Accessible by Design: Research Everyone Can Take Part In

Roughly one in six people lives with a significant disability.¹ Many use a keyboard instead of a mouse, a screen reader instead of a screen, or need captions instead of audio. When an interface quietly assumes none of that, those respondents do not file a complaint. They drop out, or they never start, and the people who remain in your sample are the ones your design happened to suit.

That is the part teams miss. An inaccessible study does not leave an obvious hole you can patch later. It leaves a silent skew in who answered, and a skew you cannot see is far more dangerous than one you can. Accessibility, in research, is not a courtesy bolted on at the end. It is a precondition for the sample being what you think it is.

Good intentions are not a method. The discipline that turns "we tried to be inclusive" into something defensible is to measure against a published standard, and the accepted one is the Web Content Accessibility Guidelines at level AA.² It is concrete and testable: text has sufficient contrast, every control has a name, the interface works from the keyboard alone, content does not rely on colour by itself.

Naming the standard matters for the same reason naming any floor matters. It lets a buyer, a panel provider, or a respondent advocate evaluate your work on the merits instead of taking a reassuring sentence on faith. WCAG AA is the target you build to and test against, not a badge you award yourself once and forget.

Three habits cover most of the ground. First, make everything operable from the keyboard: every button, field, and menu reachable in a sensible order, with a focus outline you can actually see, and no trap a keyboard user cannot escape. A great deal of research tooling fails here because custom controls are built from generic containers that a mouse can click but a keyboard cannot reach.

Second, make everything perceivable without sight. That means a real text name on every control and image, status that is announced rather than only shown, and a structure of headings and landmarks a screen reader can navigate. Third, never let meaning ride on a single channel: a red dot that means "error" needs a word too, and a video with audio needs captions. The same goes for motion. Honour a respondent’s request for reduced motion, because animation that cannot be turned down is a barrier for people with vestibular conditions.

We hold the Reactor app to WCAG AA and treat it as engineering, not decoration. The interface is fully operable from the keyboard, with a visible focus indicator and a skip link that lets keyboard and screen-reader users jump past the navigation straight to the content. Dialogs move focus into themselves, trap it while open, and return it when they close, so a screen-reader user is never stranded behind a panel they cannot find.

Beyond operability, every control carries a real accessible name, loading states announce themselves rather than spinning in silence, and errors are spoken, not merely coloured. Status is never carried by colour alone. Media that needs captions can carry them, and the whole experience respects the operating-system setting for reduced motion, so the ambient animation that gives the product its character steps aside the moment a respondent asks it to.

Accessibility, like security, is honest only when it states its limits. Conformance is assessed against a standard, not guaranteed against every possible combination of device and assistive technology, and the field of assistive tools is wide and always moving. Embedded third-party content and respondent-supplied media may not meet the same bar as the parts we control. We say so on purpose, because a claim that hides its own edge is marketing, not a commitment.

So the honest floor is this: accessibility is ongoing work, measured against WCAG AA, audited and improved rather than declared finished. If you hit a barrier, that is a defect we want to hear about and fix, the same as any other. Build for everyone, hold the result to a standard someone else can check, and be specific about where the protection stops. That pairing is the whole discipline.