8 - Laffort, P., Héricourt, P., Valentin, D., Callegari P., 2001. Disposition of 141 odorants and 16 major semantic descriptors in a 3D olfactory space. Chem Senses, 26, 742-743 (see also Chem. Senses, 24, 56-57 and Chem. Senses, 26, 1067) The name of a fifth contributor to the four ones from Dijon could be added: Andrew Dravnieks, from Chicago, deceased in 1986. His major work on olfactory profiles (1985) allowed the present study. Detailed abstract The classification of odors according to their quality discrimination is an old challenge. Before 1968, it was solved using rules that were more empirical than experimental. In 1968, two different experimental ways were proposed and continue to be fruitfully applied: 1/ Woskow (1968) applied first a multidimensional scaling (MDS) to direct measurements of similarities for a matrix of 25 x 25 odorants. Practically, the geometrical growing of experiments for this type of procedure prevents to apply it to more than 25 odorants. Therefore, the 3D space yielded, explaining 86% of the variance, could be due to the small size of the matrix. 2/ Initiated by Harper et al. (1968) the systematic characterization of odorants by a limited number of semantic descriptors was developed by Dravnieks and the ASTM (1985)] which fixed 146 standard descriptors (valid for the North-American cultural environment). However, factor analysis of the data obtained with this method gives rise to a 17-dimensional space explaining 89% of the variance, difficult to handle (Jeltema and Southwick, 1986). Callegari (1998) combined the advantages of both procedures to characterize the olfactory quality, using an algorithm based on the ratio model of categorization of Tversky (1977) which from semantic profiles, leads to similarities comparable to that obtained directly by human subjects. The adapted Tversky algorithm applied to binarized profiles data is as follows:

S ( A, B ) =

AI B AU B

in which: S(A,B) stands for the similarity between odorants A and B A B stands for the sum of common elements of profiles for A and B A U B stands for the sum of elements of profiles used to characterize A or B The final 3D space results from the MDS applied to the similarity data derived from the semantic profiles of Dravnieks for 141 odorants, using this algorithm (about 10 000 similarities). It explains 88% of the experimental psychophysical variance, similarly to that obtained by Woskow for a matrix of 25 x 25 odorants.

References Woskow M. H., 1968. In N. Tanyolaç ed., Theories of odor and odor measurements, Istanbul, 147-188 Harper R., Bate-Smith E.C., Land D.G., Griffiths N.M., 1968. Perf. Essent. Oil Rec., 59, 22-37 Dravnieks A., 1985. Atlas of odor character profiles. ASTM Data Series 61. Philadelphia, 353 pp. Jeltema M.A., Southwick E.W., 1986. J. of Sensory Studies, 123-136 Callegari P., 1998. Thesis University of Bourgogne, Dijon, France, 201 pp. Tversky A., 1977. Psychol. Review, 84, 327-352