Abstract
The processing that underlies the subordinate level recognition of faces is a point of great debate in cognitive neuroscience. However, arguments derived from investigations attempting to discriminate between processing strategies assume a controversial premise, because it is not known what categories of stimuli are face-like. One theoretical model used to explain face recognition is the heuristic concept of “face space”. Face space presumes that identity judgements are based on the distance between an exemplar and the average of all faces experienced. The ‘average’ face sits in the centre of an array of dimensions that span out in an infinite number of directions towards a boundary that has seldom been tested. In a series of three experiments the faces of nonhumans were used as stimuli to probe the limits of face space, using both the inversion and composite effect as behavioural indices of configural/holistic processing. Since the initial data suggested that primate, but not avian faces were being processed configurally, we tested the idea that the boundary of face space might fall between different face categories, classed by species. The final experiment examined more precisely where evidence of face-like processing disappeared along a continuum of face categories that were chosen as a function of relative phylogenetic distance, to represent departures from typical human face morphology.