In this project, we investigate three constructions that have proven problematic for standard formulations of grammatical constraints on dependencies – these constructions should be ruled out by standard formulations of the constraints, but have been reported to be (relatively) acceptable. In this project we ask two questions. First, we ask whether these constructions show the pattern of unacceptability that is typically taken to indicate the presence of a constraint (i.e., island effects). We use a 7-point acceptability judgment study, and the factorial definition of island effects that is standard in the literature (in which the island effect appears as a superadditive interaction). We test both the three apparently exceptional constructions in (1) and their structurally-similar, standard island effect counterparts in (2) so that we can compare effect sizes and relative acceptability. Second, we ask whether the island effects that we see for these three exceptional constructions can be explained by the working-memory capacity approach to island effects (Kluender and Kutas 1993, Hofmeister and Sag 2010) by correlating participants’ judgment responses to their performance on a standard reading span (working memory) task.