Octopuses are soft-bodied marine molluscs with eight arms, around 300 species, and a nervous system that evolved complexity entirely independently of ours — most of their neurons are in their arms rather than their brain.
The octopus is the standard example of intelligence arriving by another route. Our common ancestor with them lived more than 500 million years ago and was, as far as anyone can tell, something like a flatworm. Everything that came afterwards — the large nervous system, the camera eye, the problem-solving — evolved twice, separately. That makes octopuses the best available test of which features of a mind are inevitable and which are accidents of our own lineage. It also makes them exceptionally easy to over-interpret: they are unlike us in enough ways that any similarity feels like a revelation. This page tries to keep the findings and the enthusiasm apart, particularly around sleep, colour vision and the widely repeated claim that they learn by watching each other.
In-depth record · 98% complete · reviewed 2026-08-09
What this page covers
Around 300 species. Every major finding on this page comes from a different one, which is worth remembering when reading "the octopus" anywhere.
Often confused with: Sepia officinalis (common cuttlefish — has an internal shell and eight arms plus two tentacles); Loligo (squid — streamlined, with fins and two feeding tentacles)
Quick facts
Species
Around 300
Nervous system
Roughly 500 million neurons, about two-thirds of them in the arms
Colour vision
One visual pigment — they are almost certainly colour-blind, yet match coloured backgrounds
Sleep
Two stages, one of them resembling REM sleep
Lifespan
Typically one to two years — short, for an animal this complex
The last common ancestor of humans and octopuses lived over half a billion years ago and had nothing resembling a complex brain. Everything either lineage built afterwards was built separately. When an octopus and a crow and a person all turn out to solve a problem, that convergence is informative in a way that similarity between two mammals is not.
The octopus nervous system is a second, independent invention of complex intelligence
Well supported
Good evidence backs this, though some details remain open.
The Octopus bimaculoides genome shows a large independent expansion of protocadherin gene families associated with neural wiring, on a scale otherwise seen mainly in vertebrates, in a lineage separated from ours by more than 500 million years.
Who this applies to
sequenced in one species, applied to cephalopods generally
Studied in
Octopus bimaculoides
Why we rate it this way, and what the caveats are
Well supportedHigh confidence
Genome sequence is about as solid as evidence gets. The interpretive step — that gene family expansion underlies neural complexity — is an association rather than a demonstration.
How far it can be extended
Later cephalopod genome work has found comparable features, and the neural anatomy that motivated the question is shared across coleoid cephalopods.
Caveats
Gene family expansion is associated with neural complexity, not shown to cause it.
One sequenced species.
Still unanswered
Which of these genetic changes actually matter for the behaviour we find remarkable?
The result behind "an octopus has nine brains", and why the slogan undersells it.
An octopus’s brain says "reach" — the arm works out how
Well supported
Good evidence backs this, though some details remain open.
Octopus arm extension follows a stereotyped bend propagating from base to tip. Arms surgically disconnected from the brain and stimulated at the nerve cord produce the same movement, indicating the motor program resides in the arm’s peripheral nervous system, which holds roughly two thirds of the animal’s neurons.
Who this applies to
the common octopus, in which arm control has been characterised
Studied in
Octopus vulgaris
You may have heard
“An octopus has nine brains”
A memorable shorthand for a real fact — most of the neurons are in the arms, and the arms generate their own movements — but the arm ganglia are not brains. There is one brain issuing goals and eight limbs containing the machinery to achieve them, which is a genuinely novel architecture and not eight extra minds.
Why we rate it this way, and what the caveats are
Well supportedHigh confidence
A direct physiological demonstration: the movement is produced with the brain disconnected, which is about as unambiguous as a localisation result gets.
How far it can be extended
Arm anatomy and nerve cord organisation are conserved across octopuses, so peripheral movement generation is expected to be general within the group.
Caveats
Reaching is one movement; how much other behaviour is peripherally generated is unclear.
Severed-arm preparations are far from a normal physiological state.
Peripheral control coexists with central control, which takes over for visually guided tasks.
Still unanswered
How does the brain select among the arm’s available motor primitives?
What, if anything, does the central brain know about where the arms are?
Shows the brain can guide an arm visually when the task requires it, so peripheral control is not the whole story.
How we know
An octopus arm that reaches with no brain attached
An octopus arm has no joints and effectively infinite ways to bend. Is the brain computing every reach, or is the arm doing it?
Reaching movements in intact octopuses were filmed and characterised: a bend that propagates smoothly from the base of the arm to the tip. The test was then to remove the brain from the equation entirely — arms disconnected from the central nervous system were stimulated directly at the nerve cord, and the resulting movement compared with the intact animal’s.
What happened
Denervated arms produced the same stereotyped reaching movement as intact ones.
What it shows
The motor program lives in the arm. The brain issues a goal; the limb contains the machinery to achieve it. That is an engineering answer to a problem vertebrates never had — a boneless arm cannot be micromanaged centrally, so the movement primitive was pushed down into the limb.
What it does not show
It does not mean the arms are intelligent or independent, and "nine brains" overstates it — arm ganglia are not brains. Reaching is also one movement; how much other behaviour is peripherally generated is unclear, and when a task needs visual guidance the central brain takes over. A severed-arm preparation is far from a normal physiological state.
The controls — what makes this evidence rather than a story
Intact animals provide the movement signature the denervated preparation is measured against.
Stimulating the arm’s own nerve cord shows the pattern is generated locally rather than relayed.
Movement was characterised quantitatively — the propagating bend — not judged by eye.
The experiment is blunt and decisive: an arm disconnected from the brain, stimulated at its own nerve cord, produces the same stereotyped reaching movement as an intact one — a bend travelling from base to tip. The program is in the arm.
This is not arms being intelligent. It is an engineering solution to a problem vertebrates do not have. A limb with joints has a handful of degrees of freedom and can be commanded directly; a boneless arm has effectively infinite ones and cannot. Evolution pushed the movement primitive down into the limb so the brain only has to specify a goal, and roboticists have been trying to copy the arrangement ever since.
It is also not the whole story. When an octopus has to guide an arm through a maze to a visible target, the central brain takes over and steers it by eye. Peripheral control is the default, not the only mode.
Words used here
Muscular hydrostat
A structure that moves by squeezing fluid-filled muscle against itself, with no skeleton. An octopus arm, an elephant trunk and a tongue are all examples.
Motor program
A stored pattern of muscle activation that produces a movement once triggered, without moment-to-moment supervision.
Octopus skin produces some of the most sophisticated colour matching in the animal kingdom. Octopus retinas contain a single visual pigment, which should make colour vision impossible. Nobody has established how the animal resolves this.
Nobody has established how a colour-blind octopus matches coloured backgrounds
Contested
Researchers actively disagree, and the disagreement is substantive.
Cephalopod retinas contain a single visual pigment, which should preclude colour vision, yet the animals produce and match complex colour patterns. A proposed mechanism using chromatic aberration and unusual pupil shape is theoretically viable but has not been tested behaviourally or physiologically.
Who this applies to
octopuses, cuttlefish and squid generally
Studied in
Octopus, Sepia, Loligo
Why we rate it this way, and what the caveats are
ContestedModerate confidence
The puzzle is agreed by everyone; the solution is not. The chromatic-aberration model is elegant and untested, and competing explanations including skin-based light sensing remain live.
How far it can be extended
Single-pigment retinas are documented across coleoid cephalopods, and the puzzle applies to all of them.
Caveats
The leading proposal is a model with no experimental test.
Where researchers disagree
The chromatic-aberration mechanism requires focus adjustments that may be too slow for the speed at which cephalopods change pattern.
Skin-based light sensing could supply wavelength information without any eye-based mechanism, and the two proposals have not been tested against each other.
Still unanswered
Can any behavioural experiment distinguish the proposed mechanisms?
Do cephalopods match colour as well as human observers assume they do?
Ramirez and Oakley, 2015 · Journal of Experimental Biology
Offers an alternative or complementary route to wavelength information, through the skin.
Octopus skin can detect light on its own
Well supported
Good evidence backs this, though some details remain open.
Isolated skin from Octopus bimaculoides expands its chromatophores in response to light, most strongly to blue wavelengths, and expresses rhodopsin — indicating light detection independent of the eyes and brain.
Who this applies to
the California two-spot octopus, in isolated skin preparations
Studied in
Octopus bimaculoides
You may have heard
“Octopuses see with their skin.”
The skin detects light and reacts locally. There is no evidence it produces an image or that the animal experiences anything through it. "Sees" implies vision; "detects brightness and wavelength across its whole surface" is what was shown, and is unusual enough on its own.
Why we rate it this way, and what the caveats are
Well supportedHigh confidence
A direct physiological result: tissue disconnected from the nervous system responded to light and contained the relevant photopigment. What it means for the whole animal is a separate question.
How far it can be extended
Light-sensitive opsins have been found in the skin of several cephalopod species, so the capacity is probably widespread in the group. The behavioural response was measured in this species.
Caveats
Detecting light is not the same as forming an image.
Measured in isolated tissue rather than in a behaving animal.
Still unanswered
Does skin light-sensing actually contribute to camouflage in a living octopus?
Ramirez and Oakley, 2015 · Journal of Experimental Biology
Isolated skin responded to light and expressed rhodopsin.
Words used here
Chromatophore
A pigment-filled sac in the skin that muscles can stretch open or let contract, changing the colour of that patch of skin in milliseconds.
Chromatic aberration
The tendency of a lens to focus different colours at slightly different distances. Usually a flaw; the proposed octopus mechanism would make it useful.
The body a very different nervous system is attached to.
Almost everything structural about an octopus is unfamiliar, and the oddities are connected rather than a list of trivia. It is a mollusc — closer kin to a snail than to anything with a backbone — that abandoned its shell, and nearly every strange feature follows from that decision.
The parts people ask about
Feature
What is actually true
Hearts
Three. Two pump blood through the gills, one pumps it to the body — and that one stops when the animal swims, which is why octopuses prefer crawling.
Blood
Blue. It carries oxygen with haemocyanin, which uses copper, rather than haemoglobin, which uses iron. Copper works better in cold, low-oxygen water.
Bones
None. The only hard part is the beak — which is why an octopus can pass through any gap its beak fits through.
Beak
Parrot-like, made of chitin, at the centre where the arms meet. It bites prey and delivers venom.
Arms
Eight arms, not tentacles. Squid and cuttlefish have eight arms plus two longer tentacles; octopuses have no tentacles at all.
Ink
Melanin plus mucus, fired as a decoy shaped roughly like the animal, which also dulls a predator’s sense of smell.
Breathing
Gills. An octopus out of water is suffocating, though some species cross wet ground briefly.
Regeneration
Arms regrow fully after loss, including the nerve cord.
The escapes people film follow directly from the lack of a skeleton: nothing about an octopus limits which gap it fits through except the beak. A large animal squeezing through a hole the width of a coin is not doing anything clever with the hole. It has no bones.
One correction worth making plainly: octopuses are venomous, all of them, and only the blue-ringed octopuses of the Indo-Pacific carry venom dangerous to a person. The rest can deliver a bite that hurts and nothing more, and none of them are aggressive towards people.
Words used here
Haemocyanin
A copper-based oxygen carrier, blue when oxygenated. Less efficient than haemoglobin in warm water and better in cold.
Mantle
The bulbous part behind the eyes, containing the organs. Not the head, which is the small part where the arms and eyes meet.
The three hearts exist because octopus blood is bad at carrying oxygen
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Octopus blood transports oxygen using haemocyanin, a copper-based pigment dissolved in the plasma rather than contained in cells, which is blue when oxygenated. Haemocyanin carries less oxygen per unit volume than haemoglobin and performs comparatively well in cold, low-oxygen water. Two branchial hearts pump blood through the gills and a systemic heart drives it around the body; the systemic heart ceases beating during jet-propelled swimming, which limits sustained fast locomotion and favours crawling.
Who this applies to
octopuses, with the physiology characterised mainly in Octopus vulgaris
Studied in
Octopus vulgaris, Octopoda
You may have heard
“Octopuses have three hearts and blue blood”
Both true, and they are the same fact told twice. Haemocyanin is dissolved in the plasma rather than packed into cells and carries less oxygen than haemoglobin, so the animal needs extra pumping to compensate — and still cannot swim fast for long, because the main heart stops when it jets. The trivia describes a constraint rather than a superpower.
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
Circulatory anatomy is directly observable and the oxygen-carrying properties of haemocyanin are measured biochemistry.
How far it can be extended
Haemocyanin and the three-heart arrangement are general to cephalopods; performance details are best measured in one species.
Caveats
Haemocyanin outperforms haemoglobin in cold, oxygen-poor water, so "worse" is conditional.
Most measurements come from one temperate species.
Deep-sea octopuses live under conditions where the trade-offs differ.
Still unanswered
How do octopuses in warm shallow water manage the oxygen limitation haemocyanin imposes?
The circulatory arrangement, haemocyanin transport, and the systemic heart stopping during jetting.
The three hearts and the blue blood are usually told as two separate marvels. They are one fact stated twice. Haemocyanin, the copper pigment that makes the blood blue, is dissolved loose in the plasma rather than packed into cells, and that sets a hard ceiling on how much of it there can be — too much would make the blood thick enough to be difficult to pump. So an octopus carries less oxygen per litre of blood than a vertebrate does, and compensates by moving the blood faster and by running it through the gills under dedicated pressure. Two branchial hearts push blood through the gills; the systemic heart pushes it round the body. Three hearts is what carrying an inefficient oxygen carrier costs.
The consequence shows up in how the animal moves. The systemic heart stops beating while an octopus jets — the contraction of the mantle that produces the thrust also compresses the heart — so fast swimming is done with the main circulation interrupted, and it exhausts the animal within a short distance. This is why an octopus that can jet chooses to crawl, and why chases are brief. It is a sprinter with no cardiac output during the sprint.
None of this makes haemocyanin a poor choice. In cold water with little dissolved oxygen it outperforms haemoglobin, which is the environment most cephalopods evolved in and where several of them still live. The trade-off only bites in warm, well-oxygenated shallows, and warming seas are expected to press on exactly that margin.
The same pigment is what oxygenates a spider, a crab and a snail. It is the standard invertebrate solution; the octopus is unusual in needing so much of it.
An octopus eats animals with hard parts: crabs, clams, snails, shrimp, small fish, and other octopuses. The arms do most of the work and they do not need to see. An octopus hunting a reef spreads its web of arms over a coral head and gropes into every crevice at once, feeling and tasting as it goes, then contracts the web around whatever it has found. In the giant Pacific octopus, chemical receptors in the suckers detect prey by contact at concentrations low enough that the animal effectively tastes the rock.
The crab is the instructive case, because a crab is armoured and an octopus has no jaws to crush it. The bite delivers venom from the salivary glands, which paralyses the crab within a minute or so. Then the shell is opened — pulled apart at the joints if that works, and drilled if it does not. The radula, a rasping tongue, bores a hole through shell, and a second secretion is injected that loosens the muscle from its attachments so the flesh can be pulled out whole. A clam eaten this way is found afterwards with a neat hole a millimetre or two across, and the position of that hole is not random: an octopus drills over the muscle it needs to reach.
Prey is usually carried back to the den and eaten there, which is why an occupied den is surrounded by a midden of shells.
That midden is how researchers find dens, and how diet is reconstructed without watching a hunt.
Cannibalism is common enough that octopuses are cautious around each other.
Feeding is slow: an octopus with a large crab may spend an hour on it inside the den.
Waste leaves the same way water does. The gut ends in an anus inside the mantle cavity, beside the base of the funnel, so faeces are flushed out with the exhalant jet and directed away from the den. It is the same opening used for ink, for jetting and for the exhaled water from the gills — one funnel, several jobs, and a reason an octopus is fastidious about where its den entrance points.
Words used here
Radula
The rasping, toothed tongue-like organ molluscs feed with. In an octopus it drills through shell.
Midden
The pile of discarded shells outside an octopus den. The most reliable sign one is occupied.
Funnel
The muscular tube below the head. Used for breathing out, jetting, inking and waste.
All three plurals appear in dictionaries and one of them is built on a mistake. "Octopi" applies a Latin rule to a word that is Greek: *oktopous* would give *octopodes* if the Greek were followed through. "Octopuses" applies the ordinary English rule to a word that has been in English for centuries, which is what English does with borrowed nouns once they settle. Scientific writing uses octopuses almost universally, octopodes survives as a pedant’s flourish, and octopi is common, understood and etymologically confused. NatureHQ uses octopuses.
The confusion with squid is more consequential, because it changes what you are looking at. Both are cephalopods and they are built for opposite lives.
Octopus and squid
Feature
Octopus
Squid
Limbs
Eight arms, no tentacles
Eight arms plus two longer feeding tentacles
Body
Soft throughout; the beak is the only hard part
Stiffened by an internal pen, a remnant of the shell
Where it lives
On the bottom, in dens and crevices
Open water, often in shoals
How it moves
Crawls by preference; jets to escape
Jets as its normal means of travel
Life
Solitary; most species semelparous, one to two years
Often social; also short-lived and mostly semelparous
Cuttlefish are the third familiar group: eight arms and two tentacles like a squid, a broad internal cuttlebone, and hovering rather than shoaling. All three descend from shelled ancestors, and the nautilus is the one that still has its shell — which makes it the useful reminder of what the rest of the group gave up, and what they gained by it.
Words used here
Cephalopod
The mollusc group containing octopuses, squid, cuttlefish and nautiluses. Literally "head-foot".
Pen
The stiff internal remnant of the shell that gives a squid its shape. Octopuses have none.
The hardest constraint on octopus intelligence is how little time it has.
Most octopuses live a year or two, breed once, and die — the mother starves guarding her eggs
Well supported
Good evidence backs this, though some details remain open.
Most studied octopus species are semelparous, reproducing once and dying shortly after. A brooding female ceases feeding, tends her clutch continuously, and declines rapidly around hatching. Removal of the optic glands halts the decline and restores feeding, and transcriptomic work shows several distinct signalling pathways activating in sequence after laying rather than a single hormonal switch.
Who this applies to
the shallow-water octopus species that have been studied in captivity
“Octopuses would rule the world if they lived longer”
An enjoyable line that treats lifespan as an accident. Dying after one brood is not a limitation the animal is working around: it is part of a life history that also produces very fast growth and enormous clutches. The interesting question is why that combination persists in an animal this capable, not what it would do with more time.
Why we rate it this way, and what the caveats are
Well supportedHigh confidence
The behavioural pattern is consistently observed, the optic gland control point is established by direct removal, and the underlying signalling has now been characterised at the level of gene expression.
How far it can be extended
Semelparity is documented across the shallow-water species studied. Some deep-sea octopuses brood for years and are far less well characterised.
Caveats
Deep-sea species can brood for years and are poorly characterised; the generalisation is about shallow-water octopuses.
Gene expression identifies correlates of the decline rather than proving causation.
Male senescence also occurs and is much less studied.
Still unanswered
Why has semelparity persisted in a lineage whose cognition would benefit from a longer life?
What triggers the optic gland programme in the first place?
Genomic context for a lineage whose nervous system elaborated independently of vertebrates.
An octopus does not simply wear out after breeding — a gland switches it off
Well supported
Good evidence backs this, though some details remain open.
Octopuses undergo a stereotyped senescence after reproduction: appetite declines, movement becomes uncoordinated, skin lesions appear and camouflage control degrades, with death following within weeks to months. Females enter it after brooding, during which they cease feeding; males after mating. The process is governed by secretions from the optic glands, and removing those glands experimentally extends feeding and lifespan.
Who this applies to
the octopus species observed in aquaria; cephalopods generally vary
Studied in
Octopus, Enteroctopus dofleini
You may have heard
“Octopuses die after mating”
Not immediately, and not from exhaustion. Both sexes enter a controlled decline governed by the optic glands — remove them in an experiment and the animal starts eating again and lives longer. That is programmed rather than incidental, which is a stranger fact than the one usually told.
Why we rate it this way, and what the caveats are
Well supportedModerate confidence
The behavioural syndrome is consistently described and the optic-gland mechanism is supported by removal experiments. Confidence is moderate because observations are largely from captivity and species coverage is thin.
How far it can be extended
Semelparity and optic-gland control are documented across several octopus species, but not every cephalopod is semelparous and deep-sea species are almost unstudied.
Caveats
Aquarium conditions differ from the wild in diet, temperature and disturbance.
Not every cephalopod is semelparous — some squid and the larger deep-sea octopuses are poorly characterised.
Describing the process does not explain what selects for it.
Still unanswered
Why is a programmed death after one reproductive episode favoured in an animal that learns as much as an octopus does?
Wang and Ragsdale, 2018 · Journal of Experimental Biology
The optic-gland signalling changes accompanying maternal decline.
How we know
Removing two small glands, and watching a dying octopus start eating again
A female octopus stops eating while brooding her eggs and dies shortly after they hatch. Is she starving, or is something switching her off?
Starvation and programmed death predict the same thing — a thin animal that dies — so watching a brooding female cannot distinguish them. The manipulation that can is removal of the optic glands, small hormone-producing structures behind the eyes, in females that had already laid and entered the decline. If the decline is exhaustion, taking the glands out should change nothing. If the glands are driving it, the decline should stop.
What happened
Females whose optic glands were removed resumed feeding, regained condition, abandoned their eggs, and lived substantially longer than brooding females left intact.
What it shows
That post-reproductive death in octopuses is signalled rather than incidental. Two small glands hold the whole programme — the refusal to feed, the care of the eggs, and the decline — and removing them releases all three together, which is why abandoning the eggs is part of the result rather than a side effect.
What it does not show
It does not explain why a programmed death after one brood is favoured in an animal that learns as much as an octopus does. Nor does it establish that every cephalopod works this way: the manipulation has been done in a small number of species, in aquaria, and the deep-sea octopuses that brood for years are entirely unstudied. An operated animal is also not a normal one, and abandoning eggs is not evidence about what an intact female experiences.
The controls — what makes this evidence rather than a story
Surgery performed after laying, so the animals are already in the state being explained.
Feeding, movement and survival time compared against unoperated brooding females of the same species.
Both glands removed in the treatment group, since the effect had to be attributable to the structure rather than to the surgery.
Egg care recorded as well as feeding, because the behaviour that stops is as informative as the behaviour that resumes.
Most octopuses live one to two years. A female mates once, lays tens of thousands of eggs, and then stops eating: she guards and aerates the clutch continuously until it hatches, and dies around the same time. Males decline after mating too.
This is not simple exhaustion. Removing the optic glands — small structures behind the eyes — halts the decline, and a female treated that way resumes feeding and lives considerably longer. Later transcriptome work showed the glands are not a single switch but several signalling pathways firing in sequence after laying, which is why the process is so orderly.
The consequence is worth sitting with. An octopus cannot accumulate decades of experience, cannot learn from a parent it never meets, and cannot pass anything on. Whatever it works out, it works out alone and within about a year — which makes what it does manage considerably more impressive, and rules out the cultural transmission that explains so much corvid and cetacean behaviour.
Words used here
Semelparity
Reproducing once in a lifetime and then dying. Salmon and many octopuses do it; most mammals and birds do not.
Optic gland
A small hormone-producing structure behind an octopus’s eye, roughly analogous in position to a vertebrate pituitary, that controls the post-reproductive decline.
Two captive octopuses out of eight repeatedly pushed a bottle into a current to watch it come back
Preliminary
Early results only. Treat as a lead rather than a conclusion.
Giant Pacific octopuses given inedible floating objects showed exploratory handling that declined with familiarity. Two of eight individuals subsequently jetted the object into a tank inlet current so that it returned, repeatedly — behaviour meeting the standard behavioural criteria for play, occurring after habituation rather than as exploration.
Who this applies to
two individuals of one species in captivityDo not extend this beyond the taxa listed — the popular version over-reaches.
Studied in
Enteroctopus dofleini
You may have heard
“Octopuses play with toys”
Two animals did, in one study, in a tank. That is a real and carefully-designed observation and it is not a property of octopuses. The habituation step is what makes it interesting — the behaviour appeared only after the object stopped being novel.
Why we rate it this way, and what the caveats are
PreliminaryLow confidence
The design is right — habituation first, so exploration is excluded — and the result rests on two individuals in one captive study, with "play" defined behaviourally in a way that remains contested for invertebrates.
How far it can be extended
Two of eight animals in a single study is not a basis for a claim about octopuses in general, and NatureHQ records it as an individual-level observation.
Caveats
Six of the eight animals never showed the behaviour.
Captive animals with little other enrichment may behave unusually.
Whether the behavioural definition of play applies to invertebrates is disputed.
Still unanswered
Does anything resembling play occur in wild octopuses?
Is the behaviour individual temperament, or a response to a barren tank?
Mather and Anderson, 1999 · Journal of Comparative Psychology
The original observation, including the six animals that did not do it.
The study everyone cites gave eight giant Pacific octopuses inedible floating pill bottles. All eight explored them at first and lost interest as they became familiar. Two then began jetting the bottle into the tank’s inlet current so that it circled back to them, over and over.
The habituation step is what makes this more than a curiosity. Investigating a novel object is exploration and every animal does it; doing something repetitive with a boring object after exploration has stopped is the behavioural definition of play. That two animals did it and six did not is the part that usually goes missing, and it is why NatureHQ records this as an observation about individuals rather than a property of octopuses.
Words used here
Habituation
Responding less to something as it becomes familiar. The control that separates play from exploration.
Octopuses have a sleep stage that looks like REM sleep
Well supported
Good evidence backs this, though some details remain open.
Octopuses cycle between a quiet sleep state and brief active episodes involving rapid skin-pattern changes, eye and arm movement, and wake-like neural activity. The structural resemblance to vertebrate REM sleep is clear; nothing about subjective content is established.
Who this applies to
two octopus species studied so far
Studied in
Octopus insularis, Octopus laqueus
You may have heard
“Octopuses dream, and you can watch them dreaming.”
The colour changes during active sleep are real and were filmed. What they mean is unknown. The same evidence is equally consistent with the nervous system rehearsing patterns with nothing being experienced at all — and no design available today can tell the difference.
Why we rate it this way, and what the caveats are
Well supportedHigh confidence
Behavioural characterisation in one species and simultaneous neural recording in another, converging on the same picture. High confidence in the sleep state; the dream question is a separate matter and is not claimed here.
How far it can be extended
Two independent laboratories found the same two-stage pattern in different species, which suggests it is a feature of octopuses rather than of one animal.
Caveats
Small samples, as is normal for this work.
A sleep stage with REM-like structure does not establish that anything is experienced during it.
Two species; octopuses vary widely.
Still unanswered
Does active sleep serve memory consolidation, as REM sleep appears to in vertebrates?
Could any experiment address dream content in an animal that cannot report?
Showed wake-like neural activity and waking skin patterns during the active phase.
The honest position is narrower than the coverage and still remarkable: octopuses have a sleep stage with the behavioural and neural signature of REM sleep, arrived at independently of vertebrates. What happens inside it is unknown, and no current method could find out.
One octopus carries coconut shells around to hide in later
Well supported
Good evidence backs this, though some details remain open.
Veined octopuses (Amphioctopus marginatus) excavate, clean and transport coconut and clam shell halves across open sediment at a clear cost to locomotion, assembling them into shelter when threatened.
Who this applies to
veined octopuses, in Indonesian waters
Studied in
Amphioctopus marginatus
Why we rate it this way, and what the caveats are
Well supportedHigh confidence
Repeatedly filmed across twenty individuals and hundreds of diver hours. The cost of carrying is what makes the tool-use interpretation defensible rather than a matter of definition.
Caveats
One species, on soft sediment where shelter is genuinely scarce.
Whether the animal anticipates a future need, or follows a simpler rule, is unresolved.
Observational; no experimental manipulation.
Still unanswered
Do other octopus species do this where suitable objects are available?
Octopuses carried shells while walking awkwardly on stiffened arms, then assembled them as shelter.
Octopuses are less solitary than their reputation when conditions crowd them together
Emerging evidence
Real findings exist, but too few or too recent to be settled.
At a high-density site in Jervis Bay, Australia, Octopus tetricus used body colour and posture in contests: dark colour was associated with aggression and with winning, pale colour with retreat.
Who this applies to
gloomy octopuses, at one unusually dense site
Studied in
Octopus tetricus
Why we rate it this way, and what the caveats are
Emerging evidenceModerate confidence
Careful video analysis of a genuinely unusual site. It is one site, one species, and correlational — colour is associated with contest outcome rather than shown to determine it.
Caveats
A single site with atypically high density.
Correlational.
Still unanswered
Is octopus social behaviour rare, or is it simply rare to find the conditions that reveal it?
The famous result that octopuses learn by watching each other has not held up cleanly
Contested
Researchers actively disagree, and the disagreement is substantive.
A 1992 experiment reported that untrained Octopus vulgaris copied a trained demonstrator's choice after few observations. Replication attempts have been inconsistent, and simpler explanations such as local enhancement are difficult to exclude.
Who this applies to
the common octopus, in one laboratoryDo not extend this beyond the taxa listed — the popular version over-reaches.
Studied in
Octopus vulgaris
Why we rate it this way, and what the caveats are
ContestedModerate confidence
We are confident that the finding is unsettled, and that is the useful thing to record. The paper is real, influential and published in a leading journal; the field does not treat it as established.
How far it can be extended
The original result has not been reliably reproduced, so extending it to octopuses in general would compound an unsettled finding with an unsupported generalisation.
Caveats
Widely cited in popular writing as an established fact about octopus intelligence.
Where researchers disagree
Subsequent attempts to reproduce observational learning in octopuses have given inconsistent results.
Local enhancement — the demonstrator simply drawing attention to a location — can produce the same outcome without any social learning.
Octopus vulgaris is largely solitary, which makes an ability to learn from conspecifics ecologically awkward to explain.
Still unanswered
Would a modern replication with adequate controls find the effect?
Shows that octopuses do interact socially where density is high, which bears on whether social learning is ecologically plausible.
This is included deliberately. A 1992 Science paper is still cited in popular writing as established fact, and the field does not treat it that way. Recording the state of an argument is more useful than quietly dropping the paper or repeating it.
An octopus steers a single arm through an opaque maze by watching it, without the precise limb control a vertebrate would use.
Changes how the 1992 result reads
This does not overturn the 1992 result, but it changes the picture around it. The popular account of octopuses — semi-independent arms, minimal central control — turns out to be wrong in a specific way: the animal can direct an arm when it needs to. Meanwhile the social-learning finding has not replicated consistently, so the two have moved in opposite directions.
A large expansion of protocadherin genes — cell-adhesion molecules involved in wiring nervous systems — on a scale otherwise seen mainly in vertebrates.
This is a young field, and the timeline shows it: a landmark early result that has not replicated cleanly, a burst of solid mechanism work three decades later, and a recent finding whose interpretation is openly unsettled. NatureHQ keeps the 1992 study on the record precisely because its status changed — removing it would hide how the field actually moved.
Why it matters: It is one of the most conspicuous unsolved problems in sensory biology.
What would settle it: A behavioural test capable of distinguishing the chromatic-aberration proposal from skin-based light sensing.
Does anything happen during octopus active sleep?
Why it matters: It is the closest science has come to asking whether an invertebrate has experiences, and the honest answer is that current methods cannot reach it.
Why is an animal this complex dead within two years?
Why it matters: Almost every other animal with a large nervous system is long-lived. Octopuses are the conspicuous exception, and nobody has a complete explanation.