Abstract
General Recognition Theory (GRT, Ashby&Townsend, 1986) is a powerful, static, nonparametric theory used to characterize the relationship between theoretical characteristics such as perceptual independence (PI), perceptual separability (PS) and decisional separability (DS), and response-frequency-based measures such as marginal response invariance (MRI). RTGRT (Townsend, Houpt, & Silbert, 2012) is a stochastic version of GRT that addresses both response frequencies and response times (RTs). The current study examines the extent to which GRT and RTGRT consistently capture the same relationships between in data. A complete-identification experiment with a feature-complete-factorial design using perceptually integral stimulus dimensions (Macmillan & Ornstein, 1998) was conducted to achieve this goal. Width and height of square/rectangular stimuli were varied with two levels of salience for each dimension. We predicted violations of PS, evidenced by violations of marginal response invariance (MRI) in the response frequency data and violations of timed MRI (tMRI) in the RT data. Further, we predicted that fits of parametric multidimensional signal detection models to the data would confirm the inferences drawn from the non-parametric measures. Results from the non-parametric measures suggested possible violations of PS, PI, and DS, as indicated by violations of both MRI and tMRI. Results from fitting the parametric models were consistent with these inferences and further suggested violations of PS in both dimensions for all subjects, and fulfillment of DS for three out of four subjects. The results in sum document the empirical consistency of GRT and RTGRT and suggest their utility in characterizing the perception of multidimensional forms.
Meeting abstract presented at VSS 2016