REM was discovered in the 1950s and in the next few decades scientists tried to figure out why Nature made REM and what it does for us. The EEG readings of a person in REM looks like that of a waking person, and the consciousness experienced in REM (often narratively coherent dreams) is more like waking than like the comparative subjective darkness of NREM sleep.
Also called "paradoxical sleep", REM is almost a hybrid of waking the NREM sleep. Some neurons and brain areas are as active as during waking; others are silent or dormant. Even deep in the brainstem some cells (e.g. cholinergic neurons) fire away during REM while some (e.g. monoaminergic neurons) stop firing. Drug stimulation of the cholinergic neurons can increase time spent in REM, but scientists know there is a lot more to REM than simple switching of neuron firing, even if they can’t totally explain everything.
One idea about REM was that it was important in the consolidation of memories. This makes sense because the subject of dreams in REM often involve or borrow from the previous day’s events.
Further, we spend less time in REM the older we get. Do young children spend more time in REM than old people because they are learning so much? It seems plausible that there would be a connection.
Research in recent decades has found memory consolidation and translation from short-term to long-term memory occurs more in NREM sleep. Learning facts and how to do things – these are facilitated by NREM and the memories are enforced and hardened.
During REM the emotional center of the brain, the amygdala, is very active. This may explain the strong feelings that can be associated with the dreams we remember in REM, and may help connect emotions to memories consolidated in NREM. A scenario or incident transferred to long-term memory in NREM may be replayed, perhaps with a fantastic twist, during REM, and emotion associated with it.
Further, there is no correlation between time spent in REM and IQ or other measures of cognitive capability.
Another idea is that in some way REM is important in the growth and development of the person’s body, and in particular the nervous system. Newborn babies spend 8 hours per day in REM and even fetuses experience REM. Small children spend more time in total sleep and in REM than adults do. While the current thought is that REM may partially function to assist neurological development, it did not evolve for that reason and it stands separately as a form of consciousness, not as a way to help the brain grow.
Emotion is tied up with sleep and tired brains, and people are generally more upbeat in the morning after a good night’s sleep. But REM does not seem to make us optimistic or particularly happy. Indeed, REM is associated with depression. Anti-depressant drugs (the popular SSRI type) tend to suppress REM sleep and it turns out this isn’t just a side effect, but appears tied in to how they reduce depression.
People with post-traumatic stress disorder spend more time in REM than other people. Indeed, an increasingly accepted counter-measure to inhibit PTSD development is to keep the victim from going to sleep after the stressful incident. People with difficult or traumatic pasts often have psychiatric problems that are exacerbated by dreaming about them. Replaying traumatic incidents over and over during REM dreams makes waking life worse.
Scientist J. Allan Hobson published a theory that REM is a sort of virtual reality program run by the brain. In this view REM is a "protoconscious state" that helps us rehearse and mentally act out scenarios in a manner similar to waking playtime does. This is how it helps us function in the world when we are awake and this is why it evolved.
Play is well-known to assist in a number of cognitive benefits, http://udel.edu/~roberta/play/benefits.html
While this idea has not been universally accepted, the absence of psychological explanations for dreaming and REM is pretty much widely agreed to by serious scientists.
Because warm-blooded animals experience REM, it is thought that REM arose in evolution around the same time as thermal homeostasis. Brain temperature drops during NREM but rises in REM. The thermal regulation of the body during sleep differs from waking. You don't sweat or shiver during REM sleep but your body uses other mechanisms to affect internal temperature. The normal thermoregulation homeostasis is off-line, so to speak.
One theory holds that at one point our ancestors were nocturnal, sleeping in the daytime. During twilight the ambient temperature was about the same as the internal body temperature, so contraction of muscles was not needed. In this theory primitive sleep that happened during the daytime evolved into NREM sleep, while sleep that happened at twilight evolved into REM sleep. This may be partly why the EEG signature of the brain in REM looks like that of a waking brain – REM is a warm-up (thermally and electrically) for waking. This theory also provides a partial explanation for long-term memory formation. During NREM "uncoordinated reinforcement" of memories fire in the neural circuit, while in REM a coordinated stimulation, perhaps in the virtual reality of the mind, shape the memories into a coherent whole that the waking mind understands. The one-two punch of NREM followed by REM sleep is what makes the mind form long-term memories.
Neuroscientist W.R. Klemm postulates that REM helps the brain wake up. This certainly meshes with the evidence from EEG readings that show REM and waking to be similar and with the subjective reports of people woken during REM who describe their dreams as much more vivid and coherent than those woken during NREM sleep. The brain literally gets warmer during REM and activity in the cortex increases as if to imply a revving up in anticipation of waking. In his Wake-Up Hypthoseis Klemm points out that REM and waking both rely on similar brainstem arousal systems.
The periods of REM during a night therefore act partially as a test period in which the brain evaluates whether enough sleep has been experienced yet. This gives an explanation, if not a convincing one, of why REM periods increase in duration and frequency throughout the night and why sleepers go through a long REM cycle shortly before awakening.
REM is in a way a dress-rehearsal for waking life in this theory. This also explains why young children have more REM: to develop their cognitive capabilities they need additional practice and mental stimulation. Klemm points out that waking is the most biologically adaptive form of consciousness, so we have an advantage in leaving NREM as soon as our brains are rested and NREM needs have been satisfied.
Jerome Siegel at UCLA says REM has little or no physiologically critical role and cannot be considered as important as slow-wave sleep. He told the New York Times that individuals who cannot experience REM do not have cognitive or emotional handicaps. Siegel speculates that REM functions to prepare the brain for waking. The body is allowed to continue to rest but the brain and mind starts to hum. Steven Lima at Indiana State told the Times he hypothesizes REM is a way to let part of the brain sleep while other parts wake. Dividing up the period of inactivity and reduced awareness this way may have allowed our ancestors to better survive in a predator-rich environment.
Another advantage to the flip between REM and NREM during the night is that is allows different parts of the brain to rest at different times.