Differential gain, cost, and event probability were varied in a two-choice situation . Increases in the gain associated with a correct response, increases in the probability of being correct, and decreases in the cost of an incorrect response all result in increases in the probability of that response. Gain had a greater effect than cost, but both variables became less effective as event probability inc reased. Changes in these relationships were noted as a function of trials.' IntroductionThe present study was designed to investigate the effects of differential gains and costs in a two-choice situation. "Differential" means that the gain resulting from a correct de cision was a function of which choice had been made ; similarly, the cost of an incorrect choice depended upon the choice. The resultsofstudiesof·differ-entiaI'gains (Edwards, 1956; and of differential gains and costs suggest that the asymptotic probability of making the response with the highest expected value (EV) is correlated with the difference in expected values (t.EV) associated with the two responses. The acceptance of this conclusion must be qualified by noting that none of these investigations involved more than 150 trials. In a more recent experiment, with more trials and equal event probabilities, Katz (1964) investigated the effects of nine different costgain combinations. He concluded that the t.EV rule does not account for the rank ordering of choice probabilities across conditions. The present study extends Katz's research to two additional levels of event probability, and should therefore test the gene rality of his findings regarding gain-cost combinations.
MethodThe probability of the more frequent event (rr),andgain
Levels of risk were varied within Ss in a two-choice probability learning experiment. Contrary to the results of studies where risk was varied between Ss, the probability of predicting the more frequent event, P(All. was significantly greater at high than at low risk. A second finding was that P(A l ) was significantly higher for male than for female Ss, regardless of risk level.
ProblemIn a recent reviewofthe literature on animal learning, Pubols (1960) concluded that variation in incentive magnitude has no significant effect on time-independent measures of learning when an absolute method is used (i. e., each S receives only a single incentive level), but that incentive effects are obtained when a differential method is employed (i. e., each S experiences more than one incentive level). Experiments involving the effects of risk on human choice behavior have yielded results which are consistent with Pubols' conclusions regarding the absolute method. For example, Myers et al (1963) reported that the probability of choosing the more frequent of two events was not significantly greater for a 10 cent-than for a 1 cent-risk group. The purpose of the present study was to determine whether the differential method of incentive presentation would affect human choice behavior. More specifically, on each trial Ss were first informed which of two amounts was to be risked, and then required to predict which of two events would occur.
MethodIn the first of two experiments, six male and six female Ss, undergraduate volunteers from the University of Massachusetts, were required to predict which of two events would occur on each of 500 trials. The two events were low-associative nonsense syllables with probabilities of occurrence of .6 (E l ) and .4 (E 2 ). On half the trials, S was informed he would gain 10 cents for a correct event prediction, or lose 10 cents for an incorrect one. On the remaining trials, the level of risk was 1 cent. The sequence of risks was random with the restriction that half the E1 trials, and half the E2 trials were 10 cent risk and the remainder were 1 cent risk. The events were recorded on a 100-card deck which was shuffled before each trial block. Each S was given an initial stake oUl.OO in chips. He was told there was no pattern in the event sequence, and that he would receive all money won during the experimental session. The S kept track of his wins and losses by taking or
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.