The literature argues that linguistic enclaves negatively affect immigrants’ language proficiency by reducing their exposure and incentives to learn destination language. This negative association may, however, be spurious, arising due to the self-selection of immigrants into regions with larger enclaves. Exploiting the natural experiment of the German residential policy, this paper analyses the influence of linguistic enclaves on refugees’ language proficiency. We find no evidence that enclaves hamper German language learning among refugees. Our results are robust to various measurements and model specifications. We conclude that the negative relationship reported in previous research is mainly driven by immigrant’s residential sorting.