Preference without a Difference
Rousseau, B., Ennis, D. M. (2017). IFPress, 20(1) 3-4.
There are certain universally accepted tenets that support the continued adoption of sensory evaluation methods. One of these strongly held beliefs is that a person cannot have a preference between two perceptually identical products. This idea justifies the existence of difference testing sensory programs in many consumer product companies and is the basis for the argument that if an internal panel cannot perceive a difference, then a consumer will not. Consequently, a consumer will not have a preference. To use the principle requires a method to decide if two products are different. Typically the basis for that decision depends on the use of a difference testing method, such as the triangle test, followed by a statistical test of the results based on a null hypothesis of no difference. In many cases, failure to reject the null hypothesis from the results of an expert or experienced internal panel are used as an indication that consumers are unlikely to have a preference. An advantage to this type of procedure is that relatively rapid and inexpensive testing can be conducted that may obviate the need to conduct expensive consumer preference tests when a difference is not detected.
As with any declaration of a fundamental principle, it is usually in the application of the principle that the difficulties arise. In this technical report it will be shown how a frequently used difference testing method may rarely identify a difference but a consumer preference test may show that one product is preferred to another. It will be shown that although the tenet stated earlier may be abstractly justified, difference and preference test results depend critically on the methods chosen to obtain the data and the form of analysis used to interpret them. Difference testing interpretation using significance tests leads to binary decision-making (‘go/no go’). Differences in methodologies and the magnitude and precision of the detected sensory difference are not considered and may lead to inexplicable results. While internal testing may lead to a conclusion that there is no evidence of a difference, a consumer preference test may disagree.
Table 2. Triangle test results for four comparisons.