Research on multilingual speakers is often compared to monolingual baselines which are commonly treated as if they were homogeneous across speakers. Despite recent research showing that this homogeneity does not hold, these practices reproduce native-speakerism and monolingualism. Heritage language research, which established itself in the past two decades, is no exemption. Focusing on three predefined linguistic groups, namely Turkish speakers which are framed as monolingual in Turkey as well as two heritage bilingually framed groups in Germany and the USA, we ask: (1) Do heritage speakers of Turkish produce more discourse and fluency markers (FMs) than monolingual speakers? (2) Are the groups homogeneous, or is there wide variation between speakers across groups? We focus on the variation between and within groups using Bayesian Linear Regression with a multilevel model for speakers and heritage groups. Our findings confirm that the use of discourse and FMs is largely defined through individual variation, and not through the belonging to a certain speaker group. By focusing on variation across groups rather than between groups, our study design supports the growing body of literature that questions common heritage language research practices of today and shows alternative paths to understanding heritage grammars.

对多语言说话者的研究经常与单语基线进行比较，单语基线通常被视为在不同说话者之间是同质的。尽管最近的研究表明这种同质性并不成立，但这些做法再现了母语主义和单语主义。过去二十年来建立起来的遗产语言研究也不例外。我们关注三个预先定义的语言群体，即在土耳其被视为单语的土耳其语使用者，以及在德国和美国的两个传统双语群体，我们问：(1) 说土耳其语的传统使用者是否会产生更多的话语和流利度标记（FM） ）比单一语言的人？(2) 各组是否同质，或者不同群体的发言者之间是否存在很大差异？我们使用贝叶斯线性回归以及针对说话者和传统群体的多级模型来关注群体之间和群体内部的差异。我们的研究结果证实，话语和调频的使用很大程度上是通过个体差异来定义的，而不是通过属于某个特定的说话者群体来定义的。通过关注群体之间而不是群体之间的差异，我们的研究设计支持了越来越多的文献，这些文献质疑当今共同的遗产语言研究实践，并展示了理解遗产语法的替代途径。