Preference gives more consistent judgments than oddity only if the task can be modeled as forced choice.

Abstract

Earlier data showed that subjects presented with two samples of distilled water and one of tap water were significantly more consistent in choosing the tap water as preferred than in identifying it as the odd sample in the set. The results were sometimes interpreted as demonstrating greater sensitivity for hedonic judgments than for oddity judgments. They are now shown to be explained by the statistical properties of the decision rules followed in different judgment tasks. In a new experiment, oddity and preference judgements were obtained in a replication of the original task with extra conditions. In two of the new conditions, the decision structure of a preference task was the same as that for the oddity task; in these conditions, performance was no better than with explicit oddity responses. The Thurstone-Ura model of triangle judgments proposed by Frijters predicts the results as an outcome of the greater statistical power of three-alternative forced choice tests compared with triangular tests. An excellent fit to all the data is given by a model wherein all subjects have the same d' for the difference between the water types but 25% of them prefer distilled water to tap water.