Four rats were trained to lever press under a discriminated avoidance/escape schedule in which separately signalled safe and warning periods were 100 sec and 32 sec respectively. The auditory and not the visual component of the compound warning signal became associated with the discriminative control of lever pressing. Avoidance behavior also came under temporal control, in that the probability of lever pressing increasd as the warning period progressd. Timing began with the onset of the waming signal rather than the offset of the safe dgnal. However, after the warning signal had been pogrively eiminated, timing began with the offset of the safe signal. When neither signal was normally available, the temporal distribution of avoidance behavior changed markedly. Drifts in the temporal distribution of lever pressing occurred throughout the study; these were manipulated for two animals.In the conventional discriminated avoidance procedure, a safe period, during which behavior is without scheduled consequence, is followed by a warning period during which emission of behavior selected arbitrarily by the experimenter, such as pressing a lever or leaving a certain area, reinstates the safe period. If the selected behavior is not emitted during the warning period, a shock period follows and continues until the behavior is emitted, whereupon the safe period is reintroduced. If the selected behavior is lever depression, a lever press during the warning period avoids shock and a lever press during the shock period escapes shock. One environmental state, the safe signal, is correlated with the safe period, and another state, the warning signal, is correlated with the warning period. The warning signal usually continues through the shock period.