Humans have always maintained a fascinated engagement with sleep and dreams. Shakespeare as well as the ancients wrote frequently about the phenomena. Indeed Freud published The Interpretation of Dreams in which he discussed dreams as the disguised, coded replay of the day's events. They were viewed as the "royal road to the unconscious." Still farther back in history dreams were viewed as predicting the future or communicating with long gone ancestors. However, their function and meaning were not seriously investigated until the EEG was developed and REM sleep was discovered in the last half of the twentieth century. Indeed it is strange to think that humans and all warm-blooded animals sleep since this makes them more vulnerable to predators. To me this is a strong indication that sleep has an important function. But what?
Starting in the mid 1960s a hypothesis was generated linking REM sleep and learning. This has remained prominent but the evidence is mixed. The first evidence is that learning causes increased REM sleep. The second evidence comes from deprivation studies in which memories are not consolidated if sleep is prevented. The final evidence is that memory processing occurs during REM sleep. Let's look at these three ideas.
Learning produces increased REM sleep in based on the need for increased memory consolidation to support learning. Yet animal studies that involve shock avoidance or frustration involved in appetitive reinforcement have major stressful impact on the animal and thus influence sleep. These variables cannot be separated. In experiments on rats, the REM sleep appeared at erratic intervals after the learning task and were found to vary with the type of task, strain of the animal or even the vendor that supplied the animals. Some human studied involving looking through prism glasses that rotate or invert images 90 degrees or more. Little or no effect was found on REM sleep. This line of evidence is open to substantial question.
The consolidation theory required that memory formation be prevented if REM sleep is interrupted. If REM sleep deprivation does not affect memory consolidation then it is reasonable to conclude that it was not needed for the task. Most animal tests have been done on rodents using the "platform technique" in which the rats tended to fall into the water (REM sleep is accompanied by complete loss of muscle tone). Also REM deprived rats tend to be hyperphasic, hyperactice, hypersexual, anxious, and irritable. All of these changes interfere with evaluating the REM sleep-deprivation hypothesis. Many human studies do not demonstrate REM sleep deprivation causing impairment of verbal learning and retention of anagrams. Finally, monoamine oxidase inhibitors (MAOI) administered in therapeutic doses can completely suppress REM sleep and dreams for very long periods. Many thousands of people have taken these drugs and thus we have a large naturalistic experiment on the effects of the suppression of REM sleep. That naturalistic experiment is negative. Interestingly, benzodiazepines produce sleep and do not affect REM sleep time or distribution yet they are noted for their deleterious effects on memory. The sheer size of these naturalistic experiments suggests strongly that REM sleep is not central to memory or learning.