Sensitivity to Visual Gaussian Noise Patterns as Determined by Binary and Rating Procedures

Abstract

Trained Os made judgments as to the presence of a signal plus noise pattern adjacent to a noise pattern with binary and 4-category rating response procedures. Two mean differences between stimulus distributions, 2 criterion levels, and 2 response procedures were factorially combined to produce 8 experimental conditions. Individual protocols of all Os in rating sessions and of 2 Os in most binary sessions resulted in straight line z-coordinate receiver operating functions. A receiver operating function expresses the probability of a hit as a function of the probability of a false alarm. As predicted by the theory of signal detection, neither response procedures nor criterion manipulations affected d_e, but d_e was sensitive to the magnitude of the mean difference between the signal and signal plus noise distributions. However, Os were not influenced by pay-off conditions to adopt low criteria in binary sessions even though criterion shifts were produced in rating sessions. Finally, the obtained slope for the receiver operating functions was nearly twice as large as predicted by the theory. On the basis of these mixed positive and negative results, the acceptance of the theory as a general model for visual detection may be questioned even while some successful applications of the theory to visual detection problems exist.