Insight Editions · No. 292
The Ancient Origins of Consciousness
How the Brain Created Experience
Todd E. Feinberg and Jon M. Mallatt
Feinberg and Mallatt argue that sensory consciousness is an emergent property of a living, neural, complexly hierarchical brain and that it is far older and more widespread than usually assumed.
A study guide to The Ancient Origins of Consciousness by Todd E. Feinberg and Jon M. Mallatt. 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
The Ancient Origins of Consciousness (MIT Press, 2016) is neurologist Todd Feinberg and evolutionary biologist Jon Mallatt's attempt to date, locate and explain the origin of experience. Their answer: sensory consciousness arose with the first vertebrates in the Cambrian, before 520 million years ago, probably also in the first arthropods and later in cephalopods, built from ordinary neurobiology arranged in complex, mapped neural hierarchies and subcortical affective circuits. Once the evolutionary history is added to neurobiology and philosophy, the authors argue, the explanatory gaps close and the subjective-objective divide becomes a natural, dated event rather than a mystery.
Neurobiological naturalism: sensory consciousness is an emergent property of a living, neural, complexly hierarchical brain; it is ancient, widespread and architecturally diverse, and the hard problem dissolves into auto- and allo-ontological irreducibility created by evolution.
Comparative neuroanatomy, paleontology, developmental genetics, animal behaviour and philosophy of mind, covering all vertebrate clades, amphioxus and tunicates, and the main protostome invertebrates, from the Ediacaran (about 580 mya) to modern humans.
Before 520 mya: first conscious fish. 4 subjective features (NSFCs). 2 steps: tectum then pallium (mammals 220 mya, birds 165 mya). 3 types of consciousness. 3 independent origins. 13 of 18 affect structures already in lampreys.
Readers who want an evidence-heavy, comparative answer to which animals feel, and a reusable checklist (Tables 2.1, 8.2, 9.2) for judging consciousness in animals or, by analogy, in engineered systems.
Splitting the mystery. Following Nagel, the target is primary (sensory) consciousness, "something it is like to be." The single explanatory gap is broken into four neuroontologically subjective features (NSFCs): referral, mental unity, qualia and mental causation. Philosophy, neurobiology and evolution each stall alone; the method is consilience, like the blind men pooling their views of the elephant.
What makes a conscious brain. Consciousness is built in three levels: the general features of life (embodiment, process, self-organisation, hierarchy with emergence and constraint, adaptation), reflexes (fast, tightly connected), and the special neurobiological features: complex neural hierarchies, a dual nested and non-nested organisation, isomorphic maps and affective states, reciprocal interactions, attention, and diverse architectures. Amphioxus, the best living model of the prevertebrate, has only a ventral, reflexive core brain with three or four neuron levels.
Why the Cambrian. The first predators set off an arms race on a newly busy seafloor. Only mobile arthropods and vertebrates paid for expensive senses and sensory maps. In vertebrates, vision came first: eyes grew from the diencephalon, neural crest and placodes rebooted the other senses, and new dorsal brain regions (cerebrum, optic tectum, cerebellum) received isomorphically mapped input along a shared five-stage pathway.
Two steps in vertebrates. In Step 1 the first fish built mapped multisensory images in the optic tectum (all senses but smell) and smell images in the pallium, backed by arousal (reticular activating system) and attention (isthmus-tectum); lampreys have it all, so every vertebrate is conscious. In Step 2 the dorsal pallium, where all senses meet the hippocampus's memories, expanded convergently in mammals (220 mya) and birds (165 mya) and became the sole end-site of consciousness.
Feelings. Consciousness comes in three types: exteroceptive (mapped, outward), interoceptive (dual, body states) and affective (unmapped, valenced, referred to the self). Strict operant-learning criteria (trade-offs, frustration, self-medication, drug seeking) and the distribution of limbic structures show subcortical affect in all vertebrates from the start; only suffering C-fibre pain seems to have come late, with tetrapods.
Beyond the backbone. Applying seven criteria for exteroceptive consciousness and the affect criteria, arthropods are probably conscious (fifth-level retinotopy in flies; bees that hold a picture), held back only by tiny brains; cephalopods are possibly conscious, lacking proof of nonvisual mapping. Worms and gastropods are not. Consciousness therefore probably evolved three times.
The solution. Three postulates: consciousness is explained by known neurobiology; it is ancient, widespread and diverse; and its subjectivity follows from a natural confluence of neurobiological and evolutionary events. The subject cannot experience its own neurons (auto-ontological irreducibility) and an observer cannot experience the subject's experience (allo-ontological irreducibility), both created when reflexes became images and feelings. The explanatory gaps are closed; the ontological gap is bridged, not closed.
| Ch | Title | One-line payload |
|---|---|---|
| P | Preface | Two specialists propose that experience began in the Cambrian and that the hard problem was born with it. |
| 1 | The Mystery of Subjectivity | The target is primary consciousness; the gap splits into four NSFCs, attacked by three combined domains. |
| 2 | The General Biological and Special Neurobiological Features of Consciousness | Consciousness = life + reflexes + special features (hierarchies, nested binding, maps, interactions, attention, diversity). |
| 3 | The Birth of Brains | The prevertebrate brain, read from amphioxus, was a ventral reflexive core with no dorsal sensory centres. |
| 4 | The Cambrian Explosion | Predation started an arms race that rewarded sensory maps in mobile arthropods and vertebrates. |
| 5 | Consciousness Gets a Head Start | Vision first, then neural crest and placodes and new dorsal brain; one five-stage mapped pathway for every sense. |
| 6 | Two-Step Evolution of Sensory Consciousness in Vertebrates | Tectal consciousness in the first fish (lampreys included), then pallial consciousness in mammals and birds. |
| 7 | Searching for Sentience: Feelings | Exteroceptive, interoceptive, affective: three types; rival theories date affect from 560 mya to 200,000 years ago. |
| 8 | Finding Sentience | Operant criteria and limbic anatomy put affect in all vertebrates from the start; suffering pain came with tetrapods. |
| 9 | Does Consciousness Need a Backbone? | Arthropods probably, cephalopods possibly, worms and snails not: consciousness evolved about three times. |
| 10 | Neurobiological Naturalism: A Consilience | Three postulates; auto- and allo-irreducibility; explanatory gaps closed, ontological gap bridged. |
Preface
Todd Feinberg, a clinical neurologist and consciousness researcher, and Jon Mallatt, an evolutionary biologist of early animals, joined forces in 2013 to ask when experience first appeared on Earth and which animals have it. Their answer, previewed here, is bold: sensory consciousness arose in the Cambrian, roughly 560 to 520 million years ago, in the first vertebrates and probably also in the first arthropods, and the "hard problem" was born with it.
The questions and the target
The book asks four linked questions: how is consciousness created, when did it first appear, how did it evolve, and which living animals have it? The target is primary or sensory consciousness, the most basic kind, which means having any experience at all. The starting point is Thomas Nagel's criterion that a conscious animal has "something it is like to be" it, and the goal is to find the evolutionary origin of two philosophical markers of consciousness: John Searle's "ontological subjectivity" and David Chalmers's "hard problem" (how a physical brain produces personal experience).
The authors write for a broad readership but promise not to skimp: full vertebrate brain anatomy, the philosophical axioms of conscious states, the Cambrian explosion, sensory pathways, learning, the origins of pleasure and pain, and a good deal of invertebrate biology.
The argument in outline
- Part one: correlates. The opening chapters lay out the philosophical puzzles and assemble a list of neural features that seem responsible for consciousness, formalised as neurobiological naturalism, a theory Feinberg had proposed earlier and which later chapters revise.
- The Cambrian explosion. Between about 560 and 520 mya all known animal phyla appeared, along with the first complex nervous systems, the first complex behaviour and the first evidence of animal cognition.
- Mapped exteroceptive consciousness. In the first fishlike vertebrates, reflexes gave way to a unified "inner world" of mapped images from the distance senses (vision, hearing, smell). This is the first appearance of qualia.
- All vertebrates. Against long-standing taboos, the authors conclude that every fish, amphibian, reptile, bird and mammal is conscious. A large, memory-enhanced advance came in the first mammals and birds, when the main site of consciousness shifted from the midbrain tectum to the cerebrum.
- Sentience. A second aspect, affective consciousness (positive and negative feelings), is traced by behavioural and brain markers and also found in all vertebrates, past and present.
- Invertebrates. Applying the same criteria, arthropods (insects, crabs) and cephalopods (octopus) meet many of them, implying that consciousness evolved independently in the first arthropods and first vertebrates over half a billion years ago.
- The solution. The final chapter upgrades the theory and locates the answer to the mind-brain problem in two transitions: reflex to image and reflex to affect, both over 520 million years ago.
- Qualia. Why experiences have subjective felt qualities at all.
- Unity. Why experience is unified, though the brain is made of separate parts.
- Referral. Why conscious brains place experience out in the world or in the body, and never experience their own neurons firing.
- Mental causation. How an apparently immaterial consciousness can cause changes in the material world.
Why three approaches are needed
The book's claim to originality is methodological: consciousness becomes understandable only when the evolutionary, neurobiological and philosophical approaches are combined, because each has blind spots that only the other two reveal. Biology solves problems by reduction to parts and their interactions; philosophy shows that such reduction cannot dissolve subjectivity; evolutionary history then explains why consciousness is both irreducible and natural.
The preface already commits to a strong, testable thesis: consciousness is old (Cambrian), widespread (all vertebrates, probably arthropods and cephalopods) and diverse, and its philosophical mystery can be explained without new physics.
Since 2016 the wide distribution of consciousness has gained institutional support. The UK government's 2021 sentience review (led by Jonathan Birch at the LSE) concluded that decapod crustaceans and cephalopod molluscs are sentient, and they were brought into the Animal Welfare (Sentience) Act 2022. The 2024 New York Declaration on Animal Consciousness stated strong support for conscious experience in mammals and birds and a "realistic possibility" of it in all vertebrates and many invertebrates, including insects. Both fit the Feinberg-Mallatt picture, though neither endorses its specific Cambrian dating.
The Mystery of Subjectivity
The chapter fixes the target, sensory (primary) consciousness, and explains why it resists explanation. Philosophy, neurobiology and evolution each hit an "explanatory gap" when working alone, and the authors break the single gap into four neuroontologically subjective features (referral, mental unity, qualia, mental causation) that any theory must explain. Their proposal is to use the origin of the hard problem itself as a dated marker for the origin of consciousness.
"Something it is like to be"
Thomas Nagel's 1974 paper "What Is It Like to Be a Bat?" supplies the criterion used throughout: an organism has conscious states if and only if there is something it is like to be that organism, something it is like for it. The test is not intelligence or brain size but whether the animal has subjective experiences. Getting inside a bird's brain, the question would be whether anything is felt there, not how cleverly it solves problems.
The aspect of consciousness closest to Nagel's phrase, and hardest for science, is sensory consciousness, also called phenomenal, primary or perceptual consciousness, or the experiencing of qualia (perceived qualities). Antti Revonsuo's definition sets the bar low on purpose: the mere presence of any pattern of subjective experience, simple or complex, faint or vivid, fleeting or lingering, is necessary and minimally sufficient. Sensory consciousness is what lets a brain build an "inner world" with a single point of view.
- Higher consciousness. Elaborate human-style reflection on the world is set aside.
- Self-consciousness. Awareness of oneself as a self is not required for primary consciousness.
- Access consciousness. Ned Block's term for contents available for verbal report and reasoning; distinct from phenomenal experience.
- Reflective consciousness. Awareness of one's own thoughts is excluded.
- Intelligent consciousness. Problem-solving capacity is treated as a separate question.
- Theory of mind. Modelling other minds is a later, separate achievement.
The authors part company with Antonio Damasio's framework here. Damasio's four perspectives (introspective, behavioural, brain, evolutionary) centre on the self and the human brain; Feinberg and Mallatt doubt that primary consciousness needs any real sense of self (following Gerald Edelman and Anil Seth) and aim to explain experience, not selfhood.
The hard problem and the explanatory gap
David Chalmers named the "hard problem": why visual and auditory information processing is accompanied by experience at all, the quality of deep blue or the sensation of middle C. Everyone agrees experience has a physical basis, yet there is no good account of why and how it arises. Joseph Levine (1983) called the residue the explanatory gap: however detailed the third-person story of neural pathways, the personal experience itself is left out. In effect, the hard problem is the problem of explanatory gaps. (Schopenhauer's older name for it, revived by Globus, was the "world knot.")
The blind men and the elephant. In this Indian parable (one version is attributed to the Buddha), blind men each touch one part of an elephant: trunk (a snake), tail (a rope), tusk (a spear), ear (a fan). Each clings to his view and they come to blows. Feinberg and Mallatt draw two lessons: each man cannot escape his own point of view, and, more importantly, each discipline gives a partial answer. The remedy is consilience across approaches.
Three approaches, each incomplete
The hard problem stays mysterious, the authors argue, because each field attacks it with only its own questions and answers. They combine three domains: philosophical, neurobiological and neuroevolutionary, the last split into a paleobiological domain (fossils, deep time) and a domain of living animals.
The philosophical approach: four subjective features
John Searle locates the difficulty in the "first-person ontology" of consciousness ("ontology" meaning being or fundamental nature). Experiences exist only when felt by some agent, so you can neither reduce neuron firings to feelings nor feelings to firings: each reduction leaves out the objectivity or subjectivity in question. Nagel makes the same point: subjectivity is tied to a single point of view, and any objective theory must abandon that point of view.
Close analysis shows several gaps, not one. The authors call them the neuroontologically subjective features of consciousness (NSFCs) (in earlier papers, the "neuroontologically irreducible features," NOIFs). Each NSFC is (1) ontologically subjective and (2) shows an explanatory gap between how the brain experiences it and how it looks from the third person.
| NSFC | What is experienced (first person) | What is observed (third person) | The puzzle |
|---|---|---|---|
| Referral | Sights and sounds out in the world; pain in the foot; sadness "in me" | Neurons firing inside the skull | Why is no experience ever felt as brain activity? |
| Mental unity | One seamless field of experience from one viewpoint | Billions of separate, spatially extended neurons | How do divisible parts yield one undivided experience? |
| Qualia | Redness, pain, stink | Neurons that are not red, painful or stinky | How do felt qualities arise from unfelt cells? |
| Mental causation | My intention moves my arm | Only material neurons drive muscles | How can the immaterial-seeming mental cause physical change? |
Referral
No experience refers to the brain itself: you never perceive neurons firing and are not aware of firings at all. Experience is instead referred outward to the world or to somewhere in or on the body.
- Projicience. For the special senses (seeing, hearing, tasting, smelling), Charles Sherrington called this projicience: sensation is externalised to the stimulus source. You do not feel a sight in the eye, or a sound in the ear, or either in the brain.
- Interoceptive referral. Pain or the stretch of a full stomach is referred away from the brain to a body location.
- Affective referral. Positive and negative feelings are referred to the self as a whole, not to the outside world and not to the brain. When sad, you do not feel your neural circuits as sad.
Mental unity
The brain objectively consists of billions of neurons, yet consciousness is experienced as a single unified field. This is the "grain argument" (Wilfrid Sellars): the brain is like grains of sand, experience like the whole beach. The classic case is cyclopean perception: two eyes deliver two images, but we see one world as if from a single central eye. Descartes reasoned that since the sense organs are paired but the soul is one, the single pineal gland must do the unifying; it actually secretes melatonin.
The split screen. Imagine wiring each of your eyes to a separate monitor in a control room. An engineer sees two slightly offset images, two streams of numbers. You, behind the eyes, see one scene with depth. Nothing in the engineer's view looks like "one," and nothing in yours looks like "two." The unity is real in experience and absent in the hardware description, which is exactly why the authors count unity as its own gap rather than folding it into qualia.
Qualia
Qualia are the felt qualities of sensory consciousness: a colour, a sound, a smell, or a negative affect. Francis Crick and Christof Koch called them the most difficult aspect of consciousness and judged it "fruitless" to attack them head-on. Neurons are not red, painful or stinky; the states associated with them are. For most philosophers and neuroscientists this is the most perplexing gap.
Mental causation
Jaegwon Kim framed this as explaining how the mental can exercise causal influence in the physical world; Revonsuo asks how mind can change a purely material neural process. If consciousness is intangible and unobservable, how does it build the Great Wall of China or the Panama Canal?
Splitting "the" explanatory gap into four NSFCs gives the book a checklist: a successful theory must account for the neurobiology and evolution of referral, unity, qualia and mental causation, not just qualia.
The neurobiological approach: reduction or emergence
Biologists have two tools for relating wholes to parts: reduction and emergence.
Ontological reduction, in Searle's words, shows that objects of one type are "nothing but" objects of another type (chairs are nothing but molecules, genes nothing but DNA), where "molecules" includes their fields of force and interactions. Biology does this constantly. Digestion reduces to the stomach, intestines and their interactions; Watson and Crick's DNA structure de-mystified heredity. Neurology works the same way: there is no ontological mystery to epilepsy (abnormal electrical discharges) or paralysis (cut nerve fibres). Biologists do not fully understand every reaction, but they neither meet nor expect reductive gaps anywhere except consciousness.
Emergence comes in two strengths:
- Weak emergence. Complex systems have new higher-order properties that are still explained by their parts. Water is fluid though no molecule is; muscle contraction, hormonal regulation and circulation are weakly emergent. Even chaotic systems like weather are unpredictable in practice but natural in principle.
- Strong (radical) emergence. Some system properties can never in principle be reduced to the parts. If consciousness were radically emergent, dissecting it would always leave a remainder, which would be the hard problem.
The radical camp
- Roger Sperry. The Nobel laureate held that conscious phenomena are "different from, more than, and not reducible to, neural events."
- Erwin Schrödinger. The quantum pioneer doubted that any physiologist, however complete his knowledge of retina and optic nerve, could account for the sensation of colour.
- David Chalmers. Since all of physical theory is compatible with the absence of consciousness, he suggests taking experience as fundamental, alongside mass, charge and space-time.
- Quantum and exotic naturalists. Others call consciousness natural but explained only by quantum effects, magnetic fields, microtubules (Hameroff and Penrose) or single neurons.
The authors reject all of these. Their thesis is that consciousness is explained by naturally occurring but unique biological features, so no new "fundamental" or "radically emergent" ingredient is needed.
The neuroevolutionary approach
Some thinkers have brought evolution into consciousness studies (Butler, Damasio, Griffin, Llinás, Merker, Panksepp, MacLean and others), mostly with mammals as the reference point. Feinberg and Mallatt add much more hard data on non-mammals: the early evolution of brains, the comparative biology of brains and senses across phyla, and the developmental genetics of nervous systems. Their distinctive move is to search for the evolutionary origins of the gaps themselves: if the origin and basis of sensory consciousness can be explained at the date the hard problem first appeared, using conventional biology, the philosophical mystery can be resolved.
The evolutionary framework is standard Darwinism: organisms overproduce, offspring vary, and the fittest variants leave more offspring. The authors add that mutation, genetic drift, niche construction and developmental bias also shape evolution (citing Laland and colleagues, 2014).
The paleobiological domain: how old is consciousness?
Consciousness did not leap onto the stage from nowhere. At some point only animals without the neural apparatus existed; later, some could integrate working neurons into a fabric of experience.
The choice of criteria decides the answer. If consciousness tracks cognitive capacity, it is recent and confined to the smartest species. Intelligence has the advantage of being objectively measurable, but the authors list its weaknesses:
- It misses the hard problem. Cognitive scores say nothing about subjectivity or qualia.
- No a priori link. Cognition and consciousness may roughly co-vary with brain complexity, but nothing requires it.
- Seeing is not smarter in humans. There is no reason to think a human "sees" a tree more acutely than an orangutan at the level of primary consciousness.
- Babies. A human infant, with intelligence undeveloped, has sensory experience much like an adult's.
- Affect is ignored. Reasoning, communicating and remembering do not address the unpleasantness of cold, pain or hunger.
The authors' counter-thesis is that consciousness is very old, and they will argue that 560 to 520 million years ago (mya) was the critical window for the first vertebrates and conscious brains, tentatively with some invertebrates too. They claim this would be the first time anyone has dated consciousness, explained its architecture and found its most basal forms at once.
- c. 4,500 mya Earth forms (Hadean eon).
- c. 3,800 to 3,500 mya Origin of life; first prokaryotes (bacteria and archaea).
- c. 1,800 mya First eukaryotes.
- Late Proterozoic (Ediacaran) Sponges, then the first bilaterians.
- c. 560 to 520 mya The Cambrian explosion: the authors' proposed dawn of sensory consciousness.
- c. 541 mya onward Phanerozoic eon (Paleozoic, Mesozoic, Cenozoic eras).
The domain of living animals: who has it?
Which living descendants of those pioneers are conscious? A single-celled amoeba approaches food chemicals, withdraws from noxious ones, responds to light and to heat versus cold, yet has no nervous system where integrated experience could form; to most biologists there is nothing it is like to be one. Unless only humans are conscious (a view held by Peter Carruthers, Daniel Dennett and Joseph LeDoux, but increasingly rejected), consciousness lies somewhere between amoeba and human. Most researchers grant it to chimpanzees; many to all mammals and birds; some to octopuses and honeybees. The authors recast the question as: what neuronal organisation, and how much complexity, does a nervous system need for subjectivity?
Toward neurobiological naturalism
The authors start from Searle's biological naturalism (The Rediscovery of the Mind, 1992): mental phenomena are caused by neurophysiological processes and are features of the brain, as biological as digestion, mitosis, meiosis or enzyme secretion. They agree mind rests on ordinary matter, so radical emergence is wrong. But Searle leaves the key question unanswered: why can't experience be reduced to neural processes the way digestion reduces to the gut? Without that reduction barrier there would be no mind-body problem at all.
Biological naturalism (Searle). Consciousness is caused by brains and is fully biological, like digestion. The first-person/third-person barrier is asserted but not explained.
Neurobiological naturalism (Feinberg and Mallatt). Agrees, but holds that there are real differences between digestion-type processes and the neural processes of consciousness, and that these very differences both generate experience and explain the reduction barrier.
The solution, they say, requires integrating philosophy, neurobiology and neuroevolution, and the book's structure follows from this.
| Chapters | Job |
|---|---|
| 2 | Link neurobiology to philosophy: the features that separate reflexes from conscious brains |
| 3 to 8 | Date the evolution of exteroceptive, interoceptive and affective consciousness in vertebrates |
| 9 | Apply the accumulated criteria to invertebrates |
| 10 | Unite the three domains, upgrade the theory, address the hard problem |
Two framing choices deserve scrutiny. First, the NSFC list assumes the gaps are real features to be explained. Illusionists (Keith Frankish, Daniel Dennett) argue instead that qualia as philosophers describe them do not exist, so the job is to explain why we think they do. Second, recasting "which animals are conscious?" as "what neural organisation is needed?" builds in the assumption that consciousness requires neurons, excluding plants and brainless organisms by definition rather than by argument. Both choices are defensible but are premises, not findings.
Chalmers himself shifted the debate in 2018 with the "meta-problem of consciousness": explaining why we judge that there is a hard problem. This sits close to the authors' NSFC of referral and to their later claim that the gap has a natural explanation. On the "only humans" side, LeDoux (with Richard Brown, 2017) developed a higher-order theory of emotional consciousness, keeping alive the view that felt emotion depends on cortical higher-order representations that most animals may lack.
The General Biological and Special Neurobiological Features of Consciousness
Consciousness, on this account, emerged in stages, each stage keeping the features of the ones below. The authors sort the ingredients into three levels: general features of all life, the reflexes of all animals with nerves, and a set of special neurobiological features found only in conscious animals. The special features (complex neural hierarchies, a dual nested and non-nested organisation, isomorphic maps, reciprocal neural interactions, attention, and diverse architectures) are the missing ingredients that turn nonconscious nervous systems into experiencing ones.
Three levels on the road to consciousness
The staging idea comes from systems theory: in evolutionary hierarchies, later levels are added on top of earlier ones, which persist. The general features are necessary but never sufficient, singly or together. Reflexes add what nervous systems bring but create no experience. The special features are not "radically" different from other emergent properties of biology, yet in combination with the lower levels they supply the advanced and unique properties that make consciousness possible.
Level 1. General biological features (all living things) 1. Life: embodiment and process. 2. System and self-organisation. 3. Hierarchy, emergence and constraint. 4. Teleonomy and adaptation.
Level 2. Reflexes (all animals with nervous systems) 5. Rates and connectivity. Fast signalling and tight synaptic links.
Level 3. Special neurobiological features (animals with sensory consciousness) 6. Complex neural hierarchies; a brain. 7. Nested and non-nested hierarchical functions. 8. Neural hierarchies create isomorphic representations and mental images and/or affective states. 9. Neural hierarchies create unique neural-neural interactions. 10. Attention. 11. Sensory consciousness may be created by diverse neural architectures.
Level 1: general features of all living things
Life
The only things we know to be conscious are alive, and consciousness ends with the death of the brain. Ernst Mayr, in The Growth of Biological Thought (1982), called attempts to define "life" futile, because no special substance or force corresponds to it, though the process of living can be identified. (He later gave his own list: reproduction, metabolism, replication, adaptedness, growth and hierarchical organisation.) The authors think the definition is easier than Mayr said: all cells are alive; the simplest are bacteria and archaeans; all living cells use energy to sustain and organise themselves, respond, reproduce, have genes, and evolve by natural selection; and all have a membrane separating inside from outside.
Embodiment
Embodiment means a living thing is a distinct entity with an interior, separate from its surroundings. It grew more elaborate through Earth's history:
- Bacteria and archaeans. The oldest fossils, about 3.5 billion years old, are single cells in the oceans.
- Eukaryotes (about 1.8 billion years ago). A large archaean engulfed an energy-efficient bacterium that survived as the mitochondrion; the energy boost let the composite cell evolve many new functions, a clear emergent property of life.
- Single-celled eukaryotes. Amoebas and choanoflagellates (the closest single-celled relatives of animals).
- Colonies and true multicellularity. Cells clustered, partly because larger size protects against predation, then specialised into tissues, giving sponges, then the wormlike bilaterian ancestors of most animals.
Embodiment has philosophical weight. Life and consciousness both depend on a material body or brain with borders, so their features are internal and "ontologically individual": specific to that one organism. Evan Thompson argues that life already sits "beyond the gap" between internal and external, since a living being embodies an interiority of its own immanent purposiveness; for him the inside/outside border lies at the body surface, not the brain surface.
Process
Life and consciousness are processes, not things. Mayr warned that "life" and "mind" are reifications of activities, and that treating them as entities invites a soul-like separate existence. William James (1904) made the same point about consciousness: the word does not stand for an entity or "aboriginal stuff" but for a function, the function of knowing. The authors' summary: consciousness is part of what a living brain does, and every general and special feature is a functional feature of a particular embodiment.
System and self-organisation
Living organisms are self-organising systems. In the definition of Scott Camazine and colleagues, global pattern emerges solely from numerous interactions among lower-level components, not from an external ordering influence. Self-organisation marks all complex adaptive systems, so it is a necessary but insufficient condition for consciousness.
Hierarchy, emergence and constraint
Complex biological systems are hierarchies that grow more complex toward the top, and most evolutionary change happens at the higher levels. The genetic code is nearly universal; organelles are shared by all animal cells; tissues and organs vary more; whole bodies (worm, shrimp, iguana) vary most. The authors' explanation: selection acts on every level but is strongest at the top, on the individual organism that meets the environment directly; lower levels are buffered inside the body.
- Emergence (upward). Parts produce new properties at higher levels. Jaegwon Kim described emergentism as a layered world, each level emergent from and dependent on the one below. Examples: physiology from cell chemistry; multicellular bodies from single cells; an embryo becoming a newborn; the vertebrate forebrain, which evolved large new centres at the top of the neuraxis (brain atop spinal cord) in reptiles, birds, mammals and some fish, the mammalian cerebral cortex being the best-known case.
- Constraint (downward). Higher levels control lower levels. Howard Pattee (1970) called the origin and operation of hierarchical constraints "the essential and distinguishing characteristic of life", at every scale from molecule to brain. Cells constrain organelles, organs constrain cells, the body constrains organs; if constraint fails at any level, the body disassembles and dies.
- Limits on constraint. Control is not total: insulin-secreting pancreatic cells work with little top-down influence. And lower levels limit what the whole can do: a body made mostly of water and soft elements cannot survive freezing, boiling or cannon balls.
Teleonomy and adaptation
Saying "limbs evolved for locomotion" is shorthand; taken literally it is teleology, making the end result the cause of the process, a logical error. Evolution adapts to the now and seeks no future goal. Jacques Monod's term teleonomy names real function (a tooth's function is to chew) without implying design; the authors use "teleonomic" as a synonym for adaptive. (Mayr, by contrast, reserved teleonomy for programmed, goal-directed movements and kept it distinct from stationary adaptations.)
From the start, sensory consciousness is treated as an adaptation. The first argument comes from Shaun Nichols and Todd Grantham (2000): in evolutionary theory, structural complexity is evidence that an organ is an adaptation even if its causal role is unknown, because complexity is expensive to build and maintain and would be selected against if useless. Consciousness is structurally complex because its circuits combine many diverse sensory inputs into one unified scene.
Level 2: reflexes in all animals with nervous systems
Reflexes are fast, automatic responses to stimuli. They are not conscious, but their neuronal chains are the raw material from which the circuits of consciousness evolved.
- Monosynaptic reflex. Two neurons joined at a single synapse, as in the knee-jerk reflex that helps balance; it works even in coma.
- Polysynaptic reflex. Multineuron chains, as in the pupillary light reflex (a six-neuron chain, sensory neuron first, motor neuron sixth), which shrinks the pupil to protect the retina.
- Elaboration. Over evolutionary time, reflexes grew by adding links to the middle of the chain, with much cross-talk, yielding more hierarchical levels. On the motor side, chains became motor programs and central pattern generators (CPGs) for rhythmic locomotion, feeding and breathing, tuned by simple sensory feedback.
The authors call reflexes the "royal road" to consciousness for two reasons. Speed: some living hierarchies are slow (a saguaro cactus growing for centuries), but nervous systems are fast, and consciousness needed that rate. Connectivity: tight synaptic links let many signals flash back and forth through the hierarchy. Reflexes even foreshadow referral: a blink against blowing dust or a worm's withdrawal from a cut is processed inside the body but is already a response aimed at the external world.
Level 3: special neurobiological features
The special features mark the transition from reflexes and motor programs to consciousness, both historically (an evolutionary event) and comparatively (conscious versus nonconscious animals and systems). They are absent, or only primordial, in nonconscious animals and in reflexive circuits. This junction between reflex and special features is where the explanatory gaps and the hard problem are created.
Complex neural hierarchies
Because brain damage disrupts it, consciousness is unlikely to be "fundamental"; it needs a great deal of neural complexity. Herbert Simon, a founder of complexity theory, called hierarchy one of the central structural schemes "the architect of complexity uses." Hierarchies grow complex by adding levels, adding specialised components per level, and adding interactions between levels. For brains that meant more hierarchies, more levels and more fast processing within and between levels, yielding richer behaviour, more learning and, the authors claim, consciousness.
Giulio Tononi's measure fits this picture: the amount of consciousness is the integrated information generated by the system beyond that generated by its parts. With Christof Koch, he stresses organised interactions and feedback between centres, which the authors endorse.
Cellular diversity adds complexity:
- Neuron types. They differ in the number and branching of dendrites (receiving) and axons (sending), and in signalling, firing tonically (throughout a stimulus) or phasically (only at onset). Invertebrate neurons blur roles further: dendrites sometimes send and axons sometimes receive.
- Vertebrate receptors. Vertebrates have an unusually wide range of sensory receptors, most built from two vertebrate-only embryonic tissues: neural crest and ectodermal placodes.
- Nothing else comes close. The mammal brain has hundreds of neuron types; muscle, liver, kidney and ovary have under a dozen cell types each.
Nested and non-nested hierarchies
This is the book's key structural idea. All hierarchies show emergence (lower levels producing higher properties) and constraint (higher limiting lower), but they come in two kinds.
| Feature | Non-nested hierarchy | Nested (compositional) hierarchy | Neural nested hierarchy |
|---|---|---|---|
| Physical relation of levels | Separate, though connected | Higher levels made of lower levels | Physically separate levels |
| Information flow | Bottom-up convergence, top-down command | Rich interaction within and between levels | Both: convergence plus distributed binding |
| Constraint | Strong, central control | Weak, holistic, no central command | Mixed |
| Example | An army: reports flow up to the general, orders flow down | The body: cells compose tissues, tissues organs (the adrenal gland) | Sensory pathways of a conscious brain |
| Shape | Pyramid | Concentric containment | Pyramid whose upper levels functionally contain the lower |
Non-nested function: topical convergence. Sensory brain hierarchies are physically non-nested (each level sits elsewhere), and much processing flows "many onto the one" up a pyramid. In vision, low-level neurons respond to points or short lines; higher ones respond more specifically; at the top, in the temporal lobe, some respond only to a hand, a specific face, or the Leaning Tower of Pisa. These are jokingly called "grandmother cells" or "pontifical neurons" (the pope briefed by the cardinals). In reality each probably responds to a few similar objects rather than one (Rodrigo Quiroga and colleagues).
Why nesting is also needed. Convergence unifies, but streamlining discards detail, such as where in the visual field a face is. A unified awareness must also use the information held at all levels, which it does by functionally nesting lower levels within higher ones. Seeing an apple fall: the retina codes line segments; the cerebrum binds them into longer contours and red patches into larger patches; higher still, the stem is bound to the outline, colour to shape, and these to motion. The unified experience is built from anatomically dispersed representations with no central convergence point: no "grandmother region," no "Cartesian Theater" (Daniel Dennett's phrase).
A conscious nervous system combines non-nested convergence with nested binding, which may make it unique in nature. The authors propose that all conscious things have this dual property, that anything with it could be conscious, and that it is necessary for unified consciousness.
Isomorphic maps, mental images and affective states
A topographic map preserves the spatial order of the receptors through successive levels of the central nervous system.
- Somatotopic map. In the mammalian parietal cortex, the body surface is mapped point by point but distorted, with face and hand enlarged because they are most touch-sensitive.
- Retinotopic map. In the occipital cortex, the visual field is mapped point by point.
- Non-spatial maps. Smell is mapped by odour combinations and hearing by tones. These still rest on receptor arrangements (the cochlea, the olfactory bulb), but the authors use "isomorphic" as the general term for all such mapping.
Isomorphic representations matter because they sit at the heart of the brain-mind divide: physical neural maps on one side, experienced sensory mental images on the other, with objective representations mirroring subjective images. Many theorists agree. Sherrington said the brain is always "constructed upon and evolved upon the 'distance receptor' organs." Edelman defined primary consciousness as "having mental images in the present." Damasio links "mapped neural patterns" to mental images in his core consciousness. Llinás stresses sensory images within a movement-based theory. The authors take it as axiomatic that mental images belong to primary consciousness.
Three limits keep isomorphism in proportion:
- Not sufficient. Low-level maps in the retina, cochlea and spinal cord exist without creating conscious images.
- Fades at the top. At the highest levels of processing, maps become so indistinct that it is hard to say where the map is; reflective sensory consciousness is not obviously isomorphic.
- Not universal. Affective consciousness (positive and negative feelings) has its own circuits, needs no isomorphic mapping, and may be felt as mental states rather than mental images. The book's illustration uses Masaccio's 1425 Expulsion from the Garden of Eden to depict shame as a non-mapped affective state referred to the whole self.
Unique neural-neural interactions
Many theories treat reciprocal, recurrent or reentrant interactions as the key to consciousness: local circuit chatter plus extensive feedback between distant centres, organised and integrative, underlying unity, qualia and selective attention. Crick and Koch described consciousness as a "dominant coalition" of interacting neurons. They and many others place this cross-talk only in the mammalian cortex and thalamus (perhaps birds too). The authors agree interactions are essential but call the cortex-thalamus view too restrictive: rich interactions occur in many other vertebrate and invertebrate regions, some of which may support consciousness (chapters 5 to 9). Chapter 10 will argue that interactions within and between hierarchically arranged, sensory class-specific neuron chains produce the NSFCs, qualia included.
Attention
William James defined attention as the mind taking possession, in clear and vivid form, of one of several possible objects, implying withdrawal from some things to deal with others. The focus here is selective attention: selecting important stimuli, filtering out others, shifting between them.
Attention and consciousness are closely related but not identical: attended details are not always conscious, and unattended things (peripheral vision) are not wholly excluded. The literature ranges from "identical" (Bernard Baars and others) to "dissociated" (Stanislas Dehaene, Tononi and Koch and others). The authors take a middle position: consciousness depends on attention, attended objects are most likely to be conscious, so the two coevolved. They note the debate is so human-, cortex- and top-down-centred that its relevance to fish and insects is unclear.
Bottom-up (exogenous) attention. Driven by the salience of incoming stimuli; the animal orients to sudden events. Simpler, less brain-demanding, so likely to have evolved first. The book's main concern.
Top-down (endogenous) attention. Proactive anticipation, sustained concentration and goal focus. Tied to higher consciousness and largely outside the book's scope.
Attention is plainly adaptive: inattentive animals are easy prey.
Diverse neural architectures
Neuroscientists usually treat qualia as one problem, as if Crick and Koch's "redness of red" and "painfulness of pain" came from the same mechanism. The authors ask whether exteroceptive, interoceptive and affective consciousness differ in mechanism, and whether mechanisms differ between a fish, a human and an octopus. In vertebrates the olfactory pathway already differs from all other senses. Derek Denton has begun to see diversity, especially between affective and exteroceptive consciousness. If consciousness is diverse, the open question is what common features all forms share.
The mapping robot. Picture a robot vacuum that builds a precise spatial map of a room and steers by it. It has an isomorphic representation, fast signalling and a hierarchy of sensors and controllers. By the authors' list it still lacks several items: nested binding into one scene, rich reentrant interaction across many sensory hierarchies, attention among competing objects, and affect. The exercise shows how the table works as a conjunctive checklist: any one feature, maps included, is cheap; consciousness is claimed only for the combination.
For an architect, the nested/non-nested distinction is a familiar design tension. A pure command hierarchy (a pipeline that summarises upward) is efficient but throws away detail; a pure compositional system keeps detail but lacks a point of decision. Systems that must both decide and keep context, such as layered retrieval and agent stacks that pass summaries up while retaining access to source records, end up with the same dual structure Feinberg and Mallatt attribute to conscious brains.
Three claims are softer than they look. First, "all things with the dual nested/non-nested property could be conscious" is a bold sufficiency claim offered without a test; many engineered systems combine convergence and distributed binding. Second, citing Tononi's integrated information as support is selective: IIT implies that consciousness can exist in systems with no reflexes, no life and no embodiment, which conflicts with Level 1 of the authors' own table. Third, the adaptation argument from complexity shows that the neural machinery was selected; it does not by itself show that the experience was what selection acted on.
- Cell types. Single-cell transcriptomic atlases released in 2023 by the BRAIN Initiative consortia sorted mouse and human brain cells into thousands of distinct types and subtypes, far beyond the "hundreds" cited here, strengthening the authors' point about neural diversity.
- Eukaryote origins. The archaeal side of the mitochondrial merger is now firmly linked to Asgard archaea (the group that includes Lokiarchaeota, which the authors mention); a living Asgard archaeon was first cultured and described in 2020.
- IIT's standing. In 2023 more than a hundred researchers signed an open letter calling integrated information theory "pseudoscience," a charge many others rejected, and the first results of the IIT versus global neuronal workspace adversarial collaboration (announced 2023, published 2025) challenged key predictions of both theories. Tononi's measure is now a contested theory, not a settled benchmark.
The Birth of Brains
Just before the Cambrian, the invertebrate ancestor of the vertebrates already had a tiny brain, with the general features of life, some polysynaptic reflexes and basic motor programs, but few or none of the special features of consciousness. The authors reconstruct this ancestral brain from vertebrates' closest living relatives, the tunicates and especially amphioxus, using Thurston Lacalli's cell-by-cell map of the larval amphioxus brain. That brain corresponds to the ventral, homeostatic "core" of the vertebrate brain and lacks the dorsal sensory centres (cerebrum, optic tectum, cerebellum), so it sits at the nonconscious reflex level.
Where vertebrates sit
Vertebrates belong to the chordate phylum, defined by the notochord, a supporting rod along the back (not a nervous structure). In vertebrates it survives as the spongy cores of the intervertebral discs, the site of a "slipped disc."
Chordates are bilaterians (Bilateria), the group that includes almost all animals except sponges, cnidarians (jellyfish and relatives) and comb jellies. Other bilaterians include arthropods, nematodes, annelids, molluscs and echinoderms. Bilaterians primitively have:
- Body axes. An anterior and posterior end, a dorsal and ventral side, and mirror-image left and right sides (bilateral symmetry).
- A through-gut. In Peter Holland's description, food enters a mouth, moves one way and leaves by an anus, with muscle blocks on either side for active, directional locomotion.
- Three germ layers. Inner endoderm, outer ectoderm, middle mesoderm.
- A nervous system. Whether the first bilaterians had a brain is debated (chapter 4).
Because all living chordates have a brain, the authors infer that a simple brain existed already in the early nonvertebrate chordates. Brains almost never fossilise, so the reconstruction must use living relatives.
Basic chordate neuroanatomy
- CNS and PNS. The central nervous system is the brain plus the spinal (nerve) cord; the peripheral nervous system is the nerves running through the body. The chordate CNS lies dorsal to the notochord, on the back side.
- Directions. Rostral means toward the higher brain centres, caudal the opposite; in horizontal animals these equal anterior and posterior.
- Neurons. Excitable cells with several signal-receiving dendrites and one signal-sending axon (a long axon is a nerve fibre) attached to a cell body holding the nucleus. Most are multipolar; most sensory neurons have one stout process branching like a T.
- Synapses. An axon ending releases a neurotransmitter onto another neuron, pushing it toward firing (excitatory) or away from it (inhibitory).
Tunicates: the odd sister group
Tunicates are abundant ocean filter feeders with a tough tunic containing cellulose, otherwise found only in plants.
- Ascidians (sea squirts). The main group. A tiny tadpole-like larva swims for a few days without feeding, then settles and becomes a sessile, baglike adult that is mostly a water-straining pharynx, with no locomotor muscles.
- Thaliaceans (including salps). Probably derived from ascidians; a barrel open at both ends that moves by jet propulsion.
- Larvaceans (appendicularians). Tadpole-like for life and under a centimetre long, feeding on the tiniest plankton.
For most of the twentieth century, vertebrates were thought to descend from larvaceans or from ascidian larvae that became sexually mature (neoteny). Gene studies by Linda Holland and others overturned this: larvaceans have lost many genes and are specialised for fast development and short lives, so they are "weird," not primitive. All tunicates show heavy gene loss and rapid gene evolution, and their anatomy is specialised (larval "swimming muscles" are single cells, not segmented blocks; the anus opens in the pharynx region). Ironically, the same genomic work, chiefly by Frédéric Delsuc, Hervé Philippe and colleagues, shows tunicates are the vertebrates' closest relatives (sister group), closer than the more fishlike amphioxus: a sister with whom one has little in common.
Tunicate brains hold a few hundred neurons at most. Lacalli and Holland found traces of the vertebrate forebrain, midbrain and hindbrain, but almost never all three together: different groups and life stages have lost different subsets, as if a full tripartite ancestral brain had been differentially reduced. The authors offer a reason: adult ascidians feed but do not move, so they keep the feeding and visceral region (midbrain and anterior hindbrain); larval ascidians swim but do not feed, so they keep sensory and locomotor parts (a non-image-forming ocellus, a balance organ or otolith, and most of the hindbrain).
| Tunicate | Forebrain | Midbrain + anterior hindbrain | Hindbrain |
|---|---|---|---|
| Adult ascidian (Ciona) | Absent | Cerebral ganglion | Absent |
| Larval ascidian (Ciona) | Sensory vesicle with ocellus and otolith | Undeveloped ganglion | Visceral ganglion |
| Larval aplousobranch (Diplosoma) | Sensory vesicle | Ganglion rudiment | Visceral ganglion |
| Adult thaliacean (Thalia) | Ocellus only | Ganglion | Possibly |
| Adult larvacean (Oikopleura) | Sensory vesicle with otolith | Undeveloped | Visceral ganglion |
Amphioxus: the better model
Cephalochordates (amphioxus, lancelets) are less closely related to vertebrates than tunicates but more conservative: their genome, embryonic development and fishlike anatomy resemble vertebrates'. Larvae swim and filter-feed high in the water column; at about 1 cm they become adults that burrow in sand and reach 4 to 6 cm. The adult brain, a thin unbulged tube once doubted to be a brain at all, has many thousands of neurons, too many to trace. The young larva's brain is small enough that Lacalli and colleagues reconstructed most of its neurons from serial electron-microscope sections, then used the landmarks to read the adult brain.
Mapping amphioxus onto the vertebrate brain
| Amphioxus structure | Vertebrate counterpart | What it shows |
|---|---|---|
| Frontal "eye" (a few photoreceptors, no image) | Retina, which grows from the diencephalon | Brain front = diencephalon; probably no telencephalon |
| Infundibular organ (secretory neurons) | Neurohypophysis (posterior pituitary) in the hypothalamus | Amphioxus has a hypothalamus |
| Lamellar body (photoreceptors) | Pineal organ, the "third eye" | Dorsal diencephalon present |
| Ring of "clock cells" (adult) | Suprachiasmatic nucleus, identified by shared clock genes | Circadian generator present |
| Primary motor center | Midbrain locomotor center in the reticular formation | A ventral midbrain |
| Hox-expressing posterior region | Hindbrain | A hindbrain |
| Postinfundibular neuropile | Diencephalic locomotor region (DLR), paracrine core | Centre for basic behavioural programs |
Four structures (frontal eye, infundibular organ, lamellar body, clock cells) prove a diencephalon. Gene-expression landmarks back up the divisions: the anterior neural ridge (ANR) at the front of the forebrain, the zona limitans intrathalamica (ZLI) mid-diencephalon, and the midbrain-hindbrain boundary (MHB). Since the ZLI separates hypothalamus from thalamus in vertebrates, amphioxus may have a rudimentary thalamus.
Only the ventral half
Lacalli's key observation is that amphioxus's forebrain, midbrain and hindbrain correspond almost exclusively to the ventral parts of the vertebrate divisions. These ventral regions, the hypothalamus and midbrain locomotor center, run the most basic survival behaviours and homeostasis (keeping internal conditions constant and optimal). Missing is the dorsal brain for extensive sensory processing: no cerebrum, optic tectum or cerebellum.
The vertebrate core and its amphioxus twin
In the 1980s and 1990s Rudolf Nieuwenhuys and colleagues identified a primitive core of the vertebrate brain, variously called the limbic core or paracrine core, with the rest of the brain lying peripheral and dorsal to it. It comprises the hypothalamus and central parts of the midbrain and hindbrain, and in all vertebrates it controls:
- Homeostatic adjustment. Blood pressure, heartbeat, body chemistry, breathing and day-night adjustment.
- Motivated survival behaviour. Drinking, mating, aggression, chewing and swallowing, escape, many driven by central pattern generators.
Its neurons are unusual. They signal less by fast synapses and more by releasing diverse neuromodulators that raise or lower other neurons' activity over long periods, setting lasting behavioural states. The chemicals sit in vesicles inside varicosities (swellings) along very thin axons.
Lacalli's larval circuit. The postinfundibular neuropile of the larval amphioxus hypothalamus shows the classic paracrine-core signature: thin axons, varicosities with mixed vesicles, few classical synapses. It receives every kind of sensory input the larva picks up and switches among its few programs: fast escape swimming, slow upward swimming to the best feeding depth, and stopping to sink through food-rich water while passively trapping particles. Lacalli matched it to the diencephalic locomotor region of lampreys and other vertebrates.
Lacalli took this larval brain as the model for the ancestral chordate and earliest prevertebrate brain: it received light, touch and other stimuli but allowed little sensory processing and few behavioural choices. Elaborate eyes, nose and ears, and the dorsal brain regions that build images of the world, came later in the vertebrate line.
Lacalli judges the larval amphioxus brain too simple for consciousness, and the authors' own tracing agrees: most sensory pathways have only three or four neuron levels between stimulus and response, just one or two neurons before the motor center. Larval amphioxus is a living example of the authors' Level 2: a reflex arc with a simple pattern generator, which is the nonconscious platform the vertebrate brain was built on.
Two objections, and the replies
Objection 1: simple is not primitive. The old "ontogeny recapitulates phylogeny" idea (the "biogenetic law," popular 1860 to 1950) often failed; larvae evolve too, so a simple larval brain need not reflect the ancestor.
Reply. Lacalli grounds homology in matching locations and connections, and argues amphioxus keeps only structures essential for survival, which must have been present from the start of chordate history.
Objection 2: secondary simplification. Amphioxus may have degenerated after taking up burrowing in dark, protected sand (Conway Morris and Caron, 2012). Ariel Pani, Christopher Lowe and colleagues (2012) noted its embryo lacks key brain-patterning genes, and some (Suzuki and colleagues) deny it has a midbrain at all.
Reply. Holland and coworkers (2013) showed amphioxus expresses enough patterning genes to mark ANR, ZLI and MHB. Šestak and Domazet-Lošo (2015) found in the whole genome many genes expressed in the vertebrate forebrain, midbrain and hindbrain. The authors side with Lacalli and Holland: the region between the lamellar body and the hindbrain is a real midbrain.
The reconstruction leans heavily on one research programme (Lacalli's) and on young larvae. Lacalli himself told the authors that late larvae show an expanded, differentiated region where a dorsal telencephalon would be, and that "we really don't know what might be present later on." If adult amphioxus has more dorsal brain than assumed, the clean "ventral core only" picture, and with it the neat Level 2 classification, would need softening.
- Tunicate wiring mapped. In 2016 Kerrianne Ryan, Zhiyuan Lu and Ian Meinertzhagen published the complete synaptic connectome of the larval Ciona intestinalis central nervous system, about 177 neurons, confirming how tiny tunicate brains are and revealing marked left-right asymmetry.
- Yunnanozoans reconsidered. The status of Cambrian Haikouella and its relatives remains disputed. A 2022 study argued from preserved pharyngeal arches that yunnanozoans are stem vertebrates, which would place them close to the transition this chapter reconstructs; others still read them as more basal deuterostomes or chordates.
The Cambrian Explosion
Between roughly 560 or 540 and 520 million years ago, animal evolution went into hyperdrive: nearly all modern phyla appeared, along with the first complex senses, nervous systems and behaviours. The authors argue the explosion was real, that its main engine was the first predators and the arms race they started, and that a richer Cambrian seafloor rewarded animals that built sensory maps of their world. Only two mobile clades took that expensive road: arthropods (mostly predators) and vertebrates (small, filter-feeding fish whose genomes had doubled and redoubled).
Terms of the trade
| Term | Meaning |
|---|---|
| Chordate features | Notochord, pharyngeal pouches (gill pouches and slits), a tail behind the anus, and a hollow dorsal nerve cord. Nonchordates usually have a solid, ventral nerve cord. |
| Clade | A natural group: an ancestor and all its descendants. Mammals are a clade; "reptiles" are not, because birds evolved from them but are excluded. |
| Convergent evolution | Independent evolution of similar structures in distant lineages under similar pressures (insect, bird and bat wings). Such structures are analogous (homoplasious). |
| Homology | Correspondence due to common descent: tetrapod limbs and fish paired fins; bird feathers and snake scales; all vertebrate eyes. |
| Invertebrates | All animals that are not vertebrates; not a clade. |
| Phylum | A group sharing one basic body plan; 30 to 35 animal phyla exist. Some former "phyla" (peanut worms, spoon worms) turned out to be annelids. Vertebrates are a subgroup of the chordate phylum. |
| Phylogenetics | The study of evolutionary relationships among clades. |
| Selective pressure | Strong natural selection. |
| Sister groups | The two most closely related clades (humans and chimpanzees; horses and donkeys). |
| Vertebrate features | A head with distance senses (vision, smell, hearing); uniquely, vertebrae, a distinctive skull, and neural crest and ectodermal placodes. |
Darwin's dilemma
In On the Origin of Species (1859), Charles Darwin saw a problem: animal fossils (mineralised skeletons) appear abruptly and for many groups at once in Cambrian rocks, with no earlier ancestors. This is Darwin's dilemma. One camp, Darwin included, calls it an artifact: earlier animals were small and soft and did not fossilise. The other camp notes that small, soft-bodied animals do sometimes fossilise, yet animal fossils are absent until about 580 to 555 mya, in the Ediacaran period just before the Cambrian, so the explosion records a real event. Better soft-bodied fossil sites and better knowledge of living phyla now let scientists judge.
- c. 635 mya Ediacaran period begins; the first animals, probably sponges ("Wiffle-ball" candidates), may date from here.
- c. 600 mya A more convincing Ediacaran sponge fossil.
- Before 585 mya Earliest reported worm trackways (a worm under 1 cm).
- c. 580 mya The bilaterian ancestor lives on Ediacaran microbial mats.
- c. 580 to 555 mya Ediacaran soft-bodied animals appear in the fossil record.
- c. 541 mya Cambrian period begins (the book's date; runs to 488 mya).
- c. 520 mya Chengjiang shale, south China: early fish and diverse arthropods.
- c. 505 mya Burgess Shale, western Canada.
Explosion's eve: the Ediacaran seafloor
The authors compare two reconstructed seafloors. The Ediacaran world (fossils from southern Australia, Newfoundland, Namibia and Russia's White Sea) at about 555 mya was carpeted by a thick green microbial mat. Bacteria always form films (think of dental plaque), but in the Ediacaran algae joined them in the sunlight, making mats so lush they are called the "Garden of Ediacara." Its animals were mostly pleated, leaflike or flat, with no mouth or gut, many immobile; they ate the mat, not each other.
The ancestral bilaterian worms, perhaps derived from an elongated sponge, crept along grazing the mat. Their trails are fossilised behaviour following three rules: follow the food gradient; do not dig deep or vertically; plough straight ahead or in shallow waves with no side trips.
By 520 mya the scene is familiar: mats mostly grazed away, early members of most modern phyla, including arthropods, fish and shelled animals, and constant movement on, above and within the sediment.
The explosion's products
Bilaterians split into two branches:
- Deuterostomes. Chordates (including vertebrates), echinoderms (sea lilies, sea stars, urchins) and hemichordates (acorn worms).
- Protostomes. Ecdysozoans (arthropods, nematodes, priapulids) and lophotrochozoans (molluscs, annelids, brachiopods). Vertebrates are therefore only distantly related to arthropods and molluscs.
The Early Cambrian had members of most phyla, many close enough to living relatives that their lifestyles and nervous-system capacities can be inferred. Cambrian chordates are now well known: the tunicate Shankouclava, the laterally flattened eel-like Pikaia with a tiny head, the near-vertebrate Haikouella, and jawless fish Haikouichthys and Metaspriggina. Cambrian arthropods were hyperdiverse, many with antennae and compound eyes (the trilobite Olenoides, Sidneyia, Marrella, Leanchoilia). Some worms and brachiopods look almost modern.
Some oddballs have been placed:
- Anomalocaridids. Predators up to a metre long, the largest Cambrian animals, with spiny grasping appendages before the mouth; new fossils show arthropod compound eyes, so they are arthropods or their nearest relatives.
- Vetulicolians. Large, eyeless swimmers with a giant "gill basket" like a tunicate pharynx and a jointed, arthropod-like rear that may hold a notochord, so possibly tunicate relatives.
- Still enigmatic. Eldonia, Herpetogaster and Wiwaxia remain hard to place.
The bilaterian ancestor: complex or simple worm?
Fossil trails and the many worm phyla near the base of the tree both say the ancestor was a worm. But what kind?
Complex worm. Image-forming eyes in a distinct head, segments, limbs, a heart and blood vessels, a large three-part brain and nerve cord. Inspired by the traits and shared developmental gene pathways of vertebrates and arthropods.
Simple worm. A crude head with light-sensing ocelli but no image, no segments, limbs, heart or vessels, and no brain: just a nerve net at the body surface. The authors' choice.
The authors side with the simple ancestor, following R. Glenn Northcutt. Too many phyla lack brains, segments, limbs and image-forming eyes; a complex ancestor would demand an unreasonable number of independent losses. The shared genes are real, but in the ancestor they likely specified cell types (light-sensitive cells, not eyes) or positions in the embryo (not segments); separate lineages later redeployed them. Only this explains why sponges and sea anemones also carry these genes while lacking every complex bilaterian trait.
A brainless worm could survive easily on the Ediacaran buffet: follow chemical "reward gradients" to food and sex pheromones to mates. Reflexes and simple motor programs, no consciousness needed. In terms of table 2.1, it had interconnected nerve cells and reflexes, but none of the special features.
Causes of the explosion
- Oxygen. Atmospheric oxygen, accumulating for billions of years since the first photosynthetic bacteria, finally reached levels that could support large, active animals.
- Predation (the main cause). One worm began eating worms of another species, and prey responded.
- Bioturbation. Prey burrowed deeper to hide; burrowing churned the sediment, carrying in oxygen and nutrients, which fed new phyla of burrowers with more complex foraging trails.
- An arms race above the floor. Predators with spines, limbs and claws (many arthropods, spiny-headed worms) forced prey to adapt or die, and selection favoured the same traits in predators.
- A richer sensory environment. The Cambrian floor sent far more signals than the still Ediacaran mats, rewarding animals that sensed and mapped them.
- Contributing factors. New developmental gene pathways able to build complex bodies, ocean chemistry changes, and "snowball Earth" glaciations before and during the Ediacaran.
The arms race ran in both directions:
| Trait | Prey use | Predator use |
|---|---|---|
| Mineralised hard parts | Shells and armour | Piercing and cutting; defence against other predators |
| Concealment | Hiding among plants, behind rocks, camouflage | Hiding while stalking |
| Speed | Escape | Pursuit |
| Distance senses | Detect approaching predators | Detect prey from afar |
| Large size | Too big to kill | Subdue bigger, more nutritious prey |
From signals to maps
Roy Plotnick, Stephen Dornbos and Junyuan Chen (2010) argued that the moving, shifting Cambrian seafloor was a more complex environment that broadcast new tastes, odours, patterns of light and colour, and vibrations through water and sediment. The more of these an animal could sense, the better it could orient, avoid danger and find resources, so eyes, smell and taste organs, touch and vibration receptors all improved. The simplest use of sensory data is reflexes; a better one is to assemble it into maps in a brain (visual maps, smell maps, maps of the body surface) that reconstruct the surroundings and guide movement. Those maps were under constant pressure to improve, because crude perception is fatal: failing to see a hidden predator ends the lineage.
Why so few took the high road
Sensory processing is expensive in food energy, and so is constant activity. Most phyla therefore took cheaper shortcuts, sometimes with neural regression: clamlike molluscs and brachiopods sheltered in shells, and blind burrowing worms avoided predators without a full sensory kit. Only arthropods and vertebrates, the most mobile clades, invested in ever-better senses. In the predator-prey arms race, with its premium on sophisticated senses and newly cognitive brains, the authors see the key cause of the Cambrian explosion and the answer to Darwin's dilemma.
Arthropods versus vertebrates
The two groups are anatomically opposite, yet each independently evolved high mobility, advanced sensory processing and brains. They stand on opposite sides of the arms race: predators and prey, warriors and farmers, lions and lambs.
| Feature | Cambrian arthropods | Cambrian vertebrates |
|---|---|---|
| Skeleton | External cuticle (exoskeleton) | Internal, light notochord |
| Limbs | Many jointed limbs, head limbs adapted to grasp and shred | None; no jaws, no teeth |
| Role | Top predators (not all species) | Filter feeders and escapers |
| Origin | From tube-limbed lobopodians (velvet worm relatives), late Ediacaran | From fast-swimming, amphioxus-like prevertebrates |
| Eyes | Image-forming compound eyes | Camera eyes plus smell and hearing |
| Size | Up to a metre (anomalocaridids) | About 3 cm |
| Abundance | Most abundant and diverse phylum | Rare, not diverse |
| Genome | No duplication | Whole genome duplicated twice |
The arthropod path
Sanctacaris ("saintly claws," a pun on Santa Claus) had prey-grasping appendages before its mouth. Reaching, grasping and shredding prey demand spatial skill and benefit from spatial vision, and most Cambrian arthropods had image-forming eyes, far beyond the simple ocelli of lobopodians. Fuxianhuia, the most basal arthropod known, already had good eyes and an arthropod brain while keeping lobopodian tube limbs. Jointed limbs gave arthropods a food-handling edge no limbless animal could match.
The vertebrate path
The rare Early Cambrian fish, like Haikouichthys, were about 3 cm long with no teeth, jaws or limbs, so not predators. They were probably filter feeders, like all their closest living relatives: tunicates, amphioxus and larval lampreys. Larval lampreys, the least specialised living jawless fish, filter the densest, most nutrient-rich suspensions of any filter feeder (Mallatt's own research), suggesting the first fish exploited the richest surviving patches of microbial mat, travelling from patch to distant patch.
Travel came easily. Fish swim by contracting their myomeres (the muscle segments visible in a cooked trout) front to back, throwing the streamlined body into propulsive waves on a light notochord, with no heavy exoskeleton. Amphioxus swims this way too, "like a bullet," but has no image-forming eyes or smell organ. So fast myomeric swimming came before the keen distance senses. Perhaps prevertebrates were too fast to catch even when they sensed predators late, which would explain why the first fish had no armour. As predators got faster, selection favoured sensing danger from afar and finding the best food patches: vision, smell and hearing sharpened.
We descend from "hyperalert but gentle vegans": small, rare filter feeders whose entire genome duplicated, then duplicated again at the dawn of vertebrates. This genomic quadrupling supplied raw material for innovation, making vertebrates the most complex animals on Earth genetically, structurally and neurally, with room for neural hierarchies with more levels. No such duplication occurred in any invertebrate clade, not even the big-brained cephalopods.
Arthropods remain more abundant and species-rich, but vertebrates reached the largest bodies and largest absolute brains; average marine vertebrate body size has risen markedly since the Cambrian while marine arthropods' has not (Heim and colleagues, 2015).
Two cautions. First, the explosion's causes are still debated: oxygen, predation, developmental genetics, ocean chemistry and glaciation all have advocates, and the authors' emphasis on predation and sensory maps is a reasonable synthesis but not a consensus. It also fits their thesis neatly, which is a reason to hold it a little more loosely. Second, the link from "sensory maps were adaptive" to "maps were conscious" is not made in this chapter; the arms race explains better senses and brains, and whether experience came with them is the burden of the chapters that follow. (A small slip: the seafloor figure's key spells the Ediacaran animal "Dickensonia"; the genus is Dickinsonia.)
- Revised dates. The International Commission on Stratigraphy now places the base of the Cambrian at about 538.8 mya (revised in 2022 from 541) and its end at about 485.4 mya, not 488.
- Ctenophores first. A 2023 study of conserved chromosome-level gene linkages (synteny) strongly supported comb jellies, not sponges, as the sister group of all other animals, the placement the book flagged as most uncertain. It implies either that neurons evolved twice or that sponges lost them.
- A tiny Ediacaran bilaterian. In 2020, Ikaria wariootia, a rice-grain-sized worm about 555 million years old from South Australia, was described as a likely early bilaterian and a probable maker of Ediacaran burrows, consistent with the "simple worm" ancestor.
- Genome duplications refined. Chromosome-scale genomes of amphioxus, lampreys and hagfish (2020 to 2024) indicate that only the first round of whole-genome duplication preceded the split between jawless and jawed vertebrates; the second round occurred in the jawed-vertebrate line, while the jawless lineage underwent its own separate duplication. The broad claim that early vertebrate genomes multiplied stands; "duplicated twice at the dawn of vertebrates" is now too simple.
Consciousness Gets a Head Start: Vertebrate Brains, Vision, and the Cambrian
Exteroceptive consciousness, the experience of the outside world, arose in the first vertebrates during the Cambrian, before 520 million years ago. The route ran through image-forming eyes that grew out of the brain, followed by a near-total reboot of the other senses built from two new vertebrate tissues, neural crest and placodes. These fed new dorsal brain regions (cerebrum, optic tectum, cerebellum) with isomorphically mapped input, and all vertebrate senses came to share one five-step pathway plan. Counting levels, the authors estimate that a conscious sensory hierarchy needs at least three or four neuron levels.
The vertebrate brain
The Cambrian explosion that produced vertebrates also produced the vertebrate brain. The three-part chordate brain (forebrain, midbrain, hindbrain) already existed in amphioxus; in vertebrates all three parts are much enlarged, more complex and better differentiated, packed with well-developed neural hierarchies.
The parts and their jobs, which the rest of the book relies on, are laid out systematically below.
| Division | Part | Main functions |
|---|---|---|
| Spinal cord | Sensory input and motor output via spinal nerves; reflexes and simple movement commands below the head | |
| Hindbrain (rhombencephalon) | Medulla oblongata (myelencephalon) | Cranial nerves 8 to 12; reticular centres for swallowing, heart rate, blood pressure, breathing |
| Pons (metencephalon) | Cranial nerves 5 to 7; centres for breathing, swallowing, urinating; pontine nuclei work with cerebellum | |
| Cerebellum (metencephalon) | Makes movements smooth and coordinated; balance and posture | |
| Midbrain (mesencephalon) | Tegmentum | Cranial nerve 3 (eye movement); midbrain locomotor centre; caudal striatum-related motor selection |
| Optic tectum (superior colliculus in mammals) | Visual reflexes and extensive visual processing; also touch, hearing and more; isomorphic sensory maps; cranial nerve 4 | |
| Periaqueductal gray | Signals the bodily responses of panic | |
| Forebrain: diencephalon | Pretectum | Eye responses to changing light intensity |
| Thalamus | Relay to the pallium (cortex in mammals) for all other brain parts; processes what passes; alertness and wakefulness | |
| Epithalamus: pineal | Light receptors for direction, time of day and season (no images); secretes melatonin | |
| Epithalamus: habenula | Inhibits movements, in sleep and to prevent inappropriate or dangerous acts | |
| Prethalamus (subthalamus) | Part of the striatal system selecting motor programs | |
| Hypothalamus | Motor responses of affects; optic nerve (2) attaches; retina grows from it; pituitary (neurohypophysis); main visceral control; thirst, hunger, temperature, digestion; sex and biorhythms (suprachiasmatic nucleus) | |
| Forebrain: telencephalon | Striatum (basal ganglia), ventral | Selects which motor program to run (breathing, chewing, locomotion): fast, inflexible choices |
| Septal nuclei | Affects, moods, anxiety | |
| Amygdala, subpallial part | Affect-directed actions; fear behaviour | |
| Olfactory bulb and tract | Cranial nerve 1 attaches; initial smell processing | |
| Dorsal pallium | Most of the mammalian cerebral cortex; multisensory processing, elaborate isomorphic maps in mammals; complex decisions; voluntary movement; cognition; some affect | |
| Lateral pallium | Smell processing; the olfactory cortex of mammals | |
| Medial pallium (hippocampus) | Spatial memories and memories of objects and events | |
| Ventral pallium (pallial amygdala) | Affective learning, especially fear | |
| Multi-region systems | Brainstem | Medulla, pons and midbrain together |
| Ventricles | Central canal; fourth ventricle (hindbrain); cerebral aqueduct (midbrain); third ventricle (diencephalon); lateral ventricles (hemispheres) | |
| Autonomic nervous system | Motor neurons to glands, smooth and heart muscle; parasympathetic "rest and digest", sympathetic "fright, fight or flight"; contrasts with the somatic motor system to skeletal muscle | |
| Reticular formation | Brainstem core; reticular activating system for arousal, attention, emotional intensity; coarse movements; visceral control | |
| Greater limbic system | Affects, motivation for survival behaviour, homeostasis; amygdala, hippocampus, basal forebrain, septum, hypothalamus, periaqueductal gray, reticular formation |
How brains vary
The brainstem (medulla, pons, midbrain) is continuous; the forebrain has a diencephalon (mostly thalamus and hypothalamus) and a telencephalon (cerebrum, two hemispheres). When vertebrate brains enlarged, it was usually the dorsal pallium (cerebral cortex) that grew, most spectacularly in mammals and birds, but also in some jawed fish and reptiles. Any region serving a specialist skill can enlarge too: fish with acute taste or electroreception have big brainstem regions for those senses.
- Olfactory bulb. Relatively largest in fish, amphibians and smell-reliant mammals like rats.
- Optic tectum. A visual centre in all vertebrates (less so in mammals) that also processes every other sense except smell; very large in many fish. Together, the optic tectum and cerebral pallium are the main processing regions for vision, touch, hearing and smell.
- Cerebellum. Senses one's own movements and smooths them; largest in agile animals like birds and humans.
- Greater limbic system. Rudolf Nieuwenhuys's deep, medial system with parts in every division, running core-body functions, motivation and affect, and containing the primitive core of chapter 3.
What vertebrates have that amphioxus lacks
| Structure | Vertebrates | Amphioxus | Tunicates |
|---|---|---|---|
| Spinal cord | + | + | + |
| Hindbrain: myelencephalon | + | + | + |
| Hindbrain: metencephalon (pons, cerebellum) | + | - | - |
| Midbrain: tegmentum | + | + | + |
| Midbrain: tectum | + | - | - |
| Diencephalon: pretectum | + | - | - |
| Diencephalon: thalamus | + | ? | ? |
| Diencephalon: pineal | + | + | + |
| Diencephalon: prethalamus | + | - | - |
| Hypothalamus and neurohypophysis | + | + | + |
| Telencephalon (cerebral hemispheres) | + | - | - |
| Hollow centre | + | + | + |
| Autonomic nervous system | + | - (no crest input) | - (no crest input) |
| Reticular formation | + | + | ? |
| Greater limbic system | + | + (neuropile) | ? |
The decisive novelties are the cerebrum, a well-developed thalamus, the optic tectum and the metencephalon (cerebellum and pons), and all of them have major sensory-processing roles. Apart from them, brain structure is remarkably similar across vertebrates, the main exception being the more complex pallium of mammals, birds and reptiles.
The prevertebrate's senses
From amphioxus and the simple bilaterian ancestor, the authors reconstruct the prevertebrate's kit, all in the skin (plus some photoreceptors in the CNS):
- Photoreceptors partly shielded by pigment, giving light direction but no image.
- Mechanoreceptors of many kinds for touch (amphioxus has a wide variety).
- Nociceptors for harmful stimuli, as in almost all animals.
- Chemoreceptors for smell- and taste-like senses. The authors admit this is risky: no definitive chemosensory organ is known in amphioxus or tunicates, though both carry chemoreception genes (amphioxus has olfactory-receptor-like genes in scattered snout skin cells), and chemoreception is thought primitive for all bilaterians.
All of this was entirely reflexive. Consciousness required a bigger brain with complex sensory hierarchies.
Vision first
Andrew Parker's "Light-Switch" hypothesis held that the first image-forming eyes, in arthropods, created the first true predators and so triggered the Cambrian explosion. Michael Trestman refined it: the new eyes formed spatial images that individuate discrete objects and reveal their positions, distances, movements and occlusions, letting prey be spotted, tracked, struck, handled and eaten. Arthropods threw the switch; their prevertebrate prey's vision improved fast in response. Since focused vision yields the most information of any sense, it most plausibly drove the arms race, and smell, touch and hearing followed.
Nilsson's four stages of eye evolution
Dan-Eric Nilsson derived the stages from the dozens of invertebrate lineages (worms, snails) that evolved simple eyes independently, along similar paths.
- Stage I, nondirectional photoreception. Unshaded photoreceptor cells monitor light intensity: day or night, dark place or bright.
- Stage II, directional photoreception. A pigment screen shades receptors from some directions, so the animal can move toward or away from light and detect shadows of approaching predators. The bilaterian ancestor's ocelli were here; the amphioxus frontal eye (inside the brain, homologous to vertebrate eyes) is at Stage II or barely into III.
- Stage III, low-resolution vision. Cupping means each receptor sees only light from its own direction; together they form a crude image, enough to track moving objects or avoid collisions. Some add a lens that concentrates light for dim conditions without focusing (the tunicate ocellus). In the proposed "III-plus" stage, a deep cup acts like a pinhole camera with better resolution.
- Stage IV, high-resolution vision. A focusing lens and many more photoreceptors give sharp images: the vertebrate camera eye.
In vertebrates the eyecups grew as lateral outgrowths of the diencephalon and became the paired eyes. At the III-plus stage, "in the land of the blind, the one-eyed man is king": neo-arthropod predators probably saw no better. Note that the retina is part of the brain (CNS), not the peripheral nervous system, even though it sits near the body surface; every other sense is received in the PNS.
High-resolution eyes would be useless without matching visual processing in forebrain and midbrain. The authors propose that the resulting detailed, isomorphic visual representations were the brain's first conscious mental images, so vision signalled the dawn of consciousness. The images were mapped, multi-level and referred out to the world.
The smell-first rival
Smell first (Plotnick, Dornbos and Chen). Early worms lived under microbial mats where light cannot reach and odour matters. Bottom-dwelling predators like starfish and snails do without vision. The first fossils with image-forming eyes (trilobites) appear only at 520 mya, 20 million years into the Cambrian. (Lucia Jacobs and James Kohl also argue for smell or chemical ecology first.)
Vision first (Feinberg and Mallatt). Many worms grazed on top of mats, in light. Many bottom predators are visual: lobsters, horseshoe crabs, flounders, stingrays. And there are almost no body fossils from about 540 to 520 mya (only small scales, shells, fragments), so the absence of eyes then says nothing; the same gap would hide noses.
The theoretical case: light carries far more information. "Smell images" can be built from odour plumes and gradients, but they cannot give sharp object borders or precise distances, and are poor for fast-moving attackers. Murky water hampers vision, but the turbidity currents that cause murk also scramble odour signals.
Smell has one unique property: odours linger. Wastes, secretions and spoor tell of the past, so as soon as olfaction appeared selection linked it to memory ("this smell means that animal was here and may be near"). That is why smell and memory structures evolved side by side in the early telencephalon.
Three lines of physical evidence also favour vision first:
- Homology. Gene-expression cell typing shows the amphioxus frontal eye is homologous to the vertebrate eye, down to retinal and pigment cells, while amphioxus has no known organ of smell or hearing.
- Gene timing. Martin Šestak and colleagues found a surge of vision genes earlier on the chordate tree than genes for smell, hearing and balance.
- Fossils. Candidate Cambrian prevertebrates show more evidence of eyes than of nasal organs.
Neural crest and placodes: the sensory reboot
Where did the lens come from, and the neurons of the improved nonvisual senses? From two embryonic tissues found only in vertebrates: the neural crest and the ectodermal (neurogenic) placodes. They form at the edges of the neural plate as it folds into the neural tube: the ridge tops become neural crest, the lateral walls give placodes, paired left and right. Their origin is widely credited with the complexity of the vertebrate head (the classic paper is Gans and Northcutt, 1983).
| Tissue | Behaviour | What it builds |
|---|---|---|
| Neural crest | Migrates widely through the body | All trunk sensory neurons and many in the head (touch, pain, temperature); pigment cells (melanocytes); gut motor neurons; parts of the skull; gill bars |
| Olfactory placode | Stays near surface | Smell receptor cells in the nose |
| Lens placode | Stays near surface | The lens of the eye |
| Otic placodes | Stays near surface | Hearing and balance cells of the ear |
| Other placodes | With crest | Head somatic sensory neurons; visceral sensory neurons incl. taste buds; the lateral line of fish and aquatic amphibians (vibration detectors) |
Nearly the entire nonvisual sensory PNS of vertebrates comes from crest and placodes. Amphioxus and tunicates have a sensory PNS but at most rudimentary precursors: candidate homologues among pigment and ectodermal cells remain debated, and Abitua and colleagues (2015) found a credible but rudimentary olfactory placode homologue in tunicates whose neurons mainly secrete reproductive hormone, with no conventional odorant receptors in the genome. The implication is a near-complete replacement of the sensory PNS at the dawn of vertebrates, in some ways more impressive than the eyes. Since a lens is needed to go from Stage III to IV, the lens placode, and probably crest and placodes generally, existed by that transition.
New dorsal brain
The flood of new, mapped sensory input demanded multisensory processing, and Lacalli proposed that this explains the vertebrate-amphioxus brain differences. New, mainly dorsal regions appeared:
- Telencephalon (cerebrum). Grew from the dorsal, rostral forebrain to receive smell.
- Optic tectum. Enlarged in the midbrain for vision and other nonolfactory senses.
- Dorsal hindbrain. Enlarged for hearing, balance, lateral line and electroreception.
The first conscious organisms
Ann Butler proposed the two-wave idea first (2000). Her hypothetical "cephalate" sat between amphioxus and true vertebrates: paired eyes and an enlarged visual brain in diencephalon and midbrain, but no crest, placodes or telencephalon. Its visual pathways served as a circuit template that the later crest and placode senses copied.
Haikouella, the near-vertebrate. Inch-long, blade-shaped yunnanozoans from the 520-million-year-old Chengjiang Shale of Yunnan, China, the oldest known whole-body bilaterian fossils. Junyuan Chen, Mallatt and colleagues read Haikouella lanceolatum as the closest relative of vertebrates: notochord, paired eyes, a large brain with diencephalon and hindbrain, but no skull or ear capsules and only hints of a telencephalon and nostrils. It fits the vision-first cephalate stage, yet its gill bars (crest-derived) and dotlike lenses (placode-derived) put it closer to vertebrates than Butler's cephalate. Its eyes were tiny, 0.2 mm, twenty times the 0.01 mm amphioxus eye but probably too small to focus an image; clearly visual Cambrian arthropods had eyes of about 1 mm or more. Critics (Shu and colleagues; Conway Morris and Caron) say the "eyes" are fossilisation artifacts and yunnanozoans are basal deuterostomes, not chordates. Chen and Mallatt reply with muscle fibres in the myomeres and notochord segments suggesting protovertebrae.
Haikouichthys and Metaspriggina, the first fish. Described by Degan Shu and coworkers in the late 1990s, Haikouichthys (also Chengjiang) is almost universally accepted as a true jawless vertebrate: vertebrae, eyes 0.6 mm across, ear and nasal capsules, gill bars and pouches, fins and curved myomeres. ("-ichthys" means fish; "-ella" means "little bit," not fully a vertebrate.) Jean-Bernard Caron and Simon Conway Morris later described the close relative Metaspriggina from the Burgess Shale (505 mya), whose head looks almost identical from above. With image-forming eyes feeding a brain, Haikouichthys had visual consciousness in the authors' view, so consciousness has existed at least since 520 mya.
These fish are "too good" as vertebrates to reveal the prevertebrate stages, and their brains are not preserved (eyes and noses imply a vertebrate brain). The intermediate steps rest on the vision-first logic and the yunnanozoans. The authors' tentative dating: the larger-eyed Haikouichthys and Metaspriggina were conscious; the smaller-eyed yunnanozoans may not have been. Between 560 and 520 mya, the amphioxus-like frontal eye became a camera eye. In the authors' summary, the first vertebrates had the first consciousness.
One plan for every sense
Butler proposed that as vision evolved, reflex paths gained more levels in an isomorphic, hierarchical pattern, and the nonvisual senses then copied that pattern. All vertebrate sensory pathways do share it.
| Sense (1: receptor) | 2: Sensory neuron | 3: First-order multipolar | 4: Next order | 5: Next order, cerebrum |
|---|---|---|---|---|
| Vision (photoreceptors) | Retinal bipolar cells | Retinal ganglion cells | Thalamus (lateral geniculate); optic tectum: visual maps and images | Primary visual cortex |
| Somatic senses (mechanoreceptors) | Somatosensory neurons (sensory ganglia) | Dorsal column nuclei (hindbrain), dorsal horn (trunk), trigeminal nuclei (face) | Thalamus (VPL, VPM); tectum | Primary somatosensory cortex |
| Hearing (hair cells) | Spiral ganglion neurons | Cochlear nuclei (medulla) | Inferior colliculus then thalamus (medial geniculate); in fish, torus semicircularis to tectum | Primary auditory cortex |
| Equilibrium (hair cells) | Vestibular ganglion neurons | Vestibular nuclei | Thalamus; torus semicircularis to tectum: maps of body position and motion | Primary vestibular cortex |
| Taste (taste cells) | Cranial ganglion neurons | Gustatory nucleus of solitary tract | Thalamus; tectum | Insular cortex |
| Smell (olfactory receptors) | Olfactory sensory neurons in nose | Mitral cells in olfactory bulb glomeruli | Olfactory cortex (no thalamic relay) | Orbitofrontal cortex |
Column 5 applies to mammals (all senses but smell), where still higher processing up to the frontal lobes is probably needed for complete multisensory images. In fish and amphibians the images form in the tectum, helped by the thalamus, and no higher (chapter 6). Every pathway shows isomorphic order at each level: retinotopy for vision, somatotopy for touch, tonotopy for hearing (mapped by pitch along the cochlea's basilar membrane), and odour maps for smell, where each scent activates a specific "bar code" of glomeruli in the olfactory bulb.
Smell, the odd one out
The main olfactory path skips the thalamus, the gateway for every other sense. Jay Gottfried suggests olfactory pathways evolved before the others. Translated into the vision-first view, smell may have elaborated so soon after vision that it escaped the visual blueprint, or the olfactory cortex itself may do the thalamus's job for smell.
How many levels make a conscious hierarchy?
The authors use the number of neuron levels as a crude, countable marker for consciousness, admitting that processing within and between levels matters as much.
- 4 levels: human touch pathway (counting somatosensory cortex) and smell pathway (counting orbitofrontal cortex).
- 5 levels: human vision (counting photoreceptors and primary visual cortex) and hearing (counting primary auditory cortex).
- 3 levels: somatosensory path to the tectum in fish and amphibians (sensory neuron, spinal or brainstem relay, tectum).
- 4 levels: visual and hearing paths to the fish and amphibian tectum (hearing and lateral line relay through the torus semicircularis), and smell to the dorsal pallium.
The true number may be higher. In mammals, it is debated whether consciousness arises in primary sensory areas or higher in cortex, even up to the prefrontal cortex; in fish, much processing happens inside the tectum. The authors' working rule: sensory hierarchies can be conscious with four or more neuron levels up to and including the top processing area, sometimes just three. Larval amphioxus, with three or four levels ending in a motor centre rather than a mapping centre, falls short.
The one-pixel fish. Imagine two fish with identical bodies. One has a Stage II eye that reports only "darker on the left," wired in two synapses to a turn-away reflex. The other has a Stage IV eye whose image is re-mapped at four successive levels before any motor command. Both flee a shadow equally fast. Only the second has something that could be an image of the predator, located out there, with edges and motion, available to be bound with smell and touch. The authors' bet is that experience lives in the extra mapped levels, not in the escape behaviour, which is exactly why behaviour alone cannot settle the question.
Several links in the chain are inferences, not observations. Vision first is contested, and the authors' rebuttal relies partly on an absence of fossils from 540 to 520 mya, which cuts both ways. The step from "camera eye plus mapped processing" to "conscious mental image" is the book's central hypothesis, assumed here and defended later. Level counting depends on choices (whether photoreceptors or cortex count as levels) that shift the totals by one, and the minimum (three or four) was derived partly from the authors' own claim that the fish tectum is conscious, which risks circularity. Finally, the olfactory route is not wholly thalamus-free: a secondary path reaches orbitofrontal cortex via the mediodorsal thalamus, as the book's own smell figure hints.
- Where content becomes conscious. The Cogitate adversarial collaboration between integrated information and global workspace theorists (first results 2023, published in Nature in 2025) found conscious content most robustly represented in posterior cortex, with less prefrontal involvement than workspace theory predicted. This bears on the authors' uncertainty about how far "up" mammalian sensory hierarchies consciousness requires.
- Fossil dates. The Chengjiang biota is now usually dated to about 518 mya rather than 520, and the base of the Cambrian to about 538.8 mya, which slightly shortens the "fossil gap" the authors invoke.
- Yunnanozoans. Their placement is still disputed; a 2022 analysis argued they are stem vertebrates with cellular-cartilage gill arches, which would favour Chen and Mallatt's reading over Shu's (see chapter 3).
The sensory-pathway table is a reference architecture: six input channels, one five-stage pipeline, topology preserved end to end, and one legacy channel (smell) that bypasses the central router. For anyone designing multimodal AI systems, the vertebrate solution is instructive: new modalities were added by copying a proven pipeline template (Butler's hypothesis) rather than inventing new ones, which kept the outputs compatible enough to merge into one scene.
Two-Step Evolution of Sensory Consciousness in Vertebrates
Sensory consciousness evolved in two great steps. In Step 1, with the first fish before 520 million years ago, mapped mental images appeared in the optic tectum (all senses but smell) and in the pallium (smell), backed by arousal and attention systems; since lampreys have all of it, every vertebrate is conscious. In Step 2, independently in the first mammals (about 220 mya) and birds (about 165 mya), the memory-rich dorsal pallium expanded and became the sole end-site of sensory consciousness. The authors set this against the majority view that only mammals and birds (or only humans) are conscious.
The vertebrate clades
To track consciousness through vertebrate history, the authors first lay out who the vertebrates are. The earliest known, Haikouichthys and Metaspriggina, were jawless Cambrian filter feeders. Other jawless lines (conodonts, osteostracans, anaspids) did modestly from the Late Cambrian to about 360 mya, some invading fresh water around 440 mya. One line gave rise to the only jawless survivors, the eel-shaped cyclostomes ("round-mouths"): lampreys and hagfish.
| Clade | Members | Origin / key fact | Why it matters here |
|---|---|---|---|
| Cyclostomes (jawless) | Lampreys, hagfish | Branch from the most basal surviving vertebrate line | Test case for whether the ancestor of all living vertebrates was conscious |
| Gnathostomes (jawed) | All other vertebrates | About 460 mya in the ocean; 99% of living vertebrate species | Jaws (enlarged front gill bars) plus teeth let vertebrates take large prey and become top predators |
| Cartilaginous fish | Sharks, skates, rays, ratfish | Early gnathostome split | Some lines independently enlarged the pallium |
| Ray-finned fish | Teleosts (zebrafish, trout, perch, eels) | Most abundant fish | Main experimental fish |
| Lobe-finned fish (Sarcopterygii) | Lungfish, coelacanths | Gave rise to tetrapods about 360 mya | Lungfish are among the closest living relatives of amniotes |
| Amphibians | Caecilians, salamanders, frogs | Tetrapods | With all fish, the anamniotes ("not amniotes") |
| Amniotes | Reptiles, birds, mammals | About 350 mya, then split into synapsids and sauropsids | Where Step 2 happened |
- Lampreys. The adult is a free-swimming parasite that clamps its spiked oral hood to a larger fish and rasps away skin, fluids and muscle, quite unlike the gentle filter-feeding larva.
- Hagfish. Deep-sea burrowers that scavenge fallen fish and whale carcasses and prey on worms and burrowing fish in mud. Both cyclostomes feed with a pistonlike "tongue" tipped with horny blades.
- Shared senses. Fish have vision, smell, taste, touch, hearing and balance (inner-ear hair cells), plus the lateral line for vibrations and, in many, electric fields.
Step 1: consciousness appears
Why lampreys
The blind hagfish are poor models of the ancestor: life in deep-sea mud regressed their vision and electroreception and sharpened smell, touch and taste. Lampreys are less specialised apart from the sucker mouth and rasping tongue. They have all the vertebrate distance senses, an obvious forebrain, midbrain and hindbrain, neural crest and placode derivatives, and sensory pathways as complete as in any anamniote, comparable level by level with the five-stage mammalian plan of chapter 5. Their nervous system is simpler in three ways: no myelin on axons, a simpler visceral nervous system, and the smallest cerebrum of any vertebrate. The lamprey brain is well studied, but the functions of its pallium and thalamus are still poorly known.
The theory of isomorphic neural hierarchies
The working theory, refined from earlier chapters, is that chains and nets of neurons carry sensory signals through a hierarchical series of processing centres, with the information kept isomorphically (topographically) organised at every stage. Each centre receives outside input (OI) from elsewhere in the brain, levels talk back and forth, and near the top input arrives from other sensory hierarchies, so the map becomes more complex, multisensory and abstract as it rises. The most highly processed representations are experienced as conscious mental images.
Three themes run through the theory, and the authors insist none of them is confined to cortex:
- Topography. Point-for-point mapping of the sensory surface at each level.
- Hierarchy. Several levels of increasingly complex processing.
- Oscillatory binding. Coordinated, mostly gamma-frequency oscillations between regions are thought to encode sensory information and bind all neurons responding to one object into one percept, across senses too, giving the "unitary scene" (Edelman's phrase) and also focusing selective attention.
These themes are usually studied in the mammalian thalamus and cortex, but that does not rule out consciousness arising lower in brains with no cortex. Whether fish and amphibians are conscious, the authors insist, should be decided from evidence about those animals, not from the human state.
The optic tectum
The optic tectum roofs the midbrain. It is small in mammals (the superior colliculus), proportionally largest in anamniotes (sometimes larger than the cerebrum) and still large in birds and reptiles. It receives more retinal input than any other part of the fish and amphibian brain, and a fish without a tectum behaves as if blind.
In every vertebrate the tectum carries extreme isomorphic mapping of vision, touch, hearing, taste, lateral-line mechanoreception and electroreception, but not smell. Different layers (laminae) take different sensory classes, and the maps are in topographic register with one another, giving one coherent map of sensory space. The lamprey tectum is typical, not primitive.
- Sensory. Recognises and distinguishes objects by size, shape, newness, movement, orientation and rate of change, and assigns salience: high for objects that are fast, new, large or threateningly looming.
- Motor. Signals the reticular formation to turn the eyes and body toward objects, track them, approach or avoid them, and guide vision-driven movement through space (foraging, dodging obstacles), all demanding fine spatial resolution.
- Attentive. With the nearby brainstem isthmus nuclei, finds the most salient object in the mapped multisensory space and directs selective attention to it.
The tectum is built for fast rather than deep analysis. David Northmore describes it as handling moment-to-moment changes: the first evaluation, the quick decision, the direct reaction. A slower, more thorough evaluation can follow in the pallium.
Why should a visual centre take in so many other senses? The authors give four reasons:
- Completeness. More senses build a fuller world map than vision alone.
- Consistency. The senses cross-check that they support the same reconstruction of the world.
- Correction. Other senses fill in weak or unclear patches of the visual map.
- Efficiency. Stacked, aligned maps keep connections short. Jon Kaas: the alignments group the neurons that most often interact, cutting the need for long, slow, metabolically costly connections.
This is why, once vision evolved, the other senses soon projected to the tectum, and it marks the origin of mental unity for the exteroceptive senses.
Perception or reflex? Ursula Dicke and Gerhard Roth describe the amphibian tectum as the major centre for visual perception, where localisation, recognition and depth perception take place, using three retino-tectal subsystems: (i) size and shape, (ii) velocity and movement pattern, (iii) changes in ambient light. Mario Wullimann and Philippe Vernier call the fish tectum a centre for "object identification and location." Critics will call these the highest-order unconscious reflexes. Feinberg and Mallatt's criterion is the sequence: the tectum builds a multisensory map, attends to the most important objects in it, and drives active behaviours that show the animal is using the map.
Gamma oscillations have been recorded between the tectum and isthmus nuclei, and their discoverers noted how closely they resemble cortical gamma oscillations of spatial attention and consciousness in mammals. The authors flag the weakness themselves: the best recordings come from birds (amniotes with an elaborate tectum). Some tectal gamma has been seen in frogs and teleosts, and all vertebrates, lampreys included, have the tectum-isthmus wiring, but more recordings in anamniotes are needed.
Telencephalon and pallium
The pallium of the cerebrum has four basic parts, present in all vertebrates.
The anamniote telencephalon was once dismissed as a "smell brain". It is now known to receive visual, auditory, touch and other input, and it has the same nonsensory parts as amniotes: the hippocampus (medial pallium) for memory and the striatum for choosing motor programs. Still, smell was probably its main original job, since in the most basal lines (lampreys, hagfish, sharks) the pallium is dominated by olfactory input.
The decisive question is whether the anamniote pallium maps the nonsmell senses. Apparently not. Walter Wilczynski, writing on amphibians (and the point holds for fish studied so far): there is no evidence for separate unimodal representations and no topographically preserved projection to any telencephalic area, with a possible exception in the olfactory-recipient lateral pallium. So in fish and amphibians every isomorphic representation except smell lies in the tectum, and since all vertebrates share the same olfactory path to the telencephalon, any olfactory consciousness must arise in the pallium.
Optic tectum in anamniotes. Mapped, multisensory (all but smell), layered in register. Fast, crude first assessment. Directs gaze, orientation, approach and avoidance. Selective attention via the isthmus nuclei. End-site of nonolfactory consciousness.
Dorsal-medial pallium in anamniotes. Not mapped for nonsmell senses; receives every sense including smell, already highly processed. Slow, thorough evaluation. Stores memories. Can override the tectum as a "wiser" motor controller. End-site of olfactory consciousness only.
The two talk through the diencephalon, with especially abundant tectum-diencephalon links, which may unite olfactory and nonolfactory images. The pallium's other role shows up in lesion studies: a fish with its pallium removed can see, catch prey and act normally, but it cannot learn from experience or consequences, nor learn the locations of objects. That is a memory deficit. The authors conclude that the anamniote dorsal-medial pallium, though small and not itself the seat of mapped images, has two advantages, memory storage and convergence of all fully processed senses, which later made it the main site of brain expansion.
Memory and arousal
Gerald Edelman stressed that perception constantly calls up images from memory and updates them with new input, so the brain never rebuilds the scene from scratch; and almost nothing can be recognised without prior learning. The hippocampus builds episodic memories and memories of mapped space for finding the way, and it is found in every clade of jawed vertebrates. In lampreys it is undocumented except for hippocampus-defining genes (Lhx1/5 and Lhx2/9) in the developing brain.
The smell-memory link of chapter 5 is anatomical too: the hippocampus and the amygdala (emotional memory) lie next to the olfactory lateral pallium, and the olfactory path projects heavily to both. Hence smells evoke deep memories, and by the authors' reasoning smell-evoked memories are conscious in all vertebrates.
Arousal sets the level of consciousness (its degree or intensity). All vertebrates share the same arousal machinery: the reticular activating system (RAS) and the basal forebrain including the septal nuclei, both projecting widely and releasing norepinephrine and acetylcholine. In mammals these also help cortex direct attention; in fish and amphibians the isthmus-tectal system does that.
- Topographic maps for mental images, mostly in the optic tectum, in the pallium for smell.
- A place to form, store and retrieve memories, in the pallium (hippocampus).
- Selective attention, in the isthmus-tectal system.
- Arousal, in the reticular activating system and basal forebrain, using norepinephrine and acetylcholine.
Is amphioxus conscious?
With these vertebrate-based markers, amphioxus fails on every count.
| Line of evidence | Vertebrates | Larval amphioxus |
|---|---|---|
| 1. Sensory neurons | From neural crest and placodes | Not from crest or placodes, so not homologous |
| 2. Pathway depth | Several levels to mapping centres | One or two neurons to a single neuropile centre |
| 3. Topography | Mostly head-to-tail body maps | Holland and Yu found dorsal and ventral body axons grouped together, apart from midbody ones: unlike vertebrate mapping |
| 4. Arousal chemistry | Norepinephrine and acetylcholine in forebrain | Has a reticular "primary motor centre" but lacks both in the forebrain |
The adult is also ruled out: no distinct tectum, a frontal eye that cannot form images and sits at the very front of the brain (so probably no telencephalon). No tectum, no camera eye, no telencephalon, no consciousness.
Genes of consciousness
The authors' 2013 survey found many genes common to all vertebrates for building the three-part brain, crest, placodes and eyes, but none for isomorphic organisation across all senses, which their theory requires. They have since found it in the Eph/ephrin gene family, which signals topographic wiring in the visual, auditory and somatosensory systems. In amphioxus, Eph/ephrin genes do not seem to take part in CNS development, nor do any other consciousness-linked genes.
Step 1 in summary
As image-forming eyes evolved, visual images arose in the new optic tectum from a mapped hierarchy running from retina to tectum; these carried the first visual qualia, soon enriched by mapped input from hearing, balance, lateral line and the rebooted touch and taste. In parallel, smell images arose in the pallium. The authors give three adaptive reasons:
- Targeted action. Accurate reconstructions let the first fish aim actions at any location in a fast, dangerous Cambrian sea to feed, escape and mate.
- Short-range prediction. The brain could project its images a second or so ahead (where prey or predator will be), a version of the popular "prediction device" view. Unlike Rodolfo Llinás's motor-centred version with simplified simulations, theirs is sensory-centred with highly detailed ones.
- Evolvability. Detail could be increased for any sense simply by adding neurons to existing hierarchies.
Step 2: consciousness jumps ahead in mammals and birds
The jump
Early amniotes split about 350 mya into synapsids (mammal-like reptiles, then mammals) and sauropsids (reptiles, then birds). In both, at the origin of the crown group, the dorsal pallium became the predominant sensory processor and the end-site of sensory consciousness: first in mammals about 220 mya, then in birds, which arose from dinosaurs about 165 mya.
- 10 times: brain size of living mammals and birds relative to living reptiles, adjusted for body weight.
- 3 to 5 times (300 to 500%): the jump at the origin of the first mammals and birds over their reptile-grade ancestors, achieved in only tens of millions of years.
- Under 2 times: living reptile brains relative to living amphibians, so the jump was not a gradual climb from the first amniotes.
- 10 kg to 0.8 kg: body shrinkage across the bird-line dinosaur radiation (toddler-sized to crow-sized), within about 5 million years (Michael Lee and colleagues).
- 190 to 220 mya: last common ancestor of monotremes, marsupials and placentals.
Because the growth was in the sensory pallium, many researchers conclude consciousness began in mammals and birds and exists only there: Ann Butler, Rodney Cotterill, Melanie Boly, Bernard Baars, David Edelman and Anil Seth among them. This mammal-bird hypothesis has more supporters than the all-vertebrates view.
The homeothermy shortcut fails
One tempting common cause is that mammals and birds are both homeothermic (warm-blooded): a faster metabolism would let neurons signal faster and meet the computational demands of richer consciousness. The problem is timing. Warm-bloodedness may have evolved gradually, or earlier, in pre-mammal synapsids or in the bird-ancestor dinosaurs called coelurosaurs. If brain expansion and homeothermy do not coincide, homeothermy cannot explain the shared advance, so the clades must be examined one at a time.
Mammals: from vision to smell and touch
The mammalian case is "not so puzzling." Timothy Rowe, Francisco Aboitiz, Margaret Hall and others documented that, in the transition about 220 mya, cortical growth came with a shift from vision to smell and touch as the main senses, an adaptation to night life on the ground where odours are rich and body hairs (and later vibrissae) sense light touch. Fossil brain casts show enlarged olfactory bulbs, lateral pallium and somatosensory cortex. Declining vision explains the small tectum of all mammals. The authors infer that the cortex became the end-site for the senses that had reached consciousness in the tectum (touch, hearing, vision, balance): a shift from a shrinking region to a growing one that enriched smell images with the other senses.
All three living mammal clades, monotremes (platypus, echidna), marsupials and placentals, have a large cortex with an isomorphic primary sensory area for each conscious sense plus association areas, and large olfactory cortex in basal members. The common ancestor date (190 to 220 mya) sits near the first fossil mammals, supporting that as when mammalian consciousness became fully cerebral.
Birds: the harder case
Birds pose the opposite profile: from dinosaur ancestors their already fine vision sharpened and smell stayed the same. So why move visual consciousness into the largely olfactory pallium when the tectum already did the job? The answer must be a built-in advantage of the pallium: it alone receives all senses including smell, processes them more thoroughly, and compares them with memories recalled by the nearby hippocampus, allowing better decisions from past learning.
More memory changed experience in four ways:
- Less fleeting experience. Consciousness became less "in the moment" and carried more of the recalled past: more smell memories in mammals, more visual memories in birds.
- More recognition. More sensed objects could be recognised and attended to.
- Learned attention. An advanced selective attention emerged in which the pallium attends to stimuli whose importance has been learned.
- Cognition. More interpretation of primary images and accumulated knowledge of the world, setting the stage for higher consciousness.
Special habitats paid for an expensive brain
Bigger pallial processing costs time and energy, so it could only evolve under special conditions. The first mammals and birds both lived in forests or under cover, both were miniaturised, and both were hunted by larger predatory dinosaurs (reconstructions by Michael Lee, Michael Benton, Zofia Kielan-Jaworowska and colleagues).
- Nocturnal mammals. Scurried through forest-floor litter rich in smells and textures, sensed with whiskers, cloaked by darkness and cover.
- Daytime birds. Mastered a three-dimensional tree world with keen vision, flying between branches without collisions or falls; Amy Balanoff, Lee and colleagues showed brain enlargement spread across the whole feathered, partly flying bird-line dinosaur radiation, not only the first true birds.
- Safe neighbourhoods. Escape by flight or concealment gave both time to weigh options, so slow, elaborate processing was no handicap.
- Rich fuel. Both ate insects, abundant high-energy food for an energy-hungry brain.
The pallial advance in mammals and birds is a case of convergent evolution: unrelated clades in similar habitats, under similar predators, evolved the same solution, memory-enhanced, pallium-centred consciousness, by different anatomical routes.
The same story could serve the rival "pallial" hypothesis in which the tectum was never conscious and birds and mammals each evolved consciousness from scratch. The authors find two separate origins more unwieldy than one origin in the first fish. But they concede their own weak point: birds would then have two dominant sites of detailed visual images, tectal and pallial, and it is unclear how such distant images could be united into one.
Pallial diversity
Mammal and bird pallia look utterly different. The mammalian cortex has six layers (I to VI) across its whole cap. The bird dorsal pallium has a large unlayered dorsal ventricular ridge (DVR) plus an outer hyperpallium (Wulst) whose thick layers do not resemble mammalian laminae.
Comparing axon connections, microcircuits and gene expression, Harvey Karten, Jennifer Dugas-Ford, Erich Jarvis, Ann Butler and colleagues resolved the puzzle: the bird equivalents of the six laminae are scattered as nuclei and thick bands rather than stacked, and the bird visual area lies far forward. Yet many functional neuron groups, microcircuits and long sensory connections are the same, so the two clades must share many ways of generating images, by different layouts.
Which arrangement is ancestral is unknown. Living reptiles have both a DVR and a three-layered cortex, but all are sauropsids, so they cannot show whether premammal synapsids had a DVR. Judging from amphibians and lungfish, the first amniote had a one-layered cortex, and whether it had multiple isomorphic sensory regions is unknown; the reptile pallium is badly understudied.
Diversity goes further. The dorsal pallium enlarged independently in hagfish, certain sharks, skates, rays and some teleost clades, each time by adding novel nuclei or laminae, always correlated with an especially acute sense. All the jawed fish on that list arose in the Mesozoic, the same era as the mammal and bird expansions. The authors ask whether the Mesozoic was an age of widely advancing consciousness, an echo of the Cambrian arms race.
Conclusion: older, wider, more diverse
Isomorphic sensory consciousness is more widespread (all vertebrates), older (over 520 million years) and more diverse (tectal and pallial; different pallial designs in mammals and birds) than most researchers assume. That diversity is the "diverse architectures" feature postulated in Table 2.1. Hagfish, despite their oddities, should also be conscious: their brain is three times bigger than a lamprey's and bulges with smell, touch and taste areas.
- Bjorn Merker: trailblazer of tectal consciousness in all vertebrates; unlike the authors, holds that consciousness evolved to solve the logistics of self-motion and that the superior colliculus remains a major conscious centre in mammals.
- Antonio Damasio: suggested the human superior colliculus has crude images, which might explain blindsight and vision in hydranencephalic children born without cortex. The authors see the mammalian colliculus as having lost almost all its conscious role.
- Georg Northoff: proposes that consciousness rests on a stimulus-imprinted spatiotemporal structure of brainwaves and time-difference-based coding; promising, but too early for the authors' universal criteria in Table 2.1 and Table 9.2.
- Fabbro and colleagues (2015): independently argued for consciousness in every vertebrate.
- Mehta and Mashour: classify theories like this one as first-order representational; their own version requires cortex.
The tectal-consciousness claim rests on inference, not demonstration. Words like "perception" and "recognition" in Dicke and Roth are descriptive conventions, not evidence of experience, and tectal gamma oscillations are documented mainly in birds, not the anamniotes the argument needs. Fish consciousness remains fiercely contested (Rose and colleagues, 2014, among the critics cited). The lamprey verdict leans on brain structures whose functions are, by the authors' admission, poorly known. The dismissal of homeothermy shows only that timing is uncertain, not that it is wrong, and the "Mesozoic age of consciousness" is speculation. The authors also concede the unresolved double-image problem in birds. Finally, the rival single-origin versus double-origin choice is decided by parsimony, which is weak where convergence is admitted to be common.
- Maps in the lamprey pallium. In 2017 Sten Grillner's group reported that the lamprey dorsal pallium contains a visual area and a somatosensory area receiving topographically organised input via the thalamus. If confirmed broadly, this undercuts Wilczynski's "no mapped unimodal pallial representations" and suggests mapped pallial sensing is as old as vertebrates, which would soften the tectum-only story for anamniotes.
- Bird pallium and consciousness. Two 2020 papers in Science strengthened the bird side of Step 2: Andreas Nieder's lab found neurons in the crow pallium whose activity tracked the bird's subjective report of near-threshold stimuli, a neural correlate of sensory consciousness in a bird; and Martin Stacho, Onur Güntürkün and colleagues described cortex-like columnar circuits running across the avian pallium. In 2016 Olkowicz and colleagues showed parrot and corvid pallia pack primate-scale neuron numbers at very high density.
- Cell-type homology. Single-cell transcriptomic work since 2018 (starting with turtle and lizard pallia from Gilles Laurent's lab) has challenged a strict layer-to-nucleus homology between bird and mammal pallia, favouring similar circuits built partly from differently derived cell types. The authors' conclusion of shared function by different anatomy survives; the specific Karten mapping is disputed.
- Animal sentience declarations. The 2024 New York Declaration on Animal Consciousness holds that there is strong scientific support for conscious experience in mammals and birds and at least a realistic possibility in all vertebrates (and many invertebrates), close to the authors' position.
The two-step model reads like a platform migration. A fast, mapped, low-latency engine (the tectum) handled real-time decisions; a slower, memory-backed service (the pallium) first ran one legacy workload (smell) and then, where the environment rewarded considered decisions over raw speed, took over everything. For AI architects, the lesson is that the expensive deliberative tier wins only where the cost of latency is low: the "safe neighbourhood" condition. Real-time fraud screening and an analyst-assist copilot sit on opposite sides of exactly that trade-off.
Searching for Sentience: Feelings
"Something it is like to be" is not one problem but three. Besides exteroceptive consciousness of mapped, outward-referred images, there is interoceptive consciousness of body states and affective consciousness of good and bad feelings, the core of sentience. Interoception is the intermediate, dual kind, partly mapped and suffused with affect; affect is global, unmapped and referred to the self. Six earlier theories of where and when affect evolved disagree so widely, from over 560 million years ago to 200,000 years ago, that they settle nothing, which sets up the authors' own answer in chapter 8.
What sentience means
Sentience comes from the Latin sentiens, "feeling." It can mean all sensory consciousness, but in ethics and the animal-rights literature it means the capacity to experience pleasure or pain. The authors adopt the narrow sense: an animal is sentient if it can experience an affective state.
- Valence. Affective experiences have valence: the negativity (aversiveness) or positivity (attractiveness) of a stimulus or event. The most basic affects are feeling good or bad.
- Affect is conscious by definition. The authors, like many, define affect as a conscious experience. Joseph LeDoux and others hold that affects can be unconscious.
- Avoiding "emotion." Definitions of emotion range from the feeling itself, to a cognitive reaction, to sensing a bodily imbalance, to the physiological and motor responses; so the authors use "affect," "affective state" or "affective feeling", keeping "emotion" for human affective feelings.
The key difference is that exteroceptive consciousness assigns no valence to its images. The sight of a lion or the smell of a rose evokes fear or bliss only after further processing by systems that supply the affective and inner-body aspects.
Three types of sensory consciousness
| Feature | 1. Exteroceptive | 2. Interoceptive (and pain) | 3. Affective / limbic |
|---|---|---|---|
| Senses or contents | Distance receptors: vision, smell, hearing, taste, discriminative touch, sharp skin pain | Skin pain, visceral pain, mechanical senses in the body (stretch or irritation of organ walls), chemoreception (taste, smell) | Pleasure and displeasure, all the felt emotions |
| A. Name | "Isomorphic sensory consciousness" | "Interoceptive sensory consciousness" | "Affective awareness" |
| B. Reference | Projicient: refers to the external environment; about external mapped objects | "Dual aspect": both projicient and visceral (inner body and outside world); the input for homeostatic responses | Nonprojicient, referred to the "self" |
| C. Experienced as | Local "mental images" | Both local mental images and global inner-body, visceral and affective states | Global affective states (feelings, emotions); assigns values to internal and external sensory processes |
| D. Isomorphism and valence | High isomorphism, low intrinsic affect | Moderate isomorphism; sensations suffused with affect | Low or no isomorphism, high affect |
| E. Anatomy | Visual, discriminative touch and auditory paths are purely exteroceptive; pain, smell and taste are both; paths reach cerebrum and optic tectum | Spinothalamic and trigeminothalamic paths plus insula and anterior cingulate cortex | Mostly limbic: reticular formation, periaqueductal gray, mesolimbic reward system, basal forebrain, amygdala, insula, anterior cingulate, parts of prefrontal cortex |
Exteroception
Exteroceptive qualia are projicient and experienced as externalised images. They are subjectively "local": points in the image correspond to points in the environment (vision, touch) or on a chemical map (smell, taste). They are highly isomorphic and largely affect-free, gaining feeling only when relayed to affective brain regions.
Interoception and pain
Interoceptive consciousness is the intermediate type. Widely distributed endings measure deep and shallow pain, temperature, chemical states and sensations from the gut, lungs, bladder, heart and blood vessels: from simple stretch of a distended organ, to the tickle before a sneeze, to hunger, thirst, nausea, "air hunger" and fatigue. These signals are the input for homeostasis (homeo, constant; stasis, state). Their paths are isomorphic but more diffuse and crudely organised than exteroceptive paths, except that sharp skin pain is highly somatotopic. Interoceptive feelings may be localised (stomach, lungs, muscles) yet are often body-wide, affective and motivational.
Two blends make interoception "dual" in many ways:
- Skin pain, temperature and itch. Traditionally counted as exteroceptive, but A. D. (Bud) Craig classes them as interoceptive and homeostatic because they serve body maintenance and carry strong affect. A pinprick gives both an exquisitely precise image of the pin's tip and an unpleasant feeling. Pain's subtypes split the same way: sharp pricking pain is precisely localised (exteroceptive-like); slow, dull, burning pain spreads over larger areas (interoceptive-like).
- Chemosenses. Taste and smell are exteroceptive (stimuli from outside, mapped pathways) and interoceptive (receptors inside the body, instantly judged good or bad, quickly motivating).
The mammalian interoceptive and pain pathways
The sensory neurons have thin C and Aδ fibres: in nociception, C fibres carry slow, dull pain linked to prolonged suffering, and Aδ fibres carry the fast, sharp pain of the initial injury. Spinal input ascends in the spinothalamic tract; cranial input in the solitary tract or trigeminothalamic tract. All relay through the thalamus to cortex, as exteroceptive paths do, but they send many more branches to subcortical limbic centres: hypothalamus, parabrachial nucleus (pons-midbrain junction), limbic reticular formation, ventral tegmental area and amygdala. That heavy inflow is the physical link between interoception and affect.
| Pathway | Relay and tracts | Fibres; mapping | Carries | Final targets |
|---|---|---|---|---|
| A1. Spinal interoceptive (neck, trunk, limbs) | Lamina I of spinal cord (HPC: heat, cold, pinch neurons; visceroceptive neurons); lateral spinothalamic tract ("paleospinothalamic", spinoreticular, spinomesencephalic, spinoparabrachial, spinohypothalamic); then solitary nucleus, parabrachial, medullary reticular formation, PAG, tectum, VMpo of thalamus, hypothalamus, amygdala | C and Aδ; slow, somatotopic (viscerotopic, organotopic) | Mostly tissue damage and organ fullness | Insula, anterior cingulate, somatosensory S3a, 3b, S1 (pain), orbitomedial prefrontal; in fish and amphibians, parts of pallium and telencephalon |
| A2. Cranial interoceptive (trigeminal, facial, glossopharyngeal, vagus nerves) | Nucleus of the solitary tract, then parabrachial, reticular formation, PAG, VMpo, hypothalamus, amygdala | C and Aδ; slow, somatotopic | Chemo-, mechano-, baroreceptors; hunger, thirst, nausea, respiration, taste | Insula, anterior cingulate, orbitomedial prefrontal; telencephalic targets in anamniotes |
| B1a. Affective pain (medial; old "paleospinothalamic") | Lamina I HPC neurons; the lateral spinothalamic tract family | C; slower, somatotopic | Second (burning) pain, heat, pinch, cold | Insula, anterior cingulate, orbitomedial prefrontal, S3a; pallium in anamniotes |
| B1b. Discriminative pain (lateral; old "neospinothalamic") | Lamina I NS (nociceptive-specific) and Lamina V WDR (wide-dynamic-range) neurons; spinothalamic tract | Aδ; faster, highly somatotopic (Lamina I) or not (Lamina V) | First (sharp) pain; Lamina V many modalities | Somatosensory cortex S3b and S1 |
| B2. Head pain (trigeminal) | Lamina I equivalent in spinal trigeminal nucleus; trigeminothalamic tract | C and Aδ; slow and fast, somatotopic | First and second pain, heat, pinch, cold | Insula, anterior cingulate, orbitomedial prefrontal; pallium in anamniotes |
The shading shows the dual nature: paths start isomorphic and lose isomorphism in subcortical limbic structures. The branch reaching cortex keeps its mapping to the primary viscerosensory cortex in the insula (and, for pain, to primary somatosensory cortex); beyond the insula (anterior cingulate, orbitomedial prefrontal cortex), interoceptive isomorphism is lost.
Interoceptive pathways in other vertebrates
Less is known, especially for the spinal parts, but the plan is basically the same. A mammal-like spinothalamic tract is not universal, yet all vertebrates have the spinoreticular pathway; the taste path has the same plan throughout; the trigeminal path everywhere has a spinal trigeminal nucleus (nociceptive in mammals); and all vertebrate groups have a nucleus of the solitary tract, parabrachial nucleus and brainstem reticular formation. The highest levels in fish and amphibians are poorly studied and probably simpler, given their simpler cerebrum.
The affective / limbic system
Affective consciousness is global and involves the whole self. It assigns the basic good and bad states and underlies complex human emotions (sadness, joy, shame, despair, fear). It is tied more directly than the other two types to motivation, drives and behaviour: positive affects (liking, pleasure) drive approach, negative affects (dislike, discomfort) drive avoidance or escape. It is not isomorphic and not in direct contact with the environment, yet exteroceptive experiences (a rainbow, a concerto) gain feeling through it, and it is even more closely tied to interoception.
How do stimuli become feelings?
View 1: James-Lange (William James 1884, Carl Lange 1885). Valent stimuli first change the body; interoceptors sense the change, and the brain builds the affect. Exteroceptive stimuli must be double-translated (body response, then interoception). Support: raising blood pressure experimentally amplifies affective responses to many experiences.
View 2: direct limbic route (Walter Cannon 1927 onward). Interoceptive and exteroceptive inputs reach the limbic system directly, without first changing the body. Support: people still react emotionally after a severed spinal cord, a cut vagus nerve or bilateral removal of the insula, all of which block most interoceptive signals. Georg Northoff calls this "relational coding" against James-Lange "translational coding".
The authors combine the two: inputs reach the affective brain both directly and indirectly through sensed bodily change.
Past theories of interoceptive and affective consciousness
Interoceptive theories
Bud Craig: the "sentient self" in the insula. All interoceptive and homeostatic information (pain, itch, temperature) is mapped somatotopically in the dorsal posterior insula, then re-represented in the mid and anterior insula, culminating in a "cinemascopic representation" of the sentient self, the "material me" (Sherrington's term). Craig holds that the von Economo neurons concentrated in the anterior insula and anterior cingulate of the biggest-brained mammals are critical, so only humans, chimps, gorillas, whales, porpoises and elephants have sentience, self-awareness and presumably conscious pain. He sees this as consistent with James and with Damasio's body-based feelings. Appeal: simple localisation and a recent origin (within 35 million years). Problems: an insula may exist in all amniotes, implying 350 mya; and men whose insulae were destroyed by herpes simplex tested emotionally normal, while stimulation and recording studies suggest the insula serves pain and homeostasis without accounting for sentience.
Charles Vierck and colleagues: pain in SI. Working on monkeys and humans, they found the C pathway for slow burning pain projects not only to the insula but also to primary somatosensory cortex (S3a of SI), where it integrates with Aδ fast-pain input to produce the fully conscious pain experience. Locating pain perception in SI rather than the insula sets them apart from Craig. They gave no date; all mammals have SI (220 mya), but only monkeys and humans were studied (common ancestor about 35 mya), so between 220 and 35 mya.
Affective theories
- Derek Denton: primordial emotions. The Australian physiologist holds that the first consciousness of any kind was "primordial emotions", the felt side of instinctual drives (thirst, hunger, air hunger, hunger for specific minerals), which evolved from complex reflexes into feelings. Unmet needs are "imperious sensations" producing "compelling intentions" toward "gratification." The centres are in brainstem and hypothalamus; complex emotions, the insula and limbic forebrain, exteroception and exteroceptively induced affect came later. His dating wavers: decorticate mammals show emotion via brainstem regions shared even with cephalochordates (560 mya); yet he names the full cerebrum and neocortex as early substrate (220 mya); yet he ties it to true thirst in the first land amniotes (350 mya).
- Jaak Panksepp: primal affects. Emotion extends homeostasis. Some nonhuman amniotes have "primal affects", valenced qualia from "intrinsic brain value systems" in caudal (brainstem) and medial (limbic forebrain) structures, the first seat of subjective experience. Panksepp argues that an affectively responsive organism came before sophisticated distance receptors, because animals first had to feel homeostatic needs and be driven by affect to forage. Douglas Watt concurs. Panksepp long assigned consciousness to mammals (220 mya) or amniotes (350 mya), but his subcortical structures exist in all vertebrates, and in 2015 (with Fabbro and colleagues) he endorsed 560 to 520 mya; Andrew Packard and Jonathan Delafield-Butt also claimed a Cambrian origin.
- Mark Solms (with Panksepp): affect is subcortical. Against the dominant view that affect needs cortex, Solms gathers evidence for subcortical origin: deep-brain electrode or drug stimulation evokes emotional reactions (distinct facial expressions in rats); removing a rat's cortex makes it more emotional, not less; humans with cortical lesions keep their primal affects. The cortex only fine-tunes, inhibits or amplifies emotions generated subcortically.
- Michel Cabanac: pleasure as common currency. Sensations have four dimensions: quality, intensity, duration and affect. Affect assigns a degree of pleasure to each stimulus, ranking stimuli by salience so the most survival-relevant win, which spares the brain from storing millions of preprogrammed reflexes: affect promotes survival by increasing neural efficiency. His 1996 paper addressed only why, but its logic reaches back past 560 mya to choice-making bilaterian worms, and one of its tables assigned consciousness to the first vertebrates. In 2009, using trade-off decisions, play and sleep brain waves, Cabanac and colleagues dated affect to the first amniotes, 350 mya.
- Joseph LeDoux: survival circuits. From fear conditioning in rodents (a tone paired with shock), LeDoux shows that the amygdala and its connections drive defensive behaviours (fleeing, freezing), circuits ancient and conserved across vertebrates. But he is agnostic or sceptical that nonhuman mammals feel anything: conscious feelings need cognitive self-observation and appraisal, which only humans have. He now calls these "defensive survival circuits" (also for nutrients, fluid balance, temperature, reproduction) that produce "global organismic states" without needing awareness, present even in invertebrates, with non-neural defences in single cells. Fully cognitive emotions date only to Homo sapiens, 200,000 years ago.
There is no need for conscious feelings of fear to intervene. The circuit function is the intervening variable.
- SEEKING: to meet homeostatic needs (food, water, body temperature).
- RAGE: to avoid bodily harm.
- FEAR: to avoid bodily harm.
- LUST: to promote reproduction.
- CARE: a later, uniquely social-emotional system.
- GRIEF (separation distress): social.
- PLAY: social.
The first four are the older "reptilian emotions"; the social three appeared, Panksepp says, before birds and mammals diverged. All are carried by an "affective core self".
Hydranencephaly. In this congenital disorder a blood-vessel blockage before birth destroys nearly all the cerebral cortex, sparing brainstem, diencephalon and basal telencephalon. Bjorn Merker reports that these children smile and laugh with pleasure, show aversion by fussing, back-arching and crying, and can be led by a familiar adult through play from smiling to giggling, laughter and great excitement. Despite cognitive, motor and exteroceptive limits, they learn operantly (kicking at or activating special toys for reward). Their behaviour indicates affect without cortex, implying that animals with the same subcortical regions can feel the same.
What the past theories leave unresolved
| Theory | 1. Brain site | 2. Date (mya) | 3. Affect or exteroception first? |
|---|---|---|---|
| A. Interoceptive-centred | |||
| Craig | Insula (ACC, PFC) | 350 or 35 | (not addressed) |
| Vierck | SI cortex | 220 or 35 | (not addressed) |
| B. Affective-centred | |||
| Denton | Subcortical limbic areas and cortex | about 560, 350 or 220 | Affective |
| Panksepp / Solms | Subcortical limbic areas | 350 or 220 or 560 to 520 | Affective |
| Cabanac | Cortex | before 560, 560 to 520, or 350 | Affective? |
| LeDoux | PFC, parietal cortex | 0.2 | Both |
| Feinberg and Mallatt (chapter 8) | Subcortical limbic areas | 560 to 520 | Both |
Craig's and Vierck's dates are inferred, since their theories are human-centred. Dates code for: 560 to 520 first vertebrates; 350 first amniotes; 220 first mammals; 35 human-monkey ancestor; 0.2 Homo sapiens.
Three questions stay open:
- Where? Subcortical limbic system, cortex (but insula, SI or prefrontal?), or both.
- When? Estimates run from over 560 mya to 200,000 years ago, spanning all vertebrate or even bilaterian history, a range so wide it is meaningless.
- Which came first? The table's theories favour affect first, but that reflects selection: the many exteroception-centred theories of chapter 6 (Crick and Koch, Dehaene, Tononi and Koch) would say otherwise, so the literature is unreconciled.
Sentience is a separate problem from mapped perception. Its neural basis is unmapped and largely subcortical, which is why theories that look only at cortex and those that look at the brainstem give answers hundreds of millions of years apart.
The flawless forecaster. Imagine a fish that maps its surroundings perfectly (every rock, current and predator in register) but whose images carry no valence at all. A pike looms: the fish sees it with perfect clarity and is indifferent. Without some ranking of good and bad, perfect exteroception gives no reason to move. The thought experiment shows why Cabanac calls affect a common currency: something has to convert a detailed world model into a choice, and the authors' three-part split says that conversion is done by a different, unmapped system.
Several moves here deserve scrutiny. Defining affect as conscious settles by stipulation the very point LeDoux disputes; behavioural signs of "affect" then risk being read as proof of feeling. Table 7.3 is avowedly biased toward affect-first theories. The hydranencephaly evidence is powerful but interpretive: Merker's reports come largely from caregiver observations, and some of these children retain small cortical remnants. Craig's von Economo neuron argument was already weakened by reports of these cells in other mammals, including macaques. Finally, the paleospinothalamic / neospinothalamic split is a simplification the authors themselves label "old."
- LeDoux's position hardened. With Richard Brown, LeDoux published a higher-order theory of emotional consciousness (PNAS, 2017): felt fear is a cognitive, higher-order representation, not a product of amygdala survival circuits, keeping the sharp split from Panksepp's camp.
- Solms extended the subcortical view. In The Hidden Spring (2021), Solms argues with Karl Friston's free-energy framework that consciousness itself originates in brainstem affect, a stronger affect-first claim than the one summarised here. Panksepp died in 2017.
- Sentience as valence, in policy. The 2021 LSE review for the UK government (led by Jonathan Birch) used essentially the authors' definition, sentience as the capacity for valenced experiences, and its findings led the Animal Welfare (Sentience) Act 2022 to cover decapod crustaceans and cephalopod molluscs alongside vertebrates. Birch's The Edge of Sentience (2024) develops a precautionary framework for such cases.
The three-way split maps neatly onto how decision systems are built in banking. Exteroceptive data is the mapped world model (transactions, positions, counterparties); interoceptive data is the institution's own state (liquidity, capital, limits); and affect is the valuation layer that converts both into approach or avoid: risk appetite, scoring, alerts. Cabanac's point applies directly: a single common-currency score is what lets a system choose without hard-coding a rule for every case, and getting that layer wrong produces a perfectly informed system that does nothing useful.
Finding Sentience
Two independent lines of evidence, behaviour and brain anatomy, put the origin of affective consciousness in the first vertebrates, 560 to 520 million years ago, the same window as exteroceptive consciousness. The behavioural test is strict: only operant learning and its relatives (trade-offs, frustration, self-medication, drug seeking) count, because reflexes, classical conditioning and the acts of decerebrate animals do not. Core affects are subcortical, and the lamprey's limbic system is the minimum that supports them. The exception is suffering, burning pain, which fish seem to lack and which probably arrived last, with the tetrapods.
Pain as the test case
Pain is the best-studied and most controversial window on animal feeling, and the prototype of negative affect. The standard definition (the International Association for the Study of Pain): an unpleasant sensory and emotional experience associated with actual or possible tissue damage, or described in terms of such damage; always subjective. The authors focus on prolonged, burning C-fibre pain, because that bears on suffering and on humane treatment.
The anchor is the 2009 National Research Council report Recognition and Alleviation of Pain in Laboratory Animals, a collaborative, relatively unbiased review by neuroscientists, behaviourists and pharmacologists. Its central distinction is between nociception and conscious pain.
- Nociception. Described for the skin by Charles Scott Sherrington in 1906: the response of nociceptors to stimuli that damage or threaten tissue (cuts, scrapes, burns, crushing, shocks). It can drive reflexive or even complex behaviour without any felt pain, as in single-celled organisms and nerve-net animals such as jellyfish and comb jellies.
- Decerebrate mammals. Rats and dogs whose cerebrum and diencephalon are disconnected still respond to noxious stimuli with sympathetic activation (faster heart, higher blood pressure, dilated pupils). Decerebrate cats and even spinally transected rats can learn to keep a limb withdrawn to avoid shocks. Decerebrate rats struggle and vocalise at a feeding tube, push it away, bite the syringe and the experimenter's hand, lick the injection site, and after a formalin injection shake, groom and lift the injured foot like normal rats. So painlike behaviour can occur without cortex and thus without pain.
What is lost after decerebration is the better marker. Operant (instrumental) conditioning, learning to link a behaviour to its consequence (a rat pressing a bar for food; cattle avoiding an electric fence), is largely eliminated by cortical lesions, whereas classical (Pavlovian) conditioning, linking a conditioned stimulus (a bell) to an unconditioned one (food), survives. Decorticate rats show all the nociceptive responses and classical conditioning, but cannot learn to avoid the noxious stimulus. Operant learning is demonstrated in only some phyla; classical learning in almost every bilaterian phylum.
The NRC committee added two more markers: a response must last long enough to rule out nonconscious reaction, and self-delivery of analgesics counts. Rats with ligated spinal nerves pressed a lever for the analgesic clonidine while controls did not; lame rats and arthritic chickens self-administered NSAIDs in their food while healthy ones did not. Such evidence came only from mammals and birds, and fish pain remained hotly debated.
Behavioural criteria for affect
The authors generalise the pain criteria to positive and negative affect, demanding behaviour beyond reflexes and fixed motor programs.
Table 8.1: behaviours that do NOT indicate affect - Simple approach or avoidance. Single cells without nervous systems do it. - "Spinal" learned responses. Includes a simple, incomplete operant leg flexion in the isolated spinal cord. - Classical conditioning. Mere reflexes can be conditioned. - Reflexive responses in decerebrate mammals. Pushing away syringes, licking or guarding the injured site, vocalising, jumping. - Innate, possibly reflexive behaviours. Basic motor programs such as chewing, freezing, fleeing.
Table 8.2: operant behaviours that probably DO indicate affect - Global, nonreflexive operant learning from a valenced result (adds memory of the affect to approach or avoidance). - Behavioural trade-offs: value-based cost-benefit decisions (two valences weighed against each other). - Frustration behaviour. - Successive negative contrast: degraded behaviour after a learned reward unexpectedly stops (frustration and contrast show prolonged negative affect). - Self-delivery of analgesics or rewards. - Approach to reinforcing drugs / conditioned place preference (these two go beyond approach to seeking and pursuit).
Two popular criteria were rejected. Behavioural disruption (changed activity, hiding, disrupted feeding, sleep or social life, inattention, anxiety) is not specific to affect, cannot be separated from the general effects of injury, stress or shock, and was designed for mammals. Persistence in pursuing a reward could reflect desire, but also aroused, unconscious habit cycling the pattern generators. Play might be a good positive marker (found in all amniotes and possibly amphibians and bony fish) but was not used.
Who passes
To approximate the prevertebrate stage, the poorly studied amphioxus is supplemented by protostomes at a similar level: nematodes, flatworms and gastropods. Tunicates are left out as secondarily regressed (only sensitisation and habituation).
| Criterion (Table 8.3) | Nematodes | Flatworms | Gastropods | Amphioxus | Fish | Amphibians | Reptiles | Birds | Mammals |
|---|---|---|---|---|---|---|---|---|---|
| Operant learning (rewards or punishments) | C. elegans? | - | Sea hare, pond snail | - | Carp, stingray, sea bream, others | Toad, others | Turtle, others | Pigeon | Rat, dog |
| Behavioural trade-off | - | - | Predatory sea slug | - | Trout, goldfish | Toad, frog, salamander | Lizard | Sparrow | Rat, human |
| Frustration / contrast | -/- | -/- | -/- | -/- | Trout / - | -/- | Lizard? / - | Pigeon / starling | Rat / rat |
| Self-delivery of analgesics or rewards | - | - | Land snail Helix? | - | Nine teleosts | - | Crocodile? | Pigeon, chicken | Rat |
| Drug approach / conditioned place preference | C. elegans? | Flatworm | - | - | Goldfish, zebrafish | Frog | Lizard? | Quail | Rat, monkey |
The fish column is mostly teleosts; lampreys and sharks are barely tested. The pattern is clear:
- Every vertebrate group shows most of the behaviours.
- Amphioxus and the worms do not.
- Gastropods, with some cephalisation, show two or three of the five or six.
- Only mammals and birds show behavioural contrast (agitation, aggression, leaving or stress lasting after a reward fails), which may reflect anticipation and a remembered disappointment or anger.
- The decisive jump is between the cephalochordate and vertebrate stages.
Structural criteria: the affective-limbic brain
The anatomical test assumes that homologous structures, especially when wired into similar networks, serve similar affective functions. Functions are established in mammals and partly confirmed elsewhere. Protostomes are listed for reference: they lack all vertebrate brain structures, so any affective machinery they have evolved independently.
| Table 8.4: general affect structures (mammalian function) | Protostomes | Amphioxus | Lampreys | Sharks, rays | Bony fish | Amphibians | Reptiles | Birds | Mammals |
|---|---|---|---|---|---|---|---|---|---|
| 1. Dopamine neurons: reward learning and seeking (dopamine signals reward prediction error) | + (all Bilateria) | + | + | + | + | + | + | + | + |
| 2. Nociceptors (needed, not sufficient, for pain) | + (worm, fly, leech, sea slug, squid) | - | + | + | + | + | + | + | + |
| 2A. Nociceptive fibre types | Not homologous | - | Fibres, no subtypes | Aδ only | Aδ / C sparse | Aδ / C | Aδ / C | Aδ / C | Aδ / C |
| 3. Opioid receptors (relief of nociception) | Flatworm, scallop, octopus (not C. elegans or Drosophila) | ? | + (2 types) | + (4) | + (4) | + (4) | + (4) | + (4) | + (4) |
| 4. Autonomic nervous system | - | - | + (incomplete) | + | + | + | + | + | + |
| 5. Spinoreticular, spinomesencephalic, trigeminothalamic nociceptive paths | - | ? | + | + | + | + | + | + | + |
| 6. Amygdala: fearlike avoidance, memory, learning, some feelings? | - | - | +? | ? | + | + | + | + | + |
| 7. Periaqueductal gray: panic, reproduction, aggression, calling, pain relief | - | - | + | + | + | + | + | + | + |
| 8. Lateral septum: social behaviour, territoriality, novelty, mood, stress, anxiety | - | - | + | + | + | + | + | + | + |
The mesolimbic reward system (MRS), or mesolimbic dopamine system, is the limbic division that judges the salience of stimuli and assigns positive or negative value.
| Table 8.5: mesolimbic reward system (mammalian function) | Protostomes | Amphioxus | Lampreys | Sharks, rays | Bony fish | Amphibians | Reptiles | Birds | Mammals |
|---|---|---|---|---|---|---|---|---|---|
| 1. Ventral tegmental area (VTA): dopaminergic salience, reward, analgesia, aversion | - | - | - | Shark | - | - | + | + | + |
| 1A. Diencephalic VTA equivalent (posterior tuberculum) in anamniotes | - | ? | + | + | + | + | n/a | n/a | n/a |
| 2. Nucleus accumbens: motivation into action, approach or avoid; pleasure, dread, fear | - | - | - | - | Teleosts? | + | + | + | + |
| 3. Ventral pallidum: reward processing for motor output | - | ? | ? | ? | + (zebrafish) | + | + | + | + |
| 4. Laterodorsal tegmental nucleus: emotional intensity, exploratory locomotion | - | ? | + | + | + | + | + | + | + |
| 5. Striatum: learning and reinforcing goal-directed behaviour | - | ? | + | + | + | + | + | + | + |
| 6. Pallial amygdala (basolateral in amniotes): emotional and fear learning, operant avoidance, valence | - | - | ? | ? | + | + | + | + | + |
| 7. Hippocampus: experiences and spatial maps for operant learning | - | - | ? | + | + | + | + | + | + |
| 8. Habenula: codes negative reward, suppresses movement (also in sleep) | - | ? | + | + | + | + | + | + | + |
| 9. Neocortical parts (insula, anterior cingulate, orbital and medial prefrontal): cognitive modulation of affect | - | - | - | - | - | - | +? (insula?) | + (insula?) | + |
Key: + present; ? uncertain; - absent or unreported.
- 13 of 18: affective neurostructures in lampreys.
- 14: cartilaginous fish. 16: bony fish. 17: amphibians. 18 of 18: amniotes.
- Only trait 9, the neocortical modulators, is confined to mammals, birds and possibly reptiles.
- 5 keystone "hedonic" structures (Berridge and Kringelbach's nucleus accumbens, ventral pallidum, amygdala, plus habenula and VTA): tetrapods have all five; lampreys and sharks have two for certain (VTA equivalent, habenula) and two maybes.
The structures are mostly absent in amphioxus and mostly present in every vertebrate, because they are limbic, and every vertebrate has a well-developed limbic system in the brainstem and basal forebrain. Very few lie in the expanded dorsal pallium or cortex, which fits Solms, Panksepp, Damasio and Denton: core affects are generated subcortically, and the cortex only modulates them. The counts also show progressive elaboration, with the affective system richest in the amniotes.
The adaptive behaviour network
Affects drive survival behaviour through what is usually called the "social behavior network" (SBN). Because it runs all essential behaviours (reproduction, aggression toward rivals and predators, escape, feeding), the authors rename it the adaptive behavior network (ABN). The MRS evaluates each stimulus as positive or negative and signals the ABN to act; the MRS then reinforces the act so the animal seeks more. ABN neurons carry sex-hormone receptors, which makes them easy to label in brain slices.
| Table 8.6: adaptive behaviour network (mammalian function) | Protostomes | Amphioxus | Lampreys | Sharks, rays | Bony fish | Amphibians | Reptiles | Birds | Mammals |
|---|---|---|---|---|---|---|---|---|---|
| 1. Preoptic area of hypothalamus plus paraventricular nucleus and vasotocin: sexual behaviour | ? | ? | + | ? (shark, skate) | + | + | + | + | + |
| 2. Anterior hypothalamus: aggression, sex | - | ? | + | ? | + | + | + | + | + |
| 3. Ventromedial hypothalamus: sex, energy metabolism, satiation, threat response | - | ? | + | - | + | + | + | + | + |
| 4. Periaqueductal gray: panic, reproduction, aggression, calling, pain relief | - | - | + | + | + | + | + | + | + |
| 5. Subpallial amygdala: emotion-directed responses and actions | - | - | ? | ? | + | + | + | + | + |
| 6. Lateral septum: social behaviour, territoriality, novelty, mood, stress-coping anxiety | - | - | + | + | + | + | + | + | + |
Lampreys have at least five of the six ABN traits present in mammals and all bony vertebrates, further (less direct) evidence that affect began with the first vertebrates.
The three questions left open in chapter 7 now have answers. Where: core affects are subcortical, not cortical. When: in the first vertebrates, 560 to 520 mya. Relative to exteroception: in the same interval. The lamprey brain and limbic system is the minimum level for affective consciousness among living vertebrates.
Interoception and pain
General interoception
Interoceptive pathways are similar in all vertebrates and project similarly to the limbic brain (chapter 7), and all vertebrate sensory neurons, interoceptive ones included, come from neural crest and placodes. So the affective side of interoception probably appeared at the dawn of vertebrates, though the case is not airtight given how little is known of nonmammalian pathways. Whether the mapped (somatotopic) side began then too, perhaps in the tectum, is unclear: somatotopic interoceptive input to the tectum is reported only in mammals, and nociceptive input to the tectum in sharks, teleosts and lampreys, all sparse.
Do fish feel pain?
The surprise is that agonising pain appears to be the one interoceptive affect not shared by all vertebrates, a view pressed by James D. Rose and colleagues.
- All jawed vertebrates have Aδ fibres, so all could feel fast, sharp pain.
- 0%: C fibres in cartilaginous fish.
- 4 to 5%: C fibres among nerve axons in teleosts, "far below" mammals or even amphibians.
- 25%: C-fibre proportion in rare humans so insensitive to pain that they injure themselves unawares.
The trout lip test and its critics. Lynne Sneddon, Victoria Braithwaite and Michael Gentle (2003) injected an irritant into trout lips and saw the fish rock and rub their lips on the aquarium floor. Rose and colleagues argued such responses could be artifacts of anaesthesia or other confounds, pointed out that fish often recover quickly from surgery, injury or being hooked by anglers, and showed that many "fish pain" studies demonstrated only nociception.
Rose's evolutionary explanation is that suffering pain is situationally adaptive. It tells a wounded animal to hide, rest and heal, which pays for animals that can hide: the first mammals and birds under cover, and salamanders, frogs and small lizards out of sight. Animals that must forage continually in the open gain nothing from prolonged pain, and pain behaviours attract predators. So fish can lack agonising pain and still have affects from other interoceptive senses. Suffering pain would have arrived last, with the first tetrapods about 360 mya. The authors, like others, had assumed agonising pain was the most basic negative affect; if it is the newest, then studying pain to find the primal affect may be misguided.
Genes and affect
Affect, like exteroception, reaches down to molecules. Genes are known for the neuromodulators and neurotransmitters of limbic and autonomic circuits: dopamine, serotonin, noradrenaline, acetylcholine, nitric oxide, glutamate, GABA and glycine, plus oxytocin and vasotocin, hormones that promote feelings of "liking." Damasio: affective value is expressed in the release of chemical molecules related to reward and punishment.
Conclusion: early, and purposive
As the first vertebrates built their first mental maps, they were also giving their sensations affect, value and full sentience, assigned in the limbic brain and at every level of the sensory hierarchies. Following Cabanac, affect told the animal what to approach or avoid, the ABN turned that into action, and affect also marked what to remember and learn.
- Purposive action. Following Wolfram Schultz, if the first vertebrates felt pleasure and desire, they engaged in consciously directed reward-seeking, which goes a long way toward the problem of mental causation (how the subjective acts on the world).
- Affect may be older still. Early animals with crude senses might already have felt the urge to forage for food, safety and mates, an "affective base" on which better senses were built (Panksepp, Solms, Denton). That opens the door to affective consciousness in some protostomes, the topic of chapter 9.
- Summary. Consciousness is earlier, more widespread and more architecturally diverse than neuroscience assumed: exteroceptive in the tectum of fish and amphibians as well as mammalian cortex, affective in the subcortical limbic structures of all vertebrates.
The decerebrate audit. Suppose a robot arm withdraws from heat, guards its "injured" joint, emits alarm tones and even learns, Pavlov-style, to flinch at a warning light. By Table 8.1 none of that indicates feeling. Now suppose it gives up a reward to avoid heat when the heat is mild but not when it is mild and the reward is large, sulks when an expected reward stops, and pays to dispense its own coolant. Table 8.2 says those are the signs to look for. The thought experiment shows both the strength of the criteria (they target flexible valuation, not reaction) and their limit: an engineered system could be built to pass them without anyone thinking it feels.
The criteria are carefully chosen but carry known problems. Several "absences" in Table 8.3 are absences of evidence (the authors admit this, noting lampreys and sharks are barely tested). Operant learning is used as a marker of feeling, yet invertebrates such as Aplysia and Drosophila show operant learning in tiny circuits, and isolated spinal cords show a rudimentary form. On fish pain the authors lean on Rose and colleagues, a contested camp: Sneddon showed in 2003 that morphine abolished the trout's anomalous lip-rubbing, which is hard to explain as an anaesthesia artifact. The C-fibre argument assumes that suffering requires C fibres, while some argue teleost Aδ fibres could carry slow pain. And the "fish recover quickly" observation fits suppression of pain behaviour under predation pressure as well as absence of pain. A small slip: the book lists "nitrous oxide" among the affect-related signalling molecules; the neural messenger is nitric oxide.
- Fish pain debate. It intensified after 2016: Brian Key's "Why fish do not feel pain" (Animal Sentience, 2016) drew dozens of commentaries, mostly critical, and Sneddon and colleagues continued to publish behavioural and analgesic evidence. The 2024 New York Declaration on Animal Consciousness judged conscious experience at least a "realistic possibility" in fishes, and the UK's Animal Welfare (Sentience) Act 2022 names all vertebrates, fish included, as sentient.
- Pain redefined. In 2020 the IASP revised its definition to an unpleasant sensory and emotional experience "associated with, or resembling that associated with" actual or potential tissue damage, and dropped the requirement that pain be describable in words, a change made partly with non-verbal humans and animals in mind.
- Fish cognition. Kohda and colleagues (PNAS, 2019, with follow-ups) reported that cleaner wrasse pass a version of the mirror mark test, a contested result that nonetheless widened views of fish minds.
- Invertebrate trade-offs. The criteria the authors used have since been applied to bumblebees, which make motivational trade-offs between sugar reward and noxious heat (Gibbons, Chittka and colleagues, PNAS 2022); see chapter 9.
Tables 8.1 and 8.2 are a model of evaluation design. The authors first list the behaviours a non-feeling system can already produce and exclude them, then accept only behaviours that require memory of value, trade-offs and seeking. Anyone evaluating whether an AI agent "really" plans or "really" understands should do the same: start by listing what a pattern-completing baseline can already fake, and score only the behaviours that survive.
Does Consciousness Need a Backbone?
Applying the book's criteria to the protostome invertebrates, the authors find that nematodes and flatworms show little sign of consciousness and gastropods only partial signs (more affective than exteroceptive). Arthropods meet nearly every criterion for both affective and exteroceptive consciousness, including fifth-level retinotopy in the fruit fly, and are judged "probably conscious," the main doubt being their tiny brains. Cephalopods have big brains and camera eyes but no demonstrated nonvisual mapping, so they are "possibly conscious." If so, consciousness arose independently at least three times.
The protostomes
Most protostomes have a central nervous system of two longitudinal ventral nerve cords with ganglia (expanded masses of neurons) at intervals, unlike the dorsal cord and brain of vertebrates. Chapter 4 argued that the first bilaterian had a diffuse nervous system and no brain and that most invertebrate clades never got much further, but that was not tested rigorously, and some scientists and animal-rights advocates claim that every animal with a CNS is conscious. Several protostome groups also have complex brains and may have evolved consciousness independently.
- Arthropods. (1) Insects and the related crustaceans (lobsters, crabs, shrimps, barnacles, krill); (2) myriapods (centipedes, millipedes); (3) chelicerates (spiders, horseshoe crabs, scorpions).
- Molluscs. (1) Gastropods (snails and slugs, including the much-studied sea hare Aplysia); (2) cephalopods (squid, octopus, cuttlefish, and the nautilus); (3) bivalves (clams, oysters, scallops); plus smaller groups such as chitons.
The analysis focuses on the best-studied or best-placed candidates: octopuses, squids and gastropods, insects, plus flatworms and roundworms as simpler comparisons. Affect comes first, since its behavioural tests are easier.
Affective consciousness in invertebrates
The clades are ordered by brain neuron count, keeping the two mollusc groups together even though cephalopods are far more complex than gastropods.
| Criterion (Table 9.1) | Nematodes (140 brain neurons) | Flatworms (1,000) | Gastropods (5,000 to 10,000) | Cephalopods (50 million) | Arthropods (under 100,000 to 1 million) |
|---|---|---|---|---|---|
| Operant learning from punishments or rewards | C. elegans? | - | Sea hare, pond snail | Octopus, cuttlefish | Fruit fly, bee, crayfish, lobster and crab, jumping spider |
| Behavioural trade-off | - | - | Predatory sea slug | Octopus | Crayfish, cricket, crab, jumping spider |
| Frustration / contrast | -/- | -/- | -/- | Not tested | Fruit fly? / - |
| Self-delivery of analgesics or rewards | - | - | Land snail Helix? | Not tested | Fruit fly |
| Drug approach / conditioned place preference | C. elegans? | Flatworm | - | Not tested | Fruit fly, bee, crayfish |
| Play (possible advanced marker) | - | - | - | Yes (cephalopods) | Comments on insects and crustaceans |
- Worms fail. The roundworm (about 140 brain neurons) and flatworm (about 1,000) fail most criteria.
- Gastropods are partial. Slugs and snails meet two or three.
- Arthropods pass. They meet almost all criteria, making them the best candidates for affect.
- Cephalopods are undertested. Octopuses and cuttlefish pass operant learning and trade-offs but have not been tested on the rest, though they play, a potentially enjoyable activity.
The evidence covers some arthropods and molluscs (insects, some crustaceans), not whole phyla that are huge and diverse. And nothing here shows whether any invertebrate feels suffering pain, which is hard to recognise even in vertebrates. The authors say they are more sceptical about invertebrate pain than Robert Elwood.
Exteroceptive consciousness in invertebrates
The seven criteria
Table 9.2 turns the special features of chapter 2 (Table 2.1) and the findings of chapters 5 and 6 into a working checklist of seven criteria for exteroceptive consciousness. C. elegans is left out as too small.
- Complexity. Enough neurons in the brain or nervous system (Table 2.1's complex neural hierarchies; a brain).
- Multiple sensory hierarchies with enough levels. Several distance senses, each with a chain of neuron levels before the premotor centre.
- Isomorphic organisation. Topographic mapping in the sensory pathways (Table 2.1's isomorphic representations and mental images).
- Reciprocal interactions. Two-way signalling within and between levels (Table 2.1's unique neural-neural interactions).
- Multisensory convergence. A site where separate senses merge into one image, needed for unity (the nested, non-nested binding of Table 2.1).
- A memory region. Localised brain regions for memory (from chapter 6).
- Selective attention mechanisms. (Table 2.1's attention.)
Recalibrating the level count. In vertebrates the authors knew where consciousness arises (cortex or tectum) and counted levels up to and including that region: four for most senses, perhaps three for fish somatosensation. In invertebrates the conscious regions are unknown, so they count levels before the premotor centres, where the sensory hierarchy ends. Since the vertebrate cortex and tectum also project to premotor centres (basal ganglia and motor reticular formation), the counts are comparable, except that extra levels lie inside cortex and tectum; hence vertebrate values carry a ">" sign. Rule of thumb: three, and especially four, levels before premotor suggest consciousness.
| Criterion (Table 9.2) | Arthropods (mostly insects) | Gastropods | Cephalopods | Vertebrates (comparison) |
|---|---|---|---|---|
| 1. Complexity (B brain, C CNS, N whole nervous system) | No? Crab 100,000 (C); lobster 100,000 (C); fruit fly 100,000 (B); ant 250,000 (B); bee 960,000 (B); roach 1 million (B) | No. Sea hare 20,000 (C), about 1,000 in cerebral ganglia; gastropods 5,000 to 10,000 (B) | Yes. Octopus 300 million (N), 50 million (B) | Yes. Zebrafish 10 million (N); frog 16 million (N), 10 million (B); rat 200 million (N); human 85 billion (N) |
| 2. Sensory hierarchies, levels before premotor | Yes. Visual 5n; olfactory 2 or 4n; taste 2 or 4n; hearing >2n; balance >2n; touch >2n to 4n; wind sense (lateral-line-like) >2n | Few, short. Chemosensory 3n; mechanosensory 1 or 2n; photoreception ?n | Yes. Visual 3 or 4n+; touch >3n; olfactory, taste, lateral line ?n; balance >1n | Yes. Mammals >4n to >5n per sense; fish >3n (touch, taste) to >4n |
| 3. Isomorphic organisation | Yes. Visual to the 5th level; olfactory, taste, hearing, touch to >2nd level | Yes. Mechanosensory to >2nd or >3rd level | Visual yes to 3rd or 4th level; mechanosensory no?; others unknown | Yes, all exteroceptive senses |
| 4. Reciprocal interactions | Yes (few?) | Yes, in simple circuits | Yes (unstudied deep in brain) | Yes, extensive |
| 5. Multisensory convergence | Yes. Mushroom bodies (bee); other protocerebrum (all insects) | Yes? Cerebral ganglia | Yes. Vertical and frontal lobes (visual, tactile, more); peduncle lobe (olfactory and visual) | Yes |
| 6. Memory region | Yes. Mushroom bodies | Yes? Cerebral and buccal ganglia (simple learning) | Yes. Vertical (and frontal) lobes | Yes. Hippocampus, dorsal pallium |
| 7. Selective attention | Yes. To visual targets and colours; oscillatory binding in mushroom bodies; medial protocerebrum and optic lobes | Yes. To salient stimuli; arousal of feeding and swimming in cerebral and buccal ganglia | Yes, behavioural evidence only | Yes |
Arthropods
The insect head ganglia form a three-part brain: protocerebrum, deutocerebrum and tritocerebrum. Insects behave flexibly and learn. Going criterion by criterion:
- 1. Complexity: the weak point. Insect brains hold under 100,000 to 1 million neurons against vertebrate millions, and even allowing for body size many are smaller than any vertebrate's. On a brain-body plot, three of the five arthropod points fall below the vertebrate range even when the fish line is extended down to insect size. Lobsters and crabs have under 100,000; spiders also have tiny brains.
- 2. Hierarchies: firm yes. Insects have high-resolution vision through compound image-forming eyes (the key marker in the vision-first view), plus smell, taste, hearing, balance, many touch senses and a wind sense comparable to the lateral line. Some pathways reach four or five levels before the premotor lateral horn, though some smell and taste paths reach premotor regions in two.
- 3. Isomorphism: yes, point for point. Retinotopy, tonotopy, somatotopy and odour mapping ("odortypy"). Johannes Seelig and Vivek Jayaraman showed retinotopy persists to at least the fifth level of the fruit fly visual hierarchy, into the central complex (higher multisensory processing, behavioural choice; it also works as an internal compass toward landmarks). In mammals, fifth-order processing lies well up in conscious cortex; in fish, visual consciousness emerges at the fourth level in the tectum. Retinotopy is documented in insects, crustaceans, myriapods and chelicerates.
- 4. Reciprocal interactions: yes. Lars Chittka and Jeremy Niven worried that few neurons allow few reciprocal connections, but nobody has counted them.
- 5. Convergence: yes. Chittka and Niven found no multisensory "gathering place," but convergence sites are turning up in the protocerebrum, and the large mushroom bodies of bees have long been known as one.
- 6. Memory: yes, the mushroom bodies.
- 7. Attention: yes. Fruit flies orient in flight to a trained visual cue; bees distinguish similar colours by "paying attention"; attention regions lie in the protocerebrum and use gamma-frequency oscillations, as in vertebrates.
The small-brain worry. Insect brains are "a fraction of the size of the head of a pin," some smaller than a few vertebrate neurons. One camp argues each miniaturised insect neuron is a supercomputer whose dendrites and axons divide into segments with their own inputs and outputs, so one insect neuron does the work of dozens of vertebrate neurons (Nicholas Strausfeld). The other (Chittka and Niven) finds insect neurons ordinary in transmitters, signalling and networking. The problem stands.
Bees that carry a picture. Since what finally matters is forming mental images, the authors look for a behavioural proof. Karine Fauria and colleagues trained bumblebees to fly down a long box to a feeding hole. The bees had to learn which striped pattern on a gate led toward food and which bull's-eye pattern around the hole marked the food. The gate blocked the view of the target, so the bees could not see both at once, yet they learned the correct gate-plus-target combination. To do so they must have formed a visual image of the gate pattern, held it in memory and matched it to the target. The authors' verdict: a visual image, implying consciousness, at least in bees.
From insects to all arthropods. All arthropod groups share the same distance senses and brain regions, as Nicholas Strausfeld documents in Arthropod Brains. Cambrian fossils (an anomalocarid, and Fuxianhuia) preserve brains with the same parts, especially the compound eyes and visual neuropils, already full-blown, not simple. So if living arthropods are conscious, the first arthropods probably were too.
Why did arthropods not go further? They moult their cuticle to grow, leaving them soft, unsupported and vulnerable for a time; giant, helpless moult stages could not hide and would be killed, which caps body size and so brain size.
Gastropods
Snails and slugs creep (some swim) and keep the paired ventral cords with little brain enlargement. Their behaviour is simple enough to serve as a model for decisions about approaching, ingesting or expelling food, a pond snail leaving oxygen-poor water to breathe air, and escape swimming.
- Against consciousness. Under 10,000 brain neurons (cerebral ganglia sometimes under 1,000); only mechano-, chemo- and photoreception, with small eyes mostly for light detection, not vision, the senses of the early unconscious bilaterians; sensory chains of only 1 to 3 neurons.
- For. Clean somatotopy of touch neurons (first level only; higher levels unstudied); chemosensory-mechanosensory convergence in the brain ganglia; learning tied to the cerebral and buccal ganglia; attention and arousal neurons in feeding and swimming circuits.
Verdict: probably no exteroceptive consciousness. Yet the criteria they do meet (conditioned memory, attention, arousal) are those linked to their operant learning from reward and punishment, hinting that affective consciousness can begin to evolve without, or before, exteroceptive consciousness.
Cephalopods
The coleoids (octopuses, squids, cuttlefish) show complex learning, flexibility, cognition and foresight, including tool use and trickery; the nautilus is a distant relative. Their brain circuitry is poorly known, a major handicap.
- 1. Complexity: yes. Adjusted for body size, cuttlefish and octopus brains fall in the reptile, bird and mammal range. Octopuses have over 300 million neurons, two-thirds in the arms, about 50 million in the brain: more than the entire nervous system of many fish and all amphibians.
- 2. Hierarchies: yes. Camera eyes that form sharp images, convergent with vertebrate eyes; an olfactory organ, balance organ, taste receptors and mechanoreceptors in the suckers, and a lateral line that may also "hear." Visual paths reach third or fourth level and beyond; touch probably three or four.
The isomorphism problem. Somatotopy has been sought in the octopus brain and never found. The usual explanation is the arms: eight flexible arms moving in every direction are too hard to control centrally, so each has its own nerve cord, reflexes, processing centres, movement patterns and behavioural programs. Critics conclude that touch is processed in the arms, so the octopus has no unified central representation and, lacking unity, no consciousness. The authors reply four ways:
- Wrong place. The failed searches recorded from motor regions; sensory regions need recording (only one study looked elsewhere).
- Untested senses. Smell, balance and lateral-line paths have never been checked for topography.
- Retinotopy exists. The visual path is retinotopic up to the third or fourth level, the highest examined.
- Arms are not that independent. They send sensory information to the brain, which coordinates them. Tamar Gutnick and colleagues showed octopuses using vision to guide an arm through a transparent-walled maze to visible food; since vision is processed only in the brain, the brain must hold a unified touch-plus-vision picture.
The lack of positive evidence for nonvisual mapping remains as big a hurdle for cephalopods as small brains are for arthropods. Cephalopods pass the other criteria: reciprocal cross-talk in the visual and balance paths; convergence in the vertical, frontal and peduncle lobes, with the vertical lobe the best candidate seat of unity and, from lesion studies, of sensory memory; and famous powers of focused attention, though its brain sites are unknown.
When? Nautilus, living deep and dim with lensless pinhole eyes, probably lacks visual consciousness, but it is unclear whether that is ancestral or a secondary loss. The first coleoid (common ancestor of squid, cuttlefish, octopus) had good eyes but lived about 276 mya, versus about 490 mya for the first cephalopods. So any cephalopod consciousness evolved between 490 and 276 mya.
Why not dominant? Cephalopods combine the streamlined fast swimming of fish with the manipulating limbs of arthropods, yet they hide, camouflage and are eaten by fish and whales. Candidate reasons: fish swim better (cephalopods jet backwards); fish got manipulators of their own (jaws); fish have scales and bone while early cephalopods had heavy shells and coleoids none; soft arms are tempting meat; and short life spans limit lifetime learning. The real reasons can only be guessed.
Conclusions
| Group | Affective consciousness | Exteroceptive consciousness | Verdict | Main obstacle |
|---|---|---|---|---|
| Nematodes, flatworms | Little evidence | Little evidence | Not conscious | Tiny, simple nervous systems |
| Gastropods | Partial (2 or 3 criteria) | Partial | Probably not | Few neurons and senses, 1 to 3 level paths |
| Most other invertebrate clades (D or G) | Not tested or absent | At or below gastropod level | Probably not | Diffuse nets or simple ganglia |
| Arthropods | Almost every criterion | Most criteria, multilevel retinotopy | Probably conscious, since the Cambrian | Small brains, few neurons |
| Cephalopods | The few criteria tested | All but nonvisual isomorphism | Possibly conscious, from 490 to 276 mya | No nonvisual maps found; sparse data |
If arthropods and cephalopods prove conscious, vertebrates and arthropods evolved it in the Early Cambrian arms race, cephalopods later. In all three, consciousness gave two advantages: an exteroceptive mental reconstruction of the world to guide behaviour, and affective feelings that mark which actions to seek or avoid on the basis of learned value. All three share varied distance senses, good cognition, high mobility and active lives.
Still, consciousness served vertebrates better: they include more of Earth's largest animals and brains, and no invertebrate is shown to have reached the memory-enhanced consciousness of birds and mammals, self-consciousness or theory of mind. Non-mental head starts helped: an excellent swimming mechanism, the strongest jaws, and an internal skeleton that spared vertebrates from moulting and allowed large size. The authors call their stance conservative: arthropods and cephalopods get only basic phenomenal consciousness, with reservations, and either would confirm that consciousness is more widespread, more diverse in substrate and evolved independently several times.
The minimal mind budget. Suppose an engineer must build an animal that sees, remembers and chooses, with a fixed budget of 100,000 neurons. A vertebrate solution spends millions on separate maps and redundancy; an insect solution packs five mapped visual levels, a compass, a memory centre and attention into a pinhead. If the bee in Fauria's box really holds a picture of the gate, then consciousness has no fixed neuron threshold, only an architectural one. If it does not, the same behaviour shows how much image-like competence can be had without experience. The experiment cannot tell these apart, which is exactly the problem of other minds in miniature.
Several inferences are looser than the tidy tables suggest. The Fauria bee experiment shows that bees hold a visual memory across an occlusion; a conditional association held in working memory need not be a conscious image, so "image, therefore consciousness" is the authors' interpretation. The step from insects to all arthropods rests on shared anatomy, not on tests of crabs or millipedes. The moulting explanation of small arthropod size is one of several (insect size is also linked to tracheal oxygen supply). The octopus neuron figures are on the low side: most estimates give about 500 million neurons in total, with the optic lobes alone holding well over 100 million, so "50 million in the brain" counts only the central brain. And the authors' level-counting criterion was ported from vertebrates with an admitted fudge (the ">" sign), so pass marks are approximate.
- Connectomes. The FlyWire consortium published the first whole-brain connectome of an adult fruit fly (2024, Nature): about 140,000 neurons and tens of millions of synapses. Chittka and Niven's uncounted reciprocal connections can now be counted, and recurrent loops are abundant.
- Bees. Solvi, Al-Khudhairy and Chittka (Science, 2020) showed bumblebees recognise by touch objects they had only seen, and vice versa, evidence of a cross-modal, unified object representation (Criterion 5). Bumblebees also make motivational trade-offs against noxious heat (2022) and appear to play by rolling balls without reward (Dona, Chittka and colleagues, 2022). Earlier, Perry, Baciadonna and Chittka (Science, 2016) found that an unexpected sugar reward induced a dopamine-dependent, positive emotion-like state in bumblebees. Chittka's The Mind of a Bee (2022) summarises the case.
- Octopus affect. Robyn Crook (iScience, 2021) reported that octopuses given a noxious injection avoid the place where it happened, attend to the wound, and prefer a place where they received a local anaesthetic, filling the "not tested" cells of Table 9.1 for cephalopods.
- Policy. The 2021 LSE review led by Jonathan Birch found strong evidence of sentience in decapod crustaceans and cephalopod molluscs, and the UK's Animal Welfare (Sentience) Act 2022 now covers both. The 2024 New York Declaration calls conscious experience a realistic possibility in cephalopods, decapods and insects.
- Cephalopod dates. Fossil coleoid relatives are older than the 276 mya crown estimate used here; a ten-armed vampyropod, Syllipsimopodi, was described in 2022 from Carboniferous rocks about 328 million years old.
Tables 9.1 and 9.2 are a capability matrix with a declared pass rule, the same structure an architect uses for vendor or model selection: explicit criteria, evidence per cell, "not tested" kept distinct from "fails," and a graded verdict (probably / possibly / not). The octopus case is the lesson in reading such matrices: a missing cell (no somatotopy found) is only damning if someone looked in the right place.
Neurobiological Naturalism: A Consilience
The book's findings are condensed into three postulates. Sensory consciousness is explained by known neurobiology (a living, neural, complexly hierarchical brain); it is ancient, widespread and diverse (all vertebrates, probably arthropods, possibly cephalopods, with tectal, pallial and limbic substrates); and its philosophical puzzles (subjectivity, irreducibility, the hard problem) dissolve into a natural evolutionary event once neurobiology, evolution and philosophy are combined. The authors' verdict: the explanatory gaps are closed, while the ontological gap between first and third person is bridged but not closed.
- Postulate 1. Sensory consciousness can be explained by known neurobiological principles.
- Postulate 2. Sensory consciousness is ancient and widespread in the animal kingdom, and diverse neural architectures can create it.
- Postulate 3. The philosophical issues of ontological subjectivity, neuroontological irreducibility and the "hard problem" can be explained by the nondissociable confluence of neurobiological and adaptive neuroevolutionary events.
Postulate 1: known neurobiology suffices
Consciousness obeys accepted scientific laws; no new fundamental, quantum-level or mysterious property is needed. Condensing the general and special features of Table 2.1 into one phrase: sensory consciousness is an emergent characteristic of a living, neural and complexly hierarchical brain. Each word is necessary and none is sufficient alone.
- Living. Life and consciousness are both embodied and ontologically specific to one individual, so the seeds of subjectivity go back to the origin of life. Evan Thompson argues the hard problem is an artifact of the Cartesian mental-versus-physical frame; the real task is to understand living subjectivity emerging from living being, which already has an interiority the objectivist picture misses. But a bacterium is alive, embodied and not conscious.
- Emergent. No single retinal cell sees and no auditory neuron hears; experience needs many interacting neurons in particular patterns, plus input from memory and reticular activating systems. But digestion and circulation are emergent too. An ant colony builds an anthill no ant intends, yet the colony is not conscious even if ants are, because its nervous systems are not woven into one structural and functional unity: a colony cannot be "in a state of red." Emergence is necessary, not sufficient.
- Neural. Neurons, which began as short reflex arcs, uniquely allow tight, rapid processing of many stimulus types and can be added into longer chains. Reflexes are the "royal road" but not the destination.
- Complexly hierarchical. The special features supply the missing ingredient: a unified brain with elaborate hierarchies producing unique neural-neural interactions, multiple sensory modalities, reciprocal connections and oscillatory binding for unity, isomorphic maps for the distance senses, plus attention, arousal and memory. The contrast between the short sensory chains of amphioxus and the long ones of every vertebrate shows it.
The one-sentence mechanism. Consciousness stems from unique neural interactions within and between discrete chains of neurons, for many specialised sensory modalities, each chain usually of four or more levels, arranged as hierarchical, modality-specific pathways that merge into a neural map simulating the world or feed affective circuits, ultimately serving sustained processing and behaviour, with attention, arousal and memory added.
Postulate 2: ancient, widespread and diverse
Ancient
- About 585 mya: first bilaterian animals; not conscious (their living proxies, roundworms and flatworms, show little affect or exteroception).
- 560 to 520 mya: prevertebrates become the first vertebrates during the Cambrian diversification.
- Before 520 mya: early vertebrates conscious, with distance senses and affects; arthropod consciousness probable.
- About 490 mya: oldest cephalopods (consciousness possible from here).
- About 460 mya: first jawed vertebrates; larger brains, probably more cognition, but no qualitative advance from 520 to 220 mya.
- About 330 mya: age of the coal forests; conscious vertebrates in sea and on land.
- About 276 mya: first coleoid cephalopods (latest date for any cephalopod consciousness).
- About 220 mya: first mammals; memory-enhanced consciousness (Step 2).
- About 170 to 165 mya: bird lineage makes the same advance.
- About 200,000 years ago: first modern humans; self-consciousness, language, theory of mind.
A series of reconstructed scenes marks which animals were conscious at each period: the Cambrian sea (520 to 505 mya), the Carboniferous sea and land (330 mya) and the Triassic sea and land (220 mya).
- Carboniferous seas. Ray-finned fish, sharks, a lamprey (Mayomyzon), a hagfish (Myxinikela) and a coelacanth (conscious); a mantis shrimp and a trilobite (probably); an ammonite (possibly); corals, sea lilies, bryozoans, clams, brachiopods and snails (not).
- Carboniferous land. Temnospondyl amphibians, an early amniote (Paleothyris), an early synapsid (Eothyris), a freshwater shark; probably conscious giant millipede Arthropleura, a spider relative and a sea scorpion.
- Triassic seas. Ichthyosaurs, sharks (Hybodus), ray-finned and teleost fish; possibly conscious ammonites; unconscious snails, clams, corals and sea stars.
- Triassic land. Pterosaurs, early dinosaurs (Plateosaurus, Coelophysis), a late mammal-like reptile (Placerias), an early turtle, and tiny mammals (Hadrocodium) with nest and eggs, insignificant-looking but possessing the more advanced consciousness; beetles and bee relatives probably conscious.
Few or no arthropods made the Step 2 advance, though bees, wasps and termites have enlarged mushroom bodies. Affective consciousness may also have deepened in mammals and birds, the only animals showing behavioural negative contrast ("disappointment"), and Panksepp's seven emotions may have been completed then or in the first amniotes.
Widespread
The criteria (Tables 2.1, 8.3 to 8.6, 9.1, 9.2) show consciousness is far more widespread than most thought. Tables 9.1 and 9.2 extend the original special features of Table 2.1, adding multisensory convergence, a memory region and the behavioural criteria for affect. Few investigators have held that all vertebrates are conscious in all three ways: exteroceptive, affective and probably interoceptive (though without suffering pain). Amphioxus and most invertebrates are not.
If arthropods and cephalopods are conscious, consciousness evolved three times by convergence, once in each great bilaterian branch: Deuterostomia (vertebrates), Ecdysozoa (arthropods) and Lophotrochozoa (cephalopods). The period scenes also show:
- Spread over time. Conscious vertebrates became more abundant and varied after the Cambrian and invaded more habitats; vertebrates seem to have gained more from consciousness, perhaps through better locomotion.
- The unconscious majority. Many common invertebrates never evolved it: mostly immobile, armoured or shelly types (clams, corals, sea lilies, bryozoans).
- Land over sea. A higher proportion of land animals than sea animals have always been conscious.
- Today. On a southern African grassland, mammals and birds fill the old reptile niches (birds for pterosaurs, giraffes for long-necked dinosaurs, a hippo for the hairy reptile, a real ostrich and crocodile, rodents and shrews for the first mammals), turtles and insects persist, and humans (the San people) join them. In the sea, nonconscious molluscs, corals and echinoderms remain abundant, teleosts dominate, and dolphins and whales replace ichthyosaurs (dolphins perhaps with ape-like self-awareness). Consciousness in the ocean has held its own or expanded since the Triassic.
Diverse: tectum versus cortex
The strongest diversity claim is that fish and amphibians have tectal consciousness for all nonolfactory senses, whereas mammals and birds have pallial (corticothalamic) consciousness. Most researchers reject this. The authors answer four challenges.
| Challenge from corticocentric theorists | The authors' reply |
|---|---|
| 1. Cortical blindness. Humans with destroyed visual cortex lose visual awareness, so the tectum cannot see consciously (residual blindsight may run via the superior colliculus). | That shows what human brains need, not what lampreys need. Lampreys have image-forming eyes, excellent vision and mapped multisensory tecta; the fish themselves (as in Redouan Bshary's studies of subtle social signalling) show the pallium is not necessary. |
| 2. Nonconscious perception. Primates perceive without awareness in blindsight, masking, attentional blink, binocular rivalry, implicit cognition, inattentional preconsciousness, mismatch negativity, response inhibition, nonconscious error detection and conflict resolution; fish might run on such processing. | These lab phenomena are so weak and incomplete they amount to inattention, cannot be dissociated from (or distinguished from degraded) consciousness, and a fish or human relying on them would be "senseless" and could not survive. Fish make fine-grained discriminations. |
| 3. Unconscious hierarchies. If any isomorphic hierarchy is unconscious, hierarchy cannot be the key. | The authors never claim all hierarchies are conscious. Proprioception has an isomorphic but unconscious cerebellar path and a conscious cortical one; conscious circuitry is better spent on the dangerous, unpredictable outside world, while inner states are held steady by automatic homeostasis. Tononi and Koch add that cerebellar modules interact too little to integrate. |
| 4. Cortex makes affect. | Refuted in chapters 7 and 8: basic affects are subcortical, in structures all vertebrates share, and all vertebrates pass the behavioural criteria with or without cortex. |
The practical lesson: a private mental image is hard to detect, but it is easy to see that even a fish pays attention and shows valenced preferences. Subcortical affect and selective attention are themselves indicators of primary consciousness.
Diversity within one brain
- Smell versus the rest. In every vertebrate, every conscious sense relays through the thalamus except olfaction (its minor transthalamic branch is not needed). Olfactory information is carried by gamma oscillations much like, but generated independently of, the thalamocortical gamma that binds vision, hearing, balance and touch; in mammals the two systems meet only over large areas of prefrontal cortex, as if they evolved separately.
- Three types. Exteroceptive consciousness comes from the dorsal pallium or tectum and is highly mapped; affective consciousness comes from the subcortical limbic core with little mapping; interoceptive consciousness shares features of both.
So instead of one emergent process explaining all sensory consciousness, many "emergences" contribute, each explicable in its evolutionary context.
Commonality within diversity
No single network accounts for all three types in all vertebrates. Mammals have rich cortex-limbic interconnections, but fish and amphibians lack cortex. The common threads the authors find:
- Olfactory-limbic and hippocampus-limbic connections in all vertebrates (smell and memory feeding affect). In mice, a superior colliculus to isthmus to amygdala path lets looming visual threats induce fear (Shang and colleagues, 2015); whether fish have it is unknown.
- The laterodorsal tegmental nucleus (LDN) of the reticular activating system, near the hindbrain-midbrain junction, which adjusts emotional intensity and is present in every vertebrate studied, lampreys included; since the RAS also drives exteroceptive arousal, the LDN links the two types.
- Gamma oscillations may accompany affect too (amygdala-hippocampus-prefrontal synchrony in rat fear and safety memories; beta-gamma bursts after unexpected rewards), though this needs testing beyond mammals.
- Genes, especially for dopamine (reward, appetitive surprise, motivation), shared across molluscs, arthropods and vertebrates.
- Hierarchy itself. One basic neural bioplan over 520 million years old, with top-down and bottom-up interactions, produces qualia from a pink sunset to the smell of a rose, the taste of sugar to the pain of a pinprick, the differences coming from neuron structure, neurotransmitters and connections.
A continuum to higher consciousness
The claims apply only to primary sensory consciousness, "something it is like to be." Higher forms (meta-awareness, thoughts about the self and about thoughts, mirror self-recognition, theory of mind, verbal report) came later, by naturalistic increases in brain complexity, with no mysterious ingredient added.
| Level | Damasio's selves | Tulving's consciousness (memory type) | Feinberg and Mallatt |
|---|---|---|---|
| Lowest | Proto-self: registers the organism's current state and perceptions; unconscious | Anoetic (nonknowing): registers, represents and responds to the present; procedural memory | Primary consciousness (the authors would count Damasio's proto-self perceptions here) |
| Middle | Core self: conscious, second-order awareness of the proto-self | Noetic (knowing): semantic memory, awareness of absent objects and events; many mammals and birds | Step 2, memory-enhanced consciousness of mammals and birds |
| Highest | Autobiographical self: autobiographical memory and anticipation; humans, great apes, dolphins | Autonoetic (self-knowing): episodic memory, mental time travel, identity; humans only | Humans, and perhaps great apes, dolphins, grey parrots, elephants; episodic-like memory exists in many amniotes and some fish |
Postulate 3: the hard problem as an evolutionary event
The irreducible, subjective aspects of consciousness came into being at a branch point over 520 million years ago, when reflexes became experiences that are unified, referred, qualitative and causally effective. Using evolution this way first requires showing consciousness is a real adaptation.
Consciousness and its subjective features are adaptive
- Unity. Choosing behaviour from a disjointed world map would be wasteful; failing to unify information about a stalking predator, or having disunited feelings, would mean hesitation and death.
- Referral. Survival depends on the world and body, not the irrelevant workings of one's own neurons.
- Mental causation. Lets an animal act proactively rather than only reactively.
- Qualia. Discriminate stimuli: blue from red (a guppy telling a blue heron from a red herring), a threat call from a mating call, fresh from rotten food. As simulations they support short-range prediction. Affective qualia say what to approach or avoid.
Against the view that consciousness is an epiphenomenon riding on cognition, learning or brain size:
- Nichols and Grantham: consciousness relies on too many complex, expensive structures to be useless.
- Mark Bradley: specific qualia are too tightly correlated with specific stimuli (a predator), brain activity (the amygdala) and responses (escape) to be idle by-products. Correlation is not causation, but if consciousness were idle, selection would relax and the pairings would drift: someone would react to danger with pleasure or flee food in terror. That does not happen.
- No healthy wild individual lacks attention to salient stimuli, preference for helpful over harmful stimuli, or isomorphic sensory mapping.
- Reality-tracking. Selection keeps images and affects matched to the world: the mad, the hallucinating and the inappropriately emotional quickly die in nature.
Consciousness is not universally beneficial: it is energetically expensive and absent from most invertebrates with cheaper strategies (shelled, immobile clams). No adult vertebrate clade has lost it; blind cave fish and cave salamanders compensate with an acute lateral line. Larval vertebrates (newly feeding zebrafish, tadpoles) may run on reflexes before consciousness develops. Among arthropods, some myriapods with reduced eyes may have lost visual images, and the parasitic rhizocephalans (barnacle relatives inside crabs) have stages with no nervous system, ending as a giant egg sac: consciousness can be lost by parasites in protected, food-rich, unstimulating settings.
Half-conscious. If consciousness took millions of years to evolve, intermediate animals must have been partly conscious. The idea sounds as absurd as being half pregnant, yet every human goes from an unconscious fertilised egg to a conscious newborn. The authors' proposal: transitional prevertebrates were reflexive most of the time and became conscious only when arousal and attention demanded it, with images and feelings flickering on for danger and food; as life grew more active and dangerous, consciousness became more continuous, idling only in sleep or dormancy. The thought experiment shows that a gradualist history of experience is coherent if consciousness is state-dependent rather than all-or-nothing.
Auto- and allo-ontological irreducibility
To explain subjectivity the authors add two negative features. Gordon Globus set them up: "mental events" belong to the subject's perspective, "neural events" to the observer's; the subject has no direct acquaintance with his neurons, and the observer cannot have the subject's pain; the nervous system has no sense organ directed at its own structure.
- Auto-ontological irreducibility. The subject cannot experience his own neurons. Herbert Feigl's imagined "autocerebroscope", a probe letting you watch your own cortical currents while listening to music, would only make you a third-person observer of yourself; you still could not feel neurons acting. Evolution explains why: awareness evolved to refer to the body and world, where danger lies; monitoring one's own neurons would waste energy, since they are protected by homeostatic mechanisms. The sports-team manager leaves the playing to the players. (Georg Northoff calls this the brain's "autoepistemic limitation" and blames the neural code; the authors' reason is passive and evolutionary: such access never evolved because it had no value.)
- Allo-ontological irreducibility. The observer cannot access the subject's experience. From outside the brain is visible but not the experience; from inside the experience is visible but not its construction.
Reflexes can be described fully in objective terms. When growing brain complexity turned reflexes into mental images and states before 520 mya, both irreducibilities appeared at once: the brain-mind distinction was created by a natural, historical event. The same solution holds even if consciousness began with mammals and birds.
Bridging each gap: general, reflex and special features
The authors trace each neuroontologically subjective feature of consciousness (NSFC) from Table 1.1 through the general (life), reflexive and special features ("gen-refl-spec"), showing a smooth, unbroken transition in each case.
| NSFC | Roots in general and reflex features | Special features that complete it | Why it is irreducible yet natural |
|---|---|---|---|
| Referral (to the body surface, into the body, out to the world) | Living process, embodied system separating inside from outside; hierarchy; reflexes already aimed at the world | Complex neurohierarchies, neural interactions, attention; isomorphic maps simulating the objects referred to | Auto-ontological irreducibility is itself adaptive: no energy wasted referring to one's own neurons. Feinberg (2012) calls referral the most representative NSFC |
| Mental unity | A process, not a place (no single unifying site); embodied hierarchical system | Nested hierarchy that binds percepts (the falling apple) plus downward constraint (cyclopean vision); strong ties to attention | The subject experiences processes as unified; the observer sees them distributed and gappy |
| Mental causation | The earliest to start: any moving organism changes its surroundings (sponges pumping water, no nerves); embodiment (causal efficacy issues from my own brain); self-organisation with Mayr's programmed goal-directed processes | All the special features, culminating in complex hierarchies; memory, since motor programs are learned | Goal-directed action by a conscious animal; Schultz's reward-seeking first vertebrates |
| Qualia | Systems hierarchy and adaptive evolution | Multisensory isomorphic maps and complex hierarchies from receptors up; recursive interactions between levels | Serve discrimination of stimuli and valences; all qualia are experienced only by the brain, unreachable from outside |
The character of experience
The last holdout is Chalmers's question: why is seeing red like this rather than like seeing blue, or like a trumpet? Why does pain hurt? The authors answer with three explanations, each correct and none sufficient alone, like the blind men and the elephant.
- Neurobiological. Colour processing and pain processing use different neural pathways and mechanisms, so they should not feel the same.
- Neuroevolutionary. There is strong adaptive value in responding to harm differently from colour, so selection made them distinct.
- Neurophilosophical. These qualia are ontologically subjective because the subject's point of view differs from the observer's, so no objective description will ever replace them.
All three domains came into being together, in the Early Cambrian.
Conclusion: closed, bridged, not closed
Francis Crick's "Astonishing Hypothesis" holds that you, your joys, sorrows, memories and free will are "nothing but a pack of neurons." The authors answer yes and no: yes, nothing more fundamental is found in the conscious brain than in biology generally; no, "nothing but" is too reductionist, ignoring the emergent features that make consciousness the most complex natural phenomenon and create the gaps.
Explanatory gaps: closed. The list of general and special neurobiological features, plus the principles of auto- and allo-ontological irreducibility, give a seamless reconstruction of the nature and origins of consciousness. In its essentials, the authors call the case "closed."
Ontological gap: bridged, not closed. The divide between first-person experience and third-person knowledge is a real barrier: experience is embodied, personal and irreducible both ways, so no explanation can eliminate it. But since we know how the brain creates experience, the gap is naturalised and bridged.
Perhaps nobody solved the hard problem before because it requires all three perspectives, neurobiological, evolutionary and philosophical, including what happened over half a billion years ago.
The conclusion is bolder than the argument supports at several points. Bradley's correlation argument does not defeat epiphenomenalism: if experience simply accompanies the selected neural states, the qualia-stimulus pairing stays tight without experience doing any work, so tight correlation is predicted by both views. The same applies to "no healthy individual lacks attention, affect or mapping," which tests neural functions, not experience. The auto-ontological account explains why we do not perceive our neurons, not why there is something it is like at all; the consilience answer to Chalmers explains why red and pain differ, while Chalmers's question is why either feels like anything, so many philosophers would say the hard problem is relocated rather than solved. The "living" requirement is asserted rather than argued, and it rules out machine consciousness by definition. Smaller slips: in the Cambrian scene, Haikouella carries the asterisk reserved for probably conscious arthropods, although chapter 5 judged the small-eyed yunnanozoans probably not conscious; and the key to the three-types figure glosses LHab as "left" habenula, where the lateral habenula is meant.
- Testing the cortex question. The Cogitate adversarial collaboration (first results 2023, Nature 2025) found conscious content most robustly in posterior cortex and challenged key predictions of both integrated information and global workspace theory; neither cortex-centred camp emerged clearly vindicated. The tectal hypothesis remains untested by such paradigms, which are human-only.
- Animal consciousness consensus. The 2024 New York Declaration on Animal Consciousness (strong support for mammals and birds; a realistic possibility for all vertebrates, cephalopods, decapods and insects) closely tracks the authors' Postulate 2, and UK law since 2022 recognises decapods and cephalopods as sentient.
- Machine consciousness. A 2023 report by Patrick Butlin, Robert Long and colleagues derived indicator properties from leading neuroscientific theories (recurrence, global workspace, higher-order, predictive processing, agency and embodiment) and concluded that no current AI system is a strong candidate for consciousness but that no obvious technical barrier prevents building systems that satisfy the indicators. Neurobiological naturalism, by making life and embodiment necessary, sits at the sceptical end of that debate.
For anyone building or governing AI, the three-postulate structure is a useful template for capability claims. Postulate 1 is a mechanism (which architecture produces the property); Postulate 2 is a distribution (which systems have it, judged by a published checklist); Postulate 3 is the interpretive layer (what the property means from inside versus outside). Claims about model "understanding" or "reasoning" usually skip straight to the third without the first two. Equally, the authors' auto/allo split is the governance problem of interpretability in miniature: an LLM has no access to its own weights, and an auditor with the weights still lacks the model's "point of view," so behavioural criteria (the Table 8.2 approach) remain indispensable.
Master Takeaways
Taken whole, the book is a dated natural history of experience joined to a checklist for detecting it. It argues that consciousness is old, common and built in more than one way, and that its philosophical strangeness is what you should expect from a living brain that evolved to model the world and the body, not itself. Its most durable contribution is methodological: decide which animals are conscious by explicit structural and behavioural criteria applied across the whole tree of life, not by extrapolating from the human cortex.
The takeaways
- Consciousness began in the Cambrian, with the first fish. Mapped sensory images and valenced feelings arose in the first vertebrates before 520 million years ago, in the predator-prey arms race of the Cambrian explosion; arthropods probably got there at the same time.
- Every vertebrate is conscious, lampreys included. Fish and amphibians build mapped multisensory images in the optic tectum and have the same subcortical affective machinery as mammals; a cerebral cortex is not required.
- Consciousness evolved in two steps in vertebrates. Step 1 (tectum plus olfactory pallium) in the first fish; Step 2, a memory-enhanced pallial consciousness, evolved convergently in mammals (220 mya) and birds (165 mya).
- There are three kinds of primary consciousness. Exteroceptive (mapped, outward-referred images), interoceptive (partly mapped body states suffused with feeling) and affective (unmapped good and bad feelings referred to the self) use different circuits.
- Feelings are subcortical and judged by operant behaviour. Core affects come from the limbic brainstem and basal forebrain; they are detected by global operant learning, trade-offs, frustration, self-medication and drug seeking, never by reflexes or classical conditioning.
- Consciousness probably evolved three times. Vertebrates (certain), arthropods (probable) and cephalopods (possible) meet most criteria independently, so consciousness is convergent and multiply realised.
- The hard problem is four problems. Referral, mental unity, qualia and mental causation each pose their own explanatory gap and each must be traced through life, reflexes and special neural features.
- Consciousness is life plus reflexes plus special features. The general features of life and the speed of reflexes are necessary but not sufficient; experience needs complex hierarchies, nested binding, isomorphic maps, reciprocal interactions, attention, arousal and memory.
- Conscious hierarchies are both pyramid and nest. Signals converge up physically separate levels while lower-level detail stays functionally nested in the unified scene, with no single "Cartesian theatre."
- Vision came first. Image-forming eyes, grown from the diencephalon, gave the first detailed maps; the lens and the other senses came from vertebrate-only neural crest and placodes, and copied the visual pathway template.
- Count the levels. Conscious sensory pathways usually have four or more neuron levels up to the top processing area (sometimes three); amphioxus, with one or two before the motor centre, falls short.
- The tectum is fast, the pallium is wise. The tectum makes quick, mapped, attention-directing assessments; the pallium is slow, memory-rich and multisensory, which is why it took over when habitats rewarded considered decisions.
- Memory changes what consciousness is like. Step 2 made experience less fleeting, enabled learned attention and laid the base for cognition and higher selves (Tulving's noetic and autonoetic levels).
- Smell is the odd sense out. It skips the thalamus, uses its own gamma oscillations, is wired next to memory and emotion, and may have evolved before or apart from the shared plan.
- Suffering came last. Fish have few or no C fibres and may lack burning, prolonged pain; suffering pain, adaptive only for animals that can hide and heal, probably arose with tetrapods (about 360 mya).
- Affect is the common currency of choice. Following Cabanac, valence ranks stimuli so an animal need not store a reflex for every case; the mesolimbic reward system values, the adaptive behaviour network acts, and reward reinforces.
- Small brains are the arthropod problem; missing maps the octopus problem. Insects show fifth-level retinotopy, attention with gamma oscillations, memory centres and image-holding bees, but have few neurons; octopuses have big brains and camera eyes but no proven nonvisual mapping.
- Consciousness is costly and can be lost. Most phyla took cheaper routes (shells, burrows, sessile life); parasites such as rhizocephalans shed their nervous systems entirely.
- Consciousness is an adaptation, not a by-product. Its features (unity, referral, causation, qualia) each pay their way; its expense and its tight fit to stimuli and responses count against epiphenomenalism.
- Subjectivity has a natural explanation. The subject cannot experience its own neurons (auto-ontological irreducibility) because such access had no survival value; observers cannot experience the subject's experience (allo-ontological irreducibility). Both appeared when reflexes became images.
- Gaps closed, gap bridged. The explanatory gaps are closed by the feature list and evolutionary history; the ontological gap between first and third person is real but naturalised: bridged, not closed.
- Consilience is the method. Why red does not hurt has three correct but individually insufficient answers, neurobiological, evolutionary and philosophical, and only their combination satisfies.
Frameworks & mental models
| Name | What it says | Use it when |
|---|---|---|
| Four NSFCs | Referral, mental unity, qualia and mental causation are separate subjective features, each with its own gap | Breaking "the hard problem" into answerable parts |
| Three levels of features (Table 2.1) | Life (4 features), reflexes (1), special neurobiological features (6) stack; each level necessary, only the top sufficient in combination | Checking whether a system has the ingredients for experience |
| Nested versus non-nested hierarchy | Physical convergence up a pyramid plus functional nesting of lower detail gives unity without a central site | Explaining unity, or designing systems that must decide and keep context |
| Bottom-up versus top-down attention | Stimulus-driven salience evolved first; goal-driven attention later and with higher consciousness | Asking what minimal attention a simple conscious animal needs |
| Reflexes as the "royal road" | Fast, connected reflex chains were the raw material; adding levels and cross-talk produced consciousness | Tracing a capability back to nonconscious precursors |
| Primitive (limbic) core | A ventral, homeostatic, neuromodulator-based brain core shared by amphioxus and all vertebrates | Separating survival machinery from sensory-mapping machinery |
| Simple bilaterian ancestor | The first bilaterian had a nerve net, no brain; brains and image eyes evolved independently in a few lines | Reasoning about convergent origins of brains and minds |
| Causes of the Cambrian explosion | Oxygen, predation, bioturbation, arms race, richer sensory world, new gene pathways | Explaining why senses and brains escalated |
| Nilsson's four eye stages | Nondirectional, directional, low-resolution, high-resolution vision | Placing any animal's vision on a scale of image-forming capacity |
| Vision first versus smell first | Light carries the most spatial information, so vision led; smell's link to memory explains its odd wiring | Weighing which sense drove the first mental maps |
| Two waves of sensory evolution (Butler) | Eyes and visual brain first (the "cephalate"); crest and placode senses copied the visual template | Explaining why all senses share one pathway plan |
| Five-stage sensory pathway | Receptor, sensory neuron, first-order multipolar, next order (thalamus or tectum), cortex; mapped at every level | Comparing sensory hierarchies across species |
| Level counting | Conscious hierarchies have about four or more levels to the top centre (or before premotor centres) | A rough, countable marker of potential consciousness |
| Isomorphic neural hierarchy theory | Topography, hierarchy and oscillatory (gamma) binding produce unified mental images | Explaining exteroceptive images in any brain |
| Three functions of the tectum | Sensory (recognition, salience), motor (orienting, approach, avoid), attentive (with isthmus nuclei) | Assessing tectal or subcortical vision |
| Two-step evolution | Step 1 tectal and olfactory consciousness; Step 2 pallial, memory-enhanced, convergent in mammals and birds | Placing a vertebrate's consciousness on the scale |
| Three types of consciousness (Table 7.1) | Exteroceptive, interoceptive, affective, differing in reference, mapping, valence and anatomy | Not treating "qualia" as one problem |
| James-Lange versus direct route | Affect arises from sensed bodily change and also directly from limbic input; both operate | Explaining how stimuli become feelings |
| Panksepp's seven emotions | SEEKING, RAGE, FEAR, LUST, CARE, GRIEF, PLAY from subcortical systems | Classifying primary affective systems |
| Cabanac's common currency | Sensations have quality, intensity, duration and affect; affect ranks choices efficiently | Explaining why valence is adaptive |
| LeDoux's survival circuits | Defensive circuits create global organismic states without requiring felt emotion | Testing whether behaviour implies feeling |
| Affect criteria (Tables 8.1, 8.2) | Exclude reflexes, classical conditioning and decerebrate acts; accept operant learning, trade-offs, frustration, contrast, self-medication, drug approach | Behavioural tests of sentience in any animal |
| Reward system plus adaptive behaviour network | The mesolimbic system values stimuli; the ABN executes survival behaviours; reward reinforces | Mapping valuation onto action |
| Situational adaptiveness of pain (Rose) | Prolonged suffering pays only for animals that can hide and heal | Asking which animals suffer, not just sense damage |
| Seven criteria for exteroceptive consciousness (Table 9.2) | Complexity, multiple deep hierarchies, isomorphism, reciprocal interactions, multisensory convergence, memory region, selective attention | Judging invertebrate (or engineered) candidates |
| Three postulates of neurobiological naturalism | Known neurobiology suffices; consciousness is ancient, widespread, diverse; subjectivity follows from neurobiology plus evolution | Stating the whole theory in brief |
| Auto- and allo-ontological irreducibility | Subjects cannot access their own neurons; observers cannot access subjects' experience | Explaining why first person never reduces to third |
| Damasio's selves; Tulving's consciousness levels | Proto, core, autobiographical self; anoetic, noetic, autonoetic consciousness | Separating primary consciousness from higher, self-aware forms |
| Consilience answer | Neurobiological, evolutionary and philosophical answers are each needed for "why does red feel like red?" | Handling the character-of-experience objection |
Strengths. Few books bring this much comparative anatomy, paleontology and behaviour to the question of animal consciousness. The criteria are explicit and reusable, the tables keep "absent" separate from "untested," and the authors flag their own weak points (small insect brains, missing octopus maps, the double visual image in birds). Their split of consciousness into exteroceptive, interoceptive and affective kinds is now widely used.
Blind spots. The core move, from mapped, attended, valenced processing to experience, is assumed rather than demonstrated; the criteria detect functions that a nonconscious system could in principle perform. Tectal consciousness in fish rests on inference, with key recordings from birds. The fish-pain chapter leans on Rose's sceptical camp. The anti-epiphenomenalist argument from tight correlation does not work, since epiphenomenal experience would correlate just as tightly. The "closing" of the explanatory gaps explains why we do not perceive our neurons and why qualia differ, not why there is experience at all, so most philosophers would say the hard problem is relocated, not solved. Requiring life excludes machine consciousness by stipulation.
Dated or contested. The vision-first scenario and the yunnanozoan reading remain disputed; the Cambrian base is now dated to about 538.8 mya; the claim that the vertebrate genome doubled twice at the very origin has been refined (one round before the cyclostome split). Lamprey work since 2017 suggests the pallium may hold topographic visual and somatosensory maps, which would soften the tectum-only story.
Since 2016. The field has moved toward the authors on distribution: the 2021 LSE review and the UK Animal Welfare (Sentience) Act 2022 cover decapods and cephalopods, and the 2024 New York Declaration grants a realistic possibility of consciousness to all vertebrates, cephalopods, decapods and insects. Bee studies (cross-modal recognition, positive emotion-like states, play, nociceptive trade-offs), octopus pain experiments (2021), the fruit fly connectome (2024) and crow neurons tracking subjective report (2020) have filled several of the book's empty cells. On mechanism, the IIT versus global workspace adversarial collaboration (2023 to 2025) left cortex-centred theories unsettled, and the 2023 "IIT as pseudoscience" letter showed how unsettled the field remains.
Who's who
| Name | Who | Why they matter here |
|---|---|---|
| Todd E. Feinberg | Neurologist and psychiatrist | Co-author; originated neurobiological naturalism, NSFCs and the auto/allo irreducibility framing |
| Jon M. Mallatt | Evolutionary biologist of early vertebrates | Co-author; lamprey feeding, Cambrian fossils (Haikouella), vertebrate brain evolution |
| Thomas Nagel | Philosopher | "Something it is like to be" criterion for consciousness |
| John Searle | Philosopher | Ontological subjectivity; biological naturalism, the authors' starting point |
| David Chalmers | Philosopher | Named the hard problem; the "character of experience" challenge answered in chapter 10 |
| Joseph Levine | Philosopher | Coined the explanatory gap |
| Evan Thompson | Philosopher | Life already has interiority; subjectivity emerges from living being |
| Ernst Mayr | Evolutionary biologist | Life as process, not substance; programmed goal-directed processes |
| Howard Pattee; Herbert Simon | Theoretical biologist; complexity theorist | Hierarchical constraint as the mark of life; hierarchy as the architecture of complexity |
| Giulio Tononi; Christof Koch | Neuroscientists | Integrated information; why the cerebellum is not conscious |
| Gerald Edelman | Neuroscientist | Primary consciousness as mental images in the present; memory-updated perception |
| Antonio Damasio | Neuroscientist | Mapped images, body-based feelings; proto, core and autobiographical selves |
| Charles Sherrington | Physiologist | Projicience; nociception (1906); the "material me" |
| Thurston Lacalli | Neurobiologist | Cell-level map of the larval amphioxus brain; ventral-core hypothesis |
| Rudolf Nieuwenhuys | Neuroanatomist | The primitive limbic core and greater limbic system |
| R. Glenn Northcutt | Comparative neurobiologist | Simple bilaterian ancestor; neural crest and placodes (with Gans) |
| Andrew Parker; Michael Trestman | Biologist; philosopher | Light-switch hypothesis; spatial vision and object individuation |
| Dan-Eric Nilsson | Vision biologist | Four stages of eye evolution |
| Ann Butler | Comparative neuroscientist | Two-wave sensory evolution, the "cephalate"; mammal-bird consciousness view |
| Junyuan Chen; Degan Shu | Paleontologists | Haikouella and Haikouichthys from Chengjiang |
| David Northmore; Ursula Dicke and Gerhard Roth | Neurobiologists | Fast tectal processing; tectal object recognition in amphibians |
| Walter Wilczynski | Neurobiologist | No unimodal mapped representations in the amphibian pallium |
| Bjorn Merker | Neuroscientist | Tectal consciousness in all vertebrates; hydranencephaly evidence |
| Harvey Karten; Erich Jarvis | Neuroscientists | Bird and mammal pallium homologies |
| Timothy Rowe; Amy Balanoff; Michael Lee | Paleontologists | Nocturnal smell-driven early mammals; brain growth in the bird-line radiation |
| A. D. (Bud) Craig | Neuroanatomist | Interoception and the insular "sentient self" |
| Charles Vierck | Pain researcher | Conscious pain in somatosensory cortex SI |
| Derek Denton | Physiologist | Primordial emotions (thirst, air hunger) as first consciousness |
| Jaak Panksepp; Mark Solms | Affective neuroscientists | Seven primary emotions; subcortical origin of affect |
| Michel Cabanac | Physiologist | Pleasure as common currency; dated affect to amniotes |
| Joseph LeDoux | Neuroscientist | Survival circuits; feelings require human cognition |
| James D. Rose; Lynne Sneddon | Neurobiologists | Opposite sides of the fish pain debate |
| Wolfram Schultz | Neurophysiologist | Dopamine reward prediction; reward-seeking as purposive action |
| Johannes Seelig and Vivek Jayaraman | Neuroscientists | Fifth-level retinotopy in the fruit fly central complex |
| Lars Chittka and Jeremy Niven | Neuroethologists | Are insect brains too small? |
| Karine Fauria | Behavioural biologist | Bumblebees learning occluded gate-target combinations (a held image) |
| Nicholas Strausfeld | Neuroanatomist | Arthropod Brains; shared arthropod brain plan |
| Tamar Gutnick | Neurobiologist | Octopuses guiding arms by vision through a maze |
| Gordon Globus; Herbert Feigl | Psychiatrist-philosopher; philosopher | Subject versus observer perspectives; the autocerebroscope |
| Shaun Nichols and Todd Grantham; Mark Bradley | Philosophers | Arguments that consciousness is an adaptation, not a by-product |
| Endel Tulving | Memory researcher | Anoetic, noetic, autonoetic consciousness |
| Francis Crick | Biologist | The "Astonishing Hypothesis" that you are a pack of neurons |
Connect it
- A capability checklist is better than a vibe. Tables 8.2 and 9.2 show how to judge a contested capability: list what a baseline system already produces (reflexes, classical conditioning), exclude it, then score only flexible, memory-dependent, value-weighing behaviour. The same discipline applies to claims that an AI model "reasons," "plans" or "understands," and to evaluating agentic systems in a bank before they are allowed to act.
- Fast tier, wise tier. The tectum-pallium split is the classic architecture of real-time screening plus deliberative review: a fast mapped engine for immediate decisions and a slower, memory-backed tier that can override it where latency is cheap. Fraud and sanctions screening that escalates to analyst-assist tooling follows the same pattern, and the book's "safe neighbourhood" condition is a reminder that the deliberative tier only pays where delay is affordable.
- Valuation is its own layer. The three-way split of exteroceptive, interoceptive and affective processing maps onto external data, internal state and a valuation layer (risk appetite, scoring); Cabanac's point that a single common currency makes choice tractable is the case for explicit, auditable scoring rather than rules for every case.
- Interpretability has an auto/allo problem. A model has no access to its own weights, and an auditor with the weights lacks the model's "view"; as with animals, behavioural tests under controlled conditions remain indispensable alongside mechanistic inspection.