Complement anaphora and the plurality of worlds

In a nutshell We present novel data from German revealing an striking similarity in the behavior of complement anaphora (CA) in different domains (possible worlds, times, individuals): We argue that the distribution of CA in the domain of individuals requires the antecedent quantifier to provide a plurality. We show that the correlation of quantifier types and the availability of CA with conditional structures parallels what we find in the individual domain. From these parallels, we deduce that the latter type of construction must also provide a plurality (of times, worlds) that can be picked up by a pronominal element. This provides further support for the referential analysis of conditionals, [14], [1]. Complement anaphora and pluralities in the domain of individuals CA ([11], [12] a.o.) is the phenomenon when a plural pronoun following a quantified sentences D(P)(Q), picks up a referent corresponding to the set P ∩ Q̄ (‘CompSet’), rather than the sets P ∩Q (‘RefSet’) or P (‘MaxSet’). In German, CA is restricted in two respects. First, not any determiner licenses CA (cf. [11] for English): If die meisten (‘most’) is chosen in (1), sie (‘they’) cannot pick up the CompSet – in this case, (1) can be followed coherently by (2), but not by (3). If nicht alle (‘not all’) is chosen, sie shows a strong tendency to pick up the CompSet – in this case, continuation of (1) by (3) is coherent, but continuation by (2) isn’t. (1) {die meisten/nicht alle} the most/not all [P Kinder] children [Q haben have ihr their Eis ice-cream gegessen]. eaten ‘Most / not all (of the) children ate their ice-cream. Adapted from [11] (2) Sie fanden es gut. ‘They liked it.’ √ die meisten, # nicht alle (3) Sie haben es weggeschmissen. ‘They threw it away.’ # die meisten, √ nicht alle The second restriction is that all determiners that license CA select for plural NPs: If D = jed(‘every’) with a singular NP as in (4), the plural anaphoric pronoun can only pick up MaxSet– both the continuation in (2) as well as that in (3) yield an incoherent discourse. (4) nicht not jedes every [P Kind] child [Q hat has sein its Eis ice-cream gegessen]. eaten Analysis of complement anaphora for individuals. We make three main assumptions related to the interpretation of plural quantifiers. First, we suggest that quantifiers should be explicated by using the RefSet (R) and the CompSet (C) as schematically suggested in (5). The conservativity of quantifiers ([6]) guarantees that this can be done in all cases. We exemplify this for all,some and most in (5a-c). Second, we assume that in constructions such as alle NP the plural feature is interpreted as at the level of the quantifying determiner and not at the restrictor level. I.e. alle does not range over pluralities but over atoms. Third, we assume that in any dynamic semantic framework such as DRT, DPL, FCS etc. (cf. e.g. [4], [5], [3]) the R and C are modelled as maximal pluralities iff the quantifier carries a plural feature, and as such constitute valid anaphoric antecedents for plural pronouns such as sie. (5) D(P )(Q) is true iff RelD(R,C), where R = P ∩Q and C = P \R σ( ⊕ R) and σ( ⊕ C) are discourse referents a. all(P )(Q) is true iff R = R ∪ C b. some(P )(Q) is true iff |R| ≥ 1 c. most(P )(Q) is true iff |R| > |C| With these assumptions we can account for the data in (1)–(4) above. First, we predict that