SENSORY TESTING -- A STATISTICIAN'S APPROACH 227 REFERENCES (1) Fisher, R. A. Design of experiments 6th Edition. 11 (1935) (Oliver & Boyd, Edinburgh & London). (2) Bengtsson, K. and Helm, E. Principles of taste testing. Wallerstein Lab. Comm. 9 171 (1946). (3) Lockhart, E. E. Binomial systems and organoleptic analysis. Food Technol. 5 428 (1951). (4) Green, M. W. A note on the use of triangle design in taste testing of pharmaceutical preparations. J. Am. Pharm. Assoc. Sci. Ed. $$ 380 (1955). (5) Fourman, V. G. Taste panels for pharmaceutical flavors. Dr,g Cosmetic D•d. 77 762 (1955). (6) Ostle, B. Statistics in research. 58 (1954) (Iowa State College Press, Iowa). (7) Hardes, J. M. Sensory tests and consumer acceptance. J. Sci. Food Agric. • 477 (1953). (8) Gridgeman, N. T. Sensory item sorting. Biometrics, 15 298 (1959). (9) Richardson, E.G. The science of orchestral instruments: Some recent •vork. Discovery 1• 87 (1953). DISCUSSION MR. A. ELLIOT•r: At this Symposium three distinguished speakers have recom- mended three different methods of selecting perfumes for new cosmetic products. Mr. Erni (10) suggested it should be the perfumer's responsibility to select the perfume, Mr. Landon {11) suggested the marketing manager should make the final choice, and finally you suggested that the statistician, with the aid of panel tests, should choose the perfume. The suppliers of perfume compounds are anxious to point out that a distinctive perfume will probably only be liked by about 1-o/ of the population and yet still be O/o a commercial success, whereas a perfume which is generally acceptable is not distinc- tive and hence will not sell the product. In view of this, can you tell me of a method of panel testing which will select a distinctive, rather than generally acceptable, perfume for new products? THE LECTURER: It is not my view that the statistician should have anything to do with the actual selection of the perfume. It should be the potential customer, when all is said and done--and all that the statistician can offer to do is to help both the cosmetician and the marketing man to find the best way of obtaining this information about the potential customer. The question of hoxv one goes about looking for the proportion of people that are going to find the perfume acceptable is very involved and it depends to what group you are aiming your product. It is not a statistician's function to answer (only to ask), and what he can do is simply to draw up rules for saying--if you are making decisions like this (which are, after all, commercial decisions or, possibly, perfumery decisions) what kind of samples you have to draw, what kind of numbers of people you are going to have to interview before you have some reasonable assurance of having met that target. Without talking about specific cases (and I do not think there is much point in doing that here), this is a question of laying down specifications. Ma. C. SEBLE¾: How much rechecking takes place witIx your panel, not so much from the point of view of sensory fatigue, but from a "preference" angie? (10) Erni, M. Evolution in perfumery. Presented at the Symposium on "Perfumery" at Eastbourne, on 21st April 1970. (11) Landon, M. F. The application of perfume to cosmetics and toiletries. Presented at the Symposium on "Perfumery" at Eastbourne, on 20th April 1970.

SENSORY TESTING- A STATISTICIAN'S APPROACH 229 in assessing general acceptability would be very much governed by many other things, some of which, like brand image and promise, are controlled from the commercial side. We have confirmation that the basic characteristics of the smell acceptability have remained similar in in-use situations, to the answers that we obtained in the small-scale testing such as described. What does not always follow is that this has the direct effect on the general acceptability of the product that one hoped for one finds that because the final consumer is using the product in a total use situation (as dis- tinct from just smelling it), we have a halo effect operating in both directions. Perfume acceptability may be having a "halo effect" upon the "acceptability in use" of the product but, of course, also vice versa. Therefore the fact that on some occasions we get discrepancies between the overall in-use acceptability and the perfume test is something that we have to live with. MR. P. MoxE¾: I understand from what you said that you expunge people if they are not necessarily able to give you what one might call a distinct answer. What concerns me is that a great many products are bought in this business on the strength of the fragrance, and I can understand why you want to get your statistics right to start with. I wonder if statistically you are not building a false image into your preliminary testing. TrtE L•c'ruR•R: What you say is very true. First of all I would say that we do not exclude people from the panel except in the sense that we only look at their prefer- ences if they show evidence of being able to discriminate. They do come back the next time and are again put into the test because we do not exclude the people who last time were unable to discriminate. \¾e find that you have a variable level not only of acceptability but also of discrimination - and this is something that happens in real life as you yourself have indicated. Sometimes people are strongly motivated about a product's smell, sometimes they could not care less whether it smells nice or not. What we are saying is that, in so far as they do make their choice in terms of smell, we wish to have the smell most acceptable to the majority of people. If they are not choosing on the basis of smell, that is just too bad. We can not do anything about that. What we try to make sure is that on the occasions that people are choosing on the basis of smell, this product will be as acceptable as we know how to make it.