Two nondescriptivist views of normative and evaluative statements

Abstract

The dominant route to nondescriptivist views of normative and evaluative language is through the expressivist idea that normative terms have distinctive expressive roles in conveying our attitudes. This paper explores an alternative route based on two ideas. First, a core normative term ‘ought’ is a modal operator; and second, modal operators play a distinctive nonrepresentational role in generating meanings for the statements in which they figure. I argue that this provides for an attractive alternative to expressivist forms of nondescriptivism about normative language. In the final section of the paper, I explore ways it might be extended to evaluative language.

Keywords:

• Metaethics
• expressivism
• inferentialism
• representationalism
• modals

Notes

1. See Sinclair (2009) and Chrisman (2011) for discussion of the history and some relatively recent developments.

2. Portner (2009, chs. 2–3) contains a great introductory discussion; see also Chrisman (2015) for an introduction.

3. It is the view I defended in Chrisman (2012a) and have developed in more detail in Chrisman (2016a) and (2016b).

4. For different versions of this argument see Dreier (1996), Unwin (1999), Schroeder (2008), and Blome-Tillman (2009).

5. See Humberstone (1971) and Schroeder (2011) for a view in this vein. See Chrisman (2012b) for critical discussion.

6. See Heim and Kratzer (1998, ch. 12) and Von Fintel and Heim (2007).

7. I address some of these in more detail in ch. 5 of Chrisman (2016a), where I argue that there is an agentive use of ‘ought’ embedding prescriptive content which is not propositional, and I explain a way to capture the relative weakness of ‘ought’ compared to ‘must’.

8. This two factored account is the idea pioneered by Kratzer (1981). See Portner (2009, ch. 3) for general introduction and Chrisman (2015) for introduction to the case of deontic modals.

9. Chrisman (2016a, ch. 5, 2016b). See Ridge (2014) for a sophisticated development of this approach within a ‘hybrid’ expressivist framework that is capable of embracing the modal rule for ‘ought’ and treating some ought-statements as descriptive and others as nondescriptive.

10. What about embedded uses of ‘ought’? On the one hand, this might be a question about the semantic content of complex statements, such as ‘If you ought to call your mother, then you ought to charge up your phone.’ If so, the answer is relatively easy: the modal rule predicts truth conditions for complete ought-sentences like a general truth-conditional semantic model attempts to do for any declarative sentence. These predictions must then be integrated with the model’s treatment of sentential connectives, such as ‘if-then’ to predict truth conditions for the whole complex sentence in which the simpler sentences figure. To be sure, it is a matter of considerable controversy in compositional semantics how ‘if-then’ works, but as long as it takes truth-conditional complements, the modal rule for ‘ought’ will be able to integrate with a rule for ‘if-then’ to produce truth conditions for the complex sentence. On the other hand, however, the question about embedded uses of ‘ought’ could be a metasemantic question about what grounds our semantic model’s predictions of the truth conditions for complex ‘if-then’ statements. This is a very difficult question, but one that is perhaps even more difficult for representationalists than nonrepresentationalists. It is sometimes claimed in compositional semantics that ordinary language conditionals are covert modals, in which case some of the same descriptivist and nondescriptivist ideas explored here about ‘ought’ might be explored in developing a metasemantic interpretation of ‘if-then’. But there are also other possible metasemantic accounts of ‘if-then’.

11. In Chrisman (2017), I explain how this might be incorporated into a conceptual role account of meaning for the sorts of expressions of interest to metaethicists.

12. Compare Kennedy (2007).

13. See Thomson (2008) for a worked out version of this view addressing many of the shortfalls of Geach’s own suggestion but continuing in a similar spirit.

14. Also ‘good’ unlike ‘tall’ and ‘cold’ seems to be multidimensional. I will largely ignore this here, but it provides another path to a nondescriptivim about ‘good’: perhaps some contexts do not determine how the various dimensions of value are to be weighed in determining whether something is good. In such cases, the function of good-statements might not be to describe something’s value but to set a standard for weighing competing values. See Plunkett and Sundell (2013) for further discussion of this metalinguistic use of vague and context-sensitive adjective and some of the implications it has for metaethics.

15. Compare Silk (2015) and Köhler (forthcoming).