Understanding spatial relations: Flexible infants, lexical adults

https://doi.org/10.1016/S0010-0285(02)00514-5Get rights and content

Abstract

Concepts of containment, support, and degree of fit were investigated using nonverbal, preferential-looking tasks with 9- to 14-month-old infants and adults who were fluent in either English or Korean. Two contrasts were tested: tight containment vs. loose support (grammaticized as ‘in’ and ‘on’ in English by spatial prepositions and ‘kkita’ and ‘nohta’ in Korean by spatial verbs) and tight containment vs. loose containment (both grammaticized as ‘in’ in English but separately as ‘kkita’ and ‘nehta’ in Korean). Infants categorized both contrasts, suggesting conceptual readiness for learning such spatial semantics in either language. English-speaking adults categorized tight containment vs. loose support, but not tight vs. loose containment. However, Korean-speaking adults were successful at this latter contrast, which is lexicalized in their language. The adult data suggest that some spatial relations that are salient during the preverbal stage become less salient if language does not systematically encode them.

Introduction

A persistent question in psychology concerns the role that language plays in determining thought. This question, which has been with us at least since the writings of Edward Sapir and Benjamin Whorf, has had a cyclical history over the years, varying from the assumption of strong influence (e.g., Bloom, 1981) to weak or no influence (e.g., Au, 1992). In between the highs and lows of this cyclical course there have been middle positions, such as that language plays a role in “thinking for speaking” (Slobin, 1996), leaving open the question of the extent to which language affects other kinds of thought as well. Most recently, the cycle has reached a new high point in the estimates of linguistic influence in several crosscultural studies of spatial cognition (Levinson, 1996; Lucy, 1992). The point of view offered by Levinson, for example, is that the spatial frame of reference of a given language influences spatial thought in many tasks, such as recall, recognition, and making inferences (Levinson, 1996).

Hand in hand with differing views of the relation between language and thought have been differing views of the preverbal mind that enable the learning of language, particularly the relational aspects, such as grammatical morphemes and spatial prepositions. Linguists sometimes assumed that there are semantic primitives or conceptual universals (e.g., Bierwisch, 1967) that in various combinations are used to provide the meanings of relational terms in all languages. If these concepts are present preverbally they might help young children learn abstract linguistic categories (Clark, H., 1973). In contrast, many developmental psychologists influenced by Piagetian theory assumed that infants are purely sensorimotor beings, and that at the time that language first begins to be learned have not yet developed a conceptual system. This view entailed the further assumption that language is mapped onto sensorimotor achievements instead. However, empirical studies attempting to correlate language development with scales of sensorimotor development did not provide much support for this idea (Bloom, Lifter, & Broughton, 1985; Nelson & Lucariello, 1985).

We assume that whether or not concepts result from sensorimotor activity, some conceptual basis is required to learn language. From the point of view of language acquisition, crosslinguistic studies of early language show that young children everywhere talk about simple events in which animate agents act on inanimate objects (e.g., Brown, 1973; Slobin, 1973, Slobin, 1985 ). Congruent with this finding, developmental research has shown that preverbal concepts of animacy, inanimacy, and agency begin to be formulated in the first year of life, as well as concepts of the events in which objects take part (Johnson, Slaughter, & Carey, 1998; Legerstee, 1992; Mandler & McDonough, 1996; McDonough & Mandler, 1998; Spelke, Phillips, & Woodward, 1995). These abstract characterizations of events provide a conceptual base onto which simple sentence structure can be mapped (Mandler, 1996).

A good deal of research has begun to flesh out the concepts preverbal infants have formed about animate and inanimate objects (e.g., Mandler and McDonough, 1996, Mandler and McDonough, 2000), but the concepts onto which spatial language is mapped have remained more speculative. Perhaps the best known approach was that of the Clarks (Clark, E., 1973; Clark, H., 1973), who assumed that there are cognitive universals that account not only for commonalities in spatial terms across various languages but also for the easy acquisition of simple spatial prepositions. Examples of universal primitives that have been suggested include containment, support, contact, opening and closing, horizontality, verticality, and path (e.g., Bloom et al., 1985; Clark, E., 1973; Nelson, 1974; Sinha, Thorseng, Hayashi, & Plunkett, 1994; Talmy, 1985). Many of these are similar to the image schemas underlying language understanding discussed by cognitive linguists (e.g., Lakoff, 1987; Johnson, 1987) and proposed by Mandler (1992) as developing early in infancy. Such conceptual primitives, whether innate or learned, would help constrain the possible meanings of spatial terms, making language learning easier for the child.

Much of the relevant early research was conducted in English, but as crosscultural data on acquisition became more frequent it became apparent that across languages children learn spatial relational terms of widely diffferent forms equally easily. This finding suggested that the notion of a small set of spatial primitives might not be a plausible way to explain acquisition. Bowerman, 1989, Bowerman, 1996a and Bowerman and Choi (2001) discussed some of the highly variable ways of expressing containment and support that various languages use, from the semantically rather general “in” and “on” of English to the elaborate system of body-part metaphors found in Mixtec or the obligatory requirement when talking about containment in Tzeltal to choose the spatial term that specifies the shape and orientation of the container (Brown, 1994). As they pointed out, children in the languages that have been studied pick up on varying semantic categories quite easily. For example, Choi and Bowerman (1991) described how Korean spatial morphemes specify loose- or tight-fitting relations crosscutting the categories of containment and support, and showed that young Korean children encode this variable when talking about spatial relations. In Korean, the spatial verb “kkita” means ‘fit together tightly or interlock’, regardless of whether something is contained or supported. For example, putting books into book covers, putting pegs tightly into holes, putting rings tightly on poles and putting lego blocks together are all good instances of “kkita” in Korean. Korean does not have a verb that has the same semantic content as that of “in” nor does English have a particle/preposition with a meaning comparable to that of “kkita.”

Choi, McDonough, Bowerman, and Mandler (1999) studied the onset of comprehension of “kkita” in Korean and “‘put’ in” in English, and found that at the earliest age at which reliable comprehension of such terms was shown (18 months) infants were already sensitive to the different aspects expressed by spatial morphemes in the two languages (see below for more detail on the Choi et al. experiment). Bowerman and Choi (Bowerman, 1996b; Bowerman & Choi, 2001) concluded that the language being learned is in some sense teaching children how to conceptualize spatial relations.

This discussion suggests a chicken and egg problem. On the one hand it is not clear how it would be possible to learn any spatial terms without some conceptual base on which to map the sounds of one’s native tongue. On the other hand, how can it happen that the earliest comprehension of spatial terms already shows language-specific influences if language were not influencing the infant’s conceptions of space? One possible answer would be that preverbal infants have already learned a larger or perhaps different set of spatial concepts than is usually credited to them. Perhaps infants have not only already learned concepts of containment and support, but also concepts of tight and loose fitting, so that when Korean infants, for example, hear the word “kkita,” (‘put together tightly or interlock’) and “nehta,” (‘put in, around or together loosely’), they are as prepared to understand these terms as infants learning English are to understand the terms “in” and “on.” In this view, language is not “teaching” these concepts, but instead is packaging existing concepts semantically in different ways, each of which is roughly easy to learn.1 Thus, different languages create different semantics from a common preverbal conceptual base, enabling infants to learn language-specific semantics.

The notion of semantic primitives helping with the task of language acquisition fell into disrepute because of an apparent lack of empirical support and because of the theory-laden descriptions attached to them (see Bowerman, 1996b and Carey, 1982, for discussion). One criticism was that the primitives used to distinguish various spatial categories (such as long versus tall versus wide) are too abstract and sophisticated to be plausible for preverbal children to know and so it is not clear how they would help constrain initial hypotheses about spatial word meaning. It was also suggested that there would need to be a very large number of primitives to accommodate learning all language-specific spatial concepts (Bowerman, 1996b). However, children acquire spatial terminology only gradually, with differences among such dimensional terms as tall versus wide being relatively late acquisitions. It does not seem necessary for all spatial notions to be available to infants for them to begin to learn how many common spatial notions are expressed in their native tongue. There may be a universal set of preverbal distinctions acquired by all infants in their experience with the world that can be used to compile the most common and early learned linguistic spatial categories found in different languages. Perhaps the search for conceptual primitives was abandoned too early. Given the recent research in cognitive development showing more conceptual understanding in infancy than once thought possible, it seems timely to investigate the various components that are implicated in linguistic spatial categories to see if preverbal children are already sensitive to them.

There is by now a good deal of data on the early development of concepts related to containment and support in the first year of life. For support, Baillargeon and her colleagues describe a learning sequence in which first a qualitative distinction between support and nonsupport is made, followed by gradual learning of quantitative details, such as how much overlap between objects is needed if one is to support the other (Baillargeon, 1994). For containment, there are fewer systematic developmental data in infancy, but it is known that by 5.5 months infants know that containers must have bottoms (Baillargeon, 1995), and by 8.5 months they take into account the width and compressibility of an object when determining whether it can be put into a container (Aguiar & Baillargeon, 1998). The latter finding in particular suggests a fairly sophisticated understanding of containment. As for tightness vs looseness, to our knowledge there is no work other than that presented here that has investigated these concepts in infancy.

The present research asks whether these concepts of containment and support that preverbal infants have already formed (and perhaps degree of fit as well) are sufficiently abstract and context-free to account for the understanding of the basic spatial vocabulary of either Korean or English that begins around 18 months of age (Choi et al., 1999). Secondarily, it asks whether after learning Korean or English and using these differing spatial terms for many years, adult Korean and English speakers differ in nonverbal spatial categorization and in one kind of spatial thought.

In our previous research we investigated children’s comprehension of two spatial terms: the English word “in” and the Korean word “kkita” (Choi et al., 1999). We used a preferential looking test in which two scenes were shown simultaneously and were accompanied by an audio recording which described one but not the other scene. For example, one scene showed a book placed on top of another book and the other scene showed a book placed into a tight-fitting cover. The English audio directed the child with instructions such as “Look! Where is she putting it in?.” The Korean audio directed the child with instructions (in Korean; i.e., “Eti-ey kkie” ‘where-at tight-fitting’) that would translate to “Look! Where is (she) tight-fitting (it)?” In this situation, both the English and Korean children should look at the same scene, namely putting a book in a matching book cover, since the relation depicts both containment and tight fit (see Fig. 1). Another pair of scenes showed a ring dropped into a large basket and a ring placed tightly onto a pole. Using the same audio, English children should look to the ring tossed into the basket (i.e., containment) whereas Korean children should look to the ring placed tightly on the pole (i.e., tight fit). This technique was effective in showing that by 18 months of age, children will look to the scene described in their language, indicating their comprehension of the way the language they are learning carves up spatial relations. That is, as Fig. 1 shows, 18-month-old Korean children understood ‘books in book covers’ and ‘rings on poles’ to instantiate a tight-fitting relation, whereas 18-month-old English-learning children understood that ’books in book covers’ and ’rings in basket’ instantiate containment.

The present studies again used a preferential-looking technique to study the development of concepts of containment, support, and degree of fit in preverbal infants. We studied 9- to 14-month-olds and familiarized them to varied instances of a spatial relation. Then we tested their preferential looking by giving them two test trials in which one scene showed a novel instance of the familiarized relation and the other scene showed a novel relation. We used a wide variety of objects that differed on many dimensions (shape, size, color, and texture) to make sure that any categorical responses would be based on spatial relations rather than on physical similarity or similar taxonomic relations among objects. We were particularly interested to discover whether infants in this age range have developed spatial concepts abstract enough to account for the wide range of instances that 18-month-olds include when they begin to understand the relevant relational terms. Because they apply their newly learned words even to new instances they have not seen before, it seems likely that an abstract understanding is required.

It also seems likely that such understanding develops over the course of the first year. For example, Quinn, 1999, Quinn et al., 1996 found that 3-month-olds could categorize above and below relations involving a single figure (such as a dot or geometric drawing) vis a vis a line but not until several months later could infants categorize these relations when more than one figure was used. The work of Baillargeon and her colleagues also implicates major developmental changes during the first year in understanding containment and support. However, their experiments also used a limited range of stimuli, which made the spatial changes highly salient (Baillargeon, 1994, Baillargeon, 1995). One cannot be certain from this work whether infants in the first year can categorize containment and support relations when a wide range of objects is presented.

In the present experiments, we also tested adults who are either fluent in English or Korean by giving them the same preferential looking tasks we gave to infants. We do not in fact know if even adults will abstract a common spatial relation from a series of otherwise unrelated events. Therefore, we wanted to see if adults show the same looking preferences as infants and, if so, whether they are aware of the relational distinctions being tested. To test their awareness of the category differences, we gave them an oddity task after which we asked them to explain the bases for their selections. If adults are influenced by language in their construal of events, as suggested by Whorf (1956), then evidence should be found in the adults’ verbal descriptions of the tasks. More specifically, as recent studies have suggested (i.e., Levinson, 1996; Pederson et al., 1998), it is possible that those aspects of space that are semantically relevant to a speaker’s language may become more salient over time and more available as a basis for nonlinguistic spatial categorization.

Section snippets

Experiment 1

To begin our investigation of preverbal spatial categories we needed to see if the preferential looking technique would work, so we used a distinction that seemed as salient as possible: tight-fitting containment versus loose-fitting support. In English and Korean, these two relations are subsets of two distinct semantic categories in each language: “in” and “on” in English denoting containment and support respectively (but not tight or loose fit), and “kkita” and “nohta” in Korean denoting

Experiment 2

The results from Experiment 1 indicated that the preferential looking technique is a viable method for use with 9 to 14-month-olds even when the contrasts under study are abstract spatial relations, and so more detailed crosslinguistically meaningful studies could begin. In Experiment 2, we tested a contrast between tight-fitting containment and loose-fitting containment. This contrast is saliently marked in Korean by the terms ‘kkita’ and ‘nehta’ (although neither are strictly limited to

Experiment 3

Our next question was whether infants raised in Korean-speaking environments and Korean-speaking adults would react differently to the tests used in Experiment 2. Given that Korean spatial terms saliently mark the differences between tight and loose fit, it was expected that Korean adults would succeed on both the preferential looking task (by showing a preference during the test trials) and the oddity task by choosing and describing correctly the odd relation. Because infants presumably share

Comparison analyses

In order to contrast further the results of the infants and adults in Experiments 2 and 3, we conducted two crossexperiment analyses to see if there were any interactions that would qualify our conclusions. In each of the analyses, the mean of test trials one and two was calculated and the mean looking time to the novel and familiar scenes was entered as the dependent measure. First, we analyzed the infant data by placing all the infants from Experiments 2 and 3 in one analysis and examined for

General discussion

In these experiments a preferential-looking technique was used to study how both infants and adults categorize spatial relations. The advantage of the preferential looking technique is that the same stimuli, presentation and dependent measure can be used with these widely varying populations. The stimuli were filmed events of dynamic spatial relations being demonstrated with widely varying objects. Because looking preferences to nonverbal stimuli rather than verbal responses were examined, we

Acknowledgements

This research was made possible by NSF Grant 9973399 to Laraine McDonough, NSF Grant 970895 to Jean Mandler and Laraine McDonough. We would like to thank Melissa Bowerman who stimulated our thinking on the relationship between thought and language. We also appreciate the assistance of Lisa Hoag, Laura Holmes and several undergraduate students at UCSD for all their help without which this work would not be possible.

References (58)

  • R. Baillargeon

    How do infants learn about the physical world?

    Current Directions in Psychological Science

    (1994)
  • R. Baillargeon

    A model of physical reasoning in infancy

  • M. Bierwisch

    Some semantic universals of German adjectivals

    Foundation of Language

    (1967)
  • A.H. Bloom

    The linguistic shaping of thought: A study in the impact of language on thinking in China and the west

    (1981)
  • L. Bloom et al.

    The convergence of early cognition and language in the second year of life: Problems in conceptualization and measurement

  • M. Bowerman

    Learning a semantic system: What role do cognitive predispositions play?

  • M. Bowerman

    The origins of children’s spatial semantic categories

  • M. Bowerman

    Learning how to structure space for language: A crosslinguistic perspective

  • Bowerman, M. & Choi, S. (1994). Linguistic and nonlinguistic determinants of spatial semantic development. Paper...
  • M. Bowerman et al.

    Shaping meanings for language: Universal and language-specific in the acquisition of spatial semantic categories

  • M. Bowerman et al.

    Verbs, particles, and spatial semantics: Learning to talk about spatial actions in typologically different languages

  • P. Brown

    The Ins and Ons of Tzeltal locative expressions: The semantics of static descriptions of location

    Linguistics

    (1994)
  • R. Brown

    A first language: The early stages

    (1973)
  • S. Carey

    Semantic development: The state of the art

  • S. Choi

    Caregiver input in English and Korean: Use of nouns and verbs in book-reading and toy-play contexts

    Journal of Child Language

    (2000)
  • S. Choi et al.

    Early acquisition of verbs in Korean

    Journal of Child Language

    (1995)
  • K. Emmorey

    The confluence of space and language in signed languages

  • M.A. Hunter et al.

    Effects of stimulus complexity and familiarization time on infant preferences for novel and familiar stimuli

    Developmental Psychology

    (1983)
  • M. Johnson

    The body in the mind: The bodily basis of meaning, imagination, and reasoning

    (1987)
  • Cited by (195)

    • Cognitive processing of spatial relations in Euclidean diagrams

      2020, Acta Psychologica
      Citation Excerpt :

      In cognitive psychology, researchers have investigated the mental representations of spatial relations (e.g., Bryant & Tversky, 1999; McNamara, 1986; Stevens & Coupe, 1978), the use of spatial relations in cognitive maps (e.g., Kuipers, 1982; Kuipers, 1983; Tversky, 1993), and various forms of reasoning with spatial relations (e.g., Byrne & Johnson-Laird, 1989; Knauff, 2013). In linguistics, research has focused on spatial prepositions (Coventry & Garrod, 2004; Landau & Jackendoff, 1993; Retz-Schmidt, 1988) such as ‘part of’ (Winston, Chaffin, & Herrmann, 1987), ‘in’ and ‘on’ (McDonough, Choi, & Mandler, 2003), or ‘in the front/back of’ and ‘on the left/right of’ (Franklin, Tversky, & Coon, 1992), which have been investigated through empirical studies (e.g., Hayward & Tarr, 1995; Munnich, Landau, & Dosher, 2001) and the development of computational models (e.g., Kelleher & Costello, 2009). In artificial intelligence, researchers have been concerned, among other things, with the design of programs capable of learning and recognizing spatial relations in visual scenes (e.g., Winston, 1970), the construction of mathematical models of how humans encode spatial relations in pictures (e.g., Freeman, 1975), the formal representation of topological spatial relations (e.g., Egenhofer & Franzosa, 1991), and the development of formal calculi for representing and reasoning about regions and their connections (e.g., Cohn & Hazarika, 2001; Randell, Cui, & Cohn, 1992).

    View all citing articles on Scopus
    View full text