Interpreting past epistemic modals in English, Dutch, and French

According to many authors (Cinque 1999, Hacquard 2006, a.o.), epistemic modals in sentences such as “It had to be raining last night” always scope over past tense, expressing a present epistemic claim about a past event. However, there is no consensus on this point: for others (e.g. Rullmann & Matthewson 2018), epistemics can, or even must, scope below tense. This debate has substantial consequences for theories of modals. As of now, research lacks a systematic empirical description of the crosslinguistic picture. The examples discussed in the literature vary widely between languages, and most of the data comes from researchers’ own intuitions, fieldwork involving a limited number of informants, or informal questionnaires. The goal of this paper is, first, to settle the current disagreement on the judgments, by comparing in a more controlled way judgments in three languages, English, Dutch, and French. Second, it is to assess the prediction from Van Dooren (2020b) that there is variation between languages, which can be explained by crosslinguistic differences in modals’ lexical status: (semi-)auxiliaries versus verbs. We report two acceptability judgment experiments testing past tensed epistemic modal sentences’ interpretations in English, Dutch, and French, comparing had-to/moest/devait to the verbs seemed/leek/semblait, using identical sentences and methods and controlling for various factors overlooked in previous studies. We show that (i) epi>tense is available in all languages; (ii) English had-to is overall dispreferred as compared to Dutch moest and French devait. While we only find nuanced support for van Dooren’s proposal, our results open promising avenues for more controlled experimental research on modal-temporal interaction.