Leslie B. Cohen

Labels can override perceptual categories in early infancy

An extensive body of research claims that labels facilitate categorisation, highlight the commonalities between objects and act as invitations to form categories for young infants before their first birthday. While this may indeed be a reasonable claim, we argue that it is not justified by the experiments described in the research. We report on a series of experiments that demonstrate that labels can play a causal role in category formation during infancy. Ten-month-old infants were taught to group computer-displayed, novel cartoon drawings into two categories under tightly controlled experimental conditions. Infants were given the opportunity to learn the two categories under four conditions: Without any labels, with two labels that correlated with category membership, with two labels assigned randomly to objects, and with one label assigned to all objects. Category formation was assessed identically in all conditions using a novelty preference procedure conducted in the absence of any labels. The labelling condition had a decisive impact on the way infants formed categories: When two labels correlated with the visual category information, infants learned two categories, just as if there had been no labels presented. However, uncorrelated labels completely disrupted the formation of any categories. Finally, consistent use of a single label across objects led infants to learn one broad category that included all the objects. These findings demonstrate that even before infants start to produce their first words, the labels they hear can override the manner in which they categorise objects.

Infants' object examining: Habituation and categorization

Abstract The visual habituation paradigm has dominated the study of infant object discrimination and categorization. A more active task, object examining, was used in two studies to explore early discrimination and categorization, and to validate previous findings. The object-examining task combined active exploration of real objects with some aspects of a habituation-dishabituation paradigm. The first study explored simple discrimination. Six- and 10-month-old infants were familiarized with a single object and then tested with two novel objects. The results indicated that at both ages, infants showed clear discrimination among objects. The second study explored infant categorization and revealed that when both 6- and 10-month-old infants were familiarized with a category of objects, they responded to novel objects in terms of their categorical membership. These results converge with previous findings obtained by using more traditional methods, and demonstrate the utility of an object-examining task for the exploration of cognitive abilities in the first year of life. Finally, a comparison of two different measures of attention used in this task, examining and looking, revealed that examining time was a better measure of active processing of information about objects than was looking time.

How infants process addition and subtraction events

Three experiments are described that assess 5-month-old infants’ processing of addition and subtraction events similar to those reported by Wynn (1992a). In Experiment 1, prior to each test trial, one group of infants was shown an addition event (1 + 1) while another group was shown a subtraction event (2 − 1). On test trials, all infants were shown outcomes of 0, 1, 2 and 3. The results seemed to require one of two dual-process models. One such model assumed that the infants could add and subtract but also had a tendency to look longer when more items were on the stage. The other model assumed that infants had a preference for familiarity along with the tendency to look longer when more items were on the stage. Experiments 2 and 3 examined the assumptions made by these two models. In Experiment 2, infants were given only the test trials they had received in Experiment 1. Thus, no addition and subtraction or familiarity was involved. In Experiment 3 infants were familiarized to either one or two items prior to each test trial, but experienced no actual addition or subtraction. The results of these two experiments support the familiarity plus more items to look at model more than the addition and subtraction plus more items to look at model. Taken together, these three experiments shed doubt on Wynn’s (1992a) assertion that 5-month-old infants can add and subtract. Instead they indicate the importance of familiarity preferences and the fact that one should be cautious before assuming that young infants have sophisticated numerical abilities.

Infant categorization of containment, support and tight‐fit spatial relationships

Two experiments examined infant categorization of containment, support or tight-fit spatial relationships. English-learning infants of 10 months (Experiment 1) and 18 months (Experiment 2) were habituated to four pairs of objects in one of these relationships. They were then tested with one event from habituation, one with novel objects in the familiar relationship, one with familiar objects in a novel relationship and one with novel objects in a novel relationship. Infants at both ages generalized their habituation of the containment relationship to novel objects in this relationship. In the support and tight-fit conditions, the younger infants responded only to the novel objects in the test while the older infants responded to the novel relationship, but only with familiar objects. The results indicate that infants learn to categorize containment prior to support or tight-fit relationships and suggest that infants can recognize a relationship between familiar objects prior to novel objects.

A constructivist model of infant cognition

Abstract We propose six Information-Processing Principles (IPPs) that together describe a constructive, hierarchical system by which infants come to understand objects and events in the world around them. We then demonstrate the applicability of these principles to four specific domains of infant perception and/or cognition, (i.e., form perception, object unity, complex pattern perception, and understanding of causal events). In each case empirical developmental changes appear to be consistent with the IPPs. We then present the Constructivist Learning Architecture, a computational model of infant cognitive development. This model is based on the IPPs, and uses self-organizing, neurally based techniques from Kohonen (1997) and Hebb (1949) . We then apply the model to the complex domain of infant understanding of causal events, and replicate many of the developmental changes found empirically. Finally, we discuss the applicability of this constructivist approach to infant cognitive development in general.

Precursors to infants' perception of the causality of a simple event

Abstract Three habituation experiments examined the perception of causal and non-causal events by infants at 6 1/4, 5 1/2, and 4 months of age. Experiment 1 replicated previous studies showing that by 6 1/4 months, infants begin to respond to the interaction of simple objects in Michottian type launching events on the basis of causality. In Experiments 2 and 3, however, younger infants exposed to the identical event sequences responded on the basis of simpler perceptual features of the launching events, and not on the basis of causality. 4-month-old infants responded to continuous versus non-continuous movement. 5 1/2-month-old infants also responded somewhat on the basis of continuous movement. However, in addition they responded on the basis of the spatial and temporal perceptual features of the events. This change in the pattern of results over age suggests a multi-step developmental progression.

Infant perception of angular relations

The ability of infants to perceive an angle as the relationship between two lines as opposed to independent line segments was examined in 6- and 14-week-old infants. Sixteen infants at each age were habituated to a single angle and then tested with four stimuli designed to differentiate between perception based upon angularity versus the orientation of the line segments. 14-week-old infants dishabituated only to a change in angle and not to a change in orientation. However, the 6-week-old infants did the opposite, dishabituating only to a change in orientation. These results supported the claim that 2- to 4-month-old infants perceive angular relations but tended to refute the claim that this ability is innate. Instead they may indicate an important developmental shift in perceptual ability sometime after 6 weeks of age.

An Examination of Interference Effects in Infants' Memory for Faces.

2 experiments with 18-week-old infants employed an interference paradigm to study infant visual memory. The infants were habituated to a repeated stimulus, given several trials with 1 or more interpolated stimuli, and then tested for recognition of the original stimulus. In Experiment 1, both amount of exposure and the similarity of the potentially interfering material to the original habituation stimulus were manipulated. Neither produced a significant interference effect, although some effect may have been shown by subjects who habituated to the interference stimult did demonstrate more clearly long-term retention of the originally remembered material. It was concluded from both studies that infant visual memory is a robust phenomenon, relatively immune to interference under most conditions.

Infants’ Use of Functional Parts in Basic‐like Categorization

An experiment with the sequential touching technique investigated the role of object parts on 1- to 2-year-old infants’ ability to form basic-level categories (cows and cars) from two different superordinate domains. Using the novel task design developed by Rakison and Butterworth, infants were tested with normal category exemplars as well as modified versions that were made by removing or attaching object parts (legs and wheels). Results revealed a developmental trend whereby infants’ use of object parts in categorization decreased with age. Analyses of infants’ functional responses (e.g. jumping or rolling) suggested that they might initially associate different kinds of object movement with different kinds of parts.

How Infants form Categories

Publisher Summary Categorization is an essential perceptual–cognitive activity that enables the reduction of the enormous diversity in the world to a manageable level. A typical categorization experiment involves two phases: (1) training subjects with a set of exemplars from a given category, which is followed by (2) a test with a mixture of new exemplars from the same category and nonexemplars. The degree to which subjects correctly sort these test items into category members versus nonmembers is taken as evidence for categorization. There are a number of similarities between infant and adult categorization processes. Infants, like adults, are able to abstract a prototypic representation containing the average of experienced dimensional values. By 10 months of age, infants are rather sophisticated in their categorization abilities, and these abilities develop considerably over the course of the first year. The developmental studies support the existence of two transitions across age in the perception of correlated attributes. First, there appears to be a developmental trend from the representation of feature-specific information to the representation of feature combinations. Second, there exists a transition from processing relations among features of a single pattern or object to processing correlations in the context of a category.