Humans generate knowledge by organizing perceptual information (e.g., sights, sounds and smells) into concepts (e.g., bird, music, and food) that can then be used in communication and predicting outcomes. Indeed, concepts are a major component of many cognitive processes. Accordingly, understanding how concepts are acquired has been a fundamental goal in cognitive research. An important step in achieving this goal was the discovery that humans can reduce the processing requirements in concept acquisition by restricting attention to relevant dimensions (Shepard, Hovland & Jenkins, 1961). For example, when learning to distinguish bees from wasps one can ignore their color (because their coloring is roughly the same) and attend instead to their furriness (bees are furry). In fact, the selective attention component of categorization models accounts much for their success in describing human behavior.

Selective attention may be the key to advancing categorization theory. Currently, psychologists rely almost exclusively on analyzing people's classification responses and response times to discriminate categorization models. However because concept models often predict nearly identical patterns of behavior, it is often difficult to determine which is more accurate. To overcome this problem, another source of data is necessary. My research fills this need by using eyetracking techniques to distinguish models on the basis of their different attentional predictions.

The application of eyetracking to category learning is new, and one that I have been involved in developing. As such, measuring attention independently from classification behavior is a major research goal. Importantly, my research attempts to document aspects of selective attention that have been heretofore impossible to measure. For example, I'm attempting to investigate how learners allocate attention during the first few trials of learning, how attention changes on a trial-by-trial basis (e.g., gradually or suddenly), the order in which stimulus dimensions are attended within a single trial, and (most remarkably) how attention changes AFTER an overt classification response is made (after error feedback is received).

To see an example of my current work, please download this pdf (2005)