386 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS Therefore, the data were analyzed separately according to preference. It was possible by reading the responses to discern and key punch the tone of the comments, i.e., whether a comment was positive or negative. For example: "R had a long aftertaste, W was too hot" would be punched positive for R's aftertaste and negative for W's hot factor. In addition, the response sheets usually showed a major reason in one of the seven key word categories for the preference voted and it was then possible to tally the positive and negative major reasons. This tally showed a situation that users of consumer tests are in dread of obtaining: In both panels, the R-preferrers mostly voted against Sample W, while the W-preferrers voted in favor of Sample W. That is, 56% local and 61% national R-preferrers gave negative major reasons about Sample W in backing up their preference vote and 70% local and 67% national W-preferrers gave positive major reasons for preferring Sample W. This situation presents the problem of a statistically significant prefer- ence trend based on the negative attitude toward the nonpreferred sam- ple of the pair. The problem is: Did the R-preferrers really like R but Table III Over-all Hedonic Scores About R About W Local National Local National Average 3.61 8.29 2.90 2.92 Sigma q- 1.07 q- 1.06 q- 1.28 q- 1.18 Table IV Hedonic Scores According to Preference Vote R-Preferrers About R About W Own brand Local National Local National Local National Average 4.06 3.77 2.21 2.14 4.31 4.01 Sigma q-0.76 q-0.66 4-0.95 4-0.84 4-0.81 4-0.77 W-Preferrers About W About R Own brand Local National Local National Local National Average 4.01 4.11 2.72 2,47 4.10 4.00 Sigma q-0.97 q- 0.73 q-0.97 q-0.91 q-0.81 q-0.80

COMPARISON OF CONSUMER PANELS IN PAIRED PREFERENCE TEST. II 387 take the easy way out by isolating the negatives of W ? Or did they merely vote for R as the lesser of two evils ? What are the intensities of preference ? tIedonic Responses Most consumer test operators insert a value judgment question to measure the intensity of preference. This investigation used five- point hedonic scoring for both test products and for the consumers' usual brands of toothpaste. But test operators handle the hedonic scores differently. For example, Table III shows mean hedonic scores for both prod- ucts by all respondents in each panel. Analysis for sigma may not be the most appropriate analysis, but it does show the scatter around the mean scores, which is considerable. Actually, calculating over-all hedonic scores, like other over-all tallies, is a first look-see or horizontal analysis and probably is useful mainly in showing up inconsistencies in data. Table IV begins to unravel the problem of intensities of preference by scaling the two test products in relation to own brands. This table mainly shows that the preferred products rate as well as own brands and the nonpreferred products do not. In addition, it shows that the mean hedonic scores for the preferred samples are undoubtedly higher than those for the nonpreferred samples. But in the authors' opinion, a better way of examining hedonic data in order to view intensity of pref- erence is to compare how each voter rated each test product with how he rated his own brand--the product he usually buys. These comparisons are summarized in Table V. Two important points are dearly evident: (a) More W-preferrers scored W equally with or higher than their own brand than R-preferrers did for R and (b) the W-preferrers were not so hard on R as the R-preferrers were on W. The R-preferrers' hedonic responses about R were not unexpected, [or the major reason analysis had already shown that their preference was based on a negative reac- tion to W. But the relatively generous hedonic attitude of the W-pre- ferrers toward R has not thus far been exposed. In considering the findings from Table V, one must not ignore the staunch loyalty of the W-preferrers for Toothpaste W. If, for ex- ample, this had been a test of two new toothpastes, the question to de- cide might be: should we go with the toothpaste that most people said they preferred or with the one that was liked more intensely ?