Dynamic Semantics

The term dynamic interpretation refers to a number of ap-proaches in formal semantics of natural language that arose in the 1980s and that distinguish themselves from the preceding paradigm by viewing interpretation as an inherently dynamic concept. The phrase dynamic semantics is used to denote a specific implementation of this idea, which locates the dynamic aspect in the concept of linguistic meaning proper.

The dominant view on meaning from the origins of logically oriented semantics at the beginning of the twentieth century until well into the 1980s is aptly summarized in the slogan "Meaning equals truth conditions." This formulates a static view on what MEANING is: it characterizes the meaning relation between sentences and the world as a descriptive relation, which is static in the sense that, although the meaning relation itself may change over time, it does not bring about a change itself. The slogan focuses on sentences, but derivatively the same holds for subsentential expressions: their meanings consist in the contribution they make to the truth conditions of sentences, a contribution that is usually formalized in terms of a static relation of reference. Interpretation is the recovery of the meaning of an utterance, and is essentially sentence-based. This static view on meaning and interpretation derives from the development of formal LOGIC, and lies at the basis of the framework of Montague grammar, the first attempt to apply systematically formal semantics to natural language.

Although dominant, the static view did not go unchallenged. The development of speech act theory (Austin, Searle) and work on PRESUPPOSITION (Stalnaker) and IMPLICATURE (GRICE) stressed the dynamic nature of interpretation. However, at first this just led to a division of labor between SEMANTICS and PRAGMATICS, the latter being viewed as something that works on top of the results of the former. This situation began to change in the beginning of the 1980s when people started to realize that certain empirical problems could be solved only by viewing meaning as an integrated notion that accounts for the dynamic aspects of interpretation right from the start and that is essentially concerned with DISCOURSE (or texts), and not with sentences.

A simple but illustrative example is provided by cross-sentential ANAPHORA. In a discourse such as "A man walked into the bar. He was wearing a black velvet hat," the pronoun "he" is naturally interpreted as bound by the indefinite noun phrase "a man." If interpretation proceeds on a sentence-by-sentence basis, this can not be accounted for. And "delayed" interpretation, that is, linking quantified noun phrases and pronouns only when the discourse is finished, makes empirically wrong predictions in other cases, such as: "One man was sitting in the bar. He was wearing a black velvet hat." Such examples rather suggest that interpretation has to be viewed as a dynamic process, which takes place incrementally as a discourse or text proceeds.

Further development of this idea received both an internal and an external stimulus. The main external influence came from natural language research within the context of artificial intelligence, which favored a definitely procedural view and was oriented toward units larger than sentences. The interpretation of utterances is modeled as the execution of procedures that change the state of a system as it proceeds. It took some time before this idea caught on, mainly because it seemed hard to reconcile with the core goal of formal semantics, viz., to account for logical relationships. However, the emergence of formal models within the AI paradigm, in particular the development of NONMONOTONIC LOGICS, provided the necessary link. Also, work on the semantics of programming languages turned out to be concerned with a conceptual machinery that could be applied successfully to natural language.

In the beginning of the 1980s the dynamic view on interpretation was formulated explicitly in discourse representation theory (Kamp 1981; see also Kamp and Reyle 1993) and file change semantics (Heim 1982). The work of Kamp and Heim constitutes an extension and transformation of the framework of Montague grammar. In his original paper Kamp explicitly describes his theory as an attempt to wed the static approach of the logical tradition to the procedural view of the AI paradigm. Within different settings similar ideas developed, for example within the theory of semantic syntax (Seuren 1985), and that of game theoretical semantics (Hintikka 1983).

Discourse representation theory is a dynamic theory of interpretation, not of meaning. The dynamics is located in the process of building up representational structures, so-called discourse representations. These structures are initiated by incoming utterances and added to or modified by subsequent utterances. The structures themselves are interpreted in a static way by evaluating them with respect to a suitable model. For example, anaphoric relations across sentence boundaries are analyzed as follows. A sentence containing a referential expression (such as a proper name, or a quantified term; see QUANTIFIERS) introduces a so-called discourse referent along with restrictions on its interpretation. A subsequent sentence containing an anaphoric expression (such as a pronoun) can "pick up" this referent if certain descriptive and structural conditions are met, and thus be linked to the antecedent referential expression. The semantics of the discourse representation then takes care of the coreference.

Dynamic semantics (Groenendijk and Stokhof 1991; Groenendijk, Stokhof, and Veltman 1996) takes the idea of dynamic interpretation one step further and locates the dynamics in the concept of meaning itself. The basic starting point of dynamic semantics can be formulated in a slogan: "Meaning is context-change potential." In other words, the meaning of a sentence is the change that an utterance of it brings about. And the meanings of subsentential expressions consist in their contribution to the context-change potential of the sentences in which they occur. Unlike discourse representation theory, it tries to do away with semantic representations but assigns various expressions, such as the existential quantifier associated with indefinite noun phrases, a dynamic meaning, which allows it to extend its binding force beyond its ordinary syntactic scope.

The slogan "Meaning is context-change potential" is general in at least two respects: it does not tell us what it is that is changed, and it does not say how the change is brought about. The latter question is answered by giving analyses of concrete linguistic structures. As to the former issue, it is commonly assumed that one of the primary functions of language use is that of information exchange and that, hence, information is what is changed by an utterance. Primary focus is the information state of the hearer, but in dialogical situations that of the speaker also has to be taken into account. Depending on the empirical domain, information concerns different kinds of entities. If the subject is anaphoric relations, information is about entities which are introduced and their properties; for temporal expressions one needs information about events and their location on a time axis; in the case of default reasoning expectation patterns become relevant. In other cases (question-answer dialogues, presuppositions) "higher order" information of the speech participants about each other is also at stake.

A change in the notion of meaning brings along a change in other semantic concepts, such as entailment. In static semantics truth plays a key role in defining meaning and entailment. In dynamic semantics it becomes a limit case. The central notion here is that of support: roughly, an information state s supports a sentence F iff an utterance of F does not bring about a change in s. Entailment can then be defined as follows (alternative definitions are possible as well): F ₁ . . . F _n entails Y iff for every state s it holds that updating s with F ₁ . . . F _n consecutively leads to a state that supports Y. (Cf. van Benthem 1996 for discussion of various alternatives.)

Dynamic semantics of natural language can be seen as part of a larger enterprise: the study of how information in general is structured and exchanged. Such a study brings together results from diverse fields such as computer science, cognitive psychology, logic, linguistics, and artificial intelligence. Language is one particular means to structure and exchange information, along with others such as visual representations, databases, and so on. The dynamic viewpoint has considerable merit here, and, conversely, draws on results that have been developed with an eye to other applications.

Additional links

-- Martin Stokhof and Jeroen Groenendijk

References

Benthem, J. F. A. K. van. (1996). Exploring Logical Dynamics. Stanford: CSLI .

Groenendijk, J. A. G., and M. J. B. Stokhof. (1991). Dynamic Predicate Logic. Linguistics and Philosophy 14:39-100.

Groenendijk, J. A. G., M. J. B. Stokhof, and F. J. M. M. Veltman. (1996). Coreference and modality. In S. Lappin, Ed., Handbook of Contemporary Semantic Theory. Oxford: Blackwell, pp. 179-213.

Heim I. (1982). The Semantics of Definite and Indefinite Noun Phrases. Ph.D. diss., University of Massachusetts. (Published in 1989 by Garland, New York.)

Hintikka, J. (1983). The Game of Language. Dordrecht: Reidel.

Kamp, J. A. W. (1981). A theory of truth and semantic representation. In J. A. G. Groenendijk, T. M. V. Janssen, and M. J. B. Stokhof, Eds., Formal Methods in the Study of Language. Amsterdam: Mathematical Centre, pp. 277-322.

Kamp, J. A. W., and U. Reyle. (1993). From Discourse to Logic. Dordrecht: Kluwer.

Seuren, P. A. M. (1985). Discourse Semantics. Oxford: Blackwell.

Dynamic Semantics

See also

Additional links

References

Further Readings