This paper takes the universal force of Free Choice Items (henceforth FCI) to be an implicature that arises when an assertion with existential force interacts with a set of lexically triggered exhaustified alternatives, as proposed by Chierchia (to appear). It departs from him in capturing the distribution of FCI through a Viability Constraint on alternatives, adapting the notion of Fluctuation in Dayal 2009 to the view of FCI as existential. A substantive consequence of the proposed Viability Constraint is that FCI can only ever be licensed if they take wide scope over modals (of the appropriate sort). The paper establishes that apparent counterexamples involve non-trivial interactions between FCI and modality. When the FCI is complex it can participate in splitquantification, with one part taking scope over the modal and the other taking scope under it. Alternatively, if there are two modals, FCI can be scopally lower than one but still be licensed because it has scope over the other. A striking confirmation of this approach comes from imperatives. Though long thought to unconditionally admit FCI, it is shown here that this is not so. In addition to accounting for the more nuanced empirical generalization regarding imperatives, the Viability Constraint is argued to be generally a simpler way of predicting the distribution of FCI than the one proposed by Chierchia.
SECTION 1: THE DISTRIBUTION OF FCI
The Distribution of ∀ FCIThe distribution of FCI has been much discussed (Vendler 1967, LeGrand 1975, Carlson 1980, 1981, Kadmon and Landman 1993, Dayal 1995, Giannakidou 1998, 2001, Horn 2000a, 2000b, 2003, Jayez and Tovena 2005, Chierchia 2006, to appear, among others). I will nevertheless begin by laying out the core facts in order to make the discussion here self-contained. I adopt the standard distinction between FCI any and NPI any in English, a powerful argument for which comes from cross-linguistic considerations. Italian qualunque/qualsiasi, for example, is acceptable in modal but not negative contexts, suggesting that items appearing in those contexts, while related, must be distinct: 1 1a. * Bill read any book.b. Bill didn't read any book. c. Bill can read any book.2a. *Ieri Gianni ha letto qualsiasi libro/qualsiasi dei libri "Yesterday Gianni read any FCI book/any FCI of the books." b. *Gianni non ha letto qualsiasi libro "Gianni didn't read any FCI book." c. Puoi leggere qualunque libro "You can read any FCI book."