Generated by Codex with GPT-5

A Science Built around Private Evidence

Allison Parshall’s article begins with the difficulty at the center of consciousness research: consciousness is the most immediate fact of a person’s life, yet it is directly available only to that person. Scientists can observe behavior and brain activity, but they cannot independently inspect another individual’s pain, inner speech, sense of self or experience of color. The field must therefore study a real phenomenon through indirect evidence.

Researchers often separate conscious experience into three dimensions. Wakefulness is the state supported by the brain stem that usually comes with open eyes. Internal awareness includes thoughts, mental imagery and other experiences not tied to current sensory input. Connectedness is the ability to perceive and potentially respond to the outside world. These dimensions can come apart: dreams retain awareness while reducing connectedness, and some outwardly unresponsive brain-injury patients may still possess internal awareness.

For much of the 20th century, this measurement problem pushed consciousness toward philosophy and away from experimental neuroscience. That changed around 1990, when Francis Crick and Christof Koch proposed searching for the neural correlates of consciousness just as functional magnetic resonance imaging made the working brain easier to observe. Experiments with ambiguous images, such as Rubin’s vase, offered a promising strategy: the visual input stays fixed while perception flips, allowing researchers to ask which neural activity changes with the conscious experience itself.

The approach has ruled out some simple answers but has not identified a single consciousness center. Early sensory regions process information that never reaches awareness and continue operating during anesthesia. The cerebellum contains most of the brain’s neurons but seems largely uninvolved. Other evidence points toward distributed cortical networks and possibly deep structures such as the thalamus. After three decades, the field has learned more about where consciousness is not than about exactly how the brain produces it.

Competing Maps of the Mind

The uncertainty has produced dozens of theories with different definitions, starting points and predictions. Global neuronal workspace theory treats consciousness as a shared stage: information becomes conscious when an “ignition” event makes it globally available to systems for attention, planning and action. Higher-order theories say a sensory representation becomes conscious only when frontal regions create a representation of that representation - in effect, when the brain registers that it is having a particular mental state.

Predictive-processing theories emphasize a continuous exchange between bottom-up sensation and top-down expectation. On this view, perception is the brain’s best controlled guess about the world, and consciousness emerges from the repeated correction of prediction errors. Integrated information theory starts from different premises. It argues that consciousness belongs to systems whose states are both richly differentiated and unified, and it associates those properties with a mathematical quantity called phi. Its implications are unusually broad: if the theory is correct, some nonliving systems might possess a degree of consciousness.

No theory has closed the explanatory gap between physical activity and subjective feeling. A model may explain how information becomes broadly available or how brain networks integrate, yet still not explain why a toothache feels unlike a headache or why any processing feels like something at all. The theories nevertheless matter because they generate experiments, clinical tools and rival predictions that can be tested.

One practical advance came from Marcello Massimini and colleagues, who combined transcranial magnetic stimulation with electroencephalography. A magnetic pulse acts like a knock on the cortex, while EEG records how the disturbance spreads. In an awake or dreaming brain, activity propagates through diverse networks in a complex pattern. During dreamless sleep or anesthesia, it remains local and quickly fades. The resulting perturbational complexity index can help distinguish patients who may retain consciousness despite being unable to respond. It does not reveal the content of experience, but it turns network complexity into clinically useful evidence.

When Leading Theories Met the Data

In 1998 Koch wagered philosopher David Chalmers that researchers would identify a clear neural signature of consciousness within 25 years. Koch conceded in 2023. That same year, a large adversarial collaboration reported results designed to pit integrated information theory against global neuronal workspace theory across multiple institutions and measurement methods.

Both theories came away weakened. The study found sustained activity toward the back of the brain, as integrated information theory expected, but not the predicted synchronization. It found an initial frontal signal compatible with workspace theory, but not the second ignition event that theory predicted when an image disappeared. The mixed outcome showed the value of forcing theories to make precise bets, while also revealing how far the field remains from a decisive account.

The aftermath exposed another vulnerability. More than 100 researchers signed a letter calling integrated information theory pseudoscience, objecting especially to strong public claims and its panpsychist implications. Critics of the letter feared that branding one prominent model this way could damage the hard-won legitimacy of consciousness science as a whole. Beneath the dispute was a methodological question: how should a young field separate bold but testable theory from claims that outrun the evidence?

Despite the conflict, the article does not portray the field as empty. It has developed tools for assessing patients, mapped brain systems associated with pieces of experience and made rival theories experimentally accountable. Its lack of a unifying breakthrough resembles a difficult stage in science, not proof that the subject lies beyond science.

Animals, AI and the Cost of Uncertainty

The measurement problem grows sharper beyond healthy adult humans. Evidence of play, self-recognition and flexible decision-making has widened scientific concern from mammals and birds to fish, insects, crustaceans and some mollusks. Comparing species may also reveal what consciousness is for. One proposal is that a living creature needs a unified stream of experience because it must combine many signals while choosing only one action at a time.

Artificial intelligence raises the inverse question: can a system convincingly imitate conscious speech without any experience behind it? Today’s large language models are probably such unconscious imitators, but researchers lack an agreed test that could prove the point. Workspace theorists can look for system-wide broadcasting in an AI architecture. Others argue that computation alone is insufficient because brains are living physical systems shaped by electrical fields, chemicals, metabolism and thousands of cell types. Simulating the information flow may no more create experience than simulating a storm creates rain.

Uncertainty is not a reason to abandon the problem. Decisions about anesthesia, severe brain injury, psychedelic drugs, animal welfare and intelligent machines already depend on judgments about where consciousness exists. The field’s most honest conclusion is therefore also its most useful one: there is no accepted theory or definitive instrument yet, but the phenomenon is too consequential to leave unmeasured. Progress will come from better operational definitions, sharper predictions and tools that let researchers infer private experience without pretending they can observe it directly.