Thursday, October 04, 2007


Language in Context: 'Context and Logical Form'

I'm thrilled with how much interest and discussion my first post on Jason Stanley's Language in Context has generated already. Thanks to everyone who's joined in the discussion. I'll start here by offering a reasonably substantive summary of what I think Jason's up to. It goes without saying that this is no substitute for reading Jason himself, but it does help me to get to grips with what's going on, and it gives those of us up for discussion a refresher so we're all on more or less the same page. I should add that last time I was in St Andrews I attended a very useful seminar on this paper led by Andreas - sadly I no longer have his handout, but the discussion there improved my understanding of the paper considerably.

Next up is Jason's 2000 Linguistics and Philosophy paper, 'Context and Logical Form'. I first read 'Context and Logical Form' on a flight between Houston and New Jersey. The flight left about 4am, and I hadn't managed to get any sleep the previous night. So this paper has very strong and very strange associations for me.

Last time we saw that Jason wants the process of trying to figure out what was expressed by another's utterance to be a highly constrained process. The way he thinks that a sentence constrains the interpretation of any utterance of it is simply stated as the opening line of the paper:

'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)

Jason then tells us how to understand this thesis:

LOGICAL FORM: By "logical form", Jason means to talk about the so-called "real structure" of the sentence - the claim isn't to be read as the relatively uninteresting and uncontroversial thesis that all truth-conditional effects of context can be traced to logical form, conceived of as merely a representation of each natural language construction in a notation which makes any dependence on context perspicuous for the purpose of interpretation (30).

TRUTH-CONDITIONS: For his purposes in this paper, we're to think of truth-conditions as structured propositions: 'an ordered sequence of objects and properties' (31). The choice, at least in this context, is to be regarded as a matter of convenience rather than adherence to a substantial doctrine about the metaphysics of truth-conditions.

I seem to remember that Jason used to bill 'Context and Logical Form' as his 'statement piece' on the semantics-pragmatics interface, though I wonder if he now considers his joint piece with Jeff King (which is a mere 3 papers away on our quest) to supersede it on that front. In any case, Jason spends some time in the present paper orientating himself. He considers just three ways of carving things up, though he acknowledges (33) that they don't even begin to exhaust the options here.


1. Semantics is the study of context-invariant meaning, while pragmatics is the study of the contribution of context to communication.

2. Semantic interpretation involves the evaluation of expressions relative to contexts of utterance. So even at the stage on semantic interpretation, context can play a role in assigning denotations to elements in the logical form of an expression. Pragmatic interpretation involves taking the product of semantic interpretation and evaluating it with respect to (Gricean) conversational maxims.

3. Semantic interpretation is the process of interpretation that leads to truth-conditions. Pragmatics is 'the study of those aspects of interpretation that take as input the truth-conditions of a linguistic act, and yield other propositions implicated by that speech act' (33).

Jason opts for the second way of conceiving of the division of labor between semantics and pragmatics, though he notes (33) that if his central thesis is correct, the second and third ways of drawing the distinction coincide.

We're now in a position to get clearer on Jason's master thesis. The contribution made by extra-linguistic context in the interpretation of an assertion is restricted to the assignment of values to elements in the logical form of the expression uttered (34). (Jason's thesis isn't really restricted to assertions (31), but he focuses on assertions in this paper). Associated with each such element are rules placing very tight constraints on what context is allowed to throw that element's way; hence the role of extra-linguistic context on determining the truth-conditions of an assertion is very limited. As Jason points out (35):

'If this picture of truth-conditional interpretation is correct, then it is fundamentally different from other kinds of interpretation, like the kind involved in interpreting kicks under the table and taps on the shoulder.'

This aspect of Jason's view has always really intrigued me. And it really gets to the heart of Jason's disagreement with those such as the relevance theorists, discussed in a passage in the introduction that I didn't talk about last time:

'One possible way to pursue a systematic strategy for explaining the relation between utterances and the truth-conditions they intuitively possess is to attempt a systematic explanation of rational communicative action generally. On this model, advocated for example by relevance theorists such as Dan Sperber and Deirdre Wilson and Robyn Carston, there is a systematic account of interpretation of the mental states of others. It is our ability to make reasonable inference from the observable behavior and characteristics of our interlocutors to their mental states that ultimately explains our success in acquiring information from their utterances. On this model, linguistic behavior is no different in kind from the other cues (winks and nods, half-smiles) our interlocutors give about their inner narratives.' (9, my emphasis, some references suppressed)

For Jason, this just doesn't do justice to the speed and ease with which we arrive at the specific truth-conditions of another's assertion from hearing them having made the utterance they made (10). Unlike interpreting winks, nods and kicks under the table, interpreting the utterances of another has to be a process that is subject to considerable constraints imposed by the very expressions they uttered - as Jason nicely puts it, it has to be 'under linguistic control' (10).

Jason's opponents hear aren't just the relevance theorists; he mentions also Kent Bach, Rob Stainton, Charles Travis, and Francois Recanati. What unites such theorists is their conviction that semantics doesn't deliver up anything truth-conditional; semantics is only able to offer up something contextually impoverished, since on such views 'the truth-conditions of most utterances go well beyond what semantics can legitimately assign to the logical forms of the sentences uttered' (35). We would do better to think, with Recanati, of truth-condition pragmatics rather than truth-conditional semantics.

Pushing theorists in this direction, Jason thinks, are two assumptions:

FIRST ASSUMPTION: 'In semantic interpretation, one may never postulate hidden structure that is inconsistent with correct syntactic theory.' (35)

SECOND ASSUMPTION: 'In deriving the semantic interpretation of a logical form, every feature of the semantic interpretation must be the semantic value of something in that logical form, or introduced via a context-independent construction rule.' (36)

Jason agrees with both assumptions - he just thinks that his opponents have an impoverished conception of what the elements in the logical form of a various context sensitive expressions will be. That's an empirical issue, to be settled on a construction by construction basis, and so we turn from the grand design to particular constructions; Jason here considers two, which have been thought to undermine the kind of picture of semantic interpretation he favors, namely alleged cases of non-sentential assertion and 'utterances of expressions with sentential structure, which appear to express full-blown propositions, propositions that contain constituents which do not appear to be the values of any constituent in the logical form of the expression uttered' (39).

In the comments of the previous post, Andreas suggests that it's one of the strengths of Jason's work that he's so willing to go beyond programmatic statements of his project, and get stuck into problem cases. I think that's right, but for now I want to maintain a pretty high altitude, and discuss the importance of Jason's project for a particular topic that I'm very very interested in - the inferentialism/anti-inferentialism debate in discussions of linguistic understanding.

Consider again the third distinction between semantics and pragmatics; this splits the process of interpreting another's utterances into two parts, which following some of the plethora of terminological options here we'll describe as arriving at what was said, then determining what was meant. Arriving at what was said is a matter of semantic and perhaps pragmatic interpretation, yielding something propositional, while determining what was meant involves taking the result of semantic interpretation and evaluating it with respect to the maxims governing conversation to determine what implicatures and the like are generated. Even if we want to hold that this isn't the right way to cash out the semantics-pragmatics distinction, let us just agree to regard the latter process as a matter of inference. What of the first process, arriving - by a process of semantic interpretation or a blend of semantic and pragmatic interpretation -at something propositional and truth-conditional? Is this, at least in central cases, inferential or not?

One might have thought this issue turned on the quantity and nature of the context-sensitivity one is willing to acknowledge in natural language. As Jason writes elsewhere, arguing against accounts of linguistic understanding according to which understanding is a non-inferential perceptual or 'quasi-perceptual' grasping (see Fricker 2003, Pettit 2002, 2005):

'...the pervasive context-sensitivity and ambiguity of natural language sentences forces hearers to engage in inferential reasoning about meaning in order to grasp what is said by an utterance.' (2005: 131)

'Virtually every sentence we hear contains context-dependent expressions. Therefore, virtually all of our experience as language interpreters involves making consciously accessible linguistically guided inferences about semantic content.' (2005: 132)

So here's a thought. If you want to defend anti-inferentialism, you should deny the ubiquity of context-sensitivity in natural language. If the list of context-sensitive expressions is more or less exhausted by Cappelen and Lepore's basic set, for example, that might suggest that arriving at what is said, still considered as something truth-conditional for our present purposes, need not be considered an inferential achievement. To motivate moderate or radical contextualism over semantic minimalism, then, would be to motivate inferentialism.

But arguably, things aren't so simple. In fact, it's not only arguable, it's been argued. In his 'Does Linguistic Communication Rest on Inference?', Recanati acknowledges that in general the phenomenon of what he calls 'semantic underdetermination' provides the ammunition for 'the most powerful weapon in the inferentialist's hands' (2002: 2002). Nonetheless, despite holding that reaching something truth-conditional requires a generous helping hand from pragmatic interpretation, he thinks that one we properly understand the anti-inferentialist position, it is the more plausible position to adopt. The good news for perceptual or quasi-perceptual accounts of understanding is that 'communication is 'as direct as' perception' (2002: 125). I'm obviously glossing over a ton of details here, but the upshot if Recanati is right will be that the inferentialism/anti-inferentialism debate turns out to be much richer and more complex than we'd envisaged above. And even demonstrating a ubiquity of context-sensitive expressions outside the basic set in natural language sentences won't bury the Fricker-Pettit account of linguistic understanding, or related anti-inferentialist positions concerning linguistic communication.

I should pause to note that Jason is very careful to state his target in the 2005 paper I've been referring to, and it's plausible to think that he hits it even if one accepts Recanati's point. Jason's official target is the claim that it's a phenomenological datum that, at least in central cases, speakers simply non-inferentially grasp what is said by the utterances of others. It is, of course, perfectly consistent with recognizing the complexity of the inferentialism/anti-inferentialism debate that one urge that the ubiquity of context-sensitivity and ambiguity in natural language shows that the phenomenological claim can't be regarded as a datum.

This points at another way in which the project Jason undertakes in 'Context and Logical Form' and its successors engages with this issue. One might hold that underlying the claim that the phenomenology of linguistic understanding shows that (in central cases) hearers non-inferentially grasp what is said by the utterances of others is a true datum. It's something Jason stressed in the introduction: in most cases, we smoothly, almost effortlessly and almost instantly arrive at what was said by another's utterance, even when dealing with novel sentences. As we saw in the last post, this provides the central motivation for Jason's thesis that the process of interpretation which delivers the truth-conditions of utterances must be highly constrained. So one can see Jason as attempting to do justice to what we might hold to be the real datum underlying the controversial claim that linguistic understanding is nothing more than a kind of non-inferential grasping, thereby knocking out much of the motivation for that claim.


Fricker, E. 2003. Understanding and Knowledge of What is Said. In Barber., ed. Epistemology of Language. OUP.
Pettit, D. 2002. Why Knowledge is Unnecessary for Understanding Language. Mind 111: 519-50.
- 2005. Belief and Understanding: A Rejoinder to Gross. Mind 114: 67-74.
Recanati, F. 2002. Does Linguistic Communication Rest on Inference? Mind & Language 17: 105-26.
Stanley, J. 2000. Context and Logical Form. Linguistics and Philosophy 23: 391-434.
- 2005. Hornsby on the Phenomenology of Speech. Proceedings of the Aristotelian Society 79: 131-46.

Labels: , , ,

Great Post. Here's a question: can one summarize Jason's position as a thesis about the compositionality of truth-conditional content? I once thought that Jason held just such a position. In "Context and Logical Form", he distinguishes between his second conception of semantics (according to which semantic interpretation of a sentence S involves compositionally combining the referential contents of S's constituents) and his third conception of semantics (according to which semantics yields truth-conditions). He then remarks that if his arguments are successful, these two conceptions coincide. It seems natural then to identify Jason's position as the thesis that the truth-conditional content of an utterance is the result of compositionally combining the contents of the constituents of the uttered sentence.

However, to my best knowledge, Jason never characterized himself as defending a version of the principle of compositionality. Part of the reason seems to be that there are different versions of the principle, with some being stronger than others, and Jason is unwilling to commit himself to those stronger principles (CLF, p. 396). There is also the fact that compositionality can be stated either in terms of express-type meaning or in terms of content, and usage has not been uniform.

But I think one needs a handy little phrase to summarize Jason' thesis, one that can be conveniently used for dialectical purposes. Suppose I'm going to write a paper arguing that account A of a certain phenomenon is preferable to account B of the same phenomenon, on the grounds that account B forces us to abandon Jason's thesis and become truth-conditional pragmatists, whereas account A remains neutral on the issue. Then presumably I will write something like this: "One worry with account B is that it violates a plausible thesis, due to Jason Stanley, about…….. what?" The compositionality of content? That would be misleading, if only because Jason himself never presented his position in those terms. How about the 'Syntactic Correlation Constraint'? That's misleading as well, since the person (Bach) who coined the phrase has minimal, not truth-conditional, content in mind. Maybe one can identify Jason's position with Szabo's Context Thesis, which states that the context-sensitivity of a complex expression can always be traced to that of a constituent. But maybe there are better labels. Any suggestions?
Thanks for the interesting comment (and sorry for my lateness in responding).

You're right that Jason is very reluctant to put his core claim in terms of compositionality. I take it it's just to get as many people initially on board as possible - he doesn't want his thesis to hold out too vulnerable a hostage in the debate over compositionality. It's presumably not any suspicion of the principle that he has personally:

'The view I favour is the fairly standard one that semantic competence amounts to grasp of a compositional semantic theory for that language.' (Stanley 2005: 136)

So he's just being careful in 'Context and Logical Form', I take it, so people don't dismiss the position because of a commitment to compositionality.

But what's wrong with Jason's own statement of his view in the opening line of the paper? It's primarily a thesis, as stated there, about the limited role extra-linguistic context has to play in determining truth-conditions. So the worry with position B is that it offends against Stanley's thesis that all truth-conditional effects of extra-linguistic context can be traced to logical form, and hence position B is in trouble if one buys the arguments in 'Making it Articulated' and elsewhere against that kind of view. Is there something I'm missing in your question?
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?