Rendering Tables in Audio: The Interaction of Structure and Reading Styles
Yeliz Yesilada, Robert Stevens, Carole Goble, Shazad Hussein · 2004 · Proceedings of the 6th International ACM SIGACCESS Conference on Computers and Accessibility (Assets 04) · doi:10.1145/1028630.1028635
Summary
This paper presents a comprehensive analysis of why HTML tables are problematic for screen reader users and proposes two complementary approaches for non-visual table access. The authors from the University of Manchester first characterize what makes tables useful in print — their ability to convey multi-dimensional relationships through spatial layout, visual cues, and implicit structure — and why these same features make them inaccessible when linearized by screen readers. Through a thorough survey of table designs found on the web and in literature, the authors identified 35 different structural groups of tables and catalogued the reading styles they support. They found no direct relationship between table structure and reading style: structurally similar tables can be read differently depending on content, and the same table can support multiple reading styles (by cell, by row, by column, by intersection, sequentially, reading with unit). Six key reading and browsing styles were identified, along with presentation styles (special layout, special location, spatial cues, ordering, grouping) and content styles (special symbols, complex data, empty cells). Based on this analysis, two tools were developed: EVITA (Enabling Visually Impaired Table Access), a dedicated table browser supporting active reading through keyboard navigation across rows, columns, and cells with contextual header information; and a table linearizer that transforms tables into a linear form readable by standard screen readers using XSLT stylesheets.
Key findings
A pilot evaluation of EVITA with 9 sighted and 1 visually impaired user performing five tasks on five different tables revealed several important findings. Users generally found the interaction style effective and were able to access information from tables of different structures. In a comparative evaluation, users performed tasks with both the modal version (separate modes for row, column, and cell navigation) and the non-modal version (all functions accessible via modifier keys in a single mode). The non-modal version was preferred because it allowed faster switching between row and column navigation without mode changes. Users stated that since fewer key combinations were needed in the non-modal version, they were less likely to make errors. Specifying row and column headers for the intersection mode was preferred to the other two methods. The system provided cell presentation functionalities that gave contextual information about the current cell's position (row number, column number, row and column headers) which users found valuable for orientation. Users indicated they could achieve "a high degree of control over the information flow." However, some users became disoriented because they were more aware of the table positions and structure than the reading styles the tables supported. The linearization approach, while sub-optimal, offered a cost-effective alternative requiring no additional software or learning — it works within existing screen readers using standard XSLT transformations.
Relevance
This research makes a foundational contribution to understanding table accessibility by moving beyond simple linearization to analyze the rich interaction between table structure, content, presentation, and reading intent. For accessibility practitioners, the key insight is that tables are not just data containers but information design artifacts where spatial layout carries meaning — and making them accessible requires preserving this meaning, not just serializing the content. The taxonomy of reading styles (by cell, row, column, intersection, sequential, with unit) provides a practical framework for evaluating whether an accessible table implementation actually supports the ways people need to use table data. The two-approach strategy (rich browser for power users, linearization for broad compatibility) reflects a realistic accessibility trade-off between optimal and practical solutions. A key limitation is the very small evaluation with only one visually impaired participant; the authors acknowledge that extensive evaluation with visually impaired users is needed. The work also does not address the widespread use of tables for page layout (rather than data), though it focuses on the harder problem of data tables where structure carries semantic meaning.
Tags: table accessibility · visual impairment · screen reader · web accessibility · table linearization · non-visual interaction · data table · audio interaction · web navigation
Standards referenced: WCAG 1.0