Garth J.O. Fletcher (1994) Assessing Error in Social Judgment. Psycoloquy: 5(10) Base Rate (10)

Volume: 5 (next, prev) Issue: 10 (next, prev) Article: 10 (next prev first) Alternate versions: ASCII Summary
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 5(10): Assessing Error in Social Judgment

Commentary on Koehler on Base-Rate

Garth J.O. Fletcher
Psychology Department
University of Canterbury
New Zealand



This commentary examines the base rate controversy in the wider context of the work on errors and biases in social cognition, with remarks on a theme raised by Koehler concerning the appropriate normative model for assessing base rate usage.


Base rate fallacy, Bayes' theorem, decision making, ecological validity, ethics, fallacy, judgment, probability.


1. The central point in Koehler's (1993) target article is that, contrary to previous claims, people do not typically ignore or even underutilize base rate information in making decisions or other social judgments. Part of the fight that Koehler picks has to a large extent already been won, at least in social psychology and social cognition. There is now widespread recognition in popular social psychology texts (e.g., Myers, 1993), in social cognition texts (e.g., Fiske & Taylor, 1991), and in the research literature generally, that people do not routinely ignore base rates in making social judgments, and that base rate usage is substantially moderated by a host of conditions.

2. It is nevertheless useful to put the work on base rates in the context of the general literature on errors and biases. The work by Tversky and Kahneman, and others in the 1970s, documented several fundamental errors and biases in addition to the underutilization of base rates, including the fundamental attribution error, belief perseverance, confirmatory bias, and over-confidence in social judgment. Lay individuals came to be viewed as lazy thinkers, as incompetent statisticians, and as hopelessly biased by their frequently wrong-headed social theories (Nisbett & Ross, 1980).

3. Over the last decade, however, a flood of research and theorizing on various errors and biases has produced a seachange in this rather bleak portrayal of human social intelligence (for reviews apart from Koehler, 1993; see Fletcher, 1993; Funder, 1987; Kenrick & Funder, 1988; and Klayman & Ha, 1987). In a nutshell, this research has demonstrated the same pattern of results documented by Koehler in relation to base rate usage, namely, that there exists a patchwork of conditions under which judgment biases or "errors" are reduced or eliminated, or judgment accuracy is enhanced. In short, depending on the conditions, lay social cognition can look simplistic or complex, stupid or intelligent.


4. I suspect that another reason why the zeitgeist has swung against the earlier bleak assessments of lay social judgment, is related to the difficulties involved in assessing the rationality or soundness of subjects' responses. Many of the landmark debates in the error/bias literature have revolved around this issue (e.g., Cohen, 1981); the importance of this point is also well illustrated by Koehler's analysis of base rate usage. At root, this problem comes down to selecting the appropriate normative model with which to compare lay inference.

5. In this respect, there is a crucial difference between treating a Bayesian model as a normative versus a descriptive account of lay cognition. To use the Bayesian model as a descriptive tool involves estimating how subjects correct any prior beliefs or hunches they might have in the light of the data (in this case the relevant base rates). Often, however, the Bayesian model is used to derive the "correct" solution. This solution usually assumes that subjects should give full weight to the base rate information provided, and ignore any prior hypotheses or beliefs they might have. For example in the lawyer- engineer problem (Kahneman & Tversky, 1973), subjects who fail to ignore the individuating information (which is regarded as nondiagnostic), and give less than full weight to the base rate data, are regarded as making a mistake.

6. However, the Bayesian model is a poor instrument for assessing the rationality or soundness of lay cognition. Take, for example, one of the studies upon which the base rate fallacy myth was built by Nisbett and Borgida (1975). In this study, it was found that subjects were not influenced by base rate information concerning their predictions or causal attributions concerning the behavior of hypothetical individual subjects, who were described as participating in a psychology experiment (involving willingness to take electric shocks, or to help someone apparently suffering a seizure). As demonstrated by Wells and Harvey (1977), one reason subjects ignore the base rate information in this experiment is because it is discrepant with their prior beliefs (people are much more prone to take electric shocks, and not help other people, than most lay individuals believe). In this situation, subjects tend to dismiss the base rate information as unrepresentative of the wider population.

7. Note, first, that one cannot infer irrationality or flawed thinking from the existence of false beliefs or errors. It is all too easy to arrive rationally at false theories or beliefs with the information currently at hand (scientists do it all the time). Second, it is not necessarily correct to equate a social inference bias with below par performance. For example, it has commonly been claimed that lay individuals are overconservative when revising their prior theories in the face of disconfirming evidence. However, given the ubiquity of conflicting evidence, and the desirability of retaining a reasonably stable view of the world, such theoretical conservatism could plausibly be characterized as normatively appropriate for lay individuals and scientists alike. Bias does not necessarily equal error.

8. To return to the Nisbett and Borgida (1975) study, if we were really serious about evaluating the soundness of subjects' thinking we would need to examine how their original beliefs were derived, as well as how they weighted the base rate data provided. On both counts a Bayesian model is, in itself, not much use. There are alternative models available for assessing lay cognition, such as Thagard's (1989) model of explanatory coherence, which postulates an interlocking network of epistemic criteria for evaluating theories or explanations (e.g., simplicity, breadth, internal consistency and fertility). Provisional evidence suggests that lay evaluations of social explanations are consistent with such a model (Read & Marcus-Newhall, 1993; see also Fletcher, 1993).

9. Whatever normative account is used, it is clear that evaluations of the correctness or soundness of lay social inference are indissolubly linked to the plausibility and comprehensiveness of the standards or models by which it is assessed. For this reason alone, claims about the sagacity or otherwise of lay social cognition should always be examined with a skeptical eye.


Cohen, L.J. (1981). Can Human Irrationality Be Experimentally Demonstrated? The Behavioral and Brain Sciences, 4, 317-370.

Fiske, S.T. & Taylor, S.E. (1991). Social Cognition, 2nd ed. New York: McGraw-Hill.

Fletcher, G.J.O. (1993). The Scientific Credibility of Commonsense Psychology. In K.H. Craik, R. Hogan & R.N. Wolfe (eds.), Fifty Years of Personality Psychology (pp. 251-269). New York: Plenum Press.

Funder, D.C. (1987). Errors and Mistakes: Evaluating the Accuracy of Social Judgment. Psychological Bulletin, 101, 75-90.

Kahneman, D. & Tversky, A. (1973). On the Psychology of Prediction. Psychological Review, 80, 237-251

Kenrick, D.T. & Funder, D.C. (1988). Profiting from Controversy: Lessons from the Person-Situation Debate. American Psychologist, 43, 23-34.

Klayman, J. & Ha, Y.W. (1987). Confirmation, Disconfirmation, and Information in Hypothesis Testing. Psychological Review, 94, 211-228.

Koehler, J.J. (1993). The Base Rate Fallacy Myth. PSYCOLOQUY, 4(49) base-rate.1.koehler.

Myers, D.G. (1993). Social Psychology. 4th ed. New York: McGraw-Hill.

Nisbett, R.E. & Borgida, E. (1975). Attribution and the Psychology of Prediction. Journal of Personality and Social Psychology, 32, 932-943.

Nisbett, R.E. & Ross, L. (1980). Human Inference: Strategies and Shortcomings of Social Judgment. Englewood Cliffs, NJ: Prentice-Hall.

Read, S.R. & Marcus-Newhall, A. (1993). Explanatory Coherence in Social Explanations: A Parallel Distributed Processing Account. Journal of Personality and Social Psychology, 65, 429-447.

Thagard, P. (1989). Explanatory Coherence. Behavioral and Brain Sciences, 12, 435-467.

Wells, G.L. & Harvey, J.H. (1977). Do People Use Consensus Information in Making Causal Attributions? Journal of Personality and Social Psychology, 35, 270-293.

Volume: 5 (next, prev) Issue: 10 (next, prev) Article: 10 (next prev first) Alternate versions: ASCII Summary