Statistical properties of simple sensory difference tests: Confidence limits and significance tests

The Normal approximation to the binomial distribution is used to obtain confidence limits for the proportion of correct responses in a simple sensory difference test and the proportion of assessors able to discriminate between two products. Improved approximations, which should be considered with small panels or when the proportion of correct responses is close to 0 or 1, are discussed. To permit comparison of the discriminating ability of different panels, or investigation of the effect of a change in the products assessed, tests of the significance of the difference between two proportions are presented, based on the Normal approximation and the x² statistic. The methods are illustrated with some data from an investigation on fish flavour.