Wim Gombert

CHAPTER 6. Chunks in writing 99 Table 17 presents the correlations between chunk coverage and the CAF measures under consideration. e Pearson r correlation analyses demonstrate that there were several signi cant correlations between Chunk Coverage and CAF measures, notably in terms of uency and complexity measures. Chunk Coverage and Text Length were signi cantly positively correlated, r(54)=0.28, p=0.038, two-tailed, 95% CI [0.02, 0.50]. Moreover, Average Sentence Length was signi cantly positively correlated with chunk coverage, r(54)=0.27, p=0.045, two-tailed, 95% CI [0.01, 0.49]. In terms of the other measures, no signi cant correlations were found. However, in all cases, the e ect sizes can be considered rather low (cf. Plonsky & Oswald, 2014). TABLE 17. Chunk coverage and CAF measures correlation coefficients Correlation Coefficient (r) p-value Effect size (r2) Complexity Chunk Coverage and Guiraud Index Chunk Coverage and Average Sentence Length -0.17 0.27 0.206 0.045* 0.03 0.07 Accuracy Chunk Coverage and Subject-Verb agreement Chunk Coverage and Determiner-Noun agreement 0.004 -0.04 0.979 0.798 <0.001 0.002 Fluency Chunk Coverage and Text Length 0.28 0.039* 0.08 * indicates a significant correlation, p<.05 DISCUSSION AND CONCLUSION In general, the amount of L2 exposure may be related to L2 pro ciency (Ellis & Cadierno, 2009). Speci cally, L2 exposure and level of pro ciency may be related to the use of L2 chunks. e latter has been demonstrated by Verspoor et al. (2012), Gustafsson and Verspoor (2017) and Piggott (2019) for English. e current paper examined whether this also held true for French by comparing the written texts of students who had followed a traditional low L2 exposure program and a relatively recent high L2 exposure program. In the previous chapter, we showed that the programs led to texts that were rated rather similarly in overall pro ciency scores based on human ratings. However, on some analytical measures such as Text Length ( uency) and Average Sentence Length and the Guiraud index (complexity) the DUB students performed better. Using the same texts, the current study focused on the use of chunks to see if there were di erences in chunk usage. In line with earlier work on chunks (cf. Myles et al., 1998; Piggott, 2019),