The process-disruption hypothesis: how spelling and typing skill affects written composition process and product

Abstract

This study investigates the possibility that lack of fluency in spelling and/or typing disrupts writing processes in such a way as to cause damage to the substance (content and structure) of the resulting text. 101 children (mean age 11 years 10 months), writing in a relatively shallow orthography (Norwegian), composed argumentative essays using a simple text editor that provided accurate timing for each keystroke. Production fluency was assessed in terms of both within-word and word-initial interkey intervals and pause counts. We also assessed the substantive quality of completed texts. Students also performed tasks in which we recorded time to pressing keyboard keys in response to spoken letter names (a keyboard knowledge measure), response time and interkey intervals when spelling single, spoken words (spelling fluency), and interkey intervals when typing a simple sentence from memory (transcription fluency). Analysis by piecewise structural equation modelling gave clear evidence that all three of these measures predict fluency when composing full text. Students with longer mid-word interkey intervals when composing full text tended to produce texts with slightly weaker theme development. However, we found no other effects of composition fluency measures on measures of the substantive quality of the completed text. Our findings did not, therefore, provide support for the process-disruption hypothesis, at least in the context of upper-primary students writing in a shallow orthography.

Fig. 1

Path model for predictors of composition fluency

Fig. 2

Estimated effects of typing and spelling fluency measures on the probability that an interkey interval is sufficiently long to be considered a disfluency, with 95% CI, showing just those effects appearing in best-fit models

Fig. A1

Distributions (Gaussian kernel density estimates) of keystroke latencies (interkeystroke intervals). The dark region in the positive tail represents latencies identified in our analyses as disfluencies (mid-word, > 1 s; word-initial, > 2 s)

Fig. A2

Distributions (Gaussian kernel density estimates) for by-subject medians of composition fluency measures