A tendency to make insufficient allowance for regression towards the mean when predicting from an imperfectly reliable predictor. Suppose that a student is described by a tutor as being much more intelligent, hard-working, and self-confident than the average student on the course. (a) What percentage of students on the course would you expect to be described even more favourably by the tutor? (b) What percentage of students on the course would you predict to obtain a higher aggregate mark or grade point average? There is more uncertainty about (b) than (a), because a tutor's ratings are an imperfectly reliable predictor of academic performance, and your prediction (b) should therefore be more regressive, and hence closer to 50 per cent, than your expectation (a). In practice, most people show little or no difference between the two percentages, and this is usually interpreted as being due to their use of the representativeness heuristic, in which evaluation and prediction coincide. Compare regression fallacy.