Probability-Learning with Ambiguity in the Reinforcing Stimulus
W. K. Estes, Marcia D. Johns · The American Journal of Psychology · 1958
In the experiment that has given rise to the term learning, the S is given repeated opportunities to predict which of two mutually exclusive events will occur, and on each trial S receives immediate and complete information as to the outcome.l Naturally enough, S typically learns to adjust his predictions to the relative frequencies of the events. Suppose, however, that we give S at the termination of each trial an informative signal which reduces but does not eliminate his uncertainty about the outcome. Now S must first predict whether an uncertain event will occur and then, after receiving an ambiguous signal, judge whether his prediction is confirmed. Concerning the latter, and less restrictive, conditions we wish to know not only how the individual's predictions will be influenced by a sequence of outcomes, but also how his judgments concerning outcomes will be influenced by his predictions. The experimental situation we have developed for purposes of investigating functional relationships between predictions and judgments is schematized in Fig. 1. The outcome ('reinforcing stimulus') of each trial consists in the exposure of a card containing two geometric Sgures. At the beginning of each trial, the central light on the display-panel is turned on, providing a ready-signal to S who then operates one of the two response-keys to indicate whether he expects the larger figure to appear on the left or on the right. If the figures of each pair clearly differ in area, and the larger appears on the right with some fixed probability, we can be confident that over a series of trials the mean proportion of 'right' predictions by a group of Ss will come to match this fixed probability. In the present experiment, however, the Sgures of each pair will not be clearly different in area. In preliminary studies, we have found sets of figurepairs for which, in our population of Ss, one member of each pair is judged larger than the other other with a probability (relative frequency) significantly greater than 0.5 but less than 1.0. The two types of figure-pairs to be used in this study are illustrated in Fig. 2. The lower figures differ in area, and approximately 85Sb of our Ss correctly judge the left-hand figure