- From: Matthew Wickline <aware-colorlab@hwg.org>
- Date: Fri, 13 Aug 2004 05:27:43 -0700
- To: public-comments-wcag20@w3.org
> codifying color contrast : > Any ideas are appreciated When I read that text, the first thing I thought of was finding the effective luminance of the two colors and requiring some minimum separation. Then I compare that to http://www.w3.org/TR/AERT#test-color-contrast and find that this is exactly where you're at currently. Once a minimum separation value was determined, it'd be trivial to provide a tool allowing web designers to input two colors and find out of they satisfy the requirement. However, finding the value is the hard part. I don't personally have (or know of) any data that would suggest a certain value, or make finding the value a simple matter of looking in some table. I may be mistaken, but it sounds like an empirical question. :/ If I were one of those usability gurus, I'd probably round up a nice-sized sample of users of various ages and visual accuity and have them perform tasks on web pages of various colors (maybe using a variety of monitors as well... some lcd, some crt). The tasks might be things like clicking some specific linked text in a minimal amount of time or reading content and answering comprehension questions. Hopefully this data would give some sort of contrast difference requirement to ensure that x% of viewers will be able to access the content without suffering losses in seek time or comprehension. A statistician would probably need to help out with the sample selection and results analysis to ensure that the values are statistically valid (at a high confidence level) with a sufficiently small margin of error. For higher % values, you would need lower contrast difference. The WG could select some % value (maybe 95 or 99) and then find the associated contrast differential value. Yeah, that's a bunch of work. Maybe somone like these guys could be talked into taking it on: http://psychology.wichita.edu/surl/usability_news.html I'm afraid I can't think of any way to get a good number without actually testing though :/ -matt (that guy from http://colorfilter.wickline.org/ http://colorlab.wickline.org/ )
Received on Saturday, 14 August 2004 03:56:59 UTC