual Knowledge When Li and Thompson argued for the importance of the notion of topic versus subject for describing grammatical relations, the first characteristic of topics which they listed was 'definiteness'. Considering definiteness as "one of the primary characteristic of topics", they noted that "a subject, on the other hand, need not be definite" 461). The definition they used for definiteness was from Chafe (1976:39): "A definite noun phrase is one for which I think you already know and can identify the particular referent I have in mind." In the original, Chafe follows this statement by saying: "If I think you can, I will give this item the status of definite" 39). Chafe's notion of definiteness does not appeal to syntactic or lexical forms. Instead, he explains the assumption in this case is not just "I assume you already know this referent", but also "I assume you can pick out, from all the referents that might be categorized in this way, the one I have in mind" 39). To clarify his meaning, he gives six examples of considerations which could "lead a speaker to assume that his Hearer is able to identify what referent he is talking about": uniqueness, context, confines of a particular social group family), prior mention in the discourse, modifiers of the head noun, and entailment of a particular member of a previously mentioned set, e.g.,a 'kitchen' is a member of the properties of a 'house' (pp. 39- Although I appeal to all six in my thesis, my analyses are primarily O concerned with prior mention and contextual entailment.

The iconicity of his examples is not, however, described explicitly so that (,i it can be operationalized as a definition of definiteness in the study of language.

Prince (1981) objected to a view of the speaker as "omniscient" being able to V"assume" what was in the mind of the hearer, choosing instead to construct a 00 taxonomy of linguistic structures for topics based on an empirical study. This taxonomy concentrated on the syntactic manifestations of shared knowledge, categorizing the syntactic structures into three main types of "assumed familiarity": new, inferable and evoked, which are then broken down into two more levels.

Lambrecht (1994) developed Prince's taxonomy further into the mental Nrepresentations of the syntactic types and topics. Borrowing Chafe's suggested term of "identifiable" (Chafe, 1976:39) rather than definite, Lambrecht Shighlighted the need for the hearer to be able to pick out or identify the one referent from a presumably unlimited set of objects in the communicative situation the speaker has in mind.

C&M (1981) take a different approach to the study of the concepts of mental representations, 'mutual knowledge' and definiteness. In contrast to studies of syntactic topic, they address the question of world knowledge and definite reference directly, by engaging the questions "what kind of 'shared knowledge' do [people] use, and how." Their study surveys previous proposals for establishing definite reference in works by philosophers, semanticists, psychologists, and linguists (Hawkins, 1978, 1980; Schiffer, 1972; Karttunen, 1975, 1976). They note that the sources for definiteness have many labels which are often used interchangeably within the same work by the same author: shared knowledge, common ground, contextual domain, conversational context, common set of presumptions, shared sets, pragmatic presuppositions, normal and mutual beliefs (C&M 1981:21).

C&M next carefully distinguish 'shared knowledge' from 'mutual knowledge'. 'Shared knowledge' is knowledge which people in a community share, such as the identity of the current President of the country, or current political issues. 'Mutual knowledge', on the other hand, refers specifically to the knowledge mutually known between two interlocutors at coding time.

Revising earlier definitions of Shiffer (1972), Harman, (1977), and Cohen, (1978) (cited in they further define mutual knowledge as: A knows that A and B mutually know that p. It is incumbent for both to mutually determine that they both know p. They argue that it is not required that A be omniscient, but it is also not sufficient that two people know the same facts; it is necessary for efficient communication that both know that the other interlocutor also knows the same fact. Furthermore, both interlocutors must know that the other knows that they know this information. For example, I may be a student at The Australian National University (henceforth ANU) and my interlocutor may also be a student at the ANU, but if we meet for the first time in the Canberra Center, we can both talk about the ANU without knowing that we are both students at the ANU. Alternatively, I may know that my interlocutor is also a student at the ANU, but my interlocutor may not know that I also am a student at O the ANU, or perhaps my interlocutor does not know that I know that he is a student at the ANU.

SUntil both are aware of all of these facts, mutual knowledge is not a complete. Conversation either helps make it complete, or confusion results when Sthe knowledge referred to is not mutually known by both. The conversations in my study had the advantage of establishing group identity from the beginning: 00 both interlocutors knew that they were members of the MSU-A school community in Akita. This formed the beginning of the interviews with other mutual knowledge being established to continue the conversations.

r Since C&M are interested in how mutual knowledge can be "secured in the making of direct definite reference", their classification is constructed to show Sits ground-"its sources in a person's experience" or 'the common ground', a N term introduced by Kartunnen and Peters (1975:35). Direct definite reference is then divided into four types of mutual knowledge and its associated linguistic manifestations: Mutual Knowledge Linguistic Manifestations Community membership Community membership: Proper Nouns; Universal knowledge: generics; Physical co-presence' 4 Deictics Linguistic co-presence Anaphora, demonstratives Indirect co-presence Mixture of above This classification makes it possible to treat mutual knowledge as a specifically defined linguistic entity. In addition, although it is not their stated intention, it also confines mutual knowledge to sources of definiteness and propositional knowledge at the sentence level, necessary for the study of topics and information structure. Thus, mutual knowledge can be operationalized in order to establish its behavior in identifying various constraints on definite reference within four explicit contexts. This more tightly defined notion of types of mutual knowledge at the sentence level and explicit definition of constraints required for each type of knowledge thus provides a more empirically usable model for operationalizing mutual knowledge in a set of data such as mine.

These contexts can be further subdivided into more linguistically precise sources for types of topics. For example, whereas deixis usually refers to extralinguistic/situational knowledge, I treat deictics as linguistic utterances, based on input from the sense of sight, as found in my data. What is traditionally labelled situational knowledge, I have subdivided into visuo, visuo-spatial, spatiotemporal, and temporally conditioned constraints on linguistic behavior.

Moreover, I consider anaphora to be conditioned by the sense of hearing. Since it is subject to acoustic (analogue) linear processing, it is necessarily subject to different constraints from the visual situation. Thus, I have classified C&M's linguistic co-presence/anaphora as based on one of the five senses as other sensori-motor sources of linguistic utterance. By subdividing the sources of definite reference, e.g. topics, into distinct contexts, the constraints between traditionally labelled situational and linguistic reference can be demonstrated to be similar, yet unique.

Thus, elements in all traditionally labelled deictic utterances and anaphora O can be considered linguistic topics, with similar behavior and constraints. This eliminates the overlap and blurring of the types as described in earlier approaches (cf. Himmelmann, 1996). Finally, I treat C&M's indirect co-presence as a source d for types of inferables (cf. Prince, 1981; Lambrecht, 1994), general knowledge of Sgeneric elements (Sidner, 1983), implicit inferences (Johnson-Laird, 1983), 00 mental images (Posner, 1990), and partitives (Eng, 1991). The charts below compare C&M's classification of the four types of mutual knowledge and reference types with my revisions.

C&M (1981) Fellbaum (1999) CK Basis for MutualKnowledge Reference Types Basis for Mutual Knowledge Reference Types C, 1. Community membership Universal, C Particular 1. Community membership -Universal, N 2. Physical co-presence Deixis Particular 3. Linguistic co-presence Anaphora, 2. Visible co-presence -Deixis: Visible, deixis visuo-spatial; 4Indirect co-presence 3. Acoustic co-presence -Deixis:Spatio-, Physical Temporal; Linguistic Anaphora 4. Indirect Co-presence -Inferencing Visible Properties evoked from visual objects Acoustic Properties evoked from Acoustic objects Mental schemata; nprcifiritv For a full understanding of the meaning of these tables, it is important to distinguish the basis for the knowledge and a referent of that knowledge. In so doing, it is useful to return to the famous distinction made by Saussure (1959) between "sign [signe] to designate the whole", i.e. concept and sound image, and signified [signifie] and signifier [signifiant], the concept and sound image, respectively (Saussure, 1959:67). The linguistic entity exists only through associating the concept with the linear sound, or word. When the signified is separated from the signifier, concepts, such as events (festivals), people, or snow, belong to psychology according to Saussure.

These concepts have properties. Referring expressions in language refer back to the sign, the whole concept or individual properties of the signified and/or the signifier, once the existence of the sign has been introduced into a discourse. Referring expressions have various forms and refer to signs from several sources. The classifications above have divided the sources for these signified (concepts) into four possible sources of knowledge of the signs in a situation between interlocutors. Knowledge from any one of these four sources becomes definitely known between interlocutors once it has been articulated in a discourse.

If, in addition, we refer to external objects in the speech situation as visual, acoustic, or one of the other three senses, then words can be characterized as acoustic targets which can be referred to as well as visual targets. The target is IDthe source of the referent; it is the antecedent of the speech situation. Both O participants attend to all the targets in the speech situation, visible or aural. These targets are derived from the sources of mutual knowledge in the chart above.

SI continue for the most part analysing the data per speaker and hearer.

SHowever, this is actually a misrepresentation of the speech situation which has prevented total clarity of our understanding of the relevant issues of the concept 00 of definiteness. When a speaker utters a sentence to a hearer, the speaker's words are directed to the hearer and the words become the basis for the appropriate response. However, the words also enter the speaker's brain. The speaker is also simultaneously tracking information from the speech situation. Thus, we can create a model of the entire speech event in one person's brain, and don't need to describe the speech interaction as a two-person event. I return to this point below and in Chapter 8.

IND Mne Is e Speaker /Hearer Hearer Finally, C&M describe the behavior of antecedents of anaphors at the coding time of utterances, which has long been important for describing the behavior of deixis (Fillmore, 1966; Clark, 1974). The temporal constraints for defining all of the behaviors necessary for referring expressions, and hence topics, become: Coding Time for Referring Expressions Immediate Prior Potential 'Immediate' coding time can thus be argued to contain topics which are both old and new within a context of mutual knowledge, e.g. visual objects, proper nouns, or inferrables. 'Prior' to coding time with respect to the event referred in the utterance is traditionally labelled "previously mentioned" in studies of anaphora.

'Potential' coding time, sometimes called 'backwards anaphor', contains anaphors which precede their antecedents within a context of mutual knowledge.

All mental representations of topichood can thus be constrained by a given set of sources of mutual knowledge together with the additional variable of time.

In the study of second language acquisition of definite reference, identifying the sources of mutual knowledge allows us to more precisely pinpoint ND the differences between speakers. This is especially important when languages O differ greatly in the forms that are used to express direct definite reference. In addition, the particular elicitation task (device) used may limit the occurrence of the types of utterances and the types of mutual knowledge and their respective linguistic referential forms. Simply looking for forms such as definite articles is not sufficient; we need to understand the underlying sources for establishing 00 definiteness. The more complete our understanding the better our future attempts to study the patterns of behavior that are important for understanding their acquisition.

3.1.1 Community Comembership c- This type of mutual knowledge is known by both interlocutors from membership in a social group, small, e.g. university, city, country, or large entities, such as a continent or the planet earth. Once the learners became aware of this type of Oinformation they would most likely retain it for long periods of time. Each participant in the dyads can safely assume that the participants from the same community, e.g. Japan, MSU-A, know this same information. C&M refer to this assumption as universality of knowledge (within the community). Mutual knowledge within a community can be divided between generic and particular knowledge. Both generic and particular knowledge are knowledge about objects states, events, and processes, but generic is knowledge about kinds of things and particular knowledge is about individual, unique, or particular things (see Carlson, 1982; Kadmon, 1992). Both of these types of knowledge can be shared between interlocutors, but their sources differ and they are expressed linguistically in different ways.

3.1.1.1. Generics The first kind of community knowledge is known as generic knowledge, and these nouns and their respective properties are held in long term memory.

Generic knowledge can be a source of inferrables, such as inferring the member of a set from the set itself. For example, a university is made up of a group of students and teachers; and a festival is an event with people, (participants), specific dates, places, and purposes for celebrating. In sentence processing, the speaker may give the generic set, and then later in the utterance refer to a property which the hearer must infer belongs to the set. Looking at a student holding a picture and trying to guess its contents is another type of inferencing; a listener can infer properties both while listening and while forming a picture in the mind, especially if trying to guess the name of a particular festival. Finally, we may identify a group of friends and then refer to one of my friends, a specific member of the group. Although Prince (1981) and Lambrecht (1994) recognize this type of topic, referring to it as evoked or inferred, respectively, none of the researchers mentioned so far has further subdivided the types of knowledge inferred from generic and/or particular knowledge. Nor have they tried to establish the different types of constraints on the linguistic behavior of inferrables.

IO 3.1.1.2. Proper Nouns.

The linguistic representation of mutually known particular knowledge in a community is by the use of proper names, or rigid designators, "because it designates the same thing in all possible worlds" (Kripke, 1972, 1977), also cited in C&M, p. 26). In other words, proper names are "used to refer to the same object on different occasions" (Searle, 1971:137)." Like common nouns which 00 form the basis for generic networks, proper nouns are also held in long-term memory but in different areas neuroanatomically to common nouns and their associated semantic properties.

Proper nouns are unique, identifying culturally significant entities, people or places, so they need few modifiers to further identify them. When speakers use proper nouns, they are fairly certain that these items are known to members in Sthe community and don't change from one conversation to the next. The O universally known things in a community which get proper names are: people, Splaces, and prominent events (like famous festivals). However, the particular traits of proper nouns are only known by members of the specific community from which they obtain their identification. Not only do proper nouns refer to a unique entity, but they are highly specific, since only one such entity necessarily can exist (see Searle, 1971; Kripke, 1972, 1977; Kadmon, 1990). 6 Like generics, they need no further identifying linguistic information for a hearer. Therefore, proper nouns can be distinguished grammatically in some respect in almost all languages. In English, for example, proper names are not usually preceded by other specific determiners (articles, modifiers, possessors), and don't usually occur with relative clauses, or other further identifying devices. In this respect, proper nouns behave like determiners, since they usually cannot be preceded by another determiner, in order to make an entity more specific.' 7 They are obeying constraints in the language which are defined exclusively on their pragmatic status as known entities in the discourse due to the mutual knowledge shared among members of a given community.

The proper names in my experiment were constrained by the names of other teachers or students at MSU-A, or famous characters in the festivals, e.g.

Namahage, the Devil which came across the Sea of Japan to Oga Peninsula to scare children described picture in Tasks III and IV. Since the participants in the discussion dyads which I analysed were all native speakers of Japanese and born in Japan, they should have all had equal knowledge about these national holidays for certain national festivals and certainly the larger cities of Japan. They were all living in Akita prefecture and all students in the ESL program at the same university, MSU-A, so they all shared knowledge about local villages and the Akita festivals which were currently advertised on the walls of the university.

If we define topics as "the topic referent [which] must be taken for granted at the time of the utterance", proper nouns are behaving as topics in all contexts, because they refer to entities which are mutually known between interlocutors. This approach of automatically including proper nouns in the definition of topics differs from other well-known theories of information structure. Following Chafe (1976), Lambrecht (1987) considers given referents to be those which the speaker assumes are "present in the consciousness of the Hearer at the time of the utterance", and recoverable referents as not yet given, IDbut which the speaker assumes the hearer can "recover from the discourse context." For Lambrecht, referents in an utterance are recoverable because they Sare: previously mentioned in the discourse; S(2) inferable from other previously mentioned items; 00 saliently present in or inferable from the extra-linguistic context.(p. 121).

Since he does not include proper names in any of the three sources above, proper names become "one class of exceptions to the general rule which associates Scoding with low topicality and pragmatic backgrounding" (Lambrecht 1987: l* ~249). Instead, in Lambrecht's theory, proper nouns are considered to be "lexical phrases", a strictly syntactic category, and lexical phrases are low with respect to K subjects and topicality. For Lambrecht, high topicality refers to: degrees of IDsalience of referents in a discourse based on specificity; topic in more than one clause; found in main clauses rather than subordinate clauses; or (4) N involved in some action or process rather than in a state (pp.236-237). Thus, Lambrecht uses syntactic categories which are "lexico-grammatically evoked," for identifying salience of reference/high topicality. This is not the same as defining topicality based on sources of mutual knowledge, as I am doing in this chapter.

Nonetheless, he explains that proper nouns are pragmatically more easily accessible than other referents, similar to the deictic relations which hold between speaker/hearer and the first and second person pronouns in a discourse. Proper names have a high recoverability because they are "physically present as prominent participants in the speech situation or because they stand in a particularly close relationship with the speaker/hearer and are in a sense "mentally present" (p.249). Similarly, members of a family or close friends are referred to with first names or name-like expressions (like "mommy" or "daddy") 249.) These expressions function analogously to deictic or anaphoric pronouns because they "point to" or identify their referents directly in certain contexts, rather than appealing to a category from which the listener has to infer the referent. Thus, in his model, these expressions have a topicality, which is higher than for other lexical subjects.

Likewise, proper nouns may be first mentioned or brand new, or unused, in Prince's taxonomy, but their familiarity between the interlocutors due to shared knowledge within the community nonetheless makes them topics or given information. Evidence for this can be found linguistically in the absence of linguistic structures necessary to make these definite. They have already been identified. Hence, the linguistic context does not require additional information to make them specific or identifiable.

In C&M's model, proper nouns are part of community knowledge, and proper nouns, an exception for Lambrecht's theory of topicality, become predictable as a type of extra-linguistic information not previously mentioned. In C&M's model, proper names fit naturally into a system of language which looks at their presence in a larger sphere of mutual knowledge rather than just the immediate discourse, which characterizes Lambrecht's above definition of topic expressions. According to C&M's classification for mutual knowledge and its IDsource for direct definite reference, proper nouns, do have pragmatic backgrounding as a source for topics. This explains Lambrecht's finding in his Sdata that proper nouns are associated with S-coding and low topicality. (,i a Moreover, in Lambrecht's theory this fact is an exception, whereas in the model Spresented by C&M, the behavior of proper nouns is predictable and behaves systematically and consistently with all other topics (definite reference). Thus, 00 by examining and classifying the sources for mutual knowledge for specificity and definite reference, C&M can align the above three sources as well as proper nouns into one unified model.' 8 In my data there were nine types of Proper Nouns: I) The names of the respective interlocutors or friends of each (N/A due to confidentiality) t' 2) The names of persons unknown to either one or both interlocutor (N/A due to IN confidentiality) 3) The names of the home towns and cities known to the interlocutors Akita, rKobe, Kyoto) 4) The names of cities unknown to the interlocutor(s) Cottonwood, Minnesota) The names of the respective festivals known to the interlocutors Kanto) 6) The names of the festivals unknown to the interlocutors (Amekkoichi Candy Market) 7) The names of the new ESL module, in which all of the learners were students (Focal Skills) 8) The names of months and seasons' 9 (August, Spring) 9) Festivals depicted in pictures of the of the Tohoku area (Kamakura, Kanto) Proper nouns, like all topics, were found in various positions in a sentence in my data. Since my experiment elicited proper nouns in several different contexts, I treat each occurrence in the context of the type of co-presence in which they occurred. So, some proper nouns were uttered in visual tasks, and they are analyzed under Visual Co-presence Other proper nouns were found in Indirect Co-presence during Task IV, in which the learner had a picture of a festival and the interviewer had to guess its name without seeing the picture.

Some were produced in the introductions and some in the purely oral descriptions of festivals in the interlocutors' hometowns, Tasks I and II respectively. These are analyzed under Auditory Co-presence, i.e. pure speech production without visual presence of a target to refer to in the linguistic environment. These occurrences are parallel to the types of utterances in both Lambrecht and Prince's corpora. They would also be the type of utterances found in the Pear Stories and other narratives discussed by Himmelmann (1996) in his study of anaphora.

There is one other source of proper noun that must be identified for my experiment visually induced referents of famous people or events, the ninth type listed above. My experiment had no photographs of mutually known people or pictures of famous persons in the community, but it did have pictures of famous festivals. Students knew the pictures of the famous Japanese snow festival, Kamakura, and Kanto, the Akita festival. So did the discussants, who were able to ask fairly detailed questions about the mental image of the festival about which the learner was describing. These mutually known entities of the INO community knowledge were visually-evoked, and auditorally, during oral conversation. As such, I address their linguistic behavior in the next section on Svisual co-presence.

a Therefore, it is not the morpho-syntactic properties of lexical items which Sare Proper Nouns that distinguishes them, but rather the role in the community which establishes proper nouns as important for defining their presence as salient 00 entities. The morphosyntactic coding only helps establish their uniqueness in language and the morphosyntactic differences in coding among the world's languages establish the need to carefully identify their role and function in the C grammar of a language. These similarities and differences are conceptually defined in P-structure.

Cq 3.2 Visible Linguistic Co-presence 20 cAs stated by C&M physical co-presence is "the strongest evidence for mutual Sknowledge that people are generally prepared to accept"(p. 38). Interlocutors are in physical contact with each other and viewing and talking about the same Sphysically present targets. Linguists call this type of interaction 'deixis'. In standard usage, deictic expressions (or deictics) refer to properties of the extralinguistic context of the utterance in which they occur. The extralinguistic information of utterances containing deictics makes reference to: Person (grammatical) the speaker (Sp) or the addressee (Adr) of the utterance; Spatial location (locative) a target relative to the spatial location of Sp or Adr; Time reference (temporal) the time of an event or state relative to coding time (Anderson and Keenan 1985:259) Due to the inclusion of visibility of the targets being discussed, there can be no question as to the sharing of knowledge about the existence of the targets.

In the sentence "No, don't show the pronoun is understood only if we know the identity of the speaker. Therefore, me is a deictic-"its referent is understood of necessity to be the person who utters or asserts the sentence in which it appears" (Anderson, 1985:259; Lambrecht, 1994:4). The pronoun you is a person deictic, referring to the Hearer; the demonstrative this is a spatial deictic, referring to a target, a picture held in the speaker's hands during Task 3. The adverb yesterday is a temporal deictic, referring to the time one day earlier than coding time.

2 The existence of the speakers and the visible tasks, in fact, is known from their visual presence in the context of their respective utterances. This context constrains the grammatical behavior of each utterance, whose context is further constrained by speaker hearer knowledge.

2 2 It is the context of visibility and the mutual knowledge, especially regarding the attending to that visibility of each target by the interlocutors, i.e. immediately, potentially, or prior viewing, which affects the grammatical behavior of the utterances. This context therefore makes visible knowledge both definite and thus also specific, discussed further in §3.4.

(The three different versions of visible co-presence and the time period with respect to coding time are: immediate visible co-presence when the two interlocutors are both actually Sfocusing on a visible object at the time of the utterance, i.e. coding time; rK potential visible co-presence when the hearer is not paying attention 2 3 to an object which is visible to the speaker (but easily in view of the hearer) at coding time; prior visible co-presence when the interlocutors have previously viewed an 00 object but have ceased before the time of the actual utterance, i.e. coding C time.

In my experiment, the two interlocutors were in each other's visible presence throughout the fifteen minute conversation. The tasks were in their immediate visible environment, although each of the four differed in design and c the manner in which they were performed between the interlocutors, as described below.

The final task had the learner viewing a picture of a festival which s/he attempted to describe to the interviewer. None of the contents of the target N viewed by the learner were visible to the interviewer, although the interviewer knew that the learner was viewing a picture of a festival. Therefore, the target was not visible to both interlocutors in Task IV. Unlike Task III, the learner could not guide the interviewer in his description. The learner could only look at the items visible in the picture and try to describe them to the Hearer. Thus, the speaker is attending to a visible target, whereas the listener is inferring information about the picture from the speaker. In the case of the learner, understanding the function and knowing the English word to identify the targets in the picture is critical. For the listener, being able to use the information from the speaker and being able to form a coherent mental picture of the particular festival in the picture is important. It is therefore necessary to include a concept of rational understanding as part of the constraints and behavior of linguistic utterances referring to visible targets in the situation.

A final key point in the definition of the interaction between visible targets and visible co-presence for the interlocutors revolves around the exact moment of the utterance and the spatiotemporal relation of the speakers to the visible targets in the environment. It is critical to study the precise moment of linguistic interaction with the visible target. The linguistic utterances which are constrained by the visible presence of targets in the environment are described most completely in studies of deixis.

3.2.1. Immediate Visible Linguistic Co-Presence As stated above this type of mutual knowledge occurs when the two interlocutors are both actually attending to a visible and identifiable target simultaneously at coding time.? 4 The visual target may be a person or other target in the spatial context referred to in the utterance. C&M only mention pronominal and adjectival demonstratives, which refer to entities visible to interlocutors, because they are arguing for direct definite reference of NPs.

However, this type of mutual knowledge, i.e. Visible-Linguistic Copresence, can also include locatives, or adverbs such as here or there. I refer to INDthese as spatial deictics, which obey similar constraints of mutual knowledge for O direct definite reference for identifiability by interlocutors at coding time.

(Ni 3.2.1.1 Visible Targets in Immediate Visible Context STargets which are in a non-relational configuration to other objects processed 0visually are coded linguistically with person, number, gender and (4) Ssocial rank.

3.2.1.1.1 Person Deixis (i The first type of immediate visible co-presence found in my data is person deixis, which can encode the sex, number of individuals, social status, and social and personal relationships of the referent in the speech context. As pointed out by Anderson and Keenan (1985:260): Ithe basic person deictics are expressions which necessarily refer to the speaker(s) or addressee(s) of the utterance in which they occur.

[italics mine] These utterances with first person deictics were uttered by the student and interviewer in the same interview: i2s6.txt:3 i2 uhuh <laugh> it's not.. I don't know this age i2s6.txt: I s6 I entered this school in last spring In other words, the deictic I functions to identify the speaker and it necessarily varies within the context of the interview at coding time. Their referential target is visually determined and the referential meaning of the pronoun varies from utterance to utterance throughout the interview.

This alternation of speaker and hearer in utterances also has universal implications for question-answer pairs. As Katz Postal (1967) point out, a condition must be placed on interchanges of first-person and second-person elements such that the actual referents of these "forms are contextually determined by the very use of the sentences containing them" at coding time (pp.

114-115). This distinguishes the contextual conditions from WH-questions.

Obviously the first and second person deictic can refer to either person in the interview, unlike Proper names which are unique to each person in the interview. In Japanese deictic expressions are used not only for anchoring the participants and the referent in terms of physical location but also for locating the referent in reference to the social grouping. In Japanese one must say "I am called Misako", when you meet for the first time to identify yourself. One can not say I am unless the referent is known through mutual knowledge.

25 Thus the meaning and use of person deictics in the two languages are determined by the relationship of the deictic to the speaker and the hearer's knowledge of the speaker, although the two differ in the language specific constraints on their behavior.

These deictics are used to specify a relation which holds between the participants at the precise moment and in the context the utterance is spoken. In other words, the use of person deictics is constrained by the necessity of IND simultaneity, attention, and rational behavior that two persons are engaged in during the visual situation of a conversation.

In this regard, they are topical because they are definite and behave like the proper names of the speaker and hearer respective to the speech context.

SHowever, both the first person and second person pronouns, singular and plural, do not fit the usual definition of a pronoun, a word which replaces a noun or noun 00 phrase. The referent is the speaker of the utterance, and the speaker's name is never given in the speaker's own utterance. It is an ungrammatical utterance due to constraints on the deictic behavior of the pronouns. These pronouns refer uniquely and necessarily to the Speaker(s) and Hearer(s). As such, Kaplan (1989:146) calls them "pseudo-pronouns", since they do not replace another Sperson, and never have an antecedent. Hence, the behavior of person deictics in the immediate speech situation is constrained by the mutual visible knowledge of the two interlocutors and the immediate context of their utterances.

The information expressed by these deictic pronouns is true irrespective 0 The information expressed by these deictic pronouns is true irrespective Sof their grammatical function (subject, direct object, etc.), possessive or vocative.

Cross linguistically, their function as 'pronominal' manifests itself in various forms, independent words, clitics, or inflectional affixes. Usually, however, they are relatively reduced from other nominals and often appear as one syllable, e.g.

English 'm for him or them 260). In my data of intermediate second language learners of English, they always appeared as full lexical forms.

Person deictics and their respective grammatical functions of subjects, direct objects, and indirect objects found in my data are: Example 3.1) i3-sl.txt:3 sl canyou tell me about age of this festival? ilO 0-sl 3.txt:4 i 10 okay, you can stop. Okay, maybe finally I give you a picture, so please describe the festival what you look at and then I guess the (xx) okay i3-s2 txt:l I s2 after this March, going to go back my job iI0-2s10.txt:4 ilO I will giveyou this one picture, couldyou explain this picture? Please ask me questions.

In all of these examples, the constraints of visible co-presence are maintained, irrespective of the grammatical function. Furthermore, the referent is known to the hearer, in spite of not being 'previously mentioned', a stipulation of most current theories of information structure. Each interlocutor can identify the referent by the respective visible presence and knowledge that the referent is a person. This knowledge is not evoked lexico-grammatically, but visually. The visual co-presence makes this a known entity between the interlocutors.

Traditionally, in studies of anaphora, mutual knowledge between interlocutors is confined to the auditory sense of speaking and hearing, and consequently referred to as given, requiring an antecedent. However, the person deictics visually evoked are also functioning as mutual knowledge between the interlocutors. The difficulty of coding an antecedent in this visual source of knowledge has lead to their being called extra-linguistic. They can not be tracked back to earlier IDlinguistic units. This inability to code person deictics easily in a formal linguistic O grammar has lead to generalizations which can be proven false as argued below.

SImmediate visible co-presence of person also violates proposed constraints on antecedents with respect to proper nouns and existential sentences.

SEng (1991), Milsark (19741), cited in En9, Barwise (1981) and Keenan (1987) all maintain, for differing reasons, that existential sentences either do not or with 00 marginal acceptance contain proper nouns and definite NPs. All of these analyses simply assert that this is a fact of existentials. However, it is perfectly acceptable to utter "There is John" or "There is John now" when he is in full view r'l of both interlocutors. In other words, evidence of his existence is given by his visible presence at the moment of coding time, and, therefore, is not required to have been already presupposed or mentioned in the discourse. Moreover, the (Ni statement "I am John" asserts both the existence and identity of John, and Salthough it may be both an existential and identificational, it is referred to syntactically as an equational (or identificational) sentence. However, if we Saccept the first person pronoun as visibly present to the hearer and therefore the visible situation is the source of the antecedent, this is both existential and identificational. Eng's attempt to accommodate existential sentences which do contain definite descriptions requires it to be explicitly stated "that the antecedent of the NP does not precede the definite description" by using the word following in the sentence "There are the following counterexamples to Streck's theory...." (Eng's Example 46, p. 14).6 More importantly, the morphosyntactic coding of English does not distinguish visible, or sense anaphora, from non-visible, termed 'surface anaphora' by Hankamer Sag (1976). The English lexical forms of pronouns are identical for both types of anaphoric pronouns. If we confine our analysis of topics and definite descriptions to a traditional analysis which excludes (as these authors seem to do) or separates sense anaphora as a separate process, we miss the generalizations regarding the constraints on their behavior which are possible from the non-distinctions of visibly known from auditorally known, or given mutual knowledge. Moreover, the Japanese language codes this situation quite differently. This is presented more fully in §3.2.1.1.3 together with gender.

3.2.1.1.2 Number First person deictics can also be found in the plural. These are examples found in Task IV as the interviewer hands the folder of festival pictures to the learner: Example 3.2) i 10-s12.txt:4 i 10 mm, okay, that's enough, okay, maybe we have a small sometime, this is Pick up please one pictures of here and As mentioned earlier, the pictures of festivals in my experiment also have people and targets which are plural in number. Hence, the use of the plural number marking as found in English is required, either on the verb or on the plural noun morpheme ending. In English number distinction with respect to visibly present targets is marked when functioning as subject of the verb. For language learners, this information is not part of the selectional restrictions found O on verbs which can be memorized and held in long-term memory. Instead the O plural morphemes and S-V agreement affixes must be learned but their use is determined at coding time by the real world situation. Plurality in these cases is e displayed in the pictures which the learners held in their hands during Task IV.

Example 3.3) Taeko Drum Festival 00 C, il6-sl l.txt:4 sl I eto. First there is a very big drum.

il6-sll.txt:4 il6 mmh.

il6-s Il.txt:4 sl 1 is Japanese drum. I think is a biggest in the world.

Sand some eto on the drum. eto four people hitting drum and bottom of drum three people eto three people Shitting the drum.

I In this description of the Taeko Drum Festival, sll identifies one drum and 0correctly uses the singular copula of existence. However, after the first utterance, Sthe learner only codes the numbers of people and their actions of hitting the drum; there is no evidence of S-V agreement due to a lack of subject in the next two phrases, which lack a finite verb, or auxiliary for 'hitting'. This contrasts with s13 describing the same picture also with i16: Example 3. 4) il6-sl3.txt:4 s13 uhum are playing drums but they're playing one big drum and one little guy (xx) a guy is standing in front of the drum and he is holding a kind of mm, sign pole mm, one person standing beside, beside the, the big drum and have, (xx) mm like frog, like sss (xx) (xx) mm, tch I think, uh, there are old people hitting drum, mm, in the frog Plural deictic pronouns have other implications within a grammar. For example, the notion of inclusive we and exclusive we complicates the use of the deictic verb come.

3.2.1.1.3 Gender The gender of the speaker is not coded on first and second person deictic pronouns in English.

2 However, in English, when a third person is present in the context of the conversation, or immediate speech environment, similar rules to syntactically controlled anaphor are found (Hankamer Sag, 1976). They provide these utterances, intended to illustrate their notion of "pragmatically controlled (or deictic) anaphora" 391): Example a. He's saying that your hair will fall out.

b. Her hands are trembling.

c. I hope it's a herbivore. (Hankamer and Sag 1976:391) IDThese utterances show that when referring to a person in view, the person is O referred to with a male female or neuter pronoun Since the referents Sof these utterances are visible to both interlocutors at the time of reference, the a pronoun with the appropriate gender can be determined.

28 STraditionally these have not been called anaphors, since they have no linguistic antecedent. Nonetheless, the visual environment acts as topic-creating 00 environment in which mutual knowledge of the referent is shared; the environmental context provides an antecedent for the referential deictic.

Japanese has a gender distinction in the more informal varieties of the Slanguage. The table below from Shibatani (1990) illustrates: Table 3.1. Gender distinctions in pronominal forms in Japanese C1 Formal )Informal ID I st person Male speaker watakusi watasi boku ore C Female speaker watasi atasi 2 nd person Male speaker anata kimi anta omae Female speaker anata anta 3 r d person kare 'he' kanozyo 'she' Thus, Japanese learners of English need to learn new behaviors not only regarding formality and register but the gender differences as well. Although my data do not show any problems with regard to learning the English pronoun system, my learners have already had a minimum of six years exposure to the grammatical structure, so I can not make any real claims as to evidence of learning. Intuitively, it would seem that the simplicity of the English gender system compared to the complexity of the Japanese system of gender distinctions, makes the learning system easier. However, this needs to be carefully tested to determine what the learning patterns might be.

3.2.1.1.4 Social Rank and Relationships of Participants The social relationship of the two interlocutors does not constrain the use of the first and second person deictics in English. English has no honorific pronominals, although titles are sometimes used to indicate respect, such as in the presence of an authority, e.g. Professor or Madam Chair.

Social rank is also expressed in English using the diminutive "Billy" for young boys, followed by Bill/Will, William, Mr. Clinton, and finally at the very top of the social ladder, only the surname Clinton. However, this pattern is not obligatory and nicknames, such as Bill, can continue through the entire hierarchy.

The use of a special affix to indicate respect as used in Japanese, did not surface in my data of intermediate level Japanese speakers learning English.

However, in one interview clear nonverbal behavior indicated respect a student IDhas for a higher authority, the Provost's wife, who was the interviewer. Also, an older male student uses the politest syntax of wouldyou to a student ten years his Sjunior: Example 3.6) V, i2s6.txt:3 s6 Uh..Would you tell me the meaning about this festival? 00 Unlike Japanese, English does not assign status in a discourse to age or social position in the use of deictics. This utterance is by the 28 year old Spoliceman to a 20 year old female student at the university: SExample 3.7) i2s6.txt: s6 Uh..Would you tell me the meaning about this festival? IOne might conclude that the policeman is using a polite question form in this utterance, but this would need more detailed experimentation. It is possible c that this student only knows this formula for asking questions. Or, the student may simply have a personal style which is more polite to everyone, even in Japanese. The only thing we can safely conclude is that the speaker uses a polite question form when addressing a person younger than he, unlike the custom in Japanese.

This utterance is by a JSL teacher (i 10) to a student: Example 3.8) il Osl 2.txt: 1 io10 about one year, okay, have been teaching here at MSU for four years, this is my fourth year at teaching at university level. Do you have, do you enjoy MSUA? i 14sl 0.txt:2 i14 Have you seen um this festival international festival? As seen, the use of the English person deictics is found in utterances spoken by both students and professors. Although the first language of these speakers requires honorific affixes when speaking to teachers, the English language has no such form. Alternatively, the students could have bowed or nodded their heads to indicate polite acknowledgment. This also did not occur either in these utterances.

Thus, the honorific affix and bowing are both visually determined, and are properties of the ethnosyntax of language only and these speakers have learned the difference between the two languages.

Japanese has a special infix which designates politeness to the hearer.

According to some analyses of Japanese, honorifics are one test for the existence of subject Shibatani (1977). In fact, when the subject is ellipted in Japanese, the honorific may be the only evidence of the existence of a subject in a sentence.

However, since topics can be ellipted in Japanese, this behavior is consistent with an analysis which claims that the presence of the interlocutors is real world knowledge and is therefore constrained by the deictics of the situation. The exact form of the infix is determined by the real-world knowledge of the speaker and the relationship to the Hearer at coding time. Arka (1998) has argued that IBalinese has lexical and syntactic manifestations for establishing and constraining language with respect to social ranking in discourse.

2 9 However, in both Japanese and Balinese, the social rank and required honorifics, are sometimes complicated by the semantic role of agent and patient.

In these cases argument structure seems to outrank the pragmatic constraints posited here. For more on this point, see Arka (1998) and Arka (2003) for 00 Balinese and Japanese, respectively.

To conclude this discussion regarding the language behavior determined by immediate visual mutual knowledge with respect to person, English and K ~Japanese have different constraints. The person system of English requires the use of first and second pronouns, but not proper nouns, when two people are in (each other's presence and engaged in a language task. In Japanese, on the other N hand, deictic expressions are used not only for anchoring the participants and the referent in terms of physical location but also for locating the referent in Oreference to the social grouping Shibatani (1990)." Additionally, the difference between the two languages with respect to morphological coding of each respective property is not a lexical constraint. The morphemes which are attached to parts of speech, i.e. nouns or verbs, are based on the visual knowledge known to the interlocutors. The source of this coding is not predicate-argument structure, but rather direct visual knowledge.

This knowledge is introduced into the discourse by the constraints of visible simultaneity, attention and rationality between the interlocutors. Both interlocutors must be looking at each other and attending to the situation simultaneously. They must also both be able to comprehend and infer the proper reference to each other in the situation. This requires the mental resources of visual attention and comprehension of cultural norms. This may seem trivial when discussing the use of person pronouns in a language like English, but it is nonetheless necessary for a complete description of the language.

It is especially significant to notice the similarities when addressing the processing of mutual knowledge and the implications for language learning. This is indeed a learning situation for second language learners. Each person must learn the proper linguistic forms for referring to oneself, as well as others copresent in the visible situation. In each language these forms and their use differ. Learners must develop not only the morphological properties but the mental attentiveness to the visual sources in order to produce grammatically acceptable utterances in the second language. These visual forms from language are constructed differently in different languages.

Thus, we can tentatively suggest that the mental constructs and pathways also need to be developed for learners since they can be expected to differ.

Moreover, these language constraints are subject to visual and rational understanding which exceeds the knowledge of nouns and verbs with respect to predicate argument structure.

The figure below summarizes the type of visual attention necessary to rationally communicate between two speakers using the type of constraints described in this section. All of these constraints are based on attending to real world knowledge regarding speaker identity, person, number, gender and social IDposition, while attending to singly directed visual input between participants in a proximal relationship to each other.

,1 Visuo-Linguistic Processing Needs of Stationary Target (Person) 00 00 Visuo-Linguistic Processing Needs of Plural Stationary Target (Persons)

I(N

O

0

\O

In the upper diagram, the need to attend to one target and code it linguistically while speaking is diagrammed. In the lower diagram the visual processing while attending to two (or more) targets while speaking is illustrated.

In each case, we can speculate that special visual attentional pathways view the targets while accessing the linguistic code and then produce the appropriate coding, i.e. social status ranking, singular or plural marking for nouns or subjectverb agreement. Further testing, however, is needed to study the interaction of these visual pathways and the respective linguistic code need to be acquired for language.

3.2.1.2 Visuo-Spatial Linguistic Reference According to Anderson and Keenan (1985:277), all languages have elements that designate spatial locations relative to the speech event. These deictics identify the locations by reference to the speaker at coding time.' Visible spatial reference differs from simple visible co-presence because of the addition of a third entity in the discourse, a visible target in a spatial configuration to the human participants, speakers, in the discourse. Crucially, it requires linguistic coding of this spatial relationship.

Evidence for treatment of utterances with visible targets are found in Task III and Task IV, both containing a visible target other than the speakers in the immediate environment. Task III involves a chart in which both interlocutors are focusing on the same target (chart), but one has blanks where the other has verbal information. The learner must direct the interviewer's attention to the potential blank being referred to. Once the interviewer has located the appropriate blank, the two interlocutors can continue and finish with mutual knowledge of the actual antecedent.

Task IV, on the other hand, involves a picture in the learner's hands which the interviewer is not viewing. Thus they are not both attending to the same visible target. The learner is attending to a visible target at all times, i.e. the picture, although also attending to the auditory responses of the interviewer as the ID description continues. Utterances and constraints for the interviewer, who is not directly viewing the picture in this task, are described under Indirect Visible Co- Spresence 3.2.1.2.1 Demonstrative Pronouns 'This' Versus 'That' (Visible) SDemonstrative pronouns can be any of the three types of utterances constrained 00 by visible co-presence in English with respect to coding time. In English, if the target referred to is visible to both interlocutors, and both are attending to the target, the speaker can use this and it is an act of immediate visible co-presence.

N Following Himmelmann (1996), cross-linguistically valid criteria of 'true' demonstratives are located on a distance scale, as proximal and distal.

N Pronominal and adnominal (adjectival) demonstratives are used for both spatial N deixis and anaphoric core argument structure in my data." 2 Adnominal use Sexceeded pronominal use in my data: out of seventy occurrences of demonstratives, only eleven were pronominal, including subjects of subordinate Sclauses.

33 Of the total of eleven occurrences of pronominals in the corpus of Japanese speakers, seven were uttered in Task III by the interviewers after their attention had been drawn to the desired blank on the information gap chart. Thus, both speaker and Hearer were viewing and attending to the blank on the chart together at the moment the interviewer made the respective utterances.

Example 3.9) i 14-1 I.txt:3 i14 Kodomo no hi? yeah this is to celebrate il0-sI l.txt:3 i10 ah, (xx) big type drum? okay, this is the name of the festival is (xx) i3-s6.txt:3 i3 oh, er, I think this is the (xx) shrine's end of the year festival i3-s7.txt:3 i3 what does this mean? Since the grammaticality of the utterance is determined by the pragmatic behavior at coding time with respect to visual attention directed at a definite target by both interlocutors, these interviewers appear to consider this appropriate." Although this task involved each person holding the chart with information about Japanese festivals in the Tohoku area, the learner had blanks where the interviewer had the information the learner sought. Hence the two objects they were holding in their hands were not identical. When they were talking about the target, one participant had more information than the other. So, the learner could use information from the chart to guide the interviewer to the desired information, i.e. to direct the Hearer's attention to the desired location.

Since it is the exact moment of coding time with respect to the Hearer's attention to the target to which the speaker refers which determines the linguistic behavior, the Hearer's responses would be uttered after the speaker had made the respective blank definite.

In Task IV, when the learner is holding the picture of the festival and looking at the festival, the learner is attending to the visible properties in the picture. This example illustrates the use of the demonstrative pronoun this used NO to draw the Hearer's attention to the target and the beginning of his description in 0 the initial utterance:

(N

SExample 3.10) Si3-s3.txt:4 s3 mm, this is famous (xx) festival, mm, it's very beautiful, is, it is night 00

C

N, Since the speaker was holding a picture of a festival, the visible target acted as a noun head. Thus an adnominal would have been redundant, so the Sspeaker used the pronominal deictic only. Also, the speaker could be certain that the hearer was already attending to the target because the interviewer had just offered the learner the folder from which to choose a picture. Moreover, the cvisible interaction between the two participants confirms the attention of the Shearer to the picture in the speaker's hands.

O

S3.2.1.2.2 Demonstrative Adnominals (Adjectives) -'This' Versus 'That' Adnominal use exceeded pronominal use in my data: out of seventy occurrences of demonstratives, fifty-nine were adnominal, including subjects of subordinate clauses. Example 3.11 was given as the interviewer in Task IV drew the learner's attention to the folder needed for the picture description task: Example 3.11) il -s l.txt:4 ilO 1 will give you this one picture, could you explain this picture? ilO-sl l.txt:4 il0 maybe we have just a small time. I can give you a picture.

Please describe a picture (xx) and I will guess what this one is, okay? Describe it.

il0-sl3.txt:4 il0 don't tell the name of this festival, just describe il6-sl2.txt:4 il6 Here are some pictures of famous Tohoku festivals. Please pick one and describe this festival In all of these utterances, this was used at to refer to a target immediately visible at coding time. In the preceding words leading up to the target this, the Hearer's have already been prepared both with the physical objects in the speaker's hands and the auditory signals directing their attention, e.g. "I will give you this one picture". The verb give suggests to the Hearer that a beneficiary and goal will be mentioned soon. Thus, the Hearer is already prepared when this is made definite by the reference act.

The following utterances were given during Task IV, as learners described the pictures. In these utterances the speaker directed the Hearer's attention to the picture in his hands as the speaker started to tell the Hearer the contents of the picture: Example 3.12) ilO-s l.txt:4 sl yes, I think this picture is Kamakura 14-1 I.txt:4 sl I this this drum is a very mm ve one of biggest drum in Japan il0-s l.txt:4 i10 uhuh, this is straw hat 00

N,

12.txt:4 i 6-s 10.txt:4 i16-sl l.txt:4 i16-s 2.txt:4 i 16-s 12.txt:4 i16-s 13.txt:4 into the sky, then that, this ball it is destroyed [fireworks] ah, this one is at night, the snow, the winter location. 1 think this is in a west. ((laughs)). mmm.

and this picture is night, at night and mountain winter this picture is describing about some festival or ceremony and, maybe, not maybe this is during the day time and usually, usually, you know, if somebody play the drum they're playing like this they are using this kind of sticks like this, right, but some people are playing drums like this, holding the stick like this, and I can tell this drum is huge and kind of dirty, umm However, the Hearer was not looking directly at the picture, only the speaker and the back of the picture being described. The Hearer is attending to the speaker's words, face and picture as the speaker description continues. Thus this is functioning as a demonstrative adjective for a visible target which differs between the two interlocutors, the picture in the hands of the speaker on the one hand, and a face holding an object known to be a picture, but whose contents although being described are unknown and not visible to the Hearer.

Nonetheless, this is used as reference for a definite target.

Finally, only one instance of the plural demonstrative was found in my data at immediate coding time. The interviewer is referring to the information gap activity and his own sheets with answers as opposed to the learner's.

Example 3.13) il0-sl 1.txt:3 i10 and umrn, those papers, on those sheets of papers are the missing parts Japanese is a three term person-oriented system with the middle term closer to the speaker than the third term. The difference between the two language systems can be diagrammed as below (Kuno, 1973): Japanese G is

O

0 ano 0 sono Kono indicates 'near speaker' or this', ano indicates 'far from both speaker and Hearer' or that over there' and sono indicates targets 'near to the Hearer or easily identifiable, that'. Unlike other adjectives in Japanese, these deictic terms do not indicate past or non-past tenses. They are important for information about the speaker and Hearer and the visible relationship of targets O between speaker and Hearer (Anderson and Keenan, 1985:284).36 Thus, the Japanese learner must adapt to the new spatial dimensions and visual constraints Sgoverning their linguistic behavior in the second language.

t(N S3.2.1.2.3. Spatial Locatives SDefinite reference has been examined primarily from the perspective of nominals, 00 such as generics, proper nouns, and definite descriptions." 7 In addition, the notion of topic "is assigned only to terms" (Dik, 1989; Bresnan, 1987; Siewierska, 1991; Lambrecht, 1994) and their behavior in predicate argument (i structure. However, when we look at the sources of definite reference and how they manifest themselves in language, consistencies in the constraints and behavior between these manifestations of topics with other grammatical (i categories are apparent. So, I will expand my analysis of the above topic-making Sexpressions to this type of mutual knowledge to include the entire deictic system, including spatiovisual adverbials, or spatial (locative) relations in the "extralinguistic" environment. Therefore, I adopt and expand C&M's classification of spatiovisual deictics to establish specific contexts of mutual knowledge which constrain the behavior of these utterances at coding time between speaker and hearer. Instances of direct definite reference of nouns can be shown to be part of a more general system in the pragmatics of utterances, i.e.

the exact context of mutual knowledge and coding time. These function more like nominals, since NPs may be substituted for these forms as well. In addition, they follow patterns of specificity associated with nominals.

My intention is to demonstrate that their behavior is part of a much larger system of direct definite reference, obeying similar constraints of time of the exact coding time and the linguistic form and behavior of elements in the system.

Additionally, Anderson and Keenan suggest that the locative adverbs 'here' and 'there' are probably universal to all languages. Interlanguage learners are attentive to these properties.

The English system of two deictic locative adverbs, 'here' versus 'there', differ with respect to the speech event and their reference to the speaker, as with all deictics. 'Here' is closer to the speaker and 'there' is a location away from the speaker. The utterances below are in the context of immediate visible copresence and were used by the interviewers to start the description in Task II.

Note that they refer to a larger geographical area than the interview room: Example 3.14) i I0-s 12.txt:2 i l10 My home town is Akita, so urn, Kanto festivals are very famous here in Akita ilO-sl3.txt:l S13 urn, I can't, I've been, I came here, I came here last spring, right, so I just, when I was in Tokyo I could do everything wherever I want The above sentences also offer a different analysis of deictic locative adverbs. Although grammarians usually refer to here and there as locative adverbs, the deictics can easily be seen as substitutions for NPs. In the first sentence by i 10, here substitutes either for a specific location inside Akita, e.g.

SMSU-A or emphatic here for Akita itself. In the second sentence by s 13, here can be instantiated either by MSU-A or the name of the village where it is located, SYuwamachi. The final example contrasts the city Tokyo with here, Akita. Thus, Sspatial deictics can be analysed as yet another instantiation of nominal entities by Sreferential lexemes. For this reason, I prefer to call these spatial deictics rather than locative adverbs.

00 The following utterances refer to a more confined space with an actual target present between the two interlocutors in the interview room. These utterances are found in Task IV, as the interviewer asks students to choose a C- picture of a festival from the folder.

C" Example 3.15) C-Kl il4-sl l.txt:4 i14 ok so now um final some pictures here it's all in Akita festival IND il4-s12.txt:4 i14 alright then um could you select one of the picture here i l0-s12.txt:4 i 10 mm, okay, that's enough, okay, maybe we have a small sometime, this is Pick up please one pictures of here and please describe the picture i 16-s II .txt:4 i 16 ok. here are some pictures of famous Tohoku festivals.

please pick one and describe the festival and I will try to guess.

Since it is the speaker's intention to hand the pictures to the student (hearer), the relation of here (in the possession of the speaker) to the hearer is suggested in the speaker's utterances.

Whereas English has a two term system for spatial demonstratives, Japanese has a three term system: koko 'here (near speaker)'; soko' 'there' (near the Hearer)'; and asoko 'there, (distant)'. The middle term is not just used to indicate an intermediate distance between the two speakers, but indicates a location near to the Hearer. Hence, Japanese learners of English have to learn to modify their perception of spatial relations with respect to linguistic behavior.

3.2.1.2.4 Presentatives Anderson and Keenan also include a category of presentatives which indicate a target's location or signals its appearance in (or relative to) the visual field of the speaker.

38 English uses the same lexical item and construction (There is X in both uses, and learners did not distinguish presentative from existential use (Chapter 4).

3.2.1.2.5 Visibly Evoked Interrogatives English interrogatives, like other p-structure lexemes, do not require the morpheme-labelling of the specific sensory-motor (visible) system being accessed.

9 However, special vocabulary may be used to direct the hearer to the speaker's desired target. My experiment had one task which required Whquestion words in order to find the answer to the blanks in a chart, a visual target.

Task III required learners to look at a chart with blanks and ask the interviewer for the answer. Although the interviewer did not have blanks, the learner directed the interviewer's attention to the relevant blank which did contain the answer.

00

(N

Thus, visual attention was required of both participants at coding time for both the questions and answers of the respective answers. In some cases the questions appear to be based on the graphemes contained in the chart. The following questions by learners during Task III illustrate: Example 3.16) il4-sl l.txt:3 sl I ok could you tell me what could you tell what purpose of Kodomo no hi? il4-sl 1.txt:3 sl 1 ok the another another festival um uh what participation particip-ate had ah Tsunahiki? Especially, the second utterance demonstrates the learner's difficulty with pronouncing the multi-syllabic word which was on the chart, and simply instantiated for the question regarding the empty blank. One strategy for filling in the blanks was used by s4, listing all of the names given and then by default getting the desired answer. The visual lexemes on the chart not only provided festival names for comparison, but provided a blank for the answer given by the interviewer.

Example 3.17) i3-s4.txt:3 s4 okay, could you tell me about na, festival name and (Tanabata), (Nebuta), (Ohmisokka) and (xx)? This student also used a copula and printed words on the chart to formulate a question: Example 3.18) i3-s4.txt:3 s4 i3-s4.txt:3 s4 and what is meaning of name of Tanabata festival? mm, and, ah Amekko-ichi festival, what is location of Amekko-ichi festival? Confusion caused by simply reading the visible captions on the chart which had the word purpose, location, age, meaning and date respectively and attaching the appropriate Wh-word to form a question resulted in the following interactions by sl 1: Example 3.19) Purpose ilO-sl l.txt:3 sl I il0-sl I.txt:3 il0 il0-sll.txt:3 sl Location il0-sll.txt:3 sll Age il0-s l.txt:3 s l mm, okay. Okay, I want, I want to ask about fes, Kamagura (xx) what origin, what pur, pur, power, pose okay, okay. (xx) Tell me what festival, um in location, ah sorry, what name of festival okay, okay. what age of Kamagura? IDMeaning il0-sll .txt:3 sl 1 what meaning ofNamahagi Date i l0-sl I .txt:3 sl 1 mm, okay. Okay, what date of(xx)? il0-sll .txt:3 sl 1I ah, what date 00 The learner consistently 4 0 does not use the copula to form the question asking about purpose, location, age, meaning and date. This learner's ability to provide the grammatical structure for the copula of existence is obviously still at an elementary level. However, the visible presence of properties of festivals in the ~chart is a source for the interrogative in order to fill the blanks on the chart.

Thus, the definite properties visibly provided become a source for instantiating c the appropriate indefinite Wh-variable of existence. However, the learner knows IDthat the Hearer has the answer to his question. Thus, these interrogative pronouns are derived from conscious knowledge that an answer can be provided-the identification of the information needed to fill in the blank.

In these cases, the expected target is visibly induced. The Wh-question word, although a non-referring pronominal, is posited with an expectation of knowledge from the speaker (for a different view cf. Lambrecht, 1998). It is referring to an expectation of the identification of a property of an event, a festival. Therefore the festival itself is definitely in the minds of both interlocutors. In addition, a specific Wh question word (variable in Heim and Kamp's sense) is combined with an identified property on the chart, e.g. whatlocation, what-name, what-age, what-date.

3.2.2 Prior Visual Co-Presence If both interlocutors have previously looked at a visual target, prior visible copresence, but have ceased prior to the utterance, the demonstrative 'that' is appropriate. In my data, that is triggered by the learner describing objects in a picture and then referring back to an object just mentioned while still viewing the same picture: Example 3.20) i I 0-2s.txt:4 slO yes, ah, some people have a (xx) that is a one, one stick, one long stick ilO-sl I.txt:4 s I and wearing that, (xx) that is Japanese old traditional shoes Using this schemata for describing the use of the demonstrative "that" is especially useful for describing double targets. Rather than assuming a contrastive element or "pointing back" in the visible situation, double objects can be described as a type of mutual knowledge which is consistent with all other aspects of visible co-presence, but additionally requires recallability of a target viewed immediately prior but requiring recallability of its visible co-presence when referring to it. Interlocutors must be attending to the target, but rather than immediately present they must both be simultaneously recalling its presence in O the prior situation. Contrastiveness is not an appropriate description of the rules O governing this linguistic behavior; rather behavior of attention at coding time.

STherefore, appealing to the mutual knowledge between speaker and a Hearer, i.e. pragmatic constraints of attention and definiteness as described, is Smore informative than an observation regarding its preverbal subject position.

00 3.2.3 Potential Visible Co-Presence When a target is easily in view, but the hearer is not attending to it, the speaker can get the hearer's attention by saying 'this' (+nonverbal gesture indicating the Starget) in English. In this situation the context of potential visible co-presence constrains the behavior of elements in an utterance. Two tasks in my study offer C this context: Task III, the information gap activity and Task IV, a description of Nl a picture. The learners were not initially attending to either visible target, which O were both being held by the interviewer. The interviewer had to draw the Slearners' attention to the visible targets, before the activity could begin.

3.2.3.1 Demonstrative Pronouns Keenan (1976) claims demonstrative pronouns are "highly referential" NPs which can always occur as subjects. Although this is true, the exact behavior of these pronouns like other deictics is constrained by the exact type of mutual knowledge shared between the interlocutors at coding time in the specific context, in this case visibility and attention. Thus, constraints on verbs and their respective predicate argument structure is irrelevant to the pragmatic behavior of entities in a visual context, e.g. pronouns, social status markings, and demonstratives. Task III was initiated with demonstrative adjectives, no demonstrative pronouns were found.

3.2.3.2 Demonstrative Adjectives The following utterances use demonstrative adjectives indicating the chart in the speaker's hands (also the learners') to initiate Task III: Example 3.21) i2-s5.txt:3 i2 Please show uh please look at this chart il4-si l.txt:3 i14 one night ok alright so now I go to this chart i I 0-s I 1.txt:3 i l 0 and urn, those papers, on those sheets of papers are the missing parts In my experiment this context also occurred during Task III, when the learner was drawing the interviewer's attention to the blank containing the "information gap" chart. The interviewer had the same chart as the learner, but had not yet been instructed which blank to attend to in order to provide the answer requested by the learner.

Task IV also provided a context for interviewers to use demonstrative adjectives to draw the learner's attention to the description of a picture, the last activity in the discussion: IDExample 3.22) i I0s2.txt:4 i10 I will give you this one picture. could you explain this picture? i3s6.txt:4 i3 can you pass this one, pick one of the pictures i lOs I .txt:4 i10 maybe we have just a small time. I can give you a picture. Please Sdescribe a picture (xx) and I will guess what this one is, okay? Describe it.

00 3.2.4 Summary of Visible-Linguistic Co-Presence In conclusion, this analysis of "situational uses" of pronouns and demonstratives in a visible context provides a narrower analysis of the use of the forms than any other analysis of given information in a situation. It not only includes specific Scontexts defined according to coding time with respect to attention to a target, but Sit includes the temporality of immediate, prior or potential presence in the IDinteraction. Previous attempts to explain topics in "situations" have not formally Sdistinguished coding time as constraints on behavior, although they are N sometimes included vaguely in the descriptions (Prince, 1981; Lambrecht, 1994; Himmelman, 1996).

Neither have earlier attempts subdivided the situation into individual sensory language areas. In so doing it becomes immediately obvious that simultaneous attention to a targeted item at coding time determines the grammaticality of the definiteness of the respective entities in utterances.

The representation of shared knowledge in this analysis offers an explanation for one of the more perplexing questions in information structure (Lambrecht, 1994): How can a word be both new or focused and given or topic at the same time? In the case of NPs beginning with the determiner or demonstratives of this and that, if the visible target is in the immediate copresence, it may be first mentioned in the conversation but it is nonetheless a topic, because it is visibly present between the interlocutors.

3.3 Auditory linguistic co-presence This type of mutual knowledge is based solely on auditory or textual comprehension, since we are speaking of verbal communication, usually between at least two 'hearers', as opposed to speakers and hearers in the presence of visibly present targets. Since auditory comprehension of linguistic text is necessarily linear, utterances must occur before a hearer can become conscious of their presence in a discourse.

4 Therefore, C&M claim linguistic co-presence can never be immediate, unlike visible co-presence. So, they propose only two categories of linguistic co-presence, potential and prior. C&M's claim is based on an argument that "unlike physical co-presence, linguistic co-presence can never be 'immediate', that is, simultaneous with the definite reference for which it is used"...because "a candle cannot be spoken at the same time as the candle"(p.39). Clearly, they are talking about indefinite physical targets being both indefinite and definite, so that the antecedent must come either before or after the anaphor. They do not mention the possibility of proper nouns.

Obviously, mentioning a proper noun is simultaneous with the definite reference for which it is used. Therefore, in contrast to C&M's analysis, I have included immediate auditory (linguistic) co-presence as an additional basis for mutual knowledge. Thus, as in the case of NPs beginning with the determiner or O demonstratives of this and that in immediate visible co-presence, proper nouns O also may be first mentioned solely in an auditory context in the conversation. In order to accommodate this target in an auditory linguistic context, I have added immediate to coding time of auditory linguistic co-presence.

S(I) immediate auditory linguistic co-presence when the two interlocutors COOare attending to a mentioned NP target, proper noun, in an C utterance prior linguistic co-presence when the interlocutors have previously mentioned a target but have ceased before the time of the actual Ci utterance.

potential linguistic co-presence when the two interlocutors are both attending and anticipating an antecedent in a phrase in an utterance by the speaker.

I3 3 3.3.1 Immediate Auditory-Linguistic Co-Presence As just explained, proper nouns in the auditory context of an utterance at immediate coding time are the entities which can be both definite and just mentioned in an utterance. In Task I, the introductions, and Task II, description of a festival without a picture, were both composed almost entirely of utterances based solely on auditory information, except for the visual context of person, as described above. Since these two tasks were made in each other's direct visible presence, utterances which refer to either the Speaker or Hearer contained the person deictics, I and you, lexical forms based on direct visible co-presence. All other utterances in these two tasks are considered topics in information structure only if they are proper nouns. Utterances containing proper nouns were scattered throughout these two tasks. These utterances I also consider to contain topics, in spite of the entities being introduced into the discourse for the first time, i.e. not previously mentioned.

Since proper nouns are stored in memory, albeit long-term memory, as opposed to short term memory, they are nonetheless part of a memory system and hence are that part of a sentence which is based on information known in the 'community of speakers', such as citizens of a country with well-known national figures or current students in a teacher's class. In other words, they can be assumed to be definitely known to speech participants. Since they are topics, because they are mutually known and hence mentally co-present between the interlocutors at coding time, this representation of proper nouns and shared knowledge is thus another example of how a word can be both new or focused and given or topic at the same time.

42 Personal facts which included Japanese proper nouns of personal names, city and prefectures, known to all the participants in the dyads, are typical of the first interaction in the all of the dyads: Example 3.23 ilO-slO.txt:l ilo0 My name is Manabu Watanabe, I am teaching at Japan Area Studies Department, and I am (xx) another department. May I have your name? il0-sl0.txt:l s10 My name is Yoshihito Yamashita, please call me Yoshi il0-s IO.txt:l i 10 Yoshi, okay, Yoshi, ah, are you, where are you from originally? ID i IO-slO.txt:l slO originally, ah. 1 came from Kanagawa prefecture ilO0-slO.txt:l il0 Kanagawa prefecture, okay Kanagawa is a big prefecture ilO0-slO0.txt:l slO0 mm il0-slO.txt:l i10 I'm from Akita prefecture originally VThis is slightly unusual because the student also provided a shortened given name for the teacher to use for him. This is common in English classrooms but not in N Japanese classrooms. One introduction was closer to the Japanese cultural style of introductions which include only the surname with the occupation and job title.

Example 3.24) i4-s4:txt. I i4 my name is Kudo 3.3.2 Prior Auditory-Linguistic Co-Presence IND This source of mutual knowledge, also termed 'previously mentioned', is the Smost widely known and studied type of anaphora. An antecedent has already been mentioned in the auditory-linguistic context, so it is already mentally present for both the speaker and Hearer. The target is therefore definite and mutually known, copresent, in the speech situation. Initially, Sidner (1983) uses the traditional definition of anaphora as the words in utterances to "point back" or refer to previously mentioned persons, spatial or temporal or nominal objects, events or ideas. These words or phrases which "point back" are called anaphors. Typically, researchers of anaphor study pronouns, the words which replace previously mentioned entities. Chafe (1974) argued that the referents of pronouns must be in the listener's consciousness, "on stage," at that point in the conversation. When the referents are recallable, or locatable (in auditory memory) from previously mentioned utterances in the discourse, the speaker can use a pronoun otherwise he can not p. 44). However, the analysis above which includes visible antecedents requires including more than previously mentioned utterances as sources for pronouns. Hence, this explanation of anaphora is too narrow for my analysis.

A more appropriate definition for anaphora is given by both Sidner (1983) and Webber (1981) which leads to a possible line of investigation in my thesis: The focus and the assumed shared knowledge can be used as one of the chief constraints on the choice of the co-specification of anaphoric expressions. Rules governing an anaphor interpreter can be discovered which use these two sources of constraints. In these rules the focus will play a central roles as a source of cospecification. The focus and the structure of assumed shared knowledge are significant to rules governing the choice of anaphors because they capture the effects of what has been talked about previously and what the speaker has assumed is knowledge that is shared with the hearer (Sidner, 1983:274).

However, Sidner is talking about a discourse level of anaphora and principles of being perspicuous developed by Grice (1975). For now, I am only interested in her assumption that speakers assume the Hearer has enough real-world knowledge in common with the speaker to know about the entities in the real IDworld and the cognitive elements of the discourse which the speaker refers to or mentions. In other words, the focus of one utterance constrains the form of the 0 subsequent referent, while simultaneously contributing knowledge of a discourse a entity which hence becomes mutually known. The referent, or anaphor, then is Sconstrained by the mutual knowledge which has been introduced into the discourse.

00 Anaphora can be summarized this way. It is prototypically expressed with pronouns or definite descriptions. The expression that is appropriate depends on the type of linguistic co-presence: whether it is potential or prior, C whether it 'commands' the definite reference or not, and whether it is available in immediate or long-term memory, among other things. Anaphora can also be Sexpressed with demonstratives, yet the demonstrative that is appropriate again Sdepends on whether the linguistic co-presence is potential or prior. The choice of definite reference, then, is heavily determined by the basis for the mutual knowledge which constrains it.

3.3.2.1 Anaphoric Pronouns As stated above, anaphora must co-specify with the focus, or new information, of an earlier utterance. This definition is not as straightforward as usually presented, as shown by my data. The following example introducing Task II, the description of a festival without a picture, illustrates the notion of cospecification: Example 3.25) il4-sl I.txt:2 i14 ok so I think there must be a Japanese festival some, somewhere near your home Nakayama il4-sll.txt:2 i14 and let's talk about one ofthem i14-sll.txt:2 slI one il4-sll.txt:2 i14 hmm can you describe it In this case, "a Japanese festival some somewhere near your home Nakayama" is the new information: its existence is being posited in the 'existential -there' clause, "there must be a therefore it is first mentioned.

Secondly, it contains an indefinite article before "festival", which has been restricted both by the proper adjective 'Japanese' and then by further locative information "near your home Nakayama." The referring pronoun in the following utterance is them. The lack of number agreement between the focus in the first sentence and the referring pronoun in the second sentence is one reason that Sidner and Webber claim that we should not use the traditional definition of anaphora. Obviously, a festival somewhere and them co-specify, and the sequence is not considered ungrammatical by native speakers of English.

However, the Hearer must be able to associate that the indefinite "some, somewhere" is nonetheless making the exact existence of the particular festival near Yakahama the Hearer's choice. It is presenting indefinite locative information with the word somewhere. The indefiniteness is not only constrained by its location to Yakahama, but also by the Hearer's mental image of a festival as opposed to the speaker's. The festival is simultaneously becoming a definite mental image in the Hearer's mind, as the speaker continues to restrict the O location of the festival without forming any specific mental image. So we have O differing levels of specificity developing in the mind's of the two interlocutors.

C However, in both cases, the indefinite a festival somewhere, once its existence has been introduced by the existential- there construction (with the modal of Snecessity, must) is in fact definite. The following indefinite information is thus anchored by the definiteness of this "existing entity", a festival. The exchange 00 ends with the student, the least advanced according to TOEFL scores and class performance, comprehending that a festival linked to one of them in the partitive, most likely refers to a desire by the interviewer for a description of only one, as indicated in his question. The Hearer is to scan his memory for festivals near his home and choose one.

This exchange continues showing another type of anaphora based on CN inferencing by the interviewer: tN 0Example 3.26) il4-sl I.txt:2 i14 But please describe what you know that festival what it looks like il4-sl .Itxt:2 sll Like mm (xx) dancing il4-sll.txt:2 i14 Dancing ok il4-sl I.txt:2 sll Dancing very traditional dancing il4-sl .txt:2 i14 Traditional Japanese dancing i14-sl I .txt:2 sll The (xx) il4-sl I.txt:2 i14 the (xx) ok what kind of costume do they wear? In this entire exchange, no mention has been made of people. Nonetheless, the interviewer has formed an image of people with the notion of dancing at a festival. The use of the pronoun "they" thus seems very natural and appropriate, even though there is no specified antecedent, either singular or plural. In other words, prior lexico-grammatical mention will not work to explain this type of anaphora. Neither can one subtract the assertion from presupposition or antecedent, as some researchers (Lambrecht) use as a defining characteristic topics. Finally, this is a challenge to traditional definitions of a pronoun as a word which replaces a noun. There is no nominal antecedent in this exchange.

This example is a complicated interaction of several sources of referents: Example 3.27) sl :txt:4 So there are women who_who are wearing kimono like and they also uh with ka_kasa_Japanese kasa Initially,'the learner is looking at a picture of a festival and more than one woman which triggers the plural noun. So the antecedent for the speaker are visual targets. However, the following anaphor they, once it is uttered by the speaker, is in the auditory memory of both the speaker and Hearer. Hence, the anaphor is based on a prior auditory signal for both interlocutors at coding time.

For the speaker, however, who is also still looking at the picture, the anaphor concomitantly has a visual target of the antecedent still in view.

IND For the hearer, on the other hand, the antecedent and anaphor are both O orally presented. Initially, the hearer who is forming a mental schemata of the picture of a festival in her mind, the antecedent women is presented orally e( together with the mental image of the picture of possible participants in a festival.

DHer domain of possible referents has been constrained by the knowledge of a V) picture of a festival in the speaker's hands. This aids in understanding the 00 auditory signal pronounced by the speaker. So, there are two sources for the hearer: inferential and auditory. This is especially helpful in this case if there are problems of pronunciation by the speaker who is a non-native speaker. In C1 addition, the anaphor itself is now subject to the previously uttered antecedent, its auditory signal, as well as the continuing mental schemata and the pronunciation (N of the speaker.

CAnaphoric demonstratives in Japanese are of the same series as the deictic visible demonstrative series presented above: kono this', sono that', 0and ano that over there. However, in the case of auditory-linguistic co- Spresence, mutual knowledge by the interlocutors is not based on visible knowledge, but familiarity and acquaintance. If a person is mutually known between the two interlocutors, then 'ano' is used. If the speaker feels that the Hearer does not know the person being referred to, then sono is used. The use of kono can only be used when the speaker knows the person or object well, almost as if it were visible and by the speaker's side. Once it is introduced, the a-series is preferred and the ko-series is considered unacceptable (Kuno, 1973:282-290).

3.3.2.2 Demonstrative Pronouns Demonstrative pronouns are less frequent than anaphoric pronouns in my data: Of 64 total occurrences, 44 pronominal and 20 adnominal were found. Examples of the pronominal are given below.

Example 3.28) il0-s12.txt:2 sl2 that is so, that is made of wood il0-s13.txt:2 il0 spa, that sounds great il14-sl l.txt:4 i14 aa soo desu ka oh is that so/right? i3-s7.txt:2 s7 bamboo and lantern, yes and it's very heavy, so, I didn't try it but my friend tried how to do that and she, i3-s7.txt:2 s7 and after that they had to pick up the lantern with bamboo i 16-sl I .txt: il6 I don't think that is actually the name A similar frequency distribution was also found in studies of the use of demonstratives in four corpora of narratives by Himmelmann (1996:216). In the Susanne Corpus for English, the pronominal use of that was greater than for the adnominal use out of 237 instances. However, the use of this was less than the adnominal use out of 596. Himmelman's (1996) review of demonstrative pronouns used narrative so his results are primarily found in the auditory linguistic co-presence and mixed types. In Chafe's analysis of the well-known Pear Stories, 122 adnominal uses were found compared to 52 pronominal uses out of the total of 174 total. However, this is based on narratives O and more formal varieties of demonstratives, so is therefore not directly O comparable to my corpus.

43 3.3.2.3 Demonstrative Adjectives 0 Examples of demonstrative adjectives are shown below: OO Example 3.29) il0-sl2.txt: sl2 no just a, I don't join that team C il0-sl0.txt:l il0 life, okay, so you basically help those people? i2-s2.txt:2 s2 but 1 couldn't use that way il0-sl2.txt:2 s12 mm, watching that kind of festival CII il0-sl2.txt:2 s12 because, some frenzy that Mikoshi fell down \O i3-s6.txt:2 s6 and I think that festival is very fantastique and erotic festival S3.3.3 Potential Auditory Linguistic Co-Presence C&M (1998) addressed problems with various theories of definite reference, especially location theory proposed by Hawkins. They argued that, contrary to other theories of definite reference, it was in fact possible for an anaphor to precede its referent and gave examples from natural language to illustrate Their first example contains an anaphor with a subordinate clause preceding the matrix: "Before he could steal anything, a burglar who had broken into our house was frightened away".

44 These are special cases and an examination of my data did not have any of these types of anaphora.

3.3.3.1 Pronouns To refer to something established by prior co-presence, one can use either this or that, but to refer to something yet to be established potential linguistic copresence one must use this 1981:44).

3.3.3.1.1 Demonstrative Pronouns I have only a few examples of this type of anaphora and they all use this or those.

Example 3.30) il0-2sl0.txt:l sl0 so these are good points il6-sl3.txt:3 il6 okay, these are all (xx) In discourse, Japanese uses sono, the deictic from the three term personoriented system mentioned above. This is used deictically to indicate targets near to the hearer, or easily identifiable, but anaphorically for referring to items previously mentioned. Recall that unlike other adjectives in Japanese, these deictic terms do not indicate past or non-past tenses. They are only important for information about and interactions between the speaker and Hearer. The temporality of the situation with respect to previously mentioned is key in IDdetermining their use, similar to the English situation. Thus learners are only O learning new lexical forms for referencing, but the pathways for processing are Ssimilar.

S3.3.3.1.2 Demonstrative Adjectives In this utterance, here is refers to "this college" which follows the word here.

00 rExample 3.31) C i3-s7.txt: I s7 came here to this college because ah, I wanted to know this college system The use of here followed by this college could be interpreted as an instance of Spotential anaphora. However, the here could also be interpreted as referring to ID the large geographical sense of the spatial demonstrative at the time it is spoken and this college only fotbllows for purposes of clarification. In other words, the Sfollow up with the prepositional phrase narrows the location from Akita to MSU- A, this college.

Other examples of plural demonstrative adjectives are found in my data: Example 3.32) i3-s7.txt: 1 i3 have these majors so, maybe, I'm not sure, but maybe I will go to (xx) i3-s3.txt:4 s3 th, mm, one (xx) pass these pole, pole i3-s3.txt:4 s3 and there are a lot of lantern these pole i3-s6.txt:3 i3 eat with friends, those people buy a candy 3.3.3.2 Spatial Relations As mentioned earlier under visible co-presence, the two deictic locatives in English, 'here' versus 'there', differ with respect to the speech event and their reference to the speaker. 'Here' is closer to the speaker and 'there' is a location away from the speaker. These spatial deictics may also refer to locations which are known from the linguistic utterance, i.e. they are not visible as targets, such as the charts and pictures in Tasks III and IV. Instead, the geographic locations to which they refer are determined by the previously mentioned, i.e. auditory cues, in the utterance. If the location has not yet been mentioned, the Hearer must wait for the auditory information from the utterance by the speaker to determine the exact nature of the relation.

In these utterances, the locative deictic refers to the general university location known to both interlocutors in the interviews. The words "teaching" and study(ing) appear first in the utterance which thus establishes MSU-A, the university in which the speakers either study or teach, as the referential base for the locative relation.

Example 3.33) 14-1 I.txt:l 1 114 almost one year ok I've been teaching here for four years since the beginning of this university I\ il4-s13.txt:l 113 almost I've studying here almost one year

O

SThis utterance, although found in Task IV, is part of the closing given by the interviewer to a learner, so its referential basis is again totally auditory.

C) Example 3.34) 00 C i4-1 I.txt:4 i 4 good luck on your study here These utterances refer to other cities in Japan prior to the locative deictic, which then establishes the referential base: C Example 3.35) SHere ilO-sl I.txt:l il0 Wakayama? oh, far from here K il4-sl2t.txt:4 i14 ah not many Kyoto people around here right? il0-sl2.txt:2 il0 My home town is Akita, so um, Kanto festivals are very famous here in Akita il6-s l0.txt: I il6 I'm here for two years There i3-s6.txt:4 s6 mm, so they wait there each house i2-sl2.txt:2 s12 I have not been there before 3.3.3.2.1 Verbal Roots Verbs such as come and go, take and bring are pairs based on deictic meaning." Fillmore (1966:220-223) analyses the deictic categories of certain verbs with the same properties of person, place, and time as all deictics, but with the addition of a "type of semantic rule which may be called 'supposition rule'. For example, the verbs 'come here' versus 'go there' differ in regard to movement toward or away from the speaker. The deictic verb sets up the referential base of the participants and the other targets in the utterance (Fillmore, 1966 Once the respective verb is uttered, the spatial relation with respect to the speaker is established. However, the exact meaning and usage of these two verbs is considerably more complex.

46 In these utterances from my corpus, the deictic phrase go-there refers to leaving Japan and studying in the U.S. A complete understanding of this utterance involves reference to person, place, and time. The learner, or Hearer, is currently enrolled at MSU-A in Japan and most likely wants to attend a university in the U.S. The speaker is a teacher at the same university. So both are in a different location than the destination at coding time.

Example 3.36) i14-1 l.txt:1 i14 so do you have any special reasons for going to US and study there The replacement of the NP US by there in the conjunct 'going to NP and study NP' in this utterance also offers evidence for an interpretation of the adverb IDphrase there as a substitution for the earlier NP U.S. Therefore, a proper noun of O location, has been converted to the adverbial locative there expressing the earlier target NP as a spatial relation to the geographic location of the utterance at coding time.

d In the case of go the referential base is straightforward: the speaker is referring to a goal or destination other than the location of coding time regardless 00 of tense. The referential behavior of come, however, is more complex, as Fillmore points out. In these utterances, the deictic came establishes the referential base so that the spatial locative refers to the university in which the (Ni speakers either study or teach: SExample 3.37) ilO-s3.txt:l s13 I came here last spring Ni lO-sl3.txt:l s13 urn, I can't, I've been, I came here, I came here last spring (Ni The utterance below, however, is more complex: the prepositional phrase specifying the exact referential base follows the deictic here. In addition, the demonstrative this is assuming an established referential base introduced by the deictic come.

Example 3.38) i3-s7.txt: I s7 came here to this college because ah, I wanted to know this college system The deictic relations for the verbs go iku and come kuru in Japanese differ from English, as explained by Shibatani (1990:380-383).

47 3.3.3.3 Temporal Reference Again following the organization of deixis presented by Anderson and Keenan (1985), temporal deixis makes reference to the extralinguistic context of the utterance, which relates two events "uniquely and necessarily" along a single dimension: one event is either "earlier, simultaneous with, or later than another." This is the exact division for physical co-presence which C&M also identified in their classification of mutual knowledge for definite reference: prior, immediate, or potential. In Chapters 6, I briefly mention event structure with respect to predicate argument structure. However, event structure also relates to temporal reference. This "interval of time on which the predicate occurs" the event frame, also relates to what Chung (1985) term "the event world", or the set of conditions on which the predicate occurs (p.203).

In our terms, it is directly connected to the context of mutual knowledge which is shared between two or more interlocutors with respect to the temporal reference at coding time. Temporal deixis does not necessarily control argument structure (subjects) of sentences; it does however control the organization and behavior of the information of an utterance.

O 3.3.3.3.1 The Category of Tense/Aspect O Anderson and Keenan (1985) see tense as relating events at the time of an C<1 utterance with respect to real world knowledge. Since time reference is an Sextralinguistic fact which must be coded into the utterance, time reference is a Spragmatic constraint. This question from my experiment exemplifies: Did it held in summer... this festival? The past tense auxiliary Did, a temporal deictic, refers ~0 to an earlier time for the event, a festival, in the utterance.

Tense/aspect has direct importance in determining the properties of existence and subjecthood.

4 8 As far as the tense/aspect marking of topics, Foley 0 has stated that until an entity is posited as existing it can not be doing something.

Hence, according to his theory, the perfective aspect is not found in the first C0N clause of presentational sentences.

4 9 3.3.3.3.2 Temporal Deixis in the Lexicon SDue to a rather limited or no specialized grammatical units other than tense CN morphemes, most languages use lexical items (Anderson and Keenan, 1985:300).

Besides the deictic 'now' and 'then', languages use adverbs to make finer distinctions, e.g. today, yesterday, and tomorrow, as well as proper nouns for days of the week and month. Fillmore (1966) also includes 'ago'. Like other deictics, these can be instantiated with NPs: yesterday I went to the store when uttered on Tuesday, can be instantiated with Monday.

3.3.3.3.2.1 Temporal Deictics As in many languages, English has two temporal deictics, one for proximal time and one for distal time: 'now' and 'then', respectively. Instances of direct definite reference can also be shown to be part of this pragmatic system. These examples of the deictic now were found in the introductions of Task I in my data: Example 3.39) i 4-sll.txt: sl ah now I want to study music and art il0-sl2.txt:l s12 yes, now I'm just thinking about it i2s2.txt: I i2 but now a little bit progress so 1 s5 mm, not now, I think now Japan became stronger sometimes i3-s6.txt:1 s6 ur, now I'm a policeman The examples followed answers to questions regarding the learners' intentions of study or planned occupations. So they indicated a temporal change from a previous time to coding time.

These utterances were spoken by i14 to change topics at coding time from the introductory information to the second activity of describing a festival in the learner's hometown.

Example 3.40) il4-sll.txt:2 i14 right it's a different topic now il4-sl2t.txt:2 i14 so now um I just different topic ok? IO i 14-s I .txt:3 i 14 one night ok alright so now I go to this chart il4-sl 1.txt:4 i14 ok so now um final some pictures here it's all in Akita festival These utterances indicates the use of the temporal to indicate sequence in the description of the learner's hometown festival.

00 Example 3.41) i14-sll.txt:4 i14 and then describe it i14-sll .txt:4 i14 and thenl I These temporal lexemes, like the other variables in the above discussion, are in N themselves lexically empty; they can be instantiated with nouns. They also have a proximate and distal relationship to each other, as do the spatial locatives. Like O pronominals, they normally do not attract sentence accents. Thus, they present a Spattern of regularity with respect to their pragmatic behavior at coding time.

N Chafe simply states they are not topics. Lambrecht (1998:521) explains that they are not activated and thus can serve as a scene-setting topic and go unaccented. My interpretation, however, seems to indicate that they are activated, but temporally, at coding time. However, like indefinites, they do not involve the kind of processing effort which is required in matching a lexical description with a specific referent. They are part of a more general system of pragmatic relations at coding time.

3.4 Indirect Co-Presence In the act of speaking and listening, the hearer acquires and uses knowledge from more than just the visible and auditory signals in the discourse situation. During the experiment, as an example, one learner when meeting the interviewer became excited and asked when her baby was due. He knew from her physical appearance that it was soon, the same as his wife. This is inferencing from visible evidence. It is part of community knowledge and rational understanding about expectant women that a child will soon be born. Interlocutors can also deduce knowledge from properties of visible and auditory targets, and mental images created during the speech situation."' In SLA most of the work on inferencing has been conducted with respect to reading Johnson, 1977), and negotiation of meaning studies (Hatch, 1983:73-74). Hatch cites Carton (1971) and Yorkey (1970), but examples such as "The night was so that not a sound could be heard", asking a student to choose the correct word (quiet, beautiful, dark, dangerous) are not the same as my usage of the term.

52 I am referring to making associations of properties of an event, concept, or even object and using these properties while engaging in a conversation. These studies are all based on visible-linguistic co-presence, i.e.

reading and acquiring vocabulary.

However, languages differ crosslinguistically in their morphological coding of evidence, process and results for deducing inferences. Mental models of inferencing based on predicate calculus, logic or set theory with respect to consciousness (Johnson-Laird, 1983) are too simplistic and have missed the rich subtleties displayed in natural language." O What is the nature of definiteness in this situation? How can an entity be O specific if it has not been mentioned? How can indefinites be specific and understood to exist prior to their introduction into the speech situation? How do a they differ from subjects? These are the questions I address in this section.

S3.4.1 Indirect Visible-Linguistic Co-Presence 00We may identify a group of pictures and then refer to one of the pictures, a specific member of the group. These are all types of indirect visible co-presence, when the participants are inferring knowledge from a visible target. These hearers C are behaving according to constraints of another type of behavior called inferencing.

,I 3.4.1.1 Demonstrative Pronouns O I have described the types of properties of Type IV in which the speaker holding a picture described its contents under visible co-presence. However, the Hearer Swas looking at the speaker who was holding the picture and was attending to the speaker's utterances in order to guess the name of the festival. The Hearer knew that the speaker had a picture of a festival in the Tohoku region taken from the folder before beginning the description. While the learner was describing the picture, the Hearer could both infer properties while listening to the speaker's words and also form a mental picture in order to guess the name of the particular festival, based on the general knowledge of festivals and the particular knowledge of Tohoku festivals.

Task IV involves a picture in the learner's hands which the interviewer is not viewing. Thus they are not both attending to the same visible target, although the interviewer is actively involved in forming a mental image of the picture being described, fully aware that it is a picture of a festival in the Tohoku region of Japan. For this activity, the interviewer's inferences of properties of festivals in general and also in the region are sometimes necessary when the learner has difficulty explaining the pictures. Throughout this task, the activity by the interviewer is indirectly associated with the picture, based on the auditory signals as the learner describes the picture. The learner on the other hand, is attending to a visible target at all times, although attending to the auditory responses of the interviewer as the description continues. These are examples of indirect visible co-presence, because the picture is in the hands of the speaker, but the Hearer knows that it is a picture of a festival. Therefore, using 'community knowledge' shared by the two speakers about 'is-a' properties of festivals, the Hearer can associate possible properties of a picture of festivals and infer an understanding of the contents of the picture necessary to guess its identity.

Example 3.42) i12-sl0.txt:4 i12 when does it take place? i14-s12.txt:4 i14 a lot is that all mekoshi? il4s 10O.txt:4 i14 ok so that must be ah Omagari fireworks NO 3.4.1.2 Demonstrative Adjectives Prince (1981) is one of the first researchers to notice the inferencing use of C indefinite-this NPs in English sentences, although her example does not include a Svisual target, but rather is the beginning of a narrative referring to 'a certain' person. In the utterance below, the interviewer had already handed the learner the picture for Task IV, and although the interviewer was not looking directly at 00 C the picture, the interviewer was looking at the target in the learner's hands, inferring that it was a picture ofa Tohoku festival, and said: Example 3.43) ilO0-s13.txt:4 il0 don't tell the name of this festival, just describe IO In the following utterance, the learner was holding the picture in his hands, attending to the picture and started the description of the festival by Ssaying: Example 3.44) i14-s12.txt:4 s12 ah this picture is ah describing about festival Thus, the learner was describing a target which the interviewer did not also hold, so both persons were not viewing it together. However, the interviewer knew the target was a picture of a festival. Thus, the speaker's words were determined by the visual target which then lead the Hearer (interviewer) to begin to form a mental representation of the contents of the picture of a festival (in order to be able to guess the exact identity of the picture in the learner's hands). The interviewer could infer traits likely to be seen in the picture of a festival.

3.4.1.3 Spatial Deictics During the fifteen minute dyads, when learners were required to ask questions about blanks on the information gap chart and describe a picture and its contents, the hearers had to figure out which blank the spatial directions the speakers were directing their attention. They also had to create a mental image based on the spatial relations the speakers described as they looked at the picture.

Using 'community knowledge' about the 'is-a' properties of festivals shared by the two speakers, the Hearer can infer an understanding of the contents of the picture necessary to guess its identity. In other words, the interviewer is not attending to the exact contents of the picture, but can infer that a spatial relation exists between the locations of targets in the pictures. In this dyad the learner is describing a picture of men walking under a waterfall, carrying a mikoshi (small temple carried on the shoulders of the men).

Example 3.45) ilO0-s13.txt:4 s13 yes, I can see the over there i 10-s 13.txt:4 s13 yes, see there, over there and, no, they're, they are in the water, they are walking in the water These utterances indicate that the viewed target is more distant from another target in the picture. These are examples of indirect visible co-presence because the picture is in the hands of the speaker, but the Hearer knows that it is a picture Sofa festival.

00 3.4.1.4 Partitives Partitives, such as one of the pictures, refer to NPs that are a subset of the referent of the NP set contained in the partitive, i.e. pictures. They may be either definite 0 or indefinite. In the utterance below, interviewer 3 is passing the learner, s6, a folder.

C",K Example 3.46) i3s6.txt:4 i3 can you pass this one, pick one of the pictures The learner has already participated in two earlier interviews, so this task is familiar to him, i.e. the learner should know that the folder contains pictures.

Even if the learner has forgotten, the learner can infer that the folder contains "something" as one of the properties of folders. The interviewer makes this clear as he says "pick one of the pictures", so that the learner knows from the auditory sign of the interviewer that the folder contains pictures and he should pick one and only one picture. Thus this context enables the learner to identify from prior familiarity of the task, the inferrable knowledge of the properties of a folder, and an auditory signal that a specific task with respect to pictures is being requested of him. Together the knowledge from this context combines to make the identification of the pictures by the learner very strong. Since this identification is linked to previously established discourse referents, the pictures is a definite NP and therefore a topic.

This definiteness of the NP referent makes partitives necessarily specific, as argued by En9 (1991:10-11). An NP, definite or indefinite, with a referent that is definite must be specific. In his analysis, the pictures is a weak antecedent for the lexeme one, because it is included as a property of pictures which is established in the discourse. Inclusion has a looser relationship than identification. In other words, the learner must infer that from the group of items in the folder, he is being asked to pick one. In addition, he must infer that there are pictures in the folder. However, the above context for identification and understanding the antecedent in the speaker's utterance, makes this context one of the strongest possible for understanding partitives for a second language learner.

The task itself with the visible presence of the folder, knowledge of the function of folders, and prior experience on earlier interviews all provide a highly familiar context for the utterance. Hence, the partitive can only be considered linked to previously established discourse referents, and therefore necessarily definite and specific. Partitives with indefinite NPs which have the definite NP as antecedent can also be specific.

I\ 3.4.2 Indirect Auditory-Linguistic Co-Presence The source for utterances which are inferred or deduced from the speaking C situation without visual targets are from a type of mutual knowledge which can CL be called indirect auditory-linguistic co-presence. The source of knowledge for Sthese linguistic utterances differs from direct auditory linguistic co-presence which draws solely from the lexico-grammatical mention of antecedents and the 00 i sense of hearing rather than association of properties which have not already been lexico-grammatically mentioned auditorily. Inferrables from the auditory signals are found in my data from three main sources: discourse level mental schemata developed in the introductions; during Task I, the textual festival descriptions during Task II, and inferred properties from pictures of festivals C during Task IV; clause level inferencing of properties (cf. Evoked, Prince, 1981), and phrase level inferencing of properties, i.e. backwards anaphora or N partitives.

3.4.2.1 Anaphora Potential linguistic anaphora is one kind of phrase from which a Hearer must infer the intended antecedent. I have already mentioned these under auditory copresence demonstrative pronouns. An example of this kind of inferencing is: Example 3.47) il4sl0.txt:4 14 ok so that must be ah Omagari fireworks 3.4.3 Indirect Community Knowledge and Auditory Co- Presence Inferencing occurred throughout the discussions. This was especially evident in the use of questions by interviewers.

3.4.3.1 Interrogatives Wh-Questions The first set of questions involved community knowledge about one's life in general and at MSU-A and in the introductions. Introductions normally contain mental schema of getting to know a person, name, hometown, year in school, length of time at a school or hobbies. Although the interviewers were given a list of these questions to ask the learners, they were also instructed that they could ask additional questions. (The intent was to encourage the learners to talk, not follow a rigid set of guidelines which were provided only to facilitate the interview and keep data comparable across interviews.) These examples indicate questions for the mental schema of introductions in Task I.

Example 3.48) il4-11.txt: il4 what's your name? il0-2sl0.txt:l il0 Yoshi, okay, Yoshi, ah, are you, where are you from originally? il0-2sl0.txt:l il0 life, mm, mm, what do you do actually, for example, at the weekend? il0-sl2.txt: il0 Kyoto, okay, um, what kind of Kyoto, Northern part? i3-s2.txt:\txt i3 mm, mm, what's your job?

\O

0

ID

0

(N

oo tq 112-slO.txt:l 112 il2-sl0.txt:l 112 do you have many American friends? are your roommate is American? Some questions were directed at their student life and educational goals.

Example 3.49) 14-1 1.txt: il4 why did you have any special reasons for wanting to go to America? il0-2slO.txt: IilO could you tell me why you like University? ilO-sl 1.txt:l ilO so what do you like other than studying English? il0-sl2.txt:l ilO okay, ah, maybe, what will you study in the years, in the future il4-sl2t.txt:l i 4 why you want to go to US and study there? il6-s 0.txt: 1 i6 do you know which school you would like to study i3-sl.txt:1 i3 do you know where you will study in Minnesota or America? i3-s2.txt: i3 what's you take to make your major, in major? 112-sl0.txt:l il2 'nd how long have you been studying at MSUA The third type of utterances which contains inferrables of a schemata are found in Task II, the description of the learner's hometown festival. This task was designed to elicit structures from learners, but the interviewers frequently interjected questions in order to guide or assist the learners. These questions were inferred from knowledge of properties of festivals, a universal event.

Example 3.50) il2-sl2.txt:2 i12 il2-sl2.txt:2 il2 il2-sl2.txt:2 il2 il4-sl l.txt:2 i14 il4-sl l.txt:2 il4 il0-2sl0.txt:2 il0 what is festival name? Gxx) where does it take place? yes, what happens at, what kind of things (xxx) what part of the festival did you like? the (xx) ok what kind of costume do they wear? could you explain how you, what you like at the kanto festival? The interviewers knew that festivals have names, locations, costumes, and that most people enjoy going to festival.

These utterances asked by the Hearer during Task IV again are linked to knowledge of the properties of festivals. Although the respective properties had never before been introduced into the discourse, they indicate that the speaker presupposed the existence of an answer to the question: Example 3.51) il4-sl 1 .txt:4 il0-s 2.txt:4 i12-s I .txt:4 i12-s I .txt:4 i12-sI O.txt:4 il6-sl0.txt:4 what kind of costume are they wearing? okay, what kind of hole in the sky, what is in the sky what's that snow house, do you know what will they do in Kamagura house? what kind of things do you do up there please explain why The important point about all of these interrogatives is their direct connection to the notion of festivals. They represent the network of properties or IDcharacteristics which describes the event called "festival." In this sense, they are O linked to and inferrable from the event called festival. These properties are N specific to the definite event of festivals. Consequently, following Eny, Heim, a and Kamp, since the festival is the definite event, they can be coindexed to the Snotion of festival.

Additionally, they behave as topics, since they are specific and correlated 00 with definiteness. In other theories of information structure, Wh-question words are considered in "Focus", or new information (cf. Lambrecht, 1994; Erteschik- Shir, 1986; Rochemont, 1986. My interpretation of these inferrables, in contrast, is that they are linked to a discourse topic, in this case 'festivals'. Because of this linking, they can not be considered totally new. They behave, instead, like proper Snouns from the set of community knowledge, which are both given and new at Scoding time.

SMoreover, the question word by itself does not determine the answer to the question. The question word is part of a phrase within the question which determines the appropriate answer. For example, the question What will they do in Kamagura house? requires an action as answer to "what-do", whereas the question What is name? requires an identification of the festival's name. The question word by itself does not provide sufficient information. In addition, the prosody of information questions does not normally contain the primary stress in the wh-word of the question; the question words are given secondary stress while the companion word in the phrase which directs the hearer to the desired information receives primary stress. The prosody, in other words, also does not fit the facts of wh-words as new information correlated with primary stress.

Wh-questions, then, are inferrables from a discourse event which is gradually unfolding in the minds of the speaker and Hearer. They are topics insomuch as they are linked to an already known entity or event. They are also found sentence initially in a position designated for other topics in English, common to many languages.

As I have described the various constructions with their properties, behavior and constraints, I have shown that sometimes the constituents in a context for an utterance are immediately obvious to participants, the case of visually copresent utterances) but other times the context unfolds during the utterance. In these cases, sometimes the antecedent is mentioned first and followed by an anaphor, but in others the anaphor precedes the antecedent. How are the latter understood and comprehended by the Hearer? Why can something be both New and Given simultaneously and why can NEW information precede the given information.

The answer to these questions can be found in a careful study of the context of the utterances. Sometimes the mutually known mental schema in the minds of both interlocutors creates a schema on which the conversation unfolds.

In other cases, the speaker creates the schema as the discourse develops. In still other cases, the lexical connotations of one word lead the Hearer to anticipate additional information, e.g. partitives. So, in the same way that a noun which is an agent leads the Hearer to anticipate a verb which takes an agent, the pronoun "one" leads the Hearer to anticipate the identification of the group to which the specified unique individual belongs. The semantically empty, indefinite lexeme,

IO

0

N

c/ oo indicates to the Hearer that more information, a definite NP, will follow. They behave like potential anaphors with antecedents that follow the anaphor, ie backwards anaphora.

Together these clues give a context for aiding the Hearer in decomposing and hence comprehending the overall utterance. As C&M 25) state regarding backwards pronominal anaphora: "The requirement seems to be not that Bob already have a shared set of objects, but that he be able to form one based on general or particular mutual knowledge and on the fact that the reference act occurred." Following Eng's argument regarding indefinites and specificity, whquestion words have long been known to correlate with indefinite pronouns in languages (cf. Hintikka, 1974). The behavior of wh-question words has been shown to behave in a similar manner to these indefinite pronouns. They stand for a lexically indefinite property of festivals (see Lambrecht, 1998). In this sense they behave like indefinites which are linked to definite descriptions or definites which are already familiar in the discourse.

3.4.4 Mixture of Visual and Auditory Indirect Co-Presence Often inferencing is a combination of information from all the contexts in the speech situation. The following dialogue of one interview shows the combination of the different sources of information used to determine the identity of the festival from the student's description of the picture.

Example 3.52) Discourse i4-s3:4 s3 177 ah hah they walking walking through the mountain i4 178 mmh s3 179 with fire(??) and he, they they wears like demon mask demon mask i4 180 ah hah s3 181 and they wear they wear un ara i4 182 straw s3 183 straw straw cloths i4 184 mm s3 185 and straw boots i4 186 so must be winter festival s3 187 yeah it was winter festival I repeat the table of types of contexts for establishing definiteness in the speech situation.

Table 1. Classification of Contexts for Reference Types (Fellbaum, 1999, modified) Basis for Mutual Knowledge Community membership Visible copresence Linguistic copresence Indirect Copresence Reference Types -Universal, Particular -Deixis: Visual, visuospatial; visuo-temporal -Visual (Reading, Sign Language) -Deixis: Spatio, Temporal; Anaphora -Acoustic -Deixis: Spatio, Temporal; Anaphora -Inferencing -Visible Properties evoked from visual objects -Acoustic Properties evoked from acoustic objects and visible objects IND The four types in Table 1 manifest themselves differently in the four tasks O of the experiment. In Example (3.52) the learner did not provide a discourse referent, so that no clear antecedent is available to the interviewer. However, the learner's acoustic description (Type 3) provides information from his picture that S'they' is doing the 'event' of walking and wearing the 'objects', masks and straw Sclothes and boots. From this i4 can infer that "they' refers to a person, based on 00oO universal knowledge of festivals, people, and clothing.

The fragments in this thesis are limited to the same task with two native speakers of Japanese so that the mutual knowledge, belief and suppositions in the common ground are more closely matched than those with native and non-native speakers. The interviewer can make inferences which are restricted to a particular community of Japanese members. A mixture of types and (4) ,I ~combine to help i4 infer from the acoustic elements of 'straw clothing' and 'boots' in the discourse, that this is a winter festival. In Japan three generations Iago, straw was a standard covering to keep clothing dry when venturing outside in the snow and was then later discarded. This knowledge is culturally determined from universal knowledge based on community membership among Japanese speakers. After listening to the oral description, the inference manifests itself in line 187 of Example (3.52) based on cultural knowledge of the particular event in the discourse.

The interviewer also knows that this is a picture of a festival, since she has just given the student a picture. Although this provides visible information (Type 2) which is copresent in the dyad, the information content of the visible object differs between the participants in the dyad; the mutual copresence of the visible information, is relative to the person in the dyad. The learner has a picture with visuo-spatial features of people in a line, with feet uplifted as in a walking movement; the interviewer, on the other hand, has acoustic-spatial knowledge of this event (Type 3).

By classifying the utterances in the discourse situation into various types of knowledge, we can study more carefully the types of information from the speech context and how they are being used. Thus, the researcher can better explain the phenomena involved in learning.

Four Types of Mutual Knowledge, Pscholinguistic and Linguistic Characteristics Table 3.2 is adapted from Clark and Marshall (1981:43-44). It summarizes the four types of mutual knowledge which lead to definiteness as found in the interlanguage tasks of my experiment. It also gives the psycholinguistic mechanisms associated with each type of definiteness with respect to the context in which they are found.

Following Table 3.2, I briefly describe these mechanisms, based on the data in this chapter. These are simple descriptions, closely matching the empirical description of the data. A uniquely psycholinguistic description of these mechanisms is beyond the scope of this thesis, (see Korpi, 2005).

2006222742 28 Sep 2006 Table 3.2 Four Types of Mutual Knowledge, Mental Resources, Linguistic Expression, and Associated Tasks for Assigning Definiteness Basis for Mutual Knowledge Mental Resources for Assigning Definiteness 1. Community membership Community comembership, Universality of knowledge Linguistic Expression Tasks II, III, IV Generics, Proper Names Generics 2. Visual linguistic co-presence a. Immediate Simultaneity, attention, rationality b. Potential Simultaneity, attention, rationality, locatability c. Prior Simultaneity, attention, rationality, recallability Deixis person, spatial (demonstratives) locative, temporal II, IV 3. Auditory linguistic co-presence Anaphora I, II a. Potential Simultaneity, attention, rationality, locatability, understandability pronouns, definite descriptions b. Prior Simultaneity, attention, rationality, recallability, understandability demonstratives, proper names 4. Indirect co-presence Demonstratives a. Visual b. Auditory I, 1I c. Mixture Simultaneity, attention, rationality, (locatability or recallability), person, spatial, temporal associativity community membership Simultaneity, attention, rationality, (locatability or recallability) person, spatial, temporal understandability, associativity anaphora, community membership III, IV DEFINITIONS of MENTAL RESOURCES for ASSIGNING DEFINITENESS in a CONTEXT 9. Community comembership particularly prominent people, places or events which members of a community can safely assume are known by all Universality of knowledge generic and particular knowledge that both [all] interlocutors can safely assume each other knows 11. Simultaneity interlocutor[s:both, all] looking at[/listening to] each other and target in situation simultaneously 12. Attention both [all] interlocutor[s] attending to [each other and] target copresent in the situation 13. Rationality both [all] interlocutors capable of inferring or drawing the same conclusions from the available mutual knowledge, including meaning of visual and auditory signal 2006222742 28 Sep 2006 14. Locatability-hearer is capable of finding target [incl. abstract auditory in semantic context] and bringing into view or focus [consciousness] simultaneously with speaker Recallability hearer is capable of recalling earlier interaction and targets established between both interlocutors, either from earlier visual focusing or previous mention 16. Understandability hearer is capable of"penetrating indefinite reference" and understanding that its existence is being posited, i.e. its definiteness or specificity 9. Associativity known by the individual[s] in the community that the proposition, entity, or property is certainly, probably, or possibly a particular part of, or in a particular role with the proposition or entity; that the set-subset relationship is an appropriate relationship Task Description: I Introductions; II Description of Hometown Festival; III Information Gap; IV Picture Description of Festi O 3.5.1 Psycholinguistic Correlates of Community and Universal Knowledge The basis for mutual knowledge which is the source for the respective types of knowledge are all found in a context in which the utterances are based on mutual awareness between the speaker(s) and a target in the environment. As far as language 0 is concerned, the speaker is not just looking at a picture and/or listening to a concert.

(N

The speaker(s) is actually creating a linguistic utterance about the target(s). The first two constraints, community co-membership and universality of knowledge, are the ri most enduring basis for mutual knowledge between interlocutors. The knowledge created by a community of speakers explains knowledge derived from lexemes which N are known from generic information and particular information expressed as proper (,i noun. As seen in the chart, utterances which include these lexemes are found in all Ifour tasks; they are universally applicable in conversational interactions.

In addition, the schemas they invoke are not single time-bounded entities. As a child walks through the door at 4:30 a mother may ask "So how did it go?" The child may respond, "Oh, she made it easier than I expected." Both interlocutors know the "it" refers to the test which was given by a particular teacher in school that day. The context for this utterance is created from the common knowledge within the family unit. Hence, the "it" has anaphoric reference to an antecedent created more than eight hours earlier, when the child left for school. The duration of short term memory, which only lasts milliseconds, cannot explain the origin of this antecedent.

Nonetheless, the "it" has an antecedent and the reference is understood by both interlocutors.

This source of mutual knowledge for the establishment of definite reference causes problems for communication in second language learners. As long as the language is being learned in a classroom and students are only asked to discuss topics familiar to them, there is no problem. However, if asked a question which requires knowledge of the culture and generic information, such as the function of a particular cooking utensil restricted to that culture, a student may hesitate or not be able to communicate if the relevant facts are not known. Thus, this constraint on definite reference pre-establishes a need for cultural knowledge in order to be successful.

3.5.2 Universality of Knowledge The second resource, universality of knowledge, also exceeds the limitations of short term memory. As explained earlier, individual languages differ in the membership of individuals which are given proper noun status. In addition, languages differ with respect to the coding of proper nouns. The constraints on inclusion as to proper nouns status in a language, plant names or gods, as well as the coding is language specific, based on the prominence given to these entities in the respective culture.

The constraints are determined with respect to mutual knowledge within the community of speakers and are not based on the metalinguistic category of nouns. In second language learning, the learners must be familiar not only with the particular members of a community, then, but also the linguistic coding devices of a specific language. This is one of the reasons against transcribing known proper nouns of the target language in capital letters in a transcription, until it can be ascertained that learners do in fact recognize the uniqueness of the particular entity.

I 3.4.3 Simultaneity

O

SSimultaneity requires that the target of the utterance, either visual, auditory, or inferred, be in view or auditorally or mentally present at coding time. If two people are present, then either visual or auditory targets must be simultaneously copresent for both interlocutors. Alternatively, only one person may be reacting to a visual target which is copresent simultaneous to coding time. Or, if one person is narrating a story, 0O for example, the narrator is simultaneously listening to the words and making mental constructs from memory at coding time, rather than interacting with another interlocutor. The necessary requirement of this constraint is that an interlocutor is Cinteracting with a target object about which the utterance refers. Positing a model of speech interaction in one person's head, in addition to a second interlocutor's, the rHearer, makes this inclusion possible. A mental target in the environment of the speaker must exist at coding time.

3.5.4 Attention N The fourth constraint, attention, refers specifically to the entity, proposition, or event which has become activated in the minds of the interlocutor Traditionally, this is the constraint which defines the entity, such as pronouns, in an utterance as topic because the entity has been previously mentioned. However, this description of topic assumes auditory or linguistic use in texts (oral or written). My analysis has a broader basis which includes multiple sources of knowledge from which a topic can be derived in an utterance, i.e. one's attention may be drawn to an entity from visual or spatial sources in the discourse. Thus, attention can be delineated as a technical term from multiple sources in a situation for identifying topics. Some attentional channels originate when activated by a stationary visual target, some by moving visual targets, some by spatial relations between visual targets, auditory stimulation, and still others originate by inferencing. Thus, for example, the utterance "You give me that book!" has three activated lexemes in the auditory message-you, me and that book. All can be regarded as entities which have been activated in the visual field, but relayed auditorally.

Although I am using the word attention to identify topic(s), it is not to be understood as equivalent to the term attention as used in the attention system of the human brain. The neurophysiological system is far more complex than I have explained, and that this set of data and its respective analyses permit.

4 I am analysing definiteness and specificity with respect to topics. In both cases, the attention system is involved. I am only discussing the elements which are already held in memory, since they are or have been made definite. Either the entities have been attended to, and/or they have been stored in long-term memory and have been accessed, so are currently being stored in short term memory at coding time.

In addition, this analysis is a first attempt at studying the behavior of definiteness and specificity beyond the auditory constraints found in traditional studies of topics, anaphora, and information structure. To the extent that the constraint of attention and definiteness as described here shows that it can be ascribed to visual, visuo-spatial, and temporal definiteness, as well as verbal definiteness, it is consistent with our current understanding of the cognitive functioning of attention, a network of anatomical areas from the entire sensory-motor system that performs different but interrelated functions, nonetheless maintaining its own identity like other sensory and motor systems (Posner and Petersen, 1990).

I 3.5.5 Rationality

O

The fifth constraint, rationality, requires that both interlocutors have the same N knowledge required for identification, associating properties with the target and drawing the same conclusions, or making similar inferences of an entity. This is especially important in the second language situation where non-native speakers may not understand the purpose of a target, e.g. Christmas tree stand, or even be able to identify the meaning of a target, either visually or auditorally. If mutual knowledge of either the visually present or spoken target is unknown, felicitous inferencing is not possible. In the experiment, the notion of festival was considered a universal event which could be rationally understood by both interlocutors. Hence, properties of festivals could be associated with the pictures and information gap charts. However, even though the generic properties associated with the event of a festival are universal, the particular properties of a specific festival from another culture may be Icompletely foreign to the learner.

3.5.6 Locatability The next constraint for direct definite reference to be completed between interlocutors is locatability. This requirement applies to both the visual and auditory target. The visual target must be copresent in the immediate context so that both interlocutors' attention may be drawn to it. Alternatively, both interlocutors must be able to establish the existence of a target in the discourse in order to be simultaneously conscious of its participation in the utterance. In other words, both interlocutors must have knowledge of the existence of the target in the context in order to be conscious of its presence in the discourse.

3.5.7 Recallability The seventh constraint, recallability, is traditionally described as "previously mentioned" and is required for topichood which I have reanalyzed as auditory targets.

The term recallability" is preferred, since I have also included visual co-presence as a source for topics. Recallability, then, can be applied to visual targets which have been viewed by participants during the discourse and are thus capable of being reintroduced.

3.5.8 Understandability Understandability, the eighth constraint, requires knowledge on the part of both interlocutors that a target which has not been previously identified, i.e. an indefinite and is therefore unfamiliar in the discourse, is being posited as existing. Thus, the entity posited is constrained so that it is capable of being posited in the respective context. As a result, its definiteness or specificity is felicitous either in the present context, where an entity is both given and new, or in its next occurrence. The constraints on backwards anaphora are one example of this constraint, since the hearer must be able to identify the antecedent from context after hearing the anaphor. Proper nouns must be known and recognized within the community in order to be felicitous in the context of an utterance.

3.5.9 Associativity Finally, associativity covers those interactions which require inferencing of any type in order for the discourse to continue. This constrains the type of propositions, entities, properties, and events which can be associated with targets. So, for example, D indefinite reference can only be associated with a discourse context in which it can 0 properly be contained in order to become a specific member. The set and the 0members of the set must be capable of being associated. Other researchers have referred to this as 'bridging' (Clark, 1977), or inferencing. Once the association has occurred (at coding time), the item is either made definite or specific and hence included in the set of definiteness.

0In Chapter 5, I analyse data according to the properties of predicate-argument structure and the behavior of semantic roles especially with respect to certain tasks found in the data, e.g. the imperative construction. In this case, the semantic i properties of agency and volitionality constrain the choice of verb and the possibility of omission of the subject. Thus, it was the semantic properties and context of N predicate-argument structure which constrained the behavior of these constructions.

i 3.6 Conclusion NIn this chapter I have described various contexts in the real world and showed how 0 language behaves with each respective situation. In particular, the real world knowledge must be known and available simultaneous to the speaker's utterance. In addition, I have limited the knowledge to only definite perception of targets currently being attended to. These constraints operate in four main contexts: General knowledge from the wider set of members in a community and the world from which particulars and properties of can be derived, i.e. proper nouns and generics; the visual context which includes either stationary or moving elements in the environment, as well as identification of spatial relationships between entities in the visual environment. Gestures are also included in this context.

the auditory targets produced between interlocutors and the linguistic network; the wider context of sets and the members which can be inferred and understood from all of the above environments. This includes mental imagery, such as the pictures formed in the interviewers' minds as they listened to the learners describe their visible picture, as well as the properties of generic sets, partitives, and other relations in which the anaphor precedes the antecedent.

C&M's taxonomy of four types of mutual knowledge and their connections to direct definite reference can be expanded to serve as the basis for explaining properties of nouns, verbs, adverbs, and adjectives and their behavior with respect to mutual knowledge. The constraints on their behavior all depend on either the speaker or the hearer or both participant's mutual knowledge in the verbal interaction. They are independent of and not constrained by subject-predicate relations of verbs and semantic roles, as explained and analyzed in Chapter 5. Their behavior is constrained by interactions of mutual knowledge between the speaker and Hearer or speaker- Hearer.

I have analyzed my data according to the categories found in natural languages. In some cases, English uses the same morpheme for several types of copresence. Therefore, the types of mutual knowledge is obscured. However, I have D nonetheless analyzed the data according to what is known from natural language since some languages do distinguish all of these categories morphologically. Thus, it is a necessary generalization in describing the English language to state that it does not make morphosyntactic or lexical distinctions in these types of language. Possibly it distinguishes them through the use of stress. This needs further examination.

VThe above analyses have shown that at coding time, the visible and non-visible 0linguistic expressions contain similar constraints. Hence it is not surprising that the (Ni same 'form' is sometimes used between the two types of utterances, i.e. visible and non-visible. Demonstrative and locative referring expressions have similar forms and (Ni constraints and hence operate similarly. The pronoun 'he' can also be deictic and non-deictic. The only forms which are restricted to the visible are the first and second person pronouns, although these can be non-visible over the telephone.

As linguists it is useful to study the behavior and constraints of both in order to establish the similarities and differences, since they do differ crosslinguistically.

OThe important point for my thesis is to demonstrate that they obey constraints imposed on their behavior by real-world knowledge and establish a systematic set of constraints. Thus, we can prove that they obey constraints independent of predicateargument structure and this needs to be acknowledged and represented in a grammar.

Careful study of this system may also provide insight into historical change in which it could be explained that the system itself allows the changes noticed between topics of demonstrative pronouns, and other forms in the IFS. It is helpful for second language acquisition to study these constraints and behavior not only for establishing theories of acquisition but also to lead to a better understanding of what the behaviors are in order to establish a more cohesive and direct pedagogical theory.

Most especially it is interesting that the constraints on production of sentences are organized around the same behavior and limitations of comprehension, especially working memory. The pre-frontal cortex contains visuo-spatio-temporo-acoustic centers which have been proven to function in comprehension and working memory.

It appears that these same types of attentional 'channels'/'loci' are operating in processing linguistic utterances constrained at the time of production. If we can step aside from the notion of mental representation, which links an entity to working or short term memory, and instead view topics as deriving from a behavior of drawing one's attention to an entity, the communicative act becomes part of a dynamic process involving both comprehension and production in each utterance. In other words, comprehension of a linguistically produced acoustic target a word or lexeme), requires storing it in memory in order to use it as a topic in the next utterance produced. Speech production incorporates the comprehended entities of previous utterances. The two skills cannot be truncated in the study of topic and information structure.

Furthermore, topics currently described as mental representations by some, can become a more active and integrated part of the study of communicative interactions in information structure with this view. Even if we don't consider neurophysiological sources of attention, the utterances as described and the definitions proposed in the above discussion of the bases for mutual knowledge and context in which they are found, all point to the possibility of more than one topic per utterance defined and constrained by attention. The topics are constrained by comprehension and the new information, which I leave for the next chapter, is the newly formed concepts of new or focused information.

Chapter 4 aPatterns of Meaning Through Specificity Relations 00 4.0 Introduction The complexity of definiteness and specificity and their interaction with subject and topic have long been noticed as a part of grammar. Two researchers, Keenan (1976) N' and Chafe (1976), include the notion of definiteness in their discussions of topic and Ssubject. They acknowledge that both topics and subjects may be definite. Keenan (p.

CK1 319) points out that definiteness does not distinguish topics from subjects while N' regarding it as a property necessary for describing the characteristics of topics, and IDalso being a characteristics of subjects. Instead of including definiteness in his list of O properties for subjects, he lists presupposed reference, topic, and "highly referential" as separate properties of subject. Under highly referential, Keenan mentions that subjects may be both indefinite or definite. Chafe (pp. 39-53) also recognizes definiteness as important for topics, but he does not claim that it is a necessary property of topics, since some indefinites, such as the indefinite pronoun one, are 'given', a property of topics (p.42).

Li Thompson (1976) (henceforth on the other hand, clearly list definiteness as a 'necessary' property of topics: "one of the primary characteristics of topics, then, is that they must be definite" 461). Furthermore, they state that "the functional role of the topic as setting the framework within which the predication holds precludes the possibility of an indefinite topic" 464). Although L&T agree that subjects may or may not be definite, the impossibility of indefinite topics is a direct contradiction of Chafe upon whom they base their notion of topic and definiteness.

None of these scholars develop the notion of definiteness as a basic unit within a separate level of grammar. L&T (1976) proposed a typology among languages "based on the grammatical relations subject-predicate and topic-comment" 459).

In spite of declaring at the outset of their proposed topic prominent typology that "one of the primary characteristics of topics, then, is that they must be definite (L&T 1976:461), definiteness is not included in L&T's criteria for determining basic sentences in such a topic prominent typology: i) a sentence A is more basic than a sentence B if, and only if, the syntactic form and the meaning of B are understood as a function of those of A. the form of B is some modification of [possibly addition to] that of A, and the meaning of B is some modification of that of A.) ii) a sentence is a basic sentence in L if, and only if, no other sentence of L is more basic than it (L&T 1976:471-472).

Their b-sentence for determining the behavior of topics is defined according to "syntactic form and meaning" [boldface and italics are mine]. A b-sentence is a syntactic structure and the meaning of sentences is a function of the syntactic form.

The role of definiteness as a defining property of basic sentences is not mentioned.

O Nor does definiteness enter into their subsequent discussion of the syntactic behavior O of topics.

SSince definiteness is a property of both subjects and topics, and yet both a subjects and topics have other properties which can be used to differentiate their Sbehavior, definiteness seems to be part of a different level in the grammar that interacts with other components which together differentiate subjects and topics. In 00 English, topic can be identified in a component of information structure (Gundel, 1977, 2003; Lambrecht, 1994; Vallduvi, 1992) and subjects in a level of grammatical functions (Manning, 1996; Bresnan, 2001). Notionally, subject is often considered to ri be a default topic (Huddleston, 1984:58; Bresnan 2001:98, and fn 15, p. 123).

Traditionally, definiteness has been studied from a semantic point of view by c- philosophers (Russell, 1905; Strawson, 1950). Recent theories from computational C linguistics, such as Kamp's Discourse Representation Theory (DRT) and Heim's File (Ni Change Semantics (FCS), have expanded the study of definiteness to include 0additional issues, such as general uses of pronouns including those without linguistic Santecedents when their referent is understood to be salient to the hearer. v In linguistics, the notions of definiteness and their interaction with other components of the grammar are labeled +Def in the syntactic analysis, or in the Fstructure (functional structure) in theories of grammar, such as LFG (Bresnan, 1982, 1988, 2001; Arka, 1998). Definiteness is still considered only a property or feature, not a defining characteristic of a separate level within a grammar. It is not treated as a fundamental (unifying) unit of referentiality in and of itself in the grammatical system.

In this chapter I explore an entirely separate system of grammar which can be defined exclusively with the properties of specificity and nonspecificity. Kamp and Helm's theories can be used to help define and operationalize the notions of specificity. To their conceptualization of definiteness, which includes auditory definiteness, I add auditory, spatial and visible specificity. In addition, I use Farkas' notion of specificity to further distinguish specific and non-specific properties (Farkas, 2002).

The arrangement of patterns of specificity and their semantic patterns of meaning clarifies their behavior and functions in both propositions and discourse. As shown in the preceding chapter, definiteness and specificity exist in the minds of the speakers. Its construal is totally due to mental instantiation of objects, both visual and auditory, between the speakers. The Hearer knows that something is being postulated based on the context which includes auditory, visible, visual, temporal and background information of the two persons in a dyad. In this chapter I show the clear presence of Existential and Characterizational patterns of meaning development in Task IV by one learner.

4.0.1 Propositions and the Context of Discourse Definiteness and specificity are relative to their position in the context of the discourse in which they are situated. Discourse has many different definitions and applications. I use Partee's definition of a discourse as used in Heim and Kamp's theories for analysis of reference. In these theories, "a discourse is simply a finite sequence of sentences" (Partee, 1984:248) used to exemplify a particular phenomena in a discourse. However, rather than a discourse composed of sentences, a syntactic unit, I use the notion of proposition to emphasize meaning composed of semantic properties.

I

IO Proposition has been used differently by logicians, psychologists, and O linguists. v Logicians are concerned with truth and satisfaction of syntactic structures.

Psychologists are concerned with the psychological representation of meaning reality in syntactic structures, although they acknowledge that propositions are different from Ssentences (Clark, 1977; Kintsch, 1974)'. The linguist Tomlin (1987:461) builds on V) the work of Kintsch, but adds the notion of a truth value: "proposition is defined here 00 as a semantic unit composed of a predicate plus its arguments for which a truth value can be determined.

Modem theories of meaning in the relatively new scientific fields of systematic semantics and predicate logic developed by logicians, such as C.S. Pierce and Frege, redefined the notion of logical form and the proposition. Their work lead Sto a more rigorous notion of the complexities of propositional structure Kamp and Reyle (1993:14-16) in the field of computational linguists (Heusinger, 2000), which Sstudies the phenomenon of the interpretation of truth-conditions, inference, and 0reference. In this tradition, the proposition is part of a context created in a more complex discourse structure. However, these theories still look at the syntactic structure of propositions as defined in predicate logic. Truth is defined relative to its position in these structures. If an element exists in "the real world" or is true in a possible world, the proposition is true, if the element does not or cannot exist in one of these worlds, it is false (see Chapter 8).

In this chapter, I take a different approach to determining meaning in a proposition from any of the above. When one moves to the level of intersentential propositions within a discourse, syntactic categories, such as sentence, clause, predicate, argument and infinitive, are too narrow. This narrowness limits the applicability of syntax in a discourse. At the level of the proposition, relations between individuals, between individuals and their properties, and between individuals and the denotation of their names can also be captured through definiteness and specificity relations.

In other words, rather than discussing categorial properties, such as noun and adjective, or information structure categories, such as topic or focus, I describe propositions according to the interaction of properties of specificityv"' and the semantic relations they create. This creates an independent level whose internal relationships can be defined and described wholly within a semantic level of definiteness and specificity, independently of properties of syntax and predicateargument structure.

4.0.2 Definiteness Properties of definiteness are created according to human interaction with the real world. As illustrated in the previous chapter, definiteness may be created through strictly oral interaction, as in many conversations, or it may consist of strictly visible interaction, when one person is reading a text or two participants are looking at a picture or a map, or in sign language, for example. Kamp's Discourse Representation Theory (DRT) places an importance on description and context of definite descriptions with no specific distinction of aural or visible input, as explained in Chapter 8.

Modern linguistic theories of grammar have concentrated on the sentence level in grammars. The recent advances made by semanticists in connecting sentence properties also distinguish types of definiteness, such as nominal and temporal anaphora. My emphasis is on the empirical validity and representational power of the

I

IDtheories rather than the theoretical issues. Although theoretical questions are O nonetheless involved in the representations themselves, I concentrate on describing my data using modifications of their methodology and formalism as presented in a Chapter 8.

SModifications to Kamp's Discourse Representation Structures (DRSs, Chapter S8) represent the influence of the visible context in the developing discourse. Since 00 Kamp's DRT is one attempt to unify treatments of deictic, discourse and bound anaphora (Partee, 1989), it is easily modified to accommodate the visual knowledge from the pictures and gestures as the discourse entities are entered into the interlanguage discourse.

File Change Semantics developed by Heim characterizes definiteness and indefiniteness in terms of co-indexing: a definite NP has an index that refers to a ri previously introduced, orfamiliar NP, while the index for an indefinite NP represents (,i Sa newly introduced, or novel NP into the discourse. The first index determines the definiteness of the NP; names, pronouns and definite descriptions are definite and specific due to inclusion 1991:9)'x. Once an NP is entered into a discourse, it becomes an antecedent for a referring expression pronoun). Definiteness involves a strong link, that of identity of reference, to an already established discourse antecedent. Enq calls the antecedents of a definite NP a "strong antecedent" (cf.

Milsark, 1977) strong determiners). Certain nominals are inherently definite: proper names, demonstratives, and pronouns. However, definite descriptions need to be linked to another NP as antecedent in the discourse.

Because my data relies on proper nouns for the identity of the name of a Japanese festival in the pictures in Task IV, the identification of the definite description and of a proper name are critical to completing the task. Kamp argues against Kripke whose theory seemed to imply that no descriptive condition can be imposed (Kamp, 1993:247). Additionally, he argues against Russell and Strawson who maintained that a definite description is true of only one thing. Instead, Kamp argues that both the descriptive content and information from the context together combine to provide the uniqueness of a referent (pp. 252-253). I adopt Kamp's marking of proper nouns as externally anchored and used in predicate logic and DRT.

This anchoring is meant to indicate the function of an entity within a specific discourse context as well as the context for community membership, as explained in the previous chapter. My data strongly support Kamps' arguments, but nonetheless indicate that further modifications of definite descriptions are necessary for a more complete description. Semantic patterns of meaning within the description can be further delineated, as shown below.

4.0.3 Specificity Neither Kamp nor Heim explore specificity in the early versions of their theory. Eny, (1991) develops Heim's notion of definiteness to accommodate specificity. In addition to the original definite reference index there is a second specificity index, representing a set that the discourse referent is to be chosen from. This means that "all NPs carry a pair of indices" Thus the specificity of the NP is constrained by its linkage indicated by the second index. If both indices are indefinite and not discourse-linked, the NP is indefinite and non-specific. But if the first index is indefinite while the second is linked to a definite NP, the NP is specific. The second index represents the specificity relation. If an NP is indefinite but nonetheless includes a discourse referent which is linked to a strong antecedent, it is specific. The O antecedents of specific NPs are called "weak antecedents", a modification of O Milsark's insight into the behavior of"weak determiners". However, this is still an Sindex on an entity at the level of syntax, whether it is in the phrase structure or on a constituent F structure.

SEn9's characterization of discourse-linking to describe specificity works well V)in my data for linking indefinites to definites. It accounts for Chafe's observation that 00 some indefinites are 'given'. However, two problems with Eng's definition for my analysis are: proper names are both definite and specific at the time of mention; therefore, no discourse-linking is necessary answers to questions may be linked to r a wh-question word, which by definition is indefinite. The first index on the answer cannot obtain a reading from a definite NP so determination of the specificity of the Ssecond index of the NP cannot be determined. A possible solution to this will be C given in the first question-answer pairs discussion

INO

4.0.4 Types of Semantic Definiteness Propositions SPropositions in my data can be subdivided into three basic types for semantic meaning. My formulation of these three categories is inspired by the work of Eng (1991) on specificity in existential-there sentences and the discussion of the copula in Thai identificational and characterizational sentences by Kuno and Wongkhomthong, (1981). The definitions and properties of these three categories are then based on modifications to Heim and Kamps' theories of familiarity in definiteness and discourse-linking. The semantic relations and function of these proposition types in the semantic network of a discourse context or situation create three patterns of meaning. These three patterns are summarized below.

Semantic Proposition Types 1. Existential Type where the semantic relation of the proposition is to introduce an entity or express a proposition which is linked neither to an existing discourse entity, or proposition, i.e. a discourse referent. Therefore, the semantic structure contains at least one entity, visible or acoustic, which is newly introduced and nonspecific.

2. Identificational Type where the referring expression in the proposition relates two entities, visible or acoustic, to the same already existing referent. This creates an identity relation between both referents in the proposition, with either an auditory referent in the discourse or visible antecedent external to the proposition.

Thus, the identificational relation is between a definite specific entity and another specific, entity in the proposition and discourse.

3. Characterizational Type where the referring expression further describes, or denotes a property or characteristic, visible or acoustic, of an already established discourse referent.

These three category types are based on patterns of meaning of each proposition within a discourse. The goal of this chapter is to first describe the patterns of meaning at the level of the proposition. Simultaneously this is creating a discourse, with the same patterns of meaning. I analyze propositions exclusively through the notions of specificity within the interlanguage.

The analysis uses empirical data from my study found in the contexts of visual and auditory language to establish the three basic semantic patterns in English. I will give brief statements about differences in the learning situation with respect to \O specificity for Japanese learners, but these are neither comprehensive nor inclusive to the learner situations.

SDefiniteness and specificity are relative to their position in the context in which they are situated. Syntactically, the context may be at the clause level or the Sinterclausal level. In this chapter, I show that a pattern of meaning created by ,1 specificity/nonspecificity forms patterns of meaning underlying all propositions at a 00 conceptual level in the grammar. This differs from other approaches to grammatical relations which only label definiteness and not specificity in the syntax'"' and which do not associate the features with semantic meaning.

oC In I investigate the patterns of meaning produced in Task IV over a three week period by one learner. The patterns change from a predominance of Existential- 04t there Propositions to no Existential-there Propositions and only Characterizational at rC the end of the three week period. In I investigate the patterns of meaning \O displayed Task. III. The patterns include all three patterns, although Existential-there 0predominates. No significant change in the patterns occurs.

S4.1 proposition types in task iv by s3 Section 4.1 examines a three week period of data using Task IV, a picture description of festivals in the Tohoku region of northeastern Japan; the learner describes the picture and the interviewer guesses its identity. Task IV provides excellent evidence of the development of the semantic relations of Existential and Characterizational Types by learner s3. For the learner, the referents are visible, being derived from the nonlinguistic context of the picture of a festival, although he also has access to the auditory input of the discourse. The learner selected is s3 because of his clear use of the proposition types during the three week period of data analyzed, starting with the second week through the fourth week of interviews'"' in the experiment.

The interviewer, on the other hand, has primarily auditory linguistic objects except for knowledge that the learner has a picture.

4.1.1 Week 1 Proposition Types in Task IV by s3 Section 4.1 .1 examines the first of a three week period of data analyzed using Task IV by s3. The propositions produced in the dyad are given below. The propositions by the learner, s3, are in boldface and numbered to the right.

Example 4.1) Discourse i2-s3: 4 (i interviewer, s student (learner), :4 Task IV) i2: I have some picture of festival so you can pick up one and don't show me s3: uh huh...

i2: And describe.it (change tape) (major disruption) (cont.) and I will try to guess which festival) s3: There are... lots of people on the street' (1) i2: uh huh s3: There are many lights (2) i2: many light? s3: many alot of lights... on the street. s3 Around.uh...around... the light there are.. lots of building (3) i2: uh huh s3: And center of the street i2: uh huh s3: there are...many trees... around the street (4) i2: there are any people? s3: Yes... (looking at picture) s3 some people has got light... (6) Ss3: light is light has....on the stick (looks at i2) (7) i2: uh huh s3: near the light there is a kind of car (8) 00 i2: card? C s3: car...on taiko 8 )'xiv i2: Uh huh..uh huh (long pause) s3: That's all. (9) s3: Can you guess? C i2: Hhm. I think... it is.. Kanto festival I s3: Yes. (11) i2: Okay. We gonna finish. Thank you.

Ss3: Thank you so much

O

C The learner's description is between propositions 1-9, although he produced a total of eleven propositions.

4.1.1.1 Existential Semantic Proposition Type The semantic relation of existence has several manifestations in natural language.

Two types of the Existential propositions in English are given below: Example 4.2) English: a. There are two books on the table.

b. Two books are on the table. (Kuno 1971: 333) In English, the lexico-semantic structure of the lexemes, there-is/are, can be thought of as automatically signalling the interpretation of the semantic relation of 'existence' in the hearer's mind. Since this is the easiest existential proposition to identify in the interlanguage data, I restrict the term "existential" relation to there be +NP in this thesis, the Existential-there Proposition Type.

The Japanese equivalent to the English existential-there relation is given by Kuno (1971) in the following example: Example 4.3) Japanese: a. Teiburu no ue ni koppu ga aru.

table 's top on cup exist b. Koppu ga teiburu no ue ni aru.

cup table 's top on aru (Kuno 1971:334) As shown, Japanese does not contain an expletive equivalent of 'there is'; only the lexeme aru meaning 'exist' is used. Although the analysis of the function of the ga morpheme is highly controversial and seems to signify several properties (see Kuroda, 1972-73; Kuno, 1972, 1973,1978; Shibatani, 1977, 1990), it does not have a straightforward "existential" meaning for the morpheme 'koppu' or 'cup', as in the English structure. So, a Japanese learner of English must learn the semantic meaning INDand usage of both lexemes, i.e. there and is (=exist), for the semantic structure of existence, there is +NP.

SThis learner makes extensive use of there be NP propositions while describing the picture he is holding. Five propositioms assert the meaning of Sexistence, 1, 2, 3, 4, and 8. Proposition 5 can also be considered an assertion of existence, because in the context of the preceding utterance by the interviewer, it 00 (re)confirms the existence of people.

Of the eight entities, people, street, lights, building, trees, stick, car, and taiko, only two, stick and taiko, are not introduced into the discourse with 'there be K Thus, the learner appears to have learned the lexical or morphosyntactic changes which trigger the semantic relation of existence in English.

4.1.1.1.1 Non-specific 'Weak' Determiners IDWhat is important semantically in existential propositions is not the morpho-syntactic Scoding, but rather the assertion of the existence of a new entity or individual into a text, written or oral. A review of the semantic characteristics which they code is useful to develop an understanding of the semantic relationships of the properties of existential propositions.

Extensive research on the definiteness properties of entities in the postcopular position of Existential-there sentences has shown that the usual form is an indefinite entity in this constructionxv, or Definiteness Effect (Ward Birner, 1995; Prince, 1978, 1992; Milsark, 1974; Hawkins, 1978; Reuland, 1987; Keenan, 1987).

En9 (1991:16) argues further that it is the specific use of indefinites which is not found in Existential-there propositions, and the predicted characteristic is more appropriately described as nonspecific. In fact, En suggests the term definiteness effect is inappropriate and misleading and should be replaced by the term 'specificity effect'. This feature of a 'there [-specific] postcopular, individual distinguishes the semantic relations of this proposition type from the other two types of semantic structures, Identificational and Characterizational Milsark (1974, 1977) divides determiners into two classes according to whether they may occur in the immediately postcopular NP in a there-sentence in English and labels the determiners which introduce indefinite, nonspecific entities in existential sentences 'weak determiners': singular and plural indefinite articles, number determiners, many, few, several, and no (cited in Comorovski, 1995:145) One weak determiner the learner uses is an indefinite article in the phrase a kind of car. The indefinite article a has two functions: one to denote nonspecificity and secondly to denote singularity. It is not clear in this description from just one token of an article in existential-there propositions, that the learner understands either its use as an indefinite article or its use for denoting singular entities as opposed to plurals.

It is probable that the learner is still in a state of variation because he uses no article in Proposition 6 before light, used as a count noun, and then immediately uses the definite article for first mention in Proposition 7, the stick. Nonetheless, it must still be noted that the learner produced this proposition in a context of an existentialthere proposition.

A second weak determiner found in the learner's data is a type of number determiner. In the first proposition, lots of is functioning similarly to a cardinal number, i.e. asserting that a set with lots of people exists."" As Kamp explains, cardinal numbers say something about the size of the set, "whether it consists of no

L

IN elements, of one element, of two elements, of more than two elements, etc." (Kamp, O 1993:454).

(Ni A third weak determiner is many, which occurs with variations of a lot of of three times. There is no evidence to suggest that the learner can distinguish many from a lot of When they co-occur, introduced simultaneously to modify the same Sreferent in Proposition I am treating these as an emphatic or simply as a redundant 00 usage for a nonspecific quantifier. Since the two are introduced adjacent to each other, it is clear the learner has not acquired knowledge of their usage in the target language English which constrains their distribution to only one prenominally. We C can hypothesize that either the learner is uncertain of their use as single morphemes, t- or is still experimenting with their use in the interlanguage.

eC\ However, the learning situation for a Japanese learner of English may not be strictly one of cardinality, word order and constraints on determiners. In a study of \D quantifiers and syntactic skewing in Japanese, Moore (1999) argues that the use of 'takusan', meaning 'many' or 'a lot', can act as an adjective or adverb and differs in syntactic and semantic usage from English. In particular, the split between definiteness and specificity pre- and post-nominally is skewed with respect to meaning and syntactic placement. The placement with respect to the noun being quantified determines definiteness and specificity relations 93). If the quantifier precedes the noun, it is definite in Japanese; if it follows it is indefinite (pp. 96-97).

Indefiniteness in the Japanese Existential construction is indicated with an NPexternal construction which means it is postnominal and postparticle. This position cannot express definiteness 99). More confusingly for a Japanese learner of English, she argues that in the existential-there construction, the speaker in Japanese uses the prenominal form for specificity. Hence a Japanese learner must learn that exactly the opposite meaning obtains in English 99) However, Moore's use of specificity is based on "what the speaker has in mind" (Givon, 1978, 1993), differing from my use where a specific individual is anchored to both the speaker and hearer in the discourse. Thus, a direct comparison or explanation of apparent difficulties can not be made pending further research.x"'"v Finally, the first proposition in s3's speech includes the phrase lots of in the proposition, There are lots of people on the street. Celce-Murcia (1998:326) describes partitives as "constructions denoting a part of a whole". The examples they give show that they are created in English by an NP argument in an of-of-phrase preceded by a numeral determiner two of the people). Since the partitive phrase may be either countable or noncountable, they can be used to quantify count or noncount nouns. Furthermore, partitives can 'modify' nonspecific nouns and specific nouns: I need a deck of cards to show you my new magic trick (any deck-nonspecific); Will a deck of these cards do? 326). Although 'lots a' functions primarily to designate a large amount, 'lots of has a partitive sense, specifying a specific group of people. Its use is thus quite useful for a learner who may not have mastered the distinction of count versus noncount nouns in English. This is especially confusing for a learner with a first language like Japanese which does not make the distinction.

As noted by Milsark (1974), partitive NPs are not usually found in theresentences. Keenan (1987) calls strings like some of the 'complex determiners'. Since complex determiners include the definite article, he argues that they are not existential and therefore not found in existential sentences. Enc (1991) also accepts that partitives are impossible in existential-there sentences, but does not accept Keenan's analysis, pointing out that his analysis of partitives as a "combination of two O determiners with of is syntactically unmotivated", because the string some of the O analyzed as a complex determiner does not form a constituent 14).

SLadusaw (1982) argues that a partitive NP consists of a determiner and an NP argument in the partitive of-phrase which must be group denoting. Thus, the argument of the determiner must be a set which is already part of the common ground, either Sbecause it has been mentioned or the speaker "has in mind" (see Comorovski 00 1995:147). In addition, Ladusaw (1982) argues that these NPs must be interpreted as specific.xx Using the behavior of partitives in there-sentences and the weak determiners in r Milsark's classification, Comorovski (1995:145-177) argues convincingly for the importance of the NP in distinguishing the two uses of existential and presentative partitives; the strength of the partitive phrase is more properly viewed as a property of N NPs rather than of determiners (Comorovski, 1995:146). This is contrary to Milsark, and Eng who note that partitives are not usually found in existential-there sentences. It is counter to Jackendoffs Partitive Constraint which says partitives must be definite.

SEng agrees with Milsark and Jackendoff, but argues that partitives are not found in there-sentences because they are nonspecific; they are not discourse-linked to an existing set, or definite entity. Hence, they do not have an existential meaning and one would not predict their occurrence in existential propositions.

The first proposition also includes a definite article the in the proposition, There are lots of people on the street. This use of a definite article is consistent with Jackendoff(1977) who Postulated the Partitive Constraint which says partitives must be definite. However, it is not immediately obvious why the learner refers to the street as definite.' There are several options as to why the learner uses the definite article in this small discourse, and we do not have enough controlled evidence to study his exact usage of the definite article, especially at the propositional level.i" Again, I only note its usage.

The other two uses of potential partitives, many a lot of lights (Proposition 2), and lots of buildings (Proposition 3) appear phonetically to be functioning as a single morpheme in the interlanguage. So, I have not analyzed these propositions as partitives in this construction. The cardinal quantifier and partitive construction are not clearly distinguishable in function or distribution. Instead, they appear to be behaving as a structural unit introducing a quantity of the introduced variables.

4.1.1.1.2 Properties When we sort individuals and their characteristics at a semantic level, the characteristics of objects in a situation are closely intertwined with the context of the situation. Is a red car, a car that is red, categorized as a type of adjectival phrases Or is it an object in and of itself, a specific member of a group of cars and it happens to be red not blue.""' How do we sort objects, individuals and properties according to definiteness and specificity? Do these have any correlation to the state of affairs or type of happening or eventuality? Fellbaum (2001) compared Aristotle's notion of the classification of predicates (KctrT7yop'ct) (Aristotle 'The Categories' Kneale and Kneale (1962:23) to modem day notions of predicates (Celce-Murcia, 1998; Collins-CoBuild Dictionary 1990).

Although it is not entirely clear exactly what Aristotle means in a technical sense with this list of categories, Kneale argues that we can assume that this is the classification of 'things', "whether these terms occupy subject or predicate positions in sentences' (p.29).

x My interest here is not with the specific details and controversies sO surrounding categorical usage, but rather Aristotle's classification to simply O exemplify one set of properties which can be used at a semantic level to distinguish different types of 'THINGS' in my data.'xx"v Aristotle's list can be divided into the following ten classes of properties at a semantic level: 1. Substance (ovtca) 'a man' or 'an animal' 00 2. Quantity (inocov) 'magnitude of a cubit' 3. Quality (7otov) 'white' 4. Relation (irpo t) 'same as' Place (7nov) 'street' 6. Time (7noT) '2:00 today'

C

7. Situation (KetOCat) 'event' C 8. State (eXEtv) 'is' O 9. Action (7rot6tv) 'bring' Passion (ntc tyoXv) 'love' In the semantics of existential relations, properties may be associated with an entity at the time they are first introduced into a discourse. During Week 1, Propositions 1, 2, 3, 4, 5, 6, and possibly 8 introduce 'quantity', the second 'THING' on Aristotle's list. All of these are quantified as many or a lot, with the exception of proposition 6, which uses 'some'.

Propositions 1, 3, 4, 7, 8, and 8' all introduce entities with information about 'place', Aristotle's fifth category. Finally, the general semantic meaning of the Existential-there proposition is a 'state' or stative, the eighth class.

4.1.1.1.2.1 Spatio-Temporal Anaphora The learner includes the parameter of place or location in five of the six Existentialthere propositions, as he introduces a new referent. The new entities are all introduced according to a location.

Example 4.4) s3: There are... lots of people on the street' (1) i2: uh huh s3: There are many lights (2) i2: many light? s3: many alot of lights... on the street. s3 Around.uh...around... the light there are.. lots of building (3) i2: uh huh s3: And center of the street i2: uh huh s3: there are...many trees... around the street (4) s3: near the light there is a kind of car (8) The 'street' is the anchor for describing the festival in four of the five propositions. Spatial nouns, such as 'a street', may of course be plural or singular, but they can also be divided into spatial relations. In these propositions 'the street' is functioning as the primary character in the description, and is divided with spatial relations, i.e. 'around', 'center', and 'near'. As temporality can be divided into slices of time according to properties of 'before' and 'after', the spatial anaphor can IDbe divided into 'around', 'center' and 'near'. These relations form the Sparameters of the learner's logical analysis of 'things'.

0 Existentials have long been considered to bear a close relation to other locational constructions and hence grouped with locatives and possessives (Lyons, S1977; Clark, 1978:3). In a large database of Japanese, Kuno (1971) notes that the V)locative pattern in existential Japanese structures is 17 to 1 for sentence initial 00 locatives 335, fn 9.) This semantic difference in the behavior of existential versus other types of sentences was first observed and described by Milsark (1974, 1977) as "state- CK, descriptive and "property" predicates. These predicates were later developed by Carlson (1977) and classified as either 'stage-level predicates' (SLPs) or 'individual- CI level predicates' (ILPs). In this system, verbs found in existential-therelxxv sentences ri are classified as stage-level predicates (SLPs). The semantic distinctions between O these two types of predicates and their interaction with syntax and pragmatics have 0been prominent in the development of theories about verbs and the syntax-semantics interface in the work, for example, of Diesing, 1992; Kratzer, 1995; Krifka, 1995; Femald, 2000).

Stages, as conceived by Carlson, are "spatially and temporally bounded manifestations of something" (cited in Fernald, 2000:5). Fernald's study forces the conclusion that SLPs "describe characteristics of individuals that hold in space and time-we might even say that they describe spatiotemporal slices of the world" 11) The predominance of existential-there propositions in this discourse suggests that the learner's language at this point is primarily using Carlson's notion of stage-level predicates. In this system, a learner who produces primarily there-sentences naturally produces them with these spatiotemporal anaphors. If we are talking about groups of people, we divide the groups into pluralities or singletons or plurals and mass terms.

Other approaches include events, processes, and statives, all subsumed under the notion of eventualities. In these approaches, Carlson's stages become only one subpart of eventualities, in particular statives (Bach, 1989:95; Chierchia, 1982, cited in Bach 1989). Throughout this chapter, I use the older system, which has undergone much recent development. To these approaches, I incorporate specificity.

4.1.1.1.2.2 Kinds Although there is only one instance of the lexeme 'kind' in this discourse, it is worth mentioning how 'kinds' fits into Carlson's system of this learner language. The most widely accepted analysis of 'kinds' is by Greg Carlson in his 1977 doctoral thesis, 'Reference to Kinds in English' (Carlson, 1977). Carlson objected to associating 'kinds' with properties, or characteristics. Using tests of syntactic scope and ambiguity, he argued that 'kinds' functioned more like proper names and bare plurals.

In this example, we are not interested in proper names and bare plurals per se, but rather their commonality as "manifestations" of 'stage-level predicates'. Stage-level predicates must be existentially interpreted and express temporary states.

'Kinds', on the other hand, incorporate properties as with other stage-level predicates, but according to Carlson, they do not express temporary properties; they are a characteristic of an entire class. In this example, as with proper nouns and bare plurals, 'sharks are dangerous' or 'sharks are types of fish'), a distinguishing property can be assigned to the class to which they belong. The learner's proposition asserts that 'cars' are in the parade, but they are a specific kind of car upon which one

L

IND can place a large taiko drum while instrumentalists beat on the platform of the 'kind' O of car, a ute.

~The predominance of existential-there propositions in this learner's a interlanguage discourse could be analyzed at this point as support for Carlson's notion Sof stage-level predicates, which also includes 'kinds'. The utility of Carlson's system Sis that it allows one to distinguish 'kinds' as a type of predicational property having 00 semantic meaning.

In the approach I use which includes statives, all subsumed under the notion of eventualities, the existential-there propositions are composed of nonspecific CK, entities, and 'a kind of car' is nonspecific entity in a stative proposition. Carlson's stage-level predicates do not distinguish specific from non-specific 'kinds'. In the work of Kamp and Heim, the discourse is an intermediary between the model of the C world (Task IV) and the expressions in the participants' language (the common ground of the discourse). For our purposes, we can say that the common ground of the discourse is an intermediary between Task IV, the link to the world, and groups of objects in the picture. In other words, the idea of groups and their members (partiality) can be tracked from the picture to the discourse or vice versa.

The utility of this system for the notion of 'kinds' is that it allows one to semantically distinguish 'kinds' as a type of property having semantic meaning. That the learner was grappling for a noun of a specific sort is clear from the follow-up description car..on taiko. The name of the type of car he is looking for is one that has the property of being able to put a large drum on it. Clearly this is not a passenger car; the taiko drum in the picture is too large to put through the doors in any kind of passenger car. The picture shows a small 'ute' (Australian English) or 'pick-up' or 'flatbed truck' (American English). This property identifies it quickly for a native Japanese, as these utes are not only common in Japan, but they are a frequent addition to a parade or festival.

More importantly, the learner is showing evidence of the ability to subdivide larger lexical units, i.e. cars. The learner is beginning to associate properties as a defining characteristic of one group of common nouns, cars.ixxvi The learner is beginning to denote individual instances of types of cars. We can say that the car is a nonspecific object and kind is a nonspecific member of the set of cars.

However, in the next utterance, when the learner says on taiko, the utterance refers back to the car and thus the next proposition has a property which is specific to car. We now have a series of properties about this one instance of a car; the car is on the street, transporting a taiko drum, and its time of occurrence is in a picture of the Kanto festival.

Figure 4.1 'Kinds' and Properties IDFigure 4.1 diagrams the objects which the learner is adding to the common O ground of the discourse while he describes the picture of the Kanto Festival. The large circle of the street intersects with the set of cars on the street at Time 8. Then, at Time S8' a property is added to the learner's proposition and a smaller set of taiko drums is Sadded which intersects with the larger set of cars. From this final intersection, the Hearer can infer that a car with a taiko drum is on the street in the picture of the 00 festival in the learner's hands.

From this the Hearer infers a 'kind' of car which is a member of the group of cars. The learner has divided this group into one member of the set of cars based on I its function in the festival or parade in the picture. This function, a combination of several types of'things' in Aristotle's classification and #8, helps define this notion of 'kind'. The 'things' are listed below: 1. Substance a ute 2. Quantity large enough for a taiko 3. Quality a flatbed or surface 4. Relation between the flatbed and the drum; car and the street Place on a street 6. Time during a festival 7. Situation being in a parade 8. State existence of a property The learner has not explicitly stated all of these things in the propositions at Time 8 and the only lexemes used are kind and car and on taiko. But the interviewer can use information from the entire context, both discourse and presence of the picture, as well as the intentions of the learner to give characteristics of a festival.

If she understands what the learner is producing, she can easily make these deductions in order to create a visual image in her mind and fill in the contents of the picture. This kind of mental imagery will be helpful to guess the identity of the picture in the learner's hands. It will also help keep track of the information the learner gives her. So, between Time 8 and Time she asks for clarification of the object just produced; she has understood 'card', not car. This interaction demonstrates the development of the properties of the kind of 'thing' with her visual image and the need for clarification. It also demonstrates the use of kind as a means to describe properties when the exact lexeme is not available to the learner.

4.1.1.2 Characterizational Semantic Proposition Type The two propositions which do not assert the existence of a new entity, are propositions six and seven. Instead, they refer to existing objects in the discourse.

These two propositions are a significant departure semantically from the Existentialthere propositions, since they assign characteristics or properties to the entities in the existing sets. This kind of characterizational relation can be graphically displayed as below: Figure 4.2 Quality characteristics (l3Zhti~ OSemantically, this diagram of light shows two set relations: the individual O Semantically, this diagram of light shows two set relations: the individual 'light' linked to a set of characteristics associated with 'light', on a 'stick'. The set a of places is not in an inclusion relationship to light. The set of lights and properties Sare 'linked' or relational (Eng, 1991:21); the properties are qualities of the light. A Vlearner learns the word light independently of the words to describe the light. The oO characteristics of a light are numerous and more properties can be attached to its description, such as 'electric', 'shape', and a variety of colors. So one group describes the set of properties linked to the individual 'light'.

I Characterizational constructions are composed of(l) individuals from existing sets, and properties which form subsets to or sets linked to these elements.

c- CI 4.1.1.2.1 Specific 'Strong' Determiners ISO The second class of determiners proposed by Milsark (1974), 'strong determiners', are not found in Existential-there sentences but do occur in the Characterizational I Proposition. These include the definite article, demonstratives, universals, both, (n)either and the proportional determiners most and some.'xx v The proportional use/meaning of indefinite determiners is not generally found in there-sentences, as Milsark indicated in his classificatory systemx"""'.

At T6, the learner uses a quantifier in its proportional rather than cardinal meaning, T7 some people has got light.... First the learner introduces the group of people into the discourse and then subdivides it into a smaller group.

Example 4.5) Discourse i2-s3:4 i2: there are any people? s3: Yes... (looking at picture) s3 some people has got light... (6) Thus, the second group of people with the light is discourse-linked to the first group.

In the proposition, Some people has got light, the learner tells us that a portion of the group of people have a light. This can be illustrated by the diagram below: Figure 4.3 Proportional Inclusion: Group Characteristics People T6 P+i I(1) Some people has got light In this diagram, two sets of people are shown. The smaller set of people is included in the first larger set. 'People with a light' are a subset of all the people just introduced as existing in the picture. The set of people with a light are a portion included in the larger group of people.

We can also say that the smaller set of people with a light are a specific group inside the large group of people. The learner indicates this relation with the pronoun, some, which is specific to the group of people who have/possess a light. This notion of specificity is an inclusion relationship; the group 'some people have a light' is a subset of the larger group of people introduced between propositions four and six.

O The use of some in this description is the proportional meaning; it specifies O that a subset, or portion, of the larger group of people (on the street) have lights.

Because it refers to a set of people which is already in the discussion, it is a 'specific' use of the indefinite quantifier.

SAn opportunity to use the definite article the occurs in Propositions 7-8 after Sthe learner asserts the existence of lights for one group of people, and then 00 immediately describes the light.

Example 4.6) Discourse is-s3:4 s3 some people has got light... (6) s3: light is light has....on the stick (7) I The learner does not use a definite article to refer to 'the light which is on a stick'.

NHowever, even without the article, we can follow the interlanguage description of the light. According to Kamp's theory of definiteness, the group of people with a light are discourse-linked to the previously mentioned group of people; the light is on a stick.

Tracking the semantic meaning of the learner's proposition, then, is not determined by the definite article. The definite article is a morphological marking of definiteness and specificity found in English, but not in all languages, such as Japanese. It is not a necessity in language for distinguishing specific from nonspecific entities in a discourse.

The learner has begun to break a larger group into its smaller elements and assign specific properties to these smaller elements. The distinction of [+specific] also appears to be entering the lexicon of the learner. However, it can not be established that the learner actually is cognitively aware of this difference. In English, the ambiguous use of 'some' as [+specific] does not necessarily require this knowledge in the learner's usage of this morpheme. In some languages, where case assignment, such as Turkish, is involved in distinguishing [+specific] use or even in Latin which had separate morphemes, 'quidam' and 'aliquis' for the distinction in specificity for the use of 'someone', the former a specific pronoun, the latter nonspecific,"xix it can be more easily tested.

The learner is beginning to assign properties to existing discourse referents.

This is the beginning of expanding the semantic relations in the learner's interlanguage grammar to include Characterizational Propositions. The specific use of some presupposes the existence of people. Hence, existence is beginning to act as a platform for development of entities rather than the introduction of entities.

4.1.1.2.2 Properties Referring expressions in Characterizational propositions may link properties to an individual which exists in the situation of the discourse context. Sometimes this property may be a nominal. For example, characteristics may be a set of nominal properties to describe a person: John is a teacher, a father, and a tennis player. In these cases the nominals are properties. In Aristotle's list of things, these may be called a substance The learner produces another type of semantic relation using a nominal as a property, or a place This characterizational proposition has a semantic relation using a nominal as a property of a position relative to an existing set, people: 1 Example 4.7) Discourse i3-s3:4 O s3: Light light has... on a stick SIn this proposition, the learner has produced a relation for the position of the D light relative to the set of people: some people have a light, and that light is on a stick.

00 4.1.1.3 Week 1 Summary of Semantic Proposition Types in Task IV by s3 In this discourse, the learner produces two proposition types, Existential and Characterizational. The proportion of these two relations is illustrated in the graph below: C'K Distribution of Proposition Types in Task IV in Week I by s3 9 0. 4 3 0 4 Types of Propositions Although the Existential-there proposition type is predominant in quantity, two semantic proposition types can be distinguished in the learner's grammar. Out of a total of nine propositions produced, six are the Existential-there type (labelled 'Exist') and two are the Characterizational type (labelled 'Charact'). The final proposition in this discourse, That's all, I have left unanalysed and am considering it either formulaic or an idiom.

One could argue that the learner doesn't even really have a notion of Existential-there propositions in week one of this set of dialogues because the words "there-are/is' can be considered formulaic; the learner relies on this structure to introduce new entities in the discourse. However, this learner and could have produced only nominals without the Existential-there construction. It is a relation which must be acquired, and the learner has used it a total of six times in this discourse. Therefore, the learner's Existential-there propositions can not easily be dismissed as formulaic.

What does appear to be 'formulaic' in the first week is the reliance upon spatio-temporal anaphors together with Existential-there propositions for introducing individuals moving the discourse forward. The spatial parameters of 'the street' describe the location of the new entities, thereby delimiting their domain of existence and changing the information content of the discourse.xx" \0 4.1.2. Week 2 Proposition Types in Task IV by s3 For Task IV in Week 2, the learner selected the same picture as the week before for Nl Discourse i3-s3:4, the Kanto Festival. The complete set of propositions is given below Swith the learner's in boldface italic, numbered to the right.

Example 4.8) Discourse i3-s3:4(i interviewer, s student (learner), :4 Task IV) 4 i3 so, pick one, one pictures, don't show me,

C

i then please describe it.

I will guess it.

C' 4 s3 there are a lot of lantern in the street TI 4 i3 mm, mm 4 s3 and there are a lot of people T2 4 i3 yes C(i 4 s3 on the street T2' 4 s3 and some people bring the fruit, kind of fruit T3 4 i3 yes 4 s3 and some people hit a drum T4 4 i3 yes 4 s3 th, mm, one dan pass with the pole, pole 4 i3 yes, yes 4 s3 and there are some kind of cars T6 4 i3 cards? 4 s3 car, mm T6' 4 i3 please describe more 4 s3 mm, this is famous in Akitafestival, T7 4 s3 mm, it's very beautiful, T8 4 s3 This..., it is night T9 4 i3 yeah 4 s3 can you guess? 4 i3 um, [laugh] at night, Then there are many people? 4 s3 many people T11 and there are a lot of lantern these pole T12 4 i3 yeah, I know, what people doing? 4 s3 people do, people is acting, T13 4 s3 people have the pole on the hand T14 4 i3 yeah 4 s3 or shoulder or, or head T14' 4 i3 oh 4 s3 mm 4 i3 okay 4 s3 thank you 4 i3 Kanto? 4 s3 yes 4 i3 finish 4 <s3>yes<check i3,s3?> 4 i3 thank you very much 4 s3 thank you professor, thanks As shown, in Week 2, the learner, s3, produced a total of 14 propositions.

INO 4.1.2.1 Existential Semantic Proposition Type Out of the 14 propositions in Week 2, the learner produced 5 Existential-there Proposition Types, 1, 2, 6, 11, and 12.

S4.1.2.1.1 Non-specific Determiners 00 In Week 2, s3 again introduces entities into the discourse with weak, or non-specific, CK quantifiers and Existential-there propositions. This week, the learner produces only three determiners, the complex plurality quantifier a lot of in propositions 1, 2, and 12, and some kind of in proposition 6, and indirectly, many in proposition 11. Unlike the previous week the learner has no distributional 'errors'. The quantifiers a lot of with many never appear adjacent to each other in the same proposition.

CN Example 4.9) Discourse i3-s3:4

\O

s3 there are a lot of lantern in thite street (1) S13 mm,mm s3 and there are a lot of people (2) 13 yes s3 on the street s3 and there are some kind of cars (6) 13 cards? s3 car, mm 13 Then there are many people? s3 many people (11) and there are a lot o0f lantern these pole (12) In Proposition 6, the learner uses the phrase some kind of cars with the determiner some to describe the set of cars in the Kanto festival. In the previous week, the learner described the set of cars using the singular relation a kind of car. This week the learner uses the plural determiner together with the plural set in which it is contained. The learner denotes plurality as indicated by both some and cars. The use of some in this proposition not only indicates plurality but also uncertainty about the exact identity of the 'kind' in the set of cars. In other words, the learner recognizes its membership in the class of 'cars', but the use of some kind of suggests a substitute for the precise lexeme which would denote the 'kind'.

The first proposition again contains a type of partitive in an existential-there proposition: there are a lot of lantern in the street. In Propositions 1, 2, and 12 the existential-there propositions are all preceded by a lot of Again, it is not clear if these quantifiers are single morphemes or developing partitives. The nominal entity asserted to exist, lantern, is presented as being in a specific spatial context, the street.

4.1.2.1.1.2 Properties The properties in the Existential-there propositions are still limited to only two types: quantification and spatial location.

4.1.2.1.1.2.1 Spatial Anaphora Spatial anaphora is the main property found in these Existential-there propositions.

Propositions 1, 2, and 12 have properties of place. The street is again the anchor for the entity just introduced in Propositions 1 and 2. Street is specific at each time of mention.

IDThis week the learner only produces two spatial relations, in and on: lanterns in the O street and people on the street.

Unlike the previous week, the learner produces one spatial relation without a S morpheme to specify the location. Proposition 12 introduces a lot of lantern these Spoles. Nonetheless, the juxtaposition of the two nominals, lantern and poles, suggests to the hearer that the lanterns are on the poles.

00 'Kinds' is again introduced in an Existential-there proposition. In proposition 6, the learner uses the phrase some kind of cars with the determiner some to describe the set of cars in the Kanto festival. This week the set to which it belongs is plural, the C set of cars. In the previous week, the learner described the set of cars using the singular relation a kind of car.

Cl 4.1.2.2 Characterizational Semantic Proposition Type IDIn Week 2, the learner's discourse contains 9 Characterizational Propositions with one Sinterrogative, can you guess? in Proposition 4.1.2.2.1 Proportional Specific Determiners The Characterizational Proposition Type denotes a property of an established discourse referent. Therefore, the referent is usually [+specific], such as a group denoted by some. The learner in Week 2 expands the division of existing sets of people into individual subsets. Simultaneously, referring expressions refer to different individuals in the set. The following examples show the division of the individuals in the group.

Example 4.10) Discourse i3-s3:4 s3 and there are a lot ofpeople T2 13 yes s3 on the street T2' s3 and some people bring the fruit, kind of fruit (3) i3 yes s3 and some people hit a drum (4) i3 yes s3 th, mm, one dan pass with the pole, pole i3 yes, yes First, the learner introduces people on the street at T2, and then begins to break the group of people into individual members. Some and one are both proportional determiners and referring expressions. The group of people is divided into sets of people with properties according to their actions, #9 on Aristotle's List of Properties: bringingfruit, hitting drums, and passing with poles. The learner refers to individuals in the group as Some action, and One action. This contrasts with Week 1 where no actions were associated with dividing individuals in the group.

4.1.2.2.2 Properties This week, of the nine Characterizational Propositions, properties are assigned to referents in eight propositions. The propositions and their associated properties are: fruit drum pole (5),famous Akita (7),festival beautiful night hand shoulder 14), and head In this type of proposition, the 'property' of O the referent is additional information and is usually assigned a characteristic of O specific], since it is not discourse-linked.

(N Five of these nine propositions are 'actions', the ninth property listed in Aristotle's list of things. In the interrogative, the learner is asking the interviewer to Scause the 'action' of 'guessing' the identity of the festival. So, we can also include this proposition as denoting an action.

00 In the remaining four Characterizational Propositions, the types of properties include qualities (Proposition 7 and temporal (Proposition and spatial information (Proposition 14) modifying a specific referring expression. In these N examples, the referring expression is specific due to its linking to the speaker's visible antecedent, the picture of a festival and its anchoring to the discourse.

N Example 4.11) Discourse i3-s3:4

INO

4 i3 please describe more 4 s3 mm, this is famous in Akita festival, quality S4 s3 mm, it's very beautiful, quality 4 s3 This..., it is night time The variables famous Akita and beautiful are introduced as separate properties of their discourse referent, picture of the festival in the speaker's hands.xxi The first two are qualities, and the last lexeme, night, refers to a time. The property, night, is associated with the entity of the picture. The festival time is night and it is a singly perceived entity, not attached to the entity as a noun and an adjective, but as one entity 'night'. Syntactically, the learner has moved from attributive modification to predicative. The learner uses a copula to describe properties, Aristotle's stative.

Additionally, the learner now produces non-copula verbs, bring, hit, pass, and act, with a possible attempt to produce dance (Example 4.10). Vendler's classification of verbs as eventualities, or types of events, processes, and states includes all of these types of verbs as well as Carlson's stage and individual level predicates (Vendler, 1957, 1967). Vendler's approach to verbs thus allows us to divide eventualities into partial or instances of events.

In Example 4.11, the propositions continue to describe individual people in the group, but with additional spatial properties relative to actions and their bodies.

Example 4.12) Discourse i3-s3:4 4 i3 yeah, I know, what people doing? 4 s3 people do, people is acting, (13) 4 s3 people have the pole on the hand (14) 4 i3 yeah 4 s3 or shoulder or, or head (14') The learner is beginning to modify the individuals within the groups using spatial reference to the individual. People is specific, but the individual actions and locations are non-specific within the proposition.

I

\O 4.1.2.3 Identificational Semantic Proposition Type SUnlike the previous week, the conclusion to this discourse does not end with That's Ci all, simply completing the description with the monosyllabic bilabial nasal stop mm.

O- Again there are no Identificational Propositions in his description of the picture.

4.1.2.4. Week 2 Summary of Semantic Proposition Types in Task -N IV by s3 The proportion of the Existential-there and Characterizational Proposition Types in C the discourse are seen in the graph below.

C Distribution of Proposition Types Cin Task IV in Week 2 by s3 In Week 2, during Task IV, the learner again produces two types of propositions: Existential-there and Characterizational. As shown, the learner produced a total of 14 propositions: 5 are the Existential-there Type and 9 are the Characterizational Type. Unlike the previous week, in this discourse the Characterizational is dominant in quantity, almost double the number of the Existential-there Proposition Type. As in the previous week, no Identificational o Typesproposition types are produced.of Propositions TIn Week shows2, during Task IV, the learner again referring expressions and types of proposition types for descriptions of objeCharacterizational. As shown, the learner uses the same antecedent, produced a total of 14five propositions, but divides the Existentialgroup according to their quand 9 are the Characterizational Type. Unlikebegun to use prevsingular referring expressions, ththis dis and itcourse the Characterizational is dominant in quantity, almost double the number of the introduce new qualities into the discourse. So, the Characterizk, no Identificational Proposition Type is beginning to increase both in terms of the number of utterances in theoduced.

discourse, as week shows the beginning of a change in referring expressions and their properties in the discourse.

4.1.3 WEEK 3 PROPOSITION TYPES IN TASK IV BY s3 proposition types for descriptions of objects. The learner uses the same antecedent, people, for five propositions, but divides the group according to their quantity and their actions. He has also begun to use singular referring expressions, this and ii, to introduce new qualities into the discourse. So, the Characterizational Proposition Type is beginning to increase both in terms of the number of utterances in the discourse, as well as types of referring expressions and their properties in the discourse.

4.1.3 WEEK 3 PROPOSITION TYPES IN TASK IV BY s3 This is the last of the three week period examined for propositions by s3. In Week 3, the learner chooses a different picture of a festival in the Tohoku region. The learner's propositions produced during Discourse i4-s3:4 are in boldface and numbered to the \s0 right below. In this discourse, I have also included the line numbers of utterances 171- O 200 produced by both participants in order to facilitate the analysis.

Example 4.13) Discourse i3-s3:4(i interviewer, s student (learner), :4 Task IV) 00 i4 171 ok and ah let's move onto the next task s3 173 mm i4 174 of Tohoku festival C i4 175 so could you explain this picture or this festival i4 176 so that I can ask the festivals name? l^- C s3 177 ah hah they walking walking through the mountain TI i4 178 mmh s3 179 with fire(??) and he they they wears like demon mask demon mask T2 i4 180 ah hah s3 181 and they wear they wear un ara T3 i4 182 straw s3 183 straw straw cloths T3' i4 184 mm s3 185 and straw boots T3" i4 186 so must be winter festival s3 187 yeah it was winter festival T4 s3 188 it is winter festival s3 189 it's famous in Akita T6 i4 190 ahh festival in Akita I see i4 191 have you ever seen that festival? s3 192 no T7 i4 193 so I know this festival s3 194 mm you know? T8 i4 195 Namahage s3 196 yes T9 i4 197 ok wakarimashita thank you very much s3 198 thank you very much i4 199 oh I can't pick up other festivals ((both laugh)) anywas thank you very much s3 200 thank you As shown, s3 only produces a total of seven propositions before the interviewer guesses the identity of the festival. The learner produces no Existentialthere propositions in his description of the picture during Week III. This is the first discourse which does not include the existential-there proposition type for the description of the objects in the picture. Again, no Identificational Propositions are produced during the description. All of the propositions are the Characterizational Type.

ID4.1.3.1 Characterizational Semantic Proposition The discourse produced during Week III contains seven propositions and all seven are ri the Characterizational Proposition Type. None of the propositions begin with either a weak or a strong determiner, as in the two prior weeks. Rather, every proposition begins with a referring expression: six of these propositions begin with a definite oO pronoun: three begin with they and three with it.

SAll six propositions occur in a semantic relation of characterizing a property of the referring expression. Spatial anaphors are used only twice to delimit the set of a new individual in the discourse and are not used again, through the mountain and in Akita, Propositions 1 and 6, respectively.

1 4.1.3.1. I Pronominal Anaphora r The first three propositions begin with plural anaphora. The first proposition, they ND walking through the mountain withfire, an entity, they, is assigned Property 9, an action or activity. The second and third propositions, respectively, they wears like 04 demon mask, and they wear straw cloths and boots are assigned Property 7, a situation an event. So plurality in the discourse is no longer introduced by determiners nominals, but rather through the use of a single expression which automatically assigns plurality.

The next three propositions begin with a singular referring expression, it, and all assign Property 3, qualities, winter and famous in Akita. The last thing, famous, is the learner's opinion, which also has a quality of 'passion'. So, one proposition is assigned an action or activity, two are assigned a situation or event, and three are assigned a quality.

The classification of verbs into eventualities, or types of events, processes, and states, allows an evaluation of all the propositions during Week III. Instead of stative properties of the Existential-there Proposition Type, the first proposition has individuated instances of the event of walking.

Kiss (1998) briefly entertains the types of NPs and their distribution in types of predicates (pp.159-161). She concludes that it is the distribution of the specificity] feature which constrains predicates; only the existential be "in the there is construction" disallows [+specific] (Kiss, 1998:160-162:fn The learner describes the men as walking down a mountain. The picture shows legs on the ground, up in the air, and part way between the ground and air. These are all partial, or specific, instances of leg motion in the act of walking. These are states of walking and each individual man is in one state of the act of walking. The learner is connecting these positions in his mind by describing the activity of the men as 'walking'.

We can also say that the learner is using the specific instances of leg movement to combine the action into one single behavior of walking. The use of the progressive is consistent with the learner's ability to dissect the action into small instances in the same manner he has begun to distinguish specific members of sets of individuals, i.e. people who are dancing or bringing fruit, are all members of the set of people on the street.

A complete description of this entire event, then, includes the aspect of an activity bounded by [+specific/-specific] features. Vendler's theory of eventualities works more favorably for the description of the difficulties and development of English by second language learners. It is well known that the English progressive is difficult to acquire for some learners, and this approach to the verbal system O individuates temporality into pieces, paralleling a pattern of development of O individuating nominals.

4.1.3.2 Summary of Propositional Proposition Types in Week 3 by s3 SIn Week III, the learner produces only one type of proposition, seven Characterizational Types, as shown in the graph below.

00 Week 3 Summary of Semantic Proposition Types in Task IV by s3 Distribution of Proposition Types Sin Task IV in Week 3 by s3

(NO

4

U,

Tpes of Propositions This week also shows a major change in the features of each proposition. The learner only use definite pronouns, it and they, as referring expressions and each referent is used to add characteristics to the entities as they are produced, rather than anchoring the new entities in a spatial parameter. Finally, as well dividing groups into individual members which began in the previous week, the learner is beginning to divide time into individual segments through the use of the progressive.

The general development in the entire discourse of Task IV in the three week period for this learner has been from introducing sets, dividing individuals according to spatial properties, and then according to actions. The learner seems to be moving into finer and finer distinctions of both time and nominals. This onset of the individuation of time and sets, or groups, needs further study to determine if this is a common occurrence among other learners' developing interlanguage.

4.2 Macropropositional Semantic Proposition Types In this section I examine the overall discourse for patterns of meaning which are manifested between both the learner and interviewer, or Speaker-Hearer interaction throughout the discourse of Task IV. If we consider a discourse to be a complex expression, we can see that the patterns in the whole discourse are also composed of the same basic semantic relations. The three semantic relations occur over many propositions, forming a semantic 'macroproposition' which serves as the basis for the organization of the whole of the discourse.

The difference of the macroproposition is that it is the group of semantic relations for each interlocutor which must be considered. The semantic relations for the Hearer may not be the same as for the Speaker. Thus, semantic relations of individual propositions differ between speaker and hearer at the discourse level as well as at the sentence level. We can not fully understand the meaning of each IND individual proposition without considering the role of each proposition in the structure of the overall discourse.

C4 The macrointerpropositional semantic relation is created with the communicative interaction of both the extensions and intentions of two collucators in the discourse. In the discourse created in Task IV, the 'intentions' of the two Sthroughout the discourse differ. The interviewer first introduces a picture to the 00 learner s3 and explains the task composed of two differing roles for speaker and hearer: the learner is to describe the picture; the interviewer is to listen to the description and identify the festival the learner chose.

Example 4.14) Discourse i2-s3:4 i2: I have some picture of festival so you can pick up one and...don't show me s3: uh huh...

IN i2: And describe...

Sand I will try to guess which festival In the following, I first describe the semantic relations of the learner as he orally produces a description of the picture. As the learner speaks, the interviewer listens to his choice of objects from the picture description. I then describe the semantic relations of the interviewer who guesses the identity of the festival the learner has chosen from the folder of eight pictures in the Tohoku region.

4.2.1 Week I Macropropositional Semantic Proposition Types 4.2.1.1. Learner as Speaker Characterizational Proposition Type The communicative goal of the learner is to describe the picture in his hands so that the interviewer can guess the identity of the festival. The starting point for establishing these intentions and their corresponding semantic relations for the learner is repeated below; the relevant instructions for the learner are underlined.

Example 4.15) Discourse i2-s3:4 i2: I have some picture of festival so you can pick up one and...don't show me s3: uh huh...

i2: And describe.., and I will try to guess which festival The learner takes the picture from the folder and then chooses visible elements from the picture to enter into the description he develops in the common ground.

After listing many people, walking on the street, many lights, cars with taeko drums, and lights on bamboo sticks, the learner indicates these are all of the characteristics of the festival and asks if she can guess the identity of the festival in the picture.

This semantic characterizational relation can be described as many items from one set, the picture, being articulated into another set of many auditory properties.

Common Ground Visible Picture Auditory Linguistic Objects people, lights, peoplelights, car, aeko drums, taeko drums Istreet, sticks, trees street,sticks,trees, buildings buildings MANY MANY 00 In this description many visible characteristics combine together to create an auditorally defined picture of one festival from the interviewer's folder of pictures.

The learner's group of propositions can be considered a set which consists of a C-i collection of many elements necessary for describing the many properties of the picture. The description, thus, forms a characterizational semantic relation of CI many-many.

,i We can consider this a function of many-to-many, which consists of many auditory propositions needed to describe the many visible entities and attributes found in the picture. Each proposition by the learner represents an extension, or picks out Svisible elements in the picture and brings them into existence and/or gives them auditory properties for the hearer. Each proposition assigns a different element to the discourse.

There does not seem to be any order to the choice of elements he mentions or describes. This conforms to a principle similar to what is called the principle of extensionality, i.e. random listing: each proposition differs from the others and no particular criteria determine the order or choice. However, although the order of the elements does not seem fixed, the set of properties which the learner chooses is confined to elements from the picture.

The question might arise as to whether the learner could be said to be selecting elements in order for the interviewer to be able to guess the identity of the picture.

However, in this discourse the learner has not really given the interviewer any distinguishing characteristics of the festival. Festivals have people, may have trees, buildings, lights or drums. These characteristics describe other festivals in the Tohoku area. Proposition 7 is the only real distinguishing proposition which might help the interviewer guess the identity. When the learner tells the interviewer that some people have a light on the stick, he looks directly at the interviewer. So, the verbal description is marginally meeting the goal of identifying elements in the picture which help the interviewer guess. The lack of control over delivering distinguishing characteristics during this discourse becomes clearer in the discourses in the next two weeks in which the both the learner and interviewer control the choice of items selected for the description.

Nonetheless, the combination of the two proposition types, Existential and Characterizational, succeeded in providing enough information about the picture to complete the description so the interviewer could identify the name of the festival depicted in the picture.

4.2.1.2. Interviewer as Hearer Identificational Proposition Type The communicative goal of the interviewer is quite different in this discourse. The starting point for establishing the interviewer's goal, intentions and the corresponding semantic relations for the interviewer is repeated below; the relevant instruction for the interviewer is underlined.

IND Example 4.16) Discourse i2-s3:4

O

i2: I have some picture of festival so you can pick up one and...don't show me e s3: uh huh...

i2: And describe...(change tape) (cont.)and I will try to guess which festival

(N

The hearer is given many properties and must guess which festival from the group of pictures in the folder. This requires a semantic relation for the hearer, or C1 interviewer, which is different from the learner. The interviewer must 'identify' the festival; this requires an identificational semantic relation.

C First, the learner develops a list of elements and characteristics from the 0 picture. This brings into existence a collection of auditory objects for the interviewer I\O who is listening to the learner. As the Hearer in this dyad, she only sees the back of Sthe picture in the learner's hands-a blank object. Her task is to listen to the auditory 0C objects as each descriptive element enters the common ground. As the speaker describes elements from the picture, these auditory elements create a mental image for the hearer. The interviewer's main goal of guessing the identity of the festival is accomplished in the conclusion to the discourse, repeated below: Example 4.17) Discourse s3-i2:4 s3 That's all. Can you guess? i2 Hhm. I think it is Kantofestival i2: Okay. We gonna finished. Thank you.

s3: Thank you so much.

Once the learner has finished the description, the interviewer guesses the identity of the one festival from the folder full of pictures that the learner has described. The interviewer asserts the proposition "I think it is Kanto". The small or intraproposition it is Kanto inside the full proposition I think it is Kanto, is an identificational semantic relation. The relation between the 'intraproposition' following the psychological lexeme think creates an identity relation between what the interviewer thinks and the mental representation 'it is Kanto'.

Not only do it and Kanto refer to the same object in the proposition creating the identity relation, but they both refer to the object in the hands of the learner.

There is a three way relation between the proposition's two acoustic objects, i.e. the anaphors it and Kanto, and the visual object or referent, the picture of a festival, in the learner's hands.

Identificational Relation Y/Z FESTIVAL NAME Kanto Festival Y Z It (the picture name) is the Kanto Festival IND All of these elements lead to only one answer or, referent, the name of the O festival in the picture, Kanto Festival. These relationships are graphically described below.

Visual Object Auditory Linguistic Objects Mental Image Festival Name 00 people, lights, people, lig cars, taeko dru, cars, taeko drums {<Kanto

F

ss c s set ks buildings buildings Ci BLANK picture MANY MANY One--Unique Festival O This is a more complex semantic relation of many-one. The individuals and their properties are listed orally in the common ground of the discourse by the learner and create mental images which are entered into the interviewer's mental representation of the festival. Even though there does not seem to be any order to the choice of elements the learner mentions, the interviewer nonetheless creates a collection of mental characteristics of the Kanto Festival. Thus, we can consider this a function of many-to-many-to-one=>unique, which consists of many auditory propositions needed to create the many visual images in the interviewer's mind. This leads to the interviewer creating a single entity and guessing the identity of the (unique) festival in the learner's hands.

Since the interviewer has several pictures in the folder, she initially tells the learner to choose one picture and then she will guess which festival. Choosing one festival from the group sets up a uniqueness relation. The relation of uniqueness from the set a of festivals is expressed in the formalism in set theoretic representation of functions used for proper names by Kamp and Reyle (1993:246-248) for the notion of anchoring and proper names. The curly-brackets are familiar as denoting functions and the name Kanto Festival is anchored in angled brackets inside of that function.

For Kamp this assumes that the proper name Kanto Festival actually belongs to and thus has a function in this discourse.

'Kanto Festival' is not the same as a proper adjective which simply modifies or assigns a property to the following noun, e. g. Japanese student. This variable refers to an entity which is entered into the set of community knowledge which all Japanese will recognize as a single unique entity. Thus, both Akita and Festival are anchored to this set of knowledge. This is represented by the anchoring notation including both members of the relation. This representation is necessary to distinguish this festival name from the city and prefecture also named Akita, as a proper name from the festival which also bears its name.

The representation of many-to-many-to-one=:unique is used for proper names expressed by the identificational relation in the final proposition of the DRS. For the interviewer, each proposition may be an extension of elements in the picture provided by the learner, but the intention of the interviewer is to guess the identity of the festival. So there is a clear purpose for integrating the particular elements chosen.

This synthesis can be considered the intention of the extensions picked out by the learner.

NO 4.2.1.3. Task IV Macropropositional Semantic Proposition Types The interaction between the learner and the interviewer, their differing goals, or i intentions, and the three semantic relations during the execution of Task IV is shown below. The differing intentions of the interviewer and the learner begin to develop at the start of the instructions given by the interviewer. The learner then describes elements from the picture in eight short propositions, using two semantic relations.

The learner's propositions enter the common ground of the discourse. The interviewer listens to the propositions as they are uttered. Since her intention is to guess the identity of the picture she listens while simultaneously producing a mental image of the picture. The learner's propositions in the common ground culminate in the interviewer's identification of the name of the festival held in the learner's hands.

C I have represented this interaction and the semantic relations between the two S, participants as below: Xy LEARNER s3 {<Kanto Festival>} (y) INTERVIEWER i2 picture of T1= T8 MR it Kanto Iestival {<Kan o Festival>}(X) Prop 1-8) (entities in Prop 1-8) 85objects visible only to s3 KA ____CommonGround Kanto.EestLv.al}.(.y.)__ Ky KA Y I think u is y

KA

The largest rectangle represents the context of the entire discourse, the universe. The universe of referents in this discourse structure is given at the top.

Kanto Festival is placed in the highest structure to represent the relation (function) it has over the source of all referents in the discourse: the task and visual world of the learner, the auditory world of the discourse and common ground for both interlocutors, the mental image for the interviewer. This position makes it clear that it is irrelevant if the anaphor is visual or auditory in the semantic representation of relationships.

It also represents its position as the variable for the intentions of both members in the communicative tasks of this discourse: the characterizational set for the learner and the identity relation for the hearer. Both sets of semantic relations are embedded inside the matrix DRS.

Two small universes are embedded inside the larger, or main, DRS. The left DRS is in boldface with a small Kv in the lower left comer. The boldface and Kv both represent visual knowledge, which includes the learner's source of visible referents. In this DRS, the picture in the learner's hands is represented by the visual referent This X represents a single referent for the complete set of objects composing the elements of the picture of the festival, represented in Kv. This referent is at the top of the small visible world knowledge inside the auditory world in the matrix KA.

NO A single variable y is at the top of the right DRS. It represents the referent for O the proper name Kanto Festival. The pronoun it represented by u, is a stipulated Sidentity; both interlocutors, the learner and the interviewer, must interpret this variable u as referring to the picture Thus, the visual referent x u, is the stipulated entity Sor pronoun it. The overall proposition is an assertion because the interviewer is claiming that she thinks her mental representation of the picture created by the objects 00 described orally by the learner is the Kanto Festival. This universe of discourse referents created her mental image and she is asserting these are objects in the real world of the picture. These have created a mental representation of the Kanto C Festival. Therefore, she asserts the proposition that she thinks the identity of the picture is the Kanto festival, x is y."XXXi C The relation of uniqueness in a set a of festivals is indicated in this DRS. In the proposition, I think it is Kantofestival, the interviewer correctly identifies the festival as the Kanto festival.

SExample 4.18) Discourse i2-s3:4 4 s3 That's all. Can you guess? i2: Hhm. I think it is Kanto festival i2: Okay. We gonna finished.

Thank you.

s3: Thank you so much.

The identity relation is represented after the two entities and in the relation have been introduced. The semantic relation between x and z is an identity relation. Both entities are definite descriptions, and refer to the same referent although they each have their own independently identified individuals. The first referent establishes the set of properties describing/referring to the picture of a festival in the learner's hands, which is already a familiar discourse referent from the interviewers earlier question. The second set identifies the member of that set which the interviewer considers unique among the individual members in this set.

At the beginning of this section I said that we can not fully understand the meaning of each individual proposition without considering the role of each proposition in the overall discourse.

The propositional semantic relations are embedded in the overall discourse context. This is a significant departure from Schank and Abelson's concept of a frame for the starting point of the discourse and the conclusion to this discourse are very tightly intertwined. Each proposition of the learner is an extension of visible referents to auditory objects. But these properties do not describe the semantic relation of each proposition for the Hearer, or interviewer. These are creating a unique set of properties which culminate in an identificational semantic relation for the interviewer.

The it in this proposition refers to the first propositions of the discourse produced by the interviewer which contains the antecedents picture offestival, one, it, and whichfestival: Example 4.19) Discourse i2-s3:4 i2: I have some picture of festival so you can pick up one and don't show me IND s3: uh huh...

i2: And describe.it (change tape) (cont.)and I will try to guess which festival) All of these refer to the same object, the picture in the hands of the learner.

SThe interviewer's goal is to guess the identity of the picture in the learner's 00 hands. She explicitly tells the learner not to show her the picture. She does not know the identity. Hence, the type of information available between the two interlocutors differs. The information for the interviewer available in the context for interacting in the process of guessing creates a special kind of mutual knowledge. The entire discourse is based on mutual knowledge which is fundamentally different: the learner knows the identity of the picture, the interviewer must guess, or infer, its identity.

SThe time it takes to develop the mental representation for the final it in the Sidentity relation is much longer than the intrasentential relation of a single IDproposition. This representation has taken from Time 1 to Time 9 and together create the referent for the it in the final identificational proposition. The it refers to the N picture as expressed in the directions given to the student. However, the contents of the referent have changed significantly from the first use of it until the final use by the interviewer. Throughout the entire discourse the information content between the speaker and hearer has not been the same. We can not say that this referent has been the same at any moment in the discourse for either discussant. In other words, the referent it in this discourse is not the same type of referent as found with a single item and later mentioned, such as he said", where 'he' refers back to a single individual and both participants in the discourse know to whom it refers. The difference is manifested in the time required for processing. This it requires something closer to episodic memory rather than short term memory.

Even when the learner has asserted the existence of a visual referent into the discourse, they are not identical. The speaker is limited by his command of English and also general need to fully describe the actual visual referent. When the hearer enters the acoustic information into her mental representation the difference is even greater. So, the final proposition I think it is the Kanto Festival is not a true identification between the referent x and the referent y. It is only a partial representation of the real world necessary for the identification of the name of the festival. It is a linguistic identification based on the linguistic information in the common ground, and the presuppositions which each speaker brings to the task.

Most importantly, the identification could not be achieved without the aid of memory. In particular, the interviewer formulates the identity of the picture through a mental visual image in visual memory. It is the visual memory created by a representation which is not present in the real world, but only in her mind.

Thus the Cognitive Constraint proposed by Jackendoff (1983, especially pp.16-18) predicts that there must be a level of mental representation at which information conveyed by language is compatible with information from other peripheral systems such as vision. In this discourse visible and visual information are interfacing with language.

4.2.2. Week 3 Macropropositional Semantic Relations Given the constraints on the differing roles in Task IV of describing the picture versus identifying the name of the picture, by the learner and interviewer respectively, the overall discourse relations remain the same in Week III. The goals of the two interlocutor are unchanged: the learner describes the picture and the interviewer IDguesses its identity. However, as already shown above, this discourse differs from 8 Week I and Week II in its referring expressions. Although the purpose of the learner Sand hearer has not changed, the semantic associations in the discourse text has Sbecome more complex. For this discourse, it is also essential to include a broader Snotion of inferencing from the context in order for the hearer to penetrate what is C)being posited as existing in the common ground.

00 4.2.2.1 Learner as Speaker Characterizational Proposition Type The starting point for establishing the learner's purpose in the Task is repeated below.

As indicated below, the communicative goal of the learner is to describe the picture.

Again, I have underlined the interviewer's instructions as well as numbered the lines of the discourse to facilitate discussion.

IN Example 4.20) Discourse i4-s3:4 Si4 171 ok and ah let's move onto the next task i4 172 I give you one picture s3 173 mm i4 174 of.Tohoku festival i4 175 so could you explain this picture or this festival i4 176 so that I can ask the festival's name? This week the interviewer does not use the word describe, but rather explain.

Although these have a slightly different connotations in the target language English, it is not clear that this has altered the learner's understanding of the goal of describing the picture. The learner has already completed the same task with other interviewers, so he is familiar with the goal expected of him. Describing the picture requires the same semantic relations and linguistic structures which the learner has used previously. As already examined above, his individual propositions all give the semantic relation required for giving characteristics of the festival.

Nonetheless, the learner presents the visual objects from the picture in a slightly different manner. We can list the objects he selects to articulate into the common ground, they, mountain, fire, demon mask, straw cloths, straw boots, winter andfamous, but the objects are not simply assertions of existence. They are either part of an ongoing process or offer an opinion of the learner.

In contrast to the first week, the learner first presents an object and then immediately gives distinguishing characteristics of the object. First, the learner presents men as doing something and then two more propositions describing their clothing. The objects selected are not common to all festivals, such as was the case in the first week describing streets, trees, and people.

Nonetheless, these first elements are confined to the picture. The last three, on the other hand, are based on a question from the interviewer and the last two offer a personal opinion and experience, respectively.

This semantic characterizational relation can be described as many items from one set, the picture, being articulated into another set of many auditory properties.

Visual Picture Auditory Linguistic Objects

OD

00

N

demon mask, straw clothes, straw boots, winter, famous

MANY

demon mask, straw clothes, straw boots, winter, famous

MANY

In terms of the mental representation, the learner has created entities in the discourse with more detail. In the first week, the entities were primarily asserted in simple Existential propositions. This week they include no Existential-there assertions, but nonetheless assert entities. The differences in the composition of the semantic patterns of meaning accomplish the same final goal of describing the picture.

4.2.2.2 Interviewer as Hearer Identificational Proposition Type The intention of the interviewer are given in the final statement, "could you explain this picture or this festival so that I can ask the festival's name?" In the instructions for Week I, the interviewer used the phrase I will guess which festival. Although which festival is more specific than the definite pronoun it, it is nonetheless clear that the intention is to identify a specific festival which the learner has selected from the group of pictures in the folder. The contextual features of the visual objects, the verbal instructions as to the intention of guessing the one picture selected, automatically ensure the specificity of the pronoun and speaker's intentions. The precise goal is to guess the name of the festival, but this is understood from the context.

Visual Object

D

Auditory Linguistic Objects Mental Image they, mountal they, mounta fire, demon ma fire, demon mask, straw clothes, straw clothes, straw boots, straw bos, wter-,amous a us Festival Name {<Namahage>} BLANK PICTURE MANY (One object) MANY One Unique Festival The mental representation, this week includes no Existential-there assertions, but nonetheless asserts entities, more detail, and describes the picture.

4.2.2.3 Task IV Macropropositional Semantic Proposition Types xy S{<Namahage 0 Namahage(X) (Prop x u {<Namahage 7 objects visible only to s3 KA {<Namahage Kv Common Ground KA

KA

00 C Again, the universe of referents in this DRS is given at the top. Namahage is placed inside the curly brackets to represent its anchoring as a proper noun to the Japanese community knowledge and also throughout the common ground of the discourse. This week the only Characterizational Propositions are produced by the learner, but the Interviewer still produces an Identificational Proposition at the end of the description. Thus, the Identificational Relation contains entities which together C, create a description for a unique, Proper Noun. The descriptive content and Sinformation from the context together combine to provide the uniqueness of the 0picture of a festival in the learner's hands.

C1 4.3 proposition types in task iii The second discourse is from Task III, the information gap chart with information about four festivals in the Tohuku region of northeastern Japan. It is designed to elicit questions from the learner to fill in the blanks of the chart to find the specific answer for the missing information. Since the pattern of proposition types for Task III, the Information Gap activity, does not change significantly between week 1 and week 2, I only present the discourses from Week 1 and Week 3.

4.3.1 Week 1 Semantic Proposition Types in Task III (Information Gap) In Week 1, the chart contained twelve blanks. The learner uses five wh-question words; the other seven questions ask either directly or indirectly if an individual exists. The complete set of propositions by i2 and s3 during Task III is given below.

Only the learner's questions, or interrogative propositions, are in boldface and labeled to the right. The propositions labelled with AT indicate information being copied orally from the answer the interviewer is offering.

Example 4.21) Discourse i2-s3: 3 i2: I have a chart about Tohoku festival so you need to ask me questions to complete your chart s3: Uh okay Uh...hum..Top of the left blank...What's the festival name? TI i2: It's uh Tsuzureko Odaiko s3: Tsuzureko daiko... ATI s3 Where is the location of Tsuzureko Odaiko? T2 i2: It's Takanosu-machi takxxx s3: It's uh.. are there special traits about Tsuzureko Odaiko? T3 i2: Special traits is 3.71 meter taiko.

s3: 7 7 points... AT3' i2: 3 point 71 s3: 71 AT3" i2: meter s3: meter AT3'" I\O i2: taiko s3: Are there special affect about that Tsuzureko Odaiko? T4 i2: Special effect s s3: Uh uh facts? T4' i2: Special fact? s3: Uh yeah T4" i2: Special fact is permanent taiko display 00 s3: Is there are a purpose of Kamakura? Si2: Purpose is to celebrate water god and ask him for water s3: What Kamakura age? T6 i2: Kamakura age? s3: uhm T6' i2: I don't know but N dash A...N s3: N? AT6 Ci i2: N slash A s3: Uh..Do you know the meaning of Namahage? T7 i2: It's uh foreign demon s3: Are there any participants., of Namahage? T8 Ci i2: Participants are men of town...

s3: men? AT8 i2: uhm s3: Do you know the meaning of Daichido...Daichido..Bugaku? T9 i2: xxx..Its meaning is small village dance....with court music s3: When is the date of Dainichido Bugaku? i2: It's Jan. 2.

s3: What How...What age of Daichiro Bugaku? Tll i2: 1000 thousand 200 years s3: Are there any activities of Daichido Bagaku? T12 i2: Activities....People dance to ancient court music s3: Asian? ATI2 i2: ancient s3: ancient ATI2' i2: court music s3: That's all T13 i2: Are you done? s3: Yes. T13' Propositions 1-12 are interrogative propositions intended to complete the task of filling in the chart. Propositions 2 and 10 use where and when, asking for spatial and temporal properties, respectively. Propositions 1, 6, and 11 use the interrogative what, asking for the name and age of specific festivals. So, five of the twelve interrogative propositions use Wh-question words to complete the task.

The other seven interrogative propositions do not use wh-question words.

Five of the interrogative propositions 3, 4, 5, 8 and 12 use a copula+there. These propositions ask the interviewer about the semantic relation of existence of an individual in the chart. Propositions 7 and 9 ask the interviewer if she knows the meaning of a particular festival, Do you know the meaning of x? Yet, the learner knows that the interviewer has the answers and that they are provided on the chart.

We might ask if the directions are not explicit; they do not specify that answers exist on the interviewer's chart for each blank on the learner's chart. The directions given by the interviewer are: I have a chart about Tohoku festival, so you need to ask me questions to complete your chart. Although the interviewer did not IND make this explicit and does focus on herself as having a chart, the learner not only O responded Okay, but has had two previous experiences with this task: a pilot one month earlier and again one week earlier with a different interviewer. So, the student Sdid have reasonable expectations that the interviewer had answers.

SIn the following I look at the interlanguage semantic patterns of meaning in the learner's grammar this week. I analyze the interrogatives for the patterns of 00 meaning in the semantic relations of specificity: Existential, Characterizational, and Identificational in his choices of question formation.

C-i 4.3.1.1 Existential Interrogative Semantic Proposition In my classification of three types of semantic relations, the Existential-there relation C- is defined as a proposition which introduces an entity or proposition, and the C- there+copula structure triggers the semantic relation of existence to the hearer.

IND However, the learning situation for an interrogative of existence differs between SJapanese and English as shown below: English: a. There are special traits about Tsuzureko Odaiko.

b. Are there special traits about Tsuzureko Odaiko? Japanese: a. ga aru.

b. special traits aru.

The difference in morphosyntactic structure of the existential semantic relation propositions in English and Japanese differ, and both the lexemes and the syntactic structural difference must be learned. The assertions in English, or the a sentences, are composed of a there +copula; Japanese has no equivalent to the there morpheme, using only the ga particle. The syntax also differs between the two languages and the word order changes in the English pair of sentences must be learned. In the English examples, the is composed of a there-sentence and the is formed by an inversion of this structure, copula+there.

However, the communicative intention of the two propositions in each language does not differ between the two languages. Semantically, proposition 'a' assert existence and queries existence in both languages. Thus, at this level of semantic interpretation, the two existential relations have an additional difference in meaning. This difference in meaning can be explained according to the assertion versus querying of existence, but the similarities can be explained according to the properties of definiteness and specificity. xXXii In this discourse, the interrogative propositions 3, 4, 5, 8 and 12 use a copula+there. The learner is asking the interviewer if information needed to complete the chart exists. These propositions ask the interviewer about the semantic relation of existence of an entity in the chart. The interviewer, i2, does not answer the question s3 poses, i.e. Does x exist? Instead, in each case she simply provides the answer for the question. In other words, the interviewer assumes that the learner knows that an entity to complete the information on the chart exists; the instructions she gave informed the learner that she had the answers, and so she immediately provides the information in spite of a mismatched question structure. The answer with the missing IND element is confirmation of the existence of the element and her recognition that this is what the learner desires.

So we might ask, "Why does the learner use the existential relation for Sinformation?" He knows the interviewer has the specific information he needs. This Sis a communicative task with a closed question: information exists and the task for Sthe learner is to fill in the chart with the missing elements from the interviewer's 00 chart.

We can not claim that the difficulty for the learner is in the morphosyntactic structures. As was demonstrated in Task IV during Week I, the learner can handle the CK, morphosyntactic there-sentence for the assertion of existence in this week's discourse.

As far as a command of the interrogative structure, half of the there-propositions use about without the morpheme any and the last three, 5, 8, and 12 use the morpheme ,i 'any' without the morpheme about. So the interrogative proposition is still in a flux Sas far as native speaker-like production in the interlanguage. However, the use of the copula+there as a trigger for existence, is in the interlanguage of the learner. After the Sfirst interrogative proposition, in which the learner stumbles and moves from an assertion to the interrogative, he demonstrates in 100% of the propositions in both discourses that he has acquired the meaning difference in the semantic relation of asserting versus questioning existence.

The question remains, why does the learner choose the existential-there in both discourses during Week II of the experiment? What the two discourses share in common for both the assertive, or declarative, and the interrogative proposition, is a reliance on the existential proposition for the development of their discourse. What these propositions share, and what the other two semantic relations do not share with the existential, are their properties of definiteness and specificity. Of the three semantic relations, the existential proposition, regardless of its assertive or interrogative status, only introduces a nominal which is nonspecific.

As demonstrated in §4.1.1.1 and the learner has a predominance of existential semantic relations in his description of the Kanto Festival picture and again in this information gap task about four Tohoku festivals. The learner has not acquired the ability to develop a discourse using the propositions of characterizational and identificational semantic relations. In this task which requires the learner to ask for a specific element missing on his chart, it appears that he does not yet have the semantic ability to recognize and use the necessary relation which requires specificity. Hence, he relies on the one semantic relation which only uses indefinite, nonspecifics. A study of the interaction of the next two semantic relations in the discourse should shed some light on this question.

4.3.1.2. Characterizational Semantic Relation This week two Characterizationals were found in this Task as Yes-No questions.

Unlike Task IV no properties are attached to any entities.

4.3.1.3. Identificational Semantic Relation Unlike the Existential relation whose sole purpose is to introduce a new individual into the discourse, and the Characterizational which introduces a property to an already existing individual, the pattern of definiteness in the Identificational semantic relation is created when a new individual is introduced which refers to an already existing individual.

N Propositions 1, 6, and 11 have an Identificational Semantic Relation. First, O they are identificational at the level of the proposition. This is the traditional idea of San identificational. In this pattern of definiteness, the identificational relation has the S equative use of the copula rather than the existential be at the level of syntax. It is Screated when two individuals flanking a copula are considered categorically equal, i.e V they are both nominals. At the semantic/referential level of the grammar, the 00 Identificational Semantic relation is defined as having two individuals which share the same referent. In this meaning, the identificational "asserts that two independently identified individuals are in fact one and the same" (Kamp, 1993:258).

Cr Kamp adds that one is discourse-linked, referring to the same individual. In other words, both sides of the copula refer to the same individual and both are C specific, since they refer to a set which is already in the domain of discourse. This relation can be displayed as below:

IND

SIdentificational Relation

X

The orning Star is the Evening Star Y Z Each side of the copula has the same referent. At the level of syntax, the entities are manifested as an equational relationship: NP=NP. However, semantically this relation is identificational, Y Z. Specificity is automatically created because one of the entities is definite, as it has already created a set, consisting of the object X, to which the second entity can refer. At the propositional level, the two entities in the proposition refer to the same object: X=Y=Z.

However, in Propositions 1, 6, and 11, the propositions do not fit this description. These three propositions are interrogatives. Expanding the Identificational semantic relations to interrogative propositions is the second reason these propositions differ from the complex relation described earlier. In Proposition 1, both sides of the copula refer to the same referent. The interrogative pronoun What refers to the festival name. Both of these refer to the same blank on the learner's chart.

A third reason these identificational relations differ from the earlier described complex relation (and also the traditional syntactic formulation), concerns the properties of definiteness. Higginbotham (1987) notes that the postcopular NP in an identificational construction should be "a singular term, i.e. names, pronouns, or definite descriptions 49-50). In the three interrogative identificational relations, 1, 6, and 11, one side of the copula is the indefinite interrogative What. Interrogative Wh-questions words are always indefinite. However, they are still pronouns; they refer to nouns.

Both sides of the copula in Proposition 1, What's the festival name, refer to the same entity, a blank for information about the 'festival name'. The same is true in Propositions 6 and 11, What and Kamakura age refer to the same blank in Proposition 6 and What and age ofDaichiro Bugaku refer to the same blank on the learner's chart.

However, one side of the copula is morphosyntactically definite in Proposition 1, but \O not in Propositions 6 and 11. In these two propositions, the definiteness/specificity is 0 established by their visible co-presence between interlocutors who share the same Schart and also as part of the context for the speaker referring to the visible words already existing in his visible environment/context of utterance. In all three 0 propositions an identificational semantic relation holds in spite of the indefinite pronoun.

0 0 In both 6 and 11 the copula is missing. This gives a fourth reason these propositions differ. In both propositions the learner does not use the verb of identification to equate the two objects. The identificational relation does not require a CN predicate at a semantic level; the identificational relation is maintained. At the level of semantic relations, the relation is created through definiteness and specificity relations Sof the individuals in the proposition. English has a copula of identification, the N equative copula. However, no verb is actually required for establishing this relation, and many languages do not use a verb to redundantly specify this relation. (In the Sthird identificational relation, the learner again does not use the copula of identification. Again, it is clear that the indefinite interrogative is specifying a relation of identity, what refers to age.) The semantic relation in the composition of Task 3 can be better understood and examined through a graphic representation of the chart itself given below. The top line with the title of the chart and leftmost side represents the chart as viewed by both the learner and interviewer. The second column with blanks represents the learner's chart and the right column the interviewer's chart which has the information the learner seeks.

FESTIVALS OF THE TOHOKU AREA FESTIVAL NAME Tsuzureko Odaik

AGE

AGE 1?00 vears- Learner's Chart Interviewer's Chart The corresponding propositions from the discourse for both the learner and interviewer are repeated below: Example 4.22) Discourse i2-s3:3 s3: Uh...hum..Top of the left blank...What's the festival name? (1) i2: It's uh Tsuzureko Odaiko s3: What Kamakura age? (6) i2: Kamakura age? s3: uhm i2: I don't know but N dash A

N

s3: N? i2: N slash A s3: What...How...What age of Daichiro Bugaku? (11) i2: 1000 thousand 200 years ITo begin the discourse with Proposition 1, the learner first directs the O interviewer to the top left blank on the chart. For the learner, the words indicate a definite visual location and can be interpreted as referring to a definite specific entity.

Yet the blank is empty; only the empty blank with the visual context aids in Sidentifying a referent. For the interviewer, visual spatial searching, as well as the oral queries from the learner, leads to the location of the blank. Until she finds the blank, 00 the spatial referent has not yet been entered into her aural knowledge. It is not yet mutual knowledge in the discourse. It is still an indefinite object. Yet it refers to a specific visual location which she knows exists on the chart.

C"1 From the left side of the chart, the learner repeats the visually presented material written as a label for the blank Festival Name" and asks: "What's the ,I festival name?" At this point, the searching is referring to abstract symbolic spatial symbols, the printed word. However, the native-speaker-like use of the definite article is referring to an empty blank. The label to the left of the blank, "Festival OName", identifies the information which belongs in the blank. It can be interpreted as, "What is the name of the festival which belongs in this blank?" Both members of the dyad can see this label and the adjacent empty rectangle. Although the referent is missing on the learner's chart, the learner knows that an answer to his question exists on the interviewer's sheet.

The point here is that a one-to-one relationship exists between the learner's chart and the interviewer's chart; however, one chart (the learner's) has a blank which is empty and the latter chart has words, or visual objects, in the exact same location on the chart. This creates a visual context with two unit sets, one with no members and the other with a member, or piece of information. The semantic relation between the sets is identificational with both collucators referring to the same object, but the items being referred to differ in their properties of definiteness. The blank on the learner's chart is obviously indefinite since it has no information but it is nonetheless specific, since the learner narrows the context to one particular blank on the chart, whereas the interviewer's matching element is definite (unique) and specific.

This creates an aural context in the common ground of the discourse with a variable referring to the unit set with no members, and a constant for the other one with visual objects in the blank. The semantic relation in both collucators' propositions is identificational with both collucators referring to the same object: What is the festival name? and It is Tsuzureko Odaiko, all refer to the same blank on each chart. However, each proposition differs in its contents and the elements used for referring differ in their features of definiteness. The learner's question word (Q morpheme) What is indefinite but specific, and festival name is visually definite and therefore specific, whereas the interviewer's matching elements are both visually and aurally definite and specific.

The interviewer's properties of definiteness match the claims by Higginbotham, that identificationals are limited to "a singular term, i.e. names, pronouns, or definite descriptions. The corresponding interrogative proposition has an indefinite, specific interrogative pronoun and definite description. Nonetheless, it can be still maintained that it ranges over individuals, i.e. all members of the same set) at the semantic level. The elements which flank the be and are both arguments of the copula, refer to the same individual, and in my data are both specific, if not definite.

Thus, the copula in this category is the identificational "be" which ranges over individuals all members of the same set).

ND 4.3.1.4 Mixed Relations C In my classification of semantic relations, a characterizational is defined as a C proposition which has the purpose of adding a property or characteristic to an already established discourse referent, or definite entity. The interrogative Propositions 2 and clearly ask for properties of a discourse referent. Proposition 2 queries the location 0 of Tsuzureko Odaiko which has just been offered as the name of a festival. The learner is inquiring about its spatial characteristics, the location, and uses the English interrogative form for location, Where. In Proposition 10 the learner asks for a property of Daichido Bugaku, a festival verbally present on the chart. In this proposition the learner uses the temporal interrogative When, a temporal interrogative in English.

C Both times the learner uses the interrogative lexical form for the relevant Sproperty in English--where for spatial properties and when for temporal properties.

I However, in each situation, the learner does not use the form to ask for the specific information of the festival. Instead, the learner uses the form in an identificational CN semantic relation. In the interrogative propositions, the learner asks where is the location and when is the date. He could have asked, What is the location of x and what is the date ofx. These are identificational semantic relations. Or, he could have asked Where is Tsuzureko held and When is Daichido Bugaku (held)? These two are characterizational semantic relations with properties directly assigned to the festivals.

We can say that these two semantic relations incorporate parts of both semantic relations, identificational and characterizational. The difference is graphically portrayed below. I am labelling this type of semantic relation mixed. In these utterances the semantic relation is a 'mixed characterizational-identificational'.

The first term of the relation is characterizational, or asking for a property, and the remainder, including the situation in the discourse where it is found, functions as an identificational relation.

O 4.3.1.5 by s3 Week 1 Summary of Semantic Proposition Types in Task III Week I Summary of Semantic Proposition Types in Task III by s3 Distribution of Propositions Types in Task Ill in Week 1 by s3 14 13

I

12 114 9 8 L 7 9 6 4 2 Types of Propositions Types of Propositions The proposition types for this task in Week I are five Existential, two characterizationals which are Yes-No questions.

4.3.2 Week 3 Semantic Proposition Types in Task III by s3 The design of this Information Gap chart is the same as before, but unlike Week I, this chart only has 8 blanks instead of 12 l xxxiv The complete set of the learner's interrogative propositions are given below.

Example 4.23) Discourse i4-s3:3 ok so let's do some task you need to ask me questions mm in order to complete your chart...please start are these any purpose about Kodomo no hi? purpose ok is to celebrate boys growth and health celebrate boys' boys' growth and health are there special facts about Kodomo no hi? Kodomo no hi is a national holiday as you know national holiday mm national holiday 2started after the World War II when is the date Tsunahiki? Tsunahiki? yeah ok it's February are there any part participant participants in Tsunahiki? of course of men from Kamimachi and Shimomachi participate I\ in this festival s3 Kamimachi i4 and Shimomachi C s3 Shimomachi what's the festival name ah mm (xx) un? i4 which one? s3 un on on right on the right side and i4 the second from the right? 00 s3 next to Tsunahiki (xxx) Momo no Sekku K i4 ee it's Tooji s3 mm do you have meaning of name about Momo no Sekku?(6) i4 meaning? yeah Momo no Sekku is Peach Festival s3 Peach i4 in English Ss3 where is the location Momo no Sekku? (7) Ci4 ah they celebrate Momo no Sekku all over Japan s3 are there any special traits Momo no Sekku? (8) I, i4 mm people display 15 dolls s3 50/15(??) Ci4 15 dolls on the 7 s3 7 i4 7 tiered doll stand s3 7 tiers i4 doll doll stand stand s3 that's all i4 mm really? s3 yes The proposition types in Task III during Week 3 are: 3 Existential-there, 1 Characterizational, 2 Mixed Characterizational/Identificational, and 1 Identificational Semantic Proposition Type.

4.3.3 Summary of the Semantic Proposition Types in Week 3 by s3 The proposition types found in Task III during the third week is given below.

Week 3 Summary of Semantic Proposition Types in Task III in Week 4 by s3 Distribution of Propositions Types in Task III in Week 4 by s3 8 S7 6 o 4 0 0 1 I-2 I o 7 I I Types of Propositions So, the learner is not changing the composition of proposition types for Task III during the course of the three weeks of the interviews. In Week 1, the learner had more Existentials than the other types, 2 Identificationals, 2 Mixed and 2 O Characterizationals. Although he used Identificationals and the Mixed type, he has not O changed his semantic pattern types, still exhibiting a preference for the Existential- I there Type.

4.4 Conclusion SThis chapter has attempted to show the clear presence of the Existential and Characterizational Proposition Types in Task IV by one learner. The lack of 00 Identificationals in Task IV but presence in Task III suggests a difference with respect to semantic relations that these two task types contain. Therefore, we can not say that the learner cannot produce Identificational Propositions. We can only say that they CN are in the minority. The Information Gap Activity seems to trigger Identificational Propositions, while the Description in Task IV elicits either Existentials or C- Characterizationals.

CI This chapter demonstrated that:

(N

S1) Three patterns of relations dependent on specificity and non-specificity Sproperties can be distinguished at the level of the proposition in interlanguage grammars which are basic for semantic meaning; 2) These patterns of proposition types develop in the interlanguage grammar.

3) Semantic patterns of meaning include visible, spatial, and auditory referents.

4) The patterns seem to be correlated with different communicative tasks.

Further research is needed to study the development of the semantic relations of specificity in interlanguage grammars, and also to increase our understanding of the patterns in each communication objective. In other words, as we decompose the tasks for patterns of semantic relations,. we will also be discovering patterns in communication. It is an empirical question whether all languages and cultures use the same set of semantic relations for the purpose of communication.

In the next chapter, I show that these same patterns manifest themselves within propositions. The different patterns of meaning are also found in tasks according to their communicative purpose.

No Chapter SSemantic Roles 00 5.0 Introduction The basis for Li Thompson's (1976) typology of topic prominence versus subject prominence was based on Keenan's 1976 list of more than thirty b-subject Cl properties"xxx'. Keenan clearly states that "the properties may be pragmatic, semantic, or syntactic" 312). Since semantic properties and behaviors can be distinguished c, from pragmatic and syntactic constraints and deletion processes in natural languages, r it is necessary to ask if they can also be distinguished in interlanguage. My focus in 0 this chapter is to study the behavior of semantic properties of subjects and their 0interaction with semantic patterns of definiteness/specificity.

In the previous chapter I argued for three basic Semantic Proposition Types (SPT): 1) Existential-there 2) Identificational and 3) Characterizational Proposition Types. Since these three types are formed by the definiteness/specificity features of the individuals in the SPT, I only briefly mentioned the role of the verbal predicate.

This chapter examines the interaction of these three SPTs with the lexical-semantic behavior of individuals and their predicate. The SPTs are shown to comprise ontological categories which underlay the structures in a semantic component.

5.0.1 Arguments Although the notion of argument dates back to the time of Aristotle, the role of argument in a predicate was introduced into moder linguistic theory in 1879 by Frege in his Begriffsschrift, eine der arithmetischen nachgebildete Formelsprache des reinen Denkens. In his work, propositional forms, arguments of functions, and 'subject' and predicate' were all distinguished Geach, (1970:12-15), Begriffsschrift, Chapter 1, §9 The Function). Frege (1879) proposed that the notion of arguments be distinct from 'subject' and also that arguments within the predicate had a special order, the order of the propositions within the predicate had important grammatical implications (Kneale and Kneale, 1962:436).

Later, in his study of English grammar, Jespersen (1924) (reprinted in Jespersen 1965:145-150, especially #8 p. 149) introduced the notion of 'logical subject', referred to by some Dixon (1979) "as the true universal concept of (deep) subject." still later, certain linguists began to study arguments and their semantic relations: the primitive notions of event logics, such as that of Davidson (1967), Gruber (1967), and Jackendoff (1972); thematic relations (a term Jackendoff credits to Richard Stanley (Jackendoff, 1972:29); Fillmore's case relations (1968); and Katz's study of semantic relations (1980).

However, Chomsky's notion of deep structure left no room for semantic facts in the system of syntactic rules of transformational grammar. In 1982, referring to these semanticists, Chomsky accepted the importance of semantic relations in theories of grammatical relations: traditionally it has "been assumed that such notions as 'agent-of-action,' 'goal of action,' etc., play an important role in semantic description". Chomsky then formally explained the position of semantic roles in moder linguistic theory using an example from Donald Davidson's event logics: IN which analyzes John ran quickly as: there is an event e which is a O running event with John as its agent, and e is quick. Let us assume that LF must be so designed that such expressions as the man, John, he are a. assigned 0-roles, that is, are assigned the status of terms in a thematic CD relation." Let us call such expressions "arguments", as distinct from idiom chunks too much in too much has been made of this 00 problem), non-argument it (as in it is certain that John will win), or existential there (as in there are believed to be unicorns in the garden), terms which assume no sort of "referential function," including names, C' variables, anaphors, pronouns; but not idiom chunks or elements inserted to occupy an obligatory position of syntactic structure (Chomsky, 1982:35).

CI Chomsky's description of arguments, 0-roles, and thematic relations in Sgenerative theory is important for an 'interpretive component' for syntactic theory. In this chapter, an interpretive 'semantic' component, I use the more general term, of

C

N 'semantic roles', instead of 0-roles, for the relationship between the predicate and its argument(s). Arguments are given a semantic role depending on the respective verbs and their function in the proposition.

Today, generative linguistics has two different approaches to the notion of semantic roles'xx'v'. One approach is that too many problems exist to give semantic roles any status in generative theory Zubizaretta, 1987). The other approach says that although the exact nature of the semantic roles is still undetermined (See Dowty, 1991), they play a significant role in grammatical description. Semantic roles have been shown to constrain certain semantic structures in natural language and can also distinguish syntactic processes, such as topicalization, a pragmatic notion, which does not need to be explained by an appeal to semantic roles. This chapter takes the latter approach: semantic roles play a significant role in grammatical description.

In this approach the predicate and its arguments have two distinct representations lexically. The first is relevant to a lexical-syntactic representation and the expression of arguments in syntax Zaenan and Maling, 1982) and their role in syntactic processes, such as control. The second representation is relevant to a lexical-semantic representation. Continuing with Jackendoff's (1972:178-179) claim that referentiality is a semantic property, and that coreference is a part of a network that is handled within the semantic component, I describe the interaction of lexicalsemantic behavior with the three patterns of referentiality established in the preceding chapter.

5.0.2 Semantic Roles Building on Gruber (1965) who first introduced 0-roles in his study of phonetic form Jackendoff (1972, 1983, 1990) developed a semantic, or interpretive, component of thematic relations, showing the relationship between semantics and syntax in generative linguistics.

Jackendoff (1972) was the first to notice that at a semantic-syntactic level, certain aspects of the syntax of arguments are constrained by a hierarchy of the position of their semantic role in a given proposition. Using the five semantic roles of Agent, Location, Source, Goal, and Theme, Jackendoff argued for a 'Thematic Hierarchy Condition' (Jackendoff, 1972:43) to constrain aspects of passives, reflexives and complementation.

IsD Bresnan and Kanerva (1989) increased the number of semantic roles to eight O and proposed their Thematic Hierarchy, ordered according to their behavior in a N predicate: Thematic Hierarchy (Bresnan and Kanerva, 1989) 00 Agent>Benefactive>Goal>Experiencer>Instrument[al]>Theme>Patient>Locative This thematic hierarchy defines the logical subject, or a-subject, as the highest C argument, or most prominent, in a predicate. The a-subject is the highest argument at argument structure of the basic form of a predicate, normally the agent or experiencer c of transitive verbs. It is this semantic hierarchy, which Manning (1996) refers to when ri proposing additional constraints on grammatical relations, organized in what he calls 0 A-structure, which "results from the grammaticization of notions of semantic 0 prominence,...."(p. 33). His model refers to the logical subject as the "Distinguished entity at a-structure", "the a-subject" (Manning, 1996: 34).

5.0.3 Lexical Conceptual Structures (LCSs) Jackendoffs 1972 sketch of semantic representations important for the functional structure of grammatical relations was restricted to semantic roles and their interaction with the thematic hierarchy in structures such as the passive, extraposition, and complementation at the level of the proposition (Jackendoff, 1972: 42-43). In 1985, Jackendoff sketched what he called lexical conceptual constituents, through the lexical decomposition of verbs and semantic meaning of prepositions subcategorized for verbs. This sketch was a combination of grammatical and conceptual structures, incorporating semantic roles and the conceptual "parts of speech", such as prepositions in a predicate. Prepositions are viewed as the grammatical link between visual spatial perception and the spatial/temporal constituents of grammar. He posits [THINGS] as projections of figures from the visual field into grammatical structures.

In order to capture the cognitive connection with grammar, he formulates lexical conceptual structures (LCSs). These LCSs encode a verb's meaning through predicate 'decomposition'. Predicate decomposition assumes a level of representation for the internal structure of verbs and their subcategorized prepositions at both a semantic and syntactic level.

5.0.4 Semantic Proposition Types Jackendoff (1972:Chapter 7 explored the role of specificity in a limited number of propositions, but did not attempt to incorporate these notions in a systematic description of grammar. This chapter revises the semantic interpretation of LCSs by expanding their interaction with conceptual constituents to include the three semantic definiteness Types. Instead of using [THING] as mental representations of the conceptual counterpart of "parts of speech", each linguistic [THING] is coded for definiteness or specificity. As definiteness represents projections from the visual and auditory fields, the treatment of [THING] is coded as a visible, visual, or auditory object (See Chapter Prepositions are still viewed as grammatical links between [THINGS], but are now treated as objects in sets or of sets with properties of definiteness. Jackendoffs LCSs then become formulae for a first step toward incorporating the psychological processes necessary for processing the linguistic structure [THING] according to the psycholinguistic processes described in Chapter 4.

O Whereas Jackendoff refrains from the use of 'proposition' because of its ties to formal O logic and argument structure, I will continue to use 'proposition' to distinguish the C< function-argument structure of Jackendoff's LCS from the three SPTs formulated as Spropositions in the previous chapter.

DThe organization of this chapter is as follows. First I explain the argument J structure and semantic roles composing the three SPTs as used in my data, focusing 0 0 on the roles of theme and agent. Building on Jackendoff's Lexical Conceptual Structures (LCSs), I then integrate semantic definiteness into more complete conceptual structures which utilize both visual and auditory definiteness in the Cl arguments in the predicate. I examine these revised LCSs for properties of definiteness in the five major constructions which Jackendoff first argued belong to a c semantic component, consistent with Manning's model of A-structure: proposition N types of imperatives, passives, and reflexives; and intraproposition types of control \and complementation. In addition, I include in order to constructions, which Farkas 0 (1988) and Kroeger (1993) have shown are also semantically constrained. The data show that these constructions are restricted by the communicative purpose of the four tasks in the experiment.

5.1 Semantic Proposition Types and Eventualities In this chapter, I explore the semantic properties of the types of propositions by going inside the proposition, the intraproposition. This requires the use of functionargument structure, as well as situations and their semantic roles. Although the definiteness patterns differ, the semantic roles of participants in the situations of the predicate-argument structure overlap. The situations of the Existential and Identificational Semantic Definiteness Types, comprise only the semantic role of theme. The Characterizational Type, on the other hand, has a wider range of situations and function-argument structures. The situations also manifest themes, but the function of the conceptual constituents manifest all the other semantic roles on Bresnan and Kanerva's Thematic Hierarchy, as well. This is due to the difference in the eventualities of the predicate.

5.1.1 The Three Proposition Types and Themes The first semantic role which is important for understanding the patterns of semantic definiteness is the semantic role of 'theme', listed in the sixth position on the Thematic Hierarchy. The definition of the semantic role of theme and examples are given below (Gruber, 1965; Jackendoff, 1972, 1983, 1990; Andrews, 2002:9). The italic lettering represents the auditory speech, consistent with the representation of all the utterances spoken in the data in my thesis. The argument(s) functioning as theme is boldfaced: Definition 1: Theme Participant being in a state or position, changing its state or position, or undergoing the effect of some action Existential-there s3:4 There are lots ofpeople on the street.

Identificational s3:2 My favorite festival is Omagari hama(xx) i4:4 I think it's Kanto s3:3 It's Tooji Characterizational s12:4 It is wonderful NO s3:4 they walking walking through the mountain withfire s12:4 That mountain is covered with snow The first conjunct in the definition of the semantic role of theme, "participant being in a state or position" applies to the first example: the participants are in a state or position. The entities from this proposition, people, are located on a street. Hence, 00 this existential-there proposition type has the semantic role of theme.

ri The participants in the tokens under the Identificational Proposition Type also correspond to the first conjunct. In the second and third examples, the participant is Sexpressing a mental state, favorite is a mental attitude and thinking is a mental state.

Although mental attitudes and states are not specifically included in the original definition of theme, Jackendoff does make use of them in his 1983 and 1990 work. In the fourth example, the participant, Tooji, is in a position on the Information Gap 0 chart. Both the it and Tooji identify the missing information for an exact position on Nthe chart, the second from the right.., next to Tsunahiki (xxx) Momo no Sekku.

Syntactically, this is one use of the equative copula; It's highly specific, with spatially guided information to one specific location on the chart. It answers the question, "Which one?" and identifies the name of the festival which is missing information in a specific blank on the learner's chart. Therefore, both it and Tooji are themes.

The tokens in the situations of the Characterizational Proposition Type express all three conjuncts: the participants in the situation were in a state of being 'wonderful', the people were changing position on the mountain, and the mountain underwent the effect of being covered with snow. The last two conjuncts differ because they involve movement. These two features of theme were first proposed by Gruber, (1965) for motion verbs like walk, with theme as "the entity which is in motion (Gruber, 1970: 29)xx"'xv. The motion could be either concrete or abstract, manifesting "a change of position, possession, class membership, activity, etc."(p.

29). Semantically the theme was the entity engaged in the activity about which the situation was concerned.

So, all three Semantic Proposition Types present situations with participants in the semantic role of theme.

5.1.2 The Three Proposition Types and Stative Eventualities In order to understand the function/situation of the semantic role of theme we need to clarify the relationship between an event versus a state, two subsets of eventualities.

Jackendoff (1983:179-183) provides the following diagnostic for distinguishing an event from a state.

Diagnostic Test I: EVENTS versus STATES What happened was that

EVENTS

a. the pig ran away.

b. she put the book on the table.

c. Fred heard about the accident.

d. Louise received a letter.

STATES

e. *the fire truck was red.

f. *Fred loved Louise. (Jackendoff 1983:179) "What happened was" serves as a diagnostic test to distinguish the events in (la-d) from the states in (1 However, Jackendoffs diagnostic test does not have s0 examples of all three types SPTs in Example 5.1. We can amend Diagnostic Test 1 by O adding these tokens to Test l a: Diagnostic Test la: EVENTS versus STATES SWhat happened was that

EVENTS

00 g. they walking walking through the mountain with fire C( h. (?)That mountain is covered with snow i. Kamakura is made by...from snow which is hard

STATES

j. *there were people on the street.

k. *my favorite festival is Omagari hama(xx).

1. *it's Tooji.

c m. *I think it's Kanto.

NO

Consider first the tokens under 'STATES' with the examples of Existential- C, there and Identificationals. Using Jackendoffs diagnostic question for *What happened was that there are [were] lots of people on the street, we see that this is not an event, but rather a state. Nothing has happened; only the existence of entities and their location are asserted. People are introduced with the existential-there simultaneously giving their location, on a street. Thus, we can say that the existential-there proposition type is a state using the copula We will label this BEE to represent the Existential use of the copula 'be'.

The Identificational Proposition Type also manifests a stative relationship between its participants. Example asserts that the speaker's favorite festival is Omagari, the International Fireworks Festival on the west coast of Japan. Using Jackendoffs test we can see that this is also not an event but a state. Having a favorite festival is not a happening; it is a personal preference and/or mental state.

The next example, asserts that it's Tooji and Kanto. Again, using Jackendoffs test, "What happened we can see that both are not events but states. This use of the copula we will label Bel.

Jackendoffs Diagnostic Test 1 for distinguishing states from events does include two kinds of states which are neither the existential nor identificational states.

His diagnostic test question, "What happened was that", distinguishes the states in (le-f) the fire truck was red and Fred loved Louise from the events in These states conform to the pattern of the characterizational type using Aristotle's list of things in Chapter 4. These states have properties assigned to individuals: (le) has a relation of quality, for color between the two things, 'truck' and 'red'; (If) has a relation of #10, passion between 'Fred' and 'Louise', state of loving Louise. The Characterizational Example (Definition 1, from our data also has a stative relation of quality between it and wonderful in the description of the Omagari Fireworks Festival. These uses of quality and/or property as in (le) and §5.1.1 with the copula we will label Bec.

Now let's consider the other Characterizational tokens of themes listed in the amended Diagnostic Testla under Events. Applying Jackendoff's test "What happened was" to the first example, they walking walking through the mountain with fire, we see that walking is an event, or activity function. Although walking is an action, Gruber, Jackendoff, Parsons (1990: 264) all consider a verb such as 'walk+ing' as a process or activity which takes theme as the participant in the situation. Their interpretation is that an animate participant may be engaged in the O process of walking, but the action does not require specific agentive influence to O accomplish the task. The primary goal is to walk from somewhere to someplace.

Thus, "the object in motion or being located" is considered a theme by both Gruber and Jackendoff (Jackendoff, 1987:378) SThe analysis proposed in Chapter 4.3.1.4 stressed that when a person is Vengaged in the process of walking from somewhere to someplace, each individual oO instance manifests itself as a state of walking. If we consider processes to be events with individual instances of an event which together make up an activity, then the person is in a state at each individual instance of the event which result in the specific i activity, such as walking. Therefore, in this short fragment, 'they' are in a state of iterative instances of the event of walking. (Recall that the learner began to divide groups of people or individuals simultaneous to developing the ability to move from existential states to process activities.) SFinally, the second example under Events, That mountain is covered with 0snow, is marked with Cover is an inchoative and fluctuates between a stative and Sevent depending on the context of use. For example, in a descriptive passage containing all final states, this would be a stative. If a temporal phrase of duration, such as '...overnight in the winter' is added, cover is an event. Hence the Diagnostic Test depends on the surrounding context of the situation and cannot be administered independently.

5.1.3 The Characterizational Semantic Proposition Type and Actors The second semantic role which is important for understanding the patterns of semantic definiteness in my data is 'agent', in first position on the Thematic Hierarchy. The role of agent is especially important for understanding the semantic patterns of the Characterizational Type, because this semantic role describes the behavior of individuals within and between the set-subset relations of action sets.

5.1.3.1 Agents or doers of actions Agent, the highest, or most prominent, role on Bresnan and Kanerva's Thematic Hierarchy, has been the most studied of all the semantic roles (See Gruber, 1965; Jackendoff, 1972, 1976, 1987; Lakoff, 1977; van Oosten, 1986; Dowty, 1991). Agents are part of a major class of verbs that have two participants which fill the roles in a particular situation, primary transitive verbs (PTVs), as explained in Andrews (1985:68, 2003:6-8). In a language like English, if one of the participants in the situation performs an action which affects the second participant, then the first is called 'Agent', and the second participant which receives the action is called 'Patient', ranked first and seventh on the Thematic Hierarchy, respectively.

In 1972, Jackendoff, closely following Gruber (1970), viewed Agent as: The Agent NP is identified by a semantic reading which attributes to the NP will or volition toward the action expressed by the sentence.

Hence only animate NPs can function as Agents (or perhaps the other way around, i.e. the semantic relation "potential agent" is a defining criterion for the feature +animate) (Jackendoff, 1972:32).

In 1990 Jackendoff reformulated his notion of agent to include the insights of Talmy's 1985 article Force Dynamics in Language and Thought with respect to agency and causation. He 'dissects' the notion Agent into 'three semi-autonomous IsO parts: doer of action; extrinsic instigator; volitional Actor (Jackendoff, O 1990:129).

The first notion of Agent, Jackendoff calls the doer, or 'Actor' (Jackendoff, S1983:180). The test frame for the notion of Actor or Doer are: SDiagnostic Test 3: Actors/Agents 00 What NP did was a. enter the room. (Bill entered the room.) b. ran down the hill. (Bill ran down the hill.) c. roll the ball down the hill. (The wind rolled the ball down the hill.) d. emit electrons. (The sodium emitted electrons.) e. absorb the water. (The sponge absorbed the water.) f. roll down the hill. (Bill rolled down the hill.) C<1 (Jackendoff 1990:126-128)

NO

In all of the above, the boldfaced NP is the doer, or actor.

C1 Hitting is a highly transitive act which involves one participant performing an action, 'actor', which affects another participant, 'patient'. In the discourse fragment below, some people are the Actors and the Patient is the drum.

Example 5.1) Discourse i3-s3: s3 Some people hit a drum The second type of doer is a 'causer', or instigator. In early versions of Jackendoff's theory of semantic relations, 'cause' was represented as CAUSE and was not always separated out from the notion of Agent. Talmy's 1985 article put forth a major conceptual organizing system for the traditional notion of "causative" (Talmy, 1985: 293), and Jackendoff separated actors as causers or instigators.

With the verb 'hit', as in Some people hit a drum, it is not always clear if the doer caused the action or the action was the result of another action. Learner s13 describes the action of hitting a drum with the verb play, where the doer is causing the action by using an instrument, a stick.

Example 5.2) Discourse i16-sl3:4 i 16 they're holding kind of long stick look like (xx xx) something and usually, usually, you know, if somebody play the drum they're playing like this they are using this kind of sticks like this, right but some people are playing drums like this, holding the stick like this, s13 mm, mm This interpretation of the doer is someone causing something to happen, namely the hitting creates a musical sound emitted from the drum. So, one participant in the situation is holding the sticks for use as a musical instrument, and the drum is 'affected' by the action. In this situation the causer participant hits the drum to 'play' the drum.

However, not all causers act intentionally. In Test 3c above, the wind is the external causer of the ball rolling down the hill, but this is not an intentional action.

The wind is inanimate. Likewise, d and e are not intentional actions.

The third subclass of agents is the class of 'willful' or 'intentional' agents, the first characteristic of proto-Agents in Dowty (1991:572). In the example above, the N0 people were hitting the drums to play music for the Kanto festival. Their actions are 0 volitional and Jackendoff labels this In his Test 3 above, a and b are +vol, c, d Sand e are -vol, and f is ambiguous. We can't tell if Bill's action was intentional or Snot. The Diagnostic alone is insufficient to determine the volitionality of this doer D and action; additional context information is needed.

00 5.1.3.2 Patients or Affected Entity Patients are closely related to the thematic role of theme, but differ with respect to the nature of the action. The traditional the role of patient is "affected entity" (Jackendoff, 1990:125).

Definition 2: Patient Cl Patient Participant being affected by an action, but which may not necessarily undergo a change of state'.

SAndrews distinguishes patients from themes as: "themes needn't be acted Supon by anything", and patients "may be affected by what is done to them, but do not necessarily undergo a consequent change of state", "such as things that are hit or kicked (1985:70). One of Jackendoff 's tests for the role of Patient, or affected entity, is: Diagnostic Test 4: Patient or Affected Entity r What happened S to NP was...

What Y did This test can be applied to the following examples (Jackendoff, 1990:125-126). The affected entity is in italics.

a. What happened to Fred was Sue hit him.

b. What happened to the tree was the car hit it.

c. What happened to the ball was Pete hit it into the field.

The affected entity in these examples also illustrate the well-known fact that direct objects are frequently Patients, as affected entities. However, the affected entity may also be the first or only argument: a. Fred was hit by Sue.

b. The tree was hit by the car.

c. The ball was hit into the field by Pete.

These examples show that Patient and affected entity are determined by semantic criteria and are independent of word order.

5.1.4 The Characterizational Semantic Proposition Type and Action Eventualities We know from the preceding chapter that only Characterizational Semantic Proposition Types include actions, although they also include the non-actions below in Diagnostic Test 5. This distinguishes Characterizational Propositional types from the Existential-there and Identificational semantic relations which are always INO states. Only Characterizational Propositions have agents or actors which are bound to the action.

-i In order to understand the function of the semantic role of agent we need to clarify the relationship between an action versus a non-action. In the preceding Ssection, I briefly mentioned that agents can not be part of the Existential and Identificational relation, both of which express states and not events. Agents, on the 00 other hand, must be a participant in an event. Jackendoff (1983) makes clear that the performer of the action and the action itself are independent of each other.

Considering the performer of the actions as 'variables' in [EVENTS] allows a 0C comparison with [STATES]. States also have variables which are bound to the situation, but the binding results in a different state of affairs (SOA).

C< 5.1.4.1 Actions versus Non-Actions ,O Events can be divided into actions and non-actions (See In Jackendoffs Ssystem, 'actions' is a major ontological category and is a subset of those that express C4 events. Jackendoffs diagnostic test, "What x did was", shows that actions can be distinguished from non-actions, although both are a subset of EVENTS. Notice this test uses do instead of happen: Diagnostic Test 5: ACTIONS versus NON-ACTIONS: What Fred did was

ACTION

a. run away.

b. put the book on the table.

NON-ACTIONS EVENTS c. *hear about the accident d. *receive a lette e. *be red.

f. *love Louise. (Jackendoff 1983:180) "What Fred did was" serves as a diagnostic test to distinguish the actions in from the non-actions in (5c-f) and non-events of Although both actions and non-actions (5c-d) can be a subset of events, non-actions do not require a 'doer' and actions do. Jackendoff labels this participant the 'actor'. Fred, the actor, did something in each of the action sentences, the doer was either a performer, instigator, or causer of the action.

In (5c-f) Fred didn't do anything. In (5c) Fred is an experiencer/goal, but not an actor and in (5d) a goal, but not an actor. We've already seen above that (5e-f) are themes. Therefore, these are not actions, and no actor, or doer, is included as part of these situations. So, Jackendoffs notion of actor is closer to Bresnan Kanerva's notion of agent, which distinguishes these respective roles. Thus, Jackendoffs use of actor is a more specific usage than the broader semantic notion of actor, which sometimes includes agents and experiencers from the Thematic Hierarchy in 2 However, the action and actor still must be distinguished at the level of Conceptual Structure. The ACTION is a major ontological category and corresponds to the constituent VP.

an [ACTION] can be identified independently of who is carrying it out (for instance,) "Joe did the same thing Harry did". Thus an [ACTION] IND is an [EVENT] from which one argument is missing, the one O corresponding to the [ACTOR] (Jackendoff 1983:180).

(N

y If a predicate is construed as an action, the argument (actor/agent) is bound to the action: 0 A VP may be construed as an [ACTION]; the argument position of the verb corresponding to the subject is occupied by the bound variable of the [ACTION].

An action "involves a character who has a special role, the performer of the action." Jackendoff not only proposes that the actor is the one performing an action, CI but also argues further that the actor is the argument which is missing from the action Sbeing carried out (cf. Frege). It is the performer of the action, and therefore the actor is not part of the action itself.

C

5.1.4.2 Causative versus Non-causative Actions The causative structure is very complex. Talmy breaks causative it into two main semantic roles on a par with Agent, Agonist or Antagonist, and their interaction with respect to force (Talmy, 1985:335, Note Jackendoff reformulates Talmy's 'force dynamics' of language into his LCS but retains the multiplicity of his own notions of semantic roles. Below is an example of Jackendoff's function-argument structures for events with causative-type agents. This LCS says that a conceptual constituent belonging to the category Event has the Event-function CAUSE and two arguments (Jackendoff, 1990:43, [EVENT] CAUSE [EVENT, EVENT] Event If the first argument is a Thing, it is an agent; if the first argument is an Event, then it is a Cause. The second argument, the Effect, is an Event (Jackendoff, 1990:44).

Japanese is an agglutinating language with the morphological coding of the causative consisting of the verb stem plus a fairly fixed order of sequencing with respect to the causative (di Biase Kawaguchi, 2002): Vstem-causative-aspect- NEG-tense 290). Kawaguchi's example is 294): Butyoo wa Ruusii-san ni kopii o sase-masi-ta department chief TOP Lucy-Miss DAT photocopy ACC do-CAUSE-POL-PAST The learning situation for a Japanese learner of English is thus both one of learning a new word order as well as morphological change.

5.1.4.3 Volitional versus Non-Volitional Actions If the actor is not part of the action, then are volitional actors also not part of the action? Jackendoff (1983:181-182) offers these two examples to test the role of the actor and action with respect to volitionality. (The examples numbers are his.) (9.43) a. The rock rolled down the hill.

b. John rolled down the hill.

IN (9.44) a. The rock broke the window.

b. John broke the window.

All of these VPs express [ACTIONS]. In Example (9.43) the actors are both themes, and in his (9.44) are both agents. However, they are ambiguous with respect to Sintentionality. We don't know from these sentence structures if the action was O0 accidental or intentional. Conceptually, the two senses can be distinguished as below: rdeliberately a. What John did was roll down the hill on purpose r1 [.accidentally (^K1deliberately CK1 b. Breaking windows accidentally is punishable by death.

kCK1on purpose (Jackendoff, 1983:182) CN These structures show that adverbs can influence the understanding of intentionality in a single proposition. More importantly, Jackendoff argues, this shows they are attached to the VP, not the subject syntactically, so they are part of the VP, or ACTION. In addition, the examples show that a theme (Example 9.43b) may have intentionality as part of its semantic meaningxx""". However, this would require an animate actor, in the same manner as Jackendoffs 1972 notion of agent and implicit animacy.

The verbal lexeme may 'require' volitionality, whether it is expressed or unexpressed in the proposition. Farkas (1988:39) presents a group of verb phrases which can be distinguished as volitional from non-volitional. I have modified the phrases to parallel a Jackendoff Diagnostic Test.

Diagnostic Test 6: Volitional versus Non-Volitional Eventualities What Fred did was

VOLITIONAL

a. be careful.

b. be on time.

c. be considerate.

d. be polite.

e. read the street signs.

f. write his mother a letter.

g. get himself elected president.

NON-VOLITIONAL

h.*be intelligent.

i. *be blue-eyed.

j. *be tall.

k.*be male.

1. *be elected president m.*bleed n. *resemble his father Although a-d and h-k are statives, they differ in their properties of [+Vol].

This is important in English where volitionality constrains certain grammatical constructions, such as imperatives, control and complementation structures (or Equi- IN NP), and in order to constructions (Jackendoff ,1972; Farkas, 1988; Sag Pollard, 0 1991).

5.2 Conceptual Semantic Tier Structure SThe conceptual semantic tier structure represents arguments in a predicate and their J semantic role relations. As mentioned in the introduction to this chapter many problems still exist with respect to semantic roles. One of these problems is that the 00 same argument may have more than one semantic role in a predicate. Addressing this problem, Jackendoff (1990) reformulated his earlier structure to allow more than one semantic role per argument, which he described as increasing the "expressive power C1 of the theory somewhat more radically, along the lines of tier theory in recent phonology", 125). This new formulation consists of two independent levels: an Action Tier and a Thematic Tier.

5.2.1 The Action Tier in the Characterizational Proposition Type O The Action Tier deals with the semantic roles of Actor-Patient relations. Both Agent Sand Patient are given a slot in the representation regardless of their status, expressed or unexpressed. This function, which represents the arguments of the PTV, Jackendoff labels AFF ("affect"). Its two arguments and their possible roles are represented below: a. [AFF (X=Actor or Patient?) b. [AFF (X=Actor only) c. [AFF (Y=Patient only) d. [AFF (implicit Actor) e. [AFF (implicit Patient) f. [AFF (Actor and Patient) In order to avoid ambiguity as to which role, agent or patient, is expressed in a situation, as in Example Jackendoff uses the representations in When a verb takes only an actor, b is used, when the situation has only a patient c is used. When the actor and patient roles are implicit, but unexpressed, d and e respectively, are used. When both actor and patient are present, the function is written as in f (Jackendoff, 1990:126-128).

Three argument types occur in the first argument slot of the Action Tier: (1) doer of action, AFF, extrinsic instigator (first argument of C (formerly CAUSE), and "volitional Actor, AFF+)" (Jackendoff, 1990:129). The first two types may be either or volitional (Vol). If the actor is inanimate, as in c-e in Diagnostic Test 3, [-vol] is used (Jackendoff, 1990:129).

Let's return to the first discourse fragment from the preceding discussion found in Discourse i3-s3:4, repeated below: Example 5.3) Discourse i3-s3:4 s3 and there are a lot of people (T2) i3 Yes...

s3 on the street (T2') s3 Some people hit a drum (T3) The information regarding the individuals in the proposition can be schematically diagrammed as below: Proposition 2-3, Time 2-3 Street(s) d -le 00

(N

The first proposition at Time 2 in the discourse introduces the set 'people' which includes a subset on a street and a subset not on the street. Then, at Time 3 the learner subdivides the group of people into those 'which hit a drum'. Jackendoffs system would describe this as 'actors' (people) performing the 'action' of hitting a drum.

C Thus, the diagram depicts the group of people hitting the drum as a subset of the 0 group of people on the street.

I But this diagrammatic representation does not show how the sets of people are Slinked. The set-subset relation does not tell us anything about the nature of the linking c0 between street-people set-subset and the following people-hit a drum set-subset. It also does not tell us if it hitting the drum is accidental, intentional or volitional.

Jackendoffs functional decomposition for this proposition is: [EVENT] [Action ([THING [THINCY])] This transliterates from Jackendoffs LCS to the data in Example 5.3 above as [EVENT] [,evenHit ([people], [drum])] where THING x people and THING y=drum.

5.2.2 Thematic Tier The Thematic Tier deals with motion and location and specifies "what moves where and under whose agency" (Jackendoff, 1990:126). In Diagnostic Test 3, all of the actors have other semantic roles: Bill is Theme in a, b, and f, as the participant moving; wind and sodium are Source, and sponge is Goal. Jackendoff puts these semantic roles in the Thematic Tier. This results in a two-tiered level of semantic roles, dividing the overlap in semantic roles of actor-patient with the thematic roles of motion and location. Thus, arguments may be described with more than one semantic role.

The Thematic Tier also encodes Talmy's force-dynamics notion of causation and is represented with C for his earlier representation of CAUSE. Causation is then further broken down into a success parameter represented as S, which is further broken down into superscripts of (successful),u (indeterminate), or (unsuccessful) on the function CS. The standard representation for CAUSE is CS Jackendoff (1990:132) gives this example of the Thematic Tier for the verb 'try': Harry tried to go away.

GO HARRY], [AWAY]) CSU Harry], AFF ([HARR) AFF ([HARRY], 0 This LCS of the Thematic Tier shows the causation and success parameter on the first line: the Actor Harry is 'exerting effort' to go away, but his action has an a indeterminate outcome This is because Jackendoff distinguishes 'try' as a verb Swith an indeterminate outcome from verbs, such as 'manage' and 'succeed', which V) always have a successful outcome.

00 As already explained, Actors are independent of actions, but nonetheless bound to the action. The last line 'binds' the Thematic Tier to the Action Tier which formally specifies that Harry is an Actor. The patient slot is generally left blank for C the verb 'try'. This suggests a possible implicit patient of something that has to be overcome, functioning as some kind of OBSTACLE but not necessarily encoded in N, the representation (Jackendoff, 1990: 130-133).

Thus the arguments in the Action Tier always have a role of agent, patient, or Spossibly theme which is bound to the Thematic Tier, which represents causatives and other semantic roles. This allows an argument to have more than one semantic role.

Inasmuch as causatives are actions and have agents as their semantic role, these structures are Characterizational Propositions according to our classification based on semantic definiteness and referentiality.

5.2.3 Semantic DefinitenessTemplate In we demonstrated the interaction of the semantic roles and eventuality with the patterns of semantic definiteness; they are an integral part of conceptual structure and are bound to the meaning of each individual proposition.

Thus, in the same manner that the Action Tier and the Thematic Tier are bound to the proposition, each argument also satisfies a Definiteness function at the conceptual semantics level.

These functions, defined by their definiteness pattern, were diagrammed as set-subset, or set inclusion, and set-intersection relations in the preceding chapter. In this chapter I attempt to reduce the set relations to a 'linear' proposition using Jackendoff's LCSs which more closely matches the syntax and still maintains the spatial and visual relations. The relations between the set collections and unit sets in the various Proposition Types can be better defined by the semantic role relations of their individual members. The sets in the various Proposition Types can then be linked to a linear processing mode in the syntax.

The Referentiality Template is a complement to the LCSs but integrally integrated with the other Tiers. It manifests the independent Semantic Patterns of referentiality described in Chapter 4. Jackendoff recognizes the need for a system of referentiality and appeals to it throughout his work. Moreover, he includes referentiality as a necessary component of Conceptual Structure and Conceptual Semantics. However, his theory is concerned with semantic roles within the predicate argument structure, so does not incorporate definiteness directly into his theory of LCSs. The bracketed items are meant to refer to referential objects but differ in that they do not necessarily include definiteness and specificity.

Since my system of the three semantic relations differs slightly from Jackendoff's use of the terms, I will highlight the differences below. Not only does he not appeal to definiteness and specificity to define the proposition types, he also slightly alters the composition of the propositions to more closely match Gruber's use of the thematic roles, which changes the composition within the types I use. I am not going to argue against his use of the terms in my thesis. I am simply showing how IN my system of referentiality based on semantic definiteness differs from his use at the 0 level of the predicate argument structure.

5.2.3.1 LCS of Existential Proposition Type SJackendoff of course recognizes the well-known Existential Be and even addresses definiteness and specificity in his theory. However, he is concerned with a TYPE- 0 TOKEN distinction which does not concern us here (Jackendoff, 1983: Chapter 6:95- 106). Our approach offers a principled and systematic representation of specificity directly into the LCS; the arguments of the Existential-there proposition are nonspecifics. In my system of conceptual definiteness, the Existential Be is thus not a stipulated but a derived proposition.

Cr Two Existential-there semantic propositions are produced at the beginning of

C

s3's discourse fragment i2s3:4 during Week 1: s3: There are... lots of people on the street' (TI) i2: uh huh N s3: There are many lights (T2) i2: many light? s3: many alot of lights... on the street. (T2') Proposition Time 1-2' OnSt TI Peo le T 2-2' Lights In this diagram we see a street and a set ('lots of) of people at Time 1 (TI).

We can see that on street and people have a set-subset relation at Ti; things are on an entity, the street. The [THINGS] are people. Usually the spatial relation on tells us that a smaller entity (people) is on (top of) another larger entity, (streets). At T2-2' a set of lights is added to the larger set on street. Now we have two smaller sets which are on one larger set, street. In this representation of sets, both people and lights are in an inclusion relation to the street; they are on the street. This labels the sets, and specifies the inclusion relation.

However, this creates a 'boundedness' that we may not want to specify for the set later on. For example, later in the discourse at T5, the learner returns to an entity, people, in response to the interviewer's question, and then immediately reintroduces lights at T6: Example 5.3a) Discourse i2-s3:4 i2: there are any people? s3: Yes... (looking at picture) s3 some people has got light... (6) s3: light is light has....on the stick (looks at i2) (7) Both entities are introduced without mention of the street. In addition, the group of people are subdivided into a group that have a light. Are these different sets of people and lights? No mention is made of either set being on the street, and a new O intersection of the set of people and lights is created. Technically, this means they are not necessarily members of the original set on street. Alternatively, we could divide Q the sets differently and show a different grouping, maybe reverting back to TI and T2.

Jackendoffs system allows us to view the same entity from different Sperspectives and with spatial parameters independently of the entity; the spatial 00 parameters are connected to entities or [THINGS] by prepositions. In Jackendoffs system of LCSs, people on the street is "a two-place relation between a [THING] and a [PLACE] which are 'mediated' by the functional category of the verb BE. The C preposition "on" is "a one-place function mapping a reference object, a [THING], into a [PLACE]" (1983:69). The overall function is a state (ontological category for c Jackendoff).

C, Jackendoffs conceptual formalization for showing the binding of the variable 'thing', or lights, to the Existential BE (cf. Jackendoff, 1983:180) describes a state: Thing State X i [state F(i, Y This 'well-formedness rule' says that in the State function F, with two arguments, i and Yj, the Thing, Xi is a variable bound to the function F. This binding relationship prohibits an 'action' and any semantic relation other than 'theme'. In this proposition, a theme bears the semantic relation of a "participant..., being in a state or position", as stated in the second conjunct in Definition 1: Theme, provided above.

(The dotted lines are an early attempt by Jackendoff to express recursion.) Jackendoff's functional (lexical) decomposition for this proposition is: [STATE] [state BE ([THING THINGXi, [PLACE The function is STATE with the verb BE. The [THING] is the argument and is bound to the verb BE and is in the position of PLACE. This transliterates from Jackendoffs LCS to the data in Example 5.2 above as [STATE] [state BE ([lights], [street])] where the Function is Stative BE and THING X, lights and PLACE y= street. We know from the definition of 'state' and the semantic role of 'theme', that lights (Thing Xi) must be a 'theme,' since state cannot have an agent or actor. Only actions can have agents or actors.

However, we can specify directly what kind of feature of specificity is mediated by the existential state BE, by adding these features to the semantic description of the existential-there proposition and its 'state' relation of the [THING] and [PLACE]: -spec +spec [State BEE lights, I streetYj)].

Combining this with facts about definiteness as provided in Example 5.3a above, we know that we have a semantic relation of existence as the modified rule shows in the same example. This LCS is both a Type, Existential-there and a token of O an Existential-there theme. Although Jackendoff acknowledges the Existential BE O and the existence of the existential in language, his theory does not directly incorporate the means for identifying the Existential in Conceptual semantics. It must Sbe stipulated, and it serves no theoretical purpose. Moreover, the incorporation of Sspecificity (and/or definiteness) is not a necessary (obligatory) feature of Jackendoff's LCS for Existential Be, BEE.

00 5.2.3.2 LCS of Identificational Proposition Type In my formalization of the Identificational Proposition Type, the presence of two [+spec] entities referring to a third, unique, referent creates the meaning of the Identificational Proposition Type. The relation between these two referents is one of C identity; the two referents are identical, or co-refer, and in my data the pattern is Sexpressed with a copula. We have also just established that in the Identificational O Proposition the semantic role relation between the entities in the situation is a theme Sand expresses a stative relation between two entities.

SThree tasks create the identificational proposition between the dialogue partners each receiving input of auditory objects, with the visible and visual definiteness/specificity of the objects differing between the two discussants. The three types of reference objects were: Task II: auditory objects in the conversation describing a hometown festival; Task III: visible referents (and blanks) on the Information Gap chart; Task IV: group of auditory objects describing visible characteristics of picture, creating a visual image of one unique festival identical to the auditory characteristics.

A dialogue fragment followed by a diagram of each of these situations is presented below. The three reference objects are in boldface in each of the dialogue fragments followed by a diagram of the identificational relation.

This following identificational is from Task II.

Auditory Processing Task II: Description of Hometown Festival Example 5.4) Discourse i2s6:2 i2: Uhm..is there any Jap..uh your favorite festival? s6: Uh huh...oh...my favorite festival is Omagari hama(xx) Set of Festivals My favorite festival Omagari Festival The learner selects one festival out of the festivals he knows (his set of festivals) and identifies the unique festival, Omagari Festival, as my favorite festival.

The learner's favorite festival is chosen from a set of conceptual festivals. We can use Jackendoff's LCSs for the Identificational Proposition Type to include two NO [THINGS], yet maintain the state relation with the Identificational BE. In addition, O the two [THINGS] must refer to the same object, X=Y.

[State BE, ([THING [THING This particular Identificational Proposition also has a plurality like the Existential 00 Proposition above. Notice that a plural noun phrase does not exist in the discourse, yet we automatically presuppose the existence of a plurality from which a 'favorite' festival is chosen; the adjective favorite implies a relation of one from a plurality.

C Conceptually, we can represent this type of Identificational from a collection of objects, with many individual members, from which one is selected as: -Aspec +Aspec +Aspec [stateBEE X your favorite festival)] ->[stateBEl my favorite festivall) .jX OmagariFestival SJackendoff's [THING] for the referents of the Identificational are directly instantiated with the lexemes from the discourse fragment, as well as with their respective definiteness properties. The two references to the antecedent, your favorite festival, are labelled +Aspec for the auditorially specific answer by learner s6 to the question by interviewer i2. The lambda extraction symbol is meant to show that all three NPs are bound together for the LCS of the Identificational Proposition. Notice that the first NP is in an identity relation to the last two NPs.

This binding relationship prohibits an 'action' and any semantic relation other than 'theme,' where a theme bears the semantic relation of a "participant..., being in a state or position," as stated in Definition 1 for Theme and Diagnostic Test 1 for State given above in §5.1.1 and 5.1.2 respectively.

The next Identificational Proposition is from Task III, which contains visible information for both discussants: an empty blank needing to be filled for the learner and a printed visible answer in the text on the chart for the interviewer. In the discourse fragment below, Lines 149 to 152 are spatially guiding the interviewer to the blank with the answer to the learner's question; the spatial language is underlined in italics and the antecedent and subsequent anaphors are in boldface italics.

Visio-Spatio-Auditory Processing Task III: Information Gap Activity Example 5.5) Discourse i4-s3:3 s3: 148 Shimomachi what's the festival name ah mm (xx) un? i4: 149 which one? s3: 150 un on on right on the right side and i4: 151 the second from the right? s3: 152 next to Tsunahiki (xxx) Momo no Sekku i4: 153 ee it's Tooji The Identificational Proposition in line 153 provides the visible referent for the missing answer for the FESTIVAL NAME, Tooji. This identificational relation is represented in the following diagram: FESTIVAL NAME Tooji IN In this diagram, we see that the Identificational in the Information Gap Activity is in a 0 unit set relation: each blank contains only one member in the set. In this discourse, Cithe only visible member in the set is Tooji.

The Identificational Proposition is based on visible, spatial information, J presented auditorially to guide the interviewer to the location of the visible referent in the blank on the interviewer's chart. This identificational proposition can be 00 represented in the following LCS.

-ASpec F+AVisiSpa ec +AVisi oaSpec (N [stateBELX( Wlat )]<->[staeBEi( It Tooji t"- In this LCS for Identificationals, Jackendoffs [THING] has again been directly Sinstantiated with the verbal elements and the definiteness properties. The antecedent, What, from line 148 is represented as auditorially nonspecific, [-Aspec]. When the interviewer says the auditory object It, she is looking at the information in the blank on the chart, a visible, spatial referent, but referring simultaneously to the antecedents What, Festival Name and articulated auditorially. Finally, the interviewer articulates Tooji, referring to the visible referent in the blank and textually present, coded as +A VisiSpaSpec.

A final example of the Identificational Proposition Type is found as the last utterance in the description of Task IV, as the interviewer guesses the identity (name) of the festival. The referent of the first entity is the set of a collection of distinguishing characteristics which help the interviewer visually identify the festival.

This requires visuo-acoustic processing based on a visual mental imagecreated by the learner's auditory description. A discourse fragment from Task IV is repeated below: Visuo-Auditory Processing Example 5.6) Discourse i2-s3:4 i2: I have some picture of festival so you can pick up one and don't show me s3: uh huh...

i2: And describe.it and I will try to guess which festival...

PICTURE DESCRIPTION by s3...

s3 That's all. Can you guess? i2 Hhm. I think it is Kanto festival A diagram of the sets of this identificational relation is given below: Auditory Linguistic Objects Visual (Mental) Image Festival Name \0 cars, taeko drunis cars, taeko drums {<Kanto 0 Festival>} Sstreet, sticks street, sticks 00 MANY MANY One Unique Cr Festival The learner's description is presented as a set of auditory linguistic objects C which create a collection of visual images as members of a set for the interviewer.

The visual image, mentally representing a festival in the Tohoku area of Japan for the 0interviewer is then produced as an auditory object in the form of an identificational Sproposition in the discourse. This identificational is presented in the final proposition I think it is Kanto.

The proposition I think it is Kanto, I think X and X= it is Kanto can be interpreted in several ways. First, the mental image in the discourse is created from the learner's description, (auditory objects), as an identificational referring back to which festival, in the introduction to the task. Or secondly, which festival is referring back to one from the set of pictures of festivals. The one then is a unit set of one festival from the folder of 'some 'pictures offestivals. The it in describe it, from the introduction to the task, becomes the referent for the it in the final proposition. In this interpretation, the entire identificational proposition in the final line of the discourse are linked auditorally to the picture, one, and it, and which festival.

A third interpetation is that the auditory objects ranging from TI until T9 result in a final image, and the interviewer produces a proposition with It created mentally as an image of the contents of the picture producing the referents it and Kanto. In this interpretation, It refers simultaneously to all of the above referents as well as the mental image of the picture in the interviewer's mind, and Kanto refers exclusively to the picture.

A final interpretation, closer to Jackendoffs, is that It and Kanto are nouns, resulting in an equational proposition It is Kanto, syntactically. If we only look at the syntactic characteristics of the relations of the nouns in these three examples of Identificational Propositions, they all match one of Jackendoffs examples of an identificational relation: 'Clark Kent is Superman' (Jackendoff, 1983:95).

Jackendoff just asserts that this is an identificational, following the traditional syntactician's notion of an equational sentence, or identificational.

However, if this proposition is in a list of characteristics in a discourse fragment about either Clark Kent or Superman, then it is not an identificational. For example, we might be describing Clark Kent by listing that he is an alien, a reporter at the Smallville newspaper, in love with Lois Lane and is Superman in disguise. Then, the proposition is an example of 'exhaustive listing' of properties or characteristics and functions as a Characterizational Proposition. The noun Superman is simply a property among several others; it does not identify him.

Jackendoff (1983) also includes properties as identificational: 'Sally is three inches shorter than Bill' (Example 10.20a, p. 197) and Bill wants Harry upset, Bill kept Harry angry (Examples 10.33c). We treat these as properties in Characterizational Proposition Types. They are identificational only in so much as O they distinguish Sally from Bill, or Bill's desires and behaviors with respect to Harry Sversus another individual. This sense of identificational is not based on referential Sidentity, as in our use.

SFinally, following Gruber (1965), Jackendoff also considers propositions, such as "Bill turned into a cook" as identificationals; their identity changes over time. We 00 would also include this as a Characterizational Proposition Type since Bill is still Bill, but has simply added a new property, or characteristic, to his list of skills. The person Bill at time A and time B are not the same. Gruber's use expresses a transition of a Cl class to which the subject belongs" and explicitly writes "It can not be used merely to signal an object previously referred to" (p.

1 1 Gruber and Jackendoff are C interested in classificatory properties of lexical items, properties that group classes of CN lexemes.

SOur classification of the Identificational Proposition is based strictly on its Sproperties of referentiality, within the proposition (or sentence, syntactically equative) Sand its relation to existing referents in the respective proposition or discourse. It is this system of referentiality that Jackendoff himself claims is a part of conceptual structure (Jackendoff, 1972:178). Appealing to the syntactic characteristics of the Identificational Proposition and not defining it according to properties of referentiality and co-reference are syntactic, not conceptual semantic structure. Gruber and Jackendoff's emphasis on semantic roles, in this structure 'theme', de-emphasizes referentiality and its importance as an underlying system (Gruber, 1970:27-28).

5.2.3.3 LCS of Characterizational Semantic The Characterizational Proposition differs from the Existential-there proposition which asserts the existence of an indefinite nonspecific entity and from the Identificational which links one specific entity it to another specific entity in the discourse network. How can we use Jackendoffs theory of LCSs to distinguish this proposition from the preceding Proposition Types composed of states and themes? The two propositions (le-f) which are states in Jackendoffs diagnostic test can be distinguished from event propositions because they both assign characteristics or properties to entities in existing sets. In Diagnostic Test 1, his example e. '*the fire truck was red', is a Characterizational Proposition.

Although Jackendoffs LCSs account for this type, he classifies it differently from our use of referentiality in the Existential and Identificational Proposition Types.

To distinguish the three proposition types, we can modify his LCSs. Jackendoff s first example of a state can be graphically displayed as below: Truck red firetr I w iretruc r e d Quality Characteristics THINGS PROPERTIES This diagram of the set relations for a 'red firetruck' shows several sets.

Semantically, these constructions are composed of individuals from existing sets

I

of THINGS, a subset of THINGS a set of PROPERTIES, a subset of STHINGS with properties. Thus, red firetrucks are Characterizational Propositions formed of an existing set, or definite entity 'The firetruck', with a set composed of Sindefinite properties, the color 'red'. Thus, we can classify the semantic relation in Sthese propositions as Characterizational Proposition Types.

SThe set of properties is not in an inclusion relationship to the firetruck. The 00 set with 'a firetruck' and the property red is linked; the properties are characteristics of the firetruck. A learner can learn the word firetruck independently of the words to describe it, especially if a picture is shown to the learner. Likewise, the characteristics Ci of a firetruck are numerous and more properties can be attached to its description, such as 'truck with ladders, 'shape', 'motor vehicle, and colors other than 'red', such C as yellow. So one set describes the set of properties linked to the entity of a firetruck.

We can use his system, but we must modify it to distinguish it as an ontological type, \distinct from the other two. In this example we have a [THING] and a S[PROPERTY] in a two-place relation mediated by the verb 'be' (Jackendoff, 1985: S69).

We need to distinguish this Characterizational state from the Existential BE and Identificational BE, including the pattern of semantic definiteness in the description, r +def rdef- [state BEc firetruck j red where thing x firetruck and property y= red. We know from Jackendoff s working definition of'state' and the semantic role of'theme', that this must be a 'theme,' since state cannot have an agent or actor. Only actions can have agents or actors 180).

Returning to the example under Jackendoff's Action Tier, we need to complete the feature of definiteness mediated by the EVENT HIT. We also need to show, or spell out, the types of definiteness in the propositions, visual, spatial, temporal, or auditory. We can add these features of definiteness/specificity to help complete the description of the Characterizational Proposition and its 'action' relation of the [THING] and [THING]: [+AVisiSp -AVisiSp [EVENT HIT peo umY We know from Jackendoffs working definition of event and action, that the first [THING] has the semantic role of 'agent', since only actions can have agents or actors 180). Since this proposition has an agent and an action, this proposition is a Characterizational Proposition with an action characterizing the agent, +specific people.

4 s3 there are a lot of lantern in the street (1) 4 i3 mm, mm 4 s3 and there are a lot of people (2) 4 i3 yes 4 s3 on the street \O 5.2.4 Action Tier, Thematic Tier and Semantic Definiteness Template LCS rC In a revised version of Jackendoff's Conceptual Tier Structure below, I have included Svisual and auditory definiteness and specificity. I have re-arranged Jackendoffs Tier structure so that the Thematic Tier is last. All positions of the first argument are marked with The token for this proposition is: I try to study, learn really English 00 pronunciation. I have only represented I try to learn English where English=Eng and pronunciation=pron.

Try

V

r to Sj r'l Def Definiteness Template:two Specific arg, one unexpressed D EventAFF ]ai, AFF Tier: one vol arg and one unexpressed arg SCS" ([/]ai,[EventAFF learn([ ]'j,Eng pron)j) THEM Tier: Undetermined complement subject -bound to Event The first two lines of this LCS say that try is a verb; the third line says that it takes a complement with to and a subject Sj The fourth line, the Definiteness Template, says that I is specific and takes a nonspecific complement. The fifth line says that the event is a voluntary action and a complement. The complement is an attempt to accomplish a certain Effect and its Actor, Sj is bound to the Actor of the main clause, signified by ac. Jackendoffs LCS is representing the obligatory subject control of try (Jackendoff, 1990:148).

The importance of this addition to the LCSs is to track the interaction and pattern of definiteness within the verb-argument structure in the proposition. The three patterns give added meaning to the proposition beyond just the semantic role structure of the verb.

5.2.5 Mental Resources for Processing Definiteness of LCSs Jackendoff claims that his conceptual structures are one part of three autonomous levels of mental representations: phonological, syntactic, and conceptual. The conceptual structure serves as an interface between linguistic information and other capacities such as vision and action (Jackendoff, 1993). As far as [THINGS] and the LCSs mapping or linking to the level of syntactic sentential constituents, my work suggests this is very useful tool for understanding the structures at the semantic level of second language acquisition and learning.

However, even though I agree with Jackendoff's autonomous levels of representation, my work shows that another level of systematic relations between [THINGS] can be distinguished. This level is composed of the three ontological patterns of semantic definiteness, and [THING] and the other eight conceptual categories are composed of the nine psycholinguistic resources for establishing_and processing definiteness.

In the Existential-there Proposition Type, instead of connecting [-def THINGS] and [+spec PLACE] directly to vision and action, I refer to these nine mental resources which form the psycholinguistic characteristics of the linguistic property of definiteness presented in Chapter 4. This token utilizes simultaneity, attention, rationality, locatability, recallability, and understandability in order to process lexical constituents as mental representations. All of the resources converge ISO to create the mental representation of [THING] and [PLACE], of [THING] and O [THING] in the Identificational Type, and [THING] and/or [THING]/[PROPERTY] Sin the Characterizational Type, whether it is a part of speech constituent, an entity, or e. an eventuality. Jackendoffs semantic structures are a convenient template for Sanchoring the respective pattern of definiteness and the converging resources into the grammar.

00 These I claim are necessary resources for processing a [THING] and other conceptual constituents. I am not claiming that these exact LCSs are the mechanism for joining the [+specific] and [-specific] NPs in the discourse. But I am suggesting C1 that the mechanism for joining the NPs is not accomplished with simplistic first order logic relations. Rather, Jackendoffs insight that semantic roles and the lexical decomposition which it provides for understanding the linking mechanisms involved Sin joining arguments is a better approach as suggested by my data. The arguments as presented in Chapter 4 are sets and subsets with properties of definiteness and 0specificity. These can be linked to mechanisms of memory and attention as I have Ssuggested in Chapter 3.

5.3 Conceptual Semantic Constraints on A-Structure Propositions As mentioned in certain syntactic structures and processes, such as the processes of imperatives (Jackendoff, 1972; Manning, 1996), passive, reflexive binding (Jackendoff, 1972; Manning, 1996), in order to (Jackendoff, 1972; Farkas, 1988), and control and complementation structures (Jackendoff, 1972; Fodor, 1974; Bresnan, 1982; Farkas, 1988; Sag Pollard, 1991) have been shown to be more clearly and succinctly explained by appealing to semantic constraints rather than to syntactic constraints.

It was the absence of the semantic role of agent in some sentences that led Keenan to state that "semantic roles cannot be used to identify subjects of sentences, [since] sentences of that sort [unexpressed, agentless] will be numerous among the bsentences in a 321). However, the unexpressed arguments of the imperative and complement structures will be shown to be semantically constrained by or dependent on the semantic role of agent. These structures were all included in Keenan's properties list, but were mixed together with his list of pragmatic and grammatical notions as well. He did not include in order to constructions, which will be shown to require similar characteristics of the a-subject to these other constructions. Although the semantic properties of passives and reflexives behave differently with respect to unexpressed arguments in these other constructions, they will also be shown to be constrained by the behavior of agents.

These structures have long been known to be sensitive to semantic constraints of co-reference and/or the Thematic Hierarchy (Jackendoff, 1972). In addition to Jackendoffs constraints according to the Thematic Hierarchy, I will add the semantic definiteness meanings in the representations with the Semantic Definiteness Template. This allows a more complete analysis of the meaning of each conceptual proposition.

5.3.1. Imperative Conceptual Structures The communicative intent of the imperative is for both eliciting information and giving commands. This is consistent with the description of the use of the imperative by (Jespersen, 1965): It [the imperative] is a will-mood in so far as its chief use is to express the will of the speaker, though only-and this is very important in so IND far as it is meant to influence the behaviour of the hearer,....

O Imperatives thus are requests, and, these range from the strictest C command to the humblest prayer (Jespersen, 1965:313).

dFarkas characterizes the purpose slightly differently, An imperative sentence Sis interpreted as the speaker requesting (or ordering) the addressee to bring a situation 00 about." Thus, the "first argument slot, is linked to the addressee" (1988:39). Both C1 agree that the first argument slot is linked to the Hearer or addressee, even though it's not expressed.

C If the first argument slot is linked to a person, then knowledge of the hearer's existence can be inferred from the speaker's utterance. The remainder of the proposition is adding information to our knowledge of that person; since the 0 imperative is a command, it is requesting or demanding future actions from the C1 addressee. Thus, we are receiving information about the addressee's potential future Ibehavior, or actions. As we've seen the Identificational Proposition is a stative relation and does not include actions. So, the imperative does not have an existential 0 or identity function. The imperative is a Characterizational Proposition Type.

However, the specificity conditions are not necessarily linked to an anaphoric situation in the discourse. Instead, the specificity of the actor position may be linked to visual definiteness created by the context of the discourse; the speaker would not be asserting instructions to a non-existing entity. In addition, this proposition does not add properties to the first argument. This characterizational proposition type is composed of a [+spec] and future characteristics.

5.3.1.1. LCS of Imperative Propositions The imperative lends support to Jackendoffs claim that thematic roles and conceptual structure can help explain otherwise adhoc principles in syntactic theory. For, the imperative, as we have already noted, must have an addressee which is capable of performing the action of the event structure of the imperative. Even though the performer may not be mentioned, it is implicitly bound to the action. In addition, the thematic role of a 'willful' performer is implicit in the construction. Jackendoff (1990:268) proposes this "interpretative" account of the subject in the imperative as the correspondence of conceptual arguments to syntactic positions in a possible nonstipulated Linking Theory: Conceptual Structure for Imperatives"'ci In a subjectless main clause with a tenseless V, satisfy the External Argument with [YOU](1990:268).

Since the subject is bound to the action in conceptual structure, special exceptions for the imperative at the level of syntax are not necessary. It is part of a more general semantic network of conditions which apply in the absence of syntactic structure.

Generally, this account works. It certainly describes the data in my study.

However, it can also be used in propositions such as "Somebody do something!" As this is the indefinite non-specific use of the imperative, this token proves that the imperative proposition can be both specific and nonspecific. Only when the proposition refers to you, is it specific to the Adr (either singular or plural) and I addresses the potential doer of the action asserted in the proposition. Hence, we need Sto code the unexpressed argument for features of definiteness and specificity.

Another semantic constraint proposed by Jackendoff, discussing Gruber's Q thematic hierarchy (1965, 1967b), is that "imperatives are permissible only for Agent subjects. This is due to the fact that the possibility of successfully carrying out an order depends on the order requiring volition agenthood) on the part of the 00 hearer" (1972:33).

To test Jackendoffs account of imperatives, let's begin with describe the festival. We already know that the verb is an ACTION verb. This is represented in CN the top two lines and bracketed on the right side with 'action verb'.

C, describe action verb (N

V

IN Event([+VisiSpec]a,[ +VisiAudSpe EVENT has two Specific args, both visible, I one auditory 0 Event ((AFF+vol ]a AFF- [fes iv EVENT has two args:(l)implicit, volitional, neutral ,[EVENT+VOL ([YOU] a, [festivalJ] bound to YOU in one volitional action The third line represents the Semantic Definiteness Template. The first argument, the Addressee who is visibly present to the speaker is implicit, although unexpressed. The representation of specificity and definiteness at the level of conceptual structure makes this clear. The second argument, a picture of the festival, is both visibly present and also auditorially present to both in the situation. So, the two arguments have to be independently coded to represent the different processing requirements.

The fourth and fifth lines represent Jackendoffs LCS Action and Theme Tiers, respectively. In the fourth line, arguments for the imperative are conceptually represented as [AFF for an action with two arguments. In this LCS, ]a YOU and is visibly present and specific to the speaker, but unexpressed, the implicit argument of the addressee of the imperative. The second argument, is represented as festival and is auditorially represented along with the visible picture in the situation to be described.

In addition, the first AFF argument must be volitional; the argument which represents the addressee must be willing to do the action. This argument is unexpressed, so no argument is represented. The second argument, festival, is neutral with respect to being affected so we can label this AFF Notice that this interpretation is consistent with the approach stated above, that the "first argument slot, is linked to the addressee" (Farkas, 1988: 39). The addressee and [YOU] are the same argument. Moreover, with the addressee or [YOU] as the satisfying condition of the imperative proposition at Conceptual Structure, any syntactic main clause which is "subjectless and has a tenseless V" is automatically linked to the imperative. The linking does not need to be stipulated.

The fifth line represents the thematic tier. I represent the C for cause and S for successful action; the action can be determined to be successful. I am representing the imperative as a CS since the action must be capable of being instigated by a volitional addressee, one type of a causative agent. In addition, the action can be determined quickly to be successful. Since the action itself must also be a volitional IND action, and not just be directed at the addressee, I label both the action and the 0 addressee +Vol. As a second language learner, the learner must not only want to Sperform the action, the learner must be capable of using whatever skills are necessary to complete the action.

DThis last interpretation, that an imperative is a type of causative in addition to j requiring a volitional action, depart from previous interpretations of the imperative 0 0 construction. In addition to volitionality, the imperative has a requirement that one must be able to cause it to happen. Jackendoffs formulation allows us to mark it CS+, meaning that the outcome of the action can be determined to be successful. The CK, seventeen verbs used in the imperative structures in my data are all actions which may have a volitional actor as well as having an outcome whose success can be determined.

ask, be, call, describe, explain, go, keep, locate, look, pick, show, take, teach, tell, try, use, Swrite.

C

Jackendoffs requirement that the addressee be an agent is not supported by my data. This utterance was found in my data at the beginning of the introduction in Task I.

i3 Be seated Nonetheless, the single imperative with a patientive addressee "Be seated" does satisfy Jackendoffs claim that the external argument be YOU. This use of the patient YOU has strong agentive characteristics and it is Remember that Jackendoff originally coded Agent as a characteristic also noted by Dowty (1991) in his proposal of an argument selection principle. In addition, this is clearly an action which the Adr is capable of causing to happen x Be seated verb

V

Event([+VisiSpec]',[ +VisiAudSpec]) EVENT has one Specific arg, visible and auditory Event ((AFF+voI( EVENT has one volitional arg ,[EVENT+VOL ([YOU] a [festi bound to YOU in one volitional action Appealing to the notion of volitionality for this structure is consistent with the analysis by Farkas (1988:38) which contrasts types of actions which are statives found with imperatives. The use of Be seated! then can be interpreted as an action which is [+Vol] and is consistent with Jackendoffs criteria of type of agenthood found in imperatives. The action imperative, the passive imperative and behavioral statives all require [+Vol] or a [+Responsible] for the action.

It is interesting nonetheless that the agentive and patient arguments are the two semantic roles which alternate in the first argument position in the two imperatives.

The examples with be could also be understood to mean "cause yourself to be come seated", which is more agentive, causing a 'semantic shift' from the patient to a more agentive argument. This kind of semantic shift is induced in contexts such as imperatives which normally require agents but instead are found with non-agentive clauses which are not sufficiently agentive IDThe addition of definiteness and specificity to Jackendoff's LCS requires 0 different psycholinguistic processing than expressed in his work. For, he sees an C, arguments as a [THING] which relate directly to visual perception and the object it represents in the LCS. By representing the first argument, YOU, as visibly present d) and therefore visibly definite, YOU is bound to both the action of the token and to a Vvisible real world entity.

00 The use of definiteness into the LCS brings in a completely different aspect to the structure, processing load. The first argument requires visible processing of each other for both speakers at the time of the utterance. The second argument, the picture of the festival, requires both participants to be visibly attending to the same object.

These two arguments require access to the mental resources of simultaneity, attention, rationality, and locatability in order to process the visible definiteness situation of the CK1 utterance (Chapter 4).

SThe second argument, the picture of the festival, requires both participants to 0be visibly attending to the same object and understanding that the picture is the argument referred to. Since the second argument is orally presented, the learner must be able to understand that the picture is of a festival. The learner must understand that he is to describe the festival. This requires the learner not only to understand the lexeme 'describe', but also to be capable of producing a description orally and linguistically in the situation. The learner must be able to both know words and to be able to pronounce the words to the hearer. This involves the constraints of simultaneity, attention, rationality, locatability and understandability for identifying the objects in the oral situation.

Thus, the addition of definiteness to the LCS brings psycholinguistic processing at a more detailed level than simply to claim that [THING] is an object that must be visually perceived.

5.3.1.2 Interpropositions of Imperatives Since the imperative is an Characterizational Proposition with the addressee being [+Spec], and the action being indefinite, we might ask if the possibility of an interpropositional imperative semantic form is possible. Interpropositons cross between propositons receiving properties of specificity from an earlier mentioned definite entity. In other words, how can the argument receive specificity properties due to prior mention in the discourse.

Below is one example from the experiment that shows it is possible to have the imperative linked to the another proposition in the discourse. The imperative referring to prior action is boldfaced.

i2 I have uh some picture of festival, so pick up one...don't show me...don't show me....aaaa!<laugh>..try once again you look then uhm describe it In this short fragment, the interviewer is giving the learner instructions and the learner doesn't realize that it's a guessing game. So, the learner started to show the picture of the festival to the interviewer before starting to describe it. The try once again means 'You try one more time to choose a picture (pick up one), look at the picture but don't show me'. This refers back to each of the actions necessary for the process of looking and describing the picture, but in particular not showing the interviewer. So the events leading up to the description are previously mentioned. Each event has an actor, namely YOU, bound to the action. The repetition of action is indicated by the

I

I\0 adverb 'again' which indicates to repeat the action. The information of repetition is O [+spec].

try action verb V L once again 00 Event([+Aud/+Spec],[+Aud/+pe:]) EVENT has two Specific args, both uditory SEventAFF+vol([ EVENT has one volitional actor C argument CS [ActionAFF([a], Argument (CS bound to Event

(N-

So it is possible to have interpropositional imperative structures, but this is not 0 common and not a predicted semantic rule of the structure. I return to the discussion Sof interpropositions of try below in §5.4.1.1.2.

5.3.1.3 Imperatives and Macropropositional SPTs The imperatives of the learners were found predominantly in Task III, the Information Gap Activity, produced by only four students, s2, s5, s6, and sl0, with s5 producing seven out of the overall 13. This activity required students to ask for information in order to complete the chart of festivals.

The interviewers, on the other hand, used the imperative throughout the four tasks in the experiment to direct the use of the tasks, but less frequently in Task III, where they were being asked questions by the students. In almost all cases, the imperative is used as the opening statement to the respective task, instructing students how to perform it. In other words, these were not interrogatives, but rather imperative structures used as commands or directives in the context immediately preceding each task"".

When the imperative was located within a task, it continued to function as a directive. These occurred when the student was prompted to provide more information internal to the description of a hometown festival in Task II: i2 please explain the activity of the festival i3 tell me another festivals, famous festivals The interrogative which immediately follows the imperative at the beginning of Task II by i3, please tell me mikoshi...what's mikoshi?, is used for clarification of the instructions: i3 wants information describing mikoshi, a festival near the hometown of the learner (s8).

So the communicative intentions of the speaker/hearer affect the macrolevel of the propositions produced throughout the entire discourse as described in the previous chapter. In these discourses, neither the Existential Proposition Type nor the Identificational Type is included to perform the actions of the task, especially Task III for the learner and the directives given by the interviewers at the beginning of each Task. All of the proposition types within the discourse are Characterizational with an unexpressed Actor who is visibly and auditorially present during the completion of the task.

NO

0, oo 5.3.2 Passive Conceptual Structures.

Passives can be identified when the normal subject position is assigned the semantic role of patient by the verb, which Jackendoff terms the grammatical patient (Jackendoff, 1990: 294, ftnte. Since the patient role precedes the agent in the proposition, the patient is more prominent than the agent role in the passive proposition. As such, Jackendoff (1972:43) proposed the following condition on the application of the passive transformation: Thematic Hierarchy Condition The passive by-phrase must be higher on the Thematic Hierarchy than the derived subject." ci Jackendoff's 1972 condition concerns the order of semantic roles with respect to passive by-phrases and is a distributional fact for propositional structures at the Lexical Conceptual Level, as well as the level of syntax. In spite of Jackendoff's use of the theoretical notion based on transformation grammar and derived subjects, the Bresnan and Kanerva Thematic Hierarchy (1989) applies as a distributional fact about the passive by-phrasex".

5.3.2.1 LCS of Passives The use of the passive in interlanguage is not always as straightforward as the imperative structure in the interlanguage. Lexical usage by the language learner may complicate the interpretation of the semantic role of patient, as seen below.

Example 5.7) Discourse il0-sl2:4 s12 night, yes, and on the dark sky il0 do you mean ball s12 no, ah, ah one ball, some balls ilO mm,mm sl2 is a, a jumping il0 mm, mm s12 on the sky il0 into the sky s12 into the sky, s12 then that, this ball it is destroyed il0 destroyed, mm, mm s12 then fired il0 fire, okay fired, so small ball burnt into the sky (2) (7) (7,")xcvi (8) (9) An explosion occurs and creates an image that the student describes as jumping, destroyed and thenfired. In the TL, destroyed is a transitive verb and we expect two arguments with the first being an instigator of the action. The second action, fired, does not have any argument in the IL, but we can understand that the ball is still the object being referred to. This discourse fragment from T is graphically illustrated below.

Propositions 2, 7-10, Time 7-9 T2 onalls jumping into the sky T7 Sb balls jumping into the sky 00

I

T8 This ball d o O T 9 fired V, This ball is jumping into the sky at T7-7' and at T8 is destroyed. The sequence of 00 actions leading up to the passive at T8 all describe objects, i.e. balls, doing something.

The S expresses the auditory object balls. Until we have confirmation from H we don't actually know if H has heard, comprehended or even understood. However, between T7-T9, the H is repeating the s12's auditory objects so we can assume that he has understood and comprehended. Most importantly we can assume that he was "attending" to the auditory objects since he was able to locate mentally how they were fitting together in his summarizing statement after T9.

Below is the LCS for the speaker's utterance, it is destroyed. The verb destroyed is represented in the top two lines and bracketed as 'action verb' on the Sright side.

destroy action verb

IV

Event [+VisiAudSpec]) EVENT has one arg, one suppressed.

Event (AFF+ one affected CS i, [EVENT( bound to EVENT, with one Affect arg.

The third line represents the definiteness features of the single argument, it.

This argument is in the picture so it is visibly present to the speaker as he pronounces the lexeme to H, so it is marked +VisiAudSpec. Its visible presence makes it +Spec to the speaker. It is auditorally present to both the speaker and hearer.

The fourth line represents the action tier. The action verb is represented as an Event with the function AFF This indicates that it is a function with a verb that takes a patient, or affected entity. Since the action verb, an Event, can take two arguments, the first argument slot is blank which indicates that one argument is unexpressed or implicit. The second slot is filled by it, the affected argument marked by AFF The action verb destroy has an implicit argument; someone performed the action of destroying. But, does the learner mean destroy? Is there an actor when a ball of fire is in the air? Perhaps the learner means 'explode', but did not know the English lexeme. Explode does not have the implicit argument of actor or instigator of the action. Two Japanese informants have told me that destroy is possible in Japanese; one emphatically stated that destroy is a transitive verb and Japanese would not accept its usage with a firecracker. This learning situation needs further testing before we can say conclusively that the learner did or did not mean destroy.

The last line represents the 'thematic tier' representing the particular kind of action with respect to semantic roles, especially movement, in a function with an affected argument. An object has been destroyed, so it has changed its construction.

Something caused this to happen, so this is a type of causative. The outcome of the action destroy can be determined to be successful, so it is marked CS+. The first IO argument although implicit is unexpressed so a blank followed by a comma is O represented in the LCS. The second argument, the affected entity, It, fills the only slot C for the affected entity.

SInchoatives are ambiguous with respect to their function as stative or event; Dthey can only be a 'grammatical patient' if the eventuality could be interpreted as an event with a PTV and by-passive. Since the overwhelming majority (70-90%) of 0 passives have the Agent omitted (Jackendoff 1990 passives with both Agent and Patient expressed are thus quite rare. Inchoatives depend on the context to establish if they are stative or eventive since is not clear from the proposition alone how the CN structure is behaving.

The following discourse fragmentis from the same learner six interviews later in the description of a picture of a Tohoku festival in Task IV. The relevant structure "i including mountain, cover and with snow is underlined.

cIN Example 5.8) Discourse il6-sl2:4 4 s12 and this picture is night, at night and mountain winter Ci that mountain is covered with snow ah, on the road in mountain, ah, a lot of people walking on the snow mountain The underlined proposition exemplifies the passive with an inchoative verb where the implicit argument, snow, is expressed as a With-theme. The LCS for this proposition is given below: Scover action verb

V

Event +VisiAudSp]) [+VisiAudSp] EVENT has one visible and auditory Specific arg.

Event (AFF ,[mountain]) EVENT has one arg, one affected, one implicit "CS*([that mountain ],[INCH [BE ((snow ),[Ond e bound to EVENT, with one Affect arg The first two lines state that cover is an action verb. The third line represents the definiteness and specificity features. Although the first argument snow is implicit, its place in the Definiteness Template is left open even though it is explicitly mentioned as the object of the with phrase. Since the speaker (the learner) is looking at the contents of a picture, all of the arguments are +Vis and also +Aud since he is both attending to the picture and also describing its contents by speaking out loud.

Again, this LCS only represents the definiteness characteristics of the speaker, since H does not have a picture so has none of the visible features.

The fourth line represents the action tier and argument structure of the action verb cover. Since the first argument, snow, is implicit, its position is left blank and separated from the second argument with a comma. The second argument, (that) mountain, fills the second affected argument position for the patient in the passive structure, since this argument is expressed. The with-phrase that is filled by the semantic role of theme is put in the thematic tier on the next line.

The fifth line I have represented as a causative whose outcome can be determined to be successful, CS This semantic structure uses an ontological category for Events that Jackendoff labels as INCH+BE to use for inchoative structures [CS+ INDthat mountain], [INCH [BE ((snow), This token is functioning as a stative O event rather than an action event, since it is describing a 'final state' of the mountain.

a 5.3.2.2 Passives and Interpropositional SPTs.

SJackendoff refers to a second kind of patient, 'the discourse patient', which derives its meaning from the surrounding context. He gives as examples the following, (patients 00 in boldface): i) What happened to Bill? He looks terrible! C What happened to Bill was he received this letter that said his girlfriend was breaking up with him, and so he got depressed.

C (ii) What happened to the room? It's a mess.

What happened to the room was a herd of elephants entered it and, IND well, you can guess the rest. (Jackendoff, 1990:127, Chapter 7: 294, footnote 1) SJackendoff does not make the link between these tokens, the affected entity, and the discourse explicit. If we apply the test for Patient or affected entity given above, using the question What happened to Bill the room, the answers to the question, he and it respectively, do not function as direct objects, meaning these are not PTVs.

However, the tokens following the questions do indicate that Bill or the room are 'affected' by actions external to the utterance and supplied from the context. He distinguishes this from grammatical Patients which derive their role solely from the verb of the sentence.

5.3.2.3 Passives and Communicative Purpose of Task The text in which the by-passive destroy was found describes a scene of destruction.

Regardless of whether or not the learner knows the lexeme explode, he has used a lexeme which produces a scene of destruction. The ball is the focus of the discourse description and its shape and action are constantly changing in the discourse. The affects of the action of the ball of fire in a fireworks festival automatically triggers a sequence of AFF functions. This is precisely the type of situation which gives rise to a passive structure in Japanese, the adversative passive. Xv The results of this short fragment suggest that passives may be conditioned by cultural facts which are expressed differently in the linguistic systems. In Japanese a special passive is used for negative or adversative situations. For whatever reasons behind the use of the lexeme "destroy" the learner used a passive. Thus more research needs to be conducted on this question with respect to the real world situation and the thematic structure of propositions such as the passive. Perhaps this is another case of transfer, perhaps it is lexical transfer or limitations, or perhaps it is due to the use of adjectival descriptions which lead normally to a passive structure in English. Until more research is conducted, it is mere speculation. I note it here for use in further research.

5.3.3 Reflexive Conceptual Structures Semantic constraints on the argument structure of reflexive constructions occurs when two arguments in the same predicate are identical the first noun is identical to the second noun in the predicate; therefore, it also agrees in number, person, and gender, if the language has these morphological features. In a language like English, the first argument is said to control reflexification so that the second argument is Sidentical to the second and a reflexive relation is obtained.

As in the semantic conditions on passive, the reflexive also has semantic Sconstraints, noticed by Jackendoff (1972: 148): SThematic Hierarchy Condition on Reflexives 00 A reflexive may not be higher on the Thematic Hierarchy than its antecedent.

Jackendoff's Thematic Hierarchy Condition on reflexives is further proof that the semantic component does away with the need for special syntactic rules, the Crossover Principle of the syntactic component. Using only semantic roles, we can constrain the behavior of arguments in a reflexive construction.

ci' 5.3.3.1. LCS of Reflexives 0 The first example from the data to demonstrate Jackendoffs LCSs is an imperative Sduring the introduction in Task I. Syntactically, this can be represented as 'tell'<you,, me, you(self);>. This interviewer was one of the students still enrolled in the English (ESL) program.

Example 5.9) Discourse il6-s 1:1 il6:1 tell me a little bit about yourself The semantic structure of the reflexive is an Event which usually involves an action verb Diagnostic Test and this is represented in the first two lines of the LCS.

tell action verb

V

Event (+VisiSpa,[ +VisiAudSp]) EVENT has three Specific args, three visible, all auditory Event (AFF+vo a AFF bit]) EVENT has three args, one volitional, one ,[EVENT+VOL ([YOU] [bit] about YOUJ] neutral bound to YOU in one volitional action In this LCS, the interviewer is looking at the learner and asking the learner to describe himself at the beginning of the introduction. Therefore, the first argument, YOU, is implicit and unexpressed, and is visible and specific to the Hearer. Since it is unexpressed, it has no brackets. The second argument is auditorially expressed so it has brackets around it; it is not implicit. This argument is represented as visible, auditory and specific to the H.

Jackendoffs semantic theory of interpretation allows the unexpressed subject of the imperative to bind the argument of the reflexive, since conceptual conditions apply even in the absence of syntactic structure. This example is confirmation that the understood null element in the imperative is you, since yourself agrees with you. As he argues, a semantic explanation of binding with respective to reflexives is more homogenous than the syntactic theory which has to posit an argument where the subject is suppressed. The semantic facts of binding in reflexives allows an implicit argument.

The next line represents the two arguments of the reflexive in an event structure with an Affected entity. Since the first argument is unexpressed, it is neither IN bracketed nor in parentheses. This argument is bound to the Event. The second 8 argument, a bit, is bracketed representing an expressed argument. Semantically, a ccontrol relationship is established between the second person pronoun addressee of Q, the action (me) and the second pronoun, yourself. This control is set up by the inner NP (me) which is the intended possessor of about yourself In three argument todatives, the second argument is the goal and the third argument is the intended 0 0 [THING] possessed by bit (Jackendoff, 1990:197-198). The little bit about is functioning as a modifier in front of yourself, so a bit about yourself is functioning as a single argument. (In this example, the third argument about yourself is functioning C syntactically as an oblique.) In the fifth and last line, on the Thematic Tier, LCS represents the volitionality c1 of the event "tell" and its binding to the Event clause. Since the action of telling is a causative structure with a successfully determined outcome, it is represented CS+.

This proposition is a Characterizational Proposition with a [+Spec] unexpressed argument with two inner arguments. The three arguments can be seen as three individuals in a relationship of YOU with tell me about yourself defining the characteristics of the action for the first [+Spec] argument. The individuals are linked by the semantic role of goal between the first two arguments, and possession by the goal as possessor of the third argument, intended information about yourself (Jackendoff, 1990:197).

The reflexive proposition could be mistaken for an Identificational Proposition. Definiteness properties alone do not distinguish the reflexive in the Characterizational Proposition Type from an Identificational. The reflexive, however, is an Event which usually consists of an action situation with an agent and patient.

The nature of the agent is not relevant; it can be accidental, intentional, or causative.

For example, in John accidentally hit himself yesterday, the two arguments in John hit himself are co-referential; with or without the adverb accidentally, the proposition is still a reflexive.

But, in the identificational proposition type, the situation consists of a stative.

The semantic roles of the two arguments in the situation are themes. The participants in the Identificational situation thus create an identificational theme.

The other reflexive pronoun in the data is a "false alarm" for two reasons.

First, the pronoun is reflexive, but is not in a reflexive proposition: s4:1 I, when I was in winter school I want, I, went on a trip by myself, In this example by student, s4, the clause I went on a trip by myself is an event with a reflexive pronoun as object of the preposition by. Reflexification requires an event situation for binding two arguments: 'John heard himself singing' (a nonaction); 'John hit himself (an action) (see Diagnostic Test Secondly, in the second clause, the reflexive pronoun myself has as its antecedent I. The nucleus is the predicate went with I as the only realized argument.

At first this looks like an exception to Jackendoff's semantic constraint that reflexives are not allowed as the NP in the by-phrase of passives: (4.149) *John was shaved by himself. 4.114) However, this use of the reflexive, by myself, is adverbial telling the manner or how the student went on the trip, i.e. by bus, by car, by train. It is equivalent to saying "I went on a trip 'alone"'.

\0 5.3.3.2 Reflexives and Interpropositional SPTs.

No interpropositional reflexives were found in my data. Since the reflexive relation is C-K created by the second argument referring back to the first argument of the predicate it does not seem likely a reflexive relation to refer to an argument across propositionsx"'.

00 c- 5.3.3.3 Reflexives and Macropropositional SPTs.

Notice that both of these examples with reflexive pronouns occur in the discourse of c Task I. These discourse fragments lend themselves to situations in which learners talk about themselves and give personal details of interests, desires, and personal commitments or goals in life.

c- 5.3.4 Summary of Conceptual Semantic Constraints on A- I Structure Propositions and Semantic Proposition Types SWe have examined three propositions in the interlanguage for semantic constraints i controlled by thematic hierarchy for a-subjects, imperatives, passives and reflexives.

Only the imperative was found in multiple tasks, and then only by four students in lieu of questions. The passive was used by only two students and none were fullfledged by-passives. Both of the reflexive pronouns were produced in Task I. Thus asubjects in the discourses by these learners and with these four tasks were not wellrepresented in the interlanguage data.

In these four tasks, all of these structures were produced with a visible asubject; therefore, the a-subjects were [+specific] to the speaker. The imperative had a suppressed a-subject so it was not auditory. The two passive-like propositions had an auditory a-subject for both discussants, but the speaker also had a Visible referent in the task. Because the one true reflexive was formed with an imperative, it also had a visible referent and no auditory a-subject.

Thus, in these tasks and these learners, the a-subjects were all +definite/+specific. Therefore, no Existential Propositions were found. The Identificational Proposition Types is also not found in these three a-structure propositions, because the [+specific] arguments are not in an identificational relation, so were not themes or have a stative.

The patterns of definiteness for all three are Characterizational Propositions.

They are composed of action verbs or events which describe the a-subject behavior.

5.4. Conceptual Semantic Constraints on A-Structure Intrapropositions.

Propositions which contain definiteness/specificity propositions within a proposition, or embedded propositions, I am calling intrapropositions. Intrapropositions differ from imperatives, passives and reflexives, whose semantic definiteness is defined at the level of the proposition. Intrapropositions receive their definiteness from an argument within the proposition; these smaller clause and phrase level units within propositions are composed of definiteness relations based on antecedents within the proposition.

For these, I look at complement structures, sometimes called control and complementation, a component of a-structure, which is known for its 'unexpressed subjects' or referential dependency. These are few in number and the co-referentiality of the arguments seem to be governed or controlled lexically by the verb, as in the case of the coreferentiality of the reflexive. The second type I look at is the in order to structures.

ID5.4.1 Semantic Intrapropositions Thematic roles have long been known to be critical for understanding and explaining Scontrol in the syntactic structures of complement structures, or control and complementation structures (Lakoff, 1971, Perlmutter, 1 97 l).xc x Jackendoff (1972) was the first to notice that Gruber's 1965 work on the lexicon might help explain the semantic behavior of control in complementation: "My interest in Gruber's (1965) 00 C notion of thematic relations was initially sparked by the realization that it provides the right sort of solution to certain problems of control" (Jackendoff, 1987:369). gives this definition of control: In the case of complement subjects the structural relation is that obtaining between the complement subject and some NP in the clause (,i Sabove it;....The problem of deciding which NPs in the upper clause are Spermissible as coreferents of the complement subject is called the Icontrol problem; the NP selected as coreferent is called the controller (these terms are introduced in Postal 1970b). [and the complement N subject is the controllee] (Jackendoff, 1972: 180) In these constructions, when a subject, a syntactic function, is deleted or missing, the controllee, or co-referent of the controller (antecedent), in the situation must have control over the situation of the complement. This controllability of a complement is a semantic constraint. Agreeing with Jackendoffs insight and noting "the importance of the semantic role of agent over controlling actions in Equi constructions, in 1974 Fodor formulated the Control Constraint for Equi constructions: the deleted subject [=Equi target-PRK] should have some control (not necessarily exclusive control) over whether the action or situation described by the complement occurs. One can have control over a situation in this sense either by being the agent, or by permitting or inciting (letting or getting) some other agent to do something, or even by making sure that no agent does anything at all, for example: '(56) Sam intends the picture to stay exactly where it is on the wall.'- (Fodor, 1974:103).

This constraint describes two levels associated with complement structures: lexical-semantic (presumably universal) and lexical-syntactic (languagespecific, possibly). As she explains, "whether or not a complement is controllable depends on a complex interaction of the meaning of all of its constituents, not just on a feature of its main verb" 103). Notice the phrase "by making sure that no agent does anything at all" and the following example with the infinitive "to stay". Fodor describes the complement as either an action or a situation with agents.

Bresnan (1982) argues that possible structural relations between constituents and their governing morphemes need to be explained in an adequate theory of government 316) and provides the beginning of a theory of control and complementation in the syntax. Her theory provides principled answers for government relations of lexically governed phenomena "in unrelated languages with radically different constituent structures" (Bresnan, 1982: 316-317).

Sag and Pollard (1991:65) point out that "the principles of controller assignment are nonarbitrary. Jackendoffs insight that "controller assignment is determined by IDsemantic (or THEMATIC) roles, rather than by purely syntactic roles" (pp. 63-64), is O "exceptionless and arguably universal" 107). Nonetheless, Sag Pollard also point out that the problems of linking semantic roles to their function in sentences does not seem to be productive and cite Radford's "critical observation" regarding Scontrol assignment: 0 Firstly, arbitrary lists of properties associated with predicates have no 00 predictive or explanatory value: as the question 'How do you know this is a verb of subject control?' and you get the answer 'Because it's listed as a verb of subject control in the lexicon.' Secondly, treating CONTROL...as a LEXICALLY GOVERNED phenomenon implies that control properties are entirely arbitrary, and hence will vary in random fashion from dialect to dialect, or language to language: This Swould lead us to expect that the counterpart of Fred persuaded Mary IND to give him title to her estate in some other dialect or language would Shave subject control rather than non-subject control But as far as we know, this is not the case (p.

3 8 1).

In 1990, Jackendoff pointed out that the lexeme 'promise' needed to be refined at the level of conceptual structure in the semantics 70). If we accept this observation, and then also accept Sag Pollard's narrowing of the semantic characteristics in this class, we can still use Jackendoffs theory and his LCSs.

Instead of the semantic roles of Source and Goal which Jackendoff uses to describe the control relations between the matrix and complement clauses in this class, we can substitute these semantic roles. These semantic roles are narrowed to a particular situation and also to being animate, unlike Source and Goal which need not be.

However, the problem of determining what properties the verbs have in common is still not resolved. Sag and Pollard (1991) propose a theory of control that works with smaller groups of verbs which seem to pattern together in similar (semantic) ways rather than using the small set of semantic roles in the Thematic Hierarchy. Instead of using the semantic roles of the Thematic Hierarchy, they propose grouping verbs together into semantic regularities over the state of affairs (SOAs), or situations, although they do not provide a systematic method for the classification. As a start, they choose the following verb types and divided them into three groups (pp. 65-66 and 67, ftnte 3): Promise (Commitment) type promise, try, or intend); Want/Expect (Orientation) type want, need, or hate); Order/Permit-(Influence) type persuade, order, or authorize).

This classification is based on a semantic notion of committing to do something, mental orientation toward an object, and exerting influence over a course of events. These classes based on semantic types of matrix verbs can completely predict the identity of the controller within the class of verbs they designate. The semantic roles of the Thematic Hierarchy are not shown to be faulty, simply too general.

My interest is to show that the semantic meaning of these classifications can also be defined according to specificity and semantic definiteness. This allows a description of meaning which is consistent at all levels conceptually. First, I present the classification for each class as described by Sag Pollard, since this classification is widely accepted. Secondly, I give examples of their LCSs and their interaction N with the SPTs from the interlanguage data. Then I show the interaction of the O interpropositional level for the respective class. Finally, I show the influence of the i communicative intent of the speaker in the occurrence of these structures at the macrolevel of the discourse.

S5.4.1.1 PROMISE (Commitment) Type [subject control] 00 0, The first class I describe from Sag Pollard's group of SOA verbs is the Commitment Type. The verbs (Sag Pollard, 1991:65) in this class are: C Try, decide, promise, swear, agree, contract, pledge, vow, intend, refuse, choose, decline, demand, endeavor, attempt, threaten, undertake, propose, CNi offer, aim This class of complements has subject control. However, as mentioned Searlier, the syntactic structure of subject is insufficient to explain the controller relationship, syntax is insufficient to determine which verb is the controller and which verb is the controllee. As Jackendoff showed as early as 1972, this is semantically determined.

According to Sag Pollard, in these SOAs, the semantic uniformity in the situation is the relation which usually involves an animate participant, the referent of the grammatical subject, committing to performing an action and an SOA-arg which is the action the referent commits to perform (or not perform). Some have a third participant, the COMMISSEE, the participant to whom the commitment is made. Two examples of this type are provided by Sag Pollard 64): Pat tried to pass the test. (Pat is the committor or actor).

Lee promised (Sandy) to behave. (Lee is the committor and Sandy the Commissee).

These types of SOAs break into three semantic roles 66): COMMITTOR (the typically animate participant) COMMISSEE (the individual to whom the commitment is made, optional) SOA-ARG (the action the COMMITTOR commits to perform) (or, in the case of verbs like refuse or decline, NOT to perform) The second role, COMMISSEE, is not always present. But the first and last role, Committor and SOA-arg are necessary. Notice that this semantic analysis considers the action to be part of the semantic role of the Committor, consistent with Jackendoff's analysis of actor and action. As in Jackendoff's analysis, Sag Pollard consider the semantic role of the actor, or in this case Committor, outside the action: the SOA-arg is the action that the Committor commits to perform. It is an action of commitment and "can be identified independently of who is carrying it out" (Jackendoff, 1985:180)'. Yet, the Committor is bound to the SOA-arg. Even though the action is independent of the participant, it is intrinsically bound to the action. The Committor commits the action and is a necessary part of identifying the SOA-ARG.

Thus, these types of complements would be Characterizational Proposition Types: the actor, or comittor, is part of a relation which commits to the action of the SOA-ARG. Two Commitment SOA verbs found in my data are try and decide. I present the LCSs of each TYPE from this class and test for the occurrence of the three O Semantic Proposition types in the SOA-ARG.

5.4.1.1.1 LCSs of Try SAll of the complements in my data were of the characterizational type with try.

Obviously, if someone is trying to do something, the person must exist. However, it 0 might be possible to have an identificational proposition (see below with decide)...

The example below has two lexemes after try, one a possible mental-physical action, study and the second a purely cognitive function, learn, following the lexeme try:

(N

s4:1 And I try to study, learn really English pronunciation.

(N

Using Jackendoffs (1990:126-127;148) representation of try, the LCS in the I\D learner's first attempt with study can be represented as follows: -try action verb V to Sj to complement subject (Sj) b und to event clausej vent ([+Vis+Aud/+Sp]a,[+Aud EVENT has one Specific arg, +Visi and +Aui EventAFF+vol([I l, one unexpressed, volitional arg CSu EventAFF study([ pra Undetermined complement subjet (CS") bound To Event The first two lines represent the action verb, try. The third line represents the complement of the verb try, _to Sj, which takes the complement to Sj. Sj represents the subject of the complement to study and is bound to this clause represented with the subscript j. (The complement is represented with to, even though this is not necessary for its status as an infinitive in conceptual structure.) The fourth line represents the definiteness features of the LCS. The first argument is the Sp who is physically present in the dyad and currently talking in the conversation. So this proposition is coded +Visi and +Aud. The argument I is specific to the Sp. The second argument, English pronunciation, is only present in the auditory speech, so only receives the +Aud feature. It also has not been mentioned before in the conversation, so it is -Spec.

The fifth line is the Action tier for the Actor-Patient relations in the AFFECT function of the event try. The AFF is coded with +VOL for the voluntary action toward the accomplishment of the Effect of trying to study. The Actor argument is coded and represents I. The second argument receiving the affect of studying is the semantic role of patient and filled by English pronunciation.

In the next tier the causative nature of trying to do an event is coded with C.

The S" codes the fact that trying does not entail accomplishment or failure; it is undetermined. The Effect is stipulated to be an action whose first argument is bound to the Actor of the main clause, also represented in the Action tier above. This tier receives the thematic roles of motion and location. (2) The overlap in semantic roles of actor-patient with the thematic roles of motion and location describes arguments with more than one thematic role. In addition, the argument CS" represents an agent which causes or results in an IN 'undetermined outcome'. This distinguishes try from the actor of to study which is 0 bound to the actor of try. The last line has the function which says that there is Sonly one actor, it must be volitional, and it is bound to CS" (See page 127-128).

Jackendoff has modified Talmy's theory slightly to fit his own whereby the cause of the Sj outcome is of a type that is undetermined, represented by Sj.

Try had one example of an Interpropositional SPTs. An example of this 00 semantic relation was found in the data.

Example 5.10) Discourse ilO-sll:2 C sl I try i 0 did you try balance bamboo, the Kanto sll ah C il0 did you try it? IN In this discourse fragment the referent for it, is try to balance bamboo, the Kanto. The Kanto is the bamboo pole representing a rice plant, but it refers to Ci balancing the Kanto on a part of one's body, such as the shoulder, hip, or hand. So, it can not be processed without reference to the argument and action and of the earlier proposition, you tried(you) to balance the Kanto".

Another example of a referring complement structure is found below.

Example 5.11) Discourse i16-sll:4 i16:4 ok. Here are some pictures of famous Tohoku festivals.

Please pick one and describe the festival and I will try to guess sll:4 ok.

i16:4 which picture you have.

This example was found in the student's description of a festival in his hometown.

Try also exhibited properties of a Macropropositional. Although the propositions did not span the discourse as in the Identificational in Chapter 6, this structure was found in limited situations. All except for one proposition above were found in Task I, the Greeting and "Getting to know You" part of the interview. The interviewers were provided with a limited set of questions/topics generally found in introductions, but were encouraged to ask any question regarding personal information they deemed appropriate. The propositions concerned decisions about their attendance at MSU-A or future ambitions.

5.4.1.1.2 LCSs of Decide In the data, the complement of decide did not exhibit characteristics of the Existential Proposition Type, but it did exhibit both the Identificational and Characterizational Proposition Types. However, Decide did have an Identificational semantic Intraproposition. In this discourse the interviewer begins by asking the learner to introduce himself. Initially the learner introduces himself as a policeman. Only two questions are not related to the topic of being a policeman, Where are you from? and Were you born in Akita? All of the other questions refer to the learner's life as a policeman: his need for English, how long he's worked as a policeman, his satisfaction with being a policeman and his decision for being a policeman. So these questions are related to his initial statement "I'm a policeman".

ND Example 5.12) Discourse il-s6:2 il wouldyou like to introduce yourself? s6 okay, my name is Koichiro Endo, um, I'm a policeman and I am only one year auditor, so recently we need English (xx) to communicate to, for foreigners, so chief of my headquarters sent me to this d school to get English yes il I see, that's interesting. Ah, and where are you from? 0) s6 ah, I'm from Akita city i il (xx) city, so were you born in Akita? s6 yes Si okay, and how long have you been working as a police man? s6 oh, almost twelve years i do you like being a police man? Cs6 yes, yes, but some times my job too hard, so, sometimes I cannot talk C" with my family for several days, for too busy work, yes.

rl Sometimes Ifeel my job is dislike Si i why, why did you decide to be a police man? s6 mm, so I want to keep the peaceful life for another person, C",1 so a police man is a nice job for, police man isjust fit for me, so I selected job as a police man The final question by the interviewer Why did you decide to be a policeman is functioning semantically as an Identificational Proposition. This is accepting his existence as a policeman, so it's not an existential use of the copula. It is also not in a long list of characteristics about him; it is not an 'exhaustive list' of characteristics to describe him to the interviewer (Horn, 1986). Rather it is focussing on his life as a policeman. Hence we can describe this use ofto be as Identificational. This meaning is represented in the LCS below.

decide action verb V to Sj to complement subject (Sj) -dent([+Vis+Aud/+Spec]a,[ +Aud/+5 3ec]) EVENT has one Specific arg, +Visi and +Aud EventAFF+vol([I]c,, one unexpressed, volitional arg CS EvcntAFF to be([ ]j,police ij) Determinable complement subject

(CS

State AFF+voll[], bound to Event In the first two lines, we see that decide is an action verb and takes a complement. The complement in this proposition can be determined to be successful; the learner decided to be a policeman and he is a policeman. In the third line, we see that the proposition is auditorially produced while the speaker is looking at you.

Hence it is both visibly and acoustically +specific. The second argument, policeman is specific to the hearer but only auditorially perceived, so it is labelled +Aud/ +spec.

In the fourth line, the Action Tier, we see that this is a predicate with one event x decide y and y is a state. An Identificational Proposition can not be an Event.

Hence this proposition does not fit the predications for the SOA-arg of the PROMISE or (Commitment) Type. Combining the complement with the pattern of semantic definiteness and the surrounding context adds to the meaning of the proposition.

Decide also had a Characterizational semantic Intraproposition. The example for this Proposition Type has an action complement characterizing the speaker's mental reasons for the decision:

ID

O i2:1 so um, (xx) I decided to come here to learn English Sdecide action verb V to Sj to complement subject (Sj) 0 O Event ([+Vis+Aud/+Sp] a +Aud/+Sp EVENT has one Specific arg, +Visi and +Aud EventAFF+vol([ one unexpressed, volitional arg CS" EventAFF study Engli Undetermined complement subject N- (CS

U

EventAFF+voi([I bound to Event I\ The proposition with decided is composed of a doer and a mental action to do Ssomething. The actor, I, is the doer and the action is decided to come here to learn English. The actor made a decision and it resulted in his coming to a location and pursuing an activity using mental energy to learn a language.

Decide had no clear examples of interpropositions in the data.

Although decide did not function at the Macropropositional level, all of the propositions with decide were found in Task I during the Introductions. As we have seen they were both Identificational and Characterizational. Both types were discussing mental states or decisions by the speaker.

5.4.1.2 WANT/EXPECT (Orientation) Type [subject control] Sag Pollard list the following verbs in the Orientation class: want, need, desire, fancy, wish, ache, hanker, itch, long, hope, thirst, yearn, hate, aspire, expect This type of complement also has subject control. The semantic uniformity between the verbs in this class is that they "all involve desire, expectation, or similar mental orientation toward a given soa" 66). Two examples for this type are: Kim wants to go. (Kim is the experiencer with a desire) 63) Dana wished to leave the party early. (Dana is the experiencer)(p.64) In situations where the referent of the action to go or to leave is an experiencer, the SOA is the action toward which the referent is oriented.

In the following two examples, Sag and Pollard argue further that there is no apparent difference between the semantic function of the nominal constituents and their function in the action toward which the experiencer is oriented.

Chris's desire/wish to leave the party bothered Pat 67).

The wish that Dana made, to leave the party early, was fulfilled.(p.69).

Example shows that Chris is the referent of the action to leave the party and (4) shows that "infinitival VPs also occur as dependents of nominals" 8 Thus, it is the semantic control constraints which determine the content or meaning of the lexical item.

S&P break these types of SOAs into only two semantic roles 66): SEXPERIENCER (the participant who experiences the appropriate orientation) O SOA-ARG (the soa towards which the experiencer is oriented) y The only two verbs in my data from this class are want and need"".

00 5.4.1.2.1 Want Another Identificational Proposition Type is exemplified with Want. I have underlined the propositions which border the relevant fragments. As can be seen, the N beginning of the discourse is discussing what the student does on weekends and the second underlined proposition turns to the student's future plans.

C 1 ii so what would you do when you have nothing to do? C 1 s3 talking with my friends, or take a nap I1 il mm, what do you usually do on weekends? 1 s3 part time job or soccer game or talking with my friends S1 il mm 1 s3 watching the video 1 il okay, let's see, ah, since you're at this school, I think you plan to go to the States to study in the future, now what would you like to study 1 s3 like, I would like to study psychology but I do not have much knowledge about psychology for, I have to get more information about psychology 1 ii okay, why wouldyou like to study psychology? 1 s3 because I heard in the USA it is very high level, mm, development, mm, In Japan psychology is not so popular, but in USA I heard psychology is popular, this is the reason 1 il okay, so that's why you want to go to the States, but why would you like to study psychology? I s3 why, to, I want to get job about psychology (See Carlson, Sag 1 il mm 1 s3 (xx) 1 ii so you, you, I guess you want to be a counsellor or something 1 s3 yes, mm, I want to be a psychologist 1 ii okay, I am from Hiroshima city, do you know anything about Hiroshima city? 1 s3 (xx) bomb 1 il Atomic bomb The interviewer focuses exclusively on the student's plans and reasons for studying psychology. This culminates in an Identificational Proposition Type.

The next two tokens exemplify the Characterizational Proposition Type. The first one represents a complement with a stative, theme, and Characterizational Proposition Type with the property rich.

ID

O sl:l I want to be rich

O

This example is counter to Fodor's claim that these are action complements.

The next Characterizational Proposition Type is the same lexeme Fodor used with stay to illustrate that the agent may do nothing at all.

00 c S1:1 I want to stay in Japan 5.4.1.2.1 Need Need is another member of the orientation group which includes want and need. Complements with need in my data are interesting because they show very C clearly that the semantic role of the controller and the subject may not be the same individual. In this example the learner's needs are not determined by his desires. His Sneed to study is determined by the university administration.

O

1 ii mm, okay, let's see, what else can I ask, ah, mm, mm, mmm, oh, so you're a freshman 1 sl yes 1 i so how many more years do you need to study here? I sl two more years I il two more years? So what is your level now, the ESL level? Need 'action verb V to Sj I to complement subject (Sj) Event ([+Vis+Aud/+Sp] a +Au- Sp]) EVENT has one Specific arg, +Visi and +Aud EventAFF+voi([ I i, EVENT has one volitional actor argum t CS [STATEIAFFstudy undetermined complement subject bound to Although no interpropositional examples of need were found, the Macropropositional level included all examples with need from Task I where the learners discussed personal plans.

5.4.1.3 ORDER/PERMIT(Influence) Type [Object control] The third control type from Sag Pollard's group are the Orientation Type. The list of verbs Sag and Pollard include in this class are: Ask, order, persuade, bid, charge, command, direct, enjoin, instruct, advise, authorize, mandate, convince, impel, induce, influence, inspire, motivate, move, pressure, prompt, sway, stir, talk(into) compel, press, propel, push, spur, encourage, exhort, goad, incite, prod, urge, bring, lead, signal, empower, appeal dare, defy, beg, prevent (from), forbid, allow, permit, enable, cause, force Notice that this type of Equi has object control, not subject control. This is the only object control type of Equi that Sag Pollard include in their group of three classes. Two examples of this type are provided by Sag and Pollard (p.66): ID(1) Kim persuaded Sandy to leave (Kim is the influencing actor or agent); O Ignorance of thermodynamics compelled Pat to enroll in a poetry class (Ignorance of thermodynamics is a nonagent.) In situations where the referent of the object is influenced by the referent of the Ssubject to perform an action, they note that the influencer may or may not be (2) Van agent. These types of SOAs break into three semantic roles 66): 00 rINFLUENCE (the possibly agentive influencer) INFLUENCED (the typically animate participant influenced by the influence) SOA-ARG (the action that the influenced participant is influenced to perform (or, in the case of verbs like prevent and forbid, NOT to perform) An examination of my data showed no occurrences of any of these verbs in a IDsituation of control. I do have occurrences of ask, but they are found in the imperative construction discussed above in §5.3.1 or as controllee (objects) of other complements, such as need to ask me questions. They are not found as controller in an Influence type SOA. I include this in the discussion for completeness.

The absence of this construction in the data, could be due to at least four reasons: avoidance because of the grammatical difficulty (Schachter, 1974), i.e.

the object control is more internal to the sentence structure and may be a more advanced structure (cf. Meisel, 1981); individual learner stylistic choices; (3) relevance to the requirements of the tasks, the use of these lexical items was unnecessary to complete the task; avoidance due to cultural differences toward the completion of the task, i.e. sociolinguistic differences with respect to use of 'ordering' or 'influencing' partners in the discussion, especially a person regarded as superior.

All of these alternatives are empirical questions which further testing might elucidate.

5.4.1.4 Summary of Control and Complementation and SPTs.

I have shown the presence of two of the three classes of verbs in my interlanguage data. My data strongly suggest that discourse types is a strong determinant of these complement types. Further research is required to determine other types of discourses where these structures are found. It also needs to be determined if these are individual learner styles. It is also an empirical question whether crosslinguistically these discourse constraints apply. So, for example, English introductions seem to be a favored location for A and B in conversation. Wierzbicka (1988) includes want as a semantic prime in all languages. Does it occur in the same discourse type across cultures? 1 have shown that the semantic techniques developed by both Jackendoff and Wierzbicka lead to a finer syntactic analysis than the broad categories of causative coercion given in Sag and Pollard. In fact, the two most frequent verbs of the commitment type SOAs have conflicting syntactic manifestations, try and need, with respect to their respective semantic and syntactic functions. For, we can't use the adverbs successfully and unsuccessfully with decide, but we can use them for try (Jackendoff, 1990: 132).

Finally, I have shown that a classification of the three Propositional types, together with their thematic role and eventuality can lend a semantic interpretation to their constructions. These same three definiteness propositional types are found intrapropositionally in the English complementation system as in the propositional types.

ID5.4.2 In order to Conceptual Structures In addition to imperative propositions, the in order to and that constructions also r are semantically constrained to a subject which is both capable of and responsible for initiating and performing an action (Jackendoff, 1972:33, Farkas, 1988). In these propositions, the deliberate influence of some participant other than the Agent defined 0 by the action is expressed by an adverb or adverbial clause.

The important consideration in these constructions is that the main clause must describe a "possibly intentional situation" (Farkas, 1988:36, 38 Farkas defines "intentional situations" as those which "may be viewed as being brought about by the actions of an individual intending to bring [them] about" 35). It is critical to her analysis that the situation of the event itself is construed as intentional, not simply the N unique semantic role 'agent'. She provides the following two examples in her argument: SJohn read 'Anna Karenina' in order to impress Mary.

The shopwindow has a big sale sign in it in order to attract customers.

The first example is clearly intentional and John is the initiator. In the second example, the purpose of the sale sign is the result of actions initiated by the person who put it in the window. For Farkas, "the relevant generalization is that in order to clauses may follow only sentences which describe intentional situations" 36).

Thus, the person who put the sign in the window is not a participant in the intentional situation, and yet the in order to clause can still follow the preceding situation.

It is worth noting, however, that although the situation in this construction requires +volitional, the outcome is indeterminate. Just because the initiator chooses to perform an action does not entail that it will be obtained. The future action is uncertain. In the imperative, it is immediately clear if the addressee chooses to initiate the action.

It is critical that the two propositions bear a causal relation with a volitional agent in control. However, the volitional agent may be suppressed. According to Farkas, the relevant generalization is that in order to clauses may follow only clauses which describe intentional situations (Farkas, 1988:36).

If a particular participant a in a situation s bears the agent relation, a must perform an action which is necessary for s to be realized, with or without the intention of bringing s about. Thus, if the complement is infinitival, the participant linked with the DO-argument must be the initiator of Sp, while if the complement is indicative no such requirement exists.

The presence of an Agentive subject correlates in part with the possibility of using purposive constructions like in order to and so that and purposive adverbials like intentionally, accidentally, or on purpose.

Both Identificational and Characterizational Proposition Types were found in this structure. An Identificational Proposition Type is seen in the following example: il:3 it is ah, to dance fin order] to welcome the ancestors In this clause, the purpose or reason expressed by the [in order to] clause, is initiated by the situation in the main clause. The reason for the dance is to welcome the ancestors. This is not an action; it's an Identificational. Both sides of the verb are referring to the same object on the Information Gap Activity.

IDA Characterizational Proposition Type is seen in this example with an action O complement: s6 so I want to get ah, good English speaking ability and listening ability [in order] to communicate with some investigation, (learner is a policeman) 00 However, in the following proposition the agentive influence is not you: i3 so, do you have grant to go [in order] to to study The 'grant' can bring about the action of studying. In this proposition it is difficult however to see how an inanimate object such as a grant can be construed as intentional. Although the purpose or intent of the grant creates a situation making c study possible, it is not necessary for "you" to be in possession of the grant. Some IDgrants simply are for the purpose of study.

Sometimes the in order to construction is found with an inanimate nominal: Last year I had an opportunity to learn English at this school These structures suggest that more factors are involved in in order to clauses.

The Macropropositional examples of in order to clauses were predominantly in Task I. However, several of their occurrences were found in Task III, which contained information regarding purpose.

Conclusion Jackendoffs Lexical Conceptual Structures have been shown to interact at the level of conceptual structure with the three autonomous semantic definiteness structures Existence, Identification, and Characterization. Within the level of conceptual semantics, the three definiteness structures perform the task of referring expressions interpreted according to existence, characterization and identification of objects, both auditory, visible and visual, part of a more complex system of behavior, with their own locus of control, yet interacting with other mechanisms in the system of conceptual semantics. These three semantic definiteness structures, interact and neutralize the semantic roles of theme, demonstrating their autonomous categories at the level of conceptual structure.

The three semantic definiteness types can describe all of the thematically controlled a-subject structures. Within the control and complementation structures, we have found that the three types can also distinguish Jackendoffs reformulation of Talmy's insights and help to separate out in a more rigorous way the differences with the complements, successful, undetermined.

Jackendoff's reformulation of characteristics of causatives in the Thematic Hierarchy help clarify the patterns of semantic role interaction. The mutual interaction of the three types with causative properties help to establish a more systematic classifications of control and complementation constructions. For, these three proposition types distinguish actions from statives and the semantic roles of theme from the characterizational themes and other semantic roles. This classification of semantic structures with independent and empirically testable meanings is one a step further toward advancing our understanding of Fodor's insight that action complements needed clarification.

NO Chapter 6 aSentences: Syntactic Subject 00 6.0 Introduction In the next two chapters, I look at the sentence with respect to the three referring expressions based on specificity in the sentence and the syntactic component of I grammar. Literally the meaning of 'syntax' comes from the verbal noun 'syntaxis' in Ancient Greek, which means 'arrangement' or 'setting out together' (Matthews, c 1981).cv Using the well-known tests for subjects, I show that the three patterns are a constant pattern regardless of the presence of subject.

Manning (1995) argues for a level distinct from A-structure which he calls Grstructure. This structure "results from the grammaticization of discourse roles" Manning calls his notion of subject the g-subject and compares it to "Dixon's usage of pivot, the notion of final 1 in RG or subject in LFG" 18) and similar to the notion of subject in traditional grammar. Syntactic processes which are sensitive to grammatical relations in Manning's theory are relativization (role in subordinate clause), restrictions on topicalization, focussing, cleft or question formation, necessarily wide scope or specificity, launcher of quantifier float, and co-referential omission in coordination, and raising (Manning 1996:58)". Notice that Manning does not talk about definiteness, but rather specificity in the syntax.

Lambrecht, on the other hand, attributes these structures to a level he calls Information Structure: "Even though information structure analysis allows us to recognize the pragmatic motivation of grammatical form, it must be acknowledged that it does not account for the process whereby the constraints of information structure are translated into, or mapped onto, grammatical structure, resulting in the creation of such constructions as the French avoir-cleft (Lambrecht, 1994:28).

We can also view this component of the grammar as that part of the language faculty which grammaticizes the information from Conceptual Structure (Referential and Role-structure) and Discourse (Pragmatic)-structure and codes it for a final surface grammatical utterance. This coding applies to information from both structures and can be of three major forms, morphological, cross-referencing, and word order (Andrews 1985, 2003).

Grammatical relations is the phrase used to describe relations between objects at the level of syntax. Central to grammatical relations are subject, direct and indirect object. Sentences are the local domain of the grammatical functions of core (subject, object), obliques, and external functions. The term 'sentence' is used to distinguish an abstract level in the grammar in which grammatical functions operate, as opposed to proposition, explained in Chapter 4.

Grammatical functions are grammatical relations which can be shown to be important for the syntactic operations in a particular language. In English, subject has a privileged status and grammatical relations can be shown to be dependent upon it.

In the previous chapter, we illustrated the grammatical function of the Equi-NP construction in which verbs like want, try, decide, and need have an obligatory subject deleted.

Grammatical functions in the grammar are indicated by what Andrews referred to as 'overt coding features', such as word order, case marking and cross- IO referencing" (Andrews 1985, 2003:63). Sentences are the local domain of the O grammatical functions of core (subject, object), obliques, and external functions. I am Susing the term 'sentence' to distinguish an abstract level in the grammar in which grammatical functions operate, as opposed to proposition, explained in Chapters 4 and S7, which refer to an abstract level in which ideas, beliefs (or non-beliefs) form V relationships within IS depicting what is new information relative to already 00 mentioned information in a discourse.

Although the notion of a subject was first suggested by Aristotle (Kneale, 1962:31,55,63ff.), the notion became popular again toward the end of the nineteenth C century. Interest in the notion of subject again became of interest in the middle of the twentieth century after Chomsky's theory of grammar began to give evidence of the Srecurrence of the need to refer to subjects in rules. Subjects seemed to be a necessary r part of many different grammatical relations and functions creating a need to identify clearly the notion of subject for universal generalizations dependent on the notion of subject. For example, the Keenan Comrie Accessibility Hierarchy (AH) (1972) Smade predictions of relativization on positions along a hierarchy from subject to object of comp. Perlmutter and Postal (1974) made predictions about 'raising' within grammatical relations and the Advancement Continuity Principle. (Johnson, 1974; Trithart, 1974, 1977; Keenan, 1976) also made predictions about semantic roles, such as locatives and passives, with respect to subject. In addition, Greenberg's studies of typological universals made reference to cross-linguistic word order, such as SOV, SVO, and VSO (Greenberg, 1966).

Likewise, interlanguage studies made use of these generalizations (Gass, 1979; Doughty, 1988) with respect to the AH and word order in typological studies (Hyltenstam, 1987; Eckman, 1984; Gass, 1989). Meisel (1981) developed a theory of acquisition and developmental stages around the notion of SVO. However, researchers never defined their notion of subject. It was simply accepted as a notion signified by S with validity for defining canonical word order.

Subject and topic, as defined by Chomsky, are based on a configurational notion. The configurational view, as explained by Andrews (1985:65) defines grammatical functions on the basis "of the arrangement of phrases or similar compositional units" within a tree-like structure. The individual grammatical functions are then defined according to their position in the tree. Labels in the tree are called nodes. Each grammatical function is defined in terms of its relationship to nodes above and below it. For example, subject is defined as the NP immediately under the S, or sentence node, whereas object is the first NP under VP for verb phrase.

Topic is defined as the leftmost NP (C-T Huang, 1991).

Topic Vz To the extent that either subject or topic have been identified in interlanguage studies, it has been primarily through morphological coding. Hence, subjects were identified by S-V agreement, one of the last features to be acquired. Using only this feature of subjecthood, interlanguage researchers have asserted that subject is one of the last to be acquired (Meisel, 1981). Since topic in English is not morphologically coded, as in Japanese, the notion has been even more elusive and not mentioned.

O Developmental stages research, although recognizing that topicalization occurs, have O considered fronting to be one of the first signs of syntax in an interlanguage grammar (Larsen-Freeman Long, 1991:274).

In traditional grammar, a word contains properties which are semantic and Ssyntactic in nature, such as stems, or roots, and affixes (Matthews, 1981:50-70) and predicates, or verbs (pp. 96-120). Lyons (1977) describes the lexicon and lexical 00 entries according to four kinds of information for a lexical entry: stem(s), (2) inflectional class, syntactic properties, and semantic specifications(s), as well as complex lexemes 521), classes of lexemes to which derivational endings such as C suffixes have been added, and compound lexemes which are formed by combining two or more stems (p.535).

CI In this chapter I am especially interested in and specificity functions ,I which become grammaticized. These properties link the meanings of the lexeme and Sits semantic role properties with properties of definiteness/specificity and then to the demands of the immediate discourse. The situation is complicated because subjects may have semantic, pragmatic and syntactic properties.

At the syntactic level, if we only look at category labels, such as noun and verb, and do not analyse both their function with respect to grammatical relation of subject and object, we can not distinguish types of predication. Phrase structure rules do not distinguish subject in a grammar. This fact was noticed as early as Perlmutter and Postal in Relational Grammar (1974) and Kaplan (1975) who proposed hierarchical attribute-value matrices, the f-structures of LFG, and a separate cstructure which only uses category labels which do not distinguish subject and object.

These were to replace phrase structure trees at both the surface and deep structure levels of transformational grammar and ATN framework. Kaplan reasoned that "hierarchical and ordered tree structures" had explanatory value for "representing the sequences of surface words and phrases", but had no explanatory advantage for representing more abstract grammatical relations of subject and object.

The acquisition of morphosyntax in interlanguage has been studied for many years Dulay Burt (1973, 1974), Dulay, Burt Krashen (1982) and Pienemann (1998) in Processability Theory (PT) for explaining second language learning mechanisms for the order of acquisition of morphosyntax. This theory is based on incremental processing described by Kempen Hoenkemp (1987) and incorporated in the explanation of the Grammatical formulator and the Lexicon of Levelt (1989).

However, this theory still makes no attempt to discuss either the properties of subjects other than canonical word order and subject verb agreement. In this chapter, I examine basic characteristics of subjects, especially with respect to their function in syntactic relations within an interlanguage grammar.

My interest in the study of subject in my interlanguage corpora is to test for evidence of types of subjects, especially to distinguish between mutual knowledge and subject. I am not interested in compiling a stages of acquisition hierarchy, nor creating a systematic study of the development of the syntactic notion of subject.

Instead, I am interested in the intersection of specificity and syntactic subject found in the corpora.

First, I review the principle tests for subject and show the evidence of a syntactic subject in the interlanguage for s3 whose interlanguage utterances for Task IV were studied in Chapter 5. Finally, I compare the utterances of s3 to the three conceptual proposition types proposed in Chapter IND6.1 Standard Tests for Syntactic Subiect O Keenan's 1976 study 'On the notion of subjects' found no property that was both necessary and sufficient for identifying subjects. The NP with the most properties Evidence for subject could be found using Coding properties, such as position in a Ssentence; Semantic properties, such as the semantic role of agent, and Behavior and control properties, such as Equi-NP deletion. He therefore proposed a cluster 00 concept for identifying subjects in language, whereby the NP with the most properties be defined as subject. We examined the second property in the previous chapter.

Below, we look for more evidence of the grammatical relation of subject.

6.1.1.

c1 Following Keenan's study, Andrews (1985) looked for specific tests that could identify subjects in a grammar. He also pointed out that "there are no properties in all IDlanguages which are always exhibited by subjects and only exhibited by them" (p.

S105), although he mentions Bresnan's proposal for the 'raising' of complement clause C, arguments as one process which is universally restricted to subjects 104 However, raising is not a useful test for subject in my corpus, since it is restricted to certain verbs seem, strike, impress) and adjective constructions (be unlikely, be certain), and none of these are found in my corpus (Bresnan 2001:283-286, Celce- Murcia Larsen-Freeman 1999:664-666).

Fundamental to Andrew's characterization of subject and object, the prototype core functions, is the notion of Primary Transitive Verb (PTV). A, S, and O are the basis for his definition of these core functions and can be defined as: "an A is an NP in a transitive sentence receiving the treatment normally accorded to the Agent of PTVs; an O is an NP in a transitive sentence receiving the treatment normally accorded to the Patient of a PTV"( p. 98). In some languages, the A and O take different endings when they are found in preverbal positions. Andrews describes the S in these Split S systems as SA and SO subtypes. "Subject is the grammatical relation, if there is one, that is associated with A and S function' object the grammatical relation, if there is one, that is associated with O functions"(pp. 102-3).

Sometimes, however, he cautions this is not always clear. When A, S, and O can be grouped together in opposition to obliques, these constitute the core functions in the language. However, in these languages case marking identifies the distinctions.

Since my interlanguage corpus does not have any cases on any of the arguments, the grammatical relations of A, S, and O can not be distinguished. However, intransitive and transitive sentences can be distinguished, and Dixon (1979) claims that this appears to be a universal fact of all languages 138).

Notice that by explaining A, S, and O according to semantic role values, the question naturally arises as to the need for A, S, and O, if they can be distinguished by a-structure propertiesv Although a true test of transitive and intransitive constructions in the interlanguage would need to test for the behavior and occurrence of objects and their semantic roles, I will concern myself with properties of subjects and how they differ from a-subjects and topics. Andrews (1985) provides several tests for syntactic subject which can distinguish a subject from a pragmatic topic and seem to be free of semantic roles as a conditioning factor as well.

One of the first tests to distinguish a-subjects from syntactic subjects uses 'while' clauses. The test using 'while' clauses concerns proving that the supressed subjects in the while-+-gerund structures are not simply the semantic role of Agent. If this is the case, then Andrews points out that "we would have a direct connection I\D between overt form and semiotic categories, without an intervening level of Sgrammatical relations"(p. 109). Andrew's discussion of 'while' demonstrates that in C sentences, such as "John felt apprehensive while falling to his death", the subjects S(John) are not Agents unlike the subjects of other 'while' clauses. Since more Dsemantic roles can fit into the 'while' clause structure, it can not be a semantically conditioned structure in natural language. My interlanguage data also did not have 00 any occurrences of 'while', so we need to find a different technique to distinguish asubjects from syntactic subjects.

Another test for subject which is not useful for my study is the behavior of Ci floating quantifiers. The quantifiers that launch a test for subjecthood are those that are postverbal but which refer to the subject: The boys went all to the country of their origin; the boys were all writing letters. In each of these sentence, the quantifier 'all' i refers to the subject. The semantic.role of the quantified NP is an independent variable and does not restrict the NP which can be subject. In other words, semantic roles are Sirrelevant to which semantic role can "launch" quantifiers (Kroeger, 1993: 29-30). No evidence of this use of all was found in the corpus.

If in interlanguage data, we find only one semantic role along the developmental continuum, we would have to argue that the semantic versus syntactic properties of subject are not clearly distinguishable. Thus, in interlanguage at this stage "we would have a direct connection between overt form and semiotic categories" without an intervening level of grammatical relations, a situation which Andrews proved does not obtain for at least some adult natural languages 109). In other words, we need a test which offers evidence for distinguishing the development of semantic (a-structure) versus grammatical relations which include syntactic subject in interlanguage"' 6.1.2 Standard Test for Syntactic Subject in s3's Interlanguage First, let's review the propositions for s3 that we have already seen in Chapter 4 produced during Task IV. During Week I, s3 produced six Existential-there propositions and two characterizational propositions. In Chapter 5 we showed that the semantic role of 'theme' would apply to Existential-there and to some Characterizational Proposition Types.

The two Characterizational Propositions Types are some people has got light.. and light is light has....on the stick. Although have may be a transitive verb, both structures are stative in meaning. Statives can not take agents; they may, however, take experiencers. But the semantic role of experiencer is restricted to animate objects. The first example although animate, denotes ownership and this semantic role is not considered experiencer. The first argument in the second example is inanimate, therefore it also does not denote an experiencer. Moreover, the learner is shifts between has got, and is -has, which both take themes. So, the learner is only using the semantic role of theme in his characterizational utterances.

Earlier in the same interview, when asked to describe a festival near his hometown (Task II), the learner again uses a theme instead of an agent.

Example 6.1) Discourse i2-s3:2 i2: Uh excuse me...Is there...Is there any famous festival in Akita? i2: Could you explain about one? s3: Kamakura..

i2: Kamakura festival? s3: Kama.. Omagari fireworks Ii2: Uh huh Could you explain? s3: Kamakura is made made from snow which is hard When the opportunity arose to use a suppressed agent in a passive-by phrase, S s3 pauses and then opts instead to use a stative inchoative. The matrix clause of the inchoative uses theme for subject and object of the preposition snow. Following snow, 00 the learner again uses a theme in the relative clause, which is hard. We might argue Sthat creating a passive structure is difficult in early interlanguage. However, a pattern is emerging of a learner who seems to be constrained by semantic roles. He is overwhelmingly producing themes. Opting not to use the passive structure, may be part of his interlanguage system at this point in time.

SHence, for this learner the relevant question is not whether he can produce Ssentences without agent a-subjects, but rather, can he produce any sentences with agents? Otherwise, we could say that all his syntactic subjects are themes, and Itherefore he shows no evidence of an intervening level of grammatical relations with syntactic subjects in his interlanguage. They can all be explained with one semantic Srole.

However, we have already seen that different semantic roles may surface in different tasks. In Chapter 5, we saw that the embedded clauses in the complement clauses of want, try, decide, need, and in order to clauses occurred in Task I.

Although these clauses usually have agents for the subject of the complement clause, they don't always have agents. It depends on the semantic meaning of the complement as defined in Chapter 4, Existence, Identification or Characterization Proposition Type. So, we can use these structures as tests for agent, as well as the other semantic roles.

During Task I, s3 no complement clauses were found in this discourse. One week earlier, s3 did have complement clauses with want. The relevant sentences are underlined. The previous chapter, discussed complement clauses which have obligatory subject control. One of these is want. This learner used want in the introduction with interviewer il. Since subject is obligatory and always ellipted in these structures, s3 obviously has used a subject in this structure.

Example 6.2) Discourse il-s3:1 1 sl you know, so I want, I want to enter the school 1 il mm 1 sI to aet some experience I i l okay, let's see, ah, since you're at this school, I think you plan to go to the States to study in the future, now what would you like to study 1 s3 like, I would like to study psychology but I do not have much knowledge about psychology for, I have to get more information about psychology 1 ii okay, why would you like to study psychology? S s3 because I heard in the USA it is very high level, mm, development, mm, In Japan psychology is not so popular, but in USA I heard psychology is popular, this is the reason ii okay, so that's why you want to go to the States, but why would you like to study psychology? \N 1 s3 why, to, I want to get job about psychology 1 il mm 1 s3 (xx) 1 il so you, you, I guess you want to be a counselor or something S1 s3 yes, mm, I want to be a psychologist 00 However, what is not clear is whether he has a clear command of PTVs and agents. Notice that in these complement structures with want, the learner uses a construction with an agent. In two utterances he uses get and in a third utterance he Cuses want to enter to get. Thus, he is using agents(+goal), but it is restricted to want and get. In two other utterances, he uses want to be. The semantic roles in the Scomplements are themes, not agents. Therefore, the learner's vocabulary is limited C1 with respect to want and the complement, but he did produce agents one week earlier.

In Week 1 of the analysis, s3 produced the following short discourse during 0 Task I in the same week as the above structures with theme during Task IV.

Example 6.3) Discourse i2-s3:1 i2: Good afternoon.

s3: Good afternoon.

i2: My name is Naoko Miura.

I'm a student at MSU-A.

What's your name? s3: My name is XXXXXXXX i2: Where are you from? s3: Akita city. 2 i2: You... Akita city? s3: Yes. 3 i2: laugh I'm also Akita city.

s3: Really. (laugh) uhm What grade...are you? 4 i2: What grade? s3: xxxxx i2: Uh..Second...

How long have you been studying at MSU-A? s3: Ten months i2: Ten months s3: uh Ten months 6 i2: Do you enjoy MSU-A? s3: Yes. 7 I have a couple friends 8 I sa...

I enjoy here. 9 i2: do you have many American friends? s3: Both of Japanese and American friends i2: Uh huh What will you study in the America? s3: Psychology. 12 i2: Psychology? s3: Yeah 13 i2: You are interested in psychology? s3: Yeah. 14 i2: uh huh s3: How about you? i2: Hhm...l want to study about biology.

s3: Uh huh.. 16 Ii2: in the U.S.

SUtterances 1, 4, 8, and 9 are complete sentences. The first two both use the copula, so are stative, hence take the semantic role of theme: gives the name of the student in Sa list of attributes or characteristics, and asks for the interviewer's level in the ESL program. The last two are PTVs with the verbs have and enjoy. These two have 00 agents and experiencers, respectively, as subjects and 8 takes a patient as direct .I object. The ninth utterance uses an adverb of location, here, to denote the place s3 enjoys. Technically, this utterance can not be considered a PTV at the level of syntax, because here is an adverb of location and cannot be the object of a transitive verb.

However, in Chapter 5, we demonstrated the deictic use of here and there as pronouns. In this interpretation, these can be considered direct objects in PTVs.

NI Possibly this is a first step in acquiring direct objects for this learner.

N So, for Week I of the start of the analysis, after examining two separate Tasks, I and IV, with different communicative purposes, we discover that s3 has characterizational sentences and two PTVs in Task I and Existential- there sentences ,I and Characterizational sentences in Task IV. We can clearly identify three semantic roles for a subject, themes, and one agent, and one experiencer. The learner is beginning to develop utterances using more semantic roles than theme. Although most utterances are themes, except for two, the learner has also shown evidence of using agents in the prior week. So, we can posit a level of grammatical relations for the learner independent of a semantic level of a-structure. Since the learner has shown the use of more than one semantic role, we need a statement to capture this in his grammar.

Another set of tests apply in ordinary main clauses and a second set of tests apply to multi-clause sentence structures. At the level of ordinary main clauses, three coding properties which can be identified with subject are: word order, casemarking, and cross-referencing. For example, subjects in English have preverbal coding, and limited case marking for pronouns, and cross-referencing for person and number between the subject and verb. Other languages may have no regular obligatory word order, but extensive case marking or cross-referencing. Japanese has no regular word order, but extensive case marking. Interlanguage studies of morphosyntax have primarily used word order and cross-referencing for English as the TL (target language).

Another test for subject is ellipsis of coordinate structures. Clear tests for subject are also found in complement (see above Example 6.2 with want) and coordinate clauses. This learner showed possible evidence of the structure for the test for coordination in the preceding week in this utterance: Example 6.4) Discourse il-s3:l 1 ii so how do you, how do you know about the Hiroshima style of(xx)? 1 s3 it was very big and very delicious, The postverbal adjectives both give properties of the it, referring to the Hiroshima style of ramen The phrase very delicious, could have been part of a full sentence including the pronoun subject, it was very delicious. The ellipsis of the subject is evidence that learner can manipulate syntactic properties in the sentence according to their relation to other grammatical functions: both predicate adjectives refer to it and, IDtherefore, one pronoun can be deleted. We need, however, more evidence that the O learner is not just conjoining adjective phases.

0 Another test for ellipsis is cross-referencing. In the TL, English, crossreferencing occurs between the number marking of the subject and its verb. One Sproblem with this test is the complication of tense. Not all verbs in the past tense are marked for co-reference with the subject. The verb have and the copula be do mark 00for person number even in the past tense. Sentences 1, 4 and 8 above show evidence of cross-referencing between the subject and the verb: name is are you I have and later in the introduction with ii, the past tense it was (Example 8. Thus, we have three examples of the learner's ability to use cross-referencing with the subject and verb and also two different verbs in the same discourse.

Another test for subject which the learner displays is nominative marking.

Although English does not have an extensive system of case marking it is found in the pronoun system. Although I think it is dubious that a test for subject can be 0 constructed when the learner only uses I, a deictic marker for person as described in O Chapter 3, once the learner displays evidence of marking for more than one case, we can then set up a test. In the sentence immediately preceding (Example the learner uses the dative form together with a nominative in the same sentence, so he told me.

Example 6.5) Discourse ils3:! ii oh, okay, it is August the sixth that the bomb was dropped, so how do you, how do you know about the Hiroshima style of(xx)? s3 ah, myfriend, myfriends told me, myfriends went to Hiroshima to school to eat, so he told me it was very big and very delicious, so I heard, I heard (xx) i l do you make a lot of money? s3 mm, not so much but it's good for me because we can get money, we can learn how to contact with cust, customers As demonstrated, the learner uses the dative me, and the nominatives he and I, both preverbal, functioning as subjects. It is unclear if the speaker wants the definite pronoun he to refer to the plural NP friends or preceding NP, the singular friend.

Since the learner is still having trouble using the plural pronoun they two weeks later, as shown in Chapter 4, this discourse suggests that nominative marking precedes his mastery of plural NP marking.

Finally, this learner shows evidence of mastering the word order of English subject-verb, i.e. SVO, sentence initial subject followed by the verb. This differs from the canonical marking of the learner's NL, Japanese SOV with the verb in clause final position, following the direct object. So, we could hypothesize that the learner has mastered the non-verb final word order found in English as opposed to his native language.

This learner has possible evidence for subject because of several tests: 1. Subject ellipsis in complement clauses 2.

Subject ellipsis in co-ordinate clause 3. Cross- IN referencing between the subject and verb 4. Case-marking for two grammatical functions, nominative and objective Word Order 00 The converging of these tests seems to indicate a "clustering" of properties toward the development of grammatical subject. We still need more conclusive evidence for #2, although it is possible.

SThroughout this discussion I have mentioned the Characterizational Proposition Types of this learner. We already have seen that he produced Existential- Sthere sentences (Propositions in Chapters 4) during weeks I and II, but none in Week r III. Thus we can conclude that he can produce dummy subjects. The absence of Existential-there sentences appears to be due to a developing anaphoric system in the interlanguage.

S6.2 subject and Semantic Specificity Relations At the syntactic level, if we only look at category labels, such as noun and verb, and do not analyse both their function with respect to grammatical relation of subject and object, and also the function with respect to definiteness/specificity, we can not distinguish types of semantic predication. The hierarchical attribute-value matrices or the f-structures of LFG, proposed by Kaplan in 1975 do not include a mechanism for distinguishing semantic relations of definiteness or specificity, although a morpheme may be marked [+Def].

Some linguists have discussed distinguishing predicative from nonpredicative: John is my name versus John is a lawyer, noting they have different predicating functions (Williams, 1980; Higgenbotham, 1987:45-50). In LFG, no formalism indicates in the f-structure that in the first instance the NP following the verb is identificational, i.e. "Names and definite descriptions are non-predicative; indefinite descriptions are predicative" (Higginbotham (1987:50). This must be deduced by the grammarian. The notion of specificity relations must be inferred from an analysis which designates category labels on individual morphemes, such as +Det.

In Chapter 4 we saw that three patterns are created using specificity relations, Existential, Identificational and Characterizational, labelled as semantic definiteness relations. These are independent from the syntactic relations of subject and object, forming an entirely autonomous level in the grammar. They don't belong in the fstructures as envisioned by Kaplan for distinguishing the abstract grammatical relations of subject and object from the sequences of surface words and phrases of cstructure. However, they do intersect the at the level of the sentence.

The first semantic relation, Existential-there propositions we found Chapter 4 during week I, where the learner produced all but two propositions using this relation in Task IV, as his patterns of relations began to change. The Existential-there construction in English has been considered an obligatory subject due to the SVO status of the construction, and the requirement that English sentences must have a subject. Existential-there has a 'dummy subject' because it has a verb preceded by 'there' and followed by a noun. The subject in this semantic structure is [-spec] since it has no link to a [+def] in the discourse.

The second semantic structure, the Characterizational type, is illustrated in Example 6.4 above: it was very big and very delicious. The it, an indefinite pronoun, refers back to the Hiroshima style of noodle soup, in this discourse fragment from Task I. Therefore, this is [+spec] and the information following, very big and ,O delicious, characterizes the taste of the soup, and since it has no link to the discourse, O must be coded [-spec].

0 The third semantic structure, the Identificational, is found in Task III, which requires the learner to ask the interviewer questions for specific information to fill in Dthe blanks of the information Gap chart about four Tohoku festivals. The learner V produced three Identificational Propositions, as shown below in Example 6.6. (The oO interviewer, i2, also produces an Identificational Proposition as the answer to the first question.) C Example 6.6) Discourse i2s3:1 s3: Uh...hum..Top of the left blank...What's the festival name? (TI) i2: It's uh Tsuzureko Odaik s3: What Kamakura age? (T6) \O s3: What How...What age of Daichiro Bugaku? (TI 1)

O

We can say that the learner has produced a subject in all three of these utterances. At (TI) the WH-question word behaves as the subject and the information following behaves as a predicating property giving a property of identification. The fact that the learner is still not producing verbs in the structures at (T6) and (TI1) is a sign that his notion of an English predicate is still developing. Although this structure requires more study, the nominals following the Wh-question seem like they are behaving as a property from Aristotle's List sketched in Chapter 5. As an Identificational, this subject is also coded as [+spec], and not distinguishable from the characterizational semantic structure. However, the NP it refers to on the chart establishes an identificational relation as well as the NP in subject position, What.

6.3 Conclusion In this chapter, the tests for syntactic subject have shown the learner has subjects, as well as the three semantic relations. But the tests for subject are at a level of grammatical relations, independent from the level of semantic definiteness/specificity relations. They are two autonomous levels in the grammar. The subject in the Existential type is [-spec], in the Characterizational [+spec], and in the Identificational is [+spec].

The equational relation, the Morning Star is the Evening Star has always been an ambiguous structure as to which NP is the subject, if there is a subject. We saw in Chapter 4 and 5 that this construction does have semantic meaning at the level of the proposition and within the proposition, interpropositons of complement and in order to constructions. In addition, its semantic specificity relations distinguish it from the other two types of propositions.

In Chapter 5 we saw that the semantic role of each of the patterns can all be 'theme'. Hence, the semantic role of theme does not distinguish the specificity relations. In this chapter we have seen that the semantic structures again have an autonomous level of meaning within the local domain of the sentence and grammatical relations.

Chapter 7 aSentences: Syntactic Topic 00 7.0 Introduction The study of definiteness and topic in second language acquisition has taken several forms, and different theoretical and experimental models have been used to describe N interlanguage. Huebner (1983) summarized these models as: 17- Transformationalists have traditionally held that the structure underlying all languages is subject-predicate and that the topiccomment relation is a surface phenomenon (Chomsky, 1965). Grundel ID[sic] (Gundel, 1974) has gone to the other extreme, proposing an Sunderlying topic-comment structure for all sentences of all languages.

r (Givon, 1979b) argues that syntax arises from paratactics and points out that pidgins and child language are heavily topic prominent (Huebner, 1983:85-86).

Li Thompson (1976) take a less extreme position and say that some languages are topic prominent and others are subject prominent. Researchers of creolization and pidginization in language have declared that topic prominent languages can syntacticize topic and change it into subjects Andersen (1983b). As a result, theoretical questions and research directions in the field of second language acquisition have been slightly different. Some researchers have asked the question "Is L2 characterized by an early stage of topic comment?" Others have asked "Do learners transfer topic comment features from L1 to L2?" Still others have adopted the pidginization models and studied the processes whereby topics become subjects.

The notion of definiteness, on the other hand, is still rather new in the field of second language acquisition. Although definiteness and topic are known to have an association, the nature of this link is still unclear. The few studies on definiteness have tracked the development of the acquisition of definite articles, as a marker of definiteness. This does not directly address the question of definiteness and its connection to topic in a sentence. The issue of definiteness and topic has also recently become clouded by the notion of specificity and its relation to definiteness, discourse and the syntax of the sentence. Moreover, many languages do not have any overt morpheme marking either definiteness or specificity in the syntax.

Since topic is one area of interlanguage studies related to my thesis, I briefly review prior studies, to distinguish their research questions and results from mine.

These studies can be divided into two major theoretical paradigms: the formalist paradigm proposed by Chomsky and his associates, on the one hand, and the functionalist paradigm, on the other.

Dik (1978) provided eight criteria for a summary of the differences between the two paradigms of formalist and functionalist studies of language. I have chosen to use these criteria since they provide a clear basis for distinguishing the goals and objectives of subject/topic and topic/communication. As a result, my overview of the studies and models used in interlanguage studies is organized differently from other discussions, such as Larsen-Freeman Long, 1991; Ellis, 1994) which do not IND describe, for example, Schumann's Pidginization Hypothesis (1978a) as being interactionist.

According to Dik, formal theories view language as an "abstract object", such as a "set of sentences", at a "rather high level of abstraction", and the grammar is Sviewed as a set of formal rules independent of the meaning and use of the constructions in the sentences. Alternatively, grammars are seen as a relationship of 00 sound and meaning with an autonomous system of rules. The grammarian first establishes the system of 'arbitrary' rules, and afterward studies the meanings and uses which the constructions actually have in performance.

Syntax is thus given methodological priority over semantics, and semantics is given priority over pragmatics (where the latter may be defined as the system of rules governing the use of linguistic C expressions). This formal paradigm is, of course, the basic view IND underlying Chomskyan linguistics (Dik, 1978:1).

SIn second language research these are considered "Nativist theories", which try to explain learning based on an innate biological endowment Larsen-Freeman Long (1991:227).

The functional paradigm, on the other hand, views language as "an instrument of social interaction between human beings", whose primary purpose is to promote communication between speakers and addressees. Verbal interaction, "social interactions by means of language,... a form of structured cooperative activity", is studied to see what people actually do in these interactions, how they use language. These activities are structured in as much as they are "governed by social rules, norms, or conventions" which together form the rules of "the system underlying verbal interaction". This system manifests itself "in the form of utterances", which are themselves structured. This structured "cooperative activity" is "governed by rules which together constitute the language system" The types of rule systems which linguists in the functional paradigm are concerned with are both social in nature: (1) pragmatic rules governing the "verbal interaction as a form of cooperative activity"; and semantic, syntactic, and phonological rules governing the "structured linguistic expressions used as instruments in this activity" (Dik, 1978:1- Dik's "schema" showing the major differences in the two paradigms is given below." Table 7.1 Formalist Versus Functionalist Paradigms FORMALIST PARADIGM FUNCTIONALIST PARADIGM 1. The primary function of the expression of thoughts communication a language 2. How to define a a set of sentences an instrument of social interaction language 3. Psychological correlates Competence: the capacity to communicative competence: the ability produce, interpret, and judge to carry on social interaction by means sentences of language 4. The system and its use the study of competence has the study of the language system must logical and methodological from the very start take place within 200 00 priority over the study of the framework of the system of Language and setting 6. Language acquisition 7. Language universals 8. The relation between syntax, semantics, and pragmatics performance the sentences of a language must be described independently of the setting (context and situation) in which they are used the child constructs a grammar of the language by making use of his innate properties on the basis of a quite restricted and unstructured input of linguistic data to be regarded as innate properties of the human organism syntax is autonomous with respect to semantics; syntax and semantics are autonomous with respect to pragmatics; the priorities run from syntax via semantics to pragmatics language use the description of linguistic expressions must provide points of contact for the description of their functioning in given settings the child discovers the system underlying language and language use, aided by an extensive and highly structured input of linguistic data presented in natural settings to be explained in terms of the constraints inherent in the goals of communication, (ii) the biological and the psychological constitution of language users, (iii) the settings in which language is used pragmatics is the all-encompassing framework within which semantics and syntax must be studied; semantics is subservient to pragmatics and syntax to semantics; the priorities run from pragmatics via semantics to syntax (Dik, 1978:4-5) Even though current linguistic theory has undergone many changes since TG, Dik's formal criteria, the relation of syntax, semantics and pragmatics as autonomous systems is still true today.

In this chapter, I summarize general differences between formal and functionalist paradigms for the study of language as provided by Dik (1978). Using his criteria, I first address the formal studies, based on the subject-topic typology proposed by Li Thompson (1976). This section is preceded by a summary of their typology, since most of the formal studies use this model. Secondly, I review the functionalist-based studies related to topic. These are primarily based on the models for second language acquisition of Givon (1979, 1981, 1983, 1984, 1985) and Schumann (1975, 1978) and the pscyholinguistic models of Meisel, Clahsen and Pienemann (1981) and Pienemann (1998). Many of the functionalist studies on topic are also sociopsychological, appealing to notions from grammaticization theories for the explanation of pidgin and Creole languages proposed by (Bickerton, 1973) and (Schumann, 1978).

Finally, I conclude by implementing the three semantic definiteness patterns into the level of information structure. None of these approaches address the question of definiteness and specificity, other than Gin (1994) whose work includes definiteness as part of the notion of topic, but does not mention specificity. I demonstrate that although specificity and definiteness are properties of the notion of topic, specificity is not a sufficient feature for defining topic, a binary function defined with the notion of given and new.

I 7.1 Formalist Studies O Li Thompson (1976), trained in the tradition of Chomsky's generative grammar, proposed that some languages are topic prominent and others are subject prominent.

Q Their typology proposed a four-way distinction in languages of the world based on the Sdegree of topic-prominence versus subject-prominence in a respective language.

Several influential studies since that publication have used their typology to study the 0 0 question of whether interlanguage is topic prominent or subject prominent (Huebner, 1983; Fuller, 1987, Jin, 1994). Since the topic-prominent characteristics of language are what have been the core of grammatical studies of studies of early interlanguage, I CN briefly summarize the characteristics which they propose for all four types, with special emphasis on topic-prominent and subject prominent languages, the opposite cK, extremes of the typology, before I review these studies. I return to their model in Chapter 7 where I go into more detail regarding the differences between topic and subject in current models of grammatical structure.

SLi Thompson provide the following table of four language groups based on their typology of four basic sentence structures (b-sentences)(p. 460).x" Subject-Prominent Topic-Prominent Indo-European, Chinese, Lisu Niger-Congo, Finno-Ugric, Simitic[sic], Dyirbal (Australian) Indonesian, Malagasy Subject Prominent Neither Subiectand Topic-Prominent Prominent nor Topic-Prominent Japanese, Korean Tagalog This typology prompted researchers, including interlanguage researchers, to look for evidence of topic-comment structures as a basic structure in language. The overall emphasis of the studies was to describe characteristics of topic-comment languages and show how they differed from subject prominent languages. Of the five sections in the Li Thompson article which argue for their typology, interlanguage researchers have only addressed the section on Characteristics of Topic Prominent Language, so I confine my discussion to that section in this review. The outline of their proposal is: 1. Properties subjects and topics do not share 2. Characteristics of TP languages 3. Topic-comment structure in TP languages as a basic sentence type and its implications for the study of universal grammar 4. The typology and some diachronic implications Studies of interlanguage have drawn especially from and above (Huebner, 1983; Fuller Gundel, 1987; Jin, 1994).

As we go through a description of the above items which have influenced SLA research, I point out areas in the proposed typology which I believe are unclearly/ambiguously written thus having lead to confusion for the studies of interlanguage. Also, it is important to be aware that Li Thompson were referring to and depending heavily on other articles contained in the same volume from the seminar on the theme of subject and topic.

O 7.1.1 Properties Topics and Subjects Do Not Share This section of the proposed typology has lead to some confusion on the part of ri interlanguage studies. Li and Thompson are trying to present differences between topic and subjects, but focus the discussion on the properties of topics. The seven properties or processes which they present are not necessarily exclusive to one or the other.

00 7.1.1.1 Definiteness The first characteristic Li and Thompson discussed is 'definiteness'. Citing Chafe (1976), a definite noun phrase is one for which "I think you already know and can identify the particular referent I have in mind" (Li and Thompson, 1976), they identify (Ni definiteness as one of the primary characteristic topics must have. Using Chafe's ri definition defined in the same volume (Chafe, 1976) based on world knowledge, they Nstate that "proper and generic NPs are also understood as definite"(p. 461) and definite common NPs as definite noun phrases. They conclude this contrast of subject ri versus topic by arguing that "A subject, on the other hand, need not be definite", and cite the following two examples: A couple of people have arrived A piece of pie is on the table.

Clearly "a couple of people" and "a piece of pie", the subjects of the two sentences, are not definite, especially according to the definition given by Chafe above. Confusion about application of this definition arises because definite noun phrases can also be subjects of sentences, which Li and Thompson do not explicitly point out. A subject of a sentence need not be definite, but it easily can be. For example, the two examples above for subject can easily be made definite and still function as subjects of the respective sentences in English: The couple have arrived.

The piece of pie is on the table.

In other words, definiteness does not distinguish topics from subjects; both topics and subjects can be definite in most languages (Note The lack of definiteness would, however, according to their definition, exclude an indefinite noun phrase as being a topic.

Unfortunately, studies of interlanguage topic-comment structures have used definiteness to prove the existence of a topic and have not been careful to also count this as providing evidence for subject (Huebner, 1983), (Jin, 1994). In other words, it cannot be used as a criterion to distinguish the development of topic versus subject in interlanguage. It must be shown conclusively that all nouns in the interlanguage are definite and not simultaneously functioning as subjects, in order to raise a question about the possibility of a topic prominent language which has no subjects. This has not been shown for interlanguage.

7.1.1.2 Selectional Relations The criterion that a topic need not have a selectional relation with any verb in a sentence, whereas subjects are always selectionally related to the verb, i.e. always an argument of the verb, has not been adopted by any previous interlanguage study of ISO topic, so I do not elaborate here. However, in Chapters 5 and 6, where I discussed O arguments of a verb and properties of subjects, I briefly discussed this property.

S7.1.1.3 Verb Determines Subject But Not Topic SCorrelated with the fact that the subject must be selectionally related to the verb is that the subject can "with certain qualifications" be predicted from the verb. Again, no 00 0 interlanguage studies of topic have made use of this criterion so I will not elaborate further in this section. However, I briefly discuss the "certain qualifications" in Chapter 5, in conjunction with lexical characteristics of some arguments (selectional relations).

C( 7.1.1.4 Functional Role (Ni c- This characteristic of topics is used by Huebner (1983), so I address this property IO briefly now and then again below.

According to Li and Thompson, the topic is the "center of attention" and (N therefore it states the theme of the discourse. "The functional role of the subject differs in two main ways. First, subject-prominent languages contain "dummy" or "empty" subjects which according to Li and Thompson's interpretation do not play any semantic role.

Secondly, they explain that the function of the subject NP when it is not empty, can be defined within the confines of the sentence irrespective of the discourse 464). Citing Michael Noonan they add that the "subject can be characterized as providing the orientation or the point of view of the action, experience, state, etc., denoted by the verb." This is not explained and no examples are given, so it is not clear what exactly they mean by "orientation" or "point of view" from their personal communication with Noonan.

It appears that they may be attempting to separate "center of attention" as a characteristic of topic from subjects. They have actually made two implicit claims here: topics are a property of discourse and subjects are a property of sentences.

Regardless of the characterization of "center of attention" versus "orientation" or "point of view", they have suggested that the two are properties of two separate domains, a fact which gets buried in their attempt to differentiate the respective characteristics. This is noted by Huebner, but not elaborated. The context of his discussion and problems with his analysis are addressed later.

7.1.1.5 Verb-Agreement Li and Thompson point out that subject-verb agreement is common in languages, but topic-predicate agreement is rare if it occurs at all. Their explanation for this is that "topics...are much more independent of their comments than are subjects of their verbs" 465) and refer to the facts of 7.1.1.1 (definiteness) and 7.1.1.3 (verb determines subject), above. Subjects, on the other hand, which do have selectional relations with some verb in the sentence, commonly represent "inherent properties of the subject noun" (their emphasis) through morphological agreement affixed to verbs as a type of surface coding 465). However, it is not claimed and is not supported by evidence that all subjects in all languages must have morphological agreement markers. They are only claiming that topics almost never do, in contrast to subjects which frequently do.

Acquisition researchers have known for a long time that morphological agreement marking on subjects is difficult, as evidenced by its order of acquisition compared to other morphemes in the morpheme order studies (reviewed in Larsen- O Freeman Long, 1991). However, distinguishing subjects in the developing grammar from topics has not been as great an interest to researchers. Fuller and Gundel (1987:8-9) are one exception who study this characteristic with respect to Ssubject versus topic properties, and I discuss it below in the review of their study. In addition, this characteristic of subjects was used by (Meisel, Clahsen, Pienemann, 0 0 1981) in constructing their psycholinguistic model of stages of acquisition in interlanguage. However, their model, as explain below, was not a study of the acquisition of topic; even though they include topicalization as a property of developing word order, with a later stage of subject-verb agreement property. They never mention topics in the model and do not use or identify the distinguishing c,1 characteristics of topics and subjects (see 7.2.2) S7.1.1.6 Sentence-Initial Position SIn the languages that Li and Thompson examined, topics were always sentence-initial, even if they had morphological markers. Subjects, however, are sometimes sentence final, in languages such as Malagasy and Chumash, or sentence-internal in the case of VSO languages, such as Arabic and Jacaltec. Huebner (1983), Rutherford (1983) and Fuller (1987) mentioned this property of topics and also made reference to this property of topics and subjects.

However, again confusion has resulted in interlanguage studies because Li and Thompson do not make it clear that this property does not distinguish topic from subjects, since both can be sentence-initial. It is simply that topics were always sentence-initial in the languages they had examined. Neither Huebner nor Fuller Gundel mention that this is also a coding property of subjects in early interlanguage, especially in rigid word order languages like English. In fact, many word order studies have been conducted in second language acquisition which note that SVO, canonical word order, develops first. However, because the word order studies were not also discussing topics, but only the syntax of the clause or sentence, this overlap in sentence position/function has not been addressed.

Lambrecht's definition of topic allows for sentence final topics in some languages, e.g. French (Lambrecht, 1986). Hence, sentence position alone is insufficient as a criterion for establishing topichood (cf. Halliday, 1967). In addition, Lambrecht (1994) classifies sentence topics into four categories according to their function in discourse. He, thus, has a much tighter definition than that of Chafe (1976), upon which Li and Thompson base their discussion. However, I show that these four categories are not necessary to distinguish the notion of topic, which I analyze as a two place relation.

7.1.1.7 Grammatical Processes According to Li and Thompson, since topics are syntactically independent, they do not play a prominent role in processes such as reflexification, passivization, Equi-NP deletion, verb serialization, and imperativization. Subjects however do. Since topics are not involved in the internal syntactic structure of sentences, they play no role in the above five processes.

Of these processes, only passivization has been mentioned in interlanguage studies of topic, and then only to claim its non-existence in their respective interlanguage data (Huebner, 1983; Fuller Gundel, 1987).

IDIn my analysis, I posited two types of subjects, one in the grammatical O structure (gr-structure) and the other in the argument structure (a-structure), following Manning (1996). With this analysis, the above reasoning for the relationship of topics to their role in these five processes must be modified. It requires a description of the Srespective processes in both structural levels and the relationship of topic to either one or both of the structural levels.

00 S7.1.2 Characteristics of Topic-Prominent Languages After a comparison of the properties of topics and subjects respectively, Li and Thompson then discuss the grammatical implications of topic-prominence versus subject-prominence. This is the list which has shaped the design and analyses of C interlanguage studies within the formalist paradigm.

IND 1. Surface coding SAccording to their description, TP languages have surface coding for topics but "not necessarily" for subjects. The coding of topics can be marked either by position in the sentence, e.g. Mandarin where it is always sentence initial, or with a morphological marker (e.g.

Japanese). In languages which mark the topic, the subject is not necessarily coded, although in some cases both topic and subject are marked, such as in Japanese and Korean (p.466).

As mentioned above, subjects do in fact have coding, so this is confusing. If one admits sentence position as a type of coding, then they can both be said to be coded. In fact, Fuller Gundel (1987) use this as the coding device for topics. The other form of coding, ie morphological marker, has been posited by interlanguage researchers (Huebner 1983), but I show that this analysis is not compatible with other facts about interlanguage and can be explained in a manner which is more consistent with both interlanguage studies and natural language.

2. The passive construction In TP languages, passivization either never occurs or is a marginal construction with a special meaning, eg the adversitive passive in Japanese. Subjects, on the other hand, are so basic in subject-prominent languages, that if any changes are made to the subject which should normally hold this position grammatically this "non-normal" subject must be marked 467).

An analysis which looks for passivization in the spoken interlanguage must keep in mind current research on discourse regarding its function and distribution. For example, Yule cites the work of Brown, et al. (1980) whose corpus of 50 hours of conversational speech had very few examples of passivization, which was found primarily in the written language (Brown Yule, 1983).

Hence, the absence of"subject-creating" devices in interlanguage must be treated with caution. The few studies which have examined passivization in interlanguage and not found any examples have taken this as clear evidence against any subject-creating abilities of the learner. Moreover, complications of identifying the structure of IO passives in interlanguage data can complicate the analysis, e.g.

O Huebner's 1983 analysis, discussed below and in Chapter S3. "Dummy" Subjects '"Dummy" or "empty" subjects are never found in topic-prominent languages. The location and use of dummy subjects in English must be oO kept in mind when looking for their presence in language. Both Huebner (77-78) and Fuller Gundel report that their absence in the data was evidence for their non-existence. Schachter and Rutherford ri (1979) and Rutherford (1983) noted the presence of dummy subjects in their data. I showed that the conversational topic and discourse 0 structure must be taken into account when studying dummy subjects.

0 In fact, the presence of "dummy subject" can be used to track the development of learners and their proficiency level (Chapters 4 and 6).

NO

4. "Double subject." Li and Thompson simply state that double-subjects are pervasive in topic-comment languages and do not discuss the lack of this structure in subject-prominent languages. (Gundel, 1988) disagrees and claims they are found in all languages. An example of this structure which Li and Thompson give is (p.468): "Fish (topic mkr wa), red snapper (subj mkr ga) is delicious"(Japanese).

Although no studies in interlanguage study this phenomenon explicitly, Huebner (pp. 78-80) and Fuller Gundel briefly mention it. Huebner's explanation for the absence of double subjects in his data is its absence in "running narratives" and need for at least two conditions in order to be grammatical: 1) "shift in topic"; 2) the topic must have "no grammatical relation with the verb of the comment predicate" 79). Similarly, Fuller Gundel claim that since "topics do not have to be integrated into the case frame of the verb, in topicprominent languages "a topic and a subject can simply occur sequentially in a so-called "double-subject" construction in which the topic is not coreferential with any noun phrase inside the full sentence comment that follows it, and is not preceded by a special construction such as the as-for expression in English that sets it off from the rest of the sentence I believe that this structure is more complicated than represented here by Li and Thompson they appear to be topicalized genitival constructions and see Kroeger's 1993:32-33 discussion of possessor ascension in Tagalog). Since I have no examples of double subjects in my data, an alternative explanation to all of these interpretations would require a special study to test specifically for double subjects and is beyond the scope of this thesis.

Controlling co-reference.

According to Li and Thompson, "the topic, and not the subject, typically controls co-referential constituent deletion". An example which they provide is: "That tree (topic), the leaves are big, so I don't like it" 469).

O The fact is that pragmatic, semantic and syntactic factors can be O involved in ellipsis and deletion (mentioned in Chapters 5 and S6. V-final languages.

SLi and Thompson simply cite the work of others that Japanese, Korean, and Lisu are "mature and indisputable verb-final languages" and their 00 own work that Chinese "is in the process of becoming one"(p. 470).

They suggest that all verb final languages are Tp. To date no interlanguage researchers of topic prominence have claimed to find CK, only verb final utterances. Huebner briefly discusses this 83-85).

However, word order studies demonstrate that acquisition of ,IC subject precedes acquisition of object, which would suggest an early C1 SV structure for transitive verbs which would simultaneously mean verb final utterances. However, I have not found this analysis or interpretation in the second language studies of topic.

N 7. Constraints on topic constituent.

In this section, Li and Thompson argue that topic-comment constructions in Sp languages are "highly constrained in terms of what can serve as the topic constituent". The only example they give to demonstrate this is from Indonesian, which only allows the "surface subject" and its genitive to be topic. This contrasts with topics in Tp languages which have no constraints on topic (470-71 .)cx.

8. Basicness of topic-comment sentences.

This section simply makes the claim that the most striking difference between a Tp language and a non-Tp language is the extent to which the topic-comment sentence can be considered to be part of the repertoire of basic sentence types in the former but not in the latter. (p.

471).

There is no explanation of why this might be the case but is referred to in a later section where evidence for this fact is given. This explanation actually leads to a rather loose and personal interpretation in the study by Huebner. I return to the notion of basic sentence types proposed by Keenan (which they adopt in this discussion) in more detail below.

In summary, in topic-prominent languages according to the Li and Thompson typology: I. Topics are coded in the surface structure.

2-3. Passives and dummy subjects are rare or non-existent.

4. "Double subjects" are a basic sentence type.

Topics tend to control co-referentiality.

6. All verb final languages are Tp.

7. Topics are highly constrained in Sp languages.

8. Topic- comment sentences are basic.

IsN Not all eight criteria have been addressed by all interlanguage researchers of O topic, although Li and Thompson's proposal states that all syntactic structures need to be considered in order to establish a typology of topic prominence. I use the order of the above criteria to discuss the various studies and then look collectively at the Sresults in the final discussion of all of the studies.

Huebner's longitudinal study (Huebner, 1979, 1983) was one of the first 00 comprehensive studies of the development of topic prominence in a second language learning context to use the typology which Li Thompson propose, employing notions from Bailey (1973) and Bickerton (1973) dynamic paradigms for the C analytical procedures in his thesis.

The subject of Huebner's study was one adult male in his early twenties who 'c was a native speaker of Hmong and had lived in Hawaii for only one month before the C commencement of the study. Huebner determined Subject, Ge, had shifted "the total \4 organisation of propositions and arguments from topic-comment-like structures to more subject-predicate-like structures". Huebner's analysis of this shift was limited to his subject Ge's changing use of three forms: the form is(a); the article da; and the anaphoric devices of pronouns and zero (p.

5 3 Is(a) was found to be used as a topic marker, consistent with the need for a topic marker in topic prominent languages. Secondly he found that Ge initially treated subject NPs in English as always definite, as if they were syntactic topics (Jin, p. 103). His analysis of the developing anaphoric system examined the acquisition of the processes of the form, the assignment of person, case, gender and number to the form, and finally, the discourse rules, which he claimed govern pronominalization and zero anaphora, rather than derived structures from a transformational grammar's (pp. 86- 87). However, there are basic flaws in some of these analyses: is(a) follows presupposed information but it is from the preceding question. This results in topic markers as free morphemes sentence-initially, in spite of the fact that they are bound morphemes in natural language. It also puts them immediately in front of focused material. His discussion of discourse versus sentence rules I discuss below.

Huebner's predominant concern was the variation present in the interlanguage and he was strongly influenced by Bickerton and Bailey's recently developed models of variation. Statistical measures were made of the three structures mentioned above, especially with respect to their occurrence in obligatory environments. This is puzzling since Huebner's goal was to study the interlanguage as a system in itself.

Huebner never defined the meaning of obligatory in a developing system independent of the target language SE (Standard English).

Although this was an ambitious attempt on the part of Huebner, certain of his critical analyses have weakened his claims. I address the relevant linguistic criteria below. Finally, although future researchers have referred to his work stating that he showed that interlanguage was topic prominent, he himself is not as strong in his claims on this point.

Rutherford (1983) approached the question of topic versus subject prominence in interlanguage by asking the questions: Are there common tendencies in interlanguage development for all language learners? and Are there tendencies for the native language influence on learners' interlanguage structures?"v He predicted that both tendencies would be manifest in the interlanguage. Rutherford's data is drawn from writing samples (the entrance-examination essays) from five groups of learners of English from both TP and SP languages: Mandarin, Japanese, Korean, Arabic, and Spanish. The essay question was "How a man or woman chooses a wife IDor husband in my country."'xv Rutherford hypothesised that serial verb constructions O occur in such frequency in the data because of transfer from the native language, Mandarin, since they did not occur at all in the compositions of the other speakers.

Furthermore, he hypothesised that the influence of the mother tongue transfer was Sstrong enough "to override more general acquisitional strategies that limit the early production of such constructions." 00 A second set of data came from responses to an elicited test given to a subset of the same group of subjects who were asked "to write opening sentences", i.e. an appropriate first sentence of a composition. This test again found evidence of r overproduction of topic sentences by TP speakers, especially the Chinese speakers and although the Mandarin subjects' production of unextraposed sentential subjects Swas not statistically significant, they still produced full clauses as sentential subjects r and serial verbs."'v Nonetheless, Rutherford claimed that his Mandarin subjects followed the results found by Huebner (1979) with speakers moving from topic-comment to Ssubject predicate forms in the interlanguage. Furthermore, it followed the predictions of Givon (1979) of syntacticization "in which discourse and/or pragmatic relations are gradually reanalyzed as grammatical or syntactic relations through the emergence of grammatical machinery" (Rutherford 1983:363). Rutherford postulated that speakers of TP languages went through six stages of what he termed syntacticization in learning a SP language like English: "what looks like subject-predicate form in some learner production at the earliest stages may, I feel, be rather conceived as topiccomment". He finishes the statement, however, by adding, "although it is a very difficult claim to substantiate(p. 363)." Rutherford's cycle of the Mandarin speakers interlanguage development of syntacticization begins with "what started out as artificial subject-predicate but real topic-comment finally emerges as real subjectpredicate and artificial topic-comment 365)." The Japanese and Korean speakers produced a different set of structures which Rutherford attributed to mother tongue influence: existentials with subject-placeholder "there" and "extraposed sentential subjects with "it" as the subject-placeholder. In other words, there was a large number of "dummy subjects". Rutherford postulated that this behavior of the Japanese and Korean speakers, "partial GWO, rigid verb-final SOV languages, [in which] there are no grammatical place-holders" was also due to transfer from the mother tongue, but it was not a "transfer of crosslanguage surface features" of the "conventional language transfer 366). Instead it was due to the sensitivity of Japanese and Korean speakers to the GWO typology continuum since they are more rigid in the verb-final status in their language than the other three languages in the group. Thus, this is an example of transfer of "mother tongue typological organization, viz., a comparatively strong predilection for the use of word order to signal grammatical rather than pragmatic relationships 367)." The other three languages in the group, Mandarin, Spanish, and Arabic, which did not show the same predilection for rigid word order and the demand for "grammatical place-holders" demonstrate mother tongue transfer of pragmatic word order.

However, an examination of the actual results suggests a problem in this interpretation 367): Table 7.2 Production of Existential Constructions VLanguages Number produced Total possible Percent 00 SMandarin 19 (174) 11 Spanish 10 (120) 8 Japanese/Korean* 41 (234) 18 Arabic 44 (396) 11 Even though the percentage of utterances between the Mandarin speakers and the SArabic speakers is identical, the number of existentials is not. If we take into account IND the discourse context for existential/presentatives these numbers cannot be considered the same.

C Rutherford drew two main conclusions from his study. All learners, irrespective of mother tongue, follow common routes of the acquisition of the aspect of syntacticization process of topic-comment to subject-prominent acquisition of the sentential subjects by Mandarins and existentials by Mandarin, Japanese, and Koreans. These same acquisition routes of acquisition are influenced by differences from the mother tongue "extra-heavy topic comment influence (Mandarin) and grammatical word-order sensitivity (Japanese and Koreans). Finally, he concluded that of the three types of typological primes, only the topic vs. subject prominent and pragmatic vs. grammatical word-order figure directly in transfer. The canonical arrangements of S,V,O "does not". In other words, learners did not transfer the word order from their native language, i.e. Japanese verb-final or Arabic sentenceinitially. Since topic-prominence and pragmatic word-order are discourse phenomena, and S,V,O are syntactic phenomena, Rutherford further concluded that "it is therefore discourse v and not syntax that gives gross overall shape to interlanguage(p.368).'cxv..." A 1987 cross-sectional study by Fuller Gundel investigated the role of topic- comment structure in the second language acquisition of English by native speakers of "highly topic-prominent languages", Chinese Japanese and Korean (5)and speakers of "relatively less topic-prominent languages, Arabic Spanish and Farsi for a total of 20 subjects. A twenty minute film, "The Golden Fish", which has only music as its sound track and was similar to the film designed by Chafe (1980) was used to collect the data. After viewing the film, the subjects told the story in English to a native speaker of English. Then they told the story in their native language to another speaker of their language. Thus three sets of oral narratives were collected based on the film: L1 narratives in English L2; (2) L1 narratives by the same speakers in their respective L1; and English narratives produced by English L1 speakers (control group).

Topic is defined pragmatically in a footnote (fn 4, p. 17) as "what the speaker intends to communicate something about in using a sentence." Additionally, the authors explain that one's intuitive judgments "can be tested by asking which implicit question the sentence answers in the particular discourse context in which it occurs"(p. 17). This shows an attempt to clarify their meaning for the data analysis.

This study determined the relative topic-prominence in the interlanguage of the subjects by first setting up criteria based on shared syntactic features from studies of languages with topic-comment forms in their canonical sentence types (Li IDThompson, 1976; Givon 1979; Gundel, 1980, 1988; Fuller 1985). The criteria 8 included the following: invariant coding of surface topic, usually morphologically as in Japanese and Korean, but with invariant sentence initial position in Chinese; 2) "double-subject" constructions, where the topic is not coreferential with any NP Swithin the full sentence comment that follows or is preceded by a special expression, V) such as 'as for' which sets it off from the sentence; 3) zero-NP anaphora; 4) subject- 0 creating constructions, such as passive and raising to subject; 5) dummy subjects; and 6) subject-verb agreement.

The seven native languages of subjects in the study were ranked from one to three according to whether the language was high, intermediate in topic-prominence or low in topic-prominence with respect to the six properties listed above. In the analysis of the interlanguage data, the same rankings of high, medium and low are again used to determine the topic-prominent characteristics in the interlanguage data.

IDThe discussion section points out that the data are characterised by consistent sentence-initial placement of topic expressions, so we are to assume that on the criteria of sentence initial NPs, the discourse is topic-prominent. However, as I have already pointed out, topic expressions and subjects in English are most commonly sentence initial. This does not therefore constitute evidence against a developing subject prominent system.

The discussion of "dummy subjects" reports that thirteen subjects out of the total of twenty never used dummy subjects in any of their utterances. However, only one instance was found which would have required a dummy subject but it did not appear (Chinese speaker, p. 13). In contrast, one Arabic, two Japanese and one Korean subject used dummy "it" or "there" in the data. They therefore conclude that this is evidence against considering their interlanguage data as topic-prominent on this point.

Analyzing negative evidence is always difficult, however. We would have expected all subjects to use some device for introducing the narrative, but no discussion is given as to how the subjects which did not use "dummy subjects" actually started their discourse.

In the discussion regarding zero anaphora in the data, the authors point out that of the eight sentences which show zero NP-anaphora in syntactic positions, three of these are from Spanish and Arabic which would use clitics in the respective positions.

The process of cliticizing involves bound morphemes, clitics, and interlanguage usually has no instances of bound morphemes transferring. Hence, the lack of something in these slots could still be explained as a transfer phenomenon in which the learner had not learned how to fill the slot with a full NP and instead simply left off part of a morpheme, the clitic. Regardless, this still can be attributed to a type of transfer phenomenon rather than properties exclusive to interlanguage development.

The only unexplained case is Farsi when an obligatory NP is missing in object position sentence finally. However, they point out that all of the others could easily be due to transfer from the native language.

The discussion of results for subject verb agreement is quite thorough with many examples of the problems encountered in determining "evidence for agreement", such as "is live" and "is win" and invariant is, was and/or has without the fuller set of plural agreement morphemes. The authors conclude "therefore, that the use of agreement is not consistent enough to count as clear evidence against the hypothesis that interlanguage is topic-prominent" 14-15). However, I would argue that they give proof for a developing agreement system in the interlanguage; it is IDsimply not complete, in the same way that the topic-prominent characteristics are not O complete in their data.

SThe experimental design and statistical procedures used is a problem for the final results. Even though the list of six syntactic properties of non-topic properties d listed above were tested within the data, one property was never produced by native V) speakers and in fact its only occurrence was by a non-native speaker. This suggests 00 that the instrument for testing topicality was not sufficiently refined and controlled.

Secondly, the use of only three degrees for measuring acquisition of the six properties can skew the data to collapse inward toward a median of three.

C In the discussion of the results, the authors claim that their "findings do not support the conclusions of Schachter and Rutherford (1979) and Rutherford (1983) N, regarding the influence of native-language topic-prominence in the syntax of Sinterlanguage"(p. 16). They note that this difference could be due to the difference in Stasks: "planned writing versus spontaneous oral production; proficiency level: relatively advanced learners versus their less advanced learner; and the structures: "the relevant structures all involved complex sentences which did not occur in our data"(p. 16).

However, as I have already summarized above, Rutherford's study found differences between the speech production of the Japanese and Korean speakers versus the Chinese in the production of "dummy subjects". My data also found Japanese producers of the dummy subjects. Moreover, the other two other instances of "dummy subjects" in this study were from an Arabic speaker, also consistent with Rutherford's study, a fact and its complications which I discussed above. So, the "dummy subjects" found in the data of both studies were from the same languages; the language group whose speakers in both studies did not produce "dummy subjects" is also the same, Chinese. So Characteristic 3 listed by Li and Thompson as never existing in a topic prominent language, in fact surfaced in the interlanguage of both spoken and written language samples of speakers from these same two language groups.

This does provide some evidence against Fuller and Gundel's interpretation of the results and could falsify Hypothesis 2. As they state, it can be falsified if: topic-prominence in English interlanguage could have been attributed to native-language transfer.

However, since the six utterances were produced by only four out of a total of thirteen speakers from these three languages, there is not sufficient evidence to make a strong claim on the basis of this study alone. In addition, we do not know how the other speakers structured their narratives in order to make a judgement about the presence or absence of dummy subjects in their respective narratives. We would need a discourse anlaysis of this specific property for all the speakers, including the control group. We can conclude, however, that this is an area which needs further research in interlanguage. This study, together with Rutherford's, suggests that this could be meaningful for our understanding about the development of syntax in interlanguage.

As for Hypothesis the situation is more complicated. Li Thompson and Rutherford both group Chinese, Arabic, Spanish, English, and Japanese /Korean in three different language groups (See above, Fuller and Gundel have grouped the same set of languages into only two: Non-TP Native Languages (English, Arabic, Spanish, Farsi) and TP Native Languages (Chinese, Japanese, Korean).

O Table 7.3: Comparison of Degree of Topic-Prominence of Native Language Groups with English Interlanguage Non-TP Native Languages English 6 Arabic, Spanish, Farsi 11 00 TP Native Languages Chinese, Japanese, Korean 18 English Interlanguage In this case they state that it could be falsified if: S(I) the English interlanguage data had been less topic-prominent or equal in Stopic-prominence to non-topic-prominent native languages. 16)

INO

Because of the way they code their data, "dummy subjects" are considered to be "high Sin topic-prominence" in the characteristics of English interlanguage. This does not recognize the utterances with "dummy subjects" from the three languages above.

Subsequently in Table 3, the two-way split skews the data in favor of the TP language group. If the "dummy subjects" had been calculated into Table 3 and Chinese had been split into a third group, the Japanese and Korean subjects would have influenced the ranking to a midway between their grouping of Non-TP languages, since Arabic is the only language family which showed "dummy subjects". Chinese, then, as the only TP language, with no Non-TP properties would have been the other endpoint on their continuum. Thus, the "English interlanguage data..." may have been "equal in topic prominence to non-topic prominent languages".

Jin (1994) reported on the acquisition of topic-prominence and subjectprominence by native speakers of English acquiring Chinese to test the notion of a universal TP stage in L2 acquisition. This study is interesting because it looks at the opposite direction of acquisition from the above discussed studies, the learners were from subject prominent languages acquiring a topic prominent language, Chinese.

Jin's subjects were 46 adults, all enrolled in the Chinese Summer School at Middlebury College. For the elicitation of the data, three production tasks of an oral interview, a story retelling, and a free composition were used. The oral interview asked some general questions in order "to gather background information" but in addition asked the following questions: "Tell me in detail about the school you are attending or have attended"; "Tell me about your experience of learning Chinese either in China or in the and "Tell me about your daily life at Middlebury College". The stimulus materials for the story retelling task were two cartoon films, "The Mole in the Zoo" and "The Naughty Owlet" edited by Erica McClure and Zoltan Ujhelyi and contained no verbal soundtrack. Although Jin says that "this task assessed the participant's productive ability in a narrative context with their attention directed to the use of topic-comment structures in discourse", he gives no other information regarding the way the task was administered or exactly how "their attention was directed" to their use. The third task, free composition, was limited to a description of the participant's family, so that the content style, and vocabulary would not differ drastically among participants." No information is given regarding the instructions for performing the task or how long the subjects were allowed to write 122).

IDAll three tasks were collapsed for the purpose of the analysis because "no O significant internal difference was found in the production data among any of the Sthree tasks" 106). When one looks at the three types of tasks more closely, the reasons for this become apparent. All three tasks could be expected to include Sdescription in addition to narration producing a redundancy in discourse structure and Sthe expected topic-comment structures, so that the three tasks would not be predicted 00 to be significantly different according to analyses of text types. For example, interviews and free compositions may have story retelling and all three tasks may have narratives or descriptions. Each of these types of texts has predictable and different grammatical structures and discourse properties associated with them. This is not a particular problem for the experiment, however, as it is not clear why Jin would have needed to analyse more than one text type for his experiment. In fact, it may r have been an accidental benefit because it provides more similar opportunities for Sspecific utterance types on which to base his analysis.

Language structures analysed for evidence of topic structures were null elements, specificity marking, and double nominative constructions. Jin's study has the potential to be of interest to my thesis, as he tabulated specificity marking in his study. However, he does not give an explicit definition. The morphological forms he studies are demonstratives and bare nouns. The proportion of bare nouns increased across proficiency levels which he attributes to early reliance on demonstratives until learners acquire more lexical items. It is not clear why he chose specificity, nor why he felt it was important.

7.2 Functionalist Paradigm The characteristics of a functionalist paradigm in second language acquisition research also are a reaction against the formalist theories of Chomsky. Today an "interactionist theory" motivates much of the current research being conducted in second language acquisition Long, 1981, 1983; (Gass Varonis, 1986). These interactionist theories as described in (Larsen-Freeman Long, 1991), invoke both innate and environmental factors 249). Environmentalist theories of learning are those which claim that an "organism's nurture, or experience" is more important in its development.

Dik's comment in 1978 that one of the difficulties of this paradigm is that there is "no such thing as a 'theory of verbal interaction"' is still true today. Research being conducted in this area is attempting to develop our understanding of the instruments of interaction, but there is still no theory which encompasses all of the necessary components of an interactionist theory. By adopting Dik's two-dimensional paradigm, I have attempted to finetune the linguistic and communicative components to help differentiate between the two theories of learning mentioned above. Hence the theories and studies which are included in my division of interactionism differs from that of the above authors.

7.2.1 Early Studies One of the first groups of second language researchers to become interested in the influence of the community and the environment on learner language also developed during the 1970's. These researchers began to compare the similarities of pidginization and creolization development with that of second language learner development in the linguistic continuum of interlanguage. "XI According to (Valdman, 1983), most researchers followed the development of characteristics such as those defined by (Mihlhaiisler, 1980): morphological naturalness, increase in derivational \D depth, development of a word-formative component, and the development of O grammatical devices for nonreferential purposes (cited in Valdman 1983:215). The Sconcentration on morphology naturally resulted in few studies of the acquisition of topic since these notions were not of direct interest.

s Nonetheless, the methodology used by the respective researchers and their Sseveral attempts to define and analyze topic are of interest. Based on the language of 0 0 Schumann's informants Jose, Jorge, and Juan, Schumann coined the terms 'basilang', 'mesolang' and 'acrolang' to refer to the different stages of development on the continuum. These stages corresponded to (Bickerton, 1975) description of the C1 decreolization of the Guyanese negation system and its development from Guyanese Creole to Standard Guyanese English. (Stauble, 1978) then proposed the following c schema for a continuum for second language learners: \K Figure 7.1 Interlanguage Continuum Early Mid Late LI Stage I Stage II Stage 11 Stage II Stage III L2 Basilang Lower Mid Upper Acrolang Mesolang Mesolang Mesolang As shown, in this Interlanguage continuum, the lower segment of the continuum is termed the Basilang, in Stage I, the Mesolang, or three degrees of Stage II, and finally the Acrolang in Stage III, the closest to the L2 or target language. Both Schumann, (1975,1978) and Stauble (1978) at first confined their description of interlanguage stages in the continuum to the different types of negation found in the interlanguage. The technique for analysis and classification of the data into the respective stages, ratios of the frequency of the different types of negativing devices over the total number of negatives in the data, was adopted from Cazden et al. (1975).

Initially, researchers using this continuum thought that "basilang speech was made up of English syntax without morphology". But eventually researchers "realized that basilang English was asyntactic in a great many respects" (Schumann, 1984). So they adopted a framework for dealing with more extensive linguistic units from (Givon, 1979a). Givon also represented development on a continuum with varying degrees in a system between two extreme poles of presyntactic and syntactic communicative modes. According to Schumann, Givon predicted that the presyntactic, or nonsyntactic, mode "will prevail in early first language acquisition, early in the history of a language, in pidgins and creoles, and in informal speech among adult native speakers of a language" (Schumann, 1984). This system did not include notions of definiteness and specificity. Nor did it include a definition of topic.

Schumann and Stauble decided to try to apply it to an analysis of data of second language learners. Schumann (1975, 1978, 1982,1987) made no attempt to show the link with preceding sentences and hence discourse connections. Stauble and Schumann, (1983, 1984) looked at verb-phrase morphology in the basilang, the earliest stage of the continuum proposed by Stauble (1978). This research was based on data collected by earlier research and the subjects were six native speakers of Spanish who had lived in the United States for at least ten years, with a seventh from Stauble's 1978 study used for comparison purposes (a control) since he was a lower mesolang speaker and thus more advanced than the others. We can infer from one of the transcripts (p.77) provided that the data was collected by a one-to-one interview.

The major problem with these studies as pertains to this thesis is the lack of a clear 216 IDdefinition of definiteness or specificity and "topic-comment" versus "subject- O predicate".

SSchumann (1984) again attempted to describe the presyntactic (nonsyntactic) nature of the English basilang as produced by native speakers of Spanish, citing Givon S(1979), Rutherford (1983), and Zobl (1983) as researchers who noted that one aspect V) of nonsyntactic speech is a predominance of topic-comment structures over subject- 00 predicate structures). Schumann (1984) compared the ratio of verbs to nouns plus verbs, derived from Givon's suggestion that presyntactic speech in its most extreme form should be characterized by a 1 to 1 ration of nouns to verbs (Schumann, 1984).

r Schumann studied the percentage of Vs, showing the ratio of nonsyntactic to syntactic plus nonsyntactic utterances. From this study he concluded that the value of the verb Sto noun plus verb ratio seems to correspond with degree of nonsyntactic speech (Schumann, 1984). No conclusions about early stages of topic-comment or about Sdefiniteness and specificity were noticed.

Givon (1984) was an application of Givon's attempt to study "discoursepragmatic" or pre-syntactic communication to second language acquisition. In particular, he felt that the early Praguean works which he and other researchers had followed in the early seventies lead them to think in terms of a binary distinction for 'topic'. He had come to the conclusion that this was not the case and that it was a graded property which he termed degree of topicality. The example he gives for this is As for Joe, he gave that one to Mary 111).

He called this "topic continuity or ease of topic identification with clear psycholinguistic potential" (p.111). However, his use of topic continuity is not the same notion used in Chapter 7, which views topic-comment as a binary relation at the sentence level and a second binary relation to the discourse.

7.2.2 Psycholinguistic Models Within Functionalism One of the first interlanguage studies to make use of the notion of "topicalization" and the development of "subject" in interlanguage was the Zweitsprachenwerb Italienischer und Spanischer Arbeiter (ZISA) project conducted by a team of researchers in Germany in the late 1970s, headed by Jurgen Meisel. This study was the result of research on the acquisition of Italian and Spanish "guest" workers in Germany with no formal instruction in German. This model is a sociopsycholinguistic approach to language development, making use of cognitive skills for predicting the order of acquisition of specific linguistic structures. The sociolinguistic aspects of the model refer to an explanation of the variation among learners as they do or do not advance through the psychologically determined stages Although the model, which was developed on the basis of their empirical study, is not usually considered in studies of topic-comment, it employs the notion of topicalization in the development of syntax. In Stage x of this model, canonical word order is posited as the first step in the development of syntax after memorized "formulae". Canonical word order refers to SV(O), the predominant word order of English and German of which word order is critical in both languages for determining the grammatical relation of subject. The next stage of the Multi-dimensional Model, Stage x+ 1, is characterized by the movement of elements either to initial position or final position. Certain versions of this model refer to Fronting (Lightbown Spada, ID1993) as well as "initialization-finalization strategy", which is seen as evidence for the 0 development of syntax, since it indicates that the learner has gone beyond a basic word order structure and can manipulate parts of the sentence (Larsen-Freeman and Long 1991). This has also been referred to as "topicalization". Stage x+3 involves Sinternal movement of grammatical categories, including third person singular agreement on a verb to match the subject, one of the defining characteristics of 00 subjecthood. Thus, the first three stages of this model explain the acquisition of the development of syntax, employing notions of subject and topicalization.

However, no attempt is made to distinguish topic from subject as the model ,i concerns itself only with syntax, and topic is not a property of the sentence and its parts. In addition, definiteness and specificity are not part of the model.

Processability Theory is basically the same model of syntax as the Multi- ,i Dimensional Model. However, this theory attaches itself to LFG at theory of Sgrammar which includes two levels for the role of word order and the function of subject. Topic is included in the functional structure component of LFG, but no defining characteristics of topic are made. Although topic and definiteness are coded in LFG, both subject and topic are in the F-structure separated from the c-structure, (constituent structure), so word order coding is distinct. It is not clear how the PT distinguishes subject and topic since it nonetheless continues with the SVO coding of word order, as the former Multi-Dimensional Model. The syntactic stages described above are the same, with subject defined from word order in the early stages.

7.3 summary of previous studies The notion from the pidginization and creolization hypotheses played a strong role in the analyses of the early period of studies on topic and interlanguage. One of the shortcomings of this approach was to look at language structures as "simplified" or comparing it to the basilang, or in Huebner's case Standard English Even the recent syntactic analyses by Gundel and Jin compared their data to the L2 in order to describe the data. I take the approach that as interlanguage researchers it is more insightful and revealing to identify natural language structural varieties, such as types of predicates, and compare these within the interlanguage rather than to a simplified, fractured Standard variety of an L2. Thus many of the analyses and my discussion were preceded by a desctription of natural language examples.

Both the formalist and functionalist studies, including the Multi-Dimensional Model and Processability Theory have failed in giving a clear definition of definiteness, topic and subject, and how they function in interlanguage. Researchers still have not provided a clear definition of topic and continue to study the question of a topic becoming subject using various techniques, failing to accept the autonomy of the two notions. In addition, the role of definiteness and specificity linking the sentence topic to the discourse has not been studied in second language research.

7.4 topic and information structure The notion of topic-prominent languages as a typologically defining category gave research into specific language characteristics of topichood a new research agenda.

Previously, researchers had noticed the properties of topics and topic'markers in language, but identifying basic sentences (henceforth b-sentences) from the syntactic constructions which could be used to categorize languages that had topic markers had not been attempted. Li and Thompson were the first linguists who tried to define a bsentence type in "topic prominent" languages using either syntactic processes or pragmatic criteria correlated with topics and definiteness, and SLA researchers followed their proposal without question (see Korpi, 2005) for an alternate analysis).

218 IND Inasmuch as interlanguage has been hypothesized to obey the constraints of natural O languages (cf. Eckman's Structural Conformity Hypothesis (SCH) (Eckman, 1984, 1996) c second language researchers should attempt to uncover and describe basic Q. sentences.

SHowever, I propose that the three patterns of meaning based on specificity at V) the sentence level also hold as basic sentences for topic-comment relations. It will be 0 0 shown that topic is found in the same three proposition types as a-subject, as Equi-NP complements, as gr-subjects, and now also for topic-comment relations. Therefore, topic-comment relations have the same pattern of underlying meaning not unique to rC information structure This differs from Lambrecht's classification of topics in IS, since he claims that this is the organization of IS (Lambrecht, 1994). I claim that c-i the pattern of meaning is found in every proposition in language, including the C propositions where the topic-comment relation is found. Thus, Lambrecht's definition of topic, although different from mine, is part of a greater system in 0language. This section shows that the three semantic proposition types also appear in -the syntactic level of the grammar. However, what is important for information structure is that the word order changes depending on its role in the discourse, either given or new information or topic and focus information.

s3: There are many liihts (2) i2: many light? s3: many alot of lights... on the street. s3 Around.uh...around... the light there are.. lots of building (3) The word order for the sentence around around... the light there are lots of building can be explained by pragmatic constraints of definiteness. However, we have defined the three propositions as based on specificity. So, we have lights, which are previously mentioned in the preceding sentence D-linked (discourse-linked), hence [+spec]. Its occurrence as the first phrase of the sentence can be attributed to topicalization or fronting of the 'old' discourse information. Nonetheless, the existential-there proposition still has the base form of existence with [-spec] to which the additional [+spec] feature is added.

The next example also shows a topic as the first word of the sentence.

i2: there are any people? s3: Yes... (looking at picture) 5 )cpXi s3 some people has got light... (6) The topic Some people is what the sentence is about, people have a light. It is D-linked to people in the first utterance confirmed by the learner. Although the pronoun is indefinite, it is linked to a preceding utterance and topic gives new information about the set of people.

Topic is a two-way relation with topic being what the sentence is about and comment (or focus) being what information is added to the topic. Topic can be isolated at the level of the sentence, as claimed by Lambrecht (1994); Gundel (1974, 1988); Vallduvi (1992); Reinhart (1981); Prince (1981). It also bears a relationship to the discourse, as described in the next chapter.

By redefining Li and Thompson's notion of b-sentence so that specificity serves as a template upon which sentences can be analysed within the IS we can look more carefully at both discourse processes and the function of the notion of topic and comment within information structure and sentence grammar. I have proposed that 219 IDthe three patterns of meaning based on specificity at the sentence level also hold as 0 basic semantic structures at the level of sentences for topic-comment relations.

Therefore, topic-comment relations have the same pattern of underlying meaning not unique to information structure. Specificity at the level of the sentence is an Sautonomous level independent of both subject and topic.

~7.5 Conclusion 00 In conclusion, the same pattern of semantic meaning is not unique to information structure I claim that the pattern of meaning is found in every proposition in language, including the propositions where the topic-comment relation is found.

Ci Thus, Lambrecht's information structure categories are part of a system of semantic relations independent of information structure categories in language.

NO Chapter 8 aDiscourse Representation Theory (DRT): An application to an oo Interlanguage Corpus 00 Introduction C Developing referential competence in Interlanguage (IL) discourse consists not only of learning to use the lexical and morphosyntactic forms in the target language discourse, but also of mapping these forms to the referent in the context of the cK situation. Much of current IL research is concerned with the acquisition of lexical and 0 morphosyntactic forms and functions or negotiation of meaning between speaker and Nhearer (Long, 1983; Pica Doughty, 1985).

Mapping morphosyntactic forms and functions to the referent in the speech c situation, or discourse, however, also requires mapping the forms and functions to the real world situation. In order to be able to account for all the empirical data produced by the learner, we need a theory of meaning that considers all the conditions of the IL speech, the sentence, the discourse, the real world situation, and their associated presuppositions.

In this chapter I briefly review techniques for investigating the connection between the lexico-morphosyntax utterances and real world context of the participant speaker and hearers in IL conversational discourse using Model Theoretic Semantics.

I conclude that formal semantics by itself is observationally and descriptively inadequate for evaluating the 'truth-values' of the interlanguage utterances elicited through the four tasks of the experiment.

In addition, I review proposals from cognitive psychologists who have proposed contextual constraints for visual and auditory speech. I conclude that their observations alone are also insufficient for describing and explaining the IL data, especially with respect to referentiality and mutual knowledge in IL conversation.

Throughout this chapter I use discourse 'fragments', the conventional term for constructions used to exemplify a particular phenomena in a discourse, from a grammar of a language. The discourse fragments are from Task I (conversation without a visual aid); III (discourse with both participants using a visual aid) and IV (discourse with only one member using a visual aid). The fragments in this chapter are from the forty-eight dyads of native speakers of Japanese speaking English in the corpus of ninety-six dyads. They have been selected to test the usefulness of both model-theoretic and cognitive psychology models for IL research. It is shown that computational models of language and artificial intelligence can be used for IL research, contrary to claims by Schank and Abelson (1977) but must be modified to account for all of the data.

Section 1 describes the essential precepts of model-theoretic semantics.

Section 2 incorporates insights from Cognitive Psychology for refining the notion of context necessary for an adequate description of the interlanguage data elicited by the tasks. Section 3 further narrows the context to the 'common ground' between the interlocutors. Section 4 incorporates the work of cognitive psychology for describing the types of knowledge available in the context of the interlanguage tasks.

ID8.1 Truth Conditional, Model-theoretic, and Interlanguage Discourse Semantics S8.1.1 Truth Conditional Semantics Since the days of Aristotle, logicians and philosophers have appealed to logic to Sexplain the truth conditions of propositions in natural language. Truth in semantics is Sdefined according to the truth in a proposition. Propositions are based on relations and 0components (variables) of predicate logic. If any of the components of a propositions are false, then the proposition is considered false.

However, in the latter part of the 191 h century, (Frege, 1879), (Frege, 1891) put N forth the idea that no theory of meaning in language is complete unless it accounts for the connection between truth and the meaning of the words and larger linguistic units C and their reference to things in the world. Frege drew a distinction between the sense of an expression and its reference. The reference of an expression is what it IDstands for in the real world; its sense is what connects it to its reference. So, if I ask S"How many students in your class? And the professor replies "ten", if her class has Sten students, then it is a true sentence. The proposition "ten" is its sense and connects the number to the referents, the professor's class with ten students. But if I ask the same question the following year, and again it is "ten", the proposition is only true for that class in that year. The sense and reference of the sentence only refers to that year's class. The truth of a sentence depends on the references of its expressions; the proposition it expresses depends on their senses (cf. Johnson-Laird, 1988:107-109).] In the 1960's and 1970's Richard Montague developed a system of explaining meaning by applying techniques from mathematical logic to the semantics of natural language. This system connects truth-conditional propositions to abstract models representing the real world and also all other semantic details. The method of understanding and describing language in truth-conditional model-theoretic semantics has three underlying assumptions as shown in Figure 1, based on (Dowty, Wall, Peters, 1981).

Figure 1 1. Propositions are truth conditional (p.4) 2. Model-Theoretic: abstract models in real world connected to truth of expressions and meanings (p.

10 3. "Possible World" methodology 'complete specification' of how things are, or might be, down to the finest semantically relevant detail" 12) Dowty et al. (1981) give the following example of such a connection: "The Washington Monument is west of the Capitol Building" This proposition has two different parts: the linguistic entities, Washington Monument and Capitol Building and the states-of-affairs or configurations of objects in the world, x is west ofy. This sentence would be true just in case a certain physical object (entity) named by the words "the Washington Monument" and another entity named by "the capitol Building" stands in a certain relation named by the words "is west of'.

In model-theoretic semantics, if the proposition has a referent in the real world it is true, if it does not have a referent in the real-world of the model, then the proposition is false. This seems very straightforward.

I

ID8.1.2 File Change Semantics (FCS) and Discourse Representation STheory (DRT) However, problems soon developed with Montague's truth-conditional model theoretic framework. One significant problem is the difficulty for predicate calculus of dealing with variables bound across propositions (Geach, 1962). In his example 'A 00 man walks He whistles', cross-sentential variable binding cannot be handled in KI predicate calculus because it cannot extend across the sentence boundary: 3x (man (x) walk whistle Another problem arises when the pronouns are non-referring. For example, the proposition "If this farmer owns a donkey he feeds it", the pronoun it does not refer when the farmer does not own a donkey. Thus, the proposition lacks a truth value (Heim 1983:164-165, von Heusinger and Egli 2000:1-3).

In spite of these limitations, both linguists and philosophers have become IND increasingly interested in model-theoretic semantics because it provides a formal rigor and explicitness to semantics in grammar, in much the same way that Chomsky's ,IC approach gave to the study of syntax.

In order to account for problems like the above as well as other problems of anaphora and referring expressions, theories of discourse-oriented or dynamic semantics developed using different labels: 'situations' (Barwise and Perry, 1983), 'game-theoretical' (Hintikka, 1974; Kulas, 1983), 'file change' (Heim, 1983), 'discourse representation' (Kamp, 1981; Kamp and Reyle, 1993), and 'dynamic predicate logic' (Groenendijk and Stokhof, 1991). In these theories, a cognitive machinery connects the entities in the discourse with the world or situation of the proposition. Problems of reference in model-theoretic semantics are solved by the dimensions of cognitive 'working space', 'discourse domain' in Heim's (1983, 1988) 'File Change Semantics' (FCS) or 'discourse representation' in Kamp and Reyle's (1993) Discourse Representation Theory (DRT).

In the theories of Heim and Kamp, referring expressions first denote entities in the cognitive discourse domain and secondly real world entities. In FCS the discourse entities are introduced on file cards and in Kamp's DRT they are introduced in a discourse representation structure (DRS). Discourse semantics then solves some of the reference problems of model-theoretic semantics by appealing to mental constructs and/or the set of relevant preceding utterances. Once an entity has been entered into the discourse, they are considered 'familiar' (FCS) or 'accessible' (DRT) and are capable of future reference.

The possible world's methodology, the third essential ingredient in Montague semantics, is still a precept of these theories. Stalnaker (1978) proposed the following notion of possible worlds in an attempt to enhance the notion of context of utterance: It is PROPOSITIONS that are presupposed functions from possible worlds into truth values. But the more fundamental way of representing the speaker's presuppositions is not as a set of propositions, but rather as a set of possible worlds, the possible worlds compatible with what is presupposed. This set, which I will call the CONTEXT SET, is the set of possible worlds recognized by the speaker to be the "live options" relevant to the conversation. A proposition is presupposed if and only if it is true in all of these possible worlds (Stalnaker, 1978:321-322).

For FCS and DRT, the context set for propositions is the discourse and noun phrases are treated as variables for instantiation. These theories also distinguish sets of 7 IND worlds from domains and the assignments given to propositions with possible worlds O only used where the speaker's intentions are relevant (Kadmon, 2001:62-63). So, the phrases are 'variables' which must be assigned a truth value, if they are in the discourse.

SAn example of Kamp's Discourse Representation Structures is given below: 0 3 xy people (x) Sfruit (y) Sx bring y The rectangle is called a "universe", a structure used in discrete mathematics C(Bogart, 1988), and represents the context of the discourse. The variables x and y on the top line of the universe represent the discourse referents, in the sense of Karttunen (1976). In this usage, discourse referents are entities which can serve as an antecedent for anaphora, when entered into a discourse by an indefinite NP. At the same time, these entities need not correspond to any particular referent in the existing world or model, only the discourse (Kadmon, 2001:27) In the example above, people is said to introduce a discourse referent and potential later anaphors, such as some, are said to pick up that same discourse referent.

In the quantificational example above with donkey, it is irrelevant if the farmer doesn't own a donkey, because once donkey is introduced, it becomes a discourse referent, and anaphoric it refers to that same discourse referent.

Below these variables the content or referents people and fruit, respectively, are named. All information in the DRS below the top line is called the 'conditions' of the DRS. The final line represents the semantic content of the proposition.

The existential symbol 3 in the upper left corner of the DRS signifies that the contents of the DRS are assumed to exist in the discourse, or DRS. Since the existence of the contents of each DRS is an automatic assumption of Kamp's theory, the existential symbol 3 is often not marked. Once an entity is entered into the discourse domain, it becomes the live options for the developing conversation. Heim's FCS is similar except that file cards are used symbolizing the discourse domain instead of the rectangle signifying the universe of discourse (see Kadmon, 2001). Notice that only nominal referents are represented with variables.

A DRS is true iff"there is a value assignment which verifies the conditions" people(x) and fruit In other words, a DRS is "true iff x can be mapped onto an individual in the model", in this case people bringfruit- so this DRS is "true iff there are such an individual in the model" (Kadmon, 2001:29).

Kadmon provides two notions for truth: "circumstances of evaluation" is always relative to some facts or states of affairs, such as a simple model for extensional logic" or "a model containing possible worlds"; "context of utterance" is relative to the speech situation, "including the location, the speaker, the addressees, various salient object and more" (Kadmon, 2001:8). Verification that an utterance is consistent with the "circumstances of evaluation" or the "context of utterance" is also called its truth value or satisfaction.

These theories do not allow for differing truth assignments, or a mixture of a value of true in the discourse with false reference to objects in the states-of-affairs in 224 IDthe world. According to Stalnaker's possible worlds construct, presupposed 0 propositions are "functions from possible worlds into truth values" and are true "if and only if it is true in all of these possible worlds" 322).

a In the next two sections I compare examples from the IL data with the Sconstructs of context set and common ground and natural language assumed by Smodel-theoretic semantics. Especially the notions of truth and possible worlds context 00 sets are examined for their empirical validity.

8.1.3 Model-theoretic (DRT) and Interlanguage Discourse c Semantics In my corpus of interlanguage (IL) data, an empirical study using four controlled Ci tasks, several mismatches of non-referring reference within the discourse and between CI the propositions and the real world occur. These are not the subtle non-referring types IDof classical truth-conditional model-theory described above, but simpler Sdisconnections between propositions and the discourse and/or the real world of the Ci tasks. Moreover, in my data the truth conditions are not the same for both the speaker and hearer (henceforth Sp and Hr, respectivelyc x) in the description of a festival in the Tohoku region of Japan. This creates complications for using the construct possible worlds and context set where an assignment is true "if and only if it is true in all possible worlds".

These mismatches consist of eight types with a ninth type whose truth values are indeterminant. The first four types are all true in the discourse, i.e. they have a referent in previous propositions entered into the discourse. The last four types do not have referents in the discourse and therefore are not true in the discourse. Stalnaker's possible worlds assigns these propositions a value of false and Kamp's DRT can not process them because there is no value assignment which verifies the conditions of the

DRS.

However, all types have differing truth conditions with respect to the real world of the task and its connection to the discourse (the second assumption of Montague semantics and DRT) and also with the presuppositions which can be assumed in the possible worlds of the discourse. All of the propositions are evaluated with respect to the time of utterance, or NOW, as first proposed by Kamp (1971).

To summarize, Type 1, 2, 3, and 4 have a truth value in the discourse domain, as they have a referent in the discourse. Types 2 and 4 do not have a truth value in the real world of the task. Type 5-8, on the other hand, do not have a referent in the discourse, but do have a referent in the real world of the task or context of possible world. The last type has a referent in the discourse, but its truth value can not be determined, neither in the discourse nor the real world of the task.

An utterance in Type I is true in all of the possible worlds of the discourse; therefore, no problems exist for using model-theoretic semantics theories and IL. A discourse fragment for Type 1 where the discussants are discussing personal details in the introduction during Task I is given below. The relevant propositions discussed are underlined and in boldface.

Example 8.1) Discourse i16s12:1 i16 Where are you from? s12 I come from Kyoto i16 Kyoto, I'm from Hirosaki Tell me a little bit about yourself O s12 I have five people in my family and I am oldest Initially, the speaker s12 is introduced into the discourse by the pronoun you the referent for s12, the hearer at that point in the discourse. In the next proposition, the speaker changes and the referent for the speaker, I, exists in the discourse 00 introduced by you. In addition, the speaker introduces himself with the use of I, C referring to himself as the speaker. So, the referent, I exists in the discourse. The first proposition is simply supplying information for the open proposition, Where are you from? Acknowledgment of the newly added referent, Kyoto, is repeated in the immediately following proposition. Thus, the speaker is confirming processing of the Santecedent after it is made accessible by s12. We can say that the first two N propositions all have a truth value; they refer to antecedents introduced in the Npreceding propositions: the indefinite interrogative pronoun where is the antecedent for Kyoto; you is the antecedent for I.

N The last two propositions in Example 8.1 also provide antecedents and referents in the discourse: the indefinite phrase a little bit about yourself is the antecedent forfive people in my family and I am oldest. These anaphoric relations are said to have a value assignment which verifies the conditions of the discourse. Thus they are true, because there are 'individuals' in the discourse model, where 'individuals' are members of the set of members within each DRS universe.

Task I, the source for this discourse, is to introduce oneself. All of the information about the student and interviewer can be independently verified (although there is no reason to assume that they are not being honest with these details). The real world of the proposition may be personal information, but it is objective and verifiable. The propositions refers to the speakers; thus, a referent in the real world of the task can be mapped directly to the discourse referents. So, s12 asserts that he is from Kyoto and has five people in his family and his rank is the oldest child. In addition, as the referents i16 to s12 interact, they switch between the discourse referents I and you, but in all the propositions they are used appropriately and refer to the speaker or hearer as they alternate. The truth value of the mapping is true for all the propositions and respective speakers.

Finally, the propositions have presuppositions which hold true for all the propositions in the worlds of speaker introductions in Japan.

In these examples for Type I, no problems seem to exist for model-theoretic semantics theories and its usefulness for IL. The propositions are true and the truth of the expressions can be easily matched to the referents. The propositions in this fragment are true in the discourse, true in the task, and true in the presuppositions created by the discourse fragment. The set of three possible truth conditions for these Type I utterances, then, is A second example for Type I is given below: Example 8.2) Discourse i2s3:4 i2: there are any people? (T4) s3: Yes... s3 some people has got light... (T6) First, the speaker (i4) introduces people orally in his interrogative at Time 4 The following one word proposition Yes can be considered as the proposition INC Yes there are people, when combined with the preceding interrogative proposition by O the interviewer, i2. It confirms the existence of people as visible referents in the C picture. People is then the antecedent for establishing the identity of a collection, or a set, upon which the proposition at T6 can be constructed. Some people then picks up this antecedent in proposition T6. The indefinite proposition is specific to the Cantecedent people. The truth conditions for these propositions thus are all true because oO they have a value assignment which can be mapped onto an individual in the discourse model. Note that people is both orally present and visibly present in this discourse task.

C1 Secondly, if we look at the picture which the speaker s3 is holding we can see that the referents people and some people with a light can be verified; there are such C individuals in the model which contains the real world of the task with the picture.

C Thus the propositions of the speaker can be mapped onto visible referents in the O picture.

Finally, the presuppositions created by the propositions in this discourse fragment are also all true in all possible worlds of the task and its mapping to the discourse. We can also give these Type I utterances with both an auditory and visual referent a set of truth conditions for the discourse fragment.

In Type II the value assignment verifies the conditions of the DRS; therefore, it is true in the discourse domain and possible world of the discourse, but its reference between the real world of the tasks and discourse does not obtain. This type of utterance is given below.

Example 8.3) Discourse i3s3:4 4 s3 and there are a lot of people 4 i3 yeah 4 s3 on the street 4 s3 and some people bring the fruit, kind of fruit 4 i3 yeah The learner first asserts the existence of people. Then, in the next utterance, s3 asserts that people are bringing fruit. Therefore, the subdivision of people bringing the fruit through the use of the indefinite pronoun some can find a referent in the discourse.

The learner knows that people are bringing something, but fruit can not be seen in the picture and furthermore are not an identifying property of the Kanto festival. Only s3 knows it is not true in the real world of the task.

However, people do bring fruit to festivals, so fruit does not interfere with the image the hearer, i3, is creating in his mind. Thus, the proposition has a sense in the discourse which both matches the properties of festivals and also has entered a referent in the discourse. Only the speaker knows that fruit are not clearly visible, or present, in the picture. Fruit is not a distinguishing characteristic of the Kanto festival, but a general property of all festivals.

These two examples show that it is possible in IL discourse to assert propositions which satisfy the conditions of the DRS and the possible worlds of the discourse, but which cannot be mapped onto an individual in the model of the tasks.

The first example has an individual which is specific to a particular property of the IND festival not shown in the picture; the second example has an individual which is a O general property of many festivals, although not shown in the picture.

Type III is true in the discourse and task, but it is not a possible presupposition Sin the possible worlds for the festival in the task for both speakers.

Example 8.4) Discourse il4s10:3 00 sl0: oh I I like to talk you to about Kodomo no hi i14: Kodomo no hi ok ,i slO: what is origin in Kodomo no hi? i14: oh the origin of Kodomo no hi 0 ok so that's for um celebrating boys' growth and health celebrating i14: to celebrate boys' growth and health slO: eh what is special facts in Kodomo no hi? i14: ah ok that's a national holiday after World War II sl0: national (xx) i14: World War III think (xx) I thought it's from long time we have been doing but it says national holiday after World War II sl0: oh The speaker, i14, is questioning the data on the chart after she enters it into the discourse. Her confusion arises from the fact that it is true that the festival had been occurring in Japan for many years, but its status as a national holiday legally only occurred after World War II. Thus, although she adds the discourse referents to the common ground of the discourse and they are present in the task, she simultaneously recants and doubts its authenticity. She is speaking to the student while reading information from the chart out loud.

Reading out loud while also presenting information allows the speaker to offer a differing interpretation and therefore presuppositions to the discourse. This gives rise to a truth relation between the proposition in the discourse, task and speaker's presupposition for the proposition.

Type IV is true between the propositions and the discourse domain, but is not true between the proposition and the real world of the task. One situation for this type is created by lexical misunderstandings on the part of the hearer. These propositions exist for a short time, function as antecedents, but then are quickly removed from the discourse.

Example 8.5) Discourse i3s3:4 s3 and there are some kind of cars i3 cards? s3 car, mm The interviewer, i3, has heard cards, instead of cars. According to Kartunnen (1976), Kamp and Heim, once a referent enters the discourse, it is available as an antecedent.

Although the speaker i3 enters the lexeme cards temporarily into the discourse, and therefore cards exists in the discourse, it does not have a discourse referent compatible with the possible world of the other propositions. Thus, it is neither in the INO possible world of the festival that the speaker is developing from the task nor that the O hearer is visualizing.

SYet, the lexical misunderstanding on the part of the hearer creates a discourse referent. Hence, the truth value between the real world and the proposition in the Sdiscourse does not obtain (although it does temporarily in the mind of the hearer Vwithin the discourse domain). This is not the type of utterance that is intended for 00representation by Kamp, but it is a common occurrence in IL (as well as in conversational discourse in natural language). Such lexical misunderstandings exist only temporarily in the mind of the hearer. But they exist in the discourse and r inasmuch as they exist in the common ground of the speaker and the hearer as a proposition in the discourse, referents created by lexical misunderstanding are C accessible as an antecedent. Although they are immediately rejected as a i presupposition in a possible world, they may function as the antecedent upon which Sthe referent for rejection functions. Thus, they are true in the discourse, but do not constitute a possible truth value in the presuppositions of the discourse. Hence, they are rejected. Note that this type of utterance directly excludes an individual from the discourse context and consequently the discourse and presuppositions of the 'context set shrinks'.

Example 8.6) Discourse i14s12:2 114 91 what kind of things are there in that festival? s12 92 sings? i14 93 mm things? s12 94 ah Type V is not true in the discourse domain, but is true between the proposition and the real world of the task. Hence it is a possible property of this festival in the minds of both the speaker and hearer.

Example 8.7) Discourse i4s3:4 i4 171 ok and ah let's move onto the next task <check tape/transcript> i4 172 I give you one picture s3 173 mm i4 174 ofTohoku festival i4 175 so could you explain this picture or this festival i4 176 so that I can ask the festivals name? s3 177 ah hah they walking walking through the mountain i4 178 mmh s3 179 with fire(??) and he they they wears like demon mask demon mask DRT requires that a set for the plural pronoun, they, exist in the discourse. But the learner has created the plural pronoun without any set membership. So, not only has the learner created a pronoun that does not refer, the discourse also has no set membership for they. Yet, s3 continues to use the pronoun they as if it is a member of the discourse.

This particular Type occurs when a language learner begins to use the morpho-syntactic forms to refer but has not yet mastered their use for referring either to a discourse referent or to a referent in the 'real world' context (see Chapter 6, Week 3).

O In Type VI the_value assignment does not verify the conditions of the DRS; 0 therefore, it is not true in the discourse domain and possible world of the discourse.

Neither does its reference between the real world of the tasks and discourse obtain.

However, the mismatches nonetheless create a presupposition which is true in the Spossible world of the utterance. An example of this type of utterance is given below Cwhere the learner describes the activity of a child in the picture of a festival: 00 Example 8.8) Discourse i3s8:4 s8 oh, oh, there is, ah, in, in snow country (i i3 mm,mm s8 some, a child (xx) child make, it was child gather a lot of snow i3mm s8 and make a mou, a small mountain

INO

In this fragment snow and child are first introduced into the universe of i discourse and then the learner, s8, asserts that it is the child who gathered a lot of snow. Note that since the speaker first begins to use the indefinite plural pronoun but stops and changes it to the singular indefinite, a child has a truth value as a subsequent antecedent. If the learner had continued with some, it would not refer since no set of individuals for it to refer to would be accessible. Thus, the learner first introduces a child into the universe of discourse and then explains what the child did. A referent for both the child and snow exist and therefore a truth value between the antecedents and subsequent anaphors can be obtained in the discourse.

However, the proposition it was child gather a lot of snow does not have a discourse referent. No prior utterance has indicated the event of gather a lot of snow.

Since the learner supplies the missing referent for an open proposition that does not exist and moreover does not include the missing referent as a relative pronoun in the event of gathering snow, the structure is malformed both functionally in the discourse and also sententially. Although both of the propositions, it was child gather a lot of snow and child make a small mountain refer to a child, they have no truth value.

Although the event of gathering snow can be inferred from snow country, the referent does not exist in the model. We can say that the truth conditions for the referent child in the propositions in this discourse fragment cannot be verified and mapped onto an individual in the model.

In addition, an examination of the picture in the task shows that no child is gathering snow. All persons in the picture are sitting inside the small igloo-like structure, or kamakura, sitting around a small table-like structure, either eating, drinking, or singing/talking. We also have no evidence of who built the kamakura; we only see completed structures. The process of construction by children is not in the picture and is therefore not part of the real world between the task and the discourse proposition. So, the truth conditions between the discourse and the real world of the task cannot be verified, so are false.

The presuppositions for the propositions in the discourse fragment nonetheless are true. Today the festival is a modern day tourist attraction, and this is a festival completely run by children of the village. The village children do gather snow to build the small kamakura. After the structures are built, the children welcome the visitors to their small kamakura and the shrines inside. As the visitors leave, they make a small coin donation inside the slatted wooden Shinto donation box found in the small kamakura. Later, the money is divided up among the children of Yokote.

IDNot only is it a festival with the primary structures, kamakura, built by O children of the village, children and tourists alike sit inside the small kamakura, drinking amazake, a sweet fermented rice brew specially made for children, as shown in the picture of the task. However, the only part of this festival shown in the picture Sare children sitting inside two kamakura. We don't know if they are eating, drinking, or warming their hands over the small fire. We also don't know if the people sitting in 00 the kamakura are children of the village or visitors.

The process of construction by children is not in the picture and is therefore not part of the real world between the task and the discourse proposition. However, it N is clearly a property of this festival and the presuppositions of the speaker and hence in the possible world surrounding the utterance."' As this is one of the festivals which was in the Akita prefecture and Tohoku Ci area, MSU-A students were taken to this festival by the school (See Chapter Even Sthough the student who made this claim was from Fukasawa, an area south of Tokyo with no snow, he would have been familiar with this festival being a student at MSU- A. Therefore, from his description it appears that it is part of his possible world associated with the proposition at the time of utterance, as used by FCS and DRT.

Children building the kamakuraxxv and sitting and drinking sweet wine and rice cakes in the shrine are distinguishing characteristics which are unique to this festival.

Other Type VI utterances occurred in one dyad when the learner made a comment about an item in the picture being in the West, and then laughed. These also occur when a learner mispronounces a word, such as eating lice instead of rice. If the discourse or picture contain a presupposition of eating, for example, the hearer can make the necessary inference.

Type VII is not true in the discourse, as with the previous two, but is true in the task. In these tasks, the SLA researcher uses IL experiments with bizarre tasks which do not contain realia, as described by Nunan. If a picture from a science fiction novel or movie, for example, were the task, this is probably not in the presuppositions of propositions in possible worlds of the speaker (and hearer) at the time of utterance, but nonetheless is used to produce an utterance with no referent. A student could potentially produce an utterance which is not connected to the discourse, but is true in the weird scene of the task, but is not true in a possible world because it contains entities that do not exist.

Type VIII occurs when the speaker makes a statement which is clearly false or unintended. These are different types.

Example 8.9) Discourse ilsl:l ii and what would you like to study? sl ah, international business ii international business, mm, and why is that? sl why? ii yes sl because, I'm rich ii huh? sl I want to be rich ii oh, you want to be rich sl mm The student, sl, does not intend to say I'm rich and when corrected makes the change to a syntactically more complex structure, I want to be rich. Although I has a INC referent in the discourse, namely the speaker, the answer of being rich is not intended.

O I am rich is not the same as I want to be rich,; therefore, the I in I am rich has no antecedent. Since the task is to introduce oneself, it does not have a truth value in the task. Since the speaker is not rich, it is also not true in the possible worlds of the discourse. The set of three possible truth conditions for this utterance, then, is Type IX has an indeterminant truth value. The truth value cannot be 00 determined from the real world task, although one part of it has a truth value in the discourse. In this short discourse the learner describes a state of a child in the picture of a festival: CExample 8.10) Discourse i3s8:4 s8 s8 oh, oh, there is, ah, in, in snow country s8. i3 mm, mm s8 s8 some, a child (xx) and he is not hungry, child make, There are no words to indicate the desire of this child. How can the learner know if the child is hungry just by looking at the picture? The child is sitting at a table either eating or drinking, or just talking around the small fire. Does this mean that he is eating because he is hungry, or he is eating and therefore is not hungry? Or, is he even eating? The combination of truth values for Types I-IV which all have referents in the discourse domain are as follows. Type I is true in all of the possible worlds of the discourse and the tasks; therefore, no inconsistencies exist for model-theoretic semantics theories and IL grammars. In Type II, the proposition is true in the discourse and the presuppositions in possible worlds presupposed for festivals, but is not true in the picture in Task IV. In Type III, the proposition has a value assignment in the discourse domain and a referent obtains between the real world of the tasks and discourse. However, the facts in the task are not accepted by the interviewer. Hence, the truth value between the real world and the possible worlds of the beliefs or presupposed information by the Hearer/Speaker for the proposition in the discourse does not obtain, although it does within the discourse domain and visible Task III.

Type IV is entered into the discourse, but is not true between other propositions in the discourse domain, and also is not true between the proposition and the real world of the task or its associated presuppositions in the minds of either the Hearer or Speaker.

The next four types, i.e. Types V-VIII, do not have a value assignment which verifies the discourse conditions; all four value assignments for the discourse domain are false. Again, various other truth conditions obtain between the four types in these subclasses. Type V is not true in the discourse domain, but is true between the proposition and the real world of the task. Type VI does not have a value assignment in the discourse and also does not obtain between the discourse and task. Nonetheless, the presuppositions entailed by the proposition can be found in a possible world of the discourse. This type, however, is not entailed by other propositions in the task. Type VII does not occur in my data, because all of the tasks are about objects in the real world, festivals of the Tohoku area. I include it for completeness of the description.

Type VIII occurs when the speaker makes a statement which is clearly false or unintended. Since the task is only being viewed by the learner, the mismatches in the second, third, and fourth type are only known by Speaker. In the final type of referring expression, Type IX, the truth value cannot be determined from the discourse or real world task, although one variable may have a truth value in one or more of the fields.

232 ISO We can construct a table of the above types of propositions with respect to Struth conditions found in the IL data as below. 'Discourse' refers to the proposition uttered by the speaker. A positive sign, means that it has a referent in the discourse or is adding a new potential referent (variable) to the discourse. A negative means that no individual in the discourse is accessible which verifies the C/ conditions of the discourse referent.

oO In the second column, 'Task', contains the individuals which are potential referents. Tasks I and II are completely auditory and thus are very much like the normal conversation for natural language, envisaged and described by Stalnaker, r Kadmon, Heim and Kamp. They entail propositions and semantic details based on discourse knowledge. Tasks III and IV, on the other hand, both have potential visible CI referents as well as auditory referents. For all four tasks, a value assignment must i exist which can be mapped between the task and discourse model. A positive sign, (,i means that it has a referent in the task or is adding a new potential referent (variable) to the context. A negative means that no individual in the task is Saccessible which verifies the conditions of the discourse referent.

In the third column, 'possible worlds' refers to the set of presuppositions for propositions which a speaker may consider possible for the utterance produced. These are from the context set recognized by the speaker to be possible based on existing propositions in the discourse. This implies that they are accessible because an antecedent exists in the discourse and task. A positive sign, means that it has a referent or is adding a new potential referent (variable) to the possible worlds of the discourse. A negative means that no individual in the possible worlds of the discourse is accessible which verifies the conditions of the referent. Sometimes the proposition may be malformed, but the presupposition behind the proposition still holds. In these cases, the value of the proposition is assigned a value.

Table 1. Table of Inferencing Potential Types: Individuals and their Relationships to the Context of Utterance TYPE Discour Tas Possible se k World (s) I II III IV V VI VII VIII IX To summarize, Types I-IV can be assigned a truth value in the discourse domain, as they have a referent in the discourse. Types II and IV can not be assigned a truth value in the real world of the task. Types III and IV lack a value in the presuppositions of the possible worlds of the discourse contexts. Types V-VIII, on the other hand, do not have a referent in the discourse, but do have a referent in the real IO world of the task or possible world of discourse context. Types V and VII can be O assigned a truth value in the task, while Types VI and VIII can not. Types V and VI can be assigned a truth value in the possible worlds of the discourse context, while Types VII and VIII can not. The last type may have a referent in the discourse, but its Struth value can not be determined.

SThis table describes all of the possible occurrences in a discourse of the types 00 of truth conditions of propositions for a discourse. It is intended to describe the various types of situations found in my corpus to more clearly describe referential competence in IL. The requirement of being true in all possible worlds seems I empirically non-useful, and invalid. Thus, this approach to describe meaning in the IL grammar is meant as a first step for the study of mapping the morphosyntactic ,I forms and functions to the referent in the speech situation.

i Since all of these forms are from a developing corpus in the IL, they are O internal to their respective discourse. The claim that truth of an utterance is not independent of its context has also been made by Grace (1987) in his work "The linguistic construction of reality." We might ask then how it is possible that communication is achieved considering it has so many values which cannot be assigned in the discourse situation. For this question, I turn next to the context of utterance.

8.2 MTS, Cognitive Psychology and IL Context 8.2.1 MTS and Context The lack of sensitivity to the context of a proposition by logicians/mathematicians as a means to analyze language has been challenged for over a century. Already at the end of the 19th century, the philosopher (Sidgwick, 1895:281) objected to the lack of attention given to the influence of the special context in which a proposition was used in logic (Seuren, 1998). In order to bridge this gap, File Change Semantics and Discourse Representation Theory appeal to the notion of possible worlds for a definition of context.

Let's look at Kadmon's summary of some descriptions of context sets and Stalnaker's notion of possible worlds in FCS and DRT. Below are Kadmon's definitions of types of contexts (Kadmon, 2001:62-63): 1. Definitions of Context (Kadmon, 2001:62-63) a. A context c is a set of worlds. p is a closed atomic formula.

b. c p c rn w: p is true in w}.

(91) a. A context is a pair (Dom Sat where Dom is a set of variables, and Sat is a set of total assignment functions.

p is an atomic formula.

b. The context c p is as follows: Dom Dom u x, is a variable occurring in p} Sat Sat c n {g g verifies p}.

(92) a. A context is a set of partial assignments with the same domain.

p is an atomic formula.

b. c+p the domain of g is the union of the domain of the functions in c with the set of variables that occur in p, and g is an extension of one of the functions in c, and g verifies p}.

(93) a. A context is a set of world-assignment pairs, where all the assignments have the same domain. P is an atomic formula.

IN b. c p the domain of g is the union of the domain of the functions in c with the set of variables that occur in p, and g is an extension of one of the functions in c, and g verifies p in w}.

The first definition of context is a simple intersection of a proposition in Sthe world. The set of worlds refers to the set of propositions in which the propositions 00 are true. This definition excludes the set of worlds which includes propositions which N are not true. The second definition given in pertains to the discourse situation where not only the variables are in the world of the discourse, but each variable has been assigned a referent and they are all true in the context, i.e. discourse context which includes the propositions and the presuppositions associated with all the propositions in the discourse.

As a metalanguage, this definition of context looks very detailed and specific r, and is more detailed than previous accounts of MTS we've looked at so far. Yet, it is INO still vague. In spite of Stalnaker's linking propositions with the mathematical concept of possible worlds, the notion of context in MTS is still impoverished or insufficient N as a description of context and context set capable of being operationalized for empirical research.

Empirically this notion of context is difficult to operationalize, as potentially it could apply to and define anything. As demonstrated in the previous section, empirically my data are not true in all the possible contexts of the discourse.

Secondly, each DRS only represents the speaker's actual utterance. For assigning definiteness to an utterance in conversation, we need to know if both the speaker and hearer comprehend the utterance in the same manner. How can it be applied to empirical data from a conversation with two participants if only one is represented in the DRS? 8.2.2 Cognitive Psychology and Context The notion of 'possible' is not an empirically valid means for operationalizing context; it is powerful enough to include everything in the context without excluding anything. Noting that vagueness and generality may be virtues when one is trying to describe phenomena, Clark Carlson (1981) object to vagueness when trying to explain a process in a context 314).

In order to remove vagueness, we must distinguish the 'incidental' context from the 'intrinsic' context. Both are part of any process, but only the 'intrinsic context is part of a process needed for 'object identification'. After a review of the varieties of the notion of context used by cognitive psychologists in visual perception, learning and retention, and word and sentence perception, Clark Carlson narrow the set of criteria common to all of the uses for identifying an object, either auditory or visual in the 'intrinsic context'.

Intrinsic context is "information that is available to a particular person for interaction with a particular process on a particular occasion" (Clark Carlson, 1981:318). Appealing to information processing in cognitive psychology, semantic information includes information about objects, and includes their situation with respect to their eventuality, their genericity or particularity, direct experience and inferencing. The availability and interactibility of the information in a context is also important for most definitions of context.

Clark Carlson conclude that three features are common to most of the uses for defining the 'intrinsic context' found in studies by cognitive psychologists (Clark ID& Carlson, 1981:317). These three features are relevant to a particular: 1) person, 2) O process, and 3) occasion.

This gives an obligatory three argument function the agent A of the utterance in the context. The process p, or means of producing the utterance, e.g. written text or Sspoken conversation. And finally, the occasion or time t. We can write the function for intrinsic context as: context 317).

00 Although the theory behind DRT does not intend for the DRS to delimit the intrinsic context of the utterance, it does assume a similar notion of semantic information. As cognitive psychologists, Clark Carlson are more interested in the r study of psychological processes in experimental approaches which require a more analytic notion of context. Considerations of the truth value are not an element of SClark Carlson's notion of intrinsic context. They do not represent the utterance with Srespect to the real world phenomenon in the context surrounding the utterance.

IND

08.2.3 MTS, IL Empirical Conversational Data and Context Clark Carlson's three argument function for context is not part of the functions in the DRS shown in 2.1. In conversational discourse, the first argument, A (agent), is a person; in the DRS it is a speaker and for Clark Carlson it is a hearer. The process captured in the DRS is speech production and for Clark Carlson it is comprehension, either reading or hearing. The time in the DRS is NOW and includes anaphors and quantifiers; for Clark Carlson it is the occasion relative to the process and changes to either 'before', 'simultaneous to' or 'after' NOW, whichever is necessary for the particular occasion relative to NOW in the discourse context.

The only information in a DRS is the variables reflected as referents in a proposition and the corresponding discourse conditions. Although the propositions in a DRS include the semantic information explicated above (objects, events, states or processes, whether generic or particular), they can also be shown to develop over time (see for example Partee, 1984).

The conditions for a DRS do not include Clark Carlson's three argument function for establishing a context for the conversations in the dyads. Each member of the function is not explicitly labelled or required for the DRS; only the semantic content of the speaker's utterance is given in the DRS. We can't tell anything about the process or its 'occasion' in the discourse.

If we modify DRSs to include Clark Carlson's three arguments for the function for context, we can more easily accommodate the requirements of the intrinsic context. Using the empirical data from the types of referring expressions in the preceding section, we can begin to reconstruct a DRS to include the essentials of information in the intrinsic context, the Agent(s), Process and Time Type 1 from Table 1 above is an ideal situation to examine for modifying a DRS, with the speaker's proposition being satisfied between the real world of both the task and the discourse. That is, the referents all exist in the world of both the discourse and the experimental tasks (a spoken conversation), and the proposition connects to all of these referents, i.e. it is true in all contexts.

Example 8.1 for the first type and its fragment are repeated here: Example 8.11) Discourse i16-s12:1 I i16 Where are you from? I s12 I come from Kyoto 236 IDThe conditions for the first proposition exemplifying Type 1 create the mental O representations (MR) (or Stalnaker's presuppositions) for both persons in the Sdiscourse process diagrammed below.

Diagram 1. Speech event as a three participant event.

00 Task C~ 9 \0"n frwm Kyq C" Speaker/Hearer (s12) Hearer (i16) This diagram attempts to show that the speaker is also a hearer as he Iconstructs the utterance, an auditory object, into the conversational discourse. The speaker is both a hearer and a speaker in the context of the utterance.

SHence, in the production of speech we have two independent but highly correlated actions. One includes the intentions and competence of the speaker's speech act. The second action, that of simultaneously being a hearer, includes the performance of the speech act and also, more importantly, monitoring the speaker's knowledge of what has actually been entered into the discourse. The speaker is both participating in the task and also adding information to the discourse context. As far as the dynamicity in the conversation and its developing context, the speaker, (s12) as well as the hearer (i16), need to keep track of what has been entered into the discourse. They both contribute to the addition as well as the rejection of information in the context set.

Thus, the DRS needs to reflect the active participation of both members of the discourse. This can be easily accommodated with the following additions to the DRS.

In the partial DRS below, the learner ID (s12) is inserted on the left and the interviewer ID (i 16) is on the right. Both interlocutors are represented in the DRS.

DRS Representation of Agent(s) in Speaker Hearer Interaction I s12- i16 I Kamp's DRS represent only the speaker's assertion (see §8.1.2 above). Clark Carlson, on the other hand, are characterizing context for comprehension, not production, from a cognitive psychologists perspective. In this revised partial DRS, both interlocutors in the discourse are represented.

The DRS for the discourse fragment I come from Kyoto is shown below. The speaker identification for s12 is given with the Speaker/Hearer on one side and the Hearer, i16, on the other, in the small rectangle on top of the DRS.

Below s12, the proposition produced by s12, I come from Kyoto is given in a typical DRS. The proposition is given in a separate universe, a mini-DRS, immediately below the speaker s12. The instantiation of x=I and y=Kyoto is given in this small universe. The final line, x come from y, symbolizes the semantic content of the proposition I comefrom Kyoto.

237

INO

s12 i16 Ixy xy e, I I I Y SI(x) 00 Kyoto(y) x come trom y x come from y T1 Since both s12 and i16 are also hearing the utterance, the proposition is Sentered a second time into the matrix DRS. This represents the speaking and hearing r by the Sp, s12, and the comprehension process by the H, i16. The variables for this utterance are also given at the top of the matrix universe.

In the lower right corner opposite the utterance entered into the larger Suniverse, or matrix DRS, the proposition is labeled T1 to represent the time of the utterance with respect to the discourse. Since the processes of production and comprehension are occurring simultaneously, the T of utterance is represented in the matrix DRS.

The obligatory function for 'intrinsic context' proposed by Clark Carlson has been incorporated into the modified DRS to include: both participants as Agents in the interaction of information in the discourse context; the Processes of both speaking and hearing.

Time of processing production and comprehension Since the utterance is being comprehended simultaneously to being produced, I have increased the number of Agents to the Sp and H. The 'process' is thus both one of producing and comprehending. We have thus modified the DRS to include both participants, the Speaker and Hearer in the processing the speech act, the utterance I comefrom Kyoto, and the time of the utterance.

8.3 Common Ground 8.3.1 MTS and Common Ground We might ask why consider the Kamp-Heim theories if their notion of context is so vague and incapable of being operationalized for use in IL research. However, with respect to definiteness and topichood, the theories help fill a much needed gap between the two levels of linguistic study: the sentence level of generative grammar and the intersentential levels of discourse.

Kadmon's summary of the main theoretical motivation of the Kamp and Heim models comes from three closely related linguistic goals: giving a general treatment of indefinite NPs, giving a general treatment of definite NPs (including pronouns in their different uses), giving a principled way of predicting the range of anaphora possibilities in discourse (Kadmon, 2001:25).

238 O Topichood is inextricably linked to definiteness, referentiality, and anaphora. These 0 theories can help bridge the gap between the sentence level properties of reference and the common ground of the discourse.

Stalnaker's (1978) notion of common ground is meant to enhance the notion Sof context of utterance. Kadmon (2001) summarizes the essentials of his notion of common ground, as used in File Change Semantics and Discourse Representation 0 Theory's of, as follows: (-i the set of "speaker's presuppositions"; the set of propositions whose truth is taken for granted as part of the background of the conversation; since a common ground is a set of propositions, it can be identified with a set of worlds: the set of worlds in which the propositions of the common ground are true (Kadmon, 2001:9).

O Some logicians object to the notion of a set of possible worlds because of its appeal to logical omniscience (Chierchia and McConnell-Ginet, 2002:318-327) and its lack of 'mathematical manageability'. Linguists criticize possible worlds methodology because of the empirical problems of making a semantic analysis of noun phrases being anchored to a referring expressions with properties (Seuren, 1998:389-392). The linguist Grace (1987) also severely criticizes truth-conditional semantics.

Although the notion of possible worlds and common ground is too powerful, these theories are nonetheless impoverished. As discussed in the previous section, the notion of context is not explicit with respect to referentiality and the tasks. They do not adequately encompass the discourse: the speakers, process, and time are not specifically encoded in the context, three essentials of a "context of utterance".

Common ground is an inherently interactive notion. Both Kamp and Heim's theories are based on the speaker's presuppositions. A 'common' set of 'shared' beliefs, assumptions and presumptions assumes more than one person.""cxxv The precept that the speaker share a common understanding with a shared set of beliefs automatically assumes the interaction of at least two persons. Since they are interacting with a person with a different set of knowledge, the common ground needs to be able to embrace the presuppositions in the minds of both members.

In addition, FCS and DRT do not distinguish or mention the visible and auditory information as evidenced in the context of the IL tasks and discourse. For IL studies using tasks which include visual aids such as the Information Gap Activity of Task III and the Picture Identification of Task IV, we need to be able to distinguish the information available to each interlocutor in the common ground. We need to be able to study the discourse from the perspective of the learner's available knowledge from at least two sets of presuppositions in at least two 'possible' worlds of the context of utterance as they move from the visible information to the auditory information conditioned by the tasks creating the discourse.

8.3.2 Cognitive Psychology and Common Ground The cognitive psychologists, Clark Carlson (1981) object to a definition of common ground, such as Stalnaker's, because "it is infinite in length". Not only is it unnecessary for two people in a conversation to "represent in memory an infinite number of knowledge statements" for processing each proposition, but it is impossible 321). It is only the mutual knowledge at a specific time which is required to process propositions.

IND The 'intrinsic context' of the discourse situation creates the common ground O between two or more people in the discourse consisting of their mutual knowledge, mutual beliefs, and mutual suppositions. Mutual knowledge can thus be represented as an 'elementary mental representation' that is 'inductively inferred' from the limited S"grounds" necessary for a "mutual induction schema" 321). Mutual knowledge is V then treated as a single MR, or entity, instead of the infinitely complex set of worlds.

oO For this we need a manageable set of data which can be operationalized. In order to make the notion of mutual knowledge and context more manageable, Clark and Marshall (1981) delimit the source of three major types of sources for evidence N, available between speakers at the time of utterance: Community membership proper nouns C1 Physical copresence deixis Linguistic copresence anaphora (Clark and Marshall, 1981:42) The sources of knowledge are given at the right in the list above; the N functional manifestations of the three basic types of definite reference are listed to the left. Co-presence refers to knowledge present in the minds of both interlocutors at the time of utterance, or mutual knowledge.

The modifications to the context in the preceding section begin to limit the notion of time, simply by labelling each instance of additional information in the discourse. However, this is still incomplete. It only specifies the time at the exact moment of utterance by the speaker, or the one who asserts the utterance. The modification still does not distinguish types of evidence for the 'information' in the intrinsic context with respect to time. T (time) when incorporated into temporal reference, must be considered not only as what has happened previously, but also whether the exact moment of utterance, NOW, is in fact mentally represented by 'both' participants in the intrinsic context. It does not explicitly distinguish among the three types of evidence available to the common ground of the interlocutors.

Stalnaker's notion of possible worlds applies to what is available to the speaker at the time of utterance. Kadmon claims that DRT and FCS follow Stalnaker's set of the "speaker's presuppositions" considering possible worlds at the time of utterance as part of the background of the conversation. We have just seen problems with the notion of truth in possible worlds for IL in §8.13. This is equally problematic with the intersection of an atomic formula and the world when the background to the conversation is not shared. The visible knowledge and the auditory knowledge do not create a common ground for the interlocutors in the IL conversation in my corpus. How can we represent the difference in their types of evidence, or context of utterance, in the common ground? Communication continues and is successfully obtained even though certain elements of the possible worlds between the task and the proposition, or in Kadmon's terms atomic formula, does not obtain. Hence the definitions of context and possible worlds are contradictory with respect to communication. The possible worlds methodology of MTS need not be true in all possible worlds. The set of worlds in the propositions as part of the background of the_conversation of the common ground need not be true.

8.3.3 IL Conversational Discourse and Common Ground Although Kamp and Heim refer to the speaker and the reference time of NOW, it is confusing that they are positing a common ground and context set among propositions 240 IDif only the speaker's utterance is represented in the DRS. By adding another O interlocutor to the context of the utterance in the DRS, how can we determine if N mutual knowledge exists in the common ground.

Our earlier modifications accommodate the DRS to represent the utterance in Sthe common ground of both speakers as below.

00 I6 c (x x y xy {Kyoto} (y) x come from y x come irom y T1 The common ground is the utterance as represented in the middle of the matrix DRS for both speaker and hearer. In addition, I have added curly brackets to represent the proper noun, {Kyoto}. Kyoto, a city in the southern part of the main island of Honshu in Japan, is from the first source of evidence, listed by Clark Carlson above, Community Knowledge. Japanese speakers can be assumed to share mutual knowledge of its existence and location.

Common ground assumes two persons interacting, at a particular time.

Obviously, the process is not simply production or comprehension; the process is interactive and dynamic. It changes over time and the beliefs and suppositions of the participants' mutual knowledge of a proposition are both mutually exclusive as well as mutually interactive.

Clark Carlson's notion of mutual knowledge is the mental representation inductively inferred by the hearer. Since the MR is inferred by the hearer to be information that the speaker intended for the hearer to assume, we can say that mutual knowledge is based on a presupposition. This agrees with Stalnaker's point that propositions are actually presuppositions. The speaker's utterance is created from the MR (presuppositions regarding information already available in the discourse, as well as what he presupposes to exist in the mind of the .H Clark Carlson define common ground as based on the intrinsic context.

However, we will further limit the common ground to the mutual knowledge that the speaker not only intends, but also what the hearer actually infers, and the speaker accepts. Thus until the hearer acknowledges a certain inferential presupposition about the proposition and the speaker accepts that the hearer's interpretation is the one intended, mutual knowledge does not enter the discourse context.

Various techniques for establishing mutual knowledge in a conversation or discourse exist. Let's look again at the fragment for Type 1 (Example 8.1): Example 8.12) Discourse i16s12:1 1 i16 Where are you from? I s12. I come from Kyoto 1 i16 Kyoto, I'm from Hirosaki Tell me a little bit about yourself I s12 I have five people in my family and I am oldest SIn this example, the speaker i16 repeats Kyoto to acknowledge his understanding of s12's response.

r../3s12 i16 00 c-x y xy I (x) {Kyoto} (y) ~x come from y CI x come from y T1 The restatement, Kyoto, by speaker i16 is shown under the speaker's ID on the right of the DRS. It is also added to the common ground and labelled T2, to represent its occurrence in the discourse.

Different techniques for establishing mutual knowledge are used in Type VII, Example Example 8.13) Discourse ilsl:l ii and what would you like to study? sl ah, international business ii international business, mm, and why is that? sl why? ii yes sl because, I'm rich ii huh? sl I want to be rich ii oh, you want to be rich sl mm This fragment illustrates several techniques for establishing understanding between the interlocutors. The interviewer first repeats the learner's statement, ah, international business, to indicate understanding. He then asks a follow up question to the information just asserted, Why is that?" When the learner repeats the question he again uses a different technique to indicate comprehension and acceptance, a simple "Yes." But when the learner says because, I'm rich, the interviewer objects signified by "huh?" Either he objects to the meaning of the proposition, "I would like to study X because I'm rich" as not a good reason to study X, or he is objecting because he suspects the learner has misstated her intentions.

As she corrects the statement to I want to be rich he acknowledges her acceptance of his objection to the immediately prior utterance: oh, you want to be rich.

Thus, we see several techniques for establishing mutual knowledge in the common ground of the discourse. Only when the interviewer objects, do we find a proposition being removed and a restatement entered into the common ground (as shown in Examples 8.5 and 8.6 for Type IV). All of these changes are helping to stabilize and develop the mutual knowledge between the interlocutors.

Hence, it is inadequate to claim that mutual knowledge is established simultaneously to a proposition being asserted into the common ground. The process of establishing mutual knowledge in INO the common ground is more complex and is not simply accomplished at NOW. It is dynamic, 0 constantly changing and both interlocutors are mutually involved.

The hearer may indicate their interpretation of a specific utterance either by a discourse marker such as "uh-huh" or "hmm" or by making a statement or asking a question which indicates to Sthe speaker that mutual knowledge between them has been accomplished. At this point, mutual knowledge in the common ground in a conversation may be said to exist. (A conversation may 00 continue until interrupted by a revelation that the two people do not share a common ground.) This section has shown that the function for the 'intrinsic context' established in the previous section is inadequate to operationalize context. The addition of Time as part of the function of the 'intrinsic context of each utterance is insufficient. We need confirmation that mutual knowledge is achieved at the time of utterance. In addition, modifications to the DRS IL context in the previous F" section did not distinguish the three main sources of evidence available to speakers at the time of utterance. In this section we have added a conventional method to explicitly state cultural background knowledge in the community of speakers. Clark Carlson listed two other sources of knowledge that Smay exist between interlocutors for the establishment of common ground.

8.4 Sources of Knowledge in Context of Utterance C' 8.4.1 Sources of Knowledge in the Common Ground Returning to the previous section's modifications, we can make further adjustments based on the Clark Carlson (1981) and Clark and Marshall's (1981) sources of knowledge for the information in the common ground in each of the four tasks. The information of each participant in the experiment differs according to the information content available to them in performing the task. Further modifications to a DRS is required to adequately explain the interaction of the referents from the IL data.

The context of the propositions entered into the discourse is more complicated than a simple auditory versus visible referent system between two speakers/hearers. Even with the modifications to the context, the types of anaphora and definiteness are not explicit with respect to the somatosensory perception and language. They do not distinguish the types of visible, visual and auditory information as evidenced in the IL tasks. The information of each participant differs according to the content available to them in performing the task.

Fellbaum (1999, 2002) further developed the three sources of evidence for mutual knowledge listed in the preceding section. The table below is an expansion of the types of knowledge proposed by Clark Carlson and also Clark Marshall for definite reference, which requires mutual knowledge between interlocutors. The three sources of evidence with the proposed modifications for establishing mutual knowledge in the common ground is repeated in Table 2.

Table 2. Classification of Sources in the Common Ground for Reference Types Fellbaum (1999, modified) Basis for Mutual Knowledge Reference Types Community membership -Universal, Particular Visible copresence -Deixis: Visual, visuospatial; visuo-temporal Linguistic copresence -Visible(Reading, Sign Language) -Deixis: Spatio, Temporal; Anaphora -Acoustic -Deixis: Spatio, Temporal; Anaphora Indirect Copresence -Inferencing -Visible Properties evoked from visible objects -Acoustic Properties evoked from acoustic objects and visible or visual objects I\ reference by the speaker to herself is also an example of acoustic linguistic O copresence, the third source of evidence in the chart above.

SThe second source of evidence is Visible Copresence. The referents for Task III include visible objects. In Task III, both the speaker and hearer are looking at a Svisible object and the speech is referring to printed material. For this DRS, we need to modify the DRS to include visible referents. The discourse fragment below is from 00 Task III.

Example 8.14) Discourse s3i2:3 s3: Uh...hum..Top of the left blank...What's the festival name? i2: It's uh Tsuzureko Odaiko The question by s3 and answer by i2 are represented in the following DRS.

I First, the learner reads the information on his chart and asks the question and then the Ointerviewer reads the printed words on her chart and gives the student the answer. The CNI two small universes in the larger matrix universe are labelled with KV to represent linguistically meaningful visible referents required in the process of reading the Information Gap Chart. The question by s3and answer by i2 are represented below their respective ID number in the top rectangle.

Next to each ID number is the source of evidence, KA and KV. Since the speech of both participants is orally delivered, it is part of the Common Ground in the matrix KA universe.

s3 KA+KV Common Ground KA i2 I- KV 3 xy TI Uwx z 3 Uz T2 festival U=xy name Tsuzureko Odaiko}> 3 Q: What z is xy KV 0: z is xv 3 uisz KV uis z

KA

The variables in this chart are first represented in the small universes under the respective ID for each interlocutor to indicate their source as visible referents. The left chart has two variables, x and y on the top line of the learner's visible universe. The first variable represents the visible referent festival and the second name. On the bottom line the question, What's the festival name is represented using the variables, Q:z for the indefinite interrogative What and xy for festival name. The referents for this question, x and y are visibly present to the speaker, even though the question itself is auditorially presented. I have put the interrogative in the common ground with KA.

The universe on the right represents the interviewer's answer, It is Tsuzureko Odaiko. On the top line of the interviewer's universe, the capital U signifies the definite pronoun It, representing a stipulated referent. The pronoun it represented by u, is a stipulated identity; both interlocutors, the learner and the interviewer, must interpret this variable u as referring to the referent in the chart. They must understand IDthat the referent u stands for xy, festival name. Thus, the visual referent x u, is the 8 stipulated entity or pronoun it.

N The third line contains the variable z which stands for Tsuzureko Odaiko. This is enclosed in curly brackets to represent the proper name of the festival. The special Snotation represents the status of proper nouns as anchored to a special knowledge C)type. Both interlocutors have knowledge of this referent from their membership in the 00 community of Japanese festivals, as this festival is nationally known.

In the common ground of this DRS, both propositions are represented. In this task the information between the two participants has a partial overlap in the two ,I universes. Both have access to the information festival name, and its position on the information gap chart. However, only the interviewer has access to the printed ,I information, Tsuzureko Odaiko. The learner must infer that the stipulated referent, It, N in the auditory common ground refers to the visible referent on the interviewer's IDchart.

The third source of evidence, linguistic copresence, includes both visibly and auditorially linguistically symbolic knowledge. It is the latter knowledge that FCS and DRT both purport to explain. They do not distinguish the visible knowledge from reading, although the semantic components of a sentence are basically the same for each type of linguistically symbolic message. We have already explained that KA represents the linguistically symbolic auditory knowledge and KV represents the linguistically symbolic visible knowledge. Type 1 from Table 1 of the Truth Values Types, I am from Kyoto, is an auditorially linguistically symbolic utterance. The visible linguistic symbolism is exemplified when the interlocutor is reading the information on the Information Gap Activity.

The last source, indirect copresence we have not mentioned. This is a combination of any of the three types. We have limited the fragments in this thesis to native speakers of Japanese, so that the mutual knowledge, belief and suppositions from the first source of evidence, i.e. community knowledge, is equally accessible between the interlocutors.

An example of a combination of visible and auditory knowledge is found in Task IV,cxxv the last task. Task IV differs from Task III for two reasons. First, only one participant, the learner, has a visible referent in the context. The other participant has only aural knowledge in the discourse context. Secondly, the information for the learner is visible, but the referents are not linguistically symbolic. These referents are from a picture, not a chart with printed words. This source of information, nonlinguistic visible information, is labelled Ky, with a small v, next to the universe with the visible referent.

Again, the differences in the source of information are labelled at the top of the overall DRS next to the speaker ID. S3 has the picture with Ky, as well as the auditory knowledge, KA, in the Common Ground of the discourse, as the learner has access to the interviewer's speech. The interviewer, on the other hand, has access to only the auditory speech of the learner.

rw*1u. ZaD mm nn 1r-un-i(A ;rm CIW

IA

picrure ICA I w, yz 1 N 3 T1 mm, mmm people( z rl=Zz stree (y) Ky a lot of (I rll 3there are I il z on y yeah 00 KA In this DRS, the learner asserts that There are lots of people on the street is r part of the discourse fragment from Type 2 above. In the small universe under s3's ID, the representation for cardinal numbers rl=Zz is given to the left of the smaller C universe with the referent people and street The semantic information r I for a lot of is given below the referent. In the upper left comer of this universe, the Sexistential quantifier signifies that the referent exists in the context of utterance.

When the speaker asserts a proposition, the information of that assertion is Salso added to the common ground. The interviewer's comments acknowledging the learner's utterances are on the right side of the matrix DRS. These verbal comments are critical for establishing that the interviewer is comprehending the learner's utterances. It is a first step in the development of mutual knowledge. Both parties must have at least a partial agreement on what has been entered into the common ground before mutual knowledge can begin to develop between the interlocutors in the discourse. Unless the interviewer objects, it is a first step in communicating agreement and understanding of what has just been asserted.

This interaction thus represents evidence from visible and auditory knowledge.

The evidence for the learner is both non-linguistically symbolic visible information and auditorially linguistically symbolic. The interviewer, on the other hand, has only auditory linguistic knowledge.

Other combinations of knowledge are possible and appear in my corpus of interlanguage discourse. I have only selected a few examples to illustrate the modifications I propose for the use of DRT in interlanguage studies. We have already seen modified DRSs in Chapter 4.2, where I looked more deeply into the patterns of meaning constructed while developing mutual knowledge in the common ground. In this chapter, I again return to DRT for their usefulness for representing the distinctions between topic and subject in interlanguage.

8.4.3 Truth Conditions in the DRSs of the IL A final consideration in the developing common ground is the truth value of the propositions. We have seen that the truth values of propositions in IL discourse can be classified into at least nine types. The DRSs and files of FCS only track a proposition when its truth can be verified or satisfied in the discourse. Since the empirical data from my IL experiment isn't always verified in the context of utterance with respect to the real world of the tasks, we must either reject these two theories of anaphora and reference for IL discourse, or modify the fundamental principles of truth-conditional semantic theories for IL research.

However, part of developing anaphora and referentiality in second language acquisition is connecting the truth of the proposition to the referent in the real world.

It is part of developing the meaning in language. Today, second language researchers use theories of grammar such as LFG and Minimalism to explain the acquisition processes of morphosyntax, ignoring the discrepancy between anaphors and their 247 O referents. This part of the IL grammar, connecting the truth of the proposition to the O referent in the real world, lacks consistency, completeness and coherence.

C The DRSs described in this chapter offer a means of linking the propositions Sof the developing grammar to the individuals in the real world of the discourse or the Dtasks. The question is how to do this.

j As a first step, we must account for the fact that the information of each 00 participant differs according to the type of knowledge or evidence available to them in performing the task. Hence, the mutual knowledge in the common ground also differs.

Table 1: (Table of Inferencing Potential Types) distinguished three types of situations in CN the interlanguage grammar: the discourse domain the real world of the task either the auditory knowledge from Tasks I and II, or the visible referents in Task III Ci and Task IV; and presuppositions in a possible worlds (PW) for the interlocutors Ci performing the task at the time of the utterance. We need to modify the DRSs to \NO reflect these differences in the context of utterance for the IL propositions.

SExample 8.3 for Type II from Table 1 is given below.

Example 8.15) Discourse i3s3:4 s3 and there are a lot of people i3 yeah s3 on the street s3 and some people bring the fruit, kind of fruit i3 yeah The learner first asserts the existence of a lot of people. Then, in the next utterance asserts that some of these people are bringing fruit. Therefore, the use of the indefinite pronoun some can find a referent from the group of people just mentioned in the discourse.

The DRS below represents these two propositions by s3 from Type II in Table 1.

s3 picture KvlRA Common Ground KA i2 NC Ict re

KA

w, y zr S mm, mmm z I= z people(z) Kv a lot of (1 r1 3there are I rl z on y yeah Z peoe z w z e som fruit (w -z z"'bn X3w--37- ring w yeah

KA

OWe have already explained the first DRS for s3 above. The people are represented o We have already explained the first DRS for s3 above. The people are represented C with the convention for cardinal numbers in the universe for a visible referent.

The second universe under s3 represents the indefinite pronoun, some, in the Sproposition some people bringfruit... in its proportional rather than cardinal meaning.

SA subset of the set of people is introduced by the learner. The subset of people 00 are described as bringing fruit. The tripartite form shows these three units of the construction: the small DRS on the left, the 'restrictor', divides all of the people in the picture; the middle part, the 'quantifier', tells us the nature of the smaller C group, i.e. some the right DRS tells us the 'scope' of the duplex condition, bring the fruit. The 'principal' discourse referent of this construction, some, tells us Sthe proportion of the group who have lights (Kamp, 1993:311).

SHowever, in this example of Type II the learner falsely adds people bringing O fruit to the interviewer's mental image. This knowledge is consistent with Source 1, universal knowledge about festivals, so the acoustic information received by the interviewer is accepted. We can represent three different truth values within this DRS.

Notice that the matrix DRS does not have an existential marker 3. However, the existential quantifier, 3, which signifies that all contents of the DRS exist, including the two participants, s12 and i16, is given on the upper left next to the participants universe.

The first DRS asserts the existence of a lot of people inside the Ky representing the visible referent 'people' in the picture. This DRS also has an existential marker representing the verification of the existence of the visible referent.

In the proposition next to the Ky universe, the proposition in the common ground also has the existential marker verifying its existence in the common ground of the matrix

DRS.

The next KvDRS however does not have an existential marker verifying the proposition asserting the existence of people bringing the fruit. However, the proposition in the common ground of the matrix DRS does have an existential marker.

This is because the proposition can be satisfied within the aural/oral discourse. Since the DRS at T1 establishes the existence of people, the proposition at T2 with some people can be assigned a value within the discourse. It is fully accepted as a proposition within the developing common ground of the discourse.

Neither Clark Carlson (1981), Clark and Marshall (1981), nor Fellbaum (1999) include in their definitions of context and mutual knowledge the model theoretic notions of the truth of the proposition within the situation. Maintaining the existential requirements of the DRT, we can incorporate the truth conditions of the propositions. By utilizing a narrowing of the notion of context and the existential quantifier, we can begin to start tracking changes in truth or satisfaction of referentiality within the IL.

Linguistic Meaning and the Sentence In another view of meaning in grammar, the truth value of a proposition is not of any inherent interest. Linguistic description is concerned with the actual production of utterances and whether instantiation of a noun phrase is consistent among referring expressions. Finding lexical and morphosyntactic consistency, coherence, and completeness between the units of a sentence and between sentences is the primary concern. This is the traditional means of analysing IL development. The interest of the language acquisition researcher has been the developing syntax of the linguistic utterance, and the truth assignment is of no interest.

249 O Kamp's DRT considers the linguistic analysis of DRSs equally important to O the study of grammar, although his theory does not advocate any particular theory of Sgrammar, using GPSG as a convenience only. Thus, any modem generative theory can easily be attached to the modification of DRSs proposed in the previous sections.

w By utilizing a broader notion of context and the existential quantifier in DRSs, we can begin to start tracking changes in referentiality within the IL grammar. For 00 example, in Chapter 4, Example 4.13, no value can be assigned to the pronoun 'they' in the sentence and the real world of the discourse in Task IV, since it does not refer to any referent previously mentioned in the discourse (Heim, 1983:170-171).

C Although it is inferable from the context, the Heim and Kamp models require the discourse variables to already be present in the discourse. Kamp requires an 3 marker Sin each DRS to indicate existence of referents. The modifications that I have just Ssuggested allow us to more accurately represent the sentences and their referential properties with the help of the DRS attached to the LFG structures. Thus, we can 0begin to indicate the learner's level of development of referentiality in the grammar.

SBelow I use LFG as a means to provide the morphosyntactic analysis of sentences for the DRSs. Not only does LFG have special lexical capabilities for capturing the lexical functional features, it also can be easily adapted for capturing the functional structure of specificity and topic at the sentence level.

8.5.1 Discourse Semantics and Syntax: Definiteness and Specificity Three brief examples of discourse-linking and specificity below demonstrate that when Specificity is viewed as an autonomous level at the proposition or sentence level linked to Definiteness (Farkas, 2002; Heusinger, 2002), the role of topic-comment cannot be defined according to either of these notions. It must be regarded as a binary relation at the level of the sentence which receives its features from the discourse.

This represents the role of the context, the discourse and the grammatical linking of subject and topic. This Characterizational Proposition in the learner's early interlanguage is proportional specific, Definite specific, and assigns properties to a [+specific referent]. In this section, I am only interested in showing the proportional specific and the Definite specific and their interface with the Discourse.

8.5.1.1 Characterizational Semantic Structures and the Discourse The Characterizational Proposition at T3 in Example 8.17 is a specific indefinite, assigning properties to a [+specific referent]. In this section, I am only interested in showing the relation of the specificity and its interface with the Discourse.

Example 8.17 Discourse i3-s3:4 s3 there are a lot of lantern in the street TI i3 mm, mm s3 and there are a lot ofpeople T2 i3 yes s3 on the street T2' s3 and some people bring thefruit, kind offruit T3 i3 yes s3 and some people hit a drum T4 i3 yes s3 tit, mm, one dan pass with the pole, pole i3 yes, yes s3 and there are some kind of cars T6 i3 cards? O s3 car, mm T6' SFirst the learner asserts the existence of many lanterns in the street. Then, the existence of a set of people is introduced by the learner at T2 and divided into a subset a of people starting at T3 through T5. This subdivision of individual members in the set is not sequential, but rather all subdivisions refer to the group at T2.

00 This subdivision of the group who bring fruit at T3 is shown in the DRS N below. The tripartite form at T3 shows three units of the construction: the small DRS on the left, the 'restrictor', divides all of the people in the discourse; the middle part, the 'quantifier', tells us the nature of the smaller group, i.e. some; the right DRS tells us the 'scope' of the duplex condition, bring the fruit (Kamp, 1993:311). The 'principal' discourse referent of this construction, some, tells us the c proportion of the group who bring fruit. The learner has both subdivided a discourse ,i referent and assigned properties, in this case actions, to that group.

O

s3 picture Kv+ KD Common Ground KA i3 No pict.

KA

,r people street 3 rT people street T2 mm, mmm 11=Z people people people on street a lot of (I 1 3 Iri people on the street yeah I i| people on street eople(P) T3 <No fruit in the picture> fru t Some people bring the fruit yeah Nv people rnng truit The antecedent for both the larger set of people and this inclusion relationship is visible for the hearer but visual for the interviewer. Although no fruit is in the picture, the interviewer can accept this as a possible property of festivals, so 'accommodates' the learner's description. In addition, both have access to auditory information in the common ground, listening to the picture description in Task IV.

Even though fruit has not been previously mentioned, its entry can be regarded as [+specific].

The context for the description thus includes both visible and auditory stimuli.

These are brought together in the auditory context of the common ground. However, the visible referent which is the input for the learner's speech is maintained as an independent antecedent for the learner. It cannot be an antecedent for the interviewer since the interviewer is not looking at the information in the picture. The interviewer only has access to the auditory objects in the common ground""'"x. Therefore, the interviewer does not realize no fruit exist in the picture. Fruit exist in the common \D ground and universe of the discourse. I have labelled the matrix KA with 3 to represent the existence of the lexeme in the universe of the discourse, but not on Kv Sthe visible antecedent.

The universe at the top of the large KA includes all variables of the ongoing conversation, and is constantly changing over time. As the people are introduced as V being on the street, and then subdivided into smaller groups of people with lights, the 00 context gradually changes the identity of the variables (nouns) in the universe. Since identity of reference is a key ingredient of definiteness and the representation in the universe does not automatically show the identity of the variable as it moves forward C, in time, I have placed the small DRS at T3 into the matrix. This is meant to indicate that a subdivision of people exist in the universe. However, the specificity relations in CI the sentence still need to be stated at the time of utterance to indicate the subdivision CNI is occurring in the universe. The linking at T3 is to T2. This is shown below.

IND

DRS T2 O S(S=T3) 'CPECIF] r NP p SUB PRED 'people' 9t NP NP =4 QU 'Some' Some people V (TBJ )-4 SPECIF (TPRED)=4 ,I\ PRED bring <SUBJ, BJ> LPRED 'fru' bring Det N OBJ NUM L [the fruit SPECIF This description of the level of S is too complex to describe fully here. What is relevant for our purposes is the top lefthand comer which indicates that the utterance is a subset from the DRS above at T2, (Time The arrow pointing down indicates that it is the [+SPEC] features that belong to the DRS set at T2. The [+specific] features are found in the structure marked F (functional structure) below.

This represents that the [+specific] entities some and fruit are [+specific]. They receive their specificty reading from the DRS at T2. Since they are included in the discourse, they are d-linked and therefore a subset of definiteness entities already in the discourse, as per Heim 1982.

From this we can see that two items in the characterizational propositions have a [+specific] marking since they are D-linked. Since there are two items with [+specific] we can see that [+specific] does not function as a marker for topic or focus, since each sentence has only one. Thus the notion at a sentence level that the topic is what the sentence is about is preferred. The subset of people have already been established in the discourse. The comment is telling us that this subset of people are bringing fruit.

Wh questions also are indefinite/nonspecific unless they bear a relation to a previously introduced discourse referent. This use is linked to a set in the discourse, so is +specific. The next example shows a second utterance with [+specific] features for all entities in the sentence. The interaction in Example 8.18 is from Task I in the "Introductions and Getting to Know You".

252

\D

C Example 8.18) Discourse i4-s5:4 i4 Where are you from? Hokkaido i4 which part of Hokkaido are you from? s5 ah ah Asahikabusuki 00 SThe second exchange repeats all entities from the first question except for the question phrase, Which part of Hokkaido. Hence they are D-linked and hence [+specific]. The 0C question phrase can be considered [+specific] as a partitive set-subset relation linked to Hokkaido in the answer to the first question, as being in the scope of Which, an 0 inherently [+specific] question, and also as knowledge about regions of countries as rC having subparts, Asahikabusuki is a region in Hokkaido. Therefore, all entities in C this utterance are [+specific].

0Thus, the topic-comment relationship cannot be determined by +specific since 0 there is no difference between any entities. It can however be determined if we consider the comment to be the question phrase Which part of Hokkaido linked to the topic, are you from.

8.5.1.2 Identificational Semantic Structures and the Discourse In Example 8.19, we have an Identificational semantic definiteness relation.

Example 8.19) Discourse i2-s3:4 i2: 1 have some picture of festival so you can pick up one and don't show me s3: uh huh...

i2: And describe.it and I will try to guess which festival) T1-T9 description s3: Can you guess? i2: Hhm. I think... it is.. Kanto festival TI0' In the final utterance by the interviewer at T10', it is Kanto festival all entities are +specific for several reasons. The embedded clause is an Identificational which automatically includes all +specific features as explained in Chapter 5. It is also linked to the description; which has been developing over 9 utterances by the learner, producing the Identificational relation.

Therefore, the topic-comment relationship cannot be determined by specificity since the answer giving information regarding her thoughts is all specific. It does, however, offer a comment regarding the topic, her thoughts and guesses regarding the identity of the picture.

8.6 Conclusion In conclusion, we need to recognize specificity at the level of the sentence as an autonomous system. It is pervasive in language. Specificity manifests itself at the sentence level, within sentences, and between sentences. Since it can be repeated throughout the sentence, specificity can not be used as a defining feature of topiccomment, also a separate level in grammar. As Gundel (1974, 1985, 2003) has argued, topic-comment is a binary relation at the level of the sentence.

The topic-comment relation is also linked to the discourse. Specificity receives its features from definiteness, or the mental representations in the discourse (Chapter ND Further work is needed to study the exact linkage of definiteness and specificity 0 (Fellbaum, forthcoming).

As shown throughout this thesis, using specificity as a defining level in a grammar linked to the definiteness in the discourse, regardless if it is a set-subset or a J relational linking we can explain all elements in the sentence, independent of morphosyntax. This provides a new pathway to study the emerging patterns of oO meaning in interlanguage in addition to acquisition of the morphosyntactic elements of meaning.

O Chapter 9 SConclusion 00 Introduction This chapter presents a critique of the experiment and corpus, a summary of significant findings for each of the chapters of the thesis, a brief outline of the implications for theories of semantic and conceptual structure, concluding with suggestions for future research suggested by these findings.

i 9.1 Limitations of the study Corpus-based study is always under certain constraints. Although most of the experiment proceeded as planned, several items were not anticipated. First, the use of Simperatives instead of interrogative structures for requests for information by the Japanese students, especially in the information gap activity, was not anticipated.

When researchers have not found structures where expected, it has sometimes been explained as avoidance (Schachter, 1974) or linguistic transfer/interference.

However, after the analysis began and these structures appeared for some of the students, explanations for their presence was investigated and it was established that the use of imperatives in Japanese is a common linguistic structure for requesting information, and needs further research. Thus, this could be called pragmatic interference. This created an imbalance in the number of anticipated Wh-questions, so the thesis focus shifted slightly. However, sufficient data were collected to complete the analysis of subject and topic, and no immediate problems caused by the pragmatic interference for the analysis of topic was ascertained.

A second problem concerns the tasks themselves. Several of the festivals were only small local village festivals in northern Japan and the students from Chiba and the Tokyo area were not always familiar with them. Although this provided a diverse set of festivals in terms of familiarity of the task items for the elicitation of speech by the students, it also raises questions as to the exact definition of shared cultural knowledge, since it is clear from the results that not all students were familiar with all of the festivals. The notion of shared cultural knowledge and its application to pedagogy thus needs to be clarified.

In addition, the question of exactly what does it mean to be a definite description addressed in Chapter 5, where I classified types of mutual knowledge suggests a need to more finely define proper names as part of definite descriptions.

Proper names are only a label used by linguists and language teachers. Sometimes they are familiar names or descriptions between interlocutors and sometimes they are not. Therefore, the distinction between shared versus mutual knowledge may apply to proper nouns, as well as common nouns. This unexpected outcome also did not seriously jeopardize the results, but suggests further research of the exact usage and definition of proper nouns in a culture.

Another limitation of corpus-based research surfaced, when I was not able to make clear generalizations outside the bounds of this study which was limited to native Japanese speakers. Where necessary, I elicited grammaticality judgments on certain propositions, produced by the native Japanese speakers in the experiment, in order to test the intuitions regarding certain constructions from native English speakers.

I9.2 Summary of Findings This thesis presented a methodology for the study of the semantic meaning of definiteness and specificity and the behavior of subjects and topics in interlanguage.

For an adequate description and explanation of the study of definiteness, the context of utterance and its connection to the referent in the wider context of both the discourse and the surrounding situation were shown to be required.

0 Chapter 2 described the experiment used for the thesis. A background description of the geographic location and curriculum design of the American school, as well as the immediate non-academic environment for the study, provided the cK1 setting of the learners, the experimental design, and the four tasks used to elicit data.

This information is useful for describing the context of the discourse knowledge, situation, and interactions of the conversations in the study, and was later incorporated i into the analyses of every chapter in the thesis. The fact that shared mutual knowledge (Ni is an integral part of the discussion in the conversational dyads added a dimension to Othe experiment which has not been included in other task-based experimentation in second language acquisition, essential for an adequate study of definiteness and specificity. The four tasks were described according to their usefulness for creating the context of aural discourse, visual, visible, and inferred referents. This is a new technique for the use of tasks describing interlanguage and proved useful not only for a detailed analysis of definiteness but also for the description of inferencing types discovered in the interlanguage.

Chapter 3 exemplified the sources of information constructed from mutual knowledge in the interlanguage discourse assisted by the new classification of the tasks. Using Clark and Marshall's (1981) analysis of definiteness, which distinguishes shared from mutual knowledge, shared knowledge is divided into "community and universal" knowledge, which includes generic and proper nouns. If both participants are privy to the community knowledge then it may become mutual knowledge.

Mutual knowledge is also derived from the ongoing discourse between participants. It can be divided into knowledge derived from the knowledge in the somato-sensory systemxxx and includes visual, aural, tactile, taste and smell in the discourse context.

My experiment and analyses were based primarily on definiteness derived from visible, visual and acoustic objects, although community knowledge of festivals also played a significant role. Objects found in acoustic or auditory-linguistic copresence were found in multiple grammatical categories: nouns, pronouns, adjectives, deictic locatives, and temporal deictics.

The chapter also exemplified differences between Japanese and English visual deictics represented in language. This difference is not a lexical difference but a processing difference, and suggests the visual knowledge in the attention and short term memory systems of the human brain must be acquired for Japanese learners of English with respect to visible deictics in language. A final type of definiteness found in the data is derived from inferred knowledge of the properties of festivals in community knowledge, the visual setting, and the conversation. The four tasks in the experiment served as the instruments which elicited the different types of mutual knowledge and definiteness, including inferencing, found in the propositions.

Throughout the chapter, the four general types of definiteness in the context of the common ground of the utterance were described according to nine mental resources. These were given simple non-technical definitions (see Korpi, 2005) for a more technical application), but the chart of mental resources itemized their use, N distinguishing different mental resources for creating each of the four general types of 0 definiteness, considered the unifying factor in a discourse.

(Chapter 4 demonstrated that definiteness from the sources of knowledge described in Chapter 3 form three patterns of meaning at the level of a single proposition. These patterns are created by specificity relations created by each proposition interacting within the context of the discourse and situation. This expands 00 on von Heusinger's assumption regarding specificity as 'sentence bound' (von Heusinger, 2002). Task IV, a picture description, has not been used before in second language acquisition research and was used to define each of the three patterns and Sillustrated their development over a three week period by one learner. This task was also useful as an instrument for demonstrating the difference between visible and visual antecedents as sources of knowledge in the context. Task III, the Information (Gap Activity, described certain structures usually coded "not acceptable" or "inappropriate" according to mixtures of the semantic definiteness structures, giving a Onew interpretation to these utterances, dependent on a level of semantic interpretation in interlanguage.

The three pattern types were also able to describe the communicative intentions of the entire discourse from the point of view of the two person dyad, especially as speaker and the hearer. Although it has been shown by (Yeom, 1998) that information states differ between a speaker and hearer, I showed that the patterns of meaning differed between the speaker and hearer in the common ground of the discourse. This chapter further developed the difference in each information state to show that the patterns of meaning for each individual proposition combine to form mental representations. which differ in the final information state. An empirical description of Task IV, demonstrated the differences in the information states of each person in the common ground of the discourse.

The information states in turn are conditioned by differences in the sources of knowledge in the context of the common ground. The common ground shared between two interlocutors did not always have a one to one mapping between semantic structure and the propositions entered into the discourse. Depending on the communicative intent of each of the two respective speakers, they can be constructing different mental representations of objects present in the context. The modifications to the Discourse Representation Structures represented this disparity at the macrolevel (overall) of the discourse.

Chapter 5 demonstrated that the three specificity patterns of meaning form semantic definiteness conceptual structures adding another level of representation in conceptual structure together with Jackendoff's lexical-semantic representations of (Jackendoff, 1990). The patterns of meaning from Chapter 4 are based on the decomposition of properties of semantic specificity and underlie all of the a-structure constructions studied by (Manning, 1996). Jackendoff's lexical semantic representations are based on the lexical decomposition of verbs. This chapter first demonstrated that thematic roles are restrictively neutralized in the patterns, so the three specificity patterns are argued to be autonomous units at the level of Conceptual Structure. The a-structure constructions were described according to Jackendoffs conceptual structures: action tiers and thematic tiers. Propositions and their arguments and complement structures were shown to be composed of one of the three patterns of specificity presented in Chapter 4.

The data analysis in this chapter was organized around the notion of argument structure proposed by Chris Manning for several reasons. First, it groups the properties of arguments according to the semantic properties of a-subjects which can IDbe boot-strapped and then tested against syntactic subjects. Manning's portrayal of O argument structure restricts itself to semantic properties (or processes). Secondly, Ssince certain verb classes and processes which directly involve subjects, imperatives, reflexives, control and complementation, 'in order to' clauses, have been studied only Swith respect to morphosyntactic properties, appealing to the semantic properties of their argument structure provides a new and more complete approach to describing 00 these "suppressed" subjects in interlanguage. Thirdly, Manning's analysis offered a complete set of a-structures, offering a simple organization for showing that the three semantic definiteness structures are part of a system including the proposition and embedded structures, such as complement clauses, at the conceptual level.

Organizing the analyses around the notion of argument structure lead to a C revision of previously accepted constraints on certain structures, e.g. imperatives. Not C-K only can imperative constructions use the role patient as pointed out by (Andrews, c- 1985), they can also take experiencers as subjects (unexpressed). This is contrary to Jackendoff's prediction that imperative constructions require agents as subjects. Thus, I showed that adding semantic features for an adequate characterisation of properties of agent in certain constructions is necessary, e.g. +vol] where the volitionality of the agent is a critical constraint on the grammaticality of the structure.

A final description of these constructions included the different lexical constraints of event, action and states, as defined by (Jackendoff, 1983). Included in all of these discussions, I described the a-structures with respect to their respective experimental Task, paving the way for IL.

Chapter 6 tested the interlanguage for the grammatical relation of subject.

Other interlanguage studies of subject have simply assumed that subject is defined according to word order, i.e. SVO or SOV is canonical and not further defined the S in SVO, without actually stating this or looking for other characteristics (Meisel, et al, 1981; Pienemann, 1998). For subject-predicate constructions, I first looked for evidence of subjecthood, such as subject ellipsis in complement clauses, conjunction reduction, and subject verb agreement, and case marking. Secondly, I showed the interaction of subject-predicate structures with the three semantic relations and patterns of meaning. Like the subject-object distinction, this required more than one entity within a clause to be examined to determine the specificity relations. Again, the three semantic structures and their patterns of meaning were shown to be independent of the level of syntax and grammatical relations.

Chapter 7 first reviewed the SLA literature on topic and then defined topic as a binary relation of given and new information at the level of information structure within the sentence. Since previous studies of SLA which have tested the Li and Thompson hierarchy have only looked at sentence position (initial) or morphosyntax (S-V agreement), if they addressed the presence and acquisition of topic, I examined one distinguishing property of subject with topic, as mentioned in the Li Thompson topic prominence typology, namely that topics are constrained to a [+specific] property. I then showed that three semantic definiteness structures and patterns of meaning also interact at the level of information structure, yet remain distinct systems in the grammar.

Chapter 8 briefly summarized Discourse Representation Theory (DRT), a well-known theory in computational linguistics, developed by (Kamp and Reyle, 1993) and proposed that definiteness of the propositional structure and the formalism which clarifies the labelling be based on Discourse Representation Theory. The chapter then argued that DRT representations (DRSs) are inadequate and proposed necessary modifications for its application to a study of an interlanguage corpus. First, 258 D modifications to the underlying assumptions of true in all possible worlds for computational linguistics and Artificial Intelligence (AI) research, and also the necessity for an anaphoric referent to be present in the world of the tasks or discourse was challenged. The first step was to select discourse fragments which represent inconsistent and contradictory truth conditions in the interlanguage discourse generated by the tasks and organize these in a truth-table template of inferencing 0 potentials. This was seen as a necessary revision to adequately account for the interlanguage discourse data generated from the four tasks in the experiment and to accommodate and explain the individual utterances as well as the discourse. For example, learners begin the picture description in Task IV without first presenting a referent into the discourse. Nonetheless, the dyadic partner can infer that 'they' refers 0 to people.

Secondly, modifications to the Discourse Representation Structures (DRS) were made to represent the empirical information from the context created by the four Otasks. Each DRS reflects the contribution of the persons interacting, either as speaker or hearer, rather than the usual assumption that only one person contributes to the production of an utterance. Each person was given an ID number which remained stable throughout the interaction, rather than alternating labels of Speaker or Hearer.

The next modification proposed was to represent the process of comprehension or production in each interaction of each person. The third modification to the DRS was to label the Time of each utterance entering the DRS. This gives a function for the three variables in a context of utterance: person, process, and time, proposed by (Clark and Carlson, 1981).

The DRS was next modified to represent the common ground created through the interactions in the context. Finally, the DRSs were modified to incorporate the sources of mutual knowledge in the common ground and types of definiteness contributing to the information in each utterance controlled by the four tasks of the experiment as investigated in Chapters 3 and 4. Thus, the DRSs were modified to become an empirically accountable representation of the developing discourse in the interlanguage. With the modifications to the DRS, I demonstrated the differences between visible and visual definiteness within a given discourse, interference from inferencing by both the learner and hearer, and ultimately the creation of definite descriptions line by line which culminate in a final identificational proposition at the level of specificity.

Finally, Discourse Representation Theory was used as a representational technique for the study of topics in discourse. Lexical Functional Grammar, a theory of parallel constraint-based syntax, was used for the organization of the grammar and representation of sentence level properties of subjects and topics. The thesis culminated in the connection of the sentence to the discourse where the interaction of the two theories is most useful. The combination of these two theories represented the formal properties for the interaction of the sentence level subject and topic to the discourse topic and definiteness.

9.3 implications and future directions After establishing the autonomous system of definiteness and specificity and the existence of three patterns of meaning inherent in three semantic structures for referring expressions at a conceptual semantic level in interlanguage, their interaction with lexical conceptual structures, and subject and topic at a linear level were briefly sketched. This brief sketch demonstrates that topic and subject are independent phenomena but tightly intertwined with definiteness and specificity in interlanguage.

259 IND In addition, the semantic definiteness structures of the sentence were shown to interact 0 directly with the discourse, without an intervening level. It is hypothesized that the semantic relations of definiteness and specificity of these structures in the Q^ interlanguage data are the same for natural languages with respect to their function Dand the sources of knowledge in the context of utterance.

V. Although this study was an empirical description of properties of definiteness 00 and specificity, a considerable overlap exists between Jackendoff's formulation semantic structures for the decomposition of verbs, creating Lexical Conceptual Structures and Conceptual Structure. I avoided addressing the problem of the CN' difference between the semantic definiteness structures and their patterns of meaning at a conceptual level, semantic level and pragmatic level. However, this thesis has cr nonetheless provided an empirical description of certain of Jackendoff's properties for Sa proper theory of conceptual structure and semantic structure its interface with pragmatic structure.

At the intersection of the syntax, predicate argument structure and pragmatics, the patterns of meaning at the level of conceptual structure, formulated according to the three semantic relations of definiteness and specificity, are indistinguishable. The semantic definiteness structures of Existence, Characterization, and Identification are found both at the sentence level and discourse level. In fact, the definiteness relations are based on entailment relations of definiteness and specificity, according to their situation in the context of the discourse. This in turn creates a superordinalesubordinate relation, with the semantic relations at the level of specificity and the sentence subordinate to discourse knowledge.

As the utterance is being entered into the discourse, the nine mental resources outlined in Table 3.2, DEFINITIONS of MENTAL RESOURCES for ASSIGNING DEFINITENESS in a CONTEXT, create different configurations than the linear relations required by the analogue system of speech. These resources, community comembership, universality of knowledge simultaneity, attention, rationality, locatability, recallability, understandability, and, finally, associativity need further investigation before we can understand how they operate in the creation of definiteness in the local context of the utterance and the global level of the discourse.

Initial research (Korpi, 2005) has shown that they hold promise for explaining the syntactic-o-pragmatic development of sentence initial structures in interlanguage, such as the 'displaced cxx' elements of adverbs, wh-questions, and other topicalizing structures, important for establishing Stage 3 of PT (Pienemann, 1998; 2003) and the Multi-Dimensional Model (Meisel, et al (1981). Especially 'recallability', or 'shortterm memory', and 'associativity' appear to require achieving a threshold level before the learner can manipulate the linearization constraints of the sentence in order to establish a connection with preceding sentences in the discourse, using structures such as topicalization. These mental resources are clearly part of conceptual structure controlling semantic structure, intimately involved in semantic definiteness structures.

This thesis then tentatively supports Jackendoff's approach which argues against a separate level of pragmatics, and in which "linguistic inference is but a special case of more general modality-independent principles,... in which semantic structure is subsumed under conceptual structure" (Jackendoff, 1983:19). The empirical description of the semantic definiteness relations of the utterances provided in Chapters 3, 4, and 5 was done without two levels of conceptual structure and semantic structures, suggesting that these are indistinguishable and can be collapsed into one unified level. Thus, the null hypothesis of this thesis would be that there is 260

ID

t0 0

(N

no difference between the levels of conceptual and semantic structure. The implications of this with respect to the need for an intermediate level of correspondence rules needs to be clearly spelled out. This findings of this thesis would suggest they are not necessary.

However, it is not entirely consistent with his concept of conceptual structure.

His use of the notion 'Representational', or REP, to connect the linguistic utterance to the visual modality have been replaced by the notions of "Definiteness" and the entailed, subordinate relation, of 'specificity'. These structures were described at the level of conceptual semantic structure and described according to nine cognitive resources. Most importantly, these mental resources and the example utterances have been shown to account for both linguistic and non-linguistic definiteness across modalities, i.e. visual, visible, auditory, and inferential. Each of the levels is again reviewed below, using one sample utterance with its corresponding properties of each system of language.

LEVELS OF GRAMMAR Syntactic Structure: Subject: Grammatical Relations (Word Order) Pro He Subj

S

V- NP Det N is a syntactician Predicate (V-Obj) V (0) Pragmatic structure: Sentence intersects the discourse Topic (Information structure) Semantic (Conceptual) Structure: Lexical Semantic Structures: Semantic Aspectual Types Semantic Roles Semantic Definiteness Structures: Semantic Relations Type Definiteness Properties Topic Given Focus/New Stative (Es) Theme Theme Characterizational Def/(+spec) Indef(-spec) MENTAL RESOURCES for ASSIGNING DEFINITENESS in a CONTEXT: community comembership, universality of knowledge simultaneity, attention, rationality, locatability, recallability, understandability, and associativity \D0 It is at the level of Semantic definiteness that visible, visual, inferential properties 0 interface with the auditory level of speech and its associated contextual properties of the speech situation. Thus, the use of definiteness and specificity for establishing a e( level of referring expressions and referentiality together with Jackendoffs notion of Lexical Conceptual Structure is a modification of the conceptual structures of both V) Levelt (1989) and Jackendoff (1983, 1990).

00 In conclusion, the three patterns of meaning are hypothesized to be an alternative solution to Montague grammar for the Fregean principle, i.e. "The meaning of a complex expression should be a function of the meaning of its parts" CI (Allwood, Andersson, et al. 1971). Chapter 4 was an empirical description of the learner auditorally constructing a mental representation of an Identificational relation C using existential and Characterizational relations perceived visually by the ,I interviewer. Finally, the semantic definiteness structures found in this interlanguage data are hypothesized to be the same for natural languages with respect to definiteness C0 and specificity and sources of knowledge in a context of utterance.

APPENDIX I A summary of the experiment for the purpose of this study is given Sbelow. The experiment is divided into two parts, Experiment I and Experiment II.

00 SExperiment I Experiment II 1. Japanese Male Teacher 1. American Male Teacher 2. Japanese Female Student 2. Japanese Male Teacher 3. Japanese Male Student 3. American Male Student 4. Japanese Female Teacher 4. Japanese Male Student N 5. American Male Teacher 5. American Female Teacher N 6. American Female Student 6. Japanese Female Teacher O 7. American Male Student 7. American Female Student 8. American Female Teacher 8. Japanese Female Student Notice the order of appearance differs between the two experiments. Experiment I had Japanese interviewers first, followed by American interviewers. Within each subgroup, males alternated with females. Experiment II had male interviewers first, followed by female interviewers. Within each subgroup, Americans alternated with Japanese interviewers. Interviewers were each coded with I followed by their order of appearance in the experiments, e.g. ii.

Interviewers Sex and Ethnicity (both experiments): NNS-Japanese Male 4 NS-American Male 4 NNS-Japanese Female 4 NS-American Female 4 Interviewers Status/Age (both experiments): NNS-Japanese (JSL) Teachers 4 NS-American Teachers (GE) 4 NNS-Japanese students (ESL) 4 NS-American Students (JSL) 4 Both groups include an equal number of teachers and students. The American teachers are from the General Education program and the Japanese teachers are from the Japan Area Studies Department and teach Japanese as a Second Language (JSL).

Students (both experiments, all NS Japanese): Experiment I Experiment II 8 students 1 on some interviews) 4 students 2 on some interviews) The students were from two different classes. The eight students in Experiment I were in the module for ten weeks. The four students in Experiment II were in their first five weeks of the module. In each experiment a couple of students began the module and experiment but left before the experiment and their class finished. Hence, the extra students. Students were each coded with s, followed by their order of first appearance in the experiments, e.g sl. The ID s for students was selected in order to clearly distinguish I for learner from i for interviewer. Experiment I had students coded as sl s8, and Experiment II students were coded from s9 s13.

00

(N

Totals in both Experiments 8x 8 64 dyads 8x 4 32 dyads 96 dyads 12 Students 16 Interviewers 96 dyads x 15 minutes 24 hours of speech 64 32 96 dyads

Claims (4)

1. A method of computational linguistic analysis, said method comprising the Ssteps of: selecting a passage of text to be analysed, said text containing at least 00 one sentence; (ii) for each word allocating the word to one of four categories comprising: community membership visual linguistic co-presence auditory linguistic co-presence, and ND indirect co-presence; (iii) for each allocated word allocating a further sub-category comprising: "plus" if the word is already presented in the text or can be inferred, and "minus" if the word is not already presented in the text and is not able to be inferred; and (iv) generating a relationship expressed in the relations of members in sets between all the allocated words.

2. The method as claimed in claim 1 wherein said category comprises two sub-classes of a visible linguistic co-presence and a visualized linguistic co- presence.

3. The method as claimed in claim 1 or 2 including the further step of: extending said word allocation and set relationship generation to adjacent sentences in said text.

4. The method as claimed in claim 3 including the further step of: (vi) extending said word allocation and set relationship generation to substantially all said text. A method of linguistic analysis substantially as herein described with reference to either Annexure 1 or Annexure 2. 265