Reading difficulties and bilingualism: Can decoding difficulties be identified in a second language?

Abstract

The aim of this thesis was to generate new knowledge about reading difficulties in sequential bilingual adults and to investigate whether decoding difficulties can be successfully identified in a second language (Norwegian). Reading difficulties in adult sequential bilinguals are rarely identified because we lack the knowledge and testing tools to identify this group correctly. This is because appropriate tests (and testers) are often not available in a bilingual’s L1, and that performance in L2 can be influenced by the bilingual’s spoken proficiency in that language. The research involved the collection of questionnaire data about language and reading profiles. In addition, a battery of matched L1 and L2 language tests were developed to compare a bilingual’s performance in each of their languages, and a language-neutral dynamic test of decoding called DOT. The sample tested in this study comprised adult sequential bilinguals with English, Polish or Somali as L1 and Norwegian as L2 (N=80). The study examines whether various factors (socioeconomic conditions, multilingualism, Norwegian skills etc.) influence the results of decoding when assessed in L1 and L2 respectively. In addition, it investigates whether there is a correlation between L1 and L2 testing, whether spoken language proficiency in Norwegian affects the results in L2, and critically whether spoken language proficiency in Norwegian moderates the association between L1 and L2 testing. Finally, it is investigated whether it is possible to predict the group with low decoding skills (in L1) based on L2 test performance and DOT. A Principal Component Analysis (PCA) was performed on the questionnaire data. The results showed that six components from a PCA predicted decoding in some test results in both L1 and L2. L1 word reading performance was significantly predicted by components related to socioeconomic background (p< .006) and reading difficulties (p< .001). L2 word reading performance was significantly predicted by components related to "Norwegian engagement" (p< .003), and socioeconomic background (p< .05). L1 pseudoword reading performance was predicted by the L1 proficiency component (p< .007), as well as by Multilingualism (p< .04), and Socioeconomic background (p< .02). Interestingly, performance in L2 pseudoword reading was not significantly predicted by any of the six components from the PCA. …