The present paper examines a particular type of interaction between grammatical aspect and modality, namely the aspectual contraints that are pragmatically and diachronically involved in the emergence of fake (past) tense and ‘counterfactuality’. The paper thus tackles the puzzle of ‘one-past counterfactuals’ where a simple past tense conveys ‘counterfactuality’, and is usually associated with ‘fake tense’, i.e. the non-past interpretation of past morphology. I argue for the distinction between two types of ‘counterfactuality’: (i) the contrary-to-fact interpretation ⌐p (or ‘real counterfactuality’) and the interpretation ⌐ π p (or unlikelihood). Within a neo-Reichenbachian conception of past tenses, I expand on the idea that counterfactuality and unlikelihood are implicatures of scalar origin that are locally derived from Grice’s maxim of quantity, and predict that counterfactuality is restricted to imperfects (imperfective pasts) while unlikelihood is allowed by non-perfective pasts (i.e. preterits and imperfects). Finally, I explore the pragmatic origin and conventionalization of counterfactuality and unlikelihood in two uses of the French imparfait, using a diachronic model à la Heine (2002). The analysis of data from Latin and Medieval French suggests that fake tense (and aspect) is partly due to the semantic bleaching of the past tense that parallels the conventionalization of ‘counterfactual’ implicatures.