In chapter 3, Ridge discusses ecumenical cognitivist
views, which, on his view (see the intro, pages 7 and 11), have some (relatively)
promising resources with regard to accounting for the continuities and discontinuities
(discussed in chapter 2) between normative and descriptive thought. Of course
Ridge argues that these views are not, in the end, correct. More precisely, he
argues that going ecumenical won’t help the cognitivist –
the case against ecumenical cognitivism still depends in part on cognitivist
views facing certain familiar problems (e.g., in relation to explaining
normative supervenience) which are not the focus of discussion in this book. I
must say that I found the arguments in this chapter quite convincing, and so
this isn’t a very critical ‘critical summary.’ But at least it’s a summary. Well,
kind of. I also try to highlight some interesting claims which I hope others
will deem questionable, but if you’ve read the chapter you’ll know where the most
interesting action is here. So, you could just skip reading the rest of my
summary and proceed to try to explain why Ridge’s arguments against ecumenical
cognitivism are problematic (or especially, interestingly awesome, or
whatever).
Ridge first offers a neat characterization of
cognitivism and ecumenical cognitivism (p. 80):
(Cognitivism) Normative claims
express beliefs with genuinely representational contents, such that the content
in question is the same as the content of the claim which expresses the belief.
(Ecumenical Cognitivism) Normative
claims express beliefs with genuinely representational contents, such that the
content in question is the same as the content of the claim which expresses the
belief, and they also express, in some interesting sense, desire-like states.
We may (but need not) understand the sense in which
normative claims express beliefs in a ‘deflationary’ way, so that a claim ‘p’
expresses a belief that p if and only if they have the same
content.
There are two forms of ecumenical cognitivism: Implicative
Ecumenical Cognitivism and Judgment-Individuating Ecumenical
Cognitivism, and these can be further divided into two subclasses. We
get the following views:
(Conversationally Implicative Ecumenical
Cognitivism) A normative claim ‘p’
conversationally implicates (in conversational contexts of certain interesting
kinds) that the speaker is in a certain kind of desire-like state.
(Conventionally Implicative Ecumenical
Cognitivism) A normative claim ‘p’ conventionally
implicates that the speaker is in a certain kind of desire-like state.
(Non-Constitutive Judgment-Individuating
Ecumenical Cognitivism) The desire-like state expressed by a
normative claim ‘p’ is not a part of the judgment that p, but being in the relevant desire-like state is a
necessary condition for counting as making the judgment in question.
(Constitutive Judgment-Individuating
Ecumenical Cognitivism) The desire-like state expressed by a
normative claim ‘p’ is a part of the judgment that p.
I hope that’s roughly right. Also, Ridge (p. 85)
notes, in passing, yet another view, Directly Implicative Ecumenical
Cognitivism, according to which “normative utterances somehow directly
express desire-like states of mind, where direct expression is contrasted with
expression via the implicature of the content that one is in such-and-such
state of mind” (and being in the relevant desire-like state is not necessary in
order to make the corresponding normative judgment).
Ridge explains all these views in some more detail (with
the exception of the directly implicative view) and characterizes some specific
variants of them (at least those defended by Stephen Finlay, David Copp, Stephen
Barker, Daniel Boisvert, and Jon Tresan). I found nothing wrong with these
explanations and characterizations, but I should add that it’s been a while
since I read some of the relevant papers (and I’ve yet to read what seems like
a super interesting new book from Finlay). If some of you have some relevant
concerns, I hope you’ll bring them up. (I was initially surprised about Ridge’s
classifying Jackson & Pettit 1995 as defending a judgment-individuating
ecumenical view of sorts, but yeah, that paper apparently suggests, very roughly,
that normative thinkers normally accept normative claims in a way which
involves having suitable desires. Also, incidentally, I seem to recall that G.
H. von Wright makes, in a paper from 1953 or 1954 (written only in Swedish),
some remarks which suggest a ‘pluralist’ view such as the one Ridge attributes
to Edwards.) Anyway, Ridge has some interesting criticisms to offer against the
different forms of ecumenical cognitivism. So, what’s supposed to be wrong with
them?
The problem with the implicative versions of
ecumenical expressivism is, in a nutshell, that they do not explain the way
that normative thought is action-guiding (and acrimonious, and also
affect-implicating – although I think that it is a bit hard to evaluate the
competing explanations for this last feature of normative thought at this
point, given that the connection to affect is contingent and so doesn’t perhaps
seem quite as problematic for a cognitivist to account for, and that the
expressivist explanation hasn’t really been laid out in any detail, yet,
either). As Ridge (p. 87) puts it, on the implicative views, “while my publicly
making a normative claim may commit me to a suitably desire-like state of mind,
my judgment does not.”
Ridge focuses on internalism and on Copp’s attempts at
a ‘debunking explanation’ of the intuitions that would seem to support a
stronger connection between normative (or perhaps more accurately: moral) judgment
and motivation than can be captured on the implicative views. Ridge registers
five concerns with the Copp-style idea (pp. 88–89). There’s no point in my
summarizing them. It’s a rather sweeping dismissal of the implicative views,
but I find it persuasive. Obviously, these are some of the passages that I’m
hoping others will find something critical to say about. (How about
disagreement? Well, presumably the idea is that the ecumenical views face the kinds
of problems discussed in chapter 2, and that appealing to desire-like states
being expressed by the public normative claims won’t be of any help to the
cognitivist.)
Ridge then proceeds to discuss judgment-individuating
ecumenical cognitivism. Here, Ridge suggests, going ecumenical might really
provide the cognitivist with some interesting resources. He now focuses on
disagreement, but presumably the thought is that the judgment-individuating
views really might have an edge over competing forms of
cognitivism when it comes to explaining the intimate connections between
normative thought and motivation. However, disagreement is,
in any case, a source of problems for the cognitivist. As we learned in chapter
2, all the cognitivist views have problems with disagreement. Very
roughly, sui generis normative features are too weird, and
analytic reductionism cannot make adequate sense of disagreement over the
normative matters between thinkers who agree on the non-normative stuff. The
extra resources of the ecumenical views might, though, help a little bit when
it comes to synthetic reductionist (or contextualist) views’ problems with
regard to making sense of our having genuine disagreements on normative issues,
and not talking past each other, under certain sorts of circumstances (Hare’s
cannibals, Moral Twin Earth, etc.). So, let’s suppose that X’s
use of ‘good’ is in some suitable way causally regulated by Fness, and Y’s use of ‘good’ by Gness. X says (and thinks that) A is good, Y says (and thinks that) it is
not – they end up speaking past each other and not really disagreeing, it
seems. Except that according to the judgment-individuating ecumenical views, X is ‘for’ A, while Y
is in some conflicting desire-like state. So perhaps they have a ‘disagreement
in attitude.’ If expressivists can appeal to this idea, then why not others? Or
that’s roughly the idea. The problem with this move is a familiar one: Y
should be able to signal her disagreement with X by saying
that what X says/thinks is not true. However, according to
the relevant cognitivist view, it would seem to be appropriate for Y to say something along the lines of ‘What X says is true, but I disagree with her,’ which doesn’t
really seem to make any sense.
Ridge writes, on p. 92, that he has argued (in chapter
2) that “all of the synthetic reductionist semantics on offer” have the
consequence that X and Y may both
speak/think truly. Now, that seems questionable, as Ridge only very briefly
outlines a ‘causal regulation’ view in chapter 2 (pp. 75–76), and there’s no
discussion of the literature that is critical of Horgan & Timmons (with which
Ridge undoubtedly is familiar). Myself, I’d guess that it may very well be that
the relevant, more sophisticated attempts at suitable synthetic reductionist
semantics either end up being relevantly similar to the kind of causal
regulation view that Ridge discusses, or they end up being relevantly similar
to the non-naturalist views. Also, perhaps any synthetic reductionist view
which secures genuine disagreement between X and Y just in virtue of their beliefs will have trouble
accounting for the ‘acrimonious’ features of normative thought. But I thought
I’d mention this, anyway – just in case someone would be inclined to try to
defend synthetic reductionism on this score.
Suppose, though, that the synthetic reductionist
attempts to appeal to disagreement in attitude. The very interesting, final
part of this chapter discusses a possible way of trying to respond to the
problem that Ridge presses in relation to the misalignment between attributions
of truth and disagreement. The response in question is due to Gunnar Björnsson,
Stephen Finlay and Alexander Almér (ABF). The basic idea is that when Y
says/thinks that what X says is not true, she need not be
assessing as false the same proposition that X’s claim has
expressed. Rather, she is assessing as false the proposition that would be
expressed by the claim that she herself would make using the
sentence ‘A is good.’ This is so, roughly, because when we
converse or think about normative matters our interest normally is in our own
values. So, that’s why it would not be okay for Y to say/think
‘X’s belief is true, but I disagree.’ I suppose one could
raise questions about how exactly this explanation is supposed to work. Ridge’s
point, however, is that even if this explanation did work in some contexts, we’d
expect there to be some contexts in which Y could make non-deviant claims attributing truth to X’s views while signaling disagreement. Something like
‘X’s belief about A is quite
literally true, and I disagree with it,’ perhaps even coupled with
‘Oh, and by the way, let me be clear that I have no desire to promote my
values,’ should seem fine. But it doesn’t. So, ABF’s defense strategy generates
wrong kind of predictions, and isn’t satisfying. I’m quite convinced, but
although I cannot think of anything to say in defense of ABF, I’m sure there’s
more to be said here.
Hi Teemu, thanks for this. I have a clarificatory question that might turn into a substantive comment. With respect to the final argument you mention (one that revolves around the apparent impossibility of 'canceling' the pragmatic phenomena that ABF use to explain disagreement): why assume that pragmatic phenomena are always cancellable?
ReplyDeleteSome pragmatic phenomena strongly resist cancellation; for example, presupposition does. It might be thought that the general desire to promote one's values is a constitutive presupposition of normative discourse itself, and not just an implicature carried by normative utterances. This could explain why the 'let me be clear that I have no desire to promote my values' preserves the weirdness: one is attempting to cancel a presupposition, which is just weirdness-generating in and of itself. This is only a tentative suggestion, though.
I also want to ask what people think of a suggestion made by ABF, namely, that it is very difficult to study real semantic competence simply by inserting such phrases as "quite literally true" into sentences in order to 'focus our attention on semantic contents', as Ridge would have it. I share Ridge's concern that this general line might make semantic theory nearly impossible: how are we to test semantic theories except against our responses to various permutations of sentences? Then again, I feel the force of ABF's concern: potentially theory-laden terms may make for the wrong responses. And, as they argue, more 'ordinary' locutions like "what was said" are in fact conditioned by pragmatics and are not adequate for isolating semantic contents.
Ridge responds to this by noting that the conversational models ABF use for their own view can be tested for semantic content (97). But, crucially, his tests at that point simply involve adding more ordinary sentences to the conversation, not adding such phrases as "quite literally true" to already complete utterances. Perhaps ABF can accept that this is a great way to isolate contents, but argue that it is distinct from the insertion of potentially theory-laden clauses. So I suppose I'm not completely convinced that the phrase "quite literally true" does the work that Ridge needs it to do.
Thanks, Nick. I first had some very inchoate worries along the lines of your first one, myself, but I guess I just couldn't see how the story would go. I don't think that Ridge is assuming that pragmatic phenomena are always cancellable. But it seems very plausible (to me) that when we use expressions like 'What she said is true,' it is possible, in some contexts, to fix on the proposition expressed by the utterance that we are commenting on. As Ridge points out, we certainly can do this for the nice examples given by Björnsson & Finlay ('I was told that Sally stole the money' etc.).
DeleteOn your second point: It's enough (for Ridge) that there's some way of isolating the relevant content, right? I don't quite understand what reason there's to be skeptical about the possibility of focusing on the right content in the normative case with a phrase like 'quite literally true,' for instance. This particular strategy might not work in other (different) cases, but that seems fine.
If I understand the ABF-style view correctly, it says that whether some normative utterance such as 'Scalping people is good' is true (in some theoretically interesting sense) is a wholly different question from whether it is to be agreed with (that is, from the question of whether the utterance is true, when I assess its truth as a participant to normative thought and talk). This seems to me to be very costly when it comes to the distribution of plausibility points among the different options that we have. But yeah, it's a big if in the beginning of this paragraph.
Not suggesting, by the way, that ABF (and Ridge) wouldn't have done great job presenting their views. I should read the AB paper and re-read the BF one.
DeleteHi Teemu, the worry is pretty simple: there might be no locution in ordinary language which tracks the semantics/pragmatics distinction. In ordinary language "literally true" is used by speakers to deny figurative or metaphorical status to some statement. That is an important task for ordinary language to achieve, unlike the marking of the comparatively arcane semantics/pragmatics distinction.
DeleteThe worry is pretty easy to motivate. Suppose Pekka Väyrynen is right and thick concepts have evaluative meaning as a matter of their pragmatics. If you ask an ordinary person whether the sentence "courageous acts are good" is "quite literally true", they will probably say yes. If "quite literally true" focuses ordinary competent users of the language on semantic contents, then Pekka is in trouble. But I'm inclined to think that he is not.
You may be right that there are other ways to tease out semantic content (though on some views of what semantic content is, for example on Kent Bach's view, this task looks positively daunting). I was just expressing the wish that Ridge say more about his use of the phrase "quite literally true" as a device for achieving this end.
Sorry to be so slow in responding to some of these very interesting points, guys - I know the next chapter is coming up today. On the issue of cancellability, I agree that some pragmatic phenomena strongly resist cancellation - that is why I need different arguments against conventional implicature views, as opposed to various kinds of conversational implicature views. On the suggestion that it is a robust and non-cancellable presupposition of normative discourse that the speaker wants to promote his values, though, I find this specific hypothesis implausible. First, cases in which the speaker holds some agent-relative view, like egoism, doesn't fit well with this hypothesis. When a mafioso tells a rival mob boss that he (the speaker) ought to kill him, it doesn't look like he is trying to promote his values in the sense of convincing the other mob boss to endorse egoism, which is what I take it this sort of view would predict. It may be that he is trying to promote his own values - that is, his own self-interest, but then again, that will also depend on the context and doesn't seem like it should resist non-weird cancellation. The case might be one in which the speaker isn't trying to intimidate the other mob boss, but instead is just being candid in a moment of weakness - he might preface his remarks with something like, "I really shouldn't tell you this, as it can't help me for you to know this, but..." And anyway I don't think that the speaker's desire to promote his own self-interest will work to capture the relevant sentences in the right way. If my egoistic remarks focus your attention on what will promote *my* interests then you will see me as assessing the truth of someone else's normative assertion in terms of whether it promotes my interests - but that isn't what we want! We want the focus to be on what would promote the interests of the person who the utterance is about to fix the content. We need to preserve the distinction between egoism, which is a universalized agent-relative view, from what Jamie Dreier memorably called "egomania," the view by some agent A that all of the other fools ought to be promoting his (A's) interests! The latter is a god complex sort of agent-neutral view, not the needed agent-relative view. So agent-relativity makes this specific proposal tricky, at least. Second, I don't think the view works well for uses of normative predicates in embedded contexts, like the antecedent of a conditional - am I trying to get anyone to promote my values when I say something like, "If abortion were permissible then infanticide would be permissible too"? Unclear at best, and certainly seems like it should be cancellable if so. After all, a nihilist who values nothing might sensibly endorse such a conditional.
DeleteOf course I realize that the presuppositional account was just one model, but my strategy is to take these proposals as they come. I took the models that ABF offered pretty seriously in the book, and tried to argue that they aren't really adequate models for their purposes. I don't have an a priori guarantee that no pragmatic story could ever be told which could evade these kinds of worries - I just haven't seen one yet, and I think we have good reason to be suspicious that this is possible *if* I'm right about how we can use 'quite literally' style locutions in these contexts - though I know that is controversial too!
Thanks a lot for the helpful summary Teemu. I too had Nick's concern about whether all implicatures are cancellable, but it's not an area I know well and I don't have much to add to the discussion above.
ReplyDeleteSome other comments, some of which are small and most of which are sketchy:
(1) I'm not sure about the taxonomy. We could more simply define cognitivism as the view that normative judgements are beliefs. I think Ridge’s main problem with that suggestion would be that cognitivism is normally taken to involve certain views about normative language, but if our taxonomy is to carve nature at the joints it seems like we shouldn't take it that way (cf. Kalderon). We should instead say that traditionally, cognitivists (in my sense) have also been 'descriptivists' or whatever term we want to use to name the relevant commitments regarding language.
Similarly we could define non-cognitivism as the view that normative judgements are desires (or desire-like states). On the resulting taxonomy, what Ridge calls implicative ecumenical cognitivism is a version of cognitivism, and judgement individuating ecumenical cognitivism turns out to be a genuinely hybrid theory. I'm not sure that this has any substantive upshot, but it seems to me to be a simpler taxonomy and one that better carves nature at the joints.
(2) When Ridge begins to discuss judgement individuating ecumenical cognitivism, he starts by pointing out that such a view won't help non-reductionists with their metaphysical problems. This might be true, but Ridge also raised a motivation worry for non-reductionism, and perhaps that problem is solved by going ecumenical. Of course, you only need one decisive objection to a view to kill it, but in the event that the metaphysical objection is a cost but not decisive, an ecumenical non-reductionist view might end up being an attractive prospect when we compare theories overall.
(3) He similarly says that analytic reductionist views aren't helped by judgement individuating ecumenical cognitivism. This seemed pretty quick to me: if the challenge of the open question argument is that of explaining the gap between believing some descriptive proposition and forming a normative judgement, then ecumenical views seem well-placed to solve the problem for analytic reductionists. I'm not sure that this will work in the end - thoughts welcome! - but I’d liked to have heard more here.
(4) I also shared Teemu's concern that Ridge's discussion focuses on Cornell-style synthetic reductionist views, and this leaves open the possibility that some other synthetic reductionist view might fare better. (Not that I have any suggestions.)
(5) Finally, Ridge doesn't here discuss desire-as-belief views. On some such views, normative judgements are simultaneously desires and beliefs. Such views differ from judgement individuating ecumenical cognitivism in that the relevant desire is *identical* to the relevant judgement rather than merely one constituent. Such views obviously face challenges, but at least with respect to the argument in this chapter, I wondered if they might be worth exploring. But I really need to think this through further.
Hi Alex, great points. A couple of quick comments:
DeleteAs for (1) and (5), Ridge does mention besire-views, on which normative judgments (or the states that manifest, e.g., as normative judgments) are unified belief-desire states. It'd be nice to get to classify these views as cognitivist. So that'd be one reason for going for a taxonomy such as the one suggested by Ridge.
As for (2), I completely agree. I'd guess that Ridge, too, would agree, and that the fact that the judgment-individuating view helps with internalism is at least one of the reasons he thinks cognitivists should go ecumenical. I perhaps said that Ridge argues that going ecumenical doesn't help. More precisely, his suggestion is, I take it, that it does help some, but not quite enough. (He does say, in the introduction (p. 7), that "both Ecumenical Cognitivism and Ecumenical Expressivism provide more promising research programs than their Non-Ecumenical rivals".)
A related thought which I haven't given much thought: if a cognitivist could appeal to expressivist-style resources in explaining supervenience, then the judgment-individuating view might help also with regard to supervenience worries (I think this thought came up in our non-online reading group here in Helsinki). I doubt the antecedent of the previous conditional is true, though.
As for (3), I completely agree - I'd think that going ecumenical helps with this particular issue (this, too, is a point that also came up in our non-online reading group). And so the judgment-individuating view seems, again, more promising than the implicative ones. But the kinds of problems that Ridge presses against Jackson & Pettit -style normative functionalism, and which he refers to in this context, still seem worrisome (or, well, decisive, I'd say).
Thanks Teemu, these comments are helpful. Just one tiny thing: On the taxonomy I proposed, the besire theory would presumably come out as a hybrid theory, which seems reasonable to me. (I'm also not sure that the besire theory is the same as desire-as-belief, but that's probably an issue for another time.)
DeleteOn Alex's (1), I had thought that a view which holds that normative judgments are not identical to representational beliefs with normative contents (or characters), but are partly constituted by them, should be a form of cognitivism. If a necessary condition on a view's being cognitivist is that it hold that normative beliefs are type identical to representational beliefs, full stop, then this classifies such judgment-individuating constitutive views incorrectly by my lights.
DeleteToo much here to comment on without taking over the blog, but I would like to say something about anti-reductionist views and motivation. My main worry about 'going hybrid' to deal with that for anti-reductionist views is that their anti-reductionism will make Michael Smith style fetishism worries pretty powerful - since the only desire that could do the job in explaining changes in normative outlook over time would have to be a de dicto desire to do the right thing qua right thing. Reductionist views could instead make the relevant desire the descriptive one corresponding to whatever descriptive property they think constitutes rightness, and that might blunt the fetishism worries.
ReplyDeleteOn besires: Teemu is right that I mention them in a longish footnote, but I would have liked to say more. In an earlier draft I had an appendix on besires, but I cut that to make the book more concise - tough choices, I guess!
Hi Mike, we have slightly fewer participants than anticipated, so I really wouldn't worry about writing all you'd like to write! It would be great if this blog was a resource for future readers to see how you respond to various concerns.
DeleteOK, Nick! I am on holiday in Paris just now but when I get back to Edinburgh I will be less restrained in future posts!
DeletePS Of course I realize that whether the fetishism objection is any good is itself a controversial question! I'm just noting how I see the dialectic, rather than going through all the work I'd need to defend the key moves in the arguments.
ReplyDeleteSorry that I'm late to the party; I've also been wondering about the cancellation arguments, and how quickly we should accept the claim that locutions like "quite literally" should be capable of focusing our attention on the literal content of reported beliefs. Here's my attempt to put a finer point on the worry; perhaps one of you can help put it to rest.
ReplyDeleteConsider a cognitivist view that suggests that 'good' is a context-sensitive term, similar in meaning to something like 'promotes my ends'. Then it's not unreasonable to appeal to 'tall' or other similarly context-sensitive terms (or incomplete predicates, as Ridge dubs them) to see whether the proposed tests actually work the way they are supposed to: if the test is a good one, 'tall' should pass, and then if 'good' fails or struggles the test can function as some evidence that 'good' does not work like 'tall.'
So, suppose that Bob and Sally disagree over whether Frank counts as tall. That is, Bob's uses of 'tall' are relativized to a standard that counts anything over 2ft, while Sally is more demanding, relativizing her uses to a standard of '6 ft or over.' Bob and Sally agree that Frank is 5'2". Sally aims to report their disagreement with:
S1: “Bob and I disagree about the tallness of Frank. Bob’s beliefs about tallness are all true."
Ridge notes that this sentence form sounds terrible in the moral case (his CB), but argues that if ABF are right that pragmatic phenomena are responsible, CB** should be fine. The equivalent to CB** for 'tall' is S2:
S2: "Bob's beliefs about the tallness of Frank are quite literally true, and I disagree with Bob about the tallness of Frank."
Structurally, 'tall' manifests the same sort of sensitivity ABF take 'good' to have: the literal content of Bob's belief is "Frank is tall[over 2 feet]", and Sally accepts that, while refusing to count Frank as tall[over 6 feet]. Just like ABF's proposed gloss of the moral case, Sally and Bob do not strictly speaking disagree, propositionally. Their disagreement can be understood as disagreement in attitude or aim, where the relevant goal is to get others to govern their application of 'tall' by their preferred standard (in Sally's case, restricting application to objects over 6 ft tall.) The use of 'quite literally' in CB** is supposed to focus hearers on the propositional content, resolving the badness of CB. But it doesn't seem to accomplish that in the S2 case.
Two things seem to have gone wrong that undermine the test:
(i) in cases where pragmatic phenomena nearly always accompany the relevant type of utterance, tricks like 'quite literally true' may be insufficient to focus ordinary hearers' attention on just the propositional/semantic content.
(ii) phrases like 'the cannibals' beliefs about the morality of scalping' may cause the test to fail independently. The rationale for this worry is again best seen in the 'tall' case: S1 sounds terrible, but there's a clear place to look for an explanation: presumably, one of Bob's beliefs about tallness is that the use of the predicate should be relativized to his preferred standards, a set of standards Sally rejects. Plausibly it is something like this that explains the failures in the moral case, too?
Hi Renee,
DeleteThanks - that is a very interesting line of reply - in spite of being quite a natural reply, given my own context-sensitive semantics for 'good' and the like, I don't think anyone has ever pressed me in precisely this way - nice.
I like your proposed diagnosis in (ii), but I wonder if I can't use something inspired by it to make my original argument work? I think it is very natural to here the description under which a speaker who utters something like S2 implicitly invokes to settle a standard of tallness is something like 'the standard we *ought* to be using in this context' (e.g. deciding whether Frank can comfortably sit in the seat or some such). If that is the way we very naturally hear the case, then that means that S2 is coherent insofar as we assume that we have some independently plausible theory of normative disagreement. If, though, we try the same move for first-order normative claims, then we seem to be off on a regress - we explain our normative disagreement by explaining our disagreement about what concept of 'ought' we ought to use, but then we explain our disagreement on that meta-conceptual normative question in terms of disagreement on some meta-meta-conceptual normative question, and so on. If something like this is right then there is a principled reason to treat normative discourse differently - as a tool we need to explain a wide range of other disagreement phenomena, and the buck needs to stop somewhere.
In fact, even if the speaker who utters S1 (I'll come back to S2 in a moment) isn't using some normative standard to fix the semantic content of S1, it will still plausibly be true, for the reasons you offer, that he believes that his standards are the ones that we ought to be using in the current context. This, though, will then be one of his beliefs about tallness, on a natural way of glossing 'beliefs about tallness'. So this belief will fall within the extension of the set of beliefs which the second sentence of S1 glosses as being true.
In the case of S2 what is asserted concerns not Bob's beliefs about tallness tout court, but Bob's beliefs about the tallness of Frank. You might argue that his belief about what standard we ought to use is not *really* a belief about Frank's tallness, but a belief about the right way to use 'tall' in relation to talk about Frank in this context. I suspect, though, that the folk recognition of and tendency to respect the use/mention distinction when deciding whether a belief is a belief about Frank's tallness, is not anything like robust enough to make us confident that they would not implicitly count a speaker's belief that we ought to use a certain standard S when deciding whether sincerely to apply the word 'tall' to Frank as a belief about Frank's tallness. So I think this works too.
So long as we have some independent theory of why we get disagreement in the pure normative case, we can handle all the other cases, but we can't use the same moves to deal with the pure normative case, on pain of infinite regress.
We can, of course invoke the idea of disagreement in attitude to try to explain the pure normative cases, but I argue with my example CB*** that this strategy doesn't work.
OK, I need to think about this all some more; I admit that it does make the application of my test much messier than I'd hoped, even if the moves I make here all work. Thanks again for the nice comment.
Best,
Mike