Part-to-Whole Effects and Configural Processing in Faces

Abstract In two experiments, the holistic face effect (Tanaka & Farah, 1993) was investigated by using a learning paradigm with faces which differed in terms of either componential (COMP) or configural (CONF) properties. When full faces had been learned, the expected finding of advantageous recognition of holistic presentations was replicated (Exp. 1). However, when only facial parts had been learned (Exp. 2), this effect was reduced and even reversed, indicating that wholistic superiority with both sorts of faces depends on holistic learning strategies (Leder & Carbon, in press, 2005). These results are in accord with the theory of encoding specificity (Tulving & Thomson, 1973). Moreover, the effects were found for COMP and CONF faces, supporting the hypothesis that holistic and configural processing are two different aspects of face recognition. Furthermore, when the faces were inverted at test inversion effects were not only found for CONF but also for COMP faces, indicating some kind of configural processing for faces that differ in terms of facial components. Key words: whole-to-part superiority, configural, featural, holistic, face processing, face recognition Part-to-whole effects in configurai face processing Faces are amongst the most important visual patterns in our environment. They convey important information about other people, who they are, what they feel, and moreover they provide us with verbal and nonverbal signals (Bruce, 1988). Face recognition has been discussed as a candidate for a specialized processing, which allows the distinction of seemingly unlimited within-class discrimination (Schwaninger, Carbon, & Leder, 2003). In the present study we address two classes of features which are involved in face recognition: components and configurations. Components are nameable parts of the face, such as eyes, nose and mouth, that vary from face to face. They somehow correspond to geon-like geometrical features (Biederman, 1987). These features vary in respect to shape, protuberance, etc. Configural features are the spatial and metric properties, which also distinguish different faces. Examples are the eye-distance and the mouth-nose-distance. This class of information is particularly affected when faces are turned upside down (Bartlett & Searcy, 1993; Leder & Bruce, 2000). It was recently argued that the occurrence of inversion-effects might indicate that configurai processing is involved in recognition (Mondloch, Le Grand, & Maurer, 2002). A different approach in understanding the specificity of face recognition is provided by the holistic processing hypothesis (Tanaka & Farah, 1993). Holistic processing means that faces are processed holistically as templates or Gestalts. Empirical evidence for this hypothesis comes from studies in which parts of a face, such as eyes, nose and mouth, are represented as being part of a whole rather than as explicit parts (Tanaka & Farah, 1993). There is still an ongoing debate whether this kind of whole-to-part-superiority (WPS) is exclusive for faces (Donnelly & Davidoff, 1999; Tanaka & Gauthier, 1997). In a typical test setting concerning holistic representation, full faces are compared with part conditions in recognition tests and reveal a superiority of the full face condition (WPS) (Tanaka & Farah, 1993). Importantly, in these experiments the part conditions need to be distinctive to recognize a specific face. In the present study we tested whole and part conditions at study and test, using faces which differ exclusively in respect to the part versions. In order to better understand the relation between component, configurai and holistic processing, the designs of our experiments include componential and configurai manipulations with whole and part test conditions. The relationship between configurai and holistic processing approaches is not yet clear. Leder and Bruce (2000, Experiment 5) investigated the recognition of faces which differed in terms of components and configurai features. …