to alter the clinician’s own behavior or that of
others.
Prediction to Unknown Situations
Clinical inferences and predictions are likely to be
in error when clinicians are not clear about the
situations to which they are predicting. Inferring
aggression from the TAT is one thing; relating it
to specific situations is another. Furthermore, no
matter how careful and correct clinicians are, an
extraneous event can negate an otherwise perfectly
valid prediction. Take the following example from
the Office of Strategic Services (OSS, a predecessor
to the Central Intelligence Agency) assessment
program:
One high-ranking OSS officer, while
operating abroad, received a letter from a
friend of his in America informing him
that his wife had run off with the local
garage man, leaving no message or address.
As a result the officer’s morale, which had
formerly been high, dropped to zero. The
assessment staff could predict that a small
percentage of men would have to cope
with a profoundly depressing or disquiet-
ing event of this sort, but, again, it was not
possible to guess which of the assessees
would be thus afflicted. (OSS Assessment
Staff, 1948, p. 454)
Common sense should suggest that to accurately
predict a person’s behavior, the clinician must con-
sider the environment in which that behavior will
take place. This is also a tenet of behavioral assess-
ment. However, clinicians are frequently asked to
make predictions based on only imprecise and
vague information regarding the situation in which
their patient will be living or working.
In a hospital setting, a clinician may be
requested to provide a prerelease workup on a
given psychiatric patient. But the information avail-
able to the clinician will too often cover only gen-
eral background, with supplementary descriptions
of individual differences. Investigators agree that
such data are subject to a ceiling effect that will
allow correlations of no better than .30 to .40
between the data and subsequent behavior (e.g.,
Mischel, 1968). To say the least, correlations of
that magnitude leave a great deal to be desired.
Therefore, personality data alone are likely to be
insufficient in many prediction situations.Fallacious Prediction Principles
In some instances, intuitive predictions can lead
clinicians into error because they ignore the logic
of statistical prediction. Intuitive predictions often
ignore base rates, fail to consider regression effects,
and assume that highly correlated predictors will
yield higher validity (Garb, 1998; Kahneman &
Tversky, 1973). For example, suppose that a clini-
cian is assessing a patient by collecting samples of
behavior in a variety of situations. Even though
observations reveal an extremely aggressive person,
the clinician should not be surprised to learn that
eventually the person behaves in a non-aggressive
fashion. Regression concepts should lead one to
expect that exceptionally tall parents will have a
shorter child, that brilliant students sometimes do
poorly, and so on.
In addition, clinicians’ own confidence can
sometimes be misleading.For example, Kahneman
and Tversky (1973) showed that individuals are
more confident when they are predicting from cor-
related tests. More specifically, although clinicians are
often more confident of their inferences when they
stem from a combination of the Rorschach, the
TAT, and the MMPI rather than from a single test,
Golden (1964) could find no evidence to support this
confidence. The reliability and validity of clinical
interpretations did not increase as a function of
increasing amounts of test data. One should always
seek to corroborate one’sinferences,butitwouldbe
a mistake to believe that the validity of inferences is
inevitably correlated with the size of the test battery.The Influence of Stereotyped Beliefs
Sometimes clinicians seem to interpret data in terms
ofstereotyped beliefs(Chapman & Chapman, 1967).
For example, Golding and Rorer (1972) found thatCLINICAL JUDGMENT 299