Berit Gehrke (CNRS-LLF / Paris Diderot)

Title: Manner vs. clausal readings of adverbs

Abstract: Many adverbs display a duality between a VP-related and a sentence-related reading, which in English partially correlates with the adverb’s syntactic position (cf. Jackendoff 1972). For example, adverbs like cleverly or rudely have been noted to receive an agent-oriented reading in sentence-initial position (1-a), but a manner reading in sentence-final position (1-b); in a middle position both readings are available (1).

(1) Alice has rudely written the letter. AMBIGUOUS

a. Rudely, Alice has written the letter. AGENT-ORIENTED

∼ Alice rudely has written the letter.

b. Alice has written the letter rudely. MANNER

∼ Alice has written the letter rudely.

Syntactically, then, we can assume that the different readings depend on whether these adverbs appear high or low in the tree, and this is independent of whether they are taken to involve adjunction of the adverb to the projection they semantically scope over (e.g. Nilsen 2002, Ernst 2002), or to appear in the specifier position of a designated functional projection under cartographic approaches (e.g. Cinque 1999). For instance, Ernst (2002) proposes that the higher readings of the adverbs in question are basic, but that when such adverbs appear in Spec, PredP, a projection right above VP, the ‘Manner Rule’ applies and turns such adverbs into manner adverbs. Both readings, he argues, differ in the comparison classes of events involved: ‘Events’ more generally for the agent-oriented reading, and ‘Specified Events’ for the manner reading (roughly, involving a comparison with other letter-writing events in (1-b)). However, he does not make precise what the precise semantic difference between ‘Events’ and ‘Specified Events’ is under his account, and this talk aims at improving on this point.

Building on insights from the literature on nominal kinds (Carlson 1977, Zamparelli 1995, Chierchia 1998, i.a.), I assume that the VP domain correlates with the domain of event kinds (Carlson 2003, Gehrke 2015), i.e. the verbal counterpart to NP (cf. Zamparelli), whereas event tokens (i.e. instantiations of events in time and space) require further verbal functional structure, i.e. the verbal counterpart to NumP (cf. Zamparelli). With this more general picture in mind and building on the literature on comparison classes in the adjectival domain (in par- ticular Sassoon & Toledo 2011), I propose that the manner reading involves within-event-kind comparison at the level of the VP; manner adverbs more generally, then, derive an event sub- kind (see also Landman & Morzycki 2003). In contrast, the agent-oriented reading involves between-event-token comparison, since every instantiation of an event kind at a higher level of the sentence yields a different event token. This indirectly gives the impression, e.g. in (1-a), that Alice was rude during the course of the event, since Alice instantiated (brought about) the event token that is judged to be rude (cf. Wyner 1998, Geuder 2002 for the more general claim that agent-orientedness comes about indirectly). I will show how this account sheds new light on some well-known differences between the adverbs under the two readings, including their relative position with respect to other elements in the sentence, prosody, and semantics-pragmatics.