Moving Beyond Static Features: Expanding the Scope of Learned Categorical Perception Research
Sam Schwamm, Vassar College ’16, Benjamin Chin, Vassar College ’15 and Profs. Jan Andrews and Ken Livingston
Learning to categorize objects is known to have systematic effects on how those objects are judged (e.g., similarity or same-different judgments). These so-called learned categorical perception (CP) effects have been demonstrated with a wide variety of stimuli, particularly visual stimuli. However, it does not appear that they have ever been explored with visual motion features that are arguably characteristic of many real categories, such as animals and vehicles. This project’s goal is to develop visual stimuli that vary on dimensions such as speed or direction of movement in order to test for learned CP effects in categorizing such stimuli.
In an initial study, subjects were randomly assigned to learning or control conditions. Stimuli were simple line drawings, resembling tables, which moved across a computer screen in a triangle wave. They varied on two dimensions: arc height of the overall trajectory and the maximum angle of leg movement. Two categories were defined based on these two dimensions, and the learning group was trained to classify the stimuli. Both groups were then asked to perform a similarity rating task for within- and between-category pairs. Overall, category learning produced an increase in the judged similarity of within-category pairs, indicating a compression effect and thus extending the generality of previous findings. The challenge for future research will be discovering ways to manipulate the motion features of experimental stimuli in systematic yet natural ways.