Abstract:
Much concern has been raised over the issue of whether administering concurrent verbal reports (CVRs) alters the very cognitive processes that the reports are supposed to represent and keep intact (the reactivity issue) and whether they reflect the processes completely and accurately (the veridicality issue). This thesis reports on an empirical study designed to address both issues. It examines the reactive effects of concurrent verbal reporting (CVR) or thinking aloud (TA) in a second language (L2) narrative writing task, in particular, the mediation of L2 writing proficiency, working memory capacity (WMC), as well as task type, in such effects. Eight-five Chinese sophomore participants whose awareness of lexical diversity in writing in English had been raised through instruction engaged in two WMC tests, then wrote a baseline outline-given narrative silently, and then completed a main writing task under either a no-think-aloud (NTA) condition or a think-aloud (TA) condition. The TA group were also asked to provide immediate retrospective verbal reports (RVRs) on their revisions, in addition to CVRs, and answer questions which elicited their perceived reactivity and veridicality. Participants’ compositions were analysed, in terms of 20 measures of fluency, complexity, accuracy, organization, and content, to find out group differences and interactional effects. The results showed that TA caused reactivity because it significantly increased dysfluencies, worsened organization, and most noticeably, impaired lexical diversity. Reactivity was further found to be significantly different among L2 learner writers at different baseline levels of Speed Two (i.e., total number of words produced per minute), General Complexity, Correct Verb Use, or Length Two (i.e., total number of words produced) on the corresponding measure. Reactivity was also found to distinguish different WMC Average or WMC One (i.e., the operation span) learner writers on Organization, and differentiate different WMC Two (i.e., the reading span) learner writers on Length Three (i.e., total number of syllables in the final composition), Length Four (i.e., total number of syllables produced), and Lexical Diversity. The comparison of CVRs and RVRs indicated that participants ignored reporting transitory metacognitive thinking and supposedly automatized behaviours, among others, making their CVRs incomplete. Their perceived reactivity appeared to be more serious in terms of the intensity and range of effects they felt. Their perceptions on incompleteness corroborated the findings from the comparison of CVRs and RVRs, but they stood by the accuracy of their CVRs. Kellogg’s (1996) model of writing processes in relation to Baddeley’s (1986) accounts on working memory (WM), situated within the information-processing framework, on which Ericsson and Simon (1993) based their argument for the validity of CVRs, was used to explain the reactivity and veridicality problems found in this project.