Successful recovery of words from episodic memory relies strongly on semantic processes at the time of encoding. Evidence from several functional magnetic resonance imaging (fMRI) studies has shown that changes in neural activity in the left inferior prefrontal cortex (LIPFC) during semantic encoding predict subsequent memory performance. This evidence has been taken to suggest that LIPFC plays a critical role in memory formation. Functional neuroimaging findings, however, do not establish a causal brain-behavior relationship. To determine whether there is a causal link between LIPFC involvement at encoding and subsequent success in memory performance, we conducted a two-part study in which we first used fMRI to localize encoding-related activation in LIPFC and then employed repetitive transcranial magnetic stimulation (rTMS) to manipulate neural processes in LIPFC during semantic encoding. To demonstrate the neuroanatomical specificity of any observed effect and to control for nonspecific rTMS side effects, we also stimulated neural processes in two control sites. Using frameless stereotaxy, we positioned the stimulation coil to target (1) the LIPF region that was activated during fMRI (mean xyz = −48 35 5); (2) the homologous righthemisphere region; and (3) an additional left parietal control site. At each site, “stimulated” items (600 msec of 7-Hz rTMS with Cadwell Round Coil) were intermixed with items presented without concurrent stimulation. Subsequently, subjects performed a recognition memory task for the words encountered. We found support for the predicted causal brain-behavior relationship, which was specific to LIPFC. When comparing recognition scores for stimulated items, normalized for variations in performance on nonstimulated trials, we found that words encoded under LIPFC stimulation were subsequently recognized with higher accuracy than words encoded under stimulation in the two cortical control sites. By contrast, no performance difference emerged when the two control sites were compared with each other. Based on additional analyses of the rTMS effects observed directly at the time of encoding (i.e., on semantic-decision performance), we suggest that LIPFC stimulation may have produced its effect on recognition memory, at least in part, through the triggering of more extensive processing of the stimulated items and an ensuing gain in item distinctiveness. Physiological processes of facilitation probably also contributed to the observed memory benefit. Together, these findings suggest that LIPFC does play a causal role in episodic memory formation.

A plethora of studies, across many species, have now demonstrated that the hippocampal region plays a critical role in memory for spatial location. In spite of this compelling evidence, a number of important neuropsychological and neuroanatomical issues remain unresolved. In the present study, the functional anatomy of object-location memory was investigated using positron emission tomography (PET) with magnetic resonance imaging (MRI). Regional cerebral blood flow (rCBF) was measured while normal volunteers encoded, and then retrieved, the locations of eight familiar objects presented on a computer screen. In two analogous conditions, designed to fractionate object-location memory into its component processes, the subjects were simply required to encode, and then to retrieve, eight distinct locations represented by identical white boxes on the screen. An increase in rCBF was observed in the region of the right parahippocampal gyrus corresponding to entorhinal cortex when the Retrieving Location condition was subtracted from the Retrieving Object-Location condition. In contrast, when the Encoding Location condition was subtracted from the Encoding Object-Location condition, no significant rCBF changes were observed in the hippocampal region although significant activation was observed, bilaterally, in the anterior fusiform gyrus. In addition, the two encoding conditions activated left-hemisphere regions preferentially, whereas the two retrieval conditions activated right-hemisphere regions. Together, these findings suggest that the human right hippocampal region is critically involved in retrieving information that links object to place. The secondary finding that encoding and retrieval appear to be lateralized to the left and right hemispheres respectively, is discussed with reference to current models of episodic memory, and alternative hypotheses are considered.