This article addresses two issues that emerge from a close look at extraction out of Russian indicative čto-clauses – the first being that these clauses show unexpected weak-island behavior, and the second the generally problematic question of how it can be possible for weak islands to allow “marginal” extraction at all (in argument cases), a grammaticality status never traditionally explained in pre-minimalist literature and theoretically impossible on core minimalist assumptions. An approach is proposed for weak islands under Minimalism that eliminates the non-minimalist principles that were claimed to account for their behavior (especially Subjacency and the Empty Category Principle [ECP]) and also allows for an understanding of why Russian indicative čto-clauses show the partial opacity observed.