Dream Research
Do animals dream?
REM sleep is found across mammals and birds, and rats appear to replay the routes they ran that day. Whether any of that involves experience is a question the evidence cannot reach.
Anyone who has watched a dog twitch, paddle and make muffled noises in its sleep has formed a strong intuition on this question. The scientific position is more careful, and the careful version is more interesting.
What is established
REM sleep is widespread. All mammals studied show it, with the possible exception of some marine species where sleep is unihemispheric. Birds show it in short bursts. Reptiles show sleep states with some REM-like features — a 2016 study of the Australian bearded dragon found alternating states resembling REM and slow-wave sleep, which pushed the evolutionary origin considerably further back than expected. There is even evidence of REM-like states in cuttlefish and in fruit flies.
REM atonia serves a function. Michel Jouvet's lesion experiments in the 1960s removed the brainstem region responsible for muscle paralysis in cats. The animals then, during REM, got up and performed complex behaviours — stalking, pouncing, arching, apparently attacking absent prey — with their eyes open and no response to their surroundings.
This is close to a direct demonstration that the sleeping cat's brain was generating organised, species-typical behaviour sequences. It is the strongest single piece of evidence in the field.
Hippocampal replay. Matthew Wilson and Bruce McNaughton recorded place cells in rats running a track, then recorded during subsequent slow-wave sleep. The same firing sequences recurred, in order, compressed in time. Later work by Kenway Louie and Wilson found similar replay during REM, at closer to real-time speed — and the correspondence was precise enough that researchers could identify which part of the track the rat was traversing.
Songbird rehearsal. Zebra finches show, during sleep, patterns of activity in the song-control nucleus RA that correspond to the neural activity of singing. Amish Dave and Daniel Margoliash found that the sleeping bird's neurons fire as though singing a song it is not singing.
All of the above concerns neural activity and behaviour. Dreaming is a subjective experience. We know human dreams exist because humans report them. No animal can report, and no measurement distinguishes a brain generating an experience from a brain generating an equivalent pattern without one. This is the hard problem of consciousness in miniature, and it is not going to be solved by better electrodes.
What we can reason about
The argument from continuity is the strongest available. Mammalian brains are broadly similar in architecture. REM sleep is homologous across species. The neural correlates of human dreaming — limbic activation, prefrontal deactivation, hippocampal replay — are present in other mammals. If those correlates produce experience in humans, it is difficult to argue for a principled reason they would not in a dog.
The counter-argument is that human dreaming may depend on capacities other animals lack — narrative construction, autobiographical memory, a sense of self persisting through time. On this view an animal might have REM-associated imagery without anything resembling a dream in the human sense.
Both positions are defensible and neither is testable with current methods.
REM sleep behaviour disorder as evidence
One more line worth noting. In humans, REM sleep behaviour disorder involves failure of atonia, and sufferers physically enact their dreams. When woken, they report dream content that matches the observed behaviour — a person seen defending themselves reports dreaming of being attacked.
This gives us, in humans, a validated correspondence between enacted behaviour during REM and reported dream content. Jouvet's cats displayed the same class of behaviour. The inference that they too were experiencing something is not proof, and it is not nothing.
Which animals sleep strangely
Some findings are worth knowing simply because they complicate any tidy story.
Dolphins and some whales sleep unihemispherically — one hemisphere at a time — and show very little or no REM sleep, which raises awkward questions for any theory that makes REM essential.
Many birds also sleep unihemispherically, with one eye open, particularly at the edges of a flock.
Elephants sleep around two hours a day in the wild. Giraffes similarly little. Bats and some rodents sleep for very long periods.
Octopuses show two alternating sleep states, and during the active state their skin flashes through rapid colour changes resembling waking camouflage patterns. This is a mollusc with a completely independently evolved nervous system, which makes it either a remarkable case of convergent evolution or a caution against reading too much into surface resemblance.
Where that leaves the question
Mammals and birds have REM sleep. During it, their brains replay recent experience and generate organised motor programmes that are normally blocked. In the one species that can be asked, this is accompanied by vivid subjective experience.
Whether it is accompanied by experience in the others is, on present evidence, a matter of inference. The inference is reasonable. It is not a finding.
