Changes In Payoffs Research Articles

The effects of payoffs and prior probabilities on indices of performance and cutoff location in recognition memory

We compared the effects of payoffs and prior probability levels on indices of performance and cutoff location in recognition memory. The performance indices considered were d' from signal-detection theory and two corrections for guessing from simple threshold models. Shifts in cutoff location of comparable size were found for changes in priors and payoffs. The existence of a memory load and the inclusion of feedback did not affect the extent of shifts in cutoff location. Priors affected the corrections for guessing but did not affect d'. These results indicate that d' remains the preferred index of performance and that the corrections for guessing are not criterion-free measures. The problem of finding an index of recognition performance appropriate for comparing recall and recognition is discussed, as are alternative explanations for conservative cutoff placement.

The Effects of Payoffs on Sensitivity and Bias in Absolute Identification of Intensity

In our previous work on the identification of intensity [e.g., L. D. Braida, and N. I. Durlach, J. Acoust. Soc. Amer. 51, 483–502 (1972)], no attention was given to testing the implicit assumption that sensitivity and bias are independent. In the present series of experiments, sensitivity and bias were measured as a function of the payoff matrix. Maximum-likelihood estimates of sensitivity and bias were computed using an iterative procedure and the Thurstonian decision model of the preliminary theory. The results indicate that changes in payoffs cause changes in bias, but leave sensitivity invariant (consistent with the theory). Furthermore, the changes in bias were found to be similar to those that would occur for a receiver that is designed to maximize the expected value of the payoff. It is believed that these identification experiments provide a stronger test of the independence assumption than previous experiments using detection or discrimination paradigms.