The Magical Number: Seven, Plus or Minus Two
by Matthew Ellison, Puget Sound Chapter
Introduction
Anyone who has worked in technical communication long enough will have
come across the magical number "seven, plus or minus two". It's
often quoted by trainers such as myself as a guideline governing anything
from the number of categories of information that you should include in
a Web site through to the number of steps in a procedural Help topic.
Over recent years many of us have been guilty of applying it without being
fully aware of its origin; we typically have some vague idea that it concerns
a limitation of the human brain that prevents us from being able to process
more than nine items effectively, but few of us know the details of the
original research that produced the concept of this magical number. And
still fewer of us choose to question its applicability to information
design - it's far too convenient a rule of thumb to risk invalidating
it!
This article explains the origin of the magical number, "seven,
plus or minus two" and discusses the extent to which using it as
a golden rule for information design is justified.
George Miller and the Psychological Review
The term "seven, plus or minus two" was coined by George Miller
50 years ago when he wrote an article in the Psychological Review journal.
It's interesting to note that at that time computers were in their infancy,
and no one had even begun to imagine the concept of GUI-based software
applications, Help files, and the World Wide Web - so clearly the concept
of "seven, plus or minus two" was not intended as an aid for
designing such information systems. As a matter of fact, George Miller
was writing purely from the viewpoint of a cognitive psychologist, and
his article made no reference to the field of technical communication
whatsoever.
The article described a variety of psychological experiments that fell
under three distinct categories:
- Absolute judgment (the ability to distinguish between musical tones
or other stimuli)
- Subitizing (the ability to recognize numbers)
- Short-term memory
Despite the fact that these categories appear unrelated to one another,
all three areas of experimentation seemed to indicate that the number
seven was somehow significant.
Absolute judgment
George Miller cited several examples of research in this area. A typical
example is the experiment conducted by I. Pollack in 1952: he asked listeners
to correctly identify a series of tones that were evenly distributed on
a logarithmic scale. With up to five or six different tones, listeners
rarely confused them. But with significantly more tones, they made frequent
mistakes. The conclusion was that most people have an upper limit that
is the region of seven tones. Other experiments showed that the same results
hold for variations of loudness, or brightness, of saltiness (when tasting
food), and in fact of any one-dimensional variable.
Subitizing
Subitizing means being able to judge the number of a collection of randomly
arranged items more or less instantly. The common die (or dice) used in
many board games relies on a human's capacity to be able to recognize
the numbers from one to six without hesitation. Even if the dots were
not arranged in their familiar patterns, most people would still be able
to call the numbers without needing to actually count them.
Miller described the research of Kaufman, Lord, Reese, and Volkmann at
Mount Holyoke College, Massachusetts. They flashed patterns of dots on
to a screen for 1/5 of a second, and asked observers to report on the
number of dots. For up to six dots there were almost no errors. Above
that number, however, inaccuracies began to creep in, and for significantly
higher numbers observers could only provide a rough estimate. So again,
seven appears to be the significant number where change occurs.
Short-term memory
George Miller referred to a number of experiments on short-term memory,
all of which indicated a limit of around seven items. However, Miller
was anxious to point out the difference between these results and the
ones relating to absolute judgement. He explained that absolute judgement
is limited by the amount of information, whereas the span of short-term
memory is limited to the number of items or chunks. A chunk could potentially
contain multiple bits of information; for example, most of us would have
no problem remembering five three-syllable words, where the number of
actual sounds or syllables that we are remembering totals 15. By combining
(or what Miller called "recoding") information into chunks,
we can remember more than 7 bits of information. This explains why long
telephone numbers are split into groups of two, three, or four numbers,
and is also the principal behind many of the tricks and methods for performing
what appear to be fantastic feats of memory.
George Miller's conclusion
Contrary to the urban legend that has developed around the number seven
since the original article, Miller himself was far from certain about
its status as a magical number. In concluding his article he suspected
that the recurrence of the number seven in the three categories of research
described above was "only a pernicious, Pythagorean coincidence."
His summarizing remarks were actually more general observations relating
to human limitations on the amount of information that we are able to
receive, process, and remember. He expressed the view that recoding is
an important process that deserved further research, and he has since
gone on to oversee the development of WordNet, a semantic network for
the English language.
Applying "seven, plus or minus two" to navigation design
One obvious application of the magical number is as a limiting factor
for the number of items in hierarchical navigation systems such as tables
of content, menus, and Web site home pages. In 1998, two researchers from
Microsoft, Kevin Larson and Mary Czerwinski, decided to compare the effectiveness
of three different navigation structures that all led to a set of information
articles taken from the Encarta Encyclopedia. The navigation structures
were each organized as follows:
- 8 top-level menu options, each leading to 8 further menu options,
each leading to a menu of links to 8 different articles
- 16 top-level menu options, each leading to a menu of links to 32 different
articles
- 32 top-level menu options, each leading to a menu of links to 16 different
articles
The total number of available information articles was in each case 512.
Larson and Czerwinski requested their test subjects to locate specific
articles using each of the three navigation structures in turn, and measured
their performance in terms of the time taken and the number of errors
or dead-ends. To their surprise, they found that the first navigation
structure (the only one to conform to the "seven, plus or minus two"
rule for the number of menu options) was actually the least successful.
It turned out that the number of decisions that the subjects had to make
during the navigation process was actually more critical than the number
of options presented at each decision point.
Other commentators have since added fuel to the argument against over-reliance
on the rule. As D. LeCompte put it in a 1999 paper for the Human Factors
and Ergonomics Society, "knowing that most people can successfully
remember between 5 and 9 items about half of the time gives us virtually
no useful guideline for design of user interfaces."
Applying "seven, plus or minus two" to instructions
Is there any reason why we should be unhappy with a procedure containing
only three steps since it falls outside the magical range? Should we limit
the number of steps in a procedure to nine?
The answer to the first of these questions is almost certainly no. In
fact there is an argument that, if you need your audience to remember
a set of instructions in the correct order, then you are pushing your
luck if you include more than three steps. The answer to the second question
is: "it depends". The factors that influence the number of steps
that users can process successfully include:
- Whether the user's environment allows them to read the instructions
and carry out the steps simultaneously
- Whether the user needs to remember the steps, or simply carry them
out successfully as they read the instructions
- The user's familiarity with the context and subject matter
- The complexity of the language of the instructions
- The length of each step
According to Susan Harkus in a 2003 paper: "the principle is simple:
design decisions are relative to the usage context and the nature of the
information."
Information Mapping®
Information Mapping is a widely used system for structuring and presenting
information both on screen and in printed form. The reason for mentioning
it in this article is that it is heavily based around the number seven;
it identifies seven fundamental types of information, and also recommends
breaking information into no more than seven sections.
The Information Mapping® method was initially described by Robert
Horn in 1966 and thoroughly documented in 1969 in Information Mapping
for Learning & Reference (Horn, Nicol, Klienman & Grace, 1969).
Interestingly, the original documentation and justification of the Information
Mapping® method makes no reference either to Miller or to any of the
research cited by Miller in his article. However, I think it very likely
that Horn was aware of Miller's work, and may well have been inspired
by the idea of a "magical number" as he undertook the task of
devising a set of memorable guidelines for information design.
The benefits of Information Mapping® have been demonstrated though
a number of tests and surveys. As an example, in 1994 the County of San
Diego surveyed more than 480 people who had received both a traditional
and an "information mapped" document. Of those people, more
than 80% preferred the Mapped document because it was simpler, easier
to identify the main points, and more concise. However, it is debatable
whether this proves the magical properties of the number seven. It may
simply be an indication that a document that has been carefully structured,
laid out, and signposted using common sense principles is easier to use
than a less well-structured document - which would hardly be a surprise
to our technical communication community.
Conclusion
The use of the "seven, plus or minus two" guideline in the
field of navigation design and information architecture is a tempting
rule of thumb. However, the research into cognitive psychology that led
to George Miller's original article has doubtful direct relevance to our
field of technical communication, and many commentators have urged caution
in treating the number seven with undue respect. Our choice of the number
of items in a list, the number of sections in a table of contents, and
so on, should be influenced by a number of factors - not least, the nature
of the cognitive task that is presented to the user, and the existing
knowledge that the user brings to the task. If we reject the rule of thumb,
do we have a better one with which to replace it? Perhaps three (the number
of items that almost everyone can remember in any order) is a potential
candidate for "magical number" status?
Matthew Ellison has 19 years experience as a user assistance
professional in the software industry. Much of this time was spent managing
a team of writers and trainers at a UK-based consultancy company, before
enjoying a period in the U.S. as Director of the WinWriters (now WritersUA)
Conference. Matthew has been a popular speaker at WritersUA events throughout
the world since 1997, and has covered a diverse range of topics from context-sensitive
Help, to the Spice Girls. He now runs his own independent UK-based training
and consulting company that specializes in online Help design and technology.
Matthew holds a B.Sc. in Electronic Engineering and a Post-Graduate Certificate
of Education from Bristol University in the UK. He is a certified instructor
for RoboHelp, Captivate, and WebWorks ePublisher Pro. Matthew's website
is www.ellisonconsulting.com.
This article originally appeared in the Summer 2006 edition
of ISTC's Communicator Journal.
|
