The matching paradigm as a method for validating a sensory profile

Sixteen naive subjects were asked to match five coffee aromas to five consensus descriptions generated by 16 trained subjects. This task was repeated six times. After each repetition subjects were individually corrected. The same experimental paradigm was also used with nine additional naive subjects whose task was to match five salad dressings. For both products, the proportion of correct matches was at chance level at the first repetition and increased somewhat with the repetition. However, a Correspondence Analysis (CA) showed that, as soon as the second repetition, most of the samples are relatively close to their description. The experiment was also performed with the trained subjects who produced the descriptions. When these trained subjects received the consensus descriptions, their performance was very similar to that of naive subjects. But, when they received their own descriptions, their performance slightly increased. This performance increase appears clearly on the CA graphs. The potential reasons for the low level of observed performances are discussed.