The human hippocampus is essential for both encoding and recollection, but it remains controversial whether there is a functionally different involvement of anterior versus posterior parts of the hippocampus in these memory processes. In the present study, we examined encoding and retrieval processes via intrahippocampal recordings in 27 patients with unilateral temporal lobe epilepsy. Multicontact depth electrodes were implanted along the longitudinal axis of the hippocampus as part of the presurgical evaluation. In a continuous word recognition test, subjects had to indicate whether words were new or already presented. Recognized old words, as compared to new words, resulted in a larger P600 component, as well as in a larger late negative component (LNC, 600–900 msec). In addition, subsequently remembered words elicited a larger positivity (400 to 900 msec) than later forgotten words. We found differences concerning the distribution along the hippocampus for the LNC old-new effect, reflecting successful retrieval, as well as for the subsequent memory effect, reflecting successful encoding. Both effects were larger the further posterior an electrode was located in the hippocampus. Findings are suggestive for a predominant posterior hippocampal involvement in both verbal encoding and retrieval.

Up to now, two conflicting theories have tried to explain the genesis of averaged event-related potentials (ERPs): Whereas one hypothesis claims that ERPs originate from an event-related activation of neural assemblies distinct from background dynamics, the other hypothesis states that ERPs are produced by phase resetting of ongoing oscillatory activity. So far, this question has only been addressed for early ERP components. Late ERP components, however, are generally thought to represent superimposed activities of several anatomically distinct brain areas. Thus, the question of which mechanism underlies the genesis of late ERP components cannot be easily answered based on scalp recordings. In contrast, two well-investigated late ERP components recorded invasively from within the human medial temporal lobe (MTL) in epilepsy patients, the so-called MTL-P300 and the anterior MTL-N400 (AMTL-N400), are based on single source activity. Hence, we investigated whether the MTL-P300 and the AMTL-N400 are based on an event-related activity increase, a phase reset of ongoing oscillatory activity or both. ERPs were recorded from the hippocampus and rhinal cortex in subjects performing a visual oddball paradigm and a visual word recognition paradigm. With wavelet techniques, stimulus-related phase-locking and power changes were analyzed in a frequency range covering 2 to 48 Hz. We found that the MTLP300 is accompanied by both phase reset and power increase and that both effects overlap partly in time. In contrast, the AMTL-N400 is initially associated with phase locking without power increase and only later during the course of the AMTL-N400 we observed an additional power increase. In conclusion, both aspects, event-related activation of neural assemblies and phase resetting of ongoing activity seem to be involved in the generation of late ERP components as recorded in cognitive tasks. Therefore, separate analysis of event-related power and phase-locking changes might reveal specific insights into the mechanisms underlying different cognitive functions.