For Fall 2012, the lab met Fridays 10:30-Noon, with the following topics.
MesoSpace planning: Incorporating topography data (various MesoSpace team members)
Veterans of the lab will recall how MesoSpace has been trying to quantitatively determine the influence of language and language external factors (education, literacy, local topography, etc) on spatial frame of reference preferences in discourse and recall memory. (See Bohnemeyer et al 2012, a first pass at what we are trying to do for the Cognitive Science article (sorry that it’s in Spanish, but the figures should give you a good idea of what we did and what we found)).
For this meeting, Dr. David Mark from Geography will be joining us, as will MesoSpace project member and former lab leader Dr. Carolyn O’Meara, via Skype. We will discuss the various issues with the topography data with Jesse so that we can best decide on how to incorporate the data into the model. To give you an idea of how these things go behind-the-scenes, here is what Jesse said he would need to see before this meeting:
- Actual copies of all of the survey data that is presently available.
- A timeline for when the missing data is expected to be available.
- Advisories about inconsistencies in format between data sets or incompleteness in data sets.
- A summary of the theoretical proposals concerning the direction of correlation for each variable under consideration.
Since the topography data simply consists of our researchers’ responses to a questionnaire, Carolyn provided Jesse with those responses in a single spreadsheet. Data for Chol, Huave, Mixe, and Mexican Spanish are still missing, but we expect to receive them at a later date. Juergen will speak to #4 at the meeting – what theoretical proposals underlie our expectations for correlation. This will be crucial in determining how to fit this data into the model.
Field Report: MesoSpace in La Ventosa, Oaxaca (Randi Tucker)
Dry-run presentation: Modeling Folk Physics for Yucatec Maya Dispositionals (Tim Tilbe)
Abstract: The lexical semantics of Yucatec Maya dispositionals has embedded in it a particular way of classifying physical phenomena. This folk physics provides an internally coherent, partial account of human experience of the material world. Whether or not it corresponds in any particular to the psychologically real folk physics that speakers use for conceptualizing and reasoning about the material world, an explicit account of the folk physics of Yucatec dispositionals provides a basis for comparison to the semantics of dispositionals in other languages. Ultimately, such comparisons could lead to a typology of the concepts lexicalized by dispositionals. To that end, it is desirable to create a formalization of folk physics concepts. This presentation lays out a project of formalizing the semantics of Yucatec Maya dispositionals using predicate calculus.
The componential semantic analysis that is the basis for the folk physics model advanced here was developed from the results of field work. The nature of the research task and its results will be briefly reviewed.
In formal terms, each dispositional is defined as the conjunction of its semantic components. Variables represent the figure (the entity of which the disposition is predicated) and the ground (an entity that bears some relation to the figure). Each of the semantic components is in turn defined by one- and two-place predicates that represent the necessary properties and relations from the domains of topology, mereology, metric distance, force, time, order/disorder, and a folk taxonomy of material entities.
Dry-run presentation: Plenary talk for the 1st International Conference on Yucatec Linguistics: Reference frames in Yucatec and Mesoamerica: linguistic, cultural, and environmental factors (Juergen)
Abstract: This lecture presents preliminary findings of the project Spatial language and cognition in Mesoamerica (MesoSpace), a comparative study of the representation of space in 15 indigenous languages of Mexico and Central America. The overarching question MesoSpace addresses is that of linguistic and nonlinguistic determinants of frame of reference (FoR) use in language and internal cognition. FoRs are cognitive coordinate systems used to identify places (in the sense of regions) and directions, often with respect to some reference entity or ground. A growing controversy has arisen around the demonstration of a robust crosslinguistic alignment of FoR use in language, recall memory, and spatial inferences. Levinson (1996, 2003), Pederson et al (1998), and Majid et al (2004) argue that FoR use in (L1 and L2) discourse is governed by conventional practices of the speech community, which in turn influence preferences in cognitive tasks. In contrast, Li & Gleitman (2002) propose that population-specific biases in discourse and cognition are the direct result of individual-level adaptations to factors of topography, population geography, literacy, and education. This position entails that linguistic patterns of FoR use can themselves be entirely attributed to the proposed non-linguistic factors.
To test this prediction, the MesoSpace researchers collected data on FoR use in discourse from 13 indigenous Mesoamerican languages, two indigenous languages spoken north and south of the Mesoamerican area – Seri and Sumu-Mayangna – and three varieties of Spanish (Mexican, Nicaraguan and Peninsular Spanish), using five dyads of adult native speakers per variety. All of these varieties have the lexical resources for using all major FoR categories. Participants matched four sets of twelve pictures featuring a ball and a chair in varying spatial configurations. We also estimated, on a three-point scale, each participant’s education level, level of literacy, and level of Spanish usage.
A major source of crosstalk between the two camps in this debate has been the reliance on incommensurable classifications of reference frames. To overcome this problem, the MesoSpace team developed a fine-grained classification that distinguishes among eight types of FoRs which uniquely map into the different course-grained classifications assumed by Li and colleagues and Levinson and colleagues. In any given description of one of the Ball & Chair pictures, a participant could use any or all of the eight strategies. The presentation introduces the MesoSpace classification with examples from Yucatec Maya. Yucatec emerges as a referentially “promiscuous” language: all types of frames of reference occur regularly in Yucatec discourse, including in the domain of manipulable space. The use of cardinal direction terms is, however, most common among adult males. FoR selection in Yucatec is highly variable, both across and within speakers. All speakers frequently combine multiple FoRs in a single spatial description. Even terms for relations in the vertical are regularly used intrinsically, suggesting that the Principle of Canonical Orientation (Levelt 1984, 1996) is at most a tendency in Yucatec. The relatively minor role of relative FoRs supports the hypothesis that languages in which geometric (shape-based) ‘meronyms’ (object part descriptors) are a pervasive resource for the expression of projective spatial relations favor intrinsic and/or absolute FoRs over relative ones.
The lecture then turns to two multivariate statistical analyses of the Ball & Chair data collected from speakers of six of the Mesoamerican languages (Isthmus Zapotec, Mixe, Otomi, Tarascan, Tseltal and Yucatec Maya), the two non-Mesoamerican indigenous languages, and the three varieties of Spanish. These analyses depart from previous applications of multivariate statistics in semantic typology (e.g., Levinson & Meira 2003; Majid et al. 2008) in that they treat the behavior of individual participants as the principal response variable and language as one of the predictor variables, along with the regularity of use of Spanish as a second language and Li & Gleitman’s proposed explanatory variables of literacy and education level. Bohnemeyer et al. (2012) pilot a phylogenetic analysis using the Neighbor-net algorithm (Huson & Bryant, 2006). Currently the team is working on a mixed-effects linear regression analysis of the same data matrix. Preliminary results suggest a significant effect of language that cannot be reduced to a combination of the other variables, contrary to Li & Gleitman’s predictions.
Reading discussion: Kemp & Regier (2012), and Levinson’s response
Abstract: Languages vary in their systems of kinship categories, but the scope of possible variation appears to be constrained. Previous accounts of kin classification have often emphasized constraints that are specific to the domain of kinship and are not derived from general principles. Here, we propose an account that is founded on two domain-general principles: Good systems of categories are simple, and they enable informative communication. We show computationally that kin classification systems in the world’s languages achieve a near-optimal trade-off between these two competing principles. We also show that our account explains several specific constraints on kin classification proposed previously. Because the principles of simplicity and informativeness are also relevant to other semantic domains, the trade-off between them may provide a domain-general foundation for variation in category systems across languages.
Reading discussion: “The Macro-Event Property and the Layered Structure of the Clause” (draft by Juergen and Robert Van Valin)
In this paper, we present crosslinguistic evidence supporting the hypothesis of a universal isomorphism between a level of syntax uniquely identified in Role and Reference Grammar (RRG), the Core layer within the Layered Structure of the Clause (LSC) theory, and the so-called Macro-Event Property (MEP) proposed in Bohnemeyer et al (2007). The MEP is a form-to-meaning mapping property that constrains the compatibility of event descriptions with time-positional modifiers. Bohnemeyer et al (2007, 2010) instrumentalize the MEP as a heuristic in typological studies of the properties of event descriptions across languages. The MEP can be thought of as an operationalization of the more intuitive – and therefore vague – distinction be-tween descriptions of single and multiple events. But Bohnemeyer (2003) and Bohnemeyer et al (2007) also present preliminary evidence suggesting that certain linking constraints at the syntax-semantics interface – such as the well-known biuniqueness constraint on the mapping between arguments and semantic roles (Fillmore 1968; Bresnan 1980; Chomsky 1981) – are directly sen-sitive to the MEP, in the sense that all and only those constructions that have the MEP are the domains of these constraints. In this paper, we argue that all and only simple (verbal or nominal) Cores, but not clauses or verb phrases, have the MEP. Complex Cores have the MEP if and only if their constituent Cores are in a cosubordinate linkage. We support this hypothesis with data from English event nominalizations, Ewe serial verb constructions, and Japanese converb con-structions. If correct, this hypothesis suggests that LSC outcompetes phrase structure grammars and X-bar Theory in its ability to predict the behavior of constructions at the syntax-semantics interface. This is in our mind a non-trivial empirical finding, since RRG was not originally de-signed for the specific goal of optimally modeling the syntax-semantics interface. The original motivation behind RRG was to provide a model of syntax equally applicable to languages that vary on a number of fundamental typological parameters: languages that do vs. do not have verb phrases; languages that express predicate-argument structures syntactically vs. morphologically; languages that have different organizations of grammatical relations or no grammatical relations; and languages that do vs. do not have serial or multi-verb constructions or chaining construc-tions. If this model now turns out to offer more insightful representations of form-to-meaning mapping – as we believe is the case – this would strike us powerful testimony to the potential of typological research for improving linguistic theory.
Reading discussion: Iconicity and Chafe’s legacy
- A chapter from his 1980 book “The Pear Stories” entitled “The deployment of consciousness in the production of narrative”.
- His 1975 article, featured in Cole’s “Current Issues in Linguistic Theory” entitled “The recall and verbalization of past experience”.
Reading discussion: Accessibility affects on word order
Informing the article being written by Juergen, Florian Jaeger, and Lindsay Butler, we will discuss relevant typological constraints on functional processing and the “alignment” between grammatical functions and contextual/discourse-pragmatic properties such as topicality. For example, one obvious question is whether all languages have grammatical functions such as subject and object (nope!) and what functional processing actually means in languages that lack them. Other, but presumably closely related, questions concern the relevance of topic prominence in the sense of Li & Thompson’s (1976) typology of subject- vs. topic-prominent languages and that of discourse configurationality in the sense of Kiss (1995).
The article in preparation addresses a long-standing question of sentence processing research: whether conceptual properties of discourse referents such as animacy have a direct impact on where in the clause/sentence the NP referring to the referent appears. They ran some experiments with speakers of Yucatec and Spanish and found that animacy seems to play only an indirect role in these languages, a role that’s mediated by topicality. That is, the higher the referent on the animacy hierarchy, the more likely it is to be treated as the topic of an event description. And topicality in turn interacts strongly with position in the sentence:
- Spanish has SVO order. Subjects are actors in active transitive sentences and undergoers in passive sentences. If the undergoer or a dative participant is more topical than the actor, passivization or left-dislocation are used. Thus, the referent mentioned first tends to be the most topical one.
- Yucatec has VOS order. But in transitive active sentences in connected speech, the more topical of the two arguments is always left-dislocated. (If that happens to be the undergoer, the verb is always passivized.)
So both languages can be described as “topic first”. And animacy makes a referent more likely to be treated as the topic. This seems to support the view that referential properties such as animacy affect processing only indirectly, via their interaction with contextual properties such as topicality. Contextual properties in turn are assumed to influence primarily ‘functional processing’ rather than ‘positional processing’. That is to say, such properties are presumed to affect primarily the assignment of grammatical relations rather than linear order directly. Here is a paragraph from the draft of our paper concerning the distinction between the two processing modes:
“One question that has received a substantial amount of attention is how or when conceptual accessibility comes to affect sentence production. Sentence production is widely assumed to involve two distinct levels of processing: functional processing and positional processing (Bock, 1982; Bock and Levelt, 1994; Garrett, 1976, 1980; Levelt 1989; but see Kempen XXXX). Functional processing is assumed to precede the latter and involves the mapping of lexical items to grammatical functions such as subject and object. Positional processing is assumed to involve inflectional morphology and the sequentialization of words and constituent. This has caused researchers to ask whether accessibility affects word order choices directly during positional processing (availability accounts, Ferreira, 1996; Ferreira and Dell, 2000; Prat-Sala and Branigan, 2000) or indirectly through effects on grammatical function assignment (alignment, Bock and Warren, 1985). For example, the effect described above for English passives could be attributed to a preference to align animate referents with the subject function, which would be an indirect effect. But the effects could also be due to a preference to mention easy to produce material as early as possible in the sentence.”
- Bock, J. K. 1982. Toward a Cognitive Psychology of Syntax: Information processing contributions to sentence formation. Psychological review. 89(1): 1-47.
- Bock, J. K. & R. K. Warren. 1985. Conceptual accessibility and syntactic structure in sentence formation. Cognition 21: 47-67.
- Branigan, H. P., M. J. Pickering, & M. Tanaka. 2008. Contributions of animacy to grammatical function assignment and word order during production. Lingua 118: 172-189.
- Kiss, K. E. 1995. Discourse Configurational Languages. Oxford: OUP.
- Li, C. N. & S. A. Thompson. 1976. Subject and Topic: A New Typology of Language. In C. N. Li (ed.), Subject and Topic. New York: Academic Press. p. 475.
Guest speaker: Rebecca Cover
Abstract: I present recommendations for conducting fieldwork on the semantics of tense, aspect, and modality. I argue that a thorough understanding of the semantics of a given TAM category is best obtained by applying three research methods, each of which builds upon the others: semantic elicitation, text collection and analysis, and participant observation. I illustrate these methods with case studies from my own fieldwork on Badiaranke, an underdescribed Atlantic (Niger-Congo) language. Specifically, I focus on two TAM categories in Badiaranke: the modal/aspectual imperfective, and the “discontinuous past” (Plungian & van der Auwera 2006).
Reading discussion: Fortescue (2011) “Where ‘Out to Sea’ Equals ‘Towards the Fire’: The Macrocosm-Microcosm Relationship in Languages of the North Pacific Rim” (Randi)
Abstract: An unusual conflation in the meaning of directional terms whereby ‘out to sea (or the open water)’ is equated with ‘towards the fireplace’ (and the converse, ‘to shore’ with ‘away from the fireplace’) has been observed in a number of languages of the Northwest Coast of America and of the Russian Far East. The origin and motivation for this correlation has remained a mystery, being much more specific than the overlap between terms referring to the macrocosm (the geographical surroundings) and to the microcosm (the house) that is typical of Arctic and Subarctic languages. This article presents an explanation for the phenomenon, which appears–surprisingly–to have its roots deep within Siberia.
Dry-run presentation: In search of areal effects in semantic typology: Reference frames in Mesoamerica (Juergen & Kate Donelson)
We examine whether practices of using language may be (i) contact-diffused and (ii) areal features. The domain of our study is that of spatial frames of reference (FoRs), conceptual coordinate systems used to locate and orient entities and motion paths. The reference frame types available to a speech community are not generally constrained by the grammar or lexicon of the language, but are a part of the community’s practices of language use. Since these practices vary with language (Majid et al 2004; Levinson & Wilkins (eds.) 2006; O’Meara & Pérez Báez (eds.) 2011), it stands to reason that they may be contact-diffused, although it has to our knowledge never been directly demonstrated that this is the case. Our test case for areality is the Mesoamerican (MA) sprachbund (Kaufman 1973; Campbell 1979; Campbell, Kaufman, & Smith-Stark 1986; Smith-Stark 1994). We collected data on FoR use in discourse from six indigenous MA languages (Isthmus Zapotec, Mixe, Otomí, Tarascan, Tseltal and Yucatec Maya), two indigenous languages spoken north and south of the MA area – Seri and Sumu-Mayangna – and three varieties of Spanish (Mexican, Nicaraguan and Peninsular Spanish), using five dyads of adult native speakers per variety. Participants matched four sets of twelve pictures featuring a ball and a chair in varying spatial configurations. We coded descriptions of the location of the ball for eight strategies (topological, object-centered intrinsic, egocentric intrinsic, relative, intrinsic-relative ambiguity, geocentric, geocentric-vertical, ambiguous-vertical). In a given description, a participant could use any or all of the eight strategies. We also estimated, on a three-point scale, each participant’s education level, level of literacy, and level of Spanish usage. We computed for each dyad an eight-dimensional vector assigning to each strategy the frequency with which the dyad used it. Interpreting these vectors as points in an eight-dimensional space, we calculated their Manhattan distances as a measure of the similarity between them. Figure 1 shows an MDS plot of the left-triangular distance matrix. The first dimension of this plot correlates strongly with the geocentric scores (Spearman’s Rho[df=53] 0.95; p<.001; cf. Figure 2) and weakly negatively with the relative scores (Spearman’s Rho[df=53] -0.80; p<.001), while the second dimension correlates weakly with the topological scores (Spearman’s Rho[df=53] 0.79; p<.001). We then modeled the geocentric and relative FoR scores of just the speakers of the indigenous languages as a function of education level, literacy level, L2-Spanish usage level, and areal-linguistic affiliation (MA vs. non-MA). The generalized linear mixed-effects model included nested random intercepts for language and participant ID. The fitted geocentric model revealed L2-Spanish use and literacy as significant factors (Wald-p .01 each) and the relative model only L2-Spanish use (Wald-p .02). We take this as evidence that the use of relative FoRs does indeed diffuse through contact with Spanish. At the same time, our GLMMs failed to find evidence of a sprachbund effect (Wald-p .79 for the geocentric and .16 for the relative model). A hierarchical cluster analysis of the distance matrix using an agglomerative clustering method in R confirmed this: clusters showed some coherence for the individual languages, including for the three varieties of Spanish as a group, but not for the MA sprachbund (Figure 3). The coefficient of the cluster analysis was 0.74. A comparison of the mean distances among the participants sharing a particular value of any of the predictor variables shows that membership in the MA area aligns with no more than loose coherence in FoR use (Table 1). We do not interpret this negative finding as evidence against areal diffusion of practices of language use in general. The MA sprachbund is nowadays largely what one might consider a “fossilized” sprachbund, as contact among speakers of indigenous languages has been greatly reduced after the Conquista and now proceeds to a significant extent through Spanish where it occurs. By hypothesis, a fossilized area would only reflect the most time-stable among the original areal features.
Presentation & discussion: Meronymy in Yucatec and Zapotec (Tim Tilbe dissertation project)
A common trait of Mesoamerican languages is having a highly productive system of general meronyms (part terms) that are also terms for human body parts. It has been noted that languages differ in the strategies that speakers use for assigning meronyms to the parts of any given object. MacLaury has described Ayoquesco Zapotec meronymy as involving a mapping that preserves the overall abstract arrangement of the body. By contrast, Levinson has proposed an algorithm for the assignment of Tseltal meronyms, which depends only on the geometric properties of the target object and can result in non-unique assignment of terms. In all likelihood, Yucatec meronymy is broadly similar to that of Tseltal. At this point, the project I want to pursue for my dissertation is a comparison between Juchitan Zapotec and Yucatec meronymy, using a task designed to look for Whorfian effects of the difference in meronym assignment strategy. I will be discussing variations on that, and other options I am considering.