A sea lion (Zalophus californianus) was trained to report the presence of a pure tone under water in a yes–no psychophysical procedure. The payoff matrix was varied by altering the probability of fish reinforcement [P (SR)] contingent on two classes of response—hits and correct rejections (CR). The P (SR‖Hit) and the P (SR‖CR) were 1.00, 0.75, 0.50, and 0.25. In baseline, both P (SR‖Hit) and the P (SR‖CR) were = 1.00. In the asymmetrical payoff matrices P (SR) was held constant at unity for one response and varied over three values for the other response. Thus, seven different payoff matrices were used. The sea lion showed rapid acquisition of a stable response bias independent of its ability to detect the underwater auditory signal. These results demonstrate that varying reinforcement probability is functionally equivalent to varying amount of reinforcement in the control of response bias in underwater signal detection by sea lions [see R. J. Schusterman, B. Barrett, and P. Moore, J. Acoust. Soc. Am. 57, 1526–1532 (1975)]. Subject Classification: [43]80.50, [43]80.60.