Statistical bias in cross-sequential studies of aging

In an earlier published study (Botwinick & Arenberg, 1976) it was argued that cross-sequential designs used in studies of aging and intelligence bias results in favor of a larger main effects F ratio for cohort than for time of measurement. As a consequence, purely ontogenetic influences are likely to be misinterpreted as generational. The present paper gives conclusive proof of the statistical bias inherent in these designs. In addition it shows that, for any given number of measurement occasions, the degree of bias increases with the number of different cohorts tested. An explanation of the statistical problem in intuitive terms is also provided.