|
Article Excerpt
It is often assumed that processing of faces is somehow special as it involves the efficient use of relational information (Leder & Bruce, 2000). In the present study, we investigate whether processing of relational information (distances between local components) is also involved when faces differ by components, since some types of componential differences (e.g. shape changes) might affect relational information (e.g. distance between edge of nose and corner of eye) even if the location of the feature's centre does not change.
Faces contain different classes of information (Bartlett & Searcy, 1993; Leder & Bruce, 1998, 2000; Rhodes, Brake, & Atkinson, 1993). All faces consist of components such as eyes, noses, and mouths, which can differ from each other in terms of shape, size, protuberance, and so on. Due to the first order spatial arrangement of these components (i.e. eyes above nose), faces also differ in respect to an individual spatial arrangement (Rhodes, 1988), which we call relational information. The processing of such relational information (such as the eye distance or nose-mouth distance) for face recognition is often called configural processing.
Inversion specifically disrupts the processing of this kind of information in faces (Bartlett & Searcy, 1993; Carbon & Leder, 2005; Carbon, Schweinberger, Kaufmann, & Leder, 2005; Leder & Bruce, 1998, 2000; Leder & Carbon, 2004; Rhodes et al., 1993). For example, turning faces upside down reduce effects of distinctiveness based on the relation between individual features (Leder & Bruce, 1998). As a consequence, faces that appear grotesque when viewed normally (because of unusual arrangements of the features) do not look correspondingly grotesque when viewed upside down (Bartlett & Searcy, 1993; Carbon & Leder, 2005; Carbon et al., 2005; Murray, Yong, & Rhodes, 2000).
In a precursor to the present study, Leder and Bruce (2000) tested inversion effects for faces that differed either in respect to relational information or in a combination of relational and component information. They found that inversion deficits were predicted primarily by differences in relational information. As a result of this and of other studies (e.g. Mondloch, Le Grand, & Maurer, 2002), the size of inversion effects has come to be regarded as an index of the amount of configural processing that takes place in any particular task.
In the present study, we tested faces varying in respect to three different classes of information--faces that share the same shape of local parts and spacing between these parts but which differ in colour (colour); faces with common individual features in terms of shape and colour but which differ in the spacing between these features (relational); faces whose individual components (eyes, nose, and mouth) differ but whose spatial layout remains constant (component). Our intention was to compare the final condition with conditions that are known to produce large (relational) or no inversion effects (colour).
Our study was necessary because of the ambiguity of effects when faces are constructed using changed components. Some experiments in the literature have revealed that exchanging facial components with those from other faces can produce inversion effects (e.g. Rhodes et al., 1993). More recent findings by Yovel and Kanwisher (2004) have demonstrated that the processing of faces for which components or relations between these components were varied, did show highly similar activation patterns in fMRI data. However, other evidence suggests that feature-based processing of faces is hardly affected by inversion (Carbon & Leder, 2005; Collishaw & Hole, 2000; Freire, Lee, & Symons, 2000; Murray et al., 2000). Of course, faces differing in components also inevitably differ in respect to relational information. For example, a longer nose affects the relative size of the nose-mouth distance. According to Collishaw and Hole, faces differing in terms of relational information only will be processed exclusively in terms of configuration. If such faces are inverted, configural processing is strongly impaired, making it difficult to differentiate them. Faces differing in terms of components, on the other hand, will be processed both in terms of configuration and features (Leder & Carbon, 2004). Because featural processing is hardly affected by inversion, the ability to discriminate between such faces, when inverted, will remain relatively high.
In the present study, we investigate whether faces differing from each other in terms of components sometimes show...
|
