• Increase font size
  • Default font size
  • Decrease font size

Contextualism and Compositionality

E-mail Print PDF

Contextualism and Compositionality - Conference

May 17-19, 2010


May 17 & 19 : École normale supérieure 45, rue d'Ulm 75005 Paris - Salle Dussane
May 18 : Institut d'études avancées (Paris), Maison Suger 16-18, rue Suger 75006 Paris. 



It is often claimed that contextualism the view that context-sensitivity goes beyond indexicality and affects every natural language sentence is incompatible with compositionality and threatens the project of building a systematic semantics for natural language. The aim of the conference is to assess that claim and to discuss other issues related to contextualism and compositionality.


Nicholas Asher (Institut de Recherche en Informatique de Toulouse, France)
Paul Elbourne (Queen Mary University of London)
Ruth Kempson (King's College London)
Angelika Kratzer (University of Massachusetts Amherst, USA)
Peter Lasersohn (University of Illinois, USA)
Peter Ludlow (Northwestern University, USA)
Peter Pagin (Stockholm University, Sweden)
Jeff Pelletier (University of Alberta, Canada)
François Recanati (Institut Jean Nicod, France)
Zoltan Szabo (Yale University, USA)
Dag Westerståhl (Gothenburg University, Sweden)
Stephen Yablo (Massachussetts Institute of Technology, USA)


Conference Schedule

May 17th 

9.30 Welcome
9.45-10.35 Peter Pagin (Stockholm University) "Compositionality and Complexity"
10.35-11.05 Discussion
11.05-11.30 Coffee Break
11.30-12.20 Jeff Pelletier (University of Alberta) "Holism, Compositionality, and Context"
12.20-12.50 Discussion

Lunch Break

14.45-15.35 Angelika Kratzer (UMass Amherst) "Syntactic constraints on meaning composition: deriving the meanings of conditionals"
15.35-16.05 Discussion
16.05-16.30 Coffee Break
16.30-17.20 Paul Elbourne (Queen Mary's London) "Incomplete Descriptions and Sloppy Identity"
17.20-17.50 Discussion 


May 18th

9.45-10.35 Peter Ludlow (Northwestern University) “Cheap Contextualism and Meaning Underdetermination” 
10.35-11.05 Discussion
11.05-11.30 Coffee Break
11.30-12.20 Zoltan Gendler Szabo (Yale University) "Epistemic comparativism"
12.20-12.50 Discussion

Lunch Break

14.45-15.35 Dag Westerståhl (Gothenburg University) "What should be concluded from Lewis’s argument?"
15.35-16.05 Discussion
16.05-16.30 Coffee Break
16.30-17.20 Stephen Yablo (MIT) "Saying More With Less"
17.20-17.50 Discussion 


May 19th

9.45-10.35 François Récanati (IJN) "How Radical is Radical Contextualism?"
10.35-11.05 Discussion
11.05-11.30 Coffee Break
11.30-12.20 Peter Lasersohn (University of Illinois) "On the Possibility of Compositional Pragmatics"
12.20-12.50 Discussion

Lunch Break

14.45-15.35 Ruth Kempson (King's College, London) "The Challenge of Split Utterances" (with Eleni Gregoromichelaki)
15.35-16.05 Discussion
16.05-16.30 Coffee Break
16.30-17.20 Nicholas Asher (IRIT) "Context in Content Composition"
17.20-17.50 Discussion 


Nicholas Asher (IRIT)
"Context in Content Composition" 

One of the intriguing but not well-understood observations about the composition of meaning is that when word meanings are combined, the meaning of the result can differ from what standard compositional semantics has led us to expect.  In applying, for instance, a property term ordinarily denoting a property $P$ to an object term ordinarily denoting $a$, the content of the result sometimes involves a different  but related property $P'$ applied to an object $b$ that is related to but distinct from the original denotation of $a$.   While the choice of words obviously affects the content of a predication, the discourse context in which the predication occurs also affects it, and by discourse context I mean not only the predicational environment but also the discourse context to date. An important theme of current lexical and compositional semantics is how to make sense of this interaction. I give a tour of some examples and distinctions in this domain and provide a general framework for thinking about such meaning shifts.

Paul Elbourne (Queen Mary University of London)
"Incomplete descriptions and sloppy identity"

Through an examination of the binding possibilities displayed by downstressed continuations of donkey sentences, this article argues that incomplete definite descriptions are interpreted by means of situation variables.

Ruth Kempson (King's College, London)
"The Challenge of Split Utterances" (with Eleni Gregoromichelaki)

This talk will address the challenge of so-called split utterances, in which one party starts to present some structure that the other party takes on and extends.
(i) A: “Did you burn”..B: “the pan? Yes, the fat caught fire.” A: “But did you manage..” B: “not to burn myself? Fortunately, yes.” A: “That was” B “very lucky”. As (i) displays, shift from one party to another can go across any syntactic dependency whatsoever. These are omnipresent in conversational dialogue, entirely systematic, and a phenomenon which children join in on from a very early age. This seamless shift from speaker to hearer and vice versa is extremely hard to address from within standard grammar formalisms. Split utterances also pose a problem for pragmatic theories, which define communication as the grasping of some speaker’s meaning via a process of enrichment of sentence-meaning. The propositions created via split utterances may in some sense be interactively constructed, and not necessarily be the content of any one party’s intentions: indeed, the participants’ intentions may only emerge during the course of the exchange. I will argue that, with the shift in perspective defined in Dynamic Syntax, this phenomenon is predictable as an immediate consequence. Dynamic Syntax (Cann et al 2005) models the dynamics of how structural representations of content are incrementally established. But the core claim of the framework goes further than this. Natural-language grammars are defined in terms of time-linear growth of semantic representations with no other concept of syntactic structure, incorporating the concept of structural growth relative to context into the syntactic formalism itself. With this shift, which licenses incremental application of these procedures in parsing and production alike, there is a very natural basis for modelling split-utterance phenomena. The result is an integrated theory of ellipsis of which split utterances form a sub-part, in which context is defined as an evolving record of growth of structural representations. The intricate display of interaction between participants in such exchanges, as in other ellipses, is taken not as evidence of other-mind reading but as emergent from the fact that parsing and production mechanisms involve the same procedures. Accordingly, parsing and production decisions can be made incrementally relative to the individual’s own evolving context: application of jointly available mechanisms in the two ongoing activities yields the effect of sensitivity to the other party, without any need of hypothesis as to what the other party might have in mind. The resulting core concept of communication is contextualist in spirit but departs from standard assumptions, in neither invoking any concept of sentence-meaning, nor any concept of speaker’s meaning to be grasped. The underpinning theory of language-ability involves a concept of knowing-how, rather than one of knowing-that.

Angelika Kratzer (University of Massachusetts, Amherst)
"Syntactic constraints on meaning composition: deriving the meanings of conditionals"

The talk will start out by emphasizing that natural languages use different mechanisms of meaning composition, and that it's a matter of empirical investigation to find out which mechanisms are used when and what their properties are. For example, pragmatic enrichment is not just allowed, but required, for the interpretation of noun-noun compounds like "swan boat", but is much more constrained in phrasal syntax. What exactly is the difference? What kind of constraints does phrasal syntax place on meaning composition? I will not be able to give a conclusive answer, but will approach these questions with a case study built around a popular compositionality puzzle due to James Higginbotham:  What would it take to compute the meanings of conditionals like "Everyone will fail if he goofs off" and "Nobody will fail if he goofs off"  in a fully general way?

Peter Lasersohn (University of Illinois)
"On the Possibility of Compositional Pragmatics"

 I argue that compositionality (in the sense of homomorphicinterpretation) is compatible with radical and pervasive contextual effects on interpretation.  Apparent problems with this claim lose their force if we are careful in distinguishing the question of how a grammar assigns interpretations from the question of how people figure out which interpretations the grammar assigns.I demonstrate, using a simple example, that this latter task must sometimes be done not by computing a derivation defined directly by the grammar, but through the use of pragmatic background knowledge and extragrammatical reasoning, even when the grammar is designed to be fully compositional. The fact that people must sometimes use global pragmatic mechanisms to identify truth conditions therefore tells us nothing about whether the grammar assigns truth conditions compositionally.Compositional interpretation (or the lack thereof) is identifiable not by the mechanisms necessary to calculating truth conditions, but by the structural relation between the interpretation of a phrase in context and the interpretations of its parts in that same context.  Even if this relation varies by context, an invariant grammar is possible if grammars can “invoke” pragmatic concepts; but this does not imply that grammatical theory must explain these concepts or incorporate a theory of pragmatics.

Peter Ludlow (Northwestern University)
"Cheap Contextualism and Meaning Underdetermination"

The puzzle:  nothing is completely flat, 6 foot 6, etc. so how can utterances of sentences like ‘Kansas is flat’ and ‘Michael Jordan is 6 foot 6’ etc be true?
Some standard options include the following:
i)   Sentences (claims, propositions) like (1) are false, but assertion is warranted
ii)  contextualism: e.g. ‘flat’ means flat by contextually determined standards
iii) there is an implicit “roughly speaking” operator:  roughly-speaking[Kansas is flat]
In my paper “Cheap Contextualism” I argued for an alternative position in which.
i) word meanings are underdetermined and partially sharpened on a case by case basis
ii) semantic values are underdetermined (underdetermination is lifted into the metalanguage)
iii) the “absolutely flat”/”precisely 6 foot 6” meanings are not privileged, but just one sharpening among many
In this paper I go into details about the process by which word meanings are sharpened on a case by case basis and I pursue consequences of the view for formal semantic theory and topics such as vagueness.

Peter Pagin
(Stockholm University)
"Compositionality and Complexity"

Compositionality is usually justified by appeal to the requirement of computability: language users are able to work out the meanings of new sentences, so meaning must be computable. However, computability only implies that the semantics is recursive, which in turn does not entail that it is compositional. Are there reasons to think that natural language semantics is both recursive and compositional? One possible such reason would be that compositionality minimizes the computational complexity of semantic interpretation. In this talk I will investigate the relation between compositionality and complexity.

Francis Jeffry Pelletier (University of Alberta)
"Holism, Compositionality, and Context"

Just what the heck is holism (semantic, ontological and perhaps of some other type), and whatever does it have to do with context...whatever that is? Could it really be true that there is a kind of semantic holism that incorporates context and which is incompatible with compositionality?
François Recanati (Institut Jean-Nicod)
"How Radical is Radical Contextualism ?"

A number of distinct (though related) issues are raised in the debate over Contextualism. My aim in this paper is to disentangle them, so as to get a clearer view of the positions available (where a ‘position’ consists of a particular take on each of the relevant issues simultaneously). The issues I will talk about are :
- the modularity issue (is semantic competence sufficient to assign truth-conditions to arbitrary sentences of one’s language, or is pragmatic competence also needed ?)
- the extent of context-sensitivity issue (is context-sensitivity pervasive in natural language, or it is a rather restricted phenomenon ?)
- the generalization of context-sensitivity issue (are all/most expressions similar to indexicals ?)

- the pragmatic modulation issue (is the semantic contribution of expressions calibrated through the operation of ‘free’ pragmatic processes like metonymy,  narrowing or sense-extension?)
On each issue there are two sides, corresponding to the overall debate between Contextualism and Literalism. On the first two issues (the modularity issue, and the extent of context-sensitivity issue) Contextualism is the dominant position. With respect to the other two issues, it is the other way round. I myself stand on the contextualist side on all four issues. The resulting position is a fairly radical form of Contextualism, but I will argue that, contrary to what its critics often say, that position is compatible with the project of building a compositional semantics for natural language.

Zoltan Gendler Szabo (Yale Unviversity)
"Epistemic comparativism"

Epistemic contextualism is the view that knowledge ascriptions are context-sensitive due to the semantics of "know". The view has been criticized on three main grounds: that the standard examples that are supposed to provide empirical support for it (such as Cohen's airport case and DeRose's bank case) are unconvincing, that the semantic analogies for the alleged context-sensitivity of "know" (such as "I" or "flat") are far-fetched, and that the view has not been backed up by an explicit semantic proposal. All three lines of criticism are legitimate. This paper will present new examples (whose credentials have withstood genuine testing), a new analogy (adverbial quantifiers, in particular "always"), and a detailed semantics. The core idea is that knowledge-ascriptions claim that the knower has a true answer to the question under discussion that is in good epistemic standing. The work is part of a joint project with Jonathan Schaffer. 

Dag Westerståhl (Gothenburg University)
"What should be concluded from Lewis’s argument? "

David Lewis’s 1980 paper ‘Index, context, and content’ contains a simple argument against using propositions that only depend on worlds as semantic values: they would not, he claims, be compositional. The argument has been discussed by several authors, e.g. Stanley (1997), King (2003), Recanati (2007), and Cappelen & Hawthorne (2009). (See also the related discussion in Richard (1982), Salmon (1986, 1989), and Dummett (1991).) But there is no unanimity about what the argument shows. I take another look, which leads me to some general observations on semantic values, logical form, and notions of compositionality.

Stephen Yablo
"Saying More With Less"

We explore three strategies for wringing distinct assertive contents C* out of a sentence's compositional content C. Thinking of M as the subject matter under discussion and P as presupposed, the strategies are
(1) C* is the part of C that's about M.
(2) C* is the part of C that is not at all about the subject matter #whether P#. (This might be understood as C minus P.)
(3) C* is what must be added to P to reach C. (This is C minus P in a vector-addition sense.)
The third option is the most general because the C* it generates may be logically independent of C. All three are explained using a version of David Lewis's notion of subject matter in "Statements Partly About Observation."