Rats were trained in a two-way shuttle shock avoidance task, 20 trials per day for 5 consecutive days. In Experiment 1, groups were subjected to selective paradoxical sleep deprivation (PSD) for daily 4 hour periods. It was found that only PSD applied between 9-12 hours or 17-20 hours after the daily training sessions resulted in acquisition deficits. These vulnerable time periods coincided with polygraphically observed PS increases above normal levels following identical training procedures in previous studies. In Experiment 2, animals were only allowed PS during the 9-12 hour and 17-20 hour post training time "windows" observed in Experiment 1. For the remaining time each day, they were subjected to PSD. Acquisition was found to be unimpaired, despite this severe treatment. However, any other equivalent length PSD regime which overlapped with either of the time "windows" resulted in severe retardation of task acquisition. Results were believed to provide strong support for the sleep-learning hypothesis.