Bastin et al. present a model that is designed to make predictions about memory deficits in clinical populations. This model draws upon dual-process views of episodic memory. Laboratory work analyzed using advanced methods such as state-trace analysis (Dunn Reference Dunn2008) and computational modeling (Wixted Reference Wixted2007) has failed to find evidence of multiple processes, bringing into question a fundamental assumption of the model. However, in the domain of memory for the time-of-occurrence of events, there is extensive literature on multiple memory processes. In an influential review, Friedman (Reference Friedman1993) made a distinction between “distance-based” and “location-based” processes. Location-based processes involve retrieval of information associated with the available cues, which is then used to draw inferences about when the event occurred. Location-based processing, therefore, is analogous to recollection. Distance-based processes are very similar to familiarity in that they rely on some quality of memory (such as strength) to infer when the event took place. Friedman (Reference Friedman1993) concluded that location-based processes are most common.

Much of the work in the domain of memory for the time-of-occurrence of events has relied on testing people's memories for events that are part of the public record or those that have been recorded in personal diaries (Kemp Reference Kemp1999). Most of these studies used event stimuli that occurred outside of the laboratory, but which could be dated because they were part of the public record or had been recorded in personal diaries (Kemp Reference Kemp1999). Many of the existing studies also asked people to determine the exact dates of occurrence of these events. The method of reporting, however, may influence the strategy that people employ. Furthermore, using public events may tend to emphasize unique flashbulb-type memories which, in turn, may not reflect how people retrieve the time-of-occurrence of everyday mundane and personally experienced events. We conducted several studies using smartphone-based sensors to record people's everyday life events and used those events to probe how they retrieved the week and day of occurrence of these events a few weeks after they occurred (Dennis et al. Reference Dennis, Yim, Sreekumar, Evans, Garrett, Sederberg, Bunzelmann, Howes, Tenbrink and Davelaar2017; Sreekumar Reference Sreekumar2015; Yim et al. Reference Yim, Garrett, Baker, Sreekumar and Dennis2019). Using a hierarchical Bayesian model-comparison framework, we concluded that location-based processes were employed when people had to retrieve more precise information (i.e., day of occurrence) compared to distance-based strategies when asked about the week of occurrence. Therefore, experience sampling work suggests that when one looks at people's real-world memories that have not been stripped of cues necessary to form reliable inferences, one can see clear evidence of a distinction between what Friedman (Reference Friedman1993) called “distance”- and “location”-based processes. The prior difficulty in dissociating location-based and distance-based processes behaviorally also led to neuropsychological research on the contribution of various brain regions to memory for time. For example, Curran and Friedman (Reference Curran and Friedman2003) recorded event-related potentials (ERPs), where participants engaged in temporal memory tests that were designed to emphasize one of the two processes and showed greater late-frontal ERP effects under conditions that fostered location-based processing.

In memory-for-time experiments, it is easier to manipulate these different components than in a recognition memory experiment because it is possible to vary the nature of the query and the time-scale probed (e.g., month, week, day, hour, etc.). We also have access to a wider range of the ratio between retention interval and the temporal separation between probe events, which has been identified as another factor that plays a role in fostering one process over the other. Therefore, both neuropsychological and more recent behavioral experiments based on experience sampling provide evidence for multiple processes in memory for when an event occurred, where the dominant processes are very similar to recollection and familiarity in recognition memory.

While the multiple memory processes assumption has some support from the memory-of-time literature, Bastin et al. rely on findings of fMRI (functional magnetic resonance imaging) activation of brain regions in discrimination tasks to support the assumption that the perirhinal/anterolateral entorhinal cortex is specialized for pattern separation of entities (i.e., objects). Although the hippocampal circuit has known mechanisms that allow both pattern completion and pattern separation, the mechanisms that would allow the perirhinal/anterolateral entorhinal cortices to specifically pattern-separate entities are unclear. In fact, major types of computation in the brain seem to be redundant and distributed (e.g., Siegel et al. Reference Siegel, Buschman and Miller2015; Tian et al. Reference Tian, Huang, Cohen, Osakada, Kobak, Machens, Callaway, Uchida and Watabe-Uchida2016). Furthermore, assigning “entity separation” computations to a very specific brain region seems somewhat contradictory to the goal of moving away from assigning processes to brain regions, to thinking about the type and complexity of representations they are capable of. Temporal context signals, which guide memory encoding and retrieval, are found everywhere in the brain (e.g., in various regions within the temporal lobe; El-Kalliny et al. Reference El-Kalliny, Wittig, Sheehan, Sreekumar, Inati and Zaghloul2019). Folkerts et al. (Reference Folkerts, Rutishauser and Howard2018) found that even highly visually selective units participate in a gradually changing representation of temporal context. However, the rate at which these signals drift in time may depend on where the brain region lies along the representational hierarchy because temporal receptive windows follow the same hierarchy (Lerner et al. Reference Lerner, Honey, Silbert and Hasson2011). Within the lateral entorhinal cortex specifically, Tsao et al. (Reference Tsao, Sugar, Lu, Wang, Knierim, Moser and Moser2018) reported that population states encoded temporal context information. They also previously identified a population of lateral entorhinal cortex cells that encoded object-location associations (Tsao et al. Reference Tsao, Moser and Moser2013) and, importantly, these cells were different from object-specific cells. Therefore, even the anterolateral entorhinal cortex (human homolog of the rodent lateral entorhinal cortex) “entity representational core” assumed in Bastin et al.’s model seems to have an important role to play in context and associative representations that extend beyond conjunctions of simpler features. Similarly, the hippocampal formation is not required for some context-discrimination tasks. For example, such contextual discrimination tasks can be readily learned even by animals with hippocampal lesions (see Rudy [Reference Rudy2009] for a review).

In summary, Bastin et al.’s framework is motivated by dual-process accounts of memory which are well supported by both behavioral and neuroimaging data; but the distinctions made between entity and context representational systems may not accurately reflect the distributed nature of these representations in the brain.