Turkish as an immigrant/heritage language in Europe has been extensively studied in both children and adults from a diverse perspective. While previous research conducted with late bilingual Turks in Europe who left their native language (L1) environment usually in early adulthood has not pointed to any significant L1-divergent performance except for subtle changes (e.g. Yılmaz, 2013), findings coming from studies conducted with subsequent generations (early bilinguals) on the other hand, demonstrated systematic L1 divergences especially in morphosyntactic categories (e.g. Akıncı, 2003; Pfaff, 1993). Regarding the resistance of diaspora varieties of Turkish to attrition, although the importance of the generation descent has been emphasized (Huls and van de Mond, 1992), we know relatively little about any systematic roles played by age at onset of bilingualism (AaO) which might have led to these differences between the immigrant generations. The possibility that the subsequent generations might be receiving modified input from the parental generation (e.g. Verhoeven, 2004) has also been suggested several times. That notwithstanding, this has not been systematically investigated either and the exact roles played by any sort of modifications in the input quantity and quality available to these speakers through previous generations have remained understudied.

The current study addresses these gaps by investigating the spoken performances of a total of 92 adult Turkish-English bilinguals with a wide AaO range (0–42) and of 44 Turkish monolinguals. This approach allowed us to control for the quality of input available to the speakers within this community and test the impact of AaO to see whether these factors remain equally predictive of L1 knowledge across a wide range of linguistic abilities including past tense, overall structural complexity, foreign accentedness, and word formation. The synthesis of the findings obtained suggested that this was not the case. The productivity in word formation, for example, was largely independent of AaO effect and past L1 experience, while both factors were at play in the remaining properties showing a dynamic, nonlinear interaction between the two. While in older bilinguals the transfer from the L2 to L1 was mostly subtle (due to late AaOs), for younger bilinguals, L1 development was variable and affected by a range of additional factors. Findings are discussed within the premises of various theoretical approaches.