This is considered input  to the section "The figure and figcaption elements" in the ALT-techniques document.

Please state  and operate under the assumption  that the <figcaption> element captions the entire <figure> element, and describe how to use <figure> accordingly. Also, it seems 

For reference, see these bugs against the HTML5 spec:

Bug 10480    add role="presentation" on the ASCII fish image
Bug 10483   <figcaption> should be considered the caption of <figure> _itself_
Bug 10484   default roles for figcaption and caption: Make it simple to achieve accessibility

              Background:

FIRSTLY: In a blog comment this summer, you went against ARIA 1.0 when you suggested following coding style for an img element in a photo site: [1] 

     ]] I would recommend that the image have no alt attribute and an aria-labelledby that references the <figcaption> as figure/figcaption are not supported by ATs [snip] This way the image is identified as an image to the user and it is explcitly labelled. [[

Examplified in code. where the use of role's taking the A11Y TF's ARIA mapping proposal [2] as basis:

<figure role="" ><!-- no role -->
  <img src="example.png" role="img" aria-labelledby="caption" >
  <figcaption id="caption">A visible text caption labeling the image.</figcaption>
</figure>

By contrast, ARIA 1.0 recommends an empty @alt in the same use case  and it also has a different use of @role and aria-labelledby. The ARIA 1.0 example: [3]

<div role="img" aria-labelledby="caption">
  <img src="example.png" role="presentation" alt="">
  <p id="caption">A visible text caption labeling the image.</p>
</div>

If the ARIA 1.0 example were to be expressed using a <figure> element instead, then it would have looked like this:

<figure role="img" aria-labelledby="caption">
  <img src="example.png" role="presentation" alt="">
  <figcaption id="caption">A visible text caption labeling the image.</figcaption>
</figure>

SECONDLY: In the alt-techniques document, you make a similar statement:

     ]] When supported by browsers and assistive technology the content of the figcaption will be explicitly associated with the image. [[

The correct thing to say, however, is that the <figcaption> will be explicitly associated with the _figure_ element. I honestly don't know what happens when, as you suggested this summer, the <img> lists the <figcaption> as its caption  would it mean that the <figcaption> stops being an caption to the <figure> element? (See bug 10484  perhaps the default role of <figcaption> should depend on the roile of the <figure>.)

So, taking my above recast of the ARIA 1.0 example as a figure element as starting piont, what kind of explicit association would supporting browsers and assistive technology provide once <figure> is supported? Answer: the only explicit association it would offer, would assocaiton between the <figcaption> and the <figure> element. Thus, we would   in principle, but perhaps not as best practice   not need to use the aria-labelledby attribute which the ARIA 1.0 example uses (and we could also remove role="presentation" due to alt=""). Thus we end up with this:

<figure role="img" >
  <img src="example.png" alt="">
  <figcaption>A visible text caption labeling the image.</figcaption>
</figure>

This example is also pretty much in line with the HTML5's ASCII art figure example  see bug 10480.

Also: if the figure element above did _not_ have role="img", then <figcaption> would caption an element which has no role ... according to the ARIA mapping prposal[2]. (_That_ to me sounds irrantional - but I guess that's another bug against another spec - figure should have _some_ default role, I think.)

[1] http://rebuildingtheweb.com/en/no-alt-text-for-photo-sharing-sites/#c20100729051201
[2] http://www.paciellogroup.com/blog/misc/HTML5/aria-html5-proposal.html
[3] http://www.w3.org/TR/wai-aria/roles#presentation