Abstract

A basic aim in the study of vision is to determine the fundamental mechanisms that mediate discrimination performance. In color vision, the description of discrimination data by an ellipsoidal representation has been very successful but at the cost of an indeterminacy in the knowledge of which axes in color space correspond to mechanism directions. For example, in the work of Wandell and colleagues (Wandell, 1982; Poirson et al., 1990) color discrimination contours are described by a mapping, A, from receptor coordinates to a privileged coordinate frame in which threshold differences map onto a unit sphere, where v is a color difference vector. As they so indicate, the matrix A is determined by the data only up to an arbitrary orthogonal transform. If the privileged coordinate frame is interpreted as a mechanism space, then the mechanisms mediating discrimination are specified only to within rotations and mirror image reflections. Geometrically, this means that the discrimination mechanisms do not necessarily lie along the principal axes of the ellipsoid that describes threshold but could lie along any set of axes for which the threshold lengths are mapped into unit orthogonal directions by A. Consequently, discrimination thresholds would not be sufficient to specify uniquely mechanism directions. Additional factors must be taken into account to pin down which axes of an ellipsoid correspond to visual mechanisms, e.g., specifying directions along which a mechanism might be null, based on adaptational (Krauskopf et al., 1982) or perceptual (Thornton and Pugh, 1983) criteria.

© 1992 Optical Society of America