models

I work primarily within the General Recognition Theory (GRT) framework, though I am also well acquainted with a number of versions of the Similarity Choice Model. GRT is a multidimensional generalization of Signal Detection Theory, and, as such, it provides powerful tools for analyzing (failures of) independence between processing dimensions.

GRT relies on two main assumptions. First, it is assumed that perception is probabilistic such that, from one trial to another, the perceptual effect of a given stimulus varies, which, in the long run, produces perceptual distributions. Second, perceptual space is partitioned into response regions.

Thus, in GRT, there are three possible ways in which dimensions may interact (or be independent). Dimensions may interact perceptually at the individual stimulus level (i.e., within any given perceptual distribution) or in ways the involve multiple stimuli (i.e., across multiple perceptual distributions). Dimensions may also interact in decision making. Perceptual interactions are characterized via the shape and locations of perceptual distributions, while decisional interactions are characterized via the shape and location of decision bounds.

I have developed code to carry out various parametric and non-parametric analyses of GRT-based identification data. Much of this code can be accessed to the left.

More information will be added soon (1/5/10).