Plurality of words

Draft online

2008-08-29T19:43:00.000-07:00

I just posted a new draft of a paper that I've been working on for a while. The title is 'Intention-Sensitive Semantics', and there's a pdf here. Here's an abstract:

Many (perhaps all) indexicals have their reference determined, in context, at least partly by the speaker's intentions. This fact has been seen by many as presenting an insurmountable challenge to the standard, character-based semantics for indexicals. I propose a way of accommodating this intention-sensitivity within the traditional framework by treating referential intentions as parameters of context.

It's still quite a rough draft, and comments are of course more than welcome and will be greatly appreciated. The more vicious the better.

Szabolcsi on the history of linguistics

2008-08-28T13:19:00.000-07:00

Through Online Papers in Philosophy, I just learned that Anna Szabolsci (NYU) has put together a paper entitled 'From the Study of Ancient Texts to the Study of Talking People: 222 Years of Linguistics' (pdf), which as the title says tells the tale of how the field has developed from its beginnings in the 18th century to the giants of the 20th century. Among the theorists covered are Humboldt, Schlegel, Grimm, Darwin, Saussure, Chomsky and many, many more. The paper is richly illustrated with insertions from Wikipedia, and should be of interest to anyone in this area.

Heck on Context-Dependence

2008-08-25T22:40:00.000-07:00

What follows is just an edited and slightly more organized version of the notes I took at the session. It's been a while since the talk, which means that literally all I have to go on are those few pages of disordered scribblings. So there are most likely some ways of phrasing things that Heck wouldn't necessarily agree with himself, and perhaps even graver discrepancies. Be that as it may, I deem it useful to rethink Heck's points.

Heck's paper really had two aims. First, he argued that a traditional way of looking at the problem of context-dependence is wrong. Secondly, he argued for a particular view of context-dependence.

With respect to the first point, Heck began by drawing attention to a particular way of looking at the relationship between semantics and context-dependence. According to this way of looking at things, the relationship is from the very outset antagonistic. That is, the very fact that many of our expressions are context-dependent is seen as raising a particular challenge for semantic theorizing about natural languages. Theorists who adopt this point of view (Heck mentioned Putnam, Travis, Bach, relevance theorists, Grice and Montague) usually endorse the following train of thought.

A semantic theory is a meaning theory in the sense that its job is to systematically assign propositions to utterances.

A theory which systematically assigns propositions to utterances is only possible given a complete theory of the mental.

A complete theory of the mental is impossible.

Therefore, meaning theory is impossible.

Heck wanted to accept a certain kind of approach for which some have found motivation in this kind of argument, while rejecting the conclusion that meaning theory is impossible. The approach Heck wanted to accept was the Strawsonian perspective often sloganized as the view that words don't refer, speakers do. On Heck's version of this picture, there is never any rule which tells us how context determines reference; there is only negotiation and accommodation between speaker and audience.

Heck argued that the idea that there must be a simple rule fixing reference in context shows that theorists have thought that there could be an objective fact about what is referred to. By contrast, the Strawsonian holds that a referential act is needed for referring.

As an illustration, Heck considered Kaplan's well-known Carnap/Agnew Case. According to Kaplan, there are two possible reactions to this case, namely either that the demonstrative refers to what the speaker intended to refer to (Carnap), or that it refers to what people interpreted it as referring to (Agnew). But according to the Strawsonian, the very question of what the demonstrative referred to is misguided. There is no fact which could answer that question; there is just the two facts (i) that the speaker intended Carnap and (ii) that people interpreted him as intending Agnew.

Heck was willing to accept the generalization of this, namely that there are no semantic facts. Yet for Heck, the thesis that there are no semantic facts was to be understood as the thesis that there are no semantic facts over and above facts about what the speaker intends and what the audience intends. In successful communication, there's a simple convergence of these two. There is no third fact (the objective fact about what is referred to) which they both converge on. When communication does not succeed, there is no fact of the matter as to what was "really" or "objectively" referred to.

Heck now argued that even given this kind of strong Strawsonian outlook, there is still room for semantic theory. In fact, Heck claimed that the Strawsonian picture is consistent with a variety of different takes on the semantics-pragmatics divide and consequently on how semantics should be done.

First, the Strawsonian picture allows scope for traditional, recursive semantics in the following sense. Suppose we agreed that syntax delivers LFs which come with variables that need to be assigned values relative to contexts. We might even agree with theorists like Jason Stanley who believe that all contextual effects on truth-conditions must come in at the level of LF. As Heck pointed out, it is open to the Strawsonian to hold that semantics does not tell us what variables refer to; semantics tells us how propositions are calculated once the values have been set.

On the other hand, Heck maintained that his view is also consistent with dissimilar positions, like Relevance Theory in that it is independent of what the locus of context-dependence is. Even if the Relevance Theorists argues that syntax, as it were, gets added along the way, the Strawsonian can accept this too.

So, Heck's claim was that even if we accept that some referential expressions do not have their referents-in-context fixed by rules, simple or not, the threat to semantic theorizing only follows given a the preconception that such theorizing cannot find its place after contextual saturation has taken place.

Heck on Context-Dependence

2008-08-22T13:28:00.000-07:00

This post is a result of my finally getting around to something I have wanted to do for a long time (a rare occurrence), namely write-up my notes from a talk Richard Heck gave at the last Arché Academic Audit in June this year. (Heck is the auditor for the Contextualism and Relativism Project). The talk was inspiring, not least to me since it directly addressed issues which I'm working on. The talk was also brave in that Heck ventured a defense of what some might call an extreme (or even rampant) intentionalist position.

What follows is just an edited and slightly more organized version of the notes I took at the session. It's been a while since the talk, which means that literally all I have to go on are those few pages of disordered scribblings. So there are most likely some ways of phrasing things that Heck wouldn't necessarily agree with himself, and perhaps even graver discrepancies. Be that as it may, I deem it useful to rethink Heck's points.

Heck's paper really had two aims. First, he argued that a traditional way of looking at the problem of context-dependence is wrong. Secondly, he argued for a particular view of context-dependence.

With respect to the first point, Heck began by drawing attention to a particular way of looking at the relationship between semantics and context-dependence. According to this way of looking at things, the relationship is from the very outset antagonistic. That is, the very fact that many of our expressions are context-dependent is seen as raising a particular challenge for semantic theorizing about natural languages. Theorists who adopt this point of view (Heck mentioned Putnam, Travis, Bach, relevance theorists, Grice and Montague) usually endorse the following train of thought.

A semantic theory is a meaning theory in the sense that its job is to systematically assign propositions to utterances.

A theory which systematically assigns propositions to utterances is only possible given a complete theory of the mental.

A complete theory of the mental is impossible.

Therefore, meaning theory is impossible.

In particular, much attention has been paid to the problem of how the referents of context-sensitive referential expressions are fixed in context. With respect to this issue, the thought is that if there is no rule for how context fixes reference, then semantics is in danger. In turn, this idea relies on a picture according to which what's semantic is what's fixed by the rules of the language.

Heck wanted to accept a certain kind of approach for which some have found motivation in this kind of argument, while rejecting the conclusion that meaning theory is impossible. The approach Heck wanted to accept was the Strawsonian perspective often sloganized as the view that words don't refer, speakers do. On Heck's version of this picture, there is never any rule which tells us how context determines reference; there is only negotiation and accommodation between speaker and audience.

Heck argued that the idea that there must be a simple rule fixing reference in context shows that theorists have thought that there could be an objective fact about what is referred to. By contrast, the Strawsonian holds that a referential act is needed for referring.

As an illustration, Heck considered Kaplan's well-known Carnap/Agnew Case. According to Kaplan, there are two possible reactions to this case, namely either that the demonstrative refers to what the speaker intended to refer to (Carnap), or that it refers to what people interpreted it as referring to (Agnew). But according to the Strawsonian, the very question of what the demonstrative referred to is misguided. There is no fact which could answer that question; there is just the two facts (i) that the speaker intended Carnap and (ii) that people interpreted him as intending Agnew.

Heck was willing to accept the generalization of this, namely that there are no semantic facts. Yet for Heck, the thesis that there are no semantic facts was to be understood as the thesis that there are no semantic facts over and above facts about what the speaker intends and what the audience intends. In successful communication, there's a simple convergence of these two. There is no third fact (the objective fact about what is referred to) which they both converge on. When communication does not succeed, there is no fact of the matter as to what was "really" or "objectively" referred to.

Heck now argued that even given this kind of strong Strawsonian outlook, there is still room for semantic theory. In fact, Heck claimed that the Strawsonian picture is consistent with a variety of different takes on the semantics-pragmatics divide and consequently on how semantics should be done.

First, the Strawsonian picture allows scope for traditional, recursive semantics in the following sense. Suppose we agreed that syntax delivers LFs which come with variables that need to be assigned values relative to contexts. We might even agree with theorists like Jason Stanley who believe that all contextual effects on truth-conditions must come in at the level of LF. As Heck pointed out, it is open to the Strawsonian to hold that semantics does not tell us what variables refer to; semantics tells us how propositions are calculated once the values have been set.

On the other hand, Heck maintained that his view is also consistent with dissimilar positions, like Relevance Theory in that it is independent of what the locus of context-dependence is. Even if the Relevance Theorists argues that syntax, as it were, gets added along the way, the Strawsonian can accept this too.

So, Heck's claim was that even if we accept that some referential expressions do not have their referents-in-context fixed by rules, simple or not, the threat to semantic theorizing only follows given a the preconception that such theorizing cannot find its place after contextual saturation has taken place.

Two Perspectives on the Contingent A Priori

2008-08-19T23:06:00.000-07:00

I've been thinking a little about the puzzle about the contingent a priori, and at the moment my basic thought is that there are two ways of looking at the problem. I'll try to explain them here. All I really want to do is just to try to articulate the two different ways of looking at the puzzle. I'd be interested in hearing what people think about them, of course.

Given the factivity of knowledge, and hence of a priori knowledge, it is natural to think that only true sentences could qualify as contingent a priori. As a working characterization, let us understand contingent truth and apriority in the following way:

A sentence S is contingently true if S is true in the actual world and there is at least one accessible possible world at which S is false.

A sentence S is a priori if one may have a belief in S (or some appropriate semantic entity associated with S) the justificational force of which does not rest upon experimental factors, although acquiring the resources for the appropriate belief-formation may do so.

One way to see why, given this sketch of the notions, it might be found puzzling how some sentences could be both contingently true and a priori, is to introduce a notion of triviality, or uninformativeness of the sort which is intuitively associated with sentences like 'John is John'. However, approaching the puzzle of the contingent a priori through a notion of uninformativeness really gives us two ways of looking at the problem.

According the view held by Stalnaker in Inquiry (MIT Press, 1984), passing on information is a matter of distinguishing between possible worlds. On such a view, one could regard a sentence as trivial if acquiring knowledge of its truth does not require distinguishing between worlds. Consequently, since a contingently true sentence is false at some possible world w, finding out whether it is true will always require the nontrivial task of distinguishing the actual world from w.

The upshot of such a train of thought would be that contingency implies non-triviality or informativeness. The puzzle then arises from the further consideration that discovering a posteriori that, say, the sentence ‘Indonesia has over a hundred active volcanoes’ is true seems to involve distinguishing how the actual world is from other ways it might have been, namely those ways such that the sentence would be false. But since finding out that a contingent sentence is true requires exactly that kind of enterprise, this view would describe the puzzle as how we could have a priori knowledge of an informative sentence.

On the other hand, some philosophers seem to think that this way of characterizing the worry is incorrect. Indeed, Evans ('Reference and Contingency' 1979, reprinted in his Collected Papers, Clarendon 1985, p. 198) states that:

"The puzzle is not how we can know a priori something informative, but how something contingent can be uninformative."

We might conclude that what separates Evans’ view from Stalnaker’s is that for Evans, the triviality or informativeness of the sentences in question resides in the apriority, or otherwise, of the sentence.

So, the two ways of looking at the puzzle derive from two ways of looking at uninformativeness. In a framework like Stalnaker's, no contingent proposition could be uninformative, and so the puzzle becomes how we can know a priori something informative. On Evans' view, uninformativeness is tied to apriority in the first place so that the puzzle becomes something we can know a priori, and which is therefore uninformative, could be contingent.

I think this relates to the way Kripke (Naming and Necessity, Blackwell, 1981) himself describes the problem, although I'm not sure exactly how. Kripke seems to think that a proposition is a priori just in case the mere understanding of S implies knowing without relying on evidence about any distinguishing characteristic about the actual world, that S is true in the actual world. And it seems that what he takes himself as having shown is that some propositions are like this even though there is a distinction to be made between worlds in which they are true and worlds in which they are false, i.e. they are contingent.

It is interesting to note in this connection how Kripke characterizes the view he is attacking:

"I guess it’s thought that [...] if something is known a priori it must be necessary, because it was known without looking at the world. If it depends on some contingent feature of the actual world, how could you know it without looking. Maybe the actual world is one of the possible worlds in which it would have been false." (p. 38)

Kripke adds concerning the traditional view:

"This depends on the thesis that there can’t be a way of knowing about the actual world without looking that wouldn’t be a way of knowing the same thing about every possible world." (p. 38)

At the moment, I'm not entirely sure how to link this up with either of the two ways I tried to delineate above. Does this suggest that Kripke is thinking of the puzzle in ways which are more congenial to Evans' than to Stalnaker's, or the other way around, or neither?

NYC!

2008-08-16T16:38:00.000-07:00

I've arrived in New York, where I'm spending the fall semester visiting NYU. The semester hasn't started yet, so for the time being I'm playing tourist in the city while working a little bit and planning which courses to sit in on. So far, my list contains:

Stephens Schiffer and Neale's Semantics and Pragmatics and perhaps Kit Fine's Vagueness course, Philipe Schlenker's Seminar in Semantics (Linguistics Dept.), Chris Barker's Semantics I class (Linguistics Dept.).

The list could be much longer, but I'm trying to balance courses off against thesis work (and not least the NY experience).

Updates to follow.

Wright reminisces

2008-07-18T05:14:00.000-07:00

Synthese is putting together a special number of the journal collecting essays related to the work of Crispin Wright all written by former students of his. Wright's foreword, 'On Becoming a Philosopher' was published today. It makes for an interesting read about abandoning Cicero and John Donne, about supervision as dissection, IBM and Wittgenstein. A firm recommendation.

ENS lectures online

2008-07-06T05:00:00.000-07:00

Anders recently tipped me off about this quite amazing site: The Diffusion des savoirs de l'École normale supérieure. Here one can find video lectures on virtually any topic from art history to chemistry by top researchers in the fields. The site is helpfully structured, and is an incredible source of information.

In particular, I recommend the linguistics pages. For instance, watch and listen to Richard Breheny on implicatures, Danny Fox on logical form, Bart Geurts on presuppositions, Irene Heim on features of pronouns, James Higginbotham on events, Paul Horwich on truth, Angelika Kratzer on situation semantics, Friederike Moltman on sortals, Philippe Schlenker on linguistics (an introduction).

The lectures are available for download along with handouts. So, one can in fact put together a private course series to watch on those summer nights when there's nothing else to do (....). At any rate, here is certainly a great opportunity for battling the mobile sounds of Christina Aguilera on the bus. How cool are you when you're watching linguistic lectures on your iPod?

This is truly a goldmine, and I thoroughly recommend that anyone with an interest in these things take a look at this. As we all know, listening to people explain things is often a much more effective way of learning than reading. So, there's just lots to be gained from this site.

Arché/CSMN Grad Conference

2008-07-02T13:38:00.000-07:00

The call for paper for the illustrious Arché/CSMN Graduate Conference to be held in Oslo in November 2008 is out! Check it out here. The deadline is September 1st 2008. The keynotes are Carrie Jenkins, Zoltan Szabo and one 'TBA', whom I've never heard of ;-). (Actually, my sources tell me that the third one is very likely to be Cian Dorr.)

On Perry's reflexive contents

2008-06-25T05:00:00.000-07:00

I've been re-reading bits of Perry's Reference and Reflexivity (2001). Thinking about Perry's idea of a level of reflexive content, that is contents that are about the tokens which are used to express them, it struck me that the well known examples he uses seems very similar to another group of examples. Here is a typical case from Perry (2001, p. 73): You find a note with the (1) written on it, but you have no clue about what the indexicals refer to.

(1) I plan to kill him tomorrow.

Perry uses cases like this to argue that there is a level of content, which he calls 'reflexive content', and which play a key role in his strategy for mediating between the referentialists and the descriptivists.

But I can't see that there is a fundamental difference between this case and, for instance, the one discussed by Schiffer in his 'Descriptions, indexicals, and belief reports: Some dilemmas (but not the ones you expect)' (1995):

(2) Upon encountering a huge footprint in the sand: He is a giant!

Another case of this sort is Nunberg's:

(3) Walking through Versailles: Gee, he certainly spared no expense.

Both Schiffer and Nunberg take these to be cases of a distinctive descriptive use of indexicals.

Now, I tend to agree with Recanati who, in Direct Reference (1993), argued that although indexicals do have descriptive uses, (2) and (3) are not genuine cases of this kind. Recanati's reasons are, very briefly, that in cases like (2) and (3) the thought which we want to ascribe to the interpreter, "the thinking of which constitutes the hearer's understanding of the utterance." (p. 302), is a singular thought about a particular referent but where that referent is thought of under a de re mode of presentation. The important thing here is that although a de re mode of presentation does comprise some descriptive material, that material is nevertheless irrelevant for the truth conditions of the thought. Regarding (2), the descriptive material would be 'man who made the print'; and similarly for (3).

The main motivation for this analysis is that, as is intuitive, the truth conditions of the thought which we want to ascribe to the interpreter when she has understood (2) or (3) depend directly on the referent of the indexicals, independently of the descriptive complex under which she thinks of that object. For instance, the interpreter's thought in understanding (3) is true iff Louis XIV spared no expense.

I think a similar analysis is available for Perry's case. And whether or not Recanati is right about cases like (2) and (3), the fact that his analysis is applicable to (1) strongly suggests that the phenomena are in fact the same.

The natural description of (2) and (3) goes something like this. The interpreter does what she does due to the fact that there is some contextually salient property that she is able to ascribe to the referent, i.e. being the creature that made the footprint, being the monarch who built this palace.

We can assume, for both of these cases, if we like, that the interpreter is ignorant about the identity of the referent. Even if we make that assumption, the phenomenology of the case stays the same, and Recanati's direct referentialist approach is equally applicable.

So, what's to stop us from saying that the same is in play in Perry's case? Again, the interpreter is not able to identify the referent of the indexicals. But she is able to ascribe some property to it. In this case, the property is one that has to do with the token itself.

Suppose the referent of I in (1) is in fact Jack the Ripper, that of he Charles Darwin and that of tomorrow June 9 1880. I think our intuitions about the case are accommodated by saying that the thought you have when you've interpreted the discovered note is true iff Jack the Ripper planned to kill Charles Darwin on June 9 1880. But you grasp that thought under a de re mode of presentation which includes some descriptive material.

In the footprint case, the referent is Arnold and the de re mode of presentation includes the descriptive content 'creature who made this print', although it is not identical to that descriptive content. In the Versailles case, the referent is Louis XIV and the de re mode of presentation includes the descriptive content 'monarch who built this palace'. It seems like a natural step to just say now that in Perry's note case, the referents are Jack the Ripper, Charles Darwin and June 9 1880. The de re modes of presentation are as you would expect and include descriptive content like 'person who wrote this note', 'person referred to by 'he'', 'day after this note was written'.

Yet, of course, there is something different about this kind of properties and the ones the interpreter ascribes in Schiffer's and Nunberg's cases. But I don't think the difference is fundamental. The difference seems to be the following. In cases like Perry's, the property which is being ascribed has a contextual stability which does not attach to the property at stake in the other cases. If something is referred to, then trivially that thing stands in a certain relation to a token expression of some language. So, you will always be able to ascribe at least that property, no matter how impoverished your knowledge of context may be. The property the interpreter ascribes in Perry's case is, it seems, a default property that will be available in any context (presumably, this does not mean that it will be salient in any context, though).

This seems to me, at the moment, to be the only difference between the two kinds of cases. Furthermore, I don't see that this should motivate us to posit a special kind of content that we allegedly entertain in these cases. It's just the same kind of content as in the footprint and Versailles cases.