254 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS 10 5 IOO Figure 4. The pictorial representation of hard, firm, soft and pourable materials. (ii) Having completed (i) above, to use the panel to determine the optimum for each physical property, and the acceptable limits of variation for each one. It sounds very simple. In fact it is a little more difficult than it sounds, because of the large variation in sensory judgments between people. It is never possible to obtain an absolute answer from a sensory test. The best that can be achieved is a result which is 'statistically significant', at the 5•o, 1 •o, or possibly even 0.1 •o confidence level, depending on the number of people used on the panel, and how consistent they are in their judgments. It should be obvious from the preceding paragraph that a well-trained panel, carefully selected for their sensitivity, reproducibility and interest in participation will take much longer to set up but provide much more valuable information, and at a greater rate, once established. TYPES OF PANEL TEST The organization of panel tests, and the coding and presentation of

PSYCHO-RHEOLOGY 255 samples is described in (2). Types of test and their usefulness are summarized below. The triangle test Subjects are given three coded samples, one of which is different from the other two. They have to select the odd one out. This type of test is useful to establish a panel's sensitivity, or thresholds, and to define a range of physically measurable differences in a product which cannot be per- ceived by the panel. Scoring methods Several different ways exist for scoring, or ranking samples for a particular property, e.g. Scoring several samples together on a subdivided scale, with standard samples provided for reference. As a ratio of two samples, a trial sample against a reference standard. By 'paired comparison' tests, establishing which of a pair is greater in any chosen property, but not assigning any magnitude to the difference. Scoring directly on a familiar scale without reference to standards except from time to time. This may be useful for a well-trained panel only. Whichever methods are used, it is always necessary to analyse the results statistically. Analysis of variance (3) establishes whether there are significant differences either between the products or between the panellists, and how great the significance is. Scaling methods have been used very extensively by Szczesniak (4) to establish correlations between food textural propertibs and instrumental measurements. Some good correlations have been re- ported. Other people have carried out similar work (5). The techniques which have been applied to foods are equally applicable to cosmetics. Only the rheological properties of importance will be changed. CONCLUSIONS Cosmetics show complicated rheological properties, but their textures can be described, in the various usage situations, by the the correct choice of a few important properties. These properties, as measured instrumentally,