Truth Conditional Discourse Semantics for Parentheticals

It has been often argued that parentheticals, discourse adverbials and certain parts of speech like interjections do not contribute to the truth conditional content of the assertions of which they are part. In this paper I argue that many of these constructions do contribute a truth conditional content, and I propose a semantics for parentheticals and discourse adverbials that treats these constructions similarly to SDRT's treatment of presuppositions. I also point out differences between standard presupposition triggers on the one hand and parentheticals or discourse adverbials on the other. 1 I N T R O D U C T I O N There is a tradition in pragmatics going back at least to Grice according to which certain constructions and parts of speech do not contribute to the truth conditional content of the assertions of which they are part. Rather they implicate or indicate either a particular speech act or an attitude of the speaker. Examples of such items are: 9 mood indicators—questions, commands. • interjections—Oh, Gee, Too bad, Damn, etc. a so called discourse adverbials—allegedly, unfortunately, etc. This category also includes adverbial clauses—e.g. as Mary assures us. « so called pragmatic conditionals—if you know what I mean, if you see what I'm getting at. © discourse particles—tte in Japanese or re in Sissala for hearsay. • discourse connectors—but, too, hence, so, therefore, etc. « parenthetical constructions, in which full clauses missing a verbal complement occur. Wilson (1975) and others have argued that all of these phenomena exhibit a similar behavior relative to a test for non-truth conditional meaning—the 'embedding test'. Sperber & Wilson (1995) also claim that they can give a unified analysis of these phenomena. However, I will argue here that the test does not really separate out parts of speech with a non-truth conditional meaning (whatever that might be is not my concern here). Certainly, there is reason to doubt that all of these constructions fail to be amenable to truth conditional or, more generally, model theoretic analysis. 32 Truth Conditional Discourse Semantics for Parentheticals Some mood indicators have received detailed and rigorous model theoretic analysis in e.g. Hintikka (i 974) and Groenendijk & Stokhof (1984). Others have argued that discourse connectors have an important though sometimes subtle effect on the truth conditional interpretation of discourse (Asher 1993; Lascarides & Asher 1993) and that hence traditional pragmatics gives a misleading picture of discourse interpretation by separating out the contribution of discourse connectors from an account of truth conditional content. Similar remarks apply to the Japanese discourse particle tte (Hasegawa 1996). In this paper I will examine parentheticals and discourse adverbials. Some examples of parentheticals are given in (1) below (the parentheticals are underlined). I will argue that these parts of speech also have a straightforward truth conditional semantics in a theory of discourse interpretation that takes account of discourse structure. The theory of discourse interpretation that I will use is SDRT (Asher 1993, Lascarides & Asher 1993), an extension of DRT that incorporates an account of discourse structure and rhetorical function. (1) a. The party is over, I hear. b. Please leave, I beg you. c. The party, Mary assures us, is over. 2 A TEST FOR N O N TRUTH CONDITIONAL MEANING? According to Wilson (1975), there is a test for non truth conditional meaning: embed the questionable item into the antecedent of a conditional and see if the purported truth conditional contributor's meaning falls within the scope of 'if. If it does, it is truth conditional; and if not, not. Here are some examples of the test at work: (2) a. If the party, unfortunately, is over, then we should find somewhere else to get a drink. If it is unfortunate that the party is over, then we should find somewhere else to get a drink b. If the sun is shining but it's midnight, then we must be in Norway. If the sun is shining and it's midnight and that's not expected, then we must be in Norway. c. If, I'm warning you, you cross that line, I'll hit you. If I'm warning you that you cross that line, I'll hit you. According to the test, these examples appear to indicate that neither