We Know So Much About the Brain. So Why Don't We Understand It?

In 1998, at a neuroscience conference in Germany, a neuroscientist placed a bet with a philosopher. Christof Koch, one of the world's foremost researchers on the biology of consciousness, wagered philosopher David Chalmers a case of fine wine. The terms: that within 25 years (by 2023), science would discover the neural mechanisms by which the brain produces consciousness. It seemed like a reasonable bet. Brain imaging technology had just taken a great leap forward. Neuroscience was advancing faster than almost any other discipline. And the brain, after all, is just biology.

Twenty-five years later, at a conference in New York City, Koch handed over the wine. Both scientists agreed publicly, on 23 June 2023 at the annual meeting of the Association for the Scientific Study of Consciousness, that it is an ongoing quest, and declared Chalmers the winner. Chalmers, accepting his prize with the magnanimity of a man who had always suspected he'd win, put it simply: "It's clear that things are not clear."

This is not a story of scientific failure. We know more about the brain today than at any point in human history. We have mapped its anatomy in extraordinary detail, catalogued its chemistry, decoded the genetics of its development, and developed tools to watch individual neurons fire in living animals. The problem is that none of this, taken alone or in aggregate, fully explains how the brain does what it most conspicuously does: think, feel, remember, decide, and know that it exists.

The gap between what we know about the brain and what we understand about the mind is one of the great divides in all of science. With this article, I attempt to explain exactly where that gap lies, and what scientists are doing, right now, to cross it.

The Transistor Problem: Why More Parts Doesn't Mean More Understanding

There is a useful analogy for the brain problem. Suppose you are handed a modern computer, a complex machine with billions of transistors, wires, capacitors, and chips. You are given unlimited time to study it. You map every transistor. You understand exactly how electricity flows between them. You can predict, given any pattern of current, what pattern will come next. You have, in a meaningful sense, a complete physical description of the machine.

And yet you still cannot explain how Photoshop emerges from that transistor grid. You cannot explain how a web browser arises from those billions of switching gates. The gap between the physical substrate and the emergent software is not bridged by more transistor knowledge, it requires a different level of description entirely.

This is, roughly speaking, where neuroscience stands with the brain. We have an increasingly complete account of the components and their local behavior. What we lack is an account of how those components, interacting in their billions, give rise to the software of the mind.

What We Actually Know Very Well

The complaint that "we don't understand the brain" can be misleading if it is read as saying we know almost nothing. The truth is nearly the opposite. In the domains of anatomy, cellular biology, biochemistry, and genetics, the brain is among the best-understood organs in the human body. Introductory biology textbooks are not wrong about the brain. They are simply describing one level of a multi-level problem.

What the table above reveals is not a uniformly mysterious organ, but a layered problem where our understanding degrades as the questions become more emergent, more systemic, and more experiential. The further we move from cells toward experience, the more our explanatory grip loosens.

The Memory Problem: Where Do Your Memories Actually Live?

Memory is perhaps the most deceptively understood topic in all of neuroscience. Ask a layperson where memories are stored, and many will point to the brain. Ask a neuroscientist the same question, and the honest answer is: we know a lot, and we are still surprised by what we find.

The basic framework has been clear since the mid-20th century, when the famous patient H.M. had his hippocampus surgically removed and lost the ability to form new declarative memories. This established the hippocampus as central to memory consolidation, the process by which short-term experiences become long-term memories. The hippocampus, it turns out, does not store memories permanently. It acts more like a transfer station, helping to consolidate new memories by replaying them to the cortex during sleep and quiet waking, after which the memories are gradually distributed across cortical networks for long-term storage.

The more recent and exciting chapter of memory research concerns engram cells, the specific neurons whose activity is necessary and sufficient to encode a given memory. The modern engram concept, revived and validated by Tonegawa's laboratory at MIT and many others, has shown that you can optogenetically reactivate a memory by stimulating the same small population of neurons that were active during the original experience. Research published in 2025 demonstrated that new learning produces synaptic potentiation specifically onto engram neurons in the basolateral amygdala, that this potentiation lasts at least seven days, is reversed by extinction, and its disruption impairs memory recall.

But here is where the remaining mysteries cluster:

The protein paradox. Long-term memories can persist for decades, sometimes for an entire lifetime. Yet the proteins that make up synapses turn over continuously, with most synaptic proteins being replaced within days to weeks. How a memory encoded in a protein configuration survives the complete replacement of those proteins is a deep and unresolved question. Several theories have been proposed: synaptic tagging, PKMζ-mediated maintenance, structural spine changes — but none has been conclusively established as the definitive mechanism of memory permanence.

The engram distribution problem. In the week following learning, engram cells form new synaptic connections with non-engram cells, and additional neurons are incorporated into the engram through excitatory synaptic plasticity. This means memory is not a static pattern but a dynamically evolving one, expanding its footprint across the brain over time. How this distributed, ever-shifting representation maintains its identity as a coherent memory is not understood.

The retrieval problem. How does a single cue, a smell, a fragment of a song, an unexpected emotional context, instantly retrieve an entire complex experience from decades ago, complete with details you did not know you remembered? The computational mechanism of associative retrieval across the brain's entire memory store remains elusive.

The Hardest Problem in Science: What Is Consciousness?

In 1994, David Chalmers gave a lecture that changed how scientists and philosophers talk about consciousness. In it, he drew a distinction between what he called the "easy problems" and the "hard problem" of consciousness.

The easy problems, and Chalmers was careful to note that "easy" did not mean simple, are questions about how the brain performs its cognitive functions. How does the brain integrate sensory information into a coherent perception? How does it direct attention to relevant stimuli? How does it discriminate between sleeping and waking states? These questions are hard, but they seem amenable to the methods of science. We can in principle explain them in terms of neural mechanisms, information processing, and functional organization.

The hard problem is different. It asks: why is any of this processing accompanied by subjective experience at all? Why isn't the brain simply doing all of this "in the dark", processing information, producing behavior, and generating reports about its internal states, all without anyone home? Why does it feel like something to be you?

"The hard problem of consciousness is the question of why any physical process in the brain is accompanied by inner experience at all. You can map every neuron, trace how the brain discriminates colour, reports feelings and processes information, and still not explain why there is something it is like to see red rather than nothing." — David Chalmers

The hard problem of consciousness is described as a distinctive challenge: explaining why and how physical processes in the brain are accompanied by subjective, qualitative aspects of experience, often referred to as qualia, such as the felt character of pain or the perceptual vividness of color.

The distinction matters enormously. A complete neural-correlate account of visual perception, one that explains which neurons fire, in what sequence, at what frequency, to produce the perception of a red apple, still leaves open the question of why that neural firing is accompanied by a vivid, red, felt experience. Correlation is not explanation. Knowing what fires when someone sees red is not the same as knowing why that firing feels like anything.

Chalmers himself suggested that this gap might require new kinds of explanatory principles beyond those currently in physics or biology, not because he was a mystic, but because the explanatory tools we have for physical systems are tools for describing structure, dynamics, and function, and subjective experience does not seem reducible to any of those things.

The Philosophical Zombie Thought Experiment

To sharpen the hard problem, Chalmers introduced one of the most famous thought experiments in philosophy of mind: the philosophical zombie. A philosophical zombie is a being physically identical to you, same brain, same neurons, same biochemistry, same behavior, but with no inner experience whatsoever. When a zombie is cut, it says "ouch" and withdraws, but there is no pain. When it looks at the sky, it says "blue," but there is no experience of blueness. There is nobody home.

The philosophical zombie is not a claim that such creatures actually exist. The claim is merely that they seem conceivable, that we can imagine a system that does everything consciousness does, functionally and behaviorally, without there being anything it is like to be that system. And if such a thing is even conceivable, it suggests that consciousness is not automatically explained by getting the physical story right. There seems to be something extra that physical accounts leave out.

Critics disagree. Daniel Dennett, perhaps the most prominent philosophical opponent of Chalmers' framing, argues that philosophical zombies are not genuinely conceivable, that if we fully imagined a system with all the functional and behavioral properties of consciousness, we would be imagining something conscious. For Dennett, consciousness just is certain kinds of information processing. There is no extra ingredient. The hard problem, on this view, is not a real gap in our understanding but a philosophical illusion created by misunderstanding what explanation can and should look like.

This debate, between those who think subjective experience requires explanation beyond physical mechanism and those who think it is identical with physical mechanism, remains actively contested. It is not a closed question.

The Competing Theories: Four Major Contenders

Despite the unsettled nature of the hard problem, science has produced several serious theoretical frameworks for understanding consciousness. Each captures real empirical evidence; none has definitively won. The field is characterized, as one wry observation has it, by theories that are like toothbrushes — everybody has their own, and nobody wants to use anybody else's.

The Great Experiment: When Two Theories Went Head-to-Head

In 2023, something remarkable happened in consciousness science. After years of competing theories accruing independent evidence — but never being directly compared — a group of theory-neutral researchers organized what they called an "adversarial collaboration." Proponents of Global Workspace Theory and Integrated Information Theory agreed to design a shared experiment, preregister their competing predictions, and let the data decide.

In the open-science adversarial collaboration, 256 human participants viewed suprathreshold stimuli for variable durations while neural activity was measured with functional magnetic resonance imaging, magnetoencephalography, and intracranial electroencephalography. Six theory-neutral laboratories on three continents ran the experiment. The results were formally published in Nature in April 2025, after the preprint had circulated since 2023.

The verdict? The findings were presented as a result of the adversarial collaboration from a group called the Cogitate Consortium. Both theories came away bruised. The experiment found evidence that partially supported and partially challenged each theory, with neither clearly winning. Debate between proponents of different theories has been vigorous and, at times, acrimonious. At a particularly low point in 2023, many experts signed an open letter arguing that Integrated Information Theory was not only false but doesn't even qualify as scientific.

This was not a failure of science. It was science doing its job — generating real constraints on theory in a field that had previously been dominated by theories that were difficult to test. The adversarial collaboration model, whatever its immediate outcome, has pushed consciousness science toward more rigorous, pre-registered, multi-site research — a methodological advance regardless of which theory eventually wins.

Altered States as a Microscope: What Drugs, Sleep, and Anesthesia Reveal

One of the most productive strategies in consciousness research is studying how experience changes — rather than trying to study it directly. When a drug alters consciousness, when sleep transforms awareness into dreaming, when anesthesia extinguishes it entirely, these transitions provide natural experiments that reveal what the brain is doing when it is and is not conscious.

Psychedelics: Dissolving the Default Self

Few tools have proven more scientifically illuminating about consciousness than psychedelic drugs — psilocybin, LSD, DMT — which produce dramatic, reproducible alterations in subjective experience whose neural correlates can now be measured in real time.

A landmark 2024 study published in Nature tracked individual brain changes across approximately 18 neuroimaging sessions per participant, before, during, and after high-dose psilocybin. The study found that psilocybin caused significant disruptions in brain functional connectivity, particularly in the default mode network (DMN), linked to subjective experiences, lasting for weeks — disruptions that may underlie its therapeutic effects.

The default mode network is a distributed set of brain regions that activate during rest, self-reflection, autobiographical memory, and mind-wandering — it is, in a sense, the neural substrate of the self-referential mind. Research shows that classic psychedelics such as psilocybin and LSD temporarily reduce rigid connectivity within the DMN, increasing global brain network flexibility. Psychedelics don't "turn off" the DMN — they make it less rigid and less dominantly self-referential, allowing other brain networks to communicate more freely.

The therapeutic implications are striking. A single high-dose psilocybin session has been shown to produce rapid and lasting relief from treatment-resistant depression, addiction, and end-of-life anxiety. The leading hypothesis is that the temporary dissolution of the brain's habitual self-model — the loosening of the DMN's iron grip on cognition — allows patients to form new relationships with their thoughts, emotions, and sense of self. Psilocybin-driven functional connectivity changes were strongest in the default mode network, which is connected to the anterior hippocampus and is thought to create our sense of space, time and self. And crucially, the persistent reduction in hippocampal-DMN connectivity lasted for weeks after the drug had left the system entirely — raising profound questions about the mechanisms of neural plasticity and the durability of experiential change.

Sleep and Dreaming

Sleep is consciousness's most intimate laboratory. Every night, the brain cycles through states of dramatically different awareness — from the nearly no-content experience of deep slow-wave sleep to the vivid, narrative, emotionally intense experience of REM dreaming. These transitions involve large-scale reorganizations of brain activity that researchers can now characterize in detail.

What sleep research has revealed is that consciousness is not an on/off switch. It is a continuum with many dimensions — vividness, self-awareness, narrative coherence, emotional salience, temporal orientation — and different sleep stages selectively suppress or preserve different dimensions. Deep sleep eliminates most of these dimensions; REM preserves many of them while disconnecting the sleeping brain from external input and from motor action. This dissociation between internal experience and external responsiveness is itself a key clue about what consciousness requires.

Anesthesia: The Switch That Shouldn't Exist

General anesthesia remains one of the most puzzling phenomena in medicine. Anesthetics reliably eliminate conscious experience — and reliably restore it — with remarkable precision. And yet we do not fully understand the mechanisms by which they do this. We know that different anesthetic agents affect different molecular targets: GABA receptors, NMDA receptors, sodium channels, and others. But how any of these molecular actions produces the global collapse of conscious experience is not known.

What we do know is that anesthesia disrupts large-scale brain communication — particularly the kind of long-range, high-frequency synchrony between frontal and posterior cortical areas that appears to characterize conscious states. This supports the Global Workspace idea that consciousness requires widespread information sharing across the brain, not just local processing in any particular region.

Hidden Minds: Consciousness in Patients Who Cannot Speak

Perhaps no domain has produced more ethically urgent findings in consciousness research than the study of patients with disorders of consciousness — vegetative state, minimally conscious state, and related conditions. These are patients who, following severe brain injury, show no behavioral evidence of awareness. They do not respond to commands. They do not communicate. By traditional clinical standards, they are not conscious.

In 2006, neuroscientist Adrian Owen and colleagues published a stunning paper in Science. Using fMRI, a patient who appeared to be in a vegetative state was, in fact, unequivocally aware, despite showing no behavioral evidence of awareness. When asked to imagine playing tennis or imagining walking through her home, the patient activated predicted cortical areas in a manner indistinguishable from that of healthy volunteers.

This single finding overturned decades of clinical assumption. It demonstrated that behavioral non-responsiveness and cognitive non-responsiveness are not the same thing — that a patient who cannot move, blink, or squeeze a hand may nonetheless be actively thinking, imagining, and responding to instructions entirely internally. The existence of what researchers now call "covert awareness" or "cognitive motor dissociation" — consciousness without behavioral expression — has since been confirmed in study after study.

The clinical implications are staggering. A 2024 study published in the New England Journal of Medicine found that among 241 adults who could not follow bedside commands, 60 — about 25 percent — could follow cognitive tasks in the fMRI scanner or during EEG testing, suggesting they could think and understand despite being unable to demonstrate it with their bodies.

One in four. Twenty-five percent of patients diagnosed as non-conscious by standard bedside evaluation may be capable of conscious thought. The practical and ethical consequences — for clinical decisions about life support, pain management, rehabilitation, and legal personhood — are enormous. And they underscore how incomplete our understanding of consciousness remains, even in the clinical domain where it matters most.

Is Consciousness Uniquely Human?

For most of the history of Western thought, consciousness was treated as a uniquely human possession — or at best shared with our closest mammalian relatives. That assumption has been under sustained attack for the past two decades, and by 2024, it had essentially collapsed among researchers who study animal cognition.

In April 2024, nearly 40 researchers signed "The New York Declaration on Animal Consciousness," presented at a conference at New York University. The declaration states there is "strong scientific support" that birds and mammals have conscious experience, and a "realistic possibility" of consciousness for all vertebrates — including reptiles, amphibians, and fish. That possibility extends to many creatures without backbones, including insects, decapod crustaceans (including crabs and lobsters), and cephalopod mollusks, like squid, octopus, and cuttlefish.

The empirical basis for this is remarkable. Bees play by rolling wooden balls — apparently for fun. The cleaner wrasse fish appears to recognize its own visage in an underwater mirror. Octopuses seem to react to anesthetic drugs and will avoid settings where they likely experienced past pain. All three of these discoveries came in the previous five years.

Octopuses are particularly compelling. They have nine brains — one central brain and one per arm — and a nervous system so alien to ours that their sentience, if real, represents consciousness arising from a radically different architecture. Research on animal minds, including those of crayfish, octopuses, snakes, and fish, suggests that consciousness "can exist in a neural architecture that looks completely alien" to ours.

The UK was among the first to act on this science: in 2022, it amended animal welfare law to recognize octopuses, crabs, and lobsters as sentient beings. Several US states have begun considering similar protections. The question of what biological architecture is necessary and sufficient for consciousness has taken on urgent policy dimensions.

Free Will and the Brain: Libet's Unfinished Experiment

In 1983, neuroscientist Benjamin Libet ran an experiment that has never stopped reverberating. He asked participants to perform a simple voluntary movement — flick the wrist whenever they felt the urge — while recording their brain's electrical activity with EEG. He also asked them to note the position of a clock hand at the moment they first felt the conscious intention to move.

What he found: a distinctive brain signal called the "readiness potential" appeared approximately 550 milliseconds before the movement. But the participants' reported moment of conscious intention came only about 200 milliseconds before the movement. The brain had apparently started preparing the action before the person consciously decided to act.

The popular interpretation spread fast: the brain decides before you do. Consciousness is a passive observer, arriving late to a party the unconscious nervous system has already organized. Free will is an illusion.

But science has been significantly more cautious. Libet's experiment ignited intense debate, prompting researchers to explore whether the readiness potential truly reflects an unconscious decision or merely signals general motor preparation. Neuroscientist Aaron Schurger provided a powerful reinterpretation: the readiness potential may not reflect a decision at all, but rather the slow accumulation of random neural noise that eventually crosses a threshold — triggering movement and, shortly after, conscious awareness. Thanks to Schurger's critique, it became clear that the Libet experiment did not rule out free will. But it does not constitute proof of free will either.

The contemporary scientific consensus, insofar as there is one, is that the Libet experiment does not resolve the question of free will in either direction. What it does reveal is that the relationship between conscious intention and action is far more complex, temporally strange, and empirically contested than folk psychology assumes. The brain is clearly doing something before you consciously "decide" — but what that something means for agency, responsibility, and autonomy is genuinely unclear.

Mental Illness: Biology Without a Complete Map

If you want to understand the practical stakes of not understanding the brain, consider psychiatry. Depression, schizophrenia, bipolar disorder, autism spectrum disorder, ADHD, obsessive-compulsive disorder — these conditions affect hundreds of millions of people worldwide, cause immense suffering, and are treated almost entirely by trial and error.

We know that these conditions involve the brain. We know that genetic factors contribute to most of them. We know that some neurotransmitter systems are implicated — dopamine in schizophrenia and ADHD, serotonin in depression, glutamate in schizophrenia, GABA in anxiety. We know that certain regions show altered structure and function in certain conditions. And we know that certain medications — SSRIs, antipsychotics, mood stabilizers — alleviate symptoms in many patients.

What we cannot do is explain, at the level of neural circuits and mechanisms, exactly why any of these conditions arise, why individuals respond so differently to treatments, or why the same drug works brilliantly for one patient and not at all for another who presents identically. The diagnostic categories of the DSM are based on behavioral and symptomatic clusters, not on the underlying biology — and there is increasing evidence that those clusters do not map cleanly onto distinct neurobiological mechanisms.

This is not a failure of will or investment. Psychiatry has received enormous research resources. It reflects a genuine gap between component-level understanding (we know what serotonin does at a synapse) and system-level understanding (we cannot explain why increasing synaptic serotonin alleviates depression in about 50% of patients with moderate delay, while doing apparently very little for the other 50%).

The New Question on the Horizon: Could a Machine Be Conscious?

The emergence of large language models — GPT-4, Claude, Gemini, and their successors — has introduced an entirely new dimension to the consciousness debate. These systems produce text that is, by any behavioral measure, sophisticated enough to be almost indistinguishable from human output in many contexts. They reflect, appear to empathize, describe their internal states, and discuss their own processing with apparent insight.

Does any of this mean they are conscious?

Most researchers are skeptical — and for principled reasons. Intelligence is about doing; consciousness is about being, about feeling. Just because a system acts smart doesn't mean it feels anything. Large Language Models like GPT or Claude often sound conscious — they can reflect, apologize, even simulate empathy — but that's not evidence of sentience.

The deeper problem is that we have no consensus theory of what consciousness requires. Without knowing what physical or computational properties give rise to experience, we cannot assess whether any given system — biological or silicon — has those properties. This is precisely why the hard problem matters practically, and not just philosophically. A world deploying billions of AI systems of increasingly complex behavior, without understanding what consciousness requires, is a world making potentially enormous moral wagers in the dark.

Some modern large language models, starting with GPT-4, have arguably passed the behavioral threshold of indistinguishability used in Turing-style tests. But Searle's famous Chinese Room thought experiment challenges the notion that behavioral mimicry implies genuine understanding or awareness.

What is genuinely interesting is that the hard problem of consciousness — which felt like abstract philosophy for most of its existence — has suddenly become an applied ethical question. If we cannot explain what makes a biological system conscious, we also cannot explain what would prevent a non-biological system from being conscious. The question has been promoted, involuntarily, from seminar room to boardroom.

Why Is This So Hard? A Genuine Answer

The brain problem is hard for a cluster of distinct reasons that compound each other.

The measurement problem. Consciousness cannot be directly measured. It can only be inferred from behavior and report. This creates an irreducibly interpretive layer between observation and conclusion that does not exist for other biological phenomena. A physicist can watch a particle; a biologist can observe a cell dividing; but no one can directly observe your experience of seeing blue.

The scale problem. The brain operates across timescales ranging from milliseconds (ion channel kinetics) to decades (long-term memory), and across spatial scales from individual synapses (nanometers) to whole-brain networks (centimeters). No single measurement tool spans this range. fMRI has excellent spatial resolution but poor temporal resolution; EEG has excellent temporal resolution but poor spatial resolution; electrophysiology is exquisitely precise but can only record from a tiny fraction of neurons. Every tool gives a partial picture.

The complexity problem. The brain is not merely complex in a quantitative sense — there are many parts — but in a relational sense: the parts are so massively interconnected that perturbing any one element propagates effects throughout the network in ways that are extremely difficult to predict. The brain is a high-dimensional dynamical system operating far from equilibrium. We have good theories for simple physical systems and some success with moderately complex ones. We have almost no theory adequate to the full complexity of the brain.

The conceptual problem. The hard problem suggests that even a complete physical theory of the brain may leave something unexplained. If this is true, then the difficulty is not just technical — it is conceptual. We may need new concepts that do not yet exist, in the way that the concept of energy had to be invented before thermodynamics could be formulated. What those concepts are, and whether they will be biological, computational, mathematical, or something genuinely new, no one knows.

What Is Actually Being Done: The State of the Field in 2026

None of this is to say the field is stuck. Consciousness science is experiencing a coming-of-age moment. Following three decades of sustained efforts by a relatively small group of consciousness researchers, the field has seen exponential growth over the past five years.

Conclusion: The Right Frame for an Unfinished Problem

The question that opened this article — why don't we understand the brain if we know so much about it? — turns out to have a more precise answer than "it's complicated."

We understand the brain's parts exceptionally well, and increasingly understand their local functions. What we lack is understanding at higher levels of organization: how circuits give rise to cognition, how cognition gives rise to consciousness, how consciousness gives rise to the felt sense of being a self that persists through time. These are not gaps that will be filled simply by learning more facts at the cellular or molecular level. They require new conceptual frameworks — new ways of describing and explaining complex biological systems — that neuroscience is still actively developing.

The hard problem may turn out to be solvable within the framework of physical science, once our concepts are sufficiently refined. Or it may turn out that explaining consciousness requires genuinely new concepts — the way understanding electromagnetism required the concepts of fields, or understanding evolution required the concept of selection. We do not know which of these is true.

What we can say with confidence is that the questions are not unanswerable in principle, and they are not being neglected. The field of consciousness science is larger, more empirically rigorous, and more theoretically sophisticated than at any point in its history. Adversarial collaborations are setting a new methodological standard. Animal consciousness research is forcing a rethinking of what biological architectures are necessary for experience. Clinical findings in disorders of consciousness are giving the hard problem urgent real-world stakes. And the emergence of AI systems capable of language and apparent reflection has promoted what was once an academic puzzle to one of the most consequential open questions of the 21st century.

Chalmers, accepting his case of wine in New York City in 2023, offered this assessment of the field he had done so much to define: "It started off as a very big philosophical mystery. But over the years, it's gradually been transmuting into, if not a 'scientific' mystery, at least one that we can get a partial grip on scientifically."

A partial grip is not nothing. In a field defined by the depth of its open questions, it is, in fact, quite a lot.