“Theory without experiment is lame. Experiment without theory is blind.”

— anonymous

Modern consciousness science was born ~35 years ago, when Francis Crick and Christof Koch helped usher in an age of empirical neuroscientific research on the “neural correlates of consciousness” around 19901. Thirty-five years later, what role(s) should theories play in a maturing science of consciousness? This is a topic that has undergone considerate discussions, debate, and even controversy2 recently.

Theories have always played a central role in consciousness science, owing, in part, to the field being situated at the intersection of philosophy, psychology, and neuroscience. The proposals of several prominent theories of consciousness—which remain influential today—including Global (Neuronal) Workspace Theory (GWT/GNWT)3, Integrated Information Theory (IIT)4, and Recurrent Processing Theory5 around 20–25 years ago helped build a strong community of researchers who conduct empirical or theoretical work on consciousness. Today, there are about five popular theories of consciousness, including the above three, along with higher-order theory6 and predictive processing theory7,8.

The nature of theories

What is a theory? Theory is defined as a plausible principle or body of principles to explain a phenomenon. A theory should be simplifying (otherwise it would not help with understanding) and have predictive power (otherwise it would be useless). A theory does not have to be comprehensive (e.g., a psychological theory does not have to specify neural implementation). A theory should have a clearly outlined explanatory scope (i.e., the target phenomenon it aims to explain). Finally, a theory should be testable in the sense that it can generate falsifiable hypotheses.

Theory is often confused with framework. Framework is a way to organize concepts, a way to think about a phenomenon. A framework tends to be broader than a theory and may or may not be directly testable. For instance, Marr’s three levels of description (computation, algorithm, implementation) is a framework that has been extremely useful in cognitive science and neuroscience. A good example in consciousness science is the idea that sensory processing can be subliminal, preconscious, or conscious9, which is a framework. By contrast, the idea that conscious access has the constituent mechanism of sensory information crossing a nonlinear bifurcation threshold (GNWT)10 is a theory. Viewed in this light, IIT, at least in its original formulation4, might be called a framework, because it is notoriously difficult to test but provides broader perspectives that can be quite helpful for thinking about consciousness and especially its enabling conditions (for instance, the idea that consciousness has both integration and differentiation—both a unified experience and a rich repertoire). When I proposed the Joint Determinant Theory (JDT) in 202311, I deliberately labeled it a framework because it offers a way to think about the neural basis of consciousness: a state-space view and the idea that the specific circuit mechanisms underlying different aspects of conscious content can be different; nevertheless, I retained “theory” in the name with the hope that over time it becomes specified with detailed neural mechanisms that are directly experimentally testable and the thinking that many contemporary “theories” are equally vague.

Compared to other disciplines in neuroscience and psychology, theories have played an outsized role in consciousness science. Empirical studies in this domain, and even grant proposals, are often expected to make contact with existing theories. Recently, $20–30 million was invested by Templeton World Charity Foundation in five adversarial collaboration projects carried out by large, international consortia to pit theories against each other. Sociologically, the few dominant theories have perhaps become entrenched, with young consciousness researchers often aligned with one of the camps. Pitfalls of this state of affairs have been pointed out recently, including confirmation bias that appears commonplace12.

What is the best way forward? Some may advocate for empirical studies to ditch theories altogether. There are certainly great opportunities for theory-neutral empirical studies to contribute, especially at this (still young) stage of development of the field. But it is also important to remember that “experiment without theory is blind”. With that in mind, here I advocate for a middle-of-the-road, nuanced approach.

The dangers of assuming uniformity

One potential key obstacle to progress is that consciousness theories are often idolized, their explanatory scopes not clearly outlined (some of the following ideas were recently discussed in ref. 11). A common assumption is that a key principle, or set of principles, will explain all aspects of conscious awareness (the ‘uniformity assumption’). Consciousness researchers often hold up the theory of evolution or the discovery of the double-helix structure of DNA as shining examples of the amount of explanatory power that a correct theory of consciousness will have. This is a romanticist vision, and it might be true, but it might not be; the jury is still out on this. The problem is that believing the uniformity assumption automatically extends the explanatory scope of any theory related to consciousness to encompassing all aspects of awareness, regardless of the origin of that theory. As a result, theories inspired by empirical observations of visual perception, sleep, metacognitive judgment, or attention are all automatically assumed to explain all aspects of conscious awareness. Should the uniformity assumption be wrong, this would cause massive confusion and gigantic roadblocks to progress, because theories that are most naturally suited to explaining different phenomena—e.g., different aspects of conscious awareness—are pitted against each other in an endless cycle. This might also explain why which theory is supported by empirical results can be predicted by the experimental paradigm used by that empirical study12, because if different theories have different explanatory targets that would be the natural outcome.

I believe that if we relax the uniformity assumption and allow theories to be humbler in their explanatory scope, the field would be able to make meaningful progress much faster and at the same time avoid a stagnant theoretical landscape. For instance, a theory most suited to explaining how a piece of sensory information accesses awareness does not have to simultaneously explain how sleep, psychedelic state, anesthesia and disorders of consciousness are all different from each other. A theory most suited to explaining conscious object perception (e.g., grouping) can be focused on explaining that particular phenomenon and other related phenomena, and be extremely useful at that. And a theory most suited to capturing the enabling conditions for consciousness to be possible in a system does not have to also explain how every aspect of awareness (e.g., perception and emotion) arises.

Such an attitude and culture would allow consciousness theories to be more focused and likely more experimentally testable. Put concretely, we need theories to explain how conscious perception occurs, how felt emotions happen, how conscious thinking unfolds, how intentions and agency are generated by the brain, and what brain state changes render consciousness impossible or altered under sleep, anesthesia, pathological and psychedelic conditions. This “boots-on-the-ground” approach also would allow new theories to be more readily proposed. Over time, when these “humbler” theories are experimentally validated, we would be on a firmer ground to examine whether there are common principles across these different aspects of consciousness. Importantly, such a shift in mindset would not diminish the significance of theories at all; it only takes away the grandiosity of theories.

Let us examine some examples of significant theories in other fields, which can help illuminate this discussion.

One of the all-time most influential theories in neuroscience is Hebbian plasticity proposed by Donald Hebb in 1949: “Neurons that fire together wire together.” Since then, Nobel-prize winning experimental work has confirmed the existence of Hebbian plasticity in the brain. But empirical observations opposite to the predictions of Hebbian plasticity have also been found, in a phenomenon termed “anti-Hebbian plasticity”13, which is thought to serve functions of preventing runaway excitation, decorrelating neural signals and maintaining excitation-inhibition balance. If the current model in consciousness science were adopted, the theory of Hebbian plasticity would have been “falsified” by these anti-Hebbian experimental observations and that would have been a huge mistake. Instead, Hebbian plasticity remains one of the cornerstone theories in neuroscience, with the appreciation that there are important exceptions to this rule.

Let us also take a look at an example with a much narrower scope: activity-silent working memory theory. This idea was proposed in a computational model in 200814; since then it has become highly influential and has received direct experimental support15. This theory has a clear, focused explanatory scope: it proposes a specific neural implementation of a mental phenomenon (working memory), which complements rather than replaces an earlier theory on the same topic (persistent activity in working memory). I highlight this example here because the consciousness field needs more theories like this, which postulate concrete and testable neural mechanisms with clearly outlined explanatory targets. When theories are automatically assumed to explain all of consciousness, as is the case presently, they can become vague, shape-shifting, and difficult to test. Finally, from a sociological perspective, having humbler theories with clearly specified explanatory targets would also free the younger generation of consciousness researchers to come up with independent, fresh ideas instead of building their careers on testing existing dominant theories.

Consciousness science is at a time of unprecedented opportunities. No longer a tabooed topic and no longer marginalized, able to capitalize on the breakneck speed of technological advances in neuroscience, consciousness science is expected to flourish and deliver groundbreaking insights with enormous practical and clinical implications in the next few decades. This endeavor will be especially successful if we harvest the combined power of experiment and theory in a closed-loop manner: theories guide experiments and experiments lead to new or refined theories.