2018 Sat Poster 6602

Saturday, November 3, 2018 | Poster Session II, Metcalf Small | 3:15pm

A cross-linguistic study on the development of noun case marking in morphologically complex languages
S. Granlund, J. Kolak, V. Vihman, F. Engelmann, B. Ambridge, J. Pine, A. Theakston, E. Lieven

Introduction: We investigate children’s production of nominal case marking in three languages – Estonian, Finnish and Polish. These languages differ in the nature and complexity of case systems, involving suffixes, stem changes and multiple morphological classes. Estonian exhibits suffix variation and variable predictability, with some cases marked solely by stem gradation. Finnish cases are marked by unique suffixes, sometimes requiring stem changes. In Polish, suffixes vary across classes, and homophonous suffixes may mark different cases, resulting in many syncretic forms.

Research has shown that both token frequency and phonological neighbourhood density impact on rates of learning (e.g. Matthews & Theakston, 2006, for English morphology). Morphological transparency also interacts with frequency to affect acquisition (Kjærbæk et al. 2014). However, few studies have examined whether input-based accounts can explain children’s acquisition of complex morphology cross-linguistically. The current study aims to test this.

Method: 132 children aged 32 to 63 months were shown pictures of characters interacting with objects in five to six contexts targeting different case-marked forms. The experimenter named the object with its nominative form and prompted the child to elicit each case-marked form (“This is a book. The fox is waving at…”). Three nouns differing in frequency were chosen from each of 8 (Estonian), 9 (Finnish) or 10 (Polish) declension classes. Classes varied in size, based on CDS corpora counts. We predicted that greater frequency and larger class size in the input would predict greater accuracy. We also predicted a negative interaction, such that the class size effect would be larger for lower token frequency targets.

Results: Overall error rates were low (12-18%). Bayesian analysis with mixed-effects models revealed that in Polish and Estonian, children were less accurate with lower token frequency targets (Polish: mean=0.54, CI¹=0.23–0.84, Prob²=1; Estonian: mean=0.50, CI=0.007–0.96, Prob=0.99). In Estonian and Finnish, children were less accurate with nouns from smaller classes (Estonian: mean=3.90, CI=1.20–6.80, Prob=0.99; Finnish: mean=1.26, CI=0.20–2.28, Prob=0.99). An exploratory analysis was conducted by pooling the data across languages. The pooled results point to strong main effects of class size (mean=1.49, CI=0.30–2.72, Prob=0.99) and token frequency (mean=0.38, CI=0.19–0.57, Prob=1) (Fig. 1), and strong evidence for an interaction between frequency and class size (mean=-0.96, CI=-1.70– -0.20, Prob=0.99), with a greater effect of class size for lower frequency tokens (Fig. 2). We will discuss how language-specific factors – case variance, affix syncretism and stem change complexity – interact with frequency to explain these results.

In conclusion, the results suggest that token frequency and class size affect children’s knowledge of case marking cross-linguistically, but are mediated by typological features of the languages and transparency of the nominal system.

References

Kjærbæk, L., Christensen, R. & Basbøll, H. 2014. Sound structure and input frequency impact on noun plural acquisition: Hypotheses tested on Danish children across different data types. Nordic Journal of Linguistics, 37: 47-86.

Matthews, D.E. & Theakston, A.L. 2006. Errors of Omission in English-Speaking Children’s Production of Plurals and the Past Tense: The Effects of Frequency, Phonology, and Competition. Cognitive Science, 30(6):1027-52.