Capability Survey of Japanese User Agents and Its Impact on Web Accessibility
Takayuki Watanabe, Masahiro Umegaki · 2006 · Proceedings of the 2006 International Cross-Disciplinary Workshop on Web Accessibility (W4A): Building the Mobile Web: Rediscovering Accessibility? · doi:10.1145/1133219.1133227
Summary
This paper presents a systematic evaluation of four major Japanese user agents for people with visual disabilities — IBM Home Page Reader 3.04 (a voice browser), PC-Talker XP 3.04, 95 Reader 6.0, and JAWS 6.2 (Japanese edition) — tested against the W3C UAAG 1.0 Test Suite for HTML 4.01, accessible PDF and Flash files, and custom test files addressing Japanese language-specific issues. The study was motivated by a critical gap in accessibility knowledge: Japanese users who are blind cannot use English-language screen readers because they do not speak Japanese, yet the Japanese assistive technology market is small (approximately 0.3 million legally blind users) compared to the English-speaking market, making it economically difficult for Japanese developers to build fully featured user agents. The research aimed to provide objective evidence for defining a "Japanese baseline" — a concept introduced in the WCAG 2.0 working draft referring to the set of technologies that content authors can assume user agents will support. The paper also examines the shared responsibility between web content and user agents, arguing that accessibility failures should be attributed to the correct party rather than blamed ambiguously. The study contextualizes Japanese accessibility within the JIS X 8341-3 standard (Japan's industrial accessibility standard harmonized with WCAG 1.0) and notes that surveys of Japanese government websites found even the most accessible lacked fundamental accessibility considerations.
Key findings
Of 48 Priority 1 UAAG 1.0 checkpoints, 20 were met by all four user agents while 11 were failed by all — revealing a consistent floor and ceiling of capability. All user agents handled basic functions: full keyboard access, text messages, rendering content per specification, text view, toggling background images and scripts, text scaling, volume control, speech rate/volume/characteristics configuration, content and interface focus, style sheet selection, and accessible documentation. All failed on: time-independent interaction, captions/transcripts/audio descriptions, synchronization cues, toggling animated images, toggling blinking text, font family configuration, multimedia control (slow/stop/pause/navigate), caption obscuring prevention, independent volume control, and viewport state history. A stark capability divide emerged between JAWS/Home Page Reader and the domestic PC-Talker/95 Reader: only JAWS and HPR could navigate by headings, read tables using structural markup (summary, th, scope attributes), link to same-page anchors, search text, read abbreviation/acronym titles, and select arbitrary text. PC-Talker and 95 Reader relied on MSAA (Microsoft Active Accessibility) rather than DOM access, limiting their structural navigation capabilities. For PDF, no user agent distinguished headings from body text or supported heading-based navigation — all could only read sequentially. For Japanese-specific issues, pronunciation of symbols varied across user agents, the lang attribute did not change text-to-speech voice, and no user agent handled ambiguous date/time/money representations appropriately (e.g., "2006/03/22" could be a date or fraction). The price disparity was notable: PC-Talker cost $333, 95 Reader $305, Home Page Reader $131, while JAWS cost $1,243 — four times the price of popular Japanese screen readers.
Relevance
This paper provides a rare empirical window into the accessibility technology landscape outside the English-speaking world, highlighting a problem that persists today: assistive technology capabilities vary significantly across languages and markets, yet international accessibility standards like WCAG assume a relatively uniform user agent baseline. The concept of "baseline" that this paper investigates became formally embedded in WCAG 2.0 and remains relevant in WCAG 2.1 and 2.2 conformance discussions. The finding that domestic Japanese screen readers lacked structural navigation (headings, table markup) while still meeting basic UAAG requirements illustrates why content authors cannot rely solely on standards compliance — they must understand the actual capabilities of the user agents their audience uses. For practitioners working in multilingual or non-English contexts, this paper is a reminder that accessibility testing should include assistive technologies used by the target audience, not just market-leading English-language tools. The PDF accessibility findings — that even accessible PDFs were read only sequentially without heading navigation — foreshadowed ongoing concerns about PDF accessibility that remain relevant two decades later.
Tags: user agents · screen readers · UAAG · assistive technology evaluation · Japanese accessibility · accessibility standards harmonization · PDF accessibility · Flash accessibility · baseline
Standards referenced: UAAG 1.0 · WCAG 1.0 · WCAG 2.0 · JIS X 8341-3 · Section 508 · MSAA