I take the goal of contemporary semantics-pragmatics to be to provide at least a framework for a theory of cognition that explains how meaning is ascribed in the production and comprehension of utterances. I also take it that such a theory should accommodate two salient facts: that in the right circumstances, an utterance of any (constituent of any) sentence could mean anything; but that production and understanding of meaningful utterances asymmetrically depends on a compositional (or computable) semantics. These two facts normally lead to a two-theory view according to which a set of truth conditions is derived via a set of compositional rules, while (broadly Gricean) pragmatic principles account for divergences from this semantic interpretation. Here I outline one alternative to the two-theory view, the integrated view, according to which knowledge of compositional semantic properties feeds into the computation of a semantic interpretation that is also influenced by ‘Gricean’ pragmatics. I indicate how this integrated view is motivated by diverse ‘embedded pragmatic’ phenomena, as well as diverse experimental results from our own lab and elsewhere. I also highlight similarities and differences with other Contextualist (e.g. Relevance Theory) views. In the second part of the talk, I discuss what is required to build testable predictions from each of these views and sketch four models (two based on each view). I consider further experimental evidence that bears on the predictions of these models. To date, experimental support can be found for both views but I will argue that there is an emerging pattern in the experimental results that better supports an integrated theory.
Game theory offers well-understood formal models of human interaction and decision making. If used properly in linguistics, game theoretic models may fuel a formal approach to the otherwise somewhat murky, but central notion of "language use." However, although game theoretic models have been applied sporadically to linguistic concerns at least since the late 80's, it is still unclear how exactly ideas and notions of game theory are to be made applicable within linguistics. This concerns both the choice of the game model as well as the choice of the solution concept. On a methodological level, this talk tries to contribute to these questions, with a strong plea for non-equilibrium solution concepts as favored by recent epistemic approaches within game theory.
To make these methodological issues more concrete, the talk focuses on an application of game theory to formal pragmatics. I show how a version of the iterated best response model -variations of which have been proposed by, among others, Benz, van Rooij, Jäger and Stalnaker- provides an epistemic grounding of (iterated) exhaustive interpretation. Contrary to previously held convictions, exhaustive interpretation should be conceived /not/ as interpretation based on
the assumption that the speaker follows Gricean principles of Quantity (and the like), but as rational interpretation given the semantic meaning of alternative messages. It is in this exemplary way that I wish to lend credit to the claim that game theoretic approaches not only reiterate formally what has been put forward in informal terms before, but that the more prickly formulation of game theoretic approaches leads to genuine new insights and a truly deeper understanding of fundamental questions of language use.
[The epistemic grounding of exhaustive interpretation addressed here has no bearing, is in fact orthogonal to, the question of whether exhaustivity operators may apply within the scope of other logical operators.]
I plan to talk about the mechanism that is at work while interpreting metaphors and argue that the same mechanism is at work while interpreting other word combinations. I will discuss the results of a pilot experiment on the interpretation of coerced nouns (stone lion) and conflicting adjective noun combinations (halve appel 'half an apple' (is not round)). I will go into the relation with synchronic change in which metaphors and semantic weakening in general play a big role.
Modulated meanings and lexical pragmatics have been at the center of recent inquiry. I will show how, using my framework of equilibrium linguistics, it is possible to derive such meanings from first principles. In so doing, I will argue that what semantics and pragmatics have in common - the ubiquity of context - outweighs their differences and so the two should be unified into a single field.
This paper proposes a way to accommodate the phenomenon of embedded pragmatic effects, such as local enrichment, within a broadly Gricean conception of communication. On the standard Gricean conception, interpreters are taken to reason about a speaker’s communicative intentions partly on the basis of what the speaker said. I propose that embedded pragmatic effects arise as a result of reasoning about what a speaker intended with respect to the conveyed content of an embedded clause, partly on the basis of the conventional content of that clause. As I will show, the relevant reasoning may be prompted by global considerations of felicity/cooperativity, but nonetheless result in modification of embeeded content. I propose further that this conception is fairly straightforward to implement in frameworks such as DRT or theories of structured propositions, in which semantic contents are taken to be structured entities, and subordinate clauses are represented by sub-structures. Local pragmatics can then be understood as a process of modifying a sub-structure in accord with the inferred intention of the speaker.
Like NPIs, PPIs can be divided into different classes, which have been characterized in terms of which kinds of operators 'anti-license' them. For instance, while a PPI such as 'someone' is (said to be) antilicensed only by anti-additive operators, PPI adverbials such as 'relatively' and 'almost' have been said to be anti-licensed by all downward-entailing operators. There is however another important distinction to be made between various classes of PPIs which has not been noticed so far: for some PPIs, anti-licensing is strictly local, in the sense that the relevant PPIs is antilicensed only if it is in the immediate scope of the anti-licenser (for instance negation), while others are anti-licensed as soon their syntactic environment has a certain *global* semantic property (for instance, downward-entailingness). For instance, while both French simple disjunctions ('ou') and complex disjunctions ('ou...ou', 'soit...soit') are anti-licensed by negation, simple disjunctions can be interpreted in the scope of 'je doute que' or 'sans que', but complex disjunctions cannot
PPIs behaving like simple disjunctions include, e.g., singular indefinites (quelqu'un, un NP, etc.), some aspectual adverbials (already, déjà), and PPIs behaving like complex disjunctions include, e.g, approximators (à peu près, presque, more or less at least, almost), qualifiers such as 'relatively', and many others.
In this talk, I will offer a theory of the 'soit...soit' class (whose members I dub 'global PPIs', as opposed to 'local PPIs) based on the following idea: global PPIs have the distributional properties they have because they necessarily co-occur with an exhaustivity operator. It turns out that well motivated constraints regarding the distribution of the exhaustivity operator can then be shown to correctly predict the distribution of global PPIs. Importantly, this approach predicts a direct correlation between the obligatory presence of certain scalar implicatures, exhaustivity and anti-exhaustivity effects and PPI-like behavior.