Syntactic Complexity Effects of Russian Relative Clause Sentences in Children with and without Developmental Language Disorder

ABSTRACT

We investigated relative clause (RC) comprehension in 44 Russian-speaking children with typical language (TD) and developmental language disorder (DLD) (M age = 10;67, SD = 2.84) and 22 adults. Flexible word order and morphological case in Russian allowed us to isolate factors that are obscured in English, helping us to identify sources of syntactic complexity and evaluate their roles in RC comprehension by children with typical language and their peers with DLD. We administered a working memory and an RC comprehension (picture-choice) task, which contained subject- and object-gap center-embedded and right-branching RCs. The TD group, but not adults, demonstrated the effects of gap, embedding, and case. Their lower accuracy relative to adults was not fully attributable to differences in working memory. The DLD group displayed lower than TD children overall accuracy, accounted for by their lower working memory scores. While the effect of gap and embedding on their performance was not different from what was found for the TD group, children with DLD exhibited a diminished effect of case, suggesting reduced sensitivity to morphological case markers as processing cues. The implications of these results to theories of syntactic complexity and core deficits in DLD are discussed.

Acknowledgments

We thank Roman Koposov from Northern State Medical Academy (Arkhangelsk, Russia) for his help with data collection and Jodi Reich from Yale University for help with the stimuli preparation. We also gratefully acknowledge the contributions to this research by the late Dr. Maria Babyonyshev.

Funding

This research was supported by NIH grant R01 DC007665 (National Institute on Deafness and Other Communication Disorders) to Elena L. Grigorenko. Grantees undertaking such projects are encouraged to express freely their professional judgment. This article, therefore, does not necessarily reflect the position or policies of the National Institutes of Health, and no official endorsement should be inferred.

Notes

¹ We chose to use the term DLD rather than SLI because there is no consensus among language scientists on the validity of categorically distinguishing language impairment in children with IQ above and below a certain arbitrary cutoff. For this reason, SLI is not included as a specifier of a language disorder in the DSM-5, following the recommendations of ASHA. However, all other inclusionary and exclusionary criteria typically applied for the category SLI are applicable to the DLD group in our sample.

² The architecture of working memory in the model by Just and Carpenter is distinct from that in the influential Baddeley and Gathercole model. The latter posited a system with two components: a storage component comprised of two modality-specific storage systems (the phonological loop, an articulatory rehearsal buffer of limited duration, and visual-spatial sketch pad) and the central executive. In the Just and Carpenter model, working memory does not include modality-specific buffers and roughly corresponds to the part of the central executive that deals with language comprehension in the Baddeley and Gathercole model.

³ An alternative account of this phenomenon is the Perspective Shift theory (MacWhinney & Pleh 1988), according to which processing sentences with more perspective shifts requires greater resources, where the perspective is taken from the subject of a clause. Thus, a center-embedded object-gap RC requires perspective to shift from the matrix subject to the subject of the RC and from the subject of the RC back to the matrix subject. On the other hand, processing a center-embedded subject-gap RC requires no perspective shifts because the matrix subject is also the subject of the RC. In right-branching RCs, object-gap sentences have fewer perspective shifts than subject-gap. This theory makes gradient predictions with regards to difficulty of various types of RC.

⁴ The review of theoretical proposals was not meant to be exhaustive. There are a number of additional proposals not reviewed here, as we focused only on those we felt our data could have a bearing on.

⁵ There are other proposals in this class, such as those that attribute the syntactic deficit to a difference in the featural composition/specification of certain functional heads (Clahsen, Bartke, & Gollner 1997; Jakubowicz & Nash 2001; Mastropavlou & Tsimpli 2011).

⁶ Levy, Fedorenko & Gibson used the Russian Dependency Treebank, a collection of late 20th-century texts comprised of ≈ 35,000 sentences (1 million words) from a mixture of genres, such as fiction, news, and scientific literature) hand-annotated for dependency structure. For RCs with full NPs, they report the frequency of 147 for subject-gap RCs with the VO order, and 4 for subject-gap RCs with the OV order. For object-gap RCs, they report the frequency of 41 for the VS order and 29 for SV order.

⁷ An alternative label for dummy coding is treatment coding, which is a way to create dichotomous variables, where each level of the categorical variable is contrasted to a specified reference level.

⁸ Following a suggestion by an anonymous LA reviewer, we performed this analysis separating digit span forward and backward to investigate whether one of these tasks affects children’s performance to a greater extent than the other. We found that forward digit span (DSF) produced results nearly identical to those reported for the summed score, and the addition of the backward digit span (DSB) scores (and their interactions with the study variables) did not affect this pattern, was not related to performance, and did not result in the significant improvement in model fit over the model with DSF only: Chi-square(4) = 3.4761, p = .4815.

Additional information

Funding

This research was supported by NIH grant R01 DC007665 (National Institute on Deafness and Other Communication Disorders) to Elena L. Grigorenko. Grantees undertaking such projects are encouraged to express freely their professional judgment. This article, therefore, does not necessarily reflect the position or policies of the National Institutes of Health, and no official endorsement should be inferred.