We argue against the position, articulated in Reinhart and Siloni (2005) that Universal Grammar allows valence changing operations to apply in the syntax as well as in the lexicon. Focusing on reflexivization and reciprocalization, we suggest that languages which are argued to have reflexivization in the syntax are in fact languages with a syncretism between the markings for valence changing and for anaphora. We suggest that no "syntax" language marks reflexivization with derivational morphology, and all such languages have a reflexive morpheme that is a pronominal element: this follows from analyzing putative syntactic reflexivization as anaphoric binding. The paper focuses on the French clitic se, which we claim marks lexical reflexivization and reciprocalization with naturally reflexive and reciprocal predicates, anaphoric binding with all predicates, and a long distance anaphor in syntactically restricted environments. We show that all the properties that are said to follow from the syntactic reflexivization analysis follow from the anaphoric binding analysis, and that many properties which follow from the anaphoric binding analysis cannot be accounted for naturally by the syntactic reflexivization analysis. Finally, we counter arguments raised by Grimshaw (1981) against analyzing se as an anaphor.