The validity of "conceptual span" as a measure of working memory capacity

Three experiments tested whether a modified version of the Clustered Conceptual Span task (H. J. Haarmann, E. J. Davelaar, & M. Usher, 2003), which ostensibly requires active maintenance of semantic representations, predicted individual differences in higher-order cognitive abilities better than short-term memory (STM) span tasks or nonsemantic versions of the "Conceptual" task did. Nonsemantic Conceptual tasks presented short word lists clustered by color, first letter, or initial vowel sound, and cued subjects to recall only 1 of 3 clusters from each list; the Semantic task clustered words by taxonomic category. The Semantic Conceptual task generally failed to predict incremental variance in either verbal abilities or general fluid intelligence beyond the other Conceptual tasks or STM span tasks. Although the Semantic task showed a stronger relation to working memory span tasks than did the nonsemantic tasks (Experiment 3), that stronger relation did not translate into strong prediction of cognitive individual differences.