The Simplest Experimental Design that Permits Multiple Generalization

Abstract

Recently, several methodological critiques have pointed out that most experiments in verbal learning fail to present statistical evidence that their results could be replicated using a different sample of language materials. Consequently, many of the studies have little scientific point because their conclusions have to be restricted to the specific language items used in the experiment. All these critiques are summarized, the various solutions to the problem are evaluated, and procedures to arrive at the simplest solution are described. This solution should present no difficulties to anyone who has had a course in analysis of variance.