Moore's Open Question: A Bridge to Expressivism? — Epoche B2
An Inheritance That Does Not Stand: The Open Question Argument and Expressivism The story everyone tells Metaethics has a founding narrative. In 1903 G. E. Moore showed that no naturalistic definition of 'good' can be correct [1] . Naturalism collapsed; non-cognitivists arrived in the rubble and built. Ayer endorses Moore's demolition explicitly in Language [2] , Truth and Logic [3] , and R. M. Hare's The Language of Morals (1952) runs a recognisably Moorean test on proposed definitions before offering a prescriptivist account instead. Expressivism, on this telling, is the position the open question argument made room for. The narrative is wrong twice over, and this column argues both halves. Run as Moore ran it [4] , the open question argument refutes expressivist accounts of moral language just as readily as naturalistic ones [5] . And what survives W. K. Frankena's objection to the argument is not a semantic result at all but an observation about the practical role of moral judgement — an observation that cognitivists can and do accept [6] . Expressivism did not inherit the argument. It inherited a claim about motivation that belongs to nobody in particular. The argument as Moore made it Moore's target in Principia Ethica is any attempt to say what goodness is by identifying it with something else. Take any proposed definition — that 'good' means pleasant, or desired, or conducive to survival, or commanded by God. Moore's test is whether the resulting question stays open: Whatever definition be offered, it may be always asked, with significance, of the complex so defined, whether it is itself good. Compare "is this unmarried man a bachelor?", which is closed: anyone who understands the words can see the answer without inquiry. "Is what is pleasant good?" is not like that. It is a question people argue about. So, Moore concludes, 'good' does not mean 'pleasant', and by the same route it means none of the other candidates. Goodness is a simple, indefinable, non-natural quality. Two features of the argument matter for what follows. It is a test on definitions , and its evidence is what a competent speaker finds intelligible to ask. It is worth adding that Moore's target is wider than naturalism. He applies the same test to definitions that mention no natural property at all — that the good is what God commands, or what the truly real requires — and rejects them for the same reason. The argument is not an objection to science encroaching on ethics. It is an objection to identifying goodness with anything whatever, which is why Moore's own conclusion is that goodness is simple and unanalysable rather than that it is spiritual. Frankena's objection Frankena's paper 'The Naturalistic Fallacy' ( Mind , 1939) pointed out that the argument assumes what it sets out to prove. If 'good' really did mean 'pleasant', then "is what is pleasant good?" would be a closed question, and it would only seem open to someone who had not grasped the identity. Whether it seems open is therefore not independent evidence; it is the disputed point restated. To infer that the definition fails from the fact that the question feels open is to assume in advance that the definition is not correct. The difficulty deepens once one notices the paradox of analysis. Any correct analysis of a familiar concept will strike a competent speaker as saying something, otherwise there would be no point in offering it. So informativeness cannot be a mark of failure. And the argument assumes that speakers have transparent access to what their own words mean, a claim that later work on natural-kind terms has made hard to sustain. Moore's test does not decide anything by itself. The test, turned around Suppose we set Frankena aside and grant Moore the test. Then it does more damage than the founding narrative admits, because expressivist accounts fail it too. Take the simplest version: to call an act good is to express approval of it. Now ask: I approve of this act — but is it good? The question is not closed. It is the question a person asks when they suspect their own approval is a product of upbringing or interest, and asking it is not a symptom of linguistic incompetence. Refine the account and the result repeats. To call an act good is to express acceptance of a norm permitting it: I accept a norm permitting this — but is it good? To call an act good is to prefer it under conditions of full information: I would prefer it if fully informed — but is it good? Each of these questions is exactly as open as "is what is pleasant good?", and by Moore's reasoning each analysis is refuted. The natural expressivist response is that these are not proposed definitions of 'good' and so are not what the test is for. That response is available. What is not available is having it both ways, and that is the point of this column: the same response is available to the naturalist. What actually survives There is a residue in Moore's argument that no objection has removed, and it is worth stating precisely. However complete a description of an act one has — its causes, its effects, the pleasure it produces, the attitudes it attracts — a further question can still be raised, and raising it is not confusion. The further question is practical. It asks whether to go in for the thing, to pursue it, to commend it. That is an observation about the role of moral judgement in deliberation, and it is what Hare called prescriptivity. It is a real finding and it constrains theories of moral language. But it does not favour non-cognitivism, because the connection between judgement and motivation can be built in on either side. John McDowell's view is that a moral judgement is a recognition of how things are which is at the same time a recognition of a reason to act [7] , so that the perception and the motivation are not two states; David McNaughton defends a version of the same picture [8] . Whether that combination is coherent is a serious question. What is not in d