From theory to implementation: Building a multidimensional space for face recognition

Recognition, Classification and Inversion of Faces in the Multidimensional Space

“…The MDS model, which began as a metaphor for mental representations of faces (Valentine, 1991a), and has been extensively investigated ever since its conceptualization (e.g. Bruce et al, 1994;Burton & Vokey, 1998;Busey, 1998;Byatt & Rhodes, 1998;Johnston, Kanazawa et al, 1997;Lewis & Johnston, 1997, 1999aTanaka et al, 1998;Valentine, 2001;Valentine & Endo, 1992;Wickham et al, 2000), was only recently empirically validated (Catz et al, 2009). Once validated, it was possible to test some of the model's predictions.…”

“…Distance may be considered a holistic measure since it is a mathematically derived aggregate of the distances of all 21 dimensions from a prototypical face. In other words, unlike distinctiveness which may rely on a single feature, distance relies on all dimensions and is therefore a holistic measure (Catz et al, 2009). Most researchers agree that recognition of upright faces is holistic in nature, yet a controversy exists as to whether inverted faces are similarly processed holistically (Collishaw & Hole, 2000;Kanwisher et al, 1998;Nachson & Shechory, 2002;Rakover & Teucher, 1997;Valentine, 1988Valentine, , 1991aValentine & Bruce, 1988) or feature-by-feature (Bartlett & Searcy, 1993;Carey & Diamond, 1977;Farah et al, 1995;Farah et al, 1998;Leder & Bruce, 1998;Sergent, 1984).…”

“…Thus, although many studies have referred to the MDS metaphoric model (e.g., Bruce et al, 1994;Burton & Vokey, 1998;Busey, 1998;Byatt & Rhodes, 1998;Johnston, Kanazawa, Kato, & Oda, 1997;Lewis & Johnston, 1997, 1999aTanaka et al, 1998;Valentine, Chiroro, & Dixon, 1995;Valentine & Endo, 1992;Wickham et al, 2000), it is necessary to test its predictions on an empirically defined MDS. Recently, an attempt was made (Catz, Kampf, Nachson, & Babkoff, 2009) to construct an operational MDS that included 200 faces which were each rated on 21 dimensions. A factor analysis enabled the empirical establishment of six factors that distinguished among the faces.…”

“…An overall index referring to the faces' distance from the center of the MDS was also calculated on the basis of the 21 facial dimensions. This index was based on the relative distance of the faces on each of the MDS dimensions (for further details, see Catz et al, 2009). The purpose of the present study was to utilize an empirically based MDS, in which faces are located relative to six face-defining factors, to test predictions regarding both face recognition and classification.…”

“…The faces were ranked for distinctiveness in a previous study (Catz et al, 2009) by assessing the relative ease of spotting them in a crowd Tanaka et al, 1998;Valentine & Bruce, 1986a,c;Valentine et al, 1995;Valentine & Endo, 1992). On the basis of these rankings, faces above the median were considered typical, and those below the median were considered distinctive.…”