The question of whether infants and older toddlers differentiate the world of objects into fundamentally similar kinds of categories has persisted for decades, largely due to the absence of methods for assessing categorization that are effective across a wide variety of ages. The concepts of preverbal infants must be examined through techniques that are independent of language, while language production and comprehension typically are vital to the assessment of toddlers' early concepts. Determining whether younger and older infants possess similar types of concepts is relevant to broader questions concerning the degree to which cognitive development is continuous or discontinuous. On the one hand, differences between the category extensions of infants and older toddlers may be due to different methods of assessment or to different levels of experience with objects. Even within a single age group, infants may produce different patterns of category extension across different categorization tasks because the tasks vary in terms of attentional and other information-processing demands. On the other hand, fundamental differences may exist between prelinguistic infants and older infants who know names for things because they possess qualitatively different representational systems. By this account, within-group variations in categorization emerge because different procedures tap different categorization processes. This research will introduce a new method, a comprehension measure of associative generalization, that is intended to help determine the basis for between-group and within-group variations in categorization during infancy. In this method, preferred looking will serve as an index of infants' understanding of conceptual boundaries and will correct for methodological problems associated with production-based responses such as imitation and labeling.
Throughout three consecutive phases of the project, the comprehension measure of associative generalization will be used to explore infants' understanding of the toy replicas often used in studies of early categorization as symbols for the corresponding real objects, and to differentiate among three alternative developmental sequences in the emergence of concepts: (1) Domain-first: Infants' early concepts are global in nature (e.g., at the level of animal or vehicle) and broad concepts control infants' associations of category properties; (2) Basic-first: Infants' initial concepts are at the basic level (e.g., dog, cat, car, or truck), and domain-level generalizations obtained through production-based imitation tasks are attributable to overextensions of basic level categories; and (3) Variable specificity: Infants possess rudimentary knowledge that different types of properties are aligned with categories at different levels of specificity. The last would imply that certain properties are generalized at the basic level, and others are generalized at the domain level; such awareness would constitute evidence for rudimentary understanding of the asymmetry of class inclusion relations.
The development of a comprehension-based analog to productive imitation measures has tremendous potential to serve as a methodological "bridge" between tasks appropriate for younger infants (e.g., habituation) and tasks suitable for older toddlers (e.g., naming, successive touching). Such a tool will enable researchers to sensitively test the degree to which categorization processes remain stable throughout the first two years of life and will provide useful data for the development of intervention programs aimed at facilitating early lexical development.
