Page 468 -

460 CHAPTER 15 Working with human subjects

The nature of the participants required for your study often plays a role in this de-
cision. Studies that involve systems for general use by a broad range of users should
be able to attract a suitably large pool of participants, even if hundreds of people are
needed. On the other hand, research aimed at studying very specific populations may
need to rely on substantially smaller pools of participants: there simply aren't tens of
thousands of potential participants for the study of a tool for space shuttle astronauts.
Studies of domain experts often face challenges in this regard.
Finding a suitably large participant pool can be particularly challenging for research
involving people with disabilities (see Chapter 16 for more information). In addition
to being an often-overlooked segment of society, people with disabilities often face
significant challenges in transportation, making trips to research labs difficult. Studies
with these users are often smaller, tending towards observational case studies with two
or three users (Steriadis and Constantinou, 2003), rather than controlled experiments,
see Chapter 16 for more details.
The time required for each participant is another important factor. Studies that
require a single session of limited length (perhaps a few hours) can enroll larger num-
bers of participants than ethnographic observations that may involve several days
or controlled experiments that require multiple sessions conducted over a period of
weeks. As the time required from each participant—both in terms of direct involve-
ment and the elapsed interval from start to finish—increases, it becomes more difficult
to recruit and retain people who are willing to commit to that level of involvement.
How many participants should your study have? You should start by using your
design as a guide. Ethnographies and case studies can be successfully completed
with as few as two or three people. Numbers vary wildly for controlled experiments:
although studies with as few as 12 users are not uncommon in HCI, results with 20
or more users are more convincing. From that base, you might expand to involve as
many subjects as you can reasonably afford to include. You should then add a few
more for pilot tests, replacements for participants who drop out, and a margin for
error. Investigation of related work in the research literature can help in this regard:
basing your population on a population used in similar prior work can be a good
strategy. If there is no clearly related work, you might be able to use a smaller popu-
lation, and perhaps an experimental methodology isn't the most appropriate method
to start with (see Chapter 1 for more information).
15.2.3 RECRUITING PARTICIPANTS

Once you have determined who your participants are and how many you need, you
must find them and convince them to participate.
If you work for a large corporation that frequently performs user studies, you
may be able to draw upon the expertise of a dedicated group that maintains rosters of
people interested in user studies and generates participant pools for research. Those
who don't have such resources available (i.e., most of the professionals who conduct
HCI studies) generally must do their own legwork.
The characteristics of your desired participants play an important role in deter-
mining how you will go about finding them. If you have relatively few constraints,