PRIMARY IRRITATION OF THE SKIN 223 good agreement in numerical terms. This demonstrates the value of con- cordance training of observers. However, unless extensive concordance studies are undertaken to ensure fully reliable agreement between observers, it is deskable that the same observer should make all the observations dur- ing a particular test. It is also desirable that all observers should carry out the testing at frequent intervals to avoid major drift of the scoring scale. Even so, comparisons between tests carried out at different times may have dubious validity. Experiment H Serial dilution studies on water-miscible hairdressings (mouse ear test) Groups of five female CF1 mice were used for each dilution of three formulations in two separate studies, Formula A being compared with Formula B in one study and the same batch of Formula A with Formula C in the other (Fig. 4). I I IOO Formula B Formula A Formula C I I _•• Formula A 0,1 0.1 I00 I0 • Log % concentration Figure 4. Dilution curves for water-miscible hairdressings. Vertical comparisons refer to the same experiment, e.g. Formula A 0eft-hand curve) versus Formula B. Formula C is shown as broken line since the points show considerable scatter.

224 JOURNAL OF THE SOCIETY OF COSMETIC CItEMISTS The hairdressings were studied at the following dilutions in distilled water: 100, 50, 25, 18.75, 12.5, 9.38, 6.25, 3.13, 1.56 and 0.78%. At the higher concentration, 10 mg applications were made the fluid dilutions were applied in doses of 10 I•l. pH for the dilutions ranged from 4.5 to 6.1 for Formula A, 5.1 to 7.9 for Formula B, and 4.5 to 6.3 for Formula C. Observations were made by one observer throughout. The curves show: (i) By comparison of the two curves for Formula A, the reproducibility of results with the same applications in two separate series of tests. Good reproducibility in this experiment probably depended on the fact that the same sample of material was used on both occasions, mice from the same batch were used and the same observer carried out both experiments. (ii) The manner in which irritancy is reduced by dilution with water (in the case of this particular type of cosmetic preparation). (iii) The type of comparison that may be made between alternative formulae for the same kind of cosmetic product. Experiment II1 Variances in human patch test The human patch test procedure described above was used for a com- parison between two roll-on antiperspirant-deodorant formulations, each with three different perfumes (Table II). Table II Antiperspirant-deodorant formulations (human patch test) Formula Perfume Mean overall :k SE score I A 11.66 0.69 I B 10.36 0.65 I C 10.82 0.81 II A 9.61 0.76 II B 8.66 0.87 I[ D 8.30 0.79 For this experiment, Formula I with perfume C was regarded as the standard product, and tests for statistical significance of differences from this were performed using Wilcoxoh's Signed Ranks Matched Pairs test and the Student t test. The former test did not show a statistically sig-