Abstract
Some physical continua (e.g. color, speech sounds, and central to the current study, emotional facial expressions) are perceptually grouped into discrete categories. We tested the category boundary for four continua of 11 morphed photographs representing physically equidistant images with “happy” and “sad” as endpoints. Here, we compare results from a 2AFC keypress identification to categories derived from participant's anticipatory eye movement data. Experiment 1 relied on an identification task in which participants saw continuum endpoint images (Unambiguous images) and selected “happy” or “sad” via keypress. Then they were asked to make the same decision for the five middle images (Ambiguous images) from the same continuum. Participants clearly categorized emotion images as happy or sad except for one image which elicited chance performance, suggesting a category boundary at this location. In Experiment 2, naïve participants saw a subset of these same stimuli in a visual expectation paradigm. During a training phase, the emotion (happy or sad) of an Unambiguous image displayed in the center of the screen predicted the location of a visual reward (e.g. a happy image was followed by a cartoon to the left, a sad image by a cartoon to the right). This visual reward appeared 750 ms after the central face image, allowing anticipatory eye movements to be recorded with an eye-tracker. After training, participants were presented with the five Ambiguous images and anticipatory eye movements were recorded. Experiments 1 and 2 revealed similar results: The middle images of the continuum had the most variable responses. The presence of a category boundary was further suggested by the greater number of non-responses (participants failing to show anticipatory eye movements) for images near the expected category boundary. The visual expectation paradigm is a viable methodology for the study of category boundaries.