Alt verbosity [was: Example canvas element use - accessibility concerns]

Re:

http://lists.w3.org/Archives/Public/public-html/2009Apr/0230.html

John Foliot wrote:

> based upon your perspective of a sighted user who uses a
> text only browser with frequency, you appear (to me) to be invoking your
> perception into the spec unevenly.  Specifically I point to your verbose
> @alt values, both in this example as well as examples in the current draft
> specification.  However countless others have noted on numerous occasions
> that when it comes to images, many (most? all?) non-sighted users wish to
> choose between a verbose or terse description, similar to 'glancing at an
> image' vs. 'studying an image'.  This is an important distinction which
> seems to be lost here.

[snip]

> Oh, so the specification is being written then based upon your personal
> perspective and annoyance level?  Ian, you constantly talk about use-cases
> and 'proof' and examples in the wild - that opinion alone is not sufficient.
> Yet because you consider a brief descriptor of an image coupled with an
> optional longer, verbose text potentially annoying to you, you dismiss it as
> non-viable?  Outside of your personal experience, do you have any proof that
> the majority share your opinion?
>
> While not 'conclusive', WebAIM's survey results (and interpretations) of
> screen reader users state:
> "The tendency toward the briefer alternative [text] also increased slightly
> with screen reader proficiency"
> [http://webaim.org/projects/screenreadersurvey/#images]
>
> Given those findings, I will suggest my 'opinion' is based upon more than my
> personal preference, and seems to directly contradict your opinion.

John, I doubt the WebAIM survey results are worth citing as evidence on 
this particular topic.

Leaving aside the thorny questions of how far the survey is 
representative of typical screen reader users and the question of 
whether the "slight.." difference under discussion is statistically 
significant (the published data doesn't allow us to say), I think 
there's a mismatch between the question posed in the survey and how 
you're using the results.

Users were asked whether they preferred "Photo of The White House" or 
"The White House" as "alt" text.

The long text is only three syllables longer than the short text. I'd 
suggest that respondents' view of whether the word "photo" provided 
useful information was at least as much a factor in how people answered 
the question as the marginal difference in length. Some users may have 
felt that the category of image (photo? painting? sketch?) was 
irrelevant; some users may have felt that the category of image could be 
guessed from the context; some users may have felt that the image 
sounded irrelevant and that, in the absence of blank "alt", a short 
"alt" was preferable.

Also, the question did not imply that if "The White House" was supplied 
as "alt" text then "Photo of The White House" would be made available as 
a longer description. So even if this response indicated users preferred 
short alternatives, it would be problematic to use this response as 
indicating that what users want is short alternatives _and_ long 
alternatives, rather than one alternative that keeps to the point. (A 
good control question might leave out "img" and involve users choosing 
between some concise text and some prolix text that says more or less 
the same thing!)

I doubt there's an advantage for short text that leaves out useful 
detail when you're just reading through a page lineally. Why summarize 
an "img" element but not a "p" element? Most of the time, you're going 
to need to expand both to understand the page, in which case you've just 
given the user more to process.

However, I can imagine moving from "img" to "img" trying to find the 
image you want, and that short titles for those images could be useful 
so you didn't need to listen to a full alternative. But is "alt" - 
intended for alternatives - really the best attribute to use for such 
short titles? Might the "title" attribute be a better way to provide 
such them, or "aria-label" and "aria-labelledby" from ARIA?

http://dev.w3.org/html5/spec/Overview.html#the-title-attribute

http://www.w3.org/TR/wai-aria/#nameref

--
Benjamin Hawkes-Lewis

Received on Sunday, 26 April 2009 18:15:55 UTC