Re: The permissible uses of heading levels and F43

On Fri, May 4, 2012 at 4:11 PM, <noreply@w3.org> wrote:

> Name: Duff Johnson
> Email: djohnson@commonlook.com
> Affiliation: NetCentric Technologies
> Document: W2
> Item Number: Success Criterion 1.3.1
> Part of Item:
> Comment Type: general comment
> Summary of Issue: The permissible uses of heading levels and F43
> Comment (Including rationale for any proposed change):
> I am trying to reconcile the fact of F43 with a common (but by not means
> general) impression that SC 1.3.1 does not require logical hierarchy in
> headings.
>
> Success Criterion 1.3.1 requires that if headings are in the content that
> relationships are conveyed programmatically.
>
> Now, illogical heading hierarchy does not, as has often been claimed,
> leave “equivalent information” available to everyone.
>
> This claim rests on the incorrect assumption that every user will find a
> given structural anomaly disconcerting (or not) to a similar degree. This
> is obviously not the case as WAI is, I know, aware. Page-position alone can
> provide information that effectively removes an "important" but
> non-structural "structure" element from the sighted user's perception of
> the content's organization.
>
> Instead, the reality is that users who depend on heading structure for
> navigation receive a substantially less effective navigation facility when
> the structure model is abused for presentational purposes.
>
> The whole point of 1.3.1 is to eliminate practices that reliably botch the
> conveyance of “information, structure and relationships” to the user by way
> of normatively requiring “programmatic determinability”. Elegant! I applaud
> F43. It embodies 1.3.1 precisely because it fails the misuse of structure
> elements.
>
> Now, there is – of course – a big variable in impact based on the nature
> of the content. This issue matters a lot less on short (typical web-page)
> content with few headings. On longer or highly structured content it is
> entirely fair to state that misuse of section headings for emphasis or
> styling kills the utility of headings as a reliable means of navigation for
> those who depend on structure to operate their AT.
>
> When it doubt the reading of the normative text should err towards the
> side of accessibility rather than its opposite.
>
> I’m not suggesting that headings must be used – that’s absurd. I am
> suggesting that IF headings are used to structure content (ie, to serve the
> “section heading” role), then they must make sense to pass 1.3.1.
>
> In any event, we are attempting to implement F43. It seems clear that F43
> suggests that we fail content with illogical heading levels unless some
> other “programmatically determinable” means is provided to discern the
> “real” headings from the other “uses” of <H#> tags. Whatever that could be.
>
> I have discussed this precise question with at least two-dozen experts in
> ICT accessibility over the past four months and have received almost the
> same number of opinions. ☹
>
> In the use-cases of MY great interest – longer and structured documents –
> the significant majority of experts I consulted preferred to consider valid
> heading structure as normatively required, even if they were not sure
> whether or not WCAG 2.0 required it.
>
> Proposed Change:
> I think the Committee should rectify the (apparent) delta between F43 and
> the otherwise-stated view that “WCAG does not require strict hierarchy in
> headings”, including as 'represented' in H42, example 2.
>
> Further, I think the Committee should consider that very significant
> use-cases exist in which 1.3.1 becomes substantially meaningless if it is
> NOT understood to require logically structured headings.
>
> I believe it's a lot easier to write Techniques for 1.3.1 if one takes
> this view, and far easier to implement as well. More to the point, the
> quality and accessibility of content will improve at every level.
>
> ================================
Response from the Working Group
================================
There is a difference between not using headings according to hierarchy,
and having what is visually presented differ from what is programmatically
presented.

There is no requirement that headers be used in a strictly hierarchical
manner.  So if header levels are skipped  (visually and in markup) there is
not a failure.

But if header A appears visually to be the parent of header B, but  B is a
higher level header than A in markup, then what is presented in markup
clearly is not the same information as what is presented (visually) to
everyone else.

Whatever the logic of the presentation is for sighted people, that same
logic needs to be evident programmatically.

F43 clearly describes the use (abuse) of structural elements (such as
headings) to format things that are not headings, and are not visually
representing heading on the page.   The failure is not about logical
nesting, but about using heading markup (or other semantic markup) to
achieve a visual effect for something that is not the thing markup was
intended for.  (e.g. using header to bold text that is not a header.


RE Skipping levels:
Things are different for hierarchical heading structures that *skip* levels
of headings *where this is justified by the marked-up content*. To remain
with the "Fruits and vegetables" example in H42, the main h2 sections
(Fruit, Vegetables)  as well as the individual h3 sections (Apple, Orange,
Broccoli, etc) may have a subsection with "Further links". It would make
perfect sense to mark up all these sections with h4, not mainly for visual
stylistic reasons (all h4 headings would be rendered the same size) but
also for the sake of consistency when navigating the document with a screen
reader. This means that for the main sections (Fruit, Vegetables), h3 would
be skipped.  Traversing the headings listing of a screen reader on level h4
would consequently read "Further links on Fruit", "Further links on
Apples", "Further links on Oranges", and so on.

There are also well established a11y practices in HTML design that choose a
particular (low) heading level such as h6 for (hidden) section headings.
Other practices will use h1 for (hidden) section headings ("navigation",
"content") and start content headings on h2 or below. Given the diversity
of established a11y practices, it makes little sense to be too prescriptive
(such as ruling out the skipping of heading levels) regardless of the
domain context.


Loretta Guarino Reid, WCAG WG Co-Chair
Gregg Vanderheiden, WCAG WG Co-Chair
Michael Cooper, WCAG WG Staff Contact


On behalf of the WCAG Working Group