Rats responded on concurrent schedules of shock-postponement or deletion (avoidance) and timeout from avoidance. In Experiment 1, 3 rats' responses on one lever postponed shocks for 20 s and responses on a second lever produced a 1-min timeout according to a variable-interval 45-s schedule. Across conditions, a warning signal (white noise) was presented 19.5 s, 16 s, 12 s, 8 s, or 4 s before an impending shock. Raising the duration of the warning signal increased both avoidance and timeout response rates. Timeout responding, although positively correlated with avoidance responding, was not correlated with the prevailing shock rate. In Experiment 2, 3 rats' responses on one lever deleted scheduled shocks according to a variable-cycle 30-s schedule and responses on a second lever produced a 2-min timeout as described above. After this baseline condition, the avoidance lever was removed and noncontingent shocks were delivered at rates and intershock intervals yoked to the receipt of shocks in the baseline sessions. Timeout responding decreased when the avoidance lever was removed, even though the shock-frequency reduction afforded by the timeout remained constant. These results suggest that a key factor in the reinforcing efficacy of timeout is suspension of the requirement to work to avoid shock, rather than the reduction in shock frequency associated with timeout. (Contains 2 tables and 6 figures.)