Talia Isaacs: Why modelling pronunciation descriptors in an empirically-based universal framework is such a challenge
Why modelling pronunciation descriptors in an empirically-based universal framework (or even just sticking to one target language) is such a challenge
Pronunciation is increasingly acknowledged as an important component of second language (L2) oral proficiency, particularly because of its role in impeding L2 speakers' ability to be comprehensible or easily understood by their interlocutors. However, pronunciation has proved difficult to model in assessment instruments, behaving differently than other linguistic components when it is scaled from a statistical standpoint. This could be due to the tendency for pronunciation to become fossilized and subject to strong crosslinguistic effects (e.g., Milanovic et al., 1996), or to difficulties that raters may experience in using "canned" pronunciation descriptors that may be problematic either for substantive reasons, or because they are underspecified (Isaacs et al., 2015). Although on the one hand, there has been a move within productive skills assessment for rating scales to be task- and context-specific (e.g., Fulcher et al., 2011), on the other, efforts have been made to develop and validate one-size-fits-all rating scales that cut across all world languages and are not specific to any particular target language or task performance (e.g., Council of Europe, 2001). The latter goal is clearly useful as a ruler or metric to compare L2 performances across world target languages. However, in the case of pronunciation in particular, such descriptors are likely to be confined to the generic, limiting their usability with L2 learners in a a way that could be informative for diagnostic assessment purposes, for example. There is a need to develop an empirical basis for modelling pronunciation descriptors for pronunciation that improve on the quality of the intuitively-developed descriptors used in most L2 oral proficiency scales. Starting with just one target language would seem to a realistic next step, in a way that the English Profiles project has done for some linguistic components (e.g., functional aspects, vocabulary) but that has not yet extended to pronunciation (e.g., Hawkins & Filipovic, 2012). This presentation will overview the challenges in grappling with these issues, presenting data from classroom teachers and accredited examiners to characterize the nature of the problem in developing and validating fit-for-purpose pronunciation descriptors for English as an L2.
Liked this event? Spread the word :