Fig. 14.

(a) Schematic face space showing the identity axis used to generate stimuli in this study along with the axes employed by Wilson et al. (2002). Dark grey and light grey discs indicate faces with a just-noticeable-difference (JND) in identity for this and the Wilson et al. study, respectively. Note the very small JND at an intermediate level of identity for this study, and that JNDs are greater at higher levels of identity in the Wilson et al. study. Because different projections were used to generate stimuli in both studies, their findings are not mutually inconsistent. (b) Results from a control experiment conducted to examine if dips in TvI functions depended on having the target-face as a pedestal. The solid line shows the non-linear TvI function derived in the earlier experiment, and the open symbols shows data collected using stimuli that were essentially identical except that the pedestal-face was now not matched to the target. As the identity level of the mismatched pedestal-face increases we observe a steady increase in masking (with no dips).