Sleep disturbance has been implicated in cocaine use; however, the nature of the disturbance and its potential effects on cognition and learning are largely unknown. Twelve chronic cocaine users completed a 23-day inpatient study that included randomized, placebo-controlled, cocaine self-administration sessions. Six subjects received cocaine on each of days 4-6 and placebo on days 18-20, the other six received cocaine on each of days 18-20 and placebo on days 4-6. Sleep was measured by polysomnography, the Nightcap sleep monitor, and self-reported measures. Simple and vigilance reaction times were measured daily; a motor-sequence test of procedural learning was administered four times. Electrophysiological measures of sleep showed a different pattern than self-reported sleep across cocaine administration and abstinence: total sleep time and sleep latency were at their worst by 14-17 days of abstinence while self-reported sleep was at its best. Vigilance correlated positively with electrophysiologically measured sleep and negatively with self-reported measures. Similarly, sleep-dependent procedural learning correlated with total sleep time and was impaired at 17 days abstinence relative to 2- and 3-days abstinence. Slow-wave activity was lowest at days 4-9 of abstinence and highest during use and days 10-17 of abstinence. With sustained abstinence, chronic cocaine users exhibit decreased sleep, impaired vigilance and sleep-dependent procedural learning, and spectral activity suggestive of chronic insomnia. However, they report subjectively improving sleep, indicating they are unaware of this "occult" insomnia. These results suggest the possibility of homeostatic sleep drive dysregulation in chronic cocaine users.