the boundaries of language

I think Bryan was probably right in his criticisms of my last post on this topic, but let me post some further thoughts on Cappelen and Lepore's tests for context sensitive. In a footnote on page 101, they give us the 'VP-ellipsis test'. But their discussion of this test has always struck me as somewhat bizarre.

Here's the test in short. Context sensitive expressions have fixed interpretations under VP-ellipsis. So if an expression has a fixed interpretation under VP-ellipsis, it's context sensitive, otherwise not.

This seems initially very surprising. One would have thought that context insensitive expressions also had a fixed interpretation under VP-ellipsis. But Cappelen and Lepore write:

'noncontext sensitive expressions do not exhibit this feature, as in

John bought a car, and so did Bill.' (101fn7)

But here the supposedly context invariant expression is the quantifier expression 'a car' - hardly a good choice of an uncontroversially context invariant expression.

(I'm taking it that 'noncontext sensitive' just means context insensitive here. Let me know if that's wrong).

A page over (102fn8) they suggest that Stanley is hoist by his own petard, since he employs the VP-ellipsis test to argue that vague expressions aren't context sensitive, failing to realize that the test tells against his own treatment of quantifier domain restriction. This isn't the first time I've encountered this interpretation of Stanley's argument, so I think it's worth pointing out it's badly mistaken. Stanley never uses the VP-ellipsis test, and in fact, he never argues that vague expressions as a class are context insensitive. Rather, he argues that even if they are context sensitive, that doesn't serve to offer a unified solution to the Sorites paradoxes. He takes contextualists to be making the claim that vague expressions are indexicals, and constructs instances of the Sorites which employ VP-ellipsis to fix the interpretation of the vague expression in question throughout the series. The contextualist about vagueness wanted to disarm the reasoning driving the Sorites whilst explaining why we find it so compelling, and Stanley's objection is that the contextualist's story, in terms of context shifts as we run through the series, has limited scope.

So Stanley doesn't use the VP-ellipsis test to argue that vague expressions are context insensitive. That's a really bad interpretation of his argument. Rather he uses the fact that indexicals have a fixed interpretation under VP-ellipsis (a couple of tricky examples such as the now famous A - 'I love you', B - 'I do too' exchange aside) to enable him to construct instances of the Sorites which contextualist accounts of vague expressions (or at least those according to which they are indexicals) can't disarm. And since the claimed fact is explicitly restricted to indexicals (and demonstratives), we needn't expect quantifier expressions to behave similarly.

So let me boldly state my conclusion: the VP-ellipsis test provides no test for context sensitivity, and it is powerless against Stanley's acccount of quantifier domain restriction. Right?

Labels: Cappelen and Lepore, Context Dependence, Jason Stanley, Philosophy of Language

I'm getting a little out of order here, but I had to read Literal Meaning over the break, so it made sense to offer something on Jason's NDPR review just now, and then come back to 'On Quantifier Domain Restriction'.

One of the things I've noticed reading the literature on this stuff is that no one agrees about where the really interesting issues lie. Recanati, for example, spends a great deal of time putting distance between his view and Relevance Theory on issues such as whether primary pragmatic processes (those that go into determining truth-conditional content, rather than those by which what is communicated is determined) are inferential (a topic which some of you will know is becoming very close to my heart). In contrast, Jason writes:

'the genuinely important disputes in the theory of meaning are between those who maintain that the contents primarily asserted by speakers are not generally the semantic contents of the sentences they use (even relative to contexts), and those who maintain that the contents primarily asserted by speakers are generally (not always, but typically) the semantic contents of the sentences used (relative to those contexts).' (237)

Cappelen and Lepore also want to downplay the importance of debates between contextualists, but from their perspective, Jason's view is to be lumped together with Recanati and the Relevance theorists:

'The label 'the Deep/Fundamental Issues' should be awarded to a range of issues independent of any debates internal to MC or RC.'
(Insensitive Semantics: 13)

So not only do we have a range of very different takes on this set of issues, which position one takes seems to shape one's perception of where the real action is. I don't think this is at all unusual in philosophy, but here we have a particularly stark illustration of the phenomenon.

Jason's review is of particular note between it takes on directly the difficult tasks of offering a taxonomy of positions in the debate and trying to get clear on what's at stake between them (as befits a review of Recanati's book). It also raises some methodological issues about how the debate should be conducted, and it's these I want to discuss here. I'll come to them a little indirectly, however, via a discussion of the following (independently interesting) issue.

Let's follow Recanati in calling the thesis Jason wants to defend indexicalism, and consider the following question: does indexicalism start at a serious disadvantage in the nature over the nature and extent of context sensitivity in natural language? In Literal Meaning, Recanati argues that indexicalism does indeed occupy lower ground, on account of the kind of claim the indexicalist must defend:

'Without going into those [technical and empirical] details, it is fair to say that the indexicalist starts with a significant disadvantage; for he makes a universal claim while his opponent only makes an existential claim. For his opponent to win, it is sufficient to produce one convincing example of a strong pragmatic effect. But the indexicalist is condemned to deal with all putative cases, and to show that they are not what they seem to be.' (2004: 89-90)

Jason writes in response:

'Recanati misstates the dialectical situation. The contextualist's method for arguing against indexicalism is to produce a reading R of a sentence S, and argue that R is not the result of the semantic interpretation of S, relative to the context of utterance. In each such case there are three responses available to the indexicalist.......The advocate of strong pragmatic effects on intuitive truth-conditions must produce a case, and show that none of these options is available for that case. For each putative case in which it can be persuasively argued that [it is not an option to 'establish that the alleged reading is not part of the intuitive truth-conditions of an utterance of that sentence, but is instead due to the pragmatics'], the contextualist [...] must establish that there is no way of accounting for the problematic reading within the semantics. From this perspective, it is the contextualist who makes a universal claim.' (239)

It's not immediately clear what Jason is claiming here. The first and last sentences of this passage might suggest that he thinks Recanati has just got things back to front; it is the contextualist who makes the universal claim and the indexicalist who makes the existential claim. But that doesn't seem quite right. For indexicalist clearly does make a universal claim; recall Jason's statement of his core thesis from the opening lines of 'Context and Logical Form':

'My purpose in this chapter is to defend the thesis that all truth-conditional effects of extra-linguistic context can be traced to logical form.' (30)

So I think a charitable reading of Jason's response to Recanati has to credit him with recognition that at some level Recanati has correctly described the structure of the debate between them. I think that in claiming that Recanati 'misstates the dialectical situation', Jason is referring less to the claim about the structure of the debate, and more to the suggestion that the indexicalist starts with a disadvantage. The contextualist might only have to produce one persuasive case of a truth-conditional effect of extra-linguistic context which is not traceable to logical form, but Jason's real objection is that Recanati has misrepresented how easy it will be to come up with the golden ticket here.

Interestingly, Jason's response here seems to assume that indexicalism should be the default position, and that the burden is on the contextualist to provide examples that aim to move us away from indexicalism. This is a point he addresses explicitly in the review, responding to Recanati's charge that indexicalism is 'dogmatic and stipulative' (it's also a theme of the introduction to Language in Context, recall). He writes:

'Why believe the indexicalist hypothesis? Recanati finds excluding "top-down" or "strong" pragmatic effects on intuitive truth-conditions "as dogmatic and stipulative as the literalist restriction of context-sensitivity to a short list of familiar indexical expressions' (160). But the indexicalist claim is an empirical hypothesis, not a stipulation about content. Furthermore, it is an empirical hypothesis that has a reasonable basis.' (238)

'The indexicalist's position is a plausible starting hypothesis for how language is able to be sufficiently elastic as to be usable, and sufficiently rule-governed as to be useful. It is a reasonable empirical hypothesis, in advance of detailed inquiry.' (239)

But I'm not sure this really gets at Recanati's worry. Though he perhaps didn't put things very well by accusing indexicalism of being 'dogmatic and stipulative' (and hence inviting precisely the kind of response Jason offers), I think Recanati worry is methodological from the outset. The charge he wants to make isn't really that Jason's claim is empirically empty; as he admits in a passage cited above, the issues get technical and empirical very quickly. The point is rather that we shouldn't endorse such a general thesis in advance of examining the various constructions. We should instead adopt methodological contextualism:

'Any expression may, upon analysis, turn out to have a context-sensitive semantics, that is, be such that its truth-conditional contribution varies across contexts.'

Put in hippy form, Recanati's objection to Jason is that he's not going into the study of particular constructions with an open mind; he's already laid down in advance a strong generalization about the limited role extra-linguistic context can play, and that, Recanti is suggesting, is stifling.

If this is how we should take Recanati's objection (and I think it is, though I have to admit there are passages in these pages where he clearly seems to be making the kind of charge of empirical emptiness that Jason takes him to be), then Jason's comments in his review don't really seem to get to the heart of the issues; Recanati would presumably just regard it as more of what he was objecting to. That's not to say Jason has nothing to say in response; on the contrary he's going to push the line that contextualist views that allow extra-linguistic context to determine truth-conditional content in ways that go beyond providing values for items in logical form render mysterious how it is that we can interpret the utterances of others as smoothly as we are able to. And so indexicalism provides a better starting point. As he writes in the introduction to the book:

'I have great difficulty seeing alternative proposals as sufficiently constrained. Proposals that appear smoothly to account for the bewildering variety of ways in which context appears to affect the relation between utterances and their intuitive truth-conditions only do so because they appeal (whether overtly or covertly) to interpretative processes that are deeply mysterious or wildly unconstrained. I prefer to explore a view that is neither mysterious nor unconstrained, and see how far one can use it to explain our linguistic behavior. If, at the end of the day, my proposal is too constrained, perhaps the investigation into its adequacy will reveal systematic ways of modifying it or liberalizing it. This seems more promising that beginning with a mysterious and unconstrained process, and trying to add on stipulative constraints. If we do not have a clear grip on the process with which we began, why thing that adding on restrictions will yield any greater elucidation?' (21)

We'll get to assess the charge of lack of constraint when we get to 'Making it Articulated'. In any case, these methodological issues about which perspective we should adopt going into the empirical study of different constructions seem about as challenging, rich and important as the issues about who's right, and it's a virtue of Jason review that it brings the former issues into such sharp focus.

Labels: Cappelen and Lepore, Context Dependence, Jason Stanley, Language in Context, Philosophy of Language, Recanati

There's been some further discussion on the thread about Stanley's motivations concerning Cappelen and Lepore's stance on issues about context sensitivity. Here I'd like to discuss one of their tests for context dependence put forward in Insensitive Semantics, since it strikes me as deeply problematic.

Here's their Inter-Contextual Disquotation Test:

Say one wants to test whether some expression e is context sensitive. First, one choses a sentence S containing e, but no other context sensitive expressions (2004: 105fn12). Then S, and hence e is context sensitive only if there is a true utterance of the following:

(ICD) There are (or can be) false utterances of ^S^ even though S (Where ^^ should be read as corner quotes).
(2004: 105)

Cappelen and Lepore are explicit that past and future utterances can be quantified over here (see 2004: 205fn10) Problem then is that S might feature a context invariant expression which denotes a property that an object might possess at one time, but not at another. For example, the following is supposed to show that the first-person pronoun is context sensitive:

(2c) There is a false utterance of 'I'm hungry' even though I am hungry. (2004: 106)

But I don't think this shows anything at all about the context sensitivity of 'I'. For (2c) could be true even on the bizarre assumption that 'I' invariantly denotes me (i.e. Aidan McGlynn), since my being hungry now does nothing to prevent there from being false past, future or merely possible utterances of 'I'm hungry'.

Objection: The schema should really have been understood as this:

(ICD') There are (or can be) false utterances of ^S^ at t even though S at t.

So the idea is there can be false utterances of a sentence S co-temporally with S being the case. That's the real mark of context sensitivity.

Response: This obviously won't work, since sometimes a context dependence expression is sensitive to the time of utterance. For example, take Cappelen and Lepore's discussion of 'now':

'(2d) There is a false utterance of 'Tom is leaving now' even though Tom is leaving now.

Suppose Tom is leaving now. Then, obviously, any utterance of 'Tom is leaving now' made at times other than now, say, a few days into the future when Tom isn't leaving, suffices to establish that the test utterance of (2d) expresses a truth."
(2004: 106)

By the way, notice that my original point could equally have been made with this example - one could repeat just the same reasoning as Cappelen and Lepore offer even if (2d) hadn't contained 'now':

Suppose Tom is leaving. Then, obviously, any utterance of 'Tom is leaving' made at other times, say, a few days into the future when Tom isn't leaving, suffices to establish that the test utterance of (2d) expresses a truth.

Unless I'm missing something, the ICD test is pretty hopeless insofar as the aim was to find a necessary condition for an expression to count as context sensitive.

(By the way, if anyone knows any literature where this or a related point is made about this test, I'd be very interested to know. I'm still catching up on the literature generated by the book).

Labels: Cappelen and Lepore, Context Dependence, Philosophy of Language

In Language Turned on Itself, Cappelen and Lepore argue that any viable theory of quotation must preserve the following disquotational schema and strong disquotational schema:

QS: ' 'e' ' quotes 'e',

SDS: Only ' 'e' ' quotes 'e',

where both occurances of 'e' are replaced by any quotable item (25-6).

Now, Cappelen and Lepore's own minimalist theory simply takes QS as 'the fundamental axiom governing the semantics of quotation expressions' (124). So it's in the bag. Cappelen and Lepore go on to argue that QS entails SDS, so long as quotation expressions are unambiguous. Here's their argument from p128:

"- From QS it follows that for any quotable item e, ' 'e' ' quotes 'e'. If ' 'e' ' quotes some quotable other than 'e', say 'a', then it follows that the quotation expression ' 'e' ' is ambiguous (because it quotes both 'e' and 'a').

- But, as we argued in Ch. 7, quotation expressions are unambiguous (and context-insensitive and not semantically indeterminate).

- So the quotation expression ' 'e' ' can only quote 'e'.

- So SDS follows."

But notice the conclusion reached isn't SDS at all; we wanted to reach the claim that only ' 'e' ' quotes 'e', not that
' 'e' ' only quotes 'e'. And these simply aren't the same claim (even if they turn out to be equivalent).

Fortunately I think there's a quick fix to their argument so that they clearly get the result they want. For suppose that something other than ' 'e' ', let's say ' 'a' ', quotes 'e'. From QS it follows that ' 'a' ' also quotes 'a'. But then ' 'a' ' quotes both 'a' and 'e', and hence would be ambiguous. So, if quotation expressions are unambiguous, only ' 'e' ' quotes 'e', as required.

This is, I submit, how Cappelen and Lepore should have run their argument. QS is exploited to show that if something other than ' 'e' ' quotes 'e', that quotation expression would have to quote something else also, and hence would be ambiguous.

Labels: Cappelen and Lepore, Context Dependence, Philosophy of Language

I arrived in Glasgow just this morning, and I took the flight to finally get around to reading Insensitive Semantics. For now, I'll limit myself to offering a couple of brief quotes with commentary:

'Focus on this question: If a range of people can't all be ready, can they at least all be ready for an exam? Is that something a range of people can have in common? We suppose the answer must be 'yes'.' (168)

I'd have thought anyone who'd taught undergraduates would take the answer to this to be 'no'.

More seriously, we're given a 7-step program to fix the minimal proposition expressed by an utterance of sentence S on 144-5. Step 5.d is:

'Precisify every vague expression in S.' (145)

How? One of the motivations for supervaluational treatments of vagueness is that they let you reason with sentences with vague constituents as if they were perfectly precise, but without committing us to the idea that there's a uniquely privileged way of precisifying them, or even that it makes sense to think we could precisify them if we wanted to. Supervaluationism is, in part, a response to the difficulties inherent in the idea that we can sharpen vague expressions. So how are we meant to go about taking step 5.d?

Cappelen and Lepore do work through an example sentence, but sadly they simply assume 'for the sake of argument' that it contains no vague expressions. However, they weren't offering an argument; this all happens in a section headed 'Semantic Minimalism: Illustration', so I can only presume they were meant to be offering an illustration of how to apply the 7 steps to a particular sentence in order to fix the minimal proposition it expresses. So I don't see how the trick's to be pulled with 5.d, and they just glossed right over the issue in the very section supposed to illustrate how it's done.

Labels: Cappelen and Lepore, Context Dependence, Philosophy of Language, Vagueness