Encoding-related fMRI BOLD activity predicts subsequent memory for studied scenes, but not subsequent identification of perceptually similar lures.

Pattern separation is widely regarded as a hippocampally mediated process that reduces interference between memories of similar experiences. Performance on the Mnemonic Similarity Task (MST), where the requirement is to discriminate between studied items (targets) and perceptually similar lures, is commonly held to depend on pattern separation. Specifically, it has been proposed that similar lure identification is supported by a recall-to-reject strategy, whereby lures are identified as a result of the retrieval of the corresponding studied item. According to this proposal, therefore, the encoding operations that support successful target recollection and successful lure identification should be closely similar, since both mnemonic judgments depend upon subsequent recollection of the study item. Here, using fMRI, we examined this prediction. In samples of cognitively healthy young and older adults, we employed a three-choice MST (target/lure/new) with scene and object images as test items. Using ROI-based univariate and multivoxel analyses, we assessed whether encoding-related activity was predictive of the identification of target and similar lure items on the subsequent memory test. The activity elicited by scene images predicted memory performance for subsequently presented targets but not for corresponding similar lures, contrary to the recall-to-reject hypothesis. No effects could be identified for either class of object test items. The magnitude of the encoding effects for the scene targets was age-invariant, and, moreover, the univariate scene SMEs demonstrated a robust, age-invariant association with the target-lure discriminability metric.