Accessible Tables: Making Data Readable for Every Student

A table is a small piece of design that does a large amount of work. A sighted reader glances across a row, drops down a column and understands in seconds that revenue rose in the second quarter or that one drug dose differs from another. A screen reader user cannot glance. If the table is built badly, they hear a long run of numbers with no sense of which belongs to what.
Tables fill textbooks, scholarly monographs, annual reports and standards documents, and they are among the first things to break when a file is converted.
What a table needs to be understood
Three things separate a usable table from a pile of cells.
First, header cells must be marked as headers. In an EPUB or web page this means th elements, with a scope that says whether a header applies to a column or a row. In a tagged PDF it means TH tags rather than TD. When this is in place, a screen reader can announce "Q2, revenue, 4.2 million" as the reader moves through the cells. Without it, the reader hears "4.2 million" and has to remember where they are.
Second, the table needs a name. A caption, or a clear heading directly above it, tells the reader what the table is for before they enter it. For a complex table, a short summary of the main point lets readers decide whether to go cell by cell.
Third, the structure has to be real. A table is a table because the markup says so. WCAG success criterion 1.3.1 asks that the relationships conveyed visually are also available in code, and a grid of tabbed text or a screenshot fails that test.
Where tables go wrong
Tables as images are the most common failure, and the worst. A picture of a table cannot be navigated by cell, cannot be resized without blur, and cannot be searched. Alt text can summarise the content, but it cannot give a reader the freedom to move across rows and columns.
Merged cells come next. A header spanning three columns looks tidy on the page. For assistive technology it can break the link between a header and the cells beneath it. Simple tables with one row of column headers are the safest design. Where a complex table is unavoidable, headers need explicit associations, using header and id attributes, and the result needs testing with a screen reader, not only a checker.
Layout tables are a quieter problem. Using a table to position text or place an image beside a paragraph tells a screen reader that data is coming when none is. Use styling for layout and keep tables for data.
Wide tables are a particular challenge in reflowable EPUB, where the screen may be a phone. A table that cannot reflow should scroll horizontally rather than shrink into illegibility, and the page should still make sense when the text is enlarged to twice its size.
Finally, PDFs that look perfect often carry no table tags at all, and the structure has to be rebuilt row by row.
A quick test for any table
Take one table from your list and try four things. Check that the header cells are marked as headers. Move through the cells with a screen reader and listen to whether each value arrives with its row and column names. Enlarge the text to 200 percent and see whether the table still holds together. Then ask whether the table could be made simpler by splitting it in two.
Automated checkers will find a missing header. They will not judge whether a table is too complex to follow, and that judgment is the one that matters most to the reader.
Where S4Carlisle fits
Tables are one of the places where careful tagging pays back directly. S4Carlisle builds table structure into EPUB and PDF production, rebuilds it where files arrive without it, and reviews complex tables by hand. With Benetech GCA certification and more than 50 years in publishing, our team treats every table as data a student needs to use.
To test a sample chapter with our team, visit s4carlisle.com or write to sales@s4carlisle.com.





Comments