W3C home > Mailing lists > Public > public-html-a11y@w3.org > December 2009

(unknown charset) Re: CHANGE PROPOSAL: Table Summary

From: (unknown charset) Leif Halvard Silli <xn--mlform-iua@xn--mlform-iua.no>
Date: Mon, 7 Dec 2009 02:55:44 +0100
To: (unknown charset) Ian Hickson <ian@hixie.ch>
Cc: (unknown charset) Laura Carlson <laura.lee.carlson@gmail.com>, HTML Accessibility Task Force <public-html-a11y@w3.org>, Roger Johansson <roger@456bereastreet.com>
Message-ID: <20091207025544633888.29982884@xn--mlform-iua.no>
Ian Hickson, Sun, 6 Dec 2009 13:59:11 +0000 (UTC):
> On Sun, 6 Dec 2009, Laura Carlson wrote:
>> Ian wrote:
>>> Nobody has collected equivalent data showing summary="" is useful at 
>>> improving accessibility in practice,
>> Last month Roger Johansson asked on his blog, "Do you find table 
>> summaries helpful?"[1]. and received three responses from screen reader 
>> users.


> [...] I think a better way to get data 
> about this would be a set of usability studies of Web authors followed by 
> double-blind studies of the pages they write.  For example, take six to 
> nine Web developers, and give them the task of marking up some Web pages 
> that include particularly complex data tables in an accessible way that is 
> still aesthetically pleasing to them. The developers would be split into 
> three groups, one being given instructions on using summary="", one being 
> given instructions on writing paragraphs around the table, and one being 
> given no instruction at all. [...]

What would this find out? The goals must be specified. E.g. would it 
find out whether those that use @summary write worse summaries than 
those that put this stuff "nearby" the table? Specifically, I think it 
is wrong to do it just for the purpose of seeing @summary can be 
ditched or not. The whole tone of the suggested test is to create 
something that the working group _must_ believe. That's not the right 

> [...] If people agree that this is a reasonable thing to do, 
> I might (emphasis on "might") be able to  get the resources [...]

That study would take resources. And discussing the outcome of it would 
take resources. And we would not agree about the interpretation.

Rather than spending the (debating) resources on the validity of 
@summary, I think one should establish a) what kind of extra info sight 
disabled need, b) decide the ways that this could be added - were 
@summary very well could be the last option.

And, rather than only asking users, one could focus on authors. E.g. 
pick some pages with tables "out in the wild", without @summary or any 
other similar text. Then contact the authors and ask them if they would 
be willing to add @summary or change the table visual design for the 

For a study of how authors would create a new table - and how users 
would react to it, then I think the study that you suggested, Ian, 
should be changed a bit. Here is one idea:

1 Let some designers design one or several tables. Save the result.
2 Then tell them to add a @summary with info for blind. Save the result.
3 Then tell them to instead use a visual <summary> inside <caption>. 
Save the result.
4 Then tell them that they should hide the content of <summary> via 
CSS. Tell them that they may add more text to summary, if they wish. 
Save the result.
5 Then tell them to just add the summary info nearby the table. Save 
the result.
6. Analyze the results.
7. Test the the resulting tables from step 1 to 5 on users. Both 
sighted and blind.
8. Analyze the results of the user test.

Before step 2, the designers should be told what kind of info the blind 
need. And they should write the summary texts themselves. (So before 
the text can be done, it should be clear what table summary should look 

What I would like to find out is e.g. whether a dedicated slot for the 
summary would improve the result. I would also like to test if - or how 
- adding a text that is meant for summary increases the value for 
unsighted. I'm interested in a qualitative study of authors, were 
authors strive to do the best. Rather than testing different authors 
against each others.

One could also perform dedicated user tests, with some "ideal" tables 
(with "perfect" summaries) in different versions (w.r.t how the summary 
is inserted). If we know what works best, then the next problem is "how 
to get authors/user agents to do it".
leif halvard silli
Received on Monday, 7 December 2009 01:56:27 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:05:07 UTC