| People often lend excessive weight to case cues and insufficient weight to base rates, an effect known as base-rate neglect (Tversky & Kahneman, 1982), even when subjects experience base rates and case cue accuracy instead of being told about them (Goodie & Fantino, 1995). This effect disappears when the cues bear no prior relationship to the outcomes (the direct-symbolic effect, Goodie & Fantino, 1996).; In four experiments, subjects witnessed hundreds of trials in which they predicted which of two events would occur. One event occurred more often than the other (the "base rate"), and a cue provided additional information (the "case cue").; In Experiment 1, the direct-symbolic effect obtained within a single experiment for the first time. Additionally, over 1600 trials, data did not stabilize at probability-matching, but tended toward optimal proportions, challenging an accepted "fact" of experimental psychology (Myers, 1976). Largely because of this, the direct-symbolic effect was not apparent in within-subjects manipulations.; In Experiment 2, two alternative explanations for base-rate neglect and the direct-symbolic effect were tested: that cues control behavior in proportion to their salience, and that base-rate neglect depends upon pre-existing response biases. Both possibilities were rejected, bolstering confidence in the reality of base-rate neglect. Additionally, between- and within-subjects comparisons were conducted to measure subjects' belief updating within each trial. The hypothesis was supported between subjects, but updating was not found within subjects, restricting the implications of the between-subjects finding.; In Experiments 3 and 4, the posterior probabilities associated with one cue were held constant while those associated with the other cue varied. In three comparisons, overall sample accuracy outweighed base rates in determining choice proportions following the invariant cue.; Together, these findings demonstrate that with identical cues and outcomes, base rates have less power to influence choices than they should within a Bayesian model. Cue validity has consequently more power than it should. While these preparations and findings seem to exemplify a behavioral/cognitive synthesis (Rachlin, 1989), they merely extend a pattern of converging research that has been going on for a long time. |
