Effects of language experience on sentence repetition in the heritage and majority languages: Evidence from bilingual children in the UK

A speaker learns a heritage language (HL) from their caregivers in the home, distinct to the majority language (ML), which is widely spoken across the community in their place of residence (Rothman, 2009). Sentence repetition is widely treated as an index of morphosyntactic proficiency in both languages, allowing researchers to assess the learners’ ability to produce specific syntactic structures (Kałowski et al., 2025). A recent meta-analysis demonstrated how isolated language experience factors affect heritage speakers’ ability to repeat these sentences, depending on the testing language, including the participants’ age, age of onset of the ML and exposure to the HL (Bennett et al., under review). The meta-analysis’ effect of exposure on morphosyntactic proficiency included the combined effects of current (short-term) HL exposure and cumulative exposure to the HL over the lifespan. It was not possible to analyse the effect of these variables on SRT performance separately given that the number of studies analysing these variables individually was too small to carry out the analysis. In the present study, we choose to analyse the effects of age (as the child-internal developmental factor) and current HL exposure and socioeconomic status (SES) as the child external factors. We excluded a measure of cumulative exposure given that it involves HL use over the lifespan and may therefore be a proxy for age, as discussed in the meta-analysis. Age of onset of the ML was excluded as an effect given that the children in the present study are relatively homogeneous in this respect. SES can be a significant predictor of language performance in bilingual populations (e.g., Calvo & Bialystock, 2014; Chondrogianni & Marinis, 2011), yet it was not possible in the meta-analysis to test its effect on sentence repetition given inconsistencies in how SES was measured. SES has been found to affect sentence repetition performance (e.g., Soto-Corominas et al., 2022) and in some cases interact with the effect of exposure on sentence repetition in complex ways (e.g., De Cat, 2020). This is because SES also relates to the child’s language experience, where a higher SES may result in a richer language environment (De Cat, 2020). We will therefore test whether its effect changes the overall results. Aside from other tasks, such as a narrative task, sentence repetition allows researchers to manipulate the effect of target structure type. The complexity of the target structure can affect sentence repetition performance in bilingual speakers, including in English (Marinis & Armon-Lotem, 2015; De Cat, 2020) and in Portuguese (Correia et al., 2024). These studies administered the LITMUS sentence repetition task, in which complexity is operationalised based on the target syntactic structure of the sentence of the testing language (e.g., wh-object questions, conditionals, object relative clauses, etc.) (Marinis et al., 2012). The complexity level is set based on prior research of the target language showing that children are most accurate in their production of the least complex structures, complexity level 1 (C1), and least accurate in their production of the most complex structures, complexity level 3 (C3). Production accuracy of syntactic structures in complexity level 2 (C2) falls between these two levels. The complexity of the syntactic structure is relative to all other syntactic structures within the same language. As such, cross-linguistic structures may be assigned to C2 in a language containing other structures that are more complex, but C3 in different language in which the given structure is among the least accurate. Whilst De Cat (2020) and Correia et al. (2024) tested the children in one language, it is important to compare bilingual children’s performance in both languages, firstly, to establish the extent to which each child is dominant in one language. Secondly, this allows the identification of instances in which a child’s accuracy is relatively low in both languages, suggesting a more generalised difficulty in language development (Antonijevic-Elliott et al., 2020). This provides us with an exploratory analysis of language dominance and overall language development across both languages. As such, we will analyse the extent to which age, current exposure to the HL, SES and target structure complexity predict proficiency in the HL (Portuguese) and also in the ML (English).