Frequency adjectives and an ontological solution to a compositional problem
Berit Gehrke and Louise McNally Universitat Pompeu Fabra
Frequency adjectives (FAs) such as occasional, frequent, and daily, pose two sorts of challenges for semantic composition. First, though FAs intuitively describe events, they can combine not only with event-denoting nouns but also with nouns that do not denote events (see (1)). Moreover, as Stump 1981 notes, the resulting interpretation is not always uniform (contrast the paraphrases in (1a) and (1b)). The sentences in (1) illustrate the so-called generic reading of FAs (Stump 1981, Schäfer 2007). However, Stump 1981 argues that FAs have at least two additional readings – an internal reading ((2a)), which will not concern us further here, and an adverbial reading ((2b), originally observed in Bolinger 1967). The second, and more difficult, composition problem arises with this latter reading. Stump 1981, Larson 1998 and Zimmermann 2003 provide a semantics for the adverbial reading; their analyses all effectively treat the FA as a quantifier. Zimmermann‟s analysis, sketched in (3), is representative (e is a syntactically represented event argument; semantic details which are irrelevant for our purposes are omitted). Since there are certain semantic and syntactic problems for raising the adjective by itself to get scope over the verb‟s event argument (see Larson 1998, Zimmermann 2003), Zimmermann syntactically incorporates the FA into the determiner, which forms a complex quantifier with it, roughly following Larson. This analysis correctly predicts the fact that occasional on this reading occurs only with a very limited set of determiners (mainly a and the) and that these determiners appear to have no semantic import: determiners with „inherent semantic content‟ (which a and the by assumption must lack) cannot participate in complex quantifier formation because their inherent semantic content must not be overwritten. The problem with this sort of analysis is that there is no independent motivation for the determiner-like syntax and semantics for FAs. First, there is no trigger for the movement of the FA in (3a) that licenses the adverbial reading. Second, if the FA is interpreted as a complex quantifier, it is not really clear why it must be accompanied by a real determiner (cf. the ungrammaticality of (4)). Finally, it is far from obvious how to extend the semantics in (3b) to the other readings, and indeed Stump 1981 provided a radically different semantics for the generic reading. Schäfer 2007 is the first attempt we know of to provide a (relatively) unified analysis for all readings of FAs, but his analysis of the adverbial reading in particular suffers from technical problems. In this paper, we argue that these problems can be solved, and a version of Schäfer‟s unified analysis maintained, if we make one simple, novel, and easily justifiable assumption: that sentences can be used to make assertions not only about events, but also about event types. Our semantics for occasional follows Schäfer‟s in that the FA combines with a nominal to yield a description of a kind (see (5a)). The rest is different. The FA modifies the kind indirectly, by directly modifying an event kind in which the nominal kind participates ((5b)). This event is described by the variable Qi, which gets its value contextually. Though by default the value of Qi is drink for a beer, we might imagine Qi getting other values in very specific contexts. This analysis, like Zimmermann‟s, straightforwardly captures the fact that FAs almost always combine only with the definite or indefinite articles, but without requiring the stipulation that these articles are contentless. We argue that the articles that appear with FAs are precisely those that are compatible with kind terms. Moreover, the analysis avoids positing unmotivated syntactic movement and maintains a standard semantics for the adjective. Although the semantics in (5) is very similar to Schäfer‟s, as will be shown in detail in the talk, Schäfer does not show how his semantics for the nominal is combined with the semantics for the rest of the sentence. We will show that under the standard assumption that sentences quantify over event tokens, Schäfer‟s analysis produces incorrect readings.
We then derive the generic readings in (1) as well as the adverbial reading from the representation in (5c). The semantics for (1a) and (1b) appear in (6a) and (6b), respectively. The parallel in the representation of the nominal amounts to the claim that the difference in interpretation noted in (1) is illusory: the nominals in both (1a) and (1b) could be paraphrased as “a kind of thing which is beer and is drunk occasionally (by the relevant person)”. The adverbial reading for Sara drank an occasional beer (ignoring tense) appears in (7). We suggest that for such an event kind to exist, there must be appropriate event tokens that support its existence. If Qi is fixed to λxkλek[drink(ek, Sara, xk)], this will entail that there have been on occasion drinking event tokens involving beer. A final advantage of our analysis is that it correctly predicts that only non-episodically used tenses can appear on the main verb. We will argue that such tenses are only appropriate when one is making assertions about event tokens, as opposed to event types. When semantic composition becomes problematic, invariably there is some trade-off between maintaining the simplest possible assumptions for one part of the grammar (syntax, lexical semantics, or semantic composition rules) and complicating them for another. We argue in this case that the preferable analytical alternative is a small change in our assumptions about what sentences can describe – exactly parallel to one independently motivated in the nominal domain for the analysis of natural kinds – with the result that no special assumptions are needed for syntax or semantic composition. Examples (1) a. An occasional beer is good for you. = Drinking a beer occasionally is good for you. b. An occasional beer tastes good on a hot day. ≠ Drinking a beer occasionally tastes good on a hot day. (2) a. That claim was made by an occasional sailor. = …by someone who sails occasionally. b. The / An occasional sailor strolled by. = Occasionally, a sailor strolled by. (3) a. [IP[QP[Q the/an+occasional1][NP t1 sailor]][VP e strolled by] b. an/the occasional: PQ[INFREQe,x][part-of(e,e*) Q(x)][S(e,x) ….] (4) *Occasional sailor strolled by. (5) a. ocasional: λPλxkek[Qi(ek, xk) P(xk) occasional(ek)] b. occasional beer: λxkek[Qi(ek, xk) beer(xk) occasional(ek)] c. a/the occasional beer: xkek[Qi(ek, xk) beer(xk) occasional(ek)] (6) a. ek [good-for-you(ek, xkek[Qi(ek, xk) beer (xk) occasional(ek)])] b. ek [tastes-good(ek, xkek[Qi(ek, xk) beer(xk) occasional(ek)])] (7) ek[drink(ek, Sara, xkek[Qi(ek, xk) beer(xk) occasional(ek)] (8) a. ??She was reading an/the occasional book. b.??This year, she has written an/the occasional book. References Asher, Nicholas (2008) A Web of Words: Lexical Semantics in Context. Ms., U Texas Austin/CNRS. Larson, Richard K. (1998). Events and Modification in Nominals. In Strolovitch, Devon & Aaron Lawson (eds). Proceedings from SALT VIII. CLC Publication: Ithaca, NY, 145-168. Schäfer, Roland (2007). On frequency adjectives. In Puig Waldmüller, Estela (ed.). Proceedings of Sinn und Bedeutung 11. Barcelona: Universitat Pompeu Fabra, 555-567. Stump, Gregory T. (1981). The interpretation of frequency adjectives. Linguistics and Philosophy 4, 221-257. Zimmermann, Malte (2003). Pluractionality and complex quantifier formation. Natural Language Semantics 11, 249-287.