7 February 1997 On the Semantics and Pragmatics of Information Packaging Herman Hendriks Institute for Language and Speech, Utrecht In his work on information packaging---i.e., the structuring of propositional content in function of the speaker's assumptions about the hearer's information state---Vallduví (1992, 1994) identifies the informational primitives focus, ground, link and tail, which are adapted from the traditional pragmatic focus/ground and topic/comment partitions. He concludes that the exploitation of information states of hearers by the information-packaging strategies of speakers reveals that these states have at least the internal structure of the file card systems of Heim (1982, 1983), since links and tails (which make up the ground) have an ushering role with respect to the information expressed by the focus: links, which correspond to what are traditionally known as topics and which are typically marked by L+H* pitch accent in English, say where---on what file card---the focal information goes, and tails indicate how it fits there. Similar 'locational' theories of information packaging have been proposed by Reinhart (1982) and Erteschik-Shir (1997). This conclusion will be challenged in the present talk. Observe, first, that it begs the question: if file card systems are assumed, then the information-packaging strategies do seem to contribute to efficient information exchange. However, the relevant question is whether this assumption itself is justified: ushers are very useful, but there are also halls that have unnumbered seats. Besides, it can be argued that locational theories of information packaging are problematic for various reasons. First, it is unclear what is the locus of update for quantified, negative and disjunctive links (or quantified, negative and disjunctive information in general---see also Heim 1983). Second, the existence of multiple links is enigmatic. Third, the replacement triggered by tails is complicated by the use of file cards. Fourth, the approach leads to the counterintuitive conclusion that pronouns form part of the focus. In view of these considerations it seems preferable to model information states by means of the discourse representation structures of Kamp (1981) and Kamp and Reyle (1993), which are ontologically less committed than the 'dimensionally richer' file card systems, in that discourse representation structures do not come with locations. This last aspect requires a different perspective on the function of links---for what purpose do links serve if they do not specify a locus of update? I will hypothesize the following: linkhood---and hence L+H* pitch accent in English---serves to signal non-monotonic anaphora: the discourse referent Y of a link is anaphoric to an antecedent discourse marker X such that X is not a subset of Y. This idea, which is consistent with and can actually be considered a partial execution---viz., for L+H* pitch accent---of the intonational-informational research program outlined in Pierrehumbert and Hirschberg (1990), turns out to subsume a wide range of phenomena. It sheds light on so-called 'non-identity' anaphora (Van Deemter 1992, 1994), contrastive (Rooth 1992) as well as correctional (Pierrehumbert and Hirschberg 1990) stress, the effect of pitch accenting on pronoun referent resolution (Cahn 1995, Kameyama 1994, Vallduví 1994), and the distinction between non-restrictive and restrictive relatives and adjectives (Kamp and Reyle 1993). And, finally, it contributes to an account of the phenomenon known as 'focus sensitivity': Rooth (1985) and other semanticists have claimed that the quantificational structure of expressions like only and even is crucially determined by the traditional pragmatic focus-ground partition. However, there are also cases of focus insensitivity where only and even associate with 'second occurrence focus', and with links. I will argue that focus sensitivity is an instance of anaphora, thus accommodating the contextual restriction that has been noted in the literature. This accounts for the known cases of association with second occurrence focus. In addition, it can be noted that, due to the lexical semantics of only and even, this contextual restriction is non-trival iff the anaphora is non-monotonic. Given this, we can handle the phenomenon of association with focus, but also the attested cases of association with links: since they signal non-monotonic anaphora, links always come with their own suitable, non-trivial contextual restriction. References
|