How do we remember the different places that we have visited during our day? Studies of brain activity in rats suggest that each place a rat visits is represented by a different pattern of activity in a region of the brain called hippocampus. However, it is not clear what role this “spatial remapping” plays in the formation of memories in humans.

Kyle et al. used a technique called functional magnetic resonance imaging (fMRI) to investigate how our brain represents the different places we’ve visited, and how this is linked to how well we can remember these locations. For the experiments, human volunteers played a video game where they visited four virtual environments in a different order. Later, the volunteers were asked to remember details about the virtual environments they had visited while their brain activity was monitored using fMRI.

The experiments show that in order to remember distinct locations – even if they have some features in common – the hippocampus produces patterns of activity that have very little overlap with each other. This process is termed “pattern separation”. Sometimes our memory confuses different locations, which could be due to a failure of the brain to distinguish between the patterns of brain activity that represent these locations.

Kyle et al.’s findings provide the first evidence for spatial remapping in the human hippocampus and its importance in forming memories of locations. The next steps are to find out how many different environments our brain might be able to store at the same time, and to identify factors that could aid in our memory for spatial locations.

Place neurons (e.g. “place cells”) in the rodent hippocampus preferentially fire in a particular spatial location (O'Keefe and Dostrovsky, 1971), the combination of which provide a neural code for that spatial environment (O'Keefe and Nadel, 1978; Wilson and McNaughton, 1993; Muller and Kubie, 1987). The collection of active place cells in an environment is thought to serve as a “cognitive map,” providing a spatial framework for both navigation and memory more generally (O'Keefe and Nadel, 1978; Redish, 1999; Buzsáki and Moser, 2013). Two fundamental properties of place cells are their stability (Thompson and Best, 1990; Hill, 1978) and their environmental specificity, also known as “remapping” (Muller and Kubie, 1987). Without reliable recapitulation of the ensemble of place cells representing a specific “map,” spatial memory is impaired (e.g., Kentros, 1998; Morris et al., 1986; McHugh et al., 2007). Remapping, a form of reorganization of hippocampal “maps” for different environments, is theorized to be a fundamental mechanism to navigation and memory more generally. However, the exact link between memory performance and remapping has yet to be fully established (Jeffery et al., 2003; Colgin et al., 2008). In humans, invasive recordings from the hippocampus have demonstrated place-coding neurons in single environments (Ekstrom et al., 2003; Jacobs et al., 2013; Miller et al., 2013). Additionally, the human hippocampal formation is important to episodic memory more broadly (Spiers et al., 2001), with place cells activating during item recall (Miller et al., 2013) and several studies demonstrating the ability to decode both location and episode-specific details from hippocampal fMRI blood-oxygen-level-dependent (BOLD) patterns (Guterstam et al., 2015; Hassabis et al., 2009; Chadwick et al., 2010). Whether the human hippocampus represents one spatial environment as either the same or different from another, however, – a cornerstone of the idea that the hippocampus may compute spatial “maps” as part of a larger role in processing memories—remains unknown and untested.

In addition to serving as a basic marker of memory, the environmental specificity of the hippocampus is thought to elucidate critical theoretical mechanisms of hippocampal function known as pattern separation and completion. These processes were predicted by early computational models and are thought to account for the memory interference errors commonly encountered in memory research and our everyday lives (Yassa and Stark, 2011; Marr, 1971; Kohonen, 1977; Hunsaker and Kesner, 2013). This theory states that pattern separation is a process that makes memories neurally distinct during memory storage and pattern completion a process by which memories are retrieved from a neural cue. Pattern separation and completion are thus thought to be important complements to each other (Yassa and Stark, 2011). Theoretical models and several empirical findings additionally suggest that CA3/DG and CA1 subfields mediate pattern separation and completion in the hippocampus (Guzowski et al., 2004; Leutgeb et al., 2007; Leutgeb, 2004; Bakker et al., 2008). Yet exactly how these findings relate to human spatial memory remains unclear.

Pattern completion is thought to rely on neural “attraction” between the cues that precede recall and stored representations, therefore allowing the cue to trigger re-instantiation of the full memory (Rolls, 2010). This property of attraction has the important implication that memories that are neurally similar will compete, producing interference in the case that the incorrect memory wins this competition (Colgin et al., 2008; Shapiro and Olton, 1994). Theoretical models, therefore, postulate the central importance of pattern separation as critical to making memories less similar and thus avoiding interference due to neural attraction. Alternative accounts of memory interference, however, instead argue against a pure pattern separation/completion based account in favor of a model which posits inhibition of interfering memories from executive control regions during memory recall (Anderson, 2003). This account instead predicts that similar representations can co-exist, but can be selected, maintained, and strengthened by executive control centers during memory retrieval. Therefore, a definitive neural link between behavioral interference, neural pattern separation, and spatial remapping is necessary to resolve this debate and clarify the function of the hippocampus in memory.

The aims of this study, thus, were three fold. The first was to examine whether humans also recapitulate neural codes for the same environment as well as bifurcate codes for different environments using fMRI and a multivariate pattern analyses. A second and critical test of whether remapping occurs in humans, however, is whether situations involving highly interfering spatial contexts can produce remapping failures (e.g., Skaggs and McNaughton, 1998; Spiers et al., 2015), and if so, what neural mechanisms characterize these errors. A final goal was to provide a link between behavioral measures of environment knowledge and neural measures of spatial remapping in humans.

To determine whether the human hippocampus also contains environment-specific neural codes, participants first explored four cities with varying levels of shared spatial context. Two cities (Cities 1 & 2) involved the same stores arranged in the same geometry, but with two store locations swapped (also termed the similar cities). A third city (City 3) involved the same stores as Cities 1 & 2, but arranged in a novel geometry and therefore at novel locations (interference city). Finally, City 4 involved a completely novel set of stores and geometrical arrangement (distinct city; see Figure 1a, Materials and methods). Following each round of navigation, participants drew maps of the environment to ensure that they accurately encoded spatial configuration details (Figure 1b, see Materials and methods). Following navigation and map drawing of all cities, participants entered the scanner where they performed two retrieval blocks per city (Figure 1c). Participants were instructed to recall a specific city during each retrieval block, with each trial involving judgments of the relative distances between stores (please see Methods for further details).

Figure 1 with 2 supplements see all

Download asset Open asset

To better understand the extent to which the different cities involved competing representations, we compared learning rates of city-specific map-drawing performance in a separate behavioral study (Figure 1—figure supplement 1–2 and Materials and methods). We predicted that City 3 would experience slower learning relative to Cities 1 & 2 because all of City 3’s store locations were in conflict with store locations from Cities 1 & 2. Cities 1 & 2, on the other hand, had only 2 conflicting store locations, which could be learned via a simple swap. We found that the slowest learning did occur for the interference city (City 3) as well as the greatest confusion with the similar cities (Cities 1 & 2). In contrast, transitioning between the similar cities (e.g., City 1 to City 2) resulted in little detrimental effect on learning; in fact, learning one facilitated the learning of the other. Finally, no other cities facilitated learning of the distinct city (City 4) nor did learning the distinct city interfere with learning any other city. These findings suggested that City 1 and 2 (similar cities), despite being most similar, were easily distinguished from one another and that City 3 involved a representation prone to interference from similar cities (thus called the inference city). City 4, in contrast, did not interact with the other cities (distinct city) due to its novel stores and spatial geometry (Newman et al., 2007), allowing it to be readily distinguished from the other three cities.

We then tested whether participants’ accuracy during retrieval of spatial distances of landmarks within the different cities (i.e., is store X closer to store Y or store Z?). Participants performed well above chance on spatial retrieval of all four cities (Figure 1d, single sample t-test against chance performance: t(18)>4.7, p<0.0002). A one-way repeated-measures ANOVA revealed differences in performance as a function of city (F(3,54) = 20.9, p<0.001), which was driven by significantly lower retrieval performance on the interference city than all other cities (t(18)>4.65, p<0.0002). No other cities differed from each other in terms of performance, confirming the results of our navigation and map data analysis suggesting that retrieving information from the interference city (City 3) resulted in a tendency to confuse traces with those used in Cities 1 & 2. It is important to note that even though Cities 1 & 2 involved similar representations (with the difference being two swapped stores), performance on these two cities could not be explained by using the same responses between the two cities or the same responses and guessing on the swapped stores (see Materials and methods). These findings support the idea that participants were nonetheless using at least partially non-overlapping representations to retrieve details from Cities 1 & 2.

Classification of city-specific retrieval patterns in the hippocampus demonstrates successful decoding of all spatial contexts except the interfering environment

Our first and most basic prediction was that our human participants, analogous to remapping the rodent, would exhibit hippocampal voxel patterns that could uniquely identify each spatial environment. To address this prediction, we performed a searchlight classifier throughout the MTL (see Materials and methods for details and Figure 2c). This approach allowed us to naively identify MTL regions where voxel patterns carried city specific information. Our inclusion of the interference city into this analysis allowed us to address additional questions of pattern separation/completion. For instance, if the reduced retrieval performance of the interfering city could be attributed to insufficiently separated neural patterns where models predict neural competition at retrieval, the classifier should disproportionally misclassify the interference city as one of the similar cities (Cities 1&2) but not the distinct city (City4).

Figure 2 with 1 supplement see all

Download asset Open asset

The pattern classifier correctly identified three of the four cities at levels above chance, revealing a cluster in hippocampal regions left CA3/DG and CA1 that significantly classified city identity (Figure 3a–c). Analyzing this cluster in a one-way repeated-measures ANOVA, with classifier performance of each city as a separate factor, revealed significant differences between cities (Figure 3C, F(3,54) = 12.9, p<0.001). Testing each city’s classifier performance against chance revealed that the classifier performed above chance on all cities except the interference city (Cities 1, 2, & 4 above chance: t(18) > 3.2, p<0.006, two-tailed). Conversely, interference city classification performance was consistently below chance levels (t(18) = −3.2, p = 0.006, two-tailed), despite overall classification (across all cities) being well above chance (Figure 3C, t(18) = 5.6, p = 2 × 10⁻⁵, two-tailed). We note that within our search space, which included the hippocampus, parahippocampal, fusiform, lingual, and inferior gyrus, only this cluster in CA3/DG and CA1 exceeded the family-wise error rate cluster size correction (see Materials and methods). Also, a control analysis that altered the classifier training protocol produced similar results suggesting these results were robust to differing analysis approaches (Figure 3—figure supplement 1, see Materials and methods for more details). This finding confirmed our prediction that, like the rodent, the human hippocampus contains environment specific representations in the CA1 and CA3/DG subfields for the environments that were most easily retrieved (Cities 1, 2, & 4). Additionally, the below chance classifier performance for City 3 and close resemblance of the retrieval accuracy and classifier accuracy seemed to indicate that low behavioral performance on City 3 may have related to lower levels of voxel pattern remapping for this environment.

Figure 3 with 2 supplements see all

Download asset Open asset

To further explore the idea that low classification levels of the interference city (City 3) were due to the competing representations of the similar cities, we inspected interfering city classification results. Here, we predicted that the classifier would misclassify the interference city trials as either City 1 or City 2 on more than 50% of trials (chance level). This would be consistent with our behavioral results, which indicated that the interfering city was most often confused with Cities 1 and 2 during learning (see Materials and methods). This would potentially support a pattern separation/completion account of hippocampal remapping of spatial memory errors. As predicted, interference city trials were incorrectly labeled as one of the two similar cities at levels well above 50% (t(18) = 2.7, p = 0.01, two-tailed, Figure 3c). This suggested that a disproportionate amount of interference city trials resembled the similar cities. In contrast, trials for Cities 1 & 2, on which retrieval performance was well above chance, were incorrectly classified as City 3 significantly less than chance (t(18) = -8.6, p<0.001 corrected; see Figure 3—figure supplement 2). Overall, these findings are consistent with the idea that retrieval errors on City 3 could be attributed to, at least in part, insufficient differentiation of neural patterns from Cities 1&2.

A second critical prediction from the pattern separation/completion account would suggest that better individual performance on the interference city should be attributable to more distinct and therefore more readily classified representations of this city. Therefore, we also examined the link between interference city classifier performance and participant retrieval performance by seeing if the two measures were correlated. We found that interference city retrieval and classifier performance were significantly correlated (r(17) = 0.61, p = 0.006, Figure 3d), a result which persisted even when matching the number of classifier training trials for each city (Figure 3—figure supplement 1). The link between participants who performed better on the interference city and those that had more readily classified neural representations of the interference city thus suggested that our best performing participants had neural representations that were more differentiated from each other than those of poorly performing participants. Together, the findings from the searchlight classification analysis support the idea that the human hippocampus exhibits environmental specificity. Further, the low classification performance of the interference city, and correlation between behavioral and neural classifier performance, favor a pattern separation/completion based account of memory interference.

Although searchlight classifier analyses have advantages, in our case, the ability to naively identify regions within the human hippocampus exhibiting environment specific coding, they are not as well suited to hypotheses involving functional dissociations between subfields. For instance, searchlight clusters do not necessarily carry unique signals from different brain regions (Woo et al., 2014; Etzel et al., 2013). Pattern classification also cannot indicate whether neural codes are more similar within the same vs. between different environments, a cornerstone of spatial remapping findings in the rodent (Wilson and McNaughton, 1993; Muller and Kubie, 1987). We therefore employed a region-of-interest (ROI) based, multivariate pattern similarity (MPS) approach to 1) determine whether different human hippocampal subfields played different roles in spatial remapping and 2) provide more specific alignment with findings from rodents indicating higher neural similarity for the same vs. a different spatial environment.

ROI-based multivariate pattern similarity (MPS) voxel remapping suggests a functional dissociation of CA3/DG and CA1

As outlined in the prior section, our operational definition of remapping was voxel similarity within the same city (context reinstatement) and dissimilarity between different cities (remapping). To quantify this using multivariate pattern similarity (MPS), we created a voxel “remapping index” defined as within-environment similarity minus the average between-environment similarity (Figure 2d–e). Because our searchlight classifier analysis implicated CA1 and CA3/DG subfields in exhibiting voxel pattern remapping and remapping in the rodent is predominantly studied in CA1 and CA3/DG, here, we only include data from only CA1 and CA3/DG (for other subfield results, please see Figure 4—figure supplement 1). The results of this analysis are presented in Figure 4a–c. A 2x4 subfield by city repeated measures ANOVA revealed a main effect of subfield (F(1,18) = 6.93, p<0.02), and a marginal subfield by city interaction (F(3,16) = 3.14, p<0.06). This suggested that CA3/DG tended to exhibit more remapping but also that the pattern of results across cities differed for CA3/DG vs. CA1. Also of interest was whether each city remapping score was significantly different from chance. In left CA3/DG, remapping scores were significantly above zero for Cities 1 & 2 (all t-tests one tailed, t(18)>1.8, ps<0.05 corrected; see Materials and methods) and marginally above zero City 4 (t(18) = 1.6, p = 0.06 corrected). Thus, left CA3/DG MPS patterns indicated higher pattern similarity when participants retrieved spatial distances within Cities 1, 2, and 4 compared to the correlations of the patterns between different cities.

Figure 4 with 1 supplement see all

Download asset Open asset

Left CA1, in contrast, had remapping scores that were significantly above zero for City 4 (t(18) = 2.29, p<0.05, corrected), but not for Cities 1-3 (ts<−0.25, p>0.6). Consistent with the searchlight classifier results, City 3 did not exhibit significant remapping in either CA3/DG or CA1 (ts<0.98, ps>0.17). This analysis further clarified the results of the searchlight classifier by suggesting that when characterizing entire subfields, more cities showed significant remapping effects in CA3/DG than CA1, but that CA1 did exhibit remapping for the most distinct city (City 4). Together, these findings suggested that CA1 and CA3/DG showed remapping for distinct city retrieval, while only CA3/DG showed remapping between similar cities. Overall, these findings also confirm the importance of pattern separation mechanisms to remapping in both human CA3/DG and CA1, an issue we consider with greater depth in the Discussion.

Analysis of retrieval trials from the interference city suggests a partially unstable representation

Our searchlight classifier results suggested that correct trials tended to disproportionally resemble the similar cities but this effect was reduced for our best performing participants. This begged the question, if even correct interference city trials resembled the similar cities, did incorrect trials show even greater resemblance to the similar cities than the correct trials? If the representation of the interference city was unstable and easily attracted to the similar cities, then voxel patterns for incorrect interfering city trials should be highly correlated with voxel patterns of correct trials from the similar cities. Conversely, we would not expect to see high pattern similarity between similar cities and interference city trials if they were correctly answered. An important control comparison was included to make sure that this effect could be attributed to interference rather than a general property of incorrect retrieval. We would not expect the incorrect interference city trials to show similarity to the distinct city (because the distinct city was substantially different from Cities 1–3), and thus neither incorrect nor correct interfering city trials should have correlated voxel patterns with the distinct city.

To address these issues, we calculated MPS to compare correct and incorrect interference city trials with other city trials using matched visual stimuli (triads) (see Materials and methods), the results of which are presented in Figure 5. Voxel patterns in left CA3/DG (and right CA3/DG) on incorrect high interference city retrieval trials were significantly correlated with correctly retrieved voxel patterns in the similar cities (Figure 5A). Importantly, incorrect interference city trials were significantly more correlated with correct similar city (City 1 & 2) trials than were correct interference city trials with correct similar city or distinct city trials in CA3/DG (left bar greater than others, two-tailed t-test, t(18)>2.2, p<0.04). This effect was present in left CA3/DG (it was also present in right CA3/DG, two-tailed t-test, t(18)>3.7, p<0.001, Figure 5B, Figure 5—figure supplement 1A) but not present in CA1 nor any other subfield (t(18)<1.8, p>0.09, Figure 4—figure supplement 1C, Figure 5—figure supplement 1B). These results augment our searchlight classification results by demonstrating resemblance of incorrect interference city trials to the similar cities. Specifically, our findings support the idea that the unstable, weakly differentiated neural patterns of City 3 allowed the stable representations of Cities 1 & 2 to occasionally “outcompete” City 3 during retrieval. This in turn led to pattern completion of the wrong representation and selection of an incorrect response.

Figure 5 with 2 supplements see all

Download asset Open asset

We believe that four novel components of our findings aid in understanding of human hippocampal function and its relation to memory processing. First, we extend environment specific coding to the human hippocampus using voxel-pattern based analyses. Using a searchlight classifier approach, which naively identified medial temporal lobe regions carrying city specific information, we found a cluster of voxels in CA3/DG and CA1 whose patterns decoded specific cities during retrieval. Next, using an MPS ROI approach which utilized all voxels from a subfield to characterize similarity within and between cities using simple correlations, both CA3/DG and CA1 showed higher similarity within city than between cities. Although past studies in humans have confirmed the presence of location specific coding in the hippocampus (Ekstrom et al., 2003; Jacobs et al., 2013), measuring remapping typically requires a large number (>40) of simultaneously recorded cells (Wilson and McNaughton, 1993), which are difficult to obtain in most human studies. Additionally, although past fMRI studies in humans have suggested location-specific (Guterstam et al., 2015; Hassabis et al., 2009), distance-specific (Morgan et al., 2011), and episode-specific spatial coding within a single environment (Brown et al., 2014), demonstrating remapping between different spatial environments in particular has been elusive because altering the environment changes the visual scenes and trajectories experienced by the subject. Here, we dealt with this issue by minimizing visual confounds inherent in navigation by instead having subjects retrieve spatial distances from specific environments during retrieval. Thus, our findings suggest that indeed the human hippocampus contains neural codes that differentiate specific spatial environments.

Second, our results provide support for a pattern separation/completion based account of memory disambiguation. Here, we probed the neural underpinnings of both successful and unsuccessful disambiguation during memory retrieval. Retrieval of Cities 1, 2, & 4, which were more easily learned and retrieved, were shown to involve orthogonal voxel patterns, as demonstrated by a searchlight classifier and voxel similarity analyses. City 3, however, which contained repeated landmarks but in a novel arrangement from Cities 1 & 2, did not exhibit neural characteristics consistent with remapping or pattern separation, i.e. higher within than between city similarity or above chance classification. Rather, when attempting to classify City 3’s correct retrieval trials, most trials were classified as City 1 or 2, although this effect was reduced for higher performing participants, suggesting that high performers exhibited more stable hippocampal patterns than low performers. Thus, one possible explanation for the poor performance on City 3 is that its neural patterns were insufficiently separated from those of Cities 1 & 2, resulting in a tendency to incorrectly pattern complete to stable representations of City 1 & 2. An alternative interpretation could be to attribute such errors to inhibition failure, for example, insufficient inhibition of City 1 & 2 representations by prefrontal cortex could lead to those being erroneously retrieved when attempting to retrieve City 3 (Anderson, 2004). The inhibition model, though, would not appear to predict low classification of correct City 3 trials and misclassification of these trials as Cities 1 & 2 trials since correct City 3 trials should involve trials in which traces from Cities 1 & 2 were successfully suppressed (Aron et al., 2014) and thus show no correlation with Cities 1 & 2 (see Figure 5). Furthermore, it is not clear how inhibition from higher cortical areas alone could lead to different patterns of suppression across the hippocampal subfields as prefrontal cortex projects primarily to subiculum and entorhinal cortex and not differentially to the CA fields, at least in non-human primates (Goldman-Rakic et al., 1984). Thus, our findings overall support the importance of pattern completion and separation, particularly in CA3/DG, to spatial remapping and appear less easily reconciled with an inhibition-based account.

A third important insight provided by our findings is a potential link between remapping-like mechanisms in humans, spatial learning, and rodent hippocampal remapping. The relationship between hippocampal remapping and behavior, however, remains unclear from the few studies to address both (Colgin et al., 2008). Part of the issue, as acknowledged in past such studies, is that it is difficult to assay whether a rat “knows” it is in a different environment or not, although dwell time and reversing direction may be important behavioral assays (Spiers et al., 2015). Jeffery et al., (2003) show that in a hippocampally dependent place-reward discrimination task, rodents perform only slightly worse after small environment modifications that induce global (∼85% of cells) remapping (Jeffery et al., 2003), suggesting that remapping can occur quickly and have little negative effect on performance. McHugh et al., 2007, in contrast, demonstrated that dentate gyrus NMDA knockout mice experienced less hippocampal remapping between contexts and less behavioral discrimination (McHugh et al., 2007), suggesting that remapping is important to behavior. In the current study, we assessed map drawing performance after each round of spatial exploration, which provided a more direct link to the formation of a cognitive map during navigation. We found that maps of Cities 1 & 2 were easily learned because information could be readily transferred between the two cities. Later, when assessing neural patterns during retrieval, Cities 1 & 2 were shown to have mutually orthogonal hippocampal voxel patterns. In contrast, City 3’s maps were less accurate and took more trials to acquire due to interference from Cities 1 & 2. During retrieval, performance was lower and hippocampal patterns were not orthogonal to those of Cities 1 & 2. However, we found that participants who performed better on City 3 did show voxel patterns that were more readily differentiated from the other cities. Thus, our findings from Cities 1 & 2 appear consistent with the results of Jeffrey et al. 2003 as we show remapping between Cities 1 & 2 despite the map acquisition data arguing for shared information between the two distinct representations. Our results for City 3, though, appear consistent with McHugh et al., 2007, with less remapping negatively impacting memory performance. Thus, we think our data provide a potentially important link between behavioral memory performance in humans and measures of remapping and pattern separation/completion.

Finally, our findings also provide important extensions and challenges regarding the function of the human hippocampal subfields in spatial context disambiguation. Specifically, when correctly retrieving spatial distances from two overlapping cities differing only in terms of two swapped stores, neural patterns were uncorrelated to each other and all other cities, an effect primarily shown in left CA3/DG. These findings support a role for CA3/DG in differentiation of competing spatial inputs, suggesting that this subfield in particular may be important for fine-grained discriminations amongst overlapping contexts as a part of a larger role in pattern separation/completion (Yassa and Stark, 2011). We also found that CA3/DG pattern remapping was (marginally) significant for City 4, suggesting that it differentiated the distinct city as well. These findings are consistent with the idea of CA3/DG as a universal pattern separator/completer, with failures linked to low performance on City 3 (Yassa and Stark, 2011). In contrast, pattern similarity in CA1 was higher when participants correctly judged spatial distances from a distinct city featuring novel landmarks and geometry compared to retrieval-induced patterns from the other cities. These findings appear somewhat inconsistent with models that suggest CA1 serves as a complement to CA3/DG in pattern separation/completion (Guzowski et al., 2004). Instead, our data appear to be more consistent with the emerging idea that CA1 may play a specific role in detection or representation of novelty (Duncan et al., 2012; Larkin et al., 2014), possibly acting as an important hub for integrating cortical input (Stokes et al., 2015; Remondes and Schuman, 2004). Although our study cannot provide specific insight beyond this speculation regarding the functional role of CA1, our findings suggest that its role in processing spatial contexts goes beyond a pattern separation/completion function defined by its position between CA3/DG and entorhinal cortex.

One potentially perplexing aspect of our findings is that we found similar degrees of remapping, and therefore putative neural pattern separation processes, for both the similar cities (Cities 1 & 2) and City 4, despite the fact that information content was significantly different for City 4 than Cities 1&2. Past studies, for example, have found that neural pattern differences may scale as a function of environment dissimilarity, with geometrically more distinct environments showing lower neural pattern similarity than more similar-shaped environments (Stokes et al., 2015; Leutgeb et al., 2005). In contrast to these two studies, though, we did not employ a continuous measure of environmental similarity and retrieval success hinged on being able to successfully maintain separate representations for the different environments but not necessarily tracking differences in the details of the different environments themselves. Thus, it is possible that our paradigm involved a more discrete form of pattern separation (Yassa and Stark, 2011) than would be needed in experiments involving a continuous change between environments. Consistent with the discrete pattern separation interpretation, studies of human episodic memory suggest that even overlapping episodes can be successfully decoded from multivariate patterns in the hippocampus, suggesting that retrieval of even very similar episodes can involve pattern separation-like processes (Chadwick et al., 2010). Thus, one possible interpretation of our results is that pattern separation in CA3/DG may not always scale with behavioral details and thus may be a more discrete process that depends on the exact demands required by encoding and retrieval.

In conclusion, while past studies in rodents suggest that hippocampal remapping might be one possible mechanism whereby memories are differentiated, no studies have demonstrated a comparable phenomenon in humans. Our findings thus provide several important insights about the human hippocampus: 1) the hippocampus contains unique codes about specific spatial environments, with these codes showing significant neural pattern similarity within the same spatial environment and low similarity between different spatial environments 2) when interference between spatial environments is high, pattern completion/separation processes may fail, resulting in difficulties discriminating competing cities 3) remapping, in humans at least, appears tightly linked with behavioral performance at discriminating different environments during retrieval 4) human CA3/DG appears to play a fairly ubiquitous role in pattern separation/completion during retrieval of specific spatial environments while CA1 plays a more specific role in representing features of novel environments. Together, our findings provide new insight into how the human hippocampal circuit processes competing spatial information.

1. 1. Anderson M

   (2003) Rethinking interference theory: executive control and the mechanisms of forgetting

   Journal of Memory and Language 49:415–445.

   https://doi.org/10.1016/j.jml.2003.08.006

   • Google Scholar
2. 1. Anderson MC

   (2004) Neural systems underlying the suppression of unwanted memories

   Science 303:232–235.

   https://doi.org/10.1126/science.1089504

   • Google Scholar
3. 1. Aron AR
   2. Robbins TW
   3. Poldrack RA

   (2014) Inhibition and the right inferior frontal cortex: one decade on

   Trends in Cognitive Sciences 18:177–185.

   https://doi.org/10.1016/j.tics.2013.12.003

   • Google Scholar
4. 1. Avants B
   2. Epstein C
   3. Grossman M
   4. Gee J

   (2008) Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain

   Medical Image Analysis 12:26–41.

   https://doi.org/10.1016/j.media.2007.06.004

   • Google Scholar
5. 1. Bakker A
   2. Kirwan CB
   3. Miller M
   4. Stark CEL

   (2008) Pattern separation in the human hippocampal CA3 and dentate gyrus

   Science 319:1640–1642.

   https://doi.org/10.1126/science.1152882

   • Google Scholar
6. 1. Bell AJ
   2. Sejnowski TJ

   (1995) An information-maximization approach to blind separation and blind deconvolution

   Neural Computation 7:1129–1159.

   https://doi.org/10.1162/neco.1995.7.6.1129

   • Google Scholar
7. 1. Brown TI
   2. Hasselmo ME
   3. Stern CE

   (2014) A high-resolution study of hippocampal and medial temporal lobe correlates of spatial context and prospective overlapping route memory

   Hippocampus 24:819–839.

   https://doi.org/10.1002/hipo.22273

   • Google Scholar
8. 1. Buzsáki G
   2. Moser EI

   (2013) Memory, navigation and theta rhythm in the hippocampal-entorhinal system

   Nature Neuroscience 16:130–138.

   https://doi.org/10.1038/nn.3304

   • Google Scholar
9. 1. Chadwick MJ
   2. Hassabis D
   3. Weiskopf N
   4. Maguire EA

   (2010) Decoding individual episodic memory traces in the human hippocampus

   Current Biology 20:544–547.

   https://doi.org/10.1016/j.cub.2010.01.053

   • Google Scholar
10. 1. Colgin LL
    2. Moser EI
    3. Moser M-B

    (2008) Understanding memory through hippocampal remapping

    Trends in Neurosciences 31:469–477.

    https://doi.org/10.1016/j.tins.2008.06.008

    • Google Scholar
11. 1. Copara MS
    2. Hassan AS
    3. Kyle CT
    4. Libby LA
    5. Ranganath C
    6. Ekstrom AD

    (2014) Complementary roles of human hippocampal subregions during retrieval of spatiotemporal context

    Journal of Neuroscience 34:6834–6842.

    https://doi.org/10.1523/JNEUROSCI.5341-13.2014

    • Google Scholar
12. 1. Davis T
    2. LaRocque KF
    3. Mumford JA
    4. Norman KA
    5. Wagner AD
    6. Poldrack RA

    (2014) What do differences between multi-voxel and univariate analysis mean? how subject-, voxel-, and trial-level variance impact fMRI analysis

    NeuroImage 97:271–283.

    https://doi.org/10.1016/j.neuroimage.2014.04.037

    • Google Scholar
13. Website

    1. Detre G

    (2006) The multi-voxel pattern analysis (mVPA) toolbox. in poster presented at the annual meeting of the organization for human brain mapping (florence, italy)

    http://www.csbmb.princeton.edu/mvpa
14. 1. Duncan K
    2. Ketz N
    3. Inati SJ
    4. Davachi L

    (2012) Evidence for area CA1 as a match/mismatch detector: a high-resolution fMRI study of the human hippocampus

    Hippocampus 22:389–398.

    https://doi.org/10.1002/hipo.20933

    • Google Scholar
15. 1. Ekstrom AD
    2. Bazih AJ
    3. Suthana NA
    4. Al-Hakim R
    5. Ogura K
    6. Zeineh M
    7. Burggren AC
    8. Bookheimer SY

    (2009) Advances in high-resolution imaging and computational unfolding of the human hippocampus

    NeuroImage 47:42–49.

    https://doi.org/10.1016/j.neuroimage.2009.03.017

    • Google Scholar
16. 1. Ekstrom AD
    2. Kahana MJ
    3. Caplan JB
    4. Fields TA
    5. Isham EA
    6. Newman EL
    7. Fried I

    (2003) Cellular networks underlying human spatial navigation

    Nature 425:184–188.

    https://doi.org/10.1038/nature01964

    • Google Scholar
17. 1. Etzel JA
    2. Zacks JM
    3. Braver TS

    (2013) Searchlight analysis: promise, pitfalls, and potential

    NeuroImage 78:261–269.

    https://doi.org/10.1016/j.neuroimage.2013.03.041

    • Google Scholar
18. 1. Friston KJ
    2. Frith CD
    3. Frackowiak RSJ
    4. Turner R

    (1995) Characterizing dynamic brain responses with fMRI: a multivariate approach

    NeuroImage 2:166–172.

    https://doi.org/10.1006/nimg.1995.1019

    • Google Scholar
19. 1. Goldman-Rakic PS
    2. Selemon LD
    3. Schwartz ML

    (1984) Dual pathways connecting the dorsolateral prefrontal cortex with the hippocampal formation and parahippocampal cortex in the rhesus monkey

    Neuroscience 12:719–743.

    https://doi.org/10.1016/0306-4522(84)90166-0

    • Google Scholar
20. 1. Guterstam A
    2. Björnsdotter M
    3. Gentile G
    4. Ehrsson HH

    (2015) Posterior cingulate cortex integrates the senses of self-location and body ownership

    Current Biology 25:1416–1425.

    https://doi.org/10.1016/j.cub.2015.03.059

    • Google Scholar
21. 1. Guzowski JF
    2. Knierim JJ
    3. Moser EI

    (2004) Ensemble dynamics of hippocampal regions CA3 and CA1

    Neuron 44:581–584.

    https://doi.org/10.1016/j.neuron.2004.11.003

    • Google Scholar
22. 1. Hassabis D
    2. Chu C
    3. Rees G
    4. Weiskopf N
    5. Molyneux PD
    6. Maguire EA

    (2009) Decoding neuronal ensembles in the human hippocampus

    Current Biology 19:546–554.

    https://doi.org/10.1016/j.cub.2009.02.033

    • Google Scholar
23. 1. Hill AJ

    (1978) First occurrence of hippocampal spatial firing in a new environment

    Experimental Neurology 62:282–297.

    https://doi.org/10.1016/0014-4886(78)90058-4

    • Google Scholar
24. 1. Hunsaker MR
    2. Kesner RP

    (2013) The operation of pattern separation and pattern completion processes associated with different attributes or domains of memory

    Neuroscience & Biobehavioral Reviews 37:36–58.

    https://doi.org/10.1016/j.neubiorev.2012.09.014

    • Google Scholar
25. 1. Jacobs J
    2. Weidemann CT
    3. Miller JF
    4. Solway A
    5. Burke JF
    6. Wei X-X
    7. Suthana N
    8. Sperling MR
    9. Sharan AD
    10. Fried I
    11. Kahana MJ

    (2013) Direct recordings of grid-like neuronal activity in human spatial navigation

    Nature Neuroscience 16:1188–1190.

    https://doi.org/10.1038/nn.3466

    • Google Scholar
26. 1. Jeffery KJ
    2. Gilbert A
    3. Burton S
    4. Strudwick A

    (2003) Preserved performance in a hippocampal-dependent spatial task despite complete place cell remapping

    Hippocampus 13:175–189.

    https://doi.org/10.1002/hipo.10047

    • Google Scholar
27. 1. Kentros C

    (1998) Abolition of long-term stability of new hippocampal place cell maps by NMDA receptor blockade

    Science 280:2121–2126.

    https://doi.org/10.1126/science.280.5372.2121

    • Google Scholar
28. 1. Kohonen T

    (1977)

    Associative Memory

    Associative Memory.

    • Google Scholar
29. 1. Kriegeskorte N
    2. Goebel R
    3. Bandettini P

    (2006) Information-based functional brain mapping

    Proceedings of the National Academy of Sciences of the United States of America 103:3863–3868.

    https://doi.org/10.1073/pnas.0600244103

    • Google Scholar
30. 1. Kyle CT
    2. Smuda DN
    3. Hassan AS
    4. Ekstrom AD

    (2015) Roles of human hippocampal subfields in retrieval of spatial and temporal context

    Behavioural Brain Research 278:549–558.

    https://doi.org/10.1016/j.bbr.2014.10.034

    • Google Scholar
31. 1. Larkin MC
    2. Lykken C
    3. Tye LD
    4. Wickelgren JG
    5. Frank LM

    (2014) Hippocampal output area CA1 broadcasts a generalized novelty signal during an object-place recognition task

    Hippocampus 24:773–783.

    https://doi.org/10.1002/hipo.22268

    • Google Scholar
32. 1. Leutgeb JK
    2. Leutgeb S
    3. Moser M-B
    4. Moser EI

    (2007) Pattern separation in the dentate gyrus and CA3 of the hippocampus

    Science 315:961–966.

    https://doi.org/10.1126/science.1135801

    • Google Scholar
33. 1. Leutgeb JK
    2. Leutgeb S
    3. Treves A
    4. Meyer R
    5. Barnes CA
    6. McNaughton BL
    7. Moser M-B
    8. Moser EI

    (2005) Progressive transformation of hippocampal neuronal representations in “morphed” environments

    Neuron 48:345–358.

    https://doi.org/10.1016/j.neuron.2005.09.007

    • Google Scholar
34. 1. Leutgeb S

    (2004) Distinct ensemble codes in hippocampal areas CA3 and CA1

    Science 305:1295–1298.

    https://doi.org/10.1126/science.1100265

    • Google Scholar
35. 1. Marr D

    (1971) Simple memory: a theory for archicortex

    Philosophical Transactions of the Royal Society B: Biological Sciences 262:23–81.

    https://doi.org/10.1098/rstb.1971.0078

    • Google Scholar
36. 1. McHugh TJ
    2. Jones MW
    3. Quinn JJ
    4. Balthasar N
    5. Coppari R
    6. Elmquist JK
    7. Lowell BB
    8. Fanselow MS
    9. Wilson MA
    10. Tonegawa S

    (2007) Dentate gyrus NMDA receptors mediate rapid pattern separation in the hippocampal network

    Science 317:94–99.

    https://doi.org/10.1126/science.1140263

    • Google Scholar
37. 1. Miller JF
    2. Neufang M
    3. Solway A
    4. Brandt A
    5. Trippel M
    6. Mader I
    7. Hefft S
    8. Merkow M
    9. Polyn SM
    10. Jacobs J
    11. Kahana MJ
    12. Schulze-Bonhage A

    (2013) Neural activity in human hippocampal formation reveals the spatial context of retrieved memories

    Science 342:1111–1114.

    https://doi.org/10.1126/science.1244056

    • Google Scholar
38. 1. Morgan LK
    2. MacEvoy SP
    3. Aguirre GK
    4. Epstein RA

    (2011) Distances between real-world locations are represented in the human hippocampus

    Journal of Neuroscience 31:1238–1245.

    https://doi.org/10.1523/JNEUROSCI.4667-10.2011

    • Google Scholar
39. 1. Morris RGM
    2. Anderson E
    3. Lynch GS
    4. Baudry M

    (1986) Selective impairment of learning and blockade of long-term potentiation by an n-methyl-d-aspartate receptor antagonist, AP5

    Nature 319:774–776.

    https://doi.org/10.1038/319774a0

    • Google Scholar
40. 1. Mourão-Miranda J
    2. Reynaud E
    3. McGlone F
    4. Calvert G
    5. Brammer M

    (2006) The impact of temporal compression and space selection on SVM analysis of single-subject and multi-subject fMRI data

    NeuroImage 33:1055–1065.

    https://doi.org/10.1016/j.neuroimage.2006.08.016

    • Google Scholar
41. 1. Muller RU
    2. Kubie JL

    (1987)

    The effects of changes in the environment on the spatial firing of hippocampal complex-spike cells

    The Journal of Neuroscience 7:1951–1968.

    • Google Scholar
42. 1. Mumford JA
    2. Turner BO
    3. Ashby FG
    4. Poldrack RA

    (2012) Deconvolving BOLD activation in event-related designs for multivoxel pattern classification analyses

    NeuroImage 59:2636–2643.

    https://doi.org/10.1016/j.neuroimage.2011.08.076

    • Google Scholar
43. 1. Newman EL
    2. Caplan JB
    3. Kirschen MP
    4. Korolev IO
    5. Sekuler R
    6. Kahana MJ

    (2007) Learning your way around town: how virtual taxicab drivers learn to use both layout and landmark information

    Cognition 104:231–253.

    https://doi.org/10.1016/j.cognition.2006.05.013

    • Google Scholar
44. 1. O'Keefe J
    2. Dostrovsky J

    (1971) The hippocampus as a spatial map. preliminary evidence from unit activity in the freely-moving rat

    Brain Research 34:171–175.

    https://doi.org/10.1016/0006-8993(71)90358-1

    • Google Scholar
45. Book

    1. O'Keefe J
    2. Nadel L

    (1978)

    The Hippocampus as a Cognitive Map

    Oxford: Clarendon Press.

    • Google Scholar
46. Book

    1. Redish AD

    (1999)

    Beyond the Cognitive Map

    Boston, MA: MIT Press.

    • Google Scholar
47. 1. Remondes M
    2. Schuman EM

    (2004) Role for a cortical input to hippocampal area CA1 in the consolidation of a long-term memory

    Nature 431:699–703.

    https://doi.org/10.1038/nature02965

    • Google Scholar
48. 1. Rissman J
    2. Gazzaley A
    3. D'Esposito M

    (2004) Measuring functional connectivity during distinct stages of a cognitive task

    NeuroImage 23:752–763.

    https://doi.org/10.1016/j.neuroimage.2004.06.035

    • Google Scholar
49. 1. Rolls ET

    (2010) Attractor networks

    Wiley Interdisciplinary Reviews: Cognitive Science 1:119–134.

    https://doi.org/10.1002/wcs.1

    • Google Scholar
50. 1. Shapiro ML
    2. Olton DS

    (1994)

    Hippocampal function and interference

    Memory Systems pp. 87–117.

    • Google Scholar
51. 1. Skaggs WE
    2. McNaughton BL

    (1998)

    Spatial firing properties of hippocampal CA1 populations in an environment containing two visually identical regions

    The Journal of Neuroscience 18:8455–8466.

    • Google Scholar
52. 1. Spiers HJ
    2. Burgess N
    3. Hartley T
    4. Vargha-Khadem F
    5. O'Keefe J

    (2001) Bilateral hippocampal pathology impairs topographical and episodic memory but not visual pattern matching

    Hippocampus 11:715–725.

    https://doi.org/10.1002/hipo.1087

    • Google Scholar
53. 1. Spiers HJ
    2. Hayman RMA
    3. Jovalekic A
    4. Marozzi E
    5. Jeffery KJ

    (2015) Place field repetition and purely local remapping in a multicompartment environment

    Cerebral Cortex 25:10–25.

    https://doi.org/10.1093/cercor/bht198

    • Google Scholar
54. 1. Stokes J
    2. Kyle C
    3. Ekstrom AD

    (2015) Complementary roles of human hippocampal subfields in differentiation and integration of spatial context

    Journal of Cognitive Neuroscience 27:546–559.

    https://doi.org/10.1162/jocn_a_00736

    • Google Scholar
55. 1. Thompson LT
    2. Best PJ

    (1990) Long-term stability of the place-field activity of single units recorded from the dorsal hippocampus of freely behaving rats

    Brain Research 509:299–308.

    https://doi.org/10.1016/0006-8993(90)90555-P

    • Google Scholar
56. 1. Wilson M
    2. McNaughton B

    (1993) Dynamics of the hippocampal ensemble code for space

    Science 261:1055–1058.

    https://doi.org/10.1126/science.8351520

    • Google Scholar
57. 1. Woo C-W
    2. Krishnan A
    3. Wager TD

    (2014) Cluster-extent based thresholding in fMRI analyses: pitfalls and recommendations

    NeuroImage 91:412–419.

    https://doi.org/10.1016/j.neuroimage.2013.12.058

    • Google Scholar
58. 1. Yassa MA
    2. Stark CEL

    (2011) Pattern separation in the hippocampus

    Trends in Neurosciences 34:515–525.

    https://doi.org/10.1016/j.tins.2011.06.006

    • Google Scholar
59. 1. Zhang H
    2. Zherdeva K
    3. Ekstrom AD

    (2014) Different “routes” to a cognitive map: dissociable forms of spatial knowledge derived from route and cartographic map learning

    Memory & Cognition 42:1106–1117.

    https://doi.org/10.3758/s13421-014-0418-x

    • Google Scholar

Author details

1. Colin T Kyle

   Center for Neuroscience, University of California, Davis, Davis, United States

   Contribution

   CTK, Implemented the experiment, collected and analyzed the data, and wrote the manuscript.

   Competing interests

   The authors declare that no competing interests exist.

2. Jared D Stokes

   1. Center for Neuroscience, University of California, Davis, Davis, United States
   2. Department of Psychology, University of California, Davis, Davis, United States

   Contribution

   JDS, Helped to write the manuscript.

   Competing interests

   The authors declare that no competing interests exist.

3. Jennifer S Lieberman

   Center for Neuroscience, University of California, Davis, Davis, United States

   Contribution

   JSL, Assisted with data collection.

   Competing interests

   The authors declare that no competing interests exist.

4. Abdul S Hassan

   Center for Neuroscience, University of California, Davis, Davis, United States

   Contribution

   ASH, Assisted with data analysis.

   Competing interests

   The authors declare that no competing interests exist.

5. Arne D Ekstrom

   1. Center for Neuroscience, University of California, Davis, Davis, United States
   2. Department of Psychology, University of California, Davis, Davis, United States

   Contribution

   ADE, Conceived and designed the experiment and wrote the manuscript.

   For correspondence

   adekstrom@ucdavis.edu

   Competing interests

   The authors declare that no competing interests exist.

• 2,429

  views

• 494

  downloads

• 73

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.