Trivial incentives, marginal comprehension, and dubious generalizations from Prisoner's Dilemma studies.

Abstract

Examined the conclusion of several prisoner's dilemma (pd) studies that, since incentive magnitude has little effect upon average cooperation levels, the typical use of trivial payoffs does not jeopardize the meaning of pd findings. 80 male undergraduates served as ss in a 10-trial pd game. 2 levels of incentive, pennies and dollars, were crossed with 2 levels of instruction, traditional and comprehensive. Average cooperation levels did not differ across the 4 experimental conditions. High incentives, however, produced significantly greater interdyad variance than low incentives. The difference in variance between instruction conditions was not significant. Reliability coefficients for the task and measures of response latency suggest that the increase in variance for high incentives was true variance, rather than error variance. Results suggest that generalization from pd studies using trivial incentives should be limited to behavior in parlor games. (24 ref.) (PsycINFO Database Record (c) 2016 APA, all rights reserved)