E-mail List Archives

Search E-mail List Archives for

Number of posts in this thread: 3 (In chronological order)

From: Jennifer Sutton
Date: Fri, Sep 19 2014 3:26PM
Subject: Likert scale -- making radio buttons in a table include their question
No previous message | Next message →

Greetings, WebAIM:

Has anybody built a Likert scale, using radio buttons in a table? I'm
working with a client, and we're getting different results with
different screen reader/browser combos (what else is new)?

What's generally at issue is the association of the question as the
user selects the different radio buttons -- strongly disagree,
disagree, neither, and so forth. There are five choices.

It seems to me that the screen reader user should hear the question
each time. But if folks feel otherwise, I'd like to know.

Anybody got a working example to share?

I thought this article would do the trick, including use of the
radiogroup role and ARIA Describedby:
Fieldset Legends, aria-describedby, and radiogroup roles last-child
http://www.last-child.com/legend-aria-describedby/


but it's not quite getting us there.

What about fieldsets and legends in tables? They're not playing nicely.

Thanks for any thoughts.

Best,
Jennifer

From: Ryan E. Benson
Date: Fri, Sep 19 2014 4:08PM
Subject: Re: Likert scale -- making radio buttons in a table include their question
← Previous message | Next message →

This is one of the instances I say we can use the title attributes versus
labels, since these are usually laid out in a table anyway, so if the AT
fails to read the title, they can find the cell headers and make the
connection. Depending on the question type, i recommend the answer
(strongly agree) or a snippet of the answer after it as the value of the
title attribute.

I've thought about fieldsets in tables, but seemed like it would be a large
mess.

--
Ryan E. Benson

On Fri, Sep 19, 2014 at 5:26 PM, Jennifer Sutton < = EMAIL ADDRESS REMOVED = >
wrote:

> Greetings, WebAIM:
>
> Has anybody built a Likert scale, using radio buttons in a table? I'm
> working with a client, and we're getting different results with different
> screen reader/browser combos (what else is new)?
>
> What's generally at issue is the association of the question as the user
> selects the different radio buttons -- strongly disagree, disagree,
> neither, and so forth. There are five choices.
>
> It seems to me that the screen reader user should hear the question each
> time. But if folks feel otherwise, I'd like to know.
>
> Anybody got a working example to share?
>
> I thought this article would do the trick, including use of the radiogroup
> role and ARIA Describedby:
> Fieldset Legends, aria-describedby, and radiogroup roles last-child
> http://www.last-child.com/legend-aria-describedby/
>
>
> but it's not quite getting us there.
>
> What about fieldsets and legends in tables? They're not playing nicely.
>
> Thanks for any thoughts.
>
> Best,
> Jennifer
>
> > > >

From: Hans Hillen
Date: Fri, Sep 19 2014 4:48PM
Subject: Re: Likert scale -- making radio buttons in a table include their question
← Previous message | No next message

Hi Jennifer,

The grouping you want for the radio buttons (fieldset, role="radiogroup",
role="group") conflicts with the table structure, as you can't wrap the
contents of a row in a separate container and you can't give the TR a
different role without breaking the table structure.

So you can redundantly repeat the question for each radio button (using the
title attribute or a hidden label element), that will work reasonably
consistently across screen readers and will also ensure that the question
is announced for each radio button (which I don't necessarily agree with).
Or you can use aria-describedby like in the article you linked to.

Both those options lead to a lot of redundant announcement of the same
question though. Which is not ideal.

Technically speaking, having the question and individual radio button
answers marked up as row and column headers provides enough information for
the AT to work with, but this cannot be relied upon consistently between
UA/AT combinations.

So the short answer is: There is no perfect solution here.

On Sat, Sep 20, 2014 at 10:08 AM, Ryan E. Benson < = EMAIL ADDRESS REMOVED = >
wrote:

> This is one of the instances I say we can use the title attributes versus
> labels, since these are usually laid out in a table anyway, so if the AT
> fails to read the title, they can find the cell headers and make the
> connection. Depending on the question type, i recommend the answer
> (strongly agree) or a snippet of the answer after it as the value of the
> title attribute.
>
> I've thought about fieldsets in tables, but seemed like it would be a large
> mess.
>
> --
> Ryan E. Benson
>
> On Fri, Sep 19, 2014 at 5:26 PM, Jennifer Sutton < = EMAIL ADDRESS REMOVED = >
> wrote:
>
> > Greetings, WebAIM:
> >
> > Has anybody built a Likert scale, using radio buttons in a table? I'm
> > working with a client, and we're getting different results with different
> > screen reader/browser combos (what else is new)?
> >
> > What's generally at issue is the association of the question as the user
> > selects the different radio buttons -- strongly disagree, disagree,
> > neither, and so forth. There are five choices.
> >
> > It seems to me that the screen reader user should hear the question each
> > time. But if folks feel otherwise, I'd like to know.
> >
> > Anybody got a working example to share?
> >
> > I thought this article would do the trick, including use of the
> radiogroup
> > role and ARIA Describedby:
> > Fieldset Legends, aria-describedby, and radiogroup roles last-child
> > http://www.last-child.com/legend-aria-describedby/
> >
> >
> > but it's not quite getting us there.
> >
> > What about fieldsets and legends in tables? They're not playing nicely.
> >
> > Thanks for any thoughts.
> >
> > Best,
> > Jennifer
> >
> > > > > > > >
> > > >