Correlational methods in research on human learning: an amplification.

Abstract

This paper considers two recommendations regarding the use of correlational methods in research on problems of human learning. These suggestions are that residual gain scores are especially appropriate for defining gains in such studies and that the results of factor analyses of gain scores are relevant to substantive problems of learning. It is concluded, first, that no acceptable logical or empirical grounds were presented for preferring residual scores to other linearly defined gain scores for use with correlational techniques; instead, the residual scores were shown to have several serious weaknesses. The second conclusion is that a factor analysis of gain scores is inappropriate for substantive problems of human learning; however, properly designed and correctly analyzed factor studies may supplement other techniques for such problems. Some weaknesses of a single-group design for a study of learning effects were also noted.