Abstract

In a reply to L. L. Gatlin (See PA Vol 63:11111) conventional chi-square measures of bias are discussed, and it is shown that they, and presumably her D-measures, are poor measures of the basis of any mathematical estimator strategy--namely, the predictability of target sequence. A direct approach to the question of predictability is briefly described and applied to the individual target sequences in C. T. Tart's (1976) first training study, where it scored quite poorly compared to the Ss' actual scores. Gatlin's assertion that all of the above-chance scoring in the training study can be ascribed to some (unspecified) mathematical inference strategy is examined. In addition to serious methodological flaws in her analyses, it is shown that examination of the actual patterns of responses by the Ss are contrary to expectations from a mathematical inference strategy in a variety of ways. In spite of some relatively trivial deviations from randomicity in a minority of the target sequences in the training study, its significant results are best ascribed to ESP. (18 ref)