Racing towards a Grand Theory of Consciousness (2021)
This is the original 2021 article, lightly polished for English. A substantially expanded and updated version (incorporating the COGITATE results, the Koch–Chalmers bet, and the open letter controversy) can be read here. The original article stated that the cerebellum contains ‘over 100 billion neurons’; this has been corrected to roughly 69 billion, per Azevedo et al. (2009).
Propelled by modern neuroscience, the problem of consciousness has attracted sustained scientific attention. That attention has spawned numerous theories, and the conflicts between them have become the focus of intense debate. To convert theoretical dispute into forward momentum, scientists have launched a large-scale collaboration to accelerate our understanding of consciousness.
In the fifth century BCE, Democritus proposed one of the earliest atomic theories, using immutable atoms to explain the mutability of all things: not only the cosmos, the planets, and living organisms, but also consciousness itself, invisible and intangible as it is. Democritus held that consciousness was composed of ‘fire atoms’, and that thought was a product of their motion.
Twenty-four centuries separate Democritus from the present. Human civilisation has entered the information age, and communication between communities has reached extraordinary density. Yet we still know remarkably little about the 1.5-kilogram organ inside our skulls, and the existence of consciousness remains as difficult to explain as ever. Armed with the tools of the information age, some scientists have attempted an answer.
Francis Crick, who shared the Nobel Prize for elucidating the structure of DNA, was among them. In 1990, Crick and Christof Koch, then at Caltech, sounded the charge against what they called the ‘fortress of consciousness’. In a commentary published in Seminars in the Neurosciences, they argued that the problem should be investigated from a neurobiological perspective. ‘We believe the time is now ripe for an attack,’ Crick and Koch wrote. The scientist’s first task, they argued, was to find the neural correlates of consciousness: the minimal set of neural mechanisms sufficient for any one conscious experience.
Koch often returned to the problem of solipsism: many philosophers have long held that ‘I’ am merely a conscious entity trapped inside a ‘skull prison’; can ‘I’ ever truly know whether another person is conscious? The search for the neural correlates of consciousness was an assault on precisely this problem. For Koch, if one could identify the minimal unit of conscious experience, if one could somehow quantify the ‘degree of consciousness’ of every person, even every object, one would have found the key to escape the skull prison.
Three decades after Crick and Koch’s paper, the key they sought remains nowhere in sight. Many believe we understand consciousness no better than we did then, and the field has not even reached consensus on a definition.
The field has not even reached
consensus on a definition.
In 1994, four years after Crick and Koch’s foundational paper, a twenty-seven-year-old Australian philosopher named David Chalmers delivered a landmark address at the University of Arizona’s ‘Toward a Science of Consciousness’ conference. Chalmers declared that brain research has only one truly difficult problem: the Hard Problem of Consciousness. Answering it requires not only identifying the neural correlates of consciousness but also addressing a deeper question: how does the operation of those correlates give rise to subjective experience? In Chalmers’ view, the programme Crick and Koch had proposed four years earlier ignored this second question. Whatever the neural correlate turns out to be, whether heightened activity in a particular brain region or synchronised oscillation among neurons, it cannot on its own explain why subjective experience exists at all.
Koch was in the audience. He asked Chalmers how one might bridge the gap between neural mechanism and subjective consciousness. Chalmers proposed that consciousness might originate in information itself.
Chalmers drew inspiration from Claude Shannon’s information theory, developed in the 1940s, which had made it possible to treat information (like matter and energy) as a physical quantity in nature. Chalmers argued that information possesses not only physical properties but also experiential ones: an object with a great deal of information (a brain) possesses a great deal of consciousness, while an object with very little (a thermometer) possesses only a trace.
Years after the first conference, Giulio Tononi, director of the Wisconsin Institute for Sleep and Consciousness, built on Chalmers’ idea to construct an elaborate theory: Integrated Information Theory (IIT). At its centre is a measure of how much information a system integrates: the integrated information index, phi. The higher a system’s phi, the more conscious it is. Koch, who had initially been sceptical of Chalmers’ overly simple ‘information equals consciousness’ thesis, quickly embraced IIT. In a 2009 article for Scientific American, Koch drew out a striking implication: like the brain, ‘a proton, composed of just three quarks, can also integrate information and has phi.’ Consciousness, then, is not the brain’s exclusive privilege but a universal property of all things.
Can IIT find the key to the skull prison? Tononi, Koch, and their collaborators believe the answer is yes. Their research groups are searching for experimental methods to measure phi directly in the brain. Some scholars have proposed that internal connectivity within brain regions may be a crucial vehicle for information integration. For Koch, the argument is intuitively persuasive: the cerebellum contains roughly 69 billion neurons (nearly 80 per cent of the brain’s total) yet its internal connectivity is comparatively sparse, suggesting that its phi should be low. Experimental evidence bears this out: cerebellar damage causes severe motor impairment but has no significant effect on higher cognitive function or consciousness. IIT’s proponents consider this fundamental evidence for the link between phi and consciousness.
Others disagree. Cyriel Pennartz, a neuroscience professor at the University of Amsterdam, offered a counterexample in an interview: ‘When a boxer is knocked out, we say he has lost consciousness. But IIT would say that because the boxer’s brain and body still integrate information, he still has phi, and therefore still possesses consciousness. This conflicts with our intuition.’ Consciousness is a concept deeply dependent on subjectivity and intuition; such intuitive clashes cannot be dismissed. Moreover, ‘many people consider IIT unfalsifiable,’ Hakwan Lau, then a cognitive psychologist at the University of California, Los Angeles, wrote in an interview. And although IIT purports to explain the relationship between information and subjective experience, no one has yet proposed a plausible mechanism by which phi gives rise to subjective consciousness.
The most prominent theory in opposition to IIT is Global Neuronal Workspace theory, or GNW. Before Crick and Koch published their paper, the psychologist Bernard Baars at the Wright Institute in California had already proposed GNW’s precursor, Global Workspace Theory (GWT), in 1988. Baars borrowed a concept from artificial intelligence: specialised information-processing systems can collaborate briefly in a ‘global workspace’ to solve complex problems. In his view, consciousness is the brain’s method for solving complex problems, and the brain’s various regions are highly specialised independent systems that communicate efficiently, constructing a brain-wide ‘global workspace’ from which consciousness emerges.
GWT and IIT share some common ground: both concern information processing. But unlike IIT’s straightforward ‘integrated information is consciousness’, GWT holds that consciousness is a mode of information processing: information flows from the global workspace to the specialised systems (attention, memory, language production) that require it. The two theories also diverge on scope: IIT implies that everything from a proton to the entire universe can possess consciousness; GWT restricts consciousness to information-processing networks.
The cognitive scientist Stanislas Dehaene, at the Collège de France, gave GWT its neural mechanism. In a paper published in the Proceedings of the National Academy of Sciences, Dehaene and colleagues proposed that long-range pyramidal-cell connections between prefrontal, parietal, and temporal cortex form the physical substrate of the global workspace, and added ‘Neuronal’ to the name. When sensory neurons receive sufficient stimulation, they activate these long-range networks in a sudden, nonlinear ‘ignition’ that broadcasts the signal brain-wide. The contents of that broadcast are, in GNW’s account, the contents of consciousness.
IIT and GNW thus make directly opposing predictions about where consciousness resides in the brain. IIT holds that the posterior cortex (where neural connectivity supports efficient information integration) is the primary seat of consciousness. GNW holds that the prefrontal cortex (where long-range broadcasting and ignition occur) is critical. ‘We can test these two opposing predictions and find evidence supporting one or the other,’ Pennartz said in an interview.
GNW, too, has its critics. The theory addresses questions such as ‘Why does a driver become aware of a pedestrian on a zebra crossing?’ but it cannot explain the link between the global workspace and subjective experience. In the view of Michael Graziano at Princeton, both GNW and IIT will ultimately fail. The real hope, he argues, lies in metacognitive theories, the most prominent of which is Higher-Order Theory.
Higher-Order Theory, or HOT, starts from a different question entirely. David Rosenthal, a philosophy professor at the City University of New York, explained the core principle in an interview: ‘To say that a mental state is conscious requires that we are aware of that mental state itself.’ This is the transitivity principle. A mental state is conscious when and only when the subject has a thought about that state.
Unlike IIT and GNW, HOT holds that metacognition (cognition of cognition) is a necessary condition for consciousness. A central claim is that cognitive processes (memory, belief, visual processing) and metacognitive processes are mutually independent: the former can proceed unconsciously in the absence of the latter. In such cases, although one cannot report on these processes, they influence behaviour in observable ways.
HOT’s supporters have found evidence for this dissociation. In a study published in the Proceedings of the National Academy of Sciences, participants were shown a target shape for just 33 milliseconds (too brief for confident perception). The researchers found that while accuracy on identifying the shape was comparable at short and long intervals, confidence that one had ‘truly seen’ the target was significantly higher at longer intervals. Cognition, in other words, can occur without metacognition. Activity in the mid-dorsolateral prefrontal cortex reflected the strength of metacognitive ability, providing a candidate neural substrate for HOT. Multiple research groups, including Lau’s, are investigating the viability of higher-order models.
Even so, HOT, like every other theory, is far from settled. Only through fair competition with rival theories can scientists hope to identify the best account of consciousness.
In 2019, the Templeton World Charity Foundation launched a $20 million initiative called Accelerating Research on Consciousness, or ARC. The concept is ‘adversarial collaboration’: bring proponents of competing theories to the same table, force them to pre-register opposing predictions, then test those predictions using the same experimental design and the same data.
‘The days of the lone genius scientist solving big problems in the laboratory are over,’ said Dawid Potgieter, senior programme officer at the Templeton Foundation. ARC’s model enables researchers to test their disagreements directly. Participants convene for ‘adversarial seminars’ driven by evidence-based debate, striving to refute their opponents and reflect on their own theory’s weaknesses. They then return to the laboratory to collect data, funded by Templeton, which also funds replication studies and requires open data. If this model produces breakthroughs, it will simultaneously demonstrate the value of open, adversarial science.
Not everyone shares this optimism. Anil Seth, a cognitive scientist at the University of Sussex and an ARC participant, warned that the competing theories ‘have built too many different assumptions, have different levels of falsifiability, and are even trying to explain different things’. Rosenthal argued that the supposed ‘competition’ between IIT and GNW is partly illusory: ‘IIT addresses creature consciousness, whereas GNW addresses state consciousness, attempting to clarify what conditions a mental state must satisfy to be conscious.’ Theorists should also specify ‘whether they are asking “what counts as consciousness?” or “what is the mechanism of consciousness?”’ Without that clarity, Rosenthal cautioned, ‘we will not make any real progress.’ Seth and Rosenthal, among others, therefore consider ARC premature.
Some participants felt that ARC had not achieved a level playing field. The adversarial seminars were chaired by Koch, which naturally troubled scientists supporting theories other than IIT. Lau wrote in an interview that although he greatly admires Koch, ‘his attitude toward IIT is certainly not neutral’. Seminar decisions, Lau said, were ‘basically biased. We could indeed propose our own ideas, but ultimately, Tononi and Koch would directly veto proposals they disliked and suggest that everyone investigate the questions they had wanted to investigate all along.’
Others considered these concerns overstated. Liu Ling, a postdoctoral researcher in Luo Huan’s laboratory at Peking University, which participated in ARC, said that ‘different theories do have their own definitions of consciousness, but they can find common ground to some extent’. Pennartz agreed: what supporters of different theories needed to do was identify ‘differential hypotheses’, directly opposing predictions. ‘“Is consciousness located in the prefrontal cortex or in the posterior cortex?” is one such differential hypothesis,’ Pennartz said. ‘I believe the competition between IIT and GNW is real.’
ARC has nevertheless built a bridge for discussion among the field’s most powerful theories. In John Horgan’s Mind-Body Problems, Koch cited the nineteenth-century philosopher Auguste Comte to counter the claim that consciousness is beyond understanding. ‘Comte declared that we could never know what the stars are made of. But shortly after he said this, spectroscopy was born, revealing their chemical composition.’ Koch remains convinced that science can break through the problem of consciousness, and ARC’s launch demonstrates that many scientists share his ambition. Whether that ambition will prove justified is, as of early 2021, an open question, one that the coming years of adversarial data will begin to answer.
Interviews with Anil Seth, Cyriel Pennartz, Hakwan Lau, David Rosenthal, Dawid Potgieter, and Liu Ling were conducted for the original Chinese-language feature. Further reading: Crick and Koch, ‘Towards a neurobiological theory of consciousness’ (1990); Tononi and Koch, ‘Consciousness: here, there and everywhere?’, Phil. Trans. R. Soc. B (2015); Mashour, Roelfsema, Changeux, and Dehaene, ‘Conscious Processing and the Global Neuronal Workspace Hypothesis’, Neuron (2020); Brown, Lau, and LeDoux, ‘Understanding the Higher-Order Approach to Consciousness’, Trends Cogn. Sci. (2019). Originally published in Chinese in Scientific American China, March 2021 print issue.
