Why do we sleep? Because otherwise, we'd always be doing stuff.
This is the theory advanced by UCLA sleep researcher Jerome Siegel (website) in a new paper, Sleep viewed as a state of adaptive inactivity (free pdf). It's part of a Nature Reviews Neuroscience special issue on the evolution of the nervous system. Siegel proposes that the evolutionary function of sleep is simply to ensure that animals are only active when the benefits of movement (mostly access to food, and mates) outweigh the costs (activity burns calories, and puts you at risk of predation or accidents).
Sleep, in other words, is our equivalent of the inactive states into which most living things, even plants, periodically enter when it suits them. Even (deciduous) trees spend the cold, dark half of the year doing not very much. In Siegel's view, this is their equivalent of sleep.
This theory stands in contrast to the idea that sleep has a restorative function - that animals need to sleep, because some kind of important biological process can only occur while we're sleeping. This idea is intuitively appealing - it feels like we benefit from sleep, and at least in humans sleep deprivation has many well-documented negative effects.
But, as Siegel points out, we're far from any kind of a consensus on what the biological function of sleep is. It's generally assumed that there is one, and a great many have been proposed - he lists some, ranging from that sleep is important for the formation of new neural connections, to the idea that sleep is needed to reverse cellular damage caused by oxidative stress (interestingly, Siegel himself contributed to one of the papers he gives as a reference for that idea).
If a vital restorative function of sleep were to be conclusively identified, Siegel's theory would obviously be disproven. On the other hand, if Siegel is right, several things should be true. Firstly, the proportion of time that an animal spends asleep should be directly proportional to the amount of time that it is useful for it to be active.
Siegel argues that this is what we find. The big brown bat for example is the doziest of all mammals, sleeping for 20 hours per day. But it wouldn't benefit from being awake any more, because the insects it feeds on are only active for a few hours at dusk. If it were flying around during the day, it would just be wasting energy (and risking becoming lunch for a bird.)
By contrast, he says, some marine mammals (cetaceans, dolphins and whales) never sleep at all. In land mammals, sleep consists of distinct periods of neural activity such as REM and slow wave sleep. Neither, however, occurs in cetaceans. They do show a kind of neural activity called Unihemispheric Slow Waves (USWs). But these are confined to one half of the brain at a time. It's often said that this is "half the brain going to sleep". However, the animals remain moving normally, and are able to avoid obstacles, during USWs. It's not as if only half their body remains awake. As such, Siegel says, the USW state is not sleep.
If it's true that there are animals which never sleep, this is strong evidence for Siegel's theory, and against the idea that sleep plays a vital role. But not everyone agrees with his claim that dolphins and whales don't sleep. See, for example, this 2008 open-access paper, Is Sleep Essential?, which calls Siegel's theory of sleep the "null hypothesis" and then proceeds to criticize it.
In particular, the authors claim that dolphins do sleep, albeit with only one half of their brain at a time, and they make the interesting point that "the very fact that dolphins have developed the remarkable specialization that is unihemispheric sleep, rather than merely getting rid of sleep altogether, should count as evidence that sleep must serve some essential function and cannot be eliminated."
At this point the debate becomes highly technical. The sleep behaviour and neural activity of marine mammals is hardly easy to research, and it looks as though more evidence is needed before we can know for sure whether they sleep or not. This is one of those seemingly trivial questions which could end up deciding between two theories with enormous implications. There are quite a lot of them in science. We don't yet know why we sleep. But the answer may lie with the dolphins.
Link: More recently, I asked Why do we dream?
Siegel, J. (2009). Sleep viewed as a state of adaptive inactivity Nature Reviews Neuroscience, 10 (10), 747-753 DOI: 10.1038/nrn2697