Overcoming essentialism: evaluation of theories of consciousness by the lawful regularities they predict

Abstract

A common objection to consciousness research is that researchers cannot agree on what consciousness is. But how should a consensus on the nature of consciousness be reached? Does consciousness have essential features that one can be right or wrong about? Should we select theories of consciousness based on their descriptions of the essential features of experience alone? In this article, we will argue with the help of Popper that determining unquestionable essential features of consciousness is neither possible nor desirable. Instead, any account of what consciousness is, should be understood as a subjective selection of features and should not count as an insight. According to Popper, insights in science are achieved when novel regularities between observations are discovered that generalize to scientific laws. Furthermore, all results of research can be stated independently of the definitions that were used in their making. Therefore, no consensus on the nature of consciousness or on a theory of consciousness is needed for consciousness research to progress—in fact, a multitude of definitions and theories may be beneficial. Based on this, we argue that theories of consciousness be evaluated by their ability to propose concrete, testable laws rather than by their description of what consciousness is.

Highlights

• How to define consciousness and how to determine its essential features are often considered to be major obstacles in consciousness research, but we argue in this article that this view is a misconception of scientific progress.

• According to Karl Popper, there is no way to determine indisputable essential features of a concept (such as consciousness), and it is also not the task of science to find essential features of concepts.

• Theories of consciousness often regard their definition of consciousness (what they consider its essential features) as an insight in itself, but we argue that definitions constitute untestable elements of a theory and should therefore not serve as a basis for evaluation of a theory.

• According to Popper, science progresses not by revealing essential features of concepts but when scientific laws that state regularities between observations are corroborated. In the process of proposing and testing scientific laws, concepts such as consciousness play only a temporary and subordinate role.

• We conclude that theories of consciousness should be solely evaluated by the scientific laws that they propose, which includes rejecting theories of consciousness that propose no laws at all and focusing on the predictive value of theories of consciousness in experimental settings.

Introduction

What is consciousness? Various recent scientific and philosophical publications start precisely with this question (Dehaene et al. 2017, Seth 2021, Pennartz 2022, Kind and Stoljar 2023). But in what way is the question of the nature of consciousness relevant to the field of consciousness research? Do we first need to agree on what consciousness is before we can study it? Or is it the other way around—will we determine what consciousness is by studying it? Will research reveal the essential features of consciousness gradually until we have fully determined what consciousness is, just as physics has gradually revealed the nature of the atom in the 20th century? Or will multiple definitions of consciousness exist, each tailored to the field of research that employs it, such as there are different conceptualizations of life used in medicine and cell biology?

Underlying all these questions is a more fundamental one (raised for example by Bayne 2018): should we assume that consciousness (or conscious experience, we use these terms interchangeably) has essential indisputable features that we can uncover? Scientific explanations of consciousness are usually given in the form of a theory of consciousness (ToC) that describes what the essential features of consciousness are and what physical properties are associated with these features (Seth and Bayne 2022). Can we evaluate ToCs by what they consider essential features of consciousness before considering their predictions in experimental settings? Some scientific debates certainly seem to suggest so, such as when Carter et al. (2018) dismissed Dehaene et al. (2017)’s approach of identifying conscious machines based on two “essential dimensions of conscious computation” (global availability and self-monitoring), not by citing scientific evidence but by claiming that Dehaene et al. “ask and answer the wrong questions.” Related to the issue of essential features of consciousness is the question of convergence of consciousness research on a single ToC. Discussions such as the Great Consciousness Debate at ASSC 2022 (Mudrik et al. 2025) and the recent advent of adversarial collaborations (Ferrante et al. 2025, Corcoran et al. 2026) seem to suggest that there will eventually be (or at least should be) a single “winning” ToC. But, as we will argue later, it is by no means obvious that ToCs should contradict each other such that we can pick a winning ToC (cf. (Evers et al. 2024)). So why is there this underlying belief of a single winning ToC?

Here, with the help of Popper (1945), we want to examine the belief that consciousness has essential features that a ToC can either be right or wrong about and the related belief that there can only be a single correct ToC. (Popper 1935, Popper and Weiss 1959)’s thoughts on falsification are well known but Popper (1945)’s thoughts on the role of concepts in science are less prominent. We associate the view that consciousness has essential features with the broader view of essentialism. In section Methodological Essentialism of this article, we will first discuss the role that essentialism currently plays in consciousness research. Then, we will examine the flaws that Popper saw in essentialism and how these flaws adversely impact consciousness research. In section Methodological Nominalism, we will discuss Popper’s alternative to essentialism called methodological nominalism. In methodological nominalism, definitions of concepts are understood to be arbitrary and cannot be correct or incorrect. We will examine how Popper imagined concepts to serve as temporary, lightweight tools to aid in the proposal of scientific laws and how the results of research can be stated independently from the concepts used in their making. We conclude that ToCs should be evaluated not by their definitions of consciousness but by the scientific laws they propose. We end this section with a set of guidelines to increase the practical utility of ToCs in research, as well as an evaluation of Integrated Information Theory (IIT) and Global Neuronal Workspace Theory (GNWT) according to the guidelines. An evaluation of an additional ToC, Neurorepresentationalism, can be found in the appendix. Finally, in section Discussion, we will discuss criticisms of Popper’s philosophy of science and some limitations of methodological nominalism.

Methodological essentialism

The belief in a single “winning” theory of consciousness

We propose that there are two reasons why a scientist may believe that there can only be a single correct ToC. The first is that scientists are used to having multiple theories contradict each other when they refer to the same term in their name. That is, we expect two ToCs to be mutually exclusive just as we would expect two theories of gravity to be mutually exclusive. This reason is reinforced by scientific debates between theory leads (such as the Great Consciousness Debate at ASSC 2022), as well as the recent advent of adversarial collaborations in which ToC do make competing, mutually exclusive experimental predictions. The second proposed reason is the adherence to “essentialism,” i.e. the view that concepts have a true hidden nature (“essence”). When essentialism is applied to science, we arrive at “methodological essentialism,” which is the teleological view that the goal of research should be to reveal the essence of a concept. From an essentialist point of view, the true nature of consciousness can (and should) be discovered just as the structure of the atom has been discovered in the 20th century.

Theories of consciousness do not necessarily make mutually exclusive predictions

We will next argue against these notions. First, as Irvine (2013) has discussed, many phenomena have been associated with the term consciousness and it is by no means obvious why one explanation (or ToC) should cover all of them. Furthermore, it is not obvious why ToCs—even when they apply to the same phenomena—must make competing predictions rather than compatible ones (Chis-Ciure et al. 2024, Storm et al. 2024). In fact, recent adversarial collaborations have shown that it is quite difficult to come up with experiments for which the former is the case. It is therefore not surprising that studies of the commensurability of ToCs exist and that these show that some ToCs could potentially be combined (Northoff and Lamme 2020, Evers et al. 2024). We conclude that it is not obvious that consciousness research should converge on a single ToC based on experimental predictions alone.

Essentialism may support the belief in a single “winning” theory of consciousness

Next, we examine the second reason why a researcher may believe in a single correct ToC: they believe that consciousness has a set of essential features such that whichever ToC uncovers this essential set first will win (whereas all others will lose). For example, according to IIT, consciousness arises from an integrated and differentiated substrate (Albantakis et al. 2023), while according to GNWT, consciousness arises by information broadcasting to local processors through a global workspace (Mashour et al. 2020). Can we not say (perhaps even before conducting experiments): integration and differentiation are essential features of consciousness while information broadcasting is merely a correlate, an inessential part? In the following, we will argue with the aid of Popper (1945) that such an attempt at selecting ToCs is flawed because determining unquestionable essential features of a phenomenon is neither possible nor desirable in scientific research.

Popper’s rejection of essentialism

According to Popper (1945), the goal of essentialism is to reveal the hidden nature of a phenomenon and capture this nature in a definition.¹ Such a definition consists of the essential features of a thing and is supposed to be complete and precise. Popper associates essentialism with Plato’s Theory of Forms and Aristotle’s doctrine of definitions. In methodological essentialism (essentialism as a teleological view on science), a ToC should be evaluated by whether it captures the essence of consciousness fully, but we will argue that it is impossible to conduct such an evaluation conclusively based on two issues that Popper (1945) sees in essentialism: (i) its reliance on intuitive insights and (ii) an infinite regression of terms that occurs when using essentialist definitions.²

Intuitively selected features have no scientific basis

Popper’s first problem with methodological essentialism is that there is no “scientific” way to agree on the essential features of a thing because in essentialism, such features are revealed by “intellectual intuition.”³ In consciousness research, lists of essential features of consciousness are sometimes based on the argument of conceivability—that is, experience is said to be conceivable without some features but not others (Albantakis et al. 2023). But how is it decided whether an experience is truly conceivable or not? Albantakis et al. (2023) refer to “introspection and reason” but we follow Popper in that such intuitive notions cannot form an indisputable basis for feature selection as all intuitively derived features of concepts can be contested by opposing intuitive notions, which results in unresolvable disagreements.⁴ Therefore, although philosophical considerations on the essential features of experience go back as far as Kant (1781), according to Popper, an indisputable conclusion can never be reached. Tellingly, Kant’s proposed essential features of sensory experience, time and space, are not included in IIT’s “axioms of phenomenal existence” (Albantakis et al. 2023).

Defining essential features leads to an infinite regression of terms

A second issue that arises from finding essential features by intuition concerns precision in language. Like others (see, e.g. Musgrave and Lakatos 1965), we suggest that everyday labeling of objects mostly relies on pattern recognition rather than adherence to explicit features of a concept. When we are asked to find a differentiating feature between two concepts, we are likely to hesitate before answering because we need to first convert our intuitive, non-verbal patterns upon which we base our concepts into verbal features. In essentialism, these verbal features are supposed to capture our intuitive notion (“the essence”) completely and precisely. However, Popper states that this conversion of intuitive notion into fully precise language is neither possible nor desirable, which is supported by the following argument. For a definition to be understood correctly such that there is no misunderstanding of what it means to exhibit a feature f_i, an essentialist would have to define each term used when describing f_i, leading to an infinite regression of terms.⁵

Essentialist definitions are vulnerable to intuition violations

We think another reason why intuition-derived feature-based definitions can be problematic is that they can result in counter-intuitive labeling, which we refer to as intuition violation: after an intuitive notion of a term T is converted into a definition with some features, a hypothetical entity can be constructed that exhibits all such features (and thus should be labeled T) but still violates the intuitive notions behind the definition.

Essentialist definitions are generally vulnerable to intuition violation, especially when an opponent constructs edge cases that are specifically designed to be counter-intuitive. To show this, let us consider an example of a definitional issue borrowed from ) Chalmers (2012) outside the realm of consciousness research. Chalmers considers what the essential discriminating feature between whales and fish is. We argue that no matter how well a definition of what it means to be a whale is crafted, we can always construct an (at least hypothetical) example of what we intuitively understand to be a fish but that nevertheless will get labeled as a whale according to that definition. For example, an essentialist may say that the feature of having lungs is an essential discriminating feature between whales and fish—but under this definition, lungfish (Dipnoi) are whales, which would likely violate the essentialist’s intuition. Thus, we agree with Chalmers that it is impossible to construct feature-based definitions that are fully resistant to intuition violation. To conclude, as Popper has argued, when we are forced to come up with precise essential features of concepts, we do not gain insights but end up splitting hairs.⁶

The core problem of essentialism is regarding definitions as scientific insights

By now, we have considered several reasons why essentialist approaches are flawed. As we will see in the section Examples of  Counterproductive Essentialism, those flaws can affect consciousness research negatively. Still, Popper did not oppose intuitively selecting features for a definition per se.⁷ Instead, the crucial issue Popper had is that, in methodological essentialism, such intuition-derived features are understood to be scientific insights. Not only is the essentiality of features not scientifically assessable according to Popper, but the insights of science do not lie in definitions, but in finding novel regularities between observations. This view forms the basis of an alternative view on the role of concepts in science that Popper calls methodological nominalism (see section Methodological Nominalism).

Examples of counterproductive essentialism

Overloading terms with unintended meaning

A problem particular to essentialist consciousness research lies in the attribution of meaning to labels. If, for instance, a ToC provides criteria for something to be conscious and something fulfills these criteria, it will be labeled conscious. But although this constitutes a mere technical operation, an essentialist may now associate all their intuitive notions of consciousness with the labeled entity instead of the actual criteria included in the definition. In other words, there exists the tendency to overload the meaning of the label “conscious” in essentialism. This can be misleading especially in ethical debates where rights might be preferentially granted to conscious entities. A similar sentiment is shared by Pennartz et al. (2019), who caution against applying the label “conscious” to intelligent machines and consider introducing a novel term to avoid essentialist confusions.

Intuition violation leads to premature rejection of theories

At least some ToCs will assign consciousness to entities that are not commonly thought to be conscious, especially when considering minimal systems that have been specifically constructed to fulfill a ToC’s criteria for consciousness. Because this assessment may violate another researcher’s intuition about what entities are conscious, they may decide to reject the ToC under scrutiny based on the outcome of the labeling alone rather than considering the practical utility of the ToC. An example of this can be found in Fleming et al. (2023), who reject IIT in part because of its panpsychism. Researchers may thus miss out on promising predictions for experiments because they prematurely rejected a ToC, e.g. on the predictions made by IIT in recent adversarial collaborations (Ferrante et al. 2025, Corcoran et al. 2026).

Essentialism obscures substantive disagreement

Chalmers (2012) differentiates between verbal and substantive disagreements in philosophy. Based on Chalmers’ ideas, we give the following (abstract) example of a merely verbal disagreement in consciousness research:

ToC #1 defines the term “conscious” based on features F₁. ToC #2 defines the same term based on features F₂. When examining a subject, researchers adhering to either theory agree that a subject X exhibits F₁ but not F₂. A merely verbal disagreement arises as to whether the term “conscious” should be applied to X.

In essentialism, such merely verbal disagreements about labeling are significant because definitions are understood to be insights. If a researcher questions another researcher’s definition, they therefore also question the insightfulness of their work. However, following Chalmers (2012), there is no substantive disagreement here as both parties agree on all observed features of the subject.

Let us investigate the substantiveness of disagreements between ToCs more closely. If two ToCs make seemingly opposing claims on brain activity patterns evoked by a consciously perceived stimulus (say: brain activity pattern A versus pattern B), but they also use different definitions of stimulus awareness, then it is unclear whether the two theories are merely in verbal or substantive disagreement. For example, if one ToC assesses stimulus awareness by the ability of the subject to report the presence of the stimulus (detection) while the other assesses stimulus awareness by the ability to distinguish the stimulus from other stimuli (discrimination), then the disagreement on which brain pattern is evoked when consciously perceiving a stimulus may solely arise from the disagreement on the assessment of consciousness. Even if two ToCs can be brought to agree on a method of determining stimulus awareness, they have substantive disagreement only if the brain activity patterns they predict are mutually exclusive (otherwise they have no disagreement at all).

Methodological nominalism

On the term “nominalism”

Popper (1945) proposes an alternative to methodological essentialism that he calls “methodological nominalism.”⁸ The term “nominalism” is often used to denote the view that a concept or class should be solely determined by enumerating the members of that class. To avoid confusion with the methodology introduced by Popper (1945), we suggest referring to this approach as “enumerative nominalism.”⁹ Enumerative nominalism eschews lists of features as definitions and instead relies on lists of members. As Popper (1962) rejects enumerative nominalism as a useful methodology, we will not consider it any further. In the following, we make a distinction between enumerative and methodological nominalism (as defined below), the latter forming the core of this text.

Arbitrary and imprecise definitions

Methodological nominalism as proposed by Popper (1945) still uses feature-based definitions, but under this view, coming up with a definition that captures the essence of a concept is not the goal of research. Instead, both the association of a term T with a definition D, as well as the chosen features F of D are considered “arbitrary” (but see section On the Usefulness of Definitions for a more nuanced discussion of definitional preference). Features associated with a concept can be selected intuitively but are not understood to reflect its essential nature. Thus, terms are not thought to be inherently meaningful but are regarded as mere “handy shorthand labels.”¹⁰ This bars discussions on the essentiality of features of concepts.

Furthermore, in methodological nominalism, all definitions are thought to be “imprecise” and should be specified only until a good-enough point is reached to conduct research.¹¹ This second point may seem unsatisfying as it may seem desirable to use fully precise terms to avoid ambiguity, but Popper holds that imprecision is inevitable and that it is better to acknowledge it rather than deny it.

Discovering novel regularities between observations as the goal of research

This brings us to the key difference between methodological essentialism and nominalism which lies in the purpose of concepts. In methodological nominalism, concepts and their definitions do not count as insights but are used as tools or steppingstones to reveal insights¹². This is also why concepts can be changed or discarded as needed. For Popper, insights are not achieved when the essence of a thing is revealed but when a new regularity between observations has been discovered¹³. He imagined such a discovery to stand independently of the terms used in its making. Such regularities are supposed to be captured in laws or ‘universal statements’¹⁴. We will briefly present Popper’s ideas on laws and scientific discoveries before considering how a law-focused approach can be beneficial to consciousness research.

Scientific laws as generalized forms of regularities between observations

Here, we supplement Popper (1945)’s thoughts on the role of concepts in research with his earlier, well-known thoughts on scientific discoveries (Popper 1935, Popper and Weiss 1959). According to Popper, scientists should look for (novel) regularities between observations. For now, we will equate an observation with the outcome of a measurement (Tal 2020). According to Popper, scientists should present a law as a generalization of a regularity between observations. Such laws are conjectures that can only be falsified, never verified (Popper and Weiss 1959). Nevertheless, if a law survives rigorous testing, it can be considered “corroborated.”¹⁵

Using scientific laws to derive causal explanations in experimental settings

Popper (1935) (English translation: Popper and Weiss (1959)) holds that a basic procedure of research is to give what he calls a causal explanation for events. In a causal explanation of an event, a predicted observation is derived from a set of initial conditions (also observations or “singular statements”) given a law or “universal statement.”¹⁶ We refer to such laws as scientific laws in the following. (To be clear, what Popper considers “causal” may be purely correlational from a mechanistic point of view.) In its simplest form, such a law is a conditional that says: “if we make observation(s) O_initial, then we predict observation(s) O_predicted”. The regularities that laws express can be specified with mathematical equations (which is the most common form of scientific laws in fields such as physics).

To give an example of a scientific law in consciousness research, consider the following statement: “If a subject reliably reports the presence of a stimulus (the initial observation), stimulus-related activity must be amplified and broadcast by prefrontal cortex which can be measured by electroencephalography (EEG; the predicted observation).” From this law, we can derive a causal explanation for a specific experiment and subsequently test the predictions of that causal explanation. If the predictions are met, we may infer a (hopefully new) regularity between observations.

Definitions as steppingstones for concept-based laws

Extending Popper’s line of thinking, we can combine causal explanations with feature-based definitions to arrive at a concept-based law by setting the initial observations of a causal explanation as the features of a feature-based definition. Importantly, for a law to result in more than just a tautology, O_predicted needs to be distinct from O_initial. The purpose of definitions in methodological nominalism is thus to form the basis of concept-based laws.

Scientific laws should survive experimental testing and make novel, risky predictions

Whether a proposed scientific law leads to scientific insights depends on two factors: first, a law needs to survive experimental testing. That is, the regularity between observations that it describes needs to be observed in practice. Here, Popper puts strong emphasis on falsification: just because a law led to an accurate prediction in a certain experimental setting does not mean the law will always hold true (as is well known, Popper resists such efforts at induction¹⁷); however, finding a violation of a law should lead to its rejection. As Popper’s emphasis on falsification has been criticized, which we discuss in section Criticisms of Popper’s Critical Rationalism. Nonetheless, Popper admits that we can assign a degree of corroboration to laws that survive stringent experimental testing.

Second, Popper (1962) holds that predictions of newly proposed scientific laws should be novel and risky.¹⁸ That is, based on a law, we are supposed to find regularities between observations we had not considered previously. Thus, if we want to evaluate the above GNWT-inspired law about stimulus reporting and prefrontal cortex activation as measured by electroencephalography, we should consider whether a regularity between those two observations would be novel or well-known.

Scientific findings can be stated independently from the terms used in their making

To conclude our examination of the role of concepts in methodological nominalism, let us see why, in methodological nominalism, no weight is attached to linking a term with a definition. Inspired by GNWT, let us define that to be conscious of a stimulus means to be able to report the presence of the stimulus. In essentialism, such a definition would be subject to intuitive criticism (is reportability an essential or inessential feature of consciousness?). In methodological nominalism, this definition on its own is not insightful and merely exists to helps us construct concept-based laws such as: “if a subject is conscious of a stimulus, we predict activation of the prefrontal cortex.” Next, say we perform an experiment and find corroborating evidence regarding the regularity between being conscious and prefrontal cortex activation. According to Popper, even though we used a particular definition of consciousness for this finding, we can always re-state such a finding without using the term “conscious.” That is, we can always substitute a (perhaps controversial) term with its definition to express a regularity independently from the term-definition link that was used in its discovery. In the above example, we can say that when a subject is capable of reporting a stimulus, prefrontal cortex activates (see section Global Neuronal Workspace Theory for a closer examination of this particular proposed regularity). A similar approach of substituting terms with their definitions in case of doubt is advocated by Chalmers (2012) to resolve merely verbal disputes. This is what Popper (1945) means when he writes: “[…] science does not use definitions in order to determine the meaning of its terms, but only in order to introduce handy shorthand labels.” (p. 234).

Methodological nominalism and theories of consciousness

In the following, we will highlight consequences of applying methodological nominalism to consciousness research. We will first emphasize the distinction between labels of consciousness and law-derived predictions. Then, we suggest a set of guidelines that ToCs should follow under methodological nominalism. Finally, we consider two prominent ToCs (GNWT and IIT) from a methodological nominalist point of view, while paying attention to a third theory (Neurorepresentationalism, advocated by one of this paper’s authors) in the Appendix.

Linking conscious content to behavioral capabilities

Labeling operations are not predictions

ToCs often set criteria for (i) when a subject is considered conscious, (ii) which brain areas are considered important in mediating awareness, and (iii) when a subject can be said to be aware of something (e.g. a stimulus). We stress that these conditions alone only enable labeling operations, i.e. the criteria are features of a definition (and thus not insightful on their own). Inconsistent with the terminology introduced above, some ToCs may refer to the criteria for a label of consciousness as a “prediction.” For example, Tononi et al. (2016) consider it a “key prediction” of IIT that posterior cortex is the substrate of conscious experience because they propose that posterior cortex is the site of maximum intrinsic cause-effect power in the brain. Following the above distinctions (see Table 1), posterior cortex being the substrate of consciousness is neither a law nor a prediction but a mere anatomical label.

Table 1.

Definitions and laws in essentialism versus nominalism.

			Methodological
	Description	Example (fictitious)	Essentialism	Nominalism
Definition	If X exhibits features F, term T applies to X	If a human subject’s EEG data shows markers of ignition, call the subject conscious	Insightful if observations X match intuitive notions of the essence of consciousness	Not insightful
Labeling	Intradefinitory features F of definition D apply to X, therefore term T applies to X	The subject’s EEG data shows markers of ignition; therefore, the subject is conscious	As above	Not insightful
Law	If observations Oinitial occur, predict observations Opredicted	If a subject’s EEG data shows markers of ignition, the subject will respond to questions through a BMI	Not important	Potentially insightful (requires corroboration)
Concept-based law	If intradefinitory features of definition D occur, predict observations Opredicted (extradefinitory features)	If a subject is conscious, they can respond to questions through a BMI	Not important	Potentially insightful (requires corroboration)

The examples are GNWT-inspired but simplified for brevity. Note that markers of ignition are left vague here such that specificity to the level of observations is not fully achieved. EEG: electroencephalography, BMI: brain-machine interface.

A scientific law needs to connect consciousness to observable features

Even if more specifics are provided regarding posterior cortex such as in the statement “if a subject is aware of a stimulus (inobservable), we should see sustained activity in posterior cortex (observable),” we do not arrive at a law because only one half of the conditional statement is observable. This issue is specific to psychology-based research because consciousness research often considers subjective (i.e. inobservable) experience as a predicted feature. It is only when awareness is credibly linked to observable features such as behavioral capabilities (e.g. the ability to perform psychophysical tasks) that we arrive at a law. An example of such a law would be: “If a subject is aware of a stimulus (inobservable), they should be able to report the stimulus (observable). Then, we should see sustained activity in posterior cortex (observable).” We think that awareness being a requirement for certain types of behavior is too often left implicit in consciousness research while it can be a crucial component for testability.

Behavioral capabilities reflect one of many possible predictions for consciousness-based laws

While links to behavioral capabilities are often needed to transform labeling operations into laws concerning consciousness, ToCs do not necessarily need to predict behavioral capabilities from labels of consciousness for testability. Instead, predicting any type of measurable observation that is not included in the label itself can work. Assuming that a label is based on physical properties of the substrate of consciousness, a law may as well predict other, extradefinitory physical properties. For example, based on GNWT, one may propose a law in which we predict that post-stimulus ignition in prefrontal cortex (the intradefinitory observation required for consciousness of a stimulus) leads to a wave of secondary activation in early sensory areas.

Predictions on behavioral capabilities may not hold for minimal systems

If one is explicit about the assumption that awareness is a requirement to solve many kinds of tasks, artificial neural networks (ANNs) pose an issue. ANNs can solve many tasks without fulfilling the criteria for consciousness of most ToCs (Doerig et al. 2019, Pennartz et al. 2019). The consequence of this is that laws derived from ToCs that link neural activity to cognitive capabilities and consciousness may have to be restricted to animals to not be immediately refuted. In other words, the labels “subject A is conscious” and “subject B is unconscious” are meaningful if subjects A and B are animals because it is implied that subject A is capable of certain types of behavior that subject B is incapable of. But if subjects A and B are artificial systems, the labels may become meaningless: the subjects may or may not be capable of certain kinds of behavior regardless of their conscious state. We therefore recommend ToCs to be explicit about the relationship between awareness and behavior: in what kind of systems are the neural markers of consciousness necessary and/or sufficient for awareness-associated behavioral capabilities? We thus recommend ToCs to restrict their scope to systems for which they can provide more than meaningless labels. In many cases, this may mean restricting a ToC to animals (perhaps only to specific taxa, such as vertebrates or even mammals).

Nominalist guidelines for theories of consciousness

After distinguishing labels of consciousness from law-based predictions, we will now briefly develop four guidelines that we will use to assess whether two prominent ToCs adhere to methodological nominalism. Depending on the research field, these guidelines may coincide with common standards of scientific practice; however, we think that consciousness research is especially prone to essentialism and may profit from a refocusing on scientific laws rather than essentialist definitions.

G1. Definitions should not be presented as insights

We recommend not to present definitions of consciousness as scientific insights but instead as promising starting points of research. We do not think the conceivability argument carries weight in justifying the essentiality of features of consciousness. Intuitive notions of consciousness can be used to come up with feature-based definitions but it should not be assumed that an intuitive notion is shared by other researchers. Instead of the definition itself, the utility of a particular definition as a basis for concept-based law should be highlighted (if possible).

G2. The scope of a theory of consciousness should be clear and restricted

Because we can no longer assume a common essence of consciousness that every ToC must describe, we recommend authors to be explicit and restrictive about the kinds of phenomena they want a particular ToC to cover. When ToCs claim to “explain consciousness,” they may target very different goals. One ToC may try to explain how and where integrated and differentiated information arises in the brain, while another may try to explain why prefrontal cortex activity arises in a report paradigm. Important for restricting the scope of a ToC should be the novelty and riskiness of the laws it proposes (see below).

G3. Testable scientific laws should be proposed to explain phenomena in the target scope

ToCs should propose scientific laws for the phenomena that they include in their scope. Laws may characterize relationships between initial conditions and predictions in an abstract fashion (e.g. “PFC activity is required for report behavior”). However, as vagueness is seen as a major flaw of current ToCs (Mudrik et al. 2025), ideally, such laws should come with experimental test cases in which the initial conditions and predictions are described on the level of observations. Such test cases should specify important experimental variables such as subject species, measurement modality, or experimental paradigm. Relatedly, Evers et al. (2024) go further and demand that ToCs should present experimental testing that could lead to their own rejection. When it comes to characterizing the relationship between initial conditions and predictions, both mathematical formulations and determination of the necessity and sufficiency of the initial conditions for the predictions may be helpful for applying laws to novel use cases. We acknowledge that presenting such laws and test cases may not always be possible, especially for novel ToCs that may be in need of further development. In that case, ToCs should be explicit about the fact that their current effective scope does not match their desired scope.

G4. The novelty and riskiness of laws should be considered

We recommend considering the novelty and riskiness of the laws that are bundled in a ToC. Assuming that a particular law is corroborated, would the regularity that it describes be novel or already well known? For now, we make no claims as to the desired coherence of multiple laws of a ToC (but see section Observations and Subjective Experience). However, it makes sense for ToCs to emphasize laws that are unique to their theory. For example, consider a hypothetical ToC that makes novel and risky predictions about the reportable content of consciousness based on neuronal activity. We do not think it would be particularly useful to extend such a ToC, e.g. with the commonly held belief that unconscious states arise from a disconnect between thalamus and cortex (even though that would increase the scope of this ToC). Therefore, we encourage authors to highlight those scientific laws of their ToCs that make novel and risky predictions.

Evaluating two theories of consciousness with nominalist guidelines

In the following, we evaluate two prominent ToCs based on the developed guidelines. For brevity, we only provide a short summary of each ToC and refer readers to a key publication for further information. We consider only a subset of the predictions made by each ToC and refer interested readers to the ConTraSt database for an overview over experimental testing of theories (Yaron et al. 2022). The goal of this section is to show how ToCs and their presumed laws may become more productive tools of research based on methodological nominalism.

Integrated information theory

Brief summary

IIT (Albantakis et al. 2023) begins its definition of consciousness with a list of five axioms of phenomenal existence (six if you count the zeroth axiom), which are essentialist features of experience: intrinsicality, information, integration, exclusion, and composition. Corresponding to the five axioms of phenomenal existence, IIT lists five postulates of physical existence, which are requirements for physical substrates to be conscious. According to the postulates, physical substrates for consciousness are thought of as systems of units (e.g. neurons) with cause-effect power, i.e. the units have states (e.g. firing, not firing) that can influence and can be influenced by the states of other units in the system to which they are connected. If the cause-effect structure of a system follows the five postulates of physical existence, it is (by definition) conscious. To show this, it is necessary to “unfold” the cause-effect structure, which requires computing information-theoretic measures of the system in different states and assessing whether the system can be broken into smaller systems or subsumed into a larger one.

A key measure obtained from unfolding the cause-effect structure is the amount of integrated information Φ. This measure also identifies a part of the system with the “maximally irreducible cause-effect structure,” which corresponds to the part of the system that has awareness. In the human brain, this part is supposed to be posterior cortex (Boly et al. 2017). The units of the cause-effect structure do not necessarily have to be neurons, so that IIT’s definition of consciousness can be applied to non-biological systems. Regarding the content of consciousness, IIT proposes an “explanatory identity” between the cause-effect structure of the system and the phenomenal properties of the experience of the system. For example, the feeling of extendedness of space (a phenomenal property) is attributed to the fact that the units of the substrate of the experience are connected in a grid-like manner (a physical property) (Haun and Tononi 2019).

I‌IT’s definition of consciousness is presented as an insight

IIT’s definition of consciousness (the axioms and postulates) is a prime example of essentialism in consciousness research. Consider for example the phrasing from (Albantakis et al. 2023): “[..] IIT sets out to identify the essential properties of consciousness—those that are immediate and irrefutably true of every conceivable experience.” Based on the above, we argue that IIT’s essentialist tone make it vulnerable to intuitive objections from other researchers. A prominent example of this is the objection by Aaronson (2014) and others (Pennartz 2009, Fleming et al. 2023) that IIT assigns (some form of) conscious experience to entities that are not commonly regarded as conscious, such as a network of logic gates arranged in an expander graph. Here, IIT runs the essentialist risk of rejection based on intuitive notions before its practical utility is considered. We recommend IIT to express their definition of what features of experience it wants to study as a starting point rather than as an essentialist, irrefutable insight.

I‌IT’s scope is ambitious

IIT intends to cover all phenomenal aspects of experience including the state of being conscious and the content of consciousness. Specifically, Albantakis et al. (2023) write: “IIT aims to answer two questions: 1) Why is experience present vs. absent? and 2) Why do specific experiences feel the way they do?”. Having an ambitious scope for a ToC is not necessarily bad, but we recommend being explicit about the effective coverage of the desired scope with scientific laws. Regarding the state of consciousness, IIT proposes a measure of integrated information Φ that is supposed to quantify the level of experience. Regarding the content of consciousness, recent publications of IIT have explored possible relationships between features of the substrate of experience and features of the experience itself (Haun and Tononi 2019). We will now explore how well these issues are covered by scientific laws.

I‌IT provides laws but should disentangle them from labels

IIT has faced (controversial) accusations of being pseudoscientific (Fleming et al. 2023, Klincewicz et al. 2025). We distance ourselves from such accusations. Unquestionably, IIT has made concrete testable scientific predictions, including those assessed in the recent Cogitate adversarial collaboration (Ferrante et al. 2025). However, as mentioned above, IIT is probably at its most vulnerable to such accusations when it highlights labeling operations but does not mention explicit predictions that should follow from labels.

Consider for instance Φ, the proposed measure of integrated information, that should serve as a quantification of consciousness. This measure is currently not assessable in complex biological systems, but even if it were, to produce a scientific law, IIT would have to predict extradefinitory features that correlate with having low or high Φ. The utility of Φ has attracted scrutiny because a 2D system of logic gates can exhibit high Φ (but does not seem to exhibit any other noteworthy features) (Aaronson 2014). We therefore recommend IIT to limit its scope to systems for which Φ could produce a meaningful label (“conscious” or “unconscious”). Regarding potential Φ-based laws, the Perturbational Complexity Index (PCI) is a complexity measure for which regularities between intra- and extradefinitory features have been observed; however, it is only loosely connected to Φ (Casali et al. 2013, Casarotto et al. 2016). Similar arguments apply to IIT’s efforts to characterize the relationship between substrate and experience features. According to IIT, substrate features (e.g. anatomical wiring) are supposed to determine experiential features but we stress that one needs observable features associated with those experiential features to arrive at a scientific law.

I‌IT makes both common and unique predictions

The uniqueness of some of IIT’s predictions can be contested. For example, it has been argued that the proposed relationship between the complexity of neuronal responses and behavioral state is not exclusive to IIT but is also predicted by other ToCs (Farisco and Changeux 2023). Furthermore, studies on the relationship between EEG complexity and conscious states exist that preceded publications on the PCI (Zhang et al. 2001). Still, some of IIT’s predictions are clearly unique and risky. For example, a prediction unique to IIT is that optogenetic inactivation of already inactive (i.e. non-firing, silent) neurons in mouse visual cortex should negatively affect task performance in mice (Takahashi et al. 2025, Corcoran et al. 2026). This prediction directly relies on IIT’s definitional basis that holds that inactivating units would remove them from the cause-effect structure, which should in turn affect experience. Testing of this prediction (and other predictions derived from it) will thus reveal the utility of IIT’s definitional basis (which on its own cannot count as an insight in methodological nominalism).

We conclude that IIT is a ToC with a large definitional basis (unfortunately presented in an essentialist manner) and an ambitious scope. In methodological nominalism, the utility of any definitional basis needs to prove itself in practice in the form of concept-based laws. While some of IIT’s predictions have seen corroboration, the verdict on some of its riskier predictions is still out.

Global neuronal workspace theory

Brief summary

In its definition of consciousness, GNWT (Mashour et al. 2020) places the concept of “global information availability” centrally. GNWT divides the brain into two compartments: (i) localized, specialized, and modular brain areas; and (ii) a global neuronal workspace. Prefrontal cortex is considered to be a major component of the workspace. Local modules send information to the workspace where, if excitation reaches a certain threshold, that information is amplified and broadcast to all local modules. This amplification and broadcasting are referred to as ignition. By definition, a subject is aware of a stimulus when ignition is reached, otherwise that subject is unaware of the stimulus (“all-or-none”). Because of the broadcasting of information, a subject can use global information for a variety of behavioral and cognitive capabilities that are otherwise unavailable. This assumption forms the basis for GNWT’s laws between neural markers of ignition and behavioral capabilities.

Statements about consciousness: definitions or laws?

According to GNWT, “[g]lobal availability is what we experience as a conscious state” (Dehaene et al. 2001, Mashour et al. 2020). There are two ways to interpret this statement: one is as a definition, i.e. every entity that exhibits global information availability is by definition conscious. Such a definition on its own would not be insightful in methodological nominalism. The second, less direct interpretation is that global availability of information should be generally predictive of many (as yet unspecified) behavioral manifestations of consciousness. In that case, we may set out to develop more concrete scientific laws from that statement. This distinction may seem narrow-minded, but we recommend ToCs to be explicit as to whether their statements on consciousness should be interpreted as definitions or (abstract) laws. For example, from the above quotation alone, it is unclear whether encountering an entity without global information availability but displaying common behavioral markers of consciousness would lead to a refutation of GNWT or whether such entity would simply be labeled “not conscious” by it.

GNWT’s scope is more restricted than that of IIT

Like IIT, GNWT wants to explain (i) what entities can be conscious, (ii) when such entities are in a conscious (versus unconscious) state, and (iii) how information in such entities can become conscious. Unlike IIT, GNWT does not set out to explain how connectivity features of brain areas may affect what types of experiences they can elicit. Although an essentialist may assert that a ToC should explain all of the intuitive notions they associate with the term consciousness, in methodological nominalism, what counts is the effective coverage of a desired scope with scientific laws rather than the size of the scope itself.

GNWT has proposed a well-tested law (that has seen some exceptions)

This brings us to GNWT’s proposed scientific laws that center on the proposed mechanism of broadcasting by the global neuronal workspace, leading to global information availability. We have already mentioned two (abstract) laws based on this mechanism: first, broadcasting of information is needed for report capabilities. Second, broadcasting of information leads to secondary activation of specialized brain modules that do not belong to the workspace. Clearly, the first abstract law can easily be violated by minimal artificial systems and thus needs restriction; moreover, it can be argued that reportability is part of GNWT’s definition, which renders the tight relationship between broadcasting and reportability intradefinitional. However, the second abstract law may hold for biological and artificial entities alike (although one may argue that circularity arises if the workspace is defined by its ability to broadcast and thus naturally elicits secondary activation).

The first abstract law is of special interest because it has been subject to intense scrutiny and experimental testing. For testing to be possible, the law needs to specify which brain areas belong to the global neuronal workspace in a subject species, as well as what markers of broadcasting should look like. In practice, GNWT readily provides these specifications. Prefrontal cortex is a notable component of the workspace in mammals, and broadcasting (or “ignition”) is marked by measurable bifurcation dynamics. When it comes to prefrontal cortex activation and stimulus awareness (as measured by report capability), the results have been mixed. In some cases, the proposed regularity seems to hold (Dwarakanath et al. 2023, Kapoor et al. 2022). However, investigations with no-report paradigms (where report capability is first verified, but the subject is no longer asked to execute the report) and/or task-irrelevant stimuli (or stimulus features) have shown that prefrontal cortex does not always show the predicted bifurcation dynamics (Pitts et al. 2012, 2014; Frässle et al. 2014; Sergent et al. 2021).

GNWT’s emphasis on prefrontal cortex activation is risky

GNWT’s prediction of prefrontal cortex activation during all kinds of stimulus awareness (both task-relevant and irrelevant) is unique, risky, and testable, for which the authors behind GNWT should be commended. This is not true for all of GNWT’s considerations: for example, Mashour et al. (2020) link anesthetic-induced unconsciousness specifically to the disconnect of the workspace from other brain areas—but what novel predictions follow from this is unclear.

An essentialist objection to GNWT would be that because GNWT’s laws seem to work best in report paradigms, GNWT does not actually explain consciousness but decision-making or related functions. The nominalist take can be given without referring to an essentialist consciousness concept: GNWT’s laws concerning prefrontal cortex activation do not hold under all conditions. However, the report-related regularities discovered because of GNWT will endure even though the core assumption that all reportable stimuli need to evoke prefrontal cortex activation is becoming increasingly untenable.

Discussion

Summary

Here, we first discussed the essentialist belief that we can select ToCs based on whether their description of consciousness matches the true nature of consciousness. With the help of Popper (1945), we concluded that it is impossible to determine an unquestionable list of essential features of consciousness because there is no scientific but only an intuitive method of selecting them. We therefore rejected the essentialist approach to evaluating ToCs. We next examined Popper’s alternative to essentialism, called methodological nominalism. In this stance, any definition of consciousness is understood as a choice and mere starting point of research rather than an insight in itself. That is because in methodological nominalism, research is supposed to focus on laws that state expected regularities between observations. We conclude that ToCs should not be evaluated intuitively, but by the insights that their laws will offer if they are corroborated. On the side of theory development, we propose guidelines for non-essentialist consciousness research to shift the focus away from intuitive evaluations of ToCs to proposing and testing of laws.

Our discussion of essentialism versus nominalism heavily relies on Popper’s philosophy of science as specified in Popper (1935, 1945) and Popper and Weiss (1959). As this philosophy has been criticized in the past, we will now briefly discuss prominent criticisms of Popper as well as what they mean for the approach outlined here.

Criticisms of Popper’s critical rationalism

Popper’s epistemological philosophy is called critical rationalism. One aspect of critical rationalism that has been criticized is falsificationism, i.e. the belief that observations can lead to the refutation of scientific theories. The most prominent objection to that is the Duhem–Quine problem (Quine 1951) which states that, when evidence that contradicts a theory is found, such evidence may be the result of violated background assumptions (e.g. a flawed experimental design or a technical error) rather than an actual violation of a theory—and thus the rejection of a theory becomes a choice rather than a logical necessity (Lakatos 1978). More criticism of falsificationism comes from Kuhn (1970), who identifies sociological aspects of adherence to theories, besides considerations of scientific evidence. Josephson and Josephson (1994) also advocate de-emphasizing falsification in favor of selecting the most likely explanation for observations. Finally, Salmon (1981) doubts that Popper’s views on corroboration can justify selecting more corroborated theories over less corroborated ones in the context of practical decision-making without any reference to induction.

In this article, we have avoided characterizing falsification of laws as the goal of research. Instead, we intended to highlight the practical utility that laws can have for researchers that want to derive interesting predictions for their experiments or use them to design experiments. Whether falsification of theories is desirable or possible is left to the researcher and the expert community. We also avoid making claims as to whether a theory should be rejected or adjusted when contradicting evidence is found.

On the usefulness of definitions

That definitions are understood to be arbitrary in methodological nominalism may raise several objections. We do not mean to suggest that all definitions of consciousness are equally useful (they are not) or that researchers should stop defining concepts according to their particular preferences (they still can). Researchers should be encouraged to define or circumscribe consciousness, if only to clarify what they take the concept to mean. But we do argue that particular definitions of concepts such as consciousness draw their utility from their use in discovering new regularities between observations and proposing scientific laws.

Comparison to operationalism

Arguably, the role of definitions in methodological nominalism resembles that used in operationalism, where the features of definitions should be specified to the level of observations which we equate with the outcome of measurements. Compare this to Bridgman (1927), who writes that a concept should be understood to be “synonymous with [a] corresponding set of measurements.” Two remarks can be noted about the comparison of nominalist definitions and operationalism. First, in methodological nominalism, there can be no argument whether a definition is a correct operationalization of a more abstract concept—at least not by referring to essential features of the more abstract concept. The question of whether to group a definition with features on the level of observations (e.g. of working memory) under a more abstract definition can instead only be answered by considering whether such grouping would be useful for proposing scientific laws (for a different view on the relationship between concepts and operationalizations, see, e.g. Francken et al. 2022).

Second, Popper warns that defining terms by the outcome of measurements does not extinguish imprecision that is inherent to definitions. To define and delineate outcomes, interpretations are inevitable, involving theoretical concepts. According to Popper (1962), there are no fully precise descriptions of measurements that could eradicate all ambiguity because each measurement would have to be defined by another measurement, leading to another infinite regression of terms.¹⁹

Observations and subjective experience

An objection to methodological nominalism specific to consciousness research holds that it posits as the general goal of science to find novel regularities between observations, whereas subjective experience is not observable. Therefore, methodological nominalism is not suitable for consciousness research. A stronger version of this objection would be that methodological nominalism inevitably results in eliminativism, i.e. the assertion of the nonexistence of subjective experience. This is not the case: methodological nominalism makes no claims about the existence of non-observables. However, if a relationship between inobservable features of experience and observable features (such as those of the substrate of experience) is proposed, a second set of predicted observable features is needed to turn that relationship into a scientific law. Without such a second set of predicted observable features, the relationship should not be understood as an insight.

Limitations

Use cases of definitions not considered by methodological nominalism

In this text, we refer to a use case of definitions that requires specification of definitional features to the level of observations and commitment to such features for experimental testing when used in a law. However, there may be other use cases of definitions such as in modest conceptual clarifications or circumscriptions of an explanandum. In such use cases, definitional features may be proposed tentatively, i.e. without strong commitment. Although such tentative (perhaps abstract) definitions cannot form the basis of concept-based laws per se, we still think that the framing chosen by such definitions is important since it will influence what regularities between observations researchers will consider in more concrete experimental settings (similar to how a Kuhnian paradigm determines what experiments researchers consider worthwhile pursuing) (Kuhn 1970). Methodological nominalism does not address such use cases of definitions.

Coherence of laws within a theory

In this article, we did not specify the relationships between the laws of a single ToC, and we left open to what degree the laws of a theory should be logically connected. One option is that laws within a ToC do not need to have any logical relationship, instead forming a loose patchwork. Another option is the one presented by Popper (1935), who thought that theories should build “axiomatized systems” in which a set of core axioms is used to derive all further laws. We would argue that most ToCs do not follow Popper’s ideal and we are neutral as to whether this would be desirable. Nevertheless, the problem of logical relationships between laws becomes an issue if falsification of ToCs is desired (Chis-Ciure et al. 2024). Related to this is also the role of theory leads, i.e. the initial creators and/or main proponents of ToCs. If laws are logically derived from a core set of axiomatic statements, then coming up with new laws should not require the help or approval of a theory lead. If theories are loose patchworks of laws, then who should get to decide whether a new law fits a theory? This question raises the point that theories with well-connected and coherent laws may have an advantage over loose patchwork of regularities.

Coherence between definitory features

Before ToCs propose laws, i.e. expected regularities between observations, some of them “derive” features of the physical substrate of consciousness from features of conscious experience. Questioning the reasoning behind these derivations lies outside of the scope of methodological nominalism since they occur during the definition of consciousness. This leaves unanswered the question how tight the link between the definitory features of consciousness should be. Consider for example the brain areas that GNWT and IIT consider important for consciousness. How strictly does prefrontal cortex follow as the key part of the global neuronal workspace from the description of the mechanism of consciousness envisioned by Mashour et al. (2020)? Similarly, how strictly does it follow from IIT’s definition that consciousness should be attributed to a “posterior hot zone” (Boly et al. 2017)? In both cases, to us, the linkage between substrate properties and implicated brain areas constitutes an educated guess rather than a strict, logical derivation. In a final example, Bayne (2018) calls into question that IIT’s “postulates” about the physical substrate of experience follow from its “axioms” of experience. What does this mean for laws that build on the definitory foundation of a ToC? If a study would show that prefrontal cortex failed to activate during a report paradigm, does this mean that GNWT would be wrong about the link between global availability and behavioral capabilities, or does it mean that it was wrong about which brain area implements global availability? From a nominalist perspective, we can demand that laws need to describe events on the level of observations. But methodological nominalism makes no claim about whether ambiguity in the derivation of the physical properties of a substrate should play a role when considering ToCs.

Observations and theoretical assumptions

In section Nominalist Guidelines for Theories of Consciousness, we demanded that ToCs provide experimental test cases for their laws that are specified to the level of observations. As Popper intends laws to be “universal,” they should also be applicable to experimental settings that were not explicitly referenced by their original proponents. As methodological nominalism does not believe in fully precise language, there is interpretation required to apply a law to a novel experimental scenario including the assumption of auxiliary hypotheses (Quine 1951, Chis-Ciure et al. 2024, Corcoran et al. 2026). But when is an application of a law a valid interpretation of its premises, and when is its application invalid?

Related to this question is the issue of theory-ladenness of observations, i.e. our prior beliefs influence what we accept as an observation. For example, in the studies cited above, activations of prefrontal cortex have been measured with EEG, functional magnetic resonance imaging (fMRI), and intracranial electrophysiological recordings—but each of these techniques comes with its own theoretical basis that we need to accept if we want to evaluate whether laws make accurate predictions based on their findings. In another example, conflicting findings on the role of primary visual cortex in binocular rivalry processing have been made depending on whether neuronal activity was measured with fMRI or intracranial recordings (Blake et al. 2014).

We do not think that methodological nominalism can eliminate interpretations that are required to bridge the gap between abstract concepts and concrete instances. However, unlike arguments about essential features of concepts, we do not think that this gap poses a major issue in the current practice of consciousness research.

Theories of consciousness beyond falsification

Despite essentialist undertones, many ToCs are already serving an important function in consciousness research: they guide researchers to design experiments that they otherwise would not have thought of. For example, would the role of prefrontal cortex have been studied as closely as in Sergent et al. (2021) without GNWT? Would decoding of task (ir-)relevant stimulus features from EEG data have been studied as carefully as in Ferrante et al. (2025) and would the role of inactive neurons attract as much scrutiny in Takahashi et al. (2025) without adversarial collaboration between ToCs? In this sense, nominalist goals are achieved even though the involved ToCs themselves may have been envisioned as essentialist: ToCs aid in the discovery of regularities between observations we previously had not thought of considering. Such regularities will count as insights even if all ToCs involved are eventually discarded.

Conclusion

Let us conclude by revisiting the introductory question: What is consciousness? What role should this question play in consciousness research after rejecting essentialism? We leave it to Popper (1945) to explain the nominalist take on “what is” questions in research:

The methodological nominalist will never think that a question like “What is energy?” or “What is movement?” or “What is an atom?” is an important question for physics; but he will attach importance to a question like: “How can the energy of the sun be made useful?” or “How does a planet move?” or “Under what condition does an atom radiate light?” And to those philosophers who tell him that before having answered the “what is” question he cannot hope to give exact answers to any of the “how” questions, he will reply, if at all, by pointing out that he much prefers that modest degree of exactness which he can achieve by his methods to the pretentious muddle which they have achieved by theirs. (p. 30)

Author contributions

Mathis Bassler (Conceptualization [equal], Investigation [equal], Writing—original draft [equal]) and Cyriel M. A. Pennartz (Funding acquisition [equal], Supervision [equal], Writing—original draft [equal], Writing—review & editing [equal])

Conflict of interest

None declared.

Funding

This project was supported by the Human Brain Project Specific Grant Agreement 3 (Grant agreement ID: 945539, doi:10.3030/945539) and by grant TWCF#0646 from the Templeton World Charity Foundation, both awarded to C.M.A.P. The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Templeton World Charity Foundation.