[Silhouette]

You never mentioned WCAG guidelines before, you just said "making it hard to read."

The WCAG are just one authoritative source that happens to be readily available and therefore seemed a useful objective reference for the discussion.

The point I've been trying to make is that there is a very important difference between simply de-emphasizing text (which is often useful, and which a moderate change in contrast might indeed achieve) and making it hard to read (so that even someone who wants to read that information will find it more difficult). What the author is advocating in this case doesn't just do the former, which would be fine; it also potentially does the latter if taken too far, as several of the author's own examples have been, and at that point the presentation certainly is undermining usability in that respect.

Perhaps my point would be clearer if we consider the opposite effect. In a magazine or textbook, a key definition that is shown in boldface is readily picked out when scanning the page, thanks to the contrast of its darker appearance compared to the main text. Likewise, a heading in a different colour is readily located, or a pull quote set in a large, italic font. All of these typographical techniques show some form of priority in the information hierarchy and guide the reader's focus, yet none requires that any of the text, including the less emphasized main body text, be at all difficult to read.

The reason I am trying to make this point so forcefully is that this is also a classic case of a mistake where young designers with good eyesight and high-end display equipment frequently fail to realise they are doing anything wrong, and as such I agree with others commenting today that the advice in the article could be counterproductive without additional qualification. Ironically, the author did flag up the related bad practice of using lighter font weights at body sizes, which is another common usability problem with some modern UI styles for much the same reasons. It's a little hard to reconcile awareness and avoidance of one danger with actively promoting another that is so closely related.

[ryandrake]

Jeez, folks, just put both designs out there in a 50/50 A/B test and use the one that converts better, results in fewer user mistakes, or whatever your business measurement is!

[Silhouette]

Many people have. Text with insufficient contrast is a frequent cause of observed usability problems. That's the point.

[ryandrake]

Cool! Shouldn’t that be the end of the discussion then? If you can measure that one design converts or otherwise performs better then that’s the right one. What other considerations are needed?

[Silhouette]

A/B tests and their statistical brethren are useful in some contexts, but as you (almost) said, they come towards the end of a discussion. First you have to identify some variations that you want to test. And of course in some cases, particularly more complex UI designs, there might be no simple success metric to measure for an A/B test, and indeed there is probably no simple set of changes you can isolate from their context either.

In short, A/B tests are great, but they're not a substitute for design skills, they're a way of refining what you already have in some situations.