Normativity, Function, and the Limits of Physicalist Supervenience
Can physical facts alone tell you what a heart is for? This paper argues no — and that this failure runs deep enough to trouble physicalism itself.
Plain summary: The paper challenges the common assumption that biological functions ("the heart pumps blood") are fully explained by physical and evolutionary facts. It argues that singling out one effect as the function — rather than a mere byproduct — requires a normative judgment that no amount of physical or historical data can settle.
Background: What Is Physicalist Supervenience?
Physicalism holds that everything — including facts about biology, mind, and value — is ultimately fixed by physical facts. The technical expression of this is supervenience:
(S) Necessarily, if two possible worlds are physically identical — same particles, same laws, same evolutionary history — then they are identical with respect to biological functions too.
This is the claim the paper targets. If (S) fails, then either biology contains irreducible normative facts, or functional talk in biology must be abandoned altogether.
What Is "Proper Function"?
When biologists say the heart has the function of pumping blood, they mean something stronger than a causal description. They mean:
- The heart ought to pump blood
- A heart that fails to pump is defective, not merely unusual
- There is a standard of success against which the heart can be evaluated
This evaluative dimension — the ability to malfunction — is what makes biological function normative. A thermostat that reads wrong is broken. A rock that doesn't roll uphill is not.
Key concept — malfunction: Malfunction is the crucial test. Mere causal descriptions can't accommodate it. If the heart is just "the thing that pumps," then a non-pumping heart is simply not-doing-that. But biology says it's failing. That failure requires a prior standard — a norm — relative to which failure is defined.
The Main Argument: Why Physical Facts Can't Fix Functions
The core move is an underdetermination argument. For any biological trait, complete physical and evolutionary history records many effects simultaneously:
- The heart pumps blood
- The heart makes sounds (heart sounds are clinically useful)
- The heart generates heat
- The heart contributes to chest mass
All of these are causally real effects. All are equally part of the physical history. The question is: what makes "pumping blood" the function, while "making sounds" is a mere byproduct?
The paper's answer is that no physical or causal fact can answer this. To privilege one effect over others as the success condition is already to make a normative judgment — one that sits outside the purely physical description.
Modal version of the argument:
(1) If a system has a function, it is subject to correctness conditions.
(2) Correctness conditions are not fixed by physical facts alone.
(3) Therefore, functional facts are not fixed by physical facts alone.
The Best Objection: Selected-Effects Theory
The most powerful physicalist reply comes from selected-effects theories (Millikan, Neander, Godfrey-Smith). The idea:
The function of a trait is the effect for which it was selected by natural selection. Since evolutionary history is a physical fact, this grounds normativity in nature.
On this view, the heart has the function of pumping blood because that specific effect is why hearts were selected — hearts with better pumping left more descendants. The normative standard just is whatever selection tracked.
Why the paper thinks this fails
The paper offers three replies:
- Multiple covarying effects: Selection typically tracks many effects simultaneously. The heart was selected for pumping blood, yes — but in environments where pumping blood also generated those characteristic sounds, produced that heat, etc. Selection history alone doesn't mark one effect as the "real" reason rather than another.
- Causal explanation is interest-relative: Saying the heart was selected "because of" pumping blood is already an explanatory framing. It treats one causal factor as explanatorily privileged. But that framing reflects our interests and background assumptions, not a mind-independent fact about causation.
- Counterfactuals don't help: Sophisticated versions appeal to what would have been selected across nearby possible environments. But this just pushes the problem back — "which outcomes count as success across those environments" still requires a normative standard to evaluate.
The presupposition charge: The paper's sharpest move is to argue that selected-effects theories presuppose the normativity they're trying to explain. To say a trait was "selected for F" you must already treat F as the relevant success condition — the thing that mattered for selection. That evaluative judgment isn't delivered by the physical history itself.
Implications: Teleosemantics
These problems cascade into teleosemantics — the view (developed by Millikan, Dretske, Papineau) that mental representations get their content from their biological function. If the function of a mental state is whatever it was selected to track, then content is naturalised.
But if biological function itself isn't fully naturalised — if it already presupposes irreducible normativity — then teleosemantics inherits rather than solves the problem. You can't bootstrap a theory of mental content from biological norms if biological norms themselves float free of physical facts.
The Fork: Eliminativism or Irreducible Normativity
If the argument succeeds, physicalists face an uncomfortable choice:
| Option | What it means | Cost |
|---|---|---|
| Eliminativism | Functional discourse ("the heart's function is to pump blood") is false or empty — a useful fiction, not a real feature of the world | Drastic — erases most of biology as ordinarily understood |
| Irreducible normativity | Functional facts are genuinely normative and don't reduce to physical facts — physicalism is incomplete | Requires revising a foundational commitment of mainstream metaphysics |
The paper doesn't take a side — it argues only that the choice is forced.
Critical Assessment: Strengths and Problems
The argument is carefully scoped and technically competent. But several objections deserve attention:
Strengths
- The underdetermination point is genuine. Causation really doesn't privilege one effect over others in the way normative standards require. This is a real problem for purely extensional accounts.
- The presupposition charge is sharp. The claim that selected-effects theories smuggle in the normativity they're supposed to explain is a substantive dialectical move, not just scepticism.
- The modal framing is rigorous. Framing the problem as supervenience failure rather than mere epistemic underdetermination sharpens the metaphysical stakes considerably.
Potential Weaknesses
Problem 1 — Begging the question on premise (2). The argument asserts that correctness conditions are not fixed by physical facts alone — but this is close to what needs to be proved. The selected-effects theorist would say correctness conditions just are selection history, and the paper needs to engage more directly with why that definition is inadmissible, not just assert it's "interest-relative."
Problem 2 — The interest-relativity of causal explanation. The paper uses the fact that causal explanation is interest-relative to undermine selected-effects accounts. But this is a general epistemological point about explanation, not a specific metaphysical fact about function. Many things we care about (temperature, species membership) are similarly "interest-relative" in their framing yet still reduce to physical facts. The paper needs to argue that functional normativity is interest-relative in a different way — one that blocks reduction.
Problem 3 — Missing sections 2 and 3. The paper references sections on "the normative character of proper function" and "relevant supervenience commitments" but the available text omits them. The argument at Section 4 assumes that proper function is irreducibly normative — but this is precisely what a physicalist would contest. Without the earlier sections, a key premise goes undefended in the text we have.
Problem 4 — Scope of the underdetermination claim. The claim is that physical facts underdetermine which effect is "normatively privileged." But the selected-effects theorist can bite the bullet: they can say there simply is no fact about which effect is "really" the function beyond what selection tracked — and that's fine. On this eliminativist-friendly reading, the choice the paper presents is not a reductio but a feature, not a bug.
The strongest response available to physicalists: Accept that "function" talk is interest-relative and pragmatic — a tool for organizing biological inquiry — while insisting this is fully compatible with physicalism. Normativity is in the map, not the territory. The paper would need a stronger argument that this deflationary move is unacceptable, not just metaphysically unsatisfying.
Broader Context
The paper situates itself against Millikan (Language, Thought, and Other Biological Categories, 1984), Neander's selected-effects refinements, and Godfrey-Smith's environmental complexity theory. A fuller treatment would also engage Boorse's biostatistical theory (function as species-typical contribution to survival and reproduction) and Cummins' causal-role theory — both of which offer different angles on the normativity problem. The paper's argument may apply to Cummins-style accounts differently, since those explicitly don't invoke selection history at all.