Paired Comparisons in Consumer-Product Tests

Abstract

This is a study of some assumptions made in paired-comparison product tests. The assumption that order of testing creates a bias is found to have some validity. However, the findings show that a differential bias is involved—in favor of the more recently used product when comparative judgments are used, and in favor of the first product used when absolute measurements are utilized. Therefore, the author suggests that the test group should be randomly subdivided, and order of testing alternated within each subgroup.