Forced Grouping: A new methodology to measure the similarity between products and predefined references

In sensory evaluation, many discrimination tests (e.g. A-Not A, duo-trio, ABX, etc.) aim to detect whether a blind sample matches a predefined reference or not. Such tests require blind samples to either fully match the reference (e.g. A) or not (e.g. Not-A). But in some situations, the samples tested do not match exactly the references used. The question then is: How do the tested samples relate to these references? With the new methodology of Forced Grouping (FG), this question can be answered. The objective of this paper is to detail the methodology of FG and its analysis (including a binomial test, a Thurstonian model, and some tools to evaluate the panel and panellist performance). To illustrate the FG methodology, a total of 30 trained assessors (split into two panels) evaluated the same set of six mixtures of skimmed yoghurt (SY) and full fat quark (FFQ). Reference A consists of 75% SY/ 25% FFQ and reference B consists of 50 SY/50 FFQ, while the other four samples have a ratio in between the two references. The six products were tested in monadic sequence. After familiarization with the references, assessors were asked to associate each of these products to the closest reference, based on taste and mouthfeel separately. Additionally, FG was extended to other sample sets (approx. 40 studies) to generalise the results. The yoghurt samples showed a clear transitioning from reference A to reference B across samples with an increasing proportion of full fat quark within the sample composition. Similar results were obtained for most of the other studies, although some discrepancies were observed in particular for studies including samples with large batch variations, or samples that are very similar overall. FG has shown to be a quick and easy method, which extends our sensory toolbox to new situations where "new" in-between samples are associated to pre-defined references.