Considerable evidence is now available that infants can categorize during the second half of the first year of life. In addition to the previously mentioned study with pictures of stuffed animals, several studies report infant categorization of faces (e.g., Cohen & Strauss, 1979; Strauss, 1979) of three-dimensional as well as two-dimensional representations of animals (Younger & Fearing, 1998) and even adult gender categories (Leinbach & Fagot, 1993). Other studies have reported that as early as 3 or 4 months of age, infants can distinguish cats from dogs (Quinn, Eimas, & Rosenkrantz, 1993) and animals from furniture (Behl-Chadha, 1996). In addition, if one assumes that perceptual constancies may actually be a form of categorization, then there is evidence that newborns can categorize.
Consider size constancy as an example. In a newborn size constancy experiment, infants are habituated to the same object at different distances and then are tested with that same object at a new distance versus a different-sized object at an old distance. Assuming that the infant can discriminate among these distances, then the procedure amounts to a typical categorization experiment. The infants have been habituated to multiple examples of discriminably different stimuli (in this case, the same object at different distances) and then do not dishabituate to a new example (the object at a new distance) but do dishabituate to a nonexample (a new object). Of course, as adults we assume a big difference between an instance of perceptual constancy, which we interpret as different views of one item, and categorization, which we interpret as a grouping of similar but different items. It is a totally unexplored question whether infants make that same distinction and if so at what age they do it.
The Content of Infants' Categories
Although many investigators agree that even newborns may be able to categorize, they also agree that the content of those categories changes over age. It is one thing to group together different views of the same object and quite another to group together very different animals into the category of mammal or tables and chairs into the category of furniture. An important issue in this regard is the level at which infants first categorize objects. The traditional view has been that infants and young children form basic-level categories (such as dog or chair) and only later form higher-order superordinate categories (such as animal or furniture; Mervis & Rosch, 1981). Recent evidence with infants (see Quinn & Eimas, 1996, for a review) provides some support for this view. On the other hand, Mandler (2000a) has argued just the opposite. She has reported studies in which infants first appear to respond in terms of global categories (e.g., Mandler, Bauer, & McDonough, 1991). Quinn and others have also reported that infants respond more readily to global categories than to basic categories (Quinn & Eimas, 1998; Quinn & Johnson, 2000; Younger & Fearing, 2000).
To complicate matters further, one might assume Mandler would be pleased to find additional evidence supporting a priority of global categories over basic categories. However, she makes an additional distinction, also in dispute, between perceptually and conceptually based representations (Mandler, 2000b). She believes the evidence cited previously—which is based primarily upon habituation and novelty preference techniques—taps perceptual categories, whereas her studies—which use manipulation and imitation techniques—tap something independent: conceptual categories. Quinn and Eimas (1996), on the other hand, argue that there is a continuum between the two, with perceptual categories gradually developing into more abstract conceptual categories.
One of the difficulties in deciding between global versus basic levels or perceptual versus conceptual processing is that these distinctions are based upon the presumed content of the categories from the experimenter's point of view. Evidence that infants distinguish between animals and vehicles, for example, does not necessarily mean that the infants are operating at a global or superordinate level. In fact, Rakison and Butterworth (1998) have shown that in the case of toy animals versus toy vehicles, 14- and 18-month-old infants are actually responding to legs versus wheels. Much younger infants can distinguish cats from dogs, but the distinction is based primarily on features located in the face region (Quinn & Eimas, 1996). Consistent with an information-processing viewpoint, there appears to be a developmental progression from processing these features independently to processing the correlation among the features (Younger & Cohen, 1986). In fact, by 10 months of age, attention to correlations among features becomes a major factor both in the formation of categories and in the differentiation of one category versus another (Younger, 1985).
The number and variety of features to which infants attend also increases with age (Madole & Oakes, in press). One nonobvious type of feature that appears to become salient, particularly in the second year of life, is an object's function. For example, Madole, Cohen, and Bradley (1994) found that 14-month-olds but not 10-month-olds used functional information (what an object does) in their formation of categories. Madole, Oakes, and Cohen (1993) also reported a developmental shift from processing form and function information independently at 14 months to processing the relationship between form and function at 18 months, once again a developmental shift that is consistent with an information-processing view of infant perception and cognition. The increased salience of nonobvious features of objects during the second year of life, such as their function or their animacy (Rakison & Poulin-Dubois, 2001), may account at least in part for what appears to be a shift from perceptual to conceptual categorization.
Research on infant categorization is continuing at a rapid pace. Among the most exciting developments are the ties developing to other related areas of developmental and cognitive psychology. One of these ties is the relationship between infant categorization and infant language development. Lalonde and Werker (1995), for example, have shown the close tie between the use of correlated attributes in categorization and the development of speech perception. Waxman (1999) has also reported the importance of language labels in infant categorization. Close ties are also developing between infant categorization and connectionist modeling. Several attempts recently have been reported to simulate infant categorization (e.g., Mareschal & French, 2000; Quinn & Johnson, 2000). The most popular approach has been the use of simple auto-encoder models, although other more complex models are on the drawing board. These early attempts to model infant behavior in a categorization task have been remarkably successful. Future, more extensive models are likely to lead to interesting predictions as well.
Was this article helpful?