Turkish 3rd person plural subjects normally appear with verbs that are unmarked for number. Following earlier findings which indicate that Turkish heritage speakers (HS) accept overt plural marking more readily compared to monolingually raised Turkish speakers, the present study investigates to what extent bilingual speakers are sensitive to grammatical, surface-level and semantic constraints on Turkish plural agreement marking. A scalar acceptability judgement task was carried out with non-bilingual Turkish speakers residing in Turkey and Turkish-German bilinguals residing in Germany. Our experimental design involved manipulating both subject animacy and subject position. Participants’ judgement patterns confirmed Turkish speakers’ general preference for unmarked verb forms, which was modulated both by subject animacy and by subject position. Significant differences were observed between lower proficiency HS on the one hand, and monolinguals and advanced proficiency HS on the other, suggesting that the relatively subtle interplay between different types of constraint on number agreement marking is affected by heritage language conditions. We found no evidence for simplification or optionality reduction in the lower proficiency HS’ judgements, however. We innovate on previous research by using Gradient Symbolic Computation modelling to capture between-group differences in the relative weightings of the constraints under investigation.

中文翻译：

---

土耳其-德语双语者主谓一致标记的限制

土耳其语第三人称复数主语通常与未标记数字的动词一起出现。继早期的研究结果表明，与单语养成的土耳其语使用者相比，土耳其语传统使用者 (HS) 更容易接受明显的复数标记，本研究调查了双语使用者在多大程度上对土耳其语复数协议标记的语法、表面水平和语义限制敏感. 对居住在土耳其的非双语土耳其语使用者和居住在德国的土耳其-德语双语者进行了标量可接受性判断任务。我们的实验设计涉及操纵主体生命力和主体位置。参与者的判断模式证实了说土耳其语的人普遍偏爱未标记的动词形式，这受到主体生命力和主体位置的调节。一方面，在较低水平的 HS 与单语者和高级水平的 HS 之间观察到显着差异，这表明不同类型的数字协议标记约束之间相对微妙的相互作用受到传统语言条件的影响。然而，我们没有发现证据表明在低水平 HS 的判断中进行了简化或可选性减少。我们通过使用梯度符号计算建模来捕获所调查约束的相对权重的组间差异，从而对先前的研究进行了创新。这表明不同类型的数字协议标记约束之间相对微妙的相互作用受传统语言条件的影响。然而，我们没有发现证据表明在低水平 HS 的判断中进行了简化或可选性减少。我们通过使用梯度符号计算建模来捕获所调查约束的相对权重的组间差异，从而对先前的研究进行了创新。这表明不同类型的数字协议标记约束之间相对微妙的相互作用受传统语言条件的影响。然而，我们没有发现证据表明在低水平 HS 的判断中进行了简化或可选性减少。我们通过使用梯度符号计算建模来捕获所调查约束的相对权重的组间差异，从而对先前的研究进行了创新。