Group aggregates and individual reliability: the case of verbal short-term memory.

Two experiments examined the generalizability of the effects of word length and phonological similarity with visual and auditory presentation in immediate verbal serial ordered recall. In Experiment 1, data were collected from 251 adult volunteers drawn from a broad cross-section of the normal population. Word length and phonological similarity in both presentation modes significantly influenced the group means. However, 43% of the subjects failed to show at least one of the effects, and the likelihood that effects appeared was highly correlated with verbal memory span. In Experiment 2, 40 subjects of the original sample were retested, 20 of whom had failed to show one or more effects in Experiment 1. Whether or not an effect had appeared for individual subjects on the first test session was a poor predictor of whether the effect would appear on retest. Finally, an analysis of subject reports demonstrated that the patterns of experimental data could be accounted for in part by the strategies that subjects reported using, and the effect of strategy was independent of the effect of span. The implications of these findings for theories of verbal short-term memory are discussed.