WebinSitu: A Comparative Analysis of Blind and Sighted Browsing Behavior
Jeffrey P. Bigham, Anna C. Cavender, Jeremy T. Brudvik, Jacob O. Wobbrock, Richard E. Ladner · 2007 · Proceedings of the 9th International ACM SIGACCESS Conference on Computers and Accessibility (Assets '07) · doi:10.1145/1296843.1296854
Summary
This paper presents WebinSitu, a remote in situ study comparing the web browsing behavior of 10 blind and 10 sighted users over a one-week period. The study used an advanced web proxy built on UsaProxy that injected JavaScript into viewed pages to record not just HTTP requests but also user-level actions including key presses, mouse events, clicks on specific DOM elements, and interactions with dynamic content. This proxy-based approach was critical because it allowed participants to use their own computers, browsers, and assistive technology configurations in their natural environments — avoiding the artificial constraints of lab studies that struggle to replicate the diversity of screen reader setups. Participants visited 21,244 total pages (7,161 by blind participants) over approximately 325 combined hours of browsing. The study examined two broad categories: the accessibility of web content encountered by users (alternative text, heading structure, form labels, skip links, dynamic content, Flash) and differences in browsing behavior (mouse use, probing, timing). The page visit distribution followed a Zipf-like pattern, with three sites (google.com, myspace.com, msn.com) accounting for roughly 20% of all page views, highlighting that the accessibility of popular sites disproportionately affects user experience.
Key findings
The results revealed that blind users did not generally avoid pages with accessibility problems — they visited similar sites to sighted users. However, blind participants were significantly less likely to visit pages with dynamic content or AJAX requests, suggesting awareness that such content is often inaccessible. Only 56.9% of images on visited pages had appropriate alternative text, yet blind users were significantly more likely to click on images with proper alt text (72.17% vs 34.03%), demonstrating they actively relied on it. Only 53.08% of pages contained heading tags, and just 41.73% of pages with form inputs had label elements. Skip links were used occasionally — blind participants clicked 5.60% of skip links presented vs 0.07% for sighted users. Blind participants exhibited significantly more probing behavior (0.34 vs 0.12 probes per page, p < 0.01), following links to see where they lead and then returning. Blind users spent significantly more time per page, particularly for short tasks: on Google searches, blind users averaged 74.66 seconds vs 34.54 seconds for sighted users, and scanning search results took 155.06 vs 34.81 seconds. The time gap narrowed for longer tasks. Blind users also used the mouse on 25.85% of pages — a surprising finding indicating mouse use as a coping strategy when keyboard navigation failed.
Relevance
This study is important for accessibility practitioners because it measures real-world browsing behavior rather than relying on automated audits or lab tasks. The finding that blind users do not avoid inaccessible sites but instead develop coping strategies (probing, mouse use, avoiding dynamic content) challenges the assumption that accessibility barriers simply prevent access — they often create inefficiency and frustration instead. The data on alternative text usage, heading structure, and form labels provides empirical evidence for prioritizing these accessibility features. The proxy-based remote methodology itself was influential, demonstrating a scalable way to study assistive technology users in their natural environments. For developers, the stark timing differences — blind users taking 2-4 times longer on common tasks like Google searches — quantify the real cost of accessibility failures. The observation that 44.1% of Flash content was advertising underscores that not all inaccessible content is equally important to fix.
Tags: web accessibility · blind users · screen readers · browsing behavior · user research · remote study · coping strategies · alternative text · dynamic content · AJAX · skip links
Standards referenced: WCAG · WAI-ARIA