Moore Was Not a Naturalist: Lewis's Natural Properties and the Open-Question Test — Epoche B2
Lewis, Moore, and the Right Way to Fail the Open-Question Test A framing that has to be corrected first It is often said that David Lewis's work on natural properties [1] supplies a critique of G. E. Moore's non-reductive naturalism [2] . Two things are wrong with that sentence, and clearing them away is the beginning of the essay rather than a preliminary to it. First, Moore was not a naturalist of any variety. In Principia Ethica (1903) he argued that goodness is a simple, indefinable, non-natural property. Naturalism was the position he attacked, not the position he held. Second, Lewis never wrote a critique of Moore [3] , and the paper usually cited in this connection, 'New Work for a Theory of Universals' (1983), is not about ethics. It is about which properties are fundamental — which of them, in Lewis's phrase, carve nature at its joints — and it is deployed against problems in the philosophy of science, the metaphysics of laws, and the theory of mental content. So the claim that Lewis refutes Moore cannot be defended, because the two are not addressing each other. But there is a defensible claim in the same neighbourhood, and it is more interesting. Lewis did write on metaethics, in 'Dispositional Theories of Value' (1989), and there he offers exactly what Moore said could not exist: a definition of value in naturalistic terms which is meant to be knowable by reflection rather than by empirical discovery. This essay argues that Lewis's analysis meets the open-question argument head-on rather than evading it, and that seeing how it does so shows what was always wrong with Moore's test. Moore's argument Moore's question is whether 'good' can be defined. He is asking about definition in a strict sense: an analysis that gives the meaning of the term, so that the definition and the term are interchangeable without change of content. Suppose a naturalist proposes that 'good' means 'productive of pleasure'. Moore's test is to ask a question, and his flat statement of where it leaves us stands at Principia Ethica §6: If I am asked 'What is good?' my answer is that good is good, and that is the end of the matter. Or if I am asked 'How is good to be defined?' my answer is that it cannot be defined, and that is all I have to say about it. The reasoning behind that flat refusal is the open-question argument. Ask 'this is productive of pleasure, but is it good?' The question feels substantive; a competent speaker can raise it without linguistic confusion. Compare 'this is productive of pleasure, but is it productive of pleasure?', which is closed — asking it betrays a misunderstanding. Moore concludes that 'good' and 'productive of pleasure' do not mean the same, and since the same manoeuvre works against every proposed naturalistic definition, that goodness is not any natural property at all. He called the mistake of proposing such a definition the naturalistic fallacy . One clarification, because the error is nearly universal. The naturalistic fallacy is not the is–ought gap. Hume's point was about inferring a conclusion containing 'ought' from premises containing only 'is'. Moore's point is about the definability of a term. Moore thought the same mistake was made by anyone who identified goodness with any other property, natural or metaphysical — including, in the chapter on metaphysical ethics, those who identify it with the will of God. Why the argument proves too much The standard objection is that Moore's test, applied consistently, would destroy analyses that are certainly correct. This is the paradox of analysis, named by C. H. Langford in his contribution to Schilpp's The Philosophy of G. E. Moore (1942) [4] . If an analysis is correct, then the analysing expression means the same as the analysed one, and the analysis looks trivial. If it is informative, the two expressions apparently do not mean the same, and the analysis looks false. Yet informative analyses exist. That 'x is a brother' means 'x is a male sibling' is correct and unobvious enough to be worth teaching. Someone can wonder whether all male siblings are brothers without any defect of understanding, especially if the analysis is long and its parts are hard to hold in mind at once. The lesson is that openness of a question shows at most that the analysis is not obvious . Moore needs it to show that the analysis is not correct , and it does not, unless we add the assumption that competent speakers have transparent access to the meanings of their own words. That assumption is what the argument really rests on, and it is not credible. Lewis's analysis Lewis's proposal in 'Dispositional Theories of Value' has the structure the paradox of analysis predicts: naturalistic, a priori, and anything but obvious. It runs in two stages. The first stage is about valuing rather than value. To value something, Lewis says, is not simply to desire it. It is to desire to desire it — to have a second-order desire endorsing the first. This already does work: it distinguishes the addict who wants the drug and wishes he did not, and who therefore does not value it, from the person whose desires and endorsements agree. The second stage moves from valuing to value. Something is a value, roughly, if we would be disposed to value it under ideal conditions — conditions of full imaginative acquaintance with what is in question, free of error and inattention. The result is a naturalistic analysis, because desires, dispositions and conditions of information are all natural facts about human beings. It is also a rigidified analysis: what counts as a value is fixed by the dispositions we actually have, not by whatever dispositions we might have had if our upbringing had been monstrous. Lewis is explicit that the analysis is meant to be a conceptual truth, discoverable by reflection on what we mean, though not one that any speaker could produce on demand. Now apply Moore's test. Take some candidate — say, the flourishing of one's children — and ask: 'we