The present author has examined from a psychometric point of view, the classificatory systems of odors which were proposed by Henning, Zwaardemaker, and Crocker. The principal methods employed were (1) Torgerson-Indow's multidimensional scaling, and (2) a modification of semantic differential scales. The present study consists of four experiments, two of which (VIII to IX) used the first method and the other two (X and XI) utilized the second method.The purpose of experiment (VIII) was to analyze the similarity patterns and locate some of the unpleasant odors which cause troubles in industry into a more comprehensive scheme of odors. The odorous materials sampled are shown in Table 1. Five naive subjects judged the similarities among these odors. By the application of multidimensional scaling, four factors were extracted and the results are shown in Table 4 and Fig. 1. The first factor was a pleasantunpleasant dimension. Chemically, the positive pole of this axis was represented by resinterpenoids and the negative pole was represented by nitrogen compounds and higher carboxylic acids. The second factor was a sweet-pungent dimension. These pungent “odors” are sometimes called not odors but trigeminal stimulants. The third and the fourth factors were of dubious nature as compared with the first two factors.The purpose and method of experiment (IX) was the same as the experiment (VIII) except that the number of unpleasant odors was reduced. Odorous materials sampled are shown in Table 5. The condition of this experiment differed from those of experiment (VIII) in the following points. (1) The concentrations of odorous materials were calibrated to produce nearly equi-intensive sensation. (2) Subjects were two expert perfumers (A & B of Table 6) and six naive subjects. The reliability (especially inter-judge agreement) of the naive subjects was not sufficiently high. Therefore the multidimensional scaling was applied only to the data of the experts. By the electronic computor Facom 128B, six factors were extracted, but the 4th to 6th factors may not be a real solution. The results are shown in Table 7 and Fig. 2. The first factor was pungent-sweet. Chemically the positive pole of this axis was represented by resinterpenoids, but the common characteristics for the negative pole was not clear. The second factor was persistent-heavy-spicy foul. The negative pole was represented by skatol and clearly unpleasant, but the opposite pole was not necessarily a pleasant one. The third factor was hard to interpret.It may be characteristic of the expert judgments that the hedonic tone was not found as the first or the second factor, although the fact that extremely unpleasant odors were not included in this experiment. The relative weights of the factors (including references 6 and 7) are shown in Fig. 3. Of course the weights are different for each experiment depending on the range of sampling of odorous materials. The rate of convergence of factors was relatively slow in the case of complex odors like perfumery or spices, and relatively rapid in the case of chemically pure compounds.The author has proposed (see ref. 6) 25 semantic bipolar scales (7 point rating) for the description of odors. The purpose of experiment (X) was to reduce the number of these scales. From the semantic profiles for 44 odors (described in experiment VIII to IX) we have calculated the correlation matrix of these scales (Table 8). By the application of factor analysis (centroid method), three factors were extracted (Table 9). The first factor was clearly a “sensory pleasure” dimension, the second factor was a “harshness”, and the third factor was an “intensity or vividness”.An expert perfumer Mr. Kainoshow proposed a 20 semantic monopolar scales (5 point rating) for the description of odors. In experiment (XI), we have analyzed the correlation matrix for these scales (Table 10) by