Insight Editions · No. 291
Other Minds
The Octopus, the Sea, and the Deep Origins of Consciousness
Peter Godfrey-Smith
Peter Godfrey-Smith argues that subjective experience evolved gradually from the interplay of sensing and acting in animals, rather than switching on late in a few big-brained mammals.
A study guide to Other Minds by Peter Godfrey-Smith. Shared free for personal reading; not for sale and not endorsed by the authors or publishers. Please read and buy the original.
The Book on One Page
Other Minds is philosopher and diver Peter Godfrey-Smith's account of how subjective experience evolved, told through the animals that evolved large brains on a branch of life entirely separate from ours: octopuses, cuttlefish and squid. Because our last common ancestor with them was a millimetre-scale worm living about 600 million years ago, cephalopods are an independent experiment in building a mind. Godfrey-Smith's central claim is that experience did not switch on late in a few big-brained mammals but grew gradually out of the evolution of sensing, acting and signalling, arising more than once and taking forms as strange as an octopus whose arms are partly self and partly not.
Subjective experience evolved gradually from the interplay of sensing and acting in mobile animals; simple feelings came first, and consciousness is a later transformation, not the origin, of experience. Cephalopods show that minds evolved at least twice, in very different designs.
Evolution of animals and nervous systems from single cells through the Ediacaran and Cambrian; cephalopod biology, behaviour and skin; philosophy of mind (Nagel, James, Hume, Vygotsky, workspace theories); the evolution of ageing; and field observations at the octopus site Octopolis.
~600 million years to the human-octopus ancestor; ~500 million neurons in an octopus, most in the arms; ~10 million chromatophores in a giant cuttlefish; 1 to 2 year lifespans; ~270 million years since octopuses and cuttlefish split.
Anyone curious about animal minds and consciousness who wants evolution, philosophy and natural history in one place; and anyone thinking about non-human or machine minds, since the octopus is the best natural test of which features of mind are general.
Sensing came before animals, and nervous systems came from turning it inward. Bacteria like E. coli already link sensors to motors and use memory of a moment ago to swim toward food; single cells signal to one another. When cells joined into animals, sensing and signalling between organisms became sensing and signalling within them, and a fast, targeted version of that became the nervous system.
Two views of what nervous systems were first for. The sensory-motor view says they linked perception to action; the action-shaping view (Pantin, Keijzer) says they first solved the problem of coordinating many cells into one coherent action. The quiet "Garden of Ediacara" suits the second; the predator-driven Cambrian explosion, with its eyes, claws and arms races, made the first central. From then on, the mind evolved in response to other minds.
Cephalopods took a unique road. From limpet-like shellfish they became gas-floated predators, then abandoned the shell. The octopus became a body of pure possibility, with no fixed shape, ~500 million neurons, most of them in the arms, and an intelligence that is curious, opportunistic and individual rather than social. Its arms mix central steering with local autonomy, a design that fits neither the "brain as CEO" picture nor standard embodied cognition.
Experience comes before consciousness. Against latecomer views tying experience to workspaces and inner models (Dehaene, Baars, Milner and Goodale), Godfrey-Smith defends a transformation view: pain and the primordial emotions are old, basic and found in very different brains (fish, crabs, octopuses); later integration, memory and inner speech reshaped experience. It began as something like "white noise" and probably arose several times in the Cambrian.
Skin as a screen, and a mind leaking onto it. Cephalopod skin is a brain-controlled display of muscle-driven pigment sacs over reflecting layers, evolved for camouflage, co-opted for signalling, and in big cuttlefish emitting "chromatic chatter". Most cephalopods are colour-blind, but octopus skin can sense light, and chromatophores may act as filters that let a single photoreceptor detect colour.
Humans add loops. Inner speech, plausibly grown from efference copies, is a tool for broadcasting ideas, System 2 thinking and self-reflection. It is the second great internalisation of communication, after the nervous system itself. Cephalopods, who cannot see their own displays, lack such loops.
Big brains, short lives. Octopuses and cuttlefish live one or two years because shell loss made them vulnerable, and the Medawar-Williams-Hamilton theory shows that high outside mortality lets late-life decay accumulate. The same ancient event explains both their neural richness and their compressed lives.
Octopolis and parallel minds. At a shell bed seeded by one dropped object, gloomy octopuses live densely and socially, signalling with dark and pale colours and engineering their own habitat: a glimpse of a possible evolutionary road. Genetic dating and cuttlefish episodic-like memory show that large brains evolved separately even within cephalopods, as in mammals and birds. All of it began in the sea, which we are now stressing in ways that are hard to see.
| Ch | Title | One-line payload |
|---|---|---|
| 1 | Meetings Across the Tree of Life | The human-octopus ancestor lived ~600 million years ago, so cephalopod minds are an independent experiment: the closest we come to meeting an intelligent alien. |
| 2 | A History of Animals | Sensing and signalling in single cells were turned inward to make nervous systems; the action-shaping Ediacaran gave way to the sensory-motor Cambrian arms race. |
| 3 | Mischief and Craft | Cephalopods evolved from shelled mollusks to the shell-less octopus, a curious, individual, opportunistic mind with most neurons in its arms, outside the usual body/brain divide. |
| 4 | From White Noise to Consciousness | Sentience precedes consciousness: pain and primordial feelings arose early and several times, then were transformed by integration and workspace-like processing. |
| 5 | Making Colors | Cephalopod skin is a brain-wired screen for camouflage, signalling and chatter; colour-blind animals may sense colour through their skin. |
| 6 | Our Minds and Others | Inner speech, built from efference copies, broadcasts ideas and powers reflection: the second internalisation of communication, a loop cephalopods lack. |
| 7 | Experience Compressed | Medawar, Williams and Hamilton explain ageing; shell loss made cephalopods vulnerable, so they got big brains and very short lives. |
| 8 | Octopolis | A shell city of social octopuses shows ecosystem engineering and colour signalling; large brains evolved in parallel within cephalopods; the ocean that made minds is under strain. |
Meetings Across the Tree of Life
Cephalopods are an independent experiment in building a mind. The last common ancestor of humans and octopuses was a tiny, flattened, worm-like animal living about 600 million years ago, so whatever intelligence and experience an octopus has, evolution built it on a separate track from ours. Godfrey-Smith uses that fact to set up his project: a philosopher's attack on the mind-matter problem that proceeds by tracing, gradually and without jumps, how subjective experience arose in animal life.
Two Meetings
The book opens with two encounters in Australian water that frame everything after. Each shows the same odd quality: cephalopods watch you back, and the contact feels mutual.
The shell bed (2009). Matthew Lawrence, a scuba diver exploring a large bay on Australia's east coast by letting his empty boat drift above him, found a flat sandy area with a pile of thousands of scallop shells around what looked like a single rock. About a dozen octopuses each sat in a shallow excavated den, bodies football-sized or smaller, colours shifting moment by moment, eyes with dark horizontal pupils ("cats' eyes turned on their side"). They watched Matt and watched each other; some roamed out and occasionally fell into multi-armed wrestles; dozens of six-inch baby sharks lay among them. The octopuses seemed neither friends nor enemies but in "complicated coexistence". This site becomes Octopolis, the subject of the final chapter.
The second meeting came a couple of years earlier, while Godfrey-Smith was snorkelling a boulder-strewn bay in Sydney. Under a ledge he found what looked like "an octopus attached to a turtle": a flat body, a prominent head, eight arms coming straight from the head, a skirt-like fin rippling along its back, and skin that seemed to be every colour at once, red, grey and blue-green, with veins of silver like glowing power lines. It was a giant cuttlefish, about three feet long, and it came forward to look at him.
- Mutual engagement. He dived repeatedly on held breath because the animal seemed as interested in him as he was in it. That sense of two-way attention is the thread he keeps pulling for the rest of the book.
- Cuttlefish touch. Very occasionally a giant cuttlefish reaches out an arm a few inches to touch a diver's arm, usually once and no more.
- Octopus touch. Octopuses show stronger tactile curiosity: sit at a den and reach out, and an arm or two will explore you and then try to haul you into the lair. Sometimes that is an overambitious bid for lunch, but octopuses are demonstrably interested in objects they clearly know they cannot eat.
The Departure: 600 Million Years Ago
To understand these meetings we have to go back to their opposite, a departure. Around 600 million years ago a population of small marine animals split. They were perhaps flattened worms, millimetres long, swimming or crawling, possibly with light-sensitive patches on each side as the only marker of "head" and "tail". They had nervous systems, either diffuse nets or a slight clustering into a tiny brain. Diet, reproduction and lifestyle are unknown. Their one great feature is visible only in hindsight: they were the last (most recent) common ancestors of you and an octopus.
The Tree and Its Fork
The history of animals has the shape of a tree: one root, then repeated splits. Large differences between living animals (beetle versus elephant) began as tiny, insignificant speciation events that left two near-identical groups evolving independently thereafter. Picture yourself sitting on top of this irregular, cone-shaped tree because you are alive now, not because you are superior. Looking across, you see your cousins; looking down, your ancestors. For any two living animals, trace both lines down until they meet.
| Pair of living animals | Last common ancestor | Roughly when | What it was like |
|---|---|---|---|
| Human and chimpanzee | Early hominoid ape | ~6 million years ago | Close relative; a short step down the tree |
| Mammals and birds (all the "smart" vertebrates) | Lizard-like land vertebrate | ~320 million years ago | Backbone, four limbs, head, skeleton, senses like ours, well-developed central nervous system |
| Human and octopus | Small flattened worm | ~600 million years ago | Millimetres long, possibly eye-spots, simple nervous system; no life on land yet; largest neighbours sponges and jellyfish |
The step back to the human-octopus ancestor is nearly twice as long as the step to the mammal-bird ancestor. Everything we usually call "smart" (chimps, dolphins, dogs, cats, and since the last few decades of research crows and parrots) clusters within the 320-million-year vertebrate bush. Cephalopod intelligence sits in an entirely different part of the tree.
Two Branches After the Split
- Our branch leads to vertebrates and, within them, mammals and eventually humans.
- The other branch holds most familiar invertebrates: crabs, bees, spiders, centipedes, moths, many worms, and the mollusks (clams, oysters, snails, scallops). Its animals are mostly small with small nervous systems; some insects and spiders manage very complex social behaviour on tiny brains.
- The exception: cephalopods. A subgroup of the mollusks, related to clams and snails, they evolved large nervous systems and behaviour unlike any other invertebrate, entirely on their own path.
Cephalopods are an island of mental complexity in the sea of invertebrate animals. If we can make contact with them as sentient beings, it is not because of shared history or kinship but because evolution built minds twice over, which makes the octopus probably the closest we will come to meeting an intelligent alien.
The four living cephalopod groups are octopuses, cuttlefish, squid and the nautiluses (shelled deep-sea animals of the Pacific that live very differently). Cuttlefish are relatives of octopuses but closer to squid. Octopuses, cuttlefish and squid share one further feature: large and complex nervous systems.
The Project: Mind and Matter, Approached Through Evolution
Godfrey-Smith is a philosopher, and the problem he targets is philosophy's classic one: how sentience, intelligence and consciousness fit into the physical world. His method is evolutionary. Animals began as one of several "unruly clumps of cells" living together in the sea; some then took up mobility and activity, sprouting eyes, antennae and manipulators. Somewhere along that road came subjective experience: for some animals there is something it feels like to be them, a self of some kind that experiences what goes on.
Why Cephalopods Matter to the Argument
- They are remarkable in themselves. If they could talk they could tell us a great deal.
- They correct our anthropocentrism. When we imagine simpler minds, we tend to picture scaled-down humans; cephalopods force contact with something genuinely different.
- Similar eyes, different brains. The octopus eye is camera-like, with an adjustable lens focusing an image on a retina, yet the brain behind it differs "on almost every scale". Of all other minds, theirs are the most other.
- They shaped the route. Following them through the water, a deliberately bodily kind of philosophy, became the author's way into the abstract questions.
Philosophy is "among the least corporeal of callings", with no equipment or field stations. This project was different: diving gear, gases, pressure and the easing of gravity in green-blue light shaped it, and those practical demands reflect real differences between land and water. The sea is the original home of the mind, or at least of its first faint forms.
The Questions the Book Takes On
Saying "brains evolve, neurons are added, some animals get smarter" refuses to engage with the hard parts. The book asks:
- Which were the earliest and simplest animals with subjective experience of any kind?
- Which animals first felt damage as pain?
- Does it feel like something to be a large-brained cephalopod, or is it "dark inside"?
There are two sides of the world that must fit together and do not yet visibly do so: sensations felt by an agent, and the world of biology, chemistry and physics. Charting the evolution of senses, bodies and behaviour is the way to make progress. The book aims for breadth (many kinds of animals) and depth in time (the long regimes of life's history), moving in and out of philosophy "through seams that you won't much notice". Good philosophy is opportunistic, using whatever information and tools help.
William James (1890): the demand for continuity has proved prophetic in science, so we should conceive the dawn of consciousness so that it does not look like "the irruption into the universe of a new nature". He wanted a theory built from comprehensible transitions, no sudden jumps. Godfrey-Smith adopts this gradualism as his working assumption.
The Hawaiian creation story (via Roland Dixon, 1916): life arrives in successive conquests, corals, then worms and shellfish, each destroying its predecessor, with the octopus swimming on as "the lone survivor from an earlier world". Godfrey-Smith rejects the conquest picture: the octopus is not a survivor but a second expression of what was present before.
The octopus is not Ishmael from Moby-Dick, who escaped alone to tell the tale, but a distant relative who came down another line.
The second Earth. Imagine finding a planet where large-brained animals evolved from an ancestor no more complex than a flatworm. Whatever we learned about their minds would tell us which features of minds are forced by the job of controlling a body in a world, and which are accidents of our lineage. The octopus is that planet, available by scuba. Shared features (camera eyes, curiosity, play) become candidates for general laws of mind; differences (distributed control, skin that talks) show what else is possible.
Godfrey-Smith extended this project in Metazoa (2020), which surveys a wider range of animals (sponges, corals, shrimp, fish, insects) and develops his gradualist account of how experience arose, and in later work on the evolution of subjectivity. On the timing of the fork, molecular-clock estimates for the split between our lineage (deuterostomes) and the lineage leading to octopuses (protostomes) still vary widely, often falling between roughly 550 and 650 million years ago; the book's "about 600 million years" remains a fair round figure.
A History of Animals
Before there were minds there was sensing and signalling between single cells; when cells joined into animals, that outward-facing machinery was turned inward and became the nervous system. Godfrey-Smith tracks this from bacteria through the quiet "Garden of Ediacara" to the predator-driven Cambrian explosion, and frames it with two rival accounts of what nervous systems were first for: linking sensing to action, or shaping coordinated action at all.
Beginnings: Senses Before Animals
The Earth is about 4.5 billion years old; life began perhaps 3.8 billion years ago; animals arrived much later, perhaps a billion years ago but probably after that. For most of Earth's history there was life but no animals: a world of single-celled organisms in the sea, a mode of life that still dominates. That world was never a scatter of solitary islands. Single cells live in truce, coexistence and genuine collaboration, and they already sense and react.
Any organism that senses needs three things: a receptive part, an active part that makes something happen, and an arc connecting them. Even bacteria have this layout.
How E. coli finds food. E. coli has a chemical sense (taste or smell) built from sensor molecules bridging its outer membrane (input) and flagella for swimming (output). It has two motions: run (straight line) and tumble (random change of direction). It is too small to tell the direction a chemical comes from, so it uses time to deal with space: one mechanism registers conditions now, another records conditions a moment ago. If things are getting better, keep running; if not, tumble. Swimming straight merely because concentration is high could carry the cell away from "chemical nirvana".
Eukaryotes and the Beginnings of Seeing
Eukaryotes, the larger cells with elaborate internal structure that later formed animals, arose perhaps 1.5 billion years ago when one bacterium-like cell swallowed another. Many have richer taste and swimming abilities, and they edge toward vision.
- Light's two roles. Light can be an energy resource (most single-celled use of light is "sunbathing") or a source of information about other things. The second is hard for tiny organisms, which struggle even to tell light's direction.
- Eyespots. Some eukaryotes have light-sensitive patches coupled to something that shades or focuses the light, making it informative.
- Flexible policies. Some seek light, some avoid it, some switch: seek light when energy is low and avoid it when supplies are full, or approach moderate light and flee dangerous intensities. Each case needs a control system linking eyespot to swimming machinery.
From Sensing to Signalling
Bacteria are attracted even to chemicals they cannot eat. Biologists increasingly read bacterial senses as attuned to other cells, since surface receptors detect chemicals bacteria themselves excrete, sometimes as mere metabolic overflow. Once the same chemicals are both produced and sensed, coordination becomes possible: the birth of social behaviour.
Quorum sensing in a Hawaiian squid. Light-producing bacteria (Vibrio fischeri) living inside a Hawaiian squid (the bobtail squid) glow only when enough of them are present. Each senses the local concentration of an "inducer" molecule that all of them make, and follows the rule "the more you sense, the more you make". Once enough light is produced, the squid, which hunts at night, uses it to cancel the shadow moonlight would cast down to predators below (counter-illumination). The bacteria get hospitable housing.
The chemistry of life is aquatic chemistry; we survive on land only by carrying salt water around inside us. Godfrey-Smith arranges cell sensitivities in an ascending series:
- Sensitivity to external conditions. All cells met so far respond to their surroundings.
- Sensitivity to other organisms. Some cells respond specially to other cells, including their own kind.
- Sensitivity to chemicals made to be perceived. Some respond to chemicals produced because others will detect them, not as byproducts: the threshold of signalling and communication.
- Signalling within an organism. With multicellularity, sensing and signalling between organisms becomes sensing and signalling within one; a cell's "environment" is now mostly other cells.
Living Together: The Origin of Animals
Multicellularity evolved many times: once leading to animals, once to plants, and separately to fungi, seaweeds and others. Animals most likely began not from lone cells drifting together but from a cell whose daughters failed to separate after division, forming a ball of cells that probably ate bacteria. What happened next has two rival scenarios.
Sponge-first (perhaps the majority view). Balls of cells settled on the sea floor and fed by filtering water through body channels, becoming sponges. Adult sponges are fixed, but their larvae swim and sense their world without brains. Perhaps some larvae kept swimming, became sexually mature in the water, and founded all other animal lineages: "the mothers of all the other animals". This rests on sponges being the living animals most distantly related to us.
Comb-jelly-first. Recent work suggests ctenophores (comb jellies: transparent globes with bands of beating hair-like combs) may have branched off before sponges. That does not mean our ancestor looked like a modern comb jelly, but it suggests a different start: a clump of cells folding into a filmy globe swimming in a simple rhythm, a "hovering ghost-like mother" rather than a sponge larva that refused to settle.
Distant does not mean old: living sponges are the product of as much evolution as we are. Either way, new multicellular bodies need coordination, and the old behavioural capacities of free-living cells become its basis. Coordination plays two roles in animals: building the organism (shared with plants) and, on a much faster timescale, a nervous system, which in some animals concentrates into a brain, "a chemo-electrical storm of repurposed signalling".
The sponge versus comb-jelly question has largely been settled in the comb jellies' favour. In 2023, Darrin Schultz and colleagues (Nature) compared ancient chromosome-level gene linkages across animals and found patterns best explained if ctenophores are the sister group to all other animals, sponges included. Not everyone is fully persuaded, but this is now the leading view. It raises the striking possibility that neurons evolved twice, or were lost in sponges, since comb jellies have nervous systems and sponges do not. A 2023 Science study (Burkhardt and colleagues) also found that the comb-jelly nerve net is a fused, syncytial network, its neurons joined into one continuous cell membrane rather than linked by synapses, unlike any other animal's.
Neurons and Nervous Systems
A neuron's activity rests on two things:
- Electrical excitability. The action potential is an electrical spasm that travels along a cell in a chain reaction. It predates animals: the first one ever measured was in a plant, the Venus flytrap, at Charles Darwin's instigation in the nineteenth century, and some single-celled organisms have them.
- Chemical signalling. A neuron releases a spray of chemicals into the cleft between it and another cell, triggering or suppressing that cell's action potential. This is "the residue of ancient signalling between organisms, pressed inward".
What nervous systems add is a particular kind of signalling: it is fast (plants mostly act slowly), and it is targeted, since long projections let one cell influence a few distant cells rather than broadcast to all neighbours. Broadcasting becomes an organised network.
- ~4.5 billion years: age of the Earth.
- ~3.8 billion years: origin of life.
- ~1.5 billion years: origin of eukaryotic cells.
- ~700 million years: rough date of the first nervous system, from DNA evidence alone.
- 635 to ~542 million years ago: the Ediacaran period (the book's dates).
- Nearly a quarter: share of a human's food energy spent running the brain; firing neurons is like charging and discharging a battery hundreds of times a second.
- ~34: animal phyla (basic body plans); only 3 contain animals with complex active bodies.
What Are Nervous Systems For? Two Views
This is the chapter's main conceptual tool, and it returns throughout the book.
The sensory-motor view. The original and fundamental function of a nervous system is to link perception to action. The senses track the environment; the nervous system uses that information to choose what to do. There is always a go-between bridging sensors and effectors, as even E. coli shows, and on this view that bridging role was central at the start and at every stage since.
The action-shaping view. For a multicellular animal, producing an action at all is a hard problem: countless tiny contractions and twitches of parts must be shaped into one coherent whole-body movement, micro-actions into a macro-action. Early nervous systems may have existed mainly to solve this internal coordination problem. Developed by Chris Pantin (English biologist, 1950s), revived by the philosopher Fred Keijzer.
The social analogies make the difference concrete:
- Paul Revere (1775). The sexton of Boston's Old North Church watched the British army and signalled by lantern ("one if by land; two if by sea"): the sexton is a sensor, Revere a muscle, the lantern a nerve. This is the sensory-motor kind of communication.
- The rowing crew. Several rowers each pull one oar; it hardly matters when they pull as long as they pull together, so someone calls the stroke. This is purely coordinative communication.
- The coxswain does both. The cox is the crew's eyes and coordinates micro-actions. A nervous system can likewise play both roles with no conflict.
Pantin and Keijzer's point is that treating each "action" as a ready-made unit leaves only the question of when to do X rather than Y, and ignores how an organism can do X at all. The problem barely exists for bacteria or seaweed; it grows with body size and behavioural complexity. Through most of the twentieth century the sensory-motor view was simply assumed.
The orchestra with no audience. Imagine an orchestra in a sealed, soundproof hall with no score from outside and no listeners. The conductor still has a job: without a beat, eighty musicians produce noise. Now open the doors and let a critic shout requests. The conductor gains a second job, responding to the outside, but the first job never went away. Godfrey-Smith's suggestion is that the earliest nervous systems may have been conductors in the sealed hall, and that the Cambrian opened the doors.
The Garden of Ediacara
Reginald Sprigg's lunch (1946). An Australian geologist checking abandoned mines in the Ediacara Hills of South Australia, hundreds of miles from the sea, turned over a rock while eating lunch and saw what looked like fossil jellyfish. Nature rejected his paper; it finally appeared as "Early Cambrian (?) Jellyfishes" in the Transactions of the Royal Society of South Australia (1947), beside "On the Weights of Some Australian Mammals". It took about a decade for anyone to realise he had found the first fossil record of animals older than the Cambrian, and that most were not jellyfish at all.
The Ediacaran period, named after those hills, runs from 635 to about 542 million years ago. At the South Australian Museum in Adelaide, Jim Gehling, who knew Sprigg and has worked on the fossils since 1972, showed Godfrey-Smith slabs dense with life: dozens of fossils per rock. Pressing Silly Putty into nondescript patches revealed fine imprints of tiny animals.
What Ediacaran Life Was Like
- Size. Many were several inches long, some up to three feet.
- Habitat. They lived on and among microbial mats on the sea floor, "a kind of undersea swamp".
- Mobility. Many adults were anchored and motionless; some may have been early sponges and corals.
- Strange designs. Others had body plans since abandoned by evolution: three-sided and four-sided forms and quilted, frond-like arrangements.
- Kimberella. The clearest self-propelled case: an oval animal like the top half of a macaron, with front and back and perhaps a tongue-like appendage. Its traces show it pushed sediment and scratched surfaces, perhaps feeding. Often read as a mollusk or near-mollusk; if it crawled at several inches long, it almost certainly had a nervous system.
- Dickinsonia trails. Fainter traces of the same shape near Dickinsonia fossils suggest it fed at one spot, then moved on.
- Spriggina. Named after Sprigg, sometimes drawn swimming, but Gehling thinks it crawled, because its fossils are always found the same way up; a swimmer killed by some accident would sometimes land upside down.
Some biologists argue the Ediacarans were an animal-like experiment that was not properly animal, a different way for cells to make an organism. The standard view is mixed: some (like Kimberella) belong to familiar animal groups; others are abandoned evolutionary detours, alongside ancient algae.
Dickinsonia was an animal. In 2018, Ilya Bobrovskiy and colleagues (Science) extracted molecular fossils from exceptionally preserved Dickinsonia specimens in Russia and found cholesterol-type steroids characteristic of animals, strong evidence that at least this iconic Ediacaran belonged to the animal kingdom. Separately, the base of the Cambrian is now dated to about 538.8 million years ago on the international timescale, a little younger than the book's 542.
A Peaceful World, and Its Puzzle
The consistent theme is that the Ediacaran was largely without conflict or predation: Mark McMenamin's "Garden of Ediacara". "Peace" may be the wrong word, since it suggests a truce; Ediacarans simply had little to do with each other. They grazed mats, filtered water, sometimes roamed, and left almost no evidence of interaction.
- No predation evidence. There are no half-eaten remains; only Cloudina shows possible, unclear damage.
- No big sense organs. No large eyes, no antennae, though some response to light and chemicals is near certain.
- No weapons or shields. No claws, spikes or shells. "Macarons that pass in the night."
If Ediacaran lives lacked moment-to-moment engagement with other animals, their nervous systems probably fit the action-shaping view: they were for shaping movement, keeping rhythms and crawling, with only a little sensing.
Two cautions Godfrey-Smith raises himself. First, the fossil record may mislead: chemical interaction and soft sense organs leave no trace, just as single-celled life turned out to be full of hidden chemical signalling. Second, jellyfish. Cnidarian stinging cells seem to have evolved early (both sides of a major cnidarian split, dated to the Ediacaran or earlier, share the same stingers), and "a garden of stinging jellyfish, as any Australian will insist, is far from Edenic". At the Royal Society of London's 2015 meeting on early nervous systems, nobody could say whom the earliest stings were aimed at, since neither modern prey nor modern foes yet existed.
The Cambrian Explosion
From about 542 million years ago, in a relatively sudden burst, most basic animal body plans appeared: vertebrates (as fish), arthropods (external skeleton, jointed limbs, such as trilobites), worms and others. The timing may reflect changes in Earth's chemistry and climate, but the speed probably came from evolutionary feedback between organisms, especially predation.
- Budd's resource story. The British paleontologist Graham Budd argues that animals changed how food was distributed: on an endless lawn of mats, grazers ate a uniform resource; then animals themselves became concentrated patches of carbon, first eaten after death, then hunted.
- Gehling's sequence. Near the end of the Ediacaran, scavenging arose first, then predation.
- The fossil signature. From the early Cambrian we find exactly what the Ediacaran lacked: eyes, antennae and claws, plus legs and fins. Legs are ambiguous about interaction; claws are not.
- Arthropods set the pace. Trilobites, with shells, jointed legs and compound eyes, rise early in the Cambrian.
In the Ediacaran other animals might be around you without mattering. In the Cambrian each animal becomes a major part of the environment of others, and from this point on, the mind evolved in response to other minds.
The Senses Open
If a yard-long anomalocarid, a fast-swimming predator like a giant cockroach with two grasping head appendages, is swooping toward you, it pays to know. The Cambrian saw the first image-forming eyes, both compound eyes (as in insects) and camera eyes (as in us).
- Andrew Parker argues that the invention of eyes was the decisive event of the Cambrian.
- Roy Plotnick and colleagues call the result a "Cambrian information revolution": more incoming information demands more internal processing and harder decisions (which hole is the anomalocarid less likely to intercept?).
Bilaterians and the Box Jellyfish Exception
A jellyfish has radial symmetry: a top and bottom but no left and right. Humans, fish, octopuses, ants and earthworms are bilaterians, with front and back and so left and right. Godfrey-Smith pictures early bilaterians as small worms with eyespots on each side of the head (controversial, and he admits he is "being generous"). Kimberella may have been one, so bilaterians were already a little more active before the Cambrian; in the Cambrian they became "unstoppable". Walking is a very bilateral thing to do, and the body plan suits complex behaviour.
Box jellyfish (Cubozoa): the smartest non-bilaterians. Probably late arrivals (Cambrian or after), they carry venom strong enough to have killed large numbers of humans, are invisible in the water, and close northeastern Australian beaches for much of each year except inside netted enclosures. They have two dozen sophisticated eyes with lenses and retinas around the top of the body, swim at about three knots, and some steer by landmarks on the shore. They too are products of the Cambrian world.
Complex Active Bodies
The philosopher Michael Trestman picks out animals with complex active bodies (CABs): fast movement, the ability to seize and manipulate objects, appendages moving in many directions, and senses like eyes that track distant objects.
| Phylum | CAB species? | Examples |
|---|---|---|
| Arthropods | Yes | Insects, crabs, spiders, trilobites |
| Chordates | Yes | Fish, birds, mammals, us |
| Mollusks | Only cephalopods | Octopus, squid, cuttlefish |
| The other ~31 phyla | No | Sponges, cnidarians, most worms, starfish |
Of about 34 phyla, only 3 produced CABs, and within the mollusks only the cephalopods qualify. That is the company the octopus keeps.
Did the Role of Nervous Systems Shift?
Godfrey-Smith fits the two views onto the timeline as one possibility: the very first nervous systems mainly coordinated action (animating an ancient cnidarian, shaping Ediacaran crawling); by the Cambrian, with more worth watching and inattention meaning death, tracking the outside world became central. "If there was such an era, by the Cambrian it was over."
Detlev Arendt and colleagues propose that nervous systems originated twice in the same animal, in different body regions: - Top system. In a dome-shaped, jellyfish-like ancestor with a mouth underneath, a nervous system on top tracked light to regulate bodily rhythms and hormones, not behaviour. - Movement system. A second nervous system controlled movement, at first just of the mouth. - The merger. Over evolutionary time part of the motion-controlling system moved up to meet the light-sensitive organs, which became eyes. Arendt sees this joining as one of the key events that pushed bilaterians forward in the Cambrian.
The Fork
Soon after bilaterians appeared, probably still in the Ediacaran, a population of small flattened worms with neurons and perhaps simple eyes split in two. One side led to vertebrates (along with surprising companions such as starfish); the other to most invertebrates, beetles, lobsters, slugs, ants, moths. On both sides, through the Ediacaran and Cambrian, lives became entangled, senses opened and nervous systems expanded, until a rubber-suited mammal and a colour-changing cephalopod ended up staring at each other in the Pacific.
The sensory-motor versus action-shaping distinction maps neatly onto agentic AI design. Most discussion of agents treats them as perception-to-decision pipelines (the Revere model), but a large multi-component system has a second, often harder problem: making many sub-agents, services and tools produce one coherent action at all (the rowing crew). Orchestration, shared clocks and transaction coordination are the "stroke call"; a system that perceives well but cannot coordinate its parts fails the way a large animal with no nervous system would.
Mischief and Craft
Cephalopods went from limpet-like shellfish to the top predators of the ancient seas, then gave up their shells; the octopus ended with almost no hard parts and a body of pure possibility. Its intelligence is curious, opportunistic and individual rather than social, and its half-billion neurons, most of them in the arms, may exist first to make that boneless body controllable at all. The result is a mind that fits neither the "brain as CEO" picture nor the standard embodied cognition story.
Two Meetings With Octopuses
- In a sponge garden. A cat-sized octopus, tangled in orange sponge and grey-green seaweed, seems "everywhere and nowhere": only the small head and two eyes hold a definite shape. It circles to keep the sponge between you, its skin matching the seaweed exactly, while tiny skin peaks (papillae) match the orange of the sponge at their tips. Eventually it raises its head and rockets off by jet.
- At a den. A tennis-ball-sized octopus sits behind a front yard of shells and old glass. Offer a finger and one arm uncoils to touch it, grips disconcertingly hard, and tugs you in. The arm carries hundreds of sensors in each of dozens of suckers; it is tasting your finger, and the arm itself is "a nest of nervous activity", while the eyes watch the whole time.
Evolution of the Cephalopods
Octopuses are mollusks, kin to clams, oysters and snails. Mollusks became conspicuous in the Cambrian because of their shells, the molluscan answer to the invention of predation: if suddenly surrounded by creatures that can see and want to eat you, live under armour.
- Ediacaran (?) Kimberella, a crawling oval animal, may be an early mollusk (controversial).
- Early Cambrian A limpet-like mollusk crawls on a muscular foot under a cap-shaped shell, less than an inch long; the cap grows "Pinocchio-like" into a horn.
- Later Cambrian Some lift off the sea floor: gas in a chambered shell turns "a limpet into a zeppelin". The foot, useless aloft, gives rise to jet propulsion through a steerable siphon and to tentacles, some armed with hooks.
- Paleozoic Cephalopods become the most fearsome predators in the sea; straight and coiled shells; Cameroceras reaches 18 feet or more. "Hovercrafts and tanks" prowl the bottom.
- Later Mass extinctions and competition from ever larger, better-armed fish wipe out almost all shelled forms: "the zeppelins were challenged, and eventually vanquished, by airplanes."
- Before the dinosaurs Some cephalopods abandon, reduce or internalise the shell, trading protection for freedom of movement, a gamble taken several times.
- ~290 million years ago The oldest possible octopus fossil: one specimen, "little more than a smudge".
- ~164 million years ago A fossil undeniably like an octopus, with eight arms in an octopus-like pose.
- Today About 300 octopus species, from under an inch long to the giant Pacific octopus (about 100 pounds, 20 feet arm tip to arm tip).
The Survivor and the Shell-Losers
- Nautilus: the real "lone survivor". The Hawaiian myth gave that title to the octopus, but it belongs to the nautilus, little changed over 200 million years. It lives in a coiled shell as a scavenger with simple eyes and a cluster of tentacles, rising higher at night and deeper by day in a rhythm still being studied. Why it survived, no one knows.
- The great split. The modern line divided into an eight-armed group (octopuses) and a ten-armed group (cuttlefish and squid).
- Three solutions to the shell. Cuttlefish keep an internal shell that still aids buoyancy; squid retain a sword-shaped "pen"; octopuses lost the shell entirely.
With almost no hard parts (the eyes and beak are the largest), an octopus can squeeze through a hole about the size of its eyeball and reshape itself almost without limit. Forgoing both skeleton and shell is a rare move for an animal this large and complex. The evolution of cephalopods yielded, in the octopus, a body of pure possibility.
The cephalopod fossil record has filled in since 2016. The ~290-million-year "smudge" is generally identified as Pohlsepia, and the ~164-million-year octopus-like fossil as Proteroctopus from France. In 2022, Christopher Whalen and Neil Landman described Syllipsimopodi bideni, a roughly 328-million-year-old fossil from Montana that they placed as the oldest known member of the vampyropod lineage (octopuses plus the vampire squid). It had ten arms, supporting the view that the octopus line descended from ten-armed ancestors and later lost a pair.
Puzzles of Octopus Intelligence
How Big Is the Brain?
- Absolute size. A common octopus (Octopus vulgaris) has about 500 million neurons, versus roughly 100 billion in humans. That puts it in the range of smaller mammals, close to dogs, and far beyond every other invertebrate.
- Relative size. Brain weight as a fraction of body weight (counting only brain neurons) is also high: roughly in the vertebrate range, below mammals.
- Both are crude guides. Brains differ in organisation and in the number and complexity of synapses. The big surprise of recent animal research is parrots and crows: small brains in absolute terms but very high-powered.
There is no single scale of intelligence. Godfrey-Smith's analogy is brains as tool kits for controlling behaviour: all include some perception; almost all bilaterians have memory and learning; some add problem solving and planning. Different tool kits suit different ways of making a living.
The octopus comparison is especially hard. Its eyes are built on the same camera design as ours, yet beneath them "all mappings are off": vertebrate brains (fish, bird, mammal) share an architecture that allows part-by-part comparison, but octopus brains have no part-by-part correspondence with ours, and most octopus neurons are not in the brain at all but in the arms. So the only way to gauge octopus smarts is to look at what they do.
The Lab Versus the Anecdote
The core puzzle is a mismatch between laboratory results and anecdotes. In the lab octopuses do "fairly well, without showing themselves to be Einsteins": they learn simple mazes, use visual cues to identify which of two environments they are in and take the right route, and unscrew jars for food. But they learn slowly. The anecdotes suggest much more, especially an ability to adapt to captivity and turn the apparatus to their own purposes.
Peter Dews and Charles the octopus (1959). Peter Dews, a Harvard pharmacologist influenced by B. F. Skinner's operant conditioning (itself built on Edward Thorndike's law-of-effect work around 1900), tested whether octopuses at the Naples Zoological Station could learn to pull a lever for a reward (a light came on, and a piece of sardine arrived). All three learned. Albert and Bertram worked the lever gently and "reasonably consistently". Charles passed minimally but: (1) anchored himself to the tank and applied such force that he bent the lever repeatedly and broke it on day 11, ending the experiment; (2) repeatedly wrapped the lamp above the tank and tried to drag it in; (3) spent much time with eyes above water squirting jets at anyone who approached. Dews wrote drily that the variables maintaining "the lamp-pulling and squirting behavior in this animal were not apparent."
Godfrey-Smith had first read, in Roger Hanlon and John Messenger's Cephalopod Behaviour, that the experiment failed because lever-pulling is unnatural for an octopus. Going back to the paper, he found it succeeded in its main goals. The real lesson is that Dews's behaviourist framework assumed animals of a species start identical and diverge only through reinforcement history. Charles was probably not reinforced for squirting; he had a feisty temperament. Octopus work repeatedly shows great individual variability. "Unnatural" tasks are no problem (octopuses open screw-cap jars, and one has been filmed opening a jar from the inside). The trouble was assuming an octopus would keep pulling for sardine morsels like a rat or pigeon. Octopuses take time over each item, probably cannot cram, and lose interest; stealing the lamp or soaking the experimenter was more interesting.
Captive Octopuses Know Where They Are
| Report | Who / where | What it suggests |
|---|---|---|
| Lights out | University of Otago (NZ) and a German lab | Octopuses learned to squirt the bulbs and short-circuit the power, so expensively that one was released |
| The disliked staffer | Otago | One octopus soaked one particular person with half a gallon of water on the neck whenever she passed |
| Squirting strangers | Shelley Adamo, Dalhousie | A cuttlefish squirted new visitors but not regulars |
| Recognising people | 2010 experiment | Giant Pacific octopuses recognise individual humans even in identical uniforms |
| Flooding the lab | Stefan Linquist | Octopuses plugged outflow valves with their arms, perhaps to raise the water |
| The squid in the drain | Jean Boal, Millersville | An octopus held its thawed squid, crossed to the outflow watching her, and dumped it down the drain |
| Escape timing | Godfrey-Smith, David Scheel | Octopuses bolt the moment you look away |
The lights-out story has a deflationary reading: octopuses dislike bright light, squirt at anything annoying, and roam further when no humans are near. Against it, both accounts suggest the octopus learned very quickly that getting into position and aiming works. Godfrey-Smith wants an experiment that separates the explanations.
When you work with fish, they have no idea they are in a tank, somewhere unnatural. With octopuses it is totally different.
Boal matters as a witness because she is known as one of the most rigorous and sceptical cephalopod researchers, insisting that "cognition" be invoked only when nothing simpler explains the result. Watching-dependent escape makes ecological sense (bolt when the barracuda is not looking), but doing it so quickly with humans, masked or not, is impressive.
Welfare history. Difficulty motivating octopuses led some researchers to use electric shocks more freely than with other animals. Much early Naples work removed parts of the brain or cut major nerves to see what the animal did on waking, and octopuses could be operated on without anaesthetic because invertebrates fell outside cruelty rules. Over the decade before the book, octopuses became "honorary vertebrates" in experimental regulation, especially in the European Union (its 2010 Directive 2010/63/EU on animals used in science brought live cephalopods under protection). Godfrey-Smith calls this a step forward.
Play
Jennifer Mather and Roland Anderson (Seattle Aquarium) ran the first studies of octopus play: some individuals, and only some, blow pill bottles around the tank with their jets, "bouncing" them on the intake current. An octopus's first interest in a new object is gustatory (can I eat it?), but Michael Kuba's lab work confirms they quickly recognise non-food and often stay interested in exploring and manipulating it.
Visiting Octopolis
Matthew Lawrence marked his site's GPS position and kept returning (always with a fully independent second air supply; solo diving is not recommended). The octopuses explored his camera and hoses. He saw "bullying": a larger octopus would jump on a denned one and wrestle until the lower animal shot out like a rocket, pale, landing metres away. Once, at a nearby site, an octopus took his hand and led him across the sea floor for ten minutes, "like a very small eight-legged child", ending at its den. When Matt posted photos to a cephalopod enthusiasts' site, the biologist Christine Huffard asked Godfrey-Smith whether he knew the place; it was only a few hours from Sydney.
- The site. About 50 feet down, nearly invisible until close; a bed of scallop shells a couple of yards across, shells of many ages, with a foot-high encrusted rock-like object in the middle used as a den by the largest octopus.
- The species. Octopus tetricus, the gloomy octopus, medium-sized, found only in Australia and New Zealand.
- The behaviour. Octopuses eye each other from dens, haul out and roam; an arm probes, an arm answers, and either both settle or a wrestling match follows.
- The stake. Godfrey-Smith hammered seven-inch plastic stakes weighted with bolts into the sand at the compass points, nearly invisible. Months later one had been pulled out and added to the debris around a distant den, presumably for its novelty, like tape measures and cameras.
Coconut-shell houses (2009). Researchers in Indonesia saw wild octopuses (the veined octopus, Amphioctopus marginatus) carrying pairs of halved coconut shells discarded by people. One half nested in the other; the octopus carried the stack under its body in an awkward "stilt-walk", then assembled the halves into a sphere around itself. Hermit crabs use found shelters and chimps and crows use food tools, but assembling, disassembling and transporting a compound object is very rare. Nests are compound objects, but nobody takes a nest apart and reassembles it.
The coconut houses show the distinctive shape of octopus intelligence: they are smart in the sense of being curious and flexible, adventurous and opportunistic.
Three Routes to Complex Active Bodies
Of the three groups with complex active bodies, arthropods went first (early Cambrian, over 500 million years ago) and may have triggered the feedback that drew in the rest. Setting chordates aside:
Arthropods: social complexity. Most species are not social, but many of their great behavioural achievements are: ant and honeybee colonies, termites' air-conditioned cities. Complex behaviour is achieved by coordinating many individuals, each with a small brain.
Cephalopods: lone complexity. Never went onto land; started later but evolved larger brains. Some squid are social, but nothing like ants. Cephalopods, squid partly excepted, acquired a non-social form of intelligence; the octopus followed "a path of lone idiosyncratic complexity".
Nervous Evolution
The history of large brains has the shape of a letter Y. At the fork sits the vertebrate-mollusk common ancestor: probably worm-like, with neurons, perhaps simple eyes, and some bunching of neurons at the front but no real brain. From there nervous systems evolve independently.
- Our side: the chordate design. A nerve cord down the back with a brain at one end: fish, reptiles, birds, mammals.
- The cephalopod side: the ladder. Invertebrate neurons typically sit in many ganglia (small knots) spread through the body, paired and linked by connectors running along and across the body like lines of latitude and longitude: the "ladder-like" nervous system. Cephalopod neuron multiplication was built on this plan. Some ganglia grew large, new ones were added, and neurons concentrated at the front into something like a brain, but the old architecture is only partly submerged.
- The oesophagus runs through the brain. Food passes through the middle of the central brain. Swallow something sharp that pierces the "throat" and it goes straight into the brain; octopuses have been found with exactly this injury.
The Arms
- ~500 million: neurons in a common octopus.
- Nearly 2 to 1: neurons in the arms versus the central brain; the majority are in the arms.
- Up to ~10,000: neurons per sucker, handling taste and touch.
- Hundreds: chemical and touch sensors per sucker.
- 3: hearts, pumping blue-green, copper-based blood.
Each arm has its own sensors and controllers, with touch and chemical sense (smell or taste). Even a surgically removed arm can reach and grasp. Early behavioural and anatomical work made the arms seem very independent: the nerve channel from each arm to the brain looked slim, and some studies suggested octopuses did not track where their own arms were. Hanlon and Messenger called the arms "curiously divorced" from the brain in basic motor control.
- The temporary elbow. When an arm pulls in food, the grasping tip launches two waves of muscle activation, one travelling inward from the tip and one outward from the base; where they meet, a joint forms, a temporary elbow.
- Arm memory? Arm circuitry contains recurrent connections (loops) that might give each arm simple short-term memory; what it does for the octopus is unknown.
- But central control exists. Approach a wild octopus and pause, and in some species it sends out one arm first to inspect you while it watches, which suggests deliberate, brain-guided action.
Guiding one arm by eye (Hebrew University, 2011). In Binyamin Hochner's lab, Tamar Gutnick, Ruth Byrne and Michael Kuba (with Hochner) built a maze where an octopus had to steer a single arm to a food location. The arm's chemical senses could not do it alone, because the arm had to leave the water at one point; but the walls were transparent, so the goal was visible. Learning took a long time, but nearly all octopuses tested succeeded: the eyes can guide the arms. Successful arms also did their own local exploration, crawling and feeling as they went. Conclusion: central control of the overall path plus local fine-tuning by the arm.
The arms turned out to be even more self-directed. In 2020, Lena van Giesen, Nicholas Bellono and colleagues (Cell) identified a family of chemotactile receptors in octopus suckers that let the arms "taste by touch", detecting molecules on surfaces such as prey. The 2015 sequencing of the California two-spot octopus genome (Caroline Albertin and colleagues, Nature) showed large expansions of gene families linked to neural development (notably protocadherins), and later work found octopuses and squid edit their RNA extensively, especially in neural genes. None of this overturns the book's picture; it fleshes out the distributed design.
Body and Control: Why So Many Neurons?
Neurons are expensive, so why half a billion? Godfrey-Smith offers two partial answers, then combines them.
1. Foraging: Gibson's Extractive Hunter
Octopuses are mobile hunters, not ambushers, and they are not very social, so the usual explanation for big brains (the demands of social life) fits badly. He adapts the primatologist Katherine Gibson's 1980s account of large mammalian brains:
Specialised foraging. One food, handled the same way every time: a frog snapping flying insects. Little need for flexible cognition.
Extractive foraging. Varied foods, choices adapted to circumstance, food extracted from shells and casings in a context-sensitive way: a chimp cracking nuts and fishing termites. Octopuses fit: crabs first, but also scallops, fish and other octopuses, with defences to beat.
Teaching octopuses clams. David Scheel, who works on giant Pacific octopuses in Prince William Sound, Alaska, feeds whole clams, which his local octopuses do not normally eat. He first gives a partly smashed clam. Given an intact one later, the octopus knows it is food but not how to open it: it drills, chips with the beak, manipulates it every way, and finally learns that sheer strength suffices to pull the shell apart.
This fits octopus curiosity about novel objects better than it fits cuttlefish and squid, which handle food more simply. Yet some cuttlefish have very large brains, perhaps larger relative to body than octopuses, which remains a mystery. Predator-prey engagement is also "social" in a broad sense: it requires tuning actions to what others can see and are likely to do.
2. Coordination: the Action-Shaping View Returns
When part of the molluscan foot became a mass of jointless, shell-less tentacles, the result was enormously useful if controllable and very hard to control. Losing nearly all hard parts compounded both. The octopus solved this not with centralised governance but with a mix of local and central control, turning each arm into an intermediate-scale actor while imposing order top-down. The pure-coordination demands that may have shaped the first nervous systems reappear here and could account for much of the neuron count: neurons needed just to make the body controllable.
Coordination explains size but not inventiveness: a well-coordinated animal can be uninventive. So the full story is "and/or": some capacities (perhaps recognising individual humans) are by-products of a big nervous system built for the body, while others (problem solving) were shaped by the opportunistic lifestyle. On this picture neurons multiplied for the body, and later "an octopus wakes up with a brain that can do more". The mischief in captivity points to a kind of mental surplus.
Convergence and Divergence
| Feature | Vertebrates | Octopuses | Verdict |
|---|---|---|---|
| Eyes | Camera eye, lens, retina | Camera eye, lens, retina | Convergent: the ancestor had at most eyespots |
| Reward learning | Yes | Yes | Invented independently several times, or greatly elaborated on both lines |
| Short vs long-term memory | Yes | Apparently yes | Convergent |
| Play with useless objects | Yes | Yes (some individuals) | Convergent |
| Sleep | Including REM | Something like sleep; REM-like in cuttlefish | Convergent; octopus REM unclear in 2016 |
| Recognising individuals | Yes | Yes, even humans | Striking, since octopuses are not social or monogamous |
| Hearts and blood | One heart; iron, red blood | Three hearts; copper, blue-green blood | Similar function, different build |
| Nervous system | Centralised; spinal cord | Large but distributed; most neurons in arms | Divergent |
The recognition of individuals carries a general lesson: smart animals carve the world into objects that can be re-identified despite changes in appearance. Godfrey-Smith finds this feature of the octopus mind "striking in its familiarity".
Octopus sleep is now well documented. Marcos Medeiros, Sidarta Ribeiro and colleagues (iScience, 2021) described two sleep states in Octopus insularis: a pale "quiet sleep" and brief bouts of "active sleep" with rapid skin-colour changes, eye and arm twitches, recurring roughly every half hour to hour. In 2023, Aditi Pophale, Sam Reiter and colleagues (Nature) showed that in Octopus laqueus active sleep is a distinct, homeostatically regulated state whose brain activity resembles waking and which replays waking skin patterns, an intriguing parallel to REM and memory consolidation. For cuttlefish, Teresa Iglesias and colleagues (Current Biology, 2019) confirmed recurring REM-like episodes with arm twitches and skin displays. Whether any of this involves dreaming remains unknown.
The Octopus Outside the Body/Brain Divide
Embodied cognition holds that the body itself, not just the brain, carries some of an animal's "smartness": body structure encodes information about the world, as the angles and joints of legs make walking arise naturally. Hillel Chiel and Randall Beer put it as body structure creating constraints and opportunities that guide action. Benny Hochner applies this to octopuses: a different embodiment yields a different psychology.
Godfrey-Smith agrees octopuses have a different embodiment, but argues the doctrine does not fit them:
- Embodied cognition says the body's shape and organisation encode information.
- That requires a definite shape: fixed distances between parts (aiding perception), joints at fixed locations and angles.
- The octopus has no fixed distances, no joints, no natural angles; the same animal stands tall, squeezes through an eye-sized hole, streamlines into a missile, or folds into a jar.
- The standard contrast is "body rather than brain", but in an octopus the nervous system as a whole, not the brain, is the relevant unit; the brain's boundaries are unclear and nervous tissue runs through the body.
- Conclusion: the octopus fits neither the brain-as-CEO view nor the intelligence-in-the-body view. It is, in a sense, "disembodied": material, but with a protean body that has none of the costs and gains of a constraining, action-guiding shape.
"The octopus is suffused with nervousness; the body is not a separate thing that is controlled by the brain or nervous system." The octopus lives outside the usual body/brain divide.
The shapeless robot. Imagine designing a soft robot with no joints and no fixed geometry, able to become a column, a ribbon or a ball. You cannot pre-compute "walking" into its leg lengths, because it has none. Every gait must be computed live, and the cheapest way to do that is to give each limb its own local controller and keep only goals at the centre. You end up with something like an octopus: lots of compute spent on coordination, local autonomy, and a centre that steers rather than commands. The thought experiment shows why shapelessness and distributed neurons may be two sides of one design.
The octopus arm architecture (central goal-setting, local closed-loop execution, sensing embedded at the edge) is a good mental model for agentic AI systems. A central planner that tries to micromanage every tool call scales badly; one that sets the path and lets sub-agents explore locally, as the Hochner maze arms did, is more robust. The cost is the octopus's problem too: the centre may not always know where its own arms are, so observability and audit trails are the price of local autonomy, a real concern in regulated banking workflows.
From White Noise to Consciousness
Godfrey-Smith separates subjective experience (the bare fact that life feels like something) from consciousness (a narrower, more unified and integrated form of it), and argues that experience came first. Against "latecomer" views that tie feeling to global workspaces and inner models found only in mammals and birds, he defends a "transformation" view: simple feelings such as pain and thirst arose early, probably several times independently in the Cambrian, and were later reshaped by integration and memory. The octopus, with arms that are partly self and partly not, shows how strange such experience might be.
The Problem of What It Is Like
William James's demand for continuity suits most of the mind: perception, action and memory creep into existence through intermediates and grey areas. "Do bacteria really perceive?" or "Do bees really remember?" have no clean yes-or-no answers. The hard case is subjective experience. Thomas Nagel asked "What is it like to be a bat?"; the word "like" misleads, since the issue is not comparison but the fact that waking up, watching the sky and eating have a feel. The gradualist must explain how that feel can creep into being, and what it means for an animal to be halfway to life feeling like something.
Sentience Before Consciousness
- Subjective experience is the most basic thing to be explained: that life feels like something.
- Consciousness is, for Godfrey-Smith, one form of subjective experience, not the whole of it.
- Sentience is the older term for having experience at all. Asking whether squid are "conscious" asks too much; the real question is whether damage feels bad to a squid. Sentience comes before consciousness.
Sentience is neither a soul-like substance added to matter (dualism) nor something pervading all nature (panpsychism). It arises from the evolution of sensing and acting: being a living system with a point of view. The difficulty is that sensing and acting reach down to bacteria. Unless all life has a trace of experience (a view he calls "not insane" but in need of much defence), something about how animals deal with the world must make the difference. Raw complexity is too vague; he looks at specific factors.
Factor One: Loops Between Sensing and Acting
What you do next depends on what you sense; what you sense next depends on what you do. You turn the page, and that changes what you see.
Tactile vision substitution systems (TVSS). A video camera feeds a pad on a blind person's skin (for example the back), turning images into vibrations or electrical stimulation. After training, wearers report objects located in space, not touches on the back: a passing dog is felt as something moving out in front. This happens only when the wearer controls the camera (moves it closer, changes its angle), most simply by mounting it on the body so that objects loom and come and go. The moment-to-moment feedback between acting and sensing changes how the input itself feels.
Philosophy long neglected this. Mainstream work studied the link from sensory input to belief, said little about action and less about how action shapes the next input; the dissenters who disliked this "obsession with receptivity" overcorrected into stories of the self-determining subject imposing itself on the world. It is hard, apparently, to accept traffic in both directions.
- Two arcs. There is a sensory-motor arc (senses to action) and a motor-to-sensory arc (action to what is sensed next). The second runs through external, public space, so it is looser: someone may grab the book as you turn the page. Yet it is why we do much of what we do: to control what our senses will meet.
- Against the "stream of experience". A river's flow is almost entirely outside our control; real life lets us reshape the things we interact with.
- Muscle changes everything. All life affects its surroundings chemically and by growth, but muscle, built from fibre-like elements inside cells, gives rapid, coherent action on large spatial scales, and so the manipulation of objects.
The Self/Other Problem
Loops create a problem: an animal must tell changes it caused from changes the world caused.
- Electric fish. Fish that emit electric pulses for communication and also sense electric fields send a copy of each motor command to the sensory system so it can cancel the self-produced pulse. The fish registers the line between "self" and "other" (this is the efference copy, or corollary discharge, mechanism).
- Merker's worm. The Swedish neuroscientist Björn Merker notes the problem arises just from moving: an earthworm withdraws when touched, but crawling touches its own body the same way. If it withdrew at every touch it could never move; it cancels self-produced touches.
- Who faces it. Every organism has a self/world boundary that onlookers can see, but many animals register it themselves because action would otherwise be impossible. Plants sense richly but do not move; bacteria move but their simple senses do not tangle them.
Perceptual Constancy
As you approach a chair it does not seem to grow, because you tacitly compensate for changes due to your own movement (and for some not due to you, like lighting). Perceptual constancies occur in octopuses, some spiders and vertebrates, probably evolving independently in several groups. They show an animal treating external objects as external objects.
Factor Two: Integration, and Its Absence
We usually experience a unified scene binding sight, sound and touch. That seems inevitable for eyes and ears attached to one brain, but it is one way of being wired, and many animals do much less of it.
Split vision in pigeons and octopuses. In animals with eyes on the sides of the head, each eye feeds mainly one side of the brain. Mask one eye, train a task, then test with the other. In pigeons, 8 of 9 birds showed no inter-ocular transfer at all: a skill apparently learned by the whole bird belonged to half the bird. Octopuses trained with one eye at first remembered the task only with that eye, but with extended training could use the other: some transfer, though not easily.
- Lateralisation. Giorgio Vallortigara (University of Trieste) and others have found many "fissures" between brain halves: various species react more to predators on the left; several fish and even tadpoles prefer to keep a same-species individual on their left; many animals detect food better on the right. Vallortigara argues specialised hemispheres for different processing tasks can outweigh the cost of one-sided vulnerability.
- Split-brain humans. Cutting the corpus callosum to treat severe epilepsy leaves people behaving fairly normally, yet when each hemisphere gets different stimuli, two intelligent selves seem to share one skull. The left hemisphere usually speaks; the right controls the left hand, can choose objects by touch and draw, and may disagree with what the mouth reports. That kind of fragmentation seems routine in many animals.
- Pigeon retinas. Each pigeon retina has a red field (small binocular area in front) and a yellow field (larger area the other eye cannot see). Pigeons transferred poorly even between regions of one eye.
- The weaving hen. Marian Dawkins showed hens a novel red toy hammer; they approached in a weaving path that gave different parts of each eye access to it. "The weaving gaze of a bird is a technique designed to slosh the incoming information around."
Some unity is unavoidable in a living agent, which is one physical object keeping itself alive. But in other ways unity is optional, an achievement, an invention: evolution may or may not bring experience together, even the outputs of two eyes.
Latecomer Versus Transformation
The Evidence Behind Latecomer Views
Patient DF (1988). Carbon-monoxide poisoning from a faulty shower water heater damaged the brain of a woman known as DF. She felt almost blind, losing all experience of shape and layout (only vague colour patches remained), yet she could post a letter through a slot set at various angles. She could not describe or point to the angle. Studied by David Milner and Melvyn Goodale.
Ventral stream (lower path through the brain): categorising, recognising and describing objects. Milner and Goodale argue that visual experience comes only from this stream; DF lost it.
Dorsal stream (upper path toward the top of the head): real-time guidance through space, avoiding obstacles, posting the letter. It works unconsciously, in DF and in everyone.
Godfrey-Smith thinks "you need the ventral stream for any visual experience" is too simple: dorsal vision probably feels like something, just not much like seeing. The bigger lesson is that complex processing from eye through brain to hand can run without being experienced as seeing. Milner and Goodale tie experience to building a coherent inner model of the world.
Ingle's rewired frogs (1960s). David Ingle surgically crossed connections in frog visual systems (frog nerves regenerate unusually well). The frogs snapped left at prey on their right and vice versa, yet steered normally around barriers. Part of the visual world was reversed and part normal. Milner and Goodale: asking what the frog "sees" makes sense only if one visual representation governs all behaviour, and Ingle showed this cannot be true; without that assumption "the puzzle disappears". Godfrey-Smith reads them as implying it feels like nothing to be such a frog.
Stanislas Dehaene's lab near Paris studies perception at the edge of consciousness, with images flashed too fast or presented while attention is diverted. Much unexperienced information is processed sophisticatedly: masked word strings with incongruous meanings ("very happy war") are registered differently from sensible ones ("not happy war"). But some things need awareness:
- Novel, multi-step tasks cannot be done unconsciously.
- Trace conditioning needs awareness: you can learn unconsciously to blink to a light followed immediately by an air puff, but once light and puff are about a second apart the association is learned only if you are aware of it.
- Dehaene's conclusion: a particular style of processing, used for time, sequences and novelty, brings conscious awareness; much other complex processing does not.
Bernard Baars's global workspace theory (1980s): we are conscious of information brought into a centralised workspace; Dehaene developed it. A related family ties consciousness to working memory; Jesse Prinz (Godfrey-Smith's colleague at the City University of New York) defends one version.
| Latecomer view | Transformation view | |
|---|---|---|
| Claim | Experience arises with late-evolving machinery: workspace, working memory, integrated inner model, sense of time and self | A form of experience preceded that machinery, which then transformed what it feels like to be an animal |
| Supporters | Milner and Goodale, Dehaene, Baars, Prinz (in effect) | Godfrey-Smith |
| Where experience is | Primates, dolphins, perhaps other mammals, some birds | Many animals with active bodies: likely vertebrates, cephalopods, some arthropods |
| Simple animals | Attributing experience is projection of a faint human picture | Simple experience exists but is unlike ours |
| Key evidence | Blindsight-like dissociations, masked priming, DF, rewired frogs | Primordial emotions, flexible pain behaviour across very different brains |
| Origin story | Waking up came late, though not as a sudden flash | Experience arose in the Cambrian, several times, then was reshaped |
Note the terminology trap: Baars, Dehaene and Prinz offer theories of "consciousness". If consciousness is narrower than experience, a workspace could be needed for consciousness but not for the most basic feeling, which is roughly Godfrey-Smith's view. Some of these theorists, though, deny any distinction and claim to explain when anything feels like anything. He credits latecomer research with real progress and warns against clinging to a view merely because it is more generous.
The Argument From Intrusions
Godfrey-Smith's best argument for transformation, set out as premises:
- Some experiences enter our mental life as intrusions into organised, complex processing: sudden pain, and what the physiologist Derek Denton calls primordial emotions, feelings registering important bodily states such as thirst or air hunger.
- These feelings are "imperious": they press into experience and resist being ignored.
- Their character and role look old and basic, not dependent on sophisticated cognition.
- It is implausible that pain or breathlessness feel like something only because of late-evolving mammalian processing such as inner models or working memory.
- So an animal can plausibly feel pain or thirst without a workspace or inner model, and the latecomer view is wrong about the origin of experience, even if right about the origin of consciousness.
The Evidence From Pain
Squirming is not enough. Rats with a severed spinal cord still show some "pain behaviour" and even damage-related learning with no channel to the brain; empathy makes reflexes look like suffering. The telling evidence is pain-related behaviour too flexible to be reflex, in animals whose brains fail latecomer criteria.
| Animal | Experiment | Result | Why it matters |
|---|---|---|---|
| Zebrafish | Preference between two environments tested; then injected with a suspected pain-causing chemical; painkiller dissolved in the less-preferred environment | Fish switched to the less-preferred side only when it held painkiller | A novel choice evolution could not have pre-wired |
| Chickens | Birds with damaged legs offered a normally less-preferred food | Chose it when it contained painkillers | Self-medication suggests felt pain |
| Hermit crabs (Robert Elwood) | Small electric shocks to induce crabs to leave their shells | Needed stronger shocks to leave better shells; tolerated more when predator odour made the shell more valuable | Trade-off against other priorities, not reflex |
| Insects | Observations after severe injury | Behave normally as far as they physically can; no grooming or guarding of injuries | Suggests pain is not universal |
| Crabs, some shrimp | Observations | Groom injured areas | Wound-tending, a pain indicator |
Scepticism is always possible, "but you can doubt that about your next-door neighbor". The results build a case for pain as a basic, widespread form of experience present in very different brains.
Insects have since moved toward the "maybe" column. Research from Lars Chittka's group (2022) found that bumblebees trade off avoiding a noxious heat source against the reward of sweeter sugar, the same kind of motivational trade-off as Elwood's hermit crabs, and the same group reported bumblebees rolling balls with no reward, which they interpret as play. A 2022 review led by Matilda Gibbons argued that several insect orders meet a number of the criteria for pain used in the UK sentience review. The question remains open, but the book's confident contrast between crabs and insects is weaker now.
White Noise: Imagining the Earliest Experience
The chapter title comes from Simona Ginsburg and Eva Jablonka, Israeli biologists who sketched the evolutionary origins of experience and offered, for a simple distant animal, the image of white noise: "a poorly differentiated buzz as the beginning of it all". It is a metaphor of sound for animals that mostly could not hear, but it points the right way: a crackle of metabolic electricity, experience starting inchoate and becoming organised.
- Why does experience track perception and control, not cell division? In our case experience is closely bound to using what we sense to decide what to do, not brimful of bodily rhythms or "life itself".
- Godfrey-Smith's clue. Experience arises not from the mere running of the system but from the modulation of its state, from registering things that matter and require response, internal or external. "Sentience has some point to it."
- A twist on the metaphor. Perhaps white noise is the absence of experience, what came before; from it arose the old forms tied to primordial emotions, pain and pleasure.
Tentative Conclusions About Early Animals
- The action-shaping era may have felt like nothing. If early nervous systems mostly generated and maintained activity (a swimming medusa's rhythmic pulse, self-possessed Ediacarans), with little modulation, that life may have had no experience.
- Simple experience began in the Cambrian, with richer engagement with the world.
- Multiple origins. The major branches had already split in the quiet Ediacaran, so if crabs, octopuses and cats all have experience, it arose at least three times, perhaps many more.
- Consciousness, too, arose several times. As the machinery described by Dehaene, Baars, Milner and Goodale came online, experience became integrated and gained a firmer self. "Consciousness" is "a mixed-up and overused but useful term" for unified, coherent forms of experience, not one definite step.
The dimmer, not the switch. Picture a dimmer dial running from silence to full consciousness, and ask where a pain-avoiding hermit crab sits. The latecomer says the dial stays at zero until a workspace is installed, then rises. The transformation view says the dial began moving with feelings that steer action (pain, air hunger) and that a workspace rewired what the light illuminates rather than switching it on. One test: if pain in us feels as raw during absorbed, routine activity (when workspace contents are elsewhere) as during reflection, that suggests pain's feel does not depend on workspace access.
The Case of the Octopus
An octopus has a large nervous system, a complex active body, rich senses and extraordinary behaviour. If experience comes with sensing and acting, it has plenty. It also shows, in "elusive and alien form", some of the sophistications beyond the basics.
- Novelty and curiosity. Dehaene links consciousness to novelty, which "jolts us from unconscious routine". Octopuses embrace novelty, sometimes cautiously, sometimes with puzzling recklessness (the octopus that led Matt Lawrence by the hand).
- The arm from the weed. While Godfrey-Smith hovered photographing sea slugs, anchored by a few fingers, a single slender arm slid toward those fingers from a clump of seaweed where the octopus lay balled up, one eye watching through a hole: cautious exploration under close attention, with cover and a viewing port.
- Perceptual constancy. A 1956 experiment trained octopuses to approach some squares and avoid others by size (reward or electric shock). On "several" occasions the small square was shown at half the usual distance, making its retinal image larger; in every such trial the octopus responded to the true size. It was a passing aside with no numbers and no follow-up, but if accepted, octopuses have constancies (as do honeybees and some spiders).
- Wayfinding. Foraging trips of fifteen minutes or so through murky water follow loops, leaving in one direction and returning from another. Jennifer Mather charted such loops for a Caribbean octopus. How they do it (landmarks, memory) is unknown.
- Self and other. Octopuses identify and re-identify objects, seem to grasp a self/other distinction, and direct considerable attention, especially at new things. The worm-like common ancestor had none of these skills; they are convergent solutions for an active, mobile life.
Pain at Octopolis
A large aggressive male kept "standing tall" and lifting his rear high above his head, apparently to look bigger before attacks. A small, vicious fish, a leatherjacket, darted in and bit him on the rear.
He clearly felt it but showed no wound-tending, probably because the injury was minor and he was busy fighting. Jean Alupay and colleagues studied another species in which octopuses can amputate their own arms to escape predators. After a controlled arm crush, some animals amputated the arm and some did not, but all tended and guarded the injured site for some time, the classic pain indicator.
Octopus pain and the law. Robyn Crook (2021, iScience) gave octopuses an injection of dilute acetic acid and found they avoided the chamber where it happened, tended the site, and preferred a chamber where they had received local anaesthetic (lidocaine), which also silenced the nerve signals from the injured arm: evidence for an affective, not merely reflexive, pain state. In November 2021 a team at the London School of Economics led by Jonathan Birch delivered a review to the UK government finding strong scientific evidence of sentience in cephalopod mollusks and decapod crustaceans. As a result the Animal Welfare (Sentience) Act 2022 recognises all cephalopods and decapods as sentient for UK policy purposes. In April 2024 the New York Declaration on Animal Consciousness, signed by many scientists and philosophers, stated that the evidence gives strong support for conscious experience in mammals and birds and at least a realistic possibility of it in all vertebrates and many invertebrates, explicitly including cephalopods, decapod crustaceans and insects. Ginsburg and Jablonka have meanwhile developed their view into a proposed marker of minimal consciousness, unlimited associative learning, in The Evolution of the Sensitive Soul (2019).
The latecomer theories have been tested head to head. An adversarial collaboration (Cogitate) pitted global neuronal workspace theory against integrated information theory with pre-registered predictions; results first presented in 2023 and published in Nature in 2025 challenged key predictions of both (for workspace theory, the expected "ignition" at stimulus offset was not found). In 2023 an open letter from over a hundred researchers also called IIT "pseudoscience", provoking strong objections. None of this decides the latecomer debate, but it shows that workspace accounts are not settled science.
Arms: Partly Self, Partly Not
If octopuses delegate partial autonomy to their arms, then from the central brain's perspective the arms are partly self and partly agents of their own.
- The breathing analogy. Blinking and breathing run automatically but attention can take them over. Octopus arm control is similar, except that central guidance is never complete: "you would send an arm out deliberately and hope the local fine-tuning goes right."
- Blurred boundaries. For us, the self/environment border is clear: we control our arm in path and detail, and uncontrolled motion signals "not me". For an octopus, "to some extent you would guide your arms, and to some extent you would just watch them go."
- Caveat. This tells the story from the "central octopus" viewpoint, which may itself be an error.
Godfrey-Smith's first reading: octopus action blends things that are distinct in us, deliberate control and spectating, making octopus experience of action genuinely alien.
Bence Nanay's reading (philosopher, Antwerp): the octopus is an extreme case of something in us. If a reaching target suddenly shifts, the hand corrects in under a tenth of a second, unconsciously; subjects deny any change occurred. Top-down command plus fast unconscious fine-tuning is human too; in octopuses the local part is larger, slower, guided by the arm's own chemical and touch senses, and partly watchable.
Chiel and Beer contrast the old view of the nervous system as conductor with a new one: it is one player in a jazz improvisation. Godfrey-Smith thinks that understates the nervous system in most animals, but it may fit the octopus, with the contrast now between central brain and the rest of the nervous body.
In the octopus there is a conductor, the central brain, but the players are jazz musicians who accept only so much direction, or who receive rough instructions and are trusted to play something that works.
Two ideas here translate directly to AI assurance. First, Dehaene's trace-conditioning point (long-gap associations need awareness) mirrors the way capable systems need explicit working memory to link events separated in time; fraud and AML detection across long transaction gaps likewise needs state that persists, not just pattern-matching on the latest event. Second, the Ingle frog warns against asking "what does the model believe?" as if one representation drove all behaviour; modern interpretability work finds multiple partly independent pathways, so a model can "know" a fact on one route and act against it on another. Test the behaviour you care about directly.
Making Colors
A cephalopod's skin is a layered video screen wired directly to the brain, with something like ten million muscle-driven pigment "pixels" over layers of light-bending mirrors. It evolved for camouflage, was co-opted for signalling, and in big animals like the giant cuttlefish it also leaks a constant "chromatic chatter" that may simply express what is going on inside. The deep puzzle is that most cephalopods are colour-blind, and the emerging answer is that the skin itself can sense light.
The Giant Cuttlefish
A giant cuttlefish (Sepia apama) looks like "an octopus attached to a hovercraft":
- Arms. Eight flexible, unjointed, suckered arms arranged around the mouth, best thought of as "eight huge and dexterous lips", plus two longer feeding tentacles whipped out to seize prey, and a hard beak.
- Skeleton. No spine or real bones, but a stiff cuttlebone (like the inside of a surfboard) inside the shield-like back.
- Movement. A skirt-like fin a few inches wide on each side for slow undulating swimming; a steerable siphon for jet propulsion.
- Size. Most cuttlefish measure inches; the giant reaches three feet.
Colour change is common in nature (chameleons), but cephalopods are faster and broader in range, and in large cuttlefish the whole body is a screen showing moving shapes, stripes and clouds, not snapshots. Thin silver lines wander over the head as if it were electrified; red trails run from its eyes like "a spaceship crying tears of blood".
Giant cuttlefish are little studied and rarely kept captive. Alexandra Schnell, one of few to study them in the lab, reports hints of the octopus-style response to captivity, including well-aimed squirts at visitors. They have big brains, absolutely and relative to body, but have not shown octopus-style puzzle solving, tool use or object exploration, partly from lack of study and partly because they are swimmers, not clambering explorers. What they do show is friendly curiosity in some individuals: an animal approaches in a quiet "resting" pattern and hovers close, apparently trying to work you out.
How the Screen Works
The skin is a layered screen controlled directly by the brain: neurons run from the brain into the skin, where they control muscles, which control millions of pixel-like sacs of colour. The animal senses or decides something and its colour changes in an instant.
- Dermis (outer covering). A protective top layer.
- Chromatophores. Each unit is a sac of a single pigment surrounded by one or two dozen radial muscle cells under direct neural control; muscles stretch the sac to show colour or relax to hide it. Usually three kinds per species; in giant cuttlefish red, yellow and black/brown, each much less than a millimetre across. They make reds, yellows, oranges, browns, but not blue, green, violet or silver-white.
- Iridophores. Reflecting cells that bounce and filter light through tiny stacks of plates, separating wavelengths to shine back greens and blues. Not wired directly to the brain; some seem controlled more slowly by chemical signals.
- Leucophores. Simple reflectors below the iridophores that mirror light straight back, usually appearing white but able to return whatever colour surrounds them.
- Interaction. Because chromatophores sit above the reflecting cells, expanding them changes what light reaches the mirrors and so what is shone back. The same patch can produce many colours.
- Ten megapixels, roughly. If a giant cuttlefish has about ten million chromatophores, that layer is something like a ten-megapixel screen, with caveats: pixels seem controlled in local clumps, each has one colour, and some overlap.
- Thin magic. The coloured layers are fragile. Old or damaged cuttlefish show dull white patches: "the magic skin is a thin sheet on top of a plain white body".
The Repertoire
- Base colours. Reds, from maroon to fire-engine, decorated with silver-white veins and dots ("jagged flashes or a line of pearls"), plus patches of yellow, orange and olive-green.
- Dynamic patterns. The passing cloud: alternating dark and light waves moving front to back or back to front, "movies played on the screen". Godfrey-Smith once saw a cuttlefish show a passing cloud on its left side to another cuttlefish under a rock while the right side stayed camouflaged toward open sea.
- Shape changes. Dozens of papillae, boneless skin folds an inch high like stegosaurus plates, raised in a second; sculpted "eyebrow extensions" of skin above the eyes.
- Arm postures. Arms are numbered 1 to 4, left and right, from the top. Males have larger fourth arms and flatten them into broad blades when aggressive; first arms are raised as horns, sometimes wavy, or shaped into fiddleheads, hooks and clubs. The most elaborate displays stack arms at three or four levels. Some harmless fish are met with horns and hooks as if positively despised.
Individuals: Matisse, Kandinsky, the Hostile Male, the Indifferent
Re-identifying animals that change colour and shape at will is hard, but scars and some permanent white marks on the fin work as fingerprints. Individuals of the same sex, size, place and season respond very differently.
The single touch. The friendliest individuals hover at a set distance, edging back when he edges forward and vice versa, then let the gap close to a few feet and reach out the tip of an arm or two to touch his hand. Almost always just once, then back to the old distance. Inspecting him as food seems unlikely: cuttlefish eat crabs and fish caught whole, and a human is far too big.
- Matisse. A friendly cuttlefish who, while calmly hovering in reds and whites, would explode into uniform canary yellow in under a second, "a cuttlefish-shaped sun", fading back through oranges to dark red in about ten seconds. No raised arms, no other sign of alarm (overall yellow is described as alarm in other cephalopods); his yellows toward intrusive fish were deeper and came with arm displays. Mechanism: sudden expansion of the yellow chromatophores with everything else scaled back.
- Kandinsky (2009). A cuttlefish with fixed habits and a den twelve feet down; each late afternoon for about a week he would emerge to "man the top of his rock" facing the ocean, arms raised like ceremonial lances, sometimes knotted overhead. He favoured flaring reds and oranges (including a pale orange-green) with passing clouds, and tear-like patterns down his inner arms. Not friendly, but he let Godfrey-Smith follow him on looping tours of the reef.
- The jaws of hell. At a ledge known for friendly animals, a large male produced silver-white welts flaring on and off, a forest of horn-arms, burning orange, sickle arms and skin folds like bent iron armour, and charged "like a jet-propelled medieval siege device". The charges never quite reached him, suggesting bluff. When Godfrey-Smith stopped retreating and raised his own arms, the displays calmed; the cuttlefish looked off over his shoulder, relaxed, and charged again only when he moved directly in front.
- Profound indifference. Some cuttlefish on looping reef tours seem not to register a diver as a living being at all, looking past your shoulder and adjusting to maintain non-contact. In an "existentialist game of chicken", Godfrey-Smith blocked one's path; it came to a foot away, looked up "deeply unimpressed", edged past and swam on. Being ignored like that makes you feel like a ghost who does not realise it is a ghost.
Seeing Colours: the Colour-Blind Colour-Changer
Then comes "a fact that makes absolutely no sense": almost all cephalopods are colour-blind.
- Physiology. Colour vision needs photoreceptors that respond differently to different wavelengths, so brightness can be separated from hue. Most humans have three kinds; colour vision needs at least two; most cephalopods have one.
- Behaviour. Species tested cannot learn to discriminate stimuli that differ only in colour.
- Yet they produce colour lavishly and are superb at matching the colour of their surroundings.
Standard partial explanations: cephalopods may infer likely hue from subtle brightness differences given typical environmental colours, and the reflecting cells can return a colour they cannot see. But reflection fails when an animal matches a colour behind its back while different light comes from in front, and cephalopods do seem able to do this.
- 2010 Lydia Mäthger, Steven Roberts and Roger Hanlon report that the photoreceptor molecules of the eye are probably also present in cuttlefish skin. Inconclusive alone: the molecules might do something else, and one photoreceptor type still cannot yield colour.
- 2010 to ~2015 Little follow-up; Godfrey-Smith finds one graduate student in California, Desmond Ramirez, working on it quietly.
- 2015 Days after Godfrey-Smith publishes a review asking why the lead was dropped, Ramirez and Todd Oakley show that photoreceptor genes are active in the skin of Octopus bimaculoides, and that detached skin is light-sensitive and expands its own chromatophores in response to light.
An octopus can see with its skin. Not merely be affected by light, as many animals are, but respond by changing its own pixel-like colour machinery, which extends the picture of an octopus body that is partly its own controller.
Skin vision could not focus an image; it would register general washes of light. If skin signals reach the brain, visual sensitivity extends in all directions beyond the eyes. If not, each arm might see for itself and keep what it sees to itself. Either way it remains monochrome: the skin photoreceptor has the same sensitivity as the eye's.
How a Monochrome Sensor Could See Colour
- Fixed filter (Mäthger and colleagues). A chromatophore sitting over a light-sensitive cell acts as a coloured filter, so photoreceptors under red and under yellow sacs respond differently to the same light. One photoreceptor type behaves like two; some butterflies use a similar trick.
- Changing filters (Lou Jost). The ecologist, orchid expert and artist Lou Jost suggested that the act of changing colour could do it. As chromatophores of different colours expand and contract over a sensor, the light passing through changes. If the animal tracks which filters are open and how much light gets through, it can infer the incoming colour, "like a cameraman exchanging one filter for another".
- Why it is likely. If any light sensors sit beneath the chromatophores, colour changes inevitably produce effects correlated with incoming colour. The information is available, and using it would be an easy evolutionary step.
Two later proposals. In 2016, Alexander Stubbs and Christopher Stubbs (PNAS) suggested that cephalopods may extract colour from a single photoreceptor using chromatic aberration: different wavelengths focus at slightly different depths, and the U-, W- and dumbbell-shaped pupils exaggerate this, so by shifting focus an animal could tell hues apart. It remains a hypothesis and is debated. Work on the skin side continued too: Gilles Laurent's lab (Nature, 2018 and 2023) tracked tens of thousands of individual chromatophores and showed that cuttlefish camouflage is a dynamic, feedback-driven search toward a match, not a direct readout of the background. And on cuttlefish intelligence, Alexandra Schnell and colleagues reported in 2021 that common cuttlefish pass a version of the "marshmallow test", waiting for a preferred food rather than taking a lesser one immediately, with better self-control predicting faster learning.
Being Seen: Camouflage, Signals and Chatter
Octopuses are unsurpassed at camouflage: invisible to an observer looking for them a few feet away, helped by having no hard parts. Cuttlefish come close.
The reaper cuttlefish and the moray. A small (about six inches) reaper cuttlefish, soft red with "yellow eyeliner", evaded Godfrey-Smith through weeds and vanished in a channel. Knowing these cuttlefish imitate rocks, he looked straight at a small rock, decided it was just a rock, searched elsewhere, came back: it was the cuttlefish. Once spotted, it reverted to dark pink. Then a green moray eel struck. A burst of ink like black smoke; a glimpse of the cuttlefish being shaken. Yet it shot out of the cloud, wildly coloured, flattened, one large bite on its back, still in its yellow eyeliner, swam a "punch-drunk meander", then straightened and headed for another ledge. A boneless "toy hovercraft" had escaped a consummate predator.
| Function | What it is | Directed at | Example |
|---|---|---|---|
| Camouflage (original function) | Colours made not to be seen or recognised; arose as shell-less cephalopods faced toothed fish | Predators and prey | Reaper cuttlefish as rock; octopus matching weed |
| Deimatic display (intermediate) | Sudden dramatic patterns, often while fleeing, to startle or confuse; meant to be noticed but carries no information | Predators | Flash patterns during escape |
| Signalling (co-opted) | Patterns made to be seen and responded to | Rivals, mates | Male contests at Whyalla; squid "no thanks" stripe |
| Chromatic chatter (Godfrey-Smith's proposal) | Side effects of inner electrochemical activity on a skin wired to the brain; no biological function | Nobody | Facial "murmur", idling pulses, colour change when alone |
The Whyalla Mating Displays
Near the industrial town of Whyalla in South Australia, thousands of giant cuttlefish gather each winter to mate and lay eggs in shallow water; no one knows why there.
- Consorts and intruders. A large male monopolises a female. When a rival comes, the two lie side by side, each stretching as far as it can with a gentle curve, blazing with colour, then turning 180 degrees and stretching the other way: "competitive yoga" plus a turn like "a dance from the court of some civilized French king".
- Outcome. The display reveals which male is larger, and the larger almost always wins; the smaller backs off.
- Mating. Head to head and peaceful. The male envelops the female's head with his arms; a couple of minutes of stillness (apparently blowing water at her through his funnel); then he uses his left fourth arm to place a sperm packet in a receptacle below her beak and break it open.
Chatter
Godfrey-Smith accepts that camouflage and signalling are the functions that explain why colour change evolved and is maintained. But not every colour serves a function. Many patterns are neither camouflage nor aimed at any receiver: some cuttlefish and a few octopuses run an almost continual kaleidoscope that seems disconnected from the outside and instead an inadvertent expression of the electrochemical tumult inside.
- The face murmur. Between the eyes and down the first part of the arms there is often an ongoing murmur of tiny colour changes, perhaps the machinery idling; others show steady pulses beneath the eyes "like animated eye shadow".
- Brancusi. A cuttlefish that rarely produced bright colour but held sculptural arm shapes perfectly still for as long as Godfrey-Smith could stay, while the colours on his face stayed constantly restless.
- Why tolerated. Functionless side effects would be suppressed if harmful; they might attract predators to small cephalopods but cost little to a giant cuttlefish that many predators pass by.
- Alternatives. Low-level changes might be surveying the chromatic environment (via skin sensing), or responses to the diver (or to the unattended cameras at octopus dens, which still record unexplained colour sequences when no other octopus is near).
Cephalopods have a sophisticated system built for camouflage and signalling, wired to the brain in a way that produces strange expressive quirks: an ongoing chromatic chatter.
Baboon and Squid: Two Lopsided Communicators
Dorothy Cheney and Robert Seyfarth studied wild baboons in the Okavango Delta, Botswana, documented in Baboon Metaphysics. A troop of about 80; females stay in their birth troop in a hierarchy of matrilines; most males emigrate as young adults and live shorter, more violent lives. Baboons have only three or four call types (threats, friendship grunts, submissive screams), but each caller is individually recognisable.
Playback experiments. Baboons hear, from out of sight, "A threatens, B backs down". If A outranks B, it is unremarkable. If A is below B, the sequence signals a reversal in the hierarchy, and listeners become far more attentive. Baboons build a "narrative" from call sequences, a tool for social life.
| Baboons | Cephalopods | |
|---|---|---|
| Production | Very simple: 3 to 4 calls, reliably tied to interaction types | Almost indefinitely complex: millions of pixels, huge pattern space, extraordinary bandwidth |
| Interpretation | Complex: listeners assemble narratives about rank | Limited: little social life; most output goes unnoticed (and if no one interprets it, little is really "said") |
| Social life | Soap-opera complexity, stress, alliances, grooming | Much simpler; even social squid are nowhere near baboons |
| What leaks out | Little: inner life barely expressed | A great deal of what goes on inside is made available on the outside |
| Summary | Hear much, say little | "Publish or perish": say much, hear little |
The Squid "Language" Claim
In the 1970s and 1980s, Martin Moynihan and Arcadio Rodaniche followed Caribbean reef squid in Panama for years by snorkel. They charted displays (Golden Eyebrows, Dark Arms, Downward Pointing, Flecked Yellow, Upward Curl and more), about thirty ritualised displays in total, with patterned sequences, and proposed a visual language with a grammar, nouns and adjectives. Their monograph appeared in a respectable journal but was unusual, full of personal reflection and illustrated by Rodaniche, who later became an artist.
The production/interpretation gap. Communication needs two roles, sender and receiver. Moynihan and Rodaniche documented rich production but said little about effects: they found a few clear sign-response pairs in mating, but many displays occurred outside that context, and they admitted they could not tell the meaning of most sequences while assuming every distinguishable sequence must differ in function. By their own account squid interactions were not very complex, so why so much display? Later researchers have been sceptical of the language claim. Godfrey-Smith keeps the puzzle open: perhaps subtle social roles, perhaps chatter.
Both baboons and cephalopods are one-sided versions of the basic sender-receiver duality. Evolution is not goal-directed, but Godfrey-Smith "can't resist" seeing an unfinished quality in each: baboons with a soap-opera life and little means to express it; cephalopods with less to say and extraordinary means of saying it.
Symphony
A forty-minute colour concert. Late one summer afternoon a medium-sized male hovered at a favourite den, still, neither curious nor watchful. His skin ran through "rust-with-brick", grey-greens, other reds and faint pale colours, changing in a concerted way, like chords shifting over one another, many with no analogue in the surroundings, so not passive reflection. His eye toward the diver was nearly closed. Godfrey-Smith wondered if this was a cuttlefish dream, like a dog's twitching paws. Then the animal stiffened and ran the strangest sequence he had seen, aimed at no one: arms tucked, beak exposed, a missile pose, a yellow flare, the male competitive sideways stretch, a contorted all-white shape, then, after quieting, a sudden wild aggressive pose, arms like needles, body bright yellow-orange, armour-like papillae, "a wild clashing chord", displayed in all directions. It subsided to the reds, rusts and greens of the start. Then, for the first time, he turned and looked at the diver.
The leaky dashboard. Imagine a person whose face displayed, in real time, a coloured readout of every passing brain state, whether or not anyone was watching. Most of it would mean nothing to observers and would serve no purpose; some of it might be harnessed for bluffing or courtship; and an observer who learned to read it would learn a great deal about that person's inner life. That is Godfrey-Smith's picture of the giant cuttlefish. The thought experiment also shows why chatter, if real, is scientifically precious: a sentient animal whose inner processes are partly externalised on its skin is about as close as we get to a window into an alien mind.
The baboon/cephalopod contrast is a useful lens for enterprise observability and AI explainability. A system can be "cephalopod-like", emitting vast telemetry and verbose explanations that nobody interprets, or "baboon-like", emitting a few terse signals that a well-trained consumer (an analyst, a monitoring rule) turns into a rich narrative. Value lies in matching production to interpretation: in a bank's model-risk setting, a short, well-defined signal that reviewers reliably act on beats a torrent of unread feature attributions.
Our Minds and Others
Turning to the human case, Godfrey-Smith argues that inner speech is a real tool of thought, though not its necessary medium. He proposes that it grew out of efference copies, the internal predictions of our own actions' sensory effects, and that it broadcasts ideas through the brain, supports System 2 thinking and higher-order thought, and so reshapes human consciousness. Language's internalisation is the second great internalisation in cognitive evolution, after the turning-inward of cell signalling that made nervous systems; cephalopods, whose skin displays they cannot perceive, are on a different road.
From Hume to Vygotsky
In 1739 David Hume looked inside for an enduring self and found only perceptions ("heat or cold, light or shade, love or hatred, pain or pleasure"): a person is a bundle of perceptions in "perpetual flux and movement". Anyone can repeat the exercise, and Godfrey-Smith thinks Hume missed two things:
- Integration. We find not a succession of sensations but combinations present at each moment: an integrated scene of sights, sounds and body sense, which passes into the next.
- Inner speech. Most people find a flow of inner monologue: sentences, exclamations, rambling commentaries, speeches we would like to give or wish we had.
Perhaps Hume's inner speech was weak, but more likely he lumped it into the wash of sensations, which suited his agenda. Inspired by Newton, he wanted a quasi-physics of the mind in which ideas were mental atoms bound by a "power of attraction" (association) analogous to gravity; the peculiar properties of inner speech were irrelevant to that. Nearly two centuries later John Dewey retorted that Hume's flux of "ideas" was likely "a succession of words silently uttered".
- David Hume (1739): the self as a bundle of perceptions; mind as Newtonian association.
- John Dewey: American philosopher; saw Hume's ideas as silent words; treated inner speech as mostly recreational storytelling, perhaps because he was so intensely social a thinker.
- Lev Vygotsky (died 1934, aged 37): Soviet psychologist from what is now Belarus, son of a banker; worked with the Bolsheviks after 1917; theorised that thought develops through the internalisation of speech.
- Michael Tomasello: a prominent present-day psychologist who acknowledges Vygotsky's influence.
Vygotsky held that ordinary speech organises our lives: it helps put ideas together, direct attention and order actions. As children acquire language, it "branches" into inner and outer forms; inner speech is not merely silent outer speech but has its own patterns and rhythms, and it makes organised thought possible. Around 1930 he went through a personal and intellectual crisis, revised his theories, and faced dangerous charges of "bourgeois" elements. His Thought and Language reached English only in 1962; he remains somewhat fringe, yet his picture is increasingly central to understanding human versus animal minds.
Is Language Needed for Complex Thought?
Charles Darwin, in The Descent of Man (1871), claimed that a long train of thought can no more proceed without words, spoken or silent, than a long calculation without figures or algebra. That seems inevitable, and it is false.
| Evidence | Who | What it shows |
|---|---|---|
| Baboon narratives | Cheney and Seyfarth | Only 3 to 4 calls, yet listeners build understandings of rank reversals far richer than anything a baboon could say |
| Food-caching birds | Nicola Clayton and others, Cambridge | Jays store different foods in hundreds of sites and remember what is where, retrieving perishable items first |
| Insight in chimps | Wolfgang Köhler, four years on Tenerife around World War I, nine chimpanzees | Spontaneous solutions to novel problems, famously stacking boxes to reach hanging food; this shook even Vygotsky's view |
| Deaf people without sign language in preliterate cultures | Merlin Donald, Origins of the Modern Mind (1991) | Lived more normal lives than a language-is-essential view predicts |
| "Brother John" | André Roch Lecours and Yves Joanette (1980); used by Donald | During attacks of total aphasia (no speech, comprehension or inner speech), he stayed conscious and present; once, arriving by train, he found a hotel and ordered food by gesture, pointing at an illegible menu |
Extreme views on both sides are fading. Language is an important tool for thought and inner speech is not mere froth, but language is not the medium of complex thought nor essential for organising ideas. Hume's and Dewey's inventories of the mind are both incomplete.
Darwin's role for language survives in weaker form: language is a medium for arranging and manipulating ideas. In Susan Carey's Harvard lab, children appear able to use the disjunctive syllogism (A or B; not A; therefore B: "the sticker is under this cup or that cup; not this one; so...") only after they have learned the word "or". Cause and effect are hard to untangle, but it looks very Vygotskian.
Two strands of later work bear on this. In 2024, Evelina Fedorenko, Steven Piantadosi and Edward Gibson (Nature) reviewed brain-imaging and aphasia evidence and argued that language is primarily a tool for communication rather than for thought: the brain's language network is largely inactive during reasoning, arithmetic and planning, and people with severe aphasia can still reason. That supports Godfrey-Smith's rejection of Darwin's strong claim. On the other side, Johanne Nedergaard and Gary Lupyan (Psychological Science, 2024) studied people who report little or no inner speech (sometimes called anendophasia) and found they did worse on verbal working memory and rhyme judgements but not on most other tasks: inner speech helps with some things, as a Vygotskian tool view predicts, without being essential to thought in general. Russell Hurlburt's experience-sampling studies had already shown that inner speech is far less constant for many people than introspection suggests.
Word Made Flesh: How Inner Speech Works
Godfrey-Smith offers a plausible, admittedly conjectural model built on the efference copy.
Efference Copies
- In vision. When you move your eyes, the retinal image shifts but the world does not seem to move. As you send a motor command, you also send a faint copy of it to the visual system, which subtracts the expected effect of your own movement, so real external motion stands out.
- In action control. Copies also tell the brain "here is how things should look, given what I just did". A mismatch might mean the world changed or that your action failed: if pushing a table feels wrong, perhaps it is on rollers, or perhaps you did not push it at all.
- In speech. Speech is a complex action we want to come out right, so we form an inner image of the intended words and compare it with what we hear ourselves say. Ordinary speech carries a hidden quasi-saying and quasi-hearing.
From there it is a short step to sentences we never intend to say: a purely internal medium in which we can formulate sentences and experience their results. Hearing how words hang together tells us how ideas hang together; we can order, list, instruct and exhort ourselves.
What Inner Speech Does
- Executive control (Vygotsky). It sequences actions ("turn off the power first, then unplug the machine") and exerts top-down control over habits and whims ("don't eat another slice").
- Experiment. It combines ideas to see what follows ("how would things look if I could travel as fast as light?").
- System 2 thinking. In Daniel Kahneman's terms, it is a means for slow, deliberate, rule-following thought used in novel situations, as opposed to fast, habitual System 1: "ponderous but powerful", how we resist temptation and check whether a new action will work.
- The Joycean machine. Daniel Dennett, nodding to James Joyce's careening monologues, called the result "a Joycean machine" in our heads.
Why should leftover bits of language have such power? Part of the answer is that inner sentences are made available to much of the brain in the same way heard speech is, so a novel combination or exhortation gets the same processing as a spoken sentence.
"White Christmas" (2001). Participants listened to featureless random noise through headphones and were told the song "White Christmas" might be played very quietly within it; they pressed a button when sure they heard it. About a third pressed at least once. The song was never played. The standard reading: they imagined the tune and mistook their own auditory image for real sound. Self-generated sounds are broadcast in the mind much like perceptions. Such findings inform attempts to explain voice-hearing in schizophrenia, where inner speech may be misattributed to an outside source.
- Phonological loop. Plays imagined sounds, including inner speech.
- Visuospatial sketchpad. Manipulates inner pictures and shapes; perhaps rooted in efference copies from hand motions and gestures.
- Central executive. Orchestrates the other two subsystems.
- Working memory is a short-term store for a few items held or worked on, usually consciously. (Baddeley later added a fourth component, the episodic buffer, in 2000.)
Conjecture flagged by the author. That inner speech and imagery originated in efference-copy mechanisms has been hypothesised, not demonstrated. They might instead arise from the imagination itself and only coincidentally resemble products of action-control machinery.
Conscious Experience: Workspace or Broadcast?
Inner speech is as vivid as anything in inner life: you can sit motionless before an unchanging scene with a mind teeming with chatter, so much that people meditate to escape it. In the two-part framework of chapter 4 (basic experience such as pain from widespread animal features; conscious experience as a later, more sophisticated form), inner speech helps fill in the second part.
- Baars (late 1980s). Most brain activity is unconscious; a small fraction becomes conscious by entering a global workspace, like centre stage. The spatial picture invited the jibe "Is there a little man living there?", but Baars was onto something and research bore it out. His starting point: human experience is integrated into a scene.
- Dehaene and Naccache (2001). A second-generation version: conscious thought serves novel, deliberate actions that require taking the "big picture" into account, kicking in when habits break down.
The workspace metaphor. Consciousness is information arriving in a special place. This invites the questions "where is the inner space?" and "who looks at it?"
The broadcast model. Consciousness is information being broadcast throughout the brain. Godfrey-Smith argues this should replace the workspace idea, not restate it, and that "broadcast" is barely a metaphor. Those awkward questions dissolve.
Inner speech is one way broadcasting is done. It does not live "in a little box"; it is a loop intertwining the construction of thoughts and their reception, and the format of language lets ideas be combined in organised structures. Dehaene and others trace other broadcast mechanisms unrelated to speech; inner speech is part of the story, not all of it.
Higher-Order Thought
Higher-order thought is thought about your own thoughts ("Why am I in such a bad mood?", "I hardly noticed that car").
- The extreme latecomer view. Some argue higher-order thought is necessary for any experience at all; since most animals lack it, experience becomes nearly uniquely human.
- Godfrey-Smith's view. Higher-order thought is one feature that reshaped experience in us, not what brought it into being, though perhaps an especially important one.
- Inner speech's role. The most vivid conscious moments involve attending to our own thinking and experiencing it as ours, and in the "Why did I think that?" cases inner speech is prominent. These self-directed questions and exhortations are not idle: they let us do things we otherwise could not.
Full Circle: Two Internalisations
Human language may be half a million years old, perhaps less, and its evolution from simpler communication is debated. Its internalisation as a tool of thought, Vygotsky's transition, was itself a major evolutionary event.
| First internalisation | Second internalisation | |
|---|---|---|
| When | Hundreds of millions of years ago, near the start of animal evolution | Recent, within human evolution |
| What was internalised | Cell-to-cell sensing and signalling | Language |
| Result | Multicellular bodies and the nervous system | Inner speech as a tool of thought |
| How | Enlarging the organism to include formerly independent cells | Same boundaries, but a new path within them |
In both internalisations, a means of communication between organisms became a means of communication within them.
These two events bookmark cognitive evolution so far: one near the outset, one recent (not near the "end" of the process, just near the end of it as it has run so far).
Reafference: From Problem to Opportunity
In 1950 the German physiologists Erich von Holst and Horst Mittelstaedt introduced a vocabulary for action-perception loops:
- Afference. Everything taken in through the senses.
- Exafference. Incoming signals due to changes in the world ("ex" for outside).
- Reafference. Incoming signals due to your own actions; it makes perception ambiguous.
- Efference copy. The internal copy of an outgoing command that lets the system say "Don't worry, that's just me" and discount reafference.
- The problem. If actions did not change what you sense, perception would be easier. Efference copies filter out self-caused input.
- The opportunity. You can use your actions to feed future perception. Writing a note to yourself creates a reafferent loop where you want the later input to be entirely due to your earlier act, not meddling or decay. This makes external memory, almost certainly the role of much early writing (records of goods and transactions).
- Memory as communication. A note-to-self is communication between your present self and a future self, embedded in a sender-receiver system like any other.
The note-to-self also links the two roles of communication from chapter 2. It is like Revere's lantern (a mark made to be sensed later), and like calling the stroke (the present self coordinating a later self's action to complete a task), with the coordination running through the external world.
Inner speech is the same trick pulled inside the skin: copies of things you might say become silent actions of their own, raising possibilities, combining ideas and exerting self-control. It resembles talking aloud to yourself, "but this time the loop is tighter and more confined, invisible rather than public, a field for free and silent experiment."
Cephalopods on a Different Road
Cephalopods' expressive medium, colour and pattern, does not lend itself to such loops. A display is a one-way street: the animal cannot see its own patterns the way a person hears their own speech (setting aside the colour-blindness irony), so there is probably little role for skin-pattern efference copies, unless speculative skin-sensing theories hold. For a single animal these displays are not embedded in much looping feedback, "and perhaps could never be". If reafferent opportunities drive the evolution of more complicated minds, as the extreme human case suggests, cephalopods are on a different road, and that is not the only limit on their possibilities.
The silent orator. Imagine a person born able to speak fluently but with no hearing of their own voice and no inner voice: words go out, nothing comes back. They could still communicate, but they could not rehearse an argument, catch their own slip, or talk themselves out of a second slice. Godfrey-Smith's cuttlefish is this person in colour: enormous expressive output, no loop. The thought experiment isolates the claim that what matters for a reflective mind is not the richness of the output channel but whether output loops back as input.
This chapter is a surprisingly good map of reasoning in large language models. Chain-of-thought prompting (popularised in 2022) and the "reasoning models" released from 2024 onward work by having a model write intermediate steps and then read them back: a reafferent loop, a note-to-self, inner speech made explicit. The Vygotskian lesson cuts both ways. The loop gives real gains on multi-step, novel (System 2) problems, as Carey's "or" result would predict; but language is not the whole of thought, and a model's written rationale may not faithfully report the computation behind its answer. For banking use, treat visible reasoning traces as useful but not authoritative evidence in model validation.
Experience Compressed
Octopuses and giant cuttlefish, large-brained and complex, typically live one or two years. Godfrey-Smith resolves the paradox with the evolutionary theory of ageing (Medawar, Williams, Hamilton): when outside dangers make a long life unlikely, selection barely resists late-acting harm and actively favours early benefits bought with later breakdown. Losing the shell gave cephalopods both their unbounded, neuron-rich bodies and a vulnerability to predators that tuned their lives short: a rich experience, compressed.
Decline
When Godfrey-Smith started following cephalopods around 2008, he was shocked to learn how short their lives are:
- 1 to 2 years: typical lifespan of a giant cuttlefish or most octopuses.
- ~4 years: the outside limit for the giant Pacific octopus.
- 20+ years: lifespan of the nautilus, a cephalopod "smell-and-grope scavenger".
- ~200 years: a rockfish relative of a fish living in the same Sydney waters.
- 15 years: a mouse-sized monkey; 10+ years for some hummingbirds.
- Thousands of years: bristlecone pines in California and Nevada, alive since Julius Caesar's day.
He had assumed the two- and three-foot cuttlefish he met were old and worldly, having seen many seasons and many humans. In fact he had met them early in the breeding season, and all would soon die. Toward the end of that southern winter they entered a sudden decline, visible over weeks or even days:
- Falling apart. They lost arms and clumps of flesh.
- Losing the screen. The outer skin, "the living video screen", peeled away to leave plain white flesh (at first he mistook the patches for displays).
- Clouding eyes and, at the end, an inability to control their height in the water.
- Speed. Once decline starts, health "drops off a cliff".
This sharpened the puzzle of the large brain. Nervous systems are expensive to build and run, and the value of learning depends on having time to use what is learned. Cephalopods are evolution's only big-brain experiment outside the vertebrates, yet most vertebrates can live far longer (if not eaten). Two easy explanations fail:
- "The sea imposes short lives." No: a strange rock-dwelling fish in the same patch belongs to a group with members living two hundred years, "extraordinarily unfair" next to the cuttlefish.
- "The mollusk or cephalopod body makes short life inevitable." No: the nautilus lives more than twenty years.
Why Do We Age At All?
Death by accident or infection is no puzzle; death from "old age" is. The intuitive answer, that bodies wear out like cars, fails:
- A car with continually replaced parts need never stop running, and our cells are continually dividing and turning over their material.
- If cell lineages carried "age", bacteria could not exist: today's E. coli are fresh products of division, but their lineages are billions of years old.
- A clump of E. coli whose daughter cells stayed together could persist for millions of years as a kind of "body" made of brand-new cells.
- "It is our arrangement that breaks down" invites the question why new cells cannot regenerate the arrangement, as they do during development from embryo to adult.
- Wear-and-tear predicts that faster metabolism means faster ageing; this has some predictive power but fails often: marsupials (kangaroos) have lower metabolic rates than placental mammals yet age faster; bats have furious metabolisms yet age slowly.
At the level of cells, indefinite renewal is possible. What makes animals age is something about the kind of collection of cells we are: in animals, birth and death mark out an individual life, even as cells come and go around those boundaries.
The "good of the species" fallacy. It is tempting to think ageing benefits the species by clearing resources for the young. This begs the question (it assumes the young are more vigorous, which is what needs explaining) and is evolutionarily unstable: a mutant who refuses to "pass the baton" leaves extra offspring, so the refusal spreads and self-sacrifice collapses. Godfrey-Smith notes this is not the final word on hidden-benefit views, but modern theory takes another path.
A Swarm of Motorcycles: The Evolutionary Theory of Ageing
- 1940s Peter Medawar, British immunologist, gives a brief verbal argument: mutation accumulation.
- 1957 George Williams, American evolutionary biologist, adds trade-offs between early and late effects (antagonistic pleiotropy).
- 1960s William Hamilton, "probably the genius of late-twentieth-century evolutionary biology", puts the theory into rigorous mathematical form.
Step 1: Medawar's Mutation Accumulation
The ageless population. Imagine animals with no senescence: they reproduce from early life until killed by something external (predator, famine, lightning), with a constant 5% chance of death per year regardless of age. Survival falls geometrically: about 60% reach 10, 36% reach 20, 13% reach 40, 5% reach 60, and under 1% reach 90. Yet a ninety-year-old is just as likely as a youngster to see next year. Now ask how hard selection can push against a gene that harms only ninety-year-olds: almost no one carrying it ever meets its effect.
- Mutation-selection balance. Most mutations are harmful or neutral; harmful ones are weeded out, but new ones keep arriving, so every population carries some harmful variants of each gene, with removal balancing input.
- Late-acting mutations slip through. Many mutations act at particular life stages. One that acts only after many years harms almost nobody, since other causes of death get most carriers first. Selection against it is very slight, so it is cleaned out less efficiently than early-acting ones.
- Drift does the rest. Such mutations rise or fall mostly by chance, so some become common and everyone carries some.
- The illusion of a programme. A lucky long-lived individual finds things going wrong on a schedule as these lurking effects kick in. It looks "programmed to decline"; in fact the population has begun to evolve ageing.
Step 2: Williams's Trade-Offs
Saving for retirement: is it worth saving enough to live in luxury at 120? Only with unlimited income. Otherwise every dollar saved for an unlikely future is a dollar not used now. Likewise for genes with two effects at different ages:
| Early effect | Late effect | Fate under selection |
|---|---|---|
| Bad | Bad | Weeded out by its early harm |
| Good | Good | Favoured |
| Bad | Good | Weeded out: the late gain rarely materialises |
| Good | Bad | Favoured and accumulates: the early gain counts, the late cost is rarely paid |
Once many good-now, bad-later genes are common, decay appears as if scheduled, differing a bit between individuals, "not because of some hidden evolutionary benefit of the breakdown itself, but because the breakdown is the cost paid for earlier gains."
Step 3: Positive Feedback
The two effects are compatible and mutually reinforcing. Once age-related decay is established, even fewer individuals survive past the ages at which those mutations act, which further weakens selection against harm at those ages. "Once the wheel starts turning, it turns more and more quickly."
Ageing is not wear and tear and not a benefit to the species: it is what natural selection leaves behind when outside dangers make a long life unlikely, so that late-life harms go unpunished and early-life advantages are bought on credit.
Why Trees Escape
Bristlecone pines show no sign of falling apart. Trees are special in two ways:
- Fertility grows with age. The argument assumed late-life reproductive differences barely matter because few survive. But every branch is a site of reproduction, so a very old tree with many branches is far more fertile than a young one; the few survivors matter enormously.
- Trees are partly colonies. Some sea anemones form colonies of semi-independent polyps that bud and make their own sex cells; such colonies can persist indefinitely, like a human society whose members come and go. Colonies and societies do not reproduce in the right way for the Medawar-Williams logic; their members age instead. A pine or oak sits between an organism and a colony: it grows by multiplying branching units that can each reproduce (a cutting can become a new tree). Anything built that way is exempt.
Hamilton
William Hamilton, who formalised the theory, loved insects whose lives make ours look humdrum: mites whose females hang in the air with bodies packed with hatched young while brothers mate with sisters inside the mother; tiny beetles whose males make sperm longer than their bodies. He died in 2000 of malaria caught on a trip to Africa to investigate the origins of HIV. About a decade earlier he had asked to be laid out in the Brazilian forest to be eaten from within by the larvae of an enormous Coprophanaeus beetle.
I will be many, buzz even as a swarm of motorbikes, be borne, body by flying body out into the Brazilian wilderness beneath the stars.
Long and Short Lives
The thought experiment assumed lifelong reproduction. Many animals, cephalopods included, do otherwise.
Semelparous ("big bang") reproduction. Reproduce once, or in one short season, then die. Female octopuses are extreme: she may mate with many males, then settles permanently into a den, lays a clutch of many thousands of eggs, fans and tends them for one to several months (slower in cold water), and dies soon after the larvae drift off. Cuttlefish are "big bang" too but looser: active for one breeding season, both sexes mate many times, females glue batches of eggs to rocks and leave, then rapidly fall apart.
Iteroparous reproduction. Reproduce many times over an extended period: humans, most mammals, many fish. Favoured when adults are relatively safe once past a risky juvenile stage.
- The exception. A rare Panamanian octopus found by Moynihan and Rodaniche, whose females reproduce over a longer period. No one knows why.
- Why go for broke. If adult life is very risky, saving reserves for a next season is wasted on animals that rarely reach it. Selection favours whatever wins now, even at the cost of breakdown afterward: Williams's logic applied to breeding seasons.
- Three outcomes. The rockfish, spiny and venomous so "no one tries to eat" it, matures slowly and goes on for centuries; humans are intermediate; the cuttlefish races to size and fertility, mates in one season and falls to pieces.
Lifespans are set by the risk of death from external causes, how quickly reproductive age is reached, and other features of lifestyle and environment.
The octopus "self-destruct" switch has been traced to a gland. Back in 1977, Jerome Wodinsky showed that removing the optic glands (an octopus analogue of the pituitary) from brooding female octopuses made them stop brooding, resume feeding and live for months longer. In 2022, Z. Yan Wang and colleagues (Current Biology) found that after mating the optic gland of the California two-spot octopus sharply shifts its cholesterol metabolism, boosting production of steroid precursors including 7-dehydrocholesterol, a compound that is toxic in humans; they proposed this as part of the mechanism of maternal decline. This is a proximate mechanism; the Medawar-Williams account remains the evolutionary explanation of why such a programme is tolerated. Also since publication: the larger Pacific striped octopus (the Panamanian exception) was described in detail in 2015 by Roy Caldwell, Richard Ross, Arcadio Rodaniche and Christine Huffard, documenting beak-to-beak mating, shared dens and females that spawn repeatedly over a long period.
Why Cephalopods Have Big Brains and Short Lives
- Open-ended bodies. Losing the shell produced bodies of unbounded possibility, with the octopus as the extreme: neurons spread through the body in place of bones.
- Feedback with the nervous system. The body's possibilities created an opportunity for finer behavioural control; a larger nervous system then made it worthwhile to expand the body's possibilities further (arm sensors, colour machinery, a skin that can see). The shell loss did not by itself create the pressure; it started a feedback loop.
- Vulnerability. Without a shell, cephalopods became easy prey for fast fish with bones, teeth and good vision, which put a premium on camouflage and wiles.
- Limits of trickery. Tricks only save an animal so many times, and octopuses must hunt actively, exposing themselves in the open.
- Conclusion. High, constant external mortality makes them ideal targets for the Medawar and Williams effects, which compress the natural lifespan. Big nervous system (from unbounded bodies and the need to hunt while hunted) plus short life (from vulnerability): the paradox dissolves.
The deep-sea brooding record (MBARI, 2007 to ~2011). The Monterey Bay Aquarium Research Institute used remotely operated submarines to survey a rocky outcrop nearly a mile deep off central California. In 2007 they saw a deep-sea octopus, Graneledone boreopacifica; a month later she was guarding a clutch of eggs, and she was still there every visit for four and a half years. Her 53-month brood is the longest egg-brooding period reported for any animal (no fish is known to guard eggs beyond four or five months), longer than most octopuses live in total. Bruce Robison and colleagues estimate the species might live about 16 years if brooding takes the usual fraction of life.
- No physiological barrier. This octopus kills the idea that octopus bodies cannot live long.
- Robison's explanation. Cold deep water slows biological processes (letting the mother survive apparently without feeding), and long brooding lets young hatch large and advanced, a competitive edge.
- Godfrey-Smith's addition. Medawar-Williams predicts lower predation risk for a long-lived species, and there is a clue: she brooded out in the open for years without a den. Shallow-water octopuses never do this; they would be "sitting ducks". In the deep, fish are much rarer, so evolution tuned this species' lifespan differently.
Abandoning the shell set cephalopods on a path of mobility, dexterity and nervous complexity, and also onto a live-fast-die-young existence, always exposed to sharp-toothed predators.
Ghosts
The last days of the giant cuttlefish (Sydney). Diving among boulders, Godfrey-Smith swam into an ink cloud the size of a large room. Next day: dozens of cuttlefish eggs strewn on sand at the bottom of crevices (giant cuttlefish normally hang eggs from the roof of a ledge "like tulip bulbs"), and four decaying giant cuttlefish clustered under a Stonehenge-like rock arch, mostly white, skin cracked "like broken glass", eyes failing. A fifth, with yellow left in her skin but five arms largely gone, swam off. They ignored the eggs and seemed to be waiting: "cephalopod ghosts". Days later he watched one female's end: two arms gone, a feeding tentacle hanging motionless, rising slowly from the crevice. A leatherjacket, much smaller than her, began biting at her front; she flinched, then inked, then inked profusely and began a slow spiral, "like a lumbering airplane on fire" that rose rather than fell. His own attempts to fend off the fish panicked her more than the fish did. The attacks stopped; she rose through the last metre and floated motionless on the surface, sloshed by small waves.
The death is a transition from swimming deep in her quiet world, through a slow spiralling ascent, to drifting on the noisy surface of ours.
Lifespan extremes since 2016. In 2016, Julius Nielsen and colleagues (Science) used radiocarbon dating of eye-lens tissue to estimate that Greenland sharks live at least 272 years and probably close to 400, the longest known vertebrate lifespan, a striking case of a cold, slow, rarely preyed-upon animal fitting the same pattern as the deep-sea octopus. The basic Medawar-Williams-Hamilton framework remains the textbook explanation of ageing, now joined by much work on the molecular hallmarks of ageing.
The Medawar-Williams logic applies well beyond biology. In any system where most components are retired early for external reasons, there is little pressure to fix problems that appear only late: technical debt in short-lived projects, or failure modes in models that are expected to be replaced before the problems show. Banks run some systems for decades, far beyond their designers' "expected lifespan", and then meet exactly the late-acting defects nobody was selected to remove. The lesson: if you intend a long life, design and test for late-life failure explicitly, because nothing in the normal development cycle will.
Octopolis
At Octopolis, a shell bed fifty feet down off eastern Australia, normally solitary gloomy octopuses live at high density, signal with colour and posture, police territory and throw debris. Godfrey-Smith explains the site as ecosystem engineering seeded by a single dropped object and argues it offers a glimpse of one possible road for octopus evolution. He then zooms out: genetic dating shows that large brains evolved at least twice within the cephalopods, a parallel to mammals and birds, and the book closes on the ocean that made every mind, now under visible and invisible stress.
An Armful of Octopuses
Octopolis is about 50 feet down: "Oz-like emerald green" on clear days, "grey soup" on others. Since Matt Lawrence found it in 2009, octopuses have always been present; on intense days more than a dozen roam, grapple or sit in an arena a few yards across. Clumps of octopuses had been reported before, but this was the first site visitable year after year with several animals always present and often interacting.
- No harem. A single octopus sometimes seems to command the site, but only partially; there are too many to manage. Multiple males and multiple females are often present.
- Sexing without touching. The team never pulls octopuses from dens or turns them over. Males carry a groove under the third right arm (the hectocotylus), which they stretch toward a female, close up or from "cautious long range", to pass a sperm packet; females often store sperm before fertilising eggs. Watching who makes these arm stretches is the only way to sex animals.
- Unmanned cameras. GoPro cameras on tripods filmed the site with no divers present. Behaviour was much the same, with slightly more roaming and interaction when humans were away: no secret group acrobatics, but reassurance that divers do not much disturb them.
- Matthew Lawrence: scuba diver who discovered Octopolis in 2009; the octopuses seem to treat him differently from anyone else.
- David Scheel: biologist in Alaska, trained on lions in Africa (following prides by Land Rover), now an expert on the giant Pacific octopus (100 pounds or more); tracks octopuses with transmitters; brought systematic measurement and video analysis.
- Stefan Linquist: philosopher and former octopus lab researcher; joined the January 2015 intensive filming; reinterpreted "boxing" as "high fives".
- Christine Huffard: biologist who first flagged Matt's photos to Godfrey-Smith.
Cameras have enemies: octopuses inspect and attack them (perhaps mistaking them for static, three-legged cephalopods), leaving footage of suckers and bites; enormous stingrays sweep in and knock everything over.
January 2015: The Nosferatu Male
In the southern summer of 2015 the team parked a larger boat near the site and tried to cover nearly every daylight hour on video. Activity was unprecedented and earlier glimpses resolved into patterns.
- Policing. One large male controlled access through the day, chasing off some octopuses and fighting if they did not retreat, tolerating others (probably females) and sometimes herding them back into dens.
- Boxing or high fives. Roamers and denned animals probe and lash with arms. Godfrey-Smith's first paper called this "boxing"; Linquist saw many contacts as "high fives", arm-slaps aiding recognition of individuals or roles. Sometimes they settle; sometimes a fight follows.
- The Nosferatu pose. Before attacking, an aggressive male often turns dark, rises, spreads his arms to look bigger, and may raise his mantle over his head, named after the silent-film vampire's dark cloak. The 2015 male used it often; he was not always larger than his opponents but lost only once on film.
Colour as Signal
David Scheel went back through the archive, charting hundreds of aggressor-target encounters.
| Display | When it appears | Interpretation |
|---|---|---|
| Dark skin | Before advancing or standing ground | Reliable predictor of aggression; conveys seriousness |
| Nosferatu pose (dark, mantle raised, arms spread) | Bearing down on a rival | Escalated threat |
| Bland pale grey | Unwilling to fight | Non-aggression |
| Stark blotchy (deimatic) pattern | Fleeing, or returning to a den under an aggressor's watch | Normally a startle display against predators; here apparently co-opted as submission or non-aggression, since there is no flight to startle |
| Unorganised colour runs | An octopus alone, filmed by an unmanned camera | Fits the "chatter" hypothesis |
The colour findings were published, and a second city was found. Scheel, Godfrey-Smith and Lawrence reported in Current Biology (2016) that darker body colour and "stand tall" postures signal aggression, and that when a dark approacher met a dark opponent fights were more likely, while pale animals tended to retreat. In 2017 the team (led by Scheel, with Stephanie Chancellor, Lawrence, Linquist, Godfrey-Smith and others) described a second high-density site in the same bay, nicknamed "Octlantis", built around natural rocky outcrops rather than a dropped object, with up to about fifteen gloomy octopuses denning together, mating, chasing and evicting each other. That a second site exists strengthens the reading that Octopus tetricus forms such aggregations wherever food is plentiful and safe dens are scarce.
Origins of Octopolis
The Panama Precedent
- 1982 Moynihan and Rodaniche, during their reef squid study, report an undescribed striped octopus in Panama living in groups of several dozen, sometimes sharing dens. No photos or video; a fuller description is rejected; sceptics doubt the report for years.
- 2012 The animal (the larger Pacific striped octopus) appears in the aquarium trade; Richard Ross and Roy Caldwell at the Steinhart Aquarium in California keep live specimens.
- Captivity confirms that they tolerate each other, share dens, and that females mate and lay eggs over an extended period, unlike the usual brood-once-and-die pattern. A Nicaraguan collecting company knows of one site where they aggregate; a field study was being prepared.
The normal octopus pattern is very different: an individual makes a den, stays perhaps a few weeks, moves on; males mate (often at arm's length) and leave; adults barely interact. Even Octopus tetricus appears much less social elsewhere.
How a Shell City Grows
Godfrey-Smith's partly speculative account:
- The seed. A single metal object, about a foot long and high, was dropped on the sand, probably from a boat; now completely overgrown. It is prime real estate: the largest octopus usually lives under it (sometimes alongside a few fish it "pretends not to notice"). Like a speck that seeds a crystal, it seeded the site.
- First dens. One or a few octopuses denned there and brought scallops home to eat.
- Shells change the ground. Discarded scallop shells, discs a few inches across, are far better den material than fine sand, so more dens became possible nearby.
- Positive feedback. More octopuses brought more scallops, leaving more shells, allowing more dens.
- Possible kick-start. The object might have arrived with a load of shells, before scallop dredging was banned in 1984 or diver collection in 1990. Since then, most shells were probably carried in by the octopuses: by hunting and bringing food home, they transformed their own habitat.
Why Here?
- Unlimited food. The area is a scallop bed; scallops sit singly or in clumps and are easy prey. They do have eyes (with a mirror behind the retina) and can swim by clapping their shells ("swimming castanets!"), but are helpless against octopuses.
- Few safe dens. The fine sand is hard to dig a stable hole in.
- Many deadly predators. Dolphins and seals zoom in to probe dens; several shark species, including huge carpet sharks like "old bomber planes", lie on the site for long periods; car-wide stingrays sweep through and the octopuses duck.
- Leatherjackets. Piranha-like fish gathering in hundreds. Matt filmed a school that, after cautious feints, tore an octopus to pieces within a couple of minutes as it fled toward the surface. Godfrey-Smith's guess: an octopus has a safe radius from its den within which it can get home before damage is done.
Octopolis is an island of safety in a dangerous area. Why don't octopuses eat each other there? Sizes range from matchbox to over a metre in arm span, and many octopuses, including close relatives, are cannibals. Probably because food that doesn't fight back (scallops) is so abundant.
Ecosystem engineering. The team's second paper framed Octopolis as the reshaping of an environment by its inhabitants' behaviour. Shell dens can be at least 40 cm deep; some octopuses seem to live entirely buried in shells, perhaps interacting or mating beneath the surface, and their environment increasingly consists of their own debris.
- An artificial reef. Fish schools hover overhead, squid gather and signal, carpet sharks ambush fish from the site, baby sharks of another species rest there seasonally, and banjo rays sit with hermit crabs crawling over them, all at much higher densities than nearby.
What Octopolis Means
Reading 1: octopuses are more social than we thought. Their colour changes and displays suggest it. A 2011 study found a close relative of Octopus tetricus could recognise individual octopuses; a more controversial 1992 study suggested observational learning (octopuses learning by watching others).
Reading 2: the site is unusual. Intelligent animals placed in a novel, crowded context have improvised: some behaviours are opportunistic and new, worked out by individuals who "have had to work out how to get along".
Godfrey-Smith suspects a mix: long-standing behaviours plus individual improvisations. Either way, Octopolis supplies elements usually missing from octopus life that matter for the evolution of minds: constant interaction and social manoeuvring, much feedback between action and perception, an environment made complicated by other octopuses, and continual reshaping of the shell bed. Octopuses throw debris, shells and silt, which often hits other octopuses and changes their behaviour; the team was working out whether some throws are targeted.
The throws are sometimes aimed. In 2022 Godfrey-Smith, Scheel, Chancellor, Linquist and Lawrence (PLOS ONE, "In the line of fire") analysed footage from Octopolis and Octlantis. Octopuses gather shells, silt or algae in their arms and launch them with a blast from the siphon, which is repositioned under the arms for the purpose. Most throws are den-clearing, but some hit other octopuses, and those throws were more vigorous, more often made by females, more often accompanied by dark body colour, and in some cases aimed at an octopus that had been pestering the thrower (one female repeatedly hit a male who kept trying to mate with her); targets sometimes ducked or raised arms in anticipation. Throwing objects at other animals is rarely documented outside a few mammals and birds.
Could Octopolis Change Octopus Evolution?
With roughly two-year lives, several generations have passed through the site since 2009, each remaking the same semi-sociality with no parental care. Godfrey-Smith imagines extra steps: more complex interaction, refined signals, higher densities; lives increasingly entangled, showing in brain evolution; and, since lifespan is tuned by predation (chapter 7), longer lives if octopuses could reliably survive more years.
Not at this site. Godfrey-Smith is explicit: one small site is a tiny fraction of the species' range; hatchlings drift away as larvae, so today's residents are not descendants of earlier ones; "one site and a few years mean nothing in evolutionary terms". Such arrangements would need to persist at large scale for thousands of years. Octopolis is a glimpse of one possible road, not a road being travelled.
Octopolis scaled up. Suppose seabed-wide changes (say, millions of discarded shells from aquaculture) created shell beds across a whole coast for ten thousand years. Densities stay high; dens are scarce; neighbours are constant. Selection would now reward reading rivals' colours, remembering who beat whom, and tolerating close neighbours. With predation lower inside the beds, the Medawar-Williams pressure eases and longer lives could evolve, giving time for learning to pay off. The thought experiment shows how sociality, lifespan and brain size could ratchet together, the same coupled loop that may have shaped primates, starting from a pile of shells.
Parallel Lines: Minds Evolved More Than Twice
The book's arc so far: ancient sensing and acting; animals from single cells; internalised signalling as the first nervous systems; the bilaterian body plan; the fork between vertebrates and most invertebrates; the Ediacaran-to-Cambrian shift to entangled predator-prey lives; then two experiments in large nervous systems. But the Y-shape of chapter 3 was a simplification.
- On the vertebrate side. Large brains arose along several paths: mammals, and above all parrots and crows among birds, with a lizard-like common ancestor ~320 million years ago. Vertebrate brains are "variations on a theme", but the branchings are deep.
- On the cephalopod side. The question is how deep the octopus/cuttlefish split is.
The old view (fossil-based). The coleoids (octopuses, cuttlefish, squid) appeared about 170 million years ago, in the age of dinosaurs. Andrew Packard (1972) argued that they evolved in parallel with modern fish, which from about 170 million years ago competed with the old cephalopod predators; cephalopod complexity was a response. This suggested one burst of brain evolution, followed by minor variation, and encouraged an "accidental intelligence" reading: octopuses with "too much brain" for short, asocial lives.
The new view (genetics). The last common ancestor of octopuses, cuttlefish and squid lived about 270 million years ago, in the Permian, before the dinosaurs. That split produced the octopods (octopuses plus the deep-sea Vampyromorpha) and the decapods (squid and cuttlefish). Backdating by 100 million years makes it much more likely that cephalopods evolved complex nervous systems at least twice.
A sceptic can reply that the Permian common ancestor might already have been smart. Two lines of evidence pull against that:
- Protocadherins. The 2015 octopus genome showed that octopuses build nervous systems with protocadherins, the same cell-adhesion molecule family we use. That family expanded separately in octopuses and in squid, after their split: brain-building molecules diversified at least three times (once in vertebrates, twice in cephalopods).
- Cuttlefish memory. Christelle Jozet-Alves and her group in Normandy, with the bird researcher Nicola Clayton, tested "episodic-like" (what-where-when) memory in a small cuttlefish species.
- Episodic memory. Memory of particular events (your last birthday); called "episodic-like" in animals because its subjective vividness cannot be verified.
- Semantic memory. Memory of facts (where France is).
- Procedural memory. Memory of skills (how to swim).
What-where-when memory in cuttlefish. Each cuttlefish's preferred food was established (shrimp preferred over crab). Each food was linked to a distinct visual cue in the tank. Shrimp, after being eaten, was replenished only after three hours; crab after one hour. Released one hour after a shrimp meal, cuttlefish went to the crab location; after three hours, they went for shrimp. They remembered what they ate, where, and how long ago, matching the classic food-caching bird results.
Episodic-like memory in mammals, birds and cuttlefish is almost certainly parallel evolution, and the cuttlefish result shows complex cognition on the decapod branch, built in brains that evolved partly separately from octopus brains. Complex cephalopod nervous systems were no one-off accident kept with variations; the nervous system expanded in the octopus line and, in parallel, in the other cephalopods.
| Comparison | Split | Brains |
|---|---|---|
| Mammals vs birds | ~320 million years ago, from a lizard-like ancestor | Large brains evolved separately in different bodies on a shared vertebrate plan |
| Octopuses vs cuttlefish/squid | ~270 million years ago, in the Permian | Large brains evolved in parallel on a shared molluscan plan |
| Vertebrates vs cephalopods | ~600 million years ago | Independent experiments in mind |
Cephalopods and smart vertebrates are independent experiments in the evolution of the mind, and, like mammals and birds, octopuses and cuttlefish are sub-experiments within the larger one. Shell loss probably preceded the ~270-million-year split, and at least two lines then built large nervous systems separately.
Cuttlefish cognition keeps impressing. Jozet-Alves's what-where-when study appeared in Current Biology (2013). In 2021, Alexandra Schnell, Clayton and colleagues (Proceedings of the Royal Society B) showed that episodic-like memory in common cuttlefish does not decline with age, even in their final days, while other memory types do; the same group showed cuttlefish delay gratification for preferred food. On the cephalopod tree, genomic studies of squid and octopus in the 2020s have confirmed extensive RNA editing and repeated, lineage-specific expansions of neural gene families, consistent with the parallel-evolution picture. Recent molecular-clock estimates still place the octopod-decapod split in the late Paleozoic, broadly in line with the book's 270 million years, though exact dates vary between studies.
The Oceans
The mind evolved in the sea. The origin of life, the birth of animals, the first nervous systems and brains, and the complex bodies that make brains worth having all happened in water. Animals reached land by about 420 million years ago, perhaps earlier, and took the sea with them: every cell is a membrane-bound container of remnant seawater. Meeting an octopus is the closest thing to meeting an alien, "yet it's not really an alien; the Earth and its oceans made us both".
The sea's productive features exist at tiny scales, and the sea does not change visibly as we damage it: waste seems to dilute away, and protective measures achieve little that can be seen. Yet effects can be obvious once you look. Sydney's beaches were fished nearly empty by the new millennium; in 2002 one small bay became a marine sanctuary, and within a few years it was teeming, which is where Godfrey-Smith met the cephalopods behind this book.
Huxley's inexhaustible fisheries (1883). Thomas Huxley, Darwin's chief ally, asked whether North Sea fish stocks might run out, calculated sea productivity against catch and declared that the cod, herring and mackerel fisheries were "inexhaustible". Within decades, cod especially was in serious trouble, and Huxley became a villain of fisheries history. Critics often omit his qualifier, "in relation to our present modes of fishing". What doomed the forecast was technological change: mechanised gear, freezers and fish-tracking technology meant the "present modes" vanished soon after, and so did the fish.
- Overfishing began in the nineteenth century and continues, with shrinking returns; "the sea looks the same after the boats are gone".
- Chemical change is harder to see and more global. Acidification: rising atmospheric CO2 dissolves into seawater, pushing it from mild alkalinity, affecting the metabolism of many animals including cephalopods and softening and dissolving the calcium-based hard parts of corals and others.
- Dead zones: areas where loss of oxygen leaves almost nothing alive, now far larger than natural occurrences; some pulse seasonally with fertiliser runoff, others seem permanent. "Dead zone: the very opposite of an ocean."
The bee lesson. Over lunch with the bee biologist Andrew Barron (and the philosopher Colin Klein, discussing the origins of experience), Godfrey-Smith asked about colony collapse, apparent from about 2007. Barron's answer: no single cause. Over years, small stresses accumulated (pollutants, new microorganisms, lost habitat); colonies buffered them by working harder, invisibly wearing down their capacity until a critical point, then failed visibly. Farmers now truck healthy colonies thousands of miles between orchards. Godfrey-Smith now sees the ocean the same way: vast buffering capacity, slowly exhausted, failing suddenly.
Policy has moved partly in the book's direction. The Kunming-Montreal Global Biodiversity Framework (December 2022) set a target to protect 30% of land and sea by 2030, and the UN High Seas Treaty (on biodiversity beyond national jurisdiction) was adopted in 2023. Directly relevant to this book's animals: once cephalopod sentience was recognised (UK Animal Welfare (Sentience) Act 2022), proposals for the world's first commercial octopus farm in the Canary Islands drew strong scientific opposition, and Washington State (2024) and California (2024) banned octopus farming. Ocean acidification and deoxygenation have continued to worsen since 2016.
When you dive into the sea, you are diving into the origin of us all.
Two practitioner lessons. Huxley's error is a forecasting error banks know well: a projection stated "under present modes" silently assumes the technology stays fixed. Risk models for credit, fraud or operational resilience should state their technology assumptions explicitly and be re-baselined when the tools change (as generative AI is now changing fraud). The bee lesson describes operational resilience: systems absorb accumulating small stresses through extra effort (manual workarounds, overtime, retries) until buffering capacity is exhausted and failure is sudden. Monitoring buffer consumption, not just outages, is the early warning.
Master Takeaways
Taken together, Other Minds is a gradualist, evolutionary theory of experience tested against the strangest large-brained animals on Earth. Its lasting contributions are the action-shaping versus sensory-motor framing of early nervous systems, the sentience-before-consciousness (transformation) view, and a vivid, field-based picture of a distributed, curious, short-lived cephalopod mind that evolved independently of ours, and more than once. The book's deepest lesson is that minds are not a single late invention but a recurring outcome of animals that sense, act and must deal with one another, built at least twice in radically different bodies.
The takeaways
- Cephalopods are an independent experiment in mind. Our last common ancestor with an octopus was a tiny flattened worm about 600 million years ago, so cephalopod intelligence was built from scratch on a separate branch; what we share with them (camera eyes, curiosity, play, individual recognition) are candidates for general features of minds.
- Experience comes before consciousness. Sentience, the bare fact that life feels like something, is broader and older than consciousness, which is a later, more unified and integrated form. Pain and other primordial emotions probably felt like something long before workspaces and inner models existed.
- Workspaces and integration transformed experience; they did not create it. Against latecomer theories (Dehaene, Baars, Milner and Goodale, Prinz), the transformation view holds that late-evolving machinery reshaped a pre-existing subjective life. The best evidence is flexible, trade-off-sensitive pain behaviour in fish, chickens, hermit crabs and octopuses.
- Nervous systems may have begun as coordinators, not information processors. The action-shaping view (Pantin, Keijzer) says early nervous systems solved the problem of turning many micro-actions into one macro-action; the quiet Ediacaran suits it, while the Cambrian arms race made sensory-motor tracking of other animals central.
- The octopus body is pure possibility, and its mind is distributed. With almost no hard parts and most of its ~500 million neurons in its arms, the octopus combines central steering with local arm autonomy, so its arms are partly self and partly not: "a conductor leading jazz players". It sits outside the usual body/brain divide and fits standard embodied cognition poorly.
- Big brains and short lives have one cause: losing the shell. Shell loss produced unbounded bodies that rewarded large nervous systems and also exposed cephalopods to toothed fish; that vulnerability, via the Medawar-Williams logic, tuned their lives to one or two years.
- Sensing and signalling were turned inward twice. Cell-to-cell signalling became the nervous system; much later, language became inner speech. In both, communication between organisms became communication within them.
- From the Cambrian on, minds evolved in response to other minds. Predation turned each animal into a major part of others' environments, producing eyes, claws, antennae and a "Cambrian information revolution".
- Experience probably arose several times. Since vertebrates, arthropods and mollusks had already split in the Ediacaran, if cats, crabs and octopuses all feel, experience evolved at least three times, and consciousness likely more than once too.
- Sentience has a point. Experience seems to arise not from the mere running of a living system but from modulation that registers what matters and requires response; it is not "a bathing in living activity".
- Unity of experience is an achievement, not a given. Pigeons with no inter-ocular transfer, lateralised fish and tadpoles, hens that weave to slosh information between eye regions, and split-brain humans show that integration is optional and evolved.
- Action changes what you sense, and that matters for how things feel. Tactile vision substitution yields experience of distant objects only when the user controls the camera; animals must solve the self/other problem (efference copies, Merker's worm) because they move.
- Octopus intelligence is curious and individual, not social. Lab tests show slow learners; anecdotes show mischief and craft: squirting experimenters, short-circuiting lights, recognising keepers, flooding labs, coconut-shell houses. Arthropods went the social route; cephalopods took "lone idiosyncratic complexity".
- Much octopus cleverness may be surplus. Neurons may have multiplied first to control the body, with recognition of humans and other abilities arising as by-products, while problem solving was shaped by extractive, Gibson-style foraging.
- Cephalopod skin is a brain-wired screen that also leaks. Chromatophores, iridophores and leucophores produce camouflage and signals, but giant cuttlefish also show "chromatic chatter", inadvertent expression of inner activity.
- Colour-blind animals may sense colour through their skin. Octopus skin contains photoreceptors and responds to light on its own; chromatophores could act as changing filters that let a single photoreceptor type detect hue.
- Baboons and cephalopods are lopsided communicators. Baboons have simple production, complex interpretation; cephalopods have vast production, little interpretation. Each is "unfinished" in the sender-receiver duality.
- Inner speech is a tool, not the medium, of thought. Baboons, food-caching jays, Köhler's chimps, deaf people without sign language and Brother John show complex thought without language; yet children need the word "or" to use the disjunctive syllogism.
- Inner speech probably grew from efference copies and works by broadcast. A copy of intended speech used to check output became a silent loop whose sentences are processed like heard ones; "broadcast" should replace the "workspace" metaphor.
- Reafference begins as a problem and becomes an opportunity. Self-caused input must be cancelled, but deliberate loops, from notes to self (external memory, communication with your future self) to inner speech, become engines of thought. Cephalopods cannot see their own displays and lack such loops.
- Ageing is not wear and tear. Cells renew indefinitely; ageing evolves because high external mortality weakens selection late in life (Medawar's mutation accumulation) and favours early benefits with late costs (Williams's trade-offs), in a self-reinforcing loop formalised by Hamilton.
- Lifespan is tuned by danger. The rarely eaten rockfish lives ~200 years; trees escape because fertility rises with size and they grow from self-reproducing units; the deep-sea Graneledone octopus broods for 53 months in the open where predators are rare.
- Octopolis shows ecosystem engineering and latent sociality. One dropped object seeded a feedback loop of dens, scallop meals and shell debris, producing a dense, signalling octopus community: dark means aggression, pale means retreat.
- Large brains evolved in parallel even within cephalopods. Genetic dating puts the octopus-cuttlefish split at ~270 million years, separate protocadherin expansions occurred in each line, and cuttlefish show episodic-like memory, mirroring the mammal-bird case.
- The mind evolved in the sea, and the sea is under quiet strain. Overfishing, acidification and dead zones accumulate like the hidden stresses behind bee colony collapse; buffering capacity can be exhausted before failure becomes visible.
Frameworks & mental models
| Name | What it says | Use it when |
|---|---|---|
| Gradualism (James) | Consciousness should not appear as a sudden "irruption"; explain it through continuous, comprehensible transitions | Assessing any theory that posits a sharp on/off line for minds |
| Tree of life and the fork | Large brains evolved on separate branches; human-octopus ancestor ~600 mya, mammal-bird ~320 mya, octopus-cuttlefish ~270 mya | Judging whether a trait is shared by descent or convergent |
| From sensing to signalling | External sensitivity, sensitivity to others, sensitivity to signals made to be perceived, then signalling within an organism | Tracing the origins of nervous systems and communication |
| Sensory-motor vs action-shaping views | Nervous systems link perception to action (Revere's lantern) or coordinate parts into whole actions (calling the stroke); the cox does both | Analysing what any control system is really for |
| Complex active bodies (Trestman) | Only arthropods, chordates and cephalopods produced fast, manipulating, distance-sensing bodies | Explaining where complex behaviour is found |
| Brains as tool kits | No single intelligence scale; perception, memory, learning, planning combined to suit a way of life | Comparing cognition across species or systems |
| Specialised vs extractive foraging (Gibson) | Flexible extraction of protected food drives large brains | Explaining curiosity and problem solving in non-social animals |
| Coordination plus surplus | Neurons multiplied to control a boneless body; some abilities are by-products, some selected | Explaining "too much brain" for a short, asocial life |
| Octopus outside the body/brain divide | No fixed shape, so neither brain-as-CEO nor intelligence-in-the-body fits; the whole nervous system is the unit | Evaluating embodied cognition claims |
| Two causal arcs | Sensory-motor arc and motor-to-sensory arc; action shapes the next input and how it feels | Thinking about perception, TVSS, active sensing |
| Latecomer vs transformation views | Experience arrives with late machinery, or old experience is reshaped by it | Asking which animals (or systems) might feel anything |
| White noise to consciousness | Inchoate buzz, then old feelings tied to action, then unified consciousness, on several branches | Picturing what early or simple experience could be |
| Ventral/dorsal streams (Milner and Goodale) | Recognition stream supports visual experience; action-guidance stream works largely unconsciously | Interpreting dissociations such as DF or blindsight |
| Global workspace / broadcast (Baars, Dehaene) | Information becomes conscious when globally available; Godfrey-Smith prefers broadcast to "workspace" | Thinking about integration and conscious access |
| Skin layers | Neural chromatophores over chemically controlled iridophores and passive leucophores | Understanding cephalopod colour production |
| Functions of colour | Camouflage (original), signalling (co-opted), deimatic displays (intermediate), chatter (side effect) | Interpreting any animal display |
| Chromatophores as filters (Mäthger, Jost) | Known, changing coloured filters over one photoreceptor type can yield colour information | Explaining colour-matching in colour-blind animals |
| Production vs interpretation | Baboons: simple signals, rich interpretation; cephalopods: rich signals, little interpretation | Diagnosing communication systems that over-produce or under-produce |
| Efference copy and inner speech | Copies of planned speech become a silent, broadcast loop for thought | Explaining inner speech, verbal hallucination, self-talk |
| Working memory (Baddeley and Hitch) | Phonological loop, visuospatial sketchpad, central executive | Analysing deliberate, step-by-step thought |
| System 1 / System 2 (Kahneman) | Fast habitual thinking vs slow rule-following thinking; inner speech serves System 2 | Distinguishing intuitive and deliberate processing |
| Afference, exafference, reafference (von Holst and Mittelstaedt) | Input from the world vs input caused by your own action | Designing perception that must discount self-caused change |
| Two internalisations | Cell signalling became nervous systems; language became inner speech | Seeing the long arc of cognitive evolution |
| Mutation accumulation (Medawar) | Late-acting harmful mutations escape selection when few survive to old age | Explaining ageing and late-life failure |
| Antagonistic pleiotropy (Williams) | Genes with early benefits and late costs are favoured | Explaining trade-offs that produce scheduled decline |
| Semelparity vs iteroparity | Breed once and die vs breed repeatedly; set by adult mortality risk | Explaining cephalopod "big bang" reproduction |
| Ecosystem engineering (Octopolis) | Animals reshape their environment, which then reshapes their social life | Understanding self-reinforcing habitat change |
| Parallel evolution of intelligence | Large brains arose separately in mammals and birds, and in octopuses and decapods | Weighing whether intelligence is an accident or a recurring outcome |
Strengths. Godfrey-Smith combines first-hand field observation, careful reading of primary experiments (his rescue of Dews's "failed" 1959 study is exemplary) and philosophical precision. The action-shaping view, the sentience/consciousness distinction and the transformation view are clear and testable framings, and he is candid about speculation (efference-copy origins of inner speech, chromatophore colour sensing, the origins of Octopolis).
Blind spots and contested points. The transformation view rests heavily on an intuition about primordial emotions and on behavioural pain evidence that latecomer theorists can reinterpret; the book offers no positive mechanism for the earliest feeling beyond "modulation that matters". The white noise image is admittedly a metaphor. His dismissal of insect pain was stronger than later evidence supports. The cephalopod "accidental intelligence" debate is settled more by genetic dating than by cognition data on squid, which remain thin. The book mostly ignores integrated information theory and higher-order theories beyond a brief treatment, and says little about sleep, emotion research or the neuroanatomy of the octopus vertical lobe.
Dated or overtaken. Several open questions have moved since 2016: comb jellies are now the leading candidate for sister to all other animals (2023); Dickinsonia is confirmed as an animal (2018); octopuses show two-stage sleep with REM-like "active sleep" (2021, 2023); octopus affective pain has experimental support (2021); a second octopus city, Octlantis, was described (2017) and debris throwing shown to be sometimes targeted (2022); the 2021 LSE sentience review led to cephalopods and decapods being recognised in the UK Animal Welfare (Sentience) Act 2022; the 2024 New York Declaration called experience at least a realistic possibility in cephalopods and insects; and the 2023-2025 adversarial tests challenged workspace and IIT predictions alike. Colour vision in cephalopods remains unsolved, with chromatic aberration via odd pupils (2016) a further hypothesis. Overall the field has moved toward Godfrey-Smith's generous, gradualist position.
Who's who
| Name | Who | Why they matter here |
|---|---|---|
| Peter Godfrey-Smith | Philosopher of biology, diver | Author; field observer at Octopolis; transformation view |
| Matthew Lawrence | Scuba diver | Discovered Octopolis (2009); led by the hand by an octopus |
| David Scheel | Octopus biologist, Alaska | Giant Pacific octopus expert; Octopolis colour-signal analysis |
| Stefan Linquist | Philosopher | Octopus lab experience; "high fives"; captivity awareness |
| William James | Philosopher-psychologist (1890) | Demand for continuity in explaining consciousness |
| Thomas Nagel | Philosopher | "What is it like to be a bat?" |
| Reginald Sprigg | Geologist | Found the Ediacaran fossils (1946) |
| Jim Gehling | Paleontologist, South Australian Museum | Ediacaran guide; scavenging-then-predation scenario |
| Mark McMenamin | Paleontologist | Coined "Garden of Ediacara" |
| Graham Budd | Paleontologist | Animals as nutrient patches driving predation |
| Andrew Parker | Biologist | Eyes as the decisive Cambrian event |
| Roy Plotnick | Paleontologist | "Cambrian information revolution" |
| Chris Pantin / Fred Keijzer | Biologist (1950s) / philosopher | Action-shaping view of early nervous systems |
| Detlev Arendt | Biologist | Two origins of nervous systems in one animal |
| Michael Trestman | Philosopher | Complex active bodies |
| Peter Dews | Harvard pharmacologist | 1959 lever study; Charles the octopus |
| Roger Hanlon and John Messenger | Cephalopod biologists | Cephalopod Behaviour; arms "curiously divorced" from brain |
| Jean Boal | Cephalopod researcher | Rigorous sceptic; the squid-down-the-drain story |
| Jennifer Mather | Cephalopod researcher | Octopus play (with Roland Anderson); looping foraging routes |
| Binyamin Hochner, Tamar Gutnick, Michael Kuba | Hebrew University lab | 2011 eye-guided arm maze; embodiment view |
| Katherine Gibson | Primatologist | Extractive foraging and brain size |
| Hillel Chiel and Randall Beer | Embodied cognition researchers | Body as constraint and opportunity; "jazz" metaphor |
| Björn Merker | Neuroscientist | Self-produced touch problem (earthworm) |
| Giorgio Vallortigara | Neuroscientist, Trieste | Brain lateralisation in animals |
| Marian Dawkins | Animal behaviour researcher | Weaving gaze of hens |
| David Milner and Melvyn Goodale | Vision scientists | Patient DF; ventral and dorsal streams |
| David Ingle | Neuroscientist | Rewired frogs (1960s) |
| Stanislas Dehaene | Neuroscientist | Global neuronal workspace; consciousness and novelty |
| Bernard Baars | Neuroscientist | Global workspace theory (1980s) |
| Jesse Prinz | Philosopher | Working-memory theory of consciousness |
| Derek Denton | Physiologist | Primordial emotions |
| Robert Elwood | Biologist | Hermit crab shock trade-offs |
| Jean Alupay | Biologist | Octopus wound tending after arm injury |
| Simona Ginsburg and Eva Jablonka | Biologists | "White noise" origins of experience |
| Bence Nanay | Philosopher, Antwerp | Octopus arm control as an extreme of human fast correction |
| Alexandra Schnell | Cuttlefish researcher | Giant cuttlefish in captivity |
| Lydia Mäthger, Steven Roberts | Cephalopod biologists | Opsins in cuttlefish skin (2010, with Hanlon) |
| Desmond Ramirez and Todd Oakley | Biologists | Light-sensitive octopus skin (2015) |
| Lou Jost | Ecologist and artist | Chromatophores as changing colour filters |
| Dorothy Cheney and Robert Seyfarth | Primatologists | Baboon Metaphysics; playback experiments |
| Martin Moynihan and Arcadio Rodaniche | Field biologists, Panama | Squid "language"; social striped octopus |
| David Hume | Philosopher (1739) | Bundle theory; neglect of inner speech |
| John Dewey | Philosopher | Hume's ideas as silent words |
| Lev Vygotsky | Soviet psychologist | Internalisation of speech as tool of thought |
| Charles Darwin | Naturalist | Language needed for long trains of thought (1871) |
| Wolfgang Köhler | Psychologist | Chimp insight on Tenerife |
| Merlin Donald | Psychologist | Brother John and deaf-without-sign cases |
| Nicola Clayton | Comparative psychologist, Cambridge | Bird and cuttlefish episodic-like memory |
| Susan Carey | Developmental psychologist, Harvard | The word "or" and the disjunctive syllogism |
| Daniel Kahneman / Daniel Dennett | Psychologist / philosopher | System 1 and 2; the "Joycean machine" |
| Alan Baddeley and Graham Hitch | Psychologists | Working memory model (1970s) |
| Erich von Holst and Horst Mittelstaedt | Physiologists (1950) | Efference copy, reafference, exafference |
| Peter Medawar, George Williams, William Hamilton | Biologists | Evolutionary theory of ageing |
| Bruce Robison | MBARI scientist | 53-month deep-sea octopus brood |
| Andrew Packard | Biologist (1972) | Cephalopod-fish co-evolution |
| Christelle Jozet-Alves | Biologist, Normandy | Cuttlefish what-where-when memory |
| Thomas Huxley | Biologist (1883) | "Inexhaustible" fisheries |
| Andrew Barron and Colin Klein | Bee biologist / philosopher | Colony collapse as accumulated stress |
Connect it
- Distributed control for agentic AI. The octopus arm model (central goals, local closed-loop execution, edge sensing) is a sound pattern for multi-agent systems, and its cost, that the centre does not always know what the arms are doing, is the observability and audit problem regulated institutions must solve.
- Coordination is its own problem. The action-shaping view reminds architects that getting many components to produce one coherent action (transactions, orchestration, consistent state) can be harder than perception or decision, and deserves first-class design.
- Reasoning loops are reafference. Chain-of-thought and reasoning models write and re-read intermediate steps, a note-to-self loop; useful for System 2 tasks, but visible rationales are not guaranteed faithful, so validation should test behaviour directly.
- Late-life failure and buffered stress. The Medawar-Williams logic explains why long-lived bank systems accumulate defects nobody was ever pressured to fix, and the bee-collapse story describes operational resilience failures where hidden buffering runs out; monitor buffer consumption, not just outages.
- Forecasts assume "present modes". Huxley's fisheries error is the model-risk error of assuming the technology base is fixed; state and revisit those assumptions as tools (and fraudsters' tools) change.