250 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS ratio of the weights produced by the right and left side remains constant (27), provided both are untreated. A control ratio is determined, and then the antiperspirant applied to one axilla. The ratio is again measured, and the effectiveness of the antiperspirant can be determined from the change in ratios. The full method of calculating the results and the statistical analysis has been discussed by Wooding (30) and Daley (29). Other sites than the axilla have been used for gravimetric determinations (31). These sites do not produce as large a volume of perspiration as the axillae, and therefore the errors involved in measuring such small quantities tend to make these methods unsatisfactory. Although the gravimetric methods have found much favour, because of their simplicity and the ease with which numerical data for statistical analy- sis can be obtained, they too are not wholly satisfactory. The basis of the method is the assumption that the right to left ratio of sweat output is a constant (27). Under ideal conditions this may well be so, but under con- ditions encountered in laboratory tests this is not always so. The factors which affect sweating have been discussed both by Kuno (4) and by Reller (46), who have shown that certain stimuli such as posture, pressure, etc. can affect sweating unilaterally. A subject who has to spend several 1 h sessions in a hot room with pads under his arms does not remain perfectly still. Any movement can affect the relative sweat output from each axilla. The effect of leaning on one arm or the other is shown for one subject in Table I. It can be seen that the change in axillary R/L ratio can be so large as to obscure most antiperspirant effects if the subject leans one way, or make the effect look much better than it really is if the subject leans the other way. Table I The effects of posture on axillary sweating Sweat collected (mg) Ratio Right axilla Left axilla R/L 188 169 1.15a 222 155 1.36a 586 278 2.10b 260 237 1.10a 254 610 0.41½ 281 180 1.56a 290 195 1.49a a, Subject sitting upright b, subject leaning to the left c, subject leaning to the right.

EVALUATION OF METHODS FOR MEASUREMENT OF ANTIPERSPIRANCY 251 Statistical analysis of a large amount of control data (no antiperspirant applied to either axilla) in our laboratory has shown that in some subjects the variations in sweating are such that the right to left ratio cannot be said to be constant. Table II Rates of R/L axillary sweating measured using a gravimetric method Observation 1 Observation 2 Day 1 1.39 0.65 Day 2 1.58 0.88 Day 3 1.94 1.68 Day 4 1.61 1.09 Day 5 2.03 1.42 The differences between the duplicate observations taken during one day and the differences between those found on different days are both significant at the 5% level. This means that considerable effort has to be expended in selecting the people who have the most constant ratios and are therefore most suitable for these tests. Even using these subjects it is often difficult to interpret results obtained from this type of experiment if the number of subjects in the panel is small, because of the scatter of results. In some cases the reduction, due to a particular product, is very good, in others poor, whilst some subjects apparently sweat more after treatment with a product than before. A typical experimental result is shown in Table IH. There are a number of ways in which the effects of these two products can be compared from this type of test. (1) The mean results over the whole panel, including or excluding the increases can be compared. Including increases Mean reduction due to product 1 = 7.7%. Mean reduction due to product 2 = 20.4%. Excluding increases Mean reduction due to product 1 = 31.6%. Mean reduction due to product 2 = 33.1%. (2) The averages of the statistically significant changes, including or excluding the increases can be compared.