The interaction of payoff structure and regulatory focus in classification

ABSTRACT—This report brings together research on moti-vation and learning by exploring how fit of regulatory fo-cus affects people’s ability to acquire new categories. Perceptual categories were learned by people with a pro-motion focus (a situationally determined sensitivity to gains) or a prevention focus (a sensitivity to losses). Classification performance was closest to optimal (as determined by models fit to individual subjects ’ data) when the regulatory focus matched the structure of the payoffs for the catego-ries. Promotion-focus subjects performed best when the payoffs consisted of all gains. Prevention-focus subjects performed best when the payoffs consisted of all losses. This report brings together research on category learning and motivation. There has been significant progress examining how changes in the rewards associated with correct and incorrect responses affect people’s ability to identify members of per-ceptual categories (Maddox, 2002; Maddox & Dodd, 2001). In a typical classification experiment, every correct response is re-warded (or punished) in the same way regardless of the category to which the stimulus is classified; similarly, every incorrect response is rewarded (or punished) in the same way regardless of the category to which the stimulus is classified. In this case, the optimal strategy is one that maximizes accuracy of responses. If the rewards or costs associated with correct and incorrect responses for different categories are asymmetric, however, then optimizing reward requires deviating from optimal accuracy. For example, Figure 1 illustrates a simple unidimensional classification task. Each stimulus is a dot that appears in a particular location on a computer screen. As the figure shows, the categories have overlapping distributions. The bold solid line indicates the decision rule that yields optimal accuracy. If subjects respond ‘‘category A’ ’ to every stimulus to the left of this boundary and ‘‘category B’ ’ to every stimulus to the right, then they will maximize their accuracy of performance