The relative importance of featural and configural information to the perception and short-term memory for faces has been the subject of much debate. Two hypotheses regarding the utility of spatial frequency analyses in this debate have been offered. The first, a strong version of what we refer to as the correspondence hypothesis, holds that most of the information required to support facial cognition is preserved in the low spatial frequencies. A second, weaker, version suggests a correspondence between the higher spatial frequencies and the constituent features of a face and between the lower spatial frequencies and the configuration of those features. Three experiments explored these hypotheses, along with a spatial frequency-specific task dependency outlined in recent efforts by Uttal et al. (1995a; 1995b). Results provide little support for either version of the correspondence hypothesis. The findings are, in contrast, generally consistent with the spatial frequency-specific task dependency and recently established interactions among perception and memory. Implications regarding basic definitions of facial features and characteristics of processing putatively unique to faces as stimuli are discussed.