ORGANIZATIONALBEHAVIORAND HUMAN PERFORMANCE 19, 392--402 (1977)
Strategies for the Use of Base-Rate Information
JOHN S. CARROLL AND ROBERT S. SIEGLER
Carnegie-Mellon University
Kahneman and Tversky (PsychologicalReview, 1973, 80, 237-251) found that
subjects given both base-rate information about the prevalence of some
characteristic within a population and brief personality sketches of individual
members of the population base their predictions almost solely on the
sketches. Three experiments were done to determine the conditions under
which subjects would rely on base-rate information in making predictions
about individuals. In Experiment 1 it was found that the combination of a small
population, exhaustive sampling of that population, and population percen-
tages corresponding to an exact whole number partition of the sample (the
"translatability" feature), resulted in substantial reliance on base rates. The
results of Experiment 2 suggested that the translatability feature might be
especially important in encouraging such usage. However, the results of Ex-
periment 3 indicated that the enhanced use of translatable base-rates was
limited to situations in which uninformative personality sketches were used.
These findings were discussed in terms of Kahneman and Tversky's represen-
tativeness construct.
Kahneman and Tversky (1973) have uncovered a surprising aspect of
the human judgmental process: subjects who are provided both base-rate
information about the prevalence of some characteristic within a popula-
tion and brief personality sketches of individual members of the popula-
tion base their predictions almost solely on the clinical descriptions. The
finding holds true even when subjects believe that the clinical descriptions
are of little predictive value, when large monetary incentives are provided
for correct prediction, and when the base-rate percentages are quite ex-
treme. In a different domain, Nisbett and Borgida (1975) demonstrated
that base-rate information is ignored in favor of clinical descriptions both
in attributions about the causes of past behavior and in predictions of
future acts. The evidence for the strength and pervasiveness of the
phenomenon is so strong that the question has now become: are human
beings ever substantially influenced by base-rate information when clini-
cal information is also available?
The first purpose of the present study was to answer this question
positively--to create a situation in which subjects would rely heavily on
base-rate information. The second purpose, assuming that the first goal
Requests for reprints should be sent to either author at the Psychology Department,
Carnegie--Mellon University, Pittsburgh, Pennsylvania 15213. Order of authorship was ran-
domly determined.
392
Copyright'© 1977by Academic Press, Inc.
AIIrightsof reproductionin any formreserved. 1SSN 0030-5073