In Landau 2015, it is proposed that the acceptability of implicit control (i.e. control by the implicit external argument of a passivized verb into complement clauses) is not only restricted by the revised Visser’s generalization (van Urk 2013), but also depends on the type of matrix predicate involved. While attitude matrix predicates allow implicit control (implicit logophoric control), nonattitude matrix predicates do not. Landau takes this bifurcation to support his two-tiered theory of control by assuming that in the case of nonattitude matrix predicates, the control relation is essentially a predication relation, from which implicit arguments are independently excluded. In this article, we subject these claims to empirical scrutiny, showing that Landau’s generalization on implicit control holds only in a subset of languages, while other languages license implicit control with both types of matrix predicates. We investigate and reject the hypothesis that this crosslinguistic split is the consequence of different types of implicit arguments, only some of which are syntactically represented in a way that allows them to enter a predication relation. Based on an investigation of the acceptability of agent-modifying depictives in passives, we conclude that, in principle, implicit external arguments of passives in all languages under consideration can enter predication. We show, however, that there is a different correlation: languages that allow implicit control with nonattitude verbs (implicit predicative control) are exactly those languages that allow impersonal passives of unergative predicates. To account for this correlation, we argue that implicit logophoric control, but not implicit predicative control, can be construed as a personal passive.