Se out of control

Within the framework of a uniform theory of the so-called se constructions in Spanish, I propose to explain a control ban that has received almost no attention in the previous bibliography. Specifically, as long as a subject control sentence has an impersonal SE as controller, the subordinate infinitive clause cannot contain any other instance of the clitic SE, other than the so-called spurious SE. The source of this restriction follows, as I will argue, from a legibility problem at LF produced, specifically, by a failed attempt to apply Agree between PRO and the embedded SE, which, as we shall see, acts as a probe for A-movement. If the explanation that I offer is correct, it also follows a series of theoretical conclusions that directly affect the way in which we must conceive of the design of Agree in the syntax and its effect at the LF interface. In particular, the system tolerates certain Agree failures (Preminger 2014) as long as it does not affect legibility in the semantics. Indeed, the theory of SE constructions that I assume here derives the distinction between paradigmatic and non-paradigmatic SE as the result of successful or unsuccessful Agree applications, respectively. The limit of this tolerance to failed applications of Agree must be found in the type of semantic object that can be deduced at LF. This limit is illustrated with the aforementioned restriction in control and impersonal SE contexts that motivates the present study.