Biases of Probability Assessment: A Comparison of Frequency and Single-Case Judgments

Gigerenzer (1991) has recently argued that there is a fundamental distinction between judgments of relative frequency and judgments of single-case probabilities. One of Gigerenzer′s major findings is that many representativeness effects, such as the neglect of base rates and the conjunction error, disappear when questions are reworded in terms of relative frequencies. In this paper, we claim that Gigerenzer′s distinction between frequency and single-case judgments maps onto Tversky and Kahnemans (1974) distinction between representativeness and availability. Specifically, we suggest that representativeness effects occur primarily in single-case judgments, while availability effects occur primarily in judgments of relative frequency. In Experiment 1, 170 undergraduate subjects were given four "representativeness" problems and three "availability" problems. Half of the subjects made judgments of relative frequencies, while half provided single-case probabilities. The results generally support our predictions. Representativeness effects tend to be stronger in the single-case condition, while availability effects tend to be stronger in the frequency condition. In Experiment 2 ( N = 165), we used slight modifications of the original availability problems. Again, the availability effects were more prevalent in the frequency condition than in the single-case condition. The normative and descriptive implications of these findings are discussed.