Rethinking assessment measures of phonological development and their application in bilingual acquisition

Abstract

This article discusses the measurements for phonological and prosodic development put forward in Ingram [Ingram, D. (2002). The measurement of whole-word productions. Journal of Child Language, 29, 713–733.], while at the same time expanding the crucial measures, phonological mean length of utterance (PMLU) and proportion of whole-word proximity (PWP). The goal of the expansion is to accommodate a wider set of phenomena, specifically those related to bilingual acquisition of languages with different categories (e.g. closed syllables in German and open syllables in Spanish). Data from three monolingual Spanish children and from three bilingual German--Spanish children are presented as illustration of the modified measurements: expanded PMLU of features (ePMLU-F), expanded PMLU of syllables (ePMLU-S) and expanded PWP (ePWP). By means of measuring both features and prosodic positions, the expanded measurements do better justice to the various aspects of child phonology. Nonetheless, an important goal of this article is to stimulate discussion in order to bring our state-of-the-art closer to descriptive adequacy.

• Bilingual acquisition
• closed syllables
• features
• German phonology
• prosody
• Spanish phonology
• weak language
• whole-word measurements

Acknowledgements

We would like to thank the German Research Foundation (DFG) and the University of Hamburg for their support. Our special thanks to Martin Rakow, research assistant of the project, who contributed many of the calculations carried out to illustrate the expanded measurements.

Notes

¹To trace back the birth and development of MLU, see Brown (1973: 53–54); for its first applications to phonological research, see Ingram (1989); a historical overview is provided in Taelman, Durieux and Gillis (2005: 4).

²López-Ornat has presented this measurement orally in various occasions; it has been applied to the data of a child acquiring English in Aoyama, Peters and Winchester (2010).

³Most Romance languages (including Spanish) abide by the syllabic trochee, whereas German and the other Germanic languages abide by moraic trochees. We will see later on that moras may be relevant when measuring development in German.

⁴The PAIDUS corpus stems from the BIDS and PAIDUS projects, which were cross-linguistic projects directed by Conxita Lleó and supported by the DFG (German Science Foundation) and the DAAD (Acciones Integradas). It was carried out simultaneously in Hamburg and Madrid with the goal of longitudinally analysing the phonological development of monolingual German and Spanish children. The PhonBLA corpus stems from the B3/E3 project of the Research Centre on Multilingualism of the University of Hamburg. The project was also directed by Conxita Lleó and supported by the DFG and the University of Hamburg. Its goal was to analyse the phonological and early morphological acquisition of German and Spanish by bilingual children. For more details, see Lleó (2012).

⁵Carmen López Ojea’s corpus in CHILDES pertaining to child-directed speech yields 21.57% of trisyllabic words (quoted from Saceda, 2005). Despite the frequent usage of diminutives with endings like –ín, –ito, etc., in child-directed speech, there hardly is any difference with respect to the proportion of trisyllables in the adult target language.

⁶As far as vowels are concerned, Ingram (2002: 716) proposed not to take vowels into consideration, due to the “low reliability” of vowel transcriptions. However, Ingram (2002: 723) also suggests some typological flexibility, by which tone-bearing vocalic units can be taken into account in the computation in a language such as Cantonese. Our expanded ePMLU-F criterion also dismisses vowels.

⁷Theoretically, one might pinpoint any node of the Feature Geometry tree as a possible candidate for the evaluation procedure. For practical reasons, we sort out three basic features, which additionally allow us cross-linguistic comparison in a not too troublesome way.

⁸The expanded measurements, especially the ePMLU-F are undoubtedly more complex and they take more time to calculate than the original ones. However, an algorithm that checks features as to whether they are target-like or not in the child’s productions would be straightforward to develop.