Decision strategies for the A Not-A, 2AFC and 2AFC-reminder tasks: Empirical tests

Signal detection theory provides an approach to modelling sensory difference tests that separates estimates of discriminability from the effect of response bias. However, assuming an incorrect decision strategy can also lead to inaccurate estimates of sensitivity, the most common index of which is d′. Using signal detection theory, Hautus, van Hout, and Lee (2009) describe and develop a number of models for the two-alternative forced choice with reminder (2AFCR) task; a task whose trial structure is identical to that of the constant-reference duo–trio task. There are alternative decision strategies (e.g., β or τ), when judging stimuli presented in the 2AFCR, to the “comparison of distances” strategy assumed to be used in the duo–trio task in food science. We investigated the decision strategies adopted by judges in the 2AFCR task using recently developed detection-theoretic models tested on purposefully collected discrimination data. The results indicate that, in the experimental context employed, the optimal β and τ strategies are more likely to be adopted by judges in the 2AFCR task than the “comparison of distances” strategy. Findings have implications for sensory difference testing and the validity of the sensitivity indices they provide.

► We compare taste-discrimination data from A Not-A, 2AFC, and 2AFCR tasks.
► Equal-variance assumption found satisfactory for models.
► Judges shown to use a β or τ decision strategy rather than a comparison of distances decision strategy.
► Instruction did not influence performance or the decision strategy adopted.