Input Accessibility: A Large Dataset and Summary Analysis of Age, Motor Ability and Input Performance
Leah Findlater, Lotus Zhang · 2020 · Proceedings of the 22nd International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2020) · doi:10.1145/3373625.3417031
Summary
This University of Washington study makes two key contributions to understanding how age and motor ability affect computer input performance: a large open dataset of mouse and touchscreen input traces from over 700 participants, and a summary analysis of mouse input data from 318 participants aged 18-83, of whom 53% reported a motor impairment. The work addresses a significant limitation in prior research, which has typically relied on small samples of 20-40 participants and binned them into crude categories like "younger" versus "older." The dataset was collected from August 2015 to September 2016 using a web-based testbed that guided participants through four fundamental input tasks based on the ISO 9241-9 Fitts' law standard: pointing (clicking/tapping a target), dragging (moving a target to a destination), crossing (moving a cursor across a target), and steering (navigating through a column). Tasks varied in difficulty through five amplitude-by-width combinations yielding an Index of Difficulty range of 1.9-4.1. The full dataset includes 735 unique participants (367 mouse, 321 touchscreen, 47 both), with 172,285 valid trials after removing spatial outliers. Participants were recruited through campus email lists, social media, Amazon Mechanical Turk, and Cint — a panel specifically targeted to recruit older participants with motor impairments. The 318-participant mouse analysis subset included 170 with motor impairments (reporting difficulties including low strength at 50%, difficulty gripping at 48.2%, rapid fatigue at 32.9%, limited range of motion at 31.8%, lack of sensation at 31.8%, lack of coordination at 31.3%, and tremor at 31.2%) and 148 without.
Key findings
The analysis using linear mixed effects models revealed that both age and motor impairment significantly affected trial completion time across all four tasks (p < .05 for all, except motor impairment on crossing which showed only a trend at p = .053). For trial errors, motor impairment was significant across all four tasks (p < .05), but age was only significant for the crossing task. A critical finding is what the error data reveals about strategy: error rates remained flat or actually decreased with age, confirming that older adults prioritize accuracy over speed — a well-documented speed-accuracy tradeoff that means declining speed with age does not necessarily indicate declining capability. The scatterplots showing individual participant performance tell an even more important story than the statistical models. While general trends exist, there is enormous overlap between participants with and without motor impairments at every age, and the relationship between age and performance is continuous and gradual rather than a sharp divide between "young" and "old." Analysis of the extreme deciles (fastest and slowest 32 participants each) showed that the fastest users were mostly clustered in the 18-40 age range but included some participants up to age 62 and some with motor impairments. The slowest users showed an even wider age spread and included five participants without a reported motor impairment. This demonstrates that knowing a user's age and self-reported motor ability is insufficient to predict their individual input performance.
Relevance
This paper is directly relevant to anyone designing accessible interfaces or making assumptions about user capabilities based on demographics. The central message — that age and motor ability are continuous variables with enormous individual variation, not binary categories — challenges the common practice of designing for "older users" or "users with motor impairments" as homogeneous groups. For accessibility practitioners, this means that adaptive interfaces should respond to actual observed input performance rather than relying on user profiles or self-reported abilities. The open dataset (available on GitHub) is a valuable resource for researchers studying input accessibility, enabling analyses far beyond what the paper reports — including detailed movement traces that could inform adaptive pointing algorithms, personalized target sizing, or predictive models of input difficulty. The finding that older adults strategically trade speed for accuracy is important for evaluation: measuring only task completion time would unfairly penalize users who are deliberately being careful. The study's limitations — skewing toward mild-to-moderate motor impairments and ages under 85 — also highlight the continued need for inclusive research that captures the full spectrum of human motor abilities.
Tags: motor impairment · mouse input · touchscreen · older adults · physical accessibility · Fitts law · input performance · dataset · ageing
Standards referenced: ISO 9241-9