Abstract Twenty subjects identified filtered pictures of previously learned target stimuli. Five filters were utilized: 3 two-octave wide band-pass and 2 complementary (same cutoff) high- and low-pass. Response times and per cent errors were used to assess performance. The filtered pictures were presented at two sizes: to ten subjects at twice the size presented to the other ten. The results indicated that higher spatial frequencies contribute more to the identification task than do the low spatial frequencies, but also that neither low nor very high frequencies are redundant for identification. It was also seen that both the proximal (c/deg) and the distal (c/picture) scales of spatial frequency measurement are involved in the identification process.