Parsing Objects into Categories
Barbara A. Younger · 2003
Abstract How do infants form categories? Some time ago, Rosch and her colleagues (e.g., Rosch, 1978; Rosch, Mervis, Gray, Johnson, & Boyes-Braem, 1976) noted that natural object categories reflect the correlational structure of the environment. They pointed out that attributes are not randomly distributed across objects but that clusters of attributes occur with regularity. Thus, a category learner programmed in some sense to detect and utilize correlational information would tend to form distinctive categories (Medin, 1983). In this chapter, I summarize research focused on the infant as such a category learner. In contrast to the top-down or theory-driven approach to the use of correlated attributes typically presented for child and adult category learners (e.g., Barrett, Abdi, Murphy, & Gallagher, 1993; Medin, Altom, Edelson, & Freko, 1982), the view presented in this chapter is of an infant who is attuned to statistical regularities in the input (see Saffran, Aslin, & Newport, 1996). is presented indicating that infants (and preschool-aged children) are able to detect and utilize correlations among attributes in forming novel categories.