Implicit and explicit memory for verbal stimuli presented during sleep.

Abstract

The present study explored the possibility of implicit memory for verbal information presented during sleep. Nineteen subjects in the experimental condition each spent one adaptation and two experimental nights in the sleep laboratory. On one experimental night they were presented with two lists of cued homophones (e.g. tortoise-hare) over earphones during REM or Stage 2 sleep and tested immediately afterward. On the other experimental night they were presented with two lists of category-instance pairs (e.g. bird-cardinal). In most cases, the lists were presented five times. For comparison, 12 control subjects came to the laboratory during the day and while awake underwent the same procedures as the sleeping subjects. All subjects were eliminated from the data analysis who had not been presented with two cued-homophone and two category-instance lists. The final data analysis included 10 sleeping subjects and 12 waking controls. For these subjects, all items were eliminated for which the subject had shown arousal or was in an inappropriate sleep stage before, during, or in the fifteen seconds immediately after stimulus presentation. A repeated measures ANOVA followed by a posteriori comparisons indicated that, contrary to what had been predicted, subjects in the experimental condition showed no sleep learning effects on either the homophone or the category-instance tests, although control subjects did. Consistent with earlier studies, recall for words presented during sleep was found in a few cases, but only when presentation was soon followed by arousal. These findings strongly suggest that semantic priming does not occur for verbal material presented during REM or Stage 2 sleep. The possibility of structural priming during sleep is also discussed. Past sleep learning studies are critically reviewed, and recommendations are made regarding the topics and methodology of future sleep learning experiments. In particular, a recommendation is made for additional research on quasi sleep learning, that is, learning for information presented immediately prior to arousal.