UCLA

In counterfactual conditionals, speakers make (at least) two counterfactual inferences. For example, upon hearing the utterance "If it were raining, I would be wet", an interlocutor infers that it is not currently raining (CFp) and that the speaker is not currently wet (CFq). It has been shown that these inferences are not asserted but rather implicated since they can be cancelled without contradiction or redundancy (Anderson 1951).A large number of languages (EM languages) mark such counterfactuality with past-tense morphology in either clause of the subjunctive conditional (Iatridou 2000). In EM languages, CFp and CFq are cancellable, but there exists an asymmetric independence in their cancellability. In other words, CFq can be cancelled independently of CFp, but if CFp is cancelled then CFq is necessarily also cancelled. There is another group of languages (non-EM languages) where counterfactuality is marked not by past tense but by some other dedicated counterfactual morphology. In these non-EM languages, CFp is not cancellable (Nevins 2002). In this thesis, I present novel data from Mandarin and Hebrew (both of which are non-EM languages) demonstrating that although CFp is non-cancellable in non-EM languages, CFq can indeed be cancelled. Based on a broad study of Indo-European EM-languages, Iatridou (2000) proposes that counterfactuality arises as an implicature due to the presence of an exclusion marker realized as past-tense morphology. Under Iatridou's proposal, cancellability is an inherent property of exclusion, not of counterfactuality itself. Therefore, it would be predicted that non-EM languages (which do not utilize exclusion to mark counterfactuality) would not be able to cancel counterfactual inferences. This much seems plausible for CFp (Nevins 2002), but CFq would be expected to also be non-cancellable. As the novel data show, this is not the case. While Iatridou does not formalize her proposal, Tellings (2016) speculates how one might do so. I further present an extension on Tellings (2016), and show that it is not an adequate theory of counterfactuals, at least in non-EM languages.