INHERENT INVALIDITY OF ODOR CLASSIFICATION 29 essential to scientific validity. Of the second group of systems of odor classification, the earliest im- portant example was the Henning prism (25). Henning conceived of odor as being representable by a prism, at the hexagonal points of which were located six odor classes or types, namely flowery, fragrant or ethereal, resinous, spicy, burned, and putrid. By placing an odor at a given point on the surface of the prism, and at a given distance from the six corners, it was usually possible to indicate the degree to which an odor fell into one or several classes. Thus, Henning conceived of odor classification in terms of likeness and of degree of likeness. It should be clear that the entire prism concept was merely an arbi- trary geometrical device to indicate a quantitative approach within the qualitative in other words, the ex- tent, or quantity, of a particular quality. The prism, however, was an extremely inconvenient method for such communication among its other deficiencies was the fact that certain combinations of qualities could not be expressed. The Henning system was tested by several psychologists, and their results are most illuminating. Dimmick (26) repeated the Henning experiments, to determine whether some 25 students would place cer- tain stimuli in the same main group as Henning had. In a few rare cases, there was some degree of uni- formity, as when apple blossom or jasmin was being classified. Vanilla was called flowery by 6, fruity by 11, spicy by 6, resinous by one, and burned by one. Lavender was termed flowery by 6, fruity by 6, spicy by 12, resinous by one. In several cases, such as of substances as varied as musk and acetone, every one of the six classification groups was cl•osen by at least one subject as the main point on the prism for that particular odor. Further work on the Henning system was done by MacDonald (27), who reported that it was ap- parent that "variability (of results) is great...that this variability occurs between observers and be- tween the three series of a single ob- server." Findley (28), assuming as did the others, for the sake of the experi- ment, that the six groups or classes of Henning were valid, tested the system and found that, "in general, the differences are closer to 'chance' than to perfect consistency." Find- ley (28) writes: "On the average, the degree of consistency is about 65%, where 50% is chance. These results are in accord with MacDonald's find- ings, in spite of the fact that we chose from Dimmick's results stimuli which in Dimmick's experi- ment gave the greatest consistency. In other words, we prejudiced the experiment toward precision and failed to find it. It seems, there- fore, impossible with the methods available at present to get as much precision in the qualitative classifi- cation of odors as is possible, for ex- ample, with colors." The Henning prism was tested by

30 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS Bentley (29), who utilized 50 odors and three subjects, and it was re- ported that the observers agreed "fairly well" with only 10 of these materials. The studies of psychologists on the Henning system was sum- marized by Gamble (30) as follows: "Dimmick, who by his choice of scents 'prejudiced his investigation' in Henning's favor, reports some real evidence for the prism. Mac- Donald, however, concluded...that his findings correspond with Hen- ning's only to an extent which might be indicated by a product-moments correlation between .30 and .40... Findley (in spite of concessions to Henning)...found as little evi- dence for the prism as did Mac- Donald, and both Findley and Mac- Donald stress the inconsistence of their observers." Following this summary by Gam- ble, another effort to verify the Henning system was made. Haz- zard (31) reported that "the qual- ities assigned by Henning and checked by other investigators were verified for 12 of our odors," out of a total of 14. Utilizing only 6 sub- jects, and reporting the findings of 4, there appear such inconsistencies as the following: Classification of the odor of tar on the prism, as interpreted in words, according to a system of terms suggested by Hazzard: Sub- ject 1, resinous-burned 2, fragrant- like, resinous-burned 3, putrid- like, burned 4, spicy-like and burned-like, but.mainly resinous. The latest major effort at odor classification, utilizing a system based on degrees of likeness within a qualitative differentiation, is that of Crocker and Henderson (32, 33). These scientists reduced the number of odor classes to four, which they termed fragrant, acid, burned, and caprylic. The degree to which an odorous stimulus is fragrant is ex- pressed by a number ranging from zero to eight, the former meaning that it has none of the fragrance quality, and the latter the maxi- mum. The same stimulus is then given a number or digit to express its intensity of acidness, burniness, and caprylic character. The four digits may be read as one number the odor of santalol is expressed, for instance, as 5221, and that of phenyl ethyl alcohol 7423. The Crocker-Henderson system is an improved and simplified Hen- ning prism. It retains from Hen- ning the concept that an odor be- longs to several classes simultane- ously, and the concept of degree to which it might form a part of each class. The reduction of number of classes from six to four has not been verified in a test, and is no more valid, and no less, than Zwaarde- maker's nine, Rimmel's eighteen, or Henning's six. The use of a numerical rather than a geometrical method of expression is a major simplification, which constitutes an improvement in the sense that the results are more easily communi- cated, but it is a device rather than a new system. Can the Crocker-Henderson sys- tem be verified? The experience