To sleep, perchance to dream

Somebody asked me the other day whether fish sleep. My response was: “How can you tell?”. I have spent a good deal of time in aquaria watching large lugubriouslooking fish lying motionless in a tank. Even though their eyes are wide open, they could be asleep for all the notice they take of their surroundings. There are also fish that swim endlessly and pointlessly; for all I know, they swim in their sleep. Even harder to know is whether fish can dream. Yet it is probably dreaming that is the important process and sleep may just be a way of achieving dreaming without unnecessary distraction. We humans sleep and we dream. Our waking hours have produced many interpretations of the dream experience. Humans inhabit two worlds: the waking world — solid and continuous but, at the same time, narrow, routine and repetitive — and the dream world —fantastic, fleeting, and for many, an escape from the bondage of reality. In the dream world we can do amazing things and we are reluctant to consider this world as purely illusory. That is why the interpretation of dreams has been so important in many branches of psychotherapy; the belief that there are latent truths hidden in the manifest nonsense of dreams is hard to give up. In a paper in Nature in 1983, Francis Crick and Graeme Mitchison proposed a functional role for dreams. They suggested that the function of dreaming, which occurs during REM (rapid eye movement) sleep, is to remove undesirable modes of interaction in networks of cells in the cerebral cortex. Their postulate was that this is done by a reverse learning process, such that the trace in the brain of the unconscious dream is weakened and eliminated. In an accompanying paper, John Hopfield and colleagues showed that artificial neural networks can benefit from dreaming. In such associative systems, spurious memories are created at the same time as real ones during the learning process. These can be minimised if an unlearning process, which is similar to the learning one, but with the sign reversed, is applied with noise as an input. Of course, the machine does not dream any more than it thinks, but if complex associative systems require continual cleaning up, then it is the process that is important and it is a secondary feature as to how it appears to us in our consciousness. But why do we remember some dreams and not others? It could be argued that those we remember are the ones that we think of as not being totally spurious; they seem to have some significance. Their apparent significance may subsequently be completely dispelled in the cold light of our day-time existence. But sometimes, as has happened to me, remembered parts of a dream can lead to other thoughts, more useful and more rational. The Homeric poets took dreams as objective facts and dreams had a standard form in their writings. There is a visit by a dream figure who enters the room to deliver a message. The dreamer is passive but he sees the figure and hears the voice as outside of him. Waking dreams or hallucinations were treated in much the same way. Of course, these were constructions that were conditioned by the culture of the time, where everything was believed to be under supernatural control and dreams were one way of communicating with the gods. I have never had dreams like this. All my dreams have been of endless, tortuous journeys though rooms, tunnels, up and down stairs (pace Sigmund) or have been completely abstract, like Kandinsky paintings. Some are in colour but when I was young most were in black and white; perhaps it was only after Technicolor became more widespread that dreams appeared in full colour to me. I can write and talk about my dreams just as I can write and talk about my thoughts. In a curious way, dreams and thoughts are related. Both are activities of the brain that come into consciousness. Dreams, are fleeting and nonsensical ideas on the way out; thoughts are persistent and rational ideas on the way in. Both can be clothed in words or pictures or sounds. If it is true that unlearning is mechanistically the reverse of learning, and if unlearning requires a special input (noise), the role of sleep becomes understandable. It is there to switch off all structured input while the reverse gear is engaged. If that didn’t happen, we would unlearn everything we learnt, spurious or real. That this process manifests itself in the form of dreams in humans is neither here nor there. I can now attempt to answer the question we began with: do fish sleep? And do they sleep to dream? The answer would depend on how much impinges on their brains and how much they have to learn each day. If it is a lot, they would need to have an unlearning activity period and they might need sleep to achieve it. I suspect they only need a short nap, because life is much the same day in, day out. But at least they don’t have to attend lectures, for, at last, I understand why I find some lectures so irresistably soporific: I need the sleep to unlearn all the rubbish I heard during the lecture. R507