The current study aims to systematically investigate the types of resumptive pronouns (RPs) in Baha Arabic (BA), an under-studied variety of Arabic featuring a productive use of RPs across different types of wh-dependency structures. We report on an acceptability judgment experiment that measures the extent to which crossing non-island and island clause boundaries affect resumption behavior across different types of wh-dependencies. Results suggest that two types of dependency structures need to be distinguished in BA, in relation to RPs’ distribution: structures featuring illi complementizer (i.e. relative-clauses and cleft-wh-questions) vs structures not featuring illi (i.e. bare-wh-questions).Illi-structures are always rejected with gaps but, with RPs, they remain acceptable even when crossing an island boundary. This suggests that these structures are binding dependencies featuring grammatical base-generated RPs. If no island boundary is crossed in bare-wh-questions, gaps are highly accepted, but RPs receive highly variable ratings (with variation entirely within subjects). This variability is not explained by individual variation in WM nor by embedding effects. No amelioration effect of RPs was observed inside islands. These findings suggest that bare-wh-questions are movement dependencies, and that BA does not feature intrusive RPs. We discuss these findings considering theories of resumption in syntax and psycholinguistic literatures.