Abstract

Fallacies are the ‘ideal types of improper inference’, named only because they represent a common or seductive error. Naming them facilitates identification (reducing ‘false negatives’ in argument evaluation), but increases the risk of false positives; it is essentially a cost-effectiveness issue whether to introduce a new name. Statistical fallacies include errors of elementary experimental design, but also conceptual confusions, e.g. of cause with correlation, of association with guilt, where an illicit substitution is made. The focus here is on recent nationwide efforts to replace criteria of merit with correlates of success, in the evaluation of teaching. This involves a number of mistakes, including ‘precipitate decision’, confusing the normative with the descriptive, and using minimax when optimizing or maximin is appropriate, as well as various legal and ethical blunders