Odor qualities of 10 odorants were compared by 50 panelists of 4 cooperating laboratories of the ASTM E-18 Sensory Evaluation Committee. The odorants were: acetophenone, anethole, 1-butanol, carvone, p-cresylmethylether, cyclohexanol, 1-heptanol, 1-hexanol, phenylethanol, and pyridine. In one method, dissimilarity was scored directly on a 7-point scale; results from different laboratories correlated at p < 0.001, and identical odors yielded scores well-discriminable from those for non-identical odors. In another method, odors were compared indirectly, by comparing their edited multidimensional profiles obtained using Harper's scale expanded to 136 descriptors; mean profiles from different laboratories correlated at p < 0.05 in 97 percent of cases. To eliminate semantic effects, a Chi-Squared statistic was applied to the individualized descriptor usage counts, providing a dissimilarity measure, DCHI = In [(coefficient of association)]; its value from different laboratories correlated at p < 0.001. However, a distinct discrimination of identical odors was possible only after adding some descriptors specific to alcoholic odors. DCHI and dissimilarity scores related curvilinearly, but a linearity (r = 0.88) resulted for DCHI vs (score8-score). Thus, DCHI derivable from sufficiently articulate multidimensional profiles adequately substitutes for the direct dissimilarity scores; in addition, profiles indicate the nature of the dissimilarity. Fortran programs for the profile analyses are given.
No takes yet. Share an insight, caveat, or question.
Dravnieks et al. (1978) studied this question.