ABSTRACT
J. Sergent was probably the first to explicitly suggest that different levels of resolution might be involved in the performance of different facial processing tasks. An investigation by A. Fiorentini, L. Maffei, and G. Sandini, using an identification rather than a matching measure, produced a pattern of results that, broadly speaking, has been confirmed by a number of other investigators. They examined identification with low-pass and high-pass filtered facial images. Face identification or recognition under normal conditions, subjectively at least, appears to be a fast process, and certainly we are not conscious of pondering slowly about whether a face is familiar or not, although providing the name or semantic information about the individual may sometimes be slow and problematic. In conclusion, one can say there is strong evidence pointing to the relative importance of a central range of facial frequencies, between eight and sixteen cycles, in recognition and identification.
