Vocatives, Topics, and Imperatives∗Paul PortnerGeorgetown UniversityJuly 16, 2004IMS Workshop on Information Structure, Bad Teinach1 Preview and Background• VocativesThe centerpiece of this presentation is the semantics and pragmatics of vocatives. Vocativesare relevant to information structure bec ause:1. They are indexical, involving reference to the addressee.2. Their contribution to the meaning of the sentence does not seem to be truth-conditionalin nature.3. They are very similar to topics both syntactically and pragmatically.• Summary of Main Hypotheses1. The addressee is represented in its own syntactic projection, as revealed by imperativesand vocatives.2. Vocatives are separate performatives, and can be analyzed as expressive meaning (in thesense of Potts 2003a).3. Topics – similar as they are to vocatives – should be thought of as separate performativesas well.Point 3 is obviously controversial, and may be even more so in that it allows the hypothesisthat the cognitively relevant structuring of information should pushed outside the boundariesof linguistic theory.• Clause TypingThis work is part of a broader project on clause typing undertaken collaboratively withRaffaella Zanuttini. Our work on clause types begins from the following fundamental points:∗This is part of a project funded by NSF grant BCS-0234278, ‘Clause Types: Form and Force in GrammaticalTheory’, to Paul Portner and Raffaella Zanuttini. Miok Pak, and Simon Mauck, our collaborators on the project,have made many contributions.11. Clause types are to be identified by three criteria (Sadock & Zwicky 1985):(a) They form a closed system.(b) T hey are associated with a particular sentential force.(c) The major clause types – declaratives, interrogatives, and imperatives – are uni-versal.2. There are significant typological generalizations concerning clause type systems. In ad-dition to the universality of the major types, we have some types that are com mon butnot universal (exclamatives), some that are rare (promissives), and some types that,logically speaking, might have existed but apparently do not (“threatatives”, “warna-tives”). These generalizations cannot be explained in functional terms, and insteadrequire a grammatical account.3. Each major clause type is associated with a component of the discourse context, andthese components are sets of semantic values, differentiated by type:(a) Declaratives are associated with the Common Ground, a set of propositions.(b) Interrogatives are associated with a Question Set, a s et of sets of propositions.(c) Imperatives are associated with the addressee’s To-do List. A To-do List is aset of prope rties, and a To-do List function assigns one to each participant in theconversation.(Stalnaker 1974, 1978, Ginzburg 1995a,b, Roberts 1996, 2004, Han 1998, Portner &Zanuttini 2002, Potts 2003b, Portner 2004, among others.)4. Force is not syntactically represented; rather a clause’s force is inferred from its semantics(cf. Zanuttini & Portner 2000, Zanuttini & Portner 2003 on exclamatives).5. The forces associated with major types – assertion, asking, and requiring – are simplythe addition of the meaning of a clause to the type-matching component. For example:Context + It is raining = CG ∪ { [[ It is raining. ]]w,c}6. Note: We must distinguish sentential force (Chierchia & McConnell-Ginet 1990, orsentence mood, Reis 1999) from the ultimate illocutionary force.2 Imperatives• The Imperative SubjectIn many languages, the subject of an imperative must refer to, overlap with, or quantify overthe addressee or the set of addressees (Mauck et al. 2004). (There are exceptions; somelanguages have true third-person imperatives. See (8) below and Mauck et al. (2004) fordetails.)• This restriction does not mean that the s ubject is simply a null or overt second person pronoun(cf. Potsdam 1996), and is not pragmatic in nature:(1) a. Eat!b. Boys go in, girls wait outside!c. Nobody touch your pencils!d. Someone come up to the blackboard and do this problem!(2) a. (You) be kissed by Mary!2b. *Mary kiss (you)!• We propose that the addressee is represe nted in the syntax of imperatives, but not as thecanonical subject (cf. Platzack & Rosengren 1994). Suppose that the addressee is representedas a syntactic head, addr.(3) addrPQQaddr IPMore evidence for the idea that the addressee is represented in a syntactic projection of itsown will come from vocatives in Section 3.• The imperative IP denotes a property constructed by abstracting over the subject NP (it’sa variable) or a domain variable of the subject NP. (The syntactic details are not relevanthere.)• The overall me aning of the imperative addrP is a prop e rty, spe cifically the IP property withits argument restricted to the (possibly singleton) set of addressees (cf. Hausser 1980):(4) a. [[ Eat! ]]w,c= [λwλx : x ∈ addr(c) . x eats in w]b. [[ Everyone eat! ]]w,c= [λwλx : x ∈ addr(c) . [∀y : y ∈ x . y eats in w]]• Force Assignment in Imperatives1. The property in (4) is added to the addressee’s To-do list (or addressees’ To-do lists).Because of the restriction to the addressee, it is a property that only the addressee canhave, and so it would be infelicitous to add it to any other individual’s To-do list. (Theaddressee restriction also lets us distinguish imperatives from promissives, a rare butattested clause type, cf. Pak 2004 on Korean.)2. Distinctions among subtypes of imperatives – orders, requests, permissions, etc. – shouldnot be understood at the level of conversational update. They all add a property tothe addressee’s To-do list. Rather, these differences have to do with the pragmatic orsociolinguistic basis for the speaker’s attempt to add a property to the addressee ’s To-dolist.(a) Orders occur when the basis is social authority.(b) Requests occur when no social authority is invoked, and the basis is the speaker’sor addressee’s benefit.(c) Permissions occur when the property is understood to be one the addressee herselfwants on her To-do list.3. The set of addressees can include groups. In that case, it’s a pragmatic matter how thework of making the imperative property true is distributed among group members. Ifthe property is distributive, the corresponding non-group property will be added to eachindividual’s To-do list. If it’s not, what happens is not linguistics specifiable.• Some compositional details:1. [[ addr ]]w,c= [λwλx