Page 464 -

456 CHAPTER 15 Working with human subjects

of an architect. Narrowing your pool of potential participants to architects would be
your next logical step, but even this limitation may not be fine-grained enough. Are
you willing to accept architecture students? This might help if there is a school of
architecture nearby, but students may lack real-world experience. This might lead you
to insist upon professional architects, who may be hard to find. HCI researchers are
familiar with these and related challenges in finding appropriate study participants.
In the early days, in the late 1970s and early 1980s, many of the participants in
HCI research were workers in corporate computing environments. This population of
relatively early users was professionally motivated to participate in studies aimed at
improving the systems that they used. As computer use spread more broadly into so-
ciety and academic groups became active centers of HCI research, students became
available (often just walking down the hall) and easily motivated (via cash or food)
pools of participants. Countless studies involving computer science or psychology
undergraduates have been published over the years.
So, what's wrong with recruiting undergraduate students—or other easily found
subjects—in HCI research? Often, nothing. If you are interested in evaluating inter-
faces intended for use by undergraduate students, this approach is perfect. However,
tests that draw on a homogeneous, nonrepresentative group of participants may be
open to criticism: results may not apply to users from a different demographic group.
Even if a specific menu arrangement in a word-processing program works well for
(predominantly young, male) computer science students, it may not work well for
retired women. In a case like this, the mismatch may simply limit the extent to which
you can claim that your study answers the problem.
The number of participants is another crucial factor. Different forms of research
require different numbers of participants. Studies with too few participants may not
yield generalizable results, while studies with too many participants are unnecessar-
ily expensive and time consuming.

15.2.1 WHICH SUBJECTS?

In selecting participants, you should strive to find people with personal attributes
and goals appropriate for your study. By personal attributes, we mean demographic,
educational, vocational, and avocational details. Some studies may simply need com-
puter users, while others need participants of a certain gender, age range, education
level, professional background, or any combination of these characteristics.
Each individual's goals, background, and motivations may play a role in deter-
mining how appropriate they are for your study. Insufficiently interested subjects
may be unlikely to contribute constructively, no matter how well they match your
other criteria. Even with the right physical attributes, an architect who is strongly op-
posed to the use of computers for modeling would probably not make a good subject
for studying the architectural tool described above. On the other hand, some stud-
ies might benefit from the perspective of less-motivated participants, who might be
more critical and less forgiving of shortcomings than enthusiasts. The participation
of these less-motivated users can be particularly useful when studying tools that may