From Screen Reading to Aural Glancing: Towards Instant Access to Key Page Sections
Prathik Gadde, Davide Bolchini · 2014 · Proceedings of the 16th International ACM SIGACCESS Conference on Computers & Accessibility (ASSETS 2014) · doi:10.1145/2661334.2661363
Summary
This paper addresses a fundamental problem in screen reader navigation: while sighted users can glance at a web page to instantly understand its structure and locate relevant sections, screen reader users are forced to listen to content serially, making navigation of complex pages extremely inefficient. The authors adopt a three-pronged approach. First, they conducted a formative user study with eight blind/visually impaired and four sighted participants performing shopping tasks on Amazon.com, uncovering seven key navigation problems: even simple pages obstruct navigation; rigid access across page types hampers efficiency; fear of missing information forces unnecessary effort; little tolerance for navigation mistakes demands careful line-by-line reading; blurred section boundaries confuse users; "skip to content" is not always optimal (e.g., it skips important faceted search filters); and difficulty re-finding information overloads short-term memory. The performance gap was stark — screen reader users needed far more keystrokes and time than sighted users for the same tasks. Second, the authors built a custom crowd-ranking system and recruited 100 Amazon Mechanical Turk workers to prioritize the most important page sections across four page types (home, search results, product details, checkout) for common shopping tasks. This revealed that sections users deemed most critical often required the most keystrokes to reach via screen reader — for example, shipping speed on the checkout page was ranked among the top five but required 22 keystrokes to access.
Key findings
Based on these findings, the authors introduce DASX (Dialogic Augmentation of the Screen-reader Experience), a navigation approach with four simple commands: "What's there" (reads the top five most important sections), "More on this page" (reads additional sections), "Get me going" (activates the single most important action, e.g., Add to Cart), and "Go" (loads the selected section). Commands are available via keyboard shortcuts or voice. A preliminary KLM-GOMS analytical evaluation with an expert user showed dramatic improvements: DASX reduced actions by 75% on item description pages and 69% on checkout pages compared to standard screen reader navigation. Time on the checkout page fell 43%. In absolute terms, DASX users needed only 6 keystrokes on item description pages and 8 on checkout pages — approaching the 4 and 2 required by sighted users, compared to the 24 and 34 keystrokes required by current screen readers. The crowd-ranking data confirmed that section relevance is task-dependent and varies by page type, supporting the case for page-type-aware navigation strategies.
Relevance
This paper reframes screen reader navigation from a serial reading problem to a page comprehension problem, introducing the concept of "aural glancing" as an analogue to the visual scanning sighted users perform unconsciously. For accessibility practitioners and web developers, the key insight is that WCAG's "skip to content" (Guideline 2.4.1) is necessary but insufficient — users need task-aware direct access to multiple relevant sections, not just the main content area. The crowd-ranking methodology for identifying important page sections is a replicable approach for any complex web application. The finding that screen readers are page-type-agnostic (treating product pages, search results, and checkout pages identically) highlights a significant design gap in current assistive technology. Limitations include the analytical (not empirical) evaluation of DASX and the Amazon-specific focus, though the approach generalizes to any information-dense web application.
Tags: screen readers · web navigation · blindness · visual impairment · e-commerce accessibility · aural browsing · crowdsourcing · JAWS
Standards referenced: WCAG 2.0