Abstract
General recognition theory (GRT, Ashby & Townsend, 1986) is a nonparametric generalization of signal detection theory that characterizes possible dependencies in perceptual representation in terms of the presence or absence of violations of perceptual independence (PI), perceptual separability (PS) and decisional separability (DS), using response-frequency-based measures obtained in a complete identification (CID) paradigm. Systems factorial technology (SFT, Townsend & Nozawa, 1995) is a nonparametric theory that characterizes the fundamental properties of information processing (architecture, stopping rule, capacity, and independence) using reaction-time (RT) obtained in a double factorial classification paradigm (DFP). The present study applied both GRT and SFT using a set of rectangular stimuli whose width and height were varied; these stimuli have previously been shown to be perceived in a perceptually integral manner (Macmillan & Ornstein, 1998). In addition, we manipulated response bias by running both tasks first using payoff matrixes designed to encourage unbiased respond and second using payoffs biased toward specific stimuli, testing the hypothesis that the combined use of GRT and SFT would allow for converging sources of evidence regarding inferred perceptual representations and processing characteristics. Results suggest observers (N = 4) violated PS in both the unbiased and biased conditions, and that parallel exhaustive processing was implicated for all observers in the unbiased condition and for three of four observers in the biased condition. The violation in PS indicated by GRT and the parallel exhaustive processing indicated by SFT document strong evidence to support previous findings of perceptually integral processing of rectangular stimuli and provided converging inferences by both theories, including consideration of individual differences. This coherent characterization of perceptual integral manner by both theories suggests a strong potential in the combination of GRT and SFT, allowing a more complete picture of information processing to be developed, including principled accounts of individual differences.
Meeting abstract presented at VSS 2017