Test Bias. It is important to remember that sig-
nificant differences between mean scores on a test
for different groups do not in and of themselves
indicate test bias or discrimination. Rather, test
bias or discrimination is a validity issue. That is, if
it can be demonstrated that the validity of a test
(e.g., in predicting criterion characteristics or per-
formance) varies significantly across groups, then a
case can be made that the test is“biased”for that
purpose. In other words, a test is biased to the
extent that it predicts more accurately for one
group than for another group.
An example can illustrate these considerations.
Let us assume that the authors developed a person-
ality inventory measuring the trait“hostility.”As
part of the standardization project for this test, the
authors discovered that men scored significantly
higher than women on this test. Does this indicate
that the test is biased? Not necessarily. The authors
found, in a series of validity studies, that the rela-
tionship (correlation) between hostility inventory
scores and the number ofverbalfights over the suc-
ceeding 2 months was quite similar for both men
and women. In other words, the predictive validity
coefficients for the two groups were comparable;
similar hostility scores“meant”the same thing (pre-
dicted a comparable number of verbal fights) for
men and women. On the other hand, it is quite
possible that the strength of the correlation between
hostility scores and physical fights over the next
2 months is significantly greater for men than for
women. In this case, the use of the test to predict
physical aggression in women would be biased if
these predictions were based on the known associ-
ation between hostility scores and physical fights
found in men.
Several general points should be clear. First,
differences in mean scores do not necessarily indi-
cate test bias. In the previous example, there may be
good reasons men score higher on average than
women on a measure of hostility (e.g., hormonal
differences or other biological factors may lead to
higher levels of hostility for men). In fact, to find no
difference in mean scores might call into question
the validity of the test in this case. Second, the
pronouncement of a test as “valid,” although
frequently seen in the clinical psychology literature,
is incorrect. Tests may be valid (and not biased) for
some purposes but not for others. Finally, one can
“overcome”test bias by using different (and more
appropriate) prediction equations for the different
groups. In other words, bias comes into play
when the clinical psychologist makes predictions
based on empirical associations that are characteris-
tic of another group (e.g., men) but not of the
group of interest (e.g., women). The goal is to
investigate the possibility of differential validity
and, if found, to use the appropriate prediction
equation for that group.Computer-Based Assessment. Computers have
been used for years to score tests and to generate
psychological profiles. Now they are also used to
administer and interpret responses to clinical inter-
views, IQ tests, self-report inventories, and even
projective tests. The reasons given for using com-
puters include cutting costs, enhancing clients’
attention and motivation, and standardizing proce-
dures across clinicians. However, it is important to
remember that computer systems can easily be mis-
used, either by those who are poorly trained or by
those who endow computers with a sagacity that
transcends the quality and utility of the information
programmed into them.
The increasing use of Internet-mediated psy-
chological assessment raises a number of issues
(Buchanan, 2002; Naglieri et al., 2004). First, rela-
tively few studies have demonstrated that Internet-
based versus traditional psychological tests possess
the same psychometric qualities. For example,
online tests may be less reliable or less valid for
certain predictions than their traditional counter-
parts. Second, online assessment may also be subject
to confounding factors like a lack of control over
the testing situation, distractions, or technical pro-
blems. Finally, other concerns like test security and
implications for those from culturally and linguisti-
cally diverse backgrounds reinforce the ethical
responsibilities of clinical psychologists in both
administering and interpreting scores from
Internet-administered tests. In the end, it is always
necessary to establish the reliability and validity of250 CHAPTER 8