Supervenience

The basic idea

Pick any pair of properties: beauty and brushstroke, pain and neural firing, moral wrongness and bodily harm. Sometimes the second seems to fix the first without being identical to it. Two oil paintings made of identical pigment patterns must look identical; two brains in identical physical states must feel identical. Supervenience captures that intuition.

The slogan is the cleanest definition: no A-difference without a B-difference. If anything differs in some A-property, it must also differ in some B-property. The converse is not required — two B-distinct objects might still share their A-properties, since a slightly different brain state can produce the very same headache.

This asymmetry is the whole point. Supervenience marks one direction of dependence (A on B) without inflating it into identity or definitional reduction. Davidson reached for the term precisely because he wanted mental events to depend on physical events without obeying psychophysical laws.

A short genealogy

R. M. Hare's The Language of Morals (1952) imported "supervene" into philosophy: two acts identical in every natural respect must receive the same moral verdict. G. E. Moore had said much the same in Principia Ethica (1903) without the label. Donald Davidson planted the term in mind-body debates in "Mental Events" (1970), suggesting that mental characteristics are "in some sense dependent, or supervenient, on physical characteristics." Davidson's use was deliberately loose; Jaegwon Kim spent two decades formalising it, with the 1984 paper "Concepts of Supervenience" carving out weak, strong, and global versions. By the late 1980s the concept had spread to ethics (Blackburn), aesthetics (Levinson), and semantics (Putnam, Burge), becoming the lingua franca of non-reductive physicalism.

Three varieties: weak, strong, global

Kim's taxonomy is unavoidable in any serious discussion. The differences hinge on what counts as a "case" and on whether modality is involved.

• Weak supervenience. Within any single possible world, B-indiscernibles are A-indiscernibles. So in our world, no two physically identical brains differ mentally. But there could be other worlds where a different psychophysical regime holds.
• Strong supervenience. Across all possible worlds, B-indiscernibles are A-indiscernibles. If two brains are physically identical, they share mental states regardless of which world they live in. This adds modal force; it is the version most physicalists want.
• Global supervenience. Two worlds with the same total distribution of B-properties have the same total distribution of A-properties. Individuals across the worlds may not pair up, but the wholes do. This handles holistic properties — ecosystems, economies, social roles — where no single object carries the supervenient feature.

Strong supervenience entails weak. Strong does not in general entail global, nor vice versa. A clean way to see this: imagine two worlds where every brain is duplicated in some other brain at the same world but the worlds disagree on which mental state goes with that physical type. Strong supervenience fails; global supervenience can still hold.

Supervenience vs identity vs reduction

	Supervenience	Identity	Reduction
Slogan	No A without B	A is B	A-talk = B-talk
Modal force	Necessary covariation	Necessary sameness	Translation invariant
Multiple realization compatible?	Yes	No	No, unless disjunctive
Requires shared vocabulary?	No	No	Yes (bridge laws)
Example	Mental on physical	Water = H₂O	Thermodynamics → statistical mechanics
Asymmetric?	Yes — B fixes A, not vice versa	No — symmetric	Yes — higher to lower
Explains why A tracks B?	No, only that it does	Yes — they are one thing	Yes — by translation

The table makes Kim's central worry visible: supervenience is the weakest column. It tells you the dependence holds without saying what makes it hold. That is its strength when you want to stay neutral, and its weakness when you owe an explanation.

Worked example: pain in humans, octopuses, and silicon

Consider three creatures that all feel pain: a human (pain correlates with C-fibre firing in the dorsal horn), an octopus (no spinal cord, no C-fibres, but clear behavioural and physiological hallmarks of pain), and a hypothetical silicon AI whose pain-state is a particular computation. Three relations are on the table.

1. Type-identity. "Pain is C-fibre firing." Then octopuses and AIs cannot feel pain by definition. Too strong; loses multiple realization.
2. Token-identity plus supervenience. Each pain-event is identical to some physical event, and within each creature mental properties supervene on physical, but no shared physical type across species. Davidson's anomalous monism.
3. Strong supervenience without identity. Globally, any creature physically identical to a pained human must feel pain. Cross-species, no single physical kind realises pain everywhere. Functionalism's home turf.

All three options agree on supervenience and disagree on the deeper metaphysics. That is the lesson: supervenience constrains the space of theories without picking a winner.

Counterarguments

• Kim's exclusion problem. If the physical exhausts the causes of any physical effect, and the mental supervenes on the physical, the mental seems to do no causal work that is not already done by its physical base. Kim concluded — reluctantly — that non-reductive physicalism slides into either reduction or epiphenomenalism. Critics distinguish causation from constitution, but the worry is real.
• Mere covariation, not explanation. Kim (1990) himself called supervenience "a relation in search of a metaphysics." Two objects sharing every B-property and every A-property tells you nothing about why they share them. Without a deeper story (causation, grounding, identity), supervenience is descriptive, not explanatory.
• Modal scope worries. Strong supervenience invokes all metaphysically possible worlds. Sceptics about heavy-duty modality (notably Quine) reject the apparatus that gives the relation its force.
• Carnap-style anti-metaphysics. A logical-empiricist could argue the whole question is verbal: once you have specified the B-vocabulary, all A-talk that does not reduce is excess baggage. This challenges supervenience by challenging the legitimacy of irreducible higher-level discourse.
• Pre-emption from grounding. Recent metaphysics (Schaffer, Fine, Rosen, c. 2009 onward) prefers grounding — a primitive "in virtue of" relation — to supervenience. Grounding aims to capture what supervenience misses: the explanatory direction. If grounding does the work better, supervenience may be a useful diagnostic, not a fundamental relation.

Variants and refinements

• Multiple-domain supervenience. A can supervene on B globally while only partially supervening locally. Useful for context-sensitive properties (semantic content depending on environment, not just brain).
• Necessitation supervenience. Strengthens strong supervenience by requiring that B-states necessitate A-states in a metaphysically primitive sense. Often run together with strong supervenience but technically distinct.
• Humean supervenience. Lewis's claim that all truths supervene on a vast mosaic of local matters of particular fact. A grand unifying use of the relation; widely contested.
• Moral supervenience. Two acts alike in every natural respect cannot differ morally. Almost universally accepted across metaethics, even by anti-realists, which is why it is sometimes called the "non-negotiable datum" of moral discourse.
• Mereological supervenience. Whole-properties supervene on the properties of parts plus their arrangement. The natural way to formulate physicalism without atomism.

Common confusions

• Supervenience does not mean "caused by." Mental states supervene on physical states even where no causal relation runs from physical to mental — supervenience is a synchronic structural relation, not a causal one.
• Supervenience does not mean "reducible to." Reduction needs translatability; supervenience tolerates multiple realisation. Conflating the two collapses Davidson's whole programme.
• Supervenience is not symmetric. If A supervenes on B, B does not generally supervene on A. Beauty depends on canvas; canvas does not depend on beauty.
• Supervenience is not mere correlation. Correlations can be coincidental; supervenience claims a modal connection that holds across worlds (in the strong reading).
• "Higher-level" properties are not always epiphenomenal. Whether they do causal work is a substantive question, not settled by supervenience alone.
• Supervenience does not entail physicalism. A theist can accept that the mental supervenes on the physical-plus-soul. The base must be specified, and the choice of base is theoretically loaded.