90 CHAPTER 6 may help learners to be more uent and complex in their L2. Building on these initial ndings, it is important to reach a better understanding of the exact link between learner exposure to the target language and chunk use as well as how chunk use is facilitatory in enhancing L2 uency and complexity. However, most chunk research has been done on the English language and especially in the Netherlands, so it is di cult to control for extramural exposure. French L2 is interesting because it is a language frequently learned in school settings and in the Netherlands, extramural exposure to French is rare. Also, replicating ndings in languages other than English is likely to enhance the validity of chunk theory and its applicability to L2 pedagogy. THIS STUDY is study is based on the study in the previous chapter, in which the written production of 56 French learners enrolled in SB and DUB programs at a Dutch secondary school a er six years of instruction was examined. e two groups scored the same when scored holistically, but when looking at various complexity, accuracy, and uency measures, the DUB students were more complex and uent than their SB counterparts. Here, the same data are used to examine whether the DUB students use more chunks and, if so, which kinds. In addition, we will investigate to what extent chunk use can be directly related to some CAF measures. Two research questions form the basis of this study: 1. To what extent do the SB and DUB learners di er in their use of chunks? 2. To what extent does chunk use correlate with CAF measures, most notably complexity and uency? Based on the literature, the hypothesis for the rst research question is that the DUB students will use relatively more chunks than the SB counterparts. In addition, it is hypothesized that DUB students will be able to use longer chunks than their SB counterparts. Based on previous ndings, the hypothesis for the second research question is that the use of chunks may lead to greater uency and that certain types of chunks (e.g. partially schematic ones) may lead to greater complexity.