Individuation of objects and events: a developmental study

Abstract

This study investigates children's ability to use language to guide their choice of individuation criterion in the domains of objects and events. Previous work (Shipley, E. F., & Shepperson, B. (1990). Countable entities: developmental changes. Cognition, 34, 109–136.) has shown that children have a strong bias to use a spatio-temporal individuation strategy when counting objects and that children will ignore a conflicting linguistic description in favor of this spatio-temporal bias. Experiment 1 asked children (3-, 4-, and 5-year-olds) and adults to count objects and events under different linguistic descriptions. In the object task, subjects counted pictures of familiar objects split into multiple pieces (as in Shipley, E. F., & Shepperson, B. (1990). Countable entities: developmental changes. Cognition, 34, 109–136.) and described either using an appropriate kind label (e.g. “car”) or the general term “thing”. In the event task, subjects watched short animated movies consisting of a goal-oriented event achieved via multiple, temporally separated steps. The events were described either with an appropriate telic predicate targeting the goal (e.g. “paint a flower”) or with an atelic predicate targeting the steps in the process (e.g. “paint”) and the subjects' task was to count the events. Relative to adults, children preferred a spatio-temporal counting strategy in both tasks; there was no difference among the three groups of children. However, children were able to significantly change their counting strategy to follow the linguistic description in the event but not the object task. Experiment 2 extended the object task to include counting of other types of non-spatio-temporal units such as sub-parts of objects and collections. Results showed that children could use the linguistic descriptions to guide their counting strategy for these new items, though they continued to show a bias for a spatio-temporal individuation strategy with the collections. We suggest potential cognitive origins for the spatio-temporal individuation bias and how it interacts with children's developing linguistic knowledge.

Introduction

The world provides us with an undifferentiated stream of experience, full of sounds and sights, surfaces and motions. Making sense of it requires (among other things) breaking that stream into individual units such as tables, dinners and events of eating. There are many bases for creating units from the stream of experience. One can divide the world into spatio-temporally defined units such as bounded coherent objects that trace spatio-temporally continuous paths or units of sound segregated from the background by the equivalent of edge discontinuities (e.g. a clap). Alternatively, one can discover units on the basis of transitional probabilities formulated over a lower level of analysis (as in Saffran, Aslin, and Newport's (1996) demonstrations that words are segmented from a spatio-temporally continuous stream on the basis of co-occurrences of syllables or phonemes). Finally, one can also impose units on the basis of conceptually mediated kind or causal analysis. Events of building a house are complete when the house is finished. An animal ceases to exist when it dies, in spite of the spatio-temporal continuity of its body.

Languages both reflect and impose criteria of individuation: successful communication requires that speakers of a given language individuate the world into the same units as do the other speakers of that language. This observation raises several related developmental questions: how do children come to individuate the undifferentiated stream of experience into the same units as do adults, when and how do they learn their language's means of doing so, and under what circumstances are children and adults guided by others' linguistic descriptions to focus on one possible unit of experience over another? This study focuses on the third of these questions as an initial wedge into the first two, building on earlier seminal studies by Shipley and Shepperson (1990) and extending them to event representations.

Objects and events are generally encoded by different sorts of linguistic units (typically nouns and predicates, respectively). Object individuation is implicated in the count/mass distinction whereas event individuation is implicated in aspectual marking on verbs. Nonetheless, there are deep parallels between the ways humans individuate objects and events.

By late in the first year of life, infants control various different criteria for object individuation, and some aspects of language already influence the infants' division of the stream of experience into units. A wealth of data suggest that spatio-temporal criteria are important to individuation of objects, even very early in infancy. Young infants analyze spatio-temporal continuity in establishing representations of one object or two. For example, objects seen emerging, one at a time, from opposite sides of two screens but never appearing in the middle are analyzed as two objects, not one (Spelke et al., 1992, Xu and Carey, 1996). In addition, by 12 months of age, at least, infants are able to bring kind membership to bear on the problem of object individuation. Xu and Carey (1996) showed that 12-month-old infants shown objects of two different kinds emerging one at a time from opposite sides of a single screen established representations of two numerically distinct objects. In this case, there was no spatio-temporal evidence for distinct individuals. Twelve-month-olds were similarly successful at using kind information to guide the number of times they reached into an opaque box for objects they had seen emerge from and return to the box one at a time (Van de Walle, Carey, & Prevor, 2000). Xu (2002) has shown an effect of labeling on object individuation in 9-month-olds. In the paradigm of Xu and Carey (1996), if the objects are labeled with contrastive labels as they are taken from and replaced behind the screen (“Look a duck, look a car”), 9-month-olds establish representations of two objects, whereas if labeled with a single label (“Look a toy, look a toy”), infants fail, as they do with no label or with contrastive associated sounds.

Shipley and Shepperson (1990) used counting as a wedge into preschool children's criteria for individuation of objects, asking whether language influenced what units children counted. They presented children with displays of objects, some of which were broken. For example, one display might be three intact cars and one car broken into two distinct pieces. Even when asked to count “the cars”, children as old as 5 years would often separately count each separate piece of the broken car, along with each intact car. This is a surprising result; children have known the words in Shipley and Shepperson's study (e.g. “car”) for years by age 5, and the difference between cars and other things has been a cue for individuation by 12 months of age. Still, when considering a scene consisting of whole cars and split cars, children broke it into spatio-temporally determined units, not units determined by the basic level kind term “car”. Further, children were not influenced by the label – the same pattern emerged whether asked to count “the cars” or “the things”. Thus, spatio-temporal cues persist as a powerful basis of object individuation into childhood. Preschool children who have mastery of several bases of object individuation will default to spatio-temporal cues in at least one kind of difficult counting situation. Adults, in contrast, did not count pieces of cars as “cars”, and were significantly influenced by the label.

There has been much less work on event individuation than on object individuation. There is some evidence, though, for a spatio-temporal specification of individual events that is roughly parallel to the spatio-temporal specification of individual objects. Wynn (1996) showed that infants habituated to two jumps of a puppet dishabituated when shown three jumps, and vice versa. She controlled for a variety of co-occurring factors such as total event duration and inter-jump intervals, strengthening the case that it was in fact the number of individual jumps that the infants were enumerating. The spatio-temporal information infants used was subtle, going beyond mere temporal pauses between motion events: infants succeeded if the puppets were still, and if the puppets continued to move in a waggling fashion between jumps. Apparently infants analyze discontinuities in kinds of motion. Later studies by Sharon and Wynn, 1998, Sharon and Wynn, 2000 have argued for other factors such as cyclicity of motion and tangent discontinuity of motion as critical for parsing events into individual units. This information is still spatio-temporal, of course.

As in the case of objects, events are also individuated on the basis of higher level conceptual criteria, such as the intentions and goals of the actor. There is strong evidence that infants represent events in terms of goal states by the end of the first year (Gergely et al., 1995, Csibra and Gergely, 1996, Csibra et al., 1998, Woodward, 1998, Biro et al., 1996), but it is unclear when infants use this information for individuation purposes. In one suggestive study, Baldwin, Baird, Saylor, and Clark (2001) habituated 10-month-old infants to short video clips of a woman performing a goal-directed action, such as picking a towel up off of the floor. During the test trials, pauses were inserted into the movie, either at the moment the goal was achieved or else at some point before or after the goal. Infants looked longer when the breaks did not coincide with the goal state, suggesting that the goal was a natural break point in the infants' representation of the stream of experience.

The world may be ambiguous, but linguistic descriptions commit the speaker to one among all of the available possible interpretations. This generalization applies to ambiguity regarding individuation as well as ambiguity regarding categorization. Within the object domain, the role of language in individuation is relatively straightforward. Count nouns provide criteria for individuation and numerical identity (see Macnamara, 1986). Whereas Xu (2002) found that infants inferred from contrasting labels that there were two individuals in an event, the role of count nouns in providing cues for individuation is considerably more subtle. Different kinds of count nouns provide different kinds of individuation criteria – contrast “cow” and “body” and “calf” and “pet” and “herd” and “leg”. A calf is not a different individual from the cow it becomes, and a cow does not cease to exist when it ceases to be a pet, but a cow does cease to exist when it dies, in spite of the continued existence of its body. A herd is a collective individual; spatio-temporally individuated cows make up its parts. A leg is an individual part of a spatio-temporally specified cow. Experiment 2 begins to explore the flexibility with which preschool children impose different criteria of individuation upon objects, groups of objects, and parts of objects, and whether Shipley and Shepperson's finding that units preschool children count are uninfluenced by the “fork/thing” contrast extends to other linguistic contrasts (“cow/herd” and “cow/leg”).

The linguistics of event individuation differs greatly, on the surface, from the linguistics of object individuation. The event equivalent of a count noun is a telic predicate (Bach, 1986; see also Smith, 1991). Telic predicates specify the end-point (or telos) of the event they describe and this end-point defines the criteria for individuation and numerical identity for that event. Thus, the telic predicate “build a house” specifies that the event ends with a house; an event that stopped short of a completed house would not constitute an entire instance of the event described. Telic predicates can be contrasted with atelic ones, which do not specify the result of an event and tend to describe actions or processes, such as “work” or “play”. Events described with atelic predicates may be individuated into units spatio-temporally; some sort of pause between bouts of playing may serve to establish individuals. The linguistic marking of telicity depends on the interaction of many elements in a predicate, including the verb (“The vase breaks” is telic while “The vase moves” is not), the arguments of the verb (“The girl ate an ice-cream cone” is telic while “The girl ate” and “The girl ate ice-cream” are not), as well as adjuncts in the predicate, such as prepositional phrases (“The butterfly flew to a tree” is telic while “The butterfly flew” is not).

Given the greater linguistic complexity of describing event individuation, relative to object individuation, one might expect that children would be even less likely to be influenced by the linguistic description of an event in deciding how to break it into countable units than they are in the case of objects. However, there is some evidence that children know from early on that the telic/atelic distinction ought to have some sort of linguistic implications. Examinations of children's earliest productions of verb morphology (such as past tense marking and progressive “-ing” marking in English) show that children typically use these markers in a restrictive way, and that the restriction depends on the probable telicity of the verb in question. Verbs which, for adults, typically form the center of a telic predicate (“break”, “make”) appear with one type of morphology (in English, the past tense) while verbs which typically yield atelic predicates (“play”, “ride”) appear with the other (in English, the progressive) (Antinucci and Miller, 1976, Bloom et al., 1980, Bronckart and Sinclair, 1973, Shirai and Andersen, 1995). However, the morphology that children restrict in this way is not in fact dependent on the telicity of the predicate in the adult language. Thus, in English, we use past tense and progressive markers freely with both telic and atelic predicates regularly (i.e. “broke”, “breaking”, “played” and “playing” are all perfectly grammatical).

Investigations into children's knowledge of how telicity is actually marked in natural language have been very limited. There has been one set of studies which looked at telicity knowledge in children acquiring English and Dutch (van Hout, in press). van Hout and colleagues presented children with completed and incomplete versions of the same scenario (e.g. one elephant who drank all the water in her bucket and one who left the bucket half full) and then asked children whether telic or atelic descriptions could be used (e.g. “Did this elephant drink?”, “Did this elephant drink her bucket of water?”). Children as old as 5 years did not reliably apply the telic description only to the completed version of the event.

These results suggest that it may take children years between the time that they understand something about telicity and when they understand the details of how their native language marks it. Some caution is required in interpreting these results, however. van Hout used a very small set of predicates (all involved eating and drinking), and children were asked within an item to assess both scenarios with respect to a particular sentence. Children's difficulties may be restricted to just these types of verbs, or arise from some general confusion about what to do when asked the same questions repeatedly within a trial. It remains to be seen if children might succeed with telicity marking given a simpler task and different sorts of events. Moreover, it also remains to be seen when children understand the implications of the telic/atelic distinction for event individuation.

Wynn (1990) showed that 2- and 3-year-old children can count entities that are not objects (claps, beeps and jumps). These units were spatio-temporally defined. It is not known whether young preschoolers can also count more conceptually determined event units, specified by goals, for example. Nor is it known whether they can use the language with which an event is described as a basis for one construal over another. If the results with events parallel those with objects, then preschoolers should be able to count spatio-temporally defined units only, no matter how the events are described. In Experiment 1 we replicate Shipley and Shepperson's object counting studies and extend the methodology to begin to explore preschool children's event individuation and their mastery of linguistic devices that determine how the stream of experience is to be broken into event units.