The encoding of sensory input is intertwined with external attention, whereas retrieval is intrinsically related to internal attention. This study proposes a model in which the encoding of sensory input involves mainly the anterior hippocampus and the external attention network, whereas retrieval, the posterior hippocampus and the internal attention network. This model is referred to as the HERNET (hippocampal encoding/retrieval and network) model. Functional neuroimaging studies have identified two intrinsic large-scale networks closely associated with external and internal attention, respectively. The dorsal attention network activates during any externally oriented mental activity, whereas the default mode network shows increased activity during internally oriented mental activity. Therefore, the HERNET model may predict the activation of the anterior hippocampus and the dorsal attention network during the encoding and activation of the posterior hippocampus and the default mode network during retrieval. To test this prediction, this study provides a meta-analysis of three memory-imaging paradigms: subsequent memory, laboratory-based recollection, and autobiographical memory retrieval. The meta-analysis included 167 individual studies and 2,856 participants. The results provide support for the HERNET model and suggest that the anterior-posterior gradient of encoding and retrieval includes amygdala regions. More broadly, humans continuously oscillate between external and internal attention and thus between encoding and retrieval processes. These oscillations may involve repetitive and spontaneous activity switching between the anterior hippocampus/dorsal attention network and the posterior hippocampus/default mode network.
Keywords: connectivity; episodic memory; fMRI; medial temporal lobe; meta-analysis.
© 2014 Wiley Periodicals, Inc.