Syntactic flexibility and competition in sentence production: The case of English and Russian

Abstract

We analysed how syntactic flexibility influences sentence production in two different languages—English and Russian. In Experiment 1, speakers were instructed to produce as many structurally different descriptions of transitive-event pictures as possible. Consistent with the syntactically more flexible Russian grammar, Russian participants produced more descriptions and used a greater variety of structures than their English counterparts. In Experiment 2, a different sample of participants provided single-sentence descriptions of the same picture materials while their eye movements were recorded. In this task, English and Russian participants almost exclusively produced canonical subject–verb–object active-voice structures. However, Russian participants took longer to plan their sentences, as reflected in longer sentence onset latencies and eye–voice spans for the sentence-initial subject noun. This cross-linguistic difference in processing load diminished toward the end of the sentence. Stepwise generalized linear model analyses showed that the greater sentence-initial processing load registered in Experiment 2 corresponded to the greater amount of syntactic competition from available alternatives (Experiment 1), suggesting that syntactic flexibility is costly regardless of the language in use.

Keywords:

• Syntactic flexibility
• Competition
• Sentence production
• English
• Russian

Acknowledgments

This research was supported by the Economic and Social Research Council (ESRC) Grants PTA-026-27-1579 awarded to Andriy Myachykov and RES-062-23-2009 awarded to Christoph Scheepers. Authors gratefully acknowledge Victor Shklovsky and Maria Ivanova at Russian National Center of Speech Pathology and Neurorehabilitation for their help in data collection for Experiment 2 and Oliver Garrod at University of Glasgow for his help with creating the script for automatic eye–voice span extraction.

Notes

¹ Cleft constructions are also possible in Russian. However, as with English, such constructions are extremely unlikely to be considered by Russian speakers in the absence of strong contextual constraints.

² In both Russian and English, the onset of the spoken name was determined as the onset of the relevant noun in the spoken response.

³ Since it was difficult to identify a unique “event region” in the pictures, we refrained from calculating EVS values for the verb constituent.

⁴ This results in an ordered incremental modelling approach, contrasting with standard Type III decomposition, which results in simultaneous testing of model effects. Note that simultaneous testing is unsuitable for present purposes because the two most critical predictors (language and flexibility) are highly correlated with one another (simultaneous testing would not be able to reliably estimate each predictor's unique contribution to the model fit).

⁵ Indeed, Experiment 1 revealed that English speakers infrequently produced options other than SVO-active or passive voice, even though such alternatives were clearly indicated to them in the instructions.

⁶ Similar claims can be made for the pronoun manipulations in Experiment 2 of V. S. Ferreira (1996). For example, we found that gave followed by a “nonconstraining” pronoun (him or her) is about 3.6 times more likely to occur in the corpus than gave followed by a “constraining” pronoun (it). Again, syntactic flexibility seems to be confounded with a frequency advantage.