Face recognition is thought to rely on specific mechanisms underlying a perceptual bias toward processing faces as undecomposable wholes. This face-specific "holistic processing" is commonly quantified using 3 measures: the inversion, part-whole, and composite effects. Consequently, many researchers assume that these 3 effects measure the same cognitive mechanism(s) and these mechanisms contribute to the wide range of individual differences seen in face recognition ability. We test these assumptions in a large sample (N = 282), with individual face recognition abilities measured by the well-validated Cambridge Face Perception Test. Our results provide little support for either assumption. The small to nonexistent correlations among inversion, part-whole, and composite effects (correlations between -.03 and .28) fail to support the first assumption. As for the second assumption, only the inversion effect moderately predicts face recognition (r = .42); face recognition was weakly correlated with the part-whole effect (r = .25) and not correlated with the composite effect (r = .04). We rule out multiple artifactual explanations for our results by using valid tasks that produce standard effects at the group level, demonstrating that our tasks exhibit psychometric properties suitable for individual differences studies, and demonstrating that other predicted correlations (e.g., between face perception measures) are robust. Our results show that inversion, part-whole, and composite effects reflect distinct perceptual mechanisms, and we argue against the use of the single, generic term holistic processing when referring to these effects. Our results also question the contribution of these mechanisms to individual differences in face recognition. (PsycINFO Database Record

No abstract

Recent evidence has shown that face space represents facial identity information using two-pool opponent coding. Here we ask whether the shape of the monotonic neural response functions underlying such coding is linear (i.e. face space codes all equal-sized physical changes with equal sensitivity) or nonlinear (e.g. face space shows greater coding sensitivity around the average face). Using adaptation aftereffects and pairwise discrimination tasks, our results for face attributes of eye height and mouth height demonstrate linear shape; including for bizarre faces far outside the normal range. We discuss how linear coding explains some results in the previous literature, including failures to find that adaptation enhances face discrimination, and suggest possible reasons why face space can maintain detailed coding of values far outside the normal range. We also discuss specific nonlinear coding models needed to explain other findings, and conclude face space appears to use a mixture of linear and nonlinear representations.

Face aftereffects are widely studied on the assumption that they provide a useful tool for investigating face-space coding of identity. However, a long-standing issue concerns the extent to which face aftereffects originate in face-level processes as opposed to earlier stages of visual processing. For example, some recent studies failed to find atypical face aftereffects in individuals with clinically poor face recognition. We show that in individuals within the normal range of face recognition abilities, there is an association between face memory ability and a figural face aftereffect that is argued to reflect the steepness of broadband-opponent neural response functions in underlying face-space. We further show that this correlation arises from face-level processing, by reporting results of tests of nonface memory and nonface aftereffects. We conclude that face aftereffects can tap high-level face-space, and that face-space coding differs in quality between individuals and contributes to face recognition ability.