The assumption common to several models for signal detection, that increased payoff should have no effect on detection rate, was tested in a two-interval forced-choice auditory signal detection task. The signal occurred equally often in each interval, and amount of gain or loss was the same for both intervals. Payoff values (0, .1, .5, and 1 cent) were changed within sessions in 50-trial blocks. Detection rate increased noticeably in one experiment, was significant in a second experiment and had no effect in a third. With minimal payoffs of the sort typically used, significant effects may depend on convincing S of the importance of the incentive. Analysis of conditional error probabilities showed that errors were more likely to follow errors, which suggests periodic variation in level of attention.