Perceptual variation among drinking waters: the reliability of sorting and ranking data for multidimensional scaling

Twenty untrained assessors took part in repeated similarity sorting and preference ranking tasks on five separate occasions. The same three drinking water types were assessed on each trial, and the sample set to be assessed included duplicate and blended samples. A satisfactory level of agreement between assessors' responses for each testing session was obtained, indicating that these tasks can be consistently performed by untrained assessors. Multidimensional scaling of the grouping and ranking data produced reliable and interpretable solutions (although a vector model of the preference data was not satisfactory). Duplicate pairs of samples were grouped together and different water types were spatially separated in the resulting configurations. The position of blended samples in relation to the water types of which they were comprised was also often interpretable. It is concluded that simple sorting and ranking procedures can be efficient methods for the collection of consistent sensory data from untrained consumers on qualitative variation among drinking waters, and that multidimensional scaling can provide a valid model of such data.