The hippocampal-entorhinal region forms cognitive maps of space and stimulus relations in general. In this talk, I will address how cognitive maps underlie human spatial and temporal memory. First, I will show that boundary geometry distorts spatial memory in ways consistent with deformations of grid-cell firing patterns. In a highly-immersive virtual reality experiment, mnemonic distortions through environmental geometry followed predictions from a model of grid-pattern deformations based on the successor representation.These findings strengthen the putative link between the grid metric of cognitive maps and spatial memory. Turning to cognitive maps of temporal relations in the hippocampal-entorhinal region, I will argue that hippocampal sequence representations do not merely reflect the sequential order of events or elapsing time, but are flexibly scaled to an experimentally-defined temporal reference frame. Further, the hippocampus and entorhinal cortex generalize temporal relations across multiple sequences. Together, these findings show that spatial and temporal reference frames anchor cognitive maps and shed new light onto how cognitive maps underlie episodic memory.
The hippocampal-entorhinal region forms cognitive maps of learned relations. Environmental boundaries are central to spatial orientation. However, trapezoidal boundary geometry distorts the regular firing patterns of entorhinal grid cells that proposedly provide a metric for cognitive maps. In this talk, I will first show that environmental geometry affects spatial memory in ways consistent with a grid-cell model based on the successor representation. Positional memory was degraded in a trapezoid compared to a square control environment and distance estimates between remembered positions were persistently biased. The reference frame provided by environmental geometry affected human spatial memory similarly to rodent grid-cell activity — thus strengthening the putative link between the grid metric of cognitive maps and behavior. Second, I will turn to cognitive maps of temporal relations in the hippocampus and anterior-lateral entorhinal cortex. I will address the question whether we flexibly reference sequence memories to a continuous clock as opposed to merely storing the sequential order of events or representing time as a passive “absolute”, akin to the readout of a stopwatch. In a sequence learning task, participants inferred when individual events took place based on infrequent umaskings of a hidden clock. We manipulated the clock’s speed between sequences to partially dissociate event times from their sequential order and the time objectively elapsing between them. After learning, multi-voxel patterns reflected the temporal relationships of event pairs in both hippocampus and entorhinal cortex. Hippocampal sequence memories were anchored to the reference frame of the hidden clock. Further, hippocampal and entorhinal sequence representations were organized in a way that generalized temporal relations across multiple sequences. Together, these findings suggest that cognitive maps built in service of episodic memory are tied to the external world by anchoring them to spatial and temporal reference frames.
Time is the fundamental dimension along which we organize our experience. For example, sequences of events shape our episodic memory. The hippocampus and, more recently, the anterior-lateral entorhinal cortex have been implicated in memory for event sequences. Specifically, multi-voxel pattern similarity in these regions scales with learned temporal relationships of events encountered in a sequence. However, the precise nature of these representations remains to be understood: Do we merely store the sequential order of events or do we learn temporal relations along a continuous dimension? Is time represented as a passive “absolute”, akin to the readout of a stopwatch, or can temporal relationships be adjusted flexibly depending on the task at hand? Here, we combined a novel learning task with fMRI to investigate the level of complexity at which temporal relations are represented in the hippocampal-entorhinal memory system. Participants encountered four event sequences, which we refer to as virtual days, and were to infer when individual events occurred during the virtual days based on infrequent unmaskings of an otherwise hidden clock. Importantly, we manipulated the speed of this clock between virtual days to (partially) dissociate event times from their sequential order and the time objectively elapsing between them. Participants successfully learned event times and their mnemonic responses reflected information beyond sequence order and objectively elapsed time, indicating that knowledge of temporal relations was referenced to the hidden clock. We further investigated how learning changed multi-voxel pattern representations in the anterior hippocampus and the anterior-lateral entorhinal cortex. Representational change relative to a pre-learning baseline scan reflected the temporal relationships of events in both the hippocampus and entorhinal cortex. Interestingly, our findings suggest a dissociation between representations of temporal relations of events from the same or a different virtual day. Together, we show that participants’ memory scaled with the temporal structure of the virtual days with respect to the hidden clock and thereby demonstrate that event representations in the hippocampus and entorhinal cortex are shaped by learned temporal relationships within and across sequences. Thereby, our findings provide novel insights into how multiple event sequences are organized in the hippocampal-entorhinal memory system.