42 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS product and pack are necessary although they may be irrelevant to the test. It may be advisable to get some indication of the size of this unknown factor by testing the product in two different packs or two different colours or perfumes. In this type of consumer testing, products are usually presented under codes. The choice of suitable codes in order to avoid bias from this source is very important especially when there is only a small difference between products. As well as avoiding the use of letters which form well known initials, the letters A and X and the number 1 or 99, etc., it is best to choose codes of equivalent visual appeal. Horton and Mecherikoff (1) asked a sample of 100 college students to rank the alphabet according to their preferences for the appearance of the capital letter and they give a list of letters in rank order. Letters close in rank order should be selected but it is usually wise to make sure also that letters of very different shape are used. This avoids possible confusion when letters are written badly on question- naires. When a combination of letters and/or numbers are used it may be worth while testing them on a small panel for visual appeal to check that one is not more appealing than another. Another effect of codes was illustrated by Appel and Hernandez (2). Certain letters can be read either way up i.e. S H 0 WZ M I N X In a test of powdered products bearing only a code letter and no indication of which was the top of the box, boxes bearing a reversible letter were opened at the bottom more often than boxes bearing a non- reversible letter. Since the finer particles settled to the bottom, panelists opening the box at the bottom obtained a different impression of the particle size from those opening the box at the top. THE DESIGN OF THE TEST The design of the test follows from the objects. When the relative merits of a product with a special ingredient versus the product alone are to be examined, a paired comparison of the two products is usually suitable for a subjective panel test. In preliminary tests on panelists at the laboratory, the untreated substrate may also be assessed as a control. For example, if an ingredient X has been formulated into a skin cream, examination could be made of skin treated with cream con- taining X and of untreated skin to find the effects of the cream plus X.

THE ROLE OF CONSUMER STUDIES IN RESEARCH Also, skin treated with the cream alone could be examined to find out how much effect is due to the cream and how much is due to X. Where three or more different products are to be assessed, it is more satisfactory to use a Round Robin design than to compare each product with the same standard. In a Round Robin, every possible combination of the products in pairs is made and each pair is tested in a paired comparison test. For n products there are n(•-l).possible pairs. The advantages of a Round Robin design are the possibility of ranking all products and the possibility of using smaller panels. For instance, if a fairly average control product is tested against some test products which are much "better" and some which are much "worse", comparison of each test product with the control product will classify the test products as better or worse. However, it will be difficult to assess exactly the degree of difference in each comparison and it will probably not be possible to place the products in any rank order. Round Robin testing compares each product with every other product and does allow all products to be placed in a rank order. Also, in order to get a significant answer for each comparison of a test product with the control product, fairly large panels will be needed. In a Round Robin design, however, information about a product can be obtained from every comparison in which it appears. Therefore smaller panels can be used for testing each pair. This will be considered further under Analysis of results. Factorial designs are useful when variations within a given formula are being considered. In practice the test is a Round Robin but the analysis takes account of the factorial design. There are occasions when it is not reasonable to ask panelists to test more than one product. This occurs when the product has a long lasting effect and the product would not be used again for a few weeks or when the product has to be used for a period of time to show any effect. Then Monadic testing must be used. When more than one product is under consideration, balanced panels must be drawn up, one to test each product. This means that each person in one panel must be matched by someone with similar charac- teristics in every other panel. The disadvantage of Monadic testing is the lack of a reference standard for the panelists. If products are compared with panelists' "usual brand" this could be any one of a number of different brands. Also, since the "usual brand" has a brand image, advertising claim, professionally designed pack