Eliminative Materialism

Eliminative materialism holds that some familiar mental categories do not refer to real kinds and will not survive a mature science of mind. Beliefs, desires, intentions, or qualia may resemble phlogiston and demonic possession: useful elements of an inherited explanatory scheme that should be replaced rather than reduced. The view is stronger than physicalism. A physicalist may identify or functionally realise beliefs; an eliminativist says there are no beliefs of the ordinary kind to identify.

The position is easiest to caricature as denying the obvious. Its serious form is a claim about scientific change. Everyday psychology is treated as a theory whose posits earn their place through explanation and prediction, not through familiarity alone. If the theory is deeply mistaken, preserving its ontology may obstruct understanding. The corresponding burden is demanding: a replacement must explain the successes of belief-desire discourse, the practice of giving reasons, and the apparent immediacy of experience without covertly restoring what it eliminates.

Folk psychology as a theory

Paul and Patricia Churchland describe folk psychology as a body of lawlike generalisations connecting beliefs, desires, perceptions, intentions, emotions, and actions:

• if someone desires $G$ and believes action $A$ will secure $G$, they will tend to do $A$;
• seeing that $p$ tends to produce the belief that $p$;
• believing $p$ and believing that $p\to q$ tends to support believing $q$.

The terms gain content through their places in this network, much as theoretical terms in science gain content through laws. On the theory-theory view, attributing a mind is deploying a tacit theory that predicts and explains behaviour.

If folk psychology is a theory, three outcomes are possible. It may reduce to neuroscience, as temperature reduces to statistical mechanics. It may remain an autonomous higher-level theory, as non-reductive physicalism proposes. Or its central posits may fail to map onto the structure discovered by science and be eliminated.

The eliminativist stresses alleged stagnation and gaps. Folk psychology says little about sleep, mental illness, creativity, memory mechanisms, learning, and fine-grained cognitive difference. Neuroscience may classify states along dimensions that cross-cut belief, desire, and emotion rather than vindicate them.

The historical analogy

Science has eliminated entities once taken seriously: phlogiston, caloric fluid, crystalline spheres, and vital forces. Reduction is not guaranteed merely because a category is entrenched. If no lower-level kind plays the expected role, retaining the term can preserve a false picture.

The analogy has force against a conservative inference from ordinary use to ontology. It does not establish that mental categories are like failed scientific posits. Belief and desire participate in successful prediction, deliberation, interpretation, and social coordination. They may resemble biological categories such as predator or gene, which survive through revision and multiple realisation, more than phlogiston.

There is also a difference in evidential position. Phlogiston was posited to explain public phenomena and could be discarded when a better theory explained them. Pain and experience are not known only as theoretical causes of behaviour; subjects appear acquainted with them directly. This blocks a simple historical analogy for phenomenal states, though it does not guarantee that introspection classifies those states correctly.

What might be eliminated?

Eliminativism is not all-or-nothing. It can target different categories:

The first four can be revisionary without denying every mental phenomenon. The fifth is the most radical and the hardest to formulate coherently. A theory should state whether it eliminates a word, a folk essence, a scientific kind, or the phenomenon itself. Otherwise critics and defenders talk past one another.

Reduction versus elimination

A reduction preserves the truth of many higher-level claims by identifying or deriving their subject matter. Water survived identification with $H_{2}O$; caloric did not survive kinetic theory. The difference is not that one vocabulary changes. It is whether the old theory's central claims are approximately true under the new account.

For belief, a reduction might identify a functional or representational state that plays enough of the folk role. Elimination predicts instead that scientific categories will not preserve the inferential unity, determinacy, or causal structure associated with belief. Neural populations may carry graded, context-sensitive information without encoding one proposition the subject simply believes or disbelieves.

Between the extremes lies revision. Memory is not one faculty, but working, episodic, semantic, and procedural memory remain useful kinds. Similar fragmentation may refine belief or emotion without removing them. An eliminativist argument must show not only mismatch but mismatch severe enough that revision cannot preserve reference.

The self-refutation objection

The standard objection is immediate: if eliminativism is true, nobody believes that eliminativism is true. How can the view be asserted, justified, or rationally accepted? The utterance appears to presuppose the propositional attitudes it denies.

The eliminativist has several replies.

1. The theory is stated in current language during a transition, as early chemists used inherited terms while revising them.
2. Assertion and inference may eventually be explained through non-doxastic neurocomputational categories.
3. The view need eliminate only the folk conception of belief, not every information- bearing or inferential state.

The third reply is strongest but moderates the thesis. If there remain states with content, inferential relations, truth sensitivity, and causal roles in assertion, much of what belief was introduced to explain survives. The dispute may then concern the precision and architecture of belief rather than its existence.

Self-refutation is therefore not a one-line logical disproof. It sets a replacement condition: the successor vocabulary must explain how the theory itself can be represented, supported by evidence, and preferred for reasons.

Predictive success and the personal level

Belief-desire explanation predicts action across indefinitely varied contexts. Knowing that someone wants coffee and believes the cafe is east can explain walking east without specifying muscles or neurons. It also rationalises: the action becomes intelligible as a response to reasons.

The eliminativist can concede practical usefulness while denying literal truth, as one might use an idealised fiction. But successful novel prediction is ordinarily evidence that a theory captures structure. Folk psychology also supports counterfactuals: had the person learned the cafe was closed, the action would have changed.

Its failures are real. People confabulate, act from biases they disavow, hold fragmented or context-sensitive attitudes, and violate ideal rationality. These may show that belief is graded and dispositional rather than absent. A replacement must improve on the personal level without drowning its generalisations in microphysical detail.

The functionalism and non-reductive alternatives treat this predictive success as evidence for real higher-level kinds. Eliminativism treats the same success as limited and likely to be surpassed. The issue is empirical in part: does neuroscience integrate with the categories, fragment them, or make them idle?

Content and normativity

Beliefs are not grouped only by causal profile. They can be true or false, justified or unjustified, and logically related. A neurocomputational replacement must explain how physical states acquire content and how normative relations enter reasoning.

One option is to eliminate determinate content as well: cognitive systems transform distributed states without encoding propositions. This fits some connectionist models but makes communication and systematic error difficult to describe. Another option naturalises content through causal, informational, or biological function. That keeps intentional states under a revised account and is therefore less eliminative.

Normativity sharpens the problem. A causal sequence can conform to an inference without occurring because the conclusion follows. If reasons explanation disappears, the eliminativist must show how epistemic evaluation survives or explain why it too should be replaced. Otherwise scientific evidence cannot rationally favour eliminativism; it can only cause its advocates to utter it.

Consciousness and illusionism

Illusionism is sometimes described as eliminativism about qualia. It does not usually deny wakefulness, access, perception, discrimination, or the occurrence of states subjects report. It denies that introspection reveals intrinsic, private, ineffable phenomenal properties of the kind assumed in hard-problem arguments.

On this view, cognitive systems construct simplified self-models and are disposed to judge that experience has special qualitative properties. The explanatory target is the seeming that qualia exist. If the mechanisms producing that judgment are fully explained, no further phenomenal property is needed.

Critics argue that an illusion requires an appearance and that explaining why subjects judge they are conscious leaves untouched why the judgment is accompanied by anything it is like. The illusionist replies that treating the seeming as an additional ineffable object simply repeats the posit being rejected. The disagreement concerns whether "seeming" can be exhausted by representational and access functions.

This is not identical with Churchland-style elimination of propositional attitudes. An illusionist needs beliefs or belief-like representations to explain the mistaken self-model, while an attitude eliminativist may retain phenomenal consciousness. The targets must remain separate.

Neuroscience and replacement vocabularies

Eliminativism is strongest when it offers a successor rather than a blank. Candidate vocabularies classify population dynamics, predictive models, control policies, attractor states, precision weighting, and learned representations. These can reveal structure hidden by folk categories.

Scientific novelty does not itself imply replacement. Different vocabularies may answer different questions. A neural account explains mechanism; a belief-desire account explains action through content and reasons. The lower level eliminates the higher only if it shows that the higher-level generalisations are systematically false or add no explanatory control.

The method page supplies the appropriate standard: compare coverage, intervention, generalisation, and failure modes. Brain images do not show beliefs to be unreal merely because beliefs are not visibly located in one region. Distributed implementation is compatible with higher-level reality.

A spectrum of revision

The practical alternatives form a spectrum:

1. Conservative realism: folk categories are broadly accurate and will be retained.
2. Reductive revision: categories survive but are identified with physical or functional structures.
3. Pluralist revision: some categories fragment, while others remain autonomous.
4. Instrumentalism: folk psychology is useful without literally describing inner kinds.
5. Elimination: the categories are false and should be replaced.

Evidence may place different concepts at different points. Pain, belief, emotion, self, and consciousness need not share a fate. This piecemeal approach is less dramatic than global eliminativism and more responsive to how scientific taxonomies actually change.

Assessment

Eliminative materialism is valuable even where its conclusion is rejected. It denies that introspective familiarity grants permanent ontological status and forces every mental category to earn its place. Its own standard applies reflexively: elimination earns its place only when a successor explains the old theory's successes and failures better, including the reasoning by which the successor is accepted.

Selected references

• Churchland, Paul M. "Eliminative Materialism and the Propositional Attitudes" (1981).
• Churchland, Patricia S. Neurophilosophy (1986).
• Dennett, Daniel C. The Intentional Stance (1987).
• Frankish, Keith. "Illusionism as a Theory of Consciousness" (2016).
• Ramsey, William. "Eliminative Materialism" in The Stanford Encyclopedia of Philosophy.