The Reliability of Fitts's Law as a Movement Model for People with and without Limited Fine Motor Function
Ather Sharif, Victoria Pao, Katharina Reinecke, Jacob O. Wobbrock · 2020 · Proceedings of the 22nd International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2020) · doi:10.1145/3373625.3416999
Summary
This paper investigates a foundational assumption in HCI and assistive technology research: that Fitts's law reliably measures pointing performance. Fitts's law, introduced in 1954, models the relationship between the time to complete a pointing movement and the difficulty of the task (determined by target size and distance). Its key metric, throughput (measured in bits per second), combines speed and accuracy into a single efficiency measure widely used to evaluate pointing devices, interaction techniques, and software. Despite over six decades of use and more than 8,500 citations, no prior work had examined whether Fitts's law's throughput metric produces consistent results when the same person performs the same task across multiple sessions (test-retest reliability), nor whether the model fits the data well for people with limited fine motor function. The researchers conducted a rigorous study with 55 participants — 21 with limited fine motor function in their dominant hand (conditions including cerebral palsy, multiple sclerosis, essential tremor, carpal tunnel syndrome, and others) and 34 without such limitations. Each participant completed a classic 1-D reciprocal pointing task using the ISO 9241-9 standard protocol across two sessions, separated by 4-48 hours. The study used both the traditional A×W experimental design and Guiard's Form × Scale design to ensure robust results, generating 15,000 total trials across all participants.
Key findings
The central finding is that Fitts's law provides low test-retest reliability for both groups, but the problem is worse for people with limited fine motor function. Throughput values between sessions were statistically significantly different (F(1,48)=24.08, p<.001, Cohen's d=1.42), with a mean absolute difference of 0.23 bits/s (2.94%) for participants without limitations and a larger shift for those with limitations (9.16% change between sessions using Guiard's A=256 design). The average throughput for participants with limited fine motor function was 2.88 bits/s — about 39.6% lower than the 4.77 bits/s for those without. Critically, model fitness was also significantly lower for the limited fine motor function group: the average Pearson correlation coefficient was r=.81 (SD=0.09) compared to r=.89 (SD=0.08) for those without limitations — an 8.9% difference. Only 7.5% of sessions for participants with limited fine motor function achieved the r≥.90 threshold traditionally considered a 'good' model fit, compared to 65% for those without. Neither age, fatigue, nor stress level had a significant effect on the results.
Relevance
This paper delivers a critical methodological warning for anyone evaluating assistive technology or accessible input methods. Fitts's law throughput is the standard metric used to compare pointing devices and techniques in accessibility research, yet this study shows it is less reliable and fits data more poorly for the very population most often studied — people with motor impairments. The practical implication is clear: single-session Fitts's law evaluations, which are the norm in HCI research, may produce misleading results, especially for assistive technology. Researchers should conduct multi-session evaluations and report confidence intervals rather than single throughput values. For accessibility practitioners, this also means that published throughput comparisons of assistive pointing devices should be interpreted with caution. The finding that the model itself is a poor fit (r<.90) for 92.5% of sessions with motor-impaired participants raises deeper questions about whether alternative models might better serve the assistive technology community.
Tags: motor accessibility · input methods · research methods · pointing devices · assistive technology · usability testing · HCI
Standards referenced: ISO 9241-9