Incremental processing in head-final child language: online comprehension of relative clauses in Turkish-speaking children and adults

Abstract

The present study investigates the parsing of pre-nominal relative clauses (RCs) in children for the first time with a real-time methodology that reveals moment-to-moment processing patterns as the sentence unfolds. A self-paced listening experiment with Turkish-speaking children (aged 5–8) and adults showed that both groups display a sign of processing cost both in subject and object RCs at different points through the flow of the utterance when integrating the cues that are uninformative (i.e., ambiguous in function) and that are structurally and probabilistically unexpected. Both groups show a processing facilitation as soon as the morphosyntactic dependencies are completed and parse the unbounded dependencies rapidly using the morphosyntactic cues rather than waiting for the clause-final filler. These findings show that five-year-old children show similar patterns to adults in processing the morphosyntactic cues incrementally and in forming expectations about the rest of the utterance on the basis of the probabilistic model of their language.

Keywords:

• pre-nominal relative clauses
• Turkish
• child language processing
• head-final languages
• filler-gap dependencies
• morphosyntactic processing
• expectation-based parsing

Acknowledgements

Thanks to Cem Bozşahin, Vicky Chondrogianni, Belma Haznedar, Barış Kahraman, Jaklin Kornfilt, Umut Özge, Sumru Özsoy, Mark Steedman and Patrick Sturt for helpful discussion; Ökmen Yıldırım for testing the adult participants; Akman Yuva, İşçi Blokları İlköğretim Okulu for permission and space for data collection; families, children and METU students who participated in the study.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This research was supported by the ESRC [grant number RES-061-23-0137] to Theo Marinis.

Notes

1. Few studies have used the self-paced listening paradigm to evaluate the expectation-based models due mostly to the difficulty controlling for the naturalness of the prosodic structure of the sentences. Our study shows that the paradigm yields reliable results once the segments correspond to the natural prosodic breaks.

2. The genitive case creates an expectation for a head with a possessive agreement morphology. Structure-wise, it may be more likely to be followed by a possessed NP (see Appendix), but in our experiment, the adverbial prior to the embedded verb might be increasing the expectation for a verbal head. Regardless of the type of the head, we predict that the genitive case should create anticipation for an upcoming morphosyntactic dependency.

3. Headless RCs are formed via the omission of the modified NP and can be translated as ‘the one that’, as shown in (i).

(i) Headless SRC

[_RCGoril-iit-en]     fil-i öp-tü.

[_RC gorilla-Accpush-SRel]  elephant-Acc kiss-Past.3sg

‘The one that pushed the gorilla kissed the elephant’.

4. Data trimming was done with a programme in Python, we thank Umut Özge for the programme.

5. Analysing the comprehension questions with these factors might be problematic if the means are outside of the range of proportions of .3 to .7, as detailed in Jaeger (2008). However, since the interpretation of the results does not rest on the performance in the comprehension questions, we do not conduct mixed-logit models here.

6. The pattern in this segment does not accord with the Parallel Function Hypothesis (Sheldon, 1974), which would predict an RC-Type by RC-Role interaction in this segment. We omit further discussion here due to space limitations.

7. We thank Matt Wagers for suggesting this possibility.

Additional information

Funding

This research was supported by the ESRC [grant number RES-061-23-0137] to Theo Marinis.