CHAPTER 5: DISCUSSION AND IMPLICATIONS
5.1 Key Findings
5.1.3 L1 Summary and L2 Summary Task Inference Processes
Research question 3a asked if the inferences generated by skilled, adult readers in L1 differ from the inferences generated in L2 for a summary task. Statistically, the overall differences in the percentages of reported inferences by skilled, adult readers in L1 summary task as compared to L2 summary task are significant (z = 2.24, p ≤ .05, r = 0.709). Only one type of inference, however, was statistically significant: evaluative background knowledge elaborative inferences (z = 2.19, p ≤ .05, r = 0.69), and no other inference types were approaching significance. Descriptively speaking, participants reported more evaluative inferences, both text-based and background knowledge-based, while reading in the L1 summary task than in the L2 summary task. Bridging inference patterns in L2 summary were very similar to L1 summary; readers preferred context to structure analysis or background knowledge to resolve comprehension breakdowns. However, they were less likely in L2 than L1 to strategically skip an unknown word if the first inference failed. The remaining inference processes were similar and seem to have transferred from L1 to L2 for a summary task with expository texts.
The current study produced some similarities and differences to prior research that looked at inferencing process in L1 and L2. Horiba’s (2000) second experiment used expository texts and instructed participants to read ‘freely’, or for a Rauding purpose, and to read for coherence, which is
similar to a summary task. The present study revealed more background knowledge associations and evaluations in L1 summary than L2; whereas, in Horiba’s (2000) study, L1 readers reported more background knowledge-based associations and evaluations in the read freely condition. Additionally, participants in the current study tended to apply background knowledge to the text more often in L1,
which is similar to Horiba’s (2000) findings that when reading an expository text in L1, skilled readers
evaluated the text based on background knowledge and also reported background knowledge-based associations at much higher instances in L1 than in L2. Further, participants in this study reported more text-based bridging and text-based associations in L2 summary, but more text-based evaluations and predictions in L1 summary. This finding only partially supports Horiba’s (2000) results which showed that in L2, readers employed more text-based inferences such as text-based bridging inferences and text-based evaluations than in L1 (Horiba, 2000). Lastly, in Horiba’s (2000) read-for-coherence condition, L1 readers reported far more text-based backward, predictive, and associative inferences than L2 readers who reported more bridging inferences and few elaborative inferences. NSs also seemed to refrain from background knowledge associations and evaluations while reading for the coherence task with expository texts (Horiba, 2000). Neither of those patterns were found with advanced readers in the current study.
Some of the aforementioned differences may be due to language proficiency. Horiba’s (2000) participants were NNSs of Japanese in their fourth year of Japanese study at a university in the U.S., but no measure of proficiency was given, nor was it mentioned if they studied through the medium of Japanese. The expository texts were fairly short (24 and 33 sentences) newspaper articles. If their proficiency were intermediate to upper-intermediate, they may have been short-circuited and focused on lower-level processes such as decoding and lexical issues which did not allow them to allocate cognitive resources to higher-level processes. The L2 readers in the present study had an advanced proficiency and most likely had fewer lower-level processing problems which allowed for cognitive resources to be allocated to higher-level processes such as elaborations. Further, as the texts were in a Japanese rhetorical structure (Horiba, 2000), perhaps the rhetorical structure differs significantly from
English and German rhetorical structures and does not foster elaborative associations or evaluations. That said, a contrastive rhetorical analysis is beyond the scope of this paper.
There are language effects when reading a long, expository text for a summary task. When reading in L2, readers allocated more cognitive resources to bridging inferences, were more likely to employ a second bridging attempt after a failed attempt, and were less likely to skip an unknown word than when reading in L1. Further, when reading in L1, advanced readers applied their background knowledge, especially to evaluate, to the text far more often than when reading in L2. Also, the evaluations made when reading for an L1 summary tended to denote textual content as pro, con, or viable; whereas the evaluations in L2 tended to relate to the importance of the information specific to the reading purpose. Predictions, although not very common, occurred across both languages, but there were differences from L1 to L2. Advanced L1 readers preferred to predict using textual features, but L2 readers used the text and background knowledge at similar percentages. Perhaps this is due to familiarity with rhetorical structure in L1. Although the participants studied through English medium and read a lot of academic texts in English, they would have been exposed to many more German texts throughout their lives, and it would stand to reason that they be more comfortable with German
rhetorical structure.