Measuring lexical alignment during L2 chat interaction: An eye-tracking study

Learner interaction via written synchronous computer mediated communication (SCMC) in their second language (L2) may be facilitative of L2 development (Smith, 2012), as such interaction heightens the salience of specific aspects of the input learners receive, thus increasing the likelihood that aspects of this input are noticed (Schmidt, 2001). This heightened salience of input is afforded by the permanence of the written message on the screen, which allows learners more time for processing incoming messages and monitoring their own output. This exploratory study aims at establishing whether this heightened salience during SCMC supports lexical alignment. Using eye tracking technology, we explore what L2 users seem to notice, attend to and align with in the linguistic input from their L2 interlocutor. Six advanced L2 users of English (TESOL MA students) interacted in dyads via SCMC over six-weeks. They chatted each week with another student and discussed how to finish a partially completed academic abstract. Participants’ task performances were screen-recorded and their eye movements captured. Corpus methods were used to analyze chat logs for lexical overlap of three or more consecutive words (n-grams). Potential sources for lexical alignment were then marked as areas of interest (AoI’s) and various eye-gaze measures were calculated for these AoI’s and for baseline text. Comparisons between eye-gaze measures for potentially aligned text and baseline text were made to identify those n-grams that received more visual attention than baseline text (i.e., noticed text). Analysis revealed that a limited amount of potentially aligned received heightened overt visual attention. We argue that most instances of lexical overlap were likely attributable to processes other than strategic alignment. Qualitative explorations show that SCMC partners do make use of one another’s input during task-based SCMC in a way that manifests in their written output. However, it is much less than chat transcripts may suggest