Since the earliest studies on multilingual advantages, it has proved difficult to disentangle language development from general cognition. It remains unclear whether language interdependence is an independent variable or whether observable effects are mediated by cognitive ability. Measurable effects of one language on another typically go hand in hand with differences in cognitive ability. We hypothesise that high cognitive ability produces stronger language interdependence effects than low cognitive ability. We consider this problem in the context of heritage bilingualism in Germany comparing a linguistically mixed cohort of bilingual students (n = 557; i.e. Russian-German, n = 237; Turkish-German, n = 320) with a monolingual German control (n = 852) regarding their proficiencies in the foreign language English. We ask whether the bilingual students manifest an English development that is different from their monolingually socialised peers. We place the students in three different groups depending on their performance in a visual-spatial cognitive ability test. We fit structural equation models to test whether heritage language and German proficiency impact English proficiency differently across these groups while additionally controlling for language background and socio-economic status. Results reveal differences between the high cognitive ability groups, here interpreted as a conditioned bilingualism effect.