A research anomaly is pointed out: while it is habitually recognized that multivariate analysis requires the random selection of a large number of individuals, most studies that deal with the influence of a particular group or context on individuals are satisfied to deal with a nonrandom sample of ten or twenty groups, schools, or neighborhoods. Yet, there seems to be no reason why group studies require a different sampling procedure from individual studies. Problems connected with the use of a small number of contexts are examined. It is argued that these problems cannot be solved by taking group members rather than groups as units of investigation, since deviations are multiplied by the same coefficient as the number of Ss. The problem of selecting the contexts to be studied are examined. It is noted that contexts are usually not randomly selected, & that the results obtained are, therefore, not generalizable. Further aggravating circumstances are considered. Measures of context or group characteristics are not as well developed as those of individual ones. Context effect studies need to use random samples of at least 100 contexts. A. Orianne.
