Eye Gaze Behaviour and Comprehension of Colour Commentary and Gameplay Captions of Live Fast-Paced Sports for Deaf and Hard of Hearing Television Viewers
Somang Nam, Tatyana Kumarasamy, Maria Karam, Margot Whitfield, Evan Hibbard, Jenny Leung, Deborah Fels · 2026 · ACM Transactions on Accessible Computing · doi:10.1145/3806043
Summary
This ACM TACCESS study evaluates a novel captioning approach for live fast-paced sports broadcasting: 'Reduced Captioning' that displays only the colour commentary (opinions, context, analysis) and omits the verbatim play-by-play narration. The authors motivate this design around a well-documented live-captioning dilemma. Live sports speech frequently exceeds 200 WPM, well above the 135 WPM average reading speed of d/Deaf individuals aged 17–20, and produces captions that arrive 5–6 seconds late, visually crowd the screen, and often contain stenographic errors. Regulatory frameworks in Canada (CRTC with the NER model), the US (FCC), and the UK (BBC) emphasize verbatim fidelity, which constrains editing-based approaches. The authors argue that omitting play-by-play — which describes what viewers can already see — preserves meaning while substantially cutting word count. In a within-subjects study, 27 participants (16 d/Deaf, 11 Hard of Hearing) watched live CBC broadcast clips of ice hockey or basketball under both Full and Reduced Captioning conditions while being tracked with a Tobii eye tracker. Full Captioning ran at 181.5 WPM, Reduced at 150.5 WPM; captions were generated live by human captioners without post-production editing. Measures included eye-movement data across two areas of interest (captions vs. game), a 17-question questionnaire on satisfaction and perceived quality, and a novel 15-minute conversational comprehension interview designed to avoid the frustration participants had reported with multiple-choice tests in prior work. ASL-English interpreters and CART were available on request.
Key findings
Across both sports and both hearing groups, participants spent more time and visits on the game area than on the caption area, and visited the caption area more frequently under Reduced Captioning than Full Captioning (statistically significant for d/Deaf hockey viewers and HoH basketball viewers under both conditions). Median caption satisfaction, viewing-experience rating, and perceived helpfulness were consistently higher for Reduced Captioning (e.g., basketball helpfulness M=3.58 vs 3.00; hockey M=3.50 vs 3.14). Of the 26 participants who completed the preference questionnaire, 17 had no preference; among the rest, 7 preferred Reduced and only 2 preferred Full, with d/Deaf participants favouring Reduced more strongly. Thematic analysis of 172 comments found the most common complaints about Full Captioning were that it was exhausting, felt like 'reading a book', forced users to miss gameplay, and obscured on-screen stats. Participants also identified caption placement (blocking score graphics), legibility (contrast, all-caps, font size), and delay as persistent problems. Comprehension scores did not significantly differ between the two styles overall, though gameplay-question scores were slightly higher than commentary-question scores, suggesting viewers drew gameplay understanding from the visuals rather than the captions.
Relevance
This study is directly relevant to captioning practice, broadcaster accessibility policy, and regulatory discussions that currently privilege verbatim fidelity over readability. It offers empirical support for a middle path: semantically preserving reductions (dropping narration that duplicates visual information) rather than paraphrasing, which regulators disallow. Practitioners designing live-captioning workflows, sports broadcast captioning, or adaptive/configurable caption systems should note the authors' recommendation that Reduced Captioning be offered as an option alongside Full Captioning rather than as a replacement. The paper also documents concrete design faults — caption placement covering score overlays, poor contrast against dynamic backgrounds, preference for pop-on over roll-up styles — that are actionable regardless of the play-by-play question. Limitations include the small sample, novelty effect from single-period clips, interpretation fidelity between ASL and English, and the absence of longitudinal data across a full season.
Tags: closed captioning · live captioning · broadcasting · sports · eye tracking · Deaf · Hard of Hearing · readability · cognitive accessibility · user study
Standards referenced: CRTC NER Model · FCC Closed Captioning Rules · BBC Subtitle Guidelines