In this study, subjects were presented with evidence that a fire-fighter trainee’s preference for risk is predictive of his subsequent success. Half were told that high risk taking leads to success and the other half were told that it leads to failure. This ‘evidence’ was presented in two ways, either as two concrete examples, or as abstract statistical data on 20 trainees. The subjects were then told that the evidence was bogus and were asked what they thought to be the true relationship. Interestingly, the bogus information had a strong effect both when the question was asked immediately and when it was asked a week later. If told that risk aids success, then told to ignore the statement, the subject continued to believe it, and vice versa. The concrete examples (sample of two) had a much stronger impact than the statistical data (sample of 20), and this effect was just as strong 1week later. The study has implications for the presentation of forecasts. Concrete examples have a stronger impact than do statistical data, apparently because it is easier for listeners to construct causal explanations to support the prediction.
Perhaps J. B. Watson knew this in the early 1900s. He built his fame as a psychologist with a study on conditioning that used one subject, a baby known as Little Albert. Watson’s fame continued to grow over the years despite evidence that there was no Little Albert study (Samelson, 1980). An interesting extension of Anderson’s study would be to see whether the effect holds if the subjects are told in advance (as we do in scenarios) that the predictions are hypothetical.