According to the “faces are special” hypothesis, faces per se are processed differently from other objects. One alternative theory, however, is the “within-class discrimination” hypothesis, which predicts that face-like processing will be found for objects where individual level discrimination is required. We present three experiments testing this prediction. Object stimuli were Labrador retrievers. Similar to faces, these are naturally occurring, have an pre-experimental upright, and their shape is partly defined by shading. Our first experiment replicated Diamond and Carey's (1986) finding of a much larger inversion effect on memory for faces than for dogs. The second experiment examined effects of contrast reversal (photographic negative) in an identity matching task (cf. Subramaniam & Biederman, 1997). Stimulus pairs were both normal contrast or both reversed contrast. Contrast reversal effects were large for upright faces, smaller for inverted faces, and very small for dogs (in either orientation). The third experiment employed a direct test for holistic/configural processing (proposed to occur only for faces). The top half of one face (or dog) was combined with the bottom half of a different face (dog) to form a composite, with the two halves either aligned or unaligned. For upright faces, matching two top halves was harder in the aligned than the unaligned condition, indicating perceptual integration and replicating Young et al. (1987). There was no composite effect, however, for inverted faces or for dogs. We conclude that requiring within-class discrimination does not produce the same results for dogs as faces, and thus that the within-class discrimination hypothesis cannot explain “special” face processing. The present experiments also lay the groundwork for future exploration of the “expertise” hypothesis, namely the idea that face-like processing might occur when subjects are experts at within-class discrimination (e.g., dog show judges).