Measure of Proportional Non-Chance Clustering in Free Recall

Abstract

A new ratio of category clustering ( CLT) is presented which, unlike earlier measures, uses an unconfounded estimate of the number of non-chance repetitions occurring in free recall sequences. The measure permits the assessment of clustering of individual categories independently of clustering of the other categories and yields additional statistics which cannot be accurately measured by existing methods. Computer-generated data scored according to CLT and five other clustering measures indicate that the various measures are not equivalent. In addition, the five other measures are shown to be confounded by extraneous characteristics of the set of items recalled. Consequently, the validity of conclusions based on clustering measures other than CLT is questioned.