An Epidemiology-inspired Large-scale Analysis of Android App Accessibility
Anne Spencer Ross, Xiaoyi Zhang, James Fogarty, Jacob O. Wobbrock · 2020 · ACM Transactions on Accessible Computing · doi:10.1145/3348797
Summary
This paper presents the largest analysis of Android app accessibility to date, testing 9,999 free apps from the Rico repository for seven distinct accessibility barriers. Drawing on epidemiological methods, the authors frame accessibility barriers as "diseases" affecting the app population, enabling systematic analysis of prevalence, contributing factors, and potential interventions. The seven barriers tested include: few TalkBack-focusable elements (apps with 0-1 focusable elements per screen), missing labels, duplicate labels, uninformative labels, editable TextViews with contentDescription, fully overlapping clickable elements, and undersized elements (below the 48dp × 48dp minimum). The research identifies which element classes are most prone to specific barriers, with ImageButton emerging as particularly problematic—53% of ImageButtons had missing labels and 40% were undersized. The study also examines environmental factors by investigating app categories, development tools, and third-party plug-ins that correlate with barrier prevalence. Education apps were found disproportionately likely to have few TalkBack-focusable elements, representing 19% of affected apps despite being only 7% of the total population.
Key findings
Missing labels and undersized elements emerged as the most prevalent accessibility barriers across the 9,999 apps tested. For contentDescription-dependent elements, ImageView had 91.5% missing labels, while ImageButton had 66% missing. Among image-based buttons specifically, Floating Action Buttons (FABs) had 92.8% missing labels despite lower overall prevalence. Size-based barriers affected 82% of Checkboxes and 62% of Radio Buttons. Multi-platform development tools (Apache Cordova, Adobe Air, Crosswalk) and game engines (Unity, Cocos2d-x) strongly correlated with the "few TalkBack-focusable elements" barrier—64% of Adobe Air elements, 69% of Crosswalk elements, and 75% of Cordova elements appeared in barrier-affected apps. Third-party social media plug-ins (Facebook Login, Google+) had excellent labeling (0-1.3% missing) but significant size-based errors (82-100% too small in either dimension). The research found that elements too small in both dimensions often featured tightly cropped icons, while elements that were appropriately sized typically included visible text labels or padding around icons.
Relevance
This research provides practitioners with concrete data on where accessibility failures concentrate in Android development. The finding that specific UI element classes (ImageButton, Checkbox, RadioButton) and development tools (hybrid frameworks, game engines) correlate with barriers offers targeted guidance for testing and remediation efforts. The epidemiological framework—distinguishing intrinsic factors (individual app design) from extrinsic factors (tools, plug-ins, third-party code)—helps organizations prioritize interventions at the ecosystem level rather than app-by-app. Particularly valuable is the insight that automated fixes could address common patterns: icon-based buttons could inherit labels through image recognition, and undersized standalone elements could be expanded by including their text labels in the clickable region. The data showing third-party plug-ins consistently undersized suggests documentation and default sizing changes could propagate accessibility improvements across thousands of apps simultaneously.
Tags: mobile accessibility · Android · TalkBack · screen readers · automated testing · large-scale analysis · accessibility barriers
Standards referenced: Android Accessibility Guidelines