Abstract
Recently, there has been more of a focus in the face recognition literature on the perceptual representation of faces, and specifically, a focus on attempting to identify the constituent dimensions of the face space (e.g. Valentine, 1991). Of additional interest is the idea that independent face spaces may exist depending on the task at hand. For instance, is the information required to judge identity independent of the information required to judge facial expression? One approach to investigating this problem is to manipulate what range of spatial frequency information is available to the observer. In the current study, we limited spatial frequency information to one of five frequency bands, from coarse to fine, using band-pass filtering. Observers completed a match-to-sample task where they saw a sample face followed by two alternatives, which were both limited to same one of the five spatial frequency bands. Using the exact same stimuli, observers engaged in two tasks: in one task, they matched identity; in the other, they matched facial expression. This technique allows us to isolate difference between the two types of judgments and thus draw conclusions regarding the underlying representation. Various image measures were investigated to attempt to account for the results, including those based on the Fourier phase spectrum, which we believe carries the alignment information that is critical for these tasks.