Inescapable shock (IS), an uncontrollable stressor, and presentation of fearful contexts associated with IS produce prominent reductions in REM sleep. We compared sleep in animals trained with IS to that in animals trained with escapable shock (ES), a controllable stressor, in a paradigm in which animals always received shock but could terminate it by their actions. Male BALB/cJ mice were implanted with telemetry transmitters for recording EEG and activity. After recovery from surgery, baseline sleep recordings were obtained for 2 days. The mice were then randomly assigned to receive ES (n = 9) or IS (n = 9). ES mice could escape a footshock (20 trials; 0.5 mA; 5.0 sec maximum duration; 1.0 min intervals) by moving to the unoccupied chamber in a shuttlebox. Yoked-control IS mice in a separate shuttlebox received identical footshock. The mice received 2 days of shock training (ST1; ST2) and were re-exposed to the shuttlebox without footshock (context alone). NA. NA. NA. On each training and test day, the mice were returned to their home cages, and EEG and activity were recorded for 20 h. Freezing was scored in the context alone. Compared to baseline, ES mice showed significantly increased REM, and IS mice showed significantly decreased REM after ST1, ST2, and context alone. Total NREM was decreased after shock training only in IS mice. Contextual freezing was enhanced in both ES and IS mice. The directionally opposite changes in REM suggest that stressor controllability is an important factor in the effects of stress and stressful memories on sleep.