Comparing the biased choice model and multidimensional decision bound models of identification

The biased choice model (BCM) of Lute (1963) and general recognition theory (CRT) of Ashby and Townsend (1986) are compared with respect to their ability to account for data from identification experiments. Specifically, by using CRT we investigate the ability of the BCM to account for data characterized by violations of perceptual independence and perceptual and decisional separability, and by homoscedasticity of the perceptual distributions. The effects of varying the dimensionality of the perceptual space are also studied. It is shown that the BCM has extreme difficulty accounting for data from experiments in which: (1) the stimuli are constructed from two physical components; (2) the perceptual means are positioned unevenly in the perceptual space; and (3) different stimuli are characterized by different amounts of perceptual variability or perceptual dependence. In general, however, the BCM has no difficulty with violations of perceptual or decisional separability. It is shown that CRT is not a special case of the BCM but that the two models do intersect. The BCM is shown to be closely tied to a number of different forms of symmetry that can be found in identification data and these are used to construct specific GRT models that are compatible with the BCM.