What Is Consciousness: Neuroscience's Hardest Problem

Why Consciousness Is Science's Hardest Problem

You are reading these words. You are experiencing the sensation of reading—the meaning of the words unfolding in your mind, perhaps the feel of a screen under your fingers, the ambient sounds of wherever you are. This experience of being a subject—the sense of "what it is like" to be you, right now—is consciousness. And despite being the most intimate fact of your existence, consciousness remains the deepest unsolved problem in science and philosophy.

The philosopher David Chalmers coined the phrase "the hard problem of consciousness" in 1995 to distinguish between two very different questions. The "easy problems" of consciousness—not actually easy, but tractable—involve explaining cognitive functions: how the brain integrates information, directs attention, generates reports about internal states, controls behavior. These are hard scientific problems, but they are the kind of problems science knows how to approach. The hard problem is different: why does any of this processing feel like anything? Why is there subjective experience at all? Why aren't we philosophical zombies—beings that process information and behave like humans but with nobody home, no inner light of experience?

What Neuroscience Has Established

Neuroscience has made substantial progress on the "easy" problems and has established several important facts about the brain basis of consciousness. One of the most fundamental is the existence of neural correlates of consciousness (NCCs)—the minimum neural activity sufficient for a given conscious experience. Studies using functional MRI, EEG, and single-cell recordings have identified brain regions and patterns of activity that reliably correlate with conscious perception versus unconscious processing.

The prefrontal cortex and parietal lobes play particularly important roles in conscious awareness. Damage to these areas can produce profound disturbances of consciousness—neglect syndrome (in which patients are unaware of one side of their visual field), anosognosia (unawareness of one's own disability), and various disorders of self-awareness. The thalamus acts as a relay and integrator of sensory information and is critically involved in maintaining consciousness; damage to certain thalamic nuclei produces coma. And studies of the unconscious processing that precedes conscious awareness—such as the readiness potential that appears in EEG recordings hundreds of milliseconds before subjects consciously decide to move—have revealed that much of what we experience as conscious decision-making may follow rather than precede the neural events that produce it.

Global Workspace Theory

One of the leading scientific theories of consciousness is Global Workspace Theory (GWT), developed by cognitive neuroscientist Bernard Baars and expanded by Stanislas Dehaene. GWT proposes that consciousness arises when information is broadcast widely across the brain through a "global workspace"—a kind of central broadcasting system that makes information available to many different cognitive processes simultaneously.

On this view, most of the brain's processing occurs in specialized, local modules that operate below the threshold of consciousness. Information becomes conscious when it "ignites" the global workspace—a long-range network involving prefrontal and parietal cortex—and is broadcast to the entire brain. Dehaene and colleagues have identified specific neural signatures of this "ignition" in EEG and fMRI data: when a stimulus crosses the threshold into consciousness, there is a sudden, nonlinear increase in large-scale coordinated brain activity. GWT explains many features of consciousness: its limited capacity (the workspace can broadcast only one thing at a time), its global accessibility (conscious information is available to memory, language, and attention), and its relationship to attention. Critics argue that GWT describes the mechanism of conscious access but does not address why any of this broadcasting feels like anything—it solves the easy problems but not the hard one.

Integrated Information Theory

A very different approach is Integrated Information Theory (IIT), developed by neuroscientist Giulio Tononi. IIT starts from the phenomenology of conscious experience rather than from neuroscience, asking: what are the essential properties of consciousness, and what physical systems could have those properties? Tononi argues that consciousness is identical to integrated information—a measure he calls Phi (Φ) that quantifies how much a system's parts contribute to the whole over and above their contributions as independent parts.

On IIT, any physical system with high Phi has some degree of consciousness. This makes IIT a form of panpsychism—the view that consciousness is a fundamental feature of the physical world, present in varying degrees wherever integrated information processing occurs. A human brain has very high Phi; a thermostat has very low Phi; but even the thermostat might be the subject of the tiniest flicker of experience. This implication has been both IIT's most discussed feature and its most controversial: critics argue that panpsychism is an unscientific mysticism dressed in mathematical language. Supporters argue that it is the only theory that takes the hard problem seriously—that genuinely explains why there is something it is like to be a brain, rather than just describing the brain's information processing.

Disorders of Consciousness

Much of what neuroscience knows about consciousness comes from studying patients who have lost it. Coma is a state of unresponsiveness from which the patient cannot be aroused; the brain continues to function at a basic level but shows none of the large-scale, coordinated activity associated with consciousness. The vegetative state is a condition in which patients show sleep-wake cycles but no signs of awareness of themselves or their environment. Until recently, vegetative state patients were assumed to be entirely unconscious.

A landmark 2010 study by Adrian Owen and colleagues changed this picture. Using fMRI, Owen found that some patients diagnosed as vegetative could follow instructions—when told to imagine playing tennis, their brains showed the same activity as healthy volunteers imagining playing tennis. A significant minority of apparently vegetative patients, it turned out, were aware but unable to move or communicate. This finding revolutionized the clinical management of disorders of consciousness and raised profound ethical questions about how we assess and treat these patients. The locked-in syndrome—in which patients are fully conscious but paralyzed and unable to communicate except by eye movement—provides a chilling illustration of how consciousness can be entirely dissociated from any behavioral evidence of it.

Anesthesia and Altered States

General anesthesia has become one of the most productive experimental systems for studying consciousness. Anesthesia produces a reversible loss of consciousness, and the neural mechanisms by which different anesthetics work provide clues about what consciousness requires. Most general anesthetics appear to work by disrupting the large-scale integration of brain activity that GWT and IIT both identify as crucial—they create a state in which local brain regions continue to process information normally but cannot communicate effectively with each other.

Other altered states—sleep, dreaming, psychedelic experience, meditation, and hypnosis—each provide different windows on consciousness. Dreaming is consciousness with reduced sensory input and reduced prefrontal involvement, producing vivid experience with diminished reality monitoring. Psychedelic drugs dramatically increase the complexity and integration of brain activity (increasing Phi, on IIT's measure) and produce equally dramatic increases in conscious experience. These observations are consistent with the view that consciousness is intimately connected to the global integration of brain activity.

The Future of Consciousness Research

The Cogito Project, a large-scale collaboration announced in 2019, has set about empirically testing the predictions of competing theories of consciousness—particularly GWT and IIT—using identical experiments conducted across multiple laboratories worldwide. Early results have been illuminating and sometimes surprising, suggesting that both theories capture important aspects of the phenomenon but that neither is complete. The field of consciousness science has matured from a fringe interest into a mainstream area of neuroscience research, with dedicated journals, conferences, and laboratories.

Whether the hard problem will ultimately yield to scientific investigation is contested. Some philosophers argue that science can in principle explain everything about consciousness, including the subjective dimension; others argue that subjective experience is irreducibly first-personal and cannot be fully captured in the third-person terms of physical science. What is certain is that consciousness—the fact that there is something it is like to be you—is the most extraordinary and the least understood feature of the universe we know. Its investigation is simultaneously the most personal and the most cosmic question science has ever asked.

Artificial Intelligence and the Consciousness Question

The development of increasingly sophisticated artificial intelligence systems has given the consciousness question new urgency. Large language models like GPT-4 and Claude produce outputs that, to many users, feel like genuine understanding and communication—they express opinions, appear to reason, and respond to emotional content with what seems like empathy. Do these systems have any form of consciousness or subjective experience? The question is not merely philosophical; it has ethical implications for how we design, deploy, and regulate AI systems.

Most neuroscientists and philosophers of mind argue that current AI systems, whatever their capabilities, lack the biological substrates that consciousness appears to require—or at least that we have no evidence they are conscious. IIT would predict that current AI architectures have very low Phi—they are not highly integrated in the relevant sense. But this conclusion depends on assumptions about the relationship between architecture and consciousness that are not fully established. As AI systems become more powerful and more deeply integrated into human life, the question of machine consciousness will become more pressing—not as a science fiction scenario but as a live philosophical and ethical challenge that the science of consciousness will need to address.