Using Distributional Statistics to Acquire Morphophonological Alternations: Evidence from Production and Perception.

Helen Buckler,Paula Fikkert

doi:10.3389/fpsyg.2016.00540

Helen Buckler, Paula Fikkert

Open Access

https://doi.org/10.3389/fpsyg.2016.00540

Copy DOI

Abstract

Morphophonological alternations, such as the voicing alternation that arises in a morphological paradigm due to final-devoicing in Dutch, are notoriously difficult for children to acquire. This has previously been attributed to their unpredictability. In fact, the presence or absence of a voicing alternation is partly predictable if the phonological context of the word is taken into account, and adults have been shown to use this information (Ernestus and Baayen, 2003). This study investigates whether voicing alternations are predictable from the child’s input, and whether children can make use of this information. A corpus study of child-directed speech establishes that the likelihood of a stem-final obstruent alternating is somewhat predictable on the basis of the phonological properties of the stem. In Experiment 1 Dutch 3-year-olds’ production accuracy in a plural-elicitation task is shown to be sensitive to the distributional statistics. However, distributional properties do not play a role in children’s sensitivity to mispronunciations of voicing in a Preferential Looking Task in Experiment 2.

Highlights

Non-allophonic morphophonological alternations, such as the Dutch voicing alternation, are difficult for children to acquire, in part because they are unpredictable (Van Wijk, 2007; Van de Vijver and BaerHenney, 2011; Zamuner et al, 2011)
Children were more accurate in their productions of word-medial voicing in plural forms when there was no voicing alternation between the stem and plural form
As predicted on the basis of adult experimental data and child-directed speech corpus data, children produced a [d] more accurately in words where it is preceded by a sonorant than a vowel, indicating that they are sensitive to distributional statistics in determining whether a stem-final [t] alternates in the plural or not

Summary

Introduction

Non-allophonic morphophonological alternations, such as the Dutch voicing alternation (e.g., the final [t] of [bεt] bed ‘bed’ is voiced in its plural form, bedden [bεd@n]), are difficult for children to acquire, in part because they are unpredictable (Van Wijk, 2007; Van de Vijver and BaerHenney, 2011; Zamuner et al, 2011). In this paper we use corpus data to investigate whether there are regularities in the distribution of the voicing alternation in Dutch child-directed speech, that is, whether voicing alternations are more or less frequent in a given phonological context, and experimental data to test whether these regularities play a role in children’s ability to establish accurate lexical representations of voicing alternations in morphological paradigms. Both production and perception data are presented.

Objectives

Methods

Results

Conclusion