James P. Thomas Lynn A. Olzak University of California at Los Angeles
This chapter concerns the analysis of detection and identification data obtained in psychophysical experiments. The theory and analyses we present were developed in the course of investigating mechanisms underlying visual pattern perception, but are easily generalized to other sensory domains. Specifically, we describe a multidimensional vector model of visual processing and illustrate its application in identifying properties of underlying processing mechanisms. In a general discussion of the model, we consider various questions about the nature of the underlying sensory mechanisms, the internal noise process, and the decision process, including effects of response bias and stimulus uncertainty. Many of these concepts are introduced in the context of the familiar yes-no or rating procedures of Signal Detection Theory. However, the focus of the chapter is on the theoretical analysis of data from the 2 × 2 paradigm, a psychophysical procedure that simultaneously compares the ability to detect the presence of a faint visual stimulus with the ability to identify various properties of that same stimulus. We describe analyses for estimating bandwidths of underlying mechanisms, tests for independence among mechanisms, and provide a brief example in which we test whether the stimulus properties of brightness and darkness are processed via an opponent (bipolar) mechanism or by separate unipolar pathways.
The model and experiments described have the general goal of validating and refining one view, described here, of how local spatial patterns are represented in