WebAIM - Web Accessibility In Mind

E-mail List Archives

Re: [EXTERNAL] heading question

for

From: Duff Johnson
Date: Dec 18, 2019 10:51AM

Had a little time, so wanted to offer a few thoughts about what's bothering me in this discussion:
Various people are saying that headings matter, although opinions differ on whether it's the mere fact of a heading that matters, or the heading level as well.
No-one has explained how users are supposed to be able to know when heading levels are structural (i.e., when the organization of headings structures the content) rather than merely indicating "importance". These two uses are seemingly impossible for users to disambiguate. As a result, AT users doubtless fail to spot cases in which navigating headings is essential to comprehending the document. Presumably, they suffer a dramatic shortfall in equal access as a result.
In short, it feels a lot like web accessibility folk tend to assume that:
content occurs in short web-pages
content doesn't rely on nested heading levels for organization
This might be true for most web pages, but it would be wrong to premise generic content accessibility standards on these presumptions.

I come from the land of PDF. It's a land of large documents (tens, hundreds, thousands of pages) and deeply structured documents (some have thousands of headings). In PDF we sometimes see documents with 7, 8, 9 or more heading levels. We have documents with "front matter" and "backmatter" and other concepts that are foreign to web content and thus relatively less considered by WCAG.

PDF 2.0, published in 2017 and soon to be updated in 2020, introduces a <Title> structure element that functions similarly to HTML's <title> element, but occurs on the PDF page rather than in the page's <head>. Accordingly, in PDF 2.0, document titles aren't tagged <H1> as this is a heading tag, not a document title tag. This allows for clear disambiguation of titles from headings, and reserves headings for the vital task of structuring the document in an accessible manner.

A member of the PDF Association's PDF/UA Technical Working Group recently wrote an article discussing this question:

https://www.pdfa.org/how-to-tag-titles-in-pdf-documents/

NOTE: the PDF/UA Reference Suite on pdfa.org <http://pdfa.org/>; is not yet updated along these lines; this is in part because this specific suite needs to remain PDF/UA-1 conforming files, which means PDF 1.7. Nonetheless, we're currently upgrading the files in that suite to make them forward-compatible with the forthcoming PDF/UA-2, and will let the group know when available.

Duff.


> On Dec 16, 2019, at 09:41, Jonathan Avila < <EMAIL REMOVED> > wrote:
>
> It's also likely that for many simple web pages the heading level isn't as important -- especially if the heading text is explanatory and the progression of headings implies meaning. In other cases the general order of the heading levels may be enough -- h3 may not come directly have an h1 but most folks would agree that an h3 is of less importance or a sub section. I'd agree with Birkir that landmarks have the potential to be better in many situations. For example, using just heading levels the main part of a page would have an h1 and the sections of a footer would have to have h2 to conform to the level approach. However, you don't have any details that they are in the footer and they actually probably should be less important than an h2 -- for example an h4. This gets into the messy world of headings where they denote structure but also can communicate importance in terms of weight. Most web pages don't visually show the headings needed for complete levels for screen re
> aders and this would necessitate adding off-screen headings.
>
> In terms of other types of document such as legal, financial, or research documents where the heading text is not clear and the order heading levels is necessary to understand the content then the order is absolutely needed in order to understand the relationship of text and headings. So it really depends on the context and that is why the WCAG criteria is flexible in determining when it is a failure or not.
>
> Jonathan
>
>