Sensory discrimination testing and consumer relevance

- Two products will always be found to be different if the sample size is large enough.
- Discrimination testing requires the consideration of consumer relevance.
- The limitations of the proportion of discriminators concept are described·
- The idea behind the consumer rejection threshold is outlined.
- Setting relevance with a paired preference or a same-different test is described.

In order to ensure consistency in the decision making process over time, a discrimination testing program must take into account all of five elements: The testing protocol, the sample size, the Type I error (α), the Type II error (β) (power = 1 − β) and a measure of the threshold above which the scientist has established that the difference is meaningful to the consumer (δR). Two putatively different products will always be found to be different provided that the sample size is large enough. This fact underscores the need to set δR. The concept of discriminators is attractive but flawed, as the same underlying sensory difference will result in different proportions of distinguishers depending on the method used. Prescott, Leslie, Kunst, and Kim (2005) proposed the idea of consumer rejection threshold which avoids the pitfalls of the proportion of discriminators concept. However, it is limited to differences that can be linked to a specific compound, such as one responsible for a product defect or off-flavor. In this manuscript two alternative approaches are discussed. The first one uses a special feature of the same-different protocol which permits the estimation of the size of the sensory difference above which consumers would call two products “different”. The second one links the estimate of a standardized measure of sensory difference, d′, to consumer hedonic response between the product pairs and finds the threshold above which a sensory difference results in a meaningful preference result. Experimental research is needed to study the suitability of these approaches. Ultimately, establishing δR is essential to ensure that results from a discrimination testing program are actually relevant to the consumers whose behavior it is trying to predict.