Researching Abnormality 159
Random assignment
Assigning participants to each group
in a study using a procedure that relies
on chance.
Sampling bias
The distortion that occurs when the
participants in an experiment have not been
drawn randomly from the relevant population
under investigation.
Population
The complete set of possible relevant
participants.
Sample
The small portion of a population that is
examined in a study.
Internal validity
A characteristic of a study, indicating that
it measures what it purports to measure
because it has controlled for confounds.
External validity
A characteristic of a study, indicating that
the results generalize from the sample to the
population from which it was drawn and from
the conditions used in the study to relevant
conditions outside the study.
inadvertently assign the people who
smile at you to the “non-loss” group.
Whether or not it is conscious (inten-
tional) or unconscious (unintentional),
a tendency or influence that distorts
data—which ends up producing a
confound—is called bias. This is why
researchers place participants in groups
usingrandom assignment, assigning
participants to each group by a proce-
dure that relies on chance.
Even if you randomly assign par-
ticipants to groups or have them take
part in both conditions, other biases
can interfere. Perhaps all your partici-
pants come from the same city, which
recently experienced a devastating hurricane (and so the theme of loss is particu-
larly salient for all participants).Sampling bias occurs when the participants are not
drawn randomly from the relevant population. Sampling bias needs to be avoided if
you want to be able to generalize (i.e., extrapolate) from the people in your study to
the population at large. This brings us to an important distinction: The population
is the complete set of possible participants (e.g., all rats or all people, or in certain
cases all people of a particular age, gender, or race). The sample is the small portion
of the population that is examined in a study.
Many of the same points also apply to the items that are selected to be used as
stimuli in the study. For example, you might select a very powerful “loss” fi lm and a
limp “depressing” fi lm—or vice versa. In this case, an irrelevant characteristic of the
items—how powerful they are—might determine the results. Just as you need to be
careful in sampling and assigning participants to groups, you need to be careful in
selecting the items (such as the specifi c items in a questionnaire), ensuring that they
are in fact representative of the types of items to which you want to generalize.
Internal and External Validity
A study has internal validity if it controls for possible confounding variables. Inter-
nal validity means that variations in the independent variable are in fact responsible
for variations in the dependent variable (or variables, in studies in which more than
one type of measurement is taken) and that the results are not a by-product of other,
extraneous variables.
A study is said to have external validity when the results generalize from the sam-
ple (the particular participants who were tested) to the population from which it was
drawn and from the conditions used in the study (such the particular movies shown,
in the example study) to similar conditions outside the study. If a study does not have
internal validity, it cannot have external validity. In contrast, even if a study has inter-
nal validity (its results have not been confounded), this does not guarantee that it will
have external validity (that its results apply to relevant situations or other people).
Quasi-Experimental Design
Ideally, the participants in a study are randomly assigned to groups. But, in many cases,
random assignment is not ethical, desirable, or possible. For instance, when trying to
test hypotheses about why a disorder develops, researchers cannot “assign” one group
to have a particular set of genes or brain functioning, a particular way of thinking, a
particular type of traumatic experience, or particular friends or families. Therefore, in
trying to understand possible causes of psychopathology, researchers often use quasi-
experimental designs, which rely on groups that already exist. In fact, the “experiment”
that we have been discussing is—like much research on psychopathology—a blend of
experiment (manipulation of independent variable and measurement of dependent vari-
able) and quasi-experiment (the selection of participants from pre-existing groups).
Suppose you want to know what the prevalence
is in the United States of the type of depres-
sion that results from seasonal changes in the
amount of daylight (sometimes referred to as
seasonal affective disorder). You are a researcher
in Portland, Oregon, and are collaborating on
the study with researchers in Boston, New York,
Chicago, and Detroit. Participants in the study
are drawn from those fi ve cities. Do you see the
sampling bias problem? All these cities are in the
northern half of the United States and therefore
have a different number of daylight hours than
does the southern half. Whatever prevalence
rate was calculated would not refl ect the entire
United States.
Corbis/Punchstock