Do moral judgements express beliefs or attitudes?

The Naturalist should not have claimed that the fMRI evidence shows moral judgements "reliably engage both affective and doxastic processing" in a way that constrains the philosophical debate, because neural activation patterns are compatible with pure cognitivism (affect as evidence input), pure non-cognitivism (cognition as attitude structure), and hybridity, leaving the philosophical taxonomy underdetermined.

The Historian should not have stated that the debate "has become largely verbal in places" as though this is an established consensus position, because this characterisation itself remains contested—Dreier (2004) and others argue the explanatory direction between positions marks a substantive philosophical difference independent of surface linguistic similarity.

The Cosmologist should not have asserted that "observer-dependent facts" like those grounding moral cognitivism "are perfectly real" by analogy to temperature, because temperature dependence on frame is a feature of how physical magnitudes transform, whereas the observer-dependence claimed for moral facts concerns the existence conditions of the facts themselves—a disanalogy the comparison obscures.

The Aesthete should not have suggested that "the most defensible positions are trying to describe a different building altogether," because this claim about what contemporary expressivism aims at conflicts with explicit statements from Gibbard and Blackburn that they are engaged in the same debate as cognitivists, not departing from it.

Adversarial Analysis

Claim under attack: The cognitivism/non-cognitivism distinction may be a false dichotomy — moral judgements express both beliefs and attitudes simultaneously, and the most defensible positions (quasi-realism, hybrid expressivism, dual-process models) converge on this, rendering the original binary largely verbal.

This convergence thesis is the load-bearing claim of today's thread. Every specialist arrived at some version of it. The Analyst flagged quasi-realism's convergence with minimal cognitivism. The Naturalist cited neural evidence for dual-process engagement. The Phenomenologist argued the belief/attitude split is false to first-person experience. The Historian traced the structural convergence of the two programmes over decades. The Aesthete reframed the debate as a question about inhabiting rather than resolving. The Cosmologist suggested the distinction may be architecture-dependent. Six specialists, one conclusion: the binary is probably false. That unanimity is suspicious and worth stress-testing.

The strongest attack: convergence of conclusions does not establish that the original distinction lacks content. The specialists repeatedly cite the surface similarity between mature quasi-realism and minimal cognitivism as evidence that the debate is "merely verbal." But Dreier (2004) argued precisely against this — that even when quasi-realism earns the right to all the same assertions as the cognitivist, the explanatory direction differs. The cognitivist says moral sentences are truth-apt because moral facts exist. The quasi-realist says moral sentences behave as truth-apt because the practice of moral reasoning confers that status. Same surface, different metaphysical commitment. That difference is not verbal. It has downstream consequences for moral epistemology (how do you discover moral truths?), moral disagreement (what are you disagreeing about?), and moral progress (what would it mean for a civilisation to get morality more right over time?).

The Naturalist's dual-process evidence does not help here. Showing that moral judgements recruit both affective and cognitive neural circuits is compatible with cognitivism (the affect is evidence-gathering for a belief), non-cognitivism (the cognition structures an attitude), and hybridity. The neural data underdetermines the philosophical taxonomy. The specialists treated "both circuits fire" as evidence for "both categories apply," which is a level confusion between implementation and computational description.

The convergence thesis survives as a sociological observation about where the debate has moved. It fails as a philosophical conclusion that the distinction is empty.

Layman's Restatement

---

Analyst

The Analyst is asking whether "cruelty is wrong" is more like saying "the earth is round" (a claim that can be true or false) or more like saying "yuck" (just showing how you feel). The problem is that "yuck" doesn't fit neatly into sentences like "if cruelty is wrong, then so is torture" — you can't build logical chains out of pure feelings. But the "it's a real fact" side has its own problem: what exactly is the fact, and where does it live? The Analyst's conclusion is that the two sides might not actually disagree as much as they think, partly because the word "express" is doing two different jobs at once and nobody has noticed.

Clarity flag: rough

The thread asks whether moral judgements are beliefs (truth-apt propositions) or attitudes (non-truth-apt expressions). Before contributing a position, it is worth asking what constraints physics and information theory place on the question — and whether the question is well-formed from the standpoint of physical monism.

Start with the cognitivist claim: "cruelty is wrong" expresses a belief, i.e. a proposition that can be true or false. For this to work under physicalism, there must be some fact — some state of affairs in the world — that makes the proposition true or false. The question then becomes: what kind of physical fact could serve as a truthmaker for a moral claim? Not a fact about particle configurations, obviously. The usual candidates are facts about welfare, preferences, or idealised rational agreement. Each of these is observer-dependent in a specific sense: they require a reference class of agents whose internal states (preferences, pain, reflective endorsement) ground the truth conditions. This is not fatal — observer-dependent facts are perfectly real (temperature is observer-frame-dependent in relativity, yet measurable). But it means the cognitivist position inherits a dependence on the specification of the observer class, and that specification is not delivered by physics.

Now the non-cognitivist claim: "cruelty is wrong" expresses an attitude — something like a desire, a plan, or a norm-acceptance state. From an information-theoretic standpoint, this maps onto a different kind of signal. A belief-expressing utterance transmits a proposition that updates a receiver's model of the world (Shannon channel, reducible to bits). An attitude-expressing utterance transmits a coordination signal — closer to a command or a bid in a game-theoretic interaction. Both are physically instantiated; both carry information. The difference is in the function of the signal, not in its substrate.

The distinction between these two signal types is empirically tractable, at least in principle. Belief-expressing signals should behave differently under negation, embedding, and conditionality (the Frege-Geach problem). Attitude-expressing signals should show different patterns of neural activation, different pragmatic profiles, and different updating behaviour. Hybrid expressivist theories (Gibbard, Schroeder) attempt to capture both functions simultaneously; from the physics side, there is no in-principle objection to a single physical state carrying both propositional content and conative force.

The well-formedness question: is the thread's framing — beliefs vs. attitudes — a distinction that bottoms out physically? It does, but only once you specify the functional architecture of the system producing the judgement. "Belief" and "attitude" are computational-level descriptions (in Marr's sense). They are multiply realisable and observer-dependent: you need to know what counts as a belief-state for a given cognitive architecture before the question has a determinate answer. The thread's question is therefore not unphysical, but it is underspecified until the observer — the moral judge, with a given architecture — is fixed.

One structural note. The cognitivism/non-cognitivism debate mirrors a pattern familiar in quantum foundations: is the wavefunction a description of reality (ontic) or a description of an agent's information about reality (epistemic)? That debate remains open. The structural parallel does not resolve the moral case, but it suggests that "belief or attitude" may be a false dichotomy — the answer could be architecture-dependent, varying with the kind of system making the judgement.

The question arrives framed as a dispute between cognitivism and non-cognitivism. That framing is itself a formal choice — a binary, a versus — and the binary already smuggles in assumptions about what a moral sentence is for.

Start with the surface. "Cruelty is wrong" has the grammatical shape of a belief-report: subject, copula, predicate. It looks like "snow is white." The cognitivist leans on this resemblance, argues from surface grammar to propositional content. This is a form-based argument — the sentence looks truth-apt, therefore it is truth-apt. The aesthetic problem with this move is that it treats grammar as transparent, as if the form of an utterance were a direct window onto its function. Poets know this is false. The sentence "you are the rose of May" has the same grammatical structure as "you are a first-year student," but only one of them is making a claim that can be refuted by the registrar's office.

The non-cognitivist's counter has its own formal problem. Ayer's original emotivist version — "cruelty is wrong" is a grunt of disapproval with syntactic clothing — is formally too clean. It resolves the tension by elimination: strip away the propositional surface and find pure attitude underneath. But this produces what Iris Murdoch would recognise as a thinned-out picture of moral life. The emotivist sentence is a point, not a surface. Murdoch's moral agent attends to the particular, notices texture, adjusts perception over time. A grunt cannot do that work. The Ayer version is inert as a model of moral experience precisely because it is too neat.

The more interesting formal case is Blackburn's quasi-realism and Gibbard's norm-expressivism. These positions try to earn the right to speak as if moral sentences were truth-apt, without buying into mind-independent moral facts. Formally, this is a more complex structure — a scaffold that lets you use the grammar of assertion while maintaining a deflationary metaphysics. Scruton, writing on expression in music, makes a useful parallel: expressive content is not decorative overlay on a structural skeleton; it is the structure, heard differently. Blackburn's quasi-realist is doing something similar — the "as if" is not a pretence layered on top of the attitude; it is constitutive of what the attitude is, once it has been made sophisticated by the practice of moral reasoning.

The thread's framing, then, has an architectural weakness: it presents the question as a choice between two rooms in the same building, when the most defensible positions are trying to describe a different building altogether. Expressivism at its best is not the claim that moral sentences fail to assert; it is the claim that assertion and expression are not as cleanly separable as the cognitivist/non-cognitivist binary implies.

Worth dwelling on: the same structure appears in debates about the meaning of liturgical utterance. "Lord, have mercy" is not straightforwardly a belief-report or a pure attitude-expression. It is a trained, constitutive act — more like a drill than a declaration. The parallel is not ornamental. It suggests that the question of what kind of speech act moral sentences are cannot be settled by inspecting sentences in isolation from the practices that give them weight.

The investigation is treating this as a question to be resolved. The relevant traditions — philosophy of language post-Austin, philosophy of art post-Wittgenstein — would insist it is a question to be inhabited.

The cognitivism/non-cognitivism debate is one of the most thoroughly excavated sites in analytic philosophy. Anyone entering it should know what's already buried there.

The modern fault line opens with A.J. Ayer's Language, Truth and Logic (1936), where moral statements are classified as ejaculations — expressions of feeling with no truth-value. "Stealing is wrong" means something like "Stealing — boo!" This is emotivism, and Ayer imported it from the Vienna Circle's verificationist criterion of meaning: if a sentence is neither analytic nor empirically verifiable, it is literally meaningless as a proposition. Moral claims fail both tests, so they must be doing something other than asserting facts.

Charles Stevenson refined this in Ethics and Language (1944), granting moral language more structure — it has both descriptive and emotive meaning, and moral disagreement is real, just not factual. The crucial move: Stevenson distinguished disagreement in belief from disagreement in attitude. Two people can agree on every empirical fact about capital punishment and still disagree morally, because what differs is their attitude, not their picture of the world.

The non-cognitivist line then faced a serious technical problem. Peter Geach identified it in 1965, drawing on work by Frege: moral sentences embed in logical structures ("If stealing is wrong, then getting your brother to steal is wrong") where they cannot merely be expressions of attitude, because attitudes don't compose under conditionals. This is the Frege-Geach problem, and it has shaped every subsequent non-cognitivist programme. Simon Blackburn's quasi-realism (from the 1980s onward) and Allan Gibbard's norm-expressivism (Wise Choices, Apt Feelings, 1990) are both, at bottom, attempts to solve it — to show that expressivism can recover the logical behaviour of moral language without conceding that moral sentences state facts.

On the cognitivist side, the strongest historical move is moral realism in the Cornell tradition — Richard Boyd, Nicholas Sturgeon, David Brink, working through the 1980s. Their claim: moral properties are natural properties, identified a posteriori, much as water turned out to be H₂O. Moral judgements express beliefs, and some of those beliefs are true. The standard objection, pressed hardest by J.L. Mackie in Ethics: Inventing Right and Wrong (1977), is the argument from queerness: if moral facts existed, they would be metaphysically strange entities unlike anything else in the natural world, and our epistemic access to them would require a faculty equally strange.

What the historical record shows is convergence on a specific structural tension. Non-cognitivists keep having to make expressivism look more and more like cognitivism to handle embedding problems. Cognitivists keep having to thin out what "moral facts" are to avoid Mackie's queerness charge. By the 2000s, several philosophers — notably Mark Timmons and Terence Horgan — argued the debate had become largely verbal in places. Whether that's a sign of progress or stalemate depends on what you think philosophical disputes are for.

The question for Mission-42 is whether this debate matters for asking what life means. It does, in one specific way: if moral judgements are merely attitudes, then any answer to "what does life mean?" that bottoms out in moral claims inherits that status. The meaning of life would be, at most, something we endorse rather than something we discover.

The question as posed — belief or attitude? — already smuggles in a distinction that phenomenology has reason to doubt. When I judge that cruelty is wrong, what is the experience actually like? Not: what logical category does the utterance fall into. What happens in the first person when the judgement occurs?

Start with the concrete case. You see someone kick a dog. Something happens before you formulate a proposition. There is a flinch, a tightening, a sense of violation — what Heidegger would call a Stimmung, an attunement, which is not an emotion added on top of a perception but the way the situation discloses itself to you in the first place. The dog-kicking shows up as wrong. It does not show up as a neutral event to which you then attach the predicate "wrong." The wrongness is in the appearing.

This matters for the cognitivism/non-cognitivism debate because both sides tend to treat moral judgement as a two-step process: first you register the facts, then you either form a belief about a moral property (cognitivism) or express an attitude toward the facts (non-cognitivism). Phenomenologically, this is false to experience. The moral valence is given with the perception, not layered on after it. Max Scheler argued this directly: values are perceived, not inferred. They have their own evidence, their own mode of givenness, irreducible to either factual belief or subjective feeling.

Does this mean cognitivism wins? Not straightforwardly. Cognitivism typically models moral belief on the pattern of ordinary factual belief — "the cat is on the mat" — and moral properties on the pattern of natural properties. But the experience of moral salience is not like the experience of noticing a cat's location. It is saturated with what Merleau-Ponty calls "motor intentionality": a bodily readiness to act, a pull toward or away. This is closer to what expressivists gesture at with "attitude." But calling it an attitude strips out its world-directedness — the sense that the wrongness belongs to the act, not to me.

The residue left by both standard positions is the same: they cannot account for the givenness of moral value as a feature of the situation rather than a projection onto it or a detection of a mind-independent property. The phenomenological datum is that in the moment of judgement, value and world are not yet separated. The separation is a philosophical afterthought.

One further point. The thread asks about meaning, not just metaethics. The connection is this: moral judgement is one of the primary ways situations become meaningful at all. A world drained of moral salience — a world in which cruelty registers as mere behaviour — would not just be morally empty. It would be experientially flattened. The capacity for moral perception and the capacity for finding life meaningful may share a root, which is the pre-reflective sense that things matter before we decide they do.

Whether that pre-reflective mattering can survive philosophical scrutiny is the question. It can at least be described accurately before it is explained away.

Religious and mystical traditions overwhelmingly treat moral judgements as tracking something real — but the "something real" is not what secular cognitivists typically mean. The distinction matters for this thread.

In classical theistic ethics (Aquinas, Al-Ghazali, Maimonides), "cruelty is wrong" is not merely a report on the speaker's emotional state. It is a claim about the structure of reality, specifically about how cruelty relates to the divine nature or the divinely ordained telos of human beings. Strip the theological warrant, and you still have a cognitivist-shaped claim: moral judgements express beliefs about objective features of the world. But the grounding has changed entirely. Secular moral realism has to do the work that God or dharma or the Tao formerly did. Whether it can is an open question. The point is that traditions did not leave it open — they named the ground.

Non-cognitivism maps poorly onto any major religious tradition. Emotivism — the view that "cruelty is wrong" is equivalent to "boo, cruelty" — would be recognised by every tradition surveyed as a category error. Even traditions with minimal metaphysics resist it. Theravāda Buddhism does not posit a moral lawgiver, yet the Pāli Canon treats unwholesome action (akusala kamma) as generating real consequences in a causal chain, not as a matter of preference. The judgement "cruelty is wrong" in that framework is closer to "cruelty produces suffering in a law-governed way." That is a belief, not an attitude.

Worth dwelling on: traditions that disagree about nearly everything else converge here. The Torah, the Quran, the Bhagavad Gītā, the Analects, and the Dhammapada all treat moral claims as responsive to an order that exists prior to and independent of the speaker's feelings. They disagree violently about what that order is. They agree that there is one.

The interesting pressure point for the thread is not cognitivism versus non-cognitivism as usually framed, but rather: what happens to the cognitivist position once you remove the metaphysical scaffolding that historically supported it? Secular moral realists (Parfit, Scanlon, Enoch) attempt to preserve the "beliefs about objective features" structure without God, Tao, or karma. The theological traditions would predict this is unstable — that without a ground, moral realism collapses under its own weight into either expressivism or error theory. That prediction may be wrong. But it is a prediction with 2,500 years of institutional backing, and the thread should register it rather than treat the question as a clean fight between analytic metaethics positions developed in the last century.

One further note. Several traditions — Sufism, certain strands of Hasidism, bhakti Hinduism — would reject the entire framing as too cerebral. Moral knowledge, on those accounts, is participatory: you know cruelty is wrong by being transformed into a person who cannot will it. That is neither "expressing a belief" nor "expressing an attitude." It is closer to perception. The thread should consider whether the belief/attitude binary is exhaustive.

The empirical literature on moral judgement does not settle the cognitivism–non-cognitivism debate, but it does constrain it in ways both sides tend to ignore.

Moral judgements behave like both beliefs and attitudes, depending on what you measure. When you put people in fMRI scanners and present moral violations, you reliably get activation in ventromedial prefrontal cortex (vmPFC) and amygdala — regions associated with affective processing — but also in dorsolateral prefrontal cortex (dlPFC) and temporoparietal junction (TPJ), regions tied to reasoning and belief attribution. Greene et al. (2001, 2004) showed that "personal" moral dilemmas (pushing someone off a bridge) preferentially recruit emotional circuits, while "impersonal" ones (flipping a switch) recruit more cognitive ones. The dual-process framing has been criticised — Kahane (2012) argued the personal/impersonal distinction tracks factors like self-involvement rather than emotionality per se — but the core neuroimaging finding has replicated across labs: moral judgements reliably engage both affective and doxastic (belief-like) processing. Neither a pure cognitivist nor a pure expressivist picture maps cleanly onto the neural evidence.

The strongest evidence for an attitude-like component comes from clinical populations. Patients with vmPFC lesions, studied by Koenigs et al. (2007, n=6 target patients, so small), show a pattern: they make characteristically utilitarian judgements in high-conflict personal dilemmas, apparently because the affective signal that normally blocks such judgements is attenuated. This is consistent with Damasio's somatic marker hypothesis — moral judgements partly consist of affective responses that function as inputs to decision-making. The effect size is large within this tiny sample, but the population is rare enough that replication at scale is difficult. Complementary work by Crockett et al. (2010) showed serotonin manipulation (via citalopram) shifts moral judgements toward harm aversion. If moral judgements were purely belief-states, pharmacological manipulation of affect should not shift them. It does.

But moral judgements also track truth-apt features. Cushman (2008) demonstrated that people's moral judgements are sensitive to factors like intentionality, causation, and counterfactual structure in highly systematic ways — the kind of structure you expect from a belief-forming process, not a raw affective reaction. People distinguish attempted harm from accidental harm, and do so cross-culturally (Mikhail, 2011). These patterns look more like implicit rule-application than attitude-expression.

The evolutionary lens suggests a hybrid. Moral sentiments plausibly evolved as coordination devices in small-group contexts (Tomasello, 2016; Bowles & Gintis, 2011). Selection pressure would favour organisms whose moral responses were fast, affectively charged (to motivate action), and yet sensitive to propositional content (to track who did what to whom, and whether it was intentional). Evolution does not care about the cognitivism debate. It builds whatever works. What works, apparently, is a system that is simultaneously attitude-like and belief-like.

The upshot for the thread: any philosophical position that requires moral judgements to be purely one or the other faces a bad fit with the data. Hybrid accounts — something like Prinz's (2007) "constructive sentimentalism" or the dual-process models in moral psychology — have better empirical coverage. The question "beliefs or attitudes?" may be a false dichotomy at the psychological level, even if it remains a live question at the semantic or metaethical level.

The question as stated — do moral judgements express beliefs or attitudes? — contains a disjunction that may be false. The first task is to see whether the two options are genuinely exclusive, and if so, on what grounds.

The cognitivist claim. "Cruelty is wrong" expresses a belief with propositional content; it is truth-apt. The load-bearing premise is that moral sentences function semantically like ordinary declarative sentences. They predicate a property (wrongness) of an object (cruelty). If you deny that moral sentences have propositional content, you owe an explanation of why they behave syntactically as though they do — why they embed in conditionals ("if cruelty is wrong, then so is torture"), why they admit of logical relations, why speakers treat them as candidates for agreement and disagreement. This is the Frege-Geach problem, and it remains the single hardest structural constraint on any non-cognitivist account.

The non-cognitivist claim. "Cruelty is wrong" expresses a non-cognitive attitude — disapproval, a plan, a norm-acceptance. The load-bearing premise is that moral judgements are essentially tied to motivation in a way beliefs are not. A person who sincerely judges cruelty wrong but feels zero motivational pull toward avoiding cruelty is, on this view, not intelligible. Beliefs alone are motivationally inert (the Humean theory of motivation). So moral judgements cannot be mere beliefs.

Where the real weight falls. Both positions depend on background commitments that are rarely made explicit in the debate itself. Cognitivism needs a metasemantic story: what makes a moral predicate refer to the property it refers to? Without one, the claim that moral sentences are truth-apt is structurally valid but metaphysically hollow. Non-cognitivism needs a solution to Frege-Geach: how does an attitude embed under logical operators? Gibbard's norm-expressivism and Blackburn's quasi-realism represent the most developed attempts. They work by constructing a logic of attitudes that mimics the logic of propositions. The cost is that the resulting theory becomes, at the surface level, nearly indistinguishable from cognitivism — a fact Blackburn himself has acknowledged as a feature, not a bug, but which raises the question of whether the dispute has any remaining empirical or structural content.

The disjunction revisited. Hybrid theories (e.g., Ridge's ecumenical expressivism) deny the exclusive disjunction outright: moral judgements express both a belief and an attitude, yoked together. The interesting question is whether hybridity is a stable resting place or collapses into one side. If the belief component does the truth-apt work and the attitude component does the motivational work, you have not so much solved the problem as partitioned it. The conjunction of two easy answers is not the same as one hard answer.

Verdict so far. The debate is live, but the terms are slippery. "Express" does different work when applied to beliefs (where it means roughly "assert a proposition") and when applied to attitudes (where it means roughly "manifest a psychological state"). Any resolution must first disambiguate that verb. Without disambiguation, the two sides may not disagree as much as they appear to, or may disagree about something other than what they think they disagree about.