According to Montrul’s (2016) Incomplete Language Acquisition Hypothesis (ILAH) structures acquired late by monolinguals are even more vulnerable in heritage languages (HL); their acquisition will be incomplete. Crosslinguistic research identified an asymmetry between subject and direct object relatives (SR/DOR) in L1. DORs emerge later than SRs are more difficult to produce and to comprehend. However Polinsky (2008) Marinis Özge (2020) report findings which challenge ILAH. Heritage speakers (HS) of Russian (mean age 7;7) and of Turkish (age range 4;7-9;2) respectively with English as the majority language comprehend relative clauses (RC) early in spite of the differences between RCs in the HL and in the ML. The aim of this study is to investigate the use of RCs in heritage Romanian with French the ML a context which has been understudied (see Soare 2020). RCs in the HL and the ML are relatively similar. We analyze RC production with a view to testing the predictions of ILAH: (i) RCs in the HL should emerge later than in monolingual development; (ii) performance should deteriorate as an effect of increased exposure to the majority language. We used an elicited preference task (a replica of Novogrodsky Friedmann 2006) designed within COST A33 previously used for Romanian in Sevcenco et al. (2013). The task contains 20 test items (10 SR and 10 DR). 18 HSs of Romanian  (mean age 8;11) born in France in Romanian families took part in the study. Their dominant language is French (as determined by a questionnaire and the analysis of  frog stories which they produced in both languages). The results were compared to a  group of monolinguals and to data previously reported for Romanian adults. SRs and DORs are indeed acquired later by the HSs with a more significant delay for DORs. Before age 8 even SRs are problematic. The younger HSs (7 children mean age 6;6) produced 62.8% SRs (n = 44) whereas the monolinguals (7 children mean age 6;6)  produced a SR 95.7% (n=67). DORs are more delayed. HSs produced DORs 14.3% (n=10) whereas the monolinguals produced DOR 50% (n=35) of the time.  Contrary to the predictions of ILAH no deterioration of RC production was found with age. The older group of HSs (11 children mean age = 12;7) produced 95.5 % SRs and 31.8% DORs. Though the rate of DORs is lower than the one attested with the younger monolinguals the analysis of the DOR avoidance strategies revealed a high rate (49%) of (passive) SRs used instead which is similar to what was reported for Romanian monolongual adults ( Sevcenco et al. 2013). In line with previous studies our findings show that complex syntactic structures can be acquired in spite of the impoverished input which HSs receive though at a lower pace as attested with simultaneous bilinguals in a non-heritage setting. The comparison with the studies by Polinsky (2008) and Marinis Özge (2020) reveals that this is possible both in contexts of distance and of similarity between the two languages.

View poster here.