Although Suddendorf & Corballis (S&C) provide a thought-provoking framework for the cognitive processes involved in remembering the past and imagining the future, their discussion of the neural mechanisms underlying these processes fails to adequately acknowledge the important role of the medial temporal lobe (MTL) in episodic memory. The authors argue that episodic memory is the most flexible of the memory-based neural systems and also the most recently evolved, pointing to the expansion and reorganization of the prefrontal cortex throughout evolution as the driving force behind the emergence of episodic memory in humans. By limiting their discussion of neural mechanisms to the prefrontal cortex, the authors ignore the role that a much more primitive region of the brain plays in episodic memory: the hippocampus. The MTL, and in particular the hippocampus, has been repeatedly implicated in the encoding and retrieval of episodic memories (for review, see Davachi Reference Davachi2006), and recent evidence indicates that this region may also play an important role in future event construction (Addis et al. Reference Addis, Wong and Schacter2007; Hassabis et al. Reference Hassabis, Kumaran, Vann and Maguire2007; Okuda et al. Reference Okuda, Fujii, Ohtake, Tsukiura, Tanji, Suzuki, Kawashima, Fukuda, Itoh and Yamadori2003).
As is now well-established, the MTL is anatomically suited to bind information from different sensory cortices into a cohesive trace: the hippocampus, via the entorhinal cortex, receives convergent input from a variety of unimodal and polymodal association cortices in the temporal, frontal, and parietal lobes (Bunsey & Eichenbaum Reference Bunsey and Eichenbaum1995; Suzuki & Amaral Reference Suzuki and Amaral1994). Several neurocomputational models demonstrate how the unique anatomy and physiology of the hippocampus enables the creation of specific, detailed memories. O'Reilly and Rudy (Reference O'Reilly and Rudy2001) proposed the Complementary Learning Systems (CLS) model to differentiate between representations serving episodic recollection and those serving familiarity. According to the CLS model, the hippocampus is specialized for rapidly storing episodic memories, whereas the surrounding neocortex stores information about the statistical regularities in the world. As in other models, the CLS model supposes that the hippocampus uses sparse coding to store distinct, pattern-separated representations of elements, thus avoiding catastrophic interference, or overlapping representations. The neocortex, on the contrary, assigns similar patterns of activation to similar stimuli, thereby representing the shared structure of events via overlapping representations, and generalizing to novel stimuli when they share features of previously encountered stimuli.
During encoding, synapses in the perforant path projecting from the entorhinal cortex to the dentate gyrus, CA3 and CA1 fields, undergo self-organization by creating increasingly sparser representations of inputs. Meanwhile, excitatory synapses in the CA1 and CA3 regions associate these sparse representations, forming memory traces (Hasselmo et al. Reference Hasselmo, Wyble and Wallenstein1996). During retrieval, a partial pattern of previously studied material is presented to the hippocampal network. Upon reaching CA3, the partial input pattern is completed via strengthened weights in the extensive recurrent collateral network. The complete CA3 representation then activates the associated CA1 representation, which, via the subiculum, activates the original entorhinal pattern (O'Reilly & McClelland Reference O'Reilly and McClelland1994; Rolls & Kesner Reference Rolls and Kesner2006). This activation, in turn, serves to trigger reinstatement of activity in the neocortex reflecting the multi-modal content of the original memory.
Functional neuroimaging studies have demonstrated a selective role for the hippocampus in both encoding (Uncapher & Rugg Reference Uncapher and Rugg2005) and retrieving (Eldridge et al. Reference Eldridge, Knowlton, Furmanski, Bookheimer and Engel2000; Woodruff et al. Reference Woodruff, Johnson, Uncapher and Rugg2005) episodic memories. In an effort to determine activation patterns at the subregional level, Eldridge et al. (Reference Eldridge, Engel, Zeineh, Bookheimer and Knowlton2005) scanned participants using high-resolution functional magnetic resonance imaging (fMRI) of the MTL during both encoding and retrieval. Results show that the dentate gyrus and CA3 field were selectively active during episodic memory formation, whereas the subiculum was active during retrieval of the study episode.
Data from single-unit recordings in humans also support the idea that the hippocampus relies on sparse coding to function efficiently. Viskontas et al. (Reference Viskontas, Knowlton, Steinmetz and Fried2006) examined recognition memory in a group of epilepsy patients undergoing pre-surgical mapping. Patients viewed novel images and were asked to make old/new judgments for each presentation. Upon the second presentation of a given stimulus, a large subset of hippocampal neurons showed a decrease from baseline firing, while a small subset showed excitation. These excitatory responses have been shown to be specific for particular items (Quiroga et al. Reference Quiroga, Reddy, Kreiman, Koch and Fried2005). Inhibition of firing to previously seen stimuli may serve to increase signal-to-noise ratio, thereby supporting sparse coding and effective mnemonic processing. Furthermore, preliminary data suggest that some cells in the hippocampus show excitatory responses to a particular item only when that item is encountered alone or in the context in which it was originally encountered (Viskontas Reference Viskontas2006).
More recently, evidence has emerged supporting a role for the MTL in imagining future experiences. In a positron emission tomography (PET) study, Okuda et al. (Reference Okuda, Fujii, Ohtake, Tsukiura, Tanji, Suzuki, Kawashima, Fukuda, Itoh and Yamadori2003) found greater or equivalent levels of activation while subjects spoke about future prospects versus past experiences in the right hippocampus and bilateral parahippocampal cortices. Similar results were reported in a fMRI study comparing past and future event elaboration, in which the left hippocampus and bilateral parahippocampal cortices were commonly engaged by both tasks (Addis et al. Reference Addis, Wong and Schacter2007). Interestingly, the authors also report that construction of future versus past events uniquely activated the right hippocampus. Neuroimaging results are supported by reports of marked impairments in imagining future events in patients with hippocampal amnesia (Hassabis et al. Reference Hassabis, Kumaran, Vann and Maguire2007).
As the aforementioned neurocomputational, single-unit recording, functional MRI, PET, and patient studies suggest, there is a wealth of evidence supporting a role for the hippocampus in both episodic memory and in envisioning future experiences. Although we do not deny that the prefrontal cortex plays an important and complementary role in these processes, we find that S&C rely too heavily on the recent evolution of this region in humans in arguing against mental time travel in other species. In fact, a recent study of scrub jay behavior suggests that these birds are, in fact, capable of future planning without reference to their current motivational state, or fixed action patterns, challenging the notion that future planning requires a prefrontal cortex (Raby et al. Reference Raby, Alexis, Dickinson and Clayton2007). The hippocampus is phylogenetically among the oldest parts of the brain, and may therefore allow nonhuman animals to project both backwards in time and forwards into the future. Thus, as increasingly innovative nonverbal tests of episodic memory develop, we may be surprised to discover the time-traveling capabilities of a variety of species.
