Accessibility Testing
Accessibility is not the same as usability, but it overlaps because a lot of disabilities make it harder to deal with stuff that doesn't have great usability to begin with.
Some accessibility features can be tested automatically with tools, but much of it comes down to human judgment. In fact, some accessible pages will fail markup validation because some validators don't understand accessibility markup such as WAI-ARIA. The W3C explains the limitations of validation tools and how to choose one.
Jim Thatcher's analysis of what accessibility testing is possible is a good start, but is not comprehensive. For example, it misses contrast levels appropriate for dyslexics, moving image problems that people with some neuro deficits have, everything important being keyboard-only accessible, and probably other things.
Contents
General
These things are generally good things to do and also have implications for disability accessibility.
- Pages need to validate fully to whatever DTD they've specified. Note: If adding accessible markup stops a validation tool from working, it may be out-of-date. Make sure that the tool is up-to-date, but don't remove the accessible markup.
- Pages should degrade gracefully in the absence of capabilities for Flash, JavaScript, images, CSS, and so forth. Whenever possible, the absence of those capabilities should give as rich a user experience as their presence.
- Proper semantic markup should be used: headers for headings, ACRONYM and ABBR tags, EM/STRONG rather than I/B tags, and so forth. The product should never change user-entered HTML tags if avoidable, especially if doing so will move away from semantic toward formatting-based markup.
- WAI-ARIA roles and landmarks should be used.
Blindness
This is the disability people usually think about when they discuss web accessibility, but it's important to remember it's not the only one.
- Screen reader friendliness. Try very hard to get general screen reader users to do testing for us, rather than sighted people using screen readers which is a bad approximation. However, sighted developers and testers can still perform minimalist screen reader testing. See our list of screen readers.
- Check tab order for everything, especially things which are AJAXey and therefore have changing tab order.
Relevant reading
Color Blindness
Things which are usually signalled to users through colour changes in interface elements must have some other signalling mechanism (such as change of image shape) in addition to colour.
Tools & Resources
- Colorblindness Web Page Filter
- Sim Daltonism -- A program for Mac OS X that simulates different kinds of color blindness.
Deafness
Any official videos put on the site (such as "how to use" screencasts) should have captions available.
Deafblindness
This disability is a double-whammy, as many of the solutions to problems faced by blind or deaf users rely on the other sense -- e.g., audio alternatives to CAPTCHAs.
- Even captions for videos will probably be inaccessible. Separate methods of obtaining the information such as transcripts are needed.
- Information needs to be concise. Many deafblind users will be using Braille displays that give a single line of text that's 40 characters long.
Dyslexia
A too-high or too-low contrast can be a problem.
Keyboard-only users
- Make sure anything triggered by mouse movements (such as hover attributes) that's needed for site use has a keyboard-accessible alternative.
- Check tab order for everything, especially things which are AJAXey.
- There must be alternatives to places where one would need to control-click to select multiple elements of a list. For example, choosing tags, where the current keyboard alternative is to remember the names of your tags and type them in.
- Test with the mouseless browsing Firefox addon.
Low vision
- Too-low contrast is a problem.
- The default font size should be reasonable.
- Robustness of layout to font size increases (command-+ / ctrl-+) must be tested in Internet Explorer, Firefox, Safari, Opera, and Chrome.
- For those browsers that have the capability (Firefox 3 and all versions of Opera), page size increases -- scaling images as well as text -- must be tested.
- Alternate layouts with less visual "clutter"/"noise" may be needed for some people (equivalent to LJ's Lynx site scheme).
Mouse use problems
This refers to users who are able to use a mouse but have some difficulty with it. Keyboard equivalents are a solution to some users in this group, but others will have no ability to use a physical keyboard, either -- only an on-screen keyboard (again, requiring mousing) will be used.
- Mouse targets need to be as large as possible.
- Minimise/eliminate use of any elements that require clicking and dragging, as this is most difficult for many users.
- Remember Fitt's law.
Neurological problems
This category includes people on the autism spectrum, people with traumatic brain injuries, stroke survivors, and several other groups.
- Anything on the page that moves (Flash, animated GIFs, etc.) can be sufficiently distracting to make the page unreadable.
- Evaluating cognitive usability.
Speech Recognition Users
All of the issues listed for keyboard users apply to speech recognition users. See our list of speech recognition systems.
Tools to help test accessibility
Using commonly installed tools to test
Freely available specialized testing tools
- Color Contrast Analyser checks for contrast which is too high as well as contrast that is too low.
- using nVDA to evaluate screenreader usability
- Firefox accessibility extension from UIUC
- Fangs Firefox screen reader emulator
Validators
- Markup validation W3C validator
- Searchable list of evaluation tools hosted at W3C
- Wave and Wave toolbar
