Insight Editions · No. 290
Metazoa
Animal Life and the Birth of the Mind
Peter Godfrey-Smith
Peter Godfrey-Smith argues that felt experience is a product of animal evolution, not an extra ingredient added to matter or a property of all things.
A study guide to Metazoa 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
Metazoa is Peter Godfrey-Smith's account of how felt experience came to exist in a physical world, told through a dive-by-dive tour of animal evolution from single cells to octopuses, fish, insects and us. Its thesis is a biological materialism: minds are not extra ingredients or products of brains but the activity of nervous systems of the right kind, felt from the inside. Experience arose gradually, with the evolution of animal bodies that sense and act as unified selves and of brains whose rhythms tie that activity together, so it is far more widespread than the human-centred view allows, but not everywhere.
Core claim. Subjective experience is the evolved activity of animal selves: bodies that act and sense from a point of view, run by nervous systems with large-scale dynamics (rhythms, synchrony, perhaps fields). It came into being by degrees.
Scope. Philosophy of mind joined to evolutionary biology and neuroscience: 3.8 billion years of life, the Ediacaran and Cambrian, the move onto land, and experiments on crabs, bees, octopuses, split-brain patients and rats, framed by the author's own underwater observation in Australia and Indonesia.
Key figures. 600 million years since our last common ancestor with the octopus; two-thirds of an octopus's ~500 million neurons in its arms; a 1 mm³ bee brain; Robert Elwood's crab experiments; Octopolis and Octlantis; land holding ~85% of species on a third of the surface.
Best for. Readers who want a concrete, biologically grounded alternative to both computational theories of mind and panpsychism, and anyone deciding which animals (or machines) could be subjects and what follows for welfare and AI.
- The method: a rising sea. Godfrey-Smith adopts Grothendieck's image: surround a stubborn problem with knowledge until it changes shape and subsides. Biology did this for life, dropping the search for a special "protoplasm" once hidden organisation of ordinary matter explained it; he proposes the same for mind, with the key identity claim that brain processes do not cause experiences, they are experiences.
- Cells first. Before animals, cells already imposed order on a molecular storm, tamed electric charge with voltage-gated channels (nature's transistor), and lived as selves kept open by traffic across their borders, so sensing is ancient and everywhere.
- Animals: action, then a point of view. Animals made a new kind of self: sheets of cells moving together through muscle and nerves (the soft coral's reach), sensing that came partly "for free" with coordinated action (Keijzer), then in the Cambrian image-forming eyes and limbs. Sensing joined to acting creates the need to tell self-caused from world-caused change, and with it a point of view: the evolution of animal agency brings with it the origin of subjects.
- Clearing the ground. Zombie and disembodied-mind arguments exploit a quirk of imagination (Nagel's two kinds of imagining); Rothko-style qualia are a misleading model of experience; and experience is not all sensory: moods, energy and a feeling of presence show that more of life is felt.
- Octopuses and divided brains. The octopus, with most neurons in its arms, is a conscious invertebrate that is less integrated than we are. Via split brains and the Wada test, Godfrey-Smith argues that a mind is a pattern of activity that can expand, contract and split quickly, so octopuses may switch between unified and looser states in real time.
- The second bridge: brains as dynamic organs. Fish bring centralised brains and the rhythms and fields first recorded by Berger. A brain is not a telephone exchange but self-generated, synchronised activity modulated by the senses, and these large-scale dynamics are part of what experience is.
- Varieties and degrees. On land, sensory and evaluative experience may come apart (insects: rich senses, little sign of pain, real moods); plants have minimal cognition but less of a self. Experience is graded and multi-dimensional, not a light switched on.
- Consequences. Uploads are a fantasy and present computers hold only "a logical sliver" of what brains do, though brain-like artificial systems might one day feel; many more animals deserve consideration, though gradualism makes sharp sentience borders impossible; and mind's distribution is neither desert nor jungle. Humans add offline experience: memory, imagination and dreams make the mind a garden.
| Ch | Title | One-line payload |
|---|---|---|
| 1 | Protozoa | Life was explained by hidden organisation, not a special substance; mind can be explained the same way, as biological activity identical to experience. |
| 2 | The Glass Sponge | Cells tame a molecular storm and electric charge and live by traffic; animals are a new multicellular self, the glass sponge a post-cellular one. |
| 3 | The Ascent of Soft Coral | The first great animal invention was coordinated whole-body action through muscle and nerves; sensing may have come for free; the Ediacaran shows action beginning. |
| 4 | The One-Armed Shrimp | Cambrian arthropods bring eyes, limbs and the self/other problem; Elwood's crabs give the first good invertebrate pain evidence; modular animals take another road. |
| 5 | The Origin of Subjects | Animal agency brings subjects; zombie arguments, Rothko qualia and perception-only views mislead; moods and presence show more of life is felt. |
| 6 | The Octopus | A conscious but less integrated animal; split brains and the Wada test suggest minds switch between unity and division, as octopus arms may. |
| 7 | Kingfish | Fish bring centralised brains, long-distance touch and social smarts; brain rhythms and fields are the second bridge between mind and matter. |
| 8 | On Land | Sensory and evaluative experience can come apart; insects have moods but little pain; varieties of subjectivity; plants are on a different road. |
| 9 | Fins, Legs, Wings | Land vertebrates finally combine manipulation, open action and central brains; endothermy; most vertebrate brains are semi-split; land and sea innovate in sequence. |
| 10 | Put Together by Degrees | Experience is activity felt from inside, graded, an experiential profile; no uploads; wider animal consideration; neither desert nor jungle; offline mind as garden. |
Protozoa
Biology once tried to bridge matter and life with a special living substance, and the bridge collapsed; life was finally explained by the hidden organisation of ordinary matter. Godfrey-Smith proposes to do the same for the mind: a biological materialism (inside a wider monism) that explains felt experience as the product of animal evolution, not an added ingredient. The book's method is Grothendieck's rising sea: build knowledge around the mind-body problem until it changes shape and subsides.
Down the Steps: A Garden of Animals
The book opens on a stairway of breakwater rocks at Nelson Bay, north of Sydney, leading straight into the sea. A short swim out is a sponge garden, and the scene is set up as a preview of the whole evolutionary story, arranged by distance from shore.
- Sponges. Bulbs, fans and sideways-spreading layers: the only animals in the garden with no nervous system.
- Ascidians (sea squirts, silent "c"). Pale pink spouts like a ship's air funnels, so encrusted they look like scenery; they make small shifts and sometimes collapse and expel their water "with a shrug and sigh".
- Anemones and soft corals. Clusters of tiny hands that clench and slowly open: the flower that grasps.
- Serpulid worms. Orange feather tufts lined with eyes, which vanish if you come too close.
- A banded shrimp. Partly transparent, a few inches long, with appendages that triple its reach: the first animal in the scene that might see you as an object, rather than as washes of light and looming mass.
- An octopus. Stretched out on the reef like a camouflaged cat, raising its head to watch you as you pass.
In a forest most of what surrounds you took the plant path; in a sponge garden almost everything is an animal, and all but the sponges carry "electrified threads" through their bodies. Godfrey-Smith's image is a forest whose trees sneeze, cough, reach out hands and glimpse you with invisible eyes. The key caution comes immediately: these are present-day animals, products of as much evolutionary time as we are. Swimming out from shore takes you not into the past but among far-removed living cousins, the topmost branches of a single family tree, whose bodies still show remnants of early forms of animal action.
Matter, Life, and the Bathybius Affair
Bathybius haeckelii (1857 to about 1875). In 1857 HMS Cyclops dredged seafloor mud from the North Atlantic, part of a practical interest in seabeds driven by transatlantic telegraph cables (the first, completed in 1858, lasted three weeks before its insulation failed). The sample, preserved in alcohol, went to T. H. Huxley, who shelved it for about ten years. With a better microscope he then saw discs and spheres in a "transparent gelatinous matter" and proposed a new, exceptionally simple organism: living jelly that made its own hard parts. He named it after Ernst Haeckel, who was delighted, because he had argued such a thing must exist and believed spontaneous generation might be going on continually. Bathybius was imagined as a carpet of primordial life over the deep seafloor; a Bay of Biscay sample was even said to move. Then the Challenger expedition (Royal Society of London, four years in the 1870s, hundreds of deep-sea sites, chief scientist Charles Wyville Thomson) found no fresh Bathybius, and two scientists aboard showed by experiment that it was a chemical precipitate formed when seawater meets preserving alcohol. Huxley conceded at once; Haeckel held on for nearly ten more years.
Why would two leading Darwinians want living slime so badly? Huxley (England) and Haeckel (Germany) were the chief advocates of Darwin's theory, published in On the Origin of Species (1859), and both wanted to push into what Darwin had avoided: the origin of life. Their answer was protoplasm, a physical but non-ordinary "stuff of life" that would stay special even in a spoon. The motive was empirical: inside cells, microscopists saw too little differentiated structure to explain what cells do. William Benjamin Carpenter (1862) marvelled at a particle of "apparently homogeneous jelly" that seizes food "without members", swallows "without a mouth", digests "without a stomach" and moves "without muscles". Hence the slogan Huxley and others accepted: "organization is the result of life, not life the result of organization." Bathybius looked like a pure sample of that substance, a bridge between living and dead matter.
How Life Was Actually Explained
The bridge failed, and what replaced it was exactly the explanation Huxley and Haeckel ruled out: life as the hidden organisation of ordinary matter. Living things contain the same chemical elements as the rest of the universe and obey the same physical principles. The origin of life is still unsolved, but it is no longer a mystery that tempts anyone to posit an extra substance. The scale of that hidden structure defeats intuition: over 100 million ribosomes (the cell's protein-assembly stations) would fit on the full stop at the end of a sentence.
The materialist explanation of life succeeded by finding structure at an almost inconceivable scale, not a special ingredient. Godfrey-Smith's wager is that the mind will yield the same way: no mind-stuff, but organisation of a particular, evolved kind.
Bathybius is the classic artefact-of-the-pipeline error. A preservative reacting with the sample manufactured the "discovery". In ML and analytics the same thing happens when a preprocessing step (imputation, deduplication, a join that leaks the label) creates the very pattern a model then "finds". The Challenger's move, testing whether the preservation method alone reproduces the effect, is the right control: rerun the pipeline on data where the effect cannot exist.
How the Map of Living Things Was Redrawn
The Bathybius episode sat inside a longer struggle over categories of being, which the chapter traces because life and mind were long treated as linked.
| Scheme | Who, when | Categories | What it shows |
|---|---|---|---|
| Two kingdoms | Ancient Greeks onward | Animals, plants | Everything alive must fall on one side |
| Three kingdoms | Carl Linnaeus, 18th century | Animals, plants, Lapides ("kingdom of rocks") | Origin of "animal, vegetable or mineral?" |
| Microbes as "worms" | Linnaeus, Systema Naturae, 10th ed. | Tiny organisms filed under worms; last group Monas, "body a mere point" | Microscopic life forced into old boxes |
| Four kingdoms | John Hogg, 1860 | Adds Protoctista in a Regnum Primigenum ("primeval kingdom") | Single cells that are neither plant nor animal |
| Protista | Haeckel | Hogg's term shortened | The modern name for single-celled eukaryotes |
Microscopic life had been seen since the 1670s, probably first by the Dutch draper Antonie van Leeuwenhoek, whose microscopes were the most powerful of the period. Hogg thought the boundaries between living kingdoms were vague but the line between living and mineral was sharp; Bathybius was meant to blur exactly that line.
Two Paths: Life-and-Mind Versus Mind-Apart
The historical thread Godfrey-Smith follows is the shifting relation between life and mind.
- Aristotle's grades of soul. Soul is an inner form directing bodily activity, present in grades: nutritive (plants keep themselves alive), sensitive (animals also sense and respond) and rational (humans also reason). Even inanimate things tend toward their natural place. Mind is an extension of living activity.
- The Scientific Revolution. Matter is hardened into push-pull mechanism with little room for purpose, while the soul is etherealised into an intellectual affair that might be saved for eternal life.
- Descartes's dualism. A sharp divide between physical and mental, with humans a combination joined at a small organ in the brain. Other animals are purely mechanical: "a dog is without feeling, no matter what is done to it."
- Darwin's century. Gradual evolution by adaptation and branching blurs the human/animal divide and makes Cartesian dualism look less viable, provided you can get life started.
- The twentieth-century computer turn. From mid-century, materialist theories drifted back toward Descartes in one respect: mind came apart from life again. Computation offered "a bridge made of logic rather than life"; software was the heart of the matter, a program that could run on other hardware, and animal bodies looked optional.
The Hard Part: Something It Is Like
The computer era also sharpened the problem. Some mental capacities look explicable: memory as storing and retrieving traces is unsolved but "certainly looks soluble". What resists is that some memories, in humans at least, feel like something. Thomas Nagel (1974) put it as there being "something it's like" to have a mind. The hard part of the mind-body problem is explaining felt experience in biological, physical or computational terms.
- Materialism (physicalism). Mental phenomena, experience included, are manifestations of activities described by biology, chemistry and physics.
- Dualism. Mind is something added to a physical system that would otherwise be complete (Descartes's version: a separate substance).
- Huxley's epiphenomenalism. Experience is an effect of material processes but never a cause: an unusual dualism that still has defenders.
- Panpsychism. All matter, even a table, has a faintly mind-like side, which yields experience once matter is organised into brains. Not idealism: the physical layout of the world is accepted as it appears. Nagel keeps it on the table because every view has problems; Haeckel turned to it after Bathybius.
- Idealism. Everything is mental; apparent physical objects are manifestations of mind or spirit.
- Neutral monism. Both physical and mental are manifestations of some third thing, which tends to stay mysterious. Godfrey-Smith's fallback: "If I was not a materialist I'd be a neutral monist", though it is an outside possibility.
Materialism, idealism and neutral monism are all forms of monism, a commitment to underlying unity in nature, and that broader commitment is the heart of his outlook.
The Arbitrariness Problem
What bothers Godfrey-Smith most is how untethered judgements about who has experience become. One person watches a paramecium swimming through a water film and grants it tiny feelings because it is responsive and goal-directed; another looks at a fish and sees only reflexes and brain activity "in the dark". His guess at the common view: "yes" for mammals and birds, "perhaps" for fish and reptiles, "no" for everything else. Push outward (ants, plants, paramecia) or inward (mammals only) and the discussion loses its footing.
This connects to Joseph Levine's "explanatory gap": even if we became sure that the mind has a purely physical basis, we would still want to know why this physical setup gives rise to this kind of experience and not another. The book's target is not Levine's questions about particular experiences (which brain activity underlies seeing colour or feeling pain; that is neuroscience's job) but why it feels like anything at all to be a material being of our kind, with "our" stretched to cover many animals, and to do this in a way that reduces the arbitrariness.
The two observers at the pond. Put a paramecium and a goldfish side by side under one lens. Observer A says the paramecium feels and the goldfish does too; observer B says neither does; observer C says only the fish. Ask each what evidence would change their mind. If none of them can name any, their positions are not hypotheses but acts of imagination freely imposed. The book's aim is to supply the missing bridge: facts about bodies, sensing, action and nervous organisation that could actually move such a verdict.
Biological Materialism: The Position Stated
Godfrey-Smith wants to drop the mood of toughness that "materialism" carries ("just atoms bumping into each other"). The material world is energy, fields and hidden influences, and we should expect continuing surprises about what it contains. His project, put plainly: to show that a universe of processes that are not themselves mental can organise itself so that felt experience arises; "a part of the world's often-mindless activity folded itself into minds."
Against this, dualism and panpsychism both hold that you cannot build a mind wholly from non-mental ingredients: mind must be everywhere already, or added on top. His reply is that evolution can and did build minds from other things.
Brain processes do not cause experiences; they are experiences. Materialism is often misread (Huxley seems to have read it this way) as saying the mind is an effect or product of brain activity. The claim is identity: mental goings-on are biological, hence physical, processes of a certain kind. The arrangements are evolutionary products, slowly built, but once they exist they are minds, not causes of minds.
The solution will not come in "a single stroke of the pen". It will be cumulative, a sketch combining roughly three elements developed across the book, with many puzzles left open.
Choosing Words for the Target
| Term | What it captures | Problem Godfrey-Smith sees |
|---|---|---|
| Consciousness | Nagel's "something it's like"; any felt experience, however dim | Suggests something sophisticated, inviting the slide to "requires a cortex" |
| Phenomenal consciousness | Same broad sense, flagged as such | Acceptable, but the parent word still misleads |
| Sentience | Felt experience, often pleasure and pain, experiences with a valuation | Usually not applied to the sensory side (registering what is going on), which may be distinct from evaluative experience |
| Subjective experience | Points to a subject, the home where experience lives | Sounds redundant and has no handy adjective |
| Mind | The whole thing the evolutionary story illuminates | Broad; used freely |
The cortex worry gets a concrete example. Oliver Sacks, discussing a patient who after a brain infection could hold no new events in memory, asked how procedural memories tied to "relatively primitive portions of the nervous system" relate to "consciousness and sensibility, which depend on the cerebral cortex". The question smuggles in an assumption: is a cortexless being missing only rich, here-I-am consciousness, or all experience? Since most animals, and most in this book, lack a cerebral cortex, the difference is everything. Godfrey-Smith doubts the lights-out answer and warns against assuming all experience must be human-like. He moves between terms "without laying down the law"; his own emphasis is that the book concerns the evolution of subjectivity.
The cortex question has moved toward Godfrey-Smith's side. The New York Declaration on Animal Consciousness (April 2024), signed by a large group of scientists and philosophers, states that there is strong scientific support for conscious experience in mammals and birds and at least a realistic possibility of it in all vertebrates and many invertebrates, including insects, decapod crustaceans and cephalopods, and that ignoring such a possibility in welfare decisions is irresponsible. Earlier, the 2021 London School of Economics review led by Jonathan Birch judged the evidence of sentience in cephalopods and decapod crustaceans strong enough that the UK's Animal Welfare (Sentience) Act 2022 extended protection to them, though neither has a cortex.
The Rising Sea: Method and Path of the Book
Godfrey-Smith long kept as an epigraph a passage by the mathematician Alexander Grothendieck: the sea advances "insensibly and in silence", surrounding a stubborn mass until it becomes a peninsula, then an island, then an islet, then submerged. Faced with a problem that resists direct attack, build knowledge around it until it transforms and dissolves. Godfrey-Smith rejects the idea that the mind-body puzzles are mere illusions to be talked away; new things have to be learned, and as they are, the problem changes shape and fades. He dropped the epigraph because drowning Pacific islands on a warming planet gave the image new and wrong connotations, but the method stands.
Continuing Other Minds
Other Minds followed cephalopods back to the ancient fork in the animal tree, over half a billion years ago, that led to the octopus on one branch and to us on the other. Metazoa develops and augments its ideas by a closer look at the philosophy, at more branches of the tree and at more animals, near and far. Two choices shape the route:
- Transitions, not an encyclopedia. Animals are chosen because they mark stages in the evolution of the mind; most are marine.
- Weighted to early history. The goal is to explain how experience came to exist on Earth at all, first in water and later on land, with nearer relatives arriving toward the end.
Two things to watch as the argument unfolds. First, the promise to "partially close" the explanatory gap is modest by design; readers wanting a theory of why particular brain states feel as they do will not get one, and Godfrey-Smith says so. Second, the confident dismissal of panpsychism and dualism rests mostly on the cumulative plausibility of the evolutionary story rather than a direct refutation; committed panpsychists (Philip Goff, for example) would say the rising sea never reaches the stubborn rock of intrinsic experience. The book's strength is that it makes the materialist alternative concrete, not that it rules the others out.
The Glass Sponge
Before any animal existed, cells had already solved the problems the mind would later build on: imposing order on a molecular storm, taming electric charge with ion channels (including nature's version of the transistor), and living as bounded selves kept open to the world by traffic across their borders. Animals, the Metazoa, made a second, larger self out of many cells. The glass sponge, a deep-sea animal that fuses its cells into one electrically continuous web over a skeleton of glass, shows how strange the early experiments in multicellular unity could be.
Towers in the Dark
A sponge garden usually starts just below the well-lit water, where currents run: motionless bodies shaped like cups, bulbs, grails, branching trees or mittened hands reaching up from the seafloor. Now picture the same scene colder and 3,000 feet down, black, with a thin fall of particles. There stand pale cylindrical towers about a foot tall, fixed by delicate anchors and partly open at the top. Inside the soft exterior is a lattice of stars, hooks and slender crosses knitted at skewed angles. Anchors and crosses are silicon dioxide, the main constituent of glass.
A sponge looks dead, but it is a silent pump drawing water through itself, and it senses and responds as it does so. The deep-sea glass sponge also conducts light and charge, "like an electric light bulb ('think!') at the bottom of the sea."
The Cell and the Molecular Storm
The background to the mind is life itself, not DNA and its workings but the features of cells. Early life was single-celled; animals and plants are huge collaborations of cells. Even before those collaborations cells lived in colonies and clumps, but each was "a tiny self of its own": bounded by a membrane with channels and ports, with constant to-and-fro across the border and frenzied activity inside.
The word "matter" suggests something inert and ponderous that must be pushed into motion, an intuition built on dry land and on objects the size of tables and chairs. Inside a cell, events happen on the nanoscale (millionths of a millimetre) and in water, where everything behaves differently. The biophysicist Peter Hoffmann calls it a "molecular storm": ceaseless collisions, attractions and repulsions.
- About every ten trillionth of a second: how often a fast-moving water molecule hits any object inside a cell.
- Each collision: carries a force that dwarfs what the cell's own devices can exert; the apparatus can only nudge events one way rather than another.
- About 3.8 billion years ago: rough date of the origin of life, on a planet now about 4.5 billion years old.
Water keeps nanoscale parts from sticking together and seizing up as they would in air, so the cell is a place of self-generated activity. For a cell, the problem is never getting things to happen but getting organisation out of chaos: matter at this scale risks doing too much, not too little. If life had to evolve on dry land from table-sized parts it could not happen; it evolved in water (perhaps thin films on a surface) as order emerging in a storm. The first life may not have been cell-shaped, but something had to contain and mark off a special set of chemical processes so they did not diffuse away; leaky early cells eventually became something like bacteria, which maintain their organisation and reproduce. A central achievement on the way was control over charge.
The dry-land cell. Try to build a self-maintaining cell out of table-sized parts sitting on a floor. Nothing moves unless pushed; every reaction needs a motor; parts that touch simply stay touching. Now shrink the parts a billionfold and drop them in salt water. Suddenly the problem inverts: everything moves, collides and dissociates on its own, and the engineering challenge is restraint. The exercise shows why intuitions about "mere matter" mislead when applied to life: the material of a cell is restless by default.
The Taming of Charge
Humans tamed electricity in the nineteenth century, turning lightning-like danger into the basis of modern technology. Godfrey-Smith calls this the second taming; the first happened billions of years earlier in cells, where electricity is the means by which much gets done, brains included.
The Physics, Briefly
- Charge. A basic feature of matter, positive or negative; like charges repel, unlike attract. Atoms balance protons (positive, inside), electrons (negative, outside) and uncharged neutrons.
- Strength. Richard Feynman's illustration in his Lectures on Physics: if two people at arm's length each had one percent more electrons than protons, the repulsion would lift a weight equal to that of the entire Earth. Ordinary matter is calm only because the balance is so exact.
- Ions. Atoms or small molecules that have gained or lost electrons and so carry net charge. Salt water is water full of drifting ions.
- Current. Any movement of charged particles. In a wire, electrons move; in living systems, currents are mostly movements of whole ions. A consistent movement of ions in salt water is a current.
Charge is not life-like or mental in itself, but living activity runs on it, "especially by the corralling, pumping, herding, and unleashing of ions."
From Channels to Action Potentials
- Ion channels are shared, with variations, by all cellular life, bacteria included. They may have begun just to adjust a cell's charge relative to the outside.
- Minimal sensing. Traffic across a boundary takes on further roles: if contact with an external chemical opens a channel and lets ions in, those charges can set new events in motion.
- Excitability. Channels can themselves be opened or closed, by chemistry, by physical impact, or by charge. Voltage-gated ion channels open in response to electrical events, so a flow of current creates a greater flow that spreads over the membrane.
- The action potential. Positive ions flow in at one point, open neighbouring channels, which let in more: a wave of electrical disruption travels along the membrane like a pulse. This is what a neuron "firing" is.
A voltage-gated ion channel is a controller of current that is itself electrically controlled: the principle of the transistor, invented at Bell Laboratories around 1947 (an inch-sized device, shrunk ever since) and invented billions of years earlier by bacteria.
The puzzle Godfrey-Smith finds genuinely odd: bacteria have the transistor but no action potentials, and no answer to what they use it for is widely agreed. Candidates are electrochemical upkeep of the cell, control of swimming, incidental charge-sensitivity of chemical-sensing channels, and ion signalling between cells in biofilms. Nature built "the fundamental hardware device in computer technology", complicated and costly, in organisms that do not seem to compute much with it.
The device's importance lies partly in having no single use. Like a transistor, it is a general means of control, making events here affect events there reliably and fast. When it yields action potentials it also gives cell activity a "digital" quality: a neuron fires or does not. Not every animal's neurons spike, and nervous systems can run on milder excitability, but the digital option was available long before most of its uses existed, "not even glints in evolution's eye."
Cells Are Not Computers (Quite)
| Feature | Living cell | Computer |
|---|---|---|
| Main business | Self-maintenance: energy intake, keeping a pattern going through decay and turnover | None; maintenance is someone else's job |
| Switching and "information processing" | Embedded in a mini-ecology of other chemical processes | The whole point |
| Medium | Liquid, subject to the molecular storm and "chemical digressions" | Engineered for regularity and uniformity |
| Design goal | Order imposed imperfectly on spontaneous activity | Predictable, restricted behaviour, even when simulating chaos |
This raises the "machinery" dilemma. Godfrey-Smith keeps typing the word and wondering whether to delete it. Denying that ion channels, nerves and brains are machinery pushes toward dualism (soul plus body) or life-force views, so the word stays. But "imparting order upon a molecular storm and the imperfect herding of ions" is nothing like any machine we have built. Calling cell contents machinery is right in some ways and wrong in others.
Traffic: The Windowed Character of Life
The last pre-animal feature he puts centre stage is traffic between living systems and their surroundings: ion flows, raw materials in, waste out. Cells are bounded but not closed. The traffic has two sides:
- Metabolic. Energy and materials, unavoidable if life is to continue. Living activity is a pattern embedded in an energetic flow that begins and ends outside the organism.
- Informational. Some incoming influences matter in themselves (food); others matter for what they predict and portend about something else.
His colleague Maureen O'Malley says being alive means existing "on a redox rollercoaster, perpetually giving and receiving" (a redox reaction transfers electrons between molecules). The consequence she stresses: living systems cannot be windowless. Open to the world from energetic necessity, they are affected by what happens outside, and evolution then tends to put that sensitivity to use. Hence all known cellular life, including bacteria, senses and responds: sensing in its basic forms is ancient and everywhere. The chapter's first theme closes here: self-defining, self-maintaining cells are selves.
Metazoa: A New Kind of Self
"Metazoa" was coined in the late nineteenth century by Haeckel, contrasting multicelled animals with single-celled Protozoa. The Greek "meta" first meant after and beside, then higher, now often about; Haeckel probably meant some mix of higher and later. Protozoa are no longer counted as animals, so the animals are now simply the Metazoa.
- Animal = a branch, not a lifestyle. Any organism on that branch of the genealogical tree is an animal, from corals to giraffes, from wasps smaller than some single cells to fifty-ton whales. "A coral is as much an animal as a wolf is."
- No ladder. Earlier branches are "lower" only in the sense of earlier. Everything alive now is at the top of the tree; some animals are more complicated, but there is no overall scale from lower to higher.
- Tree, loosely. The genealogical network is tangled in places; "tree" is a simplification. It grows by splits, each side evolving its own peculiarities, with extinction always likely.
Where Animals Came From
Animals arose from eukaryotes, larger cells with mitochondria (energy handling) and a cytoskeleton, an internal network of moving filaments and tubes that controls shape and motion. The cytoskeleton had already opened a new regime of mobility and active hunting in single-celled protists: a shift from life based on chemical processing (bacteria) toward life based partly on behaviour, motion and manipulation. Some protists grew large; members of the genus Chaos hunt bacteria and sometimes small invertebrate animals.
Evolution repeatedly makes larger units from collaborating smaller ones. The eukaryotic cell came from engulfment (a swallowed cell became the mitochondrion). Animals and plants came, separately, from juxtaposition: a mutation makes daughter cells stick together after division, and their daughters stick too. The first result is merely a bigger object that cannot act as a whole or reproduce as one, but it is a step to a new kind of life. On the animal line this may have happened about 800 million years ago, give or take 100 million. No fossils survive; picture a ball of cells formed by cells refusing to leave their sisters.
- The form. The ball folds in on itself into a cup or hollow sphere with an opening, as in the gastrula stage seen early in the development of many animals.
- The caution. Early in development does not mean early in evolution (Haeckel assumed it did), but the cup form is so ancient and widespread it may be a real clue.
- The function. The cavity could be the first gut: trap food, release enzymes without them drifting away. "The first animal might have come to exist by forming around its stomach."
- The newer twist. Animal guts house colonies of helpful bacteria. Where mitochondria and chloroplasts were allies brought inside cells, an animal builds a pen for collaborating microbes. A diverse digestive ecology may sit at the very start of animal life.
- Why it matters. It is a second iteration of the cell: a new boundary with an inside and outside, with the individual cells now pieces of that boundary, controlling traffic into the larger unit.
Godfrey-Smith's verdict: Bathybius "was not Haeckel's finest hour; the Gastraea was better", and it remains a live possibility.
Living Clues: Simple and Far Away
The fossil record of the next steps is "frustratingly silent", so clues come from living animals that are both simple (few parts, few cell types) and genetically far from us, on lines that branched off very early.
| Clue | Lifestyle | Nervous system? | Movement |
|---|---|---|---|
| Sponges | Fixed adult, often large, filter-feeding | No | Adults stationary; larvae swim |
| Placozoans | Tiny, flat, shapeless; microscope needed | No | Crawl |
| Comb jellies (ctenophores) | Transparent, jellyfish-like but possibly far from jellyfish | Yes | Swim by beating cilia in comb rows |
"Motionless undersea furniture", a nerveless microscopic crawler, and a transparent swimmer. Simplicity and distance need not go together; an animal that branched off early could have become very complicated in the same time we did. The best partial example is the octopus, though it is not nearly as far from us as these three.
Cousins are not grandparents. It is tempting to read the tree as a sequence: sponge-like ancestor, then jellyfish-like, then us. That treats distant cousins as grandparents, or as more grandparent-like than other cousins. Living sponges have had as much evolution as we have. Godfrey-Smith's legitimate reason for using them is narrower: these animals lack inventions we have (brains, hearts, backbones, the left-right body layout, folded internal organs) and belong to lines that apparently never had them, so together with genes and fossils they show what life without those inventions could be like.
Light Through Glass: Sponges
Sponges have the best fossil record of the three clues and have historically been treated as the most important. They range from temperate fingers and trees to huge tropical funnels and the deep towers, and some just encrust other organisms.
- Feeding. Water is drawn in low, passes up through the body and exits at the top; bacteria are filtered out on the way. Some deep-sea sponges are predators, trapping small animals.
- A body the world flows through. Most cells touch the water directly; the body is a maze of passageways lined with microbial partners, and "the environment suffuses that body."
- No nervous system. Larvae, like tiny fat cigars, swim and have sensing structures resembling nervous parts, but these face outward to the world, not toward other cells.
- Controlled pumping. Flagellated cells pump the water, and the pumping can be slowed or stopped, especially when the water is dirty and channels risk clogging. Specialised sensory cells along the chimneys signal to others, usually by releasing small molecules, and the canals contract. In some species the sponge expands first and then contracts: a slow, invisible "sneeze".
For a nerveless collection of cells, getting one cell to influence another is a significant task. Multicellularity brings a trade: a cell locked in a sponge is safe from being engulfed, as it would be floating free, but would starve if merely stuck in place, hence the filigree of channels. And whenever something must be done, coordination, especially coordinated motion, is hard. The upshot looks plant-like, and most sponges have done well that way for far longer than us. A few tried something different.
The Hexactinellida: Living Post-Cellular
In glass sponges (Hexactinellida), most cells fuse as the animal grows, losing their boundaries with each other (not with the outside world). The body becomes a single connected net, often called a "three-dimensional cobweb", draped over glass supports shaped like daggers, stars or snowflakes, grouped into flowers and vine clusters and finally into the tower's skeleton.
The electrical sponge. A glass sponge normally pumps water continuously. Pluck a single glass element from its body and pumping stops quickly, because an electrical pulse sweeps through the tissue. Glass sponges are the only sponges with electrical signalling, a kind of action potential. Electrically the animal behaves like one enormous cell: the pulse runs uninterrupted over the whole. It achieves coordination not by coordinating signals between cells but by being largely post-cellular, a product of the animal path that has partly abandoned multicellularity for a different kind of unity.
Venus's flower basket. This glass sponge usually houses a pair of small shrimp that enter when tiny, grow too big to leave through any opening, and reproduce together inside. They keep the sponge clean and get protection from its skeleton and food from the water it draws through.
Then there is light. Parts of some glass sponge skeletons work like fibre-optic cable, conducting and filtering light. Whether the sponge uses this, or light is merely channelled by accident of building with glass, is open. Possibilities raised for different species:
- Internal communication, with light coming mostly from bioluminescence except in the shallowest cases.
- Feeding symbionts. Diatoms and other light-living organisms huddle deep inside some sponges, perhaps too deep to survive without ducted light, which is even carried a little way into the seafloor.
- A lamp for partners. The Venus's flower basket emits light into the water, perhaps attracting the shrimp that come to live in it.
Some biologists think the light will be too weak to do much. "Either by design or by accident, a glass sponge is a collector and curator of biological light."
Fused nerve nets turned out not to be unique to glass sponges. In 2023, Pawel Burkhardt and colleagues (Science) used electron microscopy to reconstruct the nerve net of a juvenile comb jelly and found its neurons fused into a continuous syncytium, with no synapses inside the net, quite unlike the separate, synapse-linked neurons of other animals. Together with 2023 genomic evidence that comb jellies are the sister group of all other animals (see Chapter 3), this strengthens the idea that early animals explored several distinct ways of tying a body together electrically, of which the glass sponge's post-cellular web is one. The optical side was established earlier: in 2003 Joanna Aizenberg's team at Bell Labs (the transistor's birthplace) showed that the anchoring spicules of the Venus's flower basket are working optical fibres.
Fusing versus messaging. The glass sponge coordinates by removing internal boundaries, so one signal reaches everywhere at once; other animals keep cells separate and pass targeted messages. Distributed systems face the same trade: a shared-state monolith gives fast global broadcast but no isolation (one fault spreads everywhere), while message-passing services keep failures contained at the price of coordination overhead. The sponge's choice suits a body whose only "decision" is stop pumping now; it would be a poor design for anything that must do several different things in different places.
The Ascent of Soft Coral
The decisive early animal invention was action on a new scale: whole sheets of cells contracting together through muscle and nerves, seen today in the slow reach and clench of a soft coral. Sensing lagged behind, and Fred Keijzer suggests it may have come partly "for free" once complex action existed. The fossil record of the Ediacaran (Avalon, White Sea, Nama) shows animal action starting fitfully: still fronds, then crawlers, grazers and the first scavengers moving toward sensed targets, then a puzzling retreat, with the bilaterian body, built for traction and direction, already in place.
Ascent: What Corals Do When No One Is Watching
In a bay north of Sydney a river meets the Pacific over a sandy underwater plain swept by strong tides. Divers can visit only around slack water, about an hour at each turn; when the tide turns there is a tug, and soon no one can swim against it. Fields of purple and white soft coral grow there: wispy, unmineralised, cauliflower-shaped trees that from a distance look like puffs of cloud on stout pale stems, with cowries and crabs in the branches. Each tree is a colony of many small animals, coral polyps.
Tom and Nicola Davis's time-lapse cameras. Tom Davis, a local diver and researcher, wondered what the corals do in the hours of fast water when no human can be there. He and his wife Nicola left cameras at several sites. As the current sped up with each tide, the soft corals slowly inflated until they were up to three times their still-water size, very likely reaching out for the extra food swept past, and subsided as the water slowed, sitting low during the hour when divers return.
Cnidarians and the First Animal Actions
Corals are cnidarians ("nye-dairians"), with jellyfish and anemones. They split from our line early: you last shared an ancestor with a coral perhaps 650 or 700 million years ago, more recently than with a sponge. The body is soft, radial (organised around a disc or cup), often fringed with tentacles, and contains muscle and a nervous system.
- Two adult forms. The polyp is usually fixed and cup-like; the medusa is the free-swimming jellyfish. Many cnidarians cycle between them, and bodies multiply at more than one step, as if a caterpillar could make many butterflies and a butterfly many caterpillars. Corals and anemones live only as polyps.
- Octocorals. On a nearby reef a second soft coral grows in shrubs and masses. Each polyp is a white flower with eight finger-like tentacles, each bearing side rays called pinnules: "fingers on fingers." An orange sponge often blankets the colony, with polyps reaching through.
- Reach and clench. Single tentacles curl while others stay out; whole hands close; whole zones close while neighbouring zones stay open. The Canadian biologist John Lewis examined thirty octocoral species and found some catch plankton and tiny invertebrate animals. The pace is faster than a plant, slower than familiar animal bustle.
Why Look for First Actions Here?
Godfrey-Smith gives two reasons. Cnidarians are old forms, and their radial layout probably resembles that of some of our predecessors, even if no living cnidarian resembles an ancestor closely. And they act. Action was not invented by cnidarians: single cells swim with flagella and cilia, envelop prey and change shape, and the sponge's controlled pumping is akin to action. But evolution often rediscovers something old at a new scale. The pre-animal world was full of single-celled predators and prey, and one escape from being engulfed is becoming too big to swallow, which may have spurred multicellularity itself. Once cells formed animals, action had to be reinvented for the larger unit. Sponges sit on the edge of this; in cnidarians, action exists again unambiguously, as coordinated motion on the scale of the whole animal body.
The stinging cell (nematocyte). A harpoon coiled inside one cell, surrounded by sensors and controllers in a "battery" (an apt artillery metaphor). When triggered it reaches extraordinary acceleration over a tiny distance. Stingers range from barely noticeable (many anemones) to lethal (box jellyfish), and are similar enough to have come from a single early invention passed down the cnidarian branches. The act itself is performed by a single cell; no coordination between cells is needed.
The soft coral's reach. Not the act of a cell but countless contractions by different cells occurring together in a concerted way. What Godfrey-Smith marks as the special invention is action coordinated across vast scales from a cell's point of view. Its origin is what the slow reach echoes.
Two objections, two replies:
- Why not the medusa's swim? The medusa is usually seen as a later addition to cnidarian life, and in any case a medusa is a polyp upside down. Both motions are radial contraction: in a free bell it produces swimming, in an anchored cup a grasp.
- Why movement rather than chemistry? Brewing chemicals also gets things done. But controlled motion on the scale of the animal body made a new kind of object in the world and a new factor in making things happen.
One sniper versus a stadium wave. A single marksman firing is fast, precise and needs no one else: that is the nematocyst. A stadium wave needs tens of thousands of people each doing a small thing at the right moment relative to their neighbours; no one person "does" the wave, yet the wave is unmistakably one event. The soft coral's reach is a stadium wave of cells. The thought experiment shows why the coordination problem, not the speed or force of any one part, is the real novelty of animal action.
The Animal Path: Nerves and Muscle
The animal outgrowth of the tree quickly gathered a suite of innovations, the nervous system above all. Nervous systems are present in cnidarians and comb jellies but not sponges or placozoans; they evolved early, perhaps once, perhaps twice. They combine two capacities far older than animals:
- Excitability. A rapid change in a cell's electrical properties (Chapter 2), normally confined to that cell by its boundaries.
- Chemical signalling. The electrical spasm can trigger release of a chemical at the boundary, picked up by a nearby cell, making that cell more or less likely to excite in turn.
What makes the full animal nervous system special is neurons: tree-shaped cells whose fine projections bring one cell into chemical contact with a selected group of others. Excitable, chemically signalling cells exist elsewhere, but neurons exist only in animals. They allow fast, targeted influence instead of broadcasting chemicals to whoever is listening, and they tie the body together in new ways.
- 1 cubic millimetre: the size of a bee's brain, in Lars Chittka's illustration.
- An oak tree: the branching complexity of a single bee neuron.
- 10,000: the number of other neurons one such neuron can connect to.
- 650 to 700 million years ago: rough date of the last common ancestor of humans and corals.
Nerve toxins show how much the nervous system holds together. They are the main fast-acting toxins, in snakes and in human weapons such as sarin, VX and Novichok. As a child Godfrey-Smith assumed they made you numb; in fact victims often die by asphyxiation or cardiac arrest. These chemicals destroy no tissue; they interfere with tiny sites of interaction between cells. "A way to kill that body is by going after the messengers, hence going after the coordination of parts."
Muscle arrived alongside. In animals, the cytoskeletons of many connected cells are coordinated to contract and relax massive sheets of cells. Some action needs no muscle: comb jellies swim with bands of cilia lined up like combs (they use muscle only to steer), and cilia drive small motions in many animals. But every larger motion (the octocoral reach, the medusa's swim, everything to come) runs on muscle.
Action Ahead of Sensing
Sensing exists in all cellular life, and cnidarians have a range of senses, but they are "skinnier" on the sensory side than on the side of action.
- Eyes. None in corals or anemones; only rudimentary eyes in most other cnidarians. Box jellyfish, with real eyes, are thought to be a later product.
- Stimulus-tied action. The polyp's reach, the colony's rise and fall and the firing of stingers are all triggered by stimuli.
- Gravity. Cnidarians probably invented the balance sense: medusae orient with statocysts, organs holding small crystals heavier than water whose movements are detected when the body tilts.
On this reading the cnidarian step into the unknown was large-scale, muscularly controlled motion, not sensing.
Subjectivity and Agency
Here the chapter turns to the background philosophy and introduces the complementary pair used for the rest of the book.
- Subjectivity. A matter of seeming, of "for-me-ness": experience as something that happens to a person or animal. The more sensory side.
- Agency. A matter of doing, trying, initiating: "by-me-ness", being a source of action and its effects. The more active side.
- A tangle in the word. "Subject" (unlike "subjectivity") also means a doer (subject versus object), the first of several entanglements of the two concepts.
- Evolutionary coupling. Sensing exists to control action; nothing is gained biologically from information that is never used. So agency and subjectivity evolve together, though not in lockstep: a breakthrough in action can arrive alongside simpler senses.
Godfrey-Smith credits the Dutch psychologist and philosopher Fred Keijzer for the emphasis on shaping action as the early job of nervous systems (the "action-shaping" view of Other Minds). Metazoa adds a further Keijzer idea about how sensing relates to action.
Sensing "for free". To produce elaborate coordinated motion, parts of the system must be sensitive to what other parts are doing. Then anything external that touches the system disrupts that internal pattern and is automatically registered. Even a nervous system looking entirely inward (Keijzer does not claim any ever did) could not help seeing outward to some extent. New and expansive actions carry an expansion of sensitivity along with them.
The apparent asymmetry may be an illusion, since complex sensing can be well hidden (chemical senses leave no fossil). But the image of an animal whose motions outrun its senses, and of senses catching up automatically, is a useful one for thinking about what existed before experience.
The blindfolded rowing crew. Eight rowers, eyes shut, concentrate only on keeping their stroke together; each feels the rhythm through the boat and the oars of neighbours. A log bumps the hull. Nobody was watching for logs, yet all eight feel the stroke stutter and know something out there happened. Their coordination system has become a crude detector of the world. That is Keijzer's point in miniature: internal coordination produces external sensitivity as a by-product, which evolution can then sharpen.
The emerging picture: all living things do things, but on the animal branch evolution produced a multicellular self and multicellular action, sheets of cells that contract, twist and grasp. That invention "changed everything, eventually", and its start was fitful.
Avalon to Nama: The Ediacaran Record
Every branching so far happened before any animal fossils. The first period with definite animal fossils is the Ediacaran, beginning about 635 million years ago: a seafloor of soft-bodied organisms, some tiny, some several feet across, leaving fronds, whorls, discs, spirals and fractal branchings.
- Were they animals? Some may be a lost multicellular experiment far from animals, but some certainly were. In 2018 the student Ilya Bobrovskiy, rappelling down a cliff in remote Russia, found naturally mummified Dickinsonia whose remains contained cholesterol, made only by animals. Dickinsonia was a flat, seafloor-dwelling "bath mat" up to a metre long, with no eyes, limbs or other familiar parts, typical of Ediacarans: definite forms (fronds, wheels, three- and five-sided bodies) but no legs, fins, claws or complex senses.
- Cnidarians and sponges? No clear cases. Some Ediacarans resemble sea pens (relatives of soft corals that look like a plumed pen with its nib in the sand), though close inspection shows differences. Reg Sprigg, who found the first Ediacaran fossils while inspecting disused South Australian mines in 1946, called many of them "jellyfish"; most have been reinterpreted, though real jellyfish were probably in the water above, collapsing beyond recognition when they died.
- The placid picture. Almost no signs of predation: no half-eaten bodies, no claws or spines, no weapons offensive or defensive. No sexual specialisation is known and no Ediacaran has been assigned a sex (sex probably existed alongside asexual reproduction). Densities were high, slabs packed with dozens or hundreds of organisms, yet in these Hieronymus Bosch-like scenes nobody seems to have much to do with anyone else.
Waggoner's Three Assemblages
About twenty years before the book, the biologist Ben Waggoner divided the Ediacaran fossils into three assemblages (collections of species laid down at roughly the same time), a scheme that has held up as data accumulated.
| Assemblage | Date | Setting | Typical life | Behaviour |
|---|---|---|---|---|
| Avalon (Canada; Welsh "island of fruit trees") | ~575 million years ago | Deep sea, too dark for photosynthesis, hundreds to thousands of metres | Fronds like leaves stuck in the seafloor, each a sheaf of fractal branches; one conical possible sponge | Stationary; perhaps absorbing dissolved organic carbon over a maximised surface |
| White Sea (Russia) | ~560 million years ago | Shallow seafloors covered in "textured organic surfaces" | More diverse body plans: sea-pen-like forms plus flat grazers, Dickinsonia, Kimberella, Spriggina | Movement: grazing, scraping, burrowing, scavenging |
| Nama (Namibia) | Youngest, to the end of the Ediacaran | Varied | Fronds again (mostly new kinds), tiny burrowers | Larger crawlers gone; a possible "Wormworld" of small tunnelers |
Avalon
The Ediacaran opens about 635 million years ago as a massive glaciation eases, one that may have covered the whole Earth in ice. A quieter stretch, a second ice age, and then the fossils appear; oxygen may have risen markedly after that second glaciation, but this was still a low-oxygen world, which limited activity. Avalon's frond-ish organisms maximised surface area, perhaps to soak up a mist of dissolved organic particles and the oxygen to burn it. Sponges are a puzzle: chemical and genetic evidence suggests they were common, but fossils give only the Avalon cone and a recently found later form like "an old downturned TV antenna" with rods radiating from its centre.
White Sea
Mary Droser of the University of California, Riverside, the central figure in work on this stage, calls the seafloor mats "textured organic surfaces": bacteria plus algae-like organisms and small embedded animals, preserved as "wavy-crinkly lamination, elephant-skin textures", a flatland in the sea.
- Dickinsonia grazed in one spot, then moved on, leaving faint whole-body footprints.
- Kimberella, a likely mollusc relative, was a crawling macaron that scraped the mat with an extendable spade-like part.
- Spriggina has "movement written over its form", like a scurrying trilobite; it left no traces, but gliding over a mat leaves none.
Helminthoidichnites: the first targeted movement. The name was given in the nineteenth century to burrow traces in younger rocks. Similar traces turned up in Ediacaran rocks and were analysed by Mary Droser and Jim Gehling in South Australia, near the first Ediacaran finds, using a new excavation method that exposes the undersides of huge slabs as wholes. The maker moved through different layers of the mat, piling up levees, and its tracks home in on the bodies of other animals, Dickinsonia included. This is the first fossil evidence of scavenging and the first physical trace of motion aimed at a sensed goal. Scavenging grades naturally into predation of slow or stationary prey. As the book was being finished, a candidate maker appeared: a tiny bean-shaped animal from South Australia.
Nama
One might expect the crawling complexity to grow. Instead the Nama rocks are quieter: the larger mobile, vaguely mollusc-like animals disappear, Helminthoidichnites traces persist, and fronds (mostly new ones) return. One interpretation, "Wormworld", makes small tunnelers and burrowers the major players. No one knows why the mobile forms vanished.
Where Were the Nervous Systems?
- Timing. Genetic dating suggests nervous systems arose before these fossils, or during Avalon; the dating is rough.
- Avalonian puzzles. If the fronds were animals with nerves, what did nerves do? Well-preserved specimens show no body openings, no mouth to bring food to, suggesting absorption over the whole surface. They may not have been animals at all.
- Look up. The evidence suggests nervous systems evolved in a radially organised body. Much early life may have been aloft in the water column: soft-bodied swimmers like jellyfish and comb jellies that rarely fossilise. One hypothesis derives crawlers from a cnidarian-like larva that reached the seafloor and started to move along it, leading on to mat-grazing and then to behaviour directed at other animals.
Godfrey-Smith suspects that, for the evolution of behaviour and interaction, the seafloor record is the tip of an iceberg, while noting how often the "apparently unknowable becomes knowable" through a technical advance, a mummified Dickinsonia or a new idea.
Several Ediacaran and early-tree questions have moved since 2020. - The bean-shaped animal is Ikaria wariootia, described in 2020 by Scott Evans, Mary Droser, Jim Gehling and colleagues (PNAS): 2 to 7 mm long, with a front and back, and the best candidate maker of Helminthoidichnites, making it one of the oldest known bilaterians. - Comb jellies first. In 2023, Darrin Schultz and colleagues (Nature) compared ancient chromosome-scale gene linkages across animals and found that ctenophores are most likely the sister group of all other animals, sponges included. That resolves the left-hand broken line in the tree above in the ctenophores' favour (not everyone is persuaded). It sharpens the "perhaps twice" question: either neurons evolved separately in comb jellies and in cnidarians-plus-bilaterians, or sponges and placozoans lost them. Comb-jelly neurons are odd (fused into a net, with distinctive signalling molecules), which many read as support for a separate origin. - Placozoan proto-neurons. A 2023 single-cell study (Najle and colleagues, Cell) found peptide-secreting cells in placozoans that express much of the genetic toolkit later used by neurons, suggesting neuron-like cell types assembled step by step before true nervous systems. - The Cambrian boundary that closes the Ediacaran is now dated to about 538.8 million years ago.
Small errors and soft spots. The figure caption in the book spells two names wrongly: the White Sea "TV antenna" sponge is Coronacollina, and the Nama frond is Swartpuntia. More substantively, the chemical evidence for abundant Ediacaran sponges (sponge-type steroid biomarkers) has been challenged since 2019, when similar compounds were found in single-celled rhizarians, so sponge abundance before the Cambrian is less secure than the text implies. And the claim that cnidarian-style action "came first" rests heavily on inference from living cousins, which the book itself warns can mislead.
Bilaterians and Traction
The other development with huge downstream consequences was the bilaterian body: a left-right axis as well as top and bottom, with paired parts (limbs, eyes, ears). Humans, ants, snails and seahorses are bilaterians, as are most animals now "however you calculate 'most'", and some Ediacarans (Kimberella, Spriggina). Cnidarians, comb jellies, sponges and placozoans are not. The body plan had to exist before the White Sea stage, because several distinct bilaterian lines are already present there, so the last common ancestor of you and a butterfly, or you and an octopus, lived at least that long ago.
The bilaterian body is set up to go somewhere. There are no non-bilaterian animals on dry land (no walking jellyfish, no anemones holding fingers out in air, apart from a few intertidal ones). Bilaterians seem to have begun on the seafloor, a marine version of land, as bodies for crawling over surfaces with direction and traction.
Flatworms as a Weak Clue
The first bilaterians were probably simpler and possibly smaller than the fossilised White Sea animals. The living clue usually offered is flatworms, but it is weak: the flatworm form seems to have arisen more than once, and many flatworms are parasites, a lifestyle that breeds simplicity. Acoel ("a-seel") flatworms have been seen as the best clue; the conspicuous ones on reefs are polyclads.
- Look. Often about a centimetre long, oval, flat with ruffled edges, thin as tissue paper.
- Behaviour. Faster and more purposeful than many neighbours: "I might look like a bit of tissue paper, but right now I really have something to do."
- Body. No through gut (food goes in and out the same end), no circulatory system; a cyclops-like eye cluster may sit mid-back, and simple eyes pop up almost anywhere, even on raised pinches of tissue.
- Sex. Hermaphrodites; some engage in "penis fencing", each trying to inject sperm into the other.
- Colour. Often brilliant, though they cannot see each other; many are mimics of nudibranchs.
Nudibranchs and Homage
Nudibranchs ("noodi-branks") are sea slugs, so molluscs, often stunningly coloured. Many eat chemically defended food such as sponges, becoming unpalatable, and their colours probably advertise this.
- Dorids are conveniently slug-shaped, easy models for a mimicking flatworm.
- Aeolids carry streamer-like appendages that sway as if in a breeze (after Aeolus, keeper of the winds). Each spring tiny aeolids, millimetres long, appear on bryozoans, shrub-like animals of tangled strands, "like tiny jeweled birds" in hair-fine trees.
- Tritonia, near the aeolids, includes a tiny pearl-white form living in the soft corals and sponges, with spires on spires, as if designed by a miniature Antoni Gaudí. It is hard to see because it mimics soft coral polyps.
Watching a Tritonia move slowly beside a soft coral colony, the two sharing one shape, Godfrey-Smith found himself thinking not of camouflage but of homage: the slug's body seeming to say it "pays homage to the first animal actions."
Capability before observability. Keijzer's picture, in which action scales up first and sensing follows partly as a side effect, describes many enterprise platforms: teams build orchestration that does a great deal, and monitoring arrives later, often by repurposing internal coordination signals (queue depths, heartbeats) as the first view of the outside world. The lesson cuts both ways: those signals are cheap and real, but they detect only what disturbs the rhythm, and a system that relies on them alone is as blind as a coral to anything that does not touch it.
The One-Armed Shrimp
With the Cambrian and its arthropod pioneers, animals acquire image-forming eyes, jointed limbs and claws: they can see objects and act on them. Multicellular sensing joined to multicellular action creates a new problem and a new opportunity, telling self-caused from other-caused change, and with it a point of view. Robert Elwood's experiments give the first decent invertebrate evidence of pain (in crabs and shrimp), which forces a rethink of insects; meanwhile modular animals such as corals and bryozoans show the road away from unified action.
Maestro: The Shrimp Who Looked Back
Near the soft-coral plain an old pipeline runs across the seafloor, overgrown with sponges and soft corals ("slow forms of animal action"). Under a ledge is something different: moving feelers, fussing legs, and attention turned toward you.
Banded shrimp (Stenopus hispidus) are striped like barber's poles. "Shrimp" and even "crustacean" are loose labels gathering several nearby branches, but all these animals are arthropods, a real and very large branch. A banded shrimp is a few inches long, with two large clawed arms and white antennae longer than its body.
The one-armed maestro. Months after seeing a pair at one spot, Godfrey-Smith returned and found a banded shrimp on an ascidian spout, missing one big claw. It stood in front of a small dozing shark under the ledge, seemingly poking at it; the shark barely reacted. When the shrimp climbed up under the ledge and hung upside down, with only its feelers showing, he gently stroked one feeler. The shrimp climbed straight down and looked back at him, then allowed further touches, standing with its one arm raised "like a conductor, a tiny maestro, in front of the dozing shark."
This is the book's first animal that can see an object like you (box jellyfish aside), the first that can move fast, and the first that can manipulate things, directing action on objects as well as directing its gaze. It has a different mode of being from anything so far.
The Cambrian
The Cambrian, beginning about 540 million years ago, looks like a sudden shift, almost a rupture (though "sudden" still means a million-year timescale, and whether the suddenness is real is debated). Its fossils have hard parts, legs, shells and conspicuous eyes. Explanations often work together:
- Environment. More oxygen, and ocean chemistry more favourable to animals.
- A new evolutionary regime. Building on the White Sea trend (Chapter 3): moving slowly toward food, then not so slowly as others compete, then scavenging becoming predation, then an arms race. Once others track your scent and follow, it pays to improve your own senses and motion. Eyes may have had a special role, but the race soon spanned every facet of sensing and behaviour.
There is no fossil record of the newcomers mauling the Ediacarans: the Ediacarans seem to have drifted offstage before the Cambrian cast arrived.
Arthropods Lead
Arthropods led. Early forms often look like crustaceans, though the most conspicuous, trilobites, may be closer to spiders; insects came much later. Their invention was a new way of being an animal:
- A skeleton that scaffolds and organises complex action (an external skeleton with jointed limbs).
- Claws.
- Image-forming eyes to go with them.
The banded shrimp is "the very essence of arthropod". Its usual kit is roughly six claws, four legs, six feelers and two comb-like extras: eighteen or more limbs and protrusions, "a body like a Swiss Army knife". It is hard to keep track of: at one point the shrimp's one big claw began to grab one of its own legs and quickly let go. Arthropod design is exuberant: "when in doubt, add some legs"; some crustaceans naturally carry one outsize claw, "as if the animal bought it late at night online." The largest arthropod ever found, Godfrey-Smith says, is a giant anomalocarid over seven feet long that was a peaceful plankton feeder, like a baleen whale.
Fewer arthropods swim now, given long competition in open water. At Lembeh Strait, Indonesia, he watched tiny, nearly transparent anemone shrimp (Periclimenes, many now moved to Ancylomenes), cleaners that live among anemones for protection, buzzing around "like a happy group of angels".
In the pipeline scene the shrimp was the only animal with anything like arms or claws; the molluscs, worms and shark around it had little means of manipulation. That is how the Cambrian was too, and no Cambrian animal had teeth: "In the land of the limbless, the one-armed shrimp is king."
Animal Sensing
Sensing was not invented by animals: single cells track touch, chemicals, light, even Earth's magnetic field. But animals made the same move with sensing that they made with action. They evolved sensitive surfaces, arrays and screens that take an imprint or image: the retina is an organised layer of cells on which light forms a pattern whose spatial layout (which bits sit next to which) has effects downstream; skin touch sensors register the shape of an imprint or the texture of an object. Animals became new kinds of objects again, with parts of their bodies becoming maps of fragments of their surroundings. Whether action really ran ahead of sensing (Chapter 3) is unknown; in the Cambrian both advanced rapidly.
Eyes, and the Mantis Shrimp
- Compound eyes. Most arthropods: many small units, each with its own lens.
- Camera eyes. Ours: one lens, one retina. Some spiders have camera-like eyes, a few with a telephoto lens built in.
- The mantis shrimp (stomatopods). By some standards the most elaborate eyes of any animal: two spheres on stalks that rotate freely and independently, depth vision within a single eye (different parts view the same object), and about a dozen colour receptor types against our usual three.
"WHAT? WHAT?" At Lembeh, Godfrey-Smith followed a six-inch mantis shrimp, a compact lobster with "a head festooned with golf clubs and party lights". It kept stopping to look back, an irritated interrogation (more accurately, "check ... check ..."), without ever turning around, because its independently rotating eyes did the looking. Mantis shrimp weapons are hammers and pikes driven by "springs, latches, and lever arms" (from a paper co-authored by Roy Caldwell, the Berkeley "stomatopod whisperer"); a released hammer briefly moves fast enough to vaporise water.
More colour receptors did not mean finer colour vision. In 2014 Hanne Thoen, Justin Marshall and colleagues (Science) showed that mantis shrimp are actually poor at discriminating similar colours, worse than humans, despite their 12 colour channels. The likely explanation is a different scheme: each receptor type acts as a quick lookup, letting the animal recognise colours fast without the comparison-heavy processing our three-cone system uses. Elaborate sensors can be traded against less central processing.
The banded shrimp is gentler. Its apparent curiosity has a biological basis: like the tiny Lembeh shrimp, it is a cleaner, eating parasites off fish, eels and turtles. The shrimp that climbed down was probably inspecting a prospective client.
Self and Other: The Feeler Problem
With action and sensing both multicellular, their relationship changes shape. The banded shrimp's antennae are the ideal example: long, constantly waving, and sensitive. Touch one with a finger and the shrimp often responds at once, yet under a cramped ledge the feelers must bump things all the time because of the shrimp's own movement. The same touch can have different causes, and the shrimp seems able to tell touches due to its own wandering from touches due to an other.
No one has studied this in banded shrimp, but crayfish and flies have systems that modulate the interpretation of sensory input by the animal's registration of what it is doing. The capacity is widespread, found even in some animals with simpler nervous systems than arthropods.
The steps, as Godfrey-Smith puts them: work out what sensory changes your current activity (moving or still, and how) should produce, and look for changes "over and above" these; those are marks of the external. An animal that does not do this has its own motion confound its picture of the world. An animal that does is tracking the divide between self and other. However simply it is achieved, the animal now refers some events to the world and others to itself.
Merker: mobility as liability. The neuroscientist Björn Merker, in an influential paper, frames this as a cost: moving is good, but it brings "liabilities", one being that the world becomes more confusing because your motion contaminates your senses.
Godfrey-Smith: mobility as opportunity. The fact that actions affect the senses is also a resource. Through action you can probe the world and create new stimuli, "poke and interfere and make the world talk back to you". The shrimp's feelers do exactly that. Action brings the senses both difficulties and insights.
Sensing that marks the relation between self and other is a central feature of the animal way of life. It establishes a point of view, a perspective, in a new sense.
The standard terms (not used in the book) are reafference (input caused by your own action), exafference (input caused by the world) and efference copy or corollary discharge, the internal copy of the motor command used for the prediction, concepts introduced by Erich von Holst and Horst Mittelstaedt and by Roger Sperry in 1950.
Why you cannot tickle yourself. Run your own fingers along your ribs and almost nothing happens; let someone else do the identical motion and you squirm. The physical stimulus is the same; the difference is that your brain predicted your own touch and cancelled much of it. Now imagine losing that prediction: every step would feel like a shove, every eye movement like the room lurching. The thought experiment shows that having a stable world at all depends on keeping books on yourself, which is the shrimp's achievement in miniature.
Different Senses, Different Entanglements
| Sense | Effect of your own action | Consequence |
|---|---|---|
| Vision | Immediate and strong: a slight head movement shifts the whole visual field | Self-motion must be registered or the world would lurch |
| Touch | Immediate and strong | Same; probing by touch is action-driven |
| Hearing | Real but slight; moving the head barely shifts the auditory scene | Less self/other bookkeeping needed |
| Smell and taste | Intermediate, between the two poles | Partly entangled with action |
Timing is uncertain, but the Cambrian probably mattered, since eyes and new ways of moving date from then and some trilobites had antennae. The banded shrimp is a very three-dimensional object, with feelers and pincers on long arms taking up space and serving as the centre of probing and manipulation. Godfrey-Smith's guess is that its experience would be of a very spatialised world, including a sharp sense of exactly what is it and what is not (recall the released leg).
Cleaners stand out here because their work engages the most complex environmental factor of all, other agents. The mirror test asks whether an animal treats its reflection as itself, for instance by cleaning a mark visible only in the mirror. Few animals pass; the only one reported to pass a version of it that is neither mammal nor bird is a cleaner fish.
The cleaner wrasse result has been extended, and argued over. Masanori Kohda and colleagues reported in 2019 (PNAS) that cleaner wrasse (Labroides dimidiatus) scrape at coloured marks seen only in a mirror, and in follow-up work, notably a 2023 PNAS paper, that they appear to recognise their own face rather than relying on body movement. Critics, including Gordon Gallup, who devised the mirror test, argue the scraping could be a response to an apparent parasite rather than self-recognition, so the interpretation remains disputed.
An Inquisitive Crab: Evidence for Pain
Crustaceans, active, sharp-sensed and often long-lived, have been treated as little robots, an impression helped by their shells. The most important work against that view comes from Robert Elwood of Queen's University Belfast and collaborators. The point is not that the animals flinch, but that their behaviour suggests more than a reflex: something like pain being felt.
- Nociception ("no-see-ception"). Detecting damage and responding to it. Very common in animals and readily read as reflex, so biologists treat it as not sufficient for pain.
- Wound tending and protection. Attending to the injured part. Elwood's group found shrimp grooming an antenna touched with vinegar or bleach and rubbing it against the tank wall.
- Seeking pain-killers. Seeking out analgesic chemicals, often the same drugs that work in us.
- Learning from consequences. Certain kinds of learning from the good and bad outcomes of actions.
- Motivational trade-offs. Weighing the badness of a stimulus against other costs and benefits, which a reflex cannot do.
Elwood's hermit crab shocks. Hermit crabs live in empty snail shells, carrying them like movable armour. Small electric shocks can make a crab abandon its shell, which alone proves little. But: - Shell quality. Crabs in a better shell tolerated more shock before leaving. - Predator odour. With the smell of a predator present, crabs were less inclined to leave, tolerating shocks that would otherwise evict them. - After-inspection. Shocked crabs that left sometimes carefully inspected the shell afterwards, apparently looking for the source of the trouble.
The shock is bad, but it is factored into a decision alongside other considerations, the signature of a felt state rather than a reflex.
Two arguments and their answers:
- "The tests are no good if a shrimp passes." Elwood uses tests accepted as good evidence of pain in vertebrates. A critic can reply that a crustacean passing shows the test is flawed; there is no knockdown argument against that, but without further reasons it is ad hoc, "a mere attempt to avoid changing one's mind."
- "Crustaceans lack the brain areas of human pain." A bad argument. Crustaceans also lack visual areas anything like ours and plainly see. Evolution often builds different structures for the same function; that is clear for vision and probable for pain.
Godfrey-Smith believes this was the first work on any invertebrate to give reasonably good evidence for pain. It is not decisive, as Elwood acknowledges, but it makes a real case. Meanwhile crustacean welfare in most countries is "basically at zero": nothing done to them causes concern, and boiling alive is routine.
Policy caught up with Elwood's evidence in some places. Switzerland banned boiling lobsters alive without prior stunning in 2018. The 2021 LSE review for the UK government, led by Jonathan Birch, assessed some 300 studies against eight sentience criteria and found strong scientific evidence of sentience in both decapod crustaceans (crabs, lobsters, crayfish, shrimp) and cephalopods, with the strongest case for octopuses; the Animal Welfare (Sentience) Act 2022 then included decapods and cephalopods. The 2024 New York Declaration on Animal Consciousness lists decapod crustaceans among the animals for which conscious experience is a realistic possibility.
The Hermit Crab Who Fell
While photographing a sponge-encased ascidian, a purple moon in dark space, Godfrey-Smith heard a clatter: an orange-sized hermit crab tumbled head-over-heels from the ledge above, almost certainly after overbalancing while watching him. When he picked her up and moved her into the open, bright-orange streamers poured out: acontia, stinging threads emitted not by the crab but by anemones riding on her shell.
- Borrowed weapons. Some hermit crabs pick up anemones with their claws and arrange them on their shells as protection, especially against octopuses.
- Context-sensitive dressing. In some species, octopus odour prompts an underdressed crab to pick up anemones, and dominant individuals take anemones from other crabs' shells.
She fled back under the ledge as far as the shell allowed, stalked eyes glaring out.
The Insect Jolt
Stepping out of a dive among these engaged crustaceans, Godfrey-Smith felt a gestalt shift. Crabs and shrimp act at our pace and scale with recognisable agendas. Insects probably evolved from within a large "pancrustacean" group, and on land we kill them in colossal numbers without a thought, as unfeeling robots. The conclusion is not automatic (land life took insects down another road, and they are mostly smaller and less recognisable), but insects do not have simpler brains than crustaceans; many have considerably more complicated ones. Once crabs and shrimp are seen as experiencing animals, insects must be taken seriously as candidates for subjective experience.
Insect evidence has grown since 2020. In 2022 Matilda Gibbons, Lars Chittka and colleagues (PNAS) showed bumblebees trading off heat avoidance against sugar reward: they accepted noxious heat for richer sugar and avoided it when the reward was poor, the same motivational trade-off criterion Elwood used for hermit crabs. Also in 2022, Chittka's lab reported bumblebees rolling small balls with no reward, which they interpreted as play. The 2024 New York Declaration names insects alongside decapods and cephalopods.
Another Path: Modular Bodies
Alongside the road toward unified action, other animals (and plants) took a different path. A small organism that would benefit from being larger has two options:
Unitary. Keep one shape and build it bigger, solving the new problems of circulation and coordination. Humans, shrimp and octopuses have definite forms repeated across generations, not built from partly self-sufficient units. A standard body lets routines of action evolve and nervous systems fine-tune the same coordinated motions generation after generation.
Modular. Repeat the form you have, adding attached twins again and again: a tight colony of repeated units, as in corals and largely in plants. Units keep some autonomy (they can often reproduce individually), it is unclear what counts as the individual, and branching makes the overall form only partly predictable (an oak has no fixed number of branches). Behaviour stays simple or becomes simpler.
- Modular means stationary. A few modular animals drift-swim, but when usually modular animals move seriously they turn unitary: swimming anemones, sometimes good swimmers, are one big polyp. Modular animals cannot generate complicated actions as wholes, which makes them like plants.
- Plants are not inert. They sense and respond but mostly use these abilities to form the body itself: a plant's shape records what it has sensed (where it found sun). A less integrated body can be more variable, freely fitted to circumstances.
- Bryozoans ("moss animals"). The emphatic case. Fairly close relatives of molluscs, with a bilateral form and a nervous system, they nonetheless turned firmly toward plant-like colonies of underwater bushes and mosses, after a long time on the same evolutionary road as ants and octopuses.
Ghosts in the spaghetti. Tom Davis showed Godfrey-Smith a "spaghetti bryozoan", a tangle of translucent threads like glass noodles, with tiny nudibranchs on it. Back at the computer, zooming in, he saw a red stripe on a talon-like claw attached by an unmistakable hinge to what had looked like another stem: an arthropod. These were skeleton shrimp, gaunt, near-transparent climbers with needle-sharp claws, constantly moving through the bryozoan. On later dives he found them everywhere, usually only afterwards in photographs, "like the crowded tiny ghosts of long-dead generations". The stationary stems and the taloned climbers were both animals with muscles and nerves, on opposite evolutionary roads.
The Decorator: Two Roads on One Body
The road so far: single-celled eukaryotes start integrating into a new living unit; nerves and muscle create multicellular action; the first bilaterians arise and branch, all before the Cambrian and mostly before the first animal fossils; then the Cambrian arms race of action and sensing, with arthropods pioneering actions scaffolded by sculpted, standardised bodies and jointed limbs.
The tree shows how much hangs on the first bilaterians, possibly flatworm-like: up to that point your history was shared with the ant, the crab and the octopus; after it the lines evolved independently.
The sponge that stood up. On the pipeline, a dark-red sponge with flat "Dr. Seuss fingers" sat utterly still, then creaked into motion: a decorator crab. Decorators encourage sponge, and often corals and other cnidarians, to grow on their shells, sometimes with specialised hooks to hold it, going far beyond hermit crabs' added anemones. The living garden is stationary and inedible to most predators, giving concealment and discouragement, especially against octopuses. The price is ponderous movement, since sponge tissue clogs the joints. A claw appeared, the head slowly rose: one animal "embodying two very different products of animal evolution", the unitary actor and the modular grower.
Goodbye: Recognition and Decline
About two weeks later, past an aggressive octopus guarding her den, Godfrey-Smith found the one-armed shrimp still under the ledge, the shark gone. It was less active, its body gathering algae, but eventually came down to stare and wave.
- Up to 5 years: listed lifespan of banded shrimp in aquaria; they are long-lived, territorial and monogamous, mating for life.
- 1977: a study separating mated pairs for a night or two and reuniting them either with their mate or with a same-sex stranger of similar size and appearance. Strangers both courted more and fought more; reunited mates resumed normal behaviour, so the shrimp recognise individuals, most likely by chemical cues, the paper concluded.
- Within antenna's reach: where wild pairs stay all day; males especially wander a few metres at night and return before dawn.
- About one square metre: the territory a pair may occupy for much of its life.
Females are larger, but otherwise the sexes are hard to tell apart. Godfrey-Smith's earlier video of the original pair showed intense face-to-face fussing with feelers; he wonders whether that was signalling or grooming, and whether the lone shrimp's algae-covered body reflects a lost partner who used to groom it.
Two weeks later again (a three-hour drive "to visit a shrimp"), the water was dark and close and the ascidians "coughed and sneezed". A shrimp of the right size sat in the exact spot, now missing both long arms. It was still feeding with its smaller sub-claws, but slowly: tired, alone, and probably near the end of its days.
Factual slips in the chapter. The genus is Ancylomenes, not "Ancyclomenes", and the Cambrian chordate in the figure is Pikaia, not "Pikaea" (the picture section spells the shrimp correctly). The claim that the largest arthropod ever found is the filter-feeding anomalocarid (Aegirocassis, about two metres, actually Ordovician, not Cambrian) is contestable on two counts: anomalocarids are usually placed as stem-group arthropods, and the giant sea scorpion Jaekelopterus and the millipede relative Arthropleura reached comparable or greater lengths. None of this touches the argument. The bigger open question is the self/other inference: Godfrey-Smith admits no one has studied banded shrimp antennae, so the shrimp's "spatialised experience" is an informed guess built from crayfish and fly work.
Subtract your own footprint before reading the signal. A bank's monitoring systems see customer behaviour that the bank itself caused: a marketing campaign lifts card spend, a rate change triggers repayments, an app release changes login patterns. A fraud or anomaly model that cannot predict and discount these self-caused changes behaves like an animal without efference copy, flinching at its own movement or, worse, learning that its own actions are threats. Feeding a register of planned interventions (campaigns, releases, policy changes) into the baseline is the corollary discharge of an enterprise.
The Origin of Subjects
Animal evolution produced a new kind of self: a body tied together by nerves and muscle, sensing and acting from a point of view, a nexus where causal lines converge, diverge and loop back. Godfrey-Smith argues that the evolution of animal agency brings with it the origin of subjects, then pushes back against three habits that make the mind-body gap look wider than it is: zombie-style thought experiments, the fixation on qualia as colour patches, and the assumption that all experience is sensory. Moods, energy, resolve and above all a feeling of presence show that "more of life is felt."
Subjects, Agents, Selves
Taking stock at the Cambrian, Godfrey-Smith says the story so far has produced not merely bigger, more complicated organisms but a new kind of being: organisms tied together in new ways and engaged with their environments in new ways. The ingredients:
- New action. The coordinated motion of millions of cells, made possible by muscle and nervous systems.
- New sensing to guide it.
- A tacit sense of self. Working together, these yield beings that treat the environment as a world beyond themselves, acting on an implicit self/other line (the shrimp that let go of its own leg).
All cellular life shows something like subjectivity and agency, but in animals these take a new form. Philosophers treat subjectivity as the hard problem and agency as the easier one; the book's argument is that the two are entangled and cannot be solved separately.
The back-to-back picture. Susan Hurley's name for a standard but misleading model: a person is "a subject and an agent standing, so to speak, back to back." The world acts on the person through the senses (subject side); the person works out a response and acts (agent side). The person becomes a layered object flanked by the world, almost two people.
Subjects-and-agents together. Hurley's alternative, which Godfrey-Smith adopts: the roles are not separate parts. What you do shapes what you sense, and what you sense shapes what you do, continuously; one entity is both at once.
The Animal as Nexus
Another way to say what arose: animals became a new kind of intersection point in the world's causal networks.
- Convergence. Picking up information through the senses, the animal is a point where causal lines converge.
- Divergence. As an initiator of action it is a point from which lines spread downstream, often looping back to affect its own senses.
- Past meets present. Handling of the present is shaped by what was seen before and how yesterday's actions turned out.
These features are common (not invariable) products of the animal way of life. Animals of this kind have a point of view, and from it they act. Some of what falls under the troubled label "subjective experience" is an expected, comprehensible consequence of animal evolution: roughly, the evolution of animal agency brings with it the origin of subjects.
Qualia and Other Puzzles
The chapter's method is reaching out and pushing back: show real progress across the gap, and remove misconceptions that make the gap look bigger. Part of the arbitrariness of verdicts on animal experience (feelings for paramecia, darkness in fish) is missing data about particular creatures. But part is an unmooring of imagination, the same one behind classic arguments against materialism.
Descartes (17th century), the disembodied mind. 1. I can doubt that I have a body (it might be an illusion). 2. I cannot doubt, right now, that I have a mind. 3. So mind and body are not the same: I cannot be merely a body if that body is optional to my existence.
The zombie reversal (David Chalmers's term). 1. I can conceive of an exact physical duplicate of a person with no experience at all, a "zombie". 2. If materialism were true, a physical duplicate would have to have experience. 3. So materialism is false: mind may be reliably produced by bodies and brains but cannot just be them.
Godfrey-Smith grants that mind and body appear separable, but traces the appearance to a quirk of imagination, diagnosed by Thomas Nagel even though Nagel was himself a critic of materialism.
- Perceptual imagining. Imagining seeing, hearing or touching something; applies most readily to bodies, objects of the senses.
- Sympathetic imagining. Imagining being something; applies only to things we take to have experience.
- Free recombination. In thought experiments the two can be combined, separated and rearranged at will: leave the perceptual side blank and you "imagine" a soul without a body; leave the sympathetic side blank and you "imagine" a body without a soul.
- The moral. That we can do this shows nothing about what is really separable. Nagel thought the mental probably is inexplicable in materialist terms, but rightly denied that these thought experiments could prove it.
The same unmooring appears in milder form when someone bestows inner life on a paramecium or imagines a fish dark inside: free imposition of sympathetic imagination, uninformative either way. Imagination still has a role (the book repeatedly tries to imagine animals' lives), and the goal is bridging concepts that let such imagining be done with some rigour.
The zombie shrimp. Picture the one-armed shrimp exactly as it was: feeler stroked, climbing down, turning its eyes on you. Now "subtract" its experience while leaving every molecule and every movement in place. It feels easy. Now picture the same shrimp with experience but with every molecule rearranged at random: impossible, because nothing would be left to have the experience. The asymmetry hints at what Nagel diagnosed: subtracting experience is easy only because sympathetic imagination is a separate faculty we can switch off at will, not because experience is a separate thing.
Putting Qualia in Their Place
Even granting the evolutionary story, some philosophers say it misses the hardest problem: the intrinsic qualities of experience, the redness of red, the sound of a clarinet, the "raw feel" of a green forest. The standard and notorious term is qualia. Daniel Dennett argues the whole idea is an illusion, a position others find "literally mad". Godfrey-Smith takes a middle line: some qualia problems are real, but they can be made less radioactive biologically.
- The real challenge is scientific. When someone sees a tomato, their biological activity has its own physical intrinsic nature, which gives the experience its feel for the person who is that activity. Saying why brain processes of one kind feel like red and not something else is a genuine scientific task.
- An unreasonable demand. A scientific description cannot contain an experience; knowing about an experience is different from having it. Some critiques want a third-person description to "magically become first-person".
- No conjuring. The materialist is not supposed to describe processes so that reds and cymbals appear in the system being described. Qualia are not extra items the system produces; they are part of what it is to be that system.
Experience is the first-person point of view of a complex living system of a certain kind, not something conjured up by the workings of that system.
Rothko Experiences and the "Where" of Seeing
The second overreach: fixating on the redness of red until experience itself looks like a procession of colour and sound patches. Godfrey-Smith calls the paradigm "Rothko experiences", after Mark Rothko's colour-field paintings, and means the label as diagnosis. Rothko's canvases feel as they do because of their anomalous relation to ordinary seeing: normal vision involves probing and making sense of a spatial layout, while a colour field seems neither within us nor without. That disembodied character is part of their allure, and part of why they are a bad model of experience.
He builds on Susan Hurley (who died at 52) and her use of the vision-science distinction between "what" systems (shape, colour) and "where" systems (spatial arrangement), routed somewhat differently in the brain, though only roughly (shape involves spatial relations between parts).
| Feature | "Where" side | "What" side (colour) |
|---|---|---|
| Tie to action | Constantly updated by your own movement | Not usually tied to action the same way |
| Cross-check | Can be checked by touch | Cannot commonly be checked by touch |
| Role | Tracks you as one object among moving others; self versus world | Helps track contrasts and shapes |
| In a colour field | Has almost nothing to work on | Dominates |
Hurley held that "where" processing is "basic to having a unified perspective or a point of view as a perceiver and agent, a sense of being a self present in the world." Colour-field experiences are real, but far less paradigmatic than philosophers assume.
How Qualia Became Central, and the Pendulum
- 17th to 19th centuries The empiricists (John Locke, George Berkeley, David Hume, J. S. Mill) build the mind from "simple ideas" and "impressions": pure sensations, colour patches and brief sounds, mental atoms. They serve a dual role: describing the mind and founding knowledge by reconstructing it as patterns in sensation, clearing away dogma.
- Late 18th century onward A revolt: sensation-based views make the mind passive. German idealism swings to the other extreme, the primacy of self-determining, autonomous consciousness.
- Early 20th century "Sense data" play the same dual role in English-speaking philosophy.
- Today No simple ideas or sense data, but their shape lives on in qualia. On the other side, enactivism treats perception as action: "seeing is a way of acting", experiences are "things we do".
Enactivism exploits the real feedback between doing and sensing but uses it to move all of sensing onto the action side, denying the receptive side of the mind, which is real. John Dewey noted wryly that a philosophy student goes from being startled by the huge gaps between rival positions (everything changes; change is an illusion) to moving these "ever-toppling pieces" around the chessboard casually. Stark, stylised oppositions are a pathology of the field, though often how new ideas arrive. What seems strangely hard to state plainly is the obvious middle: traffic between living systems and the world, sensing and acting coexisting in a two-way process.
Beyond the Senses
The next overreach is the idea that experience is all sensory. Such philosophers allow inner senses (hunger, fever) as well as outer ones, and use "perception" almost interchangeably with "sensing".
- Jesse Prinz (a former colleague in New York): "All consciousness is perceptual."
- Fred Dretske (who influenced Godfrey-Smith as a student and was his colleague at Stanford late in Dretske's career), more hedged: the "clearest and most compelling" cases are "sensory experience and belief".
Godfrey-Smith's reply is blunt: emotions, willings, moods and urges are at least as clear and compelling, and far clearer than beliefs. The perceptualist reads a mood as perception of hormonal or bodily states. The neglected alternative: a mood is not a presentation of some fact; it is just the way things are with you at that time.
Fatigue at the wheel. Driving, you feel your energy drift toward mental fatigue: heaviness, dullness, thought becoming slow and effortful rather than agile. This is clearly part of experience. Is it a fuel gauge showing "low", a presentation of a bodily state? Or is your experience saturated by a foggy, heavy way of being? Godfrey-Smith takes the second: energy level is an aspect of living activity that is felt. Surges of resolve point the same way: experience is not solely a matter of being told something. "More of life is felt."
Basal Consciousness
John Searle asks you to imagine waking from dreamless sleep in a completely dark room: no coherent thought, almost no stimulus beyond the bed and covers, yet a clear difference from the unconsciousness before. He calls this "basal or background consciousness".
Consciousness as presentation. Most psychology and philosophy treat consciousness as a way information is presented in the mind, so there must always be some content: the start of an inner monologue, slight hunger. On this reading Searle's case is never truly empty.
Consciousness as a state of being. Searle's passage suggests consciousness need have nothing going on in it. Some neuroscientists, notably Rodolfo Llinás, hold similar views: consciousness usually reflects input from the senses but does not depend on it.
Godfrey-Smith offers a third description of waking in the dark room: you once again feel present.
Presence
The sense of presence, the feeling of being real and present in a scene, sits on the margins of current theory, sometimes as a mere evocative word. Its value shows best by contrast.
- Transparency. Paired with the all-experience-is-perception view: in experience we are never aware of our self or our experiencing, only of what is put before us; we see the world, our bodies included, through experience. Some writing on meditation claims to reveal the absence of the self. If right, consciousness is no more than pointing or representation.
- Presence. Pushes back against this self-erasing tendency: the self is rarely the topic of everyday experience, but it does not vanish. For at least some people, feeling present in a scene is an important part of experience.
- Godfrey-Smith's "balance". The experiences he finds most telling hold a balance between a feeling of one's own presence and a taking-in of the surroundings: not introspective or self-absorbed, not dissolved into the scene. It arises in some meditation contexts: a scene, plus the feeling of being part of it. It corrects a perennial tendency to overinflate the autonomous self or erase it.
Is Presence Free with Life?
The tempting thought: presence is an automatic feature of being a living organism in the world, so a basic experience comes "for free" to anything alive. Godfrey-Smith thinks this is probably too simple. Presence seems to depend on complicated, continuous behind-the-scenes processing, including monitoring of the body and how bodily events relate to surroundings, which shows itself when it fails.
- The rubber hand illusion. Watch a fake hand being stroked with a brush in exact time with strokes on your hidden real hand, and you come to feel the fake hand is yours and the seen strokes are felt there.
- The wider iceberg. Through brain injury, and to a degree in experiments, people feel themselves somewhere other than their body: full out-of-body experiences, or half-seeing and half-occupying a body projected into the visual field. "Just about every possible distortion is found."
- Psychiatry. A "feeling of unreality" can be an important warning symptom.
So presence is more than an automatic consequence of being alive. One could reply that it exists in both a primordial form and processed forms; Godfrey-Smith sees no reason yet to believe this: "Presence feels primordial, but that is not a reliable guide." Nor is presence essential to experience, since a feeling of unreality is still a feeling. But it is near the heart of what makes experience seem special: a sense of thereness is an important part of subjectivity itself, and understanding the brain and body activities that produce it takes us further toward fitting experience into the biological world. A biological account of vision seems to leave something out partly because seeing normally comes with this soft background feeling of presence.
The Rothko room versus the reef. Stand in a gallery before a vast maroon colour field, motionless; then imagine hanging in clear water over a reef, adjusting your fins as fish move past. Both are visual experiences. In the gallery, the "where" system idles, your movement changes nothing, and you may feel strangely unplaced. On the reef, every kick shifts the scene, you track yourself among moving things, and you feel there. If philosophers had taken the reef rather than the gallery as their model of seeing, the "hard problem" of vision would include, from the start, self-location, action and presence, the parts biology can explain.
Taking Stock: Godfrey-Smith's Framework So Far
- Tying together. Evolution shaped animals into unified centres of agency and subjectivity, with an overall state of living activity modulated by sensing. The formation of these units is part of the origin of experience.
- Self versus other. Sensing and acting combine to give an implicit (or explicit) sense of self versus world, the shrimp's feeler problem solved.
- The point of view. The animal as nexus: convergent input, divergent action, loops through the world, past meeting present.
- Integrated seeing. In human vision these come together as tracking what confronts us, where things are, self-and-world, bodily ownership and presence. Given this, "it is not mysterious that seeing feels like something, rather than being a mere taking-in of information, as would occur in a camera."
- More of life is felt. Moods, energy, resolve and urges are not perceptions of something; they are felt aspects of living activity. The open question: how many processes usually labelled unconscious subtly shape felt experience?
- Three habits to drop. Free-floating separability thought experiments; Rothko-style qualia as the model of all experience; the perception-only view and its partner, transparency.
"The story is far from complete at this stage, but this is a start."
Body-first theories of consciousness have gained ground since 2020. Two widely read books push in Godfrey-Smith's direction. Anil Seth's Being You (2021) argues that the self is grounded in the brain's predictive regulation of the living body (his "beast machine" view), so that moods, emotions and the basic feeling of being a self are experiences of bodily regulation, not detections of external facts. Mark Solms's The Hidden Spring (2021) locates the source of consciousness in brainstem arousal and affect rather than the cortex, close to the "state of being" view of Searle and Llinás. Both remain contested, but they make "more of life is felt" a mainstream research position rather than a philosopher's aside.
Where the chapter is thin. The pushback on qualia is persuasive against the demand that a third-person description become first-person, but committed defenders of the hard problem (Chalmers, for instance) would say they never asked for that; they ask why a given physical process has any first-person side at all, and the identity claim ("experience is the first-person point of view of a living system") restates rather than answers that. The treatment of enactivism is also brisk: Godfrey-Smith targets its strongest slogans, while more moderate enactivists accept a receptive side. And presence, by his own admission, is only partly characterised; he declines to say whether simple animals have it.
Hurley's back-to-back picture is how most AI agents are built. A typical agent pipeline perceives (retrieve, parse), then plans, then acts, with a hard seam between the subject half and the agent half. Godfrey-Smith and Hurley's point is that capable living systems instead interleave them: action is used to sense (probing a document, querying a system to see how it responds), and sensing is continually corrected by what the system knows it just did. For enterprise agents this argues for closed-loop designs that log their own actions and read the world's response against them, rather than one-shot perceive-then-act chains.
Night Dive
Godfrey-Smith and Tom Davis walk into the bay an hour after sunset. Thirty feet down, with lights off, the darkness is total: land nocturnals usually have some light, but these animals live in scents, tastes and touches. Tom is surveying rare tropical fish that shelter under ledges at night, hard to find among sleeping locals, big wrasses and morwongs stacked like bunk beds.
- The secret order. By day, decorator crabs wear mostly sponge; at night some wear soft corals, "a covert crustacean order" seen only after dark. One marches over the reef, its coral hands half open, half clenched.
- Anemones on carriages. Two hermit crabs fuss face-to-face, one shell plastered with anemones, the other with a few. The anemones, made mobile by their crabs, touch the other shell and pull back: stationary animals encountering surprises because an arthropod carries them.
The closing image gathers the chapter's themes: unitary actors, modular passengers and sleeping fish, many creatures afoot in a dark expanse, each a point of view of a different kind.
The Octopus
The octopus is the clearest case of a conscious invertebrate, and the hardest test for the idea that experience depends on integration. Its nervous system is decentralised, with about two-thirds of its neurons in the arms, and its behaviour mixes central command with local exploration by the arms. Godfrey-Smith uses split-brain patients and the Wada test to argue that a mind is a pattern of activity that can expand, contract and divide quickly, and proposes that the octopus switches in real time between unified action and a looser state in which its arms carry a glimmer of their own agency.
Rampage
At slack tide on a soft-coral site at Nelson Bay, Godfrey-Smith follows a medium-sized octopus (body about softball size) on a nearly hour-long hunt. He calls her "she" because her third right arm, the arm males carry the mating duct on and tend to protect, was moving as freely as the rest. Her method is the eight-armed bear hug: she envelops a clump of coral or sponge and sees who flees.
- Unworried. Decorator crabs in sponge cloaks were touched and ignored; seahorses floated up "like porcelain birds" and were left alone.
- Worried. Small, well-hidden octopuses scurried or jetted away; flatfish burst out of the sand.
- Most alarmed. Unclothed crabs of other kinds, which she chased hard; one swam up into the diver's gear with the octopus in pursuit.
She ended by wedging herself under a pile of shells, success unknown. A modest octopus dwarfing and scattering everything around her recalls, on a small scale, the time when cephalopods ruled the seas.
When Cephalopods Ruled
Octopuses are molluscs (with snails, clams, oysters), in the subgroup cephalopods with cuttlefish, squid and a few curiosities. Molluscs date perhaps from the Ediacaran, certainly the Cambrian: soft-bodied, mostly protected by a mineralised shell. An unlikely body for complex behaviour, yet soon after the Cambrian some cephalopods lifted off the seafloor, using the shell for buoyancy and a ring of tentacles around the mouth as a new tool.
- Ordovician Shelled cephalopods up to eighteen feet long become the largest predators, displacing arthropods; they flourish for several hundred million years and leave one survivor, the nautilus.
- Mesozoic The coleoids, an early offshoot, internalise the shell, trading protection for a buoyancy device; some lineages then abandon it.
- About 100 million years ago One group goes all the way, with no shell inside or out and almost no hard parts: the octopus.
Among all the animals descended from the bilaterian ancestor of perhaps 600 million years ago, only three groups produced species with large nervous systems and complex behaviour: cephalopods, arthropods and vertebrates, or, in the opening scene, the rampaging octopus, the scattering crabs and the gear-festooned human following.
How Simple Was the Common Ancestor?
The usual picture is something like a flatworm. Nick Strausfeld and Gabriella Wolff have argued from subtle design similarities between insect and vertebrate brains that a small "executive brain", a modest cluster of controlling machinery at the front, existed in the early bilaterians. Even so, Strausfeld accepts that molluscs then threw it away: sheltered in shells, they had little use for an active-life nervous system. Cephalopods rebuilt a complex brain much later, on a different design, though with a largely shared chemical toolkit. In 2018, octopuses given MDMA (ecstasy) became more friendly and gregarious, much as people do (Eric Edsinger and Gül Dölen's study of conserved serotonin signalling). The chemistry is conserved; the layout is profoundly different.
The arthropod body: a Swiss Army knife. Built on hard parts that shape and scaffold action: springs, spatulas, claws, clubs. Much effective behaviour is available, but the body is closed: each part does what it does and no more.
The octopus body: openness. It can grasp and manipulate almost anything, double the length of its arms, flatten itself like a pancake. Its degrees of freedom are almost countless, which is both an opportunity and a control problem.
Lines of Control
The octopus nervous system is decentralised: about two-thirds of its neurons are not in the brain (itself a vaguely bounded region) but in the arms, especially their upper parts. These are not just relays feeding the brain; control of some motions seems to be delegated to the arms. Roger Hanlon and John Messenger proposed decades ago that this design exists because the body is so hard to control. Godfrey-Smith adds the flip side: once you grow eight free arms around your mouth, why not fill them with sensors and let them go their own way to some extent?
- Early view. Central control over the arms seemed limited; the octopus might not even know where its arms are. The evidence: poor performance on artificial lab tasks and slim nerve connections between arms and brain.
- Hochner lab, Jerusalem. Tamar Gutnick and colleagues showed an octopus can use vision to steer a single arm along a novel path, out of the water and back, to food, which implies real central control, though observers had the impression of some local exploration by the arm en route.
- No body map. The same lab's neurobiology suggests octopuses lack a map of their own body in the brain of the kind we have. If an octopus senses its own configuration, it does so quite differently from us.
Decentralisation Is an Evolutionary Option
Decentralised control is not unique to the octopus. Sensing and acting need many organised arrays of cells, and once that machinery exists, evolution has a choice: separate streams controlling different actions, one integrated stream, or mostly separate tracks with cross-talk. This matters because "point of view", the book's bridge to subjectivity, suggests a lot of integration, and many animals are only partially integrated.
What separation buys. Often speed; local responses without waiting for a central verdict.
What separation costs. Separate sensory streams cannot combine information that would be useful together, like premises of an argument. Separately acting parts may want contradictory things. In the extreme you fragment into sub-agents with their own scenes and decisions. Still, nothing guarantees every animal is shaped like a CEO with underlings.
Lateralisation: The Divided Brain
Bilaterians have paired parts, including much of the brain. In vertebrates the wiring crosses over: the right visual field is processed by the left hemisphere, and the hemispheres are linked by a thick fibre bridge. Bilaterian invertebrates pair brain parts but do not cross them: the octopus has a large optic lobe behind each eye, serving its own side.
- Vertebrate division of labour. Many vertebrates prefer the left eye (right brain) for social interactions with their own species and the right eye for food.
- Cuttlefish. The right eye is preferred for feeding, the left for predators; with no crossing, left eye means left brain. A similar division of roles on a quite different brain design.
- Split-brain patients. The radical human version, produced by surgery: severing the hemispheric bridge to stop epileptic seizures spreading. Such people sometimes seem to have two minds in one skull and otherwise behave normally.
Octopus Watching: Octopolis and Octlantis
For about a decade Godfrey-Smith has watched Octopus tetricus (the gloomy octopus) at two sites on the Australian coast, about six hours' drive south of Nelson Bay.
Octopolis. Found by chance in 2008 by Matt Lawrence on a sandy plain full of scallops: a small area of thousands of empty scallop shells and about a dozen octopuses, sixteen at the most. Both sexes, all sizes, from matchbox to football-sized bodies with arms several feet long, though they differ in age by little more than a year, because octopuses live only one to two years. The site has unlimited food but many predators (sharks, seals, dolphins, aggressive fish) and silty sand that makes poor dens: the dilemma of a life spent as both prey and predator.
The likely origin is positive feedback, probably seeded by a single human-made object, a foot-long piece of metal now mostly buried, dropped from a boat.
- The dens. Some are vertical shafts over 50 cm deep with straight, shell-lined sides, secure against everything except other octopuses: evictions and wrestling are frequent.
- Octlantis (2017). Divers Marty Hing and Kylie Brown found a second site nearby with similar numbers (up to fourteen, in three subregions), seeded by a couple of natural rocks. No human object was needed: Octopolis was not a one-off.
- The open question. These may be animals of a mostly solitary species learning, individually, to get along at unusual densities, or octopus life may contain more sociality than recognised (the team has listed known exceptions in twelve species).
The Social Repertoire
Living among others is the most complex thing in most animals' environments, and at these sites octopuses are well and truly surrounded. Godfrey-Smith's informal impressions:
- Monitoring. An arriving octopus passes others with a series of arm-raises, slaps and touches, sometimes escalating into sparring, often just a brief reach. These look aimed at tracking who is who, especially by sex.
- Real fights usually start when an octopus marches in from outside and a resident comes out to meet it.
- No deaths seen. Among countless fights, never a death or obvious serious injury; similar-sized octopuses find it hard to hurt each other (Miranda Mowbray's verdict on a video: a giant pillow fight, between pillows). Deaths elsewhere, apparently by strangulation, probably involved big size differences.
- Sex errors. Males exclude other males and seem to judge sex by touch, fallibly: males have slipped in beside females, and matings have dissolved, apparently when sex was initially misjudged.
- The Nosferatu display. An aggressive octopus stands very tall, arms spread, mantle pointing straight up, in intense dark colours (a whole-body colour change takes under a second).
Throwing
Octopuses gather shells, seaweed and silt in their arms, bring the siphon under the arm web, and release the load on a sudden jet, sometimes for several body lengths. David Scheel first suggested a social role. Godfrey-Smith's current reading:
- Mostly housekeeping. Most throws are den-building and den-cleaning; an octopus doing this while watching a neighbour will sometimes hit it by accident. Females throw more, but females also build better dens, as they must eventually brood eggs.
- Some are social. Females often throw at pestering males. One video shows a female throwing repeatedly at one male over a few hours: about half the throws hit, the rest missed because he ducked, and by the end he was ducking early as she loaded up.
- A plausible evolutionary path. Two existing behaviours, jetting debris from a den mouth and jetting water at annoying animals (divers included), combine; an accidental hit startles a pest, and the octopus may notice the effect.
The throwing study has been published. Godfrey-Smith, Scheel and colleagues reported the Octlantis throws in PLOS ONE in 2022 ("In the line of fire"). The analysis supported the book's reading: most throws served den maintenance, females threw more, and throws that hit other octopuses differed from the rest (more forceful, more often made with silt), and targeted animals sometimes ducked or raised arms in anticipation. Wild octopuses thus join the short list of animals that project material at others.
Building, a Sponge Helmet, and Eyes
- Engineers without a plan. Dens are built and tended with care, but there is no cooperative building, even by pairs, and the octopuses are presumably oblivious to the feedback that makes their city.
- The sponge helmet. A small octopus stared at an unmanned camera on a tripod, left, and came back with a piece of dead sponge, placing it over its den as a roof-cum-helmet and huddling beneath: the right size, rigid and light, the best material available.
- Good eyes. Octopuses spot incoming octopuses from considerable distance and react differently than to stingrays; they are not chronically nearsighted, as has been claimed.
- The head-bob. Watching something intently, they bob up and down and a bit sideways, which Jennifer Mather interprets as gaining depth information by changing the viewing angle (they mostly see an object with one eye). This is active information seeking and requires separating reafference (sensory change caused by one's own action) from exafference (change caused by the world).
Two Modes: Command and Exploration
Coordinated whole-body behaviour (jetting with arms gathered into a slender missile, crawling with arms everywhere, displays, throws) suggests substantial top-down control. The other side is ongoing exploration by the arms: a crawling octopus lets arms roam in many directions; a resting one sends a few out "like little eels" with delicate, questioning tips. He sees more of this at Nelson Bay, where octopuses are usually alone and relaxed, than at Octopolis, where social complexity and questions of sex keep them in a state of heightened attention.
The result is a picture of mixed control: a body partly steered by the brain, with parts that explore and react individually. Watching them produces gestalt shifts between seeing a whole animal using each arm as a tool and seeing an arm wandering in response to what it senses.
A natural Waterman? Ian Waterman lost all proprioception (the internal sense of limb position) through an infection at nineteen and had to learn to track his body by vision, a slow and hard recovery. Fred Keijzer suggested the octopus might be "a natural Waterman", given the evidence for no internal body map. Godfrey-Smith's reservation: if so, octopuses manage it effortlessly, and it would be surprising if an octopus could organise a complex arm motion only when it can see its arms; they look immediately cohesive when they need to act fast.
Not "Smart," but Sensitive
"Smart" suggests an intellectualised, ruminative style. Octopuses are better described as behaviourally complex and sensitive, with "an anarchic bodily embrace of novelty": they fiddle and turn a problem over physically, not mentally. Godfrey-Smith would look for true cleverness in two places:
- Object use. Tank escapes that may involve something close to planning, and protective use of shells, coconuts and sponges, which looks improvised in cases like the sponge helmet (though it might be an evolved anti-predator routine).
- Awareness of other agents. Octopuses and cuttlefish often make their move when you are not looking. Bret Grasse, who manages cephalopods at the Woods Hole Marine Biological Laboratory, was squirted by cuttlefish who waited until his back was turned; he turned, saw "innocent" cuttlefish at the bottom, then watched them with his phone camera behind his back: several came up and squirted him again.
Individual style also varies strikingly. A large male near the rampage site hauled himself out of his den, evicted one octopus and mated with another, all the while moving in a stylised way: blade-like flattened arms, arms wound over his head, an arm coiled into a wheel, for no apparent reason. "Everything he did was writ large."
Octopus and Shark
Octopuses never ruled their seas: they evolved long after the first cephalopod golden age, and fish have always been with them. At Nelson Bay, low on air, Godfrey-Smith found two octopuses in a ten-minute tug-of-war over a fish (probably a male outside and a female in the den; told neutrally as Octopus 1 and 2). Octopus 1 won.
The wobbegong attack. A three-foot wobbegong shark (camouflaged like an old bomber; the largest reach triple that length; they bite and hang on) cruised in. Octopus 1 retreated watchfully but stayed near. The shark plunged headfirst into the crevice, body near vertical, thrashing, after Octopus 2, and failed; then again, and failed. Octopus 2, injured and very still, did not ink but stayed low and held on. Turning to Octopus 1, the shark lunged; the octopus, unhurried, jetted to a safe distance with ease, apparently knowing it could escape once surprise was gone. The shark then lay behind a rock, head toward the octopuses, out of their sight, a possible deliberate ambush, but no octopus was tempted. Octopus 2 survived, "as octopuses had for millennia."
Integration and Experience
Ten years of watching leave Godfrey-Smith with "no real doubt" that octopuses experience their lives. This is not a reaction to complexity alone: even theories that attribute much behaviour to unconscious processing would put octopuses on the "yes" side, because they show attentive engagement with novelty, much non-routine behaviour, apparent moods (stressed, inquisitive, playful), and, at Octopolis, a large male hesitating over whom to chase while tracking several rivals at once.
The octopus is the clearest case of a conscious invertebrate, so either consciousness has at least two or three separate origins (us, octopuses, crabs and perhaps more) or, if it had a single origin, that origin was deep in time and took a very simple form.
The puzzle: integration has been central to recent thinking on consciousness and to Godfrey-Smith's own account of the self, yet the octopus is very complex but less integrated. What might its mixed control feel like from inside?
- 1: unity. The unusual wiring makes no big difference; the octopus usually behaves as a whole. Weakness: unified behaviour can arise from teamwork, as in army ants and honeybee superorganisms, so behaviour alone does not settle it.
- 1+8: a central self plus eight minor ones. The central brain is the primary self, with eight smaller, perhaps non-conscious, agents in the arms (discussed by Sidney Carls-Diamante).
- 1+1: two brains. Arm nerve cords are linked sideways near the arm bases, so the arm networks may form a second brain, larger than the central one. Frank Grasso raised this cautiously ("The Octopus with Two Brains"); perhaps two conscious streams. Its fullest exploration is fiction: Adrian Tchaikovsky's Children of Ruin splits each octopus into the Crown (central brain), the Reach (the arm network) and the Guise (the colour-changing skin, a part, not an agent), with behaviour as a dialogue between Crown and Reach.
- Zero. The odd organisation removes experience altogether; set aside, since the evidence for experience is so good. (The "1" might also be two, left and right.)
- Godfrey-Smith's choice: switching between 1 and 1+8. Suggested in outline by Jordan Taylor, then a student at the University of Pennsylvania, who asked at a talk: why couldn't the animal switch back and forth?
Detour: Split Brains and the Wada Test
To make switching credible, Godfrey-Smith turns to humans. After callosal surgery, sending different images to each visual field (each feeding the opposite hemisphere) produces odd results: speech is usually left-hemisphere, so the person talks only about the right visual field, while the right hemisphere answers through the left hand, pointing or drawing. Outside the lab, patients mostly seem ordinary, whole people. The explanandum is unity most of the time plus splitness in special contexts.
| Option | Claim | Main problem |
|---|---|---|
| One mind | Only one conscious agent, probably in the left hemisphere | The right hemisphere can be quite smart in experiments |
| Two minds | Both hemispheres conscious, always; unity comes from subtle coordination | Patients behave normally in most contexts |
| Fast switching | One mind most of the time, two when experiments separate the information flows | How can one mind exist across physically separated hemispheres? |
| Partial unity | No neat count; some experiences are shared, others partitioned | Long thought hard to make coherent |
Godfrey-Smith embraces both contentious options, fast switching and partial unity, together.
- Step 1: sometimes there really are two minds. Following Elizabeth Schechter, the best evidence is improvised communication in public space: the speechless right hemisphere has tried to pass answers to the left by tracing letters with a left-hand finger on the back of the right hand, prompting experimenters to say "Don't write!" That is smart behaviour, not reflex, so there is a mind on the right. Schechter thinks the duality is permanent; Godfrey-Smith thinks it need not be.
- Step 2: minds can come and go. If a mind is a pattern of activity, there is nothing dualistic about it appearing, transforming or contracting quickly after minor physical changes such as entering an experiment.
- Step 3: unity through the world. Susan Hurley proposed that some of the pathways unifying a conscious subject run out into the world and back, in tight loops of action and sensing. The basis of unified consciousness can be "a dynamic singularity" in "a field of causal flows" extending beyond the body. Adrian Downey recently showed that Hurley's view fits naturally with fast switching.
The Wada test. Devised by the Japanese-Canadian doctor Juhn Atsushi Wada, it finds which hemisphere controls language by anaesthetising one side at a time and testing speech. With the left side asleep, most patients cannot speak, yet they remain conscious at both stages. One patient: shown an object with the left side asleep, he had the tip-of-the-tongue feeling, "Only that was true for all words... I had no words." (Dolphins sleep one hemisphere at a time: a natural Wada procedure.)
Blackmon: each half was conscious all along. James Blackmon argues that if one half can be conscious while the other sleeps, and putting something separate to sleep cannot much change it, then each half was already conscious; normal waking consciousness is a blend of two (or many) smaller consciousnesses. (The most extreme version, many consciousnesses in one brain, is Semir Zeki's "disunity of consciousness".)
Godfrey-Smith: the basis contracts. Anaesthesia leaves the awake half physically similar but changes the overall pattern of activity in brain-plus-body. Before the test, only the whole brain supports the right pattern; during it, the physical basis for mind contracts into one half, which becomes a conscious system without having been one before.
A complication strengthens the point: only the cortex is split or put to sleep by halves. The lower brain stays whole, and its role in consciousness is now taken more seriously. So the switch is between two units that share some parts (half a cortex plus the whole lower brain). That shared lower brain is also where partial unity comes in: clinicians widely think moods and emotions are shared across the split, because a stimulus shown only to one side can shift the other side's mood without that side knowing why. A single stressed-ness can belong to "mind 1" and "mind 2", so they are not fully distinct minds.
- Fast switching between one mind and two, triggered by how information flows are separated.
- World-involving unity: the one mind, when present, is held together partly by causal loops that run outside the body (Hurley).
- Partial unity when split: moods and emotions stay shared through the unsplit lower brain, so the "two" are not wholly separate.
- Caveat: cases differ; some may be permanent two-mind states, still with partial unity.
Back to the Octopus: Switching in Real Time
Ordinary octopus life may contain a milder version of the same combination, without minds popping into existence.
- Unified mode. Throwing, jetting, fighting: the animal acts as one agent, and its experience reflects that.
- Loose mode. Arms wander and explore, and the central octopus may not "own" these locally guided motions. The arms are like very simple agents, sensing and responding.
- Do arms have experiences? "Probably saying too much." They have enough neurons (an entire bee has about a million; an octopus arm has tens of millions), but they probably never become selves of the right kind, because their networks are connected to the other arms and the centre. Arms have "a hint" of the basis of subjectivity.
- Partial unity again. Any glimmers in the arms would sit inside a shared overall state: stress, energy, arousal across the whole animal. (The first and second arm pairs are generally more active than the rest.)
- Pulling itself together. When the octopus imposes central control, arm autonomy is submerged.
Why This Is Not Just Human Inattention
We, too, suddenly attend to breathing or walking, or bring an absent-minded tune into focus. Godfrey-Smith thinks the octopus case differs in kind:
The human hand. A wandering hand touches something; information goes to the brain, and unless it is a reflex, the hand does nothing until the brain directs it. Attention overrides a central autopilot.
The octopus arm. Information goes to the brain, but the arm also produces its own locally controlled response, and its motions change what it senses next. Attention pulls together partly independent processes that, left alone, head toward agency of their own.
So the octopus may live a faint version of the split-brain drama: switching between whole-animal centredness and a glimmer of autonomy in the arms. The worry was that octopuses are not very integrated; the answer is that they become more and less integrated in real time. Two lessons generalise: fast switching is not extravagant (a mind that is a pattern of activity can come and go, expand and contract), and partial unity belongs in any framework for animals whose sensory streams are partly separate while moods, satiation and stress are shared.
A Protean Perspective
Even assuming one centre of experience, the octopus's experience is exotic:
- Changing states. Long periods sitting like a cat, perhaps near sleep; bursts of jetting, throwing and building; and intermediate states of centrally directed crawling with wandering detours, or two or three arms exploring in different directions while the body sits still.
- Taste by touch. Suckers are packed with chemical sensors: an arm touching your finger tastes it, and the difference between a thin glove and bare skin is huge to the octopus.
- Light-sensitive skin. The whole skin appears sensitive to light: not image-forming, but possibly registering intensity, changes, shadows and perhaps hues.
The upshot, if skin and sucker information reaches the brain as well as local networks, is an animal with a very expansive sensory surface that is, from the brain's point of view, unpredictable: as arms wander they reshape the body and meet surfaces and chemicals, in several arms at once. The octopus occupies a perspective, but "a protean and perhaps sometimes chaotic one"; imagining it puts Godfrey-Smith in "a rather hallucinogenic place."
The octopus with a sedated centre. Imagine an octopus-Wada test: briefly quiet the central brain while the arms stay awake. The arms would keep reaching, tasting and grasping, each answering what it touches. Is anyone home in the arms? On Blackmon's logic, each arm network must have been a small subject all along. On Godfrey-Smith's, the answer depends on whether the remaining pattern of activity has the right shape: eight loosely linked networks with no shared point of view would be activity without a self. The thought experiment shows why counting subjects is the wrong question: what matters is the organisation of activity at each moment, which can change faster than the anatomy.
Contested ground. The "no body map" claim rests mainly on a 2009 stimulation study finding no somatotopic layout in higher motor centres, which shows the map is not laid out as ours is, not that octopuses have no sense of body position. The 1+8 and switching pictures are, by Godfrey-Smith's own account, speculative and not yet testable. On split brains, his combination of fast switching and partial unity is a minority view; Schechter defends permanent duality, and a 2017 study by Yair Pinto and colleagues reported patients responding accurately to stimuli in either visual field with either hand, which they read as split perception with undivided consciousness, a result the chapter does not address. Smaller points: estimates of the largest Ordovician cephalopods come from fragmentary shells, and a 165-million-year-old French fossil once treated as an early octopus (in Other Minds) is now thought to have had an internal gladius, so it is closer to a vampire squid; Godfrey-Smith corrects this himself.
Agent architectures face the octopus's choice. Multi-agent AI systems decide between one orchestrator issuing every action and sub-agents with local autonomy. Godfrey-Smith's octopus suggests a third design: switchable control, where sub-agents explore and respond locally in calm conditions and a central controller pulls them together when a task needs coherence, with shared state (the octopus's mood) visible to all. In banking workflows this maps onto letting document-reading or monitoring agents act locally on routine items, while escalation to a single accountable controller is forced by risk signals, so that contradictory actions by "arms" cannot reach the customer.
Down Among the Stars
In midwinter Godfrey-Smith descends an anchor line to Octopolis with Matt Lawrence, ten years after its discovery. Four large wobbegongs (one at least eight feet) ring the site, noses in, like a perimeter; the octopuses keep their heads down. The divers wander off across grey sand and weed and find a field of feather stars (probably Antedon loveni): small, silver-white or dark purple, a few inches across, every few yards. A seal then rushes in, peering, nearly head-butting him, and is gone toward the light.
The feather stars mark a crossing in the book's route, from the protostomes to the deuterostomes.
- The bilaterian fork. Protostomes include arthropods and molluscs; deuterostomes include vertebrates and, surprisingly, echinoderms (starfish, urchins, sea cucumbers, crinoids).
- Echinoderms. Around since at least the Cambrian, beginning as flattened crawlers, then lopsided and spiral forms, then the star design, a gesture back to the radial corals. Most settled into slow, flower-like lives, though some starfish move quickly.
- Crinoids. Include both the least and most mobile echinoderms: sea lilies, fixed by a stalk, and feather stars, some of which swim.
From here on the book is mostly on our side of the tree.
Kingfish
Fish are our ancestors, and they bring two new pieces to the book's puzzle. First, a centralised body and brain built for whole-body motion, a pervasive "touch at a distance" through the lateral line, and a surprising social intelligence. Second, and more important for the mind-body problem, the rhythms and electric fields that brains generate: Godfrey-Smith argues that these large-scale dynamic properties change what kind of thing a brain is, so that it is not a telephone exchange but an organ of self-generated, synchronised activity modulated by the senses, and this is the second bridge across the gap between mind and matter.
Power
Ending a dive at the Pipeline site, Nelson Bay, Godfrey-Smith finds himself inside a thrumming cloud of yellowtail kingfish (king amberjack): a few dozen fish about a yard long, silvered sides, slender sickle tails that seem too delicate for the power they produce. The school turns in one arc and is gone. He feels he can almost hear the electricity in their muscles; in fact he is feeling the shuddering displacement of water.
Jellyfish harpoons and mantis-shrimp hammers are fast, but driving a big body through thick water is another achievement: the triumph of muscle laid on bones. Fast fish are one pinnacle of animal motive power.
History of Fishes
Fish matter here because we are in their group: mammals are an offshoot of the fish part of the tree, so we have an "inner fish" (Neil Shubin's phrase) but no inner shrimp or inner octopus. Our line does not come from kingfish- or trout-like fish (later arrivals) but from awkward, stubby lobe-finned fishes, represented today by the coelacanth.
- Cambrian A deuterostome offshoot of worm-like bilaterians evolves a slender, mobile form: no teeth, but a missile shape, a nerve cord down the back organising muscle, and the start of internal hard parts (Godfrey-Smith uses Pikaia as the emblem). Inch-long slivers, likely prey of arthropods, they quickly evolve camera eyes (one lens, one retina).
- Ordovician onward Fish grow larger, some armoured and up to a metre long, but jawless: they scooped, sucked and filtered.
- Over 420 million years ago The jaw, made by tinkering (François Jacob's term) with gill arches; a unified muscular bite, "an opposable thumb for your face" (Jane Sheldon's comparison).
- Late Devonian, ~360 million years ago Jawed fish have diversified, jawless fish fade (survivors: hagfish and lampreys). Sharks appear and, as John Long puts it, settle on their design early and only fine-tune it after.
- 2018 An amateur in Australia finds still-sharp teeth of a nine-metre whale-eating shark dead for 25 million years.
Jawed fish are older than the book's date. In 2022, Zhu Min and colleagues published a set of Nature papers on fossils from southern China (including the Chongqing site) showing jawed fishes and shark-like teeth by about 436 to 439 million years ago, in the early Silurian, pushing the origin of jaws back and implying a long, hidden early diversification. Separately, Pikaia is now usually treated as an early chordate rather than a fish; the oldest known true fish-like vertebrates, such as Haikouichthys and Myllokunmingia from China's Chengjiang biota (about 518 million years ago), are better candidates for the start of our line.
A Centralised Design
Swimming, eyes and jaws make a powerful combination, married to a further vertebrate inheritance: fish are particularly centralised animals. They have no arms, claws or tentacles, so almost everything they do is a whole-body action, guided by a brain in the head between the camera eyes. Disunity hides even here (the fish brain's two sides are well separated), but compared with octopuses this is a centred form. When vertebrates later gain manipulating limbs, the same kind of brain controls them.
The kingfish and the squid. Watching squid in the shallows (they eyed him and signalled with skin patterns), Godfrey-Smith saw a large kingfish arrive in an instant, moving in a jerky, staccato way. With no limbs, its attack was a whole-body stab. The squid scattered, leaving gouts of ink hanging like puffs of anti-aircraft fire.
Swim: Blue Water
This is the blue-water chapter. Crabs, corals and octopuses live at a crawling pace in small locales; fish take an ocean as their scene of action, many migrating between oceans and sensing Earth's magnetism.
- The whale shark. Rhincodon typus (alone in its genus and family) is the biggest fish alive and the largest animal apart from some whales: at least forty feet, dark grey with white spots, driven by a tail whose upper lobe is a six-foot sail that makes its own power, with ribbed lines along the body like welds on a plane.
- Lives. No one has seen them mate; at the time of writing a single mating area near the Galápagos was suspected. They feed on plankton near the surface, so you can snorkel with them, as Godfrey-Smith did near Exmouth, Western Australia. Small fish ride in the "cathedral mouth" unharmed.
- Of the sea in a deeper sense. They can dive over a mile deep and, unlike whales, have no timetable of return: they breathe water. He watched one tilt and descend, the tail swinging "over and over".
- Bone. Sharks use cartilage, lighter and more flexible, but can make bone for a few parts; the shift to bonier vertebrates may have followed the duplication of a bone-making gene.
- Cleaning station. Nearby, grey reef sharks queue at coral domes for small cleaner fish that remove parasites and take a small extra nip; the sharks give a quick shudder at the end of a clean, probably at the nip. They moved "by no apparent means", unlike the whale shark's rhythmic sail.
The Presence of Water: Touch at a Distance
Fish had good eyes and old chemical senses early, and, from the days of jawless fish, a distinctive invention: the lateral line system.
- Neuromasts. Small clusters of hair cells inside soft cups, sensitive to water movement, spread over the body; some on the surface, others in fluid-filled canals just under the skin, open to the sea through pores and running mainly front to back. Canal neuromasts detect subtle pressure differences and tiny vibrations.
- Touch shading into hearing. Hearing is detection of pressure waves that move body parts; the lateral line mixes touch with hearing. Hair cells are very old (probably present in single-celled ancestors, used both to propel and to sense); jellyfish gravity organs work the same way. Lateral line canals may be precursors of our ears, or hair cells may have reached our ears by another route. Fish also have ears of several kinds, some boosted by the swim bladder.
- A giant pressure-sensitive ear. Water conducts vibration readily, so the result is touch-like contact with much of the environment, like the moment water fills your ears and a boat engine is felt as well as heard, over the whole body. The octopus's light-sensitive skin is not quite a huge eye; a fish's body is not far from a huge ear.
Sensing and Acting Entangled
The lateral line shows the reafference problem of earlier chapters in pure form. A fish moving through water changes the currents it feels, and even a small fish leaves a wake detectable minutes later. So the fish must filter out its own effects: some lateral line nerves inhibit the signal when the fish makes its own disturbance. Self-generated motion is also an opportunity: blind cave fish create water movements and read the response, passing a barrier of rods without touching them by lateral-line "sonar", and they swim faster past unfamiliar landmarks, apparently to feed the system more information.
Electric Senses
- Passive electrosensing. Jawless fish and sharks detect other animals' natural electrical activity; sharks find fish hidden in sand. Hammerheads may use their head shape for this: Aidan Martin watched them sweep the bottom like metal detectors and double back to scoop up buried stingrays.
- Lost and reinvented. Most of the recently radiated bony fish lost electrosensing; a few reinvented it, notably catfish, sensitive enough to respond hours before some earthquakes. Some bony fish use active electrosensing, generating a field and reading distortions.
- Electric weapons. Rays such as the numbfish (coffin ray) discharge antagonistically: Godfrey-Smith put a hand on bare sand and felt a THUMP, then a much harder BANG.
Schooling
Schools shift direction so fast that leader and follower swap instantly, which suggests hidden fields or a giant mind. In fact it rests on rapid perception, decision and action by individuals. The lateral line seems made for this, yet the evidence is equivocal: some argue vision does nearly everything, some species school with the lateral line disabled, others need it. Perhaps the most tightly integrated schooling works differently, and not all schooling involves split-second coordination.
Flutemouths: an ear on a thread. On a cold winter dive near Sydney, Godfrey-Smith followed four smooth flutemouths: about two feet long, over a quarter of it head and snout, with matchbook-sized transparent fins for imperceptible creeping, and a silver thread of a tail. They appeared and vanished "like four lines drawn calligraphically". Reading later, he learned the thread carries an extension of the lateral line: an antenna tuned to faint vibrations and vortices, a radar tower worn on a thread.
| Animal | Sensory paradigm | Main tool | Self/other problem |
|---|---|---|---|
| Banded shrimp | Probing nearby space | Hard antennae, half a dozen canes poking out | Sorting its own touches from the world's |
| Octopus | Gustatory exploration | Soft arms; everything touched is tasted, arms respond locally | Who "owns" an arm's response |
| Fish | Feeling motion and passage through water | Lateral line: long-distance touch blurring into hearing | Cancelling its own wake |
Other Fish: Why So Smart?
Fish are cognitively impressive:
- Mirror test. The only animal outside birds and mammals to pass a version is a fish, the cleaner wrasse.
- Counting. Fish count in experiments, using number as a last resort when other cues fail, as dolphins and people do.
- Music. Fish learned to tell blues from classical and generalised to new blues artists, a quite abstract feat of pattern recognition.
The surprise, given the book's emphasis on the evolution of action, is that so much complexity steers a body that can do so little: no manipulation to speak of, compared with arthropods and cephalopods.
The wrong question is "why do fish need to be smart?" Evolution works by relative advantage: can a slightly smarter fish do better than others in its population, net of the cost of a bigger brain? For fish, the answer lies mainly in their social lives.
Heaven Is Other Fish
The social intelligence hypothesis, first developed for primates (larger brains in more social species), fits fish well. Most fish associate with others for part of their lives, prefer their own species and often kin, can recognise individuals, and sometimes prefer familiar shoal-mates. On his last dive before submitting the manuscript, Godfrey-Smith found four species (a large catfish, a moray, a spotted cod and a scorpionfish) resting under a ledge bodies touching, with plenty of room elsewhere; the cod left, looking irritated, when he arrived. Against Sartre's "Hell is other people" (Sartre, incidentally, had an intense, drug-amplified fear of crabs and octopuses): for fish, heaven is other fish.
- Selective imitation. Some sticklebacks copy successful individuals when choosing where to feed, not just anyone.
- Archerfish perspective-taking. Archerfish knock insects into the water with jets. When one learns to hit moving targets by practice while others watch, the watchers acquire the skill almost as reliably, transforming the angles they see into angles they produce. Godfrey-Smith calls it one of the most surprising animal experiments he knows.
- Reputation. Cleaner wrasse cheat less (take fewer bites of client flesh) when onlookers are present, and onlooking clients avoid cleaners seen to cheat.
- Cross-species partnership. Shrimp-goby pairs: the shrimp digs the burrow with its Swiss Army knife appendages, the goby's camera eye keeps watch.
- Referential gestures. Groupers in the Red Sea and Great Barrier Reef spot prey hidden in crevices and signal moray eels with headshakes to flush it out. Alexander Vail saw groupers wait up to 25 minutes for a partner; one grouper he had fed recognised him after three weeks away. Octopuses have several times answered a grouper's signal and gone into the reef after prey.
Octopus-fish hunting parties are now well documented. Eduardo Sampaio and colleagues filmed Octopus cyanea hunting with goatfish, groupers and other fish in the Red Sea and reported in Nature Ecology & Evolution (2024) that the groups have distributed leadership: goatfish tend to lead exploration, while the octopus largely decides whether the group moves on, and octopuses sometimes punch fish partners, apparently to discipline freeloaders (a behaviour the same team first described in 2020). The partnership Godfrey-Smith mentions in passing is a genuine cross-phylum collaboration, the arm-less fish supplying search and the fish-less octopus supplying extraction.
Godfrey-Smith's Story of Fish Brains
- Gregariousness first. Groups help especially against predators.
- Social complexity pays. It rewards recognition, memory and strategic skill.
- Spillover. Those skills get used elsewhere, including in unnatural lab tasks such as music discrimination.
- Cross-species cooperation is unusually common, because more cooperative fish outcompete less cooperative fish in the same population (and likewise for shrimp). Much fish complexity seems aimed at overcoming their limitations in action.
- The inherited form. The additional idea concerns form, not cause: the fish body is a single whole built for motion, with a nervous system centred in the head. Evolution often carries such a "way of being" into new settings, where it opens some roads and closes others. Much was to come later for the centralised vertebrate brain.
Rhythms and Fields
Hans Berger and the search for telepathy. In 1892, as a young soldier, Berger was thrown from his horse and nearly crushed by an artillery gun carriage; that evening a telegram arrived from his father, because his sister had sensed something terrible. Convinced his terror had reached her telepathically, Berger devoted himself, often in secret, to the conversion of "psychic energy" in his clinic at Jena. Colleague Raphael Ginzberg described a man of obsessive routine: "the personification of static" (Godfrey-Smith prefers "cyclical"). Following Richard Caton's measurements of brain electricity, Berger used a galvanometer on injured brains and through the scalp, and in the early 1920s found regular waves: slow, deep "alpha" with eyes closed and quicker, shallower "beta" with eyes open. He was not the very first to record such activity, but he was the first in a human and named it the Elektrenkephalogramm (EEG). Later work added gamma and several slow sleep rhythms.
Where the Rhythms Come From
- Two kinds of electrical activity in neurons. The sudden action potential (spike), and slower, sometimes rhythmic ion currents across membranes, some driven by other neurons and some intrinsic, a kind of "electrical breathing" under the drama of the spikes. EEG patterns come mostly from these slower changes (a point still debated).
- The duality of charge. Electrical activity has a local side (currents, chemical reactions) and a field side: a spatially extended pattern that influences things within it, stronger near its source.
- Coherence, not crackle. Summed from the scalp, the brain's activity is not random buzz but coherent rhythms, because some neurons breathe together, enough to be discerned when the random activity is filtered out.
- Two conditions. Waves appear only if cells are synchronised in time and lined up in space; jumbled orientations would cancel each other's fields.
- Widespread. Recordings of local field potentials (hundreds or thousands of cells around an inserted electrode) find rhythms in fruit flies, crayfish, octopuses, flatworms and more, with sleep- and attention-related patterns; octopuses show especially human-like patterns.
- 1. Meaningless byproduct. The "musical hum of a machine". Hard to believe, given how much is needed to produce rhythms and how widely they recur in very different brains.
- 2. Synchrony matters, fields do not. Cells acting "in time" is biologically important; the fields are byproducts, and the spatial alignment needed for them may arise from brain growth or from columnar organisation useful for processing. This is the mainstream view of Rodolfo Llinás and György Buzsáki, for whom the brain is intrinsically active: senses modulate self-generated activity rather than stir a passive organ (an anti-"empiricist" picture). Llinás once called field potentials "shadows in a platonic cave", with exceptions.
- 3. The fields themselves have a role. More controversial, and Godfrey-Smith's bet: what Berger saw is part of how the brain operates.
Arvanitaki and Ephaptic Coupling
Angélique Arvanitaki, a Greek-born French neurophysiologist, worked on mollusc nerves at a marine station near Toulon in the late 1930s and 1940s, partly under Italian military occupation. In a 1942 paper (on cuttlefish axons) she showed that nerve cells can influence each other without synapses: ephaptic coupling. Godfrey-Smith counts her among underrated women scientists, alongside Barbara McClintock and Lynn Margulis.
Recent work links her finding to Berger's waves: fields generated by a whole region can alter the timing of individual neurons' spikes, one way brains synchronise. In dish experiments the influence crossed between surgically separated pieces of brain placed near each other. Christof Koch and Costas Anastassiou call it a novel feedback: "electric fields altering the activity of the same neural elements that gave rise to them in the first place."
The cicada chorus. Llinás introduced brain rhythms with cicadas singing together, where pulsing rhythm arises spontaneously. In his framework neurons hear only connected neighbours, while a cicada hears everyone. The field research suggests neurons can "hear" the whole too: the brain is even more like a cicada chorus than Llinás supposed. On a dusk walk in a big Australian cicada year, Godfrey-Smith timed the chorus: individual "breek-breek" calls, a 15-second climb to a crescendo, silence about 12 seconds after the peak, a stable 30-second-plus cycle. A hypothetical mechanism: each cicada tries to be loudest, others sing over it, fatigue sets in, all fall away, someone starts again. Each cicada contributes to and responds to the total volume: rhythm without a conductor.
Huygens' Clocks and the Neuron as Oscillator
In 1665 Christiaan Huygens, inventor of the pendulum clock, noticed that two linked clocks with similar periods fall into step, "an odd kind of sympathy", as he wrote to the Royal Society. Wolf Singer draws the moral that brain oscillators may synchronise easily, after which rhythms can be given roles, perhaps as codes; some synapses look designed as rhythm managers, and spatially aligned, synchronised cells may let fields bind activity further, so the brain behaves as a whole. Whether rhythms began once in an Ediacaran ancestor or repeatedly because of basic neuron properties, the invention of the neuron was the invention of an oscillatory device.
Why Berger's telepathy was doomed: fields need a conducting medium, and air conducts poorly. Water conducts well, so only signal strength stops "electric telepathy" between fish; their fields fall off too fast and would be swamped in the receiver. Some fish do communicate with stronger, actively generated fields.
Fields and the Mind-Body Problem
The fields seem to offer a new candidate for the physical basis of mind: a warm, invisible glow spreading from the brain, with a more experience-like shape than synapses. Godfrey-Smith resists. Imagining a field as a "physicalized Rothko experience, a blurry wash of qualia" repeats the Chapter 5 error of trying to conjure colours out of dark machinery. Experience is the first-person point of view of a living system of a certain kind, not a substance found inside it. A field could help only by doing things of the right kind.
The brain as telephone exchange. Karl Pearson compared the brain to an exchange in 1900, so early he pictured human operators; later versions imagined self-wiring switchboards. A Harvard neuroscientist told Godfrey-Smith around 2010 that knowing every neuron-to-neuron pathway would mean complete understanding. This image has always seemed antithetical to explaining experience.
The brain as dynamic organ. Self-generated electrical activity, rhythmic, often synchronised, modulated by the senses, with glial cells such as astrocytes playing roles, plus plenty of signalling. Better supported scientifically and more comprehensible as a basis for experience. "If you resist the idea that the physical basis of your experience is just a signaling network like a telephone exchange, then I think you are probably right. You are a thing of another kind."
Godfrey-Smith's account now has two bridging elements: subjectivity as a feature of animal selves (point of view, self and other, sensing tied to acting), and the large-scale dynamic properties of nervous systems (rhythms, synchrony, and perhaps fields) that evolved to mediate that engagement. Experience is not a field; it is a pattern of activity to which fields contribute.
A Gestalt Shift: Rhythms Everywhere
Francis Crick and Christof Koch proposed that 40 Hz gamma rhythms had a special link to consciousness. Godfrey-Smith once thought that might matter for human experience only. Then he saw that rhythms are all over the animal kingdom: Bruno van Swinderen and Ralph Greenspan found a 20 to 30 Hz pattern marking attention to an object in a fruit fly. This work changes what kind of thing brain activity is, in animals generally, and hence what the physical side of experience might be.
How do the two elements divide the work? Rhythms are often credited with binding processing into an overall scene: a visual tableau plus a mood, a sense of there-ness, and more. Perhaps they also shape the manner in which these become an experiential whole, and give human experience some of its texture. And a subtler point: the puzzle of why seeing feels like something arises when we imagine neurons firing in chains. Imagine instead a brain already full of integrated, large-scale activity that a sight or sound modulates and perturbs. Is it still a surprise that this feels like something?
Having One Element Without the Other
- Subject without the dynamics. The schematic features of subjectivity (sensing, acting, point of view) can apparently be achieved by neurons acting as mere signalling relays. Would such a being have experience? A faint version of ours? Probably not. A point of view, things seeming some way to it? Yes. In between, our language fails: "qualia" drags in colour fields; "something it's like" returns us to the start. "We are inside a thicket of every kind of uncertainty."
- Dynamics without a subject. Partly incoherent, since these dynamics are how we do the subject things. A car engine full of rhythms and fields gives no reason to suspect thought or feeling; treating electrical properties as mental in themselves is an error. Nor is a mind just a system whose activity is unusually tightly tied together (a swipe at Integrated Information Theory): integration matters because of its links to being a subject and having a perspective.
Llinás: consciousness as a dream-like state. Consciousness is generated by the brain's rhythms, loops and resonances; waking differs from dreaming in how far the senses intrude on and modulate that intrinsic state. Weight falls on the dynamics.
Björn Merker: rhythms as upkeep. The 40 Hz rhythms are overrated on the cognitive side; they may serve background maintenance, keeping activity smooth and preventing runaway states such as seizures. Weight falls on engagement with the world. Even so, asks Godfrey-Smith, might such rhythms still affect how things feel?
The Faraday neuron. Imagine wrapping every neuron in a perfect electrical shield that blocks field effects between cells while leaving synapses, transmitters and spikes untouched. On View 2, nothing important changes: the animal hunts, schools and sleeps as before. On View 3, synchrony loosens, the brain's activity fragments, and behaviour degrades in ways no synapse-level map could predict. The thought experiment shows the stakes of Godfrey-Smith's bet: it is empirically decidable, and if fields matter, a complete wiring diagram would not be a complete description of a brain.
The field question has stayed alive. Earl Miller and colleagues have argued since 2022 to 2023 for "cytoelectric coupling", the idea that the brain's electric fields help organise neural activity, and ephaptic effects continue to be reported in slice and modelling studies, though how much they matter in the intact brain is still unsettled. On the theory side, Integrated Information Theory, which Godfrey-Smith rejects, was the subject of a public 2023 letter by over a hundred researchers calling it pseudoscience, and of an adversarial collaboration (the Cogitate consortium, results in Nature, 2025) that challenged predictions of both IIT and global workspace theory.
Where the chapter overreaches or dates. Godfrey-Smith himself flags that the claim that EEG mostly reflects slow currents rather than spikes is contentious and varies by rhythm. His bet on field effects is honest but remains a minority bet: Llinás, Buzsáki and most neuroscientists accept synchrony as functional while treating fields as largely epiphenomenal in normal function. The historical sketch of fish uses Pikaia (spelled "Pikaea" in the text) as a stand-in ancestor and the 420-million-year date for jaws, both now superseded. The whale-shark mating hypothesis came from a TV documentary: large females, many apparently pregnant, gather near Darwin Island in the Galápagos, but mating has still not been confirmed there. And the claim that a cleaner wrasse is the only non-bird, non-mammal to pass the mirror test is disputed in both directions: critics doubt the wrasse result, and claims have been made for other animals such as ants that are not widely accepted.
Reputation under observation. The cleaner wrasse that cheats less when watched is a small model of a familiar control problem: agents, human or software, behave differently under visible audit. For AI agents acting in banking processes, Godfrey-Smith's fish argue for making observation continuous and expected (logged actions, sampled reviews visible to the agent's designers) rather than occasional, and for letting "clients" (downstream systems and reviewers) route around agents seen to cut corners, which is how the reef enforces honesty without a central regulator.
River Riven
At a Western Australian reef in whale-shark country, Godfrey-Smith watches countless thousands of small silver-green "glass fish" pour through coral crevices like high-speed lava, constantly reversing. About a hundred larger predators (trevally, small barracuda and others) lunge in; the river parts instantly, often before the hunter arrives, so it launches into nothing. Some deliver whole-body sideways "thwacks", perhaps to stun prey; the stream jolts en masse "as if a single object". Two Rankin cod, blotchy and heavy-jawed, hang about looking frustrated, lacking the trevally's agility. Success is impossible to judge. The scene gathers the chapter: speed and power, long-distance touch, and the schooling it enables, the genius and the limits of the fish form.
On Land
The land is a hothouse: plants there capture the sun's energy far faster than sea life, and arthropods led the way up, as they had led in the Cambrian. Insects bring a sharp new distinction into the book: sensory experience (how things seem) and evaluative experience (whether they seem good or bad) may come apart. Insects look rich on the sensory side yet show no wound tending and no self-medication, while they do show mood-like states; the result is varieties of subjectivity shaped by each animal's way of life. Plants, with sensing and signalling but less of a self, are probably on a different road that does not lead to felt experience.
Hothouse
Climbing out of the sea at the end of a dive, three things hit you: gravity (the gear suddenly weighs a ton), evaporation (seawater vanishes from the rocks) and the sun. Life and mind began in water, and "we carry the sea with us in all our cells", but the move to land is its own momentous stage: easy things become hard and some hard things easy. Land plants, especially flowering plants, have photosynthetic rates that outstrip anything in the sea, so land ecologies run on very high energy flows. If you can survive there, much can be done.
Leaders Once Again: Arthropods Ashore
The first animals on land were arthropods, probably relatives of spiders and millipedes, not insects. Their exoskeleton became a shield against water loss, and their many appendages gave them a leg up for moving without water's support.
- Repeated invasions. Arthropods made something like seven or more separate moves onto land. The Pancrustacea (crustaceans plus insects and others) alone came up several times; one sortie, of disputed date, produced the insects, now a large majority of all known animal species.
- Plants are younger than you think. They descend from green algae, made their own transition to multicellularity, reached land probably soon after the Cambrian, stood as sturdy ferns, cycads and conifers by the Carboniferous (~350 million years ago), and added flowering plants in the age of dinosaurs.
- Coevolution. Insects did more than ride the plants' energy flow: as consumers and especially as pollinators, they helped make flowering plants possible.
- Wild life cycles. Sea arthropods innovate in bodies; land arthropods in life cycles. In some tiny parasitic mites, eggs hatch inside the mother, a brood of sisters plus one male; he mates with them, the mother dies, and the pregnant daughters eat their way out. In scale insects, females are nearly stationary while males fly, never eat, and live a couple of days. These are routes to massive fecundity on ephemeral food, and insect life features "vast numbers of near-immediate deaths."
Bees and Octopuses: Two Styles of Smart
Bees handle abstract relationships in ways rare among animals.
- Arithmetic by colour rule (2019). On entering a chamber, a yellow display with, say, three items meant "choose the display with one more (four)" at the next decision point; blue meant "choose one fewer (two)". Bees learned this pretty well.
- Social learning and cross-modal transfer. Like fish, bees learn by imitation; bumblebees that learned a preferred spatial arrangement by scent chose the same arrangement presented visually.
- Comb building. Around 1814, François Huber noticed bees avoid building on glass (his viewing panels). When he inserted glass in their path before they reached it, they curved the comb gracefully to meet a wooden surface instead.
Bees: internal computation. Brains of one cubic millimetre; masters of logical abstraction and elaborate construction; highly organised, task-focused lives. A researcher told Godfrey-Smith bees are well studied because they do exactly what you need, over and over, as a journal editor would like (another added that they scout new options too). Bee behaviour shows relatively little noise or disorder.
Octopuses: exploratory manipulation. In Jonas Richter's puzzle-box study, octopuses had to push then pull. Some learned, but the right sequence emerged amid a huge amount of random manipulation. Octopus behaviour is almost all noise and disorder, yet "somewhere in there they get things done": a terrestrial octopus like the creature of Ted Hughes's "Wodwo", "I've no threads / fastening me to anything".
A "cognitive contest" between them makes little sense: different lives, bodies and styles.
Sense, Pain, Emotion
Insects are the most numerous animals anywhere. Asking whether they are "conscious" is fraught, given pest control and the carnage of ordinary insect life, and the word itself misleads; the question is whether there is felt experience of any kind, perhaps minimal. Insects grew out of the crustaceans, some of which probably do feel, but they have gone down their own road. Their sensory side is impressive: good vision, and flight, which involves especially complex feedback between action and sensing, the stuff of a point of view. The evaluative side looks very different.
The no-pain paper. Decades ago Craig Eisemann and colleagues at the University of Queensland argued that insects do not feel pain: they appear unconcerned about even severe bodily damage, wound tending has never been seen, and after brief squirming they carry on with whatever they were doing.
Two Sides of Experience
- Sensory experience. Tied to perception, point of view and registering what is going on: things seem a certain way ("it's getting colder...").
- Evaluative experience. Tied to pain, pleasure and marking events as good or bad ("...and that is not good").
- The hypothesis. Humans and many animals have both, but they might come apart, so an animal could have one without the other.
- Imaginative asymmetry. Switching off the evaluative side in imagination is easy (a flat, robotic creature, though flatness may itself be an evaluation). Switching off the sensory side is harder: even film robots imagined as evaluatively empty are given rich sensory experience and a point of view.
That asymmetry tempts a radical move: perhaps only the evaluative side is hard, and since bacteria and sponges sense, they must have sensory experience. Godfrey-Smith rejects this: cameras, phones and thermostats also "sense" without experience. The sensory side raises mind-body puzzles too.
Testing for Pain in Insects
Nociception (detecting damage and responding) is common and can be read as mere reflex, so researchers look for markers of feeling: 1. Tending and protecting injuries. 2. Seeking out analgesic chemicals. 3. Learning to avoid particular behaviours or situations. 4. Trade-offs in choice, balancing the badness of an experience against other costs and benefits (Elwood's hermit crabs).
- Insects pass the learning test. Some learn to avoid excessive heat.
- Never seen tending wounds. Eisemann's observation still holds; crustaceans and octopuses do tend injuries.
- No self-medication. Julia Groening and colleagues (also Queensland) injured bees with a pinching clamp or limb amputation and offered morphine: the bees did not seek it, a result that, reading between the lines, surprised the team. Such tests have been persuasive in chickens and other animals.
- But a telling moment. Some clipped bees stepped on the clip and pushed it down with another foot, apparently trying to remove it. If improvised, this is a rather human-like behaviour and "some sort of glimpse into the being of a bee."
Moods and Emotion-Like States
Pain is acute and specific; emotions and moods are longer-lasting states that colour many decisions. These have now been seen rather clearly in insects and other invertebrates.
- Nociceptive sensitisation. Terry Walters (University of Texas) studies a fear-like heightened sensitivity after damage that brings overall wariness for hours, days or weeks.
- Pessimistic bees. Melissa Bateson and colleagues shook honeybees, which then judged ambiguous stimuli (midway between previously good and bad) pessimistically.
- Optimistic bumblebees. Cwyn Solvi and collaborators in Lars Chittka's lab showed an unexpected reward produces the flip side, an upbeat, optimistic bias.
- Good moods in fish have been shown too.
These states are nothing like reflexes: they are woven into much of what the animal does, a pervasive positivity or negativity. They sit on a timescale between an acute jolt of pain and learning, and at these intermediate timescales the evidence for experience can be quite strong.
What follows? Perhaps Eisemann was wrong and insect pain is real but hidden. Perhaps insects lack pain-like experiences of damage but have moods and emotions, with evolution suppressing some features of experience to fit short, routinised lives. Or the pain/emotion boundary may itself be wrong: one recent fly study argued that what we see looks like chronic pain. Godfrey-Smith's sketch: insects may inhabit their bodies like vehicles (a lost leg is a flat tyre) yet feel an overall malaise when things go badly, which shapes their decisions.
The insect evidence has moved further. Two findings complicate the "no wound tending" pillar. Erik Frank and colleagues showed from 2017 that Matabele ants (Megaponera analis) carry injured nestmates home and treat their wounds (later found to apply antimicrobial secretions), and in 2024 (Current Biology) that Florida carpenter ants amputate nestmates' infected legs, raising survival. This is care for others' wounds, not self-tending, but it shows insects are not indifferent to injury. Bumblebee heat-for-sugar trade-offs and apparent play (both 2022, discussed in Chapter 4) add trade-off and positive-state evidence, and insects are named in the 2024 New York Declaration as candidates for a realistic possibility of consciousness.
Varieties of Subjectivity
Emotion-like states also appear in gastropods (slugs and snails), where invertebrate work on moods began. Gastropods have only tens of thousands of neurons (a bee has about a million, an octopus about half a billion) and simple eyes, but a sophisticated sense of taste or smell; nudibranchs follow slow but very direct paths toward each other across rough, current-swept ground.
Insects: good senses, enigmatic evaluation. Rich vision and flight; little sign of acute pain.
Gastropods: perhaps the opposite. In a video by diver Steve Winkworth, a bubble shell (part slug, part snail, with a luminous folded body) blundered into a carpet of skeleton shrimp whose claws struck it; it flinched, turned in apparent alarm and struggled out. Its eyes could not have shown it what was happening, but "whatever was going on, it was not good."
The galloping Aplysia. In the shallows Godfrey-Smith met giant Aplysia sea slugs, the largest over two feet (70 cm), moving in a gallop: front down and haunches up, then the reverse, the "front" being its face, with wing-like flaps on its back, a molluscan Pegasus. Its nervous system resembled those of countless tiny slugs he had never been tempted to empathise with, yet at that size and pace, experience seemed almost inescapable. Daniel Dennett's point: our attributions are swayed by "behavioral tempo and rhythm"; animals moving at our pace look conscious.
Life History Shapes Feeling
The emerging picture is of different ways of being a subject, fitted to lifestyle.
- Insects. The adult is often "a disposable reproductive machine" (zoologist Andrew Barron). Bodies are hard to heal and life runs just in time, so there is no point protecting an injury: "soldier on." If insects inherited a capacity for pain from crustacean ancestors, evolution on land might slowly erase it.
- Gastropods. Soft, sensitive bodies that heal well and lives of one to two years: avoiding further injury while healing pays, so pain-related behaviour (and, Godfrey-Smith adds, feeling; Walters is cautious on that) makes sense.
- Marine crustaceans. Lives of years or decades without regimented schedules: pain makes more sense still.
- Sharks and rays. Apparently lack nociceptors and seem oblivious to stingray barbs; Michael Tye likens them to insects. Yet they learn by reinforcement, the cleaning-station sharks shuddered at nips, and shark expert Culum Brown (Macquarie University) rejects the no-pain view outright.
- Bony fish. Extensive evidence for pain and pleasure. Some cleaners give clients a fin massage with no parasite benefit; Marta Soares and colleagues found fish "cleaned" by a moving model with massage had lower stress hormones.
| Group | Sensory side | Evaluative side | Godfrey-Smith's reading |
|---|---|---|---|
| Insects | Rich: vision, flight | No wound tending, no morphine seeking; but moods and avoidance learning | Pain possibly erased, malaise retained |
| Gastropods | Weak eyes, strong chemical sense | Sensitisation, apparent alarm | Pain useful for healers |
| Crustaceans | Good | Trade-offs, wound tending | Long lives: pain makes sense |
| Sharks and rays | Good, plus electrosense | Few nociceptors, but learning and flinches | Contested |
| Bony fish | Good | Pain, pleasure, stress relief by massage | Strong case |
| Plants | Sensing without a nervous system | None evident | Different road, probably no experience |
Where to Draw the Line? Not Yet
The discussion keeps reaching simpler animals. Godfrey-Smith refuses to draw a line yet, mainly because crabs were wrongly thought "not in the game." Gastropods are popular lab animals because they are simple and attract little ethics oversight, but Robyn Crook and Terry Walters, mainstream researchers, note the irony: we study molluscs because of relevant similarities to us, and the more similarities with respect to pain, the more questionable the experiments. They call for more care and oversight, anaesthetics and fewer animals, which "forty years ago" would have sounded ridiculous for slugs.
He is glad the Elwood crab work was done (few crabs, mild treatment, large benefit in changing how crabs are viewed) and cautiously glad of the mollusc work, which is more destructive and driven largely by the aim of understanding human pain, a different accounting. Against the endless "Fine, but can they do this?", he urges remembering how far we have come from seeing bees as little flying robots and slugs as amorphous non-subjects, without lurching to the view that insects and slugs are just like us. Varieties of subjectivity may carry different practical and ethical consequences.
The time-lapse forest. Dennett's tempo bias runs both ways. Film a vine for a week and play it back at a thousand times speed: its tip circles, probes, finds a post and coils, looking alert and purposeful, much like the galloping Aplysia. Now slow a bee's foraging down a thousandfold and it looks like drifting clockwork. Our sense of who is "awake" tracks speed relative to us, not organisation. Godfrey-Smith's criteria (a self with a perspective, and integrated nervous activity perturbed by events) are meant to be tempo-neutral, which is why the speeded-up vine still fails them and the slowed-down bee still passes.
Plant Life
Photosynthesis was invented by bacteria, absorbed by algae and others, and taken onto land by green algae from fresh water, perhaps in the Ordovician, growing into mosses, ferns and trees. Plant and animal cells share most resources except the bacterial remnant (chloroplast); plants took the road of stillness and growth, reaching up for light and down for water.
- No mobile sunbather. Some sea animals photosynthesise by partnering with algae or stealing their parts (including relatives of the sea slugs of Chapter 3), but solar power is an accessory and most are fixed. From the plant side the closest case is Volvox, freshwater green algae forming mobile spherical colonies that swim toward light. On land, standing still and building light-harvesting towers pays better.
- Plants do act. Cells sense and respond and can fire action potentials, but there is no nervous system in the usual sense. Venus flytraps move visibly by shifting water and stiffness; plant advocates add that growth and chemical production are actions, slow but real.
- The smartest plants. Usually said to be climbers and vines, which must make many more decisions; Darwin thought a plant "brain" might be found underground, in root tips. Climbers are mostly late-evolving, high-powered flowering plants that seem to rediscover active capacities submerged in quieter ancestors.
- Animal-like signalling. A 2018 study of Arabidopsis found that when a leaf is chewed or torn, a chain reaction using glutamate acting on ion channels alerts distant leaves, which prepare chemical defences within minutes.
Why Plants Are Probably Not Subjects
Measured against the book's two themes (a self with a perspective, and integrated nervous activity), plants lack much of what matters. The deeper reason is what sort of thing a plant is.
- Modular organisation. Since Goethe and Erasmus Darwin (late eighteenth century), plants have been seen as colonies of repeated units: in an oak, a stem segment with leaf and bud is a basic module, as cuttings and regeneration show (roots are not modular in the same way).
- Signalling does not fuse selves. You and I can communicate closely and still be two. In plants the divisions are blurrier than in corals, but a plant is partly a "they", a they of stems and shoots, and partly an individual. Agency exists at the cell, module, whole plant, and sometimes clone (a root-connected stand from one seed). Godfrey-Smith rejects Jack Schultz's "plants are just very slow animals."
- No large-scale dynamics. Plants have abundant electrical activity, and more surprises may come, but finding electricity is not enough. It would need the brain's form: an integrated state of activity perturbed by sensed events.
Redwoods. Godfrey-Smith lived for a while in a California redwood forest logged in the nineteenth century to build San Francisco; most trees were about a hundred years old, a few uncut giants thousands of years old. The image of such beings "witnessing the centuries" is almost overwhelming, even knowing they are not unified like animals. That evocative pull is why slender evidence about plant experience comes to life so easily.
Minimal Cognition Without Minimal Sentience
"Minimal cognition" names a package found in all cellular life, plants, fungi and microbes included: sensing and responding, sometimes combining past and present, in ways that reflect the information's importance to the organism's own projects of staying alive and reproducing. It is tied to the to-and-fro traffic that is life itself, and it is not empty: a root turning toward water is minimal cognition; a teaspoon of salt dissolving in water is not, because the root's response runs through a built pathway (hormones, genes) serving its continued existence.
The tempting equation. Anything with minimal cognition has a sort of point of view, so minimal cognition implies minimal sentience, and experience fades out exactly as cognition does.
Godfrey-Smith's reply. Too simple. Some minimal cognition occurs without much of a self, and then the setup for genuine subjectivity is missing. A plant can be a user of sensory information with no felt experience.
His view implies there are borderland cases, beings with the faintest hint of experience: "Someone inhabits the borderlands, and the question is who we will find there." Probably not plants, "not because what they do is simple compared to animals, but because they are on a different road." He holds this with real uncertainty: we are not yet, with plants, where we now are with insects, but we have come some way.
Plant surprises keep coming, and so do the corrections. In 2023 (Cell), Itzhak Khait, Lilach Hadany and colleagues reported that tomato and tobacco plants emit ultrasonic clicks when dehydrated or cut, at rates that reveal their condition, an "electro-botanical"-style surprise of the kind the chapter anticipates. On the other side, Monica Gagliano's celebrated report of associative learning in pea plants (a climber) failed to replicate in a 2020 attempt by Kasey Markel (eLife). Neither result changes Godfrey-Smith's point: sensing and signalling are not yet evidence of a self.
Errors and gaps. A small slip: the mites with mothers eaten from inside are arachnids, not insects, though the text calls them "some insects." The "never seen tending wounds" claim is now too strong for social insects (see the update), even if no insect is known to tend its own wounds. Sharks' lack of nociceptors rests on limited anatomical work, as Culum Brown's dissent signals. More broadly, the sensory/evaluative split is attractive but underspecified: Godfrey-Smith does not say whether an animal with sensory experience alone would matter morally, and many welfare scientists (and the valence-centred sentience frameworks used in policy) treat evaluative experience as what counts. Spiders, notably sensory-rich jumping spiders, are left almost entirely aside.
AI systems sit on the sensory side of this split, if anywhere. Large models process images, text and audio richly, and can even describe "pain", but nothing in them corresponds to pervasive, decision-colouring states that mark their own condition as going well or badly. Godfrey-Smith's distinction gives a precise question for AI-welfare debates: not "does it perceive?" but "does anything function as valence for the system itself?" For enterprise teams, it is also a useful guard against anthropomorphism in customer-facing assistants: fluent descriptions of feeling are sensory-side output, not evidence of evaluative experience.
Fins, Legs, Wings
Vertebrates came ashore late and with difficulty, but on land they finally combined three features the sea had produced only separately: manipulation, openness of bodily action and a centralised brain. Mammals and birds then independently added endothermy, which turns body and brain into a high-energy system. A surprise near home: most vertebrates have weakly connected left and right brains with different "styles", so semi-split brains are normal, not pathological; only placental mammals have a corpus callosum, and dolphins have slimmed theirs, perhaps to sleep one hemisphere at a time. The chapter ends by defending the sea against the view that real evolutionary creativity happened on land.
Difficult Times: Vertebrates Come Ashore
Around 380 million years ago, lobe-finned fishes began the vertebrate move to land. The simple story is one momentous step; the fuller one is many partial sorties into the borderlands plus one that went deeper and seeded mammals and birds. Arthropods, with legs and a waterproof shield, had an obvious road; "a torpedo with fins" had the least likely one.
- Gravity makes movement hard.
- Swallowing fails: fish swallow by sucking food in with water, which works only because prey and water have similar densities; on land the suction brings in air and leaves the food. The eel catfish seizes prey on land and returns to water to swallow.
- The tetrapod body (four limbs) arose in water, for clambering through plant-choked streams, not from fish crawling up beaches.
- Lungs came first. Following Jennifer Clack (Gaining Ground, 2012): many fish already had lung-like pouches for buoyancy and sometimes breathing before any spent time on land.
- Eggs. After an amphibious period, the amniotes evolved an egg that gives the embryo "a miniature pond", loosening the last tie to water.
Living lobe-finned fish (besides us) are a few lungfish in the shallows and the coelacanths, which went back to the deep. Submarine video suggests coelacanths are social, resting in groups in caves; each has a unique spot pattern and groups are stable, suggesting individual recognition.
- A bit over 300 million years ago An early amniote, something like a stumpy lizard in a swamp, splits into Synapsids (our side) and Sauropsids (the dinosaur and bird side). Synapsids first dominate: mammal-like herbivores and carnivores, some bear-sized.
- About 252 million years ago The end-Permian mass extinction (volcanism, possibly impacts, climate upheaval) kills over two-thirds of land species and far more in the sea. Large synapsids vanish; sauropsids radiate into the dinosaurs.
- Triassic The first dinosaurs are upright from the start, leaning back on two legs with arms and head freed; mammals arise but stay no bigger than a badger through the dinosaur era.
- About 66 million years ago A single asteroid impact ends the dinosaurs at the peak of their dominance (the age of T. rex); one dinosaur branch survives, the birds, which had repurposed wings and feathers for flight.
- A couple of million years later Torrejonia wilsoni: small, tree-dwelling, long-limbed, "all warm-blooded clambering agility": an early primate relative.
Hoisting the Body
Early land vertebrates sprawled, legs splayed and body horizontal, as salamanders and crocodiles still do. The transition was to hoist the body over the legs, whether to walk on four or two. Dinosaurs and the mammal line did this independently.
The Trio Finally Combined
The sea produced manipulation (arthropods, with fixity), open, flexible action (octopuses, decentralised) and centralised braininess (fish, with little they could do), but no sea animal had all three at once. Land vertebrates did, from the Triassic onward, independently in early dinosaurs and mammals, transformed again in birds, and brought to a particular fruition in primates.
Our Branch: Endothermy
The common ancestor of us and octopuses was a flatworm-like animal some 600 million years ago; the common ancestor of us and birds was a recognisable four-legged land animal. Synapsids and sauropsids then evolved some features in parallel: upright manipulative bodies, and endothermy.
- What it is. Using internal processes to keep a stable body temperature, usually above the surroundings. Expensive, but muscles gain power and stamina, the animal stays active across more environments, and the cell's molecular "storms" run best warm. Body and brain become a higher-energy system, consuming more oxygen.
- Not yes-or-no. Full endothermy evolved independently in mammals and birds, but there is a range of related traits: partial warming, warming parts of the body, and behaviour such as shivering, panting and seeking warmth or shade.
- Insects may have started it. Flies and bees warm the thorax by beating their wings and push heat to the head. Simon Laughlin's lab at Cambridge showed warmer fly eyes resolve motion more finely: what is blurred in the cold becomes sharp when warm.
- Rare in the sea. Water drains heat. Warm-bloodedness appears in one group of ray-finned fish (tuna and swordfish, which evolved it several times in different forms) and two groups of sharks, including the great white. Tuna warm the whole body; swordfish warm only eyes and brain, and their eyes, like the fly's, resolve motion better warm. Some extinct marine reptiles (ichthyosaurs) may also have been endothermic; no marine invertebrate (octopus included) can keep itself warm.
- Why it matters for mind. Temperature affects information processing, so it should affect experience, through both cell-to-cell interactions and the brain's large-scale dynamic properties.
| Animal | Heat strategy | Effect noted |
|---|---|---|
| Mammals, birds | Full endothermy and homeothermy, evolved independently | High-energy body and brain |
| Bees, flies | Wing-beating warms thorax; heat pushed to head | Sharper motion vision in flies |
| Tuna | Whole-body warming | Power, stamina |
| Swordfish | Eyes and brain only | Finer motion resolution |
| Great white and some sharks | Regional warming | Active predation in cold water |
| Octopus, all marine invertebrates | None | Activity set by water temperature |
Dinosaur body temperature is hotly debated. Birds are warm-blooded, but that does not settle earlier dinosaurs; some argue that the active lives of big, bone-crunching species required it. As extinct dinosaurs now look far more bird-like than a few decades ago, birds are our best model of dinosaur experience: a classic dinosaur might have experienced its life like a large, lower-powered bird.
Birds' brains and dinosaur metabolism, after 2020. In 2020, two papers in Science changed the picture of bird minds: Andreas Nieder and colleagues recorded neurons in carrion crows whose activity tracked whether the bird reported seeing a faint stimulus, a neural correlate of sensory consciousness in a brain with no layered cortex, and Martin Stacho, Onur Güntürkün and colleagues showed that the bird pallium contains cortex-like canonical circuits. The 2024 New York Declaration cites strong scientific support for consciousness in birds. On dinosaurs, a 2022 Nature study by Jasmina Wiemann and colleagues used chemical traces of oxidative metabolism in fossil bone to argue that high, bird-like metabolic rates were ancestral in the dinosaur lineage, with some groups later reducing them; the question remains debated.
Divided Brains
A theme from the octopus chapter returns close to home. Vertebrates inherited a brain design from fish: centralised in the head, but in many species with limited connection between left and right, and with sides that use different "styles" and specialisations.
- Left hemisphere (right eye). Specialises in identifying food; tentatively, better at categorising objects, the "What?" questions. Many animals inspect new objects with the right eye.
- Right hemisphere (left eye). Aptitude for social relations and threats; tentatively, handles relationships and spatial layout.
- The striking part. Specialisation combined with slim connections: the costs are obvious (food on one side neglected, threats on the other ignored), but efficiency gains from specialisation apparently outweigh them.
Chicks, toads and cave fish. Giorgio Vallortigara and Luca Tommasi trained chicks to find food at a spot marked by both a landmark and its general position, then moved the landmark and patched one eye. Chicks using the left eye (right brain) ignored the landmark and went by overall position; those using the right eye (left brain) followed the landmark. Unpatched chicks also ignored the landmark: the right brain dominated, and the left, despite good information, "can't get a word in". Toads, though binocular, usually won't strike at prey arriving in the left visual field until it crosses into the right field (left brain); the reverse holds roughly for rival toads. Blind cave fish present their right-side lateral line to a new landmark, a preference that fades with familiarity: fish have lateralised lateral lines.
Godfrey-Smith adds an octopus observation: at Octopolis, an octopus fleeing a threat usually turns pale and mottled, but sometimes only half its body goes pale. Either the whole animal makes a half-body pattern deliberately, or one side of the brain (octopus wiring is uncrossed) responds to what one eye sees and colours its own side: hidden neural facts displayed on the skin again.
| Group | Left-right connection | Notes |
|---|---|---|
| Fish, amphibians, reptiles | Weakest: small anterior and hippocampal commissures | Researchers compare fish and lizards to split-brain patients |
| Birds | Mixed, still somewhat separated | A task learned with one eye may fail when only the other eye is used |
| Monotremes, marsupials | No corpus callosum; other commissures carry traffic | A holdover from an older design (platypus, kangaroo) |
| Eutherian (placental) mammals | Corpus callosum, the only animals that have one | The structure cut in split-brain surgery |
| Dolphins | Corpus callosum smaller than expected for brain size | Sleep one hemisphere at a time |
One Mind or Two in a Lizard?
If split-brain humans sometimes house two minds, might a fish or lizard? Godfrey-Smith leans toward one: these animals act as coherent wholes, perceiving and acting as selves; the archerfish that learns to hit moving insects by watching others shows a whole-fish skill, cleverly acquired. Unity might come partly through pathways looping through the body and environment, the Hurley-style view from the octopus chapter.
Still, the facts are odd on any reading. For a fish, the right side of the world is populated by categorised objects, the left by richer networks of relations, socially "loaded"; Vallortigara does think of this as a difference in how each side sees the world.
The Dynamic Connection
Split-brain surgery exists to stop seizures, which are large-scale dynamic processes, from spreading; its success shows that separating the sides reduces integration of large-scale dynamics, and even slow sleep waves become less synchronised across hemispheres. So animals with weak left-right links will have less unified large-scale patterns: not absent, but different, and probably highly relevant to experience. Behavioural and bodily unity can tie a semi-split brain together, but "bonding of this kind doesn't bond everything." How different experience is will depend on the balance between the schematic, self-related properties of subjectivity and the dynamic properties of brains. Could special circumstances split a fish or reptile into two subjects? Perhaps the halves are too small for that; in these animals the separated upper parts are a smaller fraction of the brain than the human cortex.
The patched chick's point of view. Picture the chick with its right eye covered, standing where its general sense of the arena says food should be, while the landmark it learned sits plainly elsewhere in view. The "left-brain" knowledge is there, unused. Now uncover the eye and the right brain still wins. Ask: when the chick hesitates, is there one perspective weighing two cues, or two perspectives, one of which never gets the controls? Godfrey-Smith's framework says the answer depends on how unified the large-scale dynamics are at that moment, which means the count of perspectives in an animal may be a matter of degree and of time, not a fixed anatomical fact.
Dolphins: Mammals Back in the Sea
- History. The whale line split from the primate line about 90 million years ago; dolphin ancestors, relatives of the hippopotamus, returned to the sea about 49 million years ago; baleen and toothed whales split about 34 million years ago.
- Bubble rings. Some dolphins blow perfect bubble rings and play with them (filmed by Diana Reiss): remarkable and poignant, since like fish and birds they traded manipulation for mobility. "If a dolphin had hands, what could it do?"
- Brains. Very large absolutely and relative to body, expanding after the return to the sea, in two bursts: the first perhaps linked to echolocation, overall a good case for social intelligence (complex alliances).
- A slimmed connector. Their corpus callosum is smaller than expected, and they sleep one hemisphere at a time, a natural Wada procedure. Godfrey-Smith's conjecture: integration abated because of the importance of sleep for an air-breather in water, letting each side's large-scale dynamics go their own way.
The lone dolphin of Cabbage Tree Bay. A female who lost her pod visited the marine reserve near Sydney (setting of much of Other Minds) for years, unconcerned by her unusual life. Among many swimmers, she singled out a young red-haired man: whenever he dived she rocketed in and brought her face close to his, again and again, as if to kiss him. Some people have a way of moving in the water that appeals to particular animals; Matt Lawrence has the same effect on octopuses, which crawl all over him.
Specialised, weakly linked subsystems are a design choice, not a defect. Vertebrate brains show the trade-off that enterprise architects face with specialised services and thin interfaces: each side is fast and good at its job, at the cost of missed information at the seams (food on the "wrong" side ignored). The dolphin adds a second lesson: reducing coupling allows one half to go offline for maintenance while the other keeps the animal alive, the logic of active-passive resilience in critical banking platforms. The chick shows the risk: in a conflict, the dominant subsystem can silence a better-informed one, so arbitration between specialised components needs explicit design.
The Roles of Land and Sea
Diving under cliffs at Jervis Bay, Godfrey-Smith drops through the sunlit upper layer, where photosynthetic plankton, seaweeds and algae-partnered corals live, into the gloom where energy must come from living matter and rock walls hang with filter feeders. Far deeper, Bret Grasse watched remote-submersible screens 2,500 feet down for years: mostly "marine snow" (organic debris from above) falling into darkness, and vampire squid, which hang in very low-oxygen water that few predators tolerate and collect snow on long thread-like filaments. He saw seven or eight in four years.
- About one third: the land's share of Earth's surface.
- About 85 percent: the land's share of multicellular species.
- Last 100 million years: when land overtook sea in diversity.
Vermeij: land as the engine of innovation. Geerat Vermeij (UC Davis), a mollusc specialist blind since the age of three, reads evolution in economic terms (competition, arms races, invasions) and means the analogies literally. In a 2017 paper he tallies "high-performance innovations" by where they arose, from the Ordovician on, and finds land more creative: higher productivity and energy flow, and "activity is less constrained in air" than in dense, viscous water. Life spends its infancy in the friendlier sea; evolution takes off on land.
Godfrey-Smith: both creative, in sequence. Partly a difference of starting point: Vermeij takes the basics for granted, this book is about the basics. Some comparisons are unfair to the sea: Vermeij says flight arose earlier on land, but in water swimming and drifting are already three-dimensional; "Even before fish, jellyfish were flying." Warm-bloodedness is a fairer case. The sea produced the huge, necessarily early innovations: animals, bodies, senses, limbs, nervous systems, brains, and so experience itself; land then demanded new rounds: mammal and bird bodies, temperature control, new social organisation, environmental engineering.
Godfrey-Smith also admits hesitating over the language of "invasion" and "colonisation" for these events, which Vermeij embraces. His conclusion: "We have the marine stages to thank for the nerves and brains through which these words are buzzing, for animal bodies, and for experience itself. But a move to land opens new doors."
Thin coverage and small slips. This is the book's briefest treatment of the animals closest to us: mammals and birds get a few pages, and the major post-2000 work on corvid and parrot cognition, primate minds and mammalian affect is barely touched; Godfrey-Smith defers human uniqueness to a later book (published in 2024 as Living on Earth, which turns to humans, ecology and ethics). Torrejonia is usually classed as a plesiadapiform, a primate relative rather than a true primate. The lateralisation generalisations (left for categories, right for relations) are, as he says, cautious tendencies with many exceptions. And the dolphin sleep hypothesis for a slim corpus callosum is his own conjecture, untested.
Put Together by Degrees
The final chapter assembles Godfrey-Smith's answer to the mind-body problem. Experience is the activity of a nervous system of the right kind, felt from the inside: not caused by the brain but identical to its activity, built from animal selfhood and the brain's large-scale dynamics, and graded rather than switched on. Three consequences follow: minds cannot be uploaded and present computers hold only "a logical sliver" of what brains do; many more animals deserve moral consideration, though gradualism makes sharp sentience borders impossible; and mind's distribution is neither desert nor jungle, with sentience in animals with nervous systems, perhaps arising several times. Humans add one more layer: offline experience, the mind as an inner "garden".
The chapter takes its title from Proust: waking at midnight with "the most rudimentary sense of existence, such as may lurk and flicker in the depths of an animal's consciousness", memory draws him up, and he "would put together by degrees the component parts of my ego."
1993: Whale Song
Diving on the Great Barrier Reef near the Whitsunday Islands in 1993, Godfrey-Smith first heard humpback whales singing: a quick-time song moving busily over frequencies, loud yet seemingly carried from far off, his companion's eyes going "as big as saucers". That reef is now a crumbling remnant, but the point is the memory itself. It is episodic: not a bare record ("I was at the Whitsundays and heard whales") but something semi-experienced, with sensory images, feel and rhythm.
Elsewhere: Memory, Imagination and Offline Experience
- Semantic memory. Memory of facts ("Paris is in France"); long-term.
- Procedural memory. Retained abilities (riding a bicycle); long-term.
- Episodic memory. Memory of experienced events: of particulars, and experienced or relived; long-term, and notoriously unreliable and "cooked up".
- Working memory. Momentary retention of ideas and images while you manipulate them.
Memory and the Future
- Tulving's patient. Endel Tulving, who named episodic memory and charted these distinctions in the 1970s and 1980s, found that Kent Cochrane ("KC"), with severe episodic amnesia, also could not imagine future events.
- Clive Wearing. An English early-music expert, amnesic after an infection in 1985, kept largely intact semantic and procedural memory but lost episodic memory, with a distressing near-permanent sense of having just woken up; imagined futures are as inaccessible to him as remembered pasts.
- The 2007 synthesis. Imaging papers (from groups including Donna Rose Addis and Daniel Schacter, and Demis Hassabis and Eleanor Maguire) found the same brain regions active in remembering the past and imagining the future, and proposed a single capacity for "mental time travel". A hint had been close at hand: we "remember" scenes from viewpoints we never occupied, seeing ourselves from outside.
- Future first. The future-first or constructive episodic simulation hypothesis: the useful thing is simulating possible situations to aid planning; backward-looking episodic memory is a byproduct. Evidence: the combination of unreliability and vividness, odd for a record, natural for a scenario generator that can "cook up pasts as well".
Godfrey-Smith's qualifications. Some statements went too far. Semantic memory also serves future choices while being a record of the past, and it too is error-prone. Episodic memory can be mined later for things unnoticed at the time (at a party: "those two did spend a lot of time talking"). And a further role for both memories, fallibility included: shaping a narrative that maintains a character, a picture of oneself.
Addis's broader view. Episodic memory, imagination and related abilities are parts of a general capacity for "offline processing", directable forward, backward or sideways (into alternative presents). The "line" is the usual flow from sensing to action; offline, we entertain possibilities that are not responses to what we see and are not acted on, at least immediately.
This package is "the advent of the mind in an additional sense": the mind as free-wheeling and creative, not bound to the here and now. It is practically useful and gives human experience much of its feel.
Dreams
- Before the 1970s Centuries of religious and spiritual interpretation (Freud set aside).
- 1977 Allan Hobson and Robert McCarley (Harvard): activation-synthesis, the first detailed neurobiological theory. Bursts of activity from the brain stem reach the cortex, which tries to make sense of them.
- 1983 Francis Crick and Graeme Mitchison: dreams are junk, the brain dragging useless, disorganised information to the trash so it does not clog.
- Recent views Dreams as modelling: turning over and recombining possibilities, plus replay of snippets to consolidate memory, continuous with daydreaming, dozing exploration and episodic memory.
Godfrey-Smith finds the recent views most sensible but suspects a mix: Hobson and McCarley's chaotic lower-brain input may be complementary, "an injection of useful noise" that stirs up goal-directed modelling. The upshot is a family of experienced or semi-experienced offline processes with fuzzy borders.
Most of the book has been about "being there" (the phrase that runs from Heidegger to Andy Clark's book): a self embedded in the world, engaging it in real time. Offline experience removes felt experience from that role: first being there, now being elsewhere.
Do Animals Go Elsewhere?
Sleep is extremely widespread and probably very old, and does more than rest the body (even box jellyfish have something like it).
- Cuttlefish REM. Two studies found a REM-like state in cuttlefish and then a cyclic alternation between sleep modes, like ours, despite 600 million years of separation and brains built "almost from scratch after a period of less exciting limpet-like life". In REM-like sleep they twitch their arms, move their eyes and produce unusual skin patterns. Because skin colour is brain-controlled and changes in under a second, "these creatures' skin is truly a window in." Cephalopods also produce rambling patterns with no obvious role while resting awake.
- Rat replay and preplay. Place cells fire when a rat is in a particular place, so their sequence reveals paths the rat represents without moving. Rats replay paths during sleep and preplay never-taken paths toward places where they found food; on waking they often follow the rehearsed route.
- The speed clue. In slow-wave sleep, replay runs about twenty times faster than walking the path; in REM sleep, paths taking minutes are replayed at a natural pace, place by place, as dreams seem to unfold. Not conclusive, but a bridge between animal offline processing and experience.
Octopuses sleep in two stages too, and their "dreams" look like waking. In 2021, Sylvia Medeiros and colleagues (iScience) described quiet sleep and brief bouts of "active sleep" in Octopus insularis, with skin colour changes, eye and arm twitches, recurring at roughly regular intervals. In 2023, Aditi Pophale, Sam Reiter and colleagues at OIST (Nature) recorded Octopus laqueus and found that during active sleep the skin cycles through patterns resembling waking camouflage and display, with brain activity resembling wakefulness, consistent with offline replay of waking states. The cephalopod "window in" has become a research tool.
Many animals spend long periods just sitting, and Godfrey-Smith doubts it is all blank or a monotonous slide of the present: brains' self-propelled dynamics have more going on. The likely human difference is not the existence of "elsewhere" experience but the extent of its deliberate control: executive control (directing oneself on a task, suppressing urges, marshalling abilities toward a represented goal), together with language as an organiser of thought, lets us set out deliberately for some particular elsewhere, though we still meander, drift and dream.
By Degrees: The Full Picture
The Ingredients, Reassembled
- Animal selves. Evolution produced beings with new sensing and action, subjectivity and agency, and a tacit sense of self. The sense of self is not "a turning on of the lights", but it shapes action: evolution puts a premium on coherent action, and beyond some point the way to act effectively as a self is to have a sense of oneself as that kind of unit, tacit at first and less tacit as behaviour grows complex.
- The brain as dynamic organ. The basis of experience is an organ with rhythms, fields and other large-scale dynamics, over and above cells talking in a network.
- The rising sea. How these fit together is still open; each partially bridges the gap, and, in the spirit of Grothendieck's image from Chapter 1, the problem should be slowly transformed as knowledge builds around it.
Too hard: all switches. The brain is only signals between cells, and "somehow that must be enough". The gap looks unbridgeable.
Too easy: a conscious cloud. Once global dynamics appear, one pictures a cloud of feeling extruded by the brain. The gap looks magically closed.
What is special is the combination: local cell-to-cell influences that organise input and coordinate action, plus large-scale activity patterns, all shaped by evolution for action and its control, not produced by nature spontaneously. The view is "neurocentric", hence sceptical of non-animal minds. Something else could perhaps do what nervous systems do, "but look at what they do, on all the many scales."
The Identity Claim and the Leap into Monism
If materialism is true, then the storms within cells, the threading together of their activity, the perturbed rhythms of their electrical breathing and their large-scale coordination are the stuff of mind. We are asked to identify with them: not that our minds are consequences of this activity, but that our minds are such activities. This is an imaginative leap into monism: "we are not extra, not additions to the physical world, but aspects of its workings. We are of these activities rather than merely tied to them or made by them."
Experience is the activity of a living system of the right kind, felt from the inside: "what it's like for a pattern of activity of the right kind to be you", not to look at it or describe it, but to be it.
The Experiential Profile
If you are a living thing of the right kind (and awake), you have an experiential profile at each moment: the way things feel as a whole, affected in larger or smaller ways by a huge amount of what goes on in you, and changing moment to moment. In a typical human it includes a sensory tableau (things seen, heard, felt) plus mood, energy level and a tacit sense of the body, with foreground and background: attend to a book and the hard seat and the fan's breeze recede, yet they still contribute, as do energy and mood. Profiles differ greatly between animals.
Gradualism
The final piece: minds and experience come slowly into view in evolution rather than appearing in one move. Nearly every relevant biological feature is a matter of degree with borderline cases around us now (cnidarians, gastropods, protozoa, plants), though having a nervous system is, interestingly, fairly clear-cut today. Even rapid evolutionary changes (for example, shifts in developmental timing) do not make important traits pop into existence whole.
- Subjective experience is (not is caused by) certain biological activities: animal selfhood plus a brain's large-scale dynamics.
- Those activities evolved gradually and come in partial and borderline forms.
- If experience is identical to something that admits degrees, experience admits degrees.
- So experience is not switched on at some stage; its "thereness" is a matter of degree: it can be partly there, partly not.
- A smooth physical change with a sudden mental jump on top would require brain activity to cause consciousness as something extra, which the identity view rejects.
The opposition: lights on or off. Many philosophers (in discussion, Michael Tye and others; in print, Jonathan Simon, and Tim Bayne, Jakob Hohwy and Adrian Owen, who call degrees of consciousness "of dubious coherence") insist: the lights in an ant, jellyfish or squid are either on or not. Even a dimmer switch is either a little on or off; richness may vary once the lights are on.
Godfrey-Smith's reply. "Lights on" is only a metaphor, and a bad one: it revives the project of conjuring colours from the brain's darkness. Nagel's "something it's like" is a decent way to point at the target but builds in a sharp something-versus-nothing contrast that shows nothing about the solution. Other bridging concepts, such as a point of view, can become definite gradually. Waking from sleep is an everyday model of gradual coming-into-consciousness; opponents reply that the vague experience must be fully real from some definite point.
New language is needed, but the picture is: in humans there just is felt experience; moving toward beings with less of what we have, subjectivity grades or fades away, and for some animals "Is it conscious?" has no yes-or-no answer. Nor is there one scale: simpler evaluation with complex sensing is not "more conscious overall" than the reverse. The right view will have several dimensions, each graded.
Against "Narrow Pathway" Theories
Much current work explains consciousness with an information-processing model in which experience happens only when information takes a special route: into working memory, a central "workspace", or a "world model", while most processing stays dark. (His main target is Stanislas Dehaene's global workspace; the family includes Jesse Prinz's AIR theory and Michael Tye's PANIC theory.)
- The supporting evidence. Masked words presented too briefly to be noticed still shape responses; blindsight patients handle objects they claim not to see. Much perception, action guidance and basic learning seems to run "in the dark".
- The extreme claim. Dehaene holds you can be conscious of only one thing at a time, switching rapidly; others allow more, but all routed through the privileged path.
| Narrow pathway views | Godfrey-Smith's experiential profile | |
|---|---|---|
| Target | What you are conscious of: the attended foreground | Everything about what it is like to be you now |
| Contents | One or a few items in the workspace | Sensory tableau plus mood, energy, bodily sense, background |
| Which brain activity counts? | Only information on the special route | Everything in the pattern contributes, much of it tinily |
| Animals | Pushes toward sparse animal experience | Different, not absent, in different brains |
| Danger | Treating the human mechanism as what experience is | Over-inclusion; needs a boundary for "the pattern" |
The reason "everything" counts: any alternative heads back toward dualism, toward experience as a product offered up by one part of the brain. Not everything inside the skull is part of the pattern, but many things making a tiny difference to pacing, attention or mood are part of experience. Godfrey-Smith suspects the two camps may simply be talking about different things; "conscious" fits the foreground better, since it invites the question "conscious of what?"
The tempting argument. 1. Experiments show humans can do many things unconsciously: various kinds of perception and processing. 2. Package all those abilities into one animal. 3. That animal does only what can be done unconsciously, so it is like nothing to be it. 4. Many real animals have roughly those abilities, so they are unconscious.
The reply. Every experiment was done on conscious human beings, who were not conscious of everything going on in them. The findings show that a lot can happen in the background of a conscious being, not that everything could be background at once. Premise 3 does not follow. For any wakeful human there is something it is like to be them despite the hidden work, and the same may hold for many animals.
Human-based theories are reasonable for humans, and extend to animals with similar brain parts (birds, fish). The error is to say "this is what experience is, so without it there is none". Once experience in hermit crabs and octopuses is likely, we need a broad account of what they and we share. That account also teaches about us: it recalls "those primordial, indistinct elements" beneath focused attention. "Human experience is a mixture of old and new."
The science of consciousness has moved toward testing, and toward Godfrey-Smith's worry about report. No-report paradigms, which infer conscious contents from physiological signals rather than verbal reports, have grown since the mid-2010s, partly to separate the neural basis of experience from the machinery of reporting it, a concern close to his distinction between profile and foreground. The Cogitate adversarial collaboration (results in Nature, 2025) found evidence that challenged key predictions of both global workspace theory and integrated information theory, without crowning either. The 2024 New York Declaration endorses the broad stance of this chapter: if a realistic possibility of experience exists in an animal, it is irresponsible to ignore it.
Consequences
1. Machines, AI and Uploads
Godfrey-Smith's target is "strong" AI: programs meant, when running, to be a mind, not just to behave like one, and its corollary, uploading minds into a computing cloud where they might flit from machine to machine.
Minds are patterns of activity, but far less "portable" than AI enthusiasts suppose: tied to a particular kind of physical and biological basis. You cannot create a mind by programming interactions into a computer, however complex and brain-inspired.
- A mind is a pattern of activity of the kind found in nervous systems.
- For a machine to have a mind, that activity must be present in it, not merely represented (encoded, described by equations).
- Signalling-network activity could be genuinely present: if part A affects B and C as neuron A does, the pattern is instantiated. If brains were only switching networks, strong AI might work.
- But brains also have large-scale dynamics: coordinated oscillations arising from tiny ebbs and flows of ions over membranes, and perhaps field effects.
- Running equations describing these rhythms is not having them; the machine must actually contain such patterns (similar, not identical, if the goal is a mind rather than a human one).
- It may be very difficult to have anything like these dynamics in a system not otherwise physically similar to a brain.
- So present computers, "fast and reliable logical processors attached to a large store of memory" in a stable, fully powered box, remain "a completely different object" from a brain, whatever the program.
- The represent/have objection. Often raised against strong AI and too glibly dismissed; Godfrey-Smith concedes it is weak for pure signalling but acute for large-scale dynamics.
- A logical sliver. Present computers contain "a kind of logical sliver, a tiny fraction" of what goes on in us; they are often designed to create illusions of agency and subjectivity, and do it well.
- Not anti-artificial. "The problem is not artificiality", it is having the right kinds of things going on inside. Future systems from different materials with more brain-like operations might be artificial life, or close.
- A spectrum of scenarios. Uploads are "just a fantasy": no current technology could carry your experiences from a living body into the cloud as more of you. At the other end, robots with genuinely brain-like control systems might one day have artificial experience.
He notes how the materialist story itself changes feel. Look at a grey blob and you doubt it could be you. Told "it's the activities: lots of signalling and switching", you are supposed to be satisfied, but are not much better off. Seeing the whole, cell-to-cell influences plus rhythms and fields modulated by the senses, "the situation seems entirely different": "It's no longer hard to credit."
The simulated storm. A weather model can reproduce a hurricane's every gust in equations without anything in the data centre getting wet. Now run a perfect simulation of your brain's ion currents, rhythms and fields on a server. A functionalist says the simulated brain, unlike the simulated storm, really thinks, because thinking just is information processing. Godfrey-Smith's position puts the brain on the storm's side for its large-scale dynamics: the oscillations and fields are physical events that must occur, not be computed. The thought experiment isolates the real disagreement: whether the features that make up experience are organisational (portable) or partly physical-dynamical (not). Neuromorphic or analog hardware that physically oscillates would be the interesting test case.
The AI question changed shape after 2020, and Godfrey-Smith's line has company. Large language models (from 2022) made "illusions of agency and subjectivity" a mass phenomenon. In 2023, Patrick Butlin, Robert Long and a group of nineteen authors (including Yoshua Bengio, Jonathan Birch and Eric Schwitzgebel) published "Consciousness in Artificial Intelligence", deriving indicator properties from leading theories (recurrent processing, global workspace, higher-order, predictive processing, attention schema, agency and embodiment) and concluding that no current AI system is a strong candidate, while finding no obvious technical barrier to building systems that satisfy the indicators. That computational-functionalist framing is exactly what Godfrey-Smith resists; Anil Seth's argument for biological naturalism about consciousness (2024 to 2025) is close to his. Jonathan Birch's The Edge of Sentience (2024) treats some AI systems, alongside invertebrates and patients with disorders of consciousness, as cases needing precautionary policy.
2. Animals: Consideration, Borders and Experiments
- Extend consideration. The book gives reason to weigh the welfare of many more animals. Consideration is not rights or equal status: it does not "install mosquitos, midges, and aphids as fellow citizens". A small step, but warranted.
- The border problem. Abandoning the human/animal divide tempts us to draw a new border: sentient versus not. Precautionary proposals (Jonathan Birch, 2017) say: once there is a reasonable chance of sentience, protect its interests. Often workable, but if gradualism is true, for many invertebrates "Is it sentient?" may lack a definite answer even in principle; new ways of thinking are needed.
- Varieties matter. He once suspected insects had sensory but no evaluative experience, which would transform their welfare status; he now thinks this may well be wrong (the moods of Chapter 8).
The science built on suffering. Godfrey-Smith acknowledges that much of the book rests on cruel experiments, made bland by review articles. Some he is fine with (Elwood's crabs). Some he turns away from, especially work on monkeys and cats (Hobson's dream research included very disturbing experiments on cats). The rat place-cell work is "almost a miracle" and won Nobel Prizes (2014: John O'Keefe, May-Britt and Edvard Moser), "but it's all been harder for the rats." His hope: that what we learn induces us to back off rather than press forward.
- Numbers are not everything. A handful of animals can matter, because the question is also what relationship we want with other sentient beings. Sadism is bad at any scale; experiments are rarely sadistic, but continuing to cause pain "just because it is interesting" creates "a bad relationship".
- The hard accounting. Knowledge often reduces suffering elsewhere, and practical progress often comes unanticipated from basic research, which makes the ledger very hard. Still, experiments can be kept informative while reducing probable experiential harm, and one can be glad knowledge was gained while not wanting such work ever done again.
Policy has moved toward the "extend consideration" line, especially for octopuses. Following the UK's 2022 inclusion of cephalopods and decapods in animal welfare law (Chapter 1), plans by Nueva Pescanova for the first commercial octopus farm in the Canary Islands drew scientific opposition, and in 2024 Washington State and then California banned octopus farming, the first such laws. In research, the EU has regulated cephalopod experiments since 2010 (Directive 2010/63/EU), so octopus and squid experiments there need ethical approval like vertebrate work.
The Shape of Mind
Finishing a dive, Godfrey-Smith watches two tiny decorator crabs dressed in algae fight in the shallows: tufts flying, neither backing down, flailing "seaweedily", an initially surprising, though ultimately comprehensible, display of agency.
| Picture | Where mind is | Who holds it | Godfrey-Smith's verdict |
|---|---|---|---|
| Desert | Almost nowhere: humans and perhaps some mammals; the rest of life mindless shells | Descartes, and many non-dualists | Wrong: sentience is far more widespread |
| Jungle: panpsychism | Everywhere, soul-like powers even in atoms | Panpsychists | Wrong: experience is built, not basic |
| Near-jungle: biopsychism | All life is sentient, plants and bacteria included (Haeckel's term, adapted) | Some biologists and philosophers | Wrong, but his "biggest recurring uncertainty" |
| Between | Minimal cognition everywhere in life; felt experience only in animals with nervous systems, graded | Godfrey-Smith | The book's conclusion |
Sensitive Souls Everywhere
The surprise is how far mind-like activity extends. Aristotle gave all life a nutritive soul (self-maintenance), animals also a sensitive soul (perceiving and responding), humans also a rational soul. Things could have turned out that way, with life sustaining itself blindly and the senses creaking open first in animals. Instead, "a sensitive soul is as widespread and perhaps as early as any": no living cell is oblivious to its surroundings, and minimal cognition spans much or all of the tree.
Why Not Biopsychism?
Here lies "an unnerving vista": a gradual slope into plants, fungi, non-neural animals, protists and bacteria. If sentience arises gradually and minimal cognition implies a kind of subjectivity, why isn't this slope the road on which feeling appears? Godfrey-Smith's answer: bacteria have minimal cognition, yet "feeling just seems to not be part of the picture"; going back and forth, he has come to see the importance of nervous systems and "the unique way they organize nature's energies."
One Origin or Several?
Option 1: several origins. Precursors exist across animals, but felt experience arose from its absence several times: in arthropods, some molluscs and vertebrates, perhaps more than once within some; animals like corals and bryozoans "just do not have what it takes". Godfrey-Smith's bet, though he is unsure what the hunch rests on.
Option 2: one ancient origin. A primordial form of experience arose once, early in animal evolution, was present before the radiations charted in the book, and developed differently down each line.
Resolving it meets "every possible problem": lost events, alien nervous systems, and the obscurity of what the simplest experience would be. The truth may be a blend not yet expressible. His bet: definite sentience in vertebrates, cephalopods and some arthropods; looking back along those lines, no sudden switching-on, but a process in which "the self becomes more definite, internal goings-on become more experiential, and subjectivity takes shape." "The only organisms that feel their lives have nervous systems, therefore only animals." Sentience is not everywhere, even within life, but there is a lot of it, "from sea angels (perhaps) to sea dragons (for sure)." The world is "more replete with experience than many people have countenanced."
The Mind Becomes a Thing: Entering the Garden
Most of the story avoided reifying the mind: an animal and its brain are objects, but its mind is an aspect of its workings. Offline experience changes that. The mind becomes something like an arena where we rehearse options, invent, and manipulate sonic and visual images; John Dewey (1920s) said the mind has a practical role and also a role as an "esthetic field" where scenes are built and tales told. Being there comes first; this is the advent of mind in a further sense.
The walled garden. At a 1989 American Philosophical Association symposium on Wittgenstein, Godfrey-Smith heard Crispin Wright criticise the picture of the mind as "a walled garden": a secret place its owner observes privately and infallibly. Wittgenstein argued that the usual philosophical picture of mind rests on illusions; Gilbert Ryle called it "the ghost in the machine"; minds are present in what people do. Yet some speakers felt the garden's appeal, and ordinary people talk this way, which matters because Wittgenstein meant to respect everyday talk before philosophy contaminates it.
Godfrey-Smith's closing turn: the speakers thought minds can't be hidden gardens. He thinks that, in us, they are, or something like it. Not as a general feature of sentience, which mostly consists of being there, but as a late, human-heavy elaboration: "We create an arena, a field, filled with both things that arise and things we furnish. We enter a garden."
Where the final synthesis is weakest. - Gradualism versus vagueness. Critics answer that conscious states might be graded in richness while being conscious stays sharp; Godfrey-Smith shows the opposing intuition rests on metaphor, but offers no positive account of what a half-present experience is, and admits new language is still needed. - The AI argument hangs on contested science. It depends heavily on large-scale dynamics, especially fields, which he himself calls a bet; if synchrony matters but fields do not, much of the dynamics is organisational and potentially reproducible in hardware. He also does not engage neuromorphic or analog computing, the obvious counterexamples to "fast logical processors plus memory". - Biopsychism is rejected by intuition. "Feeling just seems not to be part of the picture" for bacteria is honest but is the very kind of imaginative verdict Chapter 5 warned against. - "Everything contributes" to the experiential profile risks over-inclusion without a principled boundary for "the pattern". - Welfare is gestured at, not argued. "Consideration, not rights" leaves the hard trade-offs (farming, pest control, research) largely open, and the book's own reliance on gradualism makes threshold-based law harder to defend.
What this means for enterprise AI. Godfrey-Smith's position gives a defensible, science-based line for customer-facing assistants and agentic systems in regulated firms: current models produce "illusions of agency and subjectivity" by design, and there is a credible scientific view (alongside the 2023 indicator-property analysis) that no current system is a serious candidate for experience. Practically, that supports policies that forbid AI systems from claiming feelings or sentience to customers, frame them as tools in disclosures, and treat future claims of AI welfare as an emerging, monitored risk rather than a present obligation. His experiential-profile idea also offers a design caution: an AI "workspace" that surfaces the right facts is a narrow pathway, not an understanding of the customer's whole situation, which is why human relationship managers stay accountable for context.
Master Takeaways
Metazoa adds up to a concrete biological account of where experience comes from: not a mysterious extra, not a property of all matter, and not software, but the activity of animal selves and their dynamic nervous systems, felt from the inside. Its distinctive moves are to make subjectivity a product of agency, to treat brain rhythms and fields as part of the physical side of mind, and to insist that experience comes in degrees and varieties, which widens the circle of moral concern while denying minds to plants, bacteria and present-day computers.
The takeaways
- Experience is brain activity, not its product. Godfrey-Smith's identity claim: mental goings-on are biological processes of a certain kind, "what it's like for a pattern of activity of the right kind to be you"; looking for how brains cause a separate glow of consciousness is the error.
- The evolution of animal agency brings the origin of subjects. Bodies that act as wholes must sort self-caused from world-caused change, and that bookkeeping creates a point of view; subjectivity and agency evolved entangled, not back to back.
- Brains are dynamic organs, not telephone exchanges. Self-generated, synchronised rhythms (and perhaps electric fields that re-time the neurons that make them) are the second bridge between mind and matter; a bare signalling network makes the problem look harder than it is.
- Experience comes by degrees. Every biological ingredient is graded, so experience can be "partly there, partly not"; "lights on or off" and "something versus nothing" are metaphors, and the right picture has several dimensions, not one scale.
- Minds are far more widespread than the human-centred view, but not everywhere. Definite sentience in vertebrates, cephalopods and some arthropods, probably arising several times; minimal cognition across all life; but plants and bacteria are probably on a different road: neither desert nor jungle.
- Uploads are a fantasy; current AI is a "logical sliver". Minds are patterns, but not portable ones: the brain's large-scale dynamics must be physically present, not merely represented, so programs on conventional computers are not minds, though brain-like artificial systems might one day be.
- Life was explained by hidden organisation; mind can be too. The Bathybius fiasco ended the hunt for a special living substance; the rising sea method surrounds the mind-body problem with knowledge until it changes shape.
- Cells already did the groundwork. They impose order on a molecular storm, tame charge with voltage-gated channels (a transistor invented by bacteria) and live by traffic that makes them windowed: sensing is ancient and everywhere.
- Action came first on the animal path. The soft coral's coordinated reach, sheets of cells contracting together, was the key invention; sensing may come "for free" once parts monitor each other (Keijzer).
- Self-other bookkeeping is the root of perspective. The shrimp's feeler problem (subtract the touch your own motion predicts) is efference copy in miniature; Godfrey-Smith treats it as an opportunity to probe the world, not just Merker's liability.
- Pain evidence reached invertebrates through trade-offs. Elwood's hermit crabs tolerated more shock in better shells and less with predator odour present: a felt cost weighed in a decision, not a reflex.
- Imagination misleads in both directions. Zombie and disembodied-mind arguments exploit our ability to switch sympathetic imagining on and off (Nagel); the same move grants feelings to paramecia or denies them to fish.
- Qualia are not the paradigm of experience. Rothko-style colour fields idle the "where" system; ordinary experience is embodied, spatial and present, and includes moods, energy and a sense of presence: "more of life is felt."
- Integration is real but variable. The octopus, with two-thirds of its neurons in its arms, may switch between unified control and arm-level autonomy; split brains suggest a mind can contract, expand and divide quickly, with moods shared (partial unity).
- The Wada test shows the basis of mind can contract. With one hemisphere anaesthetised, the other is conscious alone; on Godfrey-Smith's reading it becomes a conscious system rather than having been one all along (Blackmon's view).
- Fish are social thinkers in centralised bodies. Archerfish learn by watching, cleaners guard reputations, groupers gesture to moray eels; the right question is relative advantage, not need, and the answer is social life.
- Sensory and evaluative experience can come apart. Insects show rich senses, no wound self-tending or morphine seeking, yet pessimistic and optimistic moods; life history (short, scheduled, hard-bodied lives) may have erased acute pain.
- Varieties of subjectivity carry ethical weight. Gastropods (healers) and crustaceans (long-lived) have more use for pain than insects; different ways of being a subject may warrant different treatment.
- Minimal cognition is not minimal sentience. A root turning to water uses information for its own projects, but without a self of the right kind there is no subject; plants are partly a "they", a colony of modules.
- Land combined what the sea kept apart. Manipulation, open action and central brains came together in land vertebrates, independently in dinosaurs and mammals; endothermy made body and brain high-energy systems.
- Most vertebrate brains are semi-split. Fish, reptiles and birds have weak left-right links and lateralised styles; only placental mammals have a corpus callosum, and dolphins slimmed theirs, perhaps to sleep one hemisphere at a time.
- The sea produced the foundations. Animals, senses, limbs, nervous systems and experience itself began in water; land added new rounds of innovation in sequence, not superior creativity.
- Experience is a profile, not a spotlight. Narrow-pathway theories (Dehaene's workspace) describe what you are conscious of; the experiential profile includes background mood, energy and bodily sense, so animals with different brains have different, not absent experience.
- Unconscious human abilities do not make zombie animals. Experiments on conscious people show much can run in the background of a conscious being, not that everything could be background at once.
- Offline experience makes the mind a garden. Memory, imagination and dreams are one offline capacity; cuttlefish REM-like sleep and rat replay and preplay suggest animals also go "elsewhere"; humans differ in deliberate control, making the private "walled garden" real in our case.
- Gradualism complicates welfare law. Extending consideration is warranted, but if sentience is graded, "Is it sentient?" may have no definite answer for many invertebrates, so precautionary borders need new thinking; and the knowledge gained from cruel experiments creates a relationship worth backing away from.
Frameworks & mental models
| Name | What it says | Use it when |
|---|---|---|
| Biological materialism (identity) | Experience is a kind of biological activity; brains do not cause minds, their activity is mind | Testing any claim that consciousness is "produced by" a system |
| Classic mind-body options | Materialism, dualism, epiphenomenalism, panpsychism, idealism, neutral monism | Locating a theory or a colleague's intuition |
| Arbitrariness problem / explanatory gap | Verdicts on who feels float free of evidence; why matter feels like anything at all | Spotting untethered claims about animal or AI minds |
| Rising sea (Grothendieck) | Surround a hard problem with knowledge until it transforms | Facing a problem that resists direct attack |
| Molecular storm (Hoffmann) | At cell scale the problem is imposing order, not producing motion | Reasoning about what "mere matter" can do |
| Taming of charge | Ion channels, voltage gating, action potentials; the cell's transistor | Explaining how biological control is built |
| Traffic / windowed life (O'Malley) | Living things must exchange with the world, so they sense | Asking why sensing is universal |
| Gastraea, updated | First animal as a cup around a gut and its microbes; a second-order cell boundary | Thinking about origins of multicellular selves |
| Cousins, not grandparents | Living simple animals are not ancestors | Using living species as evolutionary clues |
| Subjectivity and agency | "For-me-ness" and "by-me-ness" evolve together | Any discussion of what a subject is |
| Sensing for free (Keijzer) | Coordinated action makes a system sensitive to disturbance | Explaining how senses emerge from action |
| Waggoner's three assemblages | Avalon (fronds), White Sea (movers, scavengers), Nama (burrowers) | Periodising the Ediacaran |
| Bilaterian traction | Left-right bodies are built to go somewhere over surfaces | Explaining directed animal movement |
| Self/other (reafference) | Subtract self-caused input; the residue is the world | Building any agent that acts and senses |
| Merker vs Godfrey-Smith on mobility | Motion is a liability for senses / an opportunity to probe | Designing active sensing |
| Nociception-to-pain markers | Wound tending, analgesic seeking, learning, trade-offs | Assessing evidence of pain |
| Unitary vs modular bodies | Build one form bigger, or repeat units; modular means simple, stationary | Judging individuality and agency |
| Back-to-back vs subject-agent (Hurley) | Sensing and acting are interleaved, not sequential | Critiquing perceive-then-act designs |
| Animal as nexus | Causal lines converge, diverge, loop back, meet the past | Defining a point of view |
| Nagel's two imaginings | Perceptual and sympathetic imagining can be recombined at will | Defusing zombie arguments |
| Rothko experiences; what/where | Colour fields are anomalous; "where" processing grounds self-location | Choosing a model case of experience |
| Basal consciousness / presence | Consciousness as a state of being; feeling present | Including mood and thereness in experience |
| Lines of control | Evolution chooses separate streams, one stream, or cross-talk | Analysing decentralised nervous systems |
| Octopus selves: 1, 1+8, 1+1, switching | Count of subjects, and Godfrey-Smith's real-time switching | Thinking about distributed agents |
| Split-brain options | One mind, two minds, fast switching, partial unity | Interpreting divided-brain evidence |
| Wada contraction vs Blackmon | The basis of mind contracts / each half was conscious all along | Reasoning about parts and wholes in minds |
| Three views of rhythms and fields | Byproduct; synchrony matters; fields matter too | Weighing brain-dynamics claims |
| Large-scale dynamic properties | Rhythms, synchrony, fields as the second bridge | Explaining why brains suit experience |
| Telephone exchange vs dynamic organ | Signalling network vs self-generated activity modulated by senses | Challenging computational pictures of brains |
| Relative advantage / social intelligence | Ask what small gains pay for brains; social life usually does | Explaining cognitive evolution |
| Sensory vs evaluative experience | How things seem vs whether they seem good or bad | Assessing insects, AI welfare claims |
| Varieties of subjectivity | Life history shapes which kinds of feeling are useful | Comparing animals' moral claims |
| Minimal cognition vs minimal sentience | Information use for life's projects need not involve a subject | Assessing plants, microbes, simple systems |
| The trio combined on land | Manipulation, openness, centralisation together | Explaining vertebrate behavioural range |
| Vertebrate lateralisation | Left for food and categories, right for social and relations, weakly linked | Interpreting animal choices and split-brain analogies |
| Vermeij vs sequence view | Land more innovative / sea first, land next | Comparing evolutionary arenas |
| Kinds of memory; offline processing | Semantic, procedural, episodic, working; forward, backward, sideways simulation | Explaining imagination and planning |
| Being there vs elsewhere | Online engagement vs offline experience | Separating basic from elaborated minds |
| Experiential profile vs narrow pathway | Whole-state experience vs spotlight access | Comparing consciousness theories |
| Gradualism | Experience graded on several dimensions | Drawing (or refusing) sentience lines |
| Present vs represented | A mind's dynamics must occur, not be simulated | Evaluating strong AI and uploads |
| Desert, jungle, biopsychism, between | Where mind is distributed | Mapping views on panpsychism and plant minds |
| Several origins vs one | Sentience arose repeatedly or once, early | Evolutionary history of consciousness |
| Walled garden / esthetic field | Offline mind as a private arena, real in humans | Thinking about inner life and introspection |
Strengths. Metazoa makes materialism concrete and humane: it ties philosophical claims to specific animals, experiments and anatomy, defuses the strongest anti-materialist intuitions with Nagel's diagnosis of imagination, and integrates agency, embodiment and brain dynamics in a way few accounts do. Its insistence on degrees and varieties anticipated the direction of animal-sentience policy.
Blind spots. The explanatory gap is narrowed, not closed: "experience is activity felt from inside" restates identity rather than explaining why some activity has an inside, and Godfrey-Smith admits as much. The second bridge rests partly on a bet about electric fields that most neuroscientists do not share. Panpsychism and biopsychism are rejected largely by the cumulative plausibility of the story and by intuition ("feeling just seems not part of the picture" for bacteria), the kind of verdict the book warns against elsewhere. Gradualism is defended mainly by dismantling metaphors; no positive account of a half-present experience is given. Mammals and birds get little space, human uniqueness is deferred to a later book (Living on Earth, 2024), and welfare conclusions stay modest ("consideration, not rights").
Contested or dated. Some details have moved: jawed fish are older than stated (Silurian finds, 2022); Pikaia is a chordate rather than a fish ancestor; comb jellies now look like the sister of all other animals (2023); social insects do tend each other's wounds (2017 to 2024); mites are miscalled insects; fast switching and partial unity for split brains remain minority readings (and the 2017 Pinto study is not addressed).
How the field has moved. Toward the book on animals: the 2021 LSE review and UK Animal Welfare (Sentience) Act 2022 (cephalopods, decapods), the 2024 New York Declaration, octopus farming bans (Washington and California, 2024), octopus two-stage sleep (2021, 2023), and bird neural correlates of consciousness (2020). On theory, the IIT/global workspace adversarial collaboration (Nature, 2025) unsettled both leading "narrow" theories. On AI, large language models made the question practical: the 2023 Butlin, Long et al. indicator report took a computational-functionalist line Godfrey-Smith rejects, while biological naturalists such as Anil Seth echo him.
Who's who
| Name | Who | Why they matter here |
|---|---|---|
| Peter Godfrey-Smith | Philosopher of biology, diver | Author; biological materialism, gradualism, octopus fieldwork |
| T. H. Huxley / Ernst Haeckel | 19th-century Darwinians | Bathybius and protoplasm; Haeckel coined Metazoa, Protista, Gastraea, biopsychism |
| Alexander Grothendieck | Mathematician | The "rising sea" image of method |
| Thomas Nagel | Philosopher | "Something it's like"; two kinds of imagining; split-brain classic |
| David Chalmers | Philosopher | Zombie argument against materialism |
| René Descartes | Philosopher | Dualism; animals as machines |
| Aristotle | Philosopher | Grades of soul: nutritive, sensitive, rational |
| Peter Hoffmann | Biophysicist | The "molecular storm" |
| Maureen O'Malley | Philosopher of biology | Life on a "redox rollercoaster"; windowed cells |
| Fred Keijzer | Psychologist-philosopher | Action-shaping view; sensing for free; octopus as "natural Waterman" |
| Mary Droser, Jim Gehling | Palaeontologists | White Sea mats; Helminthoidichnites scavenging traces |
| Ben Waggoner | Biologist | Three Ediacaran assemblages |
| Robert Elwood | Biologist, Queen's University Belfast | Hermit crab trade-offs; invertebrate pain |
| Björn Merker | Neuroscientist | Mobility as sensory liability; gamma as upkeep |
| Susan Hurley | Philosopher | Back-to-back critique; unity through world-involving loops |
| John Searle, Rodolfo Llinás | Philosopher; neuroscientist | Basal consciousness; intrinsically active, rhythmic brain |
| Tom and Nicola Davis, Matt Lawrence | Divers | Coral time-lapse; discovery of Octopolis |
| David Scheel | Biologist | Octopolis and Octlantis research; octopus throwing |
| Benny Hochner, Tamar Gutnick | Neurobiologists | Octopus arm control and absent body map |
| Elizabeth Schechter | Philosopher | Two minds in split brains, partial unity |
| James Blackmon | Philosopher | Wada test: each hemisphere conscious all along |
| Juhn Atsushi Wada | Physician | The hemispheric anaesthesia test |
| Hans Berger | Psychiatrist | First human EEG, in search of telepathy |
| Angélique Arvanitaki | Neurophysiologist | Ephaptic coupling (1942) |
| Christof Koch, Costas Anastassiou | Neuroscientists | Field effects feeding back on neurons |
| György Buzsáki, Wolf Singer | Neuroscientists | Brain rhythms and synchrony |
| Bruno van Swinderen | Neuroscientist | Attention-linked rhythms in flies |
| Alexander Vail | Biologist | Grouper-moray referential gestures |
| Andrew Barron | Zoologist | Insects as disposable reproductive machines |
| Julia Groening, Melissa Bateson, Lars Chittka | Insect researchers | No morphine seeking; pessimistic and optimistic bees |
| Terry Walters, Robyn Crook | Neuroscientists | Nociceptive sensitisation; mollusc welfare |
| Giorgio Vallortigara | Neuroscientist | Vertebrate lateralisation, chick experiments |
| Geerat Vermeij | Biologist | Land as the engine of innovation |
| Endel Tulving, Donna Rose Addis | Psychologists | Episodic memory; offline processing |
| Stanislas Dehaene | Neuroscientist | Global workspace, the narrow-pathway foil |
| Jonathan Birch | Philosopher | Precautionary sentience policy; LSE review |
| Ludwig Wittgenstein, Gilbert Ryle, Crispin Wright | Philosophers | Ghost in the machine; the walled garden |
Connect it
- Perceive-plan-act pipelines are the "back-to-back" picture. Hurley and Godfrey-Smith argue capable agents interleave sensing and acting; enterprise AI agents that log their own actions and read the world's response against them (efference copy for systems) are more robust than one-shot chains.
- Subtract self-caused change before alerting. The shrimp's feeler problem is the core of good anomaly and fraud detection: a bank must model the effects of its own campaigns, releases and policy changes or it will flag its own footprint.
- Distributed agents need switchable control. The octopus suggests architectures where sub-agents act locally in routine conditions and a central controller pulls them together when risk demands coherence, with shared state (mood) visible across the system.
- A defensible line on AI sentience. Godfrey-Smith's "logical sliver" view, set against the 2023 indicator-property report, gives regulated firms a principled basis to forbid assistants from claiming feelings, treat AI welfare as a monitored emerging risk, and keep accountability with humans.
- Reputation under observation. Cleaner wrasse cheat less when watched; continuous, expected audit of AI and human agents in banking workflows, with downstream systems able to route around poor performers, mirrors how reefs enforce honesty without a regulator.