August 2002 - Prepared for Food Quality & Preference

Sensory segmentation uses descriptive attributes from all sensory inputs (appearance, aroma, taste/flavor, texture) to predict overall liking. It clusters panelists together on the basis of the sensory profile at which overall liking maximizes. This paper introduces metrics to analyze and better understand the segmentations that emerge when the sensory attributes are limited to one class of sensory inputs (e.g., focusing on texture attributes), versus when specific sensory input liking replaces the more general attribute of overall liking. These metrics are 1) the scatterplot relating two complementary segments on the same attribute, and the correlation between the segments, 2) the range of liking ratings as revealed by the standard deviation of the means across products, 3) the nature of the 'smoothed' sensory-liking curve , 4) area under the sensory-liking curve, 5) magnitude of the directional scales. The results show 'sharpening' in segmentation for the sensory input used for segmentation. This sharpening is clearest for texture attributes, less so for appearance, and least for flavor.

Click here for complete paper