The Category-Confound: A Design Error

• Published in: The Journal of social psychology Vol. 103; no. 1; pp. 57 - 63
• Main Authors: Kay, Edwin J., Richter, Martin L.
• Format: Journal Article
• Language: English
• Published: Worcester, Mass Taylor & Francis Group 01.10.1977; Clark University Press
• Subjects: Indexing in process
• ISSN: 0022-4545; 1940-1183
• DOI: 10.1080/00224545.1977.9713296

An experimental design problem, found to be frequent in some of the social psychology literature, occurs when only one sample from a population of possible samples is used to define a category of treatment factors. An example would be using one man and one woman as Es to test the effect of sex of E. This error, called the category-confound, leads to an inability to generalize treatment effects beyond the particular sample chosen, and to an inability to perform the appropriate significance tests. Several examples are discussed in detail, and the frequency of occurrence of this error in two recent journals in social psychology are surveyed. In most cases the category-confound can be dispelled simply by taking an adequately large sample from the population.