In the conversation that closes Chapter 2 of The Ego Tunnel — Metzinger interviewing the neurophysiologist Wolf Singer, which I wrote about in the previous piece — Singer is asked why a brain scientist cares about philosophy, and his answer is expansive and optimistic, and it contains one clause that stopped me cold:
as we learn to understand how our brains assign values and distinguish between appropriate and inappropriate conditions, we will learn more about the evolution and constitution of morality.
Read quickly it sounds unobjectionable — of course understanding moral psychology will tell us about morality. Read slowly it’s a sleight of hand, and naming exactly where the card disappears turns out to clarify a lot about what neuroscience can and can’t deliver.
Note
On how these get written: I read the source material and think it through, a good deal of it in a long back-and-forth with an LLM, and then I lean on the LLM as an editor to turn that thinking into something readable, because I’ve never written much and I haven’t got the patience to grind my own sentences into shape. The ideas, the connections, and the judgments are mine, the wrong ones included. The prose has had help. I’d rather say that plainly than pretend otherwise.
The card that disappears
The smuggle lives in two words: “appropriate” and “constitution.” A brain that “distinguishes between appropriate and inappropriate conditions” is doing valuation — computing good-for-me and bad-for-me, reward and threat, approach and avoid, the machinery of dopamine and the orbitofrontal cortex. That’s appropriateness relative to a goal, or to survival. Morality is about appropriateness in a completely different sense — what is right, what one ought to do — and those are not the same ruler. A brain can be superb at computing “appropriate for my survival” — take the rival’s food, abandon the weak member of the group — and have the answer be morally irrelevant or outright monstrous. Sliding from the first sense to the second is the oldest move in the book: David Hume pointed out, nearly three centuries ago, that writers on morality reason along in ordinary statements of what is and then, without warning or argument, switch to statements of what ought to be — as if the ought simply fell out of the is. It doesn’t, and the gap between them has never been bridged. The word “constitution” is where Singer plants the flag: it claims morality is constituted by the valuation machinery, which is precisely the step Hume’s gap forbids.
What’s fair game, and what isn’t
It’s worth being precise, because Singer bundles three claims and only one is contraband.
The evolution of moral behaviour is solid science: how our moral dispositions arose — kin selection, reciprocal altruism, the moral emotions — is a real and tractable question, and primatologists like Frans de Waal have done genuinely illuminating work on its animal roots. The mechanism of moral cognition is also solid: how we actually generate moral judgments, the dual-process tug-of-war that Joshua Greene’s brain-imaging of moral dilemmas reveals, is straightforward cognitive neuroscience. Both of those are descriptions of how the moral faculty works and where it came from.
The constitution of morality — what is actually right — is the one that can’t be read off any of that. Knowing precisely how a brain generates a moral intuition tells you nothing about whether the intuition is correct. You can have a complete causal story of why someone feels that some act is wrong, and that story is fully compatible with their being mistaken. Singer is on firm ground for two of his three claims and quietly stands on them to reach the third.
Why the diversity of human morality is the proof
Here’s the cleanest way to see that the third claim fails, and it’s the thing that first made me uneasy with Singer’s sentence: human moral codes differ enormously. Cultures and religions disagree about killing, sex, food, hierarchy, honour, purity, the boundaries of the group, the weight of the individual against the collective. And they disagree while running on essentially identical neural hardware — the same valuation circuitry in every skull. If morality were constituted by the brain’s value machinery, you’d expect convergence. You get the opposite. So the diversity tells you, flatly, that the content of morality is software, not hardware. The brain doesn’t fix it; culture writes it.
But — and this is the part that keeps the question interesting rather than collapsing into “it’s all relative” — underneath the diverse content there really does seem to be a universal substrate. Not universal rules, but a universal palette. The psychologist Jonathan Haidt’s work points to a small set of moral foundations — roughly care, fairness, loyalty, authority, sanctity, liberty — that show up across cultures, which then weight and elaborate them very differently. There’s a research program (modelled explicitly on Chomsky’s universal grammar) arguing for a universal moral grammar: a shared deep structure — intention matters, acting differs from merely allowing — with culture setting the parameters. And a handful of near-universals recur almost everywhere: incest avoidance, prohibitions on in-group murder, some norm of reciprocity.
The right picture, then, is the language analogy, and it’s exact. The brain gives you a universal capacity to acquire a moral code, exactly as it gives you a universal capacity to acquire a language — and then culture installs a specific one, the way it installs Tamil or Finnish. Thousands of mutually unintelligible languages run on identical language hardware, and nobody concludes from that either that the neuroscience of language is impossible or that one language is the “true” one. Same here. Neuroscience can illuminate the moral faculty; it can never deliver the moral facts. Asking which moral code the brain “really” supports is like asking which language is the correct one — a category error.
And notice what the diversity does to Singer’s reach. Lay the world’s moral codes side by side. Biology and neuroscience describe all of them equally well — the honour culture and the dignity culture, the warrior code and the pacifist one, are all the same machinery doing what it does. So a finished science of the brain hands you every “is” and cannot tell you which “ought” should win. The diversity isn’t just evidence that morality is partly cultural; it’s the is/ought gap made visible — a row of fully-describable facts, none of which is thereby a value.
Socket and software
The resolution, I think, is one that comes straight out of Metzinger’s own framework. Elsewhere in the book he develops the idea that the big shared fictions that hold a society together — justice, money, law — are virtual organs: real in their effects, physically instantiated in the brains that carry them, but with no correlate out in the world, installed across millions of minds by story and repetition (I wrote about this as “what a lie really is”). Moral diversity is simply the diversity of which virtual organs got installed. The brain supplies the universal socket — the foundations, the grammar, the capacity to host a norm and to feel it as genuinely binding rather than optional. Culture supplies the software — a specific code. Different cultures install different code into the same socket.
That single distinction is the whole error in Singer’s sentence: he mistook the socket for the software. Understanding the socket — the universal valuation and norm-hosting machinery — tells you that humans can have a morality and will feel it as real. It tells you nothing about which morality got installed in any particular case, and still less about which one ought to be installed.
The recurring itch
What I find most telling is that this is the same move Singer makes elsewhere in the very same interview. Earlier he folds the question of how subjective feeling arises into the engineering problem of reading the neural code — treating a question about felt experience as a hard version of a data-analysis problem. Here he folds the question of what’s right into the science of how brains assign value — treating a normative question as a hard version of a descriptive one. It’s a temperamental tic, and I suspect it’s common to the whole confident-neuroscientist genre: the residue that doesn’t fit the third-person program — the felt, the ought — keeps getting quietly absorbed into the program rather than left standing outside it. The ought is to ethics what the felt quality is to consciousness: the part that does not come along with the completed mechanism, however complete it gets.
Two honesties before I close. First, the is/ought gap isn’t a settled theorem; ethical naturalists argue moral facts just are certain natural facts (about wellbeing, say), and there’s a subtler, defensible role for the science even if you keep the gap — learning that a moral intuition is a contingual evolutionary artifact can legitimately undercut its authority, which is normative bite without deriving an ought from an is. So Singer isn’t committing a howler; he’s asserting as won a battle that’s still on. Second, the diversity I leaned on so hard doesn’t settle the meta-question either: it’s compatible with relativism (no code more correct), with realism-plus-error (some codes simply wrong — most of us do think this about, say, slavery, and we’re not obviously confused to), and with a thin universal core surrounded by a wide band of permissible variation. The descriptive fact of diversity can’t pick between those — which is, one more time, the gap declining to do philosophy’s work for it.
Metzinger himself is going to have to walk this line: the final part of his book is titled “A New Kind of Ethics,” and it tries to draw normative conclusions out of the claim that the self is a construct. When I get there, the question I’ll be holding is whether he earns the crossing from is to ought more carefully than this one clause of Singer’s waves at it — or whether the whole project of deriving an ethics from a science of mind has the same itch, all the way down.