Authier and Reed (2009: 44–45) observe that thematic constraints involving French devoir ‘must’ and falloir ‘to be necessary’ fall out from Chomsky and Lasnik’s (1995) original null Case approach to control. However, Reed (2014: Ch. 6) shows that separate data involving Romance floating quantifiers (Baltin 1995; Sportiche 1988), pronominal clitic placement (Cinque 2001; Cardinaletti and Shlonsky 2004), and the distribution of overt and covert nominals in gerunds pose serious problems for that analysis. This article opens with the novel observation that Authier and Reed’s contrasts are problematic for theories of control ranging from those couched in terms of NP-Movement (e.g. Hornstein 1999) to those postulating a syntactically implicit external argument (e.g. Jackendoff and Culicover 2003) to those formulated in terms of Agree and/or predication (Landau 2004, 2013, 2015). This article then draws on additional data and discussion in Reed (2014, 2016) to show how a Caseless approach to PRO can accommodate these facts, but only if one also recognizes with, e.g. Kamp (1975) and Cinque (1999), and contra, e.g. Kratzer (1981, 1991) and Hacquard (2010, 2011), that the closest French equivalent of must, namely, devoir, is syntactically and semantically ambiguous. This article, therefore, sheds new light on our current understanding of control and modality.