W3C home > Mailing lists > Public > w3c-wai-gl@w3.org > April to June 2002

Readability factors, re 4.1 etc.

From: john_slatin <john_slatin@forum.utexas.edu>
Date: Wed, 26 Jun 2002 14:14:43 -0500
Message-ID: <6AC4E20EED49D411941400D0B77E52F0074B92C7@forum.cc.utexas.edu>
To: "WCAG (E-mail)" <w3c-wai-gl@w3.org>

Thought this might be useful as we continue refining our understanding of
the issues raised by 4.1.


-----Original Message-----
From: Jim Allan [mailto:jimallan@tsbvi.edu]
Sent: Wednesday, June 26, 2002 1:06 PM
To: John Slatin

-----Original Message-----
From: Human Factors International [mailto:hfi@humanfactors.com]
Sent: Wednesday, June 26, 2002 12:55 PM
To: jimallan@tsbvi.edu


Insights from Human Factors International, Inc. (HFI)
Providing consulting and training in software ergonomics.

Every month HFI reviews the most useful developments in
UI research from major conferences and publications.

NOTE: Last month's newsletter on Parallel Design generated a
lot of interest. A question has arisen regarding the origin of this
type of design. One reader suggested it was invented in 1994,
while others say they were using this design process as far back
as the 1970s. Any of our readers who have insight into the origins
of parallel design are encouraged to email the editors at

In this issue:

Dr. Bob Bailey -- Readability Formulas - what to consider when
writing for the Web.

The Ergonomic Pragmatist, Dr. Eric Schaffer, gives practical advice.

"The Bollywood Technique" - Dr. Schaffer presents an innovative
approach to usability testing with subjects who don't like to criticize.

Bob Bailey, Ph.D., Chief Scientist for HFI

There is a considerable amount of information published on the Web
that is intended to be read by someone. There is evidence that much
of the information may be too hard to read and understand for typical

Baker, Wilson and Kars (1997) reported that the readability scores
of most articles in the 'Health Reference Center' ranged from 10th to
14th grade levels. Another study (Graber, Roller and Kaeble, 1999)
included text-based information from commercial, academic and
government sites. They found that the reading material averaged the
10th grade level. In a more recent study, a group of researchers
(D'Alessandro, et.al., 2001) conducted readability analyses of
pediatric patient education materials on the Web, and concluded
that the information was not written at an appropriate reading level
for typical users.

Readability Formulas
Readability formulas have been developed to assist writers in
preparing information. These formulas provide a means for
estimating the difficulty a reader may have reading and understanding
a paragraph, section or entire document on the Web.

The first readability formula was developed over 80 years ago, and
a number of formulas have been developed since that time. These
formulas originally were designed to help classroom teachers choose
textbooks for their students. Currently available, computer-based
readability formulas include:
 - Automated Readability Index
 - Dale-Chall
 - Flesch Reading Ease (included with Microsoft Word)
 - Flesch-Kincaid Grade Level (included with Microsoft Word)	
 - FOG
 - SMOG (Simple Measure of Gobbledygook)

Readability results will vary depending on which formula is used.
For example, the Flesch-Kincaid tool often returns a score two to
three grades lower than other formulas. Osborne (2000) proposed
that grade level equivalent scores tend to be accurate only by plus
or minus 1.5 grade levels.

There are numerous factors that affect how easy, or how hard, a
given document is to read and understand, including sentence length,
word choice, layout and formatting, overall organization of the content,
use of illustrations, etc.  However, most readability formulas consider
only two factors:
  (a) the number of syllables (or letters) in a word, and
  (b) the number of words in a sentence.

Because most readability formulas consider only these two factors,
these formulas do not reveal why written material is difficult to read
and comprehend. Most of the important attributes of writing that
contribute to reading difficulty have not yet been quantified. Fortunately,
many of the difficult-to-measure attributes are highly correlated with
the two factors that can be easily measured.

Readability formulas are most useful as predictors of reading difficulty.
Klare (1975), in a review of readability formulas, concluded that "as
long as predictions are all that is needed, the evidence that simple
word and sentence counts can provide satisfactory predictions for
most purposes is now quite conclusive."

A document classified as highly readable solely on the basis of a
readability formula could be a disorganized disaster -- or contain no
content at all. The following paragraph has a calculated readability of
the 12th grade:

Qwerty uiopas dfg hjkl zxcvb nmqw ertyuio pas dfghj klzxcvb nmq
werty ui opas dfgh jklzxc vbnm. Qwertyuiop as dfgh jklz xcvbn mqwe
rtyui opas dfghjk lzx cv bn m.  Qw ertyu iopas dfghj klzxcvb nmqwert
yuiopasdf ghjk lzxcv b nmqw ert yuiop asdf gh jk lzxcvbn m. Qwerty
uiop asdfg hjklz xcvbn mqwe rtyuiop asdfgh jklzxcv bnmq wert yui
opa sdfgh jklzxc vbnm qwerty uiopas dfghj klzx cvbnm.

Obviously, readability scores depend on the writing style rather than
the content of written material. These stylistic features are under the
control of the writer.

Reading Skill of the Intended Audience
As general rule, it is better to write a document at a readability level
that is below the reading skill level of the intended audience. Ideally,
the reading skill level of intended readers would be based on the
results of a standardized reading test (e.g., the Nelson-Denny
Reading Test). This is usually reported as a grade level, i.e., "95%
of the users in the target audience read at an 8th grade level or higher."

Usually it is not practical to administer a reading test to all potential
users. An estimate of the reading grade level of the intended audience
can be obtained by considering the users' education level. An average
eighth grader is assumed to read at an 8th grade reading level, and a
twelfth grader at a 12th grade level. People who have completed
college are assumed to read at the 16th grade level.

In general, people with more education have better reading skills
than people with less education. However, the actual reading ability
of a person does not always match his or her educational level. Coke
and Koether (1979) collected reading scores for over 200 company
employees. The group averaged a 12th grade education, and 95%
had reading test scores above the 10th grade reading level. Hilts and
Krilyk (1991) reported that adults read at least one or two grade
levels below their last school grade completed.

Summarizing several studies done in the United States and Canada,
the average reading skill level was estimated to be at around the 8th
to 9th grade (University of Utah Health Sciences Center). However,
this same study found that about one in five adults had a reading skill
level at the 5th grade level or below.

By comparing the calculated readability of a document to the reading
skill level of typical users, a writer can estimate whether a document
has a good chance of being read and understood. The readability
formula can be used as a predictor of difficulty, but should not be
used as a diagnostic tool. Readability formulas do not provide
information about how to make instructions more comprehensible.
For example, a document with a high readability level might be made
more readable by changing its format rather than its writing style.

To make written texts truly readable, website designers should
apply all the principles of clear and simple writing. Even though
using short words and short sentences will result in lower readability
scores, this does not guarantee that a document will be easier to read.

Finally, there are times when document readability issues are not as
important as other issues. Klare (1975) found that in circumstances
where time is not crucial and readers are highly motivated, the
readability of a document was of less importance. Coke (1976)
provided evidence that readability was not as important when readers
were looking for specific information as it was when users had to
remember that information.

Incidentally, the Flesch-Kincaid readability level for this article is
12th grade.

Note: the references for these studies are posted on the HFI Web
site at http://www.humanfactors.com/downloads/jun02.asp

The Ergonomic Pragmatist
Eric Schaffer, Ph.D., CPE, Founder and CEO of HFI

Good writing is important. Use short words and short sentences.
This will give a low reading grade level (RGL), which is good. It is
also important to use common words. Get a word frequency dictionary.
This lists how often each word is used from a large sample of
materials (e.g., High School English texts). So "Amend" may be
shorter then "Change". But use "Change" because it is more common.
It is handy to know how to calculate RGL. I've done that in many
meetings and it is a great way to flag poor writing. Microsoft Word
will give an RGL. It is in the "spelling and grammar" tool.

I would NOT increase the reading grade level to match your user
population. A given population may allow a higher reading grade
level. It also may allow use of certain jargon. But even people with
lots of Doctorates like simple writing.

This was written at 5th grade level. Is it OK?

"The Bollywood Technique"

Apala Chavan is the managing director of our office in Mumbai India.
She presented her fascinating new testing method at the Chi convention
this year. She called it "The Bollywood Technique" and  I'd like to
share it because I think we can all benefit.

What is the main challenge when you are usability testing in Asia?

In Asia it is impolite to tell someone they have a bad design. It is
embarrassing to admit that you can not find something. So it is very
hard to get feedback.

Apala tested a site that offered railroad tickets for sale. She used the
conventional simulation method and got little feedback. She could see
that users were not succeeding. But they would not willingly discus
the problems.

Apala then tried the Bollywood method. Now Bollywood is the
Hollywood of India. They make more movies then Hollywood. They
are famous for movies that have long and emotionally involved plots. The
movies have great pathos and excitement. In the Bollywood method Apala
described a dire fantasy situation. The participant's beautiful, young, and
innocent niece is about to be married. But suddenly he gets news that the
prospective groom is a member of the underground. He is a hit man! His
whole life story is a sham, AND HE IS ALREADY MARRIED! The
participant has the evidence and must book an airline ticket for himself
and the groom's current wife to Bangalore. Time is of the essence!!!

The participants willingly entered this fantasy and with great excitement
began the ticket booking process. Even minor difficulties they encountered
resulted in immediate and incisive commentary. They participants
complained about the button naming and placement. They pointed out
the number of extra steps in booking.  The fantasy situation gave them
license to communicate in a way that they never would under normal
evaluation methods.

I think this is a great method for the Asian markets. But I also expect
we might be able to generalize it to special situations in North America
and other places where participants may be hesitant to communicate

3-day Annual User Interface Update Seminar presented by
Dr. Robert Bailey.

REGISTER for Bob's UI Update Seminar:

The schedule for 2002 seminars can be downloaded at

Suggestions, comments, questions?
HFI editors at mailto:hfi@humanfactors.com.

Want past issues?

Subscribe? - http://www.humanfactors.com/library/subscribe.asp

Do NOT want this newsletter?
REPLY to this message or e-mail mailto:unsubscribe@humanfactors.com
with a Subject of: "Unsubscribe Newsletter"

Incoming mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.371 / Virus Database: 206 - Release Date: 6/13/2002

Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.371 / Virus Database: 206 - Release Date: 6/13/2002
Received on Wednesday, 26 June 2002 15:14:48 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 16 January 2018 15:33:41 UTC