562 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS Table I An Expanded List of Harper's • Descriptors for Characterizing of All Types of Odors Fragrant Oily, fatty Aromatic Frnity (citrus) Sweaty Like mothballs Meaty (cooked) Fruity (other) Almond-like Like gasoline, solvent Sickening Putrid, Foul, Decayed Burnt, smoky Cooked vegetables Musty, earthy, moldy Woody, resinous Herbal, green, cut grass S veet Sharp, pungent, acid Musk-like Etherish, anesthetic Fishy Camphor-like Soapy Sour, acid, vinegar Spicy Light Garlic, onion Like blood, raw meat Paint-like Heavy Animal Dry, powdery Rancid Cool, cooling Vanilla-like Like ammonia Minty, peppermint Warm Fecal (like manure) Disinfectant, carbolic Sulfidic Metallic Floral Perfumery Yeasty Eucalyptus Strawberry-like Malty Cheesy Buttery Stale Cinnamon-like Honey-like Like burnt paper Cork-like Popcorn Anise (licorice) Cologne Lavender Incense Turpentine (pine oil) Caraway Cat-urine-like Melony (cantaloupe, Fresh green vegetables Orange (fruit) Bark-like, birch honey-dew) bark Tar-like Medicinal Household gas Rose-like Peanut butter Celery Leather-like Nutty (walnut etc.) Violets Burnt candle Pear (fruit) Fried fat Tea-leaves Mushroom-like Stale tobacco smoke Wet paper-like Wet wool, wet dog Pineapple (fruit) Raw cucumber Coffee-like Chalky Fresh cigarette smoke Raw potatoe Peach (fruit) Mouse-like Laurel leaves Beery (beer-like) Oak wood, cognac Pepper-like Scorched milk Cedarxvood-like Grapefruit Bean-like Sewer odor Coconut-like Grape-juice-like Banana-l'fke Sooty Rope-like Eggy (fresh eggs) Burnt rubber Crushed weeds Seminal (sperm-like) Bitter Geranimn leaves Rubbery (new rubber) Like cleaning fluid Cadaverous (like (carbona) dead animal) Urine-like Bakery (fresh bread) Cardboard-like Raisin-like Lemon (fruit} Seasoning (meat) Crushed grass Maple (as in syrup) Dirty linen-like Apple (fruit) Chocolate Hay Kippery (smoked fish) Soupy Molasses Kerosene Caramel Grainy (as grain) Sauerkraut-like Clove-like aThe upper block is from Harper, with some Americanization of terms (gasoline instead of petrol, etc.).

EVALUATION OF HUMAN BODY ODOR 563 One of the most balanced scales has been advanced by Harper (24) and consists of 44 characteristics. The odor sample is rated for the applicability of each characteric on a 0 to 5 scale. Our work with Harper's scale indicated that odors which are quite dif- ferent sometimes result in rating profiles that are not significantly distin- guishable. More descriptors were needed to begin resolving such odors. Re- cently, the ASTM E18 Sensory Evaluation Committee collected a list of 817 descriptors which are in use by various authors and industrial organiza- tions, and which included flavor, fragrance, cosmetics, industrial chemicals, and air pollution research and development. This material was used to ex- pand Harper's list to 136 descriptors, Table I, where Harper's block of 44 descriptors was left intact. A perennial problem with the odor descriptor technique is that panelists differ in their use of descriptors, either because of differences in their seman- tic backgrounds or in the actual perception of certain odor notes. Expand- ing the scale provides a broader "shopping list." Pane] training is necessary to obtain stable responses on the degree of "flora]," "musty," etc. However, there is a potential danger in training the panel to respond uniformly: judgments become more and more provincial and deviate from the reality of sensory/semantic world of odor perception. One solution to this problem is to consider each panelist's judgments sep- arately and to use the response patterns simply to establish similarity or dis- similarity between odors, e.g., an odor before the treatment and another after the treatment. Thus, people differ in usage of "floral," "fragrant," "perfumery," "aroma- tic." If panelist A rates two odors, I and II, and gives a higher floral score to I, while panelist B does not utilize "floral" but gives higher fragrant score to I, etc., indications accumulate that I is higher in the floral/fragrant/per- fumery/aromatic combined dimension. This is frequently quite sufficient for guidance in the odor control efficacy evaluations. There is another form of statistical analysis that permits overall compari- sons of response patterns to two odors at at time, circumventing the need for a direct comparison of two odors. D/Sving (25) used the Chi-Squared statis- tic to compare electrophysiological responses to odors. Adaptation of this ap- proach to semantic descriptor responses is equivalent to postulating that for a panel of 9 there may be as many as 9 different "floral" concepts, each pane- list perceiving this term in a slightly different fashion. With a list of 136 de- scriptors and 9 panelists, there are 1224 sensory response channels. The simplest approach is to ignore the degree of response and consider only uses and nonuses. An example to be used is an attempt to modify the malodor of isovaleric acid (a primitive model of perspiration odor) by addition of a primitive mod- ifier, a mixture of vanillin and linalool vapors. Inspecting the distribution of