Building Learner Corpora to analyse L2 oral interlanguage in English-Mandarin bilinguals
Clare Wright
University of Reading
Learner corpora in SLA have been established over several decades, sometimes building on L1 corpora design (as in the CHILDES TALKBANK), or created for specific research questions or participants (such as the ICLE). However, many problems remain in creating robust, valid and accessible datasets across an appropriate breadth of languages, informed by SLA theories, especially in relation to oral interlanguage (notable exceptions are the FLLOC and SPLLOC corpora, created by researchers at Southampton and Newcastle Universities among others).
This talk illustrates these issues in relation to problems faced in a small-scale SLA research project on development of L2 Mandarin during a period of Study Abroad, in which we needed to transfer oral data, gathered across 4 monologic and dialogic tasks, into a researchable form to track lexical and grammatical development, as well as measure changes in fluency, using CLAN and PRAAT. Focusing here on the fluency data, we found that across tasks, measures generally showed clear improvement: in particular total mean output, pronunciation accuracy and speech rate increased significantly (p <.01), supporting the general claim that study abroad favours fluency development (Freed et al. 2004). However, other measures such as pausing, hesitation and mean length of run differed between task, and were very variable across the group, highlighting individual differences in processing the planned vs unplanned task demands. Even within task, high individual variability created difficulty in clarifying precisely how to assess acquisition in measures of oral production.
Our study, and others emerging from the Centre for Literacy and Multilingualism at Reading analysing developments in L2 fluency, highlights the value of collecting detailed longitudinal SLA-motivated datasets of language learners moving between instructed and immersed contexts (Du 2013). However, the current lack of standardised linguistically-informed measures, especially in L2 Mandarin, create methodological challenges for creating longitudinal datasets, e.g. for researching monologic vs dialogic interaction (Pallotti 2009). We believe insights from corpus research would be invaluable in addressing some of the theoretical and methodological issues involved in analysing systematic L2 linguistic development, particularly in L2 Mandarin.
References:
Du, H. (2013). The development of Chinese fluency during study abroad in China. The Modern Language Journal 97(1): 131-143.
Freed, B., Segalowitz, N. & Dewey, D. (2004). Context of learning and second language fluency in French. Studies in Second Language Acquisition 26 (2): 275-301.
Pallotti, G. (2009). Complexity, Accuracy, Fluency: Defining, refining and differentiating constructs. Applied Linguistics 30: 590-601.