Graphomotor Friction and Linguistic Complexity: A Computer Vision Analysis
Sophie Baryalai
21/07/2026
Transcription Bottleneck theory suggests that mechanical writing demands consume working memory, limiting the extent to which cognitive ability is reflected in written output. This study investigates whether graphomotor friction, measured using a computer vision–based framework, is associated with changes in linguistic complexity and output efficiency. It was hypothesized that increased motor friction would reduce linguistic complexity and transcription performance.
Using a within-subjects design (n = 30), three transcription conditions were compared: dominant-hand baseline, non-dominant handwriting, and speech-to-text production. Linguistic complexity was measured using Flesch-Kincaid (FK), Automated Readability Index (ARI), and Type-Token Ratio (TTR).
Results showed a significant decline in output efficiency, with words per minute decreasing from 147.67 to 8.88 (t(29) = 24.11, p < .001), alongside reductions in linguistic complexity, as FK scores fell from 7.55 to 3.99 (p < .001) and ARI from 6.29 to 3.77 (p = .008). These differences are consistent with a transcription bottleneck effect, though condition-level prompt variation is acknowledged as a limitation. A moderate correlation between friction and transcription errors was also observed (R = 0.37, p = .042). High-friction conditions were associated with convergence in performance across participants, suggesting motor constraints may limit the observable expression of underlying cognitive ability.