Mouse Movements of Motion-Impaired Users: A Submovement Analysis
Faustina Hwang, Simeon Keates, Patrick Langdon, John Clarkson · 2003 · Proceedings of the 6th International ACM SIGACCESS Conference on Computers and Accessibility (Assets '04) · doi:10.1145/1028630.1028649
Summary
This paper presents a detailed submovement analysis of mouse cursor trajectories from six motion-impaired users (four with cerebral palsy of varying types — athetoid, athetoid with spasm, and general — and one with Friedrich's ataxia and tremor) and three able-bodied controls performing a multidirectional point-and-click task. Building on the Optimized Submovement Model from motor performance theory, which posits that rapid aimed movements consist of a sequence of discrete submovements where each corrects the error of the previous one, the authors decompose cursor paths into component submovements separated by pauses. They define and measure several characteristics: pause frequency and duration, verification time (time the cursor is stationary inside the target before clicking), number of submovements, peak submovement speed, and two-dimensional submovement accuracy using novel "coincident error" (along the instantaneous task axis) and "perpendicular error" (deviation from the task axis). The study also introduces the concepts of overshoots (submovements that pass beyond the target) and counterproductive submovements (movements directly away from the target).
Key findings
Mean task completion times ranged from 0.89 seconds for the fastest able-bodied user to 6.88 seconds for the most impaired participant, illustrating the enormous performance variability. The most severely impaired users (PI3, PI7) required approximately five times more submovements per trial than able-bodied users, with over 90% of their trials needing seven or more submovements compared to able-bodied users completing over 90% of trials in fewer than seven. More severely impaired users paused more frequently (PI7 paused 7-13 times per trial at 109-195ms each) and for longer durations, while able-bodied users rarely paused. Verification times for some impaired users (PI3, HA1) averaged 626ms and 615ms — approximately 65% of the able-bodied users' total task completion time, and in extreme cases the verification time alone exceeded the able-bodied total task time. A critical finding was the speed-accuracy relationship: for motion-impaired users, perpendicular error increased with peak submovement speed ("fanning out" pattern), while able-bodied users maintained small errors even at high speeds. Motion-impaired users also showed significantly more target slip-offs (PI7 slipped off in nearly half of trials, up to 5 times per trial, versus 2% for able-bodied CU2) and counterproductive submovements.
Relevance
This paper advances the understanding of motor accessibility from aggregate performance measures (completion time, error rate) to the fine-grained mechanics of how motion-impaired users actually move cursors. The practical implications are substantial: the finding that verification time alone can consume 65% of an able-bodied user's total task time suggests that target confirmation (not just target acquisition) is a critical barrier deserving dedicated assistive solutions — such as multimodal feedback confirming target entry, or dwell-based selection tuned to individual pause patterns. The speed-accuracy correlation unique to impaired users means that speed-enhancing assistive techniques could actually worsen accuracy, requiring careful calibration. The authors propose that interfaces could become "perceptive" — automatically detecting movement patterns like frequent slip-offs or overshoots and adapting assistance accordingly, such as switching to dwell selection or activating gravity wells. This vision of adaptive, individually-tuned cursor assistance remains highly relevant as pointing interactions persist across devices. The detailed individual differences documented here also reinforce that motor impairment is not monolithic — each user's movement profile is distinct, demanding personalised rather than one-size-fits-all solutions.
Tags: motor disability · cursor control · pointing device · target acquisition · cerebral palsy · ataxia · user research · interaction design · motor control · human factors