Understanding Low Vision Graphical Perception of Bar Charts
Yash Prakash, Akshay Kolgar Nayak, Sampath Jayarathna, Hae-Na Lee, Vikas Ashok · 2024 · ASSETS '24: Proceedings of the 26th International ACM SIGACCESS Conference on Computers and Accessibility · doi:10.1145/3663548.3675616
Summary
This paper investigates how low-vision users who rely on screen magnifiers perceive and interact with bar charts, addressing a significant gap in data visualization accessibility research. While studies on chart perception have extensively examined sighted users and some have explored blind screen reader users, the graphical perception of low-vision screen magnifier users has been virtually unexplored. Screen magnifier users occupy a unique position: they have residual vision and interact visually with charts, but magnification fundamentally changes the viewing experience — users can only see a portion of the chart at any time and must pan to build a mental model of the whole, while also contending with blurring from excessive zoom and reduced contrast at high magnification levels. The researchers designed four experiments conducted as a laboratory study with 25 low-vision participants. Experiment 1 tested position comparison (judging which of two bars is taller within grouped bar charts). Experiment 2 tested length comparison across different stacked bar configurations — single-column stacked, adjacent stacked, and separated stacked. Experiment 3 examined the effect of chart aspect ratio on perception accuracy. Experiment 4 explored the impact of axis label density on reading comprehension. To capture the fine-grained interaction behaviours unique to screen magnifier use, the researchers built a custom logging tool that recorded micro-interactions including zoom level changes, panning movements, dwell times, and cursor trajectories — data that standard usability tools do not capture.
Key findings
Low-vision users invested substantial time and effort counteracting the visual side effects of magnification — specifically blurring at high zoom levels and reduced contrast — which significantly impacted their task completion times compared to sighted users performing the same tasks. For position comparison (Experiment 1), accuracy was generally high but declined when bars being compared were spatially separated, requiring users to pan between them and rely on working memory to compare values. This contrasts with sighted users, who can compare separated bars with minimal accuracy loss through parallel visual processing. For stacked bar charts (Experiment 2), single-column stacked configurations — where multiple segments are stacked within a single bar — were the most challenging, as users had to zoom in to distinguish segment boundaries and then zoom out or pan to see the overall bar height, creating a constant trade-off between detail and context. Adjacent stacked bars were easier because users could compare corresponding segments side-by-side. Chart aspect ratio (Experiment 3) had a modest effect, with wider charts requiring more horizontal panning. Dense axis labels (Experiment 4) significantly increased task time because users had to zoom in further to read small text, losing more chart context in the process. The micro-interaction logging revealed distinctive patterns: low-vision users frequently alternated between high zoom (for reading labels and distinguishing boundaries) and lower zoom (for overall comparison and orientation), a behaviour the authors term "zoom cycling" that adds significant cognitive and temporal overhead.
Relevance
This research provides the first empirical evidence on how screen magnifier users perceive bar charts, filling a critical gap between the well-studied extremes of sighted and blind chart interaction. The findings have direct implications for data visualization design: charts intended for low-vision audiences should minimise the need for spatial comparison across distance (use adjacent rather than separated bars), avoid single-column stacked configurations, use larger axis labels to reduce zoom-in requirements, and provide clear visual boundaries between data segments. The "zoom cycling" behaviour is a key insight — it represents a fundamental interaction cost of magnification that chart designers should minimise by ensuring that relevant information is spatially proximate. For accessibility practitioners, the work demonstrates that "making charts bigger" is insufficient for low-vision accessibility; the interaction dynamics of magnification create unique perceptual challenges that require deliberate design consideration. The custom micro-interaction logging methodology is also a contribution, providing a template for future research studying how magnifier users interact with complex visual content beyond charts.
Tags: low vision · data visualization · bar charts · screen magnifier · graphical perception · magnification · chart accessibility · user study