Assessing L2 phonemic acquisition: A normalization-independent method?

This paper aims at addressing the issues that emerge when attempting to analyse the acquisition of L2 vowels: studies usually embed vowels in the same /hVd/ template (e.g. Hillenbrand [14], Ferragne [10], Clopper [9]) in order to reduce coarticulatory effects, and the number of occurrences of each vowel is carefully controlled in order to comply with normalization constraints (e.g. Lobanov [17]). However, such methods make it difficult to test the predictions of SLA models that base phonemic acquisition on phonemic parameters only (Flege [11], Best [2],[3]). This study investigates the development of advanced French learners for the acquisition of phonemic contrasts (/I-i:/ and /U-u:/) in the longitudinal DIDEROT LONGDALE [13] corpus. 15 speakers (12 female & 3 male) were recorded in spontaneous interviews over a period of two years. To test whether the acquisitions of the /I-i:/ and /Uu:/ constrasts are similar, a metric, the Ratio of the (contrast) DIstance to the vowel space Convex Hull (RaDiCHull, /­ræd.Ik."h2l/), is explored with different normalizing procedures to measure learner input as compared to native speakers of English (the reference points are from Hillenbrand [14] for values in Hz, and from Clopper [9] for values in Bark).