Taking Stock of the Cross‐Linguistic Data: Spatial Frames

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Abarbanell and Li—1

Taking Stock of the Cross‐Linguistic Data: Spatial Frames of Reference and Their Effect on Thought

LINDA ABARBANELL

Department of Psychology San Diego State University

Email: [email protected]

PEGGY LIDepartment of Psychology, Harvard University

Laboratory for Developmental Studies Email: [email protected]

patial frames of reference (FoR) have been one of the most promising yet controversial areas to test for linguistic relativity. A primary distinction is between languages like English that

primarily use an egocentric frame (e.g., left/right), and like Tseltal Mayan that use a geocentric frame (e.g., uphill/downhill). In studies across more than 20 languages, researchers from the Cognitive Anthropology Research Group (CARG) at the Max Planck Institute for Psycholinguistics found a striking correlation between the predominant FoR used by a language and speakers’ preferences for encoding small‐scale spatial arrays in memory on certain tasks. For example, in the “animals” task, participants view an array of toy animals at one table and then turn to face a second table where they are asked to make the “same” array. When turning, your body axes turns with you but the environment does not, creating at least two viable solutions. Speakers of languages like Dutch aligned the animals from the same egocentric perspective, while speakers of languages like Tseltal maintained their same geocentric orientation (Brown & Levinson, 1993). These results led some researchers to conclude that linguistic FoR (re)shape speakers’ non‐linguistic spatial representations, making it difficult for them to use their language‐incongruent system (Levinson, 2003; Majid et al, 2004; Pederson et al., 1998). Not all researchers, however, agreed with this conclusion, taking issue in particular with the open‐ended nature of the tasks which leaves it up to participants to decide what is meant by the “same” (Li & Gleitman, 2002; Li et al., 2011; Newcombe & Huttenlocher, 2000; Pinker, 2007). Two competing accounts have been proposed: a “linguistic relativity” account, where language actually (re)shapes non‐linguistic cognition, and a “pragmatic inference” account, where language affects speakers’ interpretation of the task. In this paper, we attempt to reconcile the data collected since the original CARG tasks, which we argue supports the latter account.

Before we describe our own data, it is worth pointing out there are other studies that do not quite support a linguistic relativity account. For example, Mishra et al. (2003) found that Hindi speakers who use an egocentric FoR in their language and Hindi, Nepali and Newari speakers that use a geocentric FoR all preferred the geocentric response on the animals task involving a static array but an egocentric response on a maze task involving a motion path. The same pattern was found among Balinese speakers who use a geocentric FoR in their language (Wassman & Dasen, 1998), suggesting that we encode different types of spatial information using different reference frames. Even some of the original CARG members sometimes struggled to reconcile inconsistencies in their results. In his work with Kilivila, Senft (2001, 2007) noted that many “uncontrolled parameters” could affect the results (2007: 241). Minor differences in procedures

S


such as whether participants carry the animals to the second table (Cottereau‐Reiss, 1999) and the wait time between the two tables (Brown & Levinson, 1993) have been found to affect the results.

In our work with Tseltal and English speakers (Abarbanell, 2010; Li et al., 2011), we adapted some of the original CARG tasks to have two matched conditions, egocentric and geocentric, with clear, correct solutions. By comparing the error rates across these conditions we were able to actually test which system is easier for speakers to use. We found that Tseltal, like English‐speakers, could reason equally well using either reference frame and even did better in the egocentric condition on some tasks. Some have argued that we are merely demonstrating speakers’ competence but not their preference (Bohnemeyer & Levinson, 2011; Haun et al., 2011). However, the error pattern on our tasks illustrates that this is not the case. In our “swivel chair” task (Li et al., 2011, Exp. 3), Tseltal‐speakers watched as an experimenter hid a coin in one of two boxes to their left/north or right/south side. They were then rotated, eyes closed, to face another direction, and then asked to indicate, eyes open, the location of the coin. Importantly, the number of errors in the geocentric condition varied by the degrees of rotation, with the most errors at 180º, that is, when the view of the environment was the most mismatched from participants’ initial orientation. In contrast, their performance on the egocentric trials was robust regardless of rotation. We did not force them to reason in a way they did not prefer, rather the error pattern confirms that Tseltal speakers take in such spatial information from a body‐based perspective, the same as you or I.

A second critique (see Bohnemeyer & Levinson, 2011) concerns the fact that in some of our tasks (e.g., Li et al., 2011, Exp. 1 & 2) participants carry the stimulus array, covered, from the first to the second table, either rotating it with their bodies in the egocentric condition or holding it stable with the environment in the geocentric condition. They then uncover the array to check their responses. Might this afford an alternative strategy in the egocentric condition? For example, could the Tseltal speakers have simply tracked which item was closest to this thumb or that, which can be easily expressed in Tseltal? We note that this would be difficult to do on the multi‐legged motion paths (Exp. 2). Moreover, our results held even when participants no longer carried the array (see the “leave box” and “leave maze” trials in Exp. 1 & 2), when the experimenter moved the array (Exp. 4), and even on tasks that required no carrying at all. Recently, for example, we tested Tseltal‐speaking children on the CARG group’s more difficult transitive inference task where the relationship between three objects is revealed two at a time across two tables using the transitive property (e.g., if A is left/north of B, and B is left/north of C, then A is left/north of C). In our version, we used models of fronted buildings rather than symmetrical forms, such that their facing orientation across the two tables indicated which FoR participants were expected to use. We found no difference in performance between the two conditions (F(1,23) = 1.67, p = .21). If anything, the children did better in the egocentric than the geocentric condition (66.7% vs. 52.5% correct).

A final challenge to our results came from Haun et al. (2011) who tested Hai//om (Namibia) and German‐speaking children on an animals‐type task and found that a difference in


performance between the two groups persisted even on instructed trials that, like our tasks, had correct solutions. On closer inspection, however, their instructed trials used left/right terms that are rarely used in Hai//om, and they did not control for a perseveration effect among the German children who always received the egocentric before the geocentric trials. Using a similar task with Tseltal and English‐speaking children, we found that once we eliminated left/right language from the instructions, the Tseltal children did as well as the English speakers in the egocentric condition, and curiously, both groups did better on the geocentric trials (Abarbanell, Montana & Li, 2011; Li & Abarbanell, revise & resubmit). How do we reconcile these results with the egocentric advantage found by Li et al. (2011)? The answer involves the distance and degrees of rotation between the two tables. With a 90º rotation and the tables close together so participants can see the same environmental landmarks, the children do better in the geocentric condition, but with 180º rotation this advantage disappears (Li & Abarbanell, revise & resubmit, Exp. 3). This finding concurs with studies in the spatial cognition literature (Rieser,1989; Presson, & Montello 1994; Farrell & Robertson, 1998; Simons & Wang, 1998). It also explains the findings of Haun et al. (2006) who reported a geocentric preference among three species of great apes and prelinguistic human infants. When Rosati (2015) performed a similar experiment with the tables far apart and a hallway in‐between, rather than abutted as in Haun and colleagues’ task, non‐human primates favored the egocentric solution. We are sympathetic to the position that the habitual use of a FoR in language might influence speakers’ cognitive preferences by making the underlying concepts more salient or better packaged for use, although this seems to be more plausible for spatial relations that are less readily available, such as the development of non‐egocentric left/right (Abarbanell & Li, 2009, 2015). The body of findings as outlined here, however, argue decisively against the strong claims of linguistic relativity that were based on the original CARG tasks (Levinson, 2003). Navigation and spatial reasoning require multiple FoR (see e.g., Burgess, 2005; Gallistel, 2002). Is it really sensible to think that the lack of linguistic expressions to express a FoR means that the underlying representations are not sufficiently exercised, practiced, or used? Given the extant data, we are inclined to believe speakers of different languages encode spatial scenes in much the same way, relying on the same cognitive hardware and processes, with task structure, rather than habitual language use, determining which system is easier to use in any given context.

References Abarbanell, L. (2010). Words and Worlds: Spatial Language and Thought among the Tseltal Maya. Unpublished Doctoral Dissertation. Harvard Graduate School of Education, Cambridge, MA.

Abarbanell, L. & Li, P. (2009). Spatial frames of reference and perspective taking in Tseltal Maya, Proceedings of the 33rd Annual Boston University Conference on Language Development. Somerville, MA: Cascadilla Press.

Abarbanell, L. & Li, P. (2015). Left‐right Language and Perspective‐taking in Tseltal Mayan Children. Proceedings of the 39th Annual Boston University Conference on Language Development. Somerville, MA: Cascadilla Press.

Abarbanell, L., Montana, R., & Li, P. (2011). Revisiting the Plasticity of Human Spatial Cognition. In M.

Egenhofer et al. (Eds.). Proceedings of the Conference on Spatial Information Theory: COSIT ’11, LNCS 6899: 245–263.


Bohenmeyer, J. & Levinson, S. (2011). Framing Whorf: A response to Li et al. (2011). Unpublished Manuscript, Retrieved June, 1, 2013 from the World Wide Web: http://www.cse.buffalo.edu/~rapaport/575/S11/Bohnemeyer_Levinson_ms.pdf

Brown, P. & Levinson, S.C. (1993). Linguistic and Nonlinguistic Coding of Spatial Arrays: Explorations in Mayan Cognition. Cognitive Anthropology Research Group, Max Planck Institute for Psycholinguistics.

Burgess, N., Spiers, H., & Paleologu, E. (2004). Orientational manoeuvres in the dark: dissociating allocentric and egocentric influences on spatial memory. Cognition 94: 149–166.

Cottereau‐Reiss, P. (1999). L’espace kanak, ou comment ne pas perdre son latin [Kanak space, or how not to lose one’s Latin], Annales Fyssen 14: 34–45.

Farrell, M.J. & Robertson, I.H. (1998). Mental rotation and the automatic updating of body‐centered spatial relationships. Journal of Experimental Psychology: Learning, Memory, & Cognition 24: 227–233.

Gallistel, C.R. (2002). Language and spatial frames of reference in mind and brain. Trends in Cognitive Science 6: 321–22.

Haun, D.B.M., Rapold, C.J., Janzen, G. & Levinson, S.C. (2011). Plasticity of Human Spatial Cognition: Spatial Language and Cognition Covary across Cultures. Cognition 119(1): 70–80.

Haun, D., Rapold, C. J., Call, J., Jenzen, G., & Levinson, S. C. (2006). Cognitive cladistics and cultural override in Hominid spatial cognition. Proceedings of the National Academy of Sciences of the United States of America 103(46): 17568–17573.

Levinson, S.C. (2003) Space in Language and Cognition. Cambridge University Press, Cambridge, UK. Li, P., Abarbanell, L., Gleitman, L. & Papafragou, A. (2011). Spatial Reasoning in Tenejapan Mayans. Cognition 120(1): 33–53.

Li, P. & Gleitman, L. (2002). Turning the Tables: Language and Spatial Reasoning. Cognition 83(3): 265–294.

Li, P. & Abarbanell, L. (revise and resubmit). Competing perspectives: Frames of reference in language and thought. First submission November 11, 2015 to Cognition.

Majid, A., Bowerman, M., Kita, S., Haun, D.B.M. & Levinson, S.C. (2004). Can Language Restructure Cognition? The Case for Space. Trends in Cognitive Science 8(3): 108–114.

Mishra, R.C., Dasen, P.R. & Niraula, S. (2003). Ecology, Language, and Performance on Spatial Cognitive Tasks. International Journal of Psychology 38: 366–383.

Newcombe, N.S. & Huttenlocher, J. (2000). Making Space: The Development of Spatial Representation and

Reasoning. MIT Press, Cambridge, MA.

Pederson, E., Danziger, E., Wilkins, D., Levinson, S.C., Kita, S. & Senft, G. (1998). Semantic Typology and Spatial Conceptualization. Language 74(3): 557–589.

Pinker, S. (2007). The Stuff of Thought. Penguin Group, Inc., New York

Presson, C. C. & Montello, D. R. (1994). Updating after rotational and translational body movements:

Coordinate structure of perspective space. Perception 23: 1447–1455.

Rieser, J. J. (1989). Access to knowledge of spatial structure at novel points of observation. Journal of Experimental Psychology: Learning, Memory, and Cognition 15(6): 1157–1165.

Rosati, A. G. (2015). Context influences spatial frames of reference in bonobos (Pan paniscus). Behaviour 152: 375–406.

Senft, G. (2001). Frames of Spatial Reference in Kilivila. Studies in Language 25(3): 521–555.

Senft, G. (2007). The Nijmegen space games: Studying the interrelationship between language, culture and cognition. In J. Wassmann & K. Stockhaus (eds.), Experiencing new worlds (pp. 224–244). Oxford: Berghahn Books.

Simons, D.J. & Wang, R.F. (1998). Perceiving Real World Viewpoint Changes. Psychological Science 9: 315–320.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Hellan and Beermann—5

A System for Capturing Variation inSpatial Referencing across Languages

LARS HELLAN

Emeritus Norwegian University of Science and

Technology Email: [email protected]

DOROTHEE BEERMANN

Department of Linguistics Norwegian University of Science and

Technology Email: [email protected]

e present a system of linguistic representation residing in a spatial situation type

ontology, which provides a reference point for language comparison, independently of

the particular strategies employed in the grammars. We illustrate the design with constructions

expressing direction and location in Germanic and Kwa languages. In our analysis we highlight the

extensive use of multiverb expressions in the latter versus single‐verb‐headed constructions in

the former, expressing the same types of content. The overall design spans three levels, viz., a

spatial type ontology, a formal (large scale computational) grammar, both formalized in terms of

a typed feature system employed in constraint based grammars (cf. Pollard and Sag 1994,

Copestake 2002), and a format of online corpus annotation and databasing of natural language

hosted in TypeCraft (Beermann and Mihaylov 2014), where morpho‐syntactic information is

made accessible and interpretable at multiple analytic levels, from data‐mining to formal

analysis.

Figure 1. A type hierarchy related to arguments of prepositions.

The content of our analysis reflects traditions of cognitive semantics as represented, e.g., in

Talmy 2000, Jackendoff 1990, Pustejovsky 1993; see especially Davis 2000, and Trujillo 1995 for

approaches similar to the present in this respect. Following Jackendoff 1996, we assume a basic

distinction between prepositions residing in whether their arguments represent one‐dimensional

entities or not. We refer to one‐dimensional entities as line items, as opposed to xdim items

(subsuming both “individuals” and “events”). The distinction applies to both the spatial and the

W


temporal domain, and the following diagram indicates some of the types in the spatial domain,

the branches under oriented object representing a verb argument or “external” argument of

preposition representing a “path” role (a “Figure”), and those under spatial‐entity typical roles of

the governee of a preposition (a “Ground”):

In the domain of locative relations, features include (cf., e.g. Trujillo 1995):

Topological features Definitions:

FRONT FIG is in front of GRND

BACK FIG is behind GRND

EMBEDDED FIG is embedded in GRND

CONTAINED FIG is contained in GRND

SCALAR Relation between FIG and GRND can be quantified

TRANSITIVE If R(A,B) and R(B,C), then R(A,C)

UPSIDE‐OF FIG is upside in a vertical relation to GRND

DOWNSIDE‐OF FIG is downside in a vertical relation to GRND

INTEGRATED FIG is integrated into GRND

These are all features that may be assumed to have significant cross‐linguistic relevance. It is well

known that they can in principle be carried not only by prepositions, but also by nouns and by

verbs, and in languages where there are no prepositions, we may expect transitive verbs of

roughly the same type as reach or pass to function in similar ways as prepositions do in English. In

the following example from Akan, we see these patterns illustrated, with a relational noun mu

(“inside”) in a function similar to that of a postposition, representable as “CONTAINED +

TRANSITIVE +”:

(2) John tuu ɔboɔ no faa ntokura no mu. (“John threw the stone through the window”)

John tu‐u ɔ‐boɔ no fa‐a ntokura no mu.

John throw‐PST SG‐stone DEF pass‐PST window DEF inside

N V1 N DET V2 N DET N

The V2 in the Serial Verb Construction would translate as a preposition in English; its status as

a verb in Akan is demonstrated, for instance, by its tense marking. In terms of ROLE features as

represented in Figure1, the V2 would be encoded as having “the stone” as a locomotor item, and

“the window” as viapnt.

Within the formal model in question, the semantic analysis will involve two levels, one where

the semantic features of the word units are represented relative to the words as such, and one

where they are pooled onto a more abstract level of representation – “Situation Structure” ‐

where, e.g., the sentence in (2) and its English translation will have the same representation.

While formally well defined, the aspiration of such a level to serve as an “interlingua” of course

faces well‐known challenges relating to such a construct; however, the domain of spatial

reference is restricted and well enough understood to be modeled by such a design.


References Dorothee Beermann and Lars Hellan. 2004. A treatment of directionals in two implemented HPSG grammars. In

Stefan Müller (ed) Proceedings of the HPSG04 Conference, Katholieke Universiteit Leuven. CSLI Publications

/http://csli‐publications.stanford.edu/

Dorothee Beermann and Pavel Mihaylov. 2014. TypeCraft collaborative databasing and resource sharing for

linguists. Language Resources and Evaluation. vol. 48 (2).

Ann Copestake. 2002. Implementing Typed Feature Structure Grammars. CSLI Publications, Stanford.

Anthony Davis. 2000. The Hierarchical Lexicon. CSLI Publications, Stanford.

Lars Hellan and Dorothee Beermann. 2005. Classification of Prepositional Senses for Deep Grammar

Applications. In: Proceedings of the 2nd ACL‐Sigsem Workshop on The Linguistic Dimensions of Prepositions

and their Use in Computational Linguistics Formalisms and Applications, Colchester, United Kingdom, ACL‐

Sigsem, 2005.

Ray Jackendoff. 1996. The proper treatment of measuring out, telicity, and possibly even quantification in

English. Natural Language and LinguisticTheory. 14. 305–354.

Carl Pollard and Ivan Sag. 1994. Head‐driven Phrase Structure Grammar. Chicago: Chicago Univ. Press.

James Pustejovsky 1993. Semantics and the Lexicon, vol 49 of Studies in Linguistics and Philosophy. Kluwer,

Dordrecht, The Netherlands

Leonard Talmy. 2000. Towards a cognitive semantics. MIT Press.

Arthuro Trujillo. 1995. Lexicalist Machine Translation of Spatial Prepositions. Doctoral thesis, Cambridge

University.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Bohnemeyer—8

Spatial Reference: Studying the Interplay ofLanguage, Culture and the Environment

JÜRGEN BOHNEMEYER Department of Linguistics

University at Buffalo Email: [email protected]

ntroduction—Recent advances in the availability of affordable computational power have

made it possible to subject large‐scale crosslinguistic and crosscultural data samples to

sophisticated statistical analyses that can isolate the effects of language, culture, and the

environment in spatial cognition. The use of spatial reference frames in discourse and

nonlinguistic cognition offers an ideal test case for such studies.

Background—One of the oldest and most persistent questions in the history of science is that of

nature vs. nurture, or the respective role of genetics and personal experience in shaping the

individual’s behavior. The latter in turn presumably reflects influences of the environment, some

of which may be mediated by culture. Two widespread assumptions in the cognitive sciences

(especially artificial intelligence, psychology, and linguistics) are, first, that the study of the mind

should focus on those aspects of cognition that are determined by nature, i.e., are innate, and

secondly, that variation in cognitive traits across human populations largely responds to

environmental and cultural factors and therefore falls outside the study of “cognition proper.” As

a result, comparatively little empirical research into population‐specific cognitive traits has been

conducted. More specifically, we know a great deal about certain pieces of the larger puzzle of

the role of biology, culture, and the environment in human cognition. But these pieces have been

studied in isolation, and no studies have been carried out that examine the interaction of all of

them. In addition, much of this research has proceeded qualitatively, arguing for example for

effects of physiography on dialect differentiation on the basis of the pronunciation of individual

words and sounds, rather than quantitatively.

Recently, two collaborative projects under my direction, funded by the National Science

Foundation (NSF awards BCS‐0723694 and BCS‐1053123) and collectively known under the

acronym MesoSpace, have been investigating for the first time how a single aspect of cognitio—

the use of spatial reference frames (Carlson Radvansky & Irwin 1992; Gallistel 1990; Levelt 1990;

Levinson 1996)—is shaped by language, literacy, education, and two environmental variables,

topography and population density. The research of the MesoSpace team for the first time

presents a fine‐grained picture of how these variables interact in influencing human behavior.

The results point to a far more powerful role of culture in the mind than most cognitive scientists

have assumed.

Spatial reference frame use as a cognitive anthropology laboratory—What makes reference

frames so suitable for studying the interplay of biology, culture, and the environment in

cognition is a combination of four properties: (i) they are indispensable in identifying

I


nontopological regions of space and thus can be assumed to be evolutionarily ancient and to

have a biological basis (Gallistel 1990); (ii) yet, there is considerable variation across human

populations in the types of frames customarily used for reference at the small scale; (iii) a given

population’s linguistic preferences fairly narrowly predict its preferences in nonverbal tasks,

including in inferences and memory (Pederson et al 1998; Levinson 2003; Mishra et al 2003;

Majid et al 2004; Haun et al 2011; Le Guen 2011; Bohnemeyer et al 2014); and (iv) geocentric

frames are sensitive to the environment in that their axes are defined with respect to landmarks

or gradients of the environment (with varying levels of abstraction; Wassman & Dasen 1998;

Levinson 2003; Polian & Bohnemeyer 2011; Bohnemeyer & O’Meara 2012; Palmer 2015). It was

in the context of research probing the covariation between frame selection in discourse and

nonverbal cognition (property (iii)) that the question of the factors driving frame use across

populations first came up. Pederson et al (1998) hypothesized that this covariation was the result

of different languages influencing the nonverbal cognition of their speakers in different ways—a

language‐on‐thought effect in line with the Linguistic Relativity Hypothesis (Whorf 1957). In

contrast, Li & Gleitman (2002) argued that the alignment between frame use in language and

nonverbal cognition was epiphenomenal: participants’ behavior in both types of tasks was driven

by the same set of nonlinguistic variables, including education and literacy, but also topography

and population density.

The MesoSpace studies—Bohnemeyer et al (2014, 2015, under revision) investigate how

speakers of eight indigenous languages of Mexico and Nicaragua and L1‐speakers of the

dominant contact language, Spanish, in Mexico, Nicaragua, and Spain talk about and memorize

the location and orientation of objects in space. These studies have for the first time

demonstrated quantitatively the impact of topography and population density on frame use. The

statistical models employed by the group (mixed effects logistic regression models) also showed

that the role of the first language cannot be reduced to any combination of the other factors.

Furthermore, the indigenous participants proved to be more likely to use the “relative” subtype

of egocentric frames (Levinson 1996) in their native languages the more frequently they use

Spanish as a second language. This points to Spanish serving as a conduit for the diffusion of

egocentrism in the area.

In unpublished work, the MesoSpace team has extended this investigation to a population

sample that includes speakers of English, Vietnamese, and two Mesoamerican languages

(Isthmus Zapotec and Yucatec Maya), as well as members of two Taiwanese populations

(monolingual Mandarin speakers and Mandarin‐Taiwanese bilinguals) and four Japanese

populations (rural vs. urban speakers from Honshu vs. Okinawa). This is the largest and most

diverse study of the use of reference frames in language to date. Preliminary results confirm the

non‐epiphenomenal role of language. The linguistic study also showed effects of literacy and

population density, while the recall memory study showed in addition to language topography as

a significant factor, but this apparent effect is at present still being probed.


Evidence of a pan‐simian geocentrism bias and the rise of the small scale—Several of the

MesoSpace studies produced evidence of a cognitive geocentrism bias in populations that show

no clear preference in linguistic tasks. This surprising finding is consistent with the hypothesis of

an innate pan‐simian geocentrism bias that can be reshaped through the effect of language and

other observable cultural practices (Haun et al 2006). In the context of the other MesoSpace

findings, a possible scenario for the cultural evolution of egocentrism emerges. According to this

scenario, egocentrism has been culturally “selected for” by an ever‐expanding importance of

control of the small scale in human behavior with the advent of tool use, manufactured walled‐

off spaces, and eventually manufactured visual representations, especially writing. Observable

cultural practices such as speech, gesture, and writing serve as transmission systems that allow

the members of a community to converge on the non‐innate practice of egocentric frame use. Selected references Bohnemeyer, J., E. Benedicto, K. T. Donelson, A. Eggleston, C. K. O’Meara, G. Pérez Báez, R. E. Moore, A.

Capistrán Garza, N. Hernández Green, M. S. Hernández Gómez, S. Herrera Castro, E. Palancar, G. Polian, & R.

Romero Méndez (Under revision). The linguistic transmission of cognitive practices: Reference frames in and

around Mesoamerica. Manuscript, University at Buffalo.

* Bohnemeyer, J., K. T. Donelson, R. E. Moore, E. Benedicto, A. Capistrán Garza, A. Eggleston, N. Hernández

Green, M. S. Hernández Gómez, S. Herrera Castro, C. K. O’Meara, G. Pérez Báez, E. Palancar, G. Polian, & R.

Romero Méndez (2015). The contact diffusion of linguistic practices: Reference frames in Mesoamerica.

Language Dynamics and Change 5(2): 169–201.

* Bohnemeyer, J., K. T. Donelson, R. E. Tucker, E. Benedicto, A. Eggleston, A. Capistrán Garza, N. Hernández

Green, M. S. Hernández Gómez, S. Herrera Castro, C. K. O’Meara, E. Palancar, G. Pérez Báez, G. Polian, & R.

Romero Méndez (2014). The cultural transmission of spatial cognition: Evidence from a large‐scale study.

Proceedings of the 36th Annual Meeting of the Cognitive Science Society.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Carlson and Kolesari—11

Reference Frames as Mechanismsfor Mapping Language onto Space

LAURA CARLSONDepartment of Psychology

Vice President, Associate Provost, and Dean of the Graduate School University of Notre Dame Email: [email protected]

JENNIFER KOLESARIDepartment of Psychology University of Notre Dame Email: [email protected]

eference frames are considered mechanisms for mapping language onto space (Carlson,

1999; Landau & Jackendoff, 1993; Levelt, 1984; Levinson, 1996, Logan, 1995). A typical

mapping example involves a spatial description of a perceptual event. For an example from

Carlson (2003), imagine a speaker telling a listener who is holding a coffee pot: (1) “The coffee

mug is below the coffee pot.” Each of these objects plays a distinct role in this spatial description.

The coffee pot serves as a reference object, acting as a landmark from which to describe the

location of the coffee mug, also known as the located object. The spatial description thus assists

the listener in finding the located object by reducing the area that needs to be searched to the

space around the reference object. This space is demarcated through the application of a

reference frame, a family of representations (Shelton & McNamara, 2001) that are instantiated

through a set of parameters that apply the representation to a particular perceptual event

(Logan & Sadler, 1996). A reference frame is typically thought of as consisting of a set of

coordinate orthogonal axes whose intersection point is called the origin. The origin is imposed

upon the reference object and defines the surrounding space through the orientation and

direction parameters. Orientation specifies whether a given axis is horizontal (front/back or

left/right) or vertical; direction specifies the endpoint of a given axis (for example, left vs. right).

These axes also have a scale that indicates the units of distance applied to the space. Finally, a

spatial template further parses the space around the reference object into regions for which the

spatial terms offers a god, acceptable or unacceptable characterization of the located object’s

placement (Carlson, Regier & Covey, 2003; Carlson‐Radvansky & Logan, 1997; Logan & Sadler,

1996).

Most of the research on setting the parameters of a reference frame has been done with

English speakers. Below, I discuss the factors that impact these parameters, and address

the possibility of cross‐linguistic variation.

Origin. Previous research has shown that the identity of the located object, the reference object

and the functional interaction between the objects play a role in defining the origin—that is,

where the reference frame gets placed within the reference object (Carlson‐Radvansky et al.,

1999; Carlson & Kenny, 2003). For example, in a neutral context in which the speaker in (1) is

drawing the listener’s attention to the mug as a souvenir from a trip, its relationship to the coffee

pot is not emphasized, and the located object is assumed to be geometrically below the center of

R


mass of the coffee pot itself. However, in the context of a speaker making an indirect request of

the listener to pour a cup of coffee (Clark, 1996), the ideal location of the coffee mug is not in

fact under the pot at all, but under the spout, off to the side. Thus, the role of the objects and

their interaction may play a critical function in defining the origin. Coventry, Prat‐Sala & Richards

(2001) (see also Carlson, 2000) have shown that the strength of this functional influence may

vary across types of spatial term (for example, over vs. above). This within language variation

suggests the possibility of cross‐language variation in the relative strength of geometric and

functional information.

Orientation and Direction. Orientation and direction together assign directions to space around a

reference object. Different sources of information can be used to set these parameters, resulting

in different types of reference frames. For example, Levinson (1996) proposes that the features

in the environment define the absolute reference frame, the speaker or listener defines the

relative reference frame, and the reference object defines the intrinsic reference frame. Often

times, these sources of information are in conflict, resulting in different mappings for a given

spatial term. Some work from my lab shows that for English speakers initially all reference frame

mappings are considered, followed by an inhibition of the non‐selected reference frame

(Carlson‐Radvansky & Jiang, 1998). Looking at the locus of this inhibition, it appears that

particular preferred axes are always inhibited, and less‐preferred axes inhibited only selectively

(Carlson & van Deman, 2008). Because cross‐linguistically there are different preferences for

using different types of reference frames (Levinson, 2003), this suggests that locus of inhibition

and the manner in which a reference frame is selected, and more particularly, the locus of

inhibition, may differ cross‐linguistically.

In addition, there is evidence that orientation and direction are separate representations. For

example, Logan (1995, 1996) observed savings in response time in a spatial cueing task in which

participants could respond on the basis of orientation, with additional time needed when

direction had to be further specified. Relatedly, Hoffman et al. (2003) observed impairments in

patients with Williams Syndrome in a placement task, such that errors were more likely to occur

at the wrong endpoint of the correct axis, rather than at a random location, indicating some

preservation of axial structure without a further refinement by endpoint. These findings have

important cross‐linguistic implications, particularly for languages that do not explicitly demarcate

left vs. right, but rather use the more general “side”. For these speakers, it may be possible that

the endpoints remain unspecified until needed.

Scale and Spatial Templates. The scale and spatial extent of the region that is demarcated by a

spatial term has been woefully under‐studied, both within a given language and across

languages. Morrow and Clark (1988) observed that the size of the located and reference objects

significantly impacted the distance that was inferred between them for spatial descriptions

containing the verb “approach.” Carlson and Covey (2005) built on these findings and

demonstrated that English speakers were influenced by the properties of the objects in inferring


their distance for topological (e.g., “near”) and projective (e.g., “left”) spatial terms. Importantly,

the spatial term also influenced distance estimates. To the extent that terms are associated with

different distances cross‐linguistically, one may thus expect to see variation in the spatial extent

and scale of these regions (for example, see Regier & Carlson, 2002).

References Carlson, L. A. (1999). Selecting a reference frame. Spatial Cognition and Computation 1: 365–379.

Carlson, L. A. (2000). Object use and object location: The effect of function on spatial relations. In E. van der Zee & U. Nikanne (Eds.). Cognitive Interfaces: Constraints on linking cognitive information (pp. 94–115). Oxford: Oxford University Press.

Carlson, L. A. (2003). Using spatial language. The Psychology of Learning and Motivation 43: 127–161.

Carlson, L. A. and Covey, E. S. (2005). How far is near? Inferring distance from spatial descriptions. Language and Cognitive Processes 20: 617–631.

Carlson, L. & Kenny, R. (2006). Interpreting spatial terms involves simulating interactions. Psychonomic Bulletin & Review 13: 682–688.

Carlson, L., Regier, T., & Covey, E. (2003). Defining spatial relations: Reconciling axis and vector representations. Representing direction in language and space 1: 111–131.

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, L. A. & Van Deman, S. (2008). Inhibition within a reference frame during the interpretation of spatial language. Cognition, 106, 384–407.

Carlson‐Radvansky, L. A., & Jiang, Y. (1998). Inhibition accompanies reference‐frame selection. Psychological Science 9: 386–391.

Carlson‐Radvansky, L. A., Covey, E. S., & Lattanzi, K. M. (1999). “What” effects on “where”: Functional influences on spatial relations. Psychological Science 10: 516–521.

Clark, H. H. (1996). Using language. Cambridge: Cambridge University Press.

Coventry, K., Prat‐Sala, M., & Richards, L. (2001). The interplay between geometry and function in the comprehension of over, under, above and below. Journal of Memory and Language 44: 376–398.

Hoffman, J. E., Landau, B., & Pagani, B. (2003). Spatial breakdown in spatial construction: Evidence from eye fixations in children with Williams syndrome. Cognitive Psychology 46: 260–301.

Landau, B., & Jackendoff, R. (1993). ‘‘What’’ and ‘‘where’’ in spatial language and spatial cognition. Behavioral and Brain Sciences 16: 217–265.

Levelt, W. J. M. (1984). Some perceptual limitations on talking about space. In A. J. van Doorn, W. A., van der Grind, & J. J. Koenderink (Eds.), Limits in perception (pp. 323–358). Utrecht: Wiley.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Cashdan and Davis—14

Adaptive Flexibility in Spatial Referencing

ELIZABETH CASHDAN Department of Anthropology

University of Utah Email: [email protected]

HELEN ELIZABETH DAVIS Department of Anthropology

University of Utah Email: [email protected]

ow flexible are spatial reference frames? Do individuals within a culture shift their spatial

referencing in response to different environmental contexts? Are cultural differences in

spatial referencing adaptive responses to local navigational challenges?

We are newcomers to this field, but our impression is that the emphasis has been on the

constraining role of language in shaping spatial referencing. However, spatial frames of reference

affect navigation, and navigation is frequently a matter of life (resources from travel) and death

(costs and risks associated with travel). We should have evolved to get it right. Within some

constraints of language, therefore, we might expect individuals to use the one most suitable for

the problem at hand.

If different frames of reference are adaptive in different environmental contexts, spatial

referencing should respond flexibly in response to those environmental contexts. For example,

some interesting work has shown how people in areas of sharp topographic relief use those

topographic cues in spatial referencing. More generally, geocentric (absolute) frames of

reference facilitate a survey knowledge of the environment, and should be adaptive in areas

where appropriate cues (distal landmarks, celestial markers) exist and individuals frequently

traverse large ranges with unpredictable routes. Viewpoint-centered (relative) frames of

reference should be most adaptive where distinctive proximal landmarks are available and

individuals frequently travel along established routes.

Whether this adaptive flexibility explains existing variation in spatial referencing is an

empirical question, and we are not in a position to answer it. But as human evolutionary

ecologists, our bias is to look in that direction for an answer.

We began this in Tonga, where we elicited spatial frames of reference with an “animals in a

row” test (our version assessed absolute and relative but not intrinsic reference frames). Tongan

has native terms for left and right as well as for cardinal directions. We wanted to see whether

there were individual differences in reference frames that reflected differences in mobility

experience and local context, as well as navigational accuracy. Local context clearly matters: We

chose different conditions for testing, and all the people who were tested outdoors and did the

task on the ground used an absolute frame of reference, the same was true for most (85%) of the

people who were tested outdoors but did the task on a table, whereas those who were tested

indoors in a rectangular meeting room were split about evenly in their responses: 56% (n = 35)

answered as Westerners would, using a relative reference frame, while the rest (n = 27)

answered as the outdoor Tongans did, using an absolute reference frame. Tongans differ in their

English language competence (from none to fluent), but this did not seem to explain the variation

H

mailto:[email protected]

mailto:[email protected]

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Cashdan and Davis—15

among people tested indoors. We do not yet know whether the variation reflects functional

differences in peoples’ navigational experience, but we will have an answer by December.

We hope to explore these issues further in our other field sites, beginning with the Tsimane,

forager-horticulturalists in Bolivia. In the tropical forest habitat of the more remote Tsimane

villages, geocentric cues (sun position and distal landmarks) are often obscured due to heavy

canopy and cloud cover. Previous findings suggest that these are cues men tend to favor where

they are available, which may be why we see no sex difference in navigational ability in this

population. We may find that this also affects spatial referencing, perhaps placing a premium on

relative frames of reference in the inter-riverine villages, while villages along the rivers take

advantage of the rivers themselves as a reference frame. Navigational differences are also found

in villages near the market town, where the environment is more open due to felled trees, travel

takes place along roads, and Tsimane have greater Spanish fluency and access to formal

education. We plan to return to the field in October 2016 to investigate differences among these

villages in both navigation and spatial reference frames.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Chatterjee—16

Anatomical Spatial Referencing in Medical Education

ALLISON K. CHATTERJEEMarian University College of Osteopathic Medicine, Indiana

Adjunct, Department of Anatomy and Cell Biology Indiana University‐Purdue University Indianapolis (IUPUI)


edical education imposes a specific universal international vocabulary to describe

anatomical structures, positions, and relationships upon the students. This vocabulary is

consistent with the Terminologia Anatomica: International Anatomical Terminology and establishes

a common, consistent language among medical professionals regardless of the individual’s country

of origin or native language (Whitmore, 1999). This terminology is rooted in the Latin and Greek

languages. Very rarely do students enter medical school with a background in these languages.

Those that do typically have only had a basic medical terminology course, which is not a

requirement for matriculation into medical school. Other courses in which they would be exposed

to medical terminology – anatomy, histology, physiology, and neuroscience—also are not

prerequisites to entering medical school. As a result, students are required to learn a new language

for orienting the structures of the body in space while concurrently attempting to learn the

clinically‐relevant material necessary to be a competent physician.

This anatomical referencing system is based on a standardized orientation of the body—the

anatomical position—in which a person is standing with their arms at their sides, palms facing

forward, and toes facing forward. The terminology used to describe the spatial orientation of

anatomical structures includes both relative and absolute terms. For example, a structure that is

located above another structure can be described as either superior or more cranial. Superior

would then be a relative term; while cranial would be considered an absolute term.

The extent to which students think about the spatial organization of the body in anatomical

terminology versus their respective native languages is expected to vary significantly. One question

of interest then is: What factors determine how quickly, effectively, and efficiently students adopt

anatomical terminology and incorporate them into their schemas? Several potential factors

include: (1) students’ familiarity with the language; (2) students’ level of experience; (3) students’

measured spatial ability; and (4) students’ enrollment in an osteopathic versus allopathic medical

program.

Familiarity with Language

An individual’s native language could play a role in how well they are able to assimilate spatial

anatomical terminology into their schemas. Those whose native language is most similar to Latin

and Greek may have an easier time learning anatomical terminology. Those whose native language

is vastly different from Latin or Greek might struggle to incorporate anatomical terminology into

their schema. Instead, they might use their own language as the main reference frame, with the

anatomical spatial language simply added on. This could be seen as someone redefining the word

“medial” as “close to the middle.”

M


Novice vs Expert

Those students who have taken prior elective coursework that emphasizes Latin or Greek

languages may be more adept at navigating anatomical terminology. Those who have taken prior

courses in the anatomic disciplines—anatomy, histology, neuroscience, and embryology—may also

have an advantage compared to novice learners who might be expected to rely on memorization of

terms or the use of mnemonics in order to assimilate the new spatial terminology. Additionally, as

students progress through their medical education from pre‐clinical to resident to experienced

physician, they encounter more complex examples of spatial anatomical relationships through

clinical cases, patient interactions, and imaging studies (i.e., MRI, CT, ultrasound). Through these

experiences it is assumed that they implicitly build upon their schema and more effectively

incorporate anatomical terminology into that schema.

Spatial Ability

Those who score high on standardized assessments of spatial ability may be able to utilize new

spatial terminology more effectively than those whose scores are low. A study investigating the

differences between first‐year medical students’ mental models found that students who scored

high on the Purdue Visualization of Rotations Test consistently used more spatial anatomical

terminology in their explanations than those students who scored low on the same test (Chatterjee,

2011). Replication of these results with additional spatial tests would provide further evidence that

spatial ability is directly related to effective and efficient use of new spatial language.

Osteopathic vs Allopathic

Students attending osteopathic medical schools (schools granting a DO degree) might be able to

apply anatomical terminology faster than those attending allopathic medical schools (schools

granting an MD degree). Students at osteopathic medical schools receive additional hands‐on

training in osteopathic manual/manipulative medicine. This training reinforces understanding of

the spatial relationships of anatomical structures. This kinaesthetic approach may lead to more

robust schema that incorporate anatomical terminology.

These four factors are not likely to occur in isolation of each other. Different combinations of these

factors might account for individual differences in students’ abilities to incorporate anatomical

terminology in their understanding of anatomical spatial relationships.

The focus of this narrative was specifically on anatomical spatial referencing as opposed to

clinical terminology in general. Although Terminolgica Anatomica is considered a universal

vocabulary, it is not static. Over time there have been changes and updates to accommodate new

discoveries, common trends, and simplification. For example, a nerve in the lower extremity was

previously known as the common peroneal nerve. This has recently been renamed the common

fibular nerve to indicate its spatial location on the fibular/lateral side of the lower extremity. This

can complicate communication between those who learned the old terms and those who have

learned the new terms. Additionally, Terminologica Anatomica does not include all clinical

terminology. Clinical terminology includes many eponyms, in which the name of a structure is

derived from the name of a person. These terms are much less intuitive and do not follow a


systematic naming convention, as such it is less likely that clinical terminology has spatial

applications.

References Chatterjee, A. K. (2011). The Importance of Spatial Ability and Mental Models in Learning Anatomy. (PhD), Purdue University, West Lafayette.

Whitmore, I. (1999). Terminologia Anatomica: New Terminology for the New Anatomist. Anat Rec 257(2): 50–53.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Comrie–19

Context‐specific Inconsistencies in the Use of Spatial Reference Systems

BERNARD COMRIE Distinguished Professor Department of Linguistics

University of California, Santa Barbara Email: [email protected]

ne topic that interests me as a linguist is the interaction between the content of general concepts and deviations from that content that arise in particular contexts. A simple

example would be the meaning and use of the prefix kilo‐, namely 1000. However, a kilobyte is not 1000 bytes, but rather 1024 bytes. In this case the usage is conventionalized (kilobyte cannot mean 1000 bytes), as well as having an obvious motivation, namely the fact that bytes are counted as powers or 2 (1024 = 210, which is moreover the power of 10 closest to 1000). An example closer to the thrust of this paper is the identification of hot and cold faucets. The surest way of identifying each faucet is to run the water and test its temperature. In addition, each faucet may be identified by the word “hot” or “cold,” or an abbreviation thereof, or an icon such as red color for hot and blue color for cold—these thus stand for the result of applying the direct temperature test. In addition, many cultures have a convention for arranging the two faucets, e.g., in the U.S. the hot tap to the left, the cold tap to the right. Someone accustomed to this arrangement may well take left position to indicate hot, right position to indicate cold, even in the absence of any explicit indication on the faucet, indeed even contrary to such indication. Left‐right position thus “hijacks” conceptually the actual temperature of a faucet—until, of course, one feels the water.

With respect to space, a fair amount of attention has been paid recently to shifts between different frames of reference in shifting between different tasks, as witnessed by several contributions to this special meeting. My own interest is somewhat different, concerning as it does shifts in the orientation of an individual frame of reference in shifting between different tasks. While I have presented this phenomenon to an audience of linguists (Comrie 2003), and to a broader audience in a 3‐minute talk under the auspices of the UCSB Center for Spatial Studies, I would welcome feedback from and discussion with a broader spectrum of specialists from different disciplines with an interest in spatial reference systems. Both examples cited here come from my own experience. Although they have been confirmed anecdotally by others, suggestions for more systematic investigation are welcomed.

The first concerns left‐right orientation. In general, I do not have problems with “left” versus “right,” and can orient myself systematically and correctly when given directions in these terms. It was therefore somewhat disconcerting, when I moved from the UK to the US, to find that I would often confuse the two directions when receiving instructions while driving, turning to the right when told to turn to the left and vice versa. I did not experience the problem in contexts other than driving, nor had I experienced similar problems when driving in the U.K. A moment’s

O


thought revealed why the problem had arisen, although unfortunately it did not provide a solution.

In the UK, traffic drives on the left‐hand side. This means that a left‐hand turn is (other things equal) easier than a right‐hand turn, since the latter comprises all the factors involved in the former plus the additional factor that one is driving across the line of oncoming traffic. In the U.K., therefore, a left‐hand turn is an “easy” turn, while a right‐hand turn is a “difficult” turn. This provides a mechanism for the concepts “easy turn” and “difficult turn” to hijack the content of the terms “left” and “right” respectively when driving. The change from “left/right” to “easy/difficult turn” makes no practical difference in a context such as the U.K. where traffic drives on the left, indeed it had never occurred to me that I had made the change. However, in the context of the U.S., where traffic drives on the right, the change has the predictable outcome, namely confusion of left with right and vice versa—but only in the context of driving. Indeed, mixed contexts, e.g. describing the interacting behavior of a vehicle and pedestrians, can lead to recognition of the conflict, a perceived contradiction. Recognition of the conflict does not, however, necessarily provide a practical solution. The shift from “left/right” to “easy/difficult turn” is so powerful that, even though I have not driven in the U.K. or any other country that drives on the left for more than a decade, I still have to concentrate, when driving, on instructions that involve “left” and “right” if I am not to veer off in the wrong direction.

The second example concerns the cardinal directions. Although I do not regularly update myself on the cardinal directions of my environment, I have no problem in principle with the concepts involved and can orient myself successfully, for instance, in a grid system that is oriented to the cardinal directions. When I moved to Los Angeles I was therefore surprised to find that I systematically confused the cardinal directions, more specifically confusing “south” with “north” and “east” with “west”. I had never experienced this problem in my home region around the city of Sunderland on England’s North Sea coast, nor before or since in most other places where I have lived for an extended period at various periods of my life: Schondorf (southern Germany), Cambridge (England), Moscow (Russia), Aradip (highland Papua New Guinea), Leipzig (Germany), Tokyo (Japan).

My home region is located on a coastline that runs north‐south, with the sea to the east. This thus provides a fertile base for the sea to hijack the cardinal directions—“east” is towards the sea, “west” away from the sea, “north” to the left as one faces the sea, and “south” to the right as one faces the sea. All of this works perfectly on an east coast. But on a west coast, as in California, it systematically gives the wrong results. Interestingly, in places that are far enough from the sea, or that do not obviously orient themselves to the sea, such as the other places listed above, the problem does not arise—there is no sea to cause confusion, just as when not driving there is nothing to hijack the usual interpretations of “left” and “right.” And in Santa Barbara, where the freeway runs west‐east but is said to run north‐south, I just have to concentrate, welcoming the habitual use of explicit single directions such as to the north, towards the mountains. These observations, though surely valid, require more general investigation, more explicit grounding, and more rigorous testing.


Reference

Comrie, Bernard. 2003. Left, right, and the cardinal directions: Some thoughts on consistency and usage. In Erin Shay & Uwe Seibert, eds.: Motion, Direction and Location in Languages: In Honor of Zygmunt Frajzyngier, 51–58. Amsterdam: John Benjamins.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Cooperrider—22

Ecologies of Spatial Language

KENSY COOPERRIDERDepartment of Psychology

University of Chicago Email: [email protected]

patial language is part of a broader ecology. How we talk about space is woven in with how

we gesture and reason; with the natural and built worlds we inhabit; with our cultural

conventions and practices; with the myths we tell and metaphors we live by. My interest is in

understanding these rich ecologies—in identifying the invisible threads connecting spatial

language to everyday behavior and everyday thinking. Investigating these ecologies will

ultimately help us answer a big, two‐part question: What are the causes of the diversity we see in

spatial language across speech communities, and what are the consequences of such diversity?

But we shouldn’t get ahead of ourselves. Before we can sort out cause and consequence, we

need a better understanding of the concomitants of linguistic diversity: those aspects of everyday

behavior and thought that go hand in hand with cross‐linguistic differences in spatial language.

Spatial gestures

One concomitant of spatial language that remains to be fully understood is spatial gesture.

Spatial representations evident in gesture sometimes overlap with—but other times depart

from—those evident in language (for discussion, see Cooperrider & Goldin‐Meadow, 2016). A

major reason for this departure is that gesture has representational obligations that language

does not. When we describe a recently observed event in words (e.g., “The child ran inside”), we

do not have to specify a perspective on the event or a frame of reference (FoR). But when we

gesture about the same event, there’s no getting around it: gestures unfold in space and thus

specify a perspective and an FoR obligatorily. This fact raises a number of tantalizing questions

for researchers interested in cross‐linguistic diversity. For example, will speakers show a reliable

FoR preference in gesture even when not using overt FoR language? One of my fieldwork

projects addresses this issue by examining the gestures that bilingual speakers of Juchitán

Zapotec and Spanish produce when describing simple, table‐top motion events. My collaborators

and I ask two questions: (1) Do such gestures reliably reflect a preferred FoR, even though they

are rarely accompanied by FoR language? (2) If so, what factors predict the FoR used? The

project also examines how the same speakers perform on a variant of the “Animals in a Row”

task (Marghetis, McComsey, & Cooperrider, 2014), allowing us to compare the FoR preferences

revealed by motion gestures with those revealed by a classic memory paradigm. Ultimately, this

work will help us better understand the ecology of spatial language in Juchitán. Is it spatial

language per se that drives people to gesture and reason in different ways? Or are differences in

gesture, memory, and language all driven by some other factor?

S


Spatial concepts for abstract thinking

Another dimly understood concomitant of spatial language is abstract thinking. The canonical use

of spatial language is, of course, to communicate about the concrete—for example, to describe

the location and movement of people and objects. But spatial concepts are also pressed into

service much more broadly, for construing non‐canonical spatial settings and for structuring

purely abstract concepts. But which spatial schemas do we call on for these more abstract

purposes, and why? An emerging hypothesis is that communities will favor the same concepts for

abstract thinking that they favor for concrete spatial reference. Exploring this possibility is one of

the primary aims of my other fieldwork project, in the Yupno valley of Papua New Guinea. In

Yupno, as in many languages in the interior of New Guinea, uphill‐downhill contrasts run

throughout the core grammar, cropping up in demonstratives, in basic motion verbs, and

beyond. Using a variant of the “Man and Tree” task, my collaborators and I have investigated

how uphill‐downhill concepts are deployed for describing location and orientation outdoors,

amid the rugged terrain of the valley. But we have also used the same methods to investigate

how the system is used indoors, within the flat‐floored Yupno house. Though the houses lack

slopes of their own or views of the slopes outside, speakers nonetheless press uphill‐downhill

contrasts into service, projecting a set of conceptual “slopes” onto the house’s interior

(Cooperrider, Slotta, & Núñez, 2016). This is an example of a perhaps widespread phenomenon in

which speakers parse objects and settings by projecting their most familiar spatial schemas onto

them.

Another line of my work has focused on how people use spatial concepts to think about time,

a famously ethereal and ineffable dimension of experience. English speakers conceptualize the

past, present, and future through egocentric spatial schemas (Walker & Cooperrider, 2016), but

these patterns are not universal. Yupno speakers draw instead on their trusty uphill‐downhill

concepts, with the future conceptualized as uphill and the past as downhill (Núñez, Cooperrider,

Doan, & Wassmann, 2012). In both English and Yupno, these patterns of temporal thinking can

be glimpsed in certain linguistics expressions—but, interestingly, they are seen especially vividly

in gesture. Here again, of course, cause and consequence are tricky to sort out. Do people around

the world think about time a certain way because of how they think about space? Or do both

habits of thought spring from some other, hidden source?

Outlook

Some of the biggest questions that remain for the cross‐linguistic study of spatial language are

not about language per se. Rather, they are about why we see the diversity that we do and about

how this diversity shapes the life and mind of speakers. As we puzzle over these questions in the

coming years, a promising tack is to focus on ecologies of spatial language, on how spatial

reference is interwoven with communication, culture, and cognition. Like other researchers in

the field, I suspect that spatial language is a powerful player in these ecologies—a keystone

species, if you like. But this remains to be demonstrated.


References Cooperrider, K., & Goldin‐Meadow, S. (2016, to appear). Gesture, language, and cognition. In B. Dancygier (Ed.)

CambridgeHandbook of Cognitive Linguistics. Cambridge: New York. [available on my website]

Cooperrider, K., Slotta, J., & Núñez, R. (2016). Uphill and downhill in a flat world: The conceptual topography of

the Yupno house. Cognitive Science. DOI: 10.1111/cogs.12357

Marghetis, T., McComsey, M., & Cooperrider, K. (2014). Spatial reasoning in bilingual Mexico: Delimiting the

influence of language. In P. Bello, M. Guarini, M. McShane, & B. Scassellati (Eds.), Proceedings of the 36th

Annual Meeting of the Cognitive Science Society (pp. 940–945). Austin, TX: Cognitive Science Society.

Núñez, R., Cooperrider, K., Doan, D., & Wassmann, J. (2012). Contours of time: Topographic construals of past,

present, and future in an indigenous group of Papua New Guinea. Cognition 124 (1): 25–35.

Walker, E. & Cooperrider, K. (2016). The continuity of metaphor: Evidence from temporal gestures. Cognitive

Science 40: 481–495.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Feist–25

Finding the Universal in and around Variation MICHELE I. FEIST

Department of English University of Louisiana at Lafayette


cross a sample of languages, in and on (and their equivalents) are amongst the earliest words that a child will begin to produce (Johnston & Slobin 1979; Slobin 1982). This early

use suggests the salience of containment and support relations as a potential cognitive universal (cf., Piaget & Inhelder 1967), a suggestion that may be further strengthened amongst Western researchers by the natural coherence that they, like other native speakers, tend to attribute to the spatial concepts encoded in their languages. However, a closer look at the semantics of spatial relational terms across languages raises two problems with this interpretation. First, containment and support relations, as encoded in the English words in and on, do not emerge as coherent named concepts across a broader sample of languages (Bowerman & Choi 2001; Levinson & Meira 2003). Second, there is wide variation in the range of spatial relational situations that may be described by translation equivalents of in and on (Bowerman & Choi 2001; Feist 2008, 2013; Gentner & Bowerman 2009; Zhang, Segalowitz, & Gatbonton 2011; inter alia), suggesting that what “counts as” containment or support differs substantially from language to language.

The existence of languages that do not encode a distinction between containment and support in their spatial relational terms has long been noted. As a case in point, the Korean distinction between tight and loose fit crosscuts the containment/support distinction, categorizing both a Lego on a Lego stack and a cassette in a cassette case as examples of tight fit (Bowerman & Choi 2001). The coherence of support as a universally salient concept has been similarly contradicted by cross‐linguistic evidence. Across their sample of nine unrelated languages, Levinson and Meira (2003) found evidence for a small set of universal conceptual “attractors” – but support relations were split across two of the attractors, with relations involving small, movable figure objects supported by relatively low ground objects (such as a cup on a table) clustering with a different set of scenes from the scenes clustered with relations involving larger, more elevated figures (such as a tree on top of a hill). These results suggest that there is not a unitary support concept that is cross‐linguistically valid.

Finally, close examination of the uses of spatial language in English and Mandarin reveals that the details of the concepts of containment and support themselves may likewise vary across cultures (Feist & Zhang 2016). Like English, Mandarin encodes a contrast between terms encoding containment and terms encoding support. However, the uses of this set of terms differ from the uses of related English spatial terms, particularly with respect to the categorization of part‐whole relations and of figures which are partially embedded in a ground (Zhang et al. 2011). Indeed, looking at descriptions of 71 pictures in the two languages for which there was high within‐language agreement on the applicable spatial term, Zhang and her colleagues (2011)

A


found that 20% were categorized differently by the two languages, suggesting that the contrasts encoded in the languages, while related, are drawing upon different conceptual information regarding the nature of containment and support.

Along with the striking variation that has been noted in the meanings of spatial relational terms, there is compelling evidence that languages may overlay lexicalized distinctions upon a universal conceptual space. In addition to the universal “attractors” and the attendant conceptual space noted by Levinson and Meira (2003), Feist (2008) found that spatial descriptions across 24 languages could be fit to a two‐dimensional similarity space structured by the ground’s ability to control the location of the figure and by the relative vertical positions of the two objects, suggesting that languages may encode distinctions along dimensions that are universally salient. Taken together, these findings point to a set of broad universal constraints on the conceptual distinctions encoded in spatial language.

Comparison of the evidence of cross‐linguistic variation and the suggestion of conceptual universals highlights a gap in our knowledge about spatial reference. Whereas fine‐grained studies of the semantics of individual spatial relational terms has revealed much about cross‐linguistic variation in the encoding of meaning, more coarse‐grained studies of the conceptual space suggest constraints on that variation. To truly understand the nature of both the variation and the constraints, it is important to understand how the constraints function within the spatial reference systems of individual languages, hence bringing together the strengths and insights from both lines of research.

One universal of human experience is the need to note, remember, and communicate about the locations of objects in the environment. One of the most compelling observations about spatial language is that there is a stark contrast between the native speaker’s intuition that the categories encoded in their language are conceptually basic (and, hence, potentially universal), and the prodigious cross‐linguistic variation in the mapping of words to spatial situations. One of the most interesting challenges for us is to understand why.

References Bowerman, M., & Choi, S. (2001). Shaping meanings for languages: Universal and language‐specific in the acquisition of spatial semantic categories. In M. Bowerman & S. C. Levinson (Eds.), Language acquisition and conceptual development. Cambridge UK: Cambridge University Press.

Feist, M. I. (2008). Space between languages. Cognitive Science 32(7): 1177–1199.

Feist, M. I. (2013). Experimental lexical semantics at the crossroads between languages. In A. Rojo & I. Ibarretxe‐Antuñano (Eds.), Cognitive linguistics and translation: Advances in some theoretical models and applications. Berlin: de Gruyter Mouton.

Feist, M. I., & Zhang, Y. (2016). The shape of space in English and Mandarin. Poster presented at The Future of Language Sciences, Northwestern University, September 2016.

Gentner, D., & Bowerman, M. (2009). Why some spatial semantic categories are harder to learn than others: The typological prevalence hypothesis. In J. Guo, E. Lieven, N. Budwig, S. Ervin‐Tripp, K. Nakamura, & S. Özçaliskan (Eds.), Crosslinguistic approaches to the psychology of language: Research in the tradition of Dan Isaac Slobin. New York: Psychology Press.


Johnston, J. R., & Slobin, D. I. (1979). The development of locative expressions in English, Italian, Serbo‐Croatian and Turkish. Journal of Child Language 6: 529–545.

Levinson, S. C., & Meira, S. (2003). “Natural concepts” in the spatial topological domain—adpositional meanings in crosslinguistic perspective: An exercise in semantic typology. Language 79(3): 485–516.

Piaget, J., & Inhelder, B. (1967). The child’s conception of space. New York: Norton.

Slobin, D. I. (1982). Universal and particular in the acquisition of language. In E. Wanner & L. R. Gleitman (Eds.), Language acquisition: The state of the art. Cambridge UK: Cambridge University Press.

Zhang, Y., Segalowitz, N., & Gatbonton, E. (2011). Topological spatial representation across and within languages: IN and ON in Mandarin Chinese and English. Mental Lexicon 6(3): 414–445.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Frajzyngier–28

Cross‐linguistic Encoding of Locative Predications ZYGMUNT FRAJZYNGIER Department of Linguistics University of Colorado


he basic hypothesis of the present study is that the form of an expression describing an event in relationship to space depends on (1) whether the grammatical system of a given

language has a functional domain, conventionally called “altri‐locative point of view” that encodes movement or position in relationship to a place other than the deictic center, which is often, but not necessarily, the place of speech; and (2) whether the grammatical system has a functional domain that encodes movement in relationship to the deictic center, conventionally called “directionality from the point of view of deictic center,” “directionality” for the sake of brevity.

Some languages code only altri‐locative point of view (1), some languages code only directionality (2), some languages code both (1) and (2), and there are languages that do not code either. The description of a physically identical event, such as “movement from X to Y,” thus has different forms across languages, depending on whether the language codes locative predication, directional predication, both, or neither. The present study demonstrates the importance of the existence of functional domain only with respect to altri‐locative point of view.

Languages differ with respect to how they represent “locative events,” i.e., events involving movement to or from a place and events or states occurring at a place. The differences observed raise the following questions:

Question 1: Why, within the same language, do some clauses coding locative events involve prepositions while others do not?

Question 2: Why, in expressing the same locative event, do some languages use prepositions while others do not, even if there are prepositions in the language?

Question 3: Why do some languages have the distinct lexical category “locative predicator” while others do not?

Question 4: Why do the syntactic properties of verbs that refer to the same activities, such as equivalents of “come,” “go,” “run,” “swim,” “jump,” differ significantly across languages?

Question 5. Why do prepositions involved in coding the same locative events differ significantly, across languages, in their semantic and syntactic properties?

The aim of the present study is to explain differences across languages that have the same lexical categories, particularly prepositions and postpositions, and that have lexical items referring to the same types of events or states. The foundation of the explanation is the discovery, described in Frajzyngier with Shay (2016) and Frajzyngier (in press), that some languages have the function “locative predication” encoded in their semantic structure while others do not, and that, in a language that has encoded the function of locative predication, the formal means used to code this predication are distinct from the formal means used to code all

T


other predications in the language. As a consequence of encoding the function of locative predication, some verbs and nouns in the language are inherently locative and others are not. When the verb is inherently locative, no additional means are used to mark it as the locative predicate in a locative predication (all examples here are from Mina (Frajzyngier and Johnston with Edwards 2005), but similar phenomena have been observed in other languages; see Frajzyngier (in press)):

(1) yá í‐bə ndə tətə bíŋ call PL‐ASSC go 3PL.POSS room

“They went into their room.” (the verb ndə is inherently locative)

When the predicate is not inherently locative, as is the verb yà “call,” the verb phrase must be followed by the predicator á to mark the locative predication:

(2) nd‐á yà ngùl ngən á bìŋ go‐GO call husband 3SG PRED room

“And [she] called her husband into the room.” (nd‐á is a sequential marker and á is the locative predicator, not a preposition)

Note that the translations of (1) and (2) use the preposition “into,” evidence that the nature of the predicate plays no role in the coding of the locative expression in English.

If the complement of the locative predicate is inherently locative, i.e. if it refers to referents such as “house,” “room,” or “village,” it does not require a locative preposition in the locative predication (see examples (1) and (2)). If the complement is not inherently locative, it requires the locative preposition nə:

(3) hídì wà mə‐nd‐á‐kù dèɓ nə kítà man DEM REL‐beat‐OBJ‐1SG lead PREP justice (Fula) “It was this person who hit me. Take him to be judged.” (lit. “take him to justice”)

Note that the English version requires a locative preposition regardless of whether or not the locative complement is inherently locative, as illustrated in the translations of examples (1–3).

The preposition nə has only a locative function; it does not have a directional component, as evidenced by the fact that it can be used in clauses involving movement to a place (ex. (3)), movement from a place (ex. (4)), and in clauses involving the presence of an entity or the occurrence of an event at a place (ex. (5), where the place is represented by an inherently non‐locative complement. Note that example (5) also contains the locative predicator á, since the main predicate of the clause, ɗáhà “exist,” is inherently non‐locative. This example also shows that à is not a preposition:

(4) séy ábə nd‐á ngəŋ nə yəm zá

so ASSC go‐GO 3SG PREP water FACT “Then, he came out of the water.”


(5) séy hìdì mə sə tápá ɗáhà á

so man REL drink tobacco exist PRED nə məŋ màcíŋ

PREP ANAPH DEM“So there is a smoker among them.”

The existence of locative predications, as distinct from other types of predication, has the following implications for the questions posed above:

(1) Some nouns and verbs are inherently locative and others are not. No such distinction exists in languages that have not encoded locative predication (Question 4).

(2) Within an individual language and across languages, prepositions are not used when the complement is inherently locative (Questions 1 and 2).

(3) When the predicate is not inherently locative, a locative predicator is used in locative predications (Question 3).

(4) There exist prepositions that have only the locative function, i.e. they do not code spatial relationships with respect to the complement. Spatial relations are coded by another set of markers (not illustrated for lack of space). In languages without locative predication, the locative function is fused with spatial relationships, as in English “in,” “out,” “from,” “to” (Question 5).

References: Frajzyngier, Zygmunt (in press). Coding locative predication in Chadic. In Alessandro Mengozzi and Mauro Tosco (eds.), Afroasiatic: Data and Perspectives. Amsterdam: J. Benjamins.

Frajzyngier, Zygmunt, with Erin Shay. 2016. The role of functions in syntax: a unified approach to language theory, description, and typology. Benjamins: Amsterdam.

Frajzyngier, Zygmunt, and Eric Johnston with Adrian Edwards. 2005. A Grammar of Mina. Berlin/New York: Mouton de Gruyter.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Gudde—31

Spatial Demonstratives in English and Japanese: Universal or Variation?

HARMEN B. GUDDE University of East Anglia


KENNY R. COVENTRY University of East Anglia

patial language is crucial to almost every aspect of our lives, yet languages vary considerably in how they carve up space (Kemmerer & Tranel, 2000; Levinson, 2004). We used

demonstratives as a vehicle to explore the relation between languages and spatial cognition. Although spatial demonstratives (this, that) are a small class of referential expressions, a growing body of research shows the important role they play in language. Demonstratives are present in every language and are among the most frequent words in languages (Diessel, 1999, 2006, 2014; Heine & Kutteva, 2002). Hitherto, little empirical research has been done to experimentally determine the function of spatial demonstratives. Early research found empirical evidence for a proximal/distal contrasting function of demonstratives (Coventry, Valdés, Castillo, & Guijarro‐Fuentes, 2008), suggesting that this is used for referents in peri‐personal (near) space and that for referents in extra‐personal (far) space (cf. Clark & Sengul, 1978; Diessel, 2006; Talmy, 1983), in contrast to Peeters, Hagoort, and Ozyürek (2014). Building on this body of work, current research at the University of East Anglia is exploring how people use demonstratives, and whether they affect spatial memory. Specifically, using a memory game procedure, participants are asked to name objects, placed at various distances from them, using a demonstrative, for example “this/that black cross.” This allows us to test whether parameters that are encoded by demonstratives in other languages (e.g., distance, ownership (encoded in Supyire), visibility (West Greenlandic, Sinhala), familiarity (Yoruba), affect English demonstrative use (cf. Chandralal, 2010; Coventry, Griffiths, & Hamilton, 2014; Diessel, 1999). An adaptation of this procedure, can test the influence of object knowledge on memory for object location by analysing the memory error (the difference between the actual and memorized location).

Results showed that object knowledge affects demonstrative use and similarly influences memory for object location—even though the contrasts are not explicitly encoded in English. When participants owned, saw (during encoding), or knew an object, they were more likely to refer to the object with this than if they did not. Objects were also remembered to be closer by when they were owned, seen (during encoding), or known by the participant. In other words, referents that were preferentially referred to with this were remembered to be closer to the participant, relative to that. As such, Coventry et al. posited that memory for object location is a concatenation of the actual and the expected location of an object (the Expectation model), consistent with theories of predictive coding (Bar, 2009; Friston, 2003).

More recently, we have extended the limits of the Expectation model by investigating whether the mere use of demonstratives affects spatial memory (Gudde, Coventry, & Engelhardt, 2016). In this study, participants read out instructions for object placement (e.g., “Place

S


this/that/the [object] on the [location]”), followed by a spatial memory trial. By analysing the memory error, this study was able to tease apart different models predicting an influence of language on memory for object location. The Expectation model suggests that language elicits a prediction about the object location. The model therefore predicts a main effect of language on spatial memory, in which objects placed with that are misremembered to be further away than this, irrespective of the distance from the participant. In contrast, the Congruence model, based on the embodied cognition framework (Barsalou, 2008), assumes that an effect would be driven by an (in)congruence of language and space. A plethora of studies showed that when language is congruent with a spatial situation, participants’ responses are for example faster or more accurate (cf. Bonfiglioli, Finocchiaro, Gesierich, Rositani, & Vescovi, 2009; Stevens & Zhang, 2013). The congruence model predicts a similar interaction between language and distance in memory (Hommel, Musseler, Aschersleben, & Prinz, 2001). That is, congruent trials in which objects are placed close by with this or out of reach with that, should be remembered more accurately than incongruent trials (this for objects out of reach, that for objects within reach). In three experiments, results showed a main effect of language, but no interaction, supporting the Expectation model, and we found evidence that the effects were not driven by a difference in attention allocation.

However, in order to test whether there is a universal demonstrative system, other languages than English need to be tested. The most recent study tested Japanese vs. English (Gudde & Coventry, in preparation). Results showed that Japanese demonstratives encode distance from a speaker and the position of a hearer, and an effect of position was found in English as well. Furthermore, gender seemed to influence the weight of the parameters; men (both in Japanese and English) were more strongly influenced by an interlocutors’ position, women by distance. The fact that English demonstrative use is affected by position supports the notion that demonstrative systems are reliant on a universal set of underlying non‐linguistic parameters, even though these parameters are not explicitly coded in all languages. However, explicit encoding, for example of position of an interlocutor, could lead to a slightly different weighting across languages as a function of the parameters that are explicit.

Bar, M. (2009). The proactive brain: memory for predictions. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 364(1521), 1235–43. doi:10.1098/rstb.2008.0310

Barsalou, L. W. (2008). Grounded cognition. Annual Review of Psychology, 59, 617–645. Bonfiglioli, C., Finocchiaro, C., Gesierich, B., Rositani, F., & Vescovi, M. (2009). A kinematic approach to the conceptual representations of this and that. Cognition, 111(2), 270–4. doi:10.1016/j.cognition.2009.01.006

Chandralal, D. (2010). Sinhala (15th ed.). John Benjamins Publishing Company. Clark, E. V., & Sengul, C. J. (1978). Strategies in the acquisition of deixis. Journal of Child Language, 5(03), 457–475.

Coventry, K. R., Griffiths, D., & Hamilton, C. J. (2014). Spatial demonstratives and perceptual space: describing and remembering object location. Cognitive Psychology, 69, 46–70. doi:10.1016/j.cogpsych.2013.12.001

Coventry, K. R., Valdés, B., Castillo, A., & Guijarro‐Fuentes, P. (2008). Language within your reach: near‐far perceptual space and spatial demonstratives. Cognition, 108(3), 889–895. doi:10.1016/j.cognition.2008.06.01

Diessel, H. (1999). Demonstratives: form, function and grammaticalization. Amsterdam/Philadephia: John Benjamins Publishing Company.


Diessel, H. (2006). Demonstratives, joint attention, and the emergence of grammar. Cognitive Linguistics, 17(4), 463–489.

Diessel, H. (2014). Demonstratives, Frames of Reference, and Semantic Universals of Space. Language and Linguistics Compass, 8(3), 116–132. doi:10.1111/lnc3.12066

Friston, K. (2003). Learning and inference in the brain. Neural Networks : The Official Journal of the International Neural Network Society, 16(9), 1325–52. doi:10.1016/j.neunet.2003.06.005

Gudde, H. B., Coventry, K. R., & Engelhardt, P. E. (2016). Language and memory for object location. Cognition, 153, 99–107. Retrieved from http://www.sciencedirect.com/science/article/pii/S0010027716301044

Heine, B., & Kutteva, T. (2002). World Lexicon of grammaticalization. Cambridge, UK: Cambridge University Press.

Hommel, B., Musseler, J., Aschersleben, G., & Prinz, W. (2001). The theory of event‐coding (TEC): A framework for perception and action planning. Behavioural and Brain Sciences, 24, 849–937. doi:10.1017/S0140525X01000103

Kemmerer, D., & Tranel, D. (2000). A double dissociation between linguistic and perceptual representations of spatial relationships. Cognitive Neuropsychology, 17(5), 393–414.

Levinson, S. C. (2004). Space in Language and Cognition: Explorations in Cognitive Diversity. Cambridge, UK: Cambridge University Press.

Peeters, D., Hagoort, P., & Ozyürek, A. (2014). Electrophysiological evidence for the role of shared space in online comprehension of spatial demonstratives. Cognition, 136, 64–84. doi:10.1016/j.cognition.2014.10.010

Stevens, J., & Zhang, Y. (2013). Relative distance and gaze in the use of entity‐referring spatial demonstratives: An event‐related potential study. Journal of Neurolinguistics, 26(1), 31–45. doi:10.1016/j.jneuroling.2012.02.005

Talmy, L. (1983). How Langauge Structures Space. In H. L. Pick Jr. & L. Acredolo (Eds.), Spatial Orientation (pp. 225–282). New York and London: Plenum Press.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Hirtle–34

The Use of Informational References in Spatial Descriptions

STEPHEN C. HIRTLE

School of Informational Sciences University of Pittsburgh Email: [email protected]

ne of the persistent difficulties in generating spatial descriptions comes from desire to create compact, efficient directions that are accurate, complete and accessible to the navigator. This

can be particularly difficult when there are different cultural and linguistic norms, even within the same general geographical area. In previous research, we examined this issue by documenting cases where published directions are flagged as being “tricky” due to inherent navigational difficulties[1]. The potential confusions could be a result of a number of reasons, including an unusual geometry of the space, inconsistent labels assigned to roads and surrounding landmarks, or simply the failure of meeting certain expectations. Consideration of this kind of expectancy is a major factor in the engineering standards for location and style of warning signs, while at the same time expectation is dependent on the navigator’s culture and background[1].

Of further note, spatial expectations often build on local knowledge of the design options. For example, directions in northern New Jersey might not flag taking a left turn from the right lane using a jug handle as unusual, given its common use for connecting roads. In other areas, where such junctures are relatively rare, this kind of intersection would most likely be a candidate for being tagged as an unusual (or tricky) connection. Thus, the role of expectation varies greatly from region to region and country to country.

In the same spirit, Firth[2] used the term configurational grasp mapping for the process of articulating how the structure of a road network works at a macro level to both provide and restrict access to a given area. Tomko, Winter, and Claramunt[3] took a similar approach to defining what they call an experiential hierarchy of streets. The resulting representations would suggest that further investigation of route directions aimed at systematic differentiation of conceptual aspects is needed. This might include establishing a consistent set of cognitive elements to use as building blocks in route directions, involving start and end points, route segments, action and movement descriptions, reorientations, landmarks, regions and areas, and distances. Further research has highlighted different levels of granularity and the impact of relevance on the ways in which these elements are chosen and represented in a description. It is argued that these various influences can be comprehensively captured by understanding the activity at hand[4].

Thus appropriate and suitable route directions are not created independent of the navigational task. In this spirit, Hirtle, Timpf and Tenbrink[4], argued that from an ontological point of view, activities and tasks produce partitions of reality. For example, in changing mode of transportation (bus, tram, subway), the rider needs additional information, which ideally is in the form of signage at the exchange points. Moreover, it is often easier to change to a more global mode (e.g., bus to subway) than to a more local mode (e.g, subway to bus), in part due to the number of options at the transfer point. Even in straight‐forward travelling situations, there are often what Hirtle et al[1] called

O

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Hirtle–35

the persistent endpoint problems. That is, while the general location of the final destination may be clear, the appropriate office, parking lot, entrance way, etc., may not be well‐marked.

Together, the vocabulary, amount of detail, presentation mode, and completeness are dependent as much on human and cultural factors, as they are on the geographic features of the space. When an geographic information broker adds the ability to provide spatial location in terms of verbal, visual and spatial information, one is faced with the potential of information overload[5].

Understanding these stated limitations can lead to the development of more user‐friendly directions, such as “follow the winding road until you get to the city center,” instead of detailed directions that specify every bend and turn. Likewise identifying the notable dimension, such as visual, verbal, or spatial uniqueness, would allow for compact directions that are fined‐tuned along the individual differences (both linguistic and cultural) that exist among wayfinders.

[1] Hirtle, S. C., Richter, K.‐F., Srinivas, S., & Firth, R. (2010). This is the tricky part: When directions become difficult. Journal of Spatial Information Science, 1(1), 53–73.

[2] Firth, R. Configurational‐grasp mapping. In You‐Are‐Here Maps: Creating a Sense of Place Through Map‐like

Representations (Freiburg, Germany, 2008).

[3] Tomko, M., Winter, S., & Claramunt, C. Experiential hierarchies of streets. Computers, Environment and

Urban Systems 32, 1 (2008), 41–52.

[4] Hirtle, S. C., Timpf, S., & Tenbrink, T. (2011). The effect of activity on relevance and granularity for navigation. In International Conference on Spatial Information Theory (pp. 73–89). Springer: Berlin Heidelberg.

[5] Ellard, C. (2009). You are here: Why we can find our way to the moon, but get lost in the mall. Anchor.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Hoffmann—36

Universals and Variation in Spatial Referencing

DOROTHEA HOFFMANNDepartment of Linguistics University of Chicago


Frames of Reference

ifferent aspects of Frames of Reference (FoR) have been analyzed in detail since (at least)

the early 1990s. Researchers have studied cross‐linguistic variety (Levinson, 1996; Levinson

and Wilkins, 2006; Pederson et al., 1998), given detailed accounts of individual languages

(François, 2003; Haviland, 1993; Hoffmann, 2011; Schultze‐Berndt, 2006), considered the impact

of landscape and cognition on FoRs (Bohnemeyer and O’Meara, 2012; Danziger, 2010; Levinson,

2003, 2008; Palmer, 2015), explored usage patterns (Hoffmann, sub), and provided regional

overviews (Bohnemeyer, 2013; François, 2004, 2015).

I am interested in how languages express absolute Frames of Reference focusing particularly

on the Australian continent. How do genetic a liation, location and landscape features, climate

zone, typological area, and cultural overlap influence choice and usage of these types of spatial

reference? What happens at boundary areas and which factors are the most influential?

Absolute Frames of Reference in Australia

Absolute FoR requires fixed bearings that are instantly available to all members of the community

(Levinson and Wilkins, 2006, 21). There is a wide variety of absolute systems in Australian

languages, including compass‐ (e.g., Warlpiri (Laughren, 1978)), wind‐ (Kala Lagaw Ya (Stirling,

2011, 182; Bani, 2001)), river drainage‐ (Dyirbal (Dixon, 1972)), ocean‐ (Iwaidja (EdmondsWathen,

2011); Edmonds‐Wathen, 2012, 142–143), and tide‐based (Bardi (Bowern, 2012, 30). Recently,

Blythe et al. (2016) described the Murrinh‐Patha system where no absolute directionals are

present, but speakers instead utilize named landmarks, demonstratives and pointing in spatial

descriptions. Finally, for some languages, a number of systems overlap, for example in Gurindji

sun‐ and river‐drainage‐based systems as in (1) (Meakins, 2011) and in MalakMalak wind‐, sun‐,

and riverbank‐based systems (2) (Hoffmann, 2016). (1) Gurindji: River drainage and sun‐based system

ngu‐rnalu ya‐ni kanimparra, kaarnimpa nyawa. Nangala‐lu paraj pu‐nya cat‐1pl.excl go‐pst downstream east.along this subsect‐erg find pierce‐pst ngu‐ø‐ø ngarlu.

cat‐3sg.sbj‐3sg.obj honey

“We came downstream along the eastern side here. (Then) Nangala found some bush honey.”1

(VD: FM07a021: Narrative) (Meakins, 2011, 49)

(2) MalakMalak: Wind‐ and sun‐based system

yinya nende dangid‐en pud wu‐rrunguny, miri‐nen

man thing/person southeasterly‐dir chest 3pl‐go/be.ipfv, sun‐dir payi‐ga‐ma change.location‐come‐cont

1 Abbreviations used are the following: cat = catalyst, cont = continuous, dir = directional, erg = ergative, excl = exclusive,

ipfv = imperfective, obj = object, pst = past, pl = plural, sg = singular, sbj = subject

D


“the men are facing towards the southeasterly wind direction, towards the east (where the sun comes up)” (Hoffmann fieldwork, 2012)

Furthermore, there is variation within absolute systems. For example, wind‐based systems

can either be seasonally dependent or fixed. In Kala Lagaw Ya there is evidence for a system

bound to winds at specific times of the year, even though the younger generation is now shifting

towards a fixed system (Stirling, 2011, 186–87; Bani, 2001, 477). On the other hand, the system

employed by MalakMalak and Matngele winds from the (distant) sea and inland blowing at

distinct times of the year is used year‐round with fixed reference‐points (Hoffmann, 2016).

Additionally, such wind‐based systems are common in small island communities, especially in

Oceanic (François, 2003, 2004, 2015) and Austronesian languages (Adelaar, 1997), but also in

Polynesian (Svorou, 1994), some African (Brauner, 1998; Mietzner and Pasch, 2007) and a Tibeto‐

Burman language (Post, 2011). However, their existence in mainland Australia has been largely

overlooked with the exception of Hoffmann (2016) and Nash (2013).

Finally, usage patterns for these absolute directionals have only been tentatively described.

Edmonds‐Wathen (2012, 90) determines that non‐Pama‐Nyungan languages, such as Jaminjung

and Warrwa use the absolute frame in small scale space only when other resources are not

available, while Pama‐Nyungan languages make widespread use of absolute systems in large‐ and

small scale descriptions. Furthermore, in some languages with a number of absolute systems

there is systematic usage variation based on different contexts. For example, in MalakMalak the

wind‐based system is limited to motion and orientation settings (in Terrill and Burenhult (2008)’s

sense), the sun‐based system can additionally be used in deictic FoR settings, and the riverbank

system is most flexible in allowing for deictic and non‐deictic FoR settings in addition to

orientation and motion descriptions (Hoffmann, sub).

Questions to address

In previous studies of spatial language and absolute directionals in Australia, much attention has

been paid to compass‐directions and their usage in either only large‐ or both small‐ and large‐

scale settings. Other systems have been described for individual languages, but so far no

comparative study has been conducted taking into account what possible influences on absolute

systems across linguistic and cultural areas exist. I am currently preparing a manuscript on this

subject.

Another area of particular interest are absolute systems employed by newly emerging

languages such as the varieties of Kriol, an English‐lexified creole spoken across indigenous

language boundaries all across northern Australia, or mixed languages such as Gurindji Kriol and

Light Warlpiri (Meakins, 2011; OShannessy and Meakins, 2016). Are the systems used similar to

those of their substrate or superstrate languages? To what extend are they dependent on extra‐

linguistic factors? What cognitive and linguistic shifts are taking place?

Additionally, are there universal uses of absolute directionals across the Australian continent or

can systematic variation be found with regards to genetic and/or typological variation? What

roles do landscape and geographic conditions, climate and seasonal patterns play (Palmer,

2015)? Finally, what wider implications for cross‐linguistic generalization are evident from


observing spatial frames of reference in a typologically and culturally relatively homogeneous,

but genetically diverse area such as Australia? References

Adelaar, K. A. (1997). An exploration of directional systems in west Indonesia and Madagascar. In Senft, G., editor, Referring to Space: Studies in Austronesian and Papuan languages, pages 53–81. Clarendon Press, Oxford.

Bani, E. (2001). The morphodirectional sphere. In Forty Years on: Ken Hale and Australian Languages, pages 477–480. Pacific Linguistics, Canberra.

Blythe, J., Mardigan, K. C., Perdjert, M. E., and Stoakes, H. (2016). Pointing out directions in Murrinhpatha. Open Linguistics 2: 132–159. doi:10.1515/opli‐2016‐0007.

Bohnemeyer, J. (2013). Frames of reference in language, culture, and cognition: The Mesoamerican evidence. Annual Meeting of the Berkeley Linguistic Society.

Bohnemeyer, J. and O’Meara, C. (2012). Vectors and frames of reference: Evidence from Seri and Yucatec. In Filipovi´c, L. and Jaszczolt, K. M., editors, Space and Time across Languages and Cultures: Language, Culture, and Cognition, pages 217–249. John Benjamins, Amsterdam.

Bowern, C. (2012). A grammar of Bardi, volume 57. Walter de Gruyter.

Brauner, S. (1998). Directions/Spatial orientations in African languages: Further cases. In Zima, P., editor, Language and Location in Space and Time, pages 66–78. Institute of Advanced Studies, Prague.

Danziger, E. (2010). Deixis, gesture, and cognition in spatial frame of reference typology. Studies in language 34(1): 167–185.

Dixon, R. M. (1972). The Dyirbal language of north Queensland, volume 9. CUP Archive.

Edmonds‐Wathen, C. (2011). What comes before?: Understanding spatial reference in Iwaidja. In ICMI Study 21 Conference: Mathematics and language diversity, pages 89–97. ICMI.

Edmonds‐Wathen, C. (2012). Frame of reference in Iwaidja: Towards a culturally responsive early years mathematics program. Ph.D., RMIT University.

Fran¸cois, A. (2003). Of men, hills, and winds: Space directionals in Mwotlap. Oceanic Linguistics, pages 407–437. Fran¸cois, A. (2004). Reconstructing the geocentric system of Proto‐Oceanic. Oceanic linguistics, pages 1–31.

Fran¸cois, A. (2015). The ins and outs of ‘up’ and ‘down’: Disentangling the nine geocentric space systems of Torres and Banks languages. The Languages of Vanuatu: Unity and Diversity 5: 137–195.

Haviland, J. B. (1993). Anchoring, iconicity, and orientation in Guugu Yimithirr pointing gestures. Journal of Linguistic Anthropology 3(1): 3–45.

Hoffmann, D. (2011). Descriptions of Motion and Travel in Jaminjung and Kriol. Ph.D., University of Manchester, Manchester.

Hoffmann, D. (2016). Mapping worlds: Frames of Reference in MalakMalak. In Proceedings to the 39th Meeting of the Berkeley Linguistic Society, pages 380–395, Berkeley. University of California at Berkeley.

Hoffmann, D. (sub). Usage Patterns of Spatial Frames of Reference and Orientation: Evidence from three Australian Languages. submitted.

Laughren, M. N. (1978). Directional Terminology in Warlpiri (a Central Australian Language). Working Papers in Language and Linguistics 8: 1–16.

Levinson, S. C. (1996). Frames of Reference and Molyneux’s question: Crosslinguistic evidence. In Language and Space, pages 109–169. MIT Press, Cambridge, MA.

Levinson, S. C. (2003). Space in language and cognition: Explorations in cognitive diversity, volume 5. Cambridge University Press, Cambridge.

Levinson, S. C. (2008). Landscape, seascape and the ontology of places on Rossel Island, Papua New Guinea. Language Sciences 30(2): 256–290.


Levinson, S. C. and Wilkins, D. P. (2006). Grammars of Space: Explorations in cognitive diversity, volume 6. Cambridge University Press, Cambridge.

Meakins, F. (2011). Spaced out: Intergenerational changes in the expression of spatial relations by Gurindji people. Australian Journal of Linguistics 31(1): 43–77.

Mietzner, A. and Pasch, H. (2007). Expressions of Cardinal Directions in Nilotic and in Ubangian Languages. SKASE Journal of Theoretical Linguistics 4: 17–31.

Nash, D. (2013). Wind direction words in the Sydney language: A case study in semantic reconstitution. Australian Journal of Linguistics 33(1): 51–75.

OShannessy, C. and Meakins, F. (2016). Australian language contact in historical and synchronic perspective. In OShannessy, C. and Meakins, F., editors, Loss and Renewal: Australian Languages Since Colonisation, volume 13, pages 3–26. Walter de Gruyter GmbH & Co KG.

Palmer, B. (2015). Topography in language. In Busser, R. D. and LaPolla, R. J., editors, Language Structure and Environment: Social, cultural, and natural factors, volume 6, pages 179–226. John Benjamins Publishing Company, Amsterdam/Philadelphia.

Pederson, E., Danziger, E., Wilkins, D., Levinson, S., Kita, S., and Senft, G. (1998). Semantic typology and spatial conceptualization. Language 74: 557–589.

Post, M. W. (2011). Topographical deixis and the Tani languages of North‐East India. North East Indian Linguistics 3: 137–154.

Schultze‐Berndt, E. (2006). Sketch of a Jaminjung Grammar of Space. In Levinson, S. C. and Wilkins, D. P., editors, Grammars of Space: Explorations in cognitive diversity, volume 6 of Language, Culture and Cognition, pages 63–113. Cambridge University Press, Cambridge.

Stirling, L. (2011). Space, time and environment in Kala Lagaw Ya. In Indigenous language and social identity: Papers in honour of Michael Walsh, pages 179–203. Pacific Linguistics, Canberra.

Svorou, S. (1994). The Grammar of Space, volume 25. John Benjamins Publishing.

Terrill, A. and Burenhult, N. (2008). Orientation as a strategy of spatial reference. Studies in Language 32(1): 93– 136.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Kelleher—40

The Role of Perception in Situated Spatial Reference

JOHN D. KELLEHERAdapt Research Centre School of Computing

Dublin Institute of Technology, Ireland Email: [email protected]

y research is inspired by exploring the interface between language and perception, and

spatial reference in situated dialog is a natural area of study for this topic of research. My

Ph.D. [20] studied spatial reference in situated dialogue. Part of this research focused on

modelling visual attention as a mechanism to help resolve underspecified references [11, 9].

Another theme of this research, informed by the Logan and Sadler’s spatial template concept [25],

involved designing and running psycholinguistic experiments to study the spatial templates of

projective prepositions and the impact of frame of reference ambiguity on these spatial

templates [10]. Using the results of these experiments a computational model was developed that

modelled the geometric semantics of projective prepositions and that was able to accommodate

the impact of frame of reference ambiguity [18].

My Post‐doctoral research at DFKI (Saarbruecken) was on the CoSy project

(http://cognitivesystems.org/) where I worked in the area of Human‐Robot Interaction

and Dialogue Systems. Continuing the theme of spatial language in situated dialogue and

the role of perception on linguistic reference I did work on computational models of

multimodal information fusion, including the integration of spatial information as

expressed through spatial linguistic references and visual perceptual information [23, 24].

Questions relating to the grounding of spatial reference in perception were at the core of

this work and for the CoSy project a key task was to develop computational models that

would enable the qualitative information expressed in linguistic spatial references to be

grounded within the quantitative representations of the environment that a

computational agent, such as a robot, would construct information it received through its

(perceptual) sensors [1]. During this research I became more interested in exploring the

spatial semantics of topological prepositions. For me a key problem with previous

computational models of the spatial semantics of topological prepositions had been the

relatively arbitrary mechanisms used to define the maximum extent of the spatial

template of these prepositions. This led me to consider the impact of distractor objects1

defining the extent of the spatial templates of these topological prepositions. To explore

the impact of these distractor objects I (and my co‐authors) designed and ran some

psycholingustic experiments [2]. The results of these experiments indicated that distractor

objects did impact on the spatial templates of topological prepositions. Building on these

results I developed a computational model of proximity that was sensitive to distractors

1 By the term distractor objects I am denoting the set of objects that are in the perceptual frame of an agent but which are neither the landmark nor the located object in a locative expression the agent in currently resolving)

M


objects [14, 12]. In my opinion one of the most interesting aspects of my work on distractor

objects and locative expressions is that it provides another illustration of how perceptual

information affects the semantics of the spatial reference; in this case the perceptual

information related to objects that are in the context but which are not mentioned in the

spatial references. Integrating the computational models of the semantics of projective

prepositions that I developed during my Ph.D. with the computational models of

topological prepositions that I developed during my Post‐doctoral research I developed

an algorithm for generating locative expressions [13, 21]. This algorithm extended

Incremental Algorithm of Dale and Reiter2 in two ways: (1) It defined a preference order

over spatial relationships based on Cognitive Load; (2) it integrated a mechanism for

utilising visual attention to help generate linguistically underspecified but contextual clear

spatial references. Again, this work illustrated the role of a perception, in this case an

attention mechanism, on spatial reference.

In my more recent research I have explored a number of other aspects of spatial reference.

For example, semantics of topological prepositions in spatial reference [19]; the impact of the

topological prepositions on the semantics of composite spatial terms (e.g., the difference

between at the front of versus on the front of etc.) [15]; the role of analogical reasoning in spatial

reference [6]; resolving frame of reference ambiguity [28, 5]; and using corpus based analytics to

explore the functional and geometric semantics of prepositions in visually situated spatial

reference [4]. However, throughout this time I have continued to explore perceptual factors on

spatial reference. Indeed, some of my most recent work has explored the impact of perceptual

errors on spatial reference and the mechanisms people use in dialogue to repair these

communication breakdowns [29]. Other examples of recent research on perception and spatial

reference include experiments that examined the role of perspective on the semantics of

projective prepositions [16] and the preposition between [27] and, also, the role of perceptual

occlusion on the semantics of projective terms [17].

In conclusion, I believe that an interesting avenue of exploration on universals and variation

in spatial reference is to address this topic in terms of the universals in human perception and

attention and to explore how these universals impact on spatial reference across cultures and

languages.

Acknowledgements: The ADAPT Centre for Digital Content Technology is funded under the SFI Research Centres Programme (Grant 13/RC/2106) and is co‐funded under the European Regional Development Fund.

References[1] M. Brenner, N. Hawes, J.D. Kelleher, and J. Wyatt. Mediating between qualitative and quantitative

representations for task‐oriented human‐robot interaction. In Proceedings of the 20th International Conference on Artificial Intelligence (IJCAI‐07), pages 2072–2077, 2007.

2 The Incremental Algorithm is a well‐known algorithm in Natural Language Generation research that generates referring expressions.


[2] F. Costello and J.D. Kelleher. Spatial prepositions in context: The semantics of Near in the presence of distractor objects. In Proceedings of the 3rd ACL‐Sigsem Workshop on Prepositions, pages 1–8, 2006.

[3] F. Costello, J.D. Kelleher, and M. Volk, editors. Proceedings of the 4th ACL‐SIGSEM Workshop on Prepositions at ACL‐2007. Association for Computational Linguistics, 2007.

[4] S. Dobnik and J. D. Kelleher. Exploration of functional semantics of prepositions from corpora of descriptions of visual scenes. In Proc. of the Workshop on Vision and Language, pages 33–37, 2014.

[5] S. Dobnik, J. D. Kelleher, and C. Koniaris. Priming and alignment of frame of reference in situated conversation. Proceedings of Dial‐Watt‐Semdial, pages 43–52, 2014.

[6] N. Hawes, M. Klenk, K. Lockwood, G. S Horn, and J. D. Kelleher. Towards a cognitive system that can recognize spatial regions based on context. In Proceedings of the 26th National Conference on Artificial Intelligence (AAAI’12), 2012, 2012.

[7] J. Hois, R.J. Ross, J.D. Kelleher, and J.A. Bateman, editors. Proceedings of the 2nd Workshop on Computational Models of Spatial Language Interpretation and Generation (CoSLI‐2) at Cognitive Science, 2011.

[8] S. Dobnik J.D. Kelleher, R.J. Ross, editor. Proceedings of the 3rd Workshop on Computational Models of Spatial Language Interpretation and Generation (CoSLI‐3) at the International Workshop on Computational Semantics (IWCS), 2013.

[9] J.D. Kelleher. Attention driven reference resolution in multimodal contexts. Artificial Intelligence Review, 25(1): 21–35, 2006.

[10] J.D. Kelleher and F. Costello. Cognitive representations of projective prepositions. In Proceedings of the Second ACL‐Sigsem Workshop of the Linguistic Dimensions of Prepositions and their Use in Computational Linguistic Formalisms and Applications, 2005.

[11] J.D. Kelleher, F. Costello, and J. van Genabith. Dynamically structuring, updating and interrelating representations of visual and linguistic discourse context. Artificial Intelligence, 167(1): 62–102, 2005.

[12] J.D. Kelleher and F.J. Costello. Applying computational models of spatial prepositions to visually situated dialog. Computational Linguistics, 35(2): 271–306, June 2009.

[13] J.D. Kelleher and G.J. Kruijff. A context‐dependent algorithm for generating locative expressions in physically situated environments. In Proceedings of the 10th European Workshop on Natural Language Generation (ENLG‐05), pages 68–74, Aberdeen, Scotland, August 2005.

[14] J.D. Kelleher and G.J. Kruijff. A context‐dependent model of proximity in physically situated environments. In Proceedings of the 2nd ACL‐SIGSEM Workshop on The Linguistic Dimensions of Prepositions and their use in Computational Linguistic Formalisms and Applications, 2005.

[15] J.D. Kelleher and R.J. Ross. Topology in compositie spatial terms. In D.N. Rapp, editor, Proceedings of Spatial Cognition 2010: Poster Presentations, pages 46–50. SFB/TR8 Spatial Cognition, 2010.

[16] J.D. Kelleher, R.J. Ross, B. Mac Namee, and C. Sloan. Situating spatial templates for human‐robot interaction. In Proceedings of the AAAI Symposium on Dialog with Robots, 11–13 Nov. 2010.

[17] J.D. Kelleher, R.J. Ross, C. Sloan, and B. Mac Namee. The effect of occlusion on the semantics of projective spatial terms: a case study in grounding language in perception. Cognitive Processing 12(1): 95–108, 2010.

[18] J.D. Kelleher and J. van Genabith. A computational model of the referential semantics of projective prepositions. In Syntax and Semantics of Prepositions, pages 211–228. Springer, 2006.

[19] J.D. Kelleher, Colm Sloan, and Brian Mac Namee. An investigation into the semantics of English topological prepositions. Cognitive processing, 10: 233–236, 2009.

[20] J.D. Kelleher. A perceptually based computational framework for the interpretation of spatial language. Ph.D. thesis, Dublin City University, 2003.

[21] J.D. Kelleher and G‐J. M. Kruijff. Incremental generation of spatial referring expressions in situated dialog. In Proceedings of the 21st International Conference on Computational Linguistics and the 44th annual meeting of the Association for Computational Linguistics, pages 1041–1048. Association for Computational Linguistics, 2006.


[22] J.D. Kelleher, B. Mac Namee, and A. D’Arcy. Fundamentals of Machine Learning for Predictive Data Analytics: Algorithms, Worked Examples, and Cast Studies. MIT Press, 2015.

[23] G‐J. M. Kruijff, J.D. Kelleher, and N. Hawes. Information fusion for visual reference resolution in dynamic situated dialogue. In International Tutorial and Research Workshop on Perception and Interactive Technologies for Speech‐Based Systems, pages 117–128. Springer Berlin Heidelberg, 2006.

[24] G.J. Kruijff, J.D. Kelleher, G. Berginc, and A Leonardis. Structural descriptions in human‐assisted robot visual learning. In Proceedings of the 1st ACM SIGCHI/SIGART Conference on Human‐Robot Interaction, pages 343–344. ACM, 2006.

[25] G.D. Logan and D.D. Sadler. A computational analysis of the apprehension of spatial relations. In P. Bloom, M.A. Peterson, L. Nadel, and M.F. Garret, editors, Language and Space, pages 493–530. MIT Press, 1996.

[26] J.D. Kelleher, R.J. Ross, J. Hois, editor. Proceedings of the 1st Workshop on Computational Models of Spatial Language Interpretation and Generation (CoSLI) at Spatial Cognition, 2010.

[27] R.J. Ross and J.D. Kelleher. Putting things “between” perspective. In Proceedings of the 21st Irish Conference on Artificial Intelligence and Cognitive Science conference (AICS), September 2010.

[28] R.J. Ross and J.D. Kelleher. Using the situational context to resolve frame of reference ambiguity in route descriptions. In Proceedings of the Second Workshop on Action, Perception and Language (APL’2), Uppsala, Sweden, November 2014.

[29] N. Schutte, J. Kelleher, and B. Mac Namee. Reformulation strategies of repeated references in the context of robot perception errors in situated dialogue. In Proceedings of the Workshop on Spatial Reasoning and Interaction in Real‐World Robotics at the International Conference on Intelligent Robots and Systems, 2015.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Khachaturyan—44


MARIA KHACHATURYAN Department of Anthropology

University of California, Berkeley Email: [email protected]

Recognitional function of (distal) demonstratives cross‐linguistically

n recent years, a growing number of works have appeared which critically analyze deictic

markers (e.g. English this and that, Russian etot and tot, Mano tɔɔ, dı a, ɓɛ, ya . . .) beyond

mere spatiality (Hanks 2005, 2011; Enfield 2003; Jarbou 2010, inter alia; for a recent

overview, see Peeters & Özyürek 2016). Instead of a linear approach to deictic markers as

encoding relative distance, a multimodal approach was suggested, where the psychological

proximity (Peeters & Özyürek 2016), or accessibility (Hanks 2011), replaces mere physical

proximity. Crucially, according to this theory of deixis, the speaker and the addressee

establish referents’ accessibility jointly, through joint attention focus (Clark et al. 1983,

Diessel 2006), which is opposed to the widespread egocentric approach to deixis.

That deictic markers often have anaphoric (discourse‐referential) function is well known.

However, this function is often seen as a metaphorical extension of spatial deixis proper

(Anderson & Keenan 1985). On the contrary, Hanks (2011, inter alia) argues that anaphora,

being an instance of cognitive access to the referents, functions alongside with perceptual

access, which includes, but is not restricted to, spatial (visual) access.

In this paper, I explore the function of recognitional deixis, which has not received much

scholarly attention, on a preliminary cross‐linguistic sample. Deictic markers in the

recognitional function (Schegloff 1972) serve to identify referents which are not immediately

accessible on the interactive scene. These referents are rather accessible cognitively. Unlike

anaphora, however, they are not directly mentioned in the discourse immediately prior to

the utterance in question. The access is enabled via the common ground of the interlocutors,

which is constructed in previous interaction experience. Therefore, recognitional function is a

primary example of cognitive access to referents, established jointly. See an example of

English demonstrative that in the recognitional function:

(1) (A yoga teacher to her students): Widen those collar bones.

“Widening” the collarbones (and opening the chest) is a common element of yoga postures. Without

pointing to any of the students’ collar bones, and without mentioning them in the discourse immediately

preceding the utterance, the teacher assumes that the students, already familiar with the practice, will

recognize and implement the movement.

In what follows, (1) I introduce recognition among other functions of cognitive access to

referents (anaphora, and also bridging); (2) I then argue that there is a cross‐linguistic

tendency for the recognitional function to be expressed by deictic markers which, in the

spatial axis, express relative distance from the object (distal deictics); (3) The common

I


ground being a socially constructed phenomenon, I introduce social‐pragmatic usages of

recognitional function.

Cognitive access to referents: recognition, anaphora, bridging

Recognitional function is a function of cognitive access to the referents, along with the widely‐studied function of tracking reference in discourse, or anaphora (Fox 1996), and also bridging (Clark 1975). In case of anaphora, the referent is mentioned in the prior discourse. Thus, in Crow narratives salient characters are often referred to with the use of proximal marker hinné:

(2) hinne iisaakshee‐sh hinne bachee‐sh duuxalu‐ok bin‐naaske aa‐ii‐ak

this young.man‐DET this man‐DET drag‐SS water‐edge PORT‐reach‐SS

“this young man dragged this man and brought him to the bank of the stream.” (Crow; Graczyk 2009: 70)

In the case of bridging, the referent is inferable by proxy: from a frame or other bridging context mentioned in the immediate prior discourse. Consider ex. 3 from Mandarin, where the distal demonstrative nei functions as a marker of bridging:

(3) Zuotian wanshang wo shui bu zhao

yesterday evening I sleep not achieve

Gebi de nei tiao gou jiao de lihai

next.door ATT that CLS dog bark RES terribly

“I couldn’t sleep last night. The dog next door was barking.” (Mandarin, Crosthwaite 2014:463).

In this example the dog is a discourse‐new referent. However, it is inferable by a cause‐consequence

bridging context: “I couldn’t sleep because the neighbor has a dog, and that dog was barking.”

Both anaphora and bridging involve previous discourse; bridging also involves some broader contextual knowledge shared by the interlocutors. However, as illustrated in ex. 1 and in the forthcoming examples, recognition is based solely on common ground. Scheme 1 illustrates the types of knowledge involved in anaphoric, bridging and recognitional expressions:

Scheme 1. Types of knowledge involved in cognitive access to referents

anaphora bridging recognition

previous discourse previous discourse

common ground

(frames, contextual knowledge)

common ground

Recognitional function and distal deictic markers

Recognitional function, as well as other functions of cognitive access, is in many languages

conveyed by the same means as the more straightforward function of demonstratives, namely,

visual access. The recognitional function has a chance to be universal: it is possible that in any

language at least some (demonstrative) marker will have the recognitional function (possibly,

among others). Moreover, medial and distal demonstratives seem to be preferred for the


recognitional function. See example from Crow, where the demonstrative eehk (medial, close to

addressee) is used in recognitional function:

A boy is missing, and his mother cannot find him.

(4) bacheeıtche bachee xaxua bachaahii‐ak ooppii‐ak “eehk shikaak‐kaata‐m

chief man all gathertogether‐SS smoke‐SS “that boy‐DIMIN‐DET

xapıi‐o‐k”

lost‐CAUS.PL‐DECL

“the chief gathered all the men together,” he smoked, “they lost that little boy” [he said] (Crow, Graczyk

2009: 72)

Yucatec Maya is another case at hand: the recognitional function is often marked with the non‐

immediate enclitic –o’:

(5) Father arrives home from travel and notices that one of his four children is not around and so asks his

spouse:

kux tuún le pàal o’, tz’ú chan xantal má’a tinwilik

“How about that kid, it’s been a while since I’ve seen him” (Hanks 2016)

Russian has a specialized marker –to, restricted to recognition. This marker grammaticalized from

the distal demonstrative tot:

(6) A conversation overheard in a bus: ‐ Pomniš? On‐to. Znaeš, pravda! ‐ Da? remember.2SG he‐TO know.2SG truth yes“‐ Do you remember? The man. You know, it’s true! – Really?” (Bonnot 1986:115)

Recognition in interactional context

In Yucatec Maya, stereotypical referents are often introduced with the non‐immediate marker ‐o’.

(7) chokow le k’ıin o’ “The sun is hot” (Hanks 2005:207).

Compare ex. (7) with a stereotypical referent “the sun” with (8), where “this wind” encoded by

use of the immediate marker –a’ has a value of focus associated with it: the speaker emphasizes

that the wind is stronger than the days before (and takes it as a sign of the approach of the hot

season):

(8) k’aam e ‘ıik’ a’ (pointing up) astah bey u taal camyon e’

“This wind is loud. It’s as if a truck were approaching” (Hanks 2005:207).

Therefore, in Maya recognized and stereotypical referents are contrasted with truly discourse‐

new and focalized ones, which is supported by the paradigmatic opposition between immediate

and non‐immediate deictic markers.

Deictic markers in the recognitional function can be used strategically, as a means to

formulate an utterance that would presuppose the existence of common ground between the

interlocutors and, in some cases, their joint community membership. This is the strategy that I

frequently observed in the native Bible translations by the Catholic community of Mano (Mande;

Guinea). These translations occur spontaneously, during the Sunday service, when the catechists


orally translate from the French Bible. Note the translation introducing (Lc 4:21) below. There

was no previous discourse on the subject, so the usage of the distal demonstrative ya “that”

cannot be explained by the anaphoric function; neither the objects in question were present at

the interaction scene or otherwise accessible.

(9) walaka lɛ e kɛ Nazareth ya wı a, ye e ɲɛ gbaaɓo Isaïe la sɛɓɛ ya gee a ka a . . .

“(In) the house of God that was in THAT Nazareth, when he finished reading THAT book of Esaiah...”

(own data [service_nzao_0131_00:40:28])

By using the demonstrative ya in the recognitional function, the catechist, who was orally

translating the Bible, backgrounded the referents and made them appear as already known by

the congregation. The effect was that the listeners were stepping into an ongoing story

(Clark&Haviland 1977:37). Common ground is a property of a social community (Enfield 2006); in

this case, a religious community. Thus, presupposing the common ground, the catechist

simultaneously presupposes the existence of a religious community.

References Anderson, Stephen and Edward Keenan. 1985. Deixis. In: Timothy Shopen (ed.), Language Typology and

Syntactic Description, vol. 3: Grammatical Categories and the Lexicon, pp. 259–308. Cambridge: Cambridge

University Press.

Bonnot, Christine. 1986. TO Particule de rappel et de thématisation. In Les particules énonciative en russe

contemporain, vol 2, pp. 113–171. Paris : Institut d’Études Slaves.

Clark, Herbert H. 1975. Bridging. In R. C. Schank & B. L. Nash‐Webber (Eds.), Theoretical issues in natural

language processing, pp. 169–174. New York: Association for Computing Machinery.

Clark, Herbert H. & Susan E. Haviland. 1977. Comprehension and the given‐new contract. In R. O. Freedle

(Ed.), Discourse production and comprehension, pp. 1–40. Hillsdale, NJ: Erlbaum.

Clark, H.H., R. Schreuder and S. Buttrick. 1983. Common ground and the understanding of demonstrative

reference. Journal of Verbal Learning and Verbal Behavior 22: 245–258

Crosthwaite, Peter Robert. 2014. Definite Discourse–New Reference in L1 and L2: A Study of Bridging in

Mandarin, Korean, and English. Language Learning 64:3: 456–492.

Diessel, Holger. 2006. Demonstratives, joint attention, and the emergence of grammar. Cognitive Linguistics 17:

463–489.

Enfield, Nick J. 2003. Demonstratives in space and interaction: data from Lao speakers and implications for

semantic analysis. Language 79: 82–117.

Enfield, Nick J. 2006. Social Consequences of Common Ground. In Nick J. Enfield and Stephen C. Levinson, eds,

Roots of Human Sociality, pp. 399–430. Oxford: Berg.

Fox, Barbara A. 1996. Studies in anaphora. Amsterdam: John Benjamins.

Graczyk, Randolph. 2007. A grammar of Crow. Lincoln: University of Nebraska Press.

Hanks, William F. 2005. Explorations in the deictic field. Current Anthropology 46(2): 191–220.

Hanks, William F. 2011. Deixis and indexicality. In Wolfram Bublitz and Neal R. Norrick. Foundations of

pragmatics, pp. 315–346. Berlin: Mouton de Gruyter.

Hanks, William F. 2016. Deixis, translation and relativity. Paper presented at Translating Worlds, workshop held

at Berkeley, January 15–16, 2016.


Jarbou, S.O. 2010. Accessibility vs. physical proximity: an analysis of exophoric demonstrative practice in Spoken

Jordanian Arabic. Journal of Pragmatics 42, 3078–3097.

Peeters, David and Aslı Özyürek. 2016. This and That Revisited: A Social and Multimodal Approach to Spatial

Demonstratives. Frontiers of Psychology 7:222.

Schegloff, Emanuel. 1972. Note on a conversational practice: Formulating place. In: David Sudnow (ed.), Studies

in Social Interaction, pp. 75—119. New York: Free Press.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Klippel—49


ALEXANDER KLIPPELDepartment of Geography

The Pennsylvania State University Email: [email protected]

eferencing something in the world always relates one entity to another. While possible to

develop artificial systems such as latitude and longitude to locate an entity at an absolute

location on earth, in natural language and in human concepts of location a ground object is

either implicit or explicit. This means that the question turns more directly to the relation

between entities. The simplest case is the relation between two entities such as the “the gas

station north of campus” or “I am at the coffee shop.”

Research in my group has addressed questions of the relation between spatial entities in

language and from the perspective of conceptualization both in controlled experiments as well as

in a crowdsourcing environment. Both approaches have shown to be of value for a deeper

understanding of spatial referencing, identifying universals, and in exploring cross‐cultural and

cross‐regional variation. I see my contributions to the specialist meeting as both topical as well as

methodological.

Brief overview of topics addressed in my team

Our research has addressed spatial relations between spatial entities of different dimensions for

both static and dynamic scenarios in the geo‐spatial realm (i.e., we are usually not focusing on

table top spaces). For controlled behavioral experiments addressing spatial relational concepts

and spatial relational language we often select qualitative spatial formalism as a starting point. As

briefly discussed below, the advantage is the identification of spatial equivalence classes which

lend itself to the formulation of hypotheses. Hypotheses can be tested and correspondence

between formal and cognitive equivalence classes can be established. Additionally, availability of

equivalence classes / formalisms for various types of spatial relations such as topology,

directions, entities of different dimensions such as line‐line or line‐region relations, or the

possibility to characterize trajectories allow for covering a variety of possible spatial relations.

One example (see Appendix A) are direction relations between entities such as airplanes. In our

research (Klippel, Wallgrün, Yang, & Sparks, 2015) we systematically looked at direction concepts

that people apply to characterize the direction relation between the two airplanes, how direction

concepts are linguistically externalized, and how they differ at the level of individuals. Our results

clearly show that humans have mutually exclusive direction concepts, that is, some of them apply

sector‐based direction concepts while others use cone‐shaped approaches. Only acknowledging

these differences allows for validating results from a statistical perspective. In other words, there

are no intuitive universal direction concepts.

Other examples revealed through controlled behavioral experiments are linguistic and non‐

linguistic turning concepts at intersections showing that they are not identical (Klippel &

Montello, 2007), the influence of domain semantics on relational movement concepts for both

R


translation and scaling and how semantics changes the importance of certain topological

relations (Klippel, 2012; Yang, Klippel, & Li, 2015), the cross‐linguistic comparison of overlap

relations showing a universal distinction of non‐overlapping, overlapping, and proper‐part

relations in English, Korean, and Chinese (Klippel, Wallgrün, Yang, Mason, & Kim, 2013), or the

comparison of conceptualizing static versus dynamic path relations revealing a stronger focus on

ending relations in the static case (Li, Klippel, & Yang, 2011).

A second line of research in my team has addressed the interpretation of natural language in

a crowdsourcing, often corpus‐building environment. Crowdsourcing, or what we call passive

crowdsourcing (collecting data available on websites, tweets, etc.), has the advantage of creating

massive amounts of data and in combination with geo‐referencing tools can provide insight into

regional variation of referencing expressions. We studied the use of cardinal versus relative route

directions across the United States (Xu, Klippel, MacEachren, & Mitra, 2014) finding significant

differences between the coastal (more relative directions) and interior states (more cardinal

directions). We also generated a corpus approach for ground‐truthing the use of a variety of

spatial relational expressions such as near, next to, or close by (Wallgrün, Klippel, & Baldwin,

2014). This approach is using geo‐referenceable entities to create triplets such that the geo‐

location of figure and ground are known allowing for identifying a distance measure for terms

such as near. A large enough corpus then allows for identifying contextual factors such as the size

of the entities. This work is ongoing.

A brief perspective on universals

My personal perspective on universals is that they can and should be grounded in the concept of

invariance. Researchers in many scientific fields from the cognitive to the spatial sciences have

addressed the topic of invariance in the context of cognitive information processing as well as to

formally distinguish fundamental concepts of space and time. Worboys and Duckham (2004), for

example, utilize the concept of invariance as a framework for their chapter on Fundamental

Spatial Concepts. Proposed first by Felix Klein (see Erlangen program (Wikipedia, 12/3/2011)),

geometries can be distinguished based on invariant properties under certain transformations.

This approach allows for differentiating Euclidean geometry from set‐based geometry as well as

topology. Alongside formal approaches, perceptual and cognitive invariants, which we also find

to be associated with conceptual primitives, have long been of interest to the cognitive science

community. Klix (1971) states that the human mind, in adapting to its environment, identifies

invariant characteristics of information that form the basis for cognitive processes. Likewise, the

classic work by Gibson (1979) refers to temporally constant characteristics of environments as

structural invariants. If we seek to identify universals, invariants in human perception of

environmental characteristics and ideally their formal characterization seem to be an excellent

starting point.

A brief perspective on methods

Over the last 10 years we have placed substantial effort on improving methods for understanding

both referencing concepts and referencing language. We have developed a framework

comprising of a highly efficient way to collect data on human concepts and human language. This


framework consists of CatScan (for controlled behavioral data collection) and CatAnalysis (for in‐

depth data analysis). Examples: in two recent articles we developed A) an approach to validate

human cognitive concepts from a statistical perspective (Wallgrün, Klippel, & Mark, 2014). This

approach is grounded in approaches on cluster validation and uses a bootstrapping methods to

identify stable (maybe universal) concepts; B) we develop a way to quantify variation of concepts

using the statistical measure of variance. Appendix B shows two images of landscape concepts,

one with high variance and one with low variance. Quantifying variation this way may allow for

better understanding universals and variation. In the example shown in the Appendix greater

conceptual variation correspondents to greater linguistic variation.

References Gibson, J. (1979). The ecological approach to visual perception. Boston, MA: Houghton Mifflin.

Klippel, A. (2012). Spatial information theory meets spatial thinking—Is topology the Rosetta Stone of spatio‐

temporal cognition?

Annals of the Association of American Geographers 102(6): 1310–1328. doi:10.1080/00045608.2012.702481

Klippel, A., & Montello, D. R. (2007). Linguistic and nonlinguistic turn direction concepts. In S. Winter, B. Kuipers,

M. Duckham, & L. Kulik (Eds.), Spatial Information Theory. 9th International Conference, COSIT 2007,

Melbourne, Australia, September 19‐23, 2007 Proceedings (pp. 354–372). Berlin: Springer. Retrieved from

http://www.springerlink.com/content/g4l3k87k77120459/ Klippel, A., Wallgrün, J. O., Yang, J., Mason, J., &

Kim, E.‐K. (2013). Fundamental cognitive concepts of space and time: Using cross‐linguistic, crowdsourced

data to cognitively calibrate modes of overlap. In T. Tenbrink, J. Stell, A. Galton, & Z. Wood (Eds.), Spatial

Information Theory. 11th International Conference, COSIT 2013, Scarborough, UK, September 2–6, 2013,

Proceedings (pp. 377–396). Berlin: Springer.

Klippel, A., Wallgrün, J. O., Yang, J., & Sparks, K. (2015). Intuitive direction concepts. The Baltic International

Yearbook of Cognition, Logic and Communication 10: 1–30.

Klix, F. (1971). Information und Verhalten: Kybernetische Aspekte der organismischen Informationsverarbeitung;

Einführung in naturwissenschaftliche Grundlagen der allgemeinen Psychologie. Bern: Huber.

Li, R., Klippel, A., & Yang, J. (2011). Geographic event conceptualization: Where spatial and cognitive sciences

meet. In L. A. Carlson, C. Hölscher, & T. F. Shipley (Eds.), Proceedings of the 33rd Annual Conference of the

Cognitive Science Society. (pp. 3168–3173). Austin, TX: Cognitive Science Society.

Wallgrün, J. O., Klippel, A., & Baldwin, T. (2014). Building a corpus of spatial relational expressions extracted

from web documents. In : GIR ’14, Proceedings of the 8th Workshop on Geographic Information Retrieval (pp.

6:1–6:8). New York, NY, USA: ACM. Retrieved from http://doi.acm.org/10.1145/2675354.2675702

Wallgrün, J. O., Klippel, A., & Mark, D. M. (2014). A new approach to cluster validation in human studies on

(geo)spatial concepts. In K. Stewart, E. Pebesma, G. Navratil, P. Fogliaroni, & M. Duckham (Eds.), Extended

Abstract Proceedings of the GIScience 2014 (pp. 450–453). Vienna: Hochschülerschaft, TU Vienna.

Wikipedia. (2011). Erlangen program—Wikipedia, the free encyclopedia. Retrieved from

http://en.wikipedia.org/w/index.php?oldid=461022212

Worboys, M., & Duckham, M. (2004). GIS: A Computing Perspective. (2nd ed.). Boca Raton FL: CRC Press.

Xu, S., Klippel, A., MacEachren, A. M., & Mitra, P. (2014). Exploring regional variation in spatial language using

spatially‐stratified web‐sampled route direction documents. Spatial Cognition and Computation 14(4): 255–

283. doi:10.1080/13875868.2014.943904 Yang, J., Klippel, A., & Li, R. (2015). The cognition of change:

Scaling deformations in mind and spatial theories. Cartography and Geographic Information Science 42(3):

224–234. doi:10.1080/15230406.2014.991425


Appendix A: Example stimuli from two direction experiments.

Appendix B: Examples of landscape concepts (method can be applied to relational concepts, too) with high or

low variance.

These images form highly consistent categories. High variance (last value) indicates small variation, that is, images are conceptualized similarly by all participants.

These images form highly inconsistent categories. Low variance (last value) indicates small variation, that is, images are conceptualized differently across participants.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Mark and Turk—53

Some Observations on Spatial Referencing in Yindjibarndi Culture and Language

DAVID M. MARK

Department of Geography University at Buffalo (SUNY) Email: [email protected]

ANDREW G. TURKSchool of Information Technology

Murdoch University, Perth


his position paper reports on spatial referencing in the Yindjibarndi language. Yindjibarndi is

spoken by about 500 people in the Pilbara region of Northwestern Australia (Wordick 1982;

Mark and Turk 2003; Mark, Turk, and Stea 2007). Yindjibarndi belongs to the Coastal Ngayarda

language group, within the southwest group of Pama‐Nyungan languages David Mark and

Andrew Turk conducted landscape and language fieldwork with Yindjibarndi speakers on field

trips between 2002 and 2009.

Spatial Referencing

Spatial referencing in Yindjibarndi is represented by several different parts of speech and

linguistic structures, including locative suffixes, directional nouns, and other nouns.

Locative Suffixes

“Languages of the Pama‐Nyungan family are entirely suffixing in their morphology . . . . Suffixes

serve also to derive new words, and among them those meaning ‘having’ and ‘lacking’ are almost

universal in Australian languages.” (Gutman and Avanzati 2013). Anderson (1986) listed 141

suffixes in the Yindjibarndi language. Anderson classified seven of these as “locative case.” Directional nouns

Wordick (1982) and Anderson (1986) identified five classes of nouns. Two of those classes

contain only three nouns each, and were named “Directional nouns (north type) (NDN)” and

“Directional nouns (south type) (NDS) by Wordick.

The three “north type” directional nouns (from Anderson 1986) mean “in or at the north,”

upstream/interior, and downstream. The three “south type” directional nouns mean “in the

south,” “in the west,” and “in the east.” The principal river of traditional Yindjibarndi country,

known as the Fortescue River in English, has intermittent or seasonal flow. For most of its length,

the Fortescue flows west. The two groups of directional nouns are grouped in a way that is

unexpected to an English speaker: “In the north” is grouped with the two riverine directionals,

whereas the other three cardinal directions are in the other group. This begs for further

investigation. Qualitative Distance Noun:

“Wana refers to ngurra (ground) in the middle distance, such as the side of a marnda (or perhaps

a flat area)—where you can still see something (like a kangaroo) but it is much too far away to

throw a stone at it (or shoot the kangaroo). Wanangga could refer to the location of something in

the middle distance.” (Turk and Mark 2008).

T


Transcripts and word counts for locatives:

In June 2006, we showed (40) photographs of Australian landscapes, mostly from traditional

Yindjibarndi country or the area where the people now live (Roebourne), separately to four

groups of Yindjibarndi speakers. One “group” consisted of just one person, the oldest speaker

living at the time. Others were small groups of two or more speakers. People were asked to “tell

us what’s on the picture, in Yindjibarndi language” or something similar. The four sessions

produced almost five hours of audio, and were transcribed by the authors. The transcripts

contained 26,211 words. Word Counts for locatives: The Yindjibarndi dictionary (Anderson) included 19 terms ending with

the locative suffix –ngga. The combined transcripts contained at total of 64 terms ending with–

ngga. The suffix was attached to 28 different nouns within the transcripts. Only four of the 28

terms from the transcripts appeared in the dictionary. Most or all of the others presumably were

formed from –ngga as a productive locative suffix. Thungga is the most common –ngga word in

the transcripts, occurring 40 times total, and also appears in the dictionary. But thungga is said to

mean “soil, dirt, sand” and in this case, –ngga does not seem to be the locative suffix; “thu‐“ does

not seem to be a root in Yindjibarndi. Similarly, bungga means “fall”. Wirrangga is a term for Red

River Gum trees, and again the –ngga does not make sense as a locative, although since wirra

means boomerang, perhaps there is a semantic connection. The other two –ngga words that

were found both in the transcripts and in Anderson’s dictionary were: ngarlingga (downwind) and

gardangga (down, low, below). The most common other –ngga compound nouns in the

transcripts were wirgangga (wirga = gap), bawangga (bawa = water), biyungga (biyu = dry), and

thalungga (thalu = ritual site).

Summary

Yindjibarndi speakers are able to communicate locations in the environment through a variety of

elements of their language. Landscape terms themselves refer to entities in the environment.

Locative suffixes are a way to use landscape entities as grounds for locating other entities

(figures). Two closed classes of nouns denote directions. Four of the directional nouns appear to

denote the four cardinal directions while to refer to upstream and downstream. The use of these

six directional nouns would be a high priority if future research with speakers becomes possible.

References Anderson, B. (1986) Yindjibarndi Dictionary. Darwin: Summer Institute of Linguistics.

Anderson, B., and Thieberger, N. No date [accessed October 2002]. Yindjibarndi dictionary. Document 0297 of

the Aboriginal Studies Electronic Data Archive (ASEDA) Australian Institute of Aboriginal and Torres Strait

Islander Studies.

Gutman, A., and Avanzati, B., 2013. Australian Languages. The Language Gulper.

http://www.languagesgulper.com/eng/Australian.html [accessed 9‐9‐2016]

Mark, D. M., and Turk, A. G., 2003. Landscape Categories in Yindjibarndi: Ontology, Environment, and Language.

In Kuhn, W., Worboys, M., and Timpf, S., Editors, Spatial Information Theory: Foundations of Geographic

Information Science, Berlin: SpringerVerlag, Lecture Notes in Computer Science No. 2825, pp. 31–49.


Mark, D. M., Turk, A. G., and Stea, D., 2007. Progress on Yindjibarndi Ethnophysiography. In Winter, S., Duckham,

M., Kulik, L., Kuipers, A., (editors) Spatial Information Theory. Lecture Notes in Computer Science No. 4736,

pp. 1–19.

Turk, A.G. and Mark, D.M. 2008. Illustrated Dictionary of Yindjibarndi Landscape Terms.

Informal publication, Murdoch University; Western Australia.

Wordick, F. J. F., 1982. The Yindjibarndi language. Pacific linguistics. Series C., no. 71. Canberra: Dept. of

Linguistics, Research School of Pacific Studies, Australian University, 1982.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Mesh–56


KATE MESH Department of Linguistics

The University of Texas at Austin


o locate an item in the world, we typically perform two linked behaviors: we use a spoken

language expression—often a spatial deictic term like here or there that carries limited

semantic information—and we use some combination of our hands and head to point to the item

in question.1 These behaviors are linked in that they both indicate: that is, both direct the

interlocutor's attention to a more or less narrowly circumscribed search space (Clark 1996,

discussed in Cooperrider 2015). What's more, these behaviors are linked in production: they occur

together more often than alone, a fact often interpreted as evidence of their intrinsic connection

(see e.g., Diessel 2006; Levinson 2003).

Of the two types of indicating strategies, the spoken variety has received substantially more

scholarly attention in linguistics, psychology, and related disciplines. Nevertheless, a sufficient

number of studies on gestural spatial referencing have been performed to allow for early

arguments for and against possible universals in the behavior of pointing (see, e.g., Kita 2003).

The very act of manual pointing—the behavior of extending an arm and hand to designate an

intended referent—has been described as a possible universal of human communication. But

while pointing as a broad phenomenon may be universal, it remains to be shown whether any of

its forms is universal. Consider the example of pointing handshape: an extended index finger is a

form that recurs in many of the world’s pointing systems, perhaps because it is the hand

configuration motorically easiest for humans to produce (Povinelli & Davis 1994). Index finger

extension has been described as the preferred configuration in pointing produced by children

(Liszkowski et al 2012). But Wilkins (2003) argued against the notion that the index finger

handshape remains privileged in adult pointing behavior cross‐culturally. He observed that adult

speakers of Arrernte control a variety of pointing handshapes, the use of which encodes

information about the pointing referent. While the extended index finger handshape is present in

the Arrernte system, it does not have special status within it. In making this observation, Wilkins

drew attention to the potential conflict between (motoric or cognitive) motivations shaping

pointing and the demands of a semi‐arbitrary semantic encoding system.

1 Users of signed languages, of course, locate objects exclusively in the visual‐manual modality. Their indicating

behavior remains complex: like speakers, signers typically combine pointing strategies with gaze direction to

pick out their target.

T


I am interested in the question of whether some motivations are sufficient to ensure

morphological universals in manual pointing because my own work investigates one such

potentially universal feature. Since 2012 I have worked with speakers of San Juan Quiahije

Chatino (Zapotecan, Oto‐Mangeuan) in Oaxaca, Mexico. Using the ‘local environment interview’

method of Kita (2001), I have video recorded over 11 hours of multimodal spatial referencing

behavior used to locate landmarks inside the village of San Juan Quiahije and in its environs.

Analysis of these videos reveals that Chatino speakers use the height feature of manual pointing

gestures to encode information about the distance of the intended referent (Hou & Mesh 2015;

Mesh in prep). Referents located inside the community are indicated with a low or unelevated

elbow height, while referents outside the community are indicated with an elevated elbow. When

pointing accompanies the name of a landmark, this height modulation is perceptible but subtle.

When speakers make spatial referencing the focus of their talk (as for example when giving route

directions or answering a question about a landmark’s location), the height modulation is

amplified.

Meaningful modulation of pointing height to convey information about referent distance has

been documented for a handful of other pointing systems. Kendon (1988) observed this type of

distance‐indexing in the pointing of Warlpiri and Warramungu speakers. Wilkins (2003) found

evidence for a three‐way distinction in Arrernte pointing height that maps to the three‐way

distance distinction in the spoken Arrernte demonstrative system. de Vos (2014) found that users

of the signed language Kata Kolok use height to distinguish distal and proximal referents of

pointing signs. And Eco (1976), without connecting the observation to a specific cultural pointing

system, noted that points to distal targets require greater “energy” (p. 119), describing energy in

a way that suggests that the term is a proxy for pointing height.

That this phenomenon occurs across manual pointing systems is not coincidental: distant

objects generally appear higher in the visual field (Gibson 1950) and this sensory experience can

be reflected in pointing height. Additionally, pointing height mirrors a feature of one human

practical action: propelling an item to a distant location by hand. To throw an item in this way

necessitates raising the arm, and the increased arm elevation required to launch an item farther

can be reflected in pointing gestures that ‘throw’ nothing more than an object‐indicating vector.

Could pointing height index referent distance cross‐culturally? Phrased differently: is this

motivated feature of manual pointing robust enough to persist cross‐culturally—even when

pointing height is exploited to express additional meanings?2 To address this question, and many

analogous ones about motivation, arbitrariness, and universals in pointing behavior, requires the

collection of substantially more data than we have at present. The responsive plan of action is

2 Enfield et al. (2007) observe that pointing height distinguishes location‐focus from other pragmatic features in

Lao discourse. They are silent, however, on the question of whether pointing height also encodes referent

distance in Lao pointing, and on the potential for interaction between these two semiotic dimensions of

pointing.


clear: a full account of spatial referencing requires a rigorous program of cross‐cultural and

multimodal data collection—a program that the language documentation community has the

tools, and the incentive, to begin.

References Clark, H. H. (1996). Using language. Cambridge, UK: Cambridge University Press. Cooperrider, K. (2015). The co‐organization of demonstratives and pointing gestures. Discourse Processes 1–25.

Diessel, H. (2006). Demonstratives, joint attention, and the emergence of grammar. Cognitive Linguistics 17(4):

463–489.

Eco, U. (1976). A Theory of Semiotics (Vol. 217). Indiana University Press.

Enfield, N., Kita, S., & DeRuiter, J. P. (2007). Primary and secondary pragmatic functions of pointing gestures.

Journal of Pragmatics 39(10): 1722–1741.

Gibson J.J. (1950). The Perception of the Visual World. Boston: Houghton‐Mifflin.

Hou, L. & Mesh, K. (2015). Conventionalization of pointing & gestural motion descriptors in a Chatino speech

ecology. Invited talk. Workshop: Emerging Sign Languages & The Big Picture. Sponsored by the Center for

Cognitive Studies at Tufts University. Medford, MA. May 8–9.

Kendon, A. (1988.) Kendon, A. Sign Languages of Aboriginal Australia: Cultural, Semiotic and Communicative

Perspectives. Cambridge: Cambridge University Press.

Kita, Sotaro. (2001). Locally‐anchored spatial gestures, version 2: historical description of the local environment

as a gesture elicitation task. In Stephen C. Levinson & N.J. Enfield (eds.), Manual for the field season 2001.

Nijmegen: Max Planck Institute for Psycholinguistics. 132–135.

Kita, S. (2003). Pointing: Where Language, Culture and Cognition Meet. Lawrence Erlbaum & Associates.

Levinson, S. C. (2003). Deixis and pragmatics. In L. R. Horn & G. L. Ward (Eds.), The Handbook of Pragmatics (pp.

97–121). Malden, MASS: Blackwell Publishing.

Liszkowski, U., Brown, P., Callaghan, T., Takada, A., & de Vos, C. (2012). A prelinguistic gestural universal of

human communication. Cognitive Science 36: 698–713.

Mesh, K. (in prep.) Spatial Referencing in Speech, Gesture and Sign in the San Juan Quiahije Municipality.

(Unpublished doctoral dissertation.) The University of Texas at Austin. Austin, Texas.

Povinelli, D. J., & Davis, D. R. (1994). Differences between chimpanzees (Pan troglodytes) and humans (Homo

sapiens) in the resting state of the index finger: implications for pointing. Journal of Comparative Psychology

108(2): 134–139. de Vos, C. (2014). The Kata Kolok pointing system: Morphemization and syntactic

integration. Topics in Cognitive Science 7(1): 150–168.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Montello—59

Variations in Verbal Spatial Referencing: Factors Besides Culture and Language

DANIEL R. MONTELLO

Department of Geography University of California, Santa Barbara


here is a long history of academic debate about the relationship between human language and human thought, and the debate continues. Many of the deep psychological and

philosophical issues that make up this debate are surprisingly subtle and complex, and at some point, can become tedious to some of us. We once asked if people spoke differently about core properties of the physical world. Documenting that they did, we asked whether these differences meant that people think differently or just superficially use different verbal labels for essentially equivalent thoughts. Finding evidence that these verbal differences, at least in some cases, apparently did correspond to differences in the way people think nonverbally about the world, we asked whether these nonverbal cognitive differences were profound and fundamental, or just relatively superficial consequences of the way language focuses attention. In the latter case, the cognitive differences could be overcome relatively readily by training people to focus their attention differently. For example, it is likely that people who typically use relative (egocentric) systems can be taught to use absolute (abstract allocentric) systems fairly easily; people who practice this will develop the ability to maintain orientation with respect to such absolute systems. And so on.

In this position paper, however, I want to redirect our attention to variations in linguistic spatial referencing besides the common issues of culture and language. In a given physical situation (i.e., a particular environmental setting with particular objects, actors, etc.), people vary in their use of spatial referencing terms in ways that do not correlate only with language and culture. I believe there are important and interesting things to recognize about spatial language—including its reference system(s)—besides the fact that it varies to some degree with culture and language. I develop this position here in the form of three issues for discussion:

(1) language and culture are vague concepts and have an ambiguous relation to each other; (2) there are variations in spatial reference usage between individuals that correlate with

factors besides language and culture; and (3) there are variations in spatial reference usage within individuals over time/situations that

cannot derive from cultural or linguistic variation, by definition. First, language and culture are rather vague concepts, more vague than many other natural

language concepts. Linguists and others are well aware of the ambiguity of differentiating language families, groups, languages, dialects, jargons, and so on. Likewise, culture can refer to any group characteristics that are passed on through intentional or unintentional learning, and not determined by genetic or physical environmental causes. But this leaves many ambiguities, including that the group could be as small as two people. Furthermore, the relationship of

T


language and culture is ambiguous. They only partially overlap. Members of different cultures may speak the same language; they may be different dialects or jargons, or maybe they are not even that different. One could also argue that members of the same culture may speak different languages; again, perhaps they are only different dialects or jargons, perhaps they are full‐fledged languages. Of course, a person who considers language (even dialect) to be a defining component of culture will not be able to accept this last claim.

Second, variations in spatial reference usage between individuals do not correlate only with language and culture, the ambiguity of those concepts notwithstanding. People with the same culture and language sometimes use referencing systematically differently. We have evidence that plains and mountain folks who speak the same language nevertheless tend to differ in spatial referencing. We have evidence that rural and city folks tend to differ. We have evidence that folks living in cities with orthogonal grid networks speak differently than folks living in cities with irregular networks. I believe that some important part of the evidence that is presented for cultural and linguistic differences is really evidence for environmental differences. Of course, if one wants to use the concept of culture at a higher resolution than nationality (in the sense of a nation, as opposed to a state or country) or ethnicity, then one could say that any systematic difference in the way people live is an expression of culture or “sub‐culture.” The conceptual ambiguity of culture again.

Third, there are variations in spatial reference usage within individuals. A given person does not consistently use the same reference terms in all situations. They may use relative systems indoors but absolute systems outdoors. Or they may use relative systems with figural spaces (including “table‐top” spaces) but absolute systems with environmental spaces; spatial scale likely matters to reference system use. People do not always use the same reference system in the same situation at different times. They do not always use only one system, but may redundantly say something like “turn right, toward the ocean.” It happens every day here in Santa Barbara. By definition, the intra‐individual variation in this final paragraph cannot derive from cultural or linguistic factors. It’s the same person with the same culture and language.

References: Alloway, T. P., Corley, M., & Ramscar, M. (2006). Seeing ahead: Experience and language in spatial perspective. Memory & Cognition, 34, 380–386.

Boroditsky, L. (2003). Linguistic relativity. In L. Nadel (Ed.), Encyclopedia of cognitive science (pp. 917–921). London: MacMillan Press.

Brown, C. H. (1983). Where do cardinal direction terms come from? Anthropological Linguistics, 25, 121–161.

Davies, C., & Pederson, E. (2001). Grid patterns and cultural expectations in urban wayfinding. In D. R. Montello (Ed.), Spatial information theory: Foundations of geographic information science (pp. 400–414). Berlin: Springer.

Gumperz, J. J., & Levinson, S. C. (Eds.). (1996). Rethinking linguistic relativity. Cambridge, U.K.; New York: Cambridge University Press.

Habel, C., & Werner, S. (1999). Special Issue on “Spatial reference systems.” Spatial Cognition and Computation, 1.


Hart, R. A., & Moore, G. T. (1973). The development of spatial cognition: A review. In R. M. Downs & D. Stea (Eds.), Image and environment (pp. 246‐288). Chicago: Aldine.

Hermer‐Vazquez, L., Moffet, A., & Munkholm, P. (2001). Language, space, and the development of cognitive flexibility in humans: The case of two spatial memory tasks. Cognition, 79, 263–299.

Klatzky, R. (1998). Allocentric and egocentric spatial representations: Definitions, distinctions, and interconnections. In C. Freksa, C. Habel & K. Wender (Eds.), Spatial cognition (Vol. 1404, pp. 117). Berlin/Heidelberg: Springer.

Lakoff, G. (1987). Women, fire, and dangerous things. Chicago: University of Chicago Press.

Lakusta, L., & Landau, B. (2005). Starting at the end: The importance of goals in spatial language. Cognition, 96, 1–33.

Lawton, C. A. (1994). Gender differences in wayfinding strategies: Relationship to spatial ability and spatial anxiety. Sex Roles, 30, 765–779.

Lawton, C. A. (1996). Strategies for indoor wayfinding: The role of orientation. Journal of Environmental Psychology, 16, 137–145.

Lawton, C. A. (2001). Gender and regional differences in spatial referents used in direction giving. Sex Roles, 44, 321–337.

Levinson, S. C. (1996). Frames of reference and Molyneux's question: Crosslinguistic evidence. In P. Bloom, M. A. Peterson, L. Nadel & M. F. Garrett (Eds.), Language and space (pp. 109–169). Cambridge, MA: The MIT Press.

Levinson, S. C. (2003). Space in language and cognition: Explorations in cognitive diversity. Cambridge, UK; New York: Cambridge University Press.

Levinson, S. C., Kita, S., Haun, D. B. M., & Rasch, B. H. (2002). Returning the tables: Language affects spatial reasoning. Cognition, 84, 155–188.

Li, P., & Gleitman, L. (2002). Turning the tables: Language and spatial reasoning. Cognition, 83, 265–294.

Majid, A., Bowerman, M., Kita, S., Haun, D. B., & Levinson, S. C. (2004). Can language restructure cognition? The case for space. Trends in Cognitive Sciences, 8, 108–114.

Montello, D. R. (1991). Spatial orientation and the angularity of urban routes: A field study. Environment and Behavior, 23, 47–69.

Montello, D. R. (1993). Scale and multiple psychologies of space. In A. U. Frank & I. Campari (Eds.), Spatial information theory: A theoretical basis for GIS (pp. 312‐321). Berlin: SpringerVerlag.

Montello, D. R. (1995). How significant are cultural differences in spatial cognition? In A. U. Frank & W. Kuhn (Eds.), Spatial information theory: A theoretical basis for GIS (pp. 485‐500). Berlin: Springer.

Montello, D. R., & Xiao, D. (2011). Linguistic and cultural universality of the concept of sense‐ofdirection. In M. J. Egenhofer, N. A. Giudice, R. Moratz, & M. F. Worboys (Eds.), Spatial information theory (pp. 264–282). Proceedings of COSIT ‘11. Heidelberg: Springer, Lecture Notes in Computer Science 6899.

Munnich, E., Landau, B., & Dosher, B. A. (2001). Spatial language and spatial representation: A cross‐linguistic comparison. Cognition, 81, 171–207.

Prestopnik, J. L., & Roskos‐Ewoldsen, B. (2000). The relations among wayfinding strategy use, sense of direction,sex, familiarity, and wayfinding ability. Journal of Environmental Psychology, 20, 177–191.

Taylor, H. A., Naylor, S. J., & Chechile, N. A. (1999). Goal‐specific influences on the representation of spatial perspective. Memory & Cognition, 27, 309–319.

Taylor, H. A., Naylor, S. J., Faust, R. R., & Holcomb, P. J. (1999). "Could you hand me those keys on the right?" Disentangling spatial reference frames using different methodologies. Spatial Cognition and Computation, 1, 381–397.

Tversky, B., Lee, P., & Mainwaring, S. (1999). Why do speakers mix perspectives? Spatial Cognition and Computation, 1, 399–412.


Xu, S., Jaiswal, A., Zhang, X., Klippel, A., Mitra, P., MacEachren, A. M. (2010). From data collection to analysis: Exploring regional linguistic variation in route directions by spatiallystratified web sampling. In R. J. Ross, J. Hois, J. Kelleher (Eds.), Computational Models of Spatial Language Interpretation (CoSLI) Workshop at Spatial Cognition 2010 (pp. 49–52), Mt. Hood, Oregon.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Nikitina—63

Lexical Splits and Asymmetries Spatial Referencing: Revealing Universals through the Study of Variation

TATIANA NIKITINA

Department of Languages and Cultures of Sub‐Saharan Africa Centre National de la Recherché Scientifique (CNRS)


Introduction

Early studies in the linguistic representation of space were commonly inspired by bewildering

cross‐linguistic diversity; they sought to systematize that diversity and to understand its

underlying principles. The theoretical fruitfulness of these efforts notwithstanding, much stands to

be gained by looking closely at patterns of variation that recur with surprising regularity in

language after language but have not so far received much attention from typologists. In this

paper, I explore three of the major aspects of cross‐linguistic variation, focusing on the ways

alternative strategies co‐exist within one language. I believe that by shifting the emphasis from

sweeping typologies to language‐internal variation new insight can be gained into the way

different cognitive strategies compete for becoming conventionalized in grammar and usage. In

addition to traditional typological methods, such exploration requires the use of corpora,

experimental techniques, and sometimes tools for fine‐grained quantitative data analysis. In this

talk I try to highlight the way general principles of spatial representation are reflected, in different

languages, in lexical asymmetries, usage tendencies, and regularities in language change.

Overt encoding vs. contextual inference

One important dimension along which systems of spatial representation vary across languages is

the extent of relying on overt encoding of directionality as opposed to leaving it to be inferred

from context. In the study of motion expressions, that variation has been related most famously

to the distinctions between verb‐framed and satelliteframed languages, on the one hand, and

Manner and Path languages, on the other (Talmy 1975, 1985; Levin et al. 2010). Individual

languages, however, rarely restrict their motion expressions to just one strategy, and speakers

are often offered a choice between overt vs. implicit encoding. The alternatives are illustrated by

the variation between explicitly dynamic and static prepositional phrases in (1a,b), for English

and Russian, and in (2a,b) for Ancient Greek.

(1) a. Put it into / in the box. [US English]

b. Polozhi kljuchi na stol / na stole [Russian] put:IMPER keys on table:ACC on table:PREP

“Put the keys onto / on the table.”

(2) a. hò d’ en purì bálle thuēlás [Ancient Greek; Il. 9.220]

he:NOM PRT in fire:DAT throw:IMPF.3SG offerings:ACC

“and he threw sacrificial offerings in the fire”

b. kaì tês oreías anthemórruton gános ksouthês melíssēs and

ART mountainous:GEN flowing.from.flowers:ACC pride:ACC yellow:GEN bee:GEN es

puràn balô séthen [Ancient Greek; Eur. IT 634‐635]

into fire:ACC throw:FUT.1SG you:GEN


“and I will throw into your funeral pyre the honey of tawny mountain bees that streams from flowers.”

In spite of a number of superficial differences, the variation seems to be constrained by the

same factors in all three languages. One of the most important factors is manner specificity of

the verb, as reflected in corpus distributions in English (Nikitina 2008), and in patterns of

diachronic change in Ancient Greek and Russian (Nikitina 2012; Nikitina & Maslov 2013). Manner

specificity is ultimately related to the probability of inferring the directional meaning in context.

With verbs describing specific manners, and especially manners that represent highly unusual

ways of getting to a new location (e.g., waltzing or crawling), Goals of motion are normally

marked overtly.

The same probability of inferring the argument’s meaning is reflected in a different way in

lexical restrictions on combinations of verbs and spatial adpositions with Source and Goal

arguments (Nikitina 2009). In Wan (Mande), most motion verbs can only take one argument, and

that argument can only describe the Source or the Goal of motion depending on the verb (cf. 4a

vs. b). Several verbs, however, show flexibility in combining with either a Source or a Goal, and do

not restrict their locative argument to any particular role (5).

(4) a. è go kal ɛ gó (5) a. è sia yrɛ gó s/he left forest in s/he fell tree in

“She went from / *to the forest.” “She fell from the tree.”

b. è ɓo kal ɛ gó b. è sia trɔ ɔ mì s/he arrived forest in s/he fell ground by

“She arrived in / *from the forest.” “She fell on the ground.”

The verb’s behavior can be predicted based on one particular aspect of the verb’s meaning:

flexible role assignment is only allowed with verbs that describe asymmetric motion events, such

as events of vertical motion (the Goal of motion must be located lower on the gravitational axis

than its Source) or events of deictic motion (the Goal of motion is closer to the deictic center

than its Source). The entailed asymmetries make it easier to infer the role of the particular

locative argument: in (3b), with a verb of vertical motion, the particular location can only serve as

the Goal of motion, and in (3a), it is likely its Source. Probability of inferring directional meaning is

reflected in tendencies for overt Goal marking in English and in “rigid” associations between the

verb’s meaning and its argument structure in Wan.

Static relators vs. dynamic projections

Another dimension along which languages vary is the extent to which dynamic descriptions are

used to encode static spatial relations. English makes limited use of such expressions in this

function (6a), but ancient Indo‐European languages used them more regularly, cf. the Goal and

the Source expressions in (6b,c), from Ancient Greek. (6) a. A big storage chest stood to the left

of the door. (The British National Corpus)

b. hoi tà epì deksià tôn kefaléōn

they:NOM ART:ACC.PL on right:ACC ART:GEN.PL heads:GEN

komôsi, tà d’ ep’ aristerà keípousi (Hdt. Hist. 4.191.4)

let.hair.grow:PRES.3PL ART:ACC.PL PRT on left:ACC shave:PRES.3PL

“They let their hair grow long on the right side of their heads and shave the left.”

(literally, “to the right side, “to the left”)


c. ek deksiâs d’ autôn leukádioi kaì hoi álloi bárbaroi (Thuc. 2.81.3) from right:GEN PRT

they:GEN

L.:NOM.PL and ART.NOM.PL other:NOM barbarians:NOM

“and on their right [were] Leukadians and other barbarians” (literally, “from their right”)

The dynamic expressions describe static locations by projecting paths of fictive motion (cf.

Talmy’s notion of access path) leading away from the Ground (in the case of Source expressions)

or toward it (in the case of Goal expressions). In languages where both the static and the

dynamic strategies are available, the choice between them is again constrained by the same

underlying factors. For example, the use of the dynamic strategy is correlated with the overall

frequency of a particular spatial relator. More frequent spatial relators tend not to allow for the

dynamic marking, the latter being confined to less conventional and rarely invoked spatial

relations. Best candidates for dynamic encoding are relations for which no specialized

preposition is readily available (“to the north,” “to the right,” etc.).

In Indo‐European languages, specialized static prepositions compete with the old strategy of

relying on directional adverbs, pushing it to the periphery of the spatial domain (Nikitina forthc.).

The way individual languages are affected by that change can be predicted based on the close

relationship between frequency of use and the degree of grammaticalization of spatial relators as

a conventional means for describing static relations.

Frames of spatial reference

The last dimension to be discussed is the choice of a frame of spatial reference in descriptions of

objects’ location. While speakers of European languages make extensive use of the relative frame

of reference, many languages have been reported to hardly allow their speakers such an option

(Levinson 1996, 1997; Bohnemeyer 2011). I present an experimental study of Bashkir (Turkic),

which lets its speakers choose between three reference frames where English only allows for two.

In descriptions of a ball’s relation to a chair (aligned along the horizontal axis), speakers of

Bashkir may either resort to the relative reference frame or choose between two different

subtypes of the intrinsic frame: one based on the chair’s function (the chair’s front is where the

person sitting on it would be facing), the other based on the chair’s shape (the chair’s front is its

more prominent and elevated part, i.e. the part normally used to support the person’s back). The

complex three‐way contrast leads to multiple ambiguities and extensive inter‐ and intraspeaker

variation in reference frame use. I explore the structure of that variation and argue that the same

factors are at play as previously reported for the choice of a reference frame in other languages.

In particular, that choice correlates not only with the degree to which the Ground object is

functionally asymmetric, but also with the choice of a particular lexical item used to describe the

spatial relation. Lexical items referring to more natural and functionally prominent asymmetries

(e.g., front/back) are more likely to be associated with the function‐based intrinsic reference

frame than lexical items referring to less prominent and human‐specific asymmetries (e.g.,

left/right).


Conclusion

The study of major dimensions of cross‐linguistic variation at a more fine‐grained level can lead

to insights into the way expressions of different types co‐exist within one language. The same

underlying principles can be seen at work behind different instances of language‐internal

variation. The principles discussed in this paper suggest that (i) in encoding motion events,

speakers estimate how likely directional meaning is to be inferred from context; (ii) in choosing

between more and less conventional expressions, speakers are sensitive to the spatial relation’s

frequency; and (iii) in choosing the reference frame, speakers pay close attention to the

prominence of certain aspects of the spatial relator’s lexical meaning.

The same principles show up as relevant to speakers’ contextual preferences, seemingly

categorical lexical splits, and patterns of language change. They likely reflect general principles of

human spatial cognition, and interact in complex ways with the heterogeneous lexical and

morphosyntactic resources that the particular language has developed.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Nölle—67

Variation in Spatial Language as a Form of Adaptation: An Evolutionary, Experimental Approach

JONAS NÖLLECentre for Language Evolution (CLE)

University of Edinburgh Email: [email protected]

hy do languages differ so dramatically in grammatical and conceptual forms? Usually,

linguistic diversity is attributed to random changes that accumulate and become

conventionalized. However, over the past years an increasing amount of studies emerged

suggesting that linguistic structure can additionally adapt to social, physical or technological

aspects of the external environment (see Lupyan & Dale 2016 in press for a review). This includes

climate or landscape, social factors like population size, language contact and even genetic

predisposal due to ecological adaptation.

Spatial language is arguably one of the most fundamental domains of linguistic practice, as all

humans inhabit a 3D space and therefore need to reference its qualities to interact successfully

(e.g., locations, topology, relations between objects etc.). Different speech communities have

found very different solutions to this problem leading to striking diversity in how motion,

topological or distant relation (“Frames of Reference,” FoR) are expressed (Levinson & Wilkins

2006). This has often been cited as an argument for cultural evolution leading to diversity via

conventionalization (Evans & Levinson 2009). However, since spatial language has also been used

as prime argument for the claim that language restructuring general cognition, differences have

usually been attributed to arbitrary conventions leading to language structure that, in turn,

affects cognition (Haun et al. 2011). Accounts emphasizing the possibility of ecology‐induced

cognitive styles or the environment (e.g., urban vs. rural) affecting the choice in FoR have been

criticized or accused of simplifying the issue (Levinson et al. 2002, Majid 2004). And yet, it

remains counterintuitive that clearly geographically grounded systems should have emerged

independently of the environment: Languages solely relying on the absolute FoR often directly

integrate landmarks from the surrounding environment like slopes in Tzelatal (Levinson & Wilkins

2006) or rivers in Mian (Fedden & Boroditsky). Intuitively, a certain strategy should be more

adaptive in a certain environment and get selected over time. Left/right may not be helpful in a

desert where absolute directions seem more reliable. In a dense jungle or urban area, however,

far sight is disabled; relative/intrinsic strategies might become adequate. Only recently, fieldwork

has begun to crosslinguistically test for these relationships by comparing spatial language

between related/same languages spoken in different environments and distant languages spoken

in similar environments (Palmer 2015, Bohnemeyer 2015).

In order to test for relationships between spatial language and environment, I suggest to

complement this recent correlational approach with experiments that attempt to model the

evolution of spatial language per se, possibly uncovering mechanisms that contribute to the

diversity we can observe. Recently, we conducted an experiment to test whether linguistic

W


conventions that emerge spontaneously in dialogue—the predominant mode of communication

and arguably the starting point for most language change—are affected by the environment. We

used Garrod and Doherty’s (1994) maze game, a collaborative task where dyads have to

communicate about spatial locations in order to guide each other through virtual mazes. The

original experiment showed that in this task‐oriented dialogue, dyads automatically routinize a

specific strategy to describe positions, which would even spread when partners were exchanged

leading to “communities” that adhered to a specific convention. We introduced three conditions

that systematically varied maze layouts (Fig. 1).

(a) (b) (c)

Fig. 1. Mazes from the (a) regular, (b) irregular and (c) stratified condition.

As predicted, different linguistic strategies became routinized in response to these

environmental conditions. Players in the regular condition would rely more on the MATRIX

strategy conceptualizing the maze as a coordinate system (“go to D2”), while players in the

irregular and stratified conditions relied significantly more on figural shapes or horizontal

displacements that were made salient (“Go to the head,” “The switch in the second row, left”).

This suggests that linguistic interactions and routines are not only the result of automatic priming

mechanisms, but also highly sensitive to factors of the shared task environment. Interaction in

specific environments can influence how the very same coordination problem (in this case

communicating spatial positions in the maze) is conceptualized and translated into linguistic

conventions (outcompeting equally valid alternatives).

Following these results, an additional series of experiments is planned to test whether such

mechanisms are not only at work in dialogue, but could constrain emerging communication

systems as such (Nölle, in progress): Subject pairs have to describe spatial positions to each other

to solve the same task in varying immersive, virtual environments, while the use of conventional

language is restricted. They have to develop a novel, graphical communication system.

Manipulating the environment, e.g. by contrasting flat vs. elevated, dense vs. sparse vegetation,

or landmarks vs. none, allows testing whether such differences will systematically affect the

“languages” that emerge and evolve in these settings. This work thus aims at showing that

linguistic adaptation is a possible evolutionary mechanism that contributes to the diversity in

spatial language that we can observe.

References Lupyan, G. & Dale, R. (2016). Trends in Cognitive Sciences, in press.

Levinson, S.C. & Wilkins, D. (2006). Grammars of Space. Cambridge University Press.


Haun, D.B.M., Rapold, C.J., Janzen, G., Levinson, S.C. (2011). Cognition 119: 70–80.

Levinson, S.C., Kita, S., Haun, D.B.M., Rasch, B.H. (2002). Cognition 84: 155–188.

Majid, A., Bowerman, M., Kita, S., Haun, D.B.M., Levinson, S.C. Trends in Cognitive Sciences 8(3): 108–114.

Fedden, S. & Boroditsky, L. (2012). Frontiers in psychology 3.

Palmer, B. (2015). In: Bussa & Polla (eds.), Language structure and environment. London: Benjamins.

Bohnemeyer, J., et al. (2016). Language Dynamics and Change 5(2): 169–201.

Garrod, S. & Doherty, E. (1994). Cognition 53: 81–215.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Palmer–70

The Socio‐Topographic Model: Socioculturally Mediated Responses to Environment Shaping

Universals and Diversity in Spatial Reference

BILL PALMER

Endangered Languages Documentation, Theory, and Application Research Program University of Newcastle


onsiderable diversity in spatial reference across languages is well attested (Levinson 2003; Levinson & Wilkins 2006; Pederson et al. 1998). Nonetheless, universal tendencies can be

detected within this diversity, and salient landscape and other external‐world features seem to play a role in the detail of systems involving absolute Frame of Reference (FoR) (Palmer 2002, 2015), and even in FoR choice (see Majid et al. 2004; Bohnemeyer et al. 2014). However, those aspects of the environment that are perceived as salient vary across cultures, and the nature of the interaction between humans and their environment plays a crucial role, as seen in demographic variation within individual languages in tendencies in FoR choice (e.g. Pederson 1993), and in geocentric versus egocentric strategies more generally (Palmer et al. 2016).

Spatial relations of any type can be expressed using language. However, in perhaps all languages some spatial concepts are lexicalised or expressed in a grammaticized way, while others are relegated to periphrastic expression. These lexicalized and grammaticized expressions are key to understanding the extent to which spatial reference displays universal tendencies, and the extent to which variation is systematic.

Geocentric spatial reference, including the use of absolute FoR, invokes aspects of the external world, suggesting that linguistic systems are responsive to the environment in which a language is spoken (Palmer 2002). This in turn predicts that aspects of systems of spatial reference will correlate with salient aspects of the physical environment. Palmer (2015) formulates this as the Topographic Correspondence Hypothesis (TCH), a tool to test the extent to which linguistic spatial systems correlate to environment in ways that can account for aspects of spatial reference that are universal or vary in systematic ways. To test TCH, Palmer (2015) proposes the Environment Variable Method (EVM), an approach that treats environment as a controlled variable. TCH makes predictions along two parameters: (A) that a single language spoken in diverse environments will display commensurate diversity in spatial reference; and (B) that diverse languages spoken in a single environment will display commensurate similarities in spatial reference. EVM tests (A) by holding the language constant and varying the environment. Prediction (B) is harder to test, because while the environment is to be held constant and the language varied, the environment cannot be held constant to the extent of investigating diverse languages in a single location, as it would be impossible to rule out similarities between languages arising from contact. Instead, language loci that are as similar as possible are to be used.

C


To test TCH and cast light on the relationship between spatial reference and environment, a research team comprising Palmer, Alice Gaby, Jonathon Lum and Jonathan Schlossberg are investigating spatial reference in languages spoken in the topographic environment of the atoll, in a three‐year project funded by the Australian Research Council. Atolls are an unusual environment for human habitation, comprising narrow strips of land around a central lagoon. A field‐based preliminary study of spatial reference in atoll‐based languages (Palmer 2007) found similarities in spatial systems in four languages, including an atoll‐specific lagoonside‐oceanside axis. We are testing TCH in atoll‐based languages by investigating spatial reference in Marshallese (Oceanic, Marshall Islands) and Dhivehi (Indo‐Aryan, Maldives). Following EVM, a baseline language‐environment pairing of Marshallese spoken on an atoll is compared: along one parameter with Marshallese spoken on a non‐atoll island and in urban Arkansas US; and along the other with Dhivehi spoken on an atoll topographically similar to the Marshallese site. Identical experimental elicitation techniques including several newly devised experiments were used in all locations to ensure comparability of data, and data was subject to quantitative analysis.

The study’s findings weakly support TCH. For example, both languages employ a landward‐seaward axis correlating to the boundary between land and sea. However, in Marshallese this is only used at sea, while in Dhivehi it is used on land, with only one term also used at sea. Further, the distinction between an island’s lagoonside and oceanside is lexicalised in both languages, but in Dhivehi these terms cannot participate in grammaticized constructions, while in Marshallese they frequently do. Some of our quantitative findings also support TCH. In atoll Marshallese, for example, 72% of location descriptions were geocentric or cardinal, and only 15% egocentric or intrinsic, while in urban Arkansas only 5% were geocentric or cardinal and 71% were egocentric or intrinsic, supporting earlier findings of an urban dispreference for absolute/geocentric reference.

However, our quantitative analysis revealed a more nuanced picture than TCH alone allows. While both languages provide a similar range of strategies for spatial reference, strategy preference varies significantly between the languages. For example, in atoll Marshallese, 72% of location descriptions were geocentric or cardinal and only 11% involved intrinsic FoR, while in environmentally similar Dhivehi, only 25% of location descriptions were geocentric or cardinal and 35% were intrinsic. Even more significantly, our findings introduced a crucial caveat to TCH: social and cultural factors mediate between language and environment, such that a simple predictable relationship between the two does not exist. Lexicalized and grammaticized systems of spatial reference may correlate to aspects of the environment, but the extent to which they do, and which aspects of the environment are invoked, varies on the basis of both affordance, and degree and nature of cultural interaction with the environment. For example, in Dhivehi fishing communities, 77% of orientation descriptions were geocentric or cardinal, while in non‐fishing communities, engaged primarily in white collar work, only 35% were. Significant variation was also observed on the basis of gender and age.

In response to these findings we have formulated the Socio‐Topographic Model (STM) (Palmer et al. 2016). Major environmental features tend to be salient to humans and appear


to play a role in constructing conceptual representations of space that then interact with linguistic spatial expressions. However, cultural and social factors, as well as the affordances of the environment itself, mediate in the relationship between humans and landscape. STM models the interplay of the physical environment of the language locus, sociocultural interaction with the environment, and the linguistic repertoire available to speakers (Figure 1). Socio‐Topography is defined in terms of: natural topography (broadly construed, including path of sun, prevailing winds etc); the built environment; affordance; and sociocultural interaction with the natural and built environment. Socio‐Topography is culturally “constructed”: Humans modify their environment, and conceptualise existing topography in terms of use, associations and meanings attached to it. Consequently, elements of the local landscape that are not attended to by some cultures will be prominent to others, and factors such as scale may be attended to by some cultures but less so by others.

Figure 1: The Socio‐Topographic Model

References Bohnemeyer, J, K.T. Donelson, R.E. Tucker, E. Benedicto, A.C. Garza, A. Eggleston, et al. 2014. The cultural transmission of spatial cognition: evidence from a large‐scale study. Cogsci 2014 Proceedings. 212–217 Levinson, S.C. 2003. Space in language and cognition: Explorations in cognitive diversity. Cambridge: CUP.

Levinson, S.C. & D. Wilkins eds. 2006. Grammars of space: Explorations in cognitive diversity. Cambridge: CUP. Majid, A., M. Bowerman, S. Kita, D.B.M. Haun & S.C. Levinson. 2004. Can language restructure cognition? The case for space. Trends in Cognitive Sciences 8(3): 108–114.

Palmer, B. 2002. Absolute spatial reference and the grammaticalisation of perceptually salient phenomena. In G. Bennardo ed. Representing space in Oceania: culture in language and mind. Canberra: Pacific Linguistics.

Palmer, B., 2007. Pointing at the lagoon: directional terms in Oceanic atoll‐based languages. In J. Siegel, J. Lynch & D. Eades eds. Language description, history and development. London: Benjamins.

Palmer, B. 2015. Topography in language. Absolute Frame of Reference and the Topographic Correspondence Hypothesis. In R. de Busser & R. LaPolla eds. Language structure and environment. London: Benjamins.

Palmer, B., A. Gaby, J. Lum & J. Schlossberg. 2016. Topography and frame of reference in the threatened ecological niche of the atoll. Conference paper presented at Geographic grounding. Place, direction and landscape in the grammars of the world. Copenhagen, May 2016.

Pederson, E. 1993. Geographic and manipulable space in two Tamil linguistic systems. In A.U. Frank & I. Campari eds. Spatial information theory: A theoretical basis for GIS. Berlin: Springer. 294–311.

Pederson, E., E. Danziger, D. Wilkins, S.C. Levinson, S. Kita & G. Senft. 1998. Semantic typology and spatial conceptualization. Language 74/3:557–589.

environment • natural • built

Cultural values / practices

• present & historicalinteractions with environment

• language history • conceptualization of

environment interms of the above

language use

linguistic repertoire

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Peacock—73


CANDACE PEACOCKHuman Spatial Cognition Lab University of California, Davis


patial referencing is unique to humans, as humans have the capacity to identify and

communicate the location of landmarks via the use of language. Universally, it is a necessity

that animals navigate to find landmarks important to their survival, such as food and shelter.

However, most animals rely on bottom‐up processes (i.e., their senses) to locate these landmarks

because they do not have access to language as a means of communication. For example, a rat

relies on its sense of smell to locate food but it cannot verbalize where the food source is located

to other rats. Although humans share the ability to use their senses when encoding an

environment, they also have access to top‐down processing by using language (or maps) to

communicate important geographical information to each other. Traditional navigation studies

focus on how rats and humans use their senses to spatially encode an environment (e.g., vision).

These studies depend on participants using bottom‐up processes to learn the spatial layouts of

environments. However, because humans have access to language, there must be differences in

how humans code an environment compared to rats. Thus, it is integral to study the role

language plays in human navigation.

Language is multifaceted, in that different languages rely on varying cues when describing

location. For example, Brown and Levinson (1993) found that the frame of reference dominant in

a language changed how someone who spoke that respective language coded an environment

and that people who spoke different languages coded environments differently because their

frame of reference was different. Similarly, Munnich, Landau, & Dosher (2001) found that people

who spoke different languages coded spatial representations differently. However, when these

same people learned these spatial representations non‐linguistically, there were no differences in

encoding. These studies show that our dominant language shapes how we locate landmarks in

the world. Because there are so many languages, there is divergence in how we linguistically

encode and reference landmarks. As such, some researchers have focused on how animals, who

universally use their senses, locate themselves and landmarks in the world. All animals, including

humans, must use their senses to locate themselves in the world. As such, some researchers

have opted to study spatial memory in both humans and model organisms. Rats have been a

preferred model organism of spatial memory research because of their similar neurophysiology

to humans (e.g., the demonstration of place cells in both humans and rats, see Ekstrom et al.

2015). When rats locate a landmark or reorient themselves, they take many things into account,

including the geometry and features of their environment (Cheng 1986). For example, in studies

where a rat must locate a trained corner in a rectangular enclosure, it can use wall length

(geometry) and wall color (feature) to identify the trained corner. There is an ongoing debate

that surrounds whether rats and other animals focus more on the geometry or the features of an

S


environment. Many studies support that rats use geometry because when features of an

environment are removed, rats will consistently return to a desired corner of the environment

(Cheng 1986; Margules & Gallistel, 1988; Benhamou & Poucet, 1998). Yet how linguistic codes

affect geometry versus features remains unknown and understudied.

Developmental studies attempt to contribute to the feature versus geometry debate as well.

These studies show that children generally use geometry to locate a target (Learmonth,

Newcombe, & Huttenlocher, 2001; Hermer & Spelke, 1994). There have even been cases where

children neglect to use features at all, despite remembering them in recognition memory tasks

(Hermer & Spelke, 1994). When adults are studied, however, we see more ambiguity in whether

features and geometry are used. This is partially due to language: when adults undergo a verbal

shadowing task, it is more difficult for them to use geometry to differentiate corners

(HermerVazquez, Spelke, & Katsnelson, 1999). These results imply that language is integral for

adults to orient themselves and to locate landmarks, which is where my experiments come into

play. Yet exactly how this affects top‐down coded preferences for geometry vs. features is not

known.

My experiments propose to disentangle the role of language in how human adults orient

themselves in the world with respect to geometry and features. Instead of focusing solely on

spatial memory and bottom‐up processes, these experiments attempt to see how language is

used as a tool for humans to locate themselves spatially in the world. I propose to use a simple

paradigm, where participants are instructed to use either the geometry or the features of

uniform environments (each environment has the same geometry and features) to see how

language affects how geometry and features are used. As I will be instructing participants

whether to use the geometry or features of an environment, this experiment will also inform how

verbal communication affects subsequent spatial encoding.

There are three potential interesting outcomes from this project. One is that, depending on

the linguistic code (i.e., pay attention to the square or red wall), pointing accuracy is better

depending for the attended (instructed) component. This would argue for the importance of top

down coding irrespective of feature versus geometry. Other potential outcomes are that

pointing accuracy is better for the geometry or feature regardless of linguistic codes. These two

potential outcomes would suggest that features or geometry are bound in a more bottom up

fashion and are not strong activated by top down cues. All of three of these outcomes, though,

would be important to understanding how linguistic codes can interact with geometrical and

feature binding when learning a spatial environment. These results will be ready to disseminate

by start of the Universals and Variation in Spatial Referencing across Cultures and Languages. I

am excited to receive input for future experiments to further understand how humans use

natural language to locate themselves in the world, as I am a new graduate student in the spatial

memory field.

References Benhamou, S., & Poucet, B. (1998). Landmark use by navigating rats (Rattus norvegicus) contrasting geometric and featural information. Journal of Comparative Psychology 112(3): 317.


Brown, P., & Levinson, S. (1993). Linguistic and nonlinguistic coding of spatial arrays: Explorations in Mayan cognition. Nijmegen: Cognitive Anthropology Research Group, Max Planck Institute for Psycholinguistics.

Cheng, K. (1986). A purely geometric module in the rat's spatial representation. Cognition 23(2): 149–178.

Ekstrom A.D. (2015). Why vision is important to how we navigate. Hippocampus 25: 731–735.

Hermer, L., & Spelke, E. S. (1994). A geometric process for spatial reorientation in young children. Nature.

Hermer‐Vazquez, L., Spelke, E. S., & Katsnelson, A. S. (1999). Sources of flexibility in human cognition: Dual‐task studies of space and language.Cognitive psychology 39(1): 3–36.

Learmonth, A. E., Newcombe, N. S., & Huttenlocher, J. (2001). Toddlers’ use of metric information and landmarks to reorient. Journal of experimental child psychology 80(3): 225–244.

Margules, J., & Gallistel, C. R. (1988). Heading in the rat: Determination by environmental shape. Animal Learning & Behavior 16(4): 404–410.

Munnich, E., Landau, B., & Dosher, B. A. (2001). Spatial language and spatial representation: A cross‐linguistic comparison. Cognition, 81(3), 171‐208. doi:10.1016/S0010‐0277(01)001275

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Pederson—76

Some Concerns about Methodology on Spatial Representation

ERIC PEDERSONDiscourse Lab, Department of Linguistics

University of Oregon Email: [email protected]

would like to take the opportunity of this Specialist Meeting to raise a number of

methodological concerns which have troubled me over the years. Many of these worries are

not new to the field, but they remain largely unresolved in much of the work conducted today

even while some of these concerns are strong enough to potentially invalidate the

interpretations of the results. As my own work has concerned the relationship between linguistic

representation and performance on ostensibly nonlinguistic tasks, I would address each of these

in turn.

Whether the goal is typology or comparison with non‐linguistic performance, linguistic

elicitation presumes relevant variation. As any second language learner can attest, one is biased

to assume that translation equivalents across languages are uniformly equivalent in the absence

of contrary indication. Unfortunately, when working with other languages, it is often years of

work, if ever, before one discovers that terms across two languages which seem to have the

same functional equivalence in most contexts, nonetheless may have substantially different

underlying semantics. This wouldn’t be a problem if language elicitation did not rely so heavily on

translation, but avoiding translation is a more difficult problem than just working monolingually.

For example, I was occasionally corrected in my use of the Tamil deictic verbs “go” and “come.” It

turned out that the translations are largely accurate, but the Tamils I was working with tended to

use “come” as the more generic motion verb (e.g. for describing motion passing in front of the

speaker) with “go” being more specialized for motion away from the speaker. I was sometimes

corrected by Tamils who were fluent in South Indian English and they were puzzled by my

confusion because they used English “go” and “come” in the same way as for their native Tamil—

believing, like me, that these forms were naturally equivalent across all situations.

As another example, consider the problem of translating prepositions from one language to

another. Without knowing the exact extensional set for the form in one language it is impossible

to predict which would be the appropriate spatial term in another language. Indeed, the domain

of Germanic prepositions is one of the last domains to be mastered even, e.g., by German

speakers fluent in English. Even for analysis of naturally occurring language production, we

cannot ever be certain of what an utterance can mean without translation unless we know not

only the exact context of use. A common solution to this problem is to use an “etic” stimuli set

such as the Topological Relations Picture Series or the “Tomato films” when comparing languages.

However, even this can only be a solution if we are confident of the particular intended construal

that the speaker is representing. (e.g., Is the ball in or on the field?) It is a rare study which

explains what method was used to ensure clear and consistent interpretation of that which is

being described.

I


A further problem is the nature of the usage of a form. Obviously, we can only compare

forms when we can equate or at least identify the context of use. But what constitutes context

for elicitation purposes? Are two “exact” translation equivalents identical if one is broadly used in

one language, but only used in the other language in very specific circumstances? Can the

semantics of two forms be even remotely similar if the one form has a number of

paradigmatically contrastive alternatives, while the other form alternates with only another

broadly used form? As a trivial example, English at is often used as a translation of the locative

case in many languages. However, the use of at is remarkably specific and nuanced in ways quite

distinct from the more general locative case in most languages.

This of course, leads also to the problem of translations of instructions to participants in non‐

linguistic tasks as well. I have been in countless hours of discussion with many researchers on

how one might reliably translate concepts in experimental instructions such as “same” and

“similar.” If one is uncertain how such translations are being interpreted in a task, then we must

honestly say that that we don’t actually know what task the participants believe themselves to be

engaged in.

As any experimentalist knows, much hinges on the interpretation of the task far beyond the

problem of translating instructions. Perhaps no experimental paradigm brings this more to the

fore than triads tasks in which participants are asked to group a pivot item with either of two

target items—each of which clearly relate to the pivot in distinct but valid ways. In debriefing,

participants commonly state that they alternate solutions, guess what the experimenter is really

asking for, or just pick one solution when they feel that the other was an equally acceptable

answer for them. (This is quite distinct from, e.g., AXB tasks in which the targets vary along a

single perceptual dimension such as voice onset in a phonetics perception experiment.) One

might think that this fundamental problem with triads tasks would render them nearly useless,

yet they remain one of the more popular experimental designs especially in the challenging area

of comparing linguistic and non‐linguistic representations.

I will conclude with one final example: the Animals in a Row task, described in Pederson et al.

1998 and elsewhere. This task was a memory task in which subjects were presented with three

animals (of an available four) in a line transverse to the participant on a table or mat in front of

them. They were then taken to another room or substantially away from the first table and

simply asked to rebuild the line of animals in the second location. The original design combined

two critical features: First, the participants were presented with tables in a 180‐degree rotation

from the participants such that the orientation of the line upon reconstruction could be

interpreted as consistent with either an egocentric or a geocentric memory encoding of the

original display. Importantly, the 180‐degree rotation was a simple consequence of traveling to a

different and remote table and as such was not a particularly salient aspect of the experiment for

the participants. There have been a number of quasi‐replications of this original design which

simply spun participants around 180 degrees making this rotation quite a prominent part of the

design for them and presumably inviting an interpretation of this feature.


The second critical feature is that the dependent variable was not the order of the animals,

but rather the facing orientation of the animals when the line was rebuilt. Participants clearly

understood the task as about the selection and order of the animals and not particularly about

the orientation. (I have recordings of subjects repeating “horse, pig, sheep; horse pig sheep” as

they march from one room to the other, but no one adds “all facing to the East.” In other words,

the dependent variable is covertly embedded in an otherwise transparent task, which reasonably

ensures that the response in orientation is essentially an unreflective or “natural” response,

presumably indicating an underlying representation of the orientation rather than an attempt to

guess the purpose of the design. It is remarkable how many pseudo‐replications of the Animals in

a Row task have changed these two features—without explicit motivation—and received

unsurprisingly different results. Witness the exchange between Li & Gleitman and Levinson et

alibi. (2002).

At the Specialist Meeting, I would like to discuss these methodological issues and work

toward a more common understanding to improve our future empirical designs.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Pérez Báez—79

Interactions between Strategies of Spatial Referencing:A Case Study on Diidxazá Spatial Descriptions

GABRIELA PÉREZ BÁEZCurator of Linguistics

National Museum of Natural History Smithsonian Institution


he present paper provides a synthesis of work carried out on spatial descriptions in Diidxazá

(Juchitán/Isthmus Zapotec, Otomanguean). The research takes a semantic typology

approach for the analysis of descriptions of topological and projective relations (frames of

reference, FoRs). Data from elicitation and from linguistic and non‐linguistic experimental tasks

provides a comprehensive view of the strategies used by Diidxazá speakers to organize space and

describe the location of objects in it.1 The analysis of topological descriptions provides initial data

on the use of meronyms, mainly body part‐derived meronyms, to describe the location of a

figure in relation to a part of a ground.2 Further investigation on the mechanisms enabling the

semantic extension of body part‐derived meronyms and the extent of their use in spatial

descriptions reveals an interface between a meronymic system of spatial relators and FoR

preferences. This provides a wide‐ranging understanding of spatial description strategies as

implemented by Diidxazá speakers and of their division of labor. Further, this ramifies into new

lines of inquiry such as the analysis of unexpected functions of otherwise disprefered strategies,

as well as language contact phenomena.

The use of body part‐derived meronyms for spatial description is well acknowledged as a

noteworthy feature of Mesoamerican languages (Campbell, Kaufman & Smith‐Stark 1986 inter

alia). This has been documented for Otomanguean languages including Zapotec languages

(MacLaury 1989, Lillehaugen 2006, Pérez Báez 2012; see also chapters in Lillehaugen and

Sonnenschein (eds.) 2012), Chalcatongo Mixtec (Brugman 1983, Brugman & Macaulay 1986) and

Copala Triqui (Hollenbach 1987, 1988); for the Mayan languages Tzeltal (Stross 1976, Levinson

1994) and Tzotzil (de León 1992); as well as for Tarascan (Friedrich 1969, 1970, 1971), Totonac

(Levy 1992, 2006) and Cora (Casad 1982). Pérez Báez 2012 and forthcoming, explore the process

of semantic extension of Diidxazá body part‐derived meronyms. Data from various elicitation

tasks conducted between 2003 and 2009 with over 20 native Diidxazá speakers analyzed within

the framework provided by the Structure Mapping Theory (Gentner 1983, inter alia) provide

evidence in support of an analogy‐based process of semantic extension compatible with the

1The experimental tasks include a novel objects part identification task developed by the Spatial Language and Cognition in Mesoamerica project (https://www.acsu.buffalo.edu/~jb77/MesoSpaceManual2008.pdf), and the Ball and Chair and New Animals in a Row tasks, also designed by the MesoSpace project after the Men and Tree and Animals in a Row tasks developed by the Cognitive Anthropology Research Group at the Max Planck Institute for Psycholinguistics (Danziger 1992, Levinson and Schmitt, 1993).

2Figure and ground are understood here as per Talmy 2000:184).

T


process proposed, for instance, in MacLaury 1989 for Ayoquesco Zapotec. Additional elicitation

conducted notably with tools as stimuli suggests that the analogy‐based process does not

exclude an algorithm‐based process such as the one proposed in Levinson 1994 for Tzeltal Maya.

Beyond the description of the process of semantic extension of body part‐derived meronyms,

the analysis of these relators in spatial description uncovers the interaction between the

meronymic system and a FoRs system. Data collected through referential communication tasks

conducted in the field with 12 native speakers of Diidxaxá show that the relative FoR is clearly

disprefered (Pérez Báez 2011). In orientation descriptions, the relative FoR was not used at all. In

descriptions of figure‐ground arrays, the relative FoR was used in only 3% of the documented

descriptions. In a non‐linguistic task administered to 19 native Diidxazá speakers, only one

participant in one trial produced a response consistent with the relative (or the direct) FoR. This

bias prompts the question as to what function(s) such a constrained FoR might have when used.

The notion that speakers of some Mesoamerican languages exhibit a bias against relative

FoRs has been discussed in a number of works (Brown and Levinson, 1993; Levinson, 1996, 2003;

Brown and Levinson, 2009; Pérez Báez 2011, Polian and Bohnemeyer, 2011, Hernández‐Green et

al 2011). However, there are no studies, to my knowledge, whether their constrained use might

correlate with specialized function(s). The Diidxazá data suggests that the relative FoR serves an

ambiguity resolution function: in cases where a meronym—used to refer to the part of the

ground in relation to which the figure is to be located—might refer to more than one part (as

when referring to one of the two sides of a chair, the relative FoR serves to identify the correct

part. The relative FoR is not the only disambiguiating strategy. However, it is only in this context

that the relative FoR was used.

The relative FoR and the meronymic system interact in a similar way (Pérez Báez,

forthcoming). Body part‐derived meronyms in Diidxazá can be assigned to objects even in cases

where an object might have few or no discernable parts, for example a sphere. In these cases,

the relative FoR enables a structure mapping between an abstraction of the human body –and

not the actual body– in canonical vertical position as the source domain and, say, the sphere as

the target domain (Pérez Báez forthcoming). This mapping can only be done in the context of a

projection from the observer/speaker’s perspective. In other words, on the basis of a relative FoR.

Further, the Diidxazá data shows that the relative FoRs are generally encoded by Spanish loan

words referring to the ends of the sagittal and transversal axes rather than by native Diidxazá

words. This suggests that the relative FoR has a particular function linked to the use of Spanish

loan words. This finding points to yet another line of inquiry that has received little attention:

spatial referencing in language contact situations. Hernández et al 2011 report a similar marked

function of the Spanish word lado “side” in relation to the use of the relative FoR in San Ildefonso

Tultepec Otomi (Otomanguean). McComsey 2015 reports on FoR use among speakers of Diidxazá,

Spanish and both. Bohnemeyer et al 2015 report on contact diffusion of FoR preferences. Yet,

reports on functions of linguistic spatial referencing strategies associated with language contact

phenomena are lacking in the literature.


In sum, this paper demonstrates the value of comprehensive typology‐based analysis of data

collected in the field through a combination of elicitation and experimental tasks, both linguistic

and non‐linguistic. Further, it advocates for a systemic analysis of language in spatial referencing

to uncover new lines of inquiry that may yield a more broad‐reaching understanding of the

relation between a variety of spatial referencing strategies as well as the relation between

language, cognition and social and cultural contexts.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Ragni–82

Universals and Variation in Spatial Referencing MARCO RAGNI

Cognitive Computation Lab University of Freiburg

Email: [email protected]‐freiburg.de

The interplay and conflict of spatial referencing in cultures by language and internal (preferred)

mental models: Are preferred mental models universal?

patial referencing in cognitive agents such as humans often depends on internal mental representations that are formed by experiences, the formal structure of problems, cognitive

limitations, and the interaction with other cognitive agents by using language. Such mental representations can be more visually (e.g., a landmark) or purely relational (e.g., an orientational relation between spatial objects on a map). Recent results in cognitive psychology show that using relational expressions may trigger more visual features that in turn can lead to an impedance effect (Knauff, 2013). This visual impedance effect has been often experimentally replicated in western cultures especially in English/German. A possible generalization of this probably universal effect (some explanations aim at a memory related source of additional activation) is still an open question. Currently, we are implementing this experiment in Chinese and aim to test it through Internet based experiments and to compare it with experimental results from a German sample.

To understand how such internal representations (that are influenced both by more universal features like the internal cognitive architecture and by potentially differing variable features that may depend on different cultural influences) we showed that there is a difference in topological relations (such as part of, or contained) between German and Mongolian students (Knauff & Ragni, 2011). Although topological relational information might be considered cross‐culturally universal, we did find a difference in the variation of the mental models. That clearly hints at non‐universal processes. On the other hand, the preferred mental model was stable across the cultures. These preferred mental models are formed based on spatial principles that apply for information that contains an implicit order (Ragni & Knauff, 2013). They allow, in fact, to explain systematic errors in representation and reasoning. It is important to consider different reference frames that in turn can imply different cognitive complexities. The language used was in its nature qualitative and formal features are known (Ragni & Wölfl, 2005). The advantage of a formalization of these natural language approaches was that even problems in robotic navigation could be solved (Moratz & Ragni, 2008) and so they seem to be possibly universally applicably and in some sense effective. But again, these findings are influenced by a western‐style approach. It is still an open question if these representations extend to other representations that are used in other cultures.

By using transcranial magnetic stimulation (TMS) we were able to demonstrate that in uncertain relational reasoning the construction and manipulation can be neurally localized in the posterior parietal cortex (e.g., Ragni et al., 2016).

S

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Ragni–83

The mental representations formed in cases of uncertain spatial information is especially interesting as this case may apply most often in everyday spatial descriptions, e.g., for navigation. Spatial relational information can depend on the domain it refers to: There can be small‐scale spaces (often table‐top scenarios) or large‐scale spaces that are connected to cardinal directions, for instance. Can principles for mental construction that are found for one domain easily be transferred to another one? This is another open question.

Finally, a recent analysis indicates that verbalizing spatial mental models indicate at “natural” relations as opposed to formal representations (Tenbrink & Ragni, 2012).

My core interest aims at a possible dissociation between the influence of spatial language (across cultures) and the internally built mental representations that are triggered by internal structural parts often considered to be universal. Taking these different levels together can be a necessary step towards a precise computational and cognitive theory for the human relational representation and reasoning and it is extended towards culture‐specific and universal cognitive processing principles. A related question is: How can such universal differences in spatial cognition be tested across language and cultures? This may lead to an identification of potential benchmark problems that could be focused on to test the structural/language distinction.

References Knauff, M. (2013). Space to reason: A spatial theory of human thought. MIT Press.

Knauff, M. and Ragni, M. (2011). Cross‐cultural preferences in spatial reasoning. Journal of Cognition and Culture 11: 1–21.

Moratz, R. and Ragni, M. (2008). Qualitative spatial reasoning about relative point position. Journal of Visual Languages & Computing 19(1): 75–98.

Ragni, M. and Knauff, M. (2013). A theory and a computational model of spatial reasoning with preferred mental models. Psychological Review 120(3): 561–588.

Ragni, M., Franzmeier, I., Maier, S., & Knauff, M. (2016). Uncertain relational reasoning in the parietal cortex. Brain & Cognition 104: 72–81.

Ragni, M. and Wölfl, S. (2005). Temporalizing spatial calculi: On generalized neighborhood graphs. In Furbach, U., editor, KI 2005: Advances in Artificial Intelligence, Proceedings of the 28th Annual German Conference on

AI, pages 64–78, Berlin. Springer.

Tenbrink, T. and Ragni, M. (2012). Linguistic principles for spatial relational reasoning. In Stachniss, C., Schill, K., and Uttal, D. H., editors, Spatial Cognition VIII—International Conference, Spatial Cognition 2012, Kloster Seeon, Germany, August 31–September 3, 2012. Proceedings, Lecture Notes in Computer Science, pages 279–298. Springer.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Regier–84

Universals and Variation in Spatial Referencing TERRY REGIER

Department of Linguistics and Cognitive Science Director, Cognitive Science Program University of California, Berkeley Email: [email protected]

approach cross‐language universals and variation, in the spatial semantic domain as in others, by asking what cognitive and communicative forces may give rise to observed cross‐language

patterns. My colleagues and I pursue this question in large part using computational models and methods.

Recent work along these lines in my lab has focused on topological spatial relations, of the sort investigated by Bowerman, Pederson, Levinson, and their colleagues at the MPI Nijmegen. Earlier well‐known analyses of these data have suggested a general picture of wide but constrained variation in spatial terms across languages—exemplified for instance by the in‐on continuum noted by Bowerman and Pederson. They identified a continuum of spatial relations ranging from a prototypical English “in” relation at one end (an apple in a bowl) to a prototypical English “on” relation at the other end (a cup on a table), and found that although languages differ in how and where they partition this continuum into spatial categories, the resulting spatial categories always pick out connected regions of the continuum—yielding, in effect, a semantic map for a central part of the topological spatial domain. Our group has generalized this finding to a much broader range of spatial scenes, using an algorithm for automatically inferring a semantic map from cross‐language data. Using this algorithm, we produced a novel semantic map of topological spatial relations over a larger set of stimuli than those considered by Bowerman and Pederson—and separately automatically produced the semantic map for indefinite pronouns that Haspelmath had produced by hand.

Research of this sort, based on cross‐language semantic data, has answered some important questions: it has allowed researchers to specify descriptive generalizations over cross‐language data, and to infer apparently underlying universal semantic structure. However, there is another relevant question that appears to require additional sorts of data as well: can one firmly link findings in the semantic typology of spatial relations to independently assessed non‐linguistic forces, such as those of cognition and communication? Our group has been seeking such a link.

We have been computationally testing a hypothesis that is rooted in the functionalist tradition. That hypothesis holds that the wide but constrained variation seen in spatial semantic systems across languages may reflect a functional need for efficient communication: a need to communicate informatively, but at the same time simply—that is, with minimal expenditure of cognitive resources. These two forces trade off against each other: a fine‐grained semantic system that partitions a domain using many distinct terms is highly informative in that it allows precise communication; but because it contains many terms, it is complex, not simple. In contrast, a coarse‐grained system is comparatively simple, but does not support precise,

I


informative communication. We have been exploring the proposal that semantic systems across languages navigate a near‐optimal trade‐off between these two opposing forces, and thus achieve efficient communication. Concretely, we predict that semantic systems will strongly tend to be nearly as informative as possible for their level of complexity, and nearly as simple as possible for their level of informativeness. On this view, different semantic systems may constitute different language‐specific solutions to the shared functional goal of efficiency in communication. Our group has found support for this idea in the domains of color, kinship, spatial relations, number, and artifacts. Here, I briefly sketch our findings in the spatial domain, and highlight some important questions left open for discussion and future research.

Testing the efficiency proposal in the spatial domain requires: (1) a cross‐language sample of spatial semantic systems, (2) an independently assessed cognitive account of the spatial domain, and (3) a means to test the communicative efficiency of the semantic systems in (1) relative to the cognitive structure identified in (2).

Language sample. We have worked with a language sample comprising nine languages examined in an earlier study by Levinson and colleagues (Basque, Dutch, Ewe, Lao, Lavukaleve, Tiriyó, Trumai, Yélî‐Dnye, and Yukatek), supplemented by two languages to which we had access: English, and Maijɨki, an under‐documented language of Peruvian Amazonia which is being studied by Lev Michael’s group at UC Berkeley. We thank our colleagues at the MPI and at Berkeley for providing access to these valuable data. For each of these languages, we considered naming data collected relative to the Topological Relations Picture Series or TRPS.

Cognitive characterization of the domain. We independently assessed the cognitive structure of this domain by asking speakers of English and Dutch to sort the TRPS stimuli into piles based on the similarity of the spatial relations portrayed. The resulting pile‐sorts varied widely within language, but the overall similarity structure of the domain as revealed by the pile‐sorts was broadly similar across speakers of the two languages. Still, the pile sorts did reflect the sorter’s native language to a limited extent—an interesting observation that deserves discussion in its own right and that we have pursued in a separate line of work. For present purposes however we approximated a presumed universal conceptual similarity space by averaging together the similarity structure revealed in pile sorts by speakers of these two languages. Subsequent pilesort investigations with speakers of other languages, including non‐Indo‐European languages, have revealed much the same similarity structure, so we are reasonably comfortable assuming it as an approximation to a universal space.

Testing the efficiency hypothesis. We computationally assessed the informativeness of each language’s spatial system, and compared that to the informativeness of a large number of hypothetical semantic systems, all of which had the same complexity (number of spatial terms), and the same number of spatial relations per term, as the target language. These hypothetical systems were constructed by random graph traversal of the spatial semantic map mentioned above. Informativeness was defined as the extent to which a given system supports accurate mental reconstruction by a listener of a speaker’s intended spatial meaning; accuracy was


measured in terms of the empirically derived conceptual similarity space specified above. We found that for each language in our sample, the spatial semantic system of that language was more informative than almost all hypothetical systems considered—suggesting that these attested systems are each near‐optimally informative about spatial meaning, given their level of complexity, relative to this comparison set. These findings mirror analogous results from other semantic domains such as color, kinship, and number. Our results are consistent with the view that spatial semantic categories across languages may adapt under functional pressure for efficient, informative communication.

These findings suggest certain answers, but they also raise questions. Does this account generalize to other languages? What exactly is the process by which categories (hypothetically) adapt themselves to functional needs? To what extent are communicative needs themselves culture‐specific vs. universal—and to what extent do semantic systems reflect culture‐specific communicative needs? Finally, what is the detailed character of the underlying universal spatial conceptual space, if indeed such a thing exists? We have assumed its existence and approximated it using similarity judgments—but is a more principled and firmer cognitive foundation possible? Our ongoing work is exploring some of these issues.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Rybka—87

Toward Universals and Variation:A Comparative Analysis of Noun Categorization in Spatial

Expressions

KONRAD RYBKAUniversity of California, Berkeley Email: [email protected]

ognitive geography asks the question whether “tabletop” entities (e.g. bowls, apples, tables)

are cognized differently from entities on the geographic scale (e.g. mountains, villages,

rivers) (e.g. Mark 1993; Smith & Mark 2001, 1999). Cognitive geographers such as David Mark

and colleagues suggest that the answer is YES. They claim that geographic entities “are tied

intrinsically to space in such a way that they inherit from space many of its structural properties”

(Smith and Mark 1999: 248). They go on to assert that these “structural properties” may affect

the way such entities are categorized. Independently of cognitive geography, a few semanticians,

including Lyons, have also suggested an ontological disparity between geographic and tabletop

entities. Lyons and Whorf foresaw, for instance, that the linguistic encoding of the two types of

entities could be different (e.g. Lyons 1977; Whorf 1945; Landau & Jackendoff 1993). Keeping

these theoretical points in mind, together with the participants of the meeting I would like to

look at linguistic data from three unrelated languages: Lokono (Arawakan, Suriname), Makalero

(Papuan, East Timor), and Marquesan (Oceanic, French Polynesia). The data show that in these

languages geographic and tabletop entities are grouped into distinct linguistic categories.

Importantly, spatial language—the locative phrase itself in fact—is the locus of the distinction in

question. As such, the analysis of this grammatical phenomenon offers us insights into the types

of parameters that are relevant to the cross‐linguistic encoding of spatial relations and the spatial

cognition in general.

Zooming in on the data, we observe that the three languages group nouns into two distinct

grammatical categories defined on the basis of the locative marking they receive in a spatial

expression. Let us first have a look at the central concepts of spatial language. In a spatial

description there are three indispensable elements: the Figure—the entity to be located, the

Ground, the entity with respect to which the Figure is located, and the spatial relation that holds

between the Figure and Ground. On closer inspection, the spatial relation can be split into two

elements: configuration and directionality (Lestrade 2010; Jackendoff 1990; Talmy 2000).

Configurational elements encode the spatial relation that holds between the Figure and Ground.

There are topological, relative, intrinsic, and absolute types of spatial relations. It is here that

languages show the greatest variation of spatial forms and meanings. Directionality in turn

encodes the change of configuration over time, and has only three primary distinctions:

(1) Location: the absence of change of configuration (2) Goal: the change into a configuration (3) Source: the change out of a configuration.

C


Lestrade (2010) claims that the three distinctions are universal in nature, although languages

show quite some variation in how they express them linguistically. Let us illustrate the difference

between configuration and directionality on English examples:

(1) Location: The diver is at 50 meters under the sea level. (2) Goal: The diver ascends to 50 meters under the sea level. (3) Source: The diver descended from 50 meters under the sea level.

In all three examples above diver is the Figure and the sea level is the Ground. The

configurational element under tells us where the Figure is with respect to the Ground. The

additional modifier 50 meters makes it more specific. The Location directionality element at in (1)

indicates that the Figure is at the Ground. When we change to Goal directionality, the element to

signals that the Figure moves into the configuration. When we change to the Source

directionality, the element from signals that the Figure moves out of a configuration.

Interestingly too, while configuration may be specified or not (try removing 50 meters under), a

directional choice is always required. The grammatical distinction that we will look at manifests

itself only in the directionality component of the spatial expression, not in the configurational

component. Let us look at examples from Lokono, in which there are two markers of Location

directionality, exemplified below:

(4) Dayo bithi−ka=de my.mother LOC−PFV=1SG.So “I am at my mother’s.”

(5) Kasuporhi−n−ka=de Cassipora−LOC−PFV=1SG.So “I am at Cassipora.”

In both (4) and (5), the Figure is expressed by the enclitic =de, preceded by the perfective suffix –

ka which is necessary to form a complete predicate. In both (4) and (5), there is no

configurational element, which means that the relation is unspecified. In (4), the Ground is

expressed by dayo “my mother” and the Location directionality element is bithi. In (5), the

Ground is expressed by a place name Kasuporhi and the Location directionality element is the

suffix –n. In sum, if we remove all the elements that are the same in (4) and (5), we are left with

two Location markers bithi and –n that select different types of Grounds. Both markers encode

Location directionality but select different noun types. Since bithi can combine also with the

question word hama “what,” I call nouns that combine with it what‐nouns. Since –n combines

also with the question word halo “where,” I call nouns that combine with it where‐nouns. The

question arises which nouns combine with which marker and what motivates this type of

nominal categorization? In order to answer this question, I look at two more languages, for which

there are enough data on the what/where split and inventory which nouns take which marker. I

would like to discuss with the participants of the meeting the patterns that emerge from the

comparison of the three languages—shown in the table below—as well as data from more

familiar languages such as English, in which a similar, though less conspicuous, pattern is attested.


The data suggest that a cline from tabletop entities to places (which include geographic entities)

underlies the grammatical pattern. The cline shows the likelihood of a noun being categorized as

a what‐ or where‐noun. Yet, the cut‐off point between the two categories is language‐specific.

What motivates this distribution? Is it the ontological features of the referents as Mark and Lyons

predicted? Yes, but not directly. Interestingly, in all three languages the what‐marking is always

more marked than where‐marking, suggesting that the cline ranges from nouns that are quite

marked in the function of Grounds to nouns that are unmarked as Grounds. Not forgetting that

this is a linguistic categorization, the cline should be seen as a reflection of the Figure/Ground

dichotomy, as generalized over the speakers of a language, rather than a direct translation of the

ontological features of the entities. By analyzing the cline, we can observe the types of

ontological features that change from prototypical Figures to prototypical Grounds. Instead of

defining the concepts of Figure and Ground a priori, which has been the case until now, we can

therefore let these concepts crystalize from the language data itself. Following this method and

by enlarging the language sample, we can also analyze the language‐specific cut‐off points on the

cline to investigate whether cross‐linguistically speakers of different languages prioritize different

features of Figures and Grounds, leading to the different what‐ and where‐groupings, and

whether there are any universal parameters to the categorization.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Scheider–90

Neural GIS—Computing with Cognitive Spatial References SIMON SCHEIDER

Department of Human Geography and Spatial Planning University of Utrecht


hen people refer to a location in natural language terms, they use a variety of cognitive spatial reference frames to put this location into perspective [7]. The same

place can e.g. be expressed in terms of a speaker centered, object centered or absolute frame of reference, and in a way that accounts not only for the inherent spatial vagueness, but also for the geometry of figure and ground objects [16]. For example, the expression “in front of” refers to a fuzzy location of different shape and size depending on whether the ground object is a church or a spoon on a table. Furthermore, the same location may be translated to “left of” when taking the observer as a ground object. In everyday communication, people effortlessly translate between these perspectives in order to understand what a speaker means [8]. While the different types of cognitive reference frames and their relevance for different language cultures have been studied in considerable depth [12], we still lack models that can be used to actually transform a geometric representation from one cognitive perspective to another [2,14,10,17], and thus to approximate the location that a natural language expression actually refers to [3]. We suggest one reason for this is that current Geographical Information Systems (GIS) are based on crisp reference systems [1], while cognitive reference frames require transformations that can take into account fuzzy locations, translations, rotations and scalings.

(a) Reference frame transformation for (b) A 2‐dim geographic map showing an observer, a house, and 1‐dimensional locations with neural fields [9]. a tree.

Fig.1: How to use neural field transformations in geographic maps?

W


Such transformation models can be inspired from neural fields [13,5] which are used in robotics and neural science to represent approximate relative locations in terms of arrays of interconnected firing neurons [9]. Inside a neural field, location, distance, direction (angle) and other geometric properties are not a result of crisp measurement, but of the way how neurons in this field are interconnected. For example, in Fig. 1a, the authors of [9] have wired two such fields together to transform a fuzzy absolute one‐dimensional position (target field) into a fuzzy object centered position, relative to a fuzzy reference position (reference field). In the resulting field, the dotted center line represents the ”origin,” i.e., the fuzzy location of the reference field, and the peak shows that the target location is ”left of” the reference field. How can we apply such a method to describe the spatial configuration in a geographic map, such as the one in Fig. 1b?

(a) The tree is to the right and in front of the observer (egocentric relative).

(b) The tree is to the East of the observer (egocentric absolute transformation.

(c) The tree is S‐E of the house (allocentric absolute transformation).

(d) The tree is to the left and in back of the house (intrinsic transf).

(e) The tree is to the right and in the back of the (deictic transf.)

(f) The tree is to the left and in front of the house (retinal transformation).

Fig.2: Fuzzy transformations of tree with respect to observer and house.

The trick underlying this method is that the projection takes into account all combinations

of fuzzy target and reference positions to determine the relative fuzzy position in the object‐centered field. We propose to mimic this behavior with fuzzy vector spaces [6,11,18], which allow us to compute transformations based on fuzzy translations, rotations and scalings. Using this method, we have modeled 6 well‐known types of cognitive spatial reference frames [4,7] in terms of fuzzy transformations, and applied them to the geographic map with an observer, a house and a tree (see Fig. 2). The fuzzy location of the tree can now be expressed relative to the positions, alignments and sizes of the observer and the house, without any need of prior discretization. The coordinate systems in Fig. 2 express relative frames of reference, with the


origin denoting either the observer or the house, and the vertical/horizontal axes expressing either Front‐Back/Left‐Right or (absolute) North‐South/West‐East directions. Note that the size, shape and location of the tree is transformed relative to the location, size and shape of the ground objects.

Since spatial reference systems and their transformations are fundamental for GIS [15], we suggest that being able to compute cognitive transformations may lead the way towards a Neural GIS. This is a new kind of GIS without a crisp geometry that effectively allows taking human spatial perspectives and computing with locations described in natural language texts. In a neural GIS, a coordinate field denotes fuzzy positions and directions relative to a particular cognitive reference frame, and transforming and intersecting fields allows checking to what degree spatial expressions apply.

References 1. Burrough, P.A., Frank, A.: Geographic objects with indeterminate boundaries, vol. 2. CRC Press (1996)

2. Carlson, L.A.: Selecting a reference frame. Spatial Cognition and Computation 1(4): 365–379 (1999)

3. Eschenbach, C.: Geometric structures of frames of reference and natural language semantics. SpatialCognition and Computation 1(4): 329–348 (1999)

4. Frank, A.U.: Formal models for cognition—Taxonomy of spatial location description and frames ofreference. In: Freksa, C., Habel, C., Wender, K. (eds.) Spatial Cognition, Lecture Notes in Computer

Science, vol. 1404, pp. 293–312. Springer Berlin Heidelberg (1998)

5. Johnson, J.S., Spencer, J.P., Schöner, G.: Moving to higher ground: The dynamic field theory and thedynamics of visual cognition. New Ideas in Psychology 26(2): 227–251 (2008)

6. Katsaras, A.K., Liu, D.B.: Fuzzy vector spaces and fuzzy topological vector spaces. Journal ofmathematical analysis and applications (58): 135–146 (1977)

7. Levelt, W.J.: Perspective taking and ellipsis in spatial descriptions. Language and space pp. 77–108(1996)

8. Levinson, S.C.: Space in language and cognition: Explorations in cognitive diversity, vol. 5. CambridgeUniversity Press (2003)

9. Lipinski, J., Schneegans, S., Sandamirskaya, Y., Spencer, J.P., Schöner, G.: A neurobehavioral model offlexible spatial language behaviors. Journal of Experimental Psychology: Learning, Memory, and

Cognition 38(6): 1490–1511 (2012)

10. Logan, G.D., Sadler, D.D.: A computational analysis of the apprehension of spatial relations. In:Language and space. Language, speech, and communication, pp. 493–529. MIT Press (1996)

11. Lubczonok, P.: Fuzzy vector spaces. Fuzzy sets and systems (38): 329–343 (1990)

12. Majid, A., Bowerman, M., Kita, S., Haun, D., Levinson, S.C.: Can language restructure cognition? Thecase for space. Trends in cognitive sciences 8(3): 108–114 (2004)

13. Pouget, A., Deneve, S., Duhamel, J.R.: A computational perspective on the neural basis of multisensoryspatial representations. Nature Reviews Neuroscience 3(9): 741– 747 (2002)

14. Regier, T., Carlson, L.A.: Grounding spatial language in perception: An empirical and computationalinvestigation. Journal of Experimental Psychology: General 130(2): 273–298 (2001)


15. Scheider, S., Kuhn, W.: Finite relativist geometry grounded in perceptual operations. In: Proceedings ofthe 10th International Conference on Spatial Information Theory. pp. 304–327. COSIT’11, Springer‐Verlag, Berlin, Heidelberg (2011)

16. Spencer, J.P., Simmering, V.R., Schutte, A.R., Schöner, G.: Insights from a dynamic field theory of spatial

cognition. The emerging spatial mind p. 320 (2007)

17. Tenbrink, T., Kuhn, W.: A model of spatial reference frames in language. In: Spatial Information Theory,pp. 371–390. Springer Berlin Heidelberg (2011)

18. Viertl, R., Hareter, D.: Beschreibung und Analyse unscharfer Information: statistische Methoden für

unscharfe Daten. Springer‐Verlag (2006)

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Schultheis—94

Universal Mechanisms, Variable Parameterization

HOLGER SCHULTHEIS Spatial Cognition Center University of Bremen

Email: [email protected]‐bremen.de

t seems generally accepted that spatial reference frames are crucially involved in spatial

referencing (Levinson, 2003; Logan & Sadler, 1996). Reference frames are means to mentally

organize space and they enable distinguishing and labeling different parts of space as well as

apprehending and distingushing different spatial relations. As such, reference frames are

indispensable prerequisites for spatial referencing in all cultures and languages. Previous

research has distinguished a variety of reference frames that natural cognitive agents are

assumed to draw on. In the spatial language domain an influential distinction of reference frames

into relative, absolute, and intrinsic frames has been proposed by Levinson (2003, but see also

Pederson, 2003) and much research in linguistics has focused on how languages and cultures

differ regarding their proclivity to employ these different frames.

I argue

• that differences in which reference frame is selected need not be indicative of differences in howselection is achieved,

• that the mechanisms underlying reference frame selection in English are universal as mechanismsfor reference frame selection across languages and cultures,

• more generally, that certain processing steps involved in spatial referencing and the mechanismsrealizing them are universal, but that the parameterization or tuning of these universal mechanismsvary across languages and cultures.

Studies on English spatial language use have shown (e.g., Carlson & van Deman, 2008;

Carlson, 1999) that spatial referencing requires selecting one of the available reference frames. A

recent in‐depth study of the mechanisms underlying reference frame selection has identified the

leaky competing accumulator (LCA, Usher & McClelland, 2001) model as an accurate account of

this subprocess of spatial referencing (Schultheis & Carlson, in press). LCA is a connectionist

model with a single layer of units. Each of the units represents one possible reference frame.

Each unit receives input from those available sources of information that support the frame

represented by the unit. Unit activation increases by accumulating the received input, and

decreases due to decay and lateral inhibition between units. Activation is furthermore influenced

by unsystematic fluctuations (white noise). Once any unit’s activation grows beyond a

prespecified threshold, reference frame selection stops. The selected frame is assumed to be the

one represented by the winning unit.

I propose that—insofar as the LCA is an accurate account of reference frame selection for

English—the LCA’s structure and workings are universal as a mechanism for reference frame

selection across languages and cultures. Given that (a) reference frames are crucial prerequisites

for spatial referencing and that (b) all natural cognitive agents have access to a multitude of

reference frames, the necessity to select a reference frame for spatial referencing will, arguably,

I


be of little debate. But what of the more specific claim that reference frame selection is governed

by mechanisms as implemented in the LCA? Obviously, there are differences in reference frame

selection across languages and cultures. As, for example, the seminal work of Stephen Levinson

and his colleagues has shown, there are many languages in which frames are selected for spatial

referencing that would rarely be used in English (Levinson, 2003).

However, the observed differences are differences in which frame is usually selected and

they need not be indicative of differences in how selection is achieved. Even if the mechanisms

are the same, differences can easily arise by different parameterizations of the LCA such as, for

example, different inhibitory strength, different rates of decay, and, most notably, different

salience/strength of available competing frames. Accordingly, the mechanism underlying

reference frame selection may be universal across languages and cultures, while the observed

variation across languages arises from adaptations of the universal mechanism.

Are there any reasons to believe that the LCA constitutes a universal mechanism? I think

there are, though a more direct test and corroboration certainly seems desirable (see below).

One reason is the non‐deterministic nature of the selection outcome. Even when strong

preferences have been found for one particular frame, alternative frames are also selected for a

(small) proportion of trials. This is true not only for English (Li & Gleitman, 2002), but also for

languages that exhibit a preference for an absolute frame (Levinson, 2003). Such a pattern of

selection outcomes is well in line with the type of competitive, noisy process realized by the LCA.

The second reason is that reference frame selection is an important process not only in spatial

referencing but also in spatial cognition abilities such as mental image reinterpretation (Peterson,

Kihlstrom, Rose, & Glisky, 1992) and perspective taking (May, 2004): In fact, an LCA‐like process

has been successfully employed to model perspective taking (Schultheis, 2007). Accordingly, the

LCA captures an aspect of spatial cognition that is not tightly tied to language and it seems

natural to suppose that a mechanism not specifically deployed for language is universal across

languages. It is worth noting that the applicability of the mechanisms realized in the LCA to both

spatial referencing and spatial reasoning also suggests a new view on the question of linguistic

relativity. While many proponents in the debate either argue for (some form of) the SapirWhorf

hypothesis (Majid, Bowerman, Kita, Haun, & Levinson, 2004; Levinson, 2003) or the opposite

(Gallistel, 2002; Li & Gleitman, 2002) the general applicability of the LCA indicates that observed

correlations between language and thought may arise from common mechanisms and not from a

causal influence of language on thought or vice versa.

As already stated above, it is desirable to more directly test the proposition put forth here by

investigating the universality of mechanisms involved in spatial referencing in more detail in

various languages and cultures. Questions of particular interest are, for example: (a) How well

does the LCA account for selection in other languages and cultures? (b) How well do other

mechanisms (e.g., the AVS, Regier & Carlson, 2001) transfer from English to other languages? (c)

Is the way in which mechanisms of subprocesses of spatial referencing combine (e.g., LCA and

AVS) also universal? (d) How do observed variations in spatial referencing map onto changes in

parameterizations?


ReferencesCarlson, L. A. (1999). Selecting a reference frame. Spatial Cognition and Computation 1: 365–379.

Carlson, L. A., & van Deman, S. R. (2008). Inhibition within a reference frame during the interpretation of spatial language. Cognition 106: 384–407.

Gallistel, C. R. (2002). Language and spatial frames of reference in mind and brain. Trends in Cognitive Sciences 6(8): 321–322.

Levinson, S. C. (2003). Space in language and cognition. Cambridge, UK: Cambridge University Press.

Li, P., & Gleitman, L. (2002). Turning the tables: language and spatial reasoning. Cognition 83: 265–294.

Logan, G. D., & Sadler, D. D. (1996). A computational analysis of the apprehension of spatial relations. In P. Bloom, M. Peterson, M. Garrett, & L. Nadel (Eds.), Language and space (p. 493–529). MA: M.I.T Press.

Majid, A., Bowerman, M., Kita, S., Haun, D. B. M., & Levinson, S. C. (2004). Can language restructure cognition? The case for space. Trends in Cognitive Sciences 8: 108–114.

May, M. (2004). Imaginal perspective switches in remembered environments: Transformation versus interference accounts. Cognitive Psychology 48: 163–206.

Pederson, E. (2003). How many reference frames? In C. Freksa, W. Brauer, C. Habel, & K. F. Wender (Eds.), Spatial Cognition III: Routes and navigation, human memory and learning, spatial representation and spatial learning (p. 287–304). Berlin: Springer.

Peterson, M. A., Kihlstrom, J. F., Rose, P. M., & Glisky, M. L. (1992). Mental images can be ambiguous: reconstruals and reference‐frame reversals. Memory & Cognition 20(2): 107–123.

Regier, T., & Carlson, L. A. (2001). Grounding spatial language in perception: An empirical and computational investigation. Journal of Experimental Psychology: General 130(2): 273–298.

Schultheis, H. (2007). A control perspective on imaginal perspective taking. In R. L. Lewis, T. A. Polk, & J. E. Laird (Eds.), Proceedings of the 8th international conference on cognitive modeling (iccm 2007) (p. 19–24).

Schultheis, H., & Carlson, L. A. (in press). Mechanisms of reference frame selection in spatial term use: Computational and empirical studies. Cognitive Science. doi: 10.1111/cogs.12327

Usher, M., & McClelland, J. L. (2001). The time course of perceptual choice: The leaky, competing accumulator model. Psychological Review 108: 550–592.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Spranger–97

Lessons from Robotics Experiments MICHAEL SPRANGER

Sony Computer Science Laboratories Inc. Email: [email protected]

Fig. 1: (a) spatial setup for robot‐robot interactions, (b,c) example of evolution of spatial strategies (chunks), (d) formation of spatial lexicons, (e) acquisition of the German locative system in tutorlearner interactions.

t is probably safe to assume that all languages in one way or another allow speakers to express spatial locations of relevant objects or locations, for instance, for food, good hunting grounds

or dangerous areas. It is also very conceivable that spatial language has entered our symbolic species quite early. But despite being so important in communication and also cognitively central to our experience of reality [4, 5], spatial language shows remarkable cross‐cultural variation [1, 3, 4]. Through the tedious work of many typologists we now have estimates about the degree to which human spatial languages makes use of environmental and bodily features to develop ingenious strategies for talking about locations, places and spatial configurations. Some cultures use slopes, others salient landmarks, again others rely on projective categories. But languages also differ in how these various conceptualization strategies are expressed in natural language. Spatial knowledge is communicated through all sorts of devices spatial prepositions, adpositions, morphemes, case systems etc [16].

Our understanding of how humans process language from the viewpoint of psychology, psycholinguistics and language typology is getting better, and at the same time we see more work on trying replicate and study these processes using robotic and computational models of language. In this position paper we try to extract some basic principles that have guided our work in this area and try to draw some conclusions about what computational and robotic models can contribute to our understanding of universals and diversity of spatial language. In particular, we are interested in representations, algorithms and their computational properties for processing, their ability to learn and evolve spatial language.

We have developed a series of robotic experiments [14, 7, 8, 15, 9] that show that agents can self‐organize spatial language systems similar (or at least approaching) in complexity to what has

I


been observed in human languages. The experiments show that from simple referential interactions about objects and places in the environment, paired with the right agent internal representations and algorithms various different spatial language systems can evolve. The evolving languages are influenced by environmental factors, as well as the perceptual apparatus, the cognitive capabilities and the communicative goals of the agents. Such experiments show that we can try to identify basic principles explaining spatial language diversity through robotic experiments (see Fig. 1 for some results from various experiments). The following paragraphs briefly introduce the most important principles.

Spatial language strategies are procedural Spatial utterances encode very specific instructions to the hearer that allow him to, for instance, find the referent of the utterance. We have found that such instructions are best represented as constraint‐based programs (combinations of categories and conceptualization operators) with data flow instead of instruction flow as principle [13]. This allows agents to flexibly interpret utterances even if aspects of the utterance are not yet known (e.g., in acquisition). Various strategies for conceptualizing spatial reality can be modeled using this technique [10].

Spatial conceptualization strategies are composed of cognitive building blocks Spatial language conceptualization strategies are composed of cognitive building blocks, such as categorization, perspective reversal etc. These are general mechanisms that are useful outside of spatial language. For instance, perspective reversal can be useful to compute what another agent is seeing. Spatial language exapts these cognitive building blocks, which allows agents to incrementally construct complex conceptualization strategies [9].

Syntax is a means for communicating differences in procedural spatial semantics The best way to understand spatial language syntax is to understand how differences in spatial semantics correspond with differences in syntax. For instance, in German and English adjectival use of projective categories (e.g. front) require a group‐based conceptualization strategy [6]. Prepositional use, on the other hand, requires region‐based processing [2].

Co‐development of syntax and semantics The tight connection between syntax and semantics of spatial language require that both are acquired (ontogeny) and evolved (phylogeny) at the same time. We have found in our experiment that a tight coupling between adaptation of conceptualization strategies (spatial semantics) and how they are expressed (spatial syntax) is crucial for allowing agents to learn and self‐organize complex spatial language [14, 7, 8, 15].

Spatial language evolves in a process of cultural evolution Agents develop various forms of spatial language given a sufficient perceptual apparatus, cognitive capabilities and environment. This happens without any adaptation of the agent architecture. Learning and evolution operators are implemented inside agents and lead to remarkably complex systems in relatively few iterations. Variations in the environment or the stochasticity characterizing multiple runs of the same experiment can lead to different spatial language systems. For example, in one experiment absolute spatial relations emerge, in others projective systems emerge together with landmarks


etc. This shows that cultural evolution of spatial language is a possible explanation for the cross‐cultural diversity observed in Natural language.

Spatial language is without doubt an interesting topic because it is so central to our experience of reality. Importantly, spatial language can impact other language systems and conceptualizations such as our understanding of time, but even abstract spaces such as politics and economics are in some languages closely connected with spatial language and talked about in spatial terms. Although we already have some initial computational and robotic experiments studying the emergence of such systems, much more work is needed to deepen our understanding of all the different processes involved. In particular, one area of concern at this point is how to combine the robotic models with models of grammaticalization and/or experimental semiotics (laboratory experiments on the evolution of language with people).

References [1] P. Brown. Up, down, and across the land: Landscape terms, place names, and spatial language in tzeltal.

Language Sciences 30(2–3): 151–181, 2008.

[2] J. Kelleher and F. Costello. Cognitive representations of projective prepositions. In Proceedings of the Second ACL‐Sigsem Workshop of The Linguistic Dimensions of Prepositions and their Use in Computational

Linguistic Formalisms and Applications, 2005.

[3] S. C. Levinson. Space in Language and Cognition: Explorations in Cognitive Diversity. Number 5 in Language, Culture and Cognition. Cambridge University Press, 2003.

[4] P. Li and L. Gleitman. Turning the tables: Language and spatial reasoning. Cognition 83(3): 265–294, 2002.

[5] A. Majid, M. Bowerman, S. Kita, D. Haun, and S. C. Levinson. Can language restructure cognition? The case for space. Trends in Cognitive Sciences, 8(3):108–114, 2004.

[6] R. Moratz and T. Tenbrink. Spatial reference in linguistic human‐robot interaction: Iterative, empirically supported development of a model of projective relations. Spatial Cognition & Computation 6(1): 63–107, 2006.

[7] M. Spranger. The co‐evolution of basic spatial terms and categories. In L. Steels, editor, Experiments in

Cultural Language Evolution, number 3 in Advances in Interaction Studies, pages 111–141. John Benjamins, 2012.

[8] M. Spranger. Evolutionary explanations for spatial language—A case study on landmarks. In P. Li`o, O. Miglino, G. Nicosia, S. Nolfi, and M. Pavone, editors, Advances in Artificial Life, ECAL 2013, volume 12, pages 1999–1205. MIT Press, 2013.

[9] M. Spranger. Evolving grounded spatial language strategies. KI—Ku¨nstliche Intelligenz 27(2): 97–106, 2013.

[10] M. Spranger. Procedural semantics for autonomous robots—A case study in locative spatial language. In Intelligent Robots and Systems (IROS), 2015 IEEE/RSJ International Conference on. IEEE, 2015.

[11] M. Spranger and M. Loetzsch. Syntactic Indeterminacy and Semantic Ambiguity: A Case Study for German Spatial Phrases. In L. Steels, editor, Design Patterns in Fluid Construction Grammar, volume 11 of Constructional Approaches to Language, pages 265–298. John Benjamins, 2011.

[12] M. Spranger, M. Loetzsch, and S. Pauw. Open‐ended Grounded Semantics. In H. Coelho, R. Studer, and M. Woolridge, editors, Proceedings of the 19th European Conference on Artificial Intelligence (ECAI 2010), volume Frontiers in Artificial Intelligence and Applications, pages 929–934. IOS Press, 2010.


[13] M. Spranger and S. Pauw. Dealing with Perceptual Deviation—Vague Semantics for Spatial Language and Quantification. In L. Steels and M. Hild, editors, Language Grounding in Robots, pages 173–192. Springer, 2012.

[14] M. Spranger, S. Pauw, and M. Loetzsch. Open‐ended semantics co‐evolving with spatial language. In A. D. M. Smith, M. Schouwstra, B. de Boer, and K. Smith, editors, The Evolution of Language (Evolang 8), pages 297–304, Singapore, 2010. World Scientific.

[15] M. Spranger and L. Steels. Co‐acquisition of syntax and semantics—An investigation in spatial language. In Q. Yang and M. Wooldridge, editors, IJCAI’15: Proceedings of the 24th international joint conference on Artificial intelligence, pages 1909–1905, Palo Alto, US, 2015. AAAI Press.

[16] S. Svorou. The Grammar of Space, volume 25 of Typological Studies in Language. John Benjamins, 1994.

2016 Specialist Meeting—Universals and Variation in Spatial Referencing Winter–101

On Modelling Human Place Knowledge STEPHAN WINTER

Department of Infrastructure Engineering The University of Melbourne Email: [email protected]

1. Place graphs

uman knowledge about places can be expressed in verbal place descriptions containing spatial references. Spatial references are locating something in the world, as in “I am at the

bus stop.” They can be extracted by language technologies, typically in the form of triplets of a locatum L, a relatum R, and the qualitative spatial relationship r between them [6]. In English, r can be expressed or indicated by a spatial preposition, verb or noun, or sometimes, in written text, just by a comma (as in “Cambridge, MA”).1 Accordingly, the language technologies are language‐specific: a parser developed for extracting English spatial references does not work on Dutch place descriptions. The above example can be parsed into the triplet <L: I, r: at, R: bus stop>.

The triplets representing spatial references form a directed property graph [12], and since spatial references represent human spatial knowledge this property graph is a representation of human spatial knowledge, too. We will call the location of things places, and this property graph a place graph2 [13]. Figure 1 illustrates the concept.

Figure 1. The place graph representing the spatial reference “I am at the bus stop.”

In this contribution, these place graphs will be investigated as tools for studying universals and variation in spatial referencing across cultures and languages. Even if place graphs are knowledge representations, not text, all represented knowledge has been derived from verbal descriptions.

2. Place graphs as knowledge representation

A triplet LrR establishes a fact about the world: a relationship r between two places L and R. It is a fact in the sense that some agents—those involved in the conversation the place description was taken from—believe it to be true.3

1 While triplets represent only binary relationships, ternary relations can be covered by two triplets. 2Place graphs, in one form or another, have been suggested for a while for the representation of human spatial knowledge, e.g., [8]. The presented place graph, however, is unique in its derivation from place descriptions.

3Without limiting generality, we ignore that people can make wrong spatial references, and also that parsers can establish false relationships.

H


However, the triplet is an abstract, symbolic description of a fact. It has taken both the references to the two places as well as the relationship out of the embedding conversational.

1.2 On Modelling Human Place Knowledge

context. From the triplet alone it is now impossible to reconstruct the identity of I or the bus stop, and also to interpret at spatially.

Nevertheless, collecting large sets of triplets about the same environment allows merging triplets to connected graphs, in a manner such that4 [7]:

• Nodes can be merged if their labels refer to the same place. A merged node would keep allreferences to this place (“State Library,” “library,” “the building in Gothic revival style”).This node should also keep the role in which each reference was used (R: library, L: thebuilding in Gothic revival style).

• The place graph can become a multi‐graph (e.g., with <L: café, r: at, R: library>, <L: café, r:opposite, R: library>, <L: building in Gothic revival style, r: opposite, R: café>). Since theconversational contexts are lost (and thus, the reference frames of r) even seeminglycontradicting edges (<L: café, r: near, R: library>, <L: café, r: far, R: library>) have to beaccepted since they have been true in their respective contexts.

• The relationship r can be mapped to a standardized set of labels. This further generalizationmay enable inference engines to spatial reasoning [3, 9].

Since a place graph is a knowledge representation [1], and in combination with graph traversal as inference engine even a knowledge‐based system [4], a place graph is language‐independent for all languages. It can represent spatial references from any language that locates things in relationship to something already located, i.e., by binary or ternary spatial relationships.

3. Spatial referencing across cultures and languages

The place graph introduced above is a representation of configurational knowledge extracted from language. It had been suggested for Question Answering, although their capacity in this regard has not yet been explored. But since a particular place graph is representing triplets extracted from a corpus of place descriptions in one language (due to the limitations of language technologies) such a place graph is also only intended to answer questions in the same language. From this observation two research questions can be derived:

1. Can a place graph constructed from descriptions in one language be used to answerqueries in another language?

2. Can a place graph constructed from descriptions in one language be matched with a placegraph constructed from descriptions of the same environment in another language?

The place graphs introduced above are also conceptually different to traditional spatial knowledge representations such as gazetteers (linking place names to coordinated locations; [5])

4 Again, without limiting generality, we ignore that the matching process is imperfect.


and geographic information systems (managing crisp geometries with attributes; [10]). As an alternative representation they address some of the issues with traditional representations (e.g., [2]) that are deeply rooted in Western culture (e.g., [11]). Other cultures, and especially those with oral traditions, may find place graphs more appropriate to represent their knowledge of the land. Thus:

1. Is a place graph suited to capture knowledge from cultures with oral traditions?

2. In which ways would such a place graph be different from a place graph constructed fromdescriptions of the same environment in a Western language?

© Stephan Winter; licensed under Creative Commons License CC‐BY

References

1 Ronald Brachman and Hector Levesque. Knowledge Representation and Reasoning. Morgan Kaufmann Publishers Inc., San Francisco, CA, 2004.

2 Peter A. Burrough and Andrew U. Frank. Geographic Objects with Indeterminate Boundaries, volume 2 of ESF‐GISDATA. Taylor & Francis, 1996.

3 Anthony G. Cohn and Jochen Renz. Qualitative Spatial Representation and Reasoning, book section 13, pages 551–596. Elsevier, 2008.

4 Frederick Hayes‐Roth, Donald A. Waterman, and Douglas B. Lenat. Building expert systems. Addison‐Wesley, Reading, MA, 1983.

5 Linda L. Hill. Georeferencing: The Geographic Associations of Information. Digital Libraries and Electronic Publishing. The MIT Press, Cambridge, MA, 2006.

6 Arbaz Khan, Maria Vasardani, and Stephan Winter. Extracting spatial information from place descriptions. In Simon Scheider, Benjamin Adams, Krzysztof Janowicz, Maria Vasardani, and Stephan Winter, editors, ACM SIGSPATIAL Workshop on Computational Models of Place, pages 62–69. ACM Press, 2013.

7 Junchul Kim, Maria Vasardani, and Stephan Winter. Similarity matching for integrating spatial information extracted from place descriptions. International Journal of Geographical Information Science, online 29 May 2016. URL: http://www.tandfonline.com/doi/full/10.1080/13658816.2016.1188930.

8 Benjamin J. Kuipers. Modeling spatial knowledge. Cognitive Science, 2(2):129–153, 1978. 9 Gérard Ligozat. Qualitative Spatial and Temporal Reasoning. John Wiley & Sons, Inc., Hoboken, NJ, 2013. doi:10.1002/9781118601457.

10 Paul A. Longley, Michael F. Goodchild, David J. Maguire, and David W. Rhind. Geographic Information Systems and Science. John Wiley and Sons, Chichester, 3rd ed. edition, 2010.

11 David M. Mark and Andrew G. Turk. Landscape Categories in Yindjibarndi, volume 2825 of Lecture Notes in Computer Science, pages 28–45. Springer, Berlin, 2003.

12 Marko A. Rodriguez and Peter Neubauer. Constructions from dots and lines. Bulletin of the American Society for Information Science and Technology, 36(6):35–41, 2010. doi:10.1002/bult.2010.1720360610.

13 Maria Vasardani, Sabine Timpf, Stephan Winter, and Martin Tomko. From descriptions to depictions: A conceptual framework. In Thora Tenbrink, John Stell, Antony Galton, and Zena Wood, editors, Spatial Information Theory, volume 8116 of Lecture Notes in Computer Science, pages 299–319, Cham, 2013. Springer.

Documents

Taking Stock of the Cross‐Linguistic Data: Spatial Frames