Re: Baby Steps or Backwards Steps?

Gregory J. Rosmaita wrote:
> 3. how does leaving alt out entirely when an image is not "purely 
>    decorative" "better" serve someone merely by indicating the 
>    presence of an image?

Seriously, of the following alternatives, which is better for a 
non-decorative content image, in the cases where good alternative text 
is not available?

1. No alt attribute.
2. Empty alt attribute.
3. Alt text with a redundant value (one that just repeats surrounding 
content such as its heading or caption)
4. Alt text with a value like alt="photo" or other similar value.

Given that the practical effect of requiring the alt attribute 
everywhere, including photo sites like Flickr, is typically one of those 
options, what benefits do options 2, 3 or 4 offer over option 1?

We all agree that good quality alternative text would be ideal and that 
the spec should strongly encourage authors to provide that.  However, 
this issue is explicitly about how to handle cases where that is 

Here are a few observations of what some photo sites use for alt text:

* Flickr: Redundantly repeats the image's title
* Photobucket: Provides the file name and repeats the image title and 
username of the person who submitted it.
* Facebook photo album: No alt attribute.
* Gregory's own photo album: Uses various values, depending on the page:
   [1] Appears to list the tags associated with the image.
       e.g. "perception - photography - image interpretation - blindness"
   [2] Uses the image file name
       e.g. alt="IMGRD3ONW8CG9.png"


The observations indicate that, given the current requirement for 
including alt text, some sites that are designed for hosting user 
generated content, are attempting to fulfil that requirement by 
generating alt text from other metadata, regardless of how inappropriate 
it is.

The hypothesis is the various forms of alt text that such sites generate 
is of no significant benefit, if not more harmful, than not providing 
any alt text.  Thus, by encouraging such sites to omit the alt attribute 
in the absence of any good alternative text, the user experience for 
those who cannot see the image will either be unaffected or slightly 

I believe this hypothesis can be tested using something like following 
methodology.  I'm no expert in user testing, so this could probably use 
some improvement.

1. Collect a representative sample of pages from various photo sharing
2. Gather a group of test subjects who cannot see the photos, possibly
    using a variety of different assistive technology (screen readers,
    braille, etc.).
3. The test should include various user agents, including JAWS, Windows
    Eyes, etc.
4. Perform user testing on those pages, where each test group is given
    the same set of pages, each with a particular variation of the alt
    attribute for the photo.
    a. The default alt attribute provided by those pages
    b. An empty alt attribute
    c. No alt attribute.

For the hypothesis to be confirmed, the results should indicate that the 
user experience for no alt attribute is equal or better than an empty 
alt attribute, which in turn should be better than the generated alt text.

Since Joshue O Connor offered his skills in user testing, I'm sure he'll 
have more expertise in the area and may be able to help out.

Lachlan Hunt

Received on Thursday, 16 August 2007 10:34:54 UTC