Mediators and Mechanisms

Abstract

Mediators are widely thought to be mechanisms. Mediation is to mechanism what correlation is to causation. Statistical evidence of mediation is necessary but not sufficient evidence of mechanism just as statistical evidence of correlation is necessary but not sufficient evidence of cause. Mechanisms also require evidence of causation and an explanation regarding how and why mediation works in terms of a series of causal steps in order to have explanatory value. Otgaar, Muris Howe, and Merckelbach made a case for an associative activation psychological mechanism. This article further advances the case for an associative activation mechanism by showing that it is an important parallel distributed processing mechanism in the connectionist neural network models introduced by Rumelhart and McClelland.