Refer if you can: On the decline of the stressed reflexive in the history of French

Empirically, this paper is concerned with the diachronic replacement of a reflexive pronoun with a non-reflexive pronoun for same-subject anaphora in French. In Modern French, the stressed reflexive soi is mainly restricted to non-referential subjects like on or chacun; in other contexts it has been replaced by lui/elle. In Old French however, soi was productively used in most contexts, although lui/elle was far from uncommon. In a corpus-based study, I track the gradual replacement of soi with lui/elle. In order to account for this change, the paper submits that these are alternative but fundamentally different options, soi reflecting an anaphoric relation in a predicate (binding) and lui/elle reflecting a textual relation between discourse referents (coreference). While both options are available in principle during all periods of French, they are not fully interchangeable. A comparison with Modern English confirms that the choice between bound anaphor (soi) and unbound pronoun (lui/elle) is not only governed by the syntactic configuration but that it reflects different types of coherence in discourse. With respect to an explanation of the change, it is being pointed out that it echoes a number of instances of secondary grammaticalization in the history of French, all of which consist in the replacement of a pragmatically unmarked form by a pragmatically marked competitor, un-marking the latter in the process. This is, it is suggested, driven by speakers ' strategies aimed at producing a noteworthy and expressive utterance, the 1 additional benefit of the non-reflexive over the reflexive being reference, as opposed to mere mention in a predicate implied by the reflexive. This furthermore suggests that non-complementary contexts, i.e. contexts where both the reflexive and the non-reflexive are acceptable, are in fact the core kind of contexts for the theory of same-subject reference, rather than being marginal ones as the adjective “exempt ” used for them by Pollard & Sag 1992 and Büring 2004 would imply.