Insight Editions · No. 288
Consciousness Demystified
Todd E. Feinberg and Jon M. Mallatt
Feinberg and Mallatt argue that consciousness, like life, is unique in nature yet fully natural.
A study guide to Consciousness Demystified 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
Consciousness Demystified (MIT Press, 2018) is neurologist Todd Feinberg and evolutionary biologist Jon Mallatt's compact case that subjective experience is a natural product of living, evolved brains, with no new physics and no mystery force. They break consciousness into images and affects, find it in all vertebrates, arthropods and cephalopods, list the features every conscious brain shares, date its origin to the Cambrian explosion, and explain why the first-person view can never coincide with the third-person view. The book's claim is that the explanatory gaps are real but natural: they come from life, from a stack of special neural features, and from two built-in irreducibilities, not from anything beyond biology.
Neurobiological naturalism: consciousness is unique but fully natural, created by life + reflexes + special neurobiological features, and its subjectivity follows from embodiment and auto- and allo-ontological irreducibility.
Comparative neuroscience, animal behavior, paleontology and philosophy of mind, focused on primary (sensory) consciousness across vertebrates and invertebrates, from 3.7 billion years ago to the present.
4 gap features (NSFCs); 5 behavioral criteria for affect; 3 feature levels; 560 to 520 million years ago; 3 independent origins.
Readers who want a biology-first, evidence-based answer to the hard problem, and a checklist for judging which creatures (or systems) might be conscious.
The precedent. Life once seemed to need a "life force"; Ernst Mayr showed it is unique yet natural. Feinberg and Mallatt extend John Searle's biological naturalism into neurobiological naturalism, resting on three principles: consciousness is grounded in life; it needs reflexes, a core brain and special neurobiological features; and the subjective–objective barrier has natural causes.
The target is split. Primary consciousness ("something it is like to be," after Nagel) comes in three domains, exteroceptive, interoceptive and affective, often simplified to mapped images versus valenced affects. Each domain carries four explanatory gaps: referral, mental unity, mental causation and qualia.
Who is conscious. Isomorphic sensory maps in the midbrain's optic tectum mean all vertebrates, even the lamprey, make mental images; five behavioral criteria headed by global operant learning, plus conserved subcortical circuits, mean all vertebrates feel affects. Arthropods (bees) and cephalopods (octopuses) pass the same tests; roundworms, amphioxus and tunicates fail.
What they share. Conscious brains rest on general features of life (embodiment, process, system, hierarchy, teleonomy, adaptation) and reflexes and core brains, then add the special features: neural complexity, elaborate senses, reciprocal neural hierarchies, mapped or valence-coded pathways, attention and memory.
When and why. A predator–prey arms race in the Cambrian explosion flooded the seas with signals. Vertebrates and arthropods, vision first, evolved distance senses and brains that turned them into images with affects, which beat hard-wired reflexes at organizing input, ranking importance, flexibility, learning and prediction. Cephalopods followed later and independently.
How subjectivity arises. Each gap traces down to life and reflexes. A subject cannot experience its own neurons (auto-ontological irreducibility); an observer cannot experience the subject's experience (allo-ontological irreducibility). The neurobiology of qualia and the subjectivity of qualia need different explanations, and the hard problem fuses three separate issues that must be pulled apart.
| Ch | Title | One-line payload |
|---|---|---|
| P | Preface | Goal: place consciousness among the properties of life and explain it naturally. |
| 1 | What Makes Consciousness "Mysterious"? | Like life, consciousness is unique but natural; three principles of neurobiological naturalism. |
| 2 | Approaching the Gaps: Images and Affects | Three domains of primary consciousness, each with four gaps: referral, unity, causation, qualia. |
| 3 | Naturalizing Vertebrate Consciousness: Mental Images | The mapped optic tectum gives every vertebrate, even lampreys, conscious images; amphioxus fails. |
| 4 | Naturalizing Vertebrate Consciousness: Affects | Five behavioral criteria and conserved subcortical circuits show all vertebrates feel affects. |
| 5 | The Question of Invertebrate Consciousness | Arthropods and cephalopods pass; consciousness evolved three times by convergence. |
| 6 | Creating Consciousness: The General and Special Features | Three levels of features build consciousness with no gap; box 6.3 is the checklist. |
| 7 | The Evolution of Primary Consciousness and the Cambrian Hypothesis | Consciousness arose 560 to 520 million years ago in a vision-led arms race. |
| 8 | Naturalizing Subjectivity | Gaps rooted in life; two irreducibilities; the hard problem is three problems. |
Preface
Feinberg and Mallatt set out to demystify and naturalize consciousness: to show that subjective, first-person experience is produced by ordinary brain processes, evolved by ordinary natural selection, and needs no new force, physical or nonphysical. The book's wager is that the "gaps" between neurons and experience are real but naturally explainable, once consciousness is placed among the properties of life.
Most consciousness scientists already accept that conscious animals evolved from nonconscious ones at some point in Earth's history. The puzzle is that many scholars still treat consciousness as fundamentally different from the rest of nature, with seemingly unbridgeable gaps between material neurons and immaterial-seeming experience. This short book, published by MIT Press in 2018, takes those gaps as its target.
Where this book sits
The authors' earlier book, The Ancient Origins of Consciousness (MIT Press, 2016), asked which animals are conscious and when consciousness evolved. This sequel turns to the nature of consciousness and the explanatory gaps between the objectively observed brain and subjective experience. It condenses about six years of their scientific papers into a less technical argument, adds a glossary, and folds in newer findings, especially on the neural circuits for affects (emotional feelings).
- Images were mapped first. The pathways that build mapped mental images of the world were worked out at least thirty years before this book.
- Affects are catching up. Since 2016, circuit-level work on reward, aversion, thirst and danger neurons has made the neuroscience of feeling nearly as concrete as that of perception.
Neurobiological naturalism is the authors' theory, an outgrowth of John Searle's biological naturalism. It holds that subjective experience is unique in nature yet created by life processes plus special features of evolved brains, with no new laws of physics required. It ties consciousness tightly to life: a property that evolved in living animal bodies.
The stance is deliberately two-sided. Against the mysterians, the authors insist that experience is produced by natural brain processes and evolved in natural ways. Against flat reductionists, they grant that subjectivity is genuinely unique and that a third-person description will never be the same thing as a first-person feeling. Their move is to explain why that difference exists, rather than deny it or treat it as magic.
The plan of attack
| Step | Question | Where answered |
|---|---|---|
| Name the gaps | What exactly is mysterious about experience? | Chapters 1 and 2 |
| Find the conscious animals | Which vertebrates and invertebrates have images and affects? | Chapters 3 to 5 |
| List the shared features | What do all conscious brains have in common? | Chapter 6 |
| Date the transition | When and why did consciousness evolve? | Chapter 7 |
| Close the gaps | How does subjectivity arise naturally? | Chapter 8 |
The goal is simple to state: place the most perplexing philosophical features of consciousness squarely among the properties of life, then explain when and why subjectivity evolved and how experience is created.
What Makes Consciousness "Mysterious"?
Life once looked as mysterious as consciousness does now, and biology dissolved the mystery without a "life force." Feinberg and Mallatt propose the same path for experience: a theory called neurobiological naturalism, resting on three principles (life, neural features, and a naturally explainable subjective–objective barrier). The chapter defines the target, primary consciousness, and maps the book's route to it.
The precedent: life was a mystery too
Ernst Mayr, in What Makes Biology Unique?, listed properties found only in living things: reproduction, coded information, metabolism, adaptation to the environment, and more. For centuries these looked so strange that scholars posited a special "life force" that animated matter and built embryos into bodies. Twentieth-century biochemistry removed any need for it. Biologists now agree that life is unique but fully natural: special features, no special physics.
Consciousness looks different. Nonconscious brain functions such as neural reflexes can be explained completely by physics, chemistry and biology. But whenever anyone tries to relate neurons to subjective experience, something is always "left out." The philosopher Joseph Levine named this the explanatory gap (1983): the difficulty of fully explaining, equating or reducing experience to brain function.
Feinberg and Mallatt accept that consciousness is special, as Mayr accepted that life is special. Their claim is that consciousness, like life, can be unique in nature and still be completely natural, with no new laws or forces required.
The target: primary consciousness
The authors aim at the most basic form of experience, the one Thomas Nagel pinned down in "What Is It Like to Be a Bat?" (1974): an organism has conscious states if and only if there is "something it is like to be" that organism, something it is like for the organism. Nagel called this the subjective character of experience.
- Many names, one thing. The literature calls it sensory, phenomenal, primary or perceptual consciousness; the book uses primary consciousness and sensory consciousness.
- No reflection needed. Primary consciousness does not require self-awareness or thinking about one's thoughts; those are later, "higher" forms.
- Revonsuo's minimal criterion. For Antti Revonsuo, the mere presence of any pattern of subjective experience, simple or complex, faint or vivid, meaningful or not, is the necessary and minimally sufficient condition for phenomenal consciousness.
Note what is excluded. The book is not about attention, self-consciousness, language or metacognition, and not about the clinical levels of consciousness (coma, sleep, waking) except where arousal systems matter as background conditions. It is about the contents that make up experience at all: mapped images of the world and body, and felt affects.
Even at this elemental level, the divide between brain and experience appears. The book's job is to explain and "naturalize" that gap for the simplest case, on the reasoning that if the simplest gap closes, the harder cases inherit the method.
The minimal experiencer. Imagine an animal that has no thoughts about itself, no language and no memory beyond a second or two, yet when light hits its eye there is a faint, fleeting something for it: a bright patch, out there. Revonsuo's criterion says this animal is phenomenally conscious. The thought experiment fixes the target: the book is not explaining human self-reflection but the switching-on of any experience at all.
Searle's biological naturalism, and what it leaves out
The authors start from John Searle's biological naturalism. Searle argues that conscious feeling states cannot be reduced to the brain, meaning they cannot be fully explained by the functions of brain parts. The reason is symmetrical: a third-person observation of neuron firings necessarily leaves out the first-person feeling, and you cannot reduce the feeling to firings any more than firings to feelings, because each reduction drops the objectivity or subjectivity in question. Searle called this the ontological subjectivity of consciousness.
Yet Searle insists that mental phenomena are entirely natural, caused by neurophysiological processes, and as much a part of biology as digestion, mitosis, meiosis or enzyme secretion. No undiscovered fundamental principle is needed.
Searle (biological naturalism). Consciousness is caused by the brain and is fully biological. The first-person/third-person gap is permanent and irreducible. Little is said about which features of brains make subjectivity possible.
Feinberg and Mallatt (neurobiological naturalism). Agree with Searle's premise but add that consciousness needs explanations unique within biology. They name the specific neural features that create subjectivity and explain why the first-person/third-person barrier exists.
The neurosurgeon and the patient. A surgeon watches, on a screen, every spike in a patient's visual cortex while the patient looks at a sunset. The surgeon gets a complete third-person record and still does not have the patient's orange glow. Reverse it: the patient has the glow but no idea which neurons fired. Searle's point is that each viewpoint omits what the other contains, so neither can be "reduced" to the other. Feinberg and Mallatt accept the scene but ask a further question: why is the system built so that each viewpoint is blind to the other? Chapter 8 answers with auto- and allo-ontological irreducibility.
The authors' complaint is precise: Searle does not pay enough attention to the special status of consciousness within the biological sciences. Without an account of the unique factors in conscious brains, consciousness stays a mystery to philosophers and neuroscientists alike.
Neurobiological naturalism: three principles
The authors' theory says consciousness is fully natural but requires explanations uniquely different from those used elsewhere in biology. It follows Mayr's model: accept uniqueness, keep naturalism.
- Life. Consciousness shares the features that distinguish all living things from nonlife (Mayr's list). These are consistent with standard physics and chemistry, and the unique features of consciousness are grounded in the unique features of life. Consciousness must be understood as a living process.
- Neural features. On top of life sit features unique to nervous systems. First, simple and complex reflexes and motor programs in the spinal cord and core brain, which are not conscious but are indispensable building blocks. Then a complex suite of special neurobiological features found only in conscious brains. These are new, numerous and revolutionary, but they work through no new "physics of the brain" and no mysterious "consciousness force."
- The subjective–objective barrier is naturally explainable. Once brains have the special features, there are natural reasons why one's own brain processes are inaccessible from the first-person view, and why one's experience is inaccessible to a third-person observer.
The core brain (a term from Thurston Lacalli, 2008) is the set of brain regions needed for basic survival but not for consciousness; in vertebrates it is the brain stem plus parts of the diencephalon.
The road map of the book
The outline is a sequence of moves, each meant to leave no gap in either neural structure or evolutionary time.
| Chapter | Move | Payoff |
|---|---|---|
| 2 | Split primary consciousness into exteroceptive, interoceptive and affective domains (the qualia types) | Shows each domain has multiple explanatory gaps |
| 3 | Which vertebrates make mapped sensory images | Diversity and commonality across species |
| 4 | Which vertebrates feel affects | Same pattern for feelings |
| 5 | Arthropods and cephalopods | Very different brains, same features |
| 6 | Formal lists: general, reflexive/core, special features | A checklist for consciousness in any newly found animal; three steps from nonconscious to conscious |
| 7 | Fossil record and adaptive value | Continuous steps, with a big jump about half a billion years ago |
| 8 | Integrated, hierarchical model | Why subjectivity differs from objective observation yet is natural |
- Ernst Mayr: evolutionary biologist; argued life is unique yet natural, the template for the book's stance.
- Joseph Levine: philosopher who coined the explanatory gap (1983).
- Thomas Nagel: defined consciousness as "something it is like to be" an organism (1974).
- Antti Revonsuo: gave the minimal criterion: any pattern of experience present-for-it.
- John Searle: biological naturalism; consciousness is caused by brains yet ontologically subjective.
- Thurston Lacalli: neurobiologist who named the core brain (2008).
Is the life analogy fair? The authors' opening move assumes the consciousness gap is the same kind of gap that the old life-force doctrine faced. Critics such as David Chalmers reply that the life puzzle was always about functions (reproduction, metabolism), which mechanisms explain, whereas the hard problem is about why any function is felt. The book's answer arrives in chapter 8; readers should hold the analogy as a hypothesis to test, not a settled premise.
For an architect, the method is familiar: decompose a "magical" emergent property into layered, testable components, and refuse to accept a single silver-bullet explanation. The book's later checklist (chapter 6) works like a reference architecture for consciousness: necessary layers, each with defined responsibilities.
Approaching the Gaps: Images and Affects
"Primary consciousness" is not one thing. Feinberg and Mallatt split it into three overlapping domains, exteroceptive, interoceptive and affective, arranged from outside the body to inside, and often simplified to images versus affects. They also show that the explanatory gap is not one gap but four, the neuro-ontologically subjective features of consciousness (NSFCs): referral, mental unity, mental causation and qualia, each present in every domain.
Three domains on a continuum
Researchers usually treat primary sensory consciousness as a single level. The authors divide it into three partially overlapping, interacting domains within the overall unity of consciousness, lined up on an outer-to-inner (extrapersonal-to-intrapersonal) continuum. Different theorists have championed different domains as "the" basis of consciousness:
- Exteroceptive (images of the outside world). Gerald Edelman defined primary consciousness as being "mentally aware of things in the world, of having mental images in the present"; Antonio Damasio made mental images part of core consciousness.
- Affective (feelings of good and bad). Michel Cabanac, Derek Denton and Jaak Panksepp made valenced feeling the key aspect of primary consciousness.
- Interoceptive (the inner body). A. D. Craig, Derek Denton and others put sensations from the viscera at the center. It sits in the middle, sharing features of both other domains.
The authors refuse to pick one. All three involve subjective experience and all three entail explanatory gaps, so a complete naturalistic account must cover all of them.
Shared building blocks: neurons and reflexes
All three domains are built from the same parts. Neurons form chains and networks, especially in the central nervous system (brain plus nerve cord or spinal cord), and signal one another at synapses using neurotransmitters. The simplest chain is a reflex arc: a sensory neuron, usually an interneuron, and a motor neuron. Reflexes are fast, involuntary and nonconscious: the knee jerk to a tap on the patellar tendon, the iris constricting in bright light. They operate even in people in a coma. The authors treat reflexes as the original building blocks of consciousness, though not conscious themselves.
Exteroceptive consciousness: mapped images
Exteroception processes sights, sounds, smells and touches from the outside world, mostly through the distance senses. Its end product is a sensory mental image: a simulation of one's immediate surroundings. This is not "mental imagery" or imagination, which is the more advanced ability to represent things without current perceptual input.
The signature of exteroceptive pathways is topographic (isomorphic) mapping: an orderly representation of the receptor surface is preserved point by point up the hierarchy of neurons to the brain.
- Retinotopy (vision). Points of the viewed world map onto the retina, and that map is kept in the higher visual pathways.
- Somatotopy (touch). Adjacent patches of skin are represented by adjacent neurons at successive brain levels.
- Tonotopy (hearing). Different tones vibrate different positions on the cochlea, and that tone map is kept up to the level where sound is heard.
- Smell (chemotopy). Works a little differently, mapping odor combinations into the brain, and also builds a smell map of space.
Affective consciousness: valence, whole-body, motivating
Affects differ from images in three ways that matter later.
- Intrinsic valence. Exteroceptive images carry no value in themselves; the sight of a lion or smell of a flower evokes fear or joy only after additional processing by other brain systems. Affects always carry valence: aversive (negative) or attractive (positive).
- Global, not local. Images are referred to a specific place in the world or body. Affects involve the whole bodily individual: you do not experience happiness in your foot.
- Tied to motivation. Positive affects (liking, pleasure) drive approach; negative affects (dislike, displeasure, discomfort) drive avoidance or escape.
Anatomically, the brain regions for affect are more diffuse, more numerous and less strictly hierarchical than the pathways for images (chapter 4).
Interoceptive consciousness: the middle ground
Interoception senses physiological and mechanical changes inside the body: the esophagus when you swallow hot coffee, air hunger when you hold your breath, plus global states such as hunger, thirst and fatigue. These signals tell the animal whether internal workings are running smoothly so it can hold an optimal steady state, homeostasis. Damasio and colleagues found that brain areas for interoception, homeostasis and emotion overlap closely in humans.
Interoception leans toward affect because it often carries valence: a wave of nausea or the pain of angina upsets you fast. It leans toward exteroception because it builds somatotopically mapped images of the body, though cruder ones: it can be hard to tell whether abdominal pain comes from the small intestine, appendix, gallbladder or ovary. Its signals still climb clear neural hierarchies.
Pain, temperature and itch straddle the line
A. D. Craig argues that skin receptors for pain (nociception), temperature and itch serve homeostasis and should count as interoceptive. They carry strong affect and drive, yet they also have truly exteroceptive traits: a pinprick gives an exquisitely precise somatotopic image of the pin's tip. Pain's dual nature shows in its two subtypes:
- Sharp, pricking pain is precisely localized on the skin, like exteroception.
- Slow, dull, burning pain spreads over larger areas, like interoception.
Proprioception, the sense of movement
Proprioception measures stretch in joints, muscles, tendons and skin, telling the animal how it is moving through space so it can stay balanced and reach targets. Much of it is unconscious because corrections must be too fast to think about, but joint stretch (and some muscle stretch) is consciously felt: do a full-body stretch and notice the elbows, shoulders, knees and hips. It is closer to exteroception because it has precise somatotopy and shares some exteroceptive pathways. Some theorists (Barron and Klein, Cruse and Schilling, Godfrey-Smith, Llinás, Merker, Morsella) center consciousness on movement and action; the authors focus on the sensory side but agree that directing movement through complex space is a major biological function of consciousness.
A definitional choice with consequences. The authors use "interoception" in Sherrington's classical sense (1906): receiving inner-body signals. They reject the broader usage of Ceunen, Vlaeyen and Van Diest (2016), which builds conscious awareness into the definition, because that usage is based on mammals and the cerebral cortex and would rule out interoceptive consciousness in animals without a cortex before any investigation. The choice is defensible, but it shows that how terms are defined can pre-decide the animal-consciousness question in either direction.
Multiple explanatory gaps: the four NSFCs
The explanatory gap is usually treated as a single question. The authors find four categories of gap, each present in all three domains, and call them the neuro-ontologically subjective features of consciousness (NSFCs).
- Referral. Experiences are perceived not in the brain, where they are built, but out in the world, in the body, or as a whole-self feeling. The clearest case is projicience: sights, sounds and smells projected into outer space. No brain activity is ever experienced as occurring in the brain.
- Mental unity. A divisible, discontinuous brain of separate neurons yields a unified, continuous field of awareness that holds world, location, mood and inner-body state at once. This is the grain problem: the neural grain is far coarser than the seamless grain of experience.
- Mental causation. How can a subjective, seemingly immaterial, unobservable mind cause physical effects, moving one's own body and through it objects in the world?
- Qualia. The perceived qualities of world, body or emotion: seen colors, a full stomach, anger. Many philosophers treat qualia as the holy grail: why do neural states feel the way they do, and why do brains produce feelings at all?
| Gap | Exteroceptive example | Interoceptive example | Affective example |
|---|---|---|---|
| Referral | Sunset seen beyond the horizon, not on the retina | Appendicitis felt in the abdomen, not the brain | Mood felt as a state of the whole self |
| Unity | Color, shape, motion bound into one object | Cold and congestion felt as one bodily state | Mood colors the whole scene |
| Causation | Seeing food leads you to reach for it | Air hunger makes you surface | After a snowstorm, hunger and worry empty the town's fridges |
| Qualia | Redness of red | Fullness of a stomach | The feel of anger |
The phantom sunset. Suppose a neuroscientist could record every neuron that builds your sunset and display them on a screen in the room. You would see a pattern of spikes in your brain, but you would still experience the orange glow out at the horizon. The scene isolates referral: the experience is constructed in one place and lived in another, and nothing in the spike pattern announces that relocation. The authors treat explaining this as central to their final theory.
Each of the three domains has all four explanatory gaps, so a satisfactory naturalistic account must explain referral, unity, mental causation and qualia across images, interoception and affect alike. The next step is to ask which animals have images and affects at all, and what their brains share.
Interoception has since become one of the busiest areas of consciousness science. Anil Seth's "beast machine" account (Being You, 2021) and Mark Solms's brain-stem theory of affect (The Hidden Spring, 2021) both place inner-body regulation and feeling, not exteroceptive perception, at the root of experience, which strengthens the authors' insistence that interoceptive and affective domains are not afterthoughts.
Naturalizing Vertebrate Consciousness: Mental Images
If mapped, multisensory representations mark mental images, then every vertebrate, down to the jawless lamprey, has image-based consciousness, because every vertebrate brain has a finely mapped optic tectum in the midbrain. Fish images pass all four NSFC tests (referred, unified, causal, qualitative), while the closest invertebrate relatives, amphioxus and tunicates, fail. The authors trace the jump to a sensory revolution at the dawn of vertebrates, driven by the camera eye, ectodermal placodes and neural crest, more than 520 million years ago.
The criterion and the target group
Nagel's "something it is like to be" may look like a low bar, but it is the right one when the question is whether an animal has any experience at all. Chapters 3 to 5 work out which living animals have primary consciousness and what their brains share, so that chapter 6 can name the common features. The vertebrates (fish, amphibians, reptiles, birds, mammals) come first because humans, the only animals known with certainty to be conscious, belong to them.
Beyond the cortex
Many researchers now grant consciousness to all mammals and birds, both of which have an expanded dorsal pallium: the cerebral cortex in mammals and its equivalent in birds. The standard view links consciousness to the corticothalamic system (cortex working with the thalamus), and human evidence supports it: damage to cortex causes cortical blindness, cortical deafness and loss of touch perception. But other vertebrates have a cerebrum without a cerebral cortex, and fish still look conscious: they have elaborate senses, hierarchical sensory anatomy including sharp vision, and they pay alert attention to stimuli. To find the most basic neural underpinnings, the authors look at the most basal species.
The deduction
The core inference is short, and the rest of the chapter tests it.
- Any brain that builds isomorphically mapped, multisensory representations creates mental images (from chapter 2).
- Mapped multisensory representations are found in all vertebrate brains. In mammals and birds the main maps lie in the cerebrum; in every vertebrate the midbrain's optic tectum holds a point-by-point map of the sensed environment, mostly visual but also from hearing, touch and balance.
- Therefore all vertebrates, even the earliest, have exteroceptive primary consciousness.
The optic tectum: the image-maker of fish and amphibians
The optic tectum forms most of the roof of the midbrain and is relatively huge in fish and amphibians. The authors deduce it is the site of image consciousness in those animals, so the seat of image-based consciousness shifted from tectum to cerebral cortex during mammal evolution (and independently in birds). In mammals the same structure survives as the superior colliculus but, on this view, no longer makes conscious images.
- Function. In all vertebrates the tectal map lets the animal attend to, look at and turn toward the most important object in view. Fish and amphibian researchers describe it as the center for object "recognition" and "perception", terms of consciousness, so the claim does not rest on mapping alone.
- Complexity. It has three or more layers of densely packed neurons, up to fifteen neuron types in the goldfish, many sensory inputs, and extensive connections to higher and lower motor-control centers.
- Prey recognition. In larval zebrafish, a subset of tectal neurons responds selectively to the size, speed and light or dark tone of prey (Bianco and Engert, 2015).
What the tectum does not do
The tectum is not the whole of fish consciousness. Three functions sit elsewhere.
| Function | Brain site in all vertebrates | Implication |
|---|---|---|
| Smell perception | Pallium of the cerebrum (no direct smell input to tectum) | Smell consciousness has always involved the pallium |
| Memory recall | Hippocampus (medial pallium) | Recalling a predator's look enables instant recognition and escape |
| Arousal (level of consciousness) | Reticular formation of brain stem and parts of basal forebrain | Widely projecting neurons release chemicals for alertness and vigilance |
| Selective attention | Tectum itself (plus others) | Finds the most salient object and turns the eyes to it |
Lamprey attention (Kardamakis, Pérez-Fernández and Grillner, 2016). Scientists once thought the tectum needed nearby brain-stem isthmus nuclei to direct attention. In the lamprey, when one object emits two kinds of signal (say, visual and electric-field), both converge on the same tectal neuron, which then orders the eyes to look at the exact point in space where both arose. Result: the tectum can direct attention on its own, and it is a true multisensory meeting point.
Archerfish search (Ben-Tov and colleagues; Woodruff, 2017). Micheal Woodruff argues that the best attention-based sign of image consciousness in fish is persistent visual search for an object. Archerfish spot insects in foliage above the water and knock them down with jets of water, which needs a mapped image of both background scene and attended prey. Mor Ben-Tov found the neurons that make that figure-ground distinction in the optic tectum. Archerfish can also learn individual human faces (Newport and colleagues, 2016), implying a remembered mental image.
The lamprey test
Lampreys are jawless fish in the oldest surviving vertebrate group. A lamprey has the smallest brain per body size of any vertebrate, yet it has the full vertebrate sensory kit: good eyes, sensitive touch, keen smell, hearing and electroreception. Its optic tectum receives the same mapped, hierarchical multisensory input as any fish. So if fish have image-based consciousness, lampreys do too.
The authors then ask whether these tectal images are really conscious, and run them through the four NSFCs of chapter 2.
- Referred. The lamprey responds appropriately to complex visual stimuli out in the environment.
- Unified. These "vampire killer eels" choose, approach and attach to the fish they parasitize, which suggests size, shape, movement and color are bound into one image.
- Causal. They act in complex, prolonged, planned, nonreflexive ways that change the world: in mating, they move rocks to build a nest in the streambed.
- Qualitative. Fish make high-level discriminations among sounds, visual features, electric signals, smells and tastes, so they distinguish qualia.
The conclusion: there is something it is like to be a lamprey, and so, echoing Nagel, surely something it is like to be a bat.
Comparing mammal and lamprey pathways yields the first list of shared features of image-based consciousness: many complex sensory receptors, a brain, sensory hierarchies at least three neurons long, a place where multiple senses converge, and brain sites for arousal and attention.
Weak points the authors acknowledge in their notes. Reptiles are confusing: their cerebrum shows little sensory isomorphism yet is somewhat expanded, and their tectum is large, so it is hard to tell whether reptile consciousness is tectal or cortical. The lamprey pallium (Suryanarayana and colleagues, 2017) resembles mammalian cortex in its fundamentals but looks puzzlingly simple for animals that track prey, migrate long distances and build nests. And the step from "mapped representation" to "conscious image" is an inference to the best explanation, not a demonstration; mapped visual layers also exist in systems nobody calls conscious. The behavioral evidence (recognition, search, discrimination) carries real weight here.
The cousins that fail: amphioxus and tunicates
If the features above mark consciousness, animals without them should be reflexive, using simple motor programs that need no unified world image. The test cases are the vertebrates' closest living relatives, both marine: the fish-like amphioxus (cephalochordates) and the bag-like tunicates (sea squirts). Their lineages split from vertebrates before 520 million years ago, in the Cambrian. Much of the detail comes from Thurston Lacalli's decades of work (and Ian Meinertzhagen's lab on tunicate larvae).
- Senses. Simple light receptors, mechanical touch receptors, possibly chemoreceptors; no camera eyes, sound-detecting ears or odor-smelling noses.
- Brains. Tunicate brains (cerebral ganglia) have several hundred neurons at most.
- Pathways. In amphioxus larvae, sensory pathways are just one or two neurons long before reaching a motor center: reflex arcs, not processing hierarchies. Light and touch mainly arouse the animal to swim.
- Structures. No pallium and no tectum has been found in amphioxus.
The vertebrate sensory revolution
Why do vertebrates have so many features of consciousness and their cousins none? The authors' answer begins with vision. The camera eye of the first fish forms a photograph-like, detailed image on the retina, which an elaborated optic tectum turns into a mental image of visual space. Touch, water vibrations and electric-field senses were then integrated into the visual map.
A camera eye needs a focusing lens, and the vertebrate lens comes from ectodermal placodes, embryonic thickenings of the outer surface found only in vertebrates. Placodes plus the vertebrate-only neural crest build nearly all the special sensory structures: smell receptors, the inner ear, many sensory neurons of the head, and more.
Isomorphic maps are the cornerstone of image-based consciousness; they evolved in early vertebrates more than 520 million years ago as the natural result of the camera eye, neural crest and placodes. Yet which brain region builds the maps (tectum or cortex) varies across species: commonality within diversity.
The eye without a brain to match. Give an amphioxus larva a perfect camera eye but leave its two-neuron pathway untouched. It would receive a crisp retinal image and do nothing with it except swim when the light changes. Now give it the tectum and the multisensory hierarchy but only an eye spot: a large processor with almost nothing to process. The authors' point is that eye and brain had to evolve together: rich input drove processing, and processing made the input worth having.
Bird and fish evidence has strengthened since 2018. In 2020, Andreas Nieder and colleagues reported in Science a neural correlate of sensory consciousness in crows: pallial neurons tracked what the bird reported seeing near threshold, not just what was shown, in a brain without a layered cortex. The 2024 New York Declaration on Animal Consciousness states that there is strong scientific support for conscious experience in mammals and birds and at least a realistic possibility in all vertebrates, including reptiles, amphibians and fishes. The tectal hypothesis itself remains a minority, contested position.
Naturalizing Vertebrate Consciousness: Affects
Affects leave no neat topographic maps, so Feinberg and Mallatt identify them by behavior: five criteria headed by global operant conditioning. All vertebrates pass, and all vertebrates, down to lampreys, have the subcortical structures that generate raw affect in mammals, including the medial forebrain bundle. The affective system is built differently from the image system (valence neurons, neuromodulators, no maps) but shares its hierarchies, multisensory input, attention and memory, and it feeds directly into the premotor regions that choose actions.
Finding feelings by behavior
Affects at their most basic are feelings with valence, positive or negative. (The notes split affects into emotions, short, intense and stimulus-triggered, and moods, longer background states such as optimism or irritation.) Since affects lack the topographic signature that marks images, the authors need a different marker.
Simple approach and avoidance will not do: bacteria approach nutrients and flee toxins without feeling anything. Each criterion must go further, requiring remembered affects, extended affects, or several distinct categories of affect.
- Global operant conditioning. Learning a brand-new behavior involving the whole body from its consequences.
- Behavioral trade-offs. Value-based cost-benefit decisions, weighing two valences against each other.
- Frustration behavior. Acting out (e.g., aggression) when an expected reward is denied, showing negative affect persists.
- Self-delivery of pain relievers or rewards. Actively seeking positive states.
- Approach to reinforcing drugs, or conditioned place preference. Preferring the place where drugs or rewards were received before.
Kinds of learning, ranked
| Learning type | What is learned | Example | Who shows it |
|---|---|---|---|
| Habituation / sensitization | Respond less / more to a repeated stimulus | Ignoring constant noise | Nearly all animals with neurons |
| Classical (Pavlovian) | A conditioned stimulus (bell) predicts an unconditioned one (food) | Salivating at the bell | Every animal group with a nervous system, even sea anemones |
| Operant | A behavior predicts its consequence | Pressing a bar for a pellet | Many animals, often limited to one body part |
| Global operant | A brand-new whole-body behavior from consequences | Learning a new route or manipulation for reward | Vertebrates, arthropods, cephalopods |
| Unlimited associative learning (UAL) | Value of a whole compound object, plus all steps of a complex response, open-endedly | Rat finds lever, reaches it, learns how hard and when to press | Same three groups |
Operant learning is the more demanding: it needs trial and error, more brainwork and memory. In classical conditioning the unconditioned stimulus is a reliable reminder (the bell always means food), so learning is easy. Operant conditioning is therefore more likely to rest on learned valences and memories of true affective states. The authors further insist on global operant conditioning, beyond adjusting one limb or switching between rigidly preprogrammed behaviors.
The UAL proposal and the authors' reservation
Zohar Bronfman, Simona Ginsburg and Eva Jablonka (2016) proposed unlimited associative learning as the marker of the evolutionary transition to minimal consciousness. Simple operant learning may fix on a single feature of an object; UAL learns the whole object, so the animal is never fooled by a pebble that looks like a food pellet (appearance, weight and smell are bound together).
Bronfman, Ginsburg and Jablonka. UAL is the transition marker: when it appears, minimal consciousness has arrived.
Feinberg and Mallatt. UAL is one good marker and a major driver of evolution, but it is too advanced to be the marker. Affects could exist with more limited global operant learning; UAL belongs to brains that already had minimal consciousness.
In practice both teams reach the same answer: global operant learning and UAL occur only in vertebrates, arthropods and cephalopod mollusks. All vertebrate groups pass criterion 1, and the rest follow: trade-offs are documented in all vertebrate groups (more food versus higher predation risk), frustration in fish, birds and mammals, and self-delivery and drug approach / conditioned place preference in all groups.
Is affect cortical or subcortical?
Experts split. One camp (Barrett, LeDoux, Rolls, Craig and others) puts conscious emotion in the cerebral cortex; the other (Damasio, Denton, Panksepp, Solms and others) says the cortex only adjusts or inhibits raw emotions generated subcortically. The authors side with the subcortical view, on three lines of evidence.
Hydranencephaly (Merker, 2007; Aleman and Merker, 2014). In hydranencephaly, the developing cerebral cortex is almost entirely destroyed before birth by a loss of blood supply, while the subcortical brain survives. These children smile and laugh with pleasure and show aversion by fussing, arching the back and crying. A familiar adult can build play sequences from smiling to giggling to laughter and great excitement. In a survey by Aleman and Merker, a solid majority of primary caregivers reported emotional behavior that contradicts the common belief that the condition is a vegetative state.
Deep brain stimulation of the medial forebrain bundle (Panksepp). The medial forebrain bundle (MFB) is a fiber tract interconnecting many affective regions. Stimulating one part makes humans feel fear, rage or panic, and rats try to escape. Stimulating another part makes humans feel caring, playfulness, lust and curious exploration, and rats self-activate the stimulator. Same structure, same valences, across species.
The conclusion for mammals: every mammal has affective consciousness, generated by specific subcortical structures.
Other vertebrates: an anatomy-only test
No brain-stimulation studies exist for nonmammal vertebrates, so the authors tallied affective brain structures across groups. They are highly conserved.
- 25 affective brain structures were scored across vertebrates.
- Lampreys have at least 19 of 25.
- Bony fish have 23 of 25.
- All vertebrate groups, lampreys included, have a medial forebrain bundle.
How affective circuits differ from image circuits
With images, circuitry maps onto experience fairly obviously: topographic simulations of world and body. Affective circuitry is harder to read.
- No maps. It largely lacks topographic mapping.
- Spread out. Its centers are more numerous and dispersed, and they receive more branches from ascending axons (see the ulcer pathway in chapter 2).
- Branchy. Its axons branch far more, broadcasting to many parts of the system.
- Chemical broadcast. It relies more on far-diffusing neuromodulators (dopamine, serotonin, norepinephrine) and less on point-to-point neurotransmitters.
Valence neurons and valence maps
Recent rodent work makes affect concrete. Individual neurons encode distinct valences: one class carries only reward (+), another only aversion (−). They code their sign both in firing patterns and in the chemicals they release (dopamine signals anticipated reward). They have distinctive locations and targets, and in places form "valence maps", such as the pleasure hot spots of figure 4.2B.
- Thirst neurons (William E. Allen and colleagues, 2017). Mouse hypothalamic neurons encode thirst, a primordial emotion (with food hunger and air hunger) in Derek Denton's sense: affects that drive homeostasis. This confirms the close link between interoceptive and affective consciousness.
- Danger neurons (Carlos A. Campos and colleagues, 2018). Neurons in the parabrachial nucleus carry "danger" signals from skin pain, loud noises or a dangerously full stomach and trigger fear responses in centers like the amygdala.
Both cases show neurons coding affective information far richer than simple liking or disliking.
A diffuse system with specialists
| Center | Affective specialty |
|---|---|
| Amygdala | Fear and emotional learning |
| Lateral habenular nucleus | Punishment and disappointment |
| Nucleus accumbens | Motivated seeking of rewards |
| Anterior reticular formation (laterodorsal tegmental nucleus) | Arousal and intensity of emotion |
| Orbitofrontal cortex (and other cortex) | Reasoned control of core emotions and passions |
| Periaqueductal gray | Panic motor actions: fleeing, curling up, sweating |
Affect drives action
A newer finding: affective regions are tightly wired to the premotor regions that choose behavior. In mammals, affective signals pour into the nucleus accumbens, a motivation center in the basal ganglia (premotor for body movement), and into the hypothalamus (premotor for inner-body actions such as gut movements and heartbeat).
The principle shows even in sea slugs and snails. In Pleurobranchaea californica, sensory signals are assigned valence in an "integrator circuit for incentives" that tells central pattern generators (CPGs) to start or stop feeding and locomotion. Each circuit has only a few neurons and no multiple levels, probably too simple for true affect, but a "distant mirror" of the vertebrate arrangement.
Commonality within diversity
How affect differs from images. Valence-coding neurons, many neuromodulators, nontopographic organization, centers that are diffuse and numerous.
What affect shares with images. Hierarchies of pathways, input from many senses, centers for attention and memory.
Conscious states in vertebrates therefore trace to the corticothalamic system of mammals, the optic tectum of fish and amphibians (most distance senses), and the subcortical affective structures of all vertebrates (most affects). Consciousness in vertebrates comes in different functional and anatomical forms, both within one brain and across species, yet common factors run through all of them.
Behavioral criteria can be gamed by simple mechanisms. Conditioned place preference has been reported even in planarian flatworms, whose consciousness few would grant, and a skeptic can model each criterion with nonconscious reinforcement learning. The criteria are strongest in combination and alongside the anatomical evidence. The cortical camp (notably Joseph LeDoux, whose higher-order account of emotional consciousness with Richard Brown dates from 2017) still holds that subcortical circuits produce defensive survival states, not felt fear, so the hydranencephaly and DBS evidence remains contested.
The frustrated fish. A fish learns that a lever yields food; then the food stops. If it simply stopped pressing, a thermostat-like learner would explain it. If instead it attacks a tank-mate right after the denied reward, something negative has persisted beyond the stimulus. The thought experiment shows why the authors prize extended and remembered affect: a feeling that lingers after its cause is hard to explain as mere input-output reflex.
The Question of Invertebrate Consciousness
Applying the vertebrate-derived criteria to invertebrates, Feinberg and Mallatt find that arthropods (insects, crabs, spiders) and cephalopods (octopuses, squid, cuttlefish) pass, while roundworms, flatworms and most mollusks do not. Two bee experiments carry the case for insects: one for mental images, one for positive affect via the judgment bias test. Because the three conscious groups are only distantly related, consciousness evolved independently in each: convergent evolution, or in philosophers' terms multiple realizability.
Why invertebrates matter to the argument
A growing number of researchers argue that some invertebrates are conscious. For a naturalized theory, this is valuable data: if very different brains produce consciousness, the search for common factors gets more comparison points and the set of possible neural substrates widens.
The candidates are chosen by elimination. Amphioxus and tunicates already failed (chapter 3). Most other invertebrates are protostomes: roundworms, flatworms, earthworms, mollusks, arthropods, brachiopods and others. All have a nervous system and many have a brain. Among them, arthropods and cephalopod mollusks have the most complex brains, senses and behaviors, so they get tested.
Arthropods and cephalopods against the criteria
- Affective criteria (box 4.1). Arthropods, especially insects, have been shown to meet all five: global operant learning, trade-offs, frustration, self-delivery of analgesics or rewards, and drug approach or place preference. Cephalopods, though "quirky and sometimes uncooperative" subjects, have met every criterion for which they have been tested.
- Image criteria. The vertebrate-derived checklist: elaborate distance senses (vision, hearing, smell), complex chains of sensory neurons, mapped representations, circuits for attention and memory. Insects have all of these, including image-forming compound eyes; a fly brain even holds a representation of where the fly is heading, and fly touch and taste pathways are topographically mapped. Cephalopod brains fit every neural criterion too.
- Insect brains: about 100,000 to 1 million neurons.
- Vertebrate brains: about 10 million to 85 billion neurons.
- Cephalopod brains: about 50 million neurons, the largest of any invertebrate, and as large relative to body size as a reptile's.
Two caveats apply to cephalopods: their brain circuitry is understudied, and much of their nervous system and many behavioral programs sit out in the arms, not the brain. For insects, the worry is small brain size: are there enough neurons for conscious processing?
The animals that fail
Roundworms (nematodes). Well-studied nematodes such as C. elegans (302 neurons) show classical learning, little global operant learning and no UAL, so no affective consciousness. For images they also fail: they combine touch, chemical and light inputs but lack elaborate distance senses, and their memory is limited. When a nematode loses a scent trail, it searches systematically but blindly to find it again. Conscious animals forage in a directed way that implies a spatial plan in mind. Roundworms have no blueprint of where they are.
Most mollusks other than cephalopods have little or no brain. Gastropods (snails, slugs, sea hares) have brain-like head ganglia. The best-studied, the sea hare Aplysia and sea slug Pleurobranchaea, have simple ganglia and senses and only simple, nonglobal operant learning, so primary consciousness is doubtful. A few gastropods may sit on the cusp: the great pond snail Lymnaea has image-forming eyes, and the land snail Helix shows second-order learning close to the UAL criterion.
| Group | Distance senses | Brain | Global operant / UAL | Verdict |
|---|---|---|---|---|
| Vertebrates | Camera eyes, ears, nose, electro- and lateral line | 10 million to 85 billion neurons | Yes | Conscious |
| Arthropods | Compound eyes, antennae, hearing in some | 100,000 to 1 million | Yes | Conscious |
| Cephalopods | Camera eyes, chemo- and touch senses | About 50 million | Yes, where tested | Conscious |
| Gastropods | Simple; some image eyes (Lymnaea) | Head ganglia | Simple operant; second-order (Helix) | Doubtful / cusp |
| Roundworms | Touch, chemical, light only | 302 neurons (C. elegans) | Classical only | Not conscious |
| Amphioxus, tunicates | Light and touch receptors | Hundreds of neurons | Not shown | Not conscious |
Octopus anecdotes
Cephalopods are "the darlings of the animal consciousness field." Two books summarize the case: Sy Montgomery's The Soul of an Octopus and Peter Godfrey-Smith's Other Minds, alongside studies by Jennifer Mather and Binyamin Hochner.
- Recognizing people. Octopuses appear to recognize individual humans (image-based) and take likes or dislikes to them, squirting disliked people with water from the siphon (affective).
- Play. At the Seattle Aquarium, octopuses have been seen playing with plastic bottles in a water jet; some researchers treat play as a sign of the "joy" of positive feeling.
Arthropods: controversial but defensible
Arthropod consciousness is much more controversial because of their small brains. Debates (many in the journal Animal Sentience) ask whether some arthropods are conscious and others not, and whether all must share the same kind of consciousness. The authors think all classes of arthropods are conscious because their brains share all the same parts, and fossils show these basic brain structures have been shared since the first arthropods over half a billion years ago. For the argument, though, one conscious arthropod is enough.
Bee mental images (Fauria, Colborn and Collett, 2000). A bee flies from a "nest" chamber down a long box to food behind a hole. It has been trained on which of four combinations of a striped pattern plus bull's-eye around the hole means food. In the test, the stripes are moved onto a gate that blocks the view of the bull's-eyes, so the two cues are separated in time and space. The bees still chose correctly, which the authors read as recording and recalling images of both patterns. Objection: perhaps this is just implicit associative learning. Reply: the compound pattern is complex and then unexpectedly split, so only a remembered, detailed image would solve it.
Bee positive affect (Perry, Baciadonna and Chittka, 2016). The judgment bias test, long used for emotion in humans and mammals, had never been applied to a nonmammal. Its premise: "subjects in a positive emotion state tend to respond to ambiguous stimuli as though predicting the positive event." Bees learned that blue marks a cylinder with sugar and green marks one with plain water. In the test, some bees got a small taste of sugar in the entry tunnel before meeting an ambiguous blue-green flag; controls got none. Sugar-primed bees approached the ambiguous flag more readily. The authors of the study cautiously said "emotion-like" state changes; Feinberg and Mallatt read it as positive affect, even a few moments of good mood.
Convergence and multiple realizability
The verdict: most invertebrates lack consciousness, but arthropods and cephalopods have it, as do vertebrates. Vertebrates are only distantly related to protostomes, and arthropods and mollusks are distant from each other within the protostomes. So the three groups' brains and consciousness evolved independently: biologists call this convergent evolution; philosophers call it multiple realizability, many ways to build the same thing.
Any animal with elaborate distance senses and a sufficiently complex, hierarchically organized brain, whatever its lineage, appears capable of consciousness. The three conscious groups are also the most active animals on the planet, moving through complex space with the richest sensory kits, so consciousness correlates with good locomotion and wayfinding ability.
Gaps in the invertebrate case. The "all arthropods" claim rests mainly on insects and shared brain parts, not on tests across crustaceans, spiders and myriapods. The bee image study can still be modeled by sophisticated associative learning, as the authors concede in a note. The book cites only two bee experiments in detail. And the cephalopod evidence leans on anecdote (play, squirting disliked keepers) more than on controlled tests. The nautilus, a cephalopod with a simpler brain, is flagged as possibly not conscious, which shows the group-level verdicts hide variation.
Policy and evidence have moved sharply toward the authors' position.
- 2021 LSE review. Jonathan Birch and colleagues, commissioned by the UK government, reviewed over 300 studies and concluded that all cephalopod mollusks and decapod crustaceans (crabs, lobsters, crayfish, shrimps) should be regarded as sentient. They used eight criteria, several close to box 4.1: nociceptors; integrative brain regions; connections between them; responses changed by anesthetics or analgesics; motivational trade-offs; flexible, wound-directed self-protection; associative learning beyond habituation and sensitization; and valuing analgesics when injured.
- 2022 law. The UK Animal Welfare (Sentience) Act 2022 recognizes all vertebrates, cephalopods and decapods as sentient for policy purposes.
- 2024 New York Declaration on Animal Consciousness. Signed by hundreds of researchers, it finds at least a realistic possibility of conscious experience in all vertebrates and many invertebrates, including at minimum cephalopods, decapod crustaceans and insects, and says it is irresponsible to ignore that possibility in welfare decisions.
- Bees again. The Chittka lab has since reported cross-modal object recognition in bumblebees (recognizing by sight objects only touched before, Science, 2020) and play-like ball rolling (2022). The 2016 judgment bias effect was itself blocked by a dopamine antagonist, tying it to reward chemistry.
- Connectomes. In 2024 the complete wiring diagram of an adult fruit fly brain (about 140,000 neurons) was published by the FlyWire consortium, making insect circuits for attention, spatial orientation and valence far easier to study.
The octopus arm. About two-thirds of an octopus's neurons lie in its arms, which can taste, grasp and coordinate with little input from the brain. If one arm explores a crevice while the central brain attends elsewhere, is there something it is like to be that arm? The case presses the authors' framework: their features (hierarchy, mapping, convergence, attention) are defined for centralized brains, and a distributed nervous system tests whether consciousness must have a single center.
Creating Consciousness: The General and Special Features
This is the book's engine room. Feinberg and Mallatt assemble everything shared by conscious brains into three levels of features: the general biological features of all life (box 6.1), the neuronal reflexes and core brain that are still nonconscious (box 6.2), and the special neurobiological features found only in conscious animals (box 6.3). Stacked in that order, the levels form an uninterrupted progression from living cell to experiencing brain, with no new physics and no mysterious leap anywhere along the way.
The starting point: diversity forces a feature list
Some scholars confine consciousness to humans, mammals, or mammals and birds, and tie its neural correlates to an advanced cerebral cortex and thalamus. Chapters 3 to 5 found instead that consciousness is diverse within brains (images versus affects) and across species (tectum versus cortex; vertebrates, arthropods, cephalopods). So the cortex is not required; less elaborate structures suffice.
If the substrate varies so much, the only way to explain consciousness naturally is to find what all the substrates share. The chapter catalogs those shared ingredients at three levels, and chapter 8 will use them to explain subjectivity.
Level 1: general biological features (box 6.1)
All life has features absent from inanimate nature, and each one carries forward into consciousness.
- Embodiment. Every living thing, even a single cell, has an interior separated from the exterior by a boundary. Life depends on a bounded body; consciousness on a bounded brain. So the features of both are personal and unique to the individual: the personal nature of consciousness is already rooted in the personal nature of life (a point taken from Evan Thompson, and central to chapter 8).
- Process. Life and consciousness are processes, not things. Biology calls life's processes physiology; "consciousness," "mind" and "subjective experience" name processes the brain performs. Even the nouns "primary consciousness" and "qualia" mislead slightly: they are verbs, not nouns (an idea traced to William James, 1904). Any theory of neural correlates must therefore be functional, grounded in neuroanatomy.
- System and hierarchy. Living things are systems of organized parts whose functions depend on hierarchical structure. Mayr called biology's nested series a constitutive hierarchy: atoms, molecules, macromolecules, organelles, cells, tissues, organs. The adrenal gland is the book's example: an organ made of tissues made of cells made of organelles.
- Emergence. When higher levels are added and interact with lower ones, the whole gains new or even novel (never-before-existing) features: atoms are not alive, but life is a nonmysterious emergent property of cells. The authors, with Searle (his "emergent1"), apply emergence to brains as to all nature, noting only that brains have more levels and more kinds of new features.
- Teleonomy and adaptation. Organisms perform programmed, goal-directed functions, mostly set by genes (teleonomy, a term from Jacques Monod, defined via Mayr). An adaptation is a goal-meeting function or structure evolved by natural selection, not predesigned and not consciously willed.
Life, embodiment and process - Life: using energy to sustain self, growth, responsiveness, reproduction, and adaptiveness to change. All known life is cellular. - Embodiment: a body with an interior separated from the exterior by a boundary. - Process: life functions are complex, dynamic processes, not material things.
System and self-organization - System: an entity considered as a whole, in which the arrangements and interactions of parts matter. - Self-organization: interactions of the parts organize the patterns at the global level of the whole system.
Hierarchies - Hierarchy: a complex system of interacting levels from simpler to more complex, which may be nested (macromolecules to cells to organs to organism). New emergent features may appear naturally as new levels are added and interact with lower ones.
Teleonomy and adaptation - Teleonomy: biological structures perform programmed, goal-directed functions. - Adaptation: a teleonomic structure or function evolved by natural selection.
These features are necessary but not sufficient for consciousness. Their importance, as Searle also saw, is that consciousness cannot be understood outside the rest of biology.
Level 2: neurons, reflexes and core brains (box 6.2)
Reflexes
The next step needs a multicellular body with specialized cells, including neurons for rapid communication. The simplest arrangement is the reflex arc: sensory, motor and usually interneurons. Reflexes are not conscious, and animals with only reflexes are not conscious, but reflex arcs are the building blocks of consciousness's substrate. They need only basic connections, as in the nerve net of a brainless jellyfish. What reflexes add to cellular life is the ability to signal widely and quickly across a large body, coordinating distant parts.
Reflexes come in simple and complex forms. Adding interneurons and cross talk among them turns simple arcs into complex ones; adding more levels of sensory-processing interneurons, like links in a chain, yields a multilevel hierarchy. Even the most complex reflexes are not conscious, but the elaboration of reflexes was the royal road to consciousness in the first vertebrates and arthropods, and in cephalopods.
More advanced nonconscious animals also have basic motor programs from central pattern generators (CPGs): networks that fire rhythmically to produce repetitive actions. Sensory input can modify the rhythm but does not cause it.
The sea hare's feeding CPG (Jing and colleagues, 2004). In Aplysia, feeding runs on just two brain neurons, B63 and B64. Sensory neurons detect food and signal B63, which protracts the radula (a scraping, biting mouth structure) and opens the mouth. B63 then tells B64 to signal retraction. The two pass signals back and forth to keep the protraction–retraction cycle going. Complex, adaptive, rhythmic, and entirely nonconscious.
Core brains (still not conscious)
Mobile animals meet new stimuli head first, so receptors and processing concentrate there and brains evolve there. In the primitive brain, reflex arcs elaborate into circuits for basic survival. This core brain is the brain stem plus part of the diencephalon in vertebrates and includes the protocerebrum in arthropods; its location in cephalopods is unknown.
- Homeostasis. It adjusts respiratory, heart and digestive functions.
- Complex reflexes. Swallowing, vomiting, gasping, coughing and blood-pressure control.
- Rhythms. CPGs for breathing and for the rhythmic movements of a swimming fish or walking mammal.
- Arousal, motivation, attention. Most important for later consciousness, the core sets levels of arousal and motivation and directs attention to some stimuli over others, functions the authors assume it performed since its nonconscious days.
The vertebrate reticular formation belongs to the core: its neurons receive input from and project to most of the brain, so it can arouse the whole brain at once. The nonconscious amphioxus larva has an equivalent hub, a "core integrative center" where sensory neurons converge on the motor center; Lacalli reads it as a motivational center that sets how sensitive and how readily aroused the larva is. Insects evolved the same functions independently: clusters of widely projecting dopamine neurons in the fruit fly produce brain-wide arousal and influence locomotion, and the protocerebrum modulates sensory input and motor output by motivational state.
So animals at the cusp of consciousness, in both prevertebrate and prearthropod lines, had core brains for attention, arousal, homeostasis, locomotion and survival. Yet they were not aware of what their receptors sensed and felt no affect toward what they approached or avoided. Their behavior resembled the roundworm's: follow a trail, lose it, search blindly.
- Rates. Reflexes and all neural communication are extremely fast compared with other large-scale physiology, so they can move a large body in response to a stimulus.
- Connectivity. Simple reflex arcs are chains of several neurons joined at synapses. Complex arcs have more neurons in the Chain (C) and in Networks (N), more Sensory input (S), more neuronal Interactions (I) and more capacity to Process information (P).
- Increasing complexity. A further increase in CNSIP leads to complex nervous systems and consciousness.
- Basic motor programs from central pattern generators. These nonconsciously control essential, repetitive behaviors.
- Core brain features. Modulatory, sensorimotor-integrating centers for arousal and directing attention; complex reflexes for inner-body homeostasis; rhythmic locomotion and other basic motor programs. All automatic, not conscious.
Level 3: the special neurobiological features (box 6.3)
The special features are greatly enhanced in conscious nervous systems and absent or much simpler in nonconscious ones. They arise by normal evolution, yet they are the key to consciousness and subjectivity.
Neural complexity
All three conscious groups have complex brains with at least about 100,000 neurons and many neuron subtypes with distinct shapes and roles.
Elaborated sensory organs
Without detecting many stimulus types in detail, an animal cannot build an image of world or body, or aim affect-driven responses precisely. That requires image-forming eyes, a nose for many odors, ears for hearing and balance, taste receptors and proprioceptors, each a different modality. Vertebrates, arthropods and cephalopods have them all, and aquatic members add receptors for water vibrations from moving animals. High mobility goes with them, to gather abundant sensory information.
Neural hierarchies with unique neural–neural interactions
Neural hierarchies differ from biology's constitutive hierarchies in one structural way, and the difference matters.
| Property | Constitutive hierarchy (kidney, adrenal gland) | Neural hierarchy (sensory pathway) |
|---|---|---|
| Physical nesting | Lower levels contained in higher ones | Lower levels not contained (spinal cord is not inside brain stem, brain stem not inside tectum) |
| Interconnection | Mostly local | Extreme: synaptic chains link distant levels |
| Communication | Chemical, slow | Fast, reciprocal within and between levels |
| Result | Emergent features of the organ | Exponential growth of back-and-forth signaling and novel features |
The back-and-forth traffic is called reciprocal, reentrant or recurrent communication (after Victor Lamme), and it runs between the hierarchies for different senses too, so hearing informs vision. It may be coordinated by synchronized oscillations (brain waves), thought to bind multisensory information into one image. The authors stress integration, citing Giulio Tononi's integrated information theory: images need different levels and senses integrated into one map, and affects, though less hierarchical, show heavy cross-communication too.
Multilevel hierarchy pays off in three ways:
- More processing. Higher levels added to the core brain allow the complex processing consciousness demands.
- Unity, perceptual and motivational. Senses interact more, yielding a unified world image and motivational unity: one action at a time. The convergence hubs are the optic tectum (fish, amphibians), cerebral cortex (mammals), subcortical affective regions (all vertebrates), central complex (arthropods) and superior lobe (cephalopods).
- Prediction. Hierarchies let consciousness predict events a fraction of a second ahead, a huge survival edge when anticipating a predator's attack path or a prey's escape.
Pathways for mapped images or affective states
Images and affects share complex, hierarchical, reciprocal networks but differ in topographic mapping versus valence coding, and in vertebrates they occupy different but overlapping regions.
Attention
Attention can be conscious or unconscious. In vertebrates, conscious focusing uses the optic tectum and cerebrum, aided by the brain-stem reticular formation. In insects many brain parts are tied to selective attention, so it may be more distributed. Cephalopods are plainly highly attentive, but their attention centers are unstudied.
Sensory memory
Memory capacity must have jumped when consciousness first evolved. Two lines of evidence: learning expanded from sensitization, habituation and simple association to global operant learning, and fish and arthropods react to objects as if recognizing them from recall.
Two questions follow: what minimal memory does consciousness need, and did the Cambrian memory boom come before, during or after consciousness? Christof Koch reasoned that image consciousness needs at least iconic memory; the authors prefer the umbrella term sensory memory.
| Sensory memory type | Modality | Approximate duration |
|---|---|---|
| Iconic | Vision | Under half a second |
| Echoic | Hearing | About 3 seconds |
| Tactile | Touch | About 10 seconds |
| Odor | Smell | Longer |
For Koch, sensory memory provides continuity of experience in an unbroken stream and makes brief stimuli last long enough to reach awareness. It arises from reciprocal signals reverberating up and down a hierarchy before fading, and it entails attention, so attention and sensory memory coevolved.
The authors hypothesize that longer, higher-capacity memory came soon after consciousness, using the hippocampus (vertebrates), mushroom bodies (insects) and much of the supraesophageal brain (cephalopods). They then hedge: their own vision-first theory implies that recording detailed camera-eye images instant to instant would already need considerable memory, and selection for recalling a predator, mate or food item, with its look and its felt affect, would have been immediate. Either way, memory has always contributed to consciousness.
Neural complexity (more than in the simpler core brain) - A brain with many neurons; many subtypes of neurons.
Elaborated sensory organs - Image-forming eyes; multiple mechanoreceptors for touch; separate chemoreceptors for smell and taste. - High locomotory mobility, to gather abundant sensory information.
Neural hierarchies with unique neural–neural interactions - Extensive reciprocal (reentrant, recurrent) communication within and between the hierarchies for different senses. - Synchronized brain-wave oscillations may be needed for sensory binding and generating images. - Higher levels allow complex processing and the unity of consciousness. - Hierarchies let consciousness predict events a fraction of a second ahead.
Neural pathways that create mapped mental images or affective states - Isomorphic representations: neurons in topographic maps of world or body. - Affective states: from valence coding, not mapping; more diffuse hierarchies and networks, more centers, more neuromodulators. - Both feed premotor regions to motivate and guide behavior in space.
Attention as a participant in consciousness - Selective attention focuses consciousness on salient objects; the related feature of arousal adjusts the level of consciousness.
Memory - Short-term sensory memory for continuity of experience. Longer-term, higher-capacity memory evolved soon after consciousness arose, though more memory from the start cannot be ruled out.
Summary: three levels, no gap
All vertebrates have every general feature, reflex, core-brain function and special feature required for both images and affects, and arthropods and cephalopods match them feature by feature. Instead of a mysterious leap, there is a stepwise accretion: life's building blocks, then neurons and reflexes, then novel neural features.
The checklist does double duty. It is a theory (these features jointly create consciousness) and a test kit (check any newly discovered animal against box 6.3). Because every item is an ordinary evolved trait, the explanation stays entirely natural, which is why the authors call it neurobiological naturalism.
Necessary, sufficient, or just correlated? The box 6.3 features are derived from animals already judged conscious, then used to judge which animals are conscious, so the method risks circularity unless behavioral tests stay independent. The list also gives no threshold: how much hierarchy, how many neurons (the 100,000 figure is a rough floor from insects), how much recurrence? The authors use IIT only for its emphasis on integration, without its mathematics or its controversial implications. And the notes concede that the classic role of gamma-band synchrony in binding has been challenged (Koch and colleagues, 2016).
The theory-testing scene has changed since 2018. An adversarial collaboration (the Cogitate Consortium) pitted integrated information theory (IIT) against global neuronal workspace theory (GNW) with preregistered predictions; results announced in 2023 and published in Nature in 2025 challenged key predictions of both: there was no sustained synchronization within posterior cortex as IIT predicted, and no "ignition" at stimulus offset with limited prefrontal representation of some content, against GNW. In September 2023, an open letter signed by over 100 researchers called IIT pseudoscience, provoking a sharp counter-reaction and a still-running debate about what counts as testable. Predictive processing, which the authors invoke for figure 6.5, has grown into a leading framework (Clark, Friston, Seth). None of this settles the authors' multifactor, life-based view, but it shows that single-mechanism theories have not won.
For AI architects the three-level stack reads like a capability model: substrate (life), fast reflex layer (rules and pattern generators), then a world-model layer with multimodal fusion, attention, short-term memory and prediction-error correction. Modern agent designs (perception encoders, a shared latent state, planners, memory stores) recreate several box 6.3 functions. On the authors' view they still lack Level 1: they are not alive, which chapter 8 argues matters.
The 99,999-neuron insect. Imagine removing neurons from a bee one at a time while it keeps foraging. At no single step does it switch from conscious to nonconscious; capacities degrade gradually. The thought experiment shows that box 6.3 describes graded properties, so "consciousness evolved" may mean a ramp, not a switch, which fits the authors' claim of no gap but sits awkwardly with their talk of a threshold to "cross."
The Evolution of Primary Consciousness and the Cambrian Hypothesis
Consciousness appeared during the Cambrian explosion, about 560 to 520 million years ago, in the first vertebrates and first arthropods, and later and independently in cephalopods. A predator–prey arms race flooded the seas with signals; the most active animals answered with keen distance senses (vision first) and brains that merged them into mapped images with affects. The fossil sequence runs from life (3.7 billion years ago) to multicellular animals, nervous systems and reflexes, core brains, then consciousness: continuous, but punctuated by one big jump.
The timeline
- 3.7 billion years ago First bacterium-like cell: a private body with a border to the outside world.
- 660 to 600 million years ago Cell colonies become multicellular animals; the colony, not the cell, is now the embodied organism.
- About 580 million years ago Nervous systems with reflexes and simple motor programs, attested by fossil worm trails.
- 580 to 560 million years ago Core brains for survival, arousal and homeostasis in prearthropods and prevertebrates.
- About 555 million years ago Ediacaran seafloor: simple worms with nerve nets graze microbial mats; leaf-like forms sway in currents.
- 541 million years ago The Cambrian Period begins (it lasts to 485); most diversification occurs in its first part, 541 to 518.
- About 540 million years ago First arthropods, attested by scratch marks.
- 560 to 520 million years ago Consciousness in the first arthropods and vertebrates (narrower estimate 540 to 520).
- 520 million years ago Haikouichthys, oldest known fossil vertebrate (China), with camera eyes.
- About 490 million years ago First cephalopods, after the main explosion.
- About 275 million years ago The octopus–cuttlefish–squid group appears, the only cephalopods with tested consciousness.
Before the explosion: a quiet seafloor
Just before the Cambrian, the most advanced animals were simple ocean worms with a mostly diffuse nerve net, no brain and no complex sense organs. They fed on microbial mats, the scum carpeting shallow seafloors worldwide. Their trails, preserved in fossil mats, show simple behavior: following food or mates by sensing and smell. They had reached the reflex stage of box 6.2. Other animals were simpler still, leaf-like forms swaying in the current. Few animals interacted.
Within a few million years, those worms had given rise to the thirty-plus animal phyla, locked in heavy predator–prey competition, most with living descendants.
Why the explosion: the predation arms race
The most popular explanation is the origin of predators: one worm species began eating others, perhaps after an intermediate scavenging stage on decaying carcasses in the mats. Prey responded with hiding, armor, keener senses and faster locomotion; predators responded with better senses, locomotion and weapons. Only escalating lineages survived, and since larger animals attack and defend better, body size rose too.
The busier world sent out far more signals: sights, sound vibrations, smells, things to touch and taste. The two most active groups, early arthropods and vertebrates, exploited them by evolving keen eyes, ears, noses and touch. The early arthropods, with jointed manipulative limbs, included the main predators. The first vertebrates were harmless little fish that ate the remains of microbial mats and escaped by fast swimming.
To process the flood of input, reflexes, motor programs and core brains elaborated into processors that merged all senses into conscious mental images, with affects, to drive actions in three-dimensional space. Brains and active lifestyles are energy-expensive, so most other new groups took cheaper strategies (burrowing, shells) and did not evolve consciousness.
Haikouichthys, the first known vertebrate (about 520 million years ago). An inch-long jawless fish from Chinese Cambrian rocks, with elaborate eyes, ears, a nose and a streamlined body for swimming. It could not have been a predator: any tentacles over the mouth were soft, and it had no teeth, claws or grasping arms. Consciousness in vertebrates, on this evidence, began in prey, not hunters.
Cephalopods: late and separate
Cephalopod consciousness also came with predation, but later. Fossils put the first cephalopods at about 490 million years ago, after the bang of 540 to 520. As mollusks they descend from a sluggish, heavily shelled ancestor and switched to hunting once they evolved tentacle arms and jet propulsion. Even the earliest cephalopods may not have been conscious: all the evidence comes from the octopus–cuttlefish–squid group, about 275 million years old. The authors even wonder whether the nautilus, with its simpler brain and eyes, is conscious at all.
| Lineage | When consciousness arose | Ancestor | Driver | Leading sense |
|---|---|---|---|---|
| Vertebrates | About 540 to 520 million years ago | Worm-like chordate (amphioxus-like) | Escaping predators as small, fast prey | Camera-eye vision, then other senses |
| Arthropods | About 540 to 520 million years ago | Lobopodian worm (velvet-worm relatives) | Predation with jointed limbs | Compound-eye vision, with smell early |
| Cephalopods | After 490 million years ago (tested only in a group about 275 million years old) | Shelled, inactive mollusk | Secondary switch to active hunting | Large camera eyes and visual centers |
Step by step to images
The earliest sensory responses were reflexive and innate, run by simple motor programs: the roundworm's withdrawal circuit and the sea hare's feeding CPG. Those circuits can be described entirely objectively, with no explanatory gap. In the Cambrian, added brain complexity and the special features turned reflex-level processing into images and affects, and with them came the aspects of subjectivity: experiences that are private, referred, unified, mentally causal and qualitative, carrying all four gaps.
Vision first
The authors hold that vision was the first sense to elaborate in vertebrates, for three reasons:
- The fossil. The oldest known fossil fish already has a camera eye.
- Information content. Only vision gives the positions, borders and movements of many distinct objects at the greatest distances, and which object lies in front of another.
- The relatives. Rival claims that smell came first fail for vertebrates, because their closest relatives (amphioxus and tunicates) have no obvious smell organs but do have eye rudiments.
The first vertebrate images were visual, then other senses were added into a multisensory map in the optic tectum. The distance senses of the first fish can be read off living lampreys, sharks and bony fish: detailed vision, hearing, smell, taste, plus electroreception and lateral-line mechanoreceptors that detect movements by water vibrations.
In arthropods, the nearest living relatives are the velvet worms (Onychophora) of the southern continents, kin to the Cambrian lobopodians. Their brain matches an arthropod's except for the large visual centers; they have small eye cups, not compound eyes. So vision exploded at the dawn of arthropods too, though velvet worms' good smell centers suggest smell mattered earlier in that line. In cephalopods, the brain grew from small cerebral ganglia like those of gastropods; the steps are hard to reconstruct, but the eyes and visual centers dwarf those of any other mollusk.
Step by step to affects
Affect is easiest to trace in vertebrates.
- Ancestral worm. A simple integrator circuit for incentives, like the modern sea slug's: assign + or − valence, tell the CPGs to start or stop.
- First vertebrates. With global operant learning, valences became conscious and then more varied: negative affect gained intense fear alongside dislike; positive affect split into food joy and sexual lust.
- Distribution. The single circuit split into multiple specialized centers across the brain, still interconnected and wired to premotor areas (basal ganglia, hypothalamus). Arthropod brains show the same pattern.
- Early peak. Complexity peaked early: lampreys have most affective structures humans have, and fossil brains of the earliest arthropods match living ones.
The adaptive value of consciousness
Why was consciousness worth its energy cost? The central claim is efficiency. A unified stage of images and affects organizes huge sensory input for action choice with far fewer neurons than prespecifying a response for every stimulus. No brain could hold that many prespecified reflex arcs (a point the ethologist Donald Griffin made), and each would be too rigid. Multisensory maps also resolve ambiguities between conflicting senses.
Items in the image are ranked by importance: the most important get the most attention and the strongest affect, and these decide which motor program runs (rich food triggers feeding, a predator triggers flight). Nonconscious animals relying on reflexes cannot process as much, have no map of where they are, and cannot perform guided or proactive behavior. Conscious animals can move almost anywhere and manipulate objects in jam-packed 3D habitats.
- Efficient organization of much sensory input into a set of diverse qualia for action choice, resolving conflicts among inputs.
- A unified simulation of a complex environment that directs behavior in three-dimensional space.
- Importance ranking of stimuli by assigned affects, which makes decisions easier.
- Flexible behavior.
- Much and flexible learning (UAL).
- Prediction of the near future, allowing error correction.
- Coping well with new situations.
Flexibility has two routes. Mapped images can be reprioritized by moving attention and reassigning affect from one item to another. Pure affective consciousness works even on unmapped or simple stimuli: a sudden threatening noise or smell of decay switches curious approach (+) into prolonged, directed flight (−).
Consciousness has a smooth but punctuated history: continuous steps from the first cell to reflexes to complex brains, with a burst of special features during the Cambrian arms race, and no gap in the time line.
Checks on the Cambrian story. The "all thirty-plus phyla" claim overstates: several phyla (bryozoans among them) have no secure Cambrian fossils, and soft-bodied groups barely fossilize. Molecular clocks place the origins of animal groups earlier than fossils do, which the authors acknowledge in a note; if nervous systems and even brains are older, the timing of the conscious transition could shift back. The predation trigger is the most popular, not the only, explanation (rising oxygen and ecological feedbacks are also proposed). And the inference "complex senses and brains, therefore conscious" is applied to fossils, where only anatomy, not behavior, is available.
Some dates have been revised. The International Commission on Stratigraphy now places the base of the Cambrian at about 538.8 million years ago, not 541. The 3.7-billion-year-old Isua "stromatolites" the authors cite were challenged in 2018 by Abigail Allwood and colleagues, who argued the structures are deformation features rather than biological. A 2021 report of possible cephalopod shells about 522 million years old from Newfoundland would, if confirmed, push cephalopod origins into the early Cambrian; it remains disputed. Simona Ginsburg and Eva Jablonka's The Evolution of the Sensitive Soul (2019) independently placed the emergence of minimal consciousness, marked by unlimited associative learning, in the Cambrian, so the core hypothesis now has two converging lines of argument.
Replaying the Cambrian without predators. Suppose the first worm-eating worm never evolved and the seas stayed a quiet mat-grazing world for another hundred million years. On the authors' logic there would be no signal flood, no arms race, no pressure for distance senses, and so no conscious animals; consciousness is not inevitable but an answer to a particular information problem. The counterfactual shows how strongly the theory ties experience to ecology: consciousness is what a body needs when the world becomes too busy for reflexes.
The adaptive-value argument is an architecture argument: a single integrated world model with ranked salience beats an exploding table of hard-coded stimulus–response rules. Enterprise systems hit the same wall: rule engines with thousands of special cases become brittle, while a shared representation plus a prioritization layer scales to novel situations. The Cambrian lesson is that the move is worth its cost only when the environment becomes complex and adversarial.
Naturalizing Subjectivity
The final chapter assembles the model: life, then reflexes and core brain, then the special features, then images and affects, and on top the four gap features. Each gap (referral, unity, mental causation, qualia) is traced down to roots in life and reflexes, and the first-person/third-person barrier is explained by two irreducibilities: a subject cannot experience its own neurons (auto-ontological), and an observer cannot experience the subject's experience (allo-ontological). The hard problem, the authors conclude, is really three different problems wrongly fused into one.
The whole model on one page
The three principles from chapter 1 return: consciousness is built on life; it adds reflexes, core brain and special neurobiological features; and its unique nature creates the divergence between subjective and objective viewpoints. The gradual accretion of novel properties over millions of years, from nonlife to life to reflexes to consciousness, created both consciousness and its gaps.
Life is fundamental
The first rung: consciousness is inextricably and irreducibly part of the physical life of the organism. The support is the list of shared features from chapter 6: life and consciousness are both embodied, both processes, both system features of the animal, and both the result of a system's parts and their hierarchical interactions.
One consequence carries the argument. Because life and consciousness are system features of embodied organisms, all their features are personal and unique to the living individual. So the subjectivity of later consciousness was already inherent at the level of the general features, and any explanation of the gaps must start there.
Reflexes and core brain: the bridge
Reflexes and core-brain functions are not conscious, but they link life to consciousness. They add specialized neurons, neural coordination of a whole multicellular body, speed, circuits, and the start of neurohierarchical complexity (the amphioxus level). Building consciousness on reflexes also predetermines the gaps of referral, unity and mental causation, as the next section shows.
The special features: unique, but diverse and unlocalizable
The special features (box 6.3) bring the explosion of senses and neuron types, new processing subsystems, more hierarchical levels, more multisensory integration, more reciprocal communication, more effective attention and more memory. From these arise system properties of complex brains, just as life arises from interacting cellular parts, with no new "fundamental properties" of physics (the authors' explicit contrast with Chalmers, who proposed taking consciousness as fundamental).
The basis is remarkably varied, within brains and across vertebrates, arthropods and cephalopods. The reticular activating system and thalamus, needed for attention, alertness, arousal and wakefulness, are spread so widely that the neural basis of consciousness, perhaps the broadest of all neural functions, will be at best poorly localizable.
Filling the four gaps from below
| Gap feature | Root in life (general features) | Reinforced by reflexes | Completed by special features |
|---|---|---|---|
| Referral | Embodiment sets up an inside (organism) versus outside (environment) relation | Reflexes direct responses outward: withdraw from or push against a harmful stimulus | Isomorphic maps refer images to world and body in exquisite detail |
| Unity | All physiology is integrated for homeostasis | Reflexes are prewired linked programs for unified actions | Reciprocal interactions bind sensory codes into one image or affect |
| Mental causation | Every cell acts on its environment (moves, feeds, excretes) | Reflexive actions change the world | Conscious action is still a body acting: I can move only my own arm with my thoughts |
| Qualia | Qualia are living processes of an integrated system | (Built on reflex circuitry) | Distinct special-feature architectures make distinct feels; qualia also depend on referral and unity |
The authors call mental causation the easiest gap to trace to the first life: a living cell already causes effects on its surroundings, and conscious causation fits the same mold of an individual body acting on its own environment.
Qualia as living processes
Francis Crick and Christof Koch called qualia (the redness of red, the painfulness of pain) the most difficult aspect of consciousness and judged a head-on approach fruitless. Feinberg and Mallatt think the problem is complicated but not insurmountable, and argue by analogy.
The respiration analogy. Respiration is broadly distributed. At the macroscopic level it is lungs inhaling air and a heart pumping blood; at the subcellular level it is biochemistry in the mitochondria. The two levels look nothing alike, yet both are integrated parts of one living respiratory system. Likewise the brain's macroscopic activity looks nothing like synapses and neurons, yet both are constituents of the integrated system of the animal's life. So the experience of red, or of pain, is a living process: qualia are alive in the same way a cell, a heart or a person is alive.
The authors add a guardrail in a note: life is a prerequisite for animal qualia, not sufficient for them. Qualia also inherit two other gaps already explained: they are referred (to world and body) and unified (sounds, colors, tastes and affects bound by millions of neurons across many levels).
Auto- and allo-ontological irreducibility
The third principle explains why the subjective view and the objective view can never coincide. It has two reciprocal halves.
- Auto-ontological irreducibility (auto, self). One's consciousness never refers to the neurons that create it and cannot experience how its mental events are physically produced. As Gordon Globus put it, the nervous system has no sensory apparatus directed to its own structure.
- Allo-ontological irreducibility (allo, other). An outside observer cannot directly observe or measure a subject's experiences. Once neural states become private images, interoceptive states or affects, they can never be objectively observed or experienced by someone else.
The autocerebroscope (Herbert Feigl, 1967). Imagine an instrument that lets you watch your own neurons in real time while you see a red apple. You would watch spikes flicker on a screen while also seeing red. But your view of the screen is just another third-person observation, the same one a stranger would have; the felt red stays on one side of the barrier and the spikes on the other. The device shows that auto-irreducibility is not a matter of lacking data: even perfect self-observation turns your brain into an object, and objects are not experiences.
Why would evolution build brains blind to themselves? Because there is no adaptive advantage in conscious circuits attending to their own neurons. Survival depends on attention aimed at the world and the body. Neuron maintenance is handled nonconsciously by more basic physiology, especially the "babysitter" glial cells that monitor, nourish and support neurons. Self-monitoring consciousness would be wasteful duplication.
Two questions, not one
The authors' most original analytic claim is that the neurobiology of qualia and the subjectivity of qualia have different explanations, and that past theories went wrong by conflating them.
- Qualia are created by anatomically diverse, widespread structures, so their neurobiological basis has multiple causes (the special features of box 6.3).
- Subjectivity also has multiple causes: the "interiority" of life (embodiment), allo-irreducibility and auto-irreducibility.
- The causes in (1) are not the same as the causes in (2); they overlap but are not identical.
- So "how are qualia generated?" is answered by the special features, and "how can qualia be exclusively first person?" by life plus the two irreducibilities.
- Therefore the single question "how do neurons create subjective qualia?" is misguided: it seeks one answer when two are needed. Only the combination explains the subjectivity of qualia.
The character of experience
David Chalmers asks why red looks like this rather than like blue, or like the sound of a trumpet. The authors' answer: the neural pathways for color, for sound and for affect are enormously different, evolved to process different stimuli and direct different responses, so they should not, and could not, feel the same. The qualitative differentiation of neural states lies in the neural states themselves.
Red is therefore both neurobiologically unique and subjective, with no gap between its neural creation and its subjectivity. The puzzle is not solved by some direct link between retinal cones or color areas and the feeling of red, but by the full set of contributing factors: life's contribution to subjectivity, a brain complex enough (special features) and designed appropriately (general and special features) to produce qualia, all within the bounds of the two irreducibilities. For red to feel "red," all must be present.
Why is there consciousness at all?
Chalmers's hard problem asks why all this processing does not go on "in the dark," without any inner feel. The authors offer three replies.
- The life analogy. Asking why consciousness-producing processes do not run without consciousness is like asking Mayr why life-producing processes do not run without life, or why multiplying embryonic cells do not build a body without development. Such questions are tautological at best, absurd at worst. Life has no single cause and is not simply reducible to its parts; consciousness is likewise a natural feature of life, then reflexes, then complex brains, plus subjectivity, which embodiment and the irreducibilities explain.
- Living versus computational. Chalmers's "easy problems" (discriminating, categorizing, reacting, integrating information) are explainable by computational or neural mechanisms. The authors insist on separating purely computational mechanisms (computers, nonliving devices, and information-processing cognitive theories) from theories that invoke the biological and neural properties of a living brain. Their hypothesis: experience and qualia are living processes that cannot be explained solely by nonbiological computation.
- Life is necessary, not sufficient. Complex reflexes and core-brain functions are alive yet do operate in the dark. What brings processing "into the light" is the addition of the special features.
So the scientific question is not why consciousness fails to stay dark but how neurons and brains create subjective feeling, and the answer has many variables: life and reflexes, the special features, and the two irreducibilities, across multiple integrated hierarchical levels. There is no single reason.
The mystery persists because three different issues are fused into one: the diversity of consciousness, the impossibility of treating qualia as a single localizable thing, and the irreducible difference between first-person feeling and third-person explanation. Separate them and each becomes tractable.
The three-part diagnosis of the hard problem
- Consciousness is incredibly diverse. Within brains (images versus affects, smell versus vision; the different NSFCs) and across animals (vertebrates, cephalopods, arthropods). So there is no single "fundamental force," no shared "core-conscious neurocircuit," no "single bullet" theory: consciousness is multifactorial.
- Qualia are not "one thing." Treating qualia as a unitary entity matches the diverse neurobiology only partially and broadly; localizing qualia in the brain is impossible, as is ascribing them to one neural mechanism.
- The first-person/third-person difference is permanent but not mysterious. Once brains reach the needed complexity, the philosophical gap between observing brains and feeling (the "character" of experience) appears. This is Searle's ontological subjectivity. It does not block scientific explanation of the neurobiology; it means neurobiological and philosophical problems must not be conflated.
The three are related but distinct, have different solutions, and cannot be reduced to each other. Progress requires separating them.
Has the hard problem been solved or sidestepped? A Chalmers-style critic will say the life analogy begs the question: life was always explicable as function, whereas the hard problem asks why function is felt. Saying qualia "are living processes" relocates the mystery rather than dissolving it, and the claim that different pathways "could not feel the same" assumes that pathways feel at all. The life-is-necessary thesis is also a hypothesis, not a finding: the book gives no mechanism by which metabolism or embodiment contributes to feeling, beyond analogy, and it is the claim most sharply contested by functionalists. On the other side, the irreducibility analysis is a genuinely useful clarification: it explains why the first-person/third-person gap exists without making it supernatural, and it is compatible with many physicalist views.
The life–mind link has gained followers since 2018. Anil Seth's Being You (2021) argues that experience grows out of a living body's predictive self-regulation ("beast machine"), and Mark Solms (The Hidden Spring, 2021) roots consciousness in brain-stem affect tied to homeostasis, both close to Feinberg and Mallatt's first principle. The rise of large language models has made the computational question urgent: a 2023 report by Patrick Butlin, Robert Long and colleagues derived indicator properties from scientific theories (recurrence, global workspace, higher-order representation, predictive processing, agency and embodiment) and concluded that no current AI system was a strong candidate for consciousness, while finding no obvious technical barrier to building systems that satisfy the indicators. On the authors' view, any such system would still lack the living foundation they consider necessary.
For anyone building or governing AI in a regulated setting, the book offers a clean distinction: functional capability (discrimination, integration, prediction, flexible action) is one question, subjective experience another, and the evidence for each is different. Claims that a model "feels" or "understands" should be tested against specific indicators, not inferred from fluent behavior; and on biological-naturalist views like this one, no amount of computation alone would settle it.
Master Takeaways
Taken whole, the book is a reference architecture for consciousness: a layered stack (life, reflexes and core brain, special neural features) that any conscious animal implements, a dated build history (the Cambrian), and a diagnosis of why the first-person view resists third-person explanation. Its lasting contribution is to make the hard problem smaller by splitting it: the neurobiology of experience, the diversity of experience, and the subjectivity of experience are different questions with different answers.
The takeaways
- Consciousness is unique but natural, as life is. Mayr's lesson about life (special properties, no special force) is the template: neurobiological naturalism explains subjectivity with life plus evolved neural features, invoking no new physics.
- There are four gaps, not one. Referral, mental unity, mental causation and qualia (the NSFCs) each appear in exteroceptive, interoceptive and affective consciousness, so any theory must cover all twelve combinations.
- All vertebrates are conscious, and the cortex is optional. The midbrain optic tectum gives fish and amphibians mapped multisensory images, and conserved subcortical circuits give all vertebrates affects; even the lamprey passes.
- Consciousness evolved three times. Vertebrates, arthropods and cephalopods built it independently from different brains, a case of convergent evolution and multiple realizability that rules out any single "consciousness circuit."
- A checklist exists. Box 6.3 lists the special neurobiological features (neural complexity, elaborate senses, reciprocal hierarchies, mapping or valence coding, attention, memory) that mark consciousness in any animal.
- Subjectivity comes from two irreducibilities. A subject cannot experience its own neurons (auto-ontological) and no observer can experience the subject's experience (allo-ontological); together with embodiment, these explain the first-person/third-person barrier without mystery.
- Images and affects are different machines. Images use topographic maps and tight hierarchies; affects use valence neurons, diffuse centers and neuromodulators, yet both share hierarchies, multisensory input, attention and memory.
- Affect is detected by behavior, not maps. Five criteria: global operant learning, trade-offs, frustration, self-delivery of rewards or analgesics, drug approach or place preference; simple approach and avoidance never count.
- UAL is a marker, not the threshold. The authors accept unlimited associative learning as evidence but think minimal consciousness came with more limited global operant learning.
- Reflexes are the royal road. Nonconscious reflex arcs, central pattern generators and core brains (brain stem and diencephalon; insect protocerebrum) supplied the arousal, attention and homeostasis on which consciousness was built.
- Neural hierarchies are special. Unlike biology's nested constitutive hierarchies, neural levels are not physically contained in one another but are massively and reciprocally interconnected, which yields unity and prediction a fraction of a second ahead.
- Vision led the way. The camera eye, made possible by vertebrate-only placodes and neural crest, gave the first detailed images; other senses were added to the visual map.
- The Cambrian arms race made consciousness pay. Predators flooded the seas with signals; active animals that unified them into ranked, affect-tagged images out-competed reflex-bound ones, while most phyla took cheaper routes like burrowing and shells.
- Consciousness is efficient. A unified world model with importance ranking needs far fewer neurons than a reflex for every situation, and it gives flexibility, learning, prediction and novelty handling.
- Qualia are living processes. Experience is to neurons as respiration is to mitochondria: different levels of one integrated living system. Life is necessary but not sufficient.
- Different pathways feel different because they are different. The character of red versus C-sharp versus sadness lies in the neural architectures themselves, evolved for different stimuli and responses.
- Two questions hide in "how do neurons make subjective qualia?" How qualia are generated (special features) and why they are exclusively first person (life plus irreducibilities) need separate answers.
- The hard problem is three problems. Diversity of consciousness, the non-unitary nature of qualia, and the permanent first-person/third-person difference; conflating them is what keeps consciousness mysterious.
- Brains are blind to themselves for a reason. No adaptive advantage would come from conscious circuits monitoring their own neurons; glia handle maintenance, and attention points at world and body.
- Consciousness is poorly localizable. Its contributors (tectum, cortex, subcortical affect centers, reticular activating system, thalamus) are spread so widely that it may be the broadest of all neural functions.
Frameworks & mental models
| Name | What it says | Use it when |
|---|---|---|
| Neurobiological naturalism | Consciousness = life + reflexes/core brain + special neural features; no new physics | Framing any naturalistic account of experience |
| Three principles | Life; neural features; naturally explainable subjective–objective barrier | Organizing an argument about mind and brain |
| Three domains of primary consciousness | Exteroceptive, interoceptive, affective (plus proprioception); simplified to images vs affects | Classifying what kind of experience is at issue |
| Four NSFCs (box 2.1) | Referral, mental unity, mental causation, qualia | Checking whether a theory addresses every gap |
| Isomorphic mapping criterion | Mapped multisensory representations indicate mental images | Assessing image-based consciousness from anatomy |
| Five affect criteria (box 4.1) | Global operant learning, trade-offs, frustration, self-delivery, drug approach/place preference | Testing an animal for affective consciousness |
| Unlimited associative learning (UAL) | Learning whole-object value and all response steps, open-endedly | A stricter marker of consciousness (Bronfman, Ginsburg, Jablonka) |
| Judgment bias test | Positive mood biases response to ambiguous cues toward optimism | Measuring affect objectively in nonhuman animals |
| General biological features (box 6.1) | Life, embodiment, process, system, self-organization, hierarchy, teleonomy, adaptation | Grounding consciousness in biology |
| Reflexes and core brain (box 6.2) | Rates, connectivity (CNSIP), CPGs, core arousal/attention/homeostasis | Identifying the nonconscious foundation |
| Special neurobiological features (box 6.3) | Complexity, senses, reciprocal hierarchies, mapping/valence, attention, memory | Judging whether any animal is conscious |
| Constitutive vs neural hierarchy | Nested containment vs non-nested, reciprocally linked levels | Explaining why brains yield novel features |
| Hierarchical prediction | Predictions flow down, errors flow up, top level updates | Explaining the adaptive edge of consciousness |
| Cambrian hypothesis | Consciousness arose 560 to 520 mya in a predation arms race, vision first | Dating and explaining the origin of experience |
| Adaptive advantages (box 7.1) | Organizes input, unified 3D simulation, importance ranking, flexibility, learning, prediction, novelty | Arguing why consciousness was selected |
| Auto- and allo-ontological irreducibility | Subject cannot access own neurons; observer cannot access subject's experience | Explaining the first-person/third-person barrier |
| Three issues inside the hard problem | Diversity; qualia not one thing; permanent viewpoint difference | Breaking down "why is there experience at all?" |
Strengths. A rare biology-first treatment that takes subjectivity seriously without mysticism; strong comparative evidence across vertebrates and invertebrates; clear, testable checklists; and a genuinely clarifying split between the neurobiology and the subjectivity of qualia.
Blind spots. The feature list is drawn from animals already judged conscious, which risks circularity, and gives no threshold for how much of each feature is enough. The life-is-necessary thesis rests on analogy (respiration, embryology), not mechanism. The answer to the hard problem will not satisfy committed Chalmers-style critics, who see function-to-feeling as untouched. The tectal theory of fish consciousness and the subcortical theory of affect are minority or contested positions, and alternatives (higher-order, global workspace, cortical-emotion views) get little space. Human-specific phenomena such as report, metacognition and self-consciousness are out of scope by design.
Dated or revised. The Cambrian base is now dated to about 538.8 million years ago; the 3.7-billion-year Isua fossils were challenged in 2018; some phylum counts overstate Cambrian diversity.
How the field moved. Policy has caught up with the book's conclusions: the 2021 LSE review found decapod crustaceans and cephalopods sentient, the UK Animal Welfare (Sentience) Act 2022 followed, and the 2024 New York Declaration grants at least a realistic possibility of consciousness to all vertebrates and many invertebrates, including insects. The 2023 to 2025 IIT versus GNW adversarial collaboration undercut key predictions of both leading single-mechanism theories, which suits the book's multifactor stance, and the 2023 "pseudoscience" letter on IIT exposed deep disagreement about what makes a consciousness theory testable. LLMs have turned the authors' living-versus-computational distinction into a live policy question.
Who's who
| Name | Who | Why they matter here |
|---|---|---|
| Todd E. Feinberg | Neurologist and psychiatrist | Co-author; developed neurobiological naturalism and the auto/allo-irreducibility idea |
| Jon M. Mallatt | Evolutionary biologist and anatomist | Co-author; comparative neuroanatomy and the Cambrian timeline |
| Ernst Mayr | Evolutionary biologist | Life as unique yet natural; constitutive hierarchies; teleonomy |
| John Searle | Philosopher | Biological naturalism and ontological subjectivity, the book's starting point |
| Thomas Nagel | Philosopher | "Something it is like to be" a bat: the definition of primary consciousness |
| Joseph Levine | Philosopher | Coined the explanatory gap |
| David Chalmers | Philosopher | The hard problem, easy problems, and the character of experience |
| Antti Revonsuo | Philosopher and psychologist | Minimal criterion: any pattern of experience present-for-it |
| Gerald Edelman / Antonio Damasio | Neuroscientists | Primary and core consciousness as mental images; interoception and emotion |
| Jaak Panksepp | Affective neuroscientist | Subcortical emotion; deep brain stimulation of the medial forebrain bundle |
| Björn Merker | Neuroscientist | Hydranencephaly evidence for consciousness without cortex |
| A. D. Craig | Neuroanatomist | Pain, temperature and itch as interoceptive and homeostatic |
| Bronfman, Ginsburg, Jablonka | Theorists of evolution of mind | Unlimited associative learning as a marker of consciousness |
| Thurston Lacalli | Neurobiologist | Amphioxus nervous system; the "core brain" |
| Micheal Woodruff / Mor Ben-Tov | Neuroscientists | Archerfish attention and the fish tectum |
| Perry, Baciadonna, Chittka | Bee researchers | Judgment bias test showing positive affect in bees |
| Fauria, Colborn, Collett | Bee researchers | Experiment indicating mental images in bees |
| Peter Godfrey-Smith / Sy Montgomery | Philosopher / writer | Popular cases for octopus consciousness |
| Christof Koch / Francis Crick | Neuroscientists | Sensory memory requirement; qualia as the hardest problem |
| Giulio Tononi | Neuroscientist | Integrated information theory, cited for integration |
| Gordon Globus / Herbert Feigl | Psychiatrist / philosopher | The brain does not represent its own structure; the autocerebroscope |
Connect it
- Capability is not experience. The book separates the "easy" functions (discrimination, integration, prediction) from felt experience and argues that nonliving computation alone cannot supply the latter. For AI governance in banking, this supports treating model capability claims and sentience claims as different questions with different evidence, and resisting anthropomorphic readings of fluent output.
- A layered reference architecture. The three-level stack (substrate, fast reflex/rule layer, integrated world model with attention, memory and prediction-error correction) maps cleanly onto agentic AI design: perception encoders, a shared state, a salience or prioritization layer, memory stores and feedback loops.
- Rules versus models. Box 7.1's efficiency argument (one integrated model beats a reflex for every situation) is the same trade-off as rule engines versus learned models in fraud and screening systems: rules are cheap and transparent until the environment turns complex and adversarial, then a ranked, unified representation scales better.
- Checklists for contested properties. The authors' approach (derive indicator features from known cases, test new cases against them, flag circularity) is the method the 2023 AI-consciousness indicator report adopted, and a useful template for any assurance framework judging hard-to-observe properties such as model reasoning or fairness.