Valentine (1991) proposed a theoretical framework for face recognition where faces are encoded as locations within a multidimensional space. The model assumes that distinctive faces are stored on the periphery of the space while typical ones tend to be located towards its centre. Most research on which this model is based has employed faces rated for distinctiveness on a unidimensional scale. This study requires subjects to make similarity judgements to pairs of faces rated in this manner. Multidimensional scaling techniques are used to construct possible face-spaces for these ratings. The data presented support Valentine's assumptions of the structure of this space.
|Number of pages||9|
|Publication status||Published - Mar 1997|