Abstract
Visual categorization, the ability to group objects that are visually alike, is an essential aspect of cognition that facilitates recognition, inference, and generalization from infancy onward. Several methods test the development of this ability, most notably generalized imitation, and habituation. Imitation-based methods disadvantage younger infants with less developed motor control, making habituation the standard approach. Habituation relies on the tendency for infants to look away when bored with a stimulus, with shorter looking times interpreted as reflecting greater familiarity with the stimulus. Thus, infant category learners decrease looking when presented with several examples of a particular category, and increase looking to a new object class. We offer an alternative paradigm based on the anticipatory looking procedure of McMurray and Aslin (McMurray & Aslin, 2004), which demonstrated that infants anticipate the repeated emergence of an object from behind an occluder. We extend the idea to test infants' ability to learn categories. In our design, we present infants with a large tube that contains a centralized entry point at the top and two separate left and right exits at the bottom. Over learning, infant participants learn to associate each of two different categories with specific exit points, as represented by shorter latencies and anticipatory looking to the correct exit tube. The advantage of such an approach is that unlike habituation, each familiarization trial tests category understanding. In our study, 18 infants from 5-24 months learned two perceptually similar artificial categories. 83% showed shorter latencies to fixate the emerging object on trials 5-8 versus trials 1-4. Furthermore, the percentage of correct anticipatory looks (<150ms after initial emergence from tube) increased from 20% to 31%. These results demonstrate the utility of the approach for evaluating categorization ability in infants.
Meeting abstract presented at VSS 2014