APLICACIÓN TIPOS DE ESCENARIOS

Hinrichs (1985), a dissertation supervised by Dowty, assumes a Davidsonian event- based semantics. In order to avoid the minimal-parts problem, Hinrichs relaxes the requirement thatfor-adverbials place on their predicates. Hinrichs’ proposal is complicated by the fact that he adopts Carlson (1977)’s three-tiered ontology of kinds, objects, and stages, which Carlson proposed in order to account for the

properties of bare plurals and generic sentences. Abstracting away from this and other complexities, here is howJohn slept for an hourcomes out on his proposal: (5) _JJohn slept for an hour_K

=∃e∃l[hour(l) ∧ l≤τ(e) ∧ sleep(j)(e)]∧ ∀l0_[l0 _{< l→ ∃e}0_[e0 _{< e ∧ l}0 _≤τ(e0_{) ∧}

sleep(j)(e0)]] Hinrichs comments on his translation as follows (p. 235):

The translation in [(5)] requires that . . . there has to be a spatio- temporal locationlwith the property of being one hour long such that the entire process of John’s sleeping spatio-temporally containsland for each proper sublocation l0 ofl there has to be a proper subpro- cess of John’s sleeping containingl0. . . . [W]e don’t require that each

sublocation denoted by the temporal measurement phrase has tobe

a subprocess itself. The requirement that each proper sublocation be contained in a proper subprocess of the maximal process making up the event, rules out that for each sublocation we could simply pick the maximal process itself.

The way Hinrichs avoids the minimal-parts problem does not follow Dowty’s suggestion of quantifying over less than literally all subintervals. Instead of requiring that a predicate like John sleep be true at each or most subintervals, Hinrichs requires that the runtime of every subinterval of John’s sleeping must be contained in(and not necessarily equal to) that of a sleeping event. This sleeping event must be a proper part of the event that the sentence describes. This is possible becausesleepis not quantized; if we assume that all telic predicates are quantized, they are correctly ruled out. The assumption that all telic predicates are quantized is not available to us because it is incompatible with the assumption of lexical cumulativity (see Section 2.7.2). But Hinrichs (1985) does not assume lexical cumulativity.

Krifka (1986) (p. 150) criticizes Hinrichs’ approach as arbitrary and notes that the proper-part requirement does not explain anything, since it does not serve any purpose other than that of excluding telic predicates. There is also another problem for Hinrichs’ account, which causes it to run into something very similar to the minimal-parts problem. The variablel0in (5) ranges over subintervals rather than (just) over instants. On the assumption that time is dense, it ranges over subintervals that are only minimally shorter thanl. For example, it ranges over intervals of length 58, 59, 59½. . . minutes. For each of these intervals, the definition requires there to be a proper part of the event of John’s sleeping which lasts at least as long as the interval and which qualifies as John’s sleeping. This gives the events

in the denotation of atelic predicates a dense structure. Moreover, depending on the structure of the underlying mereology, there will generally be a unique complement that results from removing the smaller sleeping event from the larger sleeping event. In classical extensional mereology (CEM), this is a consequence of Unique Separation (see Section 2.3.3). Depending on how large a sleeping event one separates off, the runtime of its complement may become arbitrary small. It follows on Hinrichs’ account that only events with arbitrarily short parts can be described by a sentence with afor-adverbial, certainly not an intuitive assumption.

At first sight it might look like Hinrichs’ approach could be salvaged by re- strictingl0 to range only over instants and not intervals. However, this requires the assumption that time is atomic, otherwise there are no instants. According to von Stechow (2009), this assumption is rejected by most semanticists, and I do not rely on it myself (see Section 2.4.4). Even if we adopt the assumption, the resulting modification of Hinrichs’ approach becomes too weak, as it ends up merely requiring that the main-clause event can be divided into at least two possibly overlapping parts that fulfill the main-clause predicate. That is,P for an houris predicted to be true of an eventeif and only ifPapplies to two or more eventse1, . . . , enwhich are distinct fromeand whose sum ise. These events may overlap and so their individual runtimes may be arbitrarily close to the runtime of e.

The analysis of Hinrichs (1985) is used in slightly adapted form by Abusch and Rooth (1990), but their adaptations do not prevent the problem from arising. A more significant reformulation of Hinrichs’ proposal is found in Rathert (2004), who renders it using the following representation:

(6) _JJohn ran for two hours_K(t) = 1

if and only if|t|=2 hours ∧ ∀m ⊆t∃n[m ⊆n⊆t ∧ _JJohn ran_K(n) = 1], withnbeing a minimal run-event (two steps, done faster than a walk) This reformulation runs into problems of its own: for example, sincemis not required to be an instant, it ranges over some subintervals which are too big to contain the runtimes of anyminimalrun-events. In fairness, I note that Rathert is less interested in the meaning offor-adverbials than in how they interact with tense and with the Perfect. Still, we see that the proposal of Hinrichs (1985) is technically flawed and has resisted several attempts to repair it.