Some cautions on interpreting qualitative indices for word-list learning tests

Abstract

The statistical performance of qualitative indices for word-list learning tests was examined using randomly generated data. These included position effects (primacy, middle recall, recency), recall consistency across trials, category clustering, and seriation. Several of the indices in current use have significant correlations with number of words recalled, and thus are not pure measures of the qualitative traits that they purportedly assess. As an alternative, Adjusted Ratio of Clustering indices generally had desirable statistical properties in random data.