Metaethics: Moral Realism, Emotivism, Expressivism, and the Naturalistic Fallacy

When someone says that “torture is wrong,” we usually go on to ask whether the claim is true, whether it fits certain moral theories, or how it applies to a concrete case. Metaethics asks something else: what are we doing when we make such a claim? Are we describing a fact, expressing an attitude, prescribing conduct? Are there moral facts independent of human minds? If so, how do we know them? If not, in what sense can morality be objective or rational?

This article presents the field: how it is delimited, the founding arguments of Hume and Moore, the main families of theories (intuitionism, emotivism, prescriptivism, expressivism, error theory, realism, constructivism, relativism), and the problems that organize them. For the normative theories, see the article on deontology, consequentialism, and virtue ethics; for the general entry, ethics.

1. Three Levels of Ethical Reflection

The usual distinction is among three levels that are related but should not be conflated:

• Normative ethics. Asks what we ought to do and what kind of person we ought to be, and seeks general principles. It is the terrain of theories such as the utilitarianism of Bentham and Mill, Kantian ethics, and Aristotelian-inspired virtue ethics.
• Applied ethics. Applies principles to specific problems: bioethics, animal ethics, the ethics of artificial intelligence, and dilemmas such as the trolley problem.
• Metaethics. Investigates the metaphysical, epistemological, semantic, and psychological presuppositions of moral thought and practice. Its objects are second-order questions: the meaning of moral terms (“good,” “right,” “ought”), the ontological status of moral properties, the possibility of moral knowledge, the relation between moral judgment and motivation.

The boundary is permeable: the choice of a normative theory may depend on metaethical theses (for example, on whether there are necessary moral truths), and vice versa. Even so, twentieth-century analytic philosophy, beginning with Moore, consolidated metaethics as an autonomous subfield.

A Map of Positions

Three axes of classification are usually combined:

1. Semantic. Cognitivism: moral judgments express beliefs and can be true or false. Non-cognitivism: they express attitudes, emotions, or prescriptions and are not truth-apt in the full sense.
2. Metaphysical. Realism: there are moral facts or properties independent of subjects’ attitudes. Anti-realism: there are none (the position may be error theory, constructivism, relativism, or expressivism). See the entries realism and anti-realism.
3. Naturalist. Naturalism: moral properties are natural properties (or reducible to them). Non-naturalism: they are sui generis.

2. Hume’s Law: From “Is” to “Ought”

In the Treatise of Human Nature (1739–1740), Book III, Part I, Section 1, David Hume observes that in every system of morality he has encountered, the author moves from propositions with the copula “is” to propositions with “ought” without explaining how this new relation follows from the former. For Hume the transition deserves attention because it expresses an entirely different relation. The observation is known as Hume’s law (or “Hume’s guillotine”): one cannot derive, by deduction, normative conclusions from purely descriptive premises.

There is debate about the reach of the passage: some commentators see a logical thesis in it; others, an argument for the claim that reason alone cannot motivate action, central to his moral sentimentalism (Hume, Treatise, III.i.1; see the article on Hume). There have also been technical objections: A. N. Prior (“The Autonomy of Ethics,” 1960) offered examples in which the inference seems valid, and John Searle (“How to Derive ‘Ought’ from ‘Is’,” 1964) tried to derive an “ought” from the institution of promising. Both proposals are controversial.

Hume’s law feeds metaethics in two directions: as a motive for non-cognitivism (if “ought” does not follow from “is,” perhaps it expresses something other than description) and as a challenge to naturalism.

3. G. E. Moore: The Naturalistic Fallacy and the Open Question

G. E. Moore (1873–1958) inaugurated modern metaethics with Principia Ethica (1903). His starting point is that ethics must begin by clarifying the question “what is good?” before asking which actions are good or obligatory.

Moore holds that “good” designates a simple and therefore indefinable property: it cannot be analyzed into other properties, just as “yellow” cannot be defined for someone who has never seen it. He called the naturalistic fallacy the mistake of identifying the good with some natural property (such as pleasure or desire) or, more broadly, of defining “good” in other terms. In this spirit Moore criticized the hedonistic utilitarianism of John Stuart Mill and the ethical evolutionism of Herbert Spencer.

The central instrument is the open question argument (Principia Ethica, §13). For any proposed definition, for example “good = pleasant,” it always makes sense to ask, “but is what is pleasant good?” If the definition were correct, the question would be trivial, like “is a bachelor a bachelor?” Since it remains open, the definition fails. Moore concluded that the good is a non-natural property, apprehended by intuition.

Criticisms and Afterlife

• William Frankena (1939) observed that the label “naturalistic fallacy” is inadequate: the error, if there is one, is that of defining one property by another, and it is not restricted to natural definitions. Hence the “definist fallacy.”
• The open question argument presupposes that true identities are conceptually transparent. After the work of Saul Kripke and Hilary Putnam on necessary a posteriori theoretical identities (such as “water = H₂O”), naturalists argue that the openness of the question shows a difference of concepts, not of properties.
• The expression “naturalistic fallacy” is now used much more loosely, including for the error of inferring what is good from what is “natural.” See the entry naturalistic fallacy.

4. Intuitionism

Ethical intuitionism descends from Moore’s legacy: basic moral truths are grasped directly, not demonstrated from other truths. Two early twentieth-century figures are central:

• H. A. Prichard, in “Does Moral Philosophy Rest on a Mistake?” (Mind, 1912), argues that moral philosophy erred in seeking a proof that we ought to act morally; obligation is apprehended immediately.
• W. D. Ross (1877–1971), in The Right and the Good (1930), proposes a pluralism of prima facie duties: fidelity, reparation, gratitude, justice, beneficence, self-improvement, and non-maleficence. None is absolute; in each situation the actual duty results from weighing the prima facie duties at stake through judgment. Ross opposes both utilitarian monism and Kantian absolutism.

Intuitionists hold that the most general moral principles are self-evident to anyone who considers them with maturity, like axioms. The classic objections: the diversity of intuitions across eras and cultures, the difficulty of explaining epistemic access to non-natural facts, and the hypothesis that moral intuitions are products of evolutionary or social processes unrelated to truth. Contemporary defenders, such as Robert Audi and Michael Huemer, have reformulated the doctrine in terms of intuitions as appearances or “seemings.”

5. Emotivism

Emotivism arises at the intersection of Hume’s critique and logical positivism, and it is the first major non-cognitivist theory.

A. J. Ayer (1910–1989), in Language, Truth and Logic (1936), chapter 6, starts from the verification principle: only analytic or empirically verifiable statements are cognitively meaningful. Ethical statements are neither. Ayer concludes that they do not state facts but express feelings and aim to arouse feelings or actions in others; they are neither true nor false. In ordinary terms, saying that someone acted wrongly in stealing would amount to stating the fact of the theft plus a tone of disapproval, and that addition asserts nothing beyond the fact.

Charles L. Stevenson (1908–1979), in “The Emotive Meaning of Ethical Terms” (Mind, 1937) and Ethics and Language (1944), offers a more refined version. He distinguishes the descriptive meaning from the emotive meaning of a word, its capacity to express attitudes and to influence those of others. Ethical terms have a strong emotive meaning. Stevenson also separates disagreement in belief (about facts) from disagreement in attitude (about what to approve), and holds that many moral disputes are disagreements in attitude, resolvable, when they are, by rational persuasion rather than proof. One tentative analysis by Stevenson is that “this is good” means roughly “I approve of this; do so as well.”

Problems

Emotivism faced the charges that it reduces moral reasoning to persuasion, that it cannot capture the possibility of rational argument in ethics, and that it cannot explain the logic of moral terms in complex contexts, which became the Frege-Geach problem (section 7). The verification principle on which Ayer relied was abandoned by analytic philosophy because of its own difficulties.

6. Universal Prescriptivism (R. M. Hare)

R. M. Hare (1919–2002), in The Language of Morals (1952), Freedom and Reason (1963), and Moral Thinking (1981), formulates a version of non-cognitivism that preserves moral rationality. For Hare, moral judgments are prescriptions, speech acts that guide action, and they have two logical properties:

1. Prescriptivity: sincerely accepting “I ought to do X” implies being disposed to do X (if able).
2. Universalizability: if a moral judgment applies to one situation, it applies to any relevantly similar situation, including those in which I would occupy another person’s place. This is a feature of Kantian inspiration.

Universalizability supplies logical constraints on moral thinking and leaves room for argument. In Moral Thinking, Hare argues that universal prescriptivism, together with facts about preferences, leads to a form of (preference) utilitarianism, in a two-level model: the critical level, at which the utilitarian principle is justified, and the intuitive level, at which one follows simple moral rules in daily life. Prescriptivism directly influenced the moral philosophy of Peter Singer, who studied with Hare at Oxford. Criticisms include that Hare’s argument for utilitarianism seems to require more than logic provides, and that the prescriptivist cannot explain the “fanatic” who universally accepts monstrous prescriptions.

7. Expressivism, Quasi-Realism, and the Frege-Geach Problem

Contemporary expressivism inherits emotivism and prescriptivism in sophisticated form: moral judgments express non-representational mental states (attitudes, plans, norms), and the task is to explain how moral discourse nonetheless functions as if it were factual.

Simon Blackburn develops quasi-realism in Spreading the Word (1984), with further development in Essays in Quasi-Realism (1993) and Ruling Passions (1998). The project is to start from an expressivist base and show that one can legitimately “earn” the right to talk of moral truth, moral facts, and error, as if realism were true, without giving up anti-realism. Allan Gibbard, in Wise Choices, Apt Feelings (1990), proposes norm-expressivism (moral judgments express acceptance of norms governing feelings such as guilt and anger) and, in Thinking How to Live (2003), plan-expressivism: thinking about what to do is deciding what to do, and moral judgments express plans.

The Frege-Geach Problem

The main objection to non-cognitivism and expressivism is the Frege-Geach problem, formulated by P. T. Geach (in “Ascriptivism,” 1960, and “Assertion,” 1965), drawing on an observation of Frege’s about the sense of propositions in non-assertoric contexts, and independently raised by John Searle (1962). The reasoning:

1. “Lying is wrong.”
2. “If lying is wrong, then getting your brother to lie is wrong.”
3. Therefore, “getting your brother to lie is wrong.”

The argument is a valid modus ponens. For it to be valid, the expression “lying is wrong” must have the same meaning in premise 1 (asserted) and as the antecedent of premise 2 (unasserted, merely supposed). But if “lying is wrong” merely expresses disapproval, then inside a conditional no one is expressing it, and the meaning seems to shift. Hence the expressivist needs a compositional semantics that covers negation, conditionals, disjunction, and quantification over moral judgments.

Responses: Blackburn and Gibbard developed logics of attitudes and plans, with substantial technical results; Mark Schroeder (Being For, 2008) showed persistent difficulties for the program. The problem remains the principal test of any expressivist theory.

8. Error Theory (J. L. Mackie)

J. L. Mackie (1917–1981), in Ethics: Inventing Right and Wrong (1977), defends a second-order moral skepticism: the error theory. It has two parts:

• A semantic (cognitivist) thesis: ordinary moral discourse purports to describe objective properties endowed with intrinsic prescriptive force; that is, it presupposes the existence of objective values.
• A metaphysical (anti-realist) thesis: there are no such properties. All positive moral judgments are therefore false (or lack truth value). The error is that of the presupposition.

Mackie offers two main arguments. The argument from relativity starts from the variation of moral codes across societies and eras, which is better explained by the hypothesis that they reflect ways of life than by the hypothesis that they are perceptions of objective values. The argument from queerness has two versions: the metaphysical, on which objective values would be entities or properties of a very peculiar kind, unlike anything else in the universe, with intrinsic prescriptive force; and the epistemological, on which we would need a special faculty of moral intuition, equally strange.

Mackie does not claim that we should abandon morality: in the second part of the book he proposes a morality as an invention serving human needs. Later versions include moral fictionalism (Richard Joyce, The Myth of Morality, 2001), on which we might continue to use moral discourse as a useful fiction. Responses to error theory come from realists (who reject the queerness charge) as well as from expressivists (who reject the semantic thesis that moral discourse purports to describe).

9. Moral Realism

Moral realism holds that there are objective moral facts, in the sense that they do not depend on what any individual or group thinks. It divides into two broad families.

9.1 Naturalist Realism: “Cornell Realism”

In the mid-1980s, philosophers associated with Cornell University, Richard Boyd (“How to Be a Moral Realist,” 1988), Nicholas Sturgeon (“Moral Explanations,” 1985), and David Brink (Moral Realism and the Foundations of Ethics, 1989), defended the view that moral properties are real natural properties that figure in causal explanations, analogously to the properties of science. The strategy: (i) the identity between moral and natural properties may be true a posteriori, like “water = H₂O,” so Moore’s open question does not refute it; (ii) moral facts explain events (for instance, the injustice of a regime explains its instability), and this justifies belief in them, in the same way as the theoretical facts of the sciences. Boyd proposed that “good” refers to a homeostatic cluster of properties that satisfy human needs.

A classic challenge is Gilbert Harman’s (The Nature of Morality, 1977): moral facts seem unnecessary for explaining our observations, unlike physical facts. Cornell realists reply that the explanation does involve moral properties. Another form of naturalism is the neo-Aristotelianism of Philippa Foot (Natural Goodness, 2001), in which the good of a living being follows from its form of life.

9.2 Non-Naturalist Realism

Non-naturalist realists, in the line of Moore and Ross, hold that normative facts are sui generis, irreducible to natural facts. Derek Parfit, in On What Matters (2011, 2 vols.), defends what he calls non-metaphysical non-naturalism: there are irreducible normative truths, but they do not exist as entities in a dense ontology, so objections from metaphysical queerness lose their footing. Parfit also argues for a convergence (the “Triple Theory”) among Kantianism, contractualism, and consequentialism. Russ Shafer-Landau, in Moral Realism: A Defence (2003), proposes a robust non-naturalist realism, holding that moral truths are attitude-independent, and defends the possibility of moral knowledge with intuitions as evidence. Other defenders include Thomas Nagel (The View from Nowhere, 1986) and, in another register, T. M. Scanlon.

9.3 Challenges to Realism

• Supervenience. Moral facts vary with natural facts (two situations identical in natural respects cannot differ morally). The non-naturalist must explain why this dependence holds if there is neither identity nor reduction. See the entry supervenience.
• Mackie’s queerness argument (section 8).
• Evolutionary debunking arguments. Sharon Street (“A Darwinian Dilemma for Realist Theories of Value,” 2006) argues that our evaluative dispositions were shaped by natural selection, which does not track attitude-independent moral truths; if realism were true, there would be an improbable coincidence between those dispositions and moral truths, or else realism would be false. Realists respond in various ways, including that selection would favor detecting moral truths, or that the argument proves too much.
• Motivation. If moral judgments are beliefs, why do they motivate? See Michael Smith’s “moral problem” (The Moral Problem, 1994) and the Humean thesis that beliefs alone do not motivate.

10. Constructivism

Constructivism occupies an intermediate position: moral truths exist and can be known, but they are the result of a procedure of construction rather than descriptions of independent facts.

John Rawls (1921–2002) presented his position in “Kantian Constructivism in Moral Theory” (Journal of Philosophy, 1980; the Dewey Lectures), in the context of justifying the principles of justice of A Theory of Justice (1971): these principles are the ones that rational and reasonable agents would choose in a fair initial situation (the original position), without treating them as truths about a prior moral order. Later, in Political Liberalism (1993), Rawls restricts constructivism to a political conception of justice, without claiming a general metaethical doctrine. See the article on Rawls and Nozick.

Christine Korsgaard, in The Sources of Normativity (1996), defends a Kantian form of constructivism: the normativity of moral reasons has its origin in our condition as reflective agents, who need a practical identity and laws for themselves; the force of moral obligations derives from the reflective endorsement of our humanity. There is also the Humean constructivism of Sharon Street, which starts from our contingent values.

Critics point out that constructivism does not solve the problem (procedures need substantive presuppositions) and that it is disputable whether it amounts to realism or anti-realism, since the answer depends on how the “independence” of truths is understood.

11. Relativism

Relativism comes at three levels that should not be confused (see the entry relativism):

• Descriptive relativism: societies in fact diverge in their moral codes. This is an empirical thesis.
• Metaethical relativism: there is no objectively true morality, and the truth or correctness of moral judgments is relative to a framework (culture, group, convention). Gilbert Harman (“Moral Relativism Defended,” 1975) holds that judgments of the form “it is wrong for A to do X” are relative to conventions or implicit agreements. David Wong (Moral Relativity, 1984) defends a form of relativism that admits limits imposed by human needs.
• Normative relativism: we ought to tolerate the moral practices of other cultures. This thesis does not follow from the previous ones and, strictly, is inconsistent with metaethical relativism, since it takes tolerance as a universal norm.

Criticisms: the argument from diversity does not prove the thesis (disagreement may reflect factual differences or error), relativism seems unable to describe internal moral criticism within cultures and moral progress, and the notion of “culture” is problematic. Mackie’s argument from relativity (section 8) shows that the empirical datum of diversity is also used by other anti-realist theories.

12. Assessment

The major problems that organize the debate are, in short: the open question argument (Moore), Hume’s law, queerness (Mackie), the Frege-Geach problem, supervenience, motivation, and evolutionary debunking. Each family of theories is a response, with costs, to this set. Metaethics also converses with epistemology, especially on how moral knowledge could be possible and what role intuitions have as evidence, and with the philosophy of language, in the analysis of the meaning of moral terms.

References

• David Hume, A Treatise of Human Nature (1739–1740), Book III, Part I, Section 1.
• G. E. Moore, Principia Ethica (1903).
• H. A. Prichard, “Does Moral Philosophy Rest on a Mistake?”, Mind, 1912.
• W. D. Ross, The Right and the Good (1930).
• A. J. Ayer, Language, Truth and Logic (1936).
• C. L. Stevenson, “The Emotive Meaning of Ethical Terms,” Mind, 1937; Ethics and Language (1944).
• R. M. Hare, The Language of Morals (1952); Freedom and Reason (1963); Moral Thinking (1981).
• P. T. Geach, “Assertion,” The Philosophical Review, 1965.
• J. L. Mackie, Ethics: Inventing Right and Wrong (1977).
• Gilbert Harman, The Nature of Morality (1977).
• Simon Blackburn, Spreading the Word (1984); Ruling Passions (1998).
• Allan Gibbard, Wise Choices, Apt Feelings (1990); Thinking How to Live (2003).
• Nicholas Sturgeon, “Moral Explanations” (1985); Richard Boyd, “How to Be a Moral Realist” (1988); David O. Brink, Moral Realism and the Foundations of Ethics (1989).
• John Rawls, A Theory of Justice (1971); “Kantian Constructivism in Moral Theory,” Journal of Philosophy, 1980.
• Christine Korsgaard, The Sources of Normativity (1996).
• Russ Shafer-Landau, Moral Realism: A Defence (2003).
• Sharon Street, “A Darwinian Dilemma for Realist Theories of Value,” Philosophical Studies 127, 2006.
• Derek Parfit, On What Matters (2011).
• Stanford Encyclopedia of Philosophy, entry “Metaethics.”

← Articles