Abstract:

The A-Not A method is a well-known discrimination method used in sensory analysis and signal detection. The method typically involves either paired or sequential presentation of two products; on each trial the subject decides whether the product presented is A or Not A (A-). A is a product familiar to the subject, sometimes called the “signal” in signal detection experiments. The A-Not A method has been used extensively to study sensitivity and bias in individual subjects and can be used with trained or expert panels. It also has important, though less well known, applications in product testing involving consumer populations where the consumer has developed familiarity with a product through continuous use. Most product testing methods currently in use do not exploit this consumer familiarity or expertise. A potential difficulty in using the A-Not A method is that individual differences may introduce extra variation that cannot be accounted for using the conventional chi-square test commonly used for the analysis of data from this method. If this extra variation is not accounted for, erroneous significance levels will be reported. In this article we discuss an adjustment to the test statistic for this method that accounts for the effect of individual differences.