Conjunctive and Item Memory Encoding

A core issue currently receiving considerable attention is whether conjunctive and item memory expression––as indexed by recollection and item familiarity, respectively––reflect quantitative or qualitative differences in the nature of the underlying memory representations, and, if the latter, whether these two forms of memory depend on separable components of the MTL memory system. From a dual-process perspective, the recollection of specific episodic details and the subjective sense of stimulus familiarity are qualitatively distinct (i.e., recollection and familiarity do not merely differ in degree of remembering, but rather differ in form). Neurally, one hypothesis is that the conjunctive representations supporting recollection depend on hippocampal auto-associative processes that encode the relation between event features and support later pattern completion, whereas item familiarity differentially depends on a match between the retrieval probe and MTL cortical representations of stored individual event elements.

In a series of fMRI experiments, we documented functional distinctions between MTL cortex and hippocampus during encoding. In an initial study [Davachi & Wagner, 2002], we reported that hippocampus is differentially engaged when encoding requires attention to the relations between items, with a subsequent memory analysis further revealing that the magnitude of hippocampal activation during relational encoding predicts later memory performance. By contrast, activation in MTL cortex (at or near PHc and ERc/PRc) is differentially greater when encoding stimuli as distinct items rather than when orienting to item-item relations. In a subsequent study of MTL encoding mechanisms [Davachi, Mitchell, & Wagner, 2003], we sorted fMRI encoding data according to subsequent item recognition and subsequent source recollection. Three critical findings were obtained. First, when sorting encoding data by subsequent item recognition (Recognized>Forgotten), independent of recollection, activation in anterior MTL cortex (at or near PRc) predicted later recognition, whereas activation in hippocampus failed to predict recognition. Second, when sorting the data according to subsequent source recollection (a measure of conjunctive memory), hippocampal activation predicted later recollection, whereas PRc activation did not. Third, this dissociation between PRc and hippocampus was accompanied by a pattern of activation in PHc that resembled that in hippocampus (rather than that in PRc), raising important questions as to the nature of PRc and PHc dissociations. These findings demonstrate that subregions within the MTL circuit support different forms of episodic encoding, and we are actively pursuing how best to characterize these distinct MTL mechanisms. Current efforts include application of high-resolution fMRI methods to target the function of specific MTL cortices and hippocampal subfields.