Abstract
Biederman & Kalocsai (1997) reported that sequential matching of faces, but not that of objects, suffers when the images to be matched are complementary, i.e. filtered so each contains every other combination of spatial frequency and orientation. They argued that faces are special in their reliance on spatial information. Last year, we reported that this complementation effect (CE) is not unique to faces and can be observed for objects such as chairs and cars, as well as upside down faces (Williams & Gauthier, VSS 2008). We also investigated the role of expertise in determining the magnitude of the CE. As in prior work, we measured car expertise by comparing performance on car matching to a baseline of bird matching, and observed no effect of expertise in two experiments. Here, we report new analyses of this dataset whereby we identify a possible source of heterogeneity in our sample. Participants showing high performance on bird matching (despite not being bird watchers) may differ from both typical novices and experts by adopting a perceptual strategy that leads to good performance for any subordinate-level discrimination, regardless of experience. Once such participants were excluded (∼23%), the magnitude of the CE for cars was predicted by perceptual expertise in the car domain, across two experiments. This relationship depended on discarding participants with high bird matching scores regardless of whether we used car matching or car-minus-bird matching to define expertise. This suggests that the excluded participants cannot easily be classified as typical car novices or car experts using our measure of expertise. Importantly, sensitivity to changes in spatial frequency is influenced by experience and the particularly large CE observed for upright faces may be explained by our expertise in this domain. In addition, characterizing expertise in a domain could be facilitated by measuring performance for several categories.