Author

Yu TianFollow

Author ORCID Identifier

0000-0003-3009-3976

Date of Award

8-8-2023

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Applied Linguistics and English as a Second Language

First Advisor

Scott Crossley

Second Advisor

Sara Cushing

Third Advisor

Diane Belcher

Fourth Advisor

Luuk Van Waes

Abstract

Recent research using keystroke logging to investigate argumentative writing in assessment settings has documented associations between features of the writing process and writing quality. However, more research is needed to examine the direct links between the dynamics of the writing process and argument development. This dissertation investigated the interplay between behavioral patterns in the writing process as manifested in keystroke activities (i.e., pause, revision, burst, process variance), the formulation of argument elements (i.e., positions, claims, evidence, counterclaims, and rebuttals), and holistic scores of writing quality in L1 and L2 adult writers. Four hundred adult writers were recruited, including both L1 (n = 200) and L2 (n = 200) English speakers. Each wrote an independent persuasive essay on a randomly assigned SAT writing prompt. Their keystroke activities during text production were recorded. The essays were annotated for argument elements and human-rated for holistic writing quality. For each writer, keystroke measures were developed for both the entire writing process and the formulation of each argument element. These measures were used in multinomial mixed effects logistic regression models to distinguish between argument element categories and in multiple linear regression models to predict writing quality. Results suggest that adult writers behaved differently in P-burst length, pause duration, and proportion of deletions when formulating different argument elements. Distinct keystroke patterns were also observed across L1 and L2 writer groups in formulating claims (compared to evidence) as well as counterclaims and rebuttals (compared to all other elements), indicating different cognitive loads the two groups underwent and different strategies they took in written argumentation. Last, for both L1 and L2 writers, higher writing scores were related to shorter pauses in general, shorter between-word pauses, lower proportion of deletions, higher proportion of insertions, and less process variance. This dissertation sheds light on the links between L1 and L2 adult writers' keystroke patterns in the writing process, the cognitive activities they engaged in, and the holistic quality of their written products. The dissertation affords important implications for writing research, writing instruction, and the development of intelligent writing support tools.

DOI

https://doi.org/10.57709/35889705

File Upload Confirmation

1

Available for download on Wednesday, July 30, 2025

Share

COinS