Metaethics

2. Language

Although Moore claimed the existence of a metaphysical property of goodness, his famous argument for it turned on an observation about moral language. This open question argument points out that for apparently any proposed definition D of the word ‘good’, one can sensibly ask of an object (even something that as a matter of fact is both D and good), ‘I know it is good, but is it D?’ or ‘I know it is D, but is it good?’ Moore believed this proves that the meaning of ‘good’ is not correctly captured by D (i.e. they are nonsynonymous), and therefore that goodness cannot be the property of being D. The first philosophers to respond to Moore accepted the negative metaphysical conclusion of the open question argument, but objected that Moore had misunderstood the import of this nonsynonymy of moral words with any string of nonmoral words. These were the noncognitivists described above, who saw the real lesson of Moore’s argument as being that moral words are categorically different from ordinary words in that their meaning and use consists not in description of the world or statement of facts, but in expression of attitudes or performance of various kinds of nondescriptive speech acts. (See Open question argument.)

Moral language presents some difficult problems for noncognitivism, however. In a variety of ways, it seems to behave like descriptive language and unlike other, less controversially nondescriptive language (like ‘ouch!’ and ‘hurray!’). A focal issue is the so-called Frege–Geach problem, which turns on the ‘embedded’ use of moral sentences, as in conditionals like the sentence, ‘If lying is wrong, then getting your little brother to lie is wrong.’ In these uses speakers seem not to be doing the kinds of things claimed by noncognitivist theories, which would therefore appear to have provided at best an incomplete account of the meaning of moral sentences (see Frege–Geach problem). Most contemporary noncognitivists pursue the programme that Simon Blackburn (1993) labelled ‘quasi-realism’, the attempt to show that all the seemingly cognitivist features of moral practice can be satisfactorily explained by purely noncognitivist means. The central issue of language-oriented metaethics is therefore whether moral words and sentences have a descriptive meaning, as cognitivism claims (e.g. Foot 2001), or a nondescriptive, ‘expressive’ meaning, as noncognitivism claims (e.g. Blackburn 1993). Hybrid theories which postulate both kinds of meaning are also possible and well-represented in the field (e.g. Copp 2001).

A different way of responding to Moore emerged in the 1970s, when philosophers of language realized that correct definitions of words do not always have to be synonymous with them as the open question argument assumes. The classic example is the definition of ‘water’ as meaning H₂O. Although this is a correct definition, one can sensibly ask, ‘Granted that this is water, is it H₂O?’ and ‘Granted that this is H₂O, is it water?’ Whereas previous metaethics was ‘analytic’, conducted through an investigation of our moral concepts or understanding of our moral language, this discovery ushered in ‘synthetic’ metaethics, which ignores moral language as largely irrelevant and seeks an understanding of moral properties like goodness and rightness through an empirical investigation into the world. For several decades these developments caused many metaethicists to dismiss the study of moral language as fruitless, but the new century has seen a revival of interest.