How People Use Presentation to Search for a Link: Expanding the Understanding of Accessibility on the Web
Caroline Jay, Robert Stevens, Mashhuda Glencross, Alan Chalmers · 2006 · Proceedings of the 2006 International Cross-Disciplinary Workshop on Web Accessibility (W4A) · doi:10.1145/1133219.1133241
Summary
This paper uses eye-tracking to investigate a fundamental question about web accessibility: what exactly does visual presentation provide to sighted users that is lost when content is accessed nonvisually (via screen readers) or on small screens? The authors argue that accessibility efforts have focused heavily on making content accessible (alt text, semantic markup) while neglecting the role that visual presentation plays in helping users navigate and complete tasks. Using the metaphor of "travel objects" — like signs and landmarks that help people navigate physical spaces — they observe that visual layout, images, color, and spatial groupings serve as navigational cues on web pages. When these are stripped away (as in text-only versions or audio screen reader presentations), users lose the ability to efficiently orient themselves and jump to relevant content. The study had 18 sighted participants search for specific links on the BBC News Manchester website in both standard formatted and text-only versions, with eye movements tracked using a Tobii x50 eye tracker.
Key findings
The eye-tracking data revealed starkly different reading behaviors between the two conditions. On the standard formatted page, participants' eyes "darted" between salient areas — headlines, images, prominent words — making large saccades (eye jumps) to navigate between visually distinct regions of the page. On the text-only page, participants read in a serial, top-to-bottom fashion similar to reading printed text, with uniform fixation patterns across the left side of the page and much smaller saccades. Quantitatively, the text-only page required significantly more fixations (F(1,32) = 2.183, p < 0.005), indicating it demanded more attention to process. However, fixation durations were significantly longer on the standard page (F(1,32) = 0.208, p < 0.005), suggesting each fixation involved more complex cognitive processing as users oriented themselves and planned navigation. The majority of participants found the standard version easier to search, and for the China link search, participants located it significantly more quickly on the standard page (t(16) = 3.696, p < 0.005). The key insight is that text-only pages (and by analogy, screen reader presentations) force users into serial reading rather than allowing the selective, visually-guided scanning that formatted pages support.
Relevance
This study provides empirical evidence for something that screen reader users have long reported: the web is "about three times more difficult to use" for them (citing Coyne and Nielsen). By demonstrating exactly how visual presentation assists sighted users — through spatial groupings, visual salience, and navigational "landmarks" that enable efficient scanning — the paper identifies what needs to be recreated in nonvisual modalities. The implication is profound: simply making content technically accessible is insufficient; the presentational cues that help sighted users navigate efficiently must be translated into equivalent nonvisual cues (audio landmarks, structural navigation shortcuts, spatial audio). This anticipates modern ARIA landmarks, heading navigation, and screen reader rotor features that attempt to provide the nonvisual equivalent of visual scanning. For practitioners, the paper reinforces that accessibility is not just about content access but about task efficiency — a page where a blind user can technically reach all content but must wade through it serially is far less usable than one that provides structural shortcuts equivalent to the visual scanning that sighted users enjoy.
Tags: eye tracking · visual presentation · screen readers · nonvisual access · web navigation · text-only pages · information scent · cognitive processing · user research
Standards referenced: WCAG 1.0