EVALUATION OF FLAVOURS IN DENTAL CREAMS 285 standard, using the Duncan multiple range test described above. A nearest match to the standard can easily be worked out on this basis. Extending the method further, even more complicated experimental designs can be used for the toothpastes, such as Simplex or star designs (14). Regression analysis can be used to fit equations relating the main rating scores to all the variable flavour component levels. The fitting of such regression equations will be valid only if a significant effect between toothpastes has been shown by the initial analysis of variance. Linear programming can be used to determine the nearest match to a standard over all the relevant flavour questions although this will only work if all the regression equations are first order, i.e. correspond to straight lines, planes or hyperplanes. Curvature due to higher order terms leads to com- plications and the best method of choosing a match in this case is by inspection. When large experimental designs are used it will be necessary to test a great number of creams. It will no longer be possible to use a complete randomized block as the basic design for the panel. This can, however, be overcome by the use of incomplete block or lattice designs in which each panel member need use only a few out of the many experimental creams, and the duration of the panel reduced once more to more manageable proportions (14). It is quite possible for any member of a simplified flavour description panel to clean his teeth five or six times a day without impairing his ability to judge flavour notes and intensities. An average panel can, therefore, be completed in three or four days. Should any panellist drop out of the test it is quite permissible to estimate his missing ratings using the recognized method of calculating missing values for the particular panel design used 02). The simplified flavour description panel is, therefore, quick and easy to run. Analysis of the results can be as complicated as desired and depends only on the experimental design chosen for the test creams. FLAVOUR ACCEPTABILITY TESTS Flavour acceptability is more difficult to test than flavour description. The test method described above can, however, be adapted for this purpose by modifying the questionnaire and ensuring that panels are larger and more representative of the toothpaste-using population. Usually only one

286 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS question will need to be asked, i.e. for flavour preference, and a rating scale can be used of the following form: 0 1 2 3 4 Very poor Poor Average Good Very good Other flavour description questions can be asked if it is wished to con- sider the influence of particular flavour sensations on acceptability. Because of the obvious variability of flavour likes and dislikes over the population, panels used for acceptability testing must be large and as representative of that population as possible. Panel size should be at least 50 and local panels confined to a single factory or workplace would seem to be unsuitable because unrepresentative, although they could be used for a first rough screening of fiavours for acceptability. Experimental design can be built into the testing by using different levels of a range of flavour components. Regression analysis can again be used and linear programming or inspection used to detect flavour combina- tions with optimum flavour preference ratings, the object of the testing being no longer to match a standard, but to optimize flavour preference. CONCLUSIONS The test procedures described above are intended to enable flavour evaluation, particularly description, to be carried out in an orderly and scientific manner. Emphasis has been placed on the generation of numerical results which can be subjected to statistical analysis. The numerical pre- scriptions applied to flavour acceptability have been deliberately left some- what vague. In this area the ideas for new flavours will usually come from a creative flavour specialist although panel testing will be necessary to judge the success of his efforts. (Received: 17th January 1975) REFERENCES (1) Fruchter, B. Introduction to factor analysis (1954) (Van Nostrand, Princeton). (2) Moncrieff, R. W. 'Gustation', Food Mfr, June, 1944, p. 203. (3) A.S.T.M. Publication S.T.P. 433. 'Basic principles of sensory evaluation' (1968). (4) Moncrieff, R. W. 'The tastes', Perrum. Essent. Oil Rev. (Oct 1950 to Feb. 1951). (5) Harris, H. Introduction to human biochemical genetics (1955) (Cambridge, London). (6) Wells, F. V. 'Duration of odorants on the skin'. S.P.C. Year Book, 107 (1966) (United Trade Press, London).