Association of keyboarding fluency and writing performance in online-delivered assessment

Keyboard-based writing has become increasingly important in everyday living and professional working. However, keyboarding literacy is presumed in most situations; there lack systematic studies on keyboarding and its association with writing processes and qualities of written products. Using the keystroke data of over 1300 middle-school students, we investigated keyboarding fluency, a key subskill of writing transcription measured by an internally reliable metric, and its association with writing process, score, and text characteristics extracted by NLP techniques. We find that: keyboarding fluency is a reliable individual-level property, subject to demographic and linguistic factors of gender, ethnicity, and grade; there exists a threshold keyboarding fluency, which varies by writing purposes and reflects the minimum level of keyboarding skill at which other writing processes are least affected; and more and less fluent typists (differentiated by the threshold) show different text production, transcription, and editing behaviors during writing processes, and distinct linguistic qualities in written products. These findings imply a keyboarding threshold hypothesis and are informative to different stakeholders including practitioners, researchers, test developers, and policy makers. They also provide practical implications for teaching and learning writing in classrooms under a digital environment and help generate personalized feedback on improving writing practice.