Network Structures in L2 Oral Performance: A Learner Corpus Study

抄録

　　Complexity, accuracy, and fluency as dimensions of L2 performance, proficiency, and development have been measured through a small number of measures that researchers arbitrarily choose. L2 performance is never a simplistic structure that can be captured through the reduction of one or two elements. Reductionism practice has overlooked perspectives of how each measure interacts and changes. The current study attempts to describe dynamic relations between various linguistic features, investigate how each element is intertwined, and explore which linguistic features play a central role in L2 oral performance, based on the complexity theory using network analysis. Spoken data were drawn from the NICT JLT Corpus, which comprises English transcriptions uttered by Japanese learners, and were attached their individual proficiency. Two networks of each proficiency group (i.e., low and high) were delineated, and centrality measures were calculated. The results reveal that the network structures in L2 oral performance change to be dependent on the learner's proficiency. Furthermore, linguistic features that play a central role in the network structure also change with learner's proficiency. These results are highly suggestive toward L2 oral performance and the relationship among various linguistic features. The significance of the network analysis is also discussed.