ODOR AND OPTICAL ACTIVITY 21 Id Tl-•d I i d-• Pinene Trd I © t-• Pinerie-... TTt t-a Pinene--,--[t-a Pinene epoxlde],• :•t 'rlTl Pinocorvyl I • propionate .SZ:da !__1© j • Pinoco rveol IL•,-J•,Myrtenøl •) Pinoacetoldehyde I '•' d•, Pinocorvyl I • propionate i • Myrteno_l • Myr. te. nol Pinocarveol 3) • oce,o, •z• d• I T•d• I Pinoacetaldehyde. _ I Pinocarvyl propionote • •,• . t Pinocorveo•11 L• '}.__..i (:i ) Myrtenol_ I _l•..Myrtenol • L• •oceto• ¸ Pinoacetaldehyde Pinocarvyl • propionate (D (D © • ocetol.o•iT 3•- I.a Pinocarveol •Trr ta L ! • Pinoacetaldehyde Key •co Fig. 1 The three materials were derived: (1) From /-citronellol ex/-pinene via/-pinane and/-dihych'o myrcene. The citronellol was hydrated to 2,6-dimethyl octan-2, 8-diol, which was oxidized to l rotatory "hydroxy citronellal" (4, 5). The distilled product was prep-trapped to 98% purity. (2) From d-citronellol, ex d-citronellal from citronella oil. The same hydration, oxidation and trapping procedure gave d "hydroxy citronellal" of 98% purity with practically the same glc pattern for the remaining impurities as in (1). (3) From d-citronellal ex citronella oil by the standard hydration pro- cedure via the bisulfite addition compound. The final 98% trapped material had a glc pattern differing from those of (1) and (2). In spite of the similarity of the glc patterns of (1) and (2), and their dissimilarity to that of (3), the panel test, conducted with experts, resulted

22 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS in the selection of sample (1), the I material, as the "odd" product with a chi square probability of less than 0.1% that this was by chance. The results, for all the test compounds, for both general and expert panelists, are shown in Table II, as probabilities (levels of significant differ- entiation) derived from a chi-square test. Table iI Probability levels of significant differentiation between pairs (d-l, d-2) and (l-l, /-2) Compound t-Pinocarvyl propionate Myrtenal diethyl acetal Pinoacetaldehyde Myrtenal t-Pinocarveol Hydroxy citronellal$ *N----number of judgements. '•P---- probability level :•----triang]e test Untrained N* Pt 32 99.9 37 25.0 42 95.0 96 99.0 99 90.0 Expert Overall N P N P •9• 99.9 51 99.9 29 98.0 66 90.0 20 95.0 62 98.0 28 99.0 124 99.9 29 99.9 128 99.9 I 99.9 The entire premise of panel testing to sort four presented samples for matching into pairs is based upon differences between the pairs (d-l, d-2) and l-1,/-2) which are significantly greater than any difference between d-1 and d-2 and between l-1 and 1-2. The latter differences, real as they no doubt are, are obviously due to impurities otherwise the samples would be identical. Description of such differences, often volunteered, particularly by the experts, was ignored, since they would have led away from the prime purpose of the experiments and into the vague area of subjectivity. Nevertheless, a number of facts obtruded which were sufficiently remarkable to warrant comment and to suggest areas for future exploration. 1. Experts did not infallibly score higher than randomly selected subjects. 2. One or another of the samples was often classified as "very different" from the other three, by members of both panels. Since there was no con- sistency as to which samples were so selected, the difference appeared to be due to variations in individual physiology, a difference apparently far greater and more widespread than generally recognized, even among experts. 3. The experts differed widely in their ability to discriminate as individuals, again pointing to the importance of the statistical method and