Here, I want to argue against Prinz’s attack on non-cognitivism. He thinks that there are good empirical reasons to reject non-cognitivism. His first argument is that cognitivism can predict the surface form of moral language better than non-cognitivism. First, I argue against this by pointing to non-cognitivist accounts of moral language that I think can predict this surface form. Second, I provide a speculative non-cognitivist theory of why moral language has the surface form we can observe. Again, I think my proposal is in agreement with Prinz’s naturalized metaethics. I do think, however, that it challenges him to explore the space of possible accounts. My proposal shows, I hope, that the empirical evidence cannot, at this point, decide this question.
The second question in Prinz’s decision tree is whether or not moral judgments are truth-apt. Can they be true or false? Theories that answer yes to this question are cognitivistic, while theories that answer negatively are non-cognitivistic. Non-cognitivism is a collective term that can refer to many different theories (Shafer-Landau 2003, p. 17). It consists of two theses (Roojen 2013, section 1.1): the first says that moral utterances do not express propositions; they’re not truth apt. This is a semantic thesis about moral language. The second thesis says that moral beliefs are not representational. They do not refer to anything in the world. This is a thesis about the mental state of the moral agent. Here Prinz wants to defend cognitivism by providing empirically-informed reasons to reject non-cognitivism. He defines expressivism in the following way (we can think of Expressivism as one form of the first, semantic thesis of non-cognitivism):
Expressivism =Df Moral assertions express mere feelings or non-assertoric attitudes, and do not purport to convey facts. (Prinz this collection, p. 7)
Prinz denies both of the two theses that make up non-cognitivism. He argues that the most obvious empirical prediction of non-cognitivism fails, as he thinks that if non-cognitivism was true we would expect our moral language to have a non-cognitive form (Prinz this collection, p. 16). But this is not the case. It seems that our moral language mostly has declarative form.
If this is correct, and if I don’t have reasons to disbelieve it, does it mean that non-cognitivism makes wrong predictions? I don’t think this is the case. Much of the work in non-cognitivism is dedicated to explaining this apparent tension. But I don’t think that this involves “elaborate logics”, as Prinz puts it (this collection, p. 16). Rather, most non-cognitivists provide theories about the nature of moral discourse that show that we should expect the surface grammar to be declarative. I don’t think that non-cognitivism has or needs to have these “obvious empirical predictions”.
The starting point is to look at the way language is used. It is not the literal meaning of ethical terms that are of interest but their function (Björnsson 2002, p. 328). Expressivism entails a pragmatist theory of moral language:
[T]he pragmatist attempts to describe the function that a word, phrase or concept plays in human life, and once he has satisfied his curiosity there, he does not think that there are any further questions to ask about utterances of that sort. (Smyth 2014, p. 608)
Arguably, such a pragmatist view is easier to naturalize because we have the social sciences, which offer large toolboxes for investigating human practices.
Although Prinz’s definition of expressivism may be at the heart of non-cognitivism, in most cases this is not the whole story. According to expressivism, moral terms are not only used to express ones attitudes but also to provoke certain attitudes in the hearer. This idea goes back to the early emotivists. The “dynamic use” of language (Stevenson 1937, p. 21) involves the manipulation of others: “[E]thical terms are instruments used in the complicated interplay and readjustment of human interests” (Stevenson 1937, p. 20; emphasis in original).
Stevenson, and many others following him, analyze expressions like “x is good” as meaning “Hooray for x! Do hooray as well!” (Stevenson 1937, p. 25).[6] It expresses the speakers attitude and the wish or the prescription that the hearer should adopt this attitude as well.
At this point Prinz could reiterate his point and simply ask: “Why then do we say ‘this is good’ and not ‘I like this, do so as well’?” Here I want to offer a speculative answer: because we don’t like to be manipulated. If the function of moral language is, at least in part, to influence the attitudes and the behavior of others, I think we should expect it to take this form. This is because a declarative sentence has more authority than a mere expressive one. If I want someone to do something it is arguably more effective to disguise it in non-subjective form, to give it the appeal of a truth-aptness.[7] I want to disguise it so that it will serve this persuasive purpose.
I don’t want to say that these ideas are correct. But they’re plausible theories that predict the surface form of moral language, and which are no worse than cognitivistic theories. Expressivism focuses on what people do with language. It focuses on the speech act, not the literal meaning. Whether people express, declare, prescribe, describe, recommend, or evaluate is nothing we can easily read from the surface form. But this is what Prinz seems to presuppose when he says the most obvious empirical prediction fails. We have to look at their behavior and the pragmatic context in which the discourse happens.
I argue that this fits even better with Prinz’s project of a naturalized metaethics. When Prinz discusses the last step in the decision tree, he writes: “Naturalizing relativism will require the marriage of cultural anthropology and sociolinguistics” (this collection, p. 24). I think this marriage could be more helpful at an earlier stage in the decision tree—to help answer the question of whether or not moral terms aim at truth.