Situated Simulation in the Human Conceptual System

1

Friday, 9:00 – Maple Leaf Room PLENARY ADDRESS

Situated Simulation in the Human Conceptual System

Lawrence W. BarsalouDepartment of Psychology

Emory University

The human conceptual system contains categorical knowledge about the world that supportsonline processing (perception, categorization, inference, action) and offline processing (memory,language, thought). Semantic memory, the classic view of the conceptual system, typicallyportrays it as modular, amodal, abstractive, and static. Alternatively, the conceptual system canbe viewed as non-modular, modal, situated, and dynamic. According to this perspective, theconceptual system is non-modular and modal because it shares representational mechanisms withmodality-specific systems in the brain. On a given occasion, modality-specific information abouta category's members is reenacted in relevant modality-specific systems to represent it.Furthermore these simulations are situated, preparing the conceptualizer for situated action withthe category. Not only do these situated simulations represent the target category, they alsorepresent background settings, actions, and mental perspectives, thereby placing theconceptualizer in the simulation, prepared for goal pursuit. Because the optimalconceptualization of a category varies across different courses of situated action, categoryrepresentations are dynamic, not static. Behavioral and neural evidence is presented to supportthis view of the conceptual system.

2

Saturday, 9:00 – Maple Leaf Room PLENARY ADDRESS

The Role of Attention in Discourse, Cognition, & Syntax:Dynamic Processes Require Dynamic Theory

Russell S. TomlinDepartment of Linguistics

University of Oregon

Among the central problems of linguistic inquiry is the problem of use, of specifying theconditions under which semantically equivalent expressions are formed or selected as speakersengage in real time discourse. Increasingly, linguists appeal to a role for cognition in capturinghow linguistic structures are deployed in discourse, with attention of particular interest. Thispaper discusses the place of attention in understanding, theoretically and empirically, howdynamic conceptual representations get mapped into linguistic structures during discourseproduction.

3

Sunday, 9:00 – Maple Leaf Room PLENARY ADDRESS

Complex Events, Predicate Overlayand the Special Status of Reciprocal Clauses

Nicholas EvansDepartment of Linguistics & Applied Linguistics

University of Melbourne

The lexicalization of event-denoting expressions has emerged as the area of semantics with themost extreme cross-linguistic variation. This makes the mapping between the semantic level of‘event’ and the syntactic level of ‘clause’ one of the greatest challenges to semantics, syntax andtypology. One dimension of this problem is working out how much complexity, in terms of eventstructure, can be accommodated within a single clause. Cross-linguistic studies of causatives,benefactive and instrumental constructions, and motion events have showed us that single-clauseEnglish expressions like I dropped the cup, I baked Mary a cake, I cut the bread with a knife andThe ball rolled down the hill are all syntactically decomposable in some languages into multi-clause expressions of the type I made/let the cup fall, I baked cake gave Mary and The balldescended the hill, rolling. A wide range of languages employ such strategies, through methodslike syntactic causatives, serial verb constructions or manner-framing through participial or othermeans; typological work here has taught us a great deal about how to motivate complex syntacticbehaviour from the semantics of event decomposition.

Reciprocal constructions are a further type of complex event that maps to a single clause inEnglish, as in many other languages. Their semantic characterisation requires two propositions:the meaning of John and Mary kissed each other will be represented, logically, by the conjoinedpropositions John kissed Mary, and Mary kissed John, with symmetric exchange of referentsbetween argument roles. Yet little work on the syntax of reciprocals has sought to ground theunusual syntactic features of these clauses in their complex event structure, in a way that relatesthese to other typological work on causatives, benefactives and motion-event-decomposition in‘event-atomizing languages’, even though a large number of languages employ serial verbconstructions for reciprocals just as they do for other types of complex event. For example inGolin, a Chimbu language of the Papuan highlands, the same overall strategy that producesmulti-verb representations for benefactives or motion events does the same for reciprocalsituations, by reporting them as complex symmetrical pairings of subevents.

In the cross-linguistic study that I report on in this paper, one finds widespreadperturbations of clause structure in reciprocal clauses, essentially resulting from an ‘overlay’ ofsymmetrically mirrored events, that include apparently contradictory signs of transitivity,apparent violations of principles producing a one-to-one mapping of NPs to thematic role,‘sesquiclausal’ constructions that appear to hover between one and two clauses in size, andviolations of binding conditions through the use of non-anaphor pronouns in the same clause astheir antecedent.

Effects like these raise the question of how many features claimed to be universals of clausestructure (such as biunique mapping between thematic roles and syntactic argument positions)are really syntactically motivated. A fuller typological stocktake of reciprocal constructionssuggests that, instead, they may be epiphenomena of limits to how much event complexity maybe mapped onto a single clause; when, as happens with reciprocals in some languages, enoughthe clause exceeds a certain level of complexity, these syntactic constraints no longer hold.

4

Sunday, 11:00 – Prairie Room MIDDLES & IMPERSONALS

A Local Account of the Unaccusative/Unergative Distinction in French

Michel AchardRice University

According to Perlmutter’s (1978) Unaccusative Hypothesis, intransitives can be separated into‘unaccusative’ and ‘unergative’ predicates on the basis of their syntactic behavior in specific testconstructions (see also Legendre 1989a). This difference in behavior is imputed to their initial syntacticstructure. Unaccusatives have an initial object (2) and no subject, whereas unergatives have an initialsubject (1). Current accounts of the unaccusative/unergative distinction all share a common ‘global’character, namely the isolation of a characteristic (or a set of characteristics) common to all unaccusativepredicates, which distinguishes them from the unergatives and explains their similar syntactic behavioracross several constructions.

This paper argues that such a global account is untenable for French because i) theunaccusative/unergative distinction receives very little structural motivation, and ii) analyses based onthat distinction wrongly predict the distribution of intransitive predicates in specific constructions. Thesyntactic behavior and distribution of intransitives is best explained by invoking the necessarycompatibility between the meaning of specific predicates and that of the constructions in which theyparticipate. This kind of treatment is called ‘local’ because it relies on the meaning of predicates in thespecific context of each construction they appear in, rather than on more global (semantic or syntactic)characteristics that they share as a class, and that operate across a whole range of constructions.

As a representative example, a ‘local’ analysis provides a more adequate picture of the distribution ofintransitive predicates with active impersonal constructions (Legendre 1989b) than an account based onthe unaccusative/unergative distinction. Unaccusatives are felicitous with impersonals, but unergatives aresubject to the “indefinite subject constraint” illustrated in (1):

(1) Il mange beaucoup de travailleurs (*Paul) dans ce restaurant‘There eat many workers (*Paul) in this restaurant’

Careful examination of corpus data shows that a global is untenable because theunaccusative/unergative distinction does not exactly correlate with the distribution of intransitives withimpersonals. The verbs of internal change are structurally unaccusative, but they behave like unergatives(i.e. they are subject to the indefinite subject constraint) in the impersonal context. Conversely, an accountbased on the semantic compatibility of the main predicate and the impersonal construction shows that therelevant factor in the distribution of predicates with impersonals is the level of salience of the setting inthe lexical semantic structure of the predicate. The more felicitous predicates in impersonal constructionsare those that most relevantly and naturally include the setting as part of their scope of predication(Langacker 1991). The benefits of this solution are twofold: i) it provides the right distribution ofintransitive predicates, and ii) it explains the presence of the indefinite subject constraint with bothunergative and change of state predicates.

References

Langacker, Ronald W. 1991. Foundations of Cognitive Grammar. Vol. 2: Descriptive Application. Stanford:Stanford University Press.

Legendre, Géraldine. 1989a. Unaccusativity in French. Lingua 79, 95-164.Legendre, Géraldine. 1989b. French impersonal constructions. NLLT: 81-127.Perlmutter, David. 1978. Impersonal passives and the unaccusative hypothesis. BLS 4: 157-189.

5

Friday, 10:30 – Aurora Room SEMANTIC MEMORY

Comparison of Feature Generation and Feature Rating Tasksas a Probe into the Cognitive Systems Underlying Semantic Memory

Carrie A. Ankerstein, Patricia E. Cowell, Rosemary A. VarleyUniversity of Sheffield

The phenomenon of category-specific semantic impairments has been the inspiration for numerousstudies investigating the nature of semantic memory. Many of the theories and models resulting fromthese studies use data from feature generation tasks to illustrate their claims, e.g. there are differentfeature profiles for different categories of objects. The idea behind the feature generation task is that theparticipant calls up a multi-sensory image of the object and reads off the features of that object andtherefore, this task provides a window into semantic memory (Cree & McRae, 2003). However, thiswidely held assumption has not been adequately tested.

Two experiments were conducted to address the following questions: (1) Is the feature generationtask a valid measure of semantic memory? and (2) Are there different feature type profiles for differentcategories of objects?

In order to address the first question, a traditional feature generation task (Experiment 1) wascompared to another task designed to tap into semantic memory: a sensory-motor rating task (Experiment2), which was based on a task from Simmons et al. (in press). In the feature generation task, 60participants were asked to describe 38 objects using short descriptive words or phrases. In the sensory-motor rating task, participants were asked to rate their knowledge for various modalities and objects. Thescale went from 1 to 5, where 1 corresponded to an irrelevant modality for the object, e.g. tasting a bus,and 5 corresponded to a highly relevant modality, e.g. tasting a strawberry. It was hypothesized that if thefeature generation task is a valid measure of semantic memory, then the results of both tasks should besimilar. In Experiment 1, the most commonly given features were visual descriptors and encyclopedicfacts, making up 74% of the total features given. In Experiment 2, all sensory-motor modalities wererated as highly relevant for at least one category of objects. These data are incompatible with theassumption that participants read features from a multi-sensory image. Indeed, the results of Experiments1 and 2 are only partially overlapping.

The second question addresses the idea of different feature type distributions for differentcategories of objects. There was a partial overlap of feature type importance in Experiments 1 and 2. Datafrom both experiments showed high levels of importance for visual information across all categories andfor tactile information for the fruit/vegetable and tool categories.

It is recommended that the feature generation task be supplemented by another task which doesnot require the explicit verbalization of individual features of an object because a conscious retrieval tasklike the feature generation task does not adequately reflect the structure of implicit semantic memory. It isalso proposed that the traditional labels of “visual” and “tactile” may be too broad, with visualencompassing color, motion and shape and tactile encompassing form, texture and manipulability, toadequately capture the differences between different categories of objects. Novel implicit memoryexperiments are being developed to further explore the types of visual and tactile information for variouscategories of objects.

References

Cree, G. S., & McRae, K. (2003). Analyzing the factors underlying the structure and computation of the meaning ofchipmunk, cherry, chisel, cheese, and cello (and many other such concrete nouns). Journal of ExperimentalPsychology-General, 132(2), 163-201.

Simmons, W. K., Pecher, D., Hamann, S. B., Zeelenberg, R., & Barsalou, L. W. (in press). fMRI evidence formodality-specific processing of conceptual knowledge on six modalities.

6


Effects of Lexical Structure on Picture Comparison

Harald Baayen & Lera BoroditskyMax Planck Institute for Psycholinguistics and Stanford University

To what extent might the contents of a speaker’s mental lexicon and the structural relations between theentries in the mental lexicon affect thought? We addressed this Whorffian question experimentally bymeans of a picture comparison task, in which participants were asked to rate, on a nine-point scale, howsimilar they judged two pictures presented side by side on a computer screen to be. There were 10 sets of6 pairs of pictures. Each group of six consisted of 3 pairs involving three pictures with names that weremorphologically related in Dutch but not in English (e.g., sink/trashcan, ashtray/trashcan, sink/ashtray).The remaining three pairs also used three pictures, but their names were morphologically unrelated inboth English and Dutch (e.g, bathtub/bowl, bathtub/wateringcan, bowl/wateringcan). We ran threeexperiments, one with native speakers of English, one with native speakers of Dutch (with instructions inDutch), and one with native speakers of Dutch with a good command of English (with instructions inEnglish).

We fitted a multilevel covariance model to the data of each experiment, with as covariates the frequenciesof the nouns for the two pictures, as well as their morphological family sizes and their number ofsynonym sets (synsets) in WordNet. For the English participants, there was no main effect ofmorphological relatedness in Dutch. The morphological family sizes of the nouns for the two pictures andthe frequency of the noun for the second picture emerged as significant predictors of perceived similarity.A larger family size and a greater frequency for the picture presented on the right led to lower ratings.

For the Dutch participants, we observed a significant main effect of morphological relatedness. Inaddition, similarity ratings decreased with log family size and the number of synsets of the noun for thefirst picture, as well as with the frequency of the noun for the second picture. Furthermore, similarityratings dropped when the noun for the second picture had many synsets, but only in the related condition.

For the Dutch participants doing the task in English, we again observed a main effect of relatedness,irrespective of whether we used English or Dutch counts as covariates. Since the model using the Englishcounts yielded a slightly better fit, the effect of relatedness seems to persist even when Dutch speakers arethinking in English.

These results support a Whorffian view of the relation between language and thought. English speakerscomparing pictures of a PRAM and a TRUCK may notice that both objects have wheels and are used fortransportation, but Dutch speakers have this shared functionality hardwired into their language, whichclassifies both objects as WAGONs (KINDER-WAGEN, ‘child-wagon’, and VRACHT-WAGEN,‘freight-wagon’), shaping the way they compare objects, even when doing the task in another language.

Finally, while the predictivity of frequency can be ascribed to language-independent conceptualfamiliarity, the predictivity of the family size measure shows that how concepts are networkedmorphologically in the mental lexicon co-determines perceived similarity.

7

Saturday, 14:30 – Aurora Room DISCOURSE PROCESSING

Exploring Attention, Memory, and Cognition in Ordinary Language Use

Marjorie BarkerUniversity of Oregon

What people ‘do’ cognitively with coherent discourse is difficult to track experimentally, because thephenomenon involves both a multifaceted stimulus and concurrent processing at several levels ofmeaning. Because of this, language processing studies have often focused on word lists and individualsentences. Such studies, though yielding important insights, are not fully representative of ordinarylanguage use. The research described here was an attempt to deal experimentally with the kind oflinguistic complexity that humans manage to make sense of every day.

On the theory that divided attention would compromise processing, subjects were asked to listen tostories under various conditions of distraction, and then to recall what they had heard. Recall for coherentspoken narrative was used as a measure of memory.

Attentional stress was operationalized by adding time pressure (speeding the stimulus to 1.5 timesoriginal rate) or requiring subjects to perform an additional verbal or visual task while listening. Anotherexperimental condition inserted 3-second pauses between clauses to test the effect of extra processingtime. Combinations of these treatments were also tested.

110 participants each listened to five brief stories and recalled the stories orally under instructionsto do so as completely and as closely to the original stimulus as possible. Recall transcripts were analyzedto determine both the amount of information recalled and the linguistic form of the recall. Results showthat distraction detrimentally affects the amount of material remembered, and that insertion of pauses inthe stories counters the effect of added tasks. However, neither attentional demand nor extra processingtime had a significant effect on reconstruction of the original linguistic structure in recall.

The linguistic data are extremely rich, allowing comparison of recall for each story across over ahundred different subjects. Results indicate that subjects constructed grammar while retelling the storiesrather than remembering the grammar of the original stimulus. Characteristics of the recall grammar mayprovide insight into the nature of mental representation. For example, given the original stimulus

‘and it so happened that he was carrying his cell phone in his pocket and while he was braggingto his buddy he inadvertently pressed one of the buttons’

where the context allows the subject of the sentence (grammatically active in the stimulus) to beconstrued as nonagentive, subjects often produced passive, middle-voice, and impersonal constructionslike the following in recall:

uh, somehow the on button was pushedhis cell phone, got turned on; and a button got pushedand it accidentally, pressed and, something accidentally pressed the buttonmeanwhile his cell phone in his pocket pushed the wrong button

Analyses of semantic and grammatical recall, including statistical comparisons, provide insightinto the cognitive resources for everyday linguistic processing.

8

Saturday, 17:30 – Aurora Room FREQUENCY EFFECTS

Usage and Frequency: Input vs. Output and Implications for Grammatical Analysis

Michael Barlow and Suzanne KemmerUniversity of Auckland and Rice University

The effects of frequency on the structuring of grammar have been widely discussed in research employing usage-based models of language (Langacker 1988, Barlow & Kemmer 2000, Bybee and Hopper 2001, Tomasello 2003). Itis argued that frequency of experienced language affects the entrenchment of patterns in the language users’grammar. On the basis of this postulated connection between usage and the form of grammar, frequency data onparticular lexico-grammatical patterns extracted from large linguistic corpora are used as a source of evidence forparticular hypotheses about the structuring of language users’ grammar.

For instance, a search for the most common three-word chunks containing thing in a corpus of spoken,professional, American English yields, in decreasing frequency order, the following phrases: sort of thing, the thingis, the only thing, the same thing, the other thing. By analyzing the data in terms of relative frequencies, it is possibleto formulate a usage-based account of the grammatical patterns involving the word thing, or of any other linguisticexpression or construction (Barlow 1996). This methodology is widespread in corpus linguistics (Biber 2001) and inCognitive Linguistics (Gries and Stefanowitsch, In Press).

However, such analyses are based on an amalgamation of language produced by different speakers, ratherthan the usage of an individual in whose mind a grammar resides. To the extent that the corpus is representative ofwhat speakers hear, corpus analyses of this sort can be taken to represent a model of general input. Little or noattention has been paid in the literature to the question of the relation between such analyses and the individual’sgrammatical knowledge. In the simplest kind of usage-based model, a speaker’s productions should more or lessmirror the patterns found in the input. High frequency in input data should generally result in high output frequency.This assumption implicitly underlies many corpus studies, as researchers use frequency information from generalcorpora to draw conclusions about ‘the grammar’.

The current study demonstrates that, contrary to this assumption, there are marked differences betweengeneral corpus frequency patterns, and patterns in the productions of individual speakers in the same corpus.Looking at thing, for example, we find that Speaker A uses the forms the thing that, that's the thing, thing that we, athing that, type of thing with the greatest frequency (compared to other expressions with thing). Speaker B, incontrast, prefers the following set: the other thing, other thing is, thing is that, thing that I, other thing that. Theresulting analyses of the grammar of thing, the composite analysis (representing input data), and the analysis ofindividual production data (representing an output grammar) are consequently different. We use corpora toinvestigate two other linguistic frequency patterns, the verbs remember and think with their associated grammaticalconstructions, and demonstrate a similar divergence between the patterns of the general corpus and those displayedin the output of individuals producing language in the same corpus.

Our findings generally support the usage-based conception of grammar stressed in cognitive and functionallinguistics. Yet they also demonstrate that a more refined model is necessary than usually assumed, specifically, amodel taking into account differences between two interlocking systems: a grammar of input and a grammar oflanguage output. On the theoretical side, we discuss some implications and subtleties of this more complex usage-based model for corpus and cognitive linguistics. On the empirical side, our contribution highlights the unexpectedlylarge divergence between general input frequency and speaker output frequency, and opens the way to a furtherexploration of speaker-specific patterns that can shed light on the way individual grammars are constructed.

References

Barlow, Michael. 1996. Corpora for theory and practice. International Journal of Corpus Linguistics 1.Barlow, Michael, and Suzanne Kemmer, eds. 2000. Usage-Based Models of Language. Stanford: CSLI.Biber, Douglas. 2001. Using corpus-based methods to investigate grammar and use: Some case studies on the use of verbs in English. Corpus

Linguistics in North America, ed. By Rita Simpson and John Swales. Ann Arbor: U. Michigan Press.Bybee, Joan, and Paul Hopper. 2001. Frequency and the Emergence of Language Structure. Amsterdam: Benjamins.Gries, Stefan, and Anatol Stefanowitsch. In Press. In Language, Culture and Mind, ed. by Michel Achard and Suzanne Kemmer. Stanford: CSLI.Langacker, Ronald. 1988. A Usage-Based Model. Topics in Cognitive Linguistics, ed. by B. Rudzka-Ostyn. Amsterdam: Benjamins.Tomasello, Michael. 2003. Constructing a Language: A Usage-based Theory of Language Acquisition. Cambridge, Mass.: Harvard University

Press.

9

Sunday, 14:00 – Prairie Room COGNITIVE GRAMMAR &CONCEPTUAL STRUCTURE

Prototypical Conceptual Types and Typological Variation in Parts of Speech Systems

David BeckUniversity of Alberta

Researchers interested in the semantic criteria for parts of speech classifications (e.g. Lyons 1977; Langacker 1987;Croft 1991, 2000) have generally observed that a) the meanings belonging to particular lexical classes are not thesame in all languages, and b) parts of speech have a robust common core and show variation on their “peripheries”.Such situations are commonly modeled in terms of prototype effects (Rosch 1978). Most research in this genre hasfocused on the properties of prototypical members of lexical classes. Broad-based, cross-linguistic comparison ofnon-prototypical members of lexical classes reveals that many of these fall into one of two types: 1) meanings thatshare features with prototypical members of more than one lexical class, and 2) meanings that have intermediatevalues for features that strongly differentiate two classes. Prototype effects seem to account not only for the fact ofvariability in lexical classification, but also for which meanings are variable and the ways in which they can vary.

Type 1 non-prototypicality is illustrated by HUMAN CHARACTERISTIC (HC) terms (Beck 2002), words thatdenote properties unique to humans. Like adjectives (Figure 1), these expressions profile an atemporal relation, butlike nouns referring to people they specify a human trajector (Figure 2):

young old

Figure 1. An atemporal relation ‘old’

young old

Figure 2. The HUMAN AGE term ‘old’

As a result, HCs show both intra- and cross-linguistic variation between adjective and noun. In English, HCs areadjectives, while in Upper Necaxa Totonac they are basically nouns, but may be used as modifiers of nounsreferring to humans and animals. In Spanish, HCs are used extensively as both adjectives and nouns referring topeople.

Type 2 non-prototypicality is illustrated by kinship terms and bodyparts, which show an intermediate degreeof conceptual autonomy, a property which distinguishes nouns and verbs (Langacker 1987). Like nouns, these referto discrete objects or THINGs, but like relational expressions, they include some other THING in their semantic base.As a result, kinship terms can be either nouns or verbs (1a), while bodyparts may do double-duty as relational,preposition-like expressions (1b):

Ilgar (Evans 2000: 117)(1a) ngan–nga–wulang

1SG.OBJ–3SG:FEM:ERG–be.mother.to:NON-PAST‘my mother’ (lit. ‘she’s my mother’)

Upper Necaxa Totonac(1b) na–k–wi:lí: x–pé'hxtún mé:sa

FUT–1SG–put 3PO–shoulder table‘I’ll put it next to the table’

Even in languages where kinship terms and bodyparts are nouns, they are frequently atypical nouns and showpatterns of inalienable or inherent possession, both of which reflect reduced degrees of conceptual autonomy.

The data presented in this paper show not only how prototype effects account for the fact that there is variationin parts of speech systems, but also demonstrate a) ways in which prototype models can identify areas where wemight expect to find variation and b) what form this variation can be expected to take. These are interesting results,and application of this methodology to other areas of typological variation in parts of speech systems is a promisingavenue for future investigation.

References

Beck, D. 2002. The typology of parts of speech systems: The markedness of adjectives. New York: Routledge.Croft, W. 1991. Syntactic categories & grammatical relations: The cognitive organization of information. Chicago: University of Chicago.–––––. 2000. Parts of speech as language universals and as language-particular categories. In P.M. Vogel & B. Comrie, Approaches to the

typology of word classes, 65 – 102. Berlin: Mouton.Evans, N. 2000. Kinship verbs. In P.M. Vogel & B. Comrie, Approaches to the typology of word classes, 103 – 72. Berlin: Mouton.Langacker, R.W. 1987. Nouns and verbs. Language 63, 53 – 94.Lyons, J. 1977. Semantics. Cambridge: Cambridge University Press.Rosch, E. 1978. Principles of categorization. In E. Rosch & B. Lloyd, Cognition and categorization, 27 – 48. Hillsdale: Erlbaum.

10

Saturday, 10:30 – Aurora Room SITUATED/SIMULATED SEMANTICS

Experimental Evidence for Simulation Semantics

Benjamin BergenUniversity of Hawaii, Manoa

Embodied approaches to language view meaning as fundamentally based on real-world perceptual andmotor knowledge that an individual accrues through interactions with the world (Lakoff 1987, Langacker1987). A unique aspect of this view is that understanding language results in internally imaging, orsimulating, the content of the utterance (Bergen et al. 2004). Early computational work (Narayanan 1997)defined ways in which such simulations could be learned, represented, and used for natural languageunderstanding. But until recently, little experimental work had evaluated the psychological reality ofmental simulation during language processing. This paper presents results from a set of studies,demonstrating such a role for mental simulation.

The experimental work was based on by Richardson et al. (2003), and tested whether processingdifferent sorts of spatial language makes use of corresponding parts of the visual system. Sentencesencoding downwards motion like The bottle fell contrast with others denoting upwards motion, like Theant climbed. If language understanders perform modal mental simulations in order to understandsentences like these, then this entails activation of the corresponding areas of their visual fields. In thiscase, hearing sentences encoding upwards or downwards motion should selectively interfere with usingthe visual system for other purposes, since it cannot efficiently process multiple similar images (real andimagined) at the same time. An object categorization task tested this idea. Subjects first heard an up ordown sentence, after which an object of varying shape appeared in the upper, lower, right, or left quadrantof a computer screen. Subjects categorized the shape of the object as quickly as possible (circle orsquare). Four types of intransitive sentences were tested: literal up or down sentences (like those above),metaphorical sentences using the same verbs (like The cost fell, and The temperature climbed), abstractdescriptions of quantity change (like The cost increased versus The temperature decreased) and finallysentences in which the subject was strongly associated with upness or downness (like The grass glistenedversus The sky darkened). We hypothesized that after up sentences (compared with down sentences),subjects would take longer to subsequently categorize an object in the upper part of the visual field, andvice versa.

Results showed that both types of concrete sentences (the ones encoding upwards versus downwardsmotion and the ones containing an up- or down-related subject noun phrase) yielded a significantinterference effect on responses in the categorization task. The abstract sentences yielded no such effect,and there was a marginal effect in the case of the metaphorical sentences. This suggests that mentalsimulation is important to literal motion sentence comprehension, but may be less so for metaphorical andabstract language. More broadly, the results support the psychological reality of mental simulation duringlanguage understanding.

References

Bergen, B., N. Chang, and S. Narayan. 2004. Simulated Action in an Embodied Construction Grammar. Proceedings CognitiveScience.

Lakoff, G.. 1987. Women, fire, and dangerous things. Chicago: University of Chicago Press.Langacker, R. 1987. Foundations of Cognitive Grammar: Stanford: Stanford University Press.Narayanan, S. (1997). KARMA: Knowledge-based active representations for metaphor and aspect. Ph.D. Diss., UC BerkeleyRichardson, D. C., Spivey, M. J., McRae, K., & Barsalou, L. W. 2003. Spatial representations activated during real-time

comprehension of verbs. Cognitive Science.

11

Friday, 13:30 – Prairie Room TAM & EVIDENTIALITY

A Contrastive Analysis of English and Lingala Verb Paradigms:Present-Time Reference and General Validity

Frank Brisard & Michael MeeuwisUniversity of Antwerp and University of Ghent

The study of tense-aspect systems in the Bantu languages has, until recently, suffered from elementary deficienciesin the empirical classification of verb-related categories, as well as in its theoretical foundation. Even now, with theavailability of functionally motivated (and cognitively informed) descriptions of Bantu grammars (see, e.g., Heine &Nurse 2000, Nurse & Philippson 2003), there is still a relative lack of agreement over basic issues in the analysis ofdifferent inflectional classes, especially at the level of their implied semantics. This rather acute problem wasaddressed in Meeuwis (1995, 1998, 2001), together with some proposals for remedying at least the terminologicalambiguities in this domain.

In this paper, we compare strategies of temporal predication in English and Lingala, a Bantu language spokenin Central Africa but also used as a lingua franca throughout large parts of Africa and in the African diaspora. Weshall challenge the received interpretation of a verbal suffix, -í, which is traditionally seen as marking a presentperfect or plain pastness, as in (1):

1. nasálí ‘I have worked/done/made’ [present perfect]

However, as is evident from example (2), this cannot be the whole story, since imperfectives sanction astraightforward present reading:

2. a) nalingí ‘I like/love’; nazalí ‘I am’; nayébí ‘I know’ [“simple” present]b) moto nyónso azalí na mabé na yé ‘everyone has his failings’ [generic]

Alternatively, we propose an analysis of this marker on partial analogy with the highly polysemous English presenttense (cf. Brisard 2002), whose various senses in Lingala are either expressed directly within the -í paradigm (e.g.,different functions of the progressive aspect with auxiliary be, ex. 3–4) or distributed over neighboring classes (aswith the generic and habitual usage types with -aka in ex. 5–7):

3. nazalí kosála ‘I am working/doing/making (right now)’ [durational]4. nazalí kobwákela yé milangi ‘I am throwing bottles at him’ [iterative]

5. nasálaka ‘I work normally/usually’ [habitual]6. moto nyónso abúngaka ‘everyone makes mistakes’ [generic]7. nalingaka musique na yé mingi ‘I like his music’ [e.g., every time I go to his show; cp. nalingí musique na

yé mingi ‘I like his music’, i.e., generally]

We shall argue that the special, perfect meaning of -í in (1) results from an inferred extension from the ‘oneinstance right now’-meaning of the present tense in combination with perfective (lexical) aspect. While in Englishthis configuration gives rise to a sense of general validity, Lingala has opted for a meaning of recent pastness withstrong current relevance to the time of speaking — yet with both interpretations featuring the same perfectivizingwork in the conceptual background. In turn, general validity for Lingala perfectives is expressed separately by -aka;imperfectives only take -aka (7) if a repetitive series of specific circumstances is understood from the context, forwhich scope the general-validity value holds. In other cases, of unqualified general validity, imperfectives take thetrue present-tense form (-í; cf. ex. 2b), as in English.

References

Brisard, Frank. 2002. “The English present”. In: Frank Brisard (ed.), Grounding, 251–297. Berlin: Mouton de Gruyter.Heine, Bernd & Derek Nurse. 2000. African Languages: An Introduction. Cambridge: Cambridge University Press.Meeuwis, Michael. 1995. “The Lingala tenses: A reappraisal”. Afrikanistische Arbeitspapiere 43: 97–118.–––––. 1998. Lingala. Munich: LINCOM Europa. (Languages of the World; Materials; Descriptive Grammars, nr. 261)–––––. 2001. “Tense and aspect in Lingala, and Lingala’s history: Some feedback on Nurse”. Afrikanistische Arbeitspapiere 67:

145–168.Nurse, Derek & Gérard Philippson. 2003. The Bantu Languages. London: Routledge.

12

Saturday, 13:30-14:30 – Main Foyer POSTER SESSION (#5)

Parsing Space Around Objects

Laura CarlsonUniversity of Notre Dame

Spatial terms denote specific regions of space (e.g., “in front of the cabinet”). Relatively little is knownempirically about the size and shape of these regions, and how they vary across contexts. In part this isdue to methodological limitations in previous studies (Carlson-Radvansky & Logan, 1997; Hayward &Tarr, 1995; Logan & Sadler, 1996) including: a) measuring spatial regions indirectly by collectingacceptability ratings for the use of a spatial term to describe various spatial relations; b) probing a limitedset of locations; and c) presenting the objects to be judged within a restricted 2D display, therebypotentially artificially limiting the size of the regions. In addition, there has been no systematic attempt toexamine how spatial regions may vary as a function of factors asserted to be relevant (Herskovits, 1986;Langacker, 1987; Miller & Johnson-Laird, 1976; Vandeloise, 1991), and known to impact other aspectsof spatial term interpretation. These factors include the identity of the objects (Carlson-Radvansky, et al,1999), the type of reference frame used to define the spatial term (Carlson-Radvansky & Irwin, 1993),and the presence of additional objects (Carlson & Logan, 2001),

We present a new methodology that addresses these limitations. Specifically, we use three directmeasurements of how “front” is defined in actual space surrounding an object. A small object (typically 5cm width X 7 cm length X 24 cm height) is placed in the middle of a 102 cm X 82 cm uniform whiteboard, and 11 lines are identified that radiate out from the reference object, three at the middle (0degrees), and 4 each at 22.5, 45, 67.5 and 90 degrees to the left and right. On each line, participantsindicate 1) the distance corresponding to the best use of “front,” 2) the farthest distance at which frontwould still be deemed acceptable; and 3) an alternative term (if applicable) that would be preferred at thisfar location, and the distance from the object at which front would be preferred to this alternative.Connecting these points across lines defines spatial regions corresponding to each of these measures (forsample data with a dollhouse cabinet, see Figure 1).

Using this basic methodology, across experiments we examined the impact of a diverse set of factors onthe size and shape of the front region, including the type of reference frame that defined the term, theidentity of the objects, and the presence of additional objects in the display. The critical findings were: 1)the spatial region defined by an intrinsic reference frame based on the object was larger than the spatialregion defined by a relative reference frame based on the viewer; 2) spatial regions were skewed in thedirection of a prominent functional part of the object; 3) the size of the region varied with the size of theobjects; 4) spatial regions around full-scale objects were larger than those around model-scaled objectsmatched in absolute size; and 5) the presence of additional objects compressed the regions.

References

Carlson-Radvansky, L. A., & Irwin, D. E. (1993). Frames of reference in vision and language: Where is above? Cognition, 46, 223-244.Carlson-Radvansky, L. A. & Logan, G. D. (1997). The influence of reference frame selection on spatial template construction. Journal of

Memory and Language, 37, 411-437.Carlson-Radvansky, L. A., Covey, E. S., & Lattanzi, K. L. (1999). “What” effects on “where”: Functional influences on spatial relations.

Psychological Science, 10, 516-521.Hayward, W. G., & Tarr, M. J. (1995). Spatial language and spatial representation. Cognition, 55, 39-84.Herskovits, A. (1986). Language and spatial cognition: An interdisciplinary study of the prepositions of English. Cambridge: Cambridge

University Press.Langacker, R. (1987). Foundations of cognitive grammar (vol. 1). Stanford: Stanford University Press.Logan, G. D., & Sadler, D. D. (1996). A computational analysis of the apprehension of spatial relations. In P. Bloom, M. A. Peterson, L. Nadel,

& M. Garrett (Eds.), Language and space (pp. 493-529). Cambridge, MA: MIT Press.Miller, G. A. & Johnson-Laird (1976). Language and perception. Cambridge: Belknap Press.Vandeloise, C. (1991). Spatial prepositions (Trans. By A. R. K. Bosch), Chicago: University of Chicago Press.

13

Friday, 17:30 – Aurora Room EYE-TRACKING STUDIES

Content and structure in anticipatory referential interpretation

Craig ChambersUniversity of Calgary

In studies of real time language comprehension, the use of nonlinguistic information istypically assumed to involve computations that are effortful, slow and computationally complex.However, there is increasing psycholinguistic evidence that nonlinguistic information is rapidlyand seamlessly integrated with the unfolding linguistic signal. This evidence appears to convergewith the theoretical perspective that few (if any) semantic values are fixed independently ofsituation-specific information or without access to general conceptual abilities.

The possibility that nonlinguistic factors play an extensive role in core interpretive processesunderscores the need to understand in greater detail what kinds of information constitute thelinguistically-relevant "context", and how these information sources interact to constrain theearliest moments of processing. In this paper, we describe a series of experiments that attempt toevaluate how some fine-grained and somewhat nonobvious nonlinguistic constraints influencethe comprehension of referring expressions. In these studies, eye movements are monitored aslisteners follow instructions to manipulate objects in a visual environment. Of interest is theexact point in the speech stream where listeners are able to identify the referent for a definitenoun phrase.

Our starting point is a study in which the visual display contained e.g., two squares (amongother things): one with a diamond pattern, and the other with a happy face pattern. Upon hearingClick on the square with the diamonds, eye movements indicated that listeners initiallyconsidered both squares, and identified the intended referent only once diamonds was heard.This pattern is expected because the unfolding NP is clearly ambiguous before the end of themodifying PP is reached. However, in a version where the second square (with the happy face)was replaced with a plain green square, identification of the intended square began immediatelyafter hearing square, i.e., before the PP modifier was heard. Listeners apparently eliminated thegreen square at this point because: (a) this object would normally be differentiated using a colourterm, (b) colour terms most naturally occur in prenominal position, and (c) the listener assumedthe speaker would adhere to conventions (a) and (b) if referring to this object. The effect thusreflects the interaction of markedly different kinds of contextual constraints, ranging fromperceptual information in the immediate environment to the listener's implicit awareness ofnongrammaticalized syntactic regularities.

We will describe additional studies that further evaluate the conditions under which listenerscan efficiently anticipate the structure of referring expressions and dynamically update thecandidacy of potential referents as an utterance proceeds. Most generally, these studies suggestthat this process depends on the ability to predict the kinds of properties that speakers are likelyto encode, which is in turn governed by ongoing conversational goals.

14


Writing Direction Influences Spatial Cognition

Ting Ting Chan & Benjamin BergenUniversity of Hawaii, Manoa

The world's languages make use of different writing system orientations, running from left to right, from right toleft, or from top to bottom. Interacting with writing systems is an important component of how literate humans gainand convey information, and as such the spatial routines we engage in while reading and writing may well have animpact on the spatial organization of other cognitive functions, like memory, visual attention, expectations about theorientations of processes, and so on. Three experiments tested for effects of writing system orientation on spatialcognition, using literate speakers of English, Mainland Chinese, and Taiwanese. The first experiment is an imagerecall task, which tested how memory (mediated by visual attention) may be influenced by writing orientation.Subjects were shown a screen with 42 images (6 per row, 7 per column) for 3 seconds and asked to remember asmany of these images as possible. It is hypothesized that the area of the screen speakers of different languages attendto will differ, depending on their writing system orientation. The results showed that English and Chinese subjectstended to scan images from left to right while Taiwanese scanned images from right to left and also from top tobottom, corresponding with their respective writing systems. The second experiment is a sentence composing task.Subjects were asked to compose simple sentences based on ten pairs of images, where each pair of objects is printedon the right and left hand side of a page. For example,

Figure 1

This experiment was to test whether the side that subjects tend to focus on first would again be influenced by writingdirection or not. English and Chinese speakers were hypothesized to take the image on the left as the subject of thesentence more frequently than the Taiwanese speakers did. A statistically significant relationship between writingsystem and visual attention was found, F (2,27)=7.275, P<0.01. On average, 6.8 and 5.9 sentences were using theobjects on the left as the subjects of the sentences for each English and Chinese speaker respectively; however, only3.3 sentences were constructed in that way by Taiwanese speakers. The third experiment was an arrangement task.Subjects were asked to spatially arrange pictures depicting three stages of development of a natural entity, from theearliest to the latest stage. It was hypothesized that the orientation of a writing system is engrained in speakers'perceptual and motor routines to the point that it surfaces when they perform these other spatial tasks. The expectedresult was subjects arranging the images in a pattern consist with the direction of their writing system. And theexperimental results showed a statistically significant relationship between writing system and the orientation inwhich sequential information was arranged. Five patterns of arrangement were found:

LR RL TB

BT CW

Figure 2: Five patterns of arrangement

For English and Chinese, the left to right pattern was dominant while for Taiwanese, the right to left and top tobottom patterns were dominant. More generally, the findings reported here support the idea that idiosyncraticcharacteristics of particular languages can influence general cognition.

1 2 3 3 2 1 1 2 3

32 1

13 2

Key:LR-left to rightRL-right to leftTB-top to bottomBT-bottom to topCW-clockwise

15

Friday, 16:00 – Prairie Room GESTURE

The Perception of the Form of Gestures as a Factor of their Function in Discourse

Alan CienkiEmory University

Previous research by the author (under review) considered the perception of the form of two typesof gestures devoid of the accompanying speech. Using a set of six elicited, videorecorded conversationsbetween pairs of students at an American college, gestures were coded by two researchers for whetherthey primarily had an abstract referential function, indicating an abstract notion as a concrete form(henceforth “A”), or primarily a discourse structuring function (“D”), such as making emphasis orpresenting an idea (adapted from Müller’s [1998] classification). The hypothesis was that whereas Agestures depict (metaphoric) source domains through various manual forms, the different functions of Dgestures involve more basic, schematic embodied patterns, namely those of image schemas. To test this,eighteen subjects were shown 20 A gestures and 20 D gestures from the data, without the accompanyingspeech. They were asked to judge for each gesture whether one of four image schemas (from the set of:CONTAINER, CYCLE, FORCE, OBJECT, PATH, SURFACE) could serve as an adequate descriptor of it, orwhether it should be categorized as “other.” Contrary to the hypothesis, the subjects reliably used imageschemas as descriptors for both D and A gestures. Thus, although the A gestures had a greater variety offorms, these forms could still be seen as schematic in structure.

The current study will test the role of the accompanying speech in the characterization of thesegestures’ forms. First we will assess the degree to which the accompanying verbal expressionsthemselves may be understood as image-schematic in content. Approximately 20 subjects will be giventhe 40 phrases affiliated with the 20 D and 20 A gestures and will be given the same coding sheet as hadbeen used for the gestures. It is predicted that subjects will not reliably use image schemas as descriptorsfor the phrases’ content. This is because, based on the assessment of two trained coders, neither theutterances with the D nor with the A gestures involve salient metaphorical or non-metaphorical schematicspatial images.

Next, approximately 20 subjects will view the gesture clips with the accompanying speech andwill be given the same coding sheet as above. It is predicted that in this condition subjects will ascribethe image schema labels with more reliable consistency to the A gestures than to the D gestures. Eventhough the utterances with the A gestures do not contain a greater amount of schematic spatial language,an A gesture depicts a concrete metaphoric source domain, and the affiliated speech indicates the relevanttarget domain for the mapping. If the A gestures are depicting conventional metaphoric mappings forthese speakers, subjects from the same lingua-cultural group (here: students from the same college) maybe influenced by these mappings, making for higher agreement in labelling the A gestures. However, thissame process should not happen with the D gestures because with them the accompanying speech is notfocussing the “search domain” in the same way. For example, most of the D gestures here (60%) aresimple palm-up open-hand presentation gestures which can be used with varying types of discoursecontent (Müller 2004); by contrast, only 10% of the A gestures here are of this type. We will conclude byconsidering the significance of these findings for our understanding of the different communicative rolesof gesture with speech.

References

Author (under review). “Image schemas and metaphoric gestures.”Müller, Cornelia (1998). Redebegleitende Gesten: Kulturgeschichte – Theorie – Sprachvergleich. Berlin: Berlin Verlag Arno

Spitz.Müller, Cornelia (2004). “The palm-up-open-hand. A case of a gesture family?” In The Semantics and Pragmatics of Everyday

Gestures, C. Müller & R. Posner (eds.). Berlin: Weidler Buchverlag, 233-256.

16


So… Discourse Particles Matter––An Empirical Approach

Nikolinka Collier, Corrina Sas, Fred CumminUniversity College Dublin

1. MotivationDo participants benefit from the use of discourse particles (DPs) in communication? There have been very fewcontributions investigating the impact of DPs on subjects’ discourse comprehension (Clark, 1997; Fischer, 2000;Fox_Tree, 1995). The work in this paper presents the results of such a study.

Here discourse particles are the words and fixed phrases (uh—huh, oh, okay, and, so, I mean ) used ininteraction to support the negotiation of understanding; to segment one utterance from another and at the same timerelate those in clause-like relationships; to give, take or hold turns; and to structure turns in topical or/andinteractional groups (Nenova, 2000).

DPs support discourse --> DPs presence facilitates discourse comprehension

This claim was the motivation for carrying out this study. It explored the impact of DPs on subjects’ comprehensionof discourse structure. Specifically, the goal was to explore how DPs influence speakers’ predictions of upcomingdiscourse events. Their prediction accuracy is used as a measure that shows the impact of DPs on an analysts’comprehension of the discourse.

2. MethodologyThe DP presence vs. absence was independent variable, while the accuracy of the annotation was the dependentvariable. The experiment comprised a group of 17 subjects divided into two groups: an experimental and a controlgroup. The subjects were given five transcribed dialogues from the TRAINS corpus (Gross, 1993). The dialogueswere modified so that some utterances were excluded from the transcriptions. The subjects had to select one of thepossible 12 labels and put it in the placeholder for the missing utterance.

The steps that were taken in conducting the pilot experiment were (a) to present the subjects with the materialsand (b) to instruct them in the annotation scheme and (c) to measure their performance based on the presence andabsence of DP interjections.

The TRAINS orthographic transcriptions were very suitable for the task because DPs are represented as separateutterances. This type of representation allowed the presence or absence of a DP to be treated as just another of themissing utterances, which needed to be predicted. Therefore there were no gaps within utterance to prompt speakersthat DPs are a possible control factor. In this way, the control group and the experimental group wereinformationally equal in the visual presentation that the dialogues allow. The first dialogue that the participantsannotated was regarded as a training dialogue and was excluded from the collected results.

3. ResultsIn order to test the impact of our independent variable upon the accuracy of prediction we performedWilcoxon signed rank test (Wilcoxon, 1945) between subjects comparing the accuracy achieved for the control andthe experimental group. There is a main effect on the accuracy of prediction (w(17), p<0.05). The difference inaccuracy of prediction between groups is significant, which supports our hypothesis that DPs constrain discourse.

4. SummaryThe first contribution of this study is the methodology. The second was that the experiment gives evidence thatdiscourse comprehension is influenced by DPs. This supports the research done so far in perusing DP functionalstructure. The design also allowed the identification of further factors that have impact on discourse comprehension.An experimental design using online data would allow stronger claims. However, the DPs and their structure andbehaviour need to be explored further in order to allow such an experiment.

ReferencesCLARK, JEAN E. FOX TREE AND HERBERT H. 1997. Pronouncing "the" as "thee" to signal problems in speaking,. Cognition, 62.151-67.FISCHER, KERSTIN. 2000. From Cognitive Semantics to Lexical Pragmatics: The Functional Polysemy of Discourse Particles. Berlin, New York.:

Mouton de Gruyter.FOX_TREE, JEAN. 1995. The effects of false starts and repetitions on the processing of subsequent words in spontaneous speech. Journal of

Memory and Language, 34.709-38.GROSS, DAVID TRAUM ; JAMES ALLEN AND DEREK. 1993. The TRAINS 91 Dialogues.: Computer Science Department, University of RochesterNENOVA, RONAN REILLY ; NIKOLINKA. 2000. Discourse Particle Taxonomy. Paper presented at Discourse Particles, Modal and Focal Particles

and all that stuff... Universitaire Stichting, Brussels, Belgium.WILCOXON, FRANK. 1945. Individual Comparisons by Ranking Methods. Biometrics, 1.80-83.

17


The Effect of Prototype on the Acquisition of English Spatial Prepositionsby Spanish Speakers: An Empirical Study

Margarita Correa-Beningfield*, Claude Vandeloise, Ignasi Navarro-Ferrando, & G. Kristiansen*Universidad Complutense de Madrid

Prepositional polysemy can be accounted for in terms of Vandeloise’s notion of ‘complex primitive’ (1991,2003) which implies prelinguistic concepts whose structure is conceived as a set of traits organised as familyresemblance sets. None of these properties is a necessary or sufficient condition. Certain combinations of the traitssanction adequate uses of a particular preposition. We will make the hypothesis that these sets can be viewed asthree-dimensional semantic structures, where each trait belongs to one of three experiential dimensions. Perception(sensory), action (motor) and interaction (purposive) contribute to conceptual development (Human experience =Perception + Body Motion/Action + Interaction/Adaptation and Assimilation).

We will test the relative importance of these three factors in the process of language acquisition. This paperoutlines the aims and methodology of the first cross-linguistic empirical study on the English and Spanish spatialprepositional systems. While the hypotheses described above do constitute a working assumption, tests are beingdesigned in such a way as to deduce conclusions from the subjects´ answers, rather than induce the subjects tochoose from predetermined options.

In this study we are systematically comparing and contrasting the English and Spanish spatial prepositionswithin the framework of prototype theory (Rosch, 1973). As both the geometrical approach (Hawkins, 1984; 1988)and the topological approach (Herweg, 1989) to the analyses of spatial prepositions within a cognitive linguisticframework do not adequately explain in all cases the different uses of a preposition we have introduced a descriptionbased on functional concepts (Vandeloise, 1991; 1994; 2003) that are determined by the dynamic interaction of thetwo arguments related by a preposition. It is a well known fact that the difficulty encountered in giving a directtranslation of prepositions is due to the fact that the semantic ranges of corresponding prepositions are usuallydifferent. The English and Spanish spatial prepositions show in a particularly striking way the absence of a one-to-one correspondence between these two languages due mainly to the fact of the significantly higher number ofEnglish spatial prepositions in comparison to the Spanish ones. This study will (1) attempt through objective testingto establish prototypes and degrees of prototypicality for the English and Spanish spatial prepositions; (2) investigatelanguage transfer from Spanish to English in spatial uses of these prepositions; and (3) examine the role of prototypein such transfer. This study will have pedagogical implications in raising questions regarding the sequencing ofmaterial for teaching English as a Second/Foreign Language.

References

Hawkins, B. W. 1984 The semantics of English spatial prepositions. Ann Arbor, MI: University MicrofilmsInternational.

–––––. 1988 The category MEDIUM. In B. Rudzka-Ostyn (ed.). Topics in Cognitive Grammar. Amsterdam &Philadelphia: John Benjamins, 231-270.

Herweg, M. 1989 Ansätze zu einer semantischen Beschreibung topologisher Präpositionen. In Habel, C., Herweg,M. & K. Rehkämper (eds.), Raumkonzepte in Verstehensprozessen. Tübingen: Max Niemeyer Verlag, 99-127

Rosch, E. 1973 On the internal structure of perceptual and semantic categories. In T.E. Moore (ed.), Cognitivedevelopment and the acquisition of language. New york: Academic Press, 111-144.

Rosch, Eleanor & Carolyn B. Mervis 1975 Family Resemblances: Studies in the Internal Structure of Categories,Cognitive Psychology 7: 573-605

Vandeloise, Claude. 1991: Spatial Prepositions: A Case Study from French. Chicago–––––. 1994: Methodology and Analyses of the Preposition in. In: Cognitive Linguistics 5 / 2, 157-184–––––. 2003 Containment, support and linguistic relativity, in Cuyckens, H., Dirven, R., & J. R. Taylor (eds.)

Cognitive Approaches to Lexical Semantics. Berlin & New York: Mouton de Gruyter, 393-425.

18


Entrenchment and the Distribution of Complement Types in English

Hubert Cuyckens and Hendrik De SmetK.U. Leuven

It has been shown that the distribution of complement types in English (gerund, to-infinitive, and that-clause) overthe range of complement-taking verbs has been unstable through time. Examples are the diffusion of to-infinitives atthe cost of subjunctive that-clauses in Old and Middle English (Fischer et al. 2000) and the diffusion of gerundclauses (partly) at the cost of to-infinitives in Modern English (Fanego 1996; Rudanko 1998).

The purpose of this paper is to look at a usage-based factor which has not yet received a great deal of attentionin diachronic accounts of the changing distribution of complement types. Making use of various synchronic anddiachronic corpora (covering the periond 1700-PDE), we examine what effect the entrenchment of a MATRIX VERB +COMPLEMENT construction at a particular synchronic stage has on the maintenance of the complement type(s) withthat matrix verb at subsequent stages. Especially, we look at the following constructions: (i) RETROSPECTIVE

VERB+TO+HAVE+VERB.OF.PERCEPTION/ENCOUNTER – a construction which no longer occurs in PDE, and (ii)LIKE+TO-INF. Examples are:

(i) He did not remember to have seen the expression on her face.(ii) Deer like to ponder the view as they graze.

In order to establish the degree of entrenchment of a construction, we make use of such quantitative data as thestring frequeny of the construction type, the transitional probability of componenti-1 relative to componenti of thestring, and the string’s collocational strength. For the two constructions under investigation, our corpus data revealthe following:

(i) In the 18th and 19th century, a high degree of entrenchment can be established for the (fairly specific) constructionRETROSPECTIVE VERB+TO+HAVE+VERB.OF. PERCEPTION/ENCOUNTER, which scores high both on collocational strength and ontransitional probability between the components RETROSPECTIVE VERB+TO+HAVE and VERB.OF.PERCEPTION/ENCOUNTER (incontrast, its string frequency is relatively low).

(ii) In the entire period 1700-PDE, the construction LIKE+TO-INFINITIVE shows (increasingly) higher entrenchment than theconstructions ‘LOVE/HATE+INF’––this is partly due to the high type frequency of the more specific WOULD.LIKE+TO-INF

construction.

On the basis of these data, it is hypothesized that:

(i) The entrenchment of the RETROSPECTIVE VERB+TO+HAVE+VERB.OF. PERCEPTION/ENCOUNTER (or PERFECT-INFINITIVE)construction is an important factor in the appearance (from 1780 on) of the RETROSPECTIVE VERB+PERF.GERUND constructionwith verbs of perception/encounter, at a time when a simple-gerund complement was already available in the language.Indeed, it appears that the PERFECT-GERUND in these cases has been calqued on the PERFECT-INFINITIVE. In that sense, theentrenchment of the PERFECT-INFINITIVE has had a progressive effect (cf. Krug 2003).

(ii) The higher entrenchment of the LIKE+TO-INF construction relative to the construction LOVE/HATE+INF has had a conservativeeffect in that has made like more resistant to taking a gerund at a time when the gerund was definitely on the increase withother emotive verbs. Especially the conservative effect of WOULD LIKE+INF is telling: even though would like is not, in fact,incompatible with a gerund, it still largely blocks complementation by the gerund. Conversely, the LIKE+TO-INF constructionhas a progressive effect on other verbs in that would hate and would love can also take the to-infinitive.

In general, this paper provides further evidence that, in addition to the domains of phonology and morphology (cf.Bybee 2001; Croft & Cruse 2004), entrenchment also impacts the domain of syntactic change (cf. Bybee andThompson 1997). In addition, it suggests that the distribution of complement types is not solely determined bysemantic/pragmatic/stylistic factors, but also by the fact that language users tend to use the complement types theyhave always used (cf. Noël 2003: 368: “most formal choices are habitual choices”).

References

Bybee, Joan. 2001. Phonology and Language Use. Cambridge: Cambridge University Press.Bybee, Joan & Sandra Thompson. 1997. “Three frequency effects in syntax”. In Berkeley Linguistics Society 23: 378-388.Croft, William & D. Alan Cruse. 2004. Cognitive Linguistics. Cambridge: Cambridge University Press.Fanego, Teresa. 1996. “The development of gerunds as objects of subject-control verbs in English”. In Diachronica 13: 29-62.Fischer, O., A. van Kemenade, W. Koopman & W. van der Wurff. 2000. The Syntax of English. Cambridge: Cambridge University Press.Krug, Manfred. 2003. “Frequency as a determinant in grammatical variation and change”. In Günter Rodenburg & Britta Mondorf (eds.),

Determinants of Grammatical Variation in English. Berlin: Mouton de Gruyter, pp. 7-67.Noël, Dirk. 2003. “Is there semantics in all syntax? The case of accusative and infinitive constructions vs. that-clauses”. In Günter Rodenburg &

Britta Mondorf (eds.), Determinants of Grammatical Variation in English. Berlin: Mouton de Gruyter, pp. 347-377.Rudanko, Juhani. 1998. Change and Continuity in the English Language. Lanham, MD: University Press of America.

19

Friday, 15:00 – Aurora Room SCHEMAS, RULES, & ANALOGY

Analysing the Grammaticalization of EDGE Noun-Expressions: A Data-Driven Approach

Kristin Davidse and Martine Vanden EyndeK.U. Leuven

This paper aims at relating theoretical insights into the properties of constructions (Langacker 1987 1991, McGregor1997) to analytical heuristics which interrogate corpus concordances about the way the component units ofconstructions combine with each other and about the lexicogrammatical selection restrictions applying to thesecomponent units.

The phenomena studied are expressions with an ‘edge noun’ (henceforth edge N) such as edge, verge, brink andpoint, which, as we will argue, can be part of three distinct constructions:

(i) a NP with the edge N as head, e.g. …on reaching the edge of the toadstool valley… (OED)(ii) a PrepP in which the edge N is part of a complex preposition, e.g. She was on the edge of tears. (COBUILD corpus)(iii) a clause in which the edge N is part of a modal-future marker of ‘imminence’ of the process depicted, e.g. …the warlord

who was also once on the brink of capturing Kabul. (COBUILD corpus)

These three constructions are linked to each other by progressive grammatical re-analysis, involvinggrammaticalization and semantic generalization of the edge N.

The data-driven argumentation for distinguishing these three constructions will refer to differentialcombinatorial properties, both internal and external to the structures being considered. Internally, for instance, thepremodification of the edge Ns changes from modification by a variety of adjectives (furthest, inner, western, etc.)in construction type (i), over the odd occurrence of fixed intensifying very in (ii) to no premodification at all in (iii).Systematic analysis of the data also reveals that the three edge N-constructions function externally in differentpositions of clausal structure: the NP with edge N as head in (i) can function in any clausal position, subject, objectand prepositional complement; the prepP with edge N as part of a complex preposition in (ii) occurs only asrelational complement of a verb or as circumstantial modifier of a clause; finally, the secondary auxiliary markerwith edge N in (iii) is found only in VPs containing verbal –ing forms.

The sorted data concordances of these three constructions will then be investigated in terms of grammatical andlexical selection restrictions applying to their component units. Important questions regarding the grammatical statusof the elements following the edge N are, for instance, their (broadly) deverbal and nominalized status in (ii) and thequestion whether in (iii) they are gerunds or clausal heads (as defined by Langacker 1991). For the investigation oflexical selection restrictions, heuristics developed by Sinclair (1991) and Stubbs (1996) can be brought to bear onthe semantic characterization of the constructions, by investigating the following questions:

• By what sort of lexical classes are the elements of structure realized? For instance, following the edge N we find NPsdesignating bounded spatial and temporal entities in (i), nominals designating states and processes in (ii), and –ing formsdesignating actions and events in (iii).

• Does the edge N command a specific collocational range in one of the constructions? By way of leftward collocates, we findbe on, teeter/hover on, be brought/led to, etc + edge N in (ii) versus no strict collocations in (i) and fixed coalesence with bein (iii).

• Does the edge N in one of the three constructions predict affectively marked (negative or positive) collocates? In (ii) itpredicts, besides the odd positive collocate such as greatness and triumph, predominantly negative collocates such asmadness, hysteria, famine, war, collapse, bankruptcy.

It is on the basis of such data-driven generalizations, as well as quantification of the relative occurrence of the threeconstruction types in diachronic and synchronic data, that we will discuss the motivating factors in this layered(Hopper 1991) grammaticalization process. We will also argue that questions relating to the as yet final stage ofgrammaticalization (iii), such as its ‘proximative’ or ‘avertive’ (Kuteva 1998) nature, can best be broached from aperspective which takes into account the whole history of the grammaticalization process, including the impact ofthe source semantics of the edge Ns and the clustering of negative collocates in some stages of thegrammaticalization process.

ReferencesHopper, P. 1991. On some principles of grammaticalization. In Traugott, E. Closs & Heine, B (eds) Approaches to Grammaticalization. Vol. I.

Amsterdam: Benjamins, 17-35.Kuteva, T. 1998. On identifying an evasive gram: Action narrowly averted. Studies in Language 22:1: 113-160.Langacker, R. 1987. Foundations of Cognitive Linguistics. Vol. 1. Theoretical Prerequisites. Stanford: Stanford University Press.Langacker, R. 1991. Foundations of Cognitive Linguistics. Vol. 2. Descriptive Application. Stanford: Stanford University Press.McGregor, W. 1997. Semiotic Grammar. Oxford: Oxford University Press.Sinclair, J. 1991. Corpus, Concordance, Collocation. Oxford: Oxford University Press.Stubbs, M. 1996. Text and Corpus Analysis. Oxford: Blackwell.

20


Finding Paradigms among the Navajo Verb Stems

David Eddington and Jordan LachlerBrigham Young University and Sealaska Heritage Institute

Navajo verbs encode lexical, aspectual and modal information in the verb stems. One difficultywith the verbs stems is that they do not lend themselves to paradigmatic characterization basedon their phonemic structure without resorting to extremely abstract representations whichessentially reconstruct their historical origins. Our goal is to flesh out some of the regularsubpatterns that do exist, basing the analysis on surface apparent traits. We also carried out aseries of analogical simulations on the stems. The analogical framework assumes that all knownstems are stored in a network in which stems are interconnected based on their phonological andsemantic similarities. Within this framework it was possible to predict the phonological shape ofa given stem by analogy to similar stems stored in the lexicon at a success rate of about 70%.Given the apparent lack of systematic relationships between the verb stems we feel that a 70%success rate is highly respectable and may relate to how the verb stems are organized in themental lexicon.

21

Sunday, 10:30 – Aurora Room COLLOSTRUCTIONS

Frame Networks for Noun Concepts: From Corpus to Cognition

Hisanori Furumaki and Maki SakomotoTokyo Insitute of Technology and University of Electro-Communications

This presentation focuses on frame networks for noun concepts from the viewpoint of the relationship between anoun and its co-occurrence phrase. We will make clear the conceptual structure of active zone in cognitive grammarand 'valence relations' based on actual use found in corpus. We aim at giving a structure based on 'frame elements'(FEs); that is, an aspect supported by the conceptually-overlapped group of co-occurrence phrases in 'frame' in thesense of Minsky (1975), Fillmore (1975, 1982), and Fontenelle (2003).

What is the relation between highlighted aspects of the concept coded by the noun book and its co-occurrencephrase with verbs such as read and buy? Chomskian approaches (e.g. Katz and Fodor 1963, Chomsky 1965) dealtwith the syntactic relationship between verb and noun as 'selectional restrictions'. In contrast, we basically regard therelation as 'valence relation' in the sense of cognitive grammar (Langacker 1987, 1999; Taylor 2002). We think,however, even valence relation in cognitive grammar cannot entirely explain the profile determinant of the book inbuy the book and translate it into French. The expression book should be explained dynamically, i.e. not statically in'selectional restrictions' approaches, considering the conceptual overlap between something bought and somethingtranslated into French. (Cf. Kunihiro 1997, Pustejovsky 1995)

We will show the characteristics of noun concepts and propose a model that accommodates it by analyzing thefollowing sentences in Japanese corpus.(1) Hon-wo yomu-koto-ga suki-da. Dekirudake ooku-no- (2) hon-wo yomi, johoushakai-ni tachi-okure-nai-yo tsutome-teiru. Ooku-no- (3) hon-wo kounyushi-te, shosai-ni obitadashii- (4) hon-ga afure-te shima-tta.(I like reading books. I make it a rule to read as many books as possible, so that I can keep up with rapid changes in information society. Ibought so many books, that my study is overflooding with them.)

The underlined part of each sentence represents a co-occurrence phrase pertaining to hon (lit. book). Hon in eachsentence activates three different FEs, i.e. "content", "commercial goods", and "physical object". Our approachallows us to determine which FE is activated by hon in each case. Furthermore, we will consider relations betweeneach FE more carefully, which show different degrees of activation. FEs are conceptually-overlapped as inWittegenstein's view of a category like Spiel.

Using computational parser for natural language, we collected about 3,000 token sets of combinations of hon-particle-verb in a Japanese computational pos-tagged corpus.

(5) Hon-wo-yomu (lit. book-particle.ACC-read) [2166 freq](6) Hon-wo-kaku (lit. book-particle.ACC-write) [343 freq](7) Hon-ni-matomeru (lit. book-particle.GOAL-collect) [307 freq](8) Hon-ga-deru (lit. book-particle.NOM-be.out) [304 freq](9) Hon-wo-kau (lit. book-particle.ACC-buy) [208 freq](10) Hon-wo-shuppansuru (lit. book-particle.ACC-publish) [179 freq](11) Hon-ga-ureru (lit. book-particle.NOM-sell) [155 freq](12) Hon-wo-tsukuru (lit. book-particle.ACC-make) [132 freq](13) Hon-wo-erabu (lit. book-particle.ACC-select) [129 freq](14) Hon-wo-miru (lit. book-particle.ACC-see) [112 freq]

These phrases can be classified on the basis of FEs. As we all know, all of the Hon in (5)-(14) have an aspect of"physical object". The hon in (5), (6), (7), (8), (9), and (12) may highlight "content". The hon in (8), (9), (10), (11),and (13) can foreground "commercial goods". Hon in (5), (6), (14) may focus on "book's letter". In other words, theconceptual noun hon consists of groups of FEs, and it forms Gestalt defined by all FEs. We will show framenetworks for noun concepts structured by FEs from the viewpoint of co-occurrence phrases, making use of corpora.FEs supported by co-occurrence phrases can be explained only by applying conceptual overlap.

ReferencesN. Chomsky (1965) Aspects of the Theory of Syntax, Cambridge, MA: The MIT Press.C. Fillmore (1975) An Alternative to Checklist Theories of Meaning, Proceedings of the 1st Annual Meeting of the BLS, 123-131.C. Fillmore (1982) Frame Semantics, In Linguistic Society of Korea (ed.) Linguistics in the morning calm, 111-137. Soel: Hanshin.T. Fontenelle (ed.) (2003) Special issue: FrameNet and Frame Semantics, International Journal of Lexicography, 16-3.J. J. Katz and J. A. Fodor (1963) The Structure of a Semantic Theory, Language 39, 170-210.T. Kunihiro (1997) Riso-no Kokugo Jiten (The Ideal Japanese Dictoinary), Tokyo: Taishukan.R. W. Langacker (1987) Foundations of Cognitive Grammar vol. 1: Theoretical Perspective, Stanford: Stanford University Press.R. W. Langacker (1999) Grammar and Conceptualization, Berlin / New York: Mouton de Gruyter.M. Minsky (1975) A framework for representing knowledge, In Winston P. (ed.) The Psychology of Computer Vision, 211-277. McGraw-Hill.J. Pustejovsky (1995) The Generative Lexicon, Mass: MIT Press.J. R. Taylor (2002) Cognitive Grammar, New York: Oxford University Press.

22


Iconicity in Expressives: An Empirical Investigation

Michael Gasser, Nitya Sethuraman, Stephen HockemaIndiana University

We suggest that different semantic domains have differing degrees of iconicity, depending uponhow crucial it is for communication. By this hypothesis, the domain of concrete nouns would have lessiconicity than, for example, the domain of manner terms.

Iconicity can be absolute or relative. Absolute iconicity is based on a similarity function relatingforms and the objects, properties, states, or events in the world (hereafter “situations”) they designate.Onomatopoeia exemplifies absolute iconicity; there is a perceived resemblance between particularlinguistic sounds and particular sounds in the world. This similarity function is psychological: whatcounts as iconic for you may not for me if we don’t measure similarity between relevant forms andsituations in the same way. Relative iconicity is based on similarity functions that relate forms to formsand situations to situations. A set of expressions embodies relative iconicity if expressions with similarmeanings tend to have similar forms, even if the relationship between form and meaning is arbitrary.

Relative iconicity can be contrasted with relative anti-iconicity. A set of expressions embodiesrelative anti-iconicity if similar meanings tend to have forms that are more dissimilar than predicted bychance. We hypothesize that relative anti-iconicity characterizes semantic domains in which boundariesbetween categories are crucial. For example, dogs resemble wolves, but the difference matters; we wouldexpect word forms for these categories to differ more than they would by chance.

If this is true, are there comparable semantic domains or lexical categories where relativeiconicity would be the norm? We hypothesize that relative iconicity will characterize “smoother”semantic domains, those with either no clear boundaries between semantic categories or with boundariesthat are not crucial for communication. Manner may be such a domain. If you misunderstand myreference to a “wobbly” gait to be a “stumbling” gait, you are still on the right track.

Many languages of the world, especially in Africa, South Asia, Southeast Asia, and the Americas,have a lexical category referred to as expressives (e.g., sorasora, ‘roughness’, Tamil). While there appearto be no necessary and sufficient conditions for defining this category cross-linguistically, a number ofphonological, morphological, syntactic, semantic, and pragmatic properties characterize the expressiveprototype (Childs, 1994). We focus on the semantic properties of expressives, specifically the extent towhich they exhibit relative iconicity.

To test these hypotheses, we examine randomly selected sets of expressives and concrete nounsin three unrelated languages, Japanese, Tamil, and Amharic. We assess similarity of form using a measureof distance based on standard phonetic analyses of the languages. Semantic similarity is assessed usingLatent Semantic Analysis (Landauer & Dumais, 1997) to compare the “meanings” of English translationsof pairs of words. We examine both the overall variability within each category (expressives and concretenouns) and the correlation of form similarity with semantic similarity, a measure of relative iconicity. Ourpreliminary results suggest that the typical semantic domains for expressives, including manner, lendthemselves to relative iconicity more than other domains.

References

Childs, G. T. (1994). African ideophones. In L. Hinton, J. Nichols, & J. Ohala (Eds.), Sound Symbolism. Cambridge: CambridgeUniversity Press.

Landauer, T.K. and Dumais, S. T. (1997). A solution to Plato's problem: The Latent Semantic Analysis theory of acquisition,induction and representation of knowledge. Psychological Review, 104, 211-24.

23

Friday, 10:30 – Prairie Room METAPHORICAL & METONYMICAL MAPPINGS

“Superschemas” and the Grammar of Metaphorical MappingsJoe Grady

Cultural Logic LLC

This paper begins with an examination of the relationship between the ‘source’ and ‘target’ concepts in threecommon (and crosslinguistic) metaphorical patterns:

DIFFICULTY AS HEAVINESS:“I’ve got quite a heavy workload to contend with.”agir Turkish ‘heavy; cumbersome, difficult’

ACHIEVING-AN-OBJECTIVE AS ARRIVING-AT-A-GOAL:“We’re almost finished but not there yet.”sampai Malay ‘arrive; achieve a goal’

CAUSES AS SOURCES:“A great deal of good came out of that meeting.”a Old Irish ‘from, out of; due to, because of’

and argues that these patterns of conceptual association are characterized neither by similarity/analogy of the sortenvisioned in traditional philosophical and psychological accounts (e.g., Searle 1979; Gentner 1988), nor by‘invariance’ as described by cognitive linguists (e.g., Brugman 1990; Lakoff 1990, 1993; Turner 1991). Cognitivelinguists since Lakoff and Johnson (Metaphors We Live By, 1980) have argued compellingly that there are manycommon metaphorical patterns which cannot be accounted for in terms of similarity (i.e. primary metaphors, Grady1999; Lakoff & Johnson, 1999) -- for instance, brightness and happiness (“a bright disposition”) are associated witheach other in experience, but not “similar” in any straightforward way. On the other hand, scholars in this samecognitive linguistic tradition have argued in favor of the ‘Invariance Principle’, which holds that source and targetmust share image-schematic structure, what Lakoff refers to as “cognitive topology” (1993: 215) - i.e. “sources willbe mapped onto sources, goals onto goals,” and so forth.

Following Grady’s (1997, 1999) argument that target concepts like Difficulty, Achieving-an-Objective andCause do not refer directly to perceptual experience, and therefore are not structured by image-schemas (rather,image-schematic structure is projected onto them from metaphorical source concepts), this paper takes the positionthat there is a large and important class of metaphorical patterns (including the three which are the focus of thepaper) to which Invariance as defined in the literature does not apply. Rather, a new constraint on metaphoricalmappings is proposed: source and target share an even more generic type of structure, at the superschematic level.Superschematic structure within the three metaphorical patterns considered includes such elements as Scalarity (i.e.both source and target can be construed as scalar properties); Aspect (e.g., source and target can be conceived aspunctual vs. durative events); and Boundedness (source and target can be construed as bounded vs. unboundedentities). Source and target must also be construable as relations involving the same number of arguments (i.e. theyshare the same ‘arity’).

The paper’s second part discusses the fact that these and other superschematic features (includingontological category, profile-base structure, trajector-landmark structure, etc.)are also central elements ofgrammatical meaning (see, e.g., Langacker’s 1987 discussion of parameters within cognitive grammar). Thisparallel adds a potentially important new dimension to the study of grammar as “the fundamental conceptualstructuring system of language” (Talmy, 2000: 21): The paper argues that metaphorical and grammatical dataconstitute converging evidence for the centrality of superschemas as cognitive organizing principles.

References

Brugman, C. 1990. What is the Invariance Hypothesis?. Cognitive Linguistics 1.Gentner, D. 1988. Metaphor as Structure Mapping: The Relational Shift. Child Development, 59: 47-59.Grady, J. 1999. A Typology of Motivation for Conceptual Metaphor: Correlation vs. Resemblance. In R. Gibbs & G. Steen (Eds.), Metaphor in

Cognitive Linguistics. Amsterdam: John Benjamins.Grady, J. 1997. Foundations of meaning: Primary metaphors and primary scenes. UC, Berkeley, Ph.D. dissertation.Lakoff, G. 1993. The contemporary theory of metaphor, In Andrew Ortony (Ed.), Metaphor and thought, 2nd edition. Cambridge: CUP.Lakoff, G. 1990. The Invariance Hypothesis: Is abstract reason based on image schemas?. Cognitive Linguistics 1.Lakoff, G. and M. Johnson. 1999. Philosophy in the flesh. New York: Basic Books.Lakoff, G. and M. Turner. 1989. More than cool reason: a field guide to poetic metaphor. Chicago: University of Chicago Press.Langacker, R.W. 1987. Foundations of cognitive grammar, Vol. I: theoretical prerequisites. Stanford: Stanford University Press.Searle, John R. 1979. Metaphor. Excerpted from Expression and Meaning. Cambridge: CUP. (Reprinted in M. Johnson, Philosophical

Perspectives on Metaphor.)Turner, M. 1991. Reading minds: The study of English in the age of cognitive science. Princeton: Princeton University Press.Talmy, L. 2000. Toward a Cognitive Semantics. Cambridge: MIT Press, 2000.

24


Clustering Collexemes to Identify Constructions’ Sense Extensions

Stefan Th. Gries & Anatol StefanowitschSouthern Denmark University and University of Bremen

Within the framework of Construction Grammar, a construction is defined as a form-meaning pair of which someaspect of form or meaning is not predictable (from the meaning of its parts or other independently establishedconstructions). This definition implies that the construction applies to the whole range of different levels oflinguistic complexity from simple morphemes over (partially) lexically filled idioms (such as the way construction)up to abstract syntactic structures and argument structure constructions. However, two interrelated questions thatarise from such a perspective are the following: (i) while the formal aspect of an abstract syntactic structure or anargument structure construction can be identified straightforwardly, how can we identify the semantics of aparticular syntactic structure or argument structure construction? (ii) which words can occur in which (slot of a)construction? As an all too familiar example, consider the V-slot of the ditransitive construction represented in (1)with some examples and their senses: what does the construction mean and which verbs can be used in it?

(1) a. [VP V [NP ] [NP ]]b. John gave Peter a suitcase. --> caused to receivec. John denied Peter a loan. --> cause not to received. John leaves Peter a present. --> cause to receive in the futuree. John told Peter a story. --> communication as transfer

In previous publications, Stefanowitsch and Gries established a variety of corpus-based techniques to address thesequestions (cf. Stefanowitsch & Gries [2003], Gries & Stefanowitsch [2004, to appear]). For example, their so-calledcollostructional analysis makes it possible to identify those words (often, as in (1), the verbs) which occur in aparticular constructional slot significantly more frequently than expected by chance alone; these are referred to ascollexemes. More importantly, it turns out that the strongest (verbal) collexemes are also most indicative of theconstruction’s sense(s). For the ditransitive construction, for example, Stefanowitsch and Gries (2003) demonstratethat all the major sense extensions described in Goldberg (1995) are represented among the its significantcollexemes (including those in (1)).

While the semantic characterization of a construction (by identifying the verb groups and the senses theyinstantiate) is a relatively straightforward matter in the case of the ditransitive, it can be much more complex withother constructions such as the into-causative (cf. (2a-b)), the way-construction (cf. (3a-b)), or the resultative (cf.(4a-b)); the problems that arise are similar to those one meets when describing the many related senses ofpolysemous words.

(2) a. He tricked him into buying this old car.b. He forced him into confessing the crime.

(3) a. He found his way into the new museum.b. They made their way across the Atlantic.

(4) a. He was making her sick.b. He shot her dead.

In this paper, we will propose an extension of collostructional analysis that is aimed at facilitating the objectiveidentification of constructional senses and the verb groups instantiating them. To that end, we will discuss the resultsof applying cluster-analytic techniques to the verbs strongly attracted to the argument structure constructionmentioned in (2) to (4) (cf. Sandra & Rice [1995] for a cognitive-linguistic use of such techniques and Schütze[1998] for a more computational background). More specifically, we will demonstrate how a hierarchical clusteranalysis can be used to group collexemes of the instances of a construction in such a way that the analyst canobjectively identify which collexemes of a construction behave similarly enough to merit grouping them together asa single (cluster of) sense(s). In a addition to the specific results for each individual constructions, we will discussthe implications this has for approaches to argument structure constructions in general.

References

Goldberg, Adele E. 1995. Constructions: A construction grammar approach to argument structure. Chicago: The University of Chicago Press.Gries, Stefan Th. & Anatol Stefanowitsch. 2004. Extending collostructional analysis: A corpus-based perspective on ‘alternations’. International Journal of Corpus

Linguistics 9.1:97-129.Gries, Stefan Th. & Anatol Stefanowitsch. to appear. Co-varying collexemes in the into-causative. In: Achard, Michel & Suzanne Kemmer (eds.). Language, Culture,

and Mind. Stanford, CA: CSLI.Sandra, Dominiek & Sally Rice. 1995. Mirroring whose mind – the linguist’s or the language user’s? Cognitive Linguistics 6.1:89-130.Schütze, Hinrich. 1998. Automatic word sense discrimination. Computational Linguistics 24.1:97-123.Stefanowitsch, Anatol & Stefan Th. Gries. 2003. Collostructions: Investigating the interaction between words and constructions. International Journal of Corpus

Linguistics 8.2:209-43.

25


Do Foreign Language Learners Also Have Constructions?Interrelated Evidence from Priming, Sorting, and Corpora

Stefan Th. Gries & Stefanie WulffSouthern Denmark University and University of Bremen

In Construction Grammar, the ultimate grammatical unit is the construction, a conventionalized form-meaningpairing. Evidence for constructions comes from the observation of systematic meaning differences between one verbin several syntactic frames (Goldberg 1995), priming of constructions (cf. Hare & Goldberg 1999), sortingexperiments (cf. Bencini & Goldberg 2000) and language acquisition (cf. Tomasello 1998).

This evidence in favor of constructions is so far exclusively based on native speaker data. This is of course not aweakness, but it would also be interesting to see to what degree such findings also apply to non-native speakers, andinvestigating constructional knowledge of non-native speakers could provide additional evidence for construction-based approaches in general. This paper explores this possibility with two case studies; one looks at the formalaspect of constructions, the other at their semantic part.

We first investigated the constructional knowledge of non-native speakers with a syntactic priming paradigm.Syntactic priming refers to the tendency to repeat syntactic structures encountered shortly ago. In a sentence-fragment completion study with native speakers of German who are advanced learners of English, we obtained ahighly significant priming effect (c2(1)=34.55; p<.001): caused-motion constructions and ditransitives primecaused-motion constructions and ditransitives respectively. Also, the priming effects correlate strongly with theverb-construction preferences in native speaker corpora (cf. Stefanowitsch & Gries 2003; Gries & Stefanowitsch2004): verbs which are strongly associated with one construction resist priming to other; more importantly, thepriming effects do not correlate with verb-construction preferences from German translation equivalents, ruling outa mere translational explanation.

The results supported a constructional approach, but one might try to explain them away with syntactic primingrather than with constructional priming. Thus, we replicated Bencini & Goldberg’s (2000) sorting experiment toinvestigate how accessible the meanings of constructions are to German learners of English. Twenty Germanstudents of English sorted 16 sentences into four piles of four sentences according to their similarity of meaning.The sentences crossed four argument structure constructions (transitives, caused-motion constructions, resultatives,and ditransitives) and four verbs (throw, take, get, and cut); the dependent variable was whether subjects performeda verb-based or a construction-based sorting. It turned out that the subjects exhibited a strong tendency towards aconstruction-based sorting (p<.01); this was supported by additional cluster-analytic results, which showed that eventhe way the constructions are grouped together are fully compatible with how some constructions are related to eachother in, say, Goldberg (1995: section 3.4.1).

In sum, we obtained interrelated (syntactic and semantic) evidence from three different methods in favor ofattributing an ontological status to constructions even for non-native speakers of a language; this in turn stronglysubstantiates constructional approaches. As to language learning, the mere fact that foreign-language learners arriveat patterns lending themselves to a construction-based explanation even in the absence of explicit constructionlearning corroborates construction-based approaches to first-language acquisition. Furthermore, it raises theinteresting question to what degree aspects of foreign-language learning can be explained with reference to(construction-based) mechanisms already known from first-language acquisition and adult usage preferences asmanifested in corpora.

References

Bencini, J. & A.E. Goldberg. 2000. The contribution of argument structure constructions to sentence meaning. Journal of Memory and Language43.4:640-51.

Goldberg, A.E. 1995. Constructions: A Construction Grammar Approach to Argument Structure. Chicago: U of Chicago Press.Gries, St.Th. & A. Stefanowitsch. 2004. Extending collostructional analysis: A corpus-based perspective on ‘alternations’. International Journal

of Corpus Linguistics 9.1:97-129.Hare, M.L. & A.E. Goldberg. 1999. Structural priming: Purely syntactic? In: Hahn, M. & S.C. Stones (eds.). Proceedings of the 21st Annual

Meeting of the Cognitive Science Society, p. 208-11.Stefanowitsch, A. & St.Th. Gries. 2003. Collostructions: Investigating the interaction between words and constructions. International Journal of

Corpus Linguistics 8.2:209-43.Tomasello, M. 1998. The return of constructions. Journal of Child Language 25.2:431-42.

26


Influence of Prosodic Weight on Sentential Discourse Structure

Mohammad Haji-AbdolhosseiniUniversity of Toronto

This study seeks to determine whether and how the prosodic weight of a discourse unit (DU) influences sententialdiscourse structure, and what implications that has for linguistic theory. In this research, prosodic weight ismeasured by syllable length, and discourse structure is defined based on the Rhetorical Structure Theory (RST,Mann 1984; Mann and Thompson 1986, 1988a, 1988b). In RST, text structure is defined in terms of nuclei,satellites, and various relations among them.

The data for this study come from the RST Discourse Treebank (Carlson et al. 2002). The corpus wasdiscourse-parsed at sentence level with Soricut and Marcu’s (2003) sentence-level discourse parser. Then 1,500sentences with Nucleus-Satellite (NS) and Satellite-Nucleus (SN) order were randomly selected. This resulted in asample of 843 NS, and 657 SN sentences. DU length, in syllables, was measured using a customized version ofTEX’s hyphenation algorithm.

Figure 1 presents a statistical summary of the data. It shows that whensentences come in the NS order, there is no significant difference between themean lengths of their DUs. In the SN order, however, a significant differencebetween the DU lengths is observed. Satellites tend to be much shorter and nucleiare often much longer in the SN order. An independent samples t-test on the meanDU lengths for the two groups showed that the differences observed in the meansare statistically significant with a 99% confidence interval.

ELABORATION and ATTRIBUTION are predominant in this corpus. Butwhat is interesting is that ELABORATION is by far the most common relation in theNS group, while ATTRIBUTION is the most common relation in the SN group.Although it may seem that the unmarked order for ATTRIBUTION is SN, statisticalanalysis reveals otherwise, suggesting that prosodic weight, at least partially,influences the order. A pattern similar to Figure 1 is also observed in theATTRIBUTION subgroup. NS sentences often have DUs with more or less the sameprosodic weight, while SN ones often have DU1s that are significantly shorter than

DU2s (cf. (1) and (2) below).

(1) [USI Far East will hold a 60% stake in Luzon Petrochemical,]N [according to papers signed with the Philippinegovernment’s Board of Investments.]S

(2) [Says Ms. Raines,]S [“judgment confirms our concern that the absence of patent lawyers on the court could provetroublesome.”]N

This research shows that the traditional constraint of requiring prosodically heavy units to come lastgenerally holds. However, no obvious weight threshold that triggers the application of the constraint is observed;rather, we see a gradual strengthening of the constraint as the difference between the DU weights gets larger.Prosodic weight is not the sole determinant of DU order, however. Obviously, other constraints from syntax,semantics, and information structure play a role here with potentially conflicting requirements. The paper argues thata soft-constraint satisfaction approach (Bistarelli 2001) to conflict resolution at interfaces is necessary for a theory ofgrammar. The incorporation of this approach in the theory is also discussed.

References

Bistarelli, S. (2001). Soft Constraint Solving and Programming: A General Framework. Ph.D. thesis, Università di Pisa.Carlson, L., D. Marcu, and M. E. Okurowski (2002). RST Discourse Treebank. FTP File http://www.ldc.upenn.edu/. LDC2002T07.Mann, W. C. (1984). Discourse structure for text generation. In Proceedings of the 22nd Annual Meeting. Association for Computational

Linguistics.Mann, W. C. and S. A. Thompson (1986). Rhetorical structure theory: description and construction of text structures. In G. Kempen (Ed.),

Natural Language Generation: New Results in Artificial Intelligence, Psychology and Linguistics, Proceedings of the ThirdInternational Workshop on Text Generation, Nijmegen, The Netherlands.

Mann, W. C. and S. A. Thompson (1988a). Rhetorical structure theory: a theory of text organization. In L. Polanyi (Ed.), The Structure ofDiscourse, Number 3. Norwood, N.J.: Ablex.

Mann, W. C. and S. A. Thompson (1988b). Rhetorical structure theory: Toward a functional theory of text organization. Text 8 (3), 243-281.Soricut, R. and D. Marcu (2003). Sentence level discourse parsing using syntactic and lexical information. In Proceedings of the Human

Language Technology & North American Association for Computational Linguistics Conference (HLT/NAACL), Edmonton, Canada.

NS SN

Order

0.0

5.0

10.0

15.0

20.0

25.0

30.0DU1 Length

DU2 Length

Figure 1: Mean DUlengths in NS and SNgroups

27


Converging Evidence II: More on the Association of Verbs and Constructions

Beate Hampe, Stefan Th. Gries, Doris SchoenefeldFriedrich-Schiller Universität Jena, Southern Denmark University, and Ruhr-Universität Bochum

Due to the usage-based orientation (Kemmer & Barlow 2000) of the constructionist approach to language,theorizing within the paradigm is increasingly based on the empirical evidence provided by naturalistic corpus data.The current paper reports on a study which represents the first attempt so far to contrast diverse kinds of empiricaldata on the use of one construction gained by both corpus research and experimental work and thus contributes toon-going discussions from a decidedly methodological angle.

The interaction between verbs and the constructions they occur in is currently one of the hotly debated issues.Regarding the use of corpus data, two different approaches can be identified in the preceding research: On the onehand, scholars (Goldberg 1999, 2004; Boas 2003, among others) rely on raw-frequency data, thereby automaticallyattributing high importance to high-frequency items. Stefanowitsch & Gries (2003) and Gries & Stefanowitsch(2004), on the other hand, develop a more refined method of quantifying a word’s importance for (investigating thesemantics of) a given construction as the degree of attraction/repulsion between verbs and constructions in terms of“collostruction-strength”.

On the basis of data from the ICE-GB as well as the BNC Sampler, we investigate a special case of complex-transitive predicates, namely the as-predicative, exemplified in (1). This construction is an interesting case in pointsince the raw-frequency approach and the collostruction-based approach make conflicting predictions concerningthe verbs most relevant to an analysis of this construction.

(1) a. Politicians are regarded – indeed, regard themselves – as being closer to actors. [ICE-GB]b. I never saw myself as a costume designer. [ICE-GB]

A first series of studies on the topic is based on the data provided by the ICE-GB. Here, see and regard, for example,were the most and third most important verb lemmas from a frequency-based perspective, but the ranks werereversed in the collostructional analysis, (thus also shifting the focus of the semantic analysis). We tested suchdiverging predictions from the corpus data by means of a sentence-completion experiment analogous to those usedin priming studies. On the basis of the complete collostruction analysis of the as-predicative in the ICE-GB, four setsof verbs, crossing high- and low-frequency verbs and high- and low-collostruction-strength verbs, were assembledand the sentence beginnings with verbs from these four sets presented to native speakers of British English forcompletion. The results clearly supported the collostruction-strength ranking, with high collostruction-strength beinga significantly better predictor of the frequency of as-predicatives in the sentence completions than high frequency(see the row totals vs. the column totals in table 2).

Frequency: high Frequency: low Row totalCollostruction strength: high 0.477 0.447 0.462Collostruction strength: low 0.164 0.119 0.14

Column total 0.328 0.281 0.304

Table 1: Mean relative frequencies of as-predicatives

In this paper, we will not only further substantiate our discussion of methodological issues in corpus-based researchby comparing and supplementing the results of the ICE-search with those gained from the BNC Sampler, but alsoamend the experimental part of our first study by a self-paced reading experiment, designed on the basis of the moreextended corpus data. We report the results of this additional experiment and relate its findings (i) to more specificquestions pertaining to the as-predicative in particular, and (ii) to the theoretical implications and methodologicalaspects following from this ‘converging evidence’ paradigm.

ReferencesBarlow, Michael and Kemmer, Suzanne. 2000. Usage-Based Models of Language. Stanford: CSLI PublicationsBoas, H.C. 2003. A Constructional Approach to Resultatives. Stanford, CA: CSLI Publications.Goldberg, A.E. 1999. The emergence of the semantics of argument structure constructions. In: B. MacWhinney (ed.). The Emergence of Language. Mahwah, NJ:

Lawrence Erlbaum, p. 197-212.Goldberg, A.E. 2004. Learning argument structure generalizations. Cognitive Linguistics.Gries, St.Th. & A. Stefanowitsch. 2004. Extending collostructional analysis: A corpus-based perspective on ‘alternations’. International Journal of Corpus Linguistics

9(1): 97-129.Gries, St. Th., Hampe, B. & Schönefeld, D. Converging Evidence: Bringing together experimental and corpus data on the interaction between verbs and constructions.

Paper to be presented at the Third International Conference on Construction Grammar.Stefanowitsch, A. & St.Th. Gries. 2003. Collostructions: Investigating the interaction between words and constructions. International Journal of Corpus Linguistics

8(2): 209-243.

28


Emotional Context is Stored with Language Forms:Evidence from Psychophysiological Studies of Bilingual Speakers

Catherine L. HarrisBoston University

Language has long been celebrated for being an abstract rule system, the natural algebra of the mind.Traditional theories propose that extralinguistic context is stripped away during learning, leaving coremeanings and context-independent grammatical rules. In contrast, the “radical contextualization view”proposes that language forms (words, phrases, grammatical constructions), are frequently stored withtheir contexts of use. On this view, context-independence is the exception, not the norm.

The importance of context for word meaning and phrasal/sentence grammar has been documentedby cognitive linguists and those working within the framework of construction grammar. This paperexamines a little-studied, but tremendously important aspect of language context: the emotional context oflanguage learning and use.

An implication of radical contextualization is that the emotional context is stored as part of words’and phrases’ long-term memory representations. It follows then that encountering emotional phrases willnot just activate their meaning, but will lead to emotional arousal. Emotional arousal can be quantified bymonitoring electrodermal response (i.e. skin conductance). Indeed, heightened skin conductance responseto emotional phrases has been found by many researchers. The interesting prediction is that skinconductance will be greater for emotional phrases heard in a first language compared to phrases heard in asecond language. A second prediction is that this will hold regardless of cultural beliefs about whichlanguage is more emotional. For example, Spanish-English bilinguals usually claim that Spanish is moreemotional than English, while Mandarin-English bilinguals often claim that English is more emotionalthan Mandarin. It was hypothesized that Mandarin speakers’ perceive North American culture to bepermissive about expressing emotions verbally. Physiological measurements were predicted to besensitive to language of childhood learning, not adult views about permissibility of emotional expression.These hypotheses were tested by studying Mandarin-English and Spanish-English speakers who hadlearned their second language (English) in early childhood or after age 12. Speakers listened to a varietyof emotional expressions and neutral phrases while their skin conductance was monitored. Skinconductance was higher for emotional expressions than for neutral expressions only for languages whichhad been acquired in childhood. The exception concerned taboo words (sexually related terms and cursewords), which demonstrated heightened skin conductance responses even in a second language.Childhood reprimands showed the strongest language effect, eliciting large skin conductance responses inthe first language, but not in the second language.

These results extend prior work suggesting that a first and a second language have separate mentalrepresentations. It contributes to the radical contextualization hypothesis because emotional arousal wasspecific to language form, not abstract meaning. This is consistent with the proposal that the emotionalcontext of utterances is stored with language forms in long term memory. The brain systems whichmediate emotional arousal are relatively well understood. Measuring the emotional arousal accompanyinglanguage is thus a valuable empirical method for studying extralinguistic emotional context andphysiological aspects of language.

29


The Middle Voice in Romanian: A Case of Conceptual Arrangement

Anne-Marie HartensteinRice University

This paper offers a systematic analysis of the two middle constructions se and isi in Romanian, as shown in (1)and (2), using a cognitive approach. In Romanian, there is more than one way to express the middle voiceconstruction and this can be considered a very significant language internal variation that needs to be accounted forin a more systematic way. The claim that the two middle voice constructions cover the same semantic spectrum, asthey were taken to be by previous linguists and traditional grammarians, is not acceptable. It will be shown that thetwo constructions differ in the semantic fields that they cover, in the grammatical associations and most important inthe way speakers conceptualize them. The main questions that this paper attempts to answer to are the followingones: What are the conceptual lines speakers follow when they use the two middle voice constructions? Secondly,what role plays the conceptualizer within the conceptualization of the two middle voice constructions in Romanian?

This paper proposes that the best way to explain the major differences between these two middle voiceconstruction is by using Langacker’s (1987, 2000) notions of Viewing (Conceptual) Arrangement and ViewingDistance. It will be shown that the middle construction se activates a more Optimal Conceptual Arrangement;whereas the middle construction isi imposes a more Egocentric Conceptual Arrangement. I have based my analysison two types of data, a self-built corpus of Romanian from two different genres, narrative and newspaper, and a self-conducted experiment. The experiment consists of 37 non-audio video-clips which were presented to 13 Romaniannative speakers. It included mainly verbs that can appear both with se and isi middle construction. The subjects hadto describe the video-clips and unconsciously make use of one or the other construction.

From the corpora, one can conclude that there are main differences in the verbal frames that each middleconstruction applies to. The isi-middle construction occurs more frequently with speech verbs, cognitive verbs,intentional perception verbs, and emotional verbs; whereas se-middle construction occurs most often in associationwith spontaneous event verbs and movement verbs including translational motion verbs. Moreover, only the se-construction appears very frequently with impersonal constructions and spontaneous events. The abovementioneddifferences in lexical associations and grammatical constructions can be seen as one type of evidence to theabovementioned difference in Viewing Arrangements of the two middles. Due to lack of visual acuity imposed bythe distance the conceptualizer in se-construction has restricted access to the immediate scope, preventing him/herfrom establishing links with ground elements; therefore impersonal and general usages will be found very often tobe constructed with se. Another main difference between these two constructions that is noticeable from both thecorpus and the experiment is that, in the case of isi-construction, the object of conceptualization is seen as anaffected participant and put in profile as a primary landmark; whereas in the case of se the object ofconceptualization is not even expressed or appears as a location. Finally, the fact that the se-construction activates amaximal scope of conceptualization (the conceptualizer modifies the whole event); whereas the isi constructionactivates an immediate scope of conceptualization (the conceptualizer modifies the profiled element in the event:location, instrumental) is another evidence in favor of the different Viewing Arrangements activated by the twomiddles [as in (3) and (4), respectively].

Hence, this paper makes two major contributions. First it takes an empirical and experimental data basedapproach to language and so to the middle voice construction finding main semantic and cognitive differencesbetween the two middle constructions. Secondly, it discusses major usages that were overlooked by the previousnon-empirical and based on introspection approaches to the notion of middle voice in Romanian. In conclusion, inorder to give a better account of the Middle voice in Romanian a usage based analysis has to be provided. Thus, areconsideration of the functions and semantics of the Middle voice in Romanian is called for.

(1) Dar cand ninsoarea [[se]] inteti si se transforma in viscolbut when snow MM got stronger and MM transform in stormBut when the snow got stronger…

(2) El [[isi ]]aduse aminte de stiuca, pe care o chema ...He MM brought mind of the pike…..He remembered the pike

(3) [[Se]] da insistent cu cremaShe puts cream on cream vigorously

(4) [[Isi]] pune crema pe tivia piciorului stang si apoi pe tivia piciorului dreptShe puts cream on the upper part of her left leg and then on the upper part of her..

ReferencesLangacker, Ronald W. 1987. Foundations of Cognitive Grammar. Vol.1. Theoretical Prerequisites. Stanford: Stanford University Press.Langacker, Ronald, W. 2000. Grammar and Conceptualization. Berlin/New York:Mouton de Gruyter.

30

Friday, 13:00 – Aurora Room REFERENCE & COHESION

A Cross-linguistic Exploration of Demonstratives in Interaction:With Particular Reference to the Context of Word Search

Makoto Hayashi and Kyung-eun YoonUniversity of Illinois, Urbana-Champaign

Demonstratives have been a popular object of inquiry in cross-linguistic/typological studies (Anderson andKeenan 1985; Himmelmann 1996; Diessel 1999; Dixon 2003). This study aims to contribute to this body ofresearch by exploring usages of demonstratives that have received little attention in the past literature; that is, uses ofdemonstratives as some sort of “filler words” in the context where speakers encounter trouble in recalling a wordduring the course of utterance production. The study documents the range of forms and functions of demonstrativesacross diverse languages that are used during word search, and explores the question of why demonstratives, amongall linguistic items, are utilized as a tool to deal with word-finding trouble in utterance production.

Based on data from spontaneous spoken discourse in 12 typologically diverse languages, we first describe twosubtypes of usage observed cross-linguistically — the “placeholder” use and the “interjective hesitator” use.Placeholder demonstratives are those used as “substitutes” that are produced to occupy the syntactic slots of thelexical items that have momentarily eluded speakers. The referents of the demonstratives are subsequently specifiedas a result of word search (cf. (1)):

(1) [Japanese]

kinooare tabeta no. nandakke, a fondu.yesterday DEM. ate PTCL what.is.it oh fondue“I ate that thing yesterday… What is it. Oh, fondue.”

We show that different languages employ different classes of demonstratives (proximal, medial, distal) for thisusage, and that some languages allow the use of placeholder demonstratives even at the level of morphology, i.e., fora missing morpheme within a morphologically complex word.

Interjective hesitator demonstratives are those used in much the same way as hesitation signals like uh inEnglish. A crucial feature that distinguishes this usage from the placeholder usage is that interjective hesitatordemonstratives are not produced as syntactic constituents of utterances-in-progress. As paralinguistic signals, theycan appear anywhere during the course of utterances irrespective of the syntactic environments in which theyhappen to be placed (cf. (2)):

(2) [Mandarin]

ta zhei-ge::, ta shuo nei-ge zhengshi de hai meiyou xia:laihe DEM. he said that formal GEN still not come.down“He, uhm, he said the official one hasn’t been issued yet.”

To answer the question of what makes demonstratives particularly useful as a resource for managing word-finding trouble, we show that the function of demonstratives to propose particular configurations of “participantaccess” to the referent (shared/non-shared knowledge) provides a useful resource for the participants to organizespeaker-hearer interaction during word search. That is, a speaker can use different forms of demonstratives toindicate whether the searched-for referent is known only to the speaker, or known to both the speaker and the hearer,and by doing so, s/he can invite or not invite the hearer’s co-participation in the search for the missing word.

We conclude by noting that, by examining “dysfluencies” in spontaneous speech, which have beenneglected as irrelevant in many approaches to grammar, we can in fact learn a great deal about how grammar worksin actual language use in interaction.

References

Anderson, S. R. and Keenan, E. L. 1985. “Deixis.” In Shopen, T. (ed.), Language Typology and Syntactic Description, Vol. 3, 259-308.Cambridge: Cambridge University Press.

Diessel, H. 1999. Demonstratives: Form, Function, and Grammaticalization. Amsterdam: John Benjamins.Himmelmann, N. 1996. “Demonstratives in narrative discourse: A taxonomy of universal uses.” In Fox, B. (ed.), Studies in Anaphora, 205-254.

Amsterdam: John Benjamins.Dixon, R. M. W. 2003. “Demonstratives: A cross-linguistic typology.” Studies in Language 27. 61-112.

31


The Effects of Cohesion Anomalies on Children’s Narrative Strategies

Maya Hickmann & Phyllis SchneiderCNRS/Université Paris 5 and University of Alberta

Previous research has examined two related but distinct aspects of children’s narratives: coherence and cohesion.Although the linguistic marking of cohesion is acquired late (Hickmann, 2003), five-year-olds can correct referentialanomalies in stories when awareness is not required (Hickmann & Schneider, 1993, 2000; Schneider, Williams &Hickmann, 1997). Children also construct story schemata (Mandler, 1978; Stein, 1982; Stein & Albro, 1997), e.g., theyre-organize structurally anomalous stories into a canonical form. Some structural anomalies, however, affect cohesion,thereby confounding these two variables (Garnham, Oakhill & Johnson-Laird, 1982). The aim of the present study is tosystematically examine children’s sensitivity to both types of anomalies.

MethodThere were three groups of 24 children each, aged five, seven, and ten years. Eighteen stories were constructed, eachabout a protagonist and another (target) referent in one episode of six units: SE=Setting, IE=Initiating Event (targetintroduction), IR=Internal Response, AT=Attempt, CN=Consequence, RE=Reaction. There were nine versions of eachstory. Three contained a displaced CN, three a displaced IE, three no coherence anomaly. Since displacements resultedin cohesion anomalies, the relative effect of cohesion/coherence anomalies was examined by constructing threecohesion versions of all stories: cohesive versions contained only appropriate expressions, low cohesion versions aninappropriate target introduction, highly disrupted versions two (first and subsequent) inappropriate target expressions.Each child heard one version of each story, two stories in each version (randomized order), retelling them to a naivelistener.

SE: Once there was a monkey who was very silly. IE: One day he was running around and he saw a balloon in the clouds. IR: Hewas in a good mood and he wanted to play with the balloon. AT: So he climbed up high and he grabbed the balloon. CN: But hisfoot slipped and he let go of the balloon. RE: He was really upset.

ResultsChildren react in three ways with displacements: maintaining the displaced unit in non-canonical position; moving itback to its canonical position; constructing two episodes. Five-year-olds often maintain or move displaced units,while older children often create two episodes. At all ages children move displaced IEs more frequently than CNs,whereas double episodes are as likely with both displacements.

Children of all ages also frequently introduce two target referents with either displacement. At all ages doubleintroductions are more frequent with highly disrupted versions than with other cohesion versions. Coherence andcohesion also interact: coherence anomalies only increase the likelihood of double introductions when cohesion ishighly disrupted.

In conclusion, children are sensitive to both coherence and cohesion. With increasing age they rely ondifferent strategies to handle displacements, but at all ages they are sensitive to the significance of referentintroductions. At all ages displacements and/or high cohesion anomalies also invite children to invent new referentsby transforming inappropriate indefinite forms into appropriate introductions. From seven years on childrensimultaneously produce double episodes and double introductions, showing that they become more efficient inrestoring coherence and cohesion.

ReferencesGarnham, A., Oakhill, J., & Johnson-Laird, P. (1982). Referential continuity and the coherence of discourse. Cognition, 11, 29-46.Hickmann, M. (2003). Children’s discourse: person, space and time across languages. Cambridge University Press.

Hickmann, M., & Schneider, P. (1993). Children's ability to restore the referential cohesion of stories. First Language, 13, 169-202.Hickmann, M., & Schneider, P. (2000). Cohesion and coherence anomalies and their effects on children's referent introduction in narrative retell.

In M. Perkins & S. Howard (Eds.), New directions in language development and disorders (pp. 251-260). New York, NY: KluwerAcademic/Plenum Publishers.

Mandler, J. M. (1978). A code in the node: the use of a story schema in retrieval. Discourse Processes, 1, 14-35.Schneider, P., Williams, B., & Hickmann, M. (1997). Awareness of referential cohesion in stories: A comparison of children with and without

language/learning disabilities. Journal of Speech-Language Pathology & Audiology, 21(1), 8-16.Stein, N. L. (1982). The definition of a story. Journal of Pragmatics, 6, 487-507.Stein, N. L., & Albro, E. R. (1997). Building complexity and coherence: children’s use of goal-structured knowledge in telling stories. In M. G.

W. Bamberg (Ed.), Narrative development: Six Approaches (pp. 5-44). Mahwah, NJ: Lawrence Erlbaum.

32


Chaining of Metonymies: A Corpus-based Approach

Martin HilpertRice University

Cognitive Linguistics has seen a recent interest in expressions in which several metonymies interact(Nerlich & Clarke 2001, Geeraerts 2002, Barcelona 2003). While these contributions have aimed at a theoreticaldescription of isolated examples, this paper offers a corpus-based approach to chained metonymies. Consider thecontrast in examples (1) and (2).

(1) Keep your eye on the paper!(2) Keep your eye on the real issue!

Example (1) means 'Keep watching!' whereas example (2) means 'Keep paying attention!', with no referenceto visual perception. This could lead the researcher to posit two different metonymies, namely EYE FORWATCHING and EYE FOR ATTENTION. In this paper I argue for a different solution, namely chainedmetonymies. In example (2) the metonymy EYE FOR WATCHING is extended by a second metonymyWATCHING FOR ATTENTION. The question what types of metonymies interact can only be settled throughinvestigation of authentic data. The present analysis takes the following steps:

a) The starting point for a corpus-linguistic study of metonymy is the choice of a productivemetonymic source domain. The present study investigates body parts, containers and timespans,since these domains are attestedly rich sources for metonymic expressions. Lexical items such aseye, cup, and hour are extracted from a 10 million word sample of the BNC.

b) A categorization of these examples into literal and figurative usages allows an assessment of theratio of literal and figurative usage. In the investigated data, figurative usages made up more than40% for all body lexemes, but less than 10% for most container and timespan lexemes.

c) The third step is the analysis of metonymy in the figurative examples. For all investigated lexicalitems, more than 65% of the data is organized in a limited array of patterns. These patterns expose aclose correlation of form and meaning. For example keep POSS eye on NP triggers EYE FORWATCHING in many cases. Thus, collocation is a major clue to the interpretation of metonymicexpressions. However, there is no one-to-one relation from pattern onto meaning. As showed above,keep POSS eye on NP can refer to both 'watching' and 'attention'.

d) An investigation of polysemous patterns like keep POSS eye on NP provides us with an insight ofwhat metonymies link up with other metonymies. With the lexical item eye, EYE FOR WATCHINGserves as a base for several extending metonymies like WATCHING FOR ATTENTION,WATCHING FOR INTENTION, and WATCHING FOR DESIRE. With the lexical item cup,CONTAINER FOR UNIT OF VOLUME is regularly extended by UNIT OF VOLUME FORCONSUMPTION.

e) Abstracting away from the particular instances, we can identify broad differences between basemetonymies as opposed to extending metonymies. In the investigated data, base metonymies are mostoften part-whole relationships. By contrast, extending metonymies tend to be categorialrelationships.

I draw three conclusions from this study. First, metonymic expressions are organized in patterns – these patternstrigger specific metonymies. Second, lexical items are associated with a limited number of base metonymies, whichform the basis for extending metonymies. Third, base metonymies and extending metonymies employ differentkinds of conceptual relations.

References

Barcelona, A. 2003. 'Metonymic chains and the inferential role of metonymy'. Paper presented at ICLC 2003 Logroño, Spain.Geeraerts, D. 2002. 'The interaction of metaphor and metonymy in composite expressions.' In Dirven & Pörings (eds), Metaphor and Metonymy

in Comparison and Contrast. Berlin & New York: Walter de Gruyter. 435-468.Nerlich, B. & D. Clarke. 2001. 'Serial Metonymy. A study of reference-based polysemisation.' Journal of Historical Pragmatics 2/2. 245-272.Seto, Ken-ichi. 1999. 'Distinguishing Metonymy from Synecdoche.' In Panther & Radden (eds), Metonymy in Language and Thought.

Amsterdam & Philadelphia: John Benjamins. 91-120.

33

Sunday, 14:00 – Aurora Room DISCOURSE CONNECTIVES

Versatile Subordinator e›ke¤ in Adioukrou: A Cognitive-Typological Analysis

Kaoru Horie, Prashant Pardeshi, and Kaul GuyTohoku University

Polyfunctionality of grammatical markers is a norm rather than exception cross-linguistically (cf. Noonan1997), as has been well demonstrated by case syncretism phenomena (e.g. English to encoding goal anddative recipient) among others. In the domain of subordinate clauses, it is rather common for the samemarkers to code 'relative clause' and 'complement clause', as evidenced by English that and French que.This follows naturally from the fact that both types of clauses occur in argument positions. However, it israther uncommon for a single marker to code 'adverbial clause', functioning as adjunct, as well as 'relative'and 'complement' clauses, both functioning as argument. This indeed is the case with the subordinatore›ke¤ in Adioukrou (a Niger-Congo language, a Kwa language of Co^te d'Ivoire). Consider (1)-(3):

(1) E›gN› a› [ e›ke¤e›b si›ka› kpa¤tro›wa› a› ]person the AOR-buy gold ring the'the person who bought the golden ring.' (Relative clause)

(2) wE›l E›sE› [e›ke¤ E›j E›s kç›tç›kç› b÷ow÷o¤w ]one say our fatherlord FUT-come'the rumor that our Lord would come.' (Complement clause)

(3) e›ke¤ si¤kplô a› a$gb a$fN$n e¤s a$ / o$w b÷i¤s(m¤) E$gN$ lu¤wsin the AOR-increase pass VerbP the it HAB-bring man death

'If the sin became great, it brings death to anyone.' (Conditional adverbial clause)

This paper argues that this apparently 'rampant' polyfunctionality of Adioukrou e›ke¤ is by no meansanomalous in view of Radical Construction Grammar (henceforth RCG; Croft 2001). RCG proposes theuniversal conceptual space (or continuum) of complex sentence types covering four pivotal constructions,i.e. relative clauses, complements, coordination, and adverbial clauses. The cross-constructionalcontinuity between these four construction types is motivated by the existence of 'intermediate'constructions observed cross-linguistically (e.g. 'internally headed relative clause' mediating 'relativeclause' and 'complement').

The Adioukrou subordinator e›ke¤ arguably covers the conceptual space demarcated by threeconstruction types, i.e. relative clause, complement, and adverbial clause, confirming that these threeconstructions do indeed form a continuum, as suggested by RCG. Its ability to code conditional adverbialclauses as well as complements is partially analogous to an English conditional marker if, which can serveto mark embedded question complements (in the sense of whether), though if doesn't code relativeclauses.

It is an intriguing question how Adioukrou native speakers resolve the potential ambiguity of e›ke¤-marked clauses contextually. Our preliminary analysis of the Adioukrou narrative texts reveals that iconicword order (cf. Haiman 1985) crucially serves to distinguish the 'conditional adverbial' (adjunct)interpretation, on the one hand, and the 'relative/complement' (argument) interpretations, on the other.Specifically, the former ('adjunct') interpretation is iconically signaled by the sentence-frontal (non-embedded) position (3), while the latter ('argument') interpretations are licensed by the sentence-medial(embedded) positions (1, 2).

References

Croft, William. 2001. Radical Construction Grammar. Oxford: Oxford UP.Haiman, John. 1985. (ed.) Iconicity in Syntax. Amsterdam: John Benjamins.Noonan, Michael. 1997. Versatile nominalizations. In: Bybee, Joan, John Haiman, and Sandra A. Thompson. Essays on Language Function and

Language Type. Amsterdam: John Benjamins, 373-394.

34


What We Can Learn from Co-occurrences of Discourse Connectives:A Corpus-based Study

Ben HutchinsonUniversity of Edinburgh

IntroductionDiscourse connectives (DCs) are expressions which explicitly signal rhetorical relations, e.g. but or despite this. Thispaper considers how the empirical study of DCs can contribute to our understanding of discourse structure. Webegin by testing the following hypothesis:

Homogeneity hypothesis: DCs are more likely to appear near DCs signalling similar relations. E.g. DCs signallingtemporal relations are more likely to occur near other DCs signalling temporal relations.

We then apply our data to evaluating claims that unless and provided that differ in polarity, and that they signalcircumstantial, rather than causal, rhetorical relations (Montolio, 2000).

MethodOur study focuses on 73 "structural connectives", which are a subclass of DCs which act syntactically likeconjunctions, e.g. because or providing that. To study these, we use a corpus compiled automatically from theBritish National Corpus and the world wide web. Automatic techniques are used to extract example sentences foreach DC.

Unlike previous related studies (for example Bestgen et. al. 2003), our study uses co-occurrences of DCs asa measure of their distributions. In particular, we use cases where DCs occur in conjoined clauses, as in (1) and (2).

(1) Although Pat was hungry, she nevertheless worked late.(2) Pat was hungry, but she worked late nevertheless.

We show that such co-occurrences of DCs can be classified into six types, depending on whether the DCs i) occur inthe same or different clauses, ii) signal relations between the same discourse units, and iii) signal the same relationor different ones. In cases where two different relations link the same clauses, the relations must be compatible, forexample compare (3) and (4).

(3) After Pat left, Chris left too.(4) #After Pat left, Chris left beforehand.

Statistics on DC co-occurrences are derived using automatic sentence parsing and analysis techniques. We show thatthe error rates in automatically extracting and analysing sentences are low enough for reliable inferences to bedrawn.

ResultsWe find that on the whole the homogeneity hypothesis is supported. However at times there is conflicting evidence,in which syntax is a relevant factor. For example, a discourse conjunction such as but that signals a negative polarityrelation (e.g. Contrast or Concession) is more likely to appear near discourse adverbials that also signal negativepolarity relations, e.g. nevertheless. However such conjunctions are actually less likely to be subordinate to anotherdiscourse conjunction that also signals a negative polarity relation.

We find that clauses with negative polarity DCs are more similar to unless-clauses than provided that-clauses. However we also find that the distribution of unless-clauses differs from that of typical clauses withnegative polarity DCs. Thus there is only partial empirical support for the previous claims. We find that both unlessclauses and provided that clauses are distributed more like temporal DCs than causal DCs. We interpret this assupport for the claim that unless and provided that signal circumstantial relations rather than causal ones.

References

Bestgen, Y., L. Degand, and W. Spooren (2003). On the use of automatic techniques to determine the semantics of connectives in largenewspaper corpora: An exploratory study. In Proceedings of the MAD 03 workshop on Multidisciplinary Approaches to Discourse.

Montolio, E. (2000). On affirmative and negative complex conditional connectives. In E. Couper-Kuhlen and B. Kortmann (Eds.), Cause -Condition - Concession - Contrast: Cognitive and Discourse Perspectives. Berlin: Mouton de Gruyter.

35


On the Conceptual Structure of Indefiniteness

Michael IsraelUniversity of Maryland

Despite years of scrutiny, the English determiner any remains mired in controversy over the distinctionbetween 'free choice' (FC) uses of any, as in (1), and polarity sensitive (PS) uses, as in (2).

(1) a. Anyone can tell you that bats have wings.b. He'll eat (almost) anything!c. It could be any minute now.

(2) a. There isn't anyone home. (but *Anyone isn't home)b. Do you want anything?c. If we had any sense, we'd have left long ago.

Much of the literature concerns whether these uses involve distinct lexical senses (Vendler 1967; Horn 1972;Ladusaw 1979; Dayal 1998), and whether these senses are properly analyzed as existential or universal quantifiers,or as a special sort of indefinite (Giannakidou 1999; Israel 1999; Tovena 1998; Horn 1999, 2000).

This paper brings together typological, lexical semantic, and usage-based evidence to argue that any in all itsuses is an indefinite scalar operator, that differences between them reflect alternate ways of construing a scale, ratherthan distinct lexical senses, and that these types of construals apply systematically to other English indefiniteconstructions, specifically to the NPIs ever and at all, and to PPIs based on some. In the spirit of Langacker (2001), Idefine indefinite constructions as referring expressions which present an entity as novel to, or otherwise unactivatedby, an addressee. Haspelmath (1997) puts the phenomenon in typological perspective, proposing a semantic mapwith nine indefinite functions (3), and showing that while forms like any are cross-linguistically common, theirprecise distributions vary widely and often overlap with other constructions.

(3)(1)specificknown

(2)specificunknown

(3)irrealisnon-specific

(4)question

(6)indirectnegation

(7) direct negation

(8)comparative (9)

free choice

(5)conditional

I examine this overlap in English indefinite constructions with any, some, ever, and at all, and show thatwhile each of these forms participates in a variety of lexically-specific, item-based constructions, they are united intheir basic scalar semantics: any and ever profile scalar endpoints and serve an emphatic function; some profiles alow scalar value and serves an attenuating function (cf. Israel 1996). The basic difference between PS and FC usesof any reflects different ways of construing the contrast set of a scalar endpoint. FC, 'quodlibetic' uses prompt theconstruction of a scale of possible types an indefinite instance might take; PS uses prompt a scale of quantities anindefinite instance might have: roughly then, PS any = 'even one' and FC any = 'of even any kind' (cf. Horn 1999).This contrast between kind and quantity scalar construals applies to determiner constructions with both some andany, but is systematically absent with adverbial indefinites like ever and at all, apparently because occasions anddegrees come in different quantities but in different kinds.

The evidence found here is consistent with a view of grammar that is massively item-based (Tomasello2003); but it also suggests that there are regular and relatively abstract patterns of usage which hold across item-based constructions. The results demonstrate the usefulness of Haspelmath's map as a basis for detailed semanticanalysis, and show it might be enriched by incorporating specific parameters of indefinite construal.

ReferencesDayal, V. 1998. “Any as Inherent Modal”, Linguistics and Philosophy 21:433-476.Giannakidou, A. 1998. Polarity Sensitivity as (Non)veridical Dependency. Amsterdam: John Benjamins.Haspelmath, M. 1997. Indefinite Pronouns. Clarendon: Oxford University Press.Horn, L. R. 1972. On the Semantic Properties of Logical Operators in English. PhD dissertation, distributed by IULC, 1976.Horn, L. R. 1999. “any and (-)ever: Free choice and free relatives.” Paper presented at IATL, Haifa, Israel.Horn, L. R. 2000. “Pick a Theory (not just any theory): Indiscriminatives and the free choice indefinite.” In Horn & Kato, eds., Negation and Polarity: syntactic and

semantic perspectives. Oxford: Oxford University Press. pp. 147-192.Israel, M. 1996. “Polarity Sensitivity as Lexical Semantics.” Linguistics & Philosophy. 19: 619-666.Israel, M. 1999. “Some and the Pragmatics of Indefinite Construal.” Berkeley Linguistics Society 25.Langacker, R. 2001. "Viewing & Experiential Reporting in Cognitive Grammar." A.S. da Silva (ed.), Linguagem e Cogniçao: A Perspectiva da Linguística Cognitiva,

pp. 19-49. Braga: Associaçao Portuguesa le Linguîstica & Universidade Católica Portuguesa, Faculdade de Filosofia de Braga.Ladusaw, W. 1979. Polarity Sensitivity as Inherent Scope Relations. New York & London: Garland Publishing. (1980)Tomasello, M. 2003. Constructing a Language: a Usage-Based Theory of Language Acquistion. Harvard University Press.Tovena, L. 1998. The Fine Structure of Polarity Items. New York: Garland Press.Vendler, Z. 1967. Linguistics in Philosophy, Ithaca, NY; Cornell University Press.

36


Does Metaphor Motivate Russian Aspect?

Laura A. JandaUniversity of North Carolina at Chapel Hill

I have developed a metaphorical model of Russian aspect (recently published as a pedagogical piecein Slavic and East European Journal and forthcoming as a theoretical piece in the journal CognitiveLinguistics) that identifies fourteen groups of properties that human beings experience when contrastingdiscrete solid objects with fluid substances, showing a strong parallelism between these properties and thedistinctions between perfective and imperfective aspect in Russian. For example, discrete solid objectshave firm edges but fluid substances do not, corresponding to the way in which perfective events aretemporally bounded, but imperfective events are not. Discrete solid objects are unique individuals withparts, but fluid substances are uniform and continuous, corresponding to the way that perfective aspectmarks unitary events, whereas imperfective typically does not. Discrete solid objects are impenetrable andthus cannot occupy the same space, but fluid substances are penetrable and can be mixed, correspondingto the perfective’s tendency to mark sequenced events as opposed to the imperfective’s capacity todescribe simultaneous events

When I have presented this model at scholarly gatherings, native speakers of Slavic languages in theaudience have declared that it represents a psychological reality. At the behest of my colleagues, I haveundertaken some preliminary investigations, the results of which I will present in this paper. Nativespeakers of Russian have been presented with various combinations of realia (a discrete solid toy on oneplate vs. a pile of sand on another), tables of properties of physical matter, and sentences containingperfective and imperfective verbs and asked to make the associations that they find appropriate. Theyhave not been told that this experiment has anything to do with aspect. One pilot study and twoexperiments have been conducted. The two experiments contrast the results with the properties of matter Ihave detailed in my model (30 participants), as opposed to a “control” group (10 participants) that wasoffered an equally large set of properties associated with the sun vs. the moon. The experiments are beingconducted by graduate students who were not told what the “correct” answers were, in hopes that theywould thus not influence the data collection. I will be working with a statistician to ensure the properhandling of the numbers, but I hope to show that the “control” group (sun vs. moon) gives random data,whereas the properties of matter group (discrete solid object vs. fluid substance) gives statisticallysignificant results suggesting that the model is indeed a reflection of psychological reality. We are stillconducting the experiments and collecting data through the month of May, so I do not have sample datato present at this time. The pilot study did produce significant results, but was flawed (mainly due to somecultural differences that needed to be worked out).

37

Saturday, 15:30 – Prairie Room GRAMMATICALIZATION OF BASIC VERBS

Semantic Extensions and Conceptual Shifts:A Corpus Study of ‘Come’ and ‘Go’ Serial Verbs in Japanese and Korean

Kaori Kabata & Jeong-Hwa LeeUniversity of Alberta and Korea Digital University

This paper deals with a corpus study that investigates how the semantic development of a linguistic itemmay bring about shifts in the conceptual saliency of its senses. As a linguistic item, be it lexical orconstructional, undergoes semantic extension, some of its extended senses may get entrenched and gainhigher status in the cognitive organization (e.g., Langacker 1987). Our discussion will be centered aroundserial verbs (SVs) containing deictic verbs ‘go’ and ‘come’, which exhibit traces of extensive semanticextension in many languages (e.g., Emanatian, 1992; Lichtenberk, 1991). Japanese and Korean are noexception.

Both Japanese iku ‘go’ and kuru ‘come’ and Korean counterparts kata and ota have acquired variousspatial and non-spatial usages when used in V-te iku/kuru or V-a/e kata/ota constructions, as illustratedbelow. They include (a) a motion and its direction in physical space, (b) spatial direction from thespeaker’s viewpoint, (c) deixis in perceptual and conceptual space. These two pairs of SVs, however,exhibit subtle differences in their distribution, due to some language-specific semantic constraints:Korean ‘come’ SVs have more limited extension of their usage in conceptual space, as shown in (d).

(a) [J] Taroo-wa gakkoo-ni {hashit-te itta / hashit-te kita}.[K] Inho-ka hakkyo-ey {ttwuy-e kassta / ttwuy-e wassta}.

‘Taro/Inho went/came to school by running.’(b) [J] Fujisan-ga {toozakat-te itta / chikazui-te kita}.

[K] Hanlasan-i {meleci-e kassta / nak-a-wassta}.‘Mt Fuji/Mt. Hanla went farther/came closer (from/to me).’

(c) [J] Taro-wa takumashiku {iki-te itta / iki-te kita}.[K] Inho-ka kanghakey {sal-a kassta / sal-a wassta}

‘Taro/Inho has lived on strongly (from/to the speaker’s viewpoint).’(d) [J] Itami-ga {kie-te itta / kie-te kita}.

[K] kotong-i {salaci-e kassta / * salaci-e wassta}.‘The pain went/is going away.’

Based on semantic analyses and data gathered from spoken and written language corpora, wedemonstrate how the dynamic nature of the semantic distribution of linguistic items may be reflected inthe distribution of their uses by native speakers. The preliminary Japanese data have indicated that‘come’ SVs are used more frequently than ‘go’ SVs both in spoken and written data, counter to ouroriginal expectations based on the cross-linguistic ‘marked’ status of the verb ‘come’ (Shibatani, inpress). Moreover, despite their claimed centrality and basicness (Hasegawa, 1993; Sohn, 1994), spatialusages may not occur more frequently than others. It has also been indicated that the usage distributionsmay vary in different modes of speech. By comparing synchronic patterns of usage distribution betweentwo languages which exhibit both similarity and difference in the semantic distributions, our studyattempts to determine to what semantic and/or conceptual factors such differences may be attributed.

ReferencesEmanatian, Michele. (1992) Chagga ‘come’ and ‘go’: metaphor and the development of tense-aspect. Studies in Language Vol. 16(1), 1-33.Hasegawa, Yoko. (1993) Prototype semantics: a case study of the k-/ik- constructions in Japanese. Language and Communication Vol. 13(1), 45-

65.Langacker, Ronald. W. (1987) Foundations of Cognitive Grammar. Vol. 1, Theoretical Prerequisites. Stanford: Stanford University Press.Lichtenberk, Frantisek. (1991) Semantic change and heterosemy in grammaticalization. Language Vol. 67(3), 475-509.Shibatani, Masayoshi. (In press) Derectional verbs in Japanese. Motion, Direction and Location in Languages: A Volume Dedicated to Zygmunt

Frajzyngier. Amsterdam: John Benjamins.Sohn, Ho-min. (1994) Korean. New York: Routledge.

38


Form Emerges from Use: Why Japanese Speakers Can’t Pass a Wug Test

Terry KlafehnUniversity of Hawaii

It is said that when the surrealist painter Magritte was asked why he had painted the image of apipe above the words “This is not a pipe” he responded, “Because I can’t use it as a pipe”. It is true thatform often follows function. A pipe looks the way it does because of the function it has and the way it isused. However, just because an image or an object looks like a pipe does not mean it is a pipe.

This paper argues that linguistic reality is also determined by function and that linguistic formemerges from language use, independent of the abstract structures of formal analysis. This paperinvestigates the processing of Japanese verbal inflection. It begins by asking the question: How doJapanese speakers mentally store and process verbal inflection? One proposal is that speakers segmentverbs into stems and endings. In the Words and Rules Model (Pinker 1999) the stems of regular verbs arestored in memory and recombined with inflectional endings by the use of formal rules. This model ofprocessing predicts that inflectional error and productivity are structurally determined.

These predictions are tested in this paper. The results of computer searches of a Japanese databasewere analyzed. It was found that; 1) Japanese children (1;5-3;0) do not overregularize irregular verbs (noJapanese *goed.) 2) Most verb errors are stem errors. 3) There are more errors with regulars than withirregulars. 4) There is no pattern of error.

Second, a modification of Vance’s (1991) nonce verb test was used to test the ability of fifty adultnative speakers and a set of fifty instructed learners to recognize inflected forms of novel Japanese verbs.The learners (76% correct) outperformed the native speakers (53% correct). For example, only 8 nativespeakers (16%), but 32 learners (64%) chose hoita as the correct past form of the nonce verb hoku. Thestrong structure-based productivity predicted by the Words and Rules Model was not confirmed.

Finally, in an original experiment, five and six year old Japanese children were provided withillustrated examples (pokemon cartoons) of nonce verbs and asked to produce novel inflected forms. Thechildren understood the task and completed all repetition tasks, but (with few exceptions) failed torespond on the production task. Again, the productivity predicted by Words and Rules was not found.

It is concluded that both children and adult native speakers of Japanese do not segment verbs intostems and endings. It is likely that fully inflected forms are stored in memory, and productivity is relatedto type and token frequency as suggested by Bybee (2001). While Japanese verbal inflection may beeconomically described in terms of stems and inflectional endings, it does not appear that native speakersmake use of verbal stems. Just as Magritte’s image of a pipe cannot be used as a pipe, it appears that thestems of linguistic analysis cannot be used as stems by Japanese speakers.

References

Bybee, J. (2001). Phonology and language use. Cambridge, UK: Cambridge University Press.

Pinker, S. (1999). Words and rules: The ingredients of language. New York: Basic Books.Vance, T. (1991). A new experimental study of Japanese verbal morphology. Journal of Japanese Linguistics. 13, 145-156.

Figure 1

39


Cherokee Positional Verbs

Chris KoopsRice University

Past research on the relation between language and space has dealt extensively with languagesthat code static topological spatial relations by means of adpositions. The conceptual organization ofspace in languages where those same meanings (e.g. ‘in’, ‘on’, etc.) are expressed primarily through verbsis less well researched and has only recently received more attention (e.g., Brown 1994, Ameka &Levinson forthc.). In this talk I will present the positional verb system of Cherokee, a languageinstantiating the latter type. In addition, at a more practical level, I will discuss the empirical methodsused in investigating these spatial relations.

Recent fieldwork on Cherokee, an Iroquoian language of Oklahoma and North Carolina, hasrevealed a formal division between the expression of non-contiguous, projective spatial relations like ‘infront of’ and topological spatial relations like containment or support. While the former relations areexpressed by postpositions, the latter notions are lexicalized in a small class of verbs which can be calledpositional verbs. Among other distinctions, Cherokee positional verbs make a three-way division betweenlocation on the ground, elevated location, and location in a container. For example, note how the first twoof these enter the descriptions in (1) and (2).

(1) gaasgiloo-gi aa-hla nvvychair-LOC 3SG-be.on.raised.surface rock‘The rock is on the chair.’

(2) gaasgiloo-gi agvydiila aa-h nvvychair-LOC in front 3SG-be.on.ground rock‘The rock is (on the ground) in front of the chair.’

The Cherokee system of positional verbs is typologically unusual in that, according to Levinson(2003), languages with small sets of contrastive locative predicates tend to derive these mostly from thehuman posture verbs (‘sit’, ‘stand’, and ‘lie’, see also Newman 2002), while in Cherokee posture verbsplay only a marginal role in the spatial relations system. Also, languages with ‘true’ positional verbs suchas Tzeltal (Brown 1994) typically do not include figure-oriented, classificatory information in the verbs.The Cherokee positional verbs, however, participate in a classificatory verb system reminiscent of thosefound in Athapaskan languages where verb choice is determined by animacy, shape, and consistency ofthe figure.

The main point of this talk are the ramifications of the Cherokee positional verb system forspeakers’ linguistic categorization of space. However, throughout the discussion I will also address theempirical problem posed by the task of investigating spatial conceptualization in a field situation.Eliciting spatial language, as opposed to other grammatical structures, allows the use of largely non-verbal stimuli. Such stimuli, if adequately designed, make it possible to record spatial descriptions inprecisely defined contexts of use. Techniques range from picture book tasks, where a speaker gives verbaldescriptions of situations depicted in drawings or photographs, to complex tasks involving promptedinteraction between two or more speakers. Data gained through such tasks can then be corroborated withdata from naturally occurring, conversational use of spatial terms.

References

Ameka, Felix, and Stephen Levinson. (to appear) Positional and postural verbs. Special issue of Linguistics.Brown, Penelope. 1994. “The INs and OUTs of Tzeltal locative expressions: The semantics of stative descriptions of location.”

Linguistics 32: 743-90.Levinson, Stephen. 2003. Space in language and cognition. Cambridge: Cambridge University Press.Newman, John (ed.). 2002. The linguistics of sitting, standing, and lying. Amsterdam/Philadelphia: Benjamins.

40


A Lot of Quantifiers

Ronald W. LangackerUniversity of California, San Diego

Nominals of the form a lot of N are described in detail using the framework of Cognitive Grammar. Afull description encompasses a synchronic characterization of the construction per se, an assessment of itsrelationship to other constructions, and consideration of its diachronic development with respect to knownprocesses of grammaticization.

One aspect of a synchronic characterization is a description of the meaning of each componentelement—notably a , lot, and of—and how it relates to the meanings it displays in other constructions.Another aspect is to show precisely how component elements combine to form the composite conceptionat each level of grammatical organization. This includes a determination of constituency, the choice ofprofile at each level, and the grounding of nominal elements (lot, N, and the expression as a whole).

The fact that a lot of is a fixed expression (hence a lexical item) does not entail that it is semanticallyor grammatically unanalyzable. Indeed, each component element has a meaning closely and naturallyrelated to its meaning in other uses. Nor does its lexical nature imply its status as a grammaticalconstituent. There is in fact evidence that the constituency is the same as for cases like a convoy of trucksor a flock of geese: [a flock [of geese]], [a lot [of geese]].

(1) (a) She loves ducks and geese. Right now she has several. Geese, that is.(b) She loves ducks and geese. Right now she has a lot. Of geese, that is.(c) *She loves ducks and geese. Right now she has a lot of. Geese, that is.

(2) (a) Geese? {Many/A lot/*A lot of} are eaten every Christmas.(b) She saw a lot, maybe even a whole flock, of geese.

These examples suggest that a lot of does not constitute an unanalyzable quantifier like many, andthat the major constituency break falls between lot and of.

This constituency does not however entail that the expression a lot of geese profiles the“lot”—instead, the composite structure profile is the geese. It contrasts in this respect with a flock ofgeese, where either the flock or the geese can be profiled. This is shown by number marking on thefollowing verb:

(3) (a) A flock of geese {was/were} flying overhead.(b) A lot of geese {*was/were} flying overhead.

Though non-canonical, it is unproblematic in Cognitive Grammar for the composite structure profileof [a lot [of N]] to be inherited from N, since there is no requirement in this theory that the compositestructure profile come from any particular structural position.

A lot of N can be seen as representing a further step along the same path of grammaticizationtraversed to some extent by many other nouns (e.g. handful, cup, barrel, ton, bunch). The steps on thispath can be characterized as instances of metonymy and subjectification. Along another axis, a lot of N isrelated to other quantifying expressions which incorporate the indefinite article: a few N, a little N/ADJ, abit ADJ.

References

Langacker, Ronald W. 1991. Foundations of Cognitive Grammar, vol. 2, Descriptive Application. Stanford: Stanford University Press.––––. 1992. ‘The Symbolic Nature of Cognitive Grammar: The Meaning of of and of of-Periphrasis.’ In Martin Pütz (ed.), Thirty Years of

Linguistic Evolution: Studies in Honour of René Dirven on the Occasion of his Sixtieth Birthday, 483-502. Philadelphia andAmsterdam: John Benjamins.

––––. 1997. ‘Constituency, Dependency, and Conceptual Grouping.’ Cognitive Linguistics 8.1-32.––––. 1998. ‘On Subjectification and Grammaticization.’ In Jean-Pierre Koenig (ed.), Discourse and Cognition: Bridging the Gap, 71-89.

Stanford: CSLI Publications.

41


Definiteness Hierarchy Effects in Referential Processing: An Eye-tracking Study

Hanjung LeeSungkyunkwan University

The present study reports results from two eye-tracking experiments investigating how the relative markedness of noun phrases(NPs) as a subject and an object influences referential processing in Korean, a verb-final language with canonical SOV order.The experiments provide new evidence about the active cognitive status of the system of markedness constraints suggested inrecent Optimality Theoretic approaches to morphosyntactic variation (Aissen 2003; Aissen and Bresnan 2002).

A great deal of research has shown that sentence complexity effects are moderated by the types of NPs in a sentence. Forinstance, in a complexity rating study, Gibson (1998) found that complexity of object-extracted relative clauses (RCs) (e.g., (1))was most reduced by having indexical pronouns (e.g., you or I) in the subject position within the RC. Gibson proposes thatindexical pronouns impose less of a load on working memory than other types of referring expressions because they refer toentities that are immediately available in the comprehender’s environment. More recently, Warren and Gibson (2002) found thatsentence complexity is more closely related to gradations of the prominence of the embedded subject, such as proposed in thedefiniteness hierarchy (2) or in the givenness hierarchy (Gundel et al. 1993) than to whether the embedded subject was indexicalpronouns or not.

The first experiment tested center-embedded complement clause structures in Korean to determine how the definiteness oftwo adjacent subject NPs contribute to processing difficulty. We varied whether the matrix subject NP and the embedded subjectNP were pronouns or descriptions as shown in (3). Forty-eight sentences were created with different matrix verbs and embeddedverbs. Eye movements were recorded as subjects read the experimental sentences. Twenty-four native speakers of Korean weretested in the experiment. There were two measures of interest for the reading of the object NP: first-pass gaze duration, which isthe sum of the duration of the first fixation on a word plus any subsequent fixations on that word that occur before a successiveword is fixated, and rereading time, which is the sum of the duration of any fixations made on a region after subsequent regionshave already fixated. Both first-pass reading and rereading time data produced a significant effect of the type of the matrixsubject NP, with the description condition taking longer to read than the pronoun condition in all critical regions.

The second experiment examined whether the types of an object NP influence processing. The experiment focused oncausative constructions which contain two objects that are accusative-marked as shown in (4). We varied whether the first objectNP (causee) and the second object NP were pronouns or descriptions. Pronouns are widely considered to be less marked andmore natural than other types of referential expressions as clausal subjects but more marked as clausal objects (Croft 1990; Givón1984; Aissen 2003). Thus, sentences containing a pronoun object should be more difficult to process than sentences containing adescription object if this type of markedness as an object affects referential processing. Significant effects of the markedness ofthe object NP were found on both reading time measures in the four critical regions: the first object NP (causee) and the first(embedded) verb were read significantly slower in the pronoun condition than in the description condition. In the second objectregion and the second verb, both reading time measures revealed a significant effect of the markedness as an object, with thesecond object (object of the embedded non-finite verb) and the verb being read more slowly in the pronoun condition than in thedescription condition. Taken together with the results of Experiment 1, this result shows that the processing difficulty associatedwith referential expressions is more closely related to markedness as a subject and an object than to the referential form of NPs.

The markedness effects can be interpreted as resulting from a preference of more prominent expressions (i.e., pronouns andnames) to be assigned a syntactically more prominent position (i.e., (matrix) subject position). Greater difficulty in the processingof disharmonic nominal expressions (e.g., low-prominence subject and high-prominence object) provides new evidence for theactive cognitive status of the system of markedness constraints suggested in Aissen and Bresnan’s (2002) stochastic OT approachto morphosyntactic variation. Although the definiteness constraint subhierarchy (see e.g., Aissen 2003; Aissen and Bresnan 2002)does not have categorical effects in the grammar of Korean, its active cognitive status can nevertheless be empirically detected bypsycholinguistic experimentation. Further the experiments show that the markedness effects over the time course ofcomprehension can be measured using eye-tracking under natural reading conditions.

(1) The reporter that you/the doctor met spoke very quickly.(2) DEFINITENESS HIERARCHY (Aissen 2003): Personal pronoun > Proper name > Definite NP > Indefinite specific NP > Non-specific NP(3) Kutul-/uysa-ka wuli-/haksayng-i silhum-ul haysstako malhayssta.

They/doctor-NOM we/student-NOM experiment-ACC ran said‘They/the doctor said that we/the student ran experiments.’

(4) Yumi-ka wulii-/uysa-lul kutul-/haksayng-ul mannakey hayssta.Yumi-NOM us/doctor-ACC them/student-ACC meet made/let‘Yumi made/let us/the doctor meet them/the student.’

ReferencesAissen, J. 2003. Differential object marking: Iconicity vs. economy. Natural Language and Linguistic Theory 21: 435-483.Aissen, J. and J. Bresnan. 2002. Optimality Theory and Typology (course material used in LSA Summer School 2002).Croft, W. 1990. Typology and Universals. Cambridge: Cambridge University Press.Gibson, E. 1998. Linguistic complexity: Locality of syntactic dependencies. Cognition 68: 1-76.Givón, T. 1984. Syntax. A Functinal-typological Introduction. Volume 1. Philadelphia: John Benjamins.Gundel, J., N. Hedberg and R. Zacharski. 1993. Cognitive status and the form of referring expressions in discourse. Language 69: 274-307.Warren, T. and E. Gibson. 2002. The influence of referential processing on sentence complexity. Cognition 85: 79-112.

42


Subject/Object Asymmetry in the Acquisition of Relative Clauses in Chinese

Kwee-Ock Lee and Sunyoung LeeKyungsung University and Kyunghee University

This paper discusses subject/object asymmetry in Chinese children’s acquisition of relative clauses based on apicture-aided comprehension task, testing the linear distance hypothesis (Hawkins 1989, Babyonyshev and Gibson1999) and the structural distance hypothesis (Hamilton 1995, Hawkins 1999, O’Grady 1999).

One of the robust findings in the study of acquisition of relative clauses concerns the sequential developmentof different types of relative clauses that conforms to Keenan and Comrie’s (1997) accessibility hierarchy (Subject >Object > Indirect object >…). In particular, ESL learners find subject relative clauses easier than object relativeclauses (Doughty 1991, Gass 1979, 1980, 1982, Hamilton 1994).

(1) English relative clausesa. The dog [that _____ [sees the rabbit]] Subject relative clauseb. The dog [that the rabbit [sees _____]] Object relative clause

The linear distance hypothesis ascribes the difficulty of relative clauses to the distance between the gap and thehead, which is counted by the number of intervening words. On the other hand, structural distance hypothesisconcerns the depth of embedding of the gap, which is counted by the number of nodes (or functional projections)intervening between the gap and the head.

In English, however, two theories predict the same results; that is, subject is closer to the head than objectlinearly as well as structurally, which makes it difficult to evaluate the two theories. In contrast, head finallanguages give us a chance to compare the two theories directly. Chinese, another SVO language like English, is ahead final language, where subject is linearly more distant from (Object > Subject) but structurally closer to (Subject> Object) the head as shown in the following examples;

(2) Chinese relative clausesa.[____ [ kan tuzi]]-de gou Subject relative clause

S V O Head‘the dog that sees the rabbit’

b. [tuzi [kan ____]]-de gou Object relative clauseS V O Head

‘the dog that the rabbit sees’

These contradictory predictions have been noticed by O’Grady, Lee and Choo (2003) with examples from Korean,who have found that subject relative clauses were easier for the second language learners of Korean, favoringstructural distance hypothesis.

Taking Chinese as a target language, this paper aims to find which theory can provide a better explanation foracquisition of relative clauses. A picture-aided comprehension task inspired by O’Grady, Lee and Choo (2003) wasconducted with a total of thirty-five Chinese kindergarten children (seventeen boys and eighteen girls) residing inShenyang, China. The results are presented in the following table:

Table 1. Chinese children’s comprehension of relative clauses (maximum mean: 5)

Group Subject relative clauses Object relative clauses6 year olds (5;6-6;6) (n=20) 4.25 .0855 year olds (4;6-5;5) (n=15) 2.87 1.47

The results reveal that subject relative clauses are easier to comprehend for the Chinese children than object relativeclauses, favoring the structural distance hypothesis.

In conclusion, it was found that children’s comprehension of relative clauses was better at subject relativeclauses than at object relative clauses in Chinese, favoring the structural distance hypothesis over the linear distancehypothesis.

43


A Reanalysis of the Aymara Verb

Erik LevinUniversity of Chicago

Up to this point, the Aymara verb has been analyzed as a matrix of tense and evidentiality (i.e. howthe speaker came to know of an event), as shown in Figure 1 below. The temporal axis consists of afuture, a present/near past, and a distant past, and the evidential axis shows personally witnessedknowledge vs. non-personal knowledge (hearsay). Under this analysis, the morphology is defective in tworegards. It collapses the distant past and the present/near past for non-personal knowledge, and there is noevidentiality distinction for the future form. Furthermore, there are significant ‘exceptions’ to the uses ofthese forms.

A more elegant, non-defective analysis without ‘exceptions’ is possible if we recognize that theprevious analyses have imposed Indo-European categories onto a language that does not give precedenceto them. Whereas most Indo-European languages are more concerned with locating an event in time,Aymara prioritizes how a speaker knows of the event. Even for instances when time is the most importantpiece of information, previous studies have neglected to incorporate the Aymara conception of time. Incontrast to a Western view, in which the speaker conceptually looks faces forward toward the future, theAymara place the past in front of the speaker, because it is ‘visible’. The unknown, unseen future liesbehind the speaker.

Aymara morphology reflects this view. Consider, for example, the words for ‘yesterday’ and‘tomorrow’, which are, respectively, masuuru and qharuuru. The morpheme mas means ‘in front’, anduuru means ‘day’, so yesterday is the ‘day in front’. In the same manner, qhar means behind, so tomorrowis the ‘day behind’ because it lies in back of the speaker, where it is not visible.

Further investigation shows that Aymara verbs reflect four grades of how sure a speaker is of theinformation that he or she reports. The passage of time is simply the prototype of any of a number ofreasons for which a speaker may choose to accept less responsibility. Other reasons include hearsay,surprise, etc. Table 2 below summarizes the new analysis by pairing verb inflections with their uses. Thisstudy suggests new approaches to analyzing tense, aspect, and modality in languages that are unrelated toAymara, too.

References

Binnick, Robert I. 1991. Time and the Verb. New York: Oxford University Press.Botne, Robert. 1999. personal communication.Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press.Comrie, Bernard. 1985. Tense. Cambridge: Cambridge University Press.Ebbing, Juan Enrique. 1965. Gramatica y Diccionario Aimara. La Paz, Bolivia: Editorial Don Bosco.Gomez, Donato Bacarreza and Cosme, José Condori. 1992. Morfoligía y Gramatica Del Idioma Aymara. La Paz: (publisher

unspecified).Hardman, M.J., et al. 1988. Aymara: Compendio de Estructura Fonologia y Gramatical. La Paz: Instituto de Lengua y Cultura

Aymara.Laprade, Richard A. 1981. “Data Source in La Paz Spanish Verb Tenses.” In: Hartman, M.J. (ed.), The Aymara Language in Its

Social and Cultural Context, 206 - 207. Gainesville: University Presses of Florida.Lakoff, George. 1987. Women, Fire, and Dangerous Things. Chicago: The University of Chicago Press.Levin, Erik. 1999. “The Influence of Aymara on Tense and Aspect in Altiplano Spanish.” unpublished manuscript.Miracle Jr., Andrew W., and de Dios Yapita Moya, Juan. 1981. “Time and Space in Aymara.” In: Hartman, M.J. (ed.), The

Aymara Language in Its Social and Cultural Context, 33 – 57. Gainesville: University Presses of Florida.Sokolowski, Robert. 2000. Introduction to Phenomenology. Cambridge: Cambridge University Press.

44

Figure 1: The traditional analysis of the Aymara TMA system

distant past near past / present future

personal knowledge puriinwa puri

non-personal knowledge puritayna purini

present deixis �

Table 2: Forms, uses, and abbreviations in the new analysis

Form UseTraditional Analysis

Name

New AnalysisName and

Abbreviation

-i

Used for events for which the speaker is willingto take high degree of responsibility in reporting:present or recent eyewitnessed events are theprototype; very well remembered eyewitnessedevents in the remote past are radial extensions.

personal knowledge -present / recent past

highresponsibility

(HR)

-inwa

Used for events for which the speaker takesslightly less responsibility for than the HR from:personally witnessed events in the remote pastare the prototype; comparatively mildExceptional Cases that fall outside of what thespeaker knows as ‘normal’ are radial extensions

personal knowledge -distant past

midresponsibility

(MR)

-itayna

Used for events for which the speaker takes littleresponsibility: indirectly acquired knowledge isthe prototype; extreme Exceptional Cases thatfall far outside of what the speaker knows as‘normal’ are radial extensions

non-personalknowledge - past

lowresponsibility

(LR)

-ini

Used for events for which the speaker takesminimal responsibility: reserved for the futurebecause by definition, absolutely no one haswitnessed it at the time of the speech event

future future (F)

The cells above are color coded – the lighter the cell, the more responsibility for a speech act its correspondingmorpheme represents.

45

Friday, 17:30 – Prairie Room ASL

The Acquisition of ASL Indicating Verbs and Pronouns:Learning to Conceptualize Blends with Real Space

Scott LiddellGallaudet University

Considerable attention has been given to the acquisition of ASL directional pronouns and indicatingverbs. Studies have been guided by assumptions such as (1) children are acquiring spatial morphemes andincorporating them into pronouns and verbs, (2) signs consist solely of morphemes, and (3) gesture is notpart of language. In analyzing adult signing, Liddell (2003) demonstrates that these assumptions areunwarranted. Indicating verbs and directional pronouns point at things. That is, in addition to expressingthe lexical meaning of a verb such as TELL, directing it toward the addressee in real space additionallyidentifies her as the recipient of the telling. Thus, ASL requires the ability to overlay pointing on thelexical production of the verb.

Complicating things further, adult signing not only directs such signs toward physically presententities, but also toward entities conceptualized as present in blends with real space. Liddell (2003)identifies several categories of real space blends, including surrogate spaces, token spaces, and depictingspaces. Thus, in addition to learning the lexical and grammatical properties of a verb such as TELL, thechild acquiring ASL must also learn that it is directable. The child must also learn multiple ways ofconceptualizing non-present entities as present, then learning to direct signs toward them.

This study examines the signing of six Deaf children with Deaf parents between the ages of 2 and 6.It is well known that when acquiring indicating verbs, children first learn non-indicating forms. Later theylearn to direct these signs. The children’s signing will be examined for signs the children know to directand their real and conceptualized-as-present entities targets. The methodology does not differ betweenchildren and adults. Video analysis and context should reveal if a sign is directed toward a physicallypresent entity. If a sign is directed toward a real space blend, the clues needed to determine the nature ofthe blend include the type of sign, overt identification of the entity, repeated references to the entity,whether topographical concepts such as near and above are important, whether, through facial expressionor body posture or movement, the child has taken on a role, and even whether the child uses the eyes orface to point. The acquisition patterns revealed should be quite complex. Liddell (2003) identifies 26distinct pronominal signs, almost all of which have the grammatical requirement of being directed towardtheir referents, and 13 categories of indicating verbs. For each individual sign being acquired childrenmust not only learn to produce the sign as part of a grammatical construction, the child must also learnhow the sign needs to be directed or placed, and must have the ability to conceptualize space in waysnecessary for the use of that sign. I expect to find a complex interaction between the acquisition of lexicalforms, the ability to identify which lexical forms can be directed and placed in space, and the ability toconceptualize space in the many ways that adults do.

References

Liddell, Scott K. 2003. Grammar, Gesture, and Meaning in American Sign Language. Cambridge: Cambridge University Press.

46

Saturday, 11:30 – Prairie Room DISCOURSE & EMERGENT GRAMMAR

Emergence of the Indefinite Article:Grammaticalization and Discourse Functions of YIGE in Spoken Mandarin

Meichun LiuNational Chiao Tung University, Taiwan

Recent studies on language use and grammaticalization have reiterated the importance of frequency in shapingprosodic and semantic changes (cf. Thompson and Mulac 1991, Bybee and Hopper 2001, Biq 2002, Tao 2002). In asimilar vein, this paper is motivated by the overwhelmingly pervasive use of YIGE in Mandarin conversation whichcalls for a re-examination of the traditional analysis of YIGE as a classifier phrase (Numeral YI ‘one’ + Classifier GE).The discourse grounding function of Numeral-Classifiers (NCs) has been thoroughly studied in Li (2000a & b), butthe frequent and non-typical NC use of YIGE indicates a departure from the classifier system and the emergency of anew category, the indefinite article, in Spoken Mandarin. A careful study of naturally-occurring data shows thatYIGE is frequently used with proper names and abstract nouns that do not require number or ‘class’ marking. It isthus suggested that the unit YIGE, as the generic and most representative member of the NCs, may be viewed as aninseparable whole, acquiring a new grammatical status as a marker of indefinite referentiality. It functions alongsimilar lines as the indefinite article ‘a’ in English, striding further than the observation stated in Li and Thompson(1981: 132): “The numeral yi ‘one’, if it is not stressed, is beginning to function as ‘a’”.

The English article a is normally used with indefinite count nouns which may be referring or non-referring(Givon 1993). The Mandarin YIGE, on the other hand, may occur with definite NPs such as proper names, as well asnon-countable NPs such as abstract nouns:

1) YIGE + Proper Noun a. zheyang YIGE [Chen Zhenan] ‘such a Zhenan Chen (Name)’

b. hui kandao YIGE [Huanan yinhang] ‘(You) will see a Huanan Bank.’

2) YIGE + Abstract Nouna. zaocheng gaoling renkou de YIGE [shiye] ‘resulted in an unemployment of the aged’b. xi-lian yao zheyang de YIGE [fanfu] ‘Washing face requires such a complicatedness.’

Huang (1998) argues that the category of articles is emerging in spoken Mandarin and that the distaldemonstrative NAGE is becoming a definite article. This paper echoes his observation and argues that the emergenceof its counterpart, the indefinite article in Mandarin, is also coming to place with a parallel function to the Englisha. However, the newly arising system of articles gets in an apparent competition with the long existing NC system.The two competing forces are at work in shaping the diverse ways YIGE are used in contemporary discourse. Bothqualitative and quantitative analyses will be done to show that Mandarin is developing a new grammatical paradigmwhere a three-way distinction may be made:

zero (non-referential) vs. YIGE (referential and indefinite) vs. NAGE (referential and definite).

The emergence of the Mandarin YIGE from a NC sequence to a grammatical marker of indefiniteness compliesnicely with the five principles of grammaticalization discussed in Hopper and Traugott (1993). It further illustrateshow a frequent collocate, with its semantic uniqueness, may be re-analyzed and re-categorized under the force ofroutinized language use.

References

Biq, Y-O. 2002. Classifier & Construction: the interaction of grammatical categories & cognitive strategies. Language & Linguistics 3.3: 521-542.

Bybee, J. & P. Hopper (eds.). 2001. Frequency & the Emergence of Linguistic Structure. Amsterdam/Philadelphia: John Benjamins.Givon, T. 1993. English Grammar: A Function-based Grammar. Vol. I. Amsterdam/Philadelphia: John Benjamins.Hopper, Paul and Elizabeth C. Traugott. 1993. Grammaticalization. Cambridge: Cambridge University Press.Li, Wendan. 2000a. Numeral-classifiers as a grounding mechanism in Mandarin Chinese. Journal of Chinese Linguistics 28.2:337-368.–––––. 2000b. The pragmatic function of numeral-classifiers in Mandarin Chinese. Journal of Pragmatics 32:1113-1133.Shuanfan, Huang. 1999. The emergency of a grammatical category ‘definite article’ in Spoken Chinese. Journal of Pragmatics 31.1.Tao, Liang. 2002. Phono-syntactic conspiracy and beyond: grammaticalization in spoken Beijing Mandarin. In New Reflections on

Grammaticalization, I. Wischer & G. Diewald (eds.), 277-292. Amsterdam/Philadelphia: John Benjamins.Thompson and Mulac. 1991. A quantitative perspective on the grammaticization of epistemic parentheticals in English. In Approaches to

Grammaticalization, E. C. Traugott and B. Heine (eds.), Vol. II. 313-329.

47


The Sentence Positions of Antecedents of the in Natural Discourse:Comparing Syntactic Parallelism and the Theory of Centering

Ring LowUniversity of Buffalo

This paper uses corpus analyses to investigate whether the grammatical positions of definite noun phrases with themay correlate with the syntactic positions of their antecedents in natural discourse, as previously suggested for thecase of pronouns. The study compares the anaphoric predictions of two theories: the Centering Theory and themodel of syntactic Parallelism. The former suggests grammatical subjects to be the default antecedents of definitenominals, while the later, on the other hand, predicts a parallel correlation between the syntactic position of anantecedent and that of its anaphor (e.g., Stevenson et al. 1995, Chambers & Smyth, 1998). In this study, I illustratethat subject position, although being ranked highly on the hierarchy of Forward Looking Center in CenteringTheory, (Grosz et al. 1995, 1983; Walker & Prince 1996; Walker et al. 1998) does not out-rank other sententialpositions to be the most frequent type of antecedents. In the mean time, a better-than-chance correlation was foundbetween the syntactic positions of noun phrases with the and those of the antecedents, suggesting that syntacticparallelism may be at play in the use of the in discourse.

The data of the study come from an analysis of 1417 definite noun phrases with the found in a corpus of28,200 words, including two interview transcriptions and twenty articles of different genres (stories, reviews,columns, and news). Each definite noun phrase was analyzed to see if it has an explicit antecedent (i.e., DiscourseOld, Prince 1981, 1992), a conceivable implicit antecedent (i.e., “Bridging”, Haviland & Clark 1974; “Inferrables”,Prince 1981, 1992;), or has no antecedent at all (i.e. Discourse New).

The syntactic positions of the 1417 definites and all antecedents (292 explicit antecedents, 112 implicitantecedents) were then classified into four categories. The categories are (1) clausal or sentential subjects; (2)objects of a verbal predicate; (3) an element embedded in an “adjunctive phrase” (e.g., after the dinner); and (4) anargument or dependent of a preposition attached to another noun phrase (e.g., the corner of the room). Thefrequencies were then tallied into these conditions: (a) when the syntactic positions of the antecedent and anaphorare parallel, but are both non-subjects (“+parallel, -centering”); (b) when the antecedent’s and the anaphor’spositions are of different categories, and the antecedent is a non-subject (“-parallel, -centering”,) (c) when theantecedent is a subject, but the anaphor isn’t (“+centering, -parallel”); and (d) when the antecedent and the anaphorboth occur in subject positions (“+centering, +parallel”). The frequency in each condition was compared to thechance probability (obtained with category frequencies (1)-(4)). Referential distance and antecedent definitenesswere controlled.

The results indicate that Discourse Old definites in natural discourse tend to occur in grammatical positionsparallel to those of their antecedents (observed: 23.3%, chance: 14.9%, condition (a)). Yet, they do not favor theirantecedents to occur in the subject position per se (observed: 15.2% chance: 23.3%, condition (c)) (all conditions:fisher exact 2x4, p<0.01). The same tendencies were also found for the category of Inferrables (observed: 30% in(a), 14% in (c), chance: 19% in (a), 20% in (c)). However, the differences there were not significant (fisher exact,p>0.05). The data suggest that Structural Parallelism, rather than Centering, may be relevant for the phenomenon ofdefinite nominals.

ReferencesClark, H. 1977. “Bridging.” In Johnson-Laird and Wason (eds) Thinking: Readings in Cognitive Science. John Wiley and Sons NYClark, H. 1992. Arenas of Language Use. The University of Chicago Press.Chambers, C. & Smyth, R. 1998. Structural parallelism & attentional focus: A test of Centering Theory. J of Memory & Language. 39, 593-608Haviland, S. & Clark, H. 1974. What's new? Acquiring new information as a process in comprehension.” JVLVB 13, 512-521.Prince, E. 1981. Toward a Taxonomy of Given-New Information.” In P. Cole (ed.), Radical Pragmatics, Academic, New YorkPrince, E. 1992. The ZPG letter: Subjects, definiteness, and information status.” In Wm. Mann & S. Thompson (eds.), Discourse Description: Diverse Linguistic

Analyses of a Fund-raising Text, 295-325.Grosz, B., Joshi, A., Weinstein, S. 1983. Providing a unified account of definite NPs in discourse. 21st Annual Meeting of the Association of Computational Linguistics;

proceedings of the conference 15-17 June 1983, MIT.Grosz, B., Joshi, A., Weinstein, S. 1995. Centering: A framework for modeling the local coherence of discourse. Computational Linguistics 21, 203-226.Sidner, C. L. 1979. Towards a computational theory of definite anaphora comprehension. English Discourse. Technical Report 537. MIT Artificial Intelligence

Laboratory, Cambridge MA.Sidner, C.. 1983. Focusing and discourse.” Discourse Processes 6, 107-130.Sidner, C. 1983a. Focusing in the comprehension of definite anaphora. In M. Brady & R. Berwick (eds.), Computational Models of Discourse, 267-330. Cambridge,

MA: MIT Press.Stevenson, R., Nelson, A., & Stenning, K. 1995.The role of parallelism in strategies of pronoun comprehension. Language & Speech, 38, 393-418.Walker, M. A., Joshi, A. K., Prince, E. F. 1998. Centering theory in discourse. Oxford, England: Clarendon Press.Walker, M A & Prince. E. 1996. A Bilateral Approach to Givenness: A hearer-status algorithm and a centering algorithm. Benjamins.

48


Subjective Progressives in Spanish

Ricardo MaldonadoUniversidad Nacional Autonoma de Mexico

This paper offers an account for Spanish Progressive Stative-Resultative constructions (SR). SR constructions takepast participle not progressive gerund as shown in (1.-2.):

1. Adrián está acostado / *acostando en la camaAdrian be lye-PST PPLE / PROG.PPLE on the bed‘Adrian is lying on the bed’

2. Valeria está parada / * parando en la esquinaValeria be stand- PST PPLE / PROG.PPLE in the corner‘Valeria is standing in the corner’

However under specific conditions the progressive in Spanish can designate an SR reading as in (3.):

3. Había un clavo saliendo de la pared [progressive]‘There was a nail sticking out from the wall’

Based on data from the Corpus del Español Mexicano Contemporáneo (3 million words) this paper offers a CGaccount of Progressive SR as a subjective construal of stative situations. The progressive SR is crosslinguisticallyuncommon since states normally reject progressive marking: *Belgium is lying between Holland and France(Langacker 1991). Japanese, Korean and Chinese are atypical in using the progressive marker to designateresultative states.

4. Ken-wa ne-te-I Japanese (Smith 1991).Ken-TOP sleep-ASP-NPST

‘Ken is sleeping / asleep’

However, verbs of position and collocation in English also allow the use of the progressive for stative situations asshown in (5.).

5. The picture is hanging on the wall

Shirai (2000) follows Smith (1983) in accounting for Japanese stative-resultatives as an interaction between actionsart and tense/aspect (viewpoint) where the progressive imposes a dynamic or a temporal view on a static situation. Afiner analysis (Langacker 1987) suggests that the progressive focuses on the commonality of the component states ofa process leading to a homogenous perfective event that is bounded in time. Thus the situation in (5.) can only bevalid during the temporal boundaries of the designated period and cannot be seen as permanent.

In this paper I propose that Spanish constitutes a further extension of Langacker’s proposal. I first showthat SR progressives only take verbs construed in a dynamic fashion. I further argue that for Spanish the temporalboundaries are not in the event itself but in the mental construal of the event by the speaker: the nail location in (3.)is permanent but the conceptualizer’s view of the scene takes the boundaries of the scene as consistently implying analternative situation. Moreover I demonstrate that the Spanish progressive SR construction involves both asubjective (Langacker 1991) and a Force-Dynamic (Talmy 1988) construal of the event: the conceptualizer is part ofthe objective scene and the event presupposes an alternative situation running against the conceptualizer’sexpectations. While both past Participle SR and Progressive SR may be temporally bounded, only the latter (7.)implies that the picture is about to fall:

6. El cuadro está colgado en la pared [past participle]7. El cuadro está colgando en la pared [progressive]

‘The Picture is hanging on the wall’

Finally I show that the internal contradiction of the Progressive SR construction iconically mirrors the ForceDynamic construal of the event.

References

Langacker, Ronald. 1987. “Nouns and Verbs”. Language 63. 53-94.–––– . 1991 “Subjectification”. Concept, Image and Symbol. Berlin: Mouton de Gruyter.315-342.Smith, Carlota, 1983 “A Theory of Aspectual Choice”. Language. 59: 479-501.Talmy, L. “Force Dynamics in Language and Cognition”. Cognitive Science 12. 1988, 49-100.

49

Saturday, 17:30 – Prairie Room JAPANESE DISCOURSE

The Japanese “Dative Subject” Construction in Written Discourse

Kyoko MasudaGeorgia Institute of Technology

The Japanese dative construction (e.g. Taroo-ni furansugo-ga wakaru. ‘Taroo understands French./It is possible forTaroo to understand French.’) has been widely discussed under different theoretical approaches due to peculiaritiesof the grammatical status of the ni-marked first noun phrase, Taroo-ni and the ga-marked second noun phrase,Furansugo ‘French’. Syntactic-based studies for this construction include Kuno (1972), Shibatani (1977, 1978),Saito (1982), Takezawa (1987), Miyagawa (1987), Dubinsky (1992), Sadakane & Koizumi (1995), Johnson & Kuno(2004), and Kishimoto (2004) while semantic-based studies include McGloin (1980), Sugimoto (1986), Kabata &Rice (1997), Kabata (1999), Kumashiro (2000), Shibatani (1999, 2001), Masuda (1999), and Shibatani & Pardeshi(2001). In spite of the semantic-based studies of Japanese dative subjects, few linguists have succeeded in evaluatingthe semantic aspect of the dative subject construction with naturally occurring discourse data. Recent studies, Sadler(2002), Ono & Sadler (to appear), and Carlee & Masuda (to appear) are exceptions. The first two studies reportedthat the dative subject is rarely used in spoken Japanese, while it is more likely to be used in written text (novels).Selecting texts from a variety of different genres is important, as each genre may exhibit certain stylisticcharacteristics, and as investigating only particular text types might lead to false generalizations (Biber, 2000). Thecurrent study, therefore, examines two different types of written texts such as novels and newspapers to examine ifthere is a prototype of the Japanese dative subject construction, and to explore the discourse function of the Japanesedative subject construction in written discourse.

The main findings are the following. First, in both novels and newspapers, the phenomenon of the Japanesedative subject construction in written discourse had a wider range of variants, such as nimo, niwa, and ni datteamong others. Second, in both novels and newspapers, the greater portion of ni/variant-marked participants werefirst person pronoun referents (38% in novel /27% in newspaper) such as (w)atashi, boku, ore, and jibun or the maincharacters. Third, the majority of ni/variant-marked participants are animate, either human or body parts. Fourth, innewspapers, in many cases, ni/variant-marked participants co-occurred with adjectives and nominal nouns ratherthan potential/possessive/perception verbs. Especially ni/variant-marked usage in the newspaper data includeabstract nouns such as kyooiku ‘education’ and azia no heiwa to antee ‘peace and stability in Asia’. There is nooccurrence of the ni/variant-marked second person pronoun participants ‘you’ in the newspaper data. Based uponthe written discourse data, this paper concludes that this use of dative-marked noun phrases in Japanese is bestunderstood as a discourse marker or as ‘frame setting’ since it is used in written texts to show that the participantsexhibit a certain perspective, regardless of the type of written text. This is also shared by the previous studies whichdeal with spoken discourse. Given that in Japanese first-person ëpronounsí may not just be doing referential work ininformal conversation (Ono & Thomson 2003), the occurrence of the first pronoun and the main characters markedby dative ni and its variants also serves as a discourse signal of a cognizer expressing the speakerís attribute ormental state in a discourse.

References

Arnett, C. & K. Masuda. To appear. On German and Japanese dative constructions. Proceedings of the Fifth Annual High DesertConference in Linguistics, University of New Mexico.

Kabata, K. 1998. On the Agent/experience continuum: A cognitive approach to the ga/ni alternation in Japanese. Unpublishedpaper.

Kabata, K. and S. Rice. 1997. Japanese ni: The particulars of a somewhat contradictory particle. Lexical and SyntacticalConstructions and the Construction of Meaning, ed. by Marjolijn Verspoor, Kee Lee, and Eve Sweetser, 109-27.Amsterdam: John Benjamins.

Masuda, K. 1999. Semantic-discourse account for dative-subject construction in Japanese. Proceedings of Second Annual HighDesert Student Conference in Linguistics, 111-122, University of New Mexico.

Sadler, M. 2002. Deconstructing the Japanese Dative Subject Construction. Doctoral dissertation. University of Arizona.Shibatani, M. 1999. Dative subject construction twenty-two years later. Studies in the Linguistic Sciences 29: 2, 45-76.Ono, T. & S.A. Thompson 2003. Japanese(w)atashi/ore/boku ‘I’: They’re not just pronouns. Cognitive Linguistics 14: 4, 321-

347.

50


Where Eye Movements Go with Fictive Motion

Teenie Matlock and Daniel RichardsonStanford University

Cognitive linguists attribute a dynamic representation to fictive motion sentences such as (1a) and (b).

(1) a. The road runs along the coastb. A rock wall crosses the mountain range

On this view, the conceptualizer is said to subjectively “scan” or “move” along the trajector (road, rock wall) eventhough the scene is static (see Langacker, 1986, 2000; Talmy, 1983, 1996, 2000). Recent psychological worksuggests that in fact there is some kind of dynamic representation involved in the processing of fictive motionsentences. For instance, in Matlock (in press), people read stories about travel followed by critical sentences such as(1a). Participants were slower to read these sentences after thinking about travel that was slow, over a long distance,or through difficult terrain, compared to travel that was quick, over a short distance, or through uncluttered terrain,respectively. Critically, the effect was not obtained for non-fictive motion sentences, such as The road is next to thecoast. Together, the results suggest that thinking about actual motion influenced the understanding of fictive motion,which is suggestive of mentally simulated motion (see also Matlock, Boroditsky, & Ramscar, 2003).

The current work examines fictive motion in the context of visual perception. Specifically, will fictive motionlanguage influence eye movements? There is evidence that when looking at a blank wall or closing their eyes,participants’ eye movements can be influenced by the literal spatial structure of a scene described in a narrative(Spivey & Geng, 2001; Spivey, Tyler, Richardson, & Young, 2000). If people simulate motion while understandingfictive motion sentences, we might also expect to see evidence of this in their eye movements. Whereas other studieshave examined eye movements coordinated with literal language (e.g., ambiguity resolution, thematic roles; forreview, see Richardson & Spivey, 2004), we implemented a novel eye-movement paradigm with which to studyfigurative language. Participants in our study passively inspected static scenes while listening to fictive and non-fictive motion sentences, such as The road runs along the lake or The road is next to the lake along with a picture ofa road and a lake. As expected, inspection times along the relevant trajector (e.g., road) in the two conditions varied:trajectors with fictive motion sentences were looked at longer than those with non-fictive motion sentences.Critically, norming studies and follow-up priming studies on fictive motion alone assured the effect was not simplydriven by the presence of a motion verb (Matlock & Richardson, in progress).

In the end, our study makes three notable contributions: (1) our eye-tracking methodology offers a new way toinvestigate the online effect of figurative language; (2) our data provide additional support for the proposal thatmentally simulated motion is involved in fictive motion processing; and (3) our findings suggest that fictive motionconstrual is embodied in perceptual processes.

References

Langacker, R.W. (1986). Abstract motion. Proceedings of the Twelfth Annual Meeting of the Berkeley Linguistics Society, 455–471.Langacker, R. W. (2000). Virtual reality. Studies in the Linguistic Sciences, 29, 77-103.Matlock, T. (2004). Fictive motion as cognitive simulation. Memory & Cognition. To appear.Matlock, T., Ramscar, M., & Boroditsky, L. (2003). The experiential basis of meaning. Proceedings of the 25th Annual Conference of the

Cognitive Science Society.Matlock, T., & Richardson, D. (in press). Do eye movements go with fictive motion? Proceedings of the 26th Annual Conference of the

Cognitive Science Society.Richardson, D. C., & Spivey, M. J. (2004). Eye movements. In Wenk. G. & Bowlin, G. (Eds.) Encyclopedia of Biomaterials and Biomedical

Engineering. Marcel Dekker, Inc.Spivey, M.J., & Geng, J.J. (2001). Oculomotor mechanisms activated by imagery and memory: Eye movements to absent objects. Psychological

Research/Psychologische Forschung, 65, 235-241.Spivey, M.J., Tyler, M.J., Richardson, D.C., & Young, E.E. (2000). Eye movements during comprehension of spoken scene descriptions. The

Proceedings of the Twenty-second Annual Cognitive Science Society Meeting, 487-492.Talmy, L. (1983). How language structures space. In H. Pick, & L.P. Acredolo (Eds.), Spatial orientation: Theory, research, and application (pp.

225-282). New York: Plenum Press.Talmy, L. (1996). Fictive motion in language and “ception”. In P. Bloom, M.A. Peterson, L. Nadel, & M.F. Garrett (Eds.), Language and space

(pp. 211-276). Cambridge, MA: MIT Press.Talmy, L. (2000). Toward a Cognitive Semantics, Volume I: Conceptual Structuring Systems. Cambridge: MIT Press.

51


The Effect of Transitivity and Argument Structure on Lexical Finite Verb Productionin Children with SLI

Ramesh Kumar MishraAll India Institute of Speech and Hearing

Specific language impairment is a developmental language disorder where grammatical morphology isseriously impaired. The production of lexical finite verbs is very important for young language learners toattain greater morpho-syntactic and semantic competence in sentence production in any language. In thispaper a group of 20 Kannada speaking children identified as SLI in the age range of 4 to 8 years arecompared with younger normal children for lexical verb production with varying degree of transitivity.The selection of children with SLI followed standard exclusionary criteria. All children were monolingualspeaker of Kannada. Kannada is a major Dravidian language spoken in the state of Karnataka in India.The specific aim was to see the performance of these children in a verb production task where they had toproduce sentences involving intransitive, transitive and intransitive verbs in declarative matrix. Therewere 40 stimuli each, line drawings, in first two categories and 20 in the last category. Productions werecoded in terms of omission of subjects in intransitive frames, omissions of direct objects in transitiveframes and omissions or substitution of argument structures in the ditransitive frame. Results indicate agradual increase in number of omissions and substitutions of obligatory arguments and necessary TENSE,AGR markers as transitivity increases. Children with SLI performed much poorly compared to normals inall three category. No such effect of transitivity if noticed in normals. The paper discusses the findings interms of a grammatical constraint that affects verb production in SLI children. The data is compared withearlier data collected by the author on Hindi speaking SLI children. A cross language effect oforganization of verbs is discussed. The outcome of the study indicates specific care to be taken whileteaching verb morphology to such children with SLI in speech language therapy. Transitivity of actionverbs as a distinct mental category for acquisition among children with SLI is proposed.

52


Preposed and Postposed Although Clauses in Discourse:A Text Based Study from English

Yuko MizunoHokkaido University

This paper attempts to clarify the difference between preposed and postposed although clauses with regard to theirdiscourse functions. As Diessel (1996:72) notes, “previous studies on adverbial clauses suggest that the order ofmain and subordinate clauses is mainly determined by discourse factors.” Ohori (2000:308) generalizes the findingsfrom previous studies as in (1):

(1) Preposed adverbial clauses have a global relation to the context, whereas postposed adverbial clauses have a local relationto the context.

Thompson (1985), Ford and Thompson (1986), and Ramsay (1987) attest that the generalization in (1) is true forpurpose clauses, if clauses, and when clauses, respectively. However, almost no study has examined whetheralthough clauses also conform to this generalization. Noordman (2001) is exceptional in this respect, but it does notspecify the patterns in which although clauses are related to the preceding discourse. The goal of the present paper isto clarify how preposed and postposed although clauses are related to the preceding discourse, and whether althoughclauses also conform to the generalization in (1), based on naturally occurring data. Since although clauses arerelatively rare in spoken English discourse (cf. Ford 1993), the data were collected from written English texts:newspapers, magazines, a novel, and an essay. The total number of collected preposed although clauses are 66,whereas that of postposed although clauses are 25.

My investigation of the data found the following three points. First, a majority of preposed althoughclauses are related to the preceding discourse in one of four ways, i.e., by representing discourse old information, asin (2); by representing “inferable” information, as in (3); by representing information which consists of an inferableopen proposition and a focus, as in (4); or by offering a contrast to an earlier statement, as in (5). Second, a majorityof postposed although clauses are not related to the preceding discourse, but just to the preceding main clause, as in(6). Third, whenever postposed although clauses are related to the preceding discourse, the preceding main clause isalso related to the prior context. Based on these findings, I conclude that although clauses are in conformity with thegeneral tendency of other adverbial clauses in (1).

(2) We were told that one of our children has a medical condition for which there is no treatment and that we should anticipateher death. … Although she is still with us, we are already grieving.

(3) Harry woke early the next morning. Although he could tell it was daylight, he kept his eyes shut tight.(4) Reservations are required for the train … Although they can be made at any railroad station in Switzerland, it is advisable

that …(5) The young man was putting his heart and soul into the match, and although he did not come in first place, he did win a

medal.(6) Harry heard something creak outside. He hoped the roof wasn't going to fall in, although he might be warmer if it did.

References

Chafe, Wallace. (1984). “How people use adverbial clauses,” BLS 10, 437-449.Diessel, Holger. (1996).“Processing factors of pre- and postposed adverbial clauses,” BLS 22, 71-82.Ford, Cecilia E. (1993). Grammar in Interaction: Adverbial Clauses in American English Conversations, CUP, Cambridge.Ford, Cecilia E. and Sandra A. Thompson. (1986). “Conditionals in discourse: A text-based study from English,” On Conditionals, ed. by

Elizabeth Closs Traugott et al., 353-372, CUP, Cambridge.Noordman, Leo G. M. (2001). “On the production of causal-contrastive although-sentences in context,” Text Representation: Linguistic and

Psycholinguistic Aspects, ed. by Ted Sanders, Joost Schilperoord, and Wilbert Spooren, 153-180, John Benjamins, Amsterdam.Ohori, Toshio. (2000). “Gengoteki chishiki to shiteno koobun--fukubun no ruikeeron ni mukete” (Constructions as linguistic knowledge--toward

a typology of complex sentences), Ninchi Gengogaku no Hatten (Advances in Cognitive Linguistics), ed. by Shigeru Sakahara, 281-315,Hituzi, Tokyo.

Prince, Ellen F. (1986). “On the syntactic marking of presupposed open propositions,” CLS 22, 208-222.Ramsay, Violeta. (1987). “The functional distribution of preposed and postposed “if” and “when” clauses in written discourse,” Coherence and

Grounding in Discourse, ed. by Russell S. Tomlin, 383-408, John Benjamins, Amsterdam/Philadelphia.Thompson, Sandra A. (1985). “Grammar and written discourse: Initial vs. final purpose clauses in English,” Text 5, 55-84.

53


Imperative Particles in Khmer

Kanako MoriTohoku University

In Khmer, one of the most common ways of making imperative sentences is by placing so-called sentence finalparticles at the end of affirmative sentences. This paper focuses on two sentence final particles phççN and coh. Bothof these particles have been described as expressing "imperative meaning" by previous researchers, particularlyUeda (2002). However, the differences between these particles have not been explicated. In light of this, the goal ofthis paper is to shed light on the subtle distinction between the particles phççN and coh and provide a principledaccount for their distribution. Specifically, we will argue that the grammaticalized meaning conveyed by theseparticles follows naturally from their respective more concrete usages (cf. Hopper and Traugott 1993).

Let us first take a look at the grammaticalized usages of the two particles. Although the two particles are usedin "imperative" sentences, there is a subtle difference. While phççN is used with the connotation of "begging afavor", coh is used in the sense of "permission".

(1) cuoj phççNhelp'Help me!'

(2) t´$v sç$mraak cohgo rest'(You may) take a rest'

We claim that the above-mentioned grammaticalized usages (1) and (2) respectively stem from the more concreteusages (3) and (4):

(3) k¯om ¯am baaj phççN m´$´l tuuE$´tç$h phççNI eat dinner watch TV'I ate dinner, and watched TV, too'

(4) me$e nuh coh pii l´$´ coN ch´$´woman that from on top tree'That woman got down from the top of the tree'

With regards to phççN the original function arguably has been "adding something parallel to the context" similarly to"too" in English. Further, its meaning can be divided into two components in (5) below.

(5) Two meanings of the original usage of phççN:a. obligatory presence of accessible parallel element in the contextb. focusing on the element immediately preceding it

The change of the functions of phççN (from (3) to (1)) can be explained as a "semantic reduction": reduction of (5a)to (5b) and thereby foregrounding of the remainder i.e. (5b). As for coh, the original usage has presumably been adirectional verb meaning "go down" having undergone metaphorical extension(cf. Lakoff and Johnson 1980) andacquired the imperative usage as a sentence final particle as in (6):

(6) Path of Metaphorical extension of cohverb: "spatial movement from upper place (generally house) to lower one (ground)" Ø

generalized to the movement from one's own sphere to outsideparticle: "relieving one's control on the addressee's act"

Based on these observations, we propose that the two particles convey different pragmatic meanings as in (7):

(7) phççN is used when the speaker wants the hearer to focus on the activity, while coh is used when the speaker releases therestraint on the hearer's activity.

Our analysis thus has been successful in disentangling the subtle semantic or pragmatic differences between the twosentence final "imperative" particles: phççN and coh. This paper further suggests the importance of investigating theway of grammaticalization as it can help us distinguish the differences clearly between apparently synonymousgrammaticalized words.

ReferencesHopper, Paul J. and Elizabeth Closs Traugott. 1993, rep. 2003. Grammaticalization. Cambridge: Cambridge University Press.Lakoff, George, and Mark Johnson. 1980. Metaphors We Live By. Chicago: University of Chicago Press.Ueda, Hiromi. 2002. "Kumeru-go no Bunmatsu-shi (Khmer final sentence particles)", Takagaki et al. (ed.) Gogaku-kenkyujyo Ronsyu (Journal of

the Institute of Language Research) No. 7, The Institute of Language Research, Tokyo University of Foreign Studies, pp. 35-48.

54


The Mental Processing of ASL Nonce Signs: The Effects of Type Frequency

Brenda Nicodemus & James MacFarlaneAlbuquerque, New Mexico

The traditional explanation of the mental representation of language is that its regularities are expressedas rules, modules, or operations that are independent of the forms to which they apply. However there isa growing body of literature which suggests that linguistic experience is the key factor in the mentalrepresentation of language. Spoken language experiments have shown that schemas gain strength by thenumber of items that exist in languages, i.e., by their type frequency (Bybee 2001). To access theseschemas, researchers utilize nonce-probe experiments. Several studies (Greenberg & Jenkins 1964, Bybee1981, Frisch et al. 2001, Pierrehumbert 1994) have found that the speakers' rankings of acceptability ofnonce forms are strongly based on the statistical tendencies among the existing words of the language.These probabilistic studies suggest that phonological preferences are driven by both similarity of givenpatterns to existing words and the frequency of these patterns in the language.

Based on the evidence from spoken language research, we hypothesized that native ASL users also relyon given patterns in the lexicon to make judgments about the “goodness” of nonce ASL signs. That is, wesuspected that subjects' judgments about nonce signs are based on actual language schemas of high typefrequency. To investigate the effect of type frequency on language processing in ASL, we devised anonce-probe experiment utilizing handshape frequency statistics. Research on ASL (Klima & Bellugi1979, Janzen 2000) has revealed that ASL handshapes occur with varying degrees of frequency in thelanguage (i.e., type frequency).

For our experiment, we videotaped a native user of ASL signing 27 nonce and 9 actual signs. The noncesigns were created by changing the handshapes of the 9 actual signs in three ways. The handshapes werereplaced with (a) handshapes with high type frequency, (b) handshapes with low type frequency, and (c)non-occurring handshapes. This process yielded the 27 nonce signs. The videotaped stimuli were thenpresented to participants (n = 24 adults) on a laptop computer. Participants were asked to rank thestimuli on a Rickert type scale ranging from 1 (impossible sign) to 7 (actual sign). The mean scores ofeach group are reported in the table below.

Mean Scores by Frequency Group

Group 1 Group 2 Group 3 Group 4

ActualSigns

Nonce Signs with FrequentHandshapes

Nonce Signs with InfrequentHandshapes

Nonce Signs with Non-occuring Handshapes

6.87 4.52 3.46 1.90

T-test results show that the difference between the frequency categories are highly significant (p < .05). Itis particularly notable that the participants ranked the signs in Group 2 and Group 3 as significantlydifferent. This result provides behavioral evidence that subjects are sensitive to low or high typefrequency of handshapes in the language. Thus, ASL users utilize their knowledge of type frequency toevaluate the acceptability of signs. These findings support a usage-based model of language whichassumes that preferences of phonetic form are driven by both similarity of given patterns in existingwords and the frequency of these patterns in the language.

55


The Actual Status of So-called Particle Ellipsis in Japanese:Evidence from Conversation, Acquisition, Diachrony, and Language Contact

Tsuyoshi OnoUniversity of Alberta

This paper challenges two commonly held views (i) an understanding of grammar should come fromgrammatical evidence and (ii) there is one grammar for spoken language and written language.

Japanese is known for its postpositional marking of NPs (e.g., by ga and o). However, in conversationNPs are often unmarked, as in 1):

(1) demo kane(o) motte nee mon anohito (ga)but money havenot final.particle that:person‘But that person doesn’t have money.’ (particles not expressed; word order inverted)

This typically leads to analyses concluding that particles are deleted on the surface (deletion analysis).Evidence from diverse sources, however, converges to support the hypothesis that in modern spokenJapanese ga and o don’t exist underlyingly in sentences like 1).

First, appearance of ga/o is rather limited in conversation, roughly 10-40% (e.g., Martin 1975; Fujii/Ono2000). Applying a deletion analysis to these results, one would have to conclude that ga/o are deleted inthe majority of cases. Further, these percentages suggest that using ga/o is a marked activity. In fact, ga/ohave always been discussed in relation to such pragmatic functions as exhaustive listing, newness, andsaliency in the literature (e.g., Kuno 1973; Masunaga 1988), suggesting that they are added to mark suchfunctions rather than simply to indicate grammatical relations.

Second, children first produce sentences without ga/o and don’t fully acquire these particles until quitelate (7th grade according to Ito et al. 1991). In fact, grade-school Japanese-language textbooks includeexercises devoted to the use of these particles. The fact that the use of ga/o has to be drilled in a mediumwhere one can take time to select appropriate expressions again supports the hypothesis that sentenceswithout ga/o are the base form and ga/o are marked expressions.

Finally, NPs without ga/o were the norm in eighth-century Japanese. Sadler (2002) reports that the use ofo increased over the years partly because of the development of a method to read Chinese texts asJapanese. Due to the differences in word order between the two languages, direct objects came to bemarked with o more frequently. Fujii (1991) reports ga increased partly because of the development of atranslation method which imitated the structure of European languages, which have grammatical case.The increases in use of ga/o are reported only for written language. NPs in modern spoken Japanese aremostly unmarked and still look very much like eighth-century Japanese, suggesting that these changestook place only in written language. This is to be expected, since it is difficult to think that spokenlanguage was influenced by the type of contact-induced conscious process described above for writtenlanguage

This study thus suggests though ga/o may have become an integral part of written Japanese partly throughcontact with other languages, their roles in spoken Japanese are limited, leading to the conclusion that inspoken Japanese ga/o are NOT ellipted but rather added for pragmatic purposes. That is, evidence fromdiverse sources shows grammatical representations for spoken language and written language can berather distinct.

56


An Anatomy of the Posture Verb ‘sit’ in Marathi: A Cognitive-Functional Account

Prashant Pardeshi, Kaoru Horie, and Shigeru SatoTohoku University

The posture verbs ‘sit’, ‘stand’ and ‘lie’—all depicting events frequently occurring in our experience—have recently receivedfocused attention in the paradigm of cognitive linguistics (especially papers reported in Newman 2002). These studies haveunraveled diversity in the morpho-syntactic realization of such verbs across languages and the cognitive-functional motivationsunderlying the polysemy exhibited by them. Inspired by these predecessors, the present study focuses on the posture verb sit inMarathi—an Indo-Aryan language of the Indian linguistic area—which exhibits interesting network of semantic extension.Specifically, it takes into account basic, grammaticalized as well as idiomatic usages of sit in Marathi and sheds light on thecognitive factors motivating such semantic extensions.

In Marathi (and other South Asian languages like Hindi, Punjabi, Kashmiri, Gujarati, and Bengali), it is well-known that sitexpresses meanings such as ‘regret; stubbornness’ when used as a vector verb in a compound verb constellation [Main verb +Vector verb] (Hook 1974, Masica 1976, Pandharipande 1990, Abbi & Gopalkrishnan 1991, among others). The main verbappears in the perfective form in such cases as exemplified below.

(1) to nako ti cuk kar-un bas-l-aa [REGRET]he unnecessary that mistake.F do-CP sit-PF-M‘He made a mistake which he should not have done.’

(2) ti aaiskrim-tsaahaTTa dhar-un bas-l-i [STUBBORNNESS]she ice cream-of obstinacy.F hold-CP sit-PF-F‘She insisted on having ice cream.’

In addition to this, sit in Marathi also functions as a marker of ‘durative/continuative’ aspect. The main verb is in imperfectiveform in such a case.

(3) to pustak waats-at bas-l-aa [DURATIVE]he book read-IMPF sit-PF-M‘He kept on reading the book.’

We claim that these twin usages of “regret, stubbornness” (having negative connotation) and “continuity” (having positiveconnotation) are grounded in the image-schematic meaning associated with the act of sitting. The “regret, stubbornness” usagecorresponds with “drooping posture” as envisaged by Lakoff & Johnson (1980) in their “BAD IS DOWN” metaphor. The“continuity” usage, on the other hand, corresponds with the “spatial stability and temporal continuity” acquired after assumingthe sitting posture and has been widely attested across languages (Newman 2002). Idiomatic usages of sit endorse our claim asshown below.

(4) tyaacaa dhandaa bas-l-aa [IDOMATIC-NEGATIVE CONNOTATION]his business.M sit-PF-M‘His business deteriorated.’

(5) tyaacaa haat bas-l-aa [IDIOMATIC-POSITIVE CONNOTATION]his hand.M sit-PF-M‘He mastered the skills.’

Further, a contrast with auxiliary verb stay—which also expresses ‘durative/continuative’ aspectual meaning like sit—lendssupport to our foregoing claim. While stay exclusively expresses continuity of events implying positive connotations, sit canexpress continuity of events with negative implications.

(6) prayatna kar-at rahaa/*bastry do-IMPF stay/sit‘Keep trying hard.’

(7) time paas kar-at basu nakos/*raahu nakostime pass do-IMPF sit not/stay notLit. ‘Do not sit/stay wasting time.’ ‘Don’t waste time.’

In sum, the insights gained through this study—in terms of both morpho-syntax and semantics—contribute new dimensions to thetypology of the posture verb ‘sit’. Furthermore, it confirms the similarities observed across various languages reported inNewman (2002).

ReferencesAbbi, A. & D. Gopalkrishnan. 1991. Semantics of Explicator Compound Verb in S. Asian languages. Studies in the Linguistic Sciences 13.2: 161-180.Hook, P. E. 1974. The Compound Verb in Hindi. Ann Arbor: Center for South & Southeast Asian Studies, University of Michigan.Lakoff, G. & M. Johnson. 1980 Metaphors We Live By. Chicago: Chicago University Press.Lakoff, G. 1990. The Invariance Hypothesis: is abstract reason based on image-schema? Cognitive Linguistics 1.1: 39-74.Masica, C. P. 1976. Defining a Linguistic Area. Chicago: University of Chicago Press.Newman, J. (ed.) (2002). The Linguistics of Sitting, Standing and Lying. Amsterdam: John Benjamins.Pandharipande, R. 1990. Serial Verb Construction in Marathi. In Joseph, B. D. and Zwicky A. M (eds.) When Verbs Collide: Papers from the 1990 Ohio State Mini-

Conference on Serial Verbs. Columbus: Dept. of Linguistics, The Ohio State University. Pp: 178-199.

57


The Hands Are Part of the Package:Gesture, Common Ground and Information Packaging

Fey ParrillUniversity of Chicago

This paper reports on experimental research investigating the relationship between common ground(Clark & Haviland 1977) and the representational features which appear in gesture, as part of continuingwork on gesture’s role in information packaging. Information packaging describes how speakers shapeutterances to meet the expected needs of their interlocutors (Chafe 1976, Prince 1981).

Levy & McNeill (Levy & McNeill 1992) have shown that gestures correlate with aspects ofinformation packaging –more complex representational gestures, for example, tend to occur withdiscourse-new elements. Thus, while information packaging has syntactic reflexes, gesture serves as anon-linguistic index of a speaker’s communicative decisions with respect to certain discourse elements,allowing us access to information which would be missed by considering the verbal channel alone.

Information packaging interacts with common ground, or knowledge, beliefs, and suppositionsshared by interactants (Clark 1992). As a simple case, information which has been introduced into adiscourse is assumed to be part of the common ground. Speakers direct attention to information which isbeing added to common ground with, for example, the use of the indefinite article. This project attemptsto address the extent to which there are gestural cues to common ground as well.

In order to investigate this relationship experimentally, we first isolated gestural elements ofinformation packaging which would be susceptible to manipulations of common ground. Participantswatched a cartoon in which a cat floats into the air over a flight of stairs. His body becomes liquid andconforms to the shape of the stairs as he moves down them. When asked to describe this event,participants received one of the following prompts:

Can you tell me how the cat melts?Can you tell me how the cat melts down the stairs?

The prompts functioned to manipulate the discourse status of the ground element –the object over whichmotion proceeds (Talmy 1985) --by either referencing it directly (making it discourse-old), or by makingno reference to it. When the ground element is discourse-old, it should be less likely to appear in gesture.While participants produced similar speech across the two conditions, their gestures varied according tothe prompt received: speakers were significantly more likely to produce a gesture depicting ground whenthey did not receive the prompt which included this element.

The current phase of the experiment manipulates common ground by having an interlocutor eitherpresent or absent while participants view the stimulus described above. Interlocutors will then requestnarrations using the same prompts. We predict that both speech and gesture will contain reference to theground element irrespective of prompt, as in the previous study, but that the observed difference ingesture across prompts will disappear. Even when the interlocutor refers to the stairs with her prompt,this information should still be considered discourse-new because the speaker knows her interlocutor hasnot seen the stimulus. Such results, in combination with our previous findings, would provide insight intoconstraints on discourse status which operate in the packaging of information in speech and gesture.

References

Chafe, Wallace. 1976. Givenness, contrastiveness, definiteness, subjects, topics, and point of view. In Charles N. Li (Ed.), Subject and Topic, 27-55. New York: Academic Press.

Clark, H.H. & Haviland, S.E. 1977. Comprehension and the given-new contract. In R. Freedle (Ed.), Discourse Production and Comprehension,1-40. Norwood, NJ: Ablex.

Clark, H.H. 1992. Arenas of Language Use. Chicago: The University of Chicago Press.Levy, Elena T., & McNeill, David. 1992. Speech, gesture, and discourse. Discourse Processes, 15, 277-301.Prince, E. 1981. Toward a taxonomy of given-new information. In P. Cole (Ed.), Radical Pragmatics, 223-255. NY: Academic Press.Talmy, Leonard. 1985. Lexicalization patterns: semantic structure in lexical forms. In Timothy Shopen (Ed.), Language Typology and Syntactic

Description, Vol. 3, 57-149. Cambridge: Cambridge University Press.

58


Bilingual Thinking for Speaking

Julia PetersUniversity of Alberta

The way in which native speakers are conditioned to think for the purposes of successful oralcommunication (i.e., 'thinking for speaking', Slobin 1991) is dependent on the internal linguistic facts of their L1.When speakers acquire a second language, they may encounter a situation in which their native thinking forspeaking does not effectively map onto the linguistic structures of the subsequent language. Therefore, the bilingualmust either learn to think in another way, or to adopt some kind of efficient translation mechanism. For example,Tatsumi (1997) found evidence of Japanese-English bilinguals simultaneously thinking in both languages whenspeaking only one. Similarly, this study is an investigation of whether French-English (F-E) bilinguals demonstratethinking in French when speaking in English in terms of how Path and Manner of Motion are expressed in oralpredicate constructions.

French, as a Romance language, falls into Talmy's (2000) categorization of 'verb-framed languages'characterized by Path of Motion being encoded in the verb root.

(1) Elle descend l'escalier. 'She descends the stairs'

Although a certain number of verb-framed constructions also exist in English (as demonstrated by the gloss in 1),the typical classification for English is that of a 'satellite-framed language' in which the Path of Motion is notencoded in the verb root but rather as a satellite constituent such as a preposition or particle.

(2) She goes down the stairs.

In the satellite-framed constructions, the verb root is 'free' of Path information so, in essence, has room forthe encoding of other semantic properties such as Manner.

(3) She walks/runs/glides/slides down the stairs.

By contrast, Manner of Motion must be expressed using a subordinate structure in French, often resulting in lessnatural sounding discourse.

(4) Elle descend l'escalier en marchant/courant/glissant

As a result, speakers of Romance languages often omit mention of Manner information altogether (see Slobin 1996,for evidence in Spanish).

If French and English are categorized accurately according to Talmy's typology, native speakers of Frenchand English will form different ways thinking for speaking in relation to Path and Manner of Motion. Do F-Ebilinguals, then, learn to think in a new way when they acquire their L2? Or, do they simply learn how to translatebetween L1 and L2 forms?

The methodology for this study is a basic oral picture description task in English involving scenes with anActor undergoing Motion. The experiment is divided into two parts with the first requiring a monolingual languagemode (Grosjean 2001) before establishing a bilingual mode for the remaining section.

The response patterns of F-E bilinguals are compared to monolinguals. The analysis will focus on (a)location of Path encoding (i.e., verb root vs. satellite), (b) frequency of Path encoding, and (c) frequency of Mannerencoding. Results will be discussed with reference to such variables as (a) speaker criteria (e.g., dominant language,age of acquisition), (b) stimulus criteria, (e.g., cognate status, verb semantics), and (c) language mode. The accuracyof Talmy's verb classifications for French and English will also be explored.

References

Grosjean, F. (2001). The bilingual’s language modes. In J. Nichol (Ed.), One mind, two languages: Bilingual language processing (pp. 1-22).Oxford: Blackwell.

Slobin, D.I. (1991). Learning to think for speaking: Native language, cognition, and rhetorical style. Pragmatics 1: 7-26.Slobin, D.I. (1996). Two ways to travel: Verbs of motion in English and Spanish. In M. Shibatani & S.A. Thompson (Eds.) Grammatical

constructions: Their form and meaning. Oxford: Clarendon Press.Talmy, L. (2000). Toward a cognitive semantics, Vol. II: Typology & process in concept structuring. Cambridge, MA: MIT Press.Tatsumi, T. (1997). The bilingual's thinking for speaking: Adaptation of Slobin's frog story experiment to Japanese-English bilinguals. Thesis,

Sophia University.

59


Discourse Frequency and Reductive Change: The Case of Lakota eta7@(ha7@)/-ta7(ha7) ‘from’

Regina PustetBoulder, Colorado

In contemporary Lakota (Siouan language family), there are four (depending on the analysis, five) different markersfor the case role 'SOURCE' which share enough phonetic similarities to be considered historically related: thepostpositions etá 4ha 4 and etá 4, and the suffixes -ta 4ha 4 and -ta 4; for all these elements, the translation 'from' isappropriate. These markers derive from the verb etá 4ha 4 'to be from'. However, the members of this diachronicallycoherent set of markers exhibit different properties on three parameters which are considered essential in theories oflanguage change: semantic content, structural complexity, and discourse usage. The semantic distinctions within theabove group of SOURCE markers are, on the whole, subtle; however, a comparison of the text collections Deloria(1932) and Pustet (forthcoming (a)), which capture developmental stages of the Lakota language which are about 70years apart, shows that the markers etá 4ha 4, etá 4, -ta 4ha 4, and -ta4 have undergone dramatic changes in their discoursefrequencies. All in all, the percentage of the longer forms etá 4ha 4 and -ta 4ha 4 decreases, while that of the shorter formsetá 4 and -ta 4 increases. Thus, in this case, in line with the observations made in the frameworks of "usage-basedmodels" (in particular, Barlow & Kemmer 2000; Bybee 2001; Bybee & Hopper 2001; Haiman 1985; Hopper 1987;Langacker 1987; also, cf. Pustet 2004, Pustet (forthcoming (b)) and grammaticalization theory (e.g. Heine at al.1991), structural change is reductive in nature.

But the most important result of the present study is that the speed at which such shortening proceeds is notthe same for all collocations the elements in question appear in. "Collocation", in this context, designates acombination of a source marker with the nominal or pronominal element it follows or is suffixed to. The decisivefactor that speeds up shortening with specific collocations and delays it with others seems to be the discoursefrequency of individual collocations. For instance, reduction of -ta 4ha 4 to -ta 4 progresses at a faster or slower rate,depending on the element which precedes the source marker in discourse. The higher the overall frequency of thecollocation 'X + -ta 4ha 4' in discourse, the more the likelihood of use of the shortened form -ta 4 increases. Analogousobservations, especially with regard to phonological reduction processes in English, have been made in Berkenfield(2001), Bush (2001), Jurafsky et al. (2001), and Krug (1998, 2001), among others.

For instance, in Deloria (1932), collocations of -ta 4ha 44 with the demonstratives hé 'that' and lé 'this' occupytop positions in the overall frequency ranking of linguistic items occurring with SOURCE markers. In Pustet(forthcoming (b)) these demonstratives still hold top positions in the frequency scale of collocations with SOURCEmarkers, but now the only structural realization of the SOURCE marker occurring in combination withdemonstratives is -ta 4. I.e., in Pustet (forthcoming (b)), -ta 4-forms have completely replaced the earlier -ta4ha4-forms inthis collocational context. ta 4ha 4-collocations encountered (1932) in Deloria which exhibit lower discourse frequencythan collocations with demonstratives show a far less pronounced tendency towards shortening in Pustet(forthcoming (b)).

References

Barlow, M. & S. Kemmer (eds.) 2000. Usage-based models of language. Stanford: CSLI.Berkenfield, C. 2001. The role of frequency in the realization of English that. In: Bybee, J. L. & P. Hopper (eds.), 281-307.Bush, N. 2001. Frequency effects and word-boundary palatalization in English. In: Bybee, J. L. & P. Hopper (eds.), 255-280.Bybee, J. L. 2001. Phonology and language use. Cambridge: Cambridge University Press.Bybee, J. L. & P. Hopper (eds.) 2001. Frequency and the emergence of linguistic structure. Amsterdam: Benjamins.Deloria, E. 1932. Dakota texts. New York: Stechert.Haiman, J. 1985. Natural syntax. Iconicity and erosion. Cambridge: Cambridge University Press.Heine, B.; U. Claudi; & F. Hünnemeyer 1991. Grammaticalization. A conceptual framework. Chicago: University of Chicago Press.Hopper, P. J. 1987. Emergent grammar. BLS 13:139-157.Jurafsky, D.; A. Bell; M. Gregory; W. D. Raymond 2001. Probabilistic relations between words: evidence from reduction in lexical production.

In: Bybee, J. L. & P. Hopper (eds.), 229-254.Krug, M. G. 1998. String frequency: a cognitive motivating factor in coalescence, language processing and linguistic change. Journal of English

Linguistics 26:286-320.Krug, M. G. 2001. Frequency, iconicity, categorization: evidence from emerging modals. In: Bybee, J. L. & P. Hopper (eds.), 309-355.Langacker, R. 1987. Foundations of cognitive grammar. Vol. 1: Theoretical prerequisites. Stanford: Stanford University Press.Pustet, R. 2004. Zipf and his heirs. Language Sciences 26: 1-25.Pustet, R. (forthcoming (a)). Lakota texts. Lincoln: University of Nebraska Press.Pustet, R. (forthcoming (b)). On discourse frequency, grammar, and grammaticalization. In: Frajzyngier, Z. & D. Rood (eds.), Linguistic

Diversity and Language Theories. Amsterdam: Benjamins.

60


Spanish Analytic Causatives with hacer: Economic and Processing Motivations

Nunghatai RangponsumritChulalongkorn University

This paper aims to account for the differences between two different constructions of analyticcausatives involving hacer in Spanish, namely those with an infinitive complement, as shown in (1), andthose with a finite clause complement, as shown in (2). Previous studies on Spanish analytic causativesexplain the differences between the two constructions in terms of syntactic properties and meanings.However, after examining a 400,000-word corpus, it is found that there are other factors that determinethe choice between the two constructions - the length and complexity of the complement. Length andcomplexity are measured by considering a) the number of arguments and modifiers of the embeddedverb, b) whether the subject of the embedded verb is realized as a full NP or a pronoun or is dropped; c)whether the complement clause is a simple or compound clause. The explanation for the importance ofthese factors can be given in terms of economy principle and limitations of short-term memory. Ingeneral, the infinitive structure is usually the preferred form for economic motivation, since it requires noconjunction or verb conjugation. On the other hand, in case of ‘heavy’ complement clauses which posedifficulty for processing, speakers tend to use the finite clause construction, as it is a more explicit formwhere the subject of the embedded verb is often present and the embedded verb is conjugated thus helpsidentify what its subject is. In summary, speakers prefer the more economical form as long as it can beprocessed with no difficulty.

As language is not determined by only syntax and semantics, but also by other factors, the presentcorpus-based study gives an account for choice between different linguistic forms from the perspective ofeconomic and language processing motivation, to supplement previous semantic explanations providedby several linguists, thus giving a broader picture of analytic causative constructions in Spanish.

(1) María hace que Pedro salga.Maria makes that Peter goes.out.SUBJ

‘Maria makes Pedro leave.’(2) María le hace salir.

Maria him makes go.out.INF

‘Maria makes him leave.’

Selected references

Curnow, Timothy Jowan.1994. Semantics of Spanish causatives involving ‘hacer’. Australian Journal of Linguistics; 13, 2, Dec,165-184.

Garcia-Miguel, José M. and Andrea Pascual. 2001. Conceptual integration in Spanish reflexive causatives. 7th InternationalCognitive Linguistics Conference (ICLC 2001). UCSB, Santa Barbara, July 22-27, 2001.

Haiman, John. 1983. Iconic and economic motivation. Language, 59(4), 781-819.Shibatani, Masayoshi and Pardeshi, Prashant. 2002. The causative continuum. In The Grammar of Causation and Interpersonal

Manipulation. Edited by Masayoshi Shibatani. Amsterdam: John Benjamins. pp 85-126.Zipf, George K. 1935. The Psychology of Language. New York: Houghton Mifflin.

Data

Oral corpora available at an ftp site of the Laboratory of Computational Linguistics of the Universidad Autonoma de Madrid:ftp://ftp.lllf.uam.es/pub/.

61


Impersonals

Nathalie SchapanskyUniversity of Alberta

The purpose of this paper is to discuss the relation between form and function with respect to impersonalconstructions. The term ‘impersonal passive’ has been applied to various constructions in which: a) thegrammatical subject, when it is expressed, is an expletive (or dummy) pronoun or an impersonalinflection on the verb, while the logical subject is demoted to some other position; b) the verb may or maynot display the passive morphology; c) the object, if there is one, is not promoted to subject; d) there mayor may not be an agentive by-phrase (Perlmutter 1978). This is illustrated in (1) for German (Blevins2001) and (2) for Breton.

1. In der Küche wurde (von vielen Leuten) geraucht.In the kitchen was (by many people) smoked‘There was smoking (by many people) in the kitchen’

2. Koumoul du ‘weler duhont (*get en dud)clouds black see.prs.imp over.there (by the people)‘One sees black clouds over there = black clouds are seen over there’

However, not all impersonal constructions (ICs) are passive. While ICs vary in form, they perform asingle function. In order to relate the forms of ICs to their function, it is necessary to take personalpassives (PPs) as a means of comparison and discuss: 1) their formal properties (valence changing,subject backgrounding, object foregrounding); 2) their semantic properties (definiteness effect andrestriction on tense and predicate type (unaccusative or unergative); 3) their syntactic structure(predicative/non-predicative) as well as their conceptual structure (internally uniform or heterogeneous);4) their function in languages having such constructions (e.g., German, Breton, Welsh, Polish, Ukrainian,Lithuanian, Estonian); and 5) their relation to other types of impersonal constructions of the French il-type, as in (3) (Legendre 1989).

3. Il déjeune beaucoup de linguistes dans ce restaurant.It lunches many of linguists in this restaurant = many linguists eat lunch in this restaurant

(3) is incorrectly predicted to be ungrammatical in any theoretical account since the unergative predicatedéjeuner lacks an internal argument to license the il.

The results show that PPs and ICs background the subject and are conceptually homogeneous(internally uniform). PPs are valence changing and foreground the object unlike ICs which show adefiniteness effect. Personal and impersonal passives cannot be constructed out of unaccusativepredicates, unlike active impersonals which can exhibit a restriction on tense. PPs have a predicative(bipartite, subject/predicate) structure while ICs have a non-predicative (unipartite) structure. Theimpersonal subject is a non-determined human and allows an agentive by-phrase depending on whetherthe IC is perceived as active or passive. ICs tend to mark evidentiality, either inferred from what isperceived, as in (1)-(3), or from hearsay, as in (4) for Lithuanian (Blevins 2001).

4. Jõn búta kareino.3s.m.gen been.imp soldier.s.m.gen‘(They say) he was a soldier’

The conclusion drawn is that impersonal constructions express evidentiality. Evidentiality is marked byan impersonal subject which prevents the foregrounding of the object if there is one, thus yielding a non-predicative structure.

References

Blevins, James P. 2001. Impersonals and passives. University of Cambridge, draft.Legendre, Geraldine. 1989. Unaccusativity in French. Lingua 79: 95-164.Perlmutter, David. 1978. Impersonal passives and the Unaccusative Hypothesis. BLS 4: 157-189

62


Non-literal Motion in Child Greek and English

Stathis Selimis and Demetra KatisUniversity of Athens

Competence in use of metaphor has been claimed to be rudimentary in preschool children, requiring along developmental period to mature (e.g. Vosniadou & Ortony 1983). Such claims may however haveunderestimated children’s ability to produce/comprehend metaphors, because of methodological biases(Colston & Kuiper 2002) as well as a narrow conception of metaphor as implicit similarity (Özc¶alıs¶kan2002). More recent research which adopts the Conceptual Metaphor Theory, has shown that metaphorcan be handled by children at an earlier age. According to Özc¶alıs¶kan, English- and Turkish-speakingchildren aged 4-5 years do understand abstract concepts structured via the common source domain ofmotion in semi-experimental situations. Based on a similar view of metaphor, this study undertakes across-linguistic comparison of metaphoric and more widely of all non-literal uses of motion verbs inGreek and English child-adult natural conversations in the age range of 1;8 to 5 years. Included areclearly metaphoric cross-domain mappings, but also akin phenomena, e.g. the presentation of (veridically)caused motion as (non-veridically) self-generated one (e.g. Greek edho tha pai i fotoghrafia/ lit. “thephoto will go here”). In fact, we may speak of a continuum of literal to non-literal uses. Though the latterare rather restricted, they appear in both child and child-directed speech even before the age of 2 years.Moreover, metaphoric uses seem earlier and more frequent in the Greek data as opposed to the Englishone, though it is uncertain whether this reflects individual or cross-linguistic variation. For instance, theSTASIS-to-MOTION metaphor appears after 3 years in Greek as opposed to later than 4 years in English,while the TIME-to-MOTION mapping is encountered only in the Greek child speech. Moreover, in bothlanguages metaphors are encountered with Path-lexicalizing verbs, in spite of the fact that English is, incontrast to Greek, a Manner-lexicalizing language in the representation of physical motion (Papafragou etal. 2002). Özc¶alıs¶kan’s claims that cross-linguistic variation in the semantic domain of motion is evidentnot only in the literal uses of the lexicon but also in the metaphorical extensions of it are not supported inearly natural speech. It may therefore be surmised that it is easier for young children across languages tomove into metaphor with Path verbs, which seem conceptually simpler than the more specific Mannerencoding verbs, even if their language is a Manner-oriented one in the representation of physical motion.

References

Colston, H.L., & M.S. Kuiper 2002. Figurative language development research and popular children’s literature: why we shouldknow, “Where the wild things are”. Metaphor and Symbol, 17(1), 27-43.

Özc¶alıs¶kan, S¶. 2002. Metaphors we move by: a crosslinguistic-developmental analysis of metaphorical motion events inEnglish and Turkish. Unpublished Ph.D. dissertation. University of California, Berkeley.

Papafragou, A., C. Massey, & L. Gleitman. 2002. Shake, rattle, ’n’ roll: the representation of motion in language and cognition.Cognition 84, 189-219.

Vosniadou, S., & A. Ortony. 1983. The emergence of the literal-metaphorical-anomalous distinction in young children. ChildDevelopment, 54, 154-161.

63


Contextualization in ASL-English Interpretation:A Question of Grammar or Discourse Strategy

Barbara Shaffer & Terry JanzenUniversity of New Mexico and University of Manitoba

When two people who share a common language engage in discourse, they make assumptionsabout information that is active within each interlocutor’s consciousness (Chafe 1994), and chooseappropriate grammatical coding mechanisms for expressing implicit versus explicit information in thatdiscourse event. Different languages, however, may encode implicit and explicit information differently.When two people who do not share a common language wish to interact, an interpreter is often employedto bridge the linguistic gap, but introducing an interpreter into a discourse event affects the very nature ofthe interchange because the interpreter will also make assumptions about each of the interlocutor’sknowledge store.

A practice that has been growing as of late in English to ASL interpretation is the use of what arebroadly called “expansions” (Lawrence 1995), which Lawrence claims are grammatical elements of ASL.In this analysis, ASL is said to be “high context” in that “information is not easily implied and in fact,must be explicit” (Lawrence 1995: 212), and this highly explicit nature is apparent in the grammaticalcode of ASL. Thus this theory claims that such explicitness is required by the grammar and must bechosen by the signer, whether primary interlocutor or interpreter.

In this paper we propose that the grammar of ASL has no such “explicitness” requirement.During interpretation, the interpreter must attend to the cognitive domain of perceived shared and non-shared information. The problem is that what the discourse participants (i.e., not the interpreter) assume tobe within the shared conscious domains of their interlocutors is not necessarily equally shared by theinterpreter. Consequently, what the interpreter chooses to make implicit or explicit in either language, andhow this is accomplished using linguistic ‘packaging strategies’ particular to each of the interpreter’s twoworking languages, is not necessarily what the interlocutors would choose to represent their ownconceptualizations of what is transpiring in the discourse. The resulting message, which should representwhat is in the minds of the interlocutors, in fact does not, and the message is skewed (Shaffer and Janzen2002, Janzen and Shaffer 2003). It is often reported by recipients of interpretation that even though thediscourse appeared to be fluent and grammatical, the resulting message is still confusing.

This study examines features of interpreted discourse where implicit and explicit coding is based onthe interpreter’s construal of the information being expressed and the interpreter’s decisions regardingwhat to profile in the target message. Our data consists of videotaped examples of actual ASL-Englishinterpretation. Here we examine the effects of prescriptive “expansions” in interpretation and the effectthese create in the overall message, regarding both Primary Information (the message) and SecondaryInformation (the effect of, for example, text structure) according to Gile (1995). We suggest thatcontextualization as a successful discourse strategy (Gile 1995) in interpretation into any target languageprovides a more correct approach to the expression of shared or non-shared knowledge.

References

Chafe, Wallace. 1994. Discourse, consciousness, and time: The flow and displacement of conscious experience in speaking and writing. Chicago:The University of Chicago Press.

Gile, Daniel. 1995. Basic Concepts & Models for Interpreter & Translator Training. Amsterdam/Philadelphia: John Benjamins.Janzen, Terry, & Barbara Shaffer. 2003. Implicit versus explicit coding across two languages: Mismatches of cognitive domains during

interpretation. Paper presented at the 8th International Cognitive Linguistics Conference, Logroño, Spain, July 20-25, 2003.Lawrence, Shelley. 1995. Interpreter discourse: English to ASL expansion. In Elizabeth A. Winston (ed.), Mapping our course: A collaborative

venture, Proceedings of the Tenth National Convention, Conference of Interpreter Trainers. October 26-29, 1994. USA: Conference ofInterpreter Trainers.

Shaffer, Barbara, & Terry Janzen. 2002. Topic marking: What signers know but interpreters don’t. Paper presented at the Association of VisualLanguage Interpreters of Canada 14th National Biennial Conference. July 22-26, 2002. Halifax, Nova Scotia.

64


A Cognitive Approach to Polysemous Extension in Verbs of Tactile Perception:Exploring a “TOUCH is COMPREHENSION and UNDERSTANDING” Idealized Cognitive Model

Christopher C. ShankUniversity of New Mexico

One of the more robust areas of research in Cognitive linguistics has been the exploration of theexperiential and cognitive dimensions of figurative language and the role it plays in areas such as categorization,language change, and semantic extension. By exploring the roles that cognitive processes such as prototypecategorization, metaphor and metonymy, polysemous extension, and semantic networking play in human cognitionand the structure of language, we have become acutely aware of the fundamentally embodied and imaginative naturehuman thought and conceptual processes.

This phenomenon is especially evident when one attempts to understand how the primary perceptual senses(i.e. seeing, hearing, touching, tasting and smelling) shape and influence the transfer, between the physical domainof perception via external interactions or experiences and the internalized conceptual domain of the mind. Onceperceptual sensory input is internally processed and categorized it is often understood and given meaning throughabstractions and / or metaphorical processes in the forms of Idealized Cognitive Models such as “SEEING IS

KNOWING” or “HEARING IS KNOWING” (Lakoff & Johnson 1980; Sweetser, 1990).This study seeks to expand our knowledge and understanding in this experiential field of inquiry by examining

the important roles that tactile verbs of perception, under the rubric of a “TOUCH is COMPREHENSION andUNDERSTANDING” I.C.M., fulfill as the expression of one of the five primary human senses. It explores theconceptual and semantic connections between the physical and metaphorical notion of touch and tactilemanipulation, and the development cognitive comprehension and understanding, through a combination of corpuslinguistic methodology and a cognitive linguistic analysis. A diachronic analysis of evolution of meaning, prototypeanalysis, polysemous extensions of following verbs (and their inflected forms) of tactile perception: TOUCH, HOLD,HANDLE, GRASP, SEIZE, and FEEL will be presented. This will be followed by an analysis of synchronic data,consisting of the same verb set, via a corpus of spoken American English. A functional / cognitive framework inconjunction with an experiential approach to the construction and understanding of meaning will be presented inorder to revel an underlying “TOUCH is COMPREHENSION and UNDERSTANDING” I.C.M. The analysis is intended tohighlight the structural multi-dimensionality and interconnection between the different senses of the lexemes oftactile perception and the larger role this network structure plays in comprehension.

In addition to examining the semantic and conceptual relationships between these tactile verbs ofperception, the role and importance of kinesthetic metaphors in cognitive processes of creating and sustainingcomprehension and understanding will also be examined in detail. Constructions taken from the spoken Englishcorpora will be analyzed in terms of their literal or increasingly metaphorical usage within an “ExtendedMetaphorical Domain” framework. This framework will also be used as a means of conceptualizing and discussingthe transitive nature of these tactile verbs of perception, drawing from (Croft, 2004; Langacker, 2000; Rice, 1987;Talmy, 2003) as the constructions move from the literal/physical realm into increasingly metaphorical and abstractusage domains. The study will conclude with some final comments concerning the role of the human body and ourperceptions as a way of grounding and explaining human conceptual systems and the construction of meaning.

References

Croft, W. and Cruse, A. (2004) Cognitive Linguistics: Cambridge University Press.Geeraerts, D. (1997). Diachronic Prototype Semantics: A Contribution to Historical Lexicology (Oxford Studies in Lexicography and

Lexicology). Clarendon Press.Johnson, M. (1987) The Body & the Mind The Bodily Basis of Meaning, Imagination, & Reason. Chicago: U of Chicago Press.Lakoff, G. (1987). Women, Fire, and Dangerous Things. Chicago: University of Chicago Press.Langacker, R (1991). Foundations of Cognitive Grammar Volume II Descriptive Application. Stanford: Stanford University Press.Langacker, R (2000). Grammar and Conceptualization. New York: Mouton de Gruyter.Nerlich, B. (Ed.) (2003). Polysemy: Flexible Patterns of Meaning in Mind and Language (Trends in Linguistics: Studies and Monographs, 142):

Walter de Gruyter, Inc.Rice, Sally A. (1987) Towards a Cognitive Model of Transitivity. Ph.D. Dissertation. University of California, San Diego. UMI.Sweetser, E. (1990). From Etymology to Pragmatics. Cambridge: Cambridge University Press.Talmy, L. (2003) Toward a Cognitive Semantics - Volume 1: Concept Structuring Systems: MIT Press.Traugott, E, & Dasher, R. (2002) Regularity in Semantic Change: Cambridge University Press.

65


Transitivity in Japanese––A Discourse View

Mitsuaki ShimojoUniversity of Buffalo

While the nominative –ga and accusative –o is the common case frame in Japanese, the languageexhibits split case marking with stative or non-volitional predicates, whose objects are marked with thenominative –ga (Kuno 1973, Tsunoda 1994). Yet the boundary between the two case frames is not clear-cut because certain predicates may take either frame, as typically observed with stative derivatives, as in(1):

(1) eega-ga/o mi-taimovie-NOM/ACC see-want‘(I) want to see a movie.’

The observed acceptability of the ga/o alternation may be captured in terms of aspectual properties orvolitionality denoted by the predicates, i.e., the co-existence of activity/volition and state/non-volition[e.g., ‘to see’ and ‘to want’ respectively for (1)]. However, this approach fails to elucidate speakers’choice of a marker in the alternation since the lexical-semantic properties per se do not single out aparticular marker.

By examining colloquial Japanese data (eight sets of two-party informal conversation, four hours intotal), this paper suggests a functional view of the ga/o alternation in order to pinpoint the contrastbetween the two markers, hence describe speakers’ choice in the alternation.

The following procedure was used for the data analysis. Among 344 utterances where the nominative–ga is possible for object marking, I obtained 40 utterances in which the objects are actually marked with–ga (35) or –o (5). I examined the acceptability of cross-case marking for these utterances, usingacceptability judgment rating by 33 native speakers. As a result, I obtained 5 nominative and 3 accusativeutterance tokens for which the ga/o alternation may be considered as possible. These 8 tokens werequalitatively examined as the basis for the discussion summarized below.

• As the nominative –ga is the unmarked case frame in terms of overall frequency, it is functionally theunmarked choice in the ga/o alternation. In all three cases, the accusative –o is associated with aspecialized context, where a speaker repeats or paraphrases a previously given proposition. None ofthe nominative tokens exhibit this property.

• Repeated/paraphrased propositions denote important information for the purpose of the conversation,as the information persists in the context. This is consistent with the overall tendency that informationmarked with the accusative –o exhibits greater referential persistence than the nominative –ga in theconversation data.

• Importance associated with the accusative marking is also consistent with the widely recognized trendthat objects, not subjects, represent pragmatic focus in transitive sentences.

• Overall, the findings suggest that a choice of a case marker in the ga/o alternation does not denote thespeaker’s foregrounding/backgrounding of certain aspectual properties/volitionality denoted by thepredicate. Instead, the choice represents certain mental processing instructions for the hearer (Givón1993) so that important information is processed and represented as such in the hearer’s mentalrepresentation of the text.

References

Givón, Talmy. 1993. Coherence in text, coherence in mind. Pragmatics & Cognition 1, 171-227.Kuno, Susumu. 1973. The structure of the Japanese language. Cambridge: MIT Press.Tsunoda, Tasaku. 1994. Transitivity. In Asher, A. E. (ed.), Encyclopedia of Language and Linguistics, 4670-77. Oxford:

Pergamon Press.

66


A Corpus-Based Study of Semantic Extensions of Sensory Adjectivesfrom Synchronic and Diachronic Perspectives

Mika ShindoNational Institute of Information and Communications Technology (JAPAN)

This paper focuses on semantic extensions of English sensory adjectives towards abstract meanings. First, Isynchronically analyze their actual usages based on corpora and lexical databases of hierarchical thesauri andsecond, I examine diachronic data of semantic changes of each word in order to explain the diversity betweensensory adjectives included in the same original perceptual domain. I collected all adjective + noun co-occurrencesfor 64 sensory adjectives (Williams 1976) from the BNC, for each attributive adjectival meaning is represented bythe noun modified (Shindo et al. 2004). All the modified nouns were classified by the WordNet 2.0 hierarchicalthesaurus to avoid individual biases. The data obtained by this corpus-based study reveal that:

1. Contrary to Sweetser’s claim (1990) that perception words metaphorically extend to abstract concepts in a systematic way, sensoryadjectives from the same original domain do not necessarily extend similarly.

2. Each adjective extends rather uniquely to abstract domains restricted by its original characteristics.

For example, brightness adjectives used for extended meanings tend to modify the following nouns (the numbersfollowing each noun indicate frequencies.):

brilliant: “Person” (e.g. man (19), player (17), student (8), scholar (5), etc.) and “Human action” (e.g. performance (21), career (17), save(11), goal (9), success (9), job (9), start (9), etc.);

bright: “Person” (e.g. boy (22), girl (12), child (11), pupil (4), scholar (5), etc.);clear: “Abstraction (Communicative content)” (e.g. evidence (156), idea (154), view (115), indication (115), understanding (103),

distinction (101), statement (81) etc.).

Sensory adjectives originating in brightness have extended to mean divergent concepts, although related tointellection (as has been commonly discussed in metaphor theory).

The contrast between bright and clear is especially noticeable from my diachronic perspective. The OEDgives almost the same etymology for these two words: (s.v. bright) “shining, emitting, reflecting, or pervaded bymuch light”; (s.v. clear) “expressing the vividness or intensity of light: brightly shining, bright, brilliant.” However,my diachronic examination based on the Helsinki Corpus, Chaucer, and Shakespeare reveals the intrinsic uniquenessinvariable in their semantic changes. Even if the domain has changed from concrete to abstract, bright describes anentity radiating light, while clear describes an unobstructed state. These two inherent meanings reflect our twoperspectives perceiving brightness: objective and subjective (Langacker 1985, 1999, Traugott 1995). In a viewingarrangement, bright merely refers to an objectively observed entity, while clear refers to subjectively construedunobstructedness in the process of viewing. A meaning of clear that does not emerge until the latter stage of itssemantic development, “free from obstruction,” prominently comes to function as a verb, as shown in “a desire tocleare all manner of scruples,” (Helsinki Corpus. EmodE II [1570-1640], CEOFFIC2). This verbalization may beexplained by the increasing intensity of involvement of the conceptualizer in the situation, that is, subjectification.By examining the diachronic semantic changes of bright and clear, we can elucidate objective and subjective humanconstruals of perception.

Analysis of the semantic changes of adjectives empirically based on synchronic and diachronic corporaclarifies the developmental processes by which humans construe and conceptualize the world.

ReferencesThe British National Corpus. (http://www.natcorp.ox.ac.uk/) [BNC]The Helsinki Corpus of English Texts (Diachronic Part), compiled by M. Rissanen, O. Ihalainen, & M. Kytö. U of Helsinki.Langacker, R. W. 1985. “Observations & Speculations on Subjectivity,” in J. Haiman (ed.) Iconicity in Syntax, 109-150. Amsterdam: Benjamins.Langacker, R. W. 1999. Grammar and Conceptualization. Berlin/New York: Mouton de Gruyter.Shindo, M. 2003. “The Role of Subjectification in Semantic Extensions of Adjectives: A Chronological Perspective.” Paper presented at Eighth

International Cognitive Linguistics Conference. Logrono: University of La Rioja.Shindo, M., M. Murata, & H. Isahara 2004. “The Role of Metaphor in Semantic Extensions of Sensory Adjectives.” Paper presented at Metaphor

Workshop, 21st Century COE Program of Japanese Ministry of Education, Culture, Sports, Science & Technology. Kyoto University.The Oxford English Dictionary 1989. 2nd ed. Oxford: OUP. [OED]Sweetser, E. 1990. From Etymology to Pragmatics. Cambridge: Cambridge University Press.Traugott, E. C. 1995. “Subjectification in Grammaticalisation.” in Dieter Stein and Susan Wright (eds.) Subjectivity and Subjectivisation:

Linguistic Perspectives, 31-54, Cambridge: Cambridge University Press.Williams, J. M. 1976. “Synaesthetic Adjectives: A Possible Law of Semantic Change.” Language 52(2), 461-478.WordNet 2.0, developed by the Cognitive Science Laboratory at Princeton University (http://www.cogsci.princeton.edu/~wn/).

67


Cognition and Conceptualization of Space in Japanese:An Empirical Study of Frames of Reference

Kazuko Shinohara and Yoshihiro MatsunakaTokyo U of Agriculture & Technology and Tokyo Polytechnic University

In the study of spatial cognition and its linguistic expression, the notion of frames of reference (FR) is oneof the most important research topics. Among its different definitions and classifications, the presentstudy follows Levinson’s (1996, 2003) framework, and aims to add new findings using an empiricalmethod.

Levinson (2003: 34-61) suggests a three-way classification of FR that seems to be appropriate forlinguistic study of spatial concepts. The three FRs are the intrinsic, the absolute, and the relative FRs.Following this framework, Shinohara et al. (2003) analyze four major Japanese spatial terms, mae (front),ushiro (back), migi (right), and hidari (left). They report a phenomenon called ‘figure-aligned mapping’,which cannot be incorporated in Levinson’s framework, and suggest a revision of his definition of theintrinsic FR. The figure-aligned mapping occurs when the figure object has intrinsic directions while theground object does not: the direction (or axis) of the figure object is projected onto the ground object.However, their research is limited to when the figure object is looking aside and located on the nearer orthe farther side of the ground object from the viewer’s viewpoint. This means that they studied only aboutthe right-left axis of the figure object. Moreover, they used only topicalized sentences. Thus, the presentstudy examines (1) whether the figure-aligned mapping applies also to its front-back axis, (2) if so, whichaxis is more likely to be mapped, and (3) whether topicalization affects this phenomenon.

In our research, 81 native speakers of Japanese participated. They were shown 24 different pictures oftwo objects (figure and ground) on the computer display. The ground object is a wooden block with nointrinsic direction, shown at the center of the display; the figure is either a model car or a directionlesswooden block, located at one of the four different sides of the block. The car faces to the right or the left.Each picture is accompanied by a Japanese sentence describing the spatial relation of the objects, ‘X is in( ) of Y’ (X=figure; Y=ground), which is given in either of the two grammatical structures, with orwithout topicalization.

Topicalized sentence:X-wa Y-no ( )-ni aru.X-Top. Y-Gen. ( )-Loc. be

Non-topicalized sentence:Y-no ( )-ni X-ga aru.Y-Gen. ( )-Loc. X-Nom. be.

The participants were asked to fill in the blank in each sentence with one of the four terms, mae, ushiro,migi, hidari. The result was analyzed statistically.

Our result shows that not only the right-left axis but also the front-back axis of the figure can beprojected onto the ground, and that topicalized sentences and the side-positions of the figure object inducesignificantly more figure-aligned mappings than non-topicalized sentences or the near-far position. Thisresult suggests that the front-back axis of the figure may be easier to be projected onto the ground, andthat topicalization of sentences may facilitate this projection in Japanese.

References

Levinson, S. 1996. Frames of reference and Molyneaux's questions: crosslinguistic evidence. In P. Bloom et al. eds Languageand Space. Mass: MIT Press, pp. 11-42.

Levinson, S. 2003. Space in Language and Cognition. Cambridge: Cambridge University Press.Shinohara, K., Matsunaka, Y., & Odate, J. 2003. Cognition and expression of FRONT and BACK in Japanese. Paper presented at

the 8th International Cognitive Linguistics Conference, Spain.

68


Reference Point Constructions, the Underspecification of Meaning,and the Conceptual Structure of Palauan ´r

Michael B. SmithOakland University

Most linguists agree that meaning is determined by a combination of a language's overt lexico-grammatical resourcestogether with various kinds of background knowledge structures (domains) and contextually inferred information (i.e. from thespeech event itself). Contextual and background knowledge play an enhanced role in meaning construction when a lexical item'sconceptual content is highly schematic, as in the case of the Palauan grammatical morpheme ´r, a preposition-like word whosewide variety of uses seem unrelated to each other and whose semantic function is often obscure. Reminiscent of English of insignifying an intrinsic relation between two entities (Langacker 1992), ´r's meaning is even more schematic. I will argue that ´r'sbasic conceptual content resides in its designation of an abstract reference point construction (cf. Langacker 1993); its apparentvariety of uses reflects instantiations of this construction when construed against different backgrounds in particular contexts.

Josephs (1975:84) claims that ´r has basically two unrelated uses: as a specifying word (1) (glossed as SPEC) or arelational word (2-6) (glossed as REL) that can also signify comparison (7). A is a phrasal introducer (PI) [e.g., for NP & VP]:

(1) A n´gl´kek a m´dakt (´r) a d´rumk (5) A r´kung a tilob´d ´r A blsibsPI child PI afraid.of SPEC PI thunder PI crab PI came.out REL PI hole'My child is afraid of thunder' 'A crab came out of the hole'

(2) Ak milsuub ´r a skuul (6) Ak smech´r ´r A t´rer´rI was.studying REL PI school I sick REL PI cold'I was studying at school' 'I'm sick with a cold/I've got a cold'

(3) A John a mo ´r A Guam ´r a klukuk (7) A Droteo a m´sisiich ´r A TokiPI John PI go REL PI Guam REL PI tomorrow PI Droteo PI stronger REL PI Toki'John is going to Guam tomorrow' 'Droteo is stronger than Toki'

(4) A ngal´k a lmang´l ´r A d´malPI child PI is.crying REL PI father'The child is crying for his father'

In (1) the absence of ´r evokes a non-specific reading where the child is afraid of thunder (d´rumk) in general, whereas thepresence of ´r indicates that a specific, definite instance of thunder is meant. Palauan relational phrases with ´r (in boldface) canevoke such diverse notions as location (2), direction and time (3), goal (4), source (5), cause (6), and comparison (7), amongothers. Examples like these will be used to show explicitly how ´r's diverse uses are related in instantiating the reference pointconstruction in different domains and contexts. The reference point construction consists of a conceptualizer establishing mentalcontact with a target entity by means of another cognitively salient entity, the reference point, within a particular contextuallydetermined dominion, a "conceptual region...to which a particular reference point affords direct access" (Langacker 1993:6).Langacker claims this construction is the abstract basis for varied grammatical phenomena, including possessive constructions,topic-like constructions, and metonymy, etc. In John's car, for example, the possessor John serves as the conceptual referencepoint for locating the target, car. Reference points are generally always specific.

I will argue that, in all of its uses, ´r is a relational predication designating a reference point relation in which its landmark(LM) (or object) is construed as a particular conceptual reference point that is used to establish mental contact with its trajector(TR), the target. The target of ´r is typically construed as a process of some type. Under this analysis, the LM of ´r in (1),d´rumk 'thunder', serves as a reference point with respect to which the target, the state of the child's being afraid, is mentallyaccessed. This use of ´r induces a specific reading of its LM, because the reference point relation requires that the reference point(in this case the thunder) be one particular entity that affords mental contact with the target (here, the child's fearful reaction).When ´r is absent, the lack of a reference point relation allows for the verbal object to be interpreted as non-specific. The specificreading of the LM of ´r in (1) thus follows as a natural consequence of ´r 's reference point sense.

The different senses of ´r in (2-7) also fall out as a natural consequence of analyzing ´r as designating a reference pointconstruction against difference knowledge domains or contexts. The paper shows that, because the meaning of ´r primarilyresides in the reference point notion, it is greatly underspecified. The precise details of its sense must be supplied by backgroundand contextual knowledge.

References

Josephs, Lewis S. 1995. Palauan Reference Grammar. The University Press of Hawaii.Langacker, Ronald W. 1992. The symbolic nature of cognitive grammar: The meaning of of and of of-periphrasis. In: Martin Phtz (ed.), Thirty

Years of Linguistic Evolution: Studies in Honour of René Dirven on the Occasion of his Sixtieth Birthday, 483-502.Philadelphia/Amsterdam: John Benjamins.

Langacker, Ronald W. 1993. Reference point constructions. Cognitive Linguistics 4:1-38.

69


“Buoys” in co-speech gesture and American Sign Language

Nathaniel SmithUniversity of California, San Diego

There is a long tradition of analyzing American Sign Language (ASL) in relation to spoken language,especially to demonstrate similarity and thus show that ASL is a language (e.g., (Hoemann 1975; Klimaand Bellugi 1979)). This particular goal is now well established, but the program as a whole has a majorflaw, in that it accepts traditional generative analyses of spoken language as its base for comparison. Suchtheories are inadequate for spoken language; it is unsurprising that attempting to force signed languagesinto that same mold causes problems. We discuss two in particular.

Firstly, it assumes that as language is an abstract formal system, the modality a language useswill not effect its structure. If one instead believes that language is a system evolved under the constraintsof brain and environment to solve certain problems, then the effect of modality deserves carefulinvestigation (Liddell 2003). Secondly, it ignores entirely the gestural track, despite strong argumentsthat speech and co-speech gesture form a tightly integrated system (McNeill 1992; McNeill 2000).It is unnatural to compare ASL to the speech stream alone, ignoring that half of the speech/gesturecomplex that is most likely to shed light on the spatial modality (Taub, Piñar, and Galvan 2002; Liddell1995). If we wish to understand the relation between sign and speech, we must also consider the relationof co-speech gesture to each.

As one contribution to this endeavor, I will examine gestural analogs of "buoys" (Liddell 2002;Liddell 2003) – grammatical signs in ASL held by the non-dominant hand (while the dominanthand continues signing) to emphasize, mark themes, and structure discourse. Analogs in spoken Englishare difficult to find; indeed, it is difficult to imagine anything of this sort in any language restricted to thepurely linear speech stream.

However, it turns out that English speakers do have analogs; I will present data from a videocorpus of naturalistic conversations demonstrating that English speakers make gestures that are similar inboth form and meaning to ASL buoys. There are differences – in particular and unsurprisingly, the ASLforms are more conventionalized and complex – but the overall structure is the same.

This, I argue, supports the claim that one should not compare ASL to spoken language directly,but rather to the entire system of speech and gesture which hearing speakers are accustomed toperforming. To do otherwise in this case would have suggested that ASL had particular sorts of linguisticstructure that were unknown in English, when in fact it seems merely that ASL's focus on space hascaused it to more fully grammaticize structure they share.

Finally, it demonstrates from a new direction the claim that language and gesture are tightlylinked; not only do speakers systematically divide content between speech and gesture, presentingpropositional content in each (McNeill 1992), but in moving between languages, the "same function" maymanifest as either linguistic or gestural.

References

Hoemann, H. (1975). The transparency of meaning of sign language gestures. Sign Language Studies 7, 151-61.Klima, E. and U. Bellugi (1979). The signs of language. Cambridge, Massachusetts: Harvard University Press.Liddell, S. K. (1995). Real, surrogate, and token space: grammatical consequences in ASL. In K. Emmorey and J. S. Reilly (Eds.), Language,

gesture and space, pp. 19-41. Hillsdale, NJ, England: Lawrence Erlbaum Associates.Liddell, S. K. (2002). Buoys in American Sign Language. Paper presented at conference, Conceptual Structure, Discourse and Language 6, Rice

University, Texas, October 11-14 2002.Liddell, S. K. (2003). Grammar, Gesture, and Meaning in American Sign Language. Cambridge: Cambridge University Press.McClave, E. Z. (2001). The relationship between spontaneous gestures of the hearing and American Sign Language. Gesture 1(1), 51-72.McNeill, D. (1992). Hand and Mind: What gestures reveal about thought. Chicago: University of Chicago Press.McNeill, D. (Ed.) (2000). Language and gesture. Cambridge: Cambridge University Press.Taub, S., P. Piñar, and D. Galvan (2002, June). Comparing spatial information in speech/gesture and sign language. In Gesture: The Living

Medium Online Proceedings.http://www.utexas.edu/coc/cms/International_House_of_Gestures/Conferences/Proceedings/Contributions/Taub/.

70


Subjectivity and Causality: A Corpus Study of Spoken Language

Wilbert Spooren & Ted SandersVU Amsterdam and Utrecht University

Introduction. Connectives and the coherence relations they express are among the building blocks ofdiscourse and therefore they are a prime object of investigation in any usage-based account of language(Sanders & Spooren, 2004). In this paper we study one set of coherence relations and their linguisticmarkers: causal relations expressed by Dutch omdat (‘because’) and want (‘for’).

(1) Jan is niet thuis omdat hij weg moest.‘Jan is not at home because he had to leave’

(2) Ik weet zeker dat Jan niet thuis is, want zijn auto is weg.‘I am sure that Jan is not at home, because his car is gone’

It has recently been claimed that the notion of subjectivity accounts for the distributional patterns of suchconnectives (Pander Maat & Degand, 2001). Subjectivity can be regarded as the degree to which to thespeaker/writer is responsible for the causal relation. In these terms, (2) is more subjective than (1). Dutchwant seems to have a preference for subjective contexts, whereas omdat prefers more objective contexts.

Empirical evidence for such claims consists of corpus studies of written texts. In this paper weinvestigate a new source of empirical evidence: the Corpus of Spoken Dutch, a 10 million word corpus,which has recently come available. As spontaneous spoken language differs from planned writtenlanguage, among others, in the direct presence of speaker and addressee, we expect the degree ofsubjectivity as encoded by the connectives want and omdat to differ in the two modalities.Method. From the Corpus of Spoken Dutch we selected 150 fragments with want and 125 fragments withomdat (from the genres interviews and spontaneous conversations). The fragments were analyzed for anumber of subjectivity features, such as modality of the related segments, the nature of the primary causalparticipant, the syntactic completeness of the utterances, the type of coherence relation, etc. For the sakeof interrater agreement each fragment was analyzed independently by two coders. Comparisons weremade with two analyses of written corpora (Degand & Pander Maat, 2003; Spooren, Bekker &Noordman, 2001).Results. Fragments with omdat more often expressed non-volitional and volitional relations, wantfragments more often expressed epistemic and speech act relations. Omdat has more 3rd person primaryparticipants in the second segment than want. Want has more explicitly realized primary participants inthe second segment than omdat. The comparison with written data shows that differences in the use ofconnectives that are relatively prominent in written language are smaller in spoken language.Discussion. Our results suggest that connectives are used less specifically in spoken language. Knott &Sanders (1998) speak of speaker-economy: In low-planning situations (such as spontaneousconversations) speakers use the least specific form to encode coherence relations, whereas in high-planning situations (like well edited newspaper articles) writers tend to accommodate the readers’informational needs and use the most specific form. Implications for a cognitive theory of discoursecoherence will be discussed.

References

Degand, Liesbeth and Henk Pander Maat (2003). A contrastive study of Dutch and French causal connectives on the Speaker Involvement Scale,A. Verhagen & J. van de Weijer (eds.) Usage based approaches to Dutch (pp. 175-199). Utrecht: LOT.

Knott, Alistair and Ted Sanders (1998). The classification of coherence relations and their linguistic markers: An exploration of two languages.Journal of Pragmatics, 30, 135-175 (1998).

Pander Maat, Henk and Liesbeth Degand (2001). Scaling causal relations and connectives in terms of speaker involvement. Cognitive Linguistics,12 (3), 211-245.

Sanders, Ted and Wilbert Spooren (2004). Discourse and text structure. In D. Geeraerts & H. Cuykens (Eds.) Handbook of Cognitive Linguistics,Oxford: Oxford University Press.

Spooren, Wilbert., Birgit Bekker and Leo Noordman (2001). Reversed order & subjectivity in different text types. In L. Degand, Y. Bestgen, W.Spooren, & L.Van Waes (Eds), Multidisciplinary approaches to discourse (pp. 61-72). Amsterdam: Stichting Neerlandistiek VU.

71


NEAR-BY Regions

Soteria Svorou and Gaston R. CangianoSan Jose State University

In describing a spatial scene, prepositions name regions associated with landmarks, within which atrajector can be found. The concept of the region of a thing was first introduced by Miller and Johnson-Laird (1976), who defined it as “a rather indeterminate penumbra surrounding it (the thing)”. Since,various studies of spatial semantics (Talmy (1983), Hawkins (1984), Langacker (1993), Landau &Jackendoff (1993), Landau (1994), Svorou (1994), Logan & Sadler (1996), Garrod et al 1999, amongothers) incorporate this notion but few have attempted to provide empirical evidence for it (cf. Logan &Sadler 1996). Listeners, in interpreting “The hotel is near the airport”, associate ‘near’ with a spatialregion, the size and shape of which depends on the following variables:

1. The size and shape of the landmark, the airport;2. The size and shape of the trajector, the hotel;3. The viewing context, i.e. declarative / map-like or procedural;4. The larger area in which the hotel and the airport are located5. The distance of the hotel from the airport;6. The existence of competing landmarks in the scene.

The hypothesis tested is that regions, although fuzzy, are determinable from the interaction of the abovevariables. Focusing on the English prepositions ‘by’ and ‘near’, we predict that the region listenersassociate with’ by’ is relatively smaller than that of’ near’, as a function of the above variables.Specifically, if variables 1-4 are kept constant, the manipulation of the distance of the trajector from thelandmark, and the interference of competing landmarks in the area, will yield a measurable difference inthe search area where ‘near’ will trigger a larger area than ‘by’.

To test this hypothesis, speakers of English are asked to follow verbal directions and use acomputer 3D graphics program to navigate through a virtual room in search of objects. The program runson a web server (http://www.gastonrcangiano.net/edu/sjsu/ling/spatial/). It depicts a library room whereseveral objects (bookcases, a table, two chairs, and a lamp) are located. Below, a set of controls allows theuser to navigate through the library. Users are distributed in twelve groups. Each group operates in anenvironment, which combines different values for three variables: the kind of preposition (‘by’ or ‘near’),the distance (short, medium, large) of a trajector (a blue marble) from a landmark (the table), and thedistance of a competing landmark (a chair) from the table (close or far). Each user reads directionsinstructing her to find a blue marble either near or by the table. As the user navigates through the roomsearching for the marble, the coordinates of the path are recorded.

We are in the process of collecting data from 120 subjects. The data are in the form ofcoordinates specifying the route each subject takes from the starting point until they find the landmark.The result of the comparison of the routes will define a spatial template, which, we claim, corresponds tothe conceptual structure on which the spatial uses of ‘by’ and ‘near’ map.

References

Garrod, S., Ferrier, G., Campbell, S. 1999. In and On: Investigating the Functional Geometry of Spatial Prepositions. Cognition 72:2, pp.167-189Hawkins, B. W. 1984. The Semantics of English Spatial Prepositions. PhD dissertation, UC, San Diego.Landau, B.. 1994. Where’s what and what’s where: The language of objects in space. Lingua 92, 259-296.Landau, B. & R. Jackendoff. 1993. ‘What’ & ‘where’ in spatial language & spatial cognition. BBS 16, 217-266.Langacker, Ronald. 1993. Reference-point constructions. Cognitive Linguistics 4-1, 1-38.Logan, G. D. and D. D. Sadler. 1996. A computational analysis of the apprehension of spatial relations. In P. Bloom, M.A. Peterson, L. Nader, &

M.F. Garrett (eds.), Language and Space. Cambridge, MA: The MIT Press.Miller, G. and P. N. Johnson-Laird. 1976. Language and Perception. Cambridge: Cambridge University Press.Svorou, S. 1994. The Grammar of Space. Amsterdam: John BenjaminsTalmy, L. 1983. How Language structures space. In Pick, H. & L. Acredolo (eds.), Spatial Orientation: Theory, Research and Application. New

York: Plenum Press, 225-282.

72


One Mechanism, Two Changes in Mandarin Chinese Discourse

Liang TaoOhio University

This study discusses changes in the singular classifier NPs of Mandarin Chinese to support theview that grammar is shaped by usage in discourse. The study examines spoken discourse from bothdiachronic records of vernacular Chinese and recorded Beijing Mandarin conversations to propose twointerrelated hypotheses: (1) usage constitutes one major cause of language change; and (2) the samemechanism may lead to different grammatical changes.

Hypothesis (1) is illustrated with the development of singular numeral NPs in three patterns: (a)yi55 (one) + Classifier + Noun, (b) Classifier + Noun, and (c) yi35 (one) + Noun. Patterns (b) and (c)differ in that (b) does not contain the overt numeral yi55, yet (c) has no overt classifier but the numeralyi35 takes a fixed tone. Previous studies differ on the origins of patterns (a) and (b). Diachronically,Wang proposes that, concerning the general classifier ge51, pattern (b) was the default singular NPpattern, and pattern (a) became widely used much later, possibly influenced by the indefinite article fromwestern languages. Lü argues that pattern (b) is a phonological reduction of pattern (a) through sounderosion in discourse. Cheng and Sybesma claim that, synchronically, patterns (a) and (b) are unrelatedbecause they have different underlying representations. Based on analysis of discourse data, this presentstudy supports Lü’s proposal that pattern (b) developed from (a) in spoken Chinese diachronically. Thedifferences of patterns (a) and (b) that Cheng and Sybesma point out are results of language development.Further, current spoken discourse data indicate that there is a change of preference in current languageusage because pattern (a) is more preferred than (b) now, contrary to data of spoken Mandarin invernacular Chinese. The frequent usage of yi35ge51 + N has lead to phonological reduction and newpattern (c) in Beijing Mandarin. The development started with a change of stress pattern in spokenBeijing Mandarin, followed by sound erosion and phonological reduction of the classifier in (c).Therefore, there are now two changes in Chinese out of the same mechanism of sound erosion indiscourse, as proposed in hypothesis (2).

In summary, discourse usage is the cause of changes in singular Mandarin numeral NPs. Thesame production mechanism in discourse has lead to two different grammatical patterns, but the differentpatterns function for the same purpose as the linguistic coding of human observations of the physicalworld. The study proposes to take discourse analysis as a fundamental means in the study of language,grammar and human cognition.

Three NP patterns:

(a) Nei51huir35 you21 yi35 ge51: (0.5) you21 yi35ge rer35 athat-time there-be one-classifier there-be a-Classifier person Int‘At that time there was a, there was this guy/person’

(b) Ta55 ba214 ge dao55 xi214 le3sg BA Classifier-knife wash Asp‘She got the cleaver washed (It was my job!)’

(c) Ba51ba, wo21 dai51 ni21 kan51 yi35 dong55::xi::Dad 1sg take 2sg look-at one thing‘Dad, I’m taking you to look at something (a thing).’

73


A Corpus-Based Semantic Analysis of Norwegian Se

Martin ThieringUniversity of Alberta

This paper is a corpus-based account of the different usages of the visual perception verb, se, incontemporary Norwegian and the construction types that it enters into. My aim is to (a) delimit the mostfrequent construction types of se, (b) explore the semantic usage potential of these types in order to (c)give evidence for ongoing grammaticalization—or at least relexicalization––processes in the language.Finally, this paper supports the idea that language structure is not independent of language use, but ratheris emergent from it.

The usage-based approach of corpus linguistics enables us to identify the various extension patternsof se and the various construction types each participates in (Bybee & Hopper 2001). The Oslo Corpusused in this survey provides both the database as well as the tool for the analysis. It is an online corpuscompiled of written texts, consisting of about 18.5 million words collected from a variety of genres, e.g.,fiction (1.7 million), factual prose (7.1 million), and newspapers/magazines (9.6 million; 1994-1999). Bysearching the corpus for instances of se followed by specific parts of speech, the most frequentconstruction types of se were determined. A close examination of a random sampling of the constructiontypes revealed the semantic range associated with each construction type. It is clear from these data that,overall, there has been a semantic shift from an initial ‘physiological’ sense of seeing to more ‘cognitive’,abstract senses. Furthermore, it is evident from this study that the more frequent the construction type, themore likely it is to have a figurative meaning.

The overall heuristics of this survey involves a focus on the role and frequency of differentconstruction types to delimit the actual usage potential of this basic verb. The extensions to figurativeusages indicate that se in contemporary Norwegian has radically departed semantically from its originalliteral meaning of physical sight of concrete, imageable entities (Sweetser 1990). Sequential semanticshifts have been observed supporting mapping processes from one cognitive domain to another, e.g., fromthe literal usage based on physical sight (spatial domain) to figurative usages whereby propositionalentities are cognitively accessed or assessed (conceptual domain). This shift is observable in the ongoingprocess of grammaticalization or relexicalization where the literal lemma se turns into figurative usages ofcomparison or experience marking and has become more evidential than perceptual.

References

Bybee, J.L & Hopper, P. (Eds.) 2001. Frequency and the Emergence of Linguistic Structure. Amsterdam/Philadelphia: JohnBenjamins.

Oslo-Corpus. www.tekstlab.uio.no/norsk/bookmaal.Sweetser, E. 1990. From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge

Studies in Linguistics, 54. Cambridge: CUP.

74


Puzzles of Joint Attention in Classic Detective Fiction and Beyond

Vera TobinUniversity of Maryland

As linguists, we tend to avoid taking complex, fictional narratives as our primary subjects ofinvestigation, in large part because these texts are so notoriously inventive and open to interpretation. Butthis productivity is not without constraints, and is in itself an important datum for our understanding ofthe conceptual structures underlying language use. In this paper, I examine this productivity in light of atype of coordination problem (Lewis 1969, Clark 1996) that is typical of, though not exclusive to, writtenfictional narratives.

Research by Michael Tomasello and others suggests that the ability to understand oneself as participatingin a joint attentional scene (Tomasello 1999, 2003) is a fundamental underpinning of the conceptual workrequired to use and acquire language. Basic language use consists of making reference to concrete orabstract objects of joint attention in an attempt to manipulate the mental states of an interlocutor. Infiction, putative joint attentional scenes can be understood to hold between the author and the reader, acharacter and the reader, the author and a character, or multiple characters. Any of these configurationsmay be exploited to encourage modes of reading that engage with absences in a joint attentional scene,creating joint attentional puzzles.

This analysis focuses on detective fiction, a genre in which the act of puzzling is particularly prominent.Being alert to clues in detective stories involves special alertness to the reason why the implied author hasdirected the (implied) reader's attention to a certain item, event, or interaction. Quite separately fromevidence within the world of the text, tracking the ways in which the implied author seems to directreaders' attention within these sequences provides the major impetus for the practiced reader of detectivestories to recognize an element as either a clue or a red herring, rather than “mere” scenery. I will providea taxonomy of these joint attentional cues and trace the history of their entrenchment as conventions ofthe genre.

Having established the importance of the joint attentional scene for characterizing this kind ofprototypical reading puzzle, I will discuss ways in which other genres – including the classically“difficult” texts of high modernism – present variations and hypertrophies of this joint attentional puzzleand present a theory for why some of these puzzles are traditionally considered so much more accessiblethan others.

Analysis of these examples will reveal that it is useful and important to consider the entrenchment ofmodels of attentional scenes in conventional frames of discourse types themselves—that is, folk modelsof narrative, conversation, instruction, and so on—in the communal common ground (Clark 1996)supporting discourse of all kinds. A classic problem for narrative theory is the difficulty of providing anaccount of narrative with any predictive power, able to exclude forms that do not occur and allow formsthat do; this study provides a model for a cognitively realistic, discourse-based account of narrativeeffects that should be of use to both linguists and scholars of literature.

References

Carroll, N. (1996). The Paradox of Suspense. In P. Vorderer, H. J. Wulff, and M. Friedrichsen (Eds.) Suspense: Conceptualization, TheoreticalAnalysies, and Empirical Explorations. Lawrence Erlbaum Associates.

Clark, H. H. (1996) Using language. Cambridge University Press.Goffman, E. (1974) Frame analysis: an essay on the organization of experience. Harper Colophon.Lewis, D. (1969) Convention: a philosophical study. Harvard University Press.Tomasello, M. (1999).The Cultural Origins of Human Cognition. Harvard University Press.Tomasello, M. (2003). Constructing a Language: A Usage-Based Theory of Language Acquisition. Harvard University Press.

75


Testing the Effects of Sign Language Phonology on Language and Motor Cognition

Meylysa TsengUniversity of Hawaii, Manoa

In this experiment I test whether the phonology of sign language affects comprehension. I base iton the study by Glenberg and Kaschak (2002) in which participants judge the acceptability of sentencesinvolving movement toward and away from the body. To respond in the affirmative or the negative, theywere asked to press a button that was either further away, or closer to their body. It was found thatparticipants were slower to respond when the direction of the action in the sentence was opposite to thedirection their hand had to move in order to press the correct button.

In investigating sign language, there is an additional variable which is not present in spokenlanguage experiments. Even though spoken language does involve movement of the tongue, mouth andother parts of the vocal tract, this movement is not perceived in the same way as that of movement ofarticulators in sign language. Sign language involves hands moving in directions and forming shapeswhich are often times iconically related to what they are depicting. Thus, in investigating the effects ofmotion events in sign language, there are two considerations: 1) the direction the sign entails semanticallyand 2) the direction the hands move in the sign.

I looked at two directions, 1) toward the body and 2) forward away from the body. Subjects wereasked to press a “ready” button to initiate a sequence of two signs, separated by a distractor. If the twosigns were the same, they lifted their finger off of the “ready” button and hit either the “same” button orthe “different” button. Depending on the condition, subjects had to move forward or backward in order tohit the button (Rotating a detachable keyboard, I used the “g” key for “ready”, the “a” key for the awaycondition and the “l” key for the toward the body condition.)

I presented native signers of ASL with one of four trials consisting of 160 sign pairs (80 signs perdirection of the “same” button, 40 with direction encoded phonologically but not semantically (p-typesigns) and 40 with direction encoded phonologically and semantically (s-type signs)). Only pairs ofidentical signs were analyzed. Reaction times were recorded at “ready” button release and when they hitthe “same” answer key correctly. To create the non-matching sign pairs, a p-sign was randomly pairedwith an s-sign of the same direction. The position of the “same” and “different” buttons was switched halfway through each trial.

I’m currently running the experiment, but I predict that the control group, signs with directionencoded semantically, will interfere with motor coordination. In addition, I hypothesize that phonologicaldirection will also have an effect on motor coordination. This is due to the fact that our cognition of motorevents is believed to be stored in the motor cortex (Buccino et. A. 2001, Gallese et al. 1996, Rizzolatti et.al. 1996), where comprehension and perception both occur for semantically meaningful and unmeaningfulactions.

References

Buccino, G, Binkofski, F, Fink, G.R., Fadiga, L, Fogassi, L, Gallese, V, Seitz, R.J., Zilles, K, Rizzolatti, G, & Freund, H.J. 2001.Action observation activates premotor and parietal areas in a somatotopic manner: an fMRI study. European Journal ofNeuroscience 13(2): 400-404. http://www.unipr.it/arpa/mirror/pubs/pdffiles/Buccino-Binkofski%202001.pdf

Gallese , V., Fadiga, L., Fogassi, L., & Rizzolatti, G.. 1996. Action recognition in the premotor cortex. Brain 119: 593-609.http://web.unife.it/progetti/neurolab/lavorinostripdf/brain1996.pdf

Glenberg, Arthur M. and Michael P. Kaschak. 2002. Grounding Language in Action. Psychonomic Bulletin & Review.Rizzolatti, G., Fadiga, L., Gallese, V., & Fogassi, L.. 1996. Premotor cortex and the recognition of motor actions. Cognitive

Brain Research, 3:131--141. http://web.unife.it/progetti/neurolab/lavorinostripdf/CognBrainRes1996.pdf

76


Eclectic and Interdisciplinary Approaches to the Empirical Study of Discourse

Linda WaughUniversity of Arizona

This study argues for an eclectic and interdisciplinary approach to the empirical study of discourse thatcombines both quantitative and qualitative methods and uses the best of various approaches to discourseanalysis. The approaches used are corpus linguistics, functional linguistics, discourse and grammar,cognitive linguistics, ethnography of communication, interactional sociolinguistics, sociohistoricallinguistics, variational sociolinguistics, grammaticalization, conversational analysis, and critical discourseanalysis.

This paper is based on a corpus of 194,000 words of Everyday Conversational European French, recordedin France and Switzerland, which will be compared with standard written French (ARTFL corpus). Usingboth quantitative results as well as qualitative analysis of selected conversations, the following issues willbe addressed empirically, using grammar (pronominal usage and negation) as the focus of analysis:(almost) completed grammatical change, change in progress, the role of vague language in discourse andlanguage change, grammatical metaphor, micro-shifts in register according to participants and topics,solidarity and distance between interlocutors, linguistic conservatism, national, ethnic, and linguisticidentity and shift of identity across topics and participants.

Starting with the change of on from its indefinite meaning to meaning ‘we’ (and thus ousting nous), itwill be shown that this change is almost completed in France (only 1% use of nous), where it ismaintained by speakers who are educators and evidence linguistic conservatism. In addition, the writtennegative element ne is gone for much of working class French (0%), but in Switzerland it still survives(2.6%) in two uses: stylistic ne, for more formal usage, imitation of written language (written legalese),and emphatic ne, to reinforce a strong negation.

Indefinite meaning is shared by on, tu ‘you’, and ils ‘they’. Tu typically includes speaker and addresseeas potential/virtual participants in the verbal process, and often connotes solidarity between andinvolvement of speaker and addressee. Ils is also used for indefinite meaning (like ‘they’ in English) andhas an exclusive meaning, excluding the speaker and addressee as potential/virtual participants in theverbal process; it thus giving distance from the perpetrators of the verbal process; and often a negativeevaluation of the topic. By comparison, on in many uses is very general and encompasses a variety ofmeanings – it is thus the unmarked indefinite pronoun.Our corpus shows that the three indefinites can be exploited for many purposes, including construction ofand shifting of identity across contexts. The use of the indefinites by a speaker with three differentidentities (French, American, Tunisian) will be analyzed to show how he shifts from one identity toanother as shown by pronoun usage as he goes from one topic to another.

The findings suggest that the various approaches used can be combined, as long as any claims are shownto be valid from a close analysis of the data. It will also be argued that the integration of these variousapproaches leads to an understanding of the subtleties of conversation that would not be possible if onlyone perspective were used. Conversation is complex and our tools for analyzing it should be equallycomplex.

77


Action Verbs and Hand Motion: An ACE Experiment

Kathryn WheelerUniversity of Hawaii, Manoa

Recent language and cognition research has explored the idea that in order to understand language aboutactions, a hearer must first simulate the action. This theory has found support from research into so-called‘mirror neurons’ that have been shown to be active both when performing and perceiving an action(Gallese et al. 1996, Rizzolatti et al. 1996). Previous studies have shown that motion and effector (bodypart) are involved in language comprehension (Glenberg & Kaschak 2002, Bergen et al. 2003) and thisstudy seeks to expand these findings by exploring the effects of specific hand-shape on understandingdifferent hand-motion verbs.

Glenberg & Kaschak (2002) proposed the Action-sentence Compatibility Effect (ACE) in which the‘implied sentence direction’ of a sentence can have a facilitory or inhibitory effect on the ‘actual responsedirection’ of an executed action. They found that a sentence such as “Open the drawer,” encoding directionaway from the body, interferes with action performance toward the body. This implies simulation effects.

Bergen et al. (2003) demonstrated effector-specific effects by showing interference arises whensubjects process a picture and verb depicting different actions that use the same bodily effector (hand,mouth, foot). This indicates comprehension of action language activates action-specific circuitry.The ACE Effect predicts that language listeners must activate not only the general effector, but also specificdetails encoded in action performance (i.e. direction, speed, and position of effector). This research seeks totest whether, in order to process an action verb (scratch, push), listeners must activate fine-grained motorfunctions (i.e. open and closed hand-shape).

By comparing response times (RTs) from sentence-response pairs where the hand shape responsesand sentential semantics may be matching or non-matching, I hypothesize that if a subject initiallysimulates the sentence, the RT will be faster when the response requires the same hand shape. In the non-matching condition, the RT should be slower than baseline due to mutual inhibition of neural machinery forthe incompatible motor actions.

Subjects make sentence sensibility judgments, responding either yes or no, by making the handmotion assigned to each response. These hand motions are either an open palm or a clenched fist. Ifunderstanding (simulating) a sentence involving a motor action uses the same neural machinery asperforming that action, we would predict that subjects should respond faster when the motor action in thesentence is compatible with the action they are asked to perform. A set of bivalent verbs (lift, carry) wasalso included such that if this interference is due to simulation, and not simply lexical priming, it shouldalso occur with these verbs. If the subject is not simulating the context of the sentence, we would predict nodifference in response time between matching and non-matching sentence-response hand shapes.

This study has the potential to support previous findings that language comprehension activates thesame neural circuitry involved in language performance. This further goes to reinforce a theory ofembodied language in which language must be facilitated by our interactions with the physical worldaround us.

References

Bergen, B., S. Narayan, and J. Feldman. 2003. Embodied verbal semantics: evidence from an image-verb matching task. In Proceedings of the 25th

Annual Conference of the Cognitive Science Society.Gallese, V., Fadiga, L., Fogassi, L., Rizzolatti, G. 1996. Action recognition in the premotor cortex. Brain 119: 593-609.Glenberg, Arthur M., and Kaschak, Michael P. 2002. Grounding Language in Action. Psychonomic Bulletin & Review, 9 (3), 558-565.Rizzolatti, G., Luciano, F., Vittorio, G., Fogassi, L. 1996. Premotor cortex and the recognition of motor actions. Cognitive Brain Research, 3:131-

141.

78


Building Anchored Blends: Gesture and Co-Gesture Speech in Instructional Discourse

Robert F. WilliamsLawrence University

In recent work, Hutchins (in press) describes how the material world is used to anchor conceptual blendsfor reasoning and problem-solving. Some artifacts are specially constructed to anchor conceptual relationsin recurring cognitive activities. The clock face, for example, supports the conceptual operations of time-telling. The material world is also exploited opportunistically to stabilize conceptual relations relevant tothe current activity. Both cases depend upon mappings that link conceptual elements to the materialstructures that anchor them. How do such mappings get set up in actual discourse? To find out, I usedcognitive ethnography to study the construction of meaning in time-telling lessons. The study includedobserving and recording time-telling instruction in elementary school classes; transcribing episodes ofinteraction for speech, gestures, and manipulations of artifacts; and analyzing step-by-step how blendedspaces are constructed and elaborated in the discourse.

As anticipated from previous studies of discourse from a mental-space perspective (e.g. Fauconnier1997), speech appears to activate conceptual content, build mental spaces, and set up relations betweenspaces. What I discovered is that gesture plays a crucial role in the construction and elaboration of theseconceptual integration networks. In episodes of instruction in clock-reading, teachers manipulate theteaching clock and other objects to align them with conceptual models. They use gestures to add image-schematic structure such as paths of motion. And they use pointing and tracing gestures (cf. Goodwin2003a, 2003b) to highlight structures on the clock face while superimposing conceptual entities (circles,dividing lines) over these structures. The coupling of these gestures with the environment and with co-gesture speech sets up the material anchors for conceptual elements. In the lessons, the teacher sets upmappings one at a time until all are in place; then she demonstrates how to use the anchored blend toconstruct a portion of a time reading. These episodes of instruction illustrate the uniquely human process Icall ‘guided conceptualization’: the teacher controls the sequence of actions while using gestures and co-gesture speech to shape the learner’s construction of meaning.

References

Fauconnier, Gilles (1997). Mappings in Thought and Language. Cambridge UP.Fauconnier, Gilles & Turner, Mark (2002). The Way We Think: Conceptual Blending and the Mind’s Hidden Complexities.

Basic Books.Goodwin, Charles (2003a). Pointing as situated practice. In Kita, Sotaro (Ed.), Pointing: Where Language, Culture, and

Cognition Meet. Lawrence Erlbaum Associates.Goodwin, Charles (2003b). Environmentally coupled gestures and the social constitution of vision. Invited lecture, University of

California, San Diego, Spring 2003.Hutchins, Edwin (in press). Material anchors for conceptual blends. Journal of Pragmatics.

Situated Simulation in the Human Conceptual System

Documents