Chapter 5: The Frege-Geach Problem

In this chapter Ridge aims to demonstrate that Ecumenical Expressivism can accommodate the kind of logical complexity needed to answer the Frege-Geach problem. Of particular concern is showing that he can give a compositional account of normative sentences of arbitrary complexity which (i) avoids rampant ambiguity, (ii) respects a formal notion of validity,which it preserves even in intensional contexts, and (iii) vindicates the epistemological soundness of inferences corresponding to valid normative arguments (as opposed to wishful thinking).

There's a lot here, so I'll focus my discussion on Ridge's positive proposals, though I should note that in section two he engages in a critical discussion of what he takes to be the most promising non-ecumenical expressivist solutions to contrast with the performance of the ecumenical expressivist story. Since many of my own concerns with this chapter have to do with the treatment of intensional contexts, I'll try to summarize the major elements well enough to set up discussion, then get to my more critical but peripheral comments towards the end.

The basic move is to offload all logical complexity to the content of the representational belief, instead of attempting to build it into the attitudes. For any logically complex sentence S in which a normative predicate is used (not merely mentioned, and bracketing intensional contexts for now):

"S expresses (a) a normative perspective, and (b) the belief that s*, where s* is what one gets when one takes 'S' and replaces all occurrences of normative predicates in 'S' with the obviously corresponding phrases about what any admissible standard would be like in the relevant way."

Accommodating the logical complexity in the belief component rather than in the attitude provides flexibility to differentiate not believing that p and believing that not p, where p involves a normative predicate. If G is the set of standards for practical reasoning that an agent is disposed to reject, then believing that "E is not valuable as an end" amounts to having (a) a stable policy of rejecting standards in G, and (b) the belief that if any standard s assigns positive weight to E then s is in G (i.e. is not an acceptable standard). By contrast, merely not believing that "E is valuable as an end" amounts to having (a) but lacking the belief in (b) above.

I'm unsure whether this account actually gets the right content for normative claims of various logically complex forms. Suppose a speaker is trying to trace the ethical implications of a framework F, and says

C: "If F is true, then stealing is never wrong."

If I've properly understood the view, C expresses something like [If F is true, then any standard that rules out stealing is in G] where G is the set of standards that the speaker in fact rejects. That picks up the wrong content, insofar as someone who in fact is disposed to reject F can coherently assert C.

Validity
Ridge offers the following characterization of validity for normative arguments:

An argument is valid iff any possible believer who accepts all the premises but at the same time denies the conclusion would thereby be guaranteed to have inconsistent beliefs, where this remains true on any acceptable substitution of the non-logical terms of the argument.

An acceptable substitution is one that preserves the grammatical and semantic kind, substituting a value that "makes the same kind of contribution to the meaning of the sentence as a whole."

To dispel the worry that this fails to explain why the validity of the arguments is transparent to speakers, Ridge distinguishes what the speakers 'understand' from what they 'know', arguing that knowledge, not understanding, is all that is required for semantic competence. They know that the arguments are valid, since they have true beliefs about their validity that are robust (would survive acquaintance with the truth) and their assessments are highly reliable, even if the speakers form their judgments on mistaken (realist) grounds (and thus fail to understand why the arguments are valid.)

Since the ecumenical expressivist casts normative content as involving a perspective/belief pair, it is natural to gloss normative arguments as moving from the propositional belief set determined by the premises to another propositional belief inferred from them. The attitudes play no essential role in the argument, and so the view neatly avoids making moral reasoning resemble wishful thinking.

Intensional Contexts
Intensional contexts (contexts where we cannot guarantee preservation of truth by substitution of co-referring terms) pose a special challenge to expressive theories. These include modal, propositional attitude ascriptions, and contexts that blur the line between using and mentioning a term. The point where I find myself least sympathetic is in Ridge's solution for attitude ascriptions. The challenge here is similar to the puzzle I raised earlier about some conditionals: though we use normative language in making attitude ascriptions, we aren't expressing our own normative perspectives and corresponding beliefs. How is a theorist to adequately accommodate these embeddings without claiming that normative language is ambiguous?

Ridge offers this analysis: a belief ascription of 'p' to S attributes to S the belief expressed by 'p' when asserted straight out. So in such contexts the contribution of a normative term does differ from ordinary contexts, but in a way that is systematically linked with the core meaning, and thus unobjectionable. It's not clear that this gloss pairs well with the expressivist element of the account. He suggests that this analysis is vindicated by the fact that it mirrors the behavior of racial epithets, which though (it looks like) he takes to also have hybrid-expressive structure, do not express the speaker's contempt in attributions like N1:

N1: "David Duke just thinks of me as a n-----r."

This isn't the best case for that point, though, since the slur looks like it functions metalinguistically in N1, and so is not an apt model for the normative embeddings Ridge is interested in. More generally, it's highly controversial whether slurs ever really embed in attributions or reports, and those who do think they embed have offered that fact as evidence against a hybrid-expressivist analysis (Kent Bach's recent paper Loaded Words runs just such an argument). If we are testing for parallel behavior in attributions, we should use something like N2, which does not single out the slur as though scare-quoting it:

N2: "John thinks I ought to be kind to tall n----rs."

N2 has the virtue of giving us several different types of context-sensitivity to compare to the behavior of 'ought'. The indexical, as Ridge notes, has its referent fixed by the report context, picking up the speaker rather than John. The incomplete predicate 'tall' could probably be understood to be relativized to standards in John's context on some reading. It's just not clear that the slur can do likewise: occurring as it does in N2, it looks to me to commit the speaker, not John, to contemptuous attitudes.

If the account of normative language is modeled off the mechanics of racial epithets, then in this case both the 'ought' and the slur should contribute the content they would have if asserted straight out by John. The fact that this doesn't occur threatens both the claim that there is a natural model for the kind of meaning shift Ridge posits between ordinary and expressive contexts, and the thought that if such a shift causes any problems, they are problems common to everyone who accommodates the behavior of slurs, and not ones that arise uniquely for the ecumenical expressivist. It may be that there are other reasons to prefer an ecumenical expressivist approach, but it looks like here the view comes up short, facing a problem that is not equally a problem for other views, and does not clearly possess the resources to dispel it.