W3C home > Mailing lists > Public > w3c-wai-ig@w3.org > January to March 1998

RE: ALT text survey [was RE: WD-WAI-PAGEAUTH-0203 and the use of ALT text

From: Charles (Chuck) Oppermann <chuckop@MICROSOFT.com>
Date: Mon, 9 Feb 1998 17:05:41 -0800
Message-ID: <E3A3FFB80F5CD1119CED00805FBECA2F013BBD91@red-msg-55.dns.microsoft.com>
To: "'Kasday, Leonard'" <kasday@att.com>, w3c-wai-ig@w3.org
Once again, I must caution.  Do not think of ALT as how it's represented in
LYNX or any other browser.  As we move to the W3C's Document Object Model,
we need to be thinking of HTML elements as objects. As such, we need to
think of how we use these attributes to property identify the object.


-----Original Message-----
From: Kasday, Leonard [mailto:kasday@att.com]
Sent: Monday, February 09, 1998 9:14 AM
To: w3c-wai-ig@w3.org
Subject: ALT text survey [was RE: WD-WAI-PAGEAUTH-0203 and the use of
ALT text



	There are differing opinions on ALT text.  About a year ago
today I sent out a survey to several listserv's asking for people's
opinions.  The survey contained both general questions and specific
questions about e.g. bullets and horizontal lines.

	The results showed that while some blind individuals preferred a
purely function approach, others said, quite emphatically, that they
wanted to know what was really there- just so long as it was brief.
I've attached some of the results of the survey to this email.

	These were the basis for the alt text guidelines now mandatory
on the AT&T web site.  These guidelines are listed at 

	http://www.att.com/style/alttext.html

	(this will need to be updated to reflect the availability of
LONGDESC in HTML 4.0)

	Len

	Survey results appear after my signature below



All opinions expressed here are my own, not necessarily those of my
employer.
=============================================================
kasday@att.com         phone 732 949 2693

Leonard R. Kasday
Room 1J-316A
AT&T Laboratories
101 Crawfords Corner Rd.
Holmdel NJ 07733


Results of survey: part 1.  this is edited version of what i put on the
webwatch list last year.


The first three questions of the  ALT text survey sent out  February 4
1997 were as follows:


1.  Should ALT text should be included for so-called "non-essential 
images"? 
2.  Should ALT text describe what's actually in an image or provide a 
functional equivalent?
3.  How should small repetitive items like fancy bullets be described?
4. How should horizontal lines be described?

There were also questions about some complicated rules I proposed for
punctuating ALT text with square brackets.   Looking back a year later,
these rules are complicated even for me and I've omitted results from
this email.  

I received 11 responses, eight from people who described themselves, 
explicitly or by implication, as blind or visually impaired, and three 
sighted individuals experienced in this field. 

The first question asked about the following rule:

    Rule 1: Designers should provide ALT text for all images.  There are
    absolutely no exceptions: it includes images the designer thinks are
    not needed for using the page, although ALT text in such cases   
    should be very brief, e.g. 2 or 3 words.  (I realize this 
    contradicts some recommendations published on the net.)

The rating scale was:

0 means very bad 
5 means neutral
10 means very good

There was general support for this rule, some of it emphatic. Here's the

responses, in a 2-row format that attempts to be universally accessible 
to people using speech, Braille, and sight.  The first row is a scale, 
consisting of the numbers 1 through 10, spaced across the page,  The 
second row is the data, consisting of a list of the ratings that people 
gave, lined up with the corresponding number on the scale above it.   
Also, to help people using Braille and sight, the number 10 is 
represented by a single letter, T.

Scale: 1     2     3     4     5     6     7     8     9    T 

data:  1     2     3                             8     9 9  T T T T

The mean is 7.2

As you can see, most responses are clustered around 10, with three 
between 1 and 3.   Both blind and sighted respondents were about 2 to 1 
in favor of ALT text for non-essential information.

The reasons for including text for all images were:

> I'd rather have access to all the information available on a page and 
> decide for myself what is irrelevant, than to have some sighted person

> deciding what I have the right to see

> When designers have to start thinking about which things to tag and 
> which not to then they are going to think it takes too much work.

> I think a user is immediately put off by images that refuse to offer 
> up any information whatever... If it's really *that* irrelevant, the 
> author shouldn't have used it in the first place


The reasons for not including the text were:

> people browsing without graphics, for whatever reason, are doing so 
> to get useful information as efficiently as possible

> anything overdone is lost to it's own importance

> You don't want to overdo it.

However, both of the blind individuals who objected to this as a general

rule were positive, later in the survey, about ALT text for images that 
might be considered nonessential.  One of these individuals, commenting 
about the ALT text

Welcome to Archimedes Plumbing [Archimedes in bathtub].

wrote:
> If I hadn't seen the tag, I wouldn't have known he was in the tub and 
> that makes all the difference between what a person looking at the web

> can see and what a person listening to the web can hear...we're almost

> equals in that case.

The other individual described the ALT text "smiling customer 
representative" as
> Very informative and yet, not too wordy.

So both of these people appreciated, in these examples, ALT text for 
images which a sighted web designer might not consider necessary for 
using a page.  Thus, overall, 8 out of the 8 blind respondents either 
demanded or at least appreciated in particular cases ALT text for such 
"non-essential" images.  This implies, at the 95% statistical confidence

level that for a large sample at least 60% of blind individuals would 
have responded this way.


I next asked about literal descriptions.  Here's the wording:

  Rule 2.  Pictures should be described literally 

   For example, if there's a picture of a customer representative, and
   there's no text in the image, a suitable ALT tag could be:

       [smiling customer representative]

    How ever, "customer service" would not be a suitable  ALT tag,    
    since, although it's related to the image, it doesn't actually    
    describe the image.


I might not have made this rule very clear, judging by some of the 
comments.  I didn't mean "literal" to mean "long".  I just meant to say 
that the actual contents of the image, rather than it's meaning, should 
be described.  For example, if a web page for a restaurant had a picture

of a doorman beckoning you in, this rule says that the ALT text should 
be something like

[Doorman beckoning you in]

not  something like "welcome", even if that's what the web designer was 
trying to convey by the image.

Here's the data, again using a universal histogram.  It includes one 
rating of "3" which I assigned for someone who didn't give a number but 
whose comment indicated that he was against the rule.

Scale: 1     2     3     4     5     6     7     8     9     T 

data:  1     2     3 3         5                       9 9 9 T T T

The mean was 6.8


As you can see, the data is clustered near 9 and 10, although there's 
about the same number of responses spread between 1 and 5.


Comments from people favoring the rule include:

> Just like descriptive video service, tell it like it is and let me
> interpret it for myself.

> As long as it's brief and to the point.

> very informative and yet, not too wordy


Comments from people opposing the rule include:

> Why have I put an image on the page?  To make it look nice - in which 
> case I would describe it literally... or, to convey a concept - in
> which case I would label it with the intended message.

> Sometimes literal is appropriate sometimes a brief concept is 
> appropriate

> This is a matter of taste. Meaning should be clear, but explicit in
> terms of the subject as well

> It would be nice if everything was described but it isn't essential.

> The shortest, the best.


It's also informative to look at what people thought about alt text for 
bullets in a bullet list which were each "tiny blue spheres".   This was

question 3 in the survey. Here's the data for four alternatives, again 
in the universal histograms format. If people didn't give numerical 
ratings on this question I converted their answers into 3, 7, or 9 
depending on their comments.  This gives some artificial clustering at 3

and 7. 

Some of these alternatives use the following punctuation characters [ ] 
*


Scale:             0   1    2    3    4    5    6    7    8    9   T   

Data for:
[tiny blue sphere] 000           3                   777       9

*                  00                 4              777       9 9  
           
[*]                0             3         5    6    777       9    T 
             
[bullet]                         3333                     88   99   T

       

Here's a table of the average ratings:

Description          average rating
___

[tiny blue sphere]   4.0
*                    5.1
[*]                  6.0
[bullet]             6.2

As you can see, all the alternatives had their fans and their opponents.

"Tiny blue spheres" had the lowest average score and the most zeros.  
Apparently, the prospect of hearing "tiny blue spheres" repeated over 
and over outweighed for some people the principle of knowing what the 
page looked like, although others gave that description high marks.  
"Bullet" had the highest average score (although this is not 
statistically significant given the small sample). As for using pure 
punctuation marks, some liked the fact that they could turn them off by 
turning off punctuation; others objected to the fact that didn't really 
describe what was being displayed.

Moving along, I asked how people thought about various ALT text
alternatives for horizontal lines that are sometimes used to break up a
page.  Of course, HTML has a logical way to do this logically, the <BR>
tag, but some authors prefer lines with specific appearance.
 
Here's the mean ratings for four alternatives.  This time I didn't
provide ratings for people who omitted numbers, since it was unclear how
to do so.  

Description              mean rating
[horizontal line]         6.3
[line]                    4.6
[break]                   4.3
[------------------------]5.8  

The ALT text "horizontal line" was the winner here, although since there
There were only 6 or 7 ratings for each, this should be taken with a
grain of salt.   Some people felt it was a bit long.

 

CONCLUSIONS.

These data imply that most or all of the blind individuals browsing the 
want to know, or at least have the option to know, the content of all 
the images on the screen, including images that a web designer might 
think they don't really need to use the page.  At the same time, it 
should be possible to browse the page quickly and efficiently, without 
spending too much time or putting up with too much clutter from 
excessive ALT text, especially for small, repetitive items like bullets.
Received on Monday, 9 February 1998 20:06:56 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 19 July 2011 18:13:38 GMT