Retour vers liste

Détail de la contribution

Auteur: Ariel COHEN

Embedded Epistemic Modals: An Experimental and Theoretical Investigation

Abstract/Résumé: Embedding presents a challenge to any theory of epistemic modals (EMs). Papafragou (2006) follows Lyons's (1977) distinction between subjective and objective EMs, and claims that the former, exemplified by (1), cannot be embedded, while the latter, exemplified by (2), can. (1) a. *If Max might be lonely, his wife will be worried. b. *It is surprising that Superman might be jealous of Lois. c. *Spiderman told me that Superman might be jealous of Lois. (2) a. If it might rain tomorrow, people should take their umbrellas. b. It is surprising that it might rain tomorrow, since there was no sign of a cloud all day today. c. The weather forecast told viewers that it might rain tomorrow. The stars in (1) indicate Papafragou's own judgments, but we tested them experimentally. Subjects were asked to judge the acceptability of embedded EMs in contexts that distinguished between subjective and objective readings. The results indicate that Papafragou is correct about the antecedents of conditionals (1a, 2a) and the complements of factives (1b, 2b); but not about verbs of saying (1c, 2c), where both types of EM are good. How can these results be explained? Von Fintel & Gillies (2011) propose that `might P' means that P is compatible with X's knowledge. When the speaker "puts into play" `might P', X usually refers to the speaker herself; but the utterance introduces a set of propositions where X can also refer to the hearer or to speaker + hearer. The introduction of this set of propositions crucially relies on the speaker putting the sentence into play, i.e. "making it available for denial and uptake" (p. 119). We propose that this is the reason why subjective EMs are ruled out in non-assertive contexts, which do not allow denial or uptake: factives (which are presupposed) and antecedents of conditionals. Subjective EMs are acceptable only in assertive contexts, e.g., when unembedded or embedded under verbs of saying. Von Fintel and Gillies do not consider objective EMs, but their theory can be naturally extended to account for them. Following approaches to objective interpretations of vagueness (Wright 1987), clarity (Wolf & Cohen 2011), and personal taste (Cohen 2010), we propose that for objective EMs, X refers not to the speaker or hearer, but to a group of the experts on the matter at hand. Crucially, no set of propositions is introduced: both speaker and hearer take the EM to be evaluated with respect to the group of experts (though they may disagree on their identity). Hence, objective EMs are not restricted to assertive contexts, and may be freely embedded. Thus, the results of our experiment are fully explained by a natural extension of von Fintel and Gillies's theory of EMs.