Midazolam amnesia and dual-process models of the word-frequency mirror effect

The word-frequency mirror effect ( Glanzer & Adams, 1985) is the finding that subjects are more accurate on low frequency words than high frequency words for old and new items in recognition memory. Recently, several theorists ( Guttentag & Carroll, 1997; Joordens & Hockley, 2000; Reder et al., 2000) have proposed dual-process accounts of the word-frequency mirror effect. These accounts hypothesize that the low frequency advantage on old items arises from increased recollection of these items, while the low frequency advantage on new items arises from reduced familiarity of these items. We tested these views using midazolam amnesia. Midazolam is a safe, fast-acting benzodiazepine that produces a dense anterograde amnesia. Based on the hypothesis that midazolam amnesia should have larger effects on recollection than familiarity, we predicted that: (1) old high frequency words should have an advantage over old low frequency words in a midazolam condition (i.e., the traditional effect should reverse); and (2) the traditional advantage of new low frequency words over new high frequency words should replicate in a midazolam condition. Both predictions of the dual-process approach were confirmed. Additional analyses demonstrated that a single-process signal detection model could not account for the current results and that midazolam amnesia produces a larger effect on recollection processes than on familiarity processes.