THE PRODUCTION-ESTIMATION METHOD: REGRESSION REDUCTION AND INDIVIDUAL DIFFERENCES IN POWER FUNCTIONS (PSYCHOPHYSICS, SCALING

While average power function exponents across groups of subjects have repeatedly yielded the same representative exponents for each of a number of modalities, large differences in individual exponents have been a consistent problem in psychophysics. Stevens and Greenbaum suggested that these differences might be due to differences in a centering tendency called the "regression effect." Using an Interspersed Production-Estimation procedure in an attempt to eliminate regression, the present study was designed to (1) evaluate this procedure, and (2) test whether individual differences remain after regression is reduced. A convenient regression score was introduced which can be used not only to compare two or more groups on the amount of regression present, but also to determine whether regression has been completely eliminated in any one group. While the Interspersed Production-Estimation procedure did not eliminate all of the regression, it reduced it very substantially at the .0001 significance level. Inter-subject variability of exponents was not reduced significantly, suggesting that other factors besides the regression effect are responsible for the persistent individual differences found in power function exponents.