Identifying reliable discourse measures for second language speakers of English
Presentation at the annual convention of the American Speech-LanguageHearing Association, Philadelphia, PA,
About one-fifth of households in the U. S. speak a language other than English at home (U.S. Census, 2010). Thus, bilingual non-native English speakers are likely to constitute a, significant proportion of most Speech Language Pathologist’s (SLP) clinical caseload. Extensive research has documented the poorer performance of bilingual speakers on language measures of word retrieval and sentence structure (e.g., Gollan et al., 2011; Ivanova & Costa, 2008). The lower language scores are not due a disorder per se, but are attributed to less frequent use of each language (Gollan et al., 2011). This poses unique challenges in accurate diagnosis of language disorders and highlights the need to identify measures that are less likely to be negatively impacted by bilingualism. There is some preliminary evidence that the so-called “bilingual language disadvantage” may not be ubiquitous. For example, bilinguals and monolinguals perform similarly on phonemic fluency (FAS, Portocarrero et al., 2007, but see Sandoval et al., 2010) and action naming (Klassert et al., 2014). However, the findings thus far are equivocal and yet to be thoroughly examined. Given that most (mid-to-high proficiency) bilinguals are conversationally competent, it is possible that measures of language use and informativeness, rather than vocabulary and language structure, are more representative of bilingual language competence. However, there is limited research on informativeness of discourse in bilinguals. One study compared correct information units (CIU) in the first versus second language of Spanish-English bilinguals (Edmonds, 2013).
The concept of CIUs was proposed to capture meaningfulness of discourse irrespective of syntactic errors or word retrieval difficulties (Nicholas and Brookshire, 1993). CIUs are words that are intelligible, informative and meaningful for the discourse context in which they are being used. Another potential measure of extracting informational content is idea density (ID, Turner and Greene, 1977). It is calculated as the number of ideas (represented as propositions) expressed per ten words in a language sample. An algorithm for automatic ID calculation (Brown et al., 2008) is available in the speech analysis software CLAN (MacWhinney, 2000). To our knowledge, CIU and ID have not been compared across bilingual and monolingual speakers as measures of informativeness.
The present study had three main goals. First we sought to report and compare performance of monolingual and bilingual (L2 English) speakers on a variety of lexical, syntactic and informational measures of discourse. Second, we aimed to identify discourse measure(s), particularly those related to informativeness and use,that are less likely to show an L2 English disadvantage. Third, we examined if particular demographic (e.g., education, age of English language acquisition and proficiency self-rating) and language (results on
standardized tests) variables influenced performance on these measures.