Abstract

Some of the selection issues that bear upon the conduct of research using multivariate, replicated, single-subject repeated measures designs are examined and their implications for the study of developmental phenomena discussed. The choices of participants, variables, occasions of measurement, etc., made in the conduct of empirical research all involve some kind of selection. These choices, therefore, introduce selection effects into collected data which, in turn, threaten the generalizability of one's conclusions. Data analyzed for single subjects, for example, are suspect regarding generalizability to other individuals. While the limits of generalizability to persons is a concern to be taken seriously, the concern properly applies to all modes of data classification. Discussions of the importance of person selection emphasize representative sampling (of persons) and generalization to populations of persons. Attention to the selection of variables has led to a focus on multivariate approaches to measurement. For the developmentalist, the occasions mode, which is either implicitly or explicitly involved in definitions of change, is especially relevant to concerns about generalizability. Whether one considers stability or change, the effects of temporal selection ought to be a central concern in designing research.