Consistency of Multidimensional Scaling Models Derived From Odor Sorting

Abstract

ABSTRACT. The odor perceptions of six groups of subjects who varied in their degree of experience and training in odor evaluation were examined using a sorting task. Data were submitted to multidimensional scaling to produce perceptual models. Regardless of training or expertise, models of different subject groups were highly similar as measured by interpoint distance correlations (median r = 0.74). The retest correlation of one group tested twice with no intervening training was 0.85. A two‐dimensional model provided a better fit to the data of fragrance descriptive panelists than to that of other subjects, but the configuration of these trained panelists was qualitatively similar to all other groups. The method of odor sorting as analyzed by multidimensional scaling was relatively unaffected by learning or other cognitive/linguistic influences on odor perception and showed a fair degree of reliability and inter‐group consistency in odor quality perception.

The role of olfaction in the elaboration and use of the Chardonnay wine concept

“…Panellists were asked to sort samples into as many groups as they wanted, based on their odour similarity. This non-verbal method has been used in several studies (Lawless & Glatter, 1990;Solomon, 1997) to assess similarities. Samples were presented according to a Latin square.…”

Sorting procedure as an alternative to quantitative descriptive analysis to obtain a product sensory map

“…Several authors showed that the dimensions of multidimensional scaling (MDS) configurations resulting from an evaluation made by untrained panelists can be interpreted using the vocabulary generated (Chollet & Valentin, 2000;Hollins et al, 1993;Lim & Lawless, 2005;Lawless, 1989;Lawless et al, 1995). Lawless and Glatter (1990) showed that perceptive maps of odors obtained with trained or untrained panelists were relatively similar. Besides Chollet and Valentin (2001) demonstrated that vocabulary used by untrained panelists was not as accurate as that of trained panelists.…”

Collection and analysis of perceived product inter-distances using multiple factor analysis: Application to the study of 10 white wines from the Loire Valley

“…Even for a relatively large set of products, this sorting can be performed during a unique session: e.g. 13 liquids flavoured with vanilla (Heymann, 1994), 18 odors (Lawless & Glatter, 1990), 17 odors (Macrae, Rawcliffe, Howgate, & Geelhoed, 1992), 13 waters (Falahee & Macrae, 1995). In practice, these data are analysed after an aggregation into a matrix whose general term (row i, column l) is the number of subjects having sorted the products i and l in the same group.…”