Assessing the dissociability of recollection and familiarity in recognition memory.

Recognition memory is often modeled as constituting 2 separate processes, recollection and familiarity, rather than as constituting a single process mediated by a generic latent strength. One way of stating evidence for the more complex 2-process model is to show dissociations with select manipulations, in which one manipulation affects recollection more than the second and the second affects familiarity more than the first. One of the best paradigms for assessing such dissociations is the confidence-ratings paradigm, because within it criterial and mnemonic effects may be separately estimated. There is, unfortunately, a relative lack of easily interpretable dissociation experiments in the confidence ratings paradigm, and those that exist do not show clear evidence for dissociations. We report the results of 3 experiments with conventional manipulations designed to maximally dissociate recollection and familiarity. To provide valid assessment, without recourse to aggregation over items or participants, we develop a hierarchical version of Yonelinas' (1994) dual-process model and a novel test of dissociability for state-trace plots. The data do not provide evidence that recollection and familiarity are dissociable. Instead, estimates of recollection and familiarity are positively related across conditions and experiments. On balance, performance is more parsimoniously accounted for by a single mnemonic process that drives both recollection and familiarity estimates.