Replication versus Tests of Equivalence

Abstract

The important concept of “replication” is critically examined within the context of current research procedure and is viewed as an unattainable ideal. The degree of success of an attempted replication may be judged in terms of equivalence; such judgments will of course differ with respect to comprehensiveness and specificity. Examination of the traditional law of the independent variable identifies two broad classes of variables; independent and controlled variables and dependent variables. It is useful then to consider attempts to replicate either or both sets. Provision is thus made for virtually all experiments in the behavioral sciences. The test of range of invariance is discussed and following Stevens, proposed as a worthy objective of behavioral research. In any experiment which tests the range of invariance, it is desirable to attempt replication of a reference experiment and to provide graduated differences on the methods variables and/or outcome variables in order to ascertain at what point invariance fails.