Cognitive mapping style relates to posterior–anterior hippocampal volume ratio

Abstract As London taxi drivers acquire “the knowledge” and develop a detailed cognitive map of London, their posterior hippocampi (pHPC) gradually increase in volume, reflecting an increasing pHPC/aHPC volume ratio. In the mnemonic domain, greater pHPC/aHPC volume ratios in young adults have been found to relate to better recollection ability, indicating that the balance between pHPC and aHPC volumes might be reflective of cross‐domain individual differences. Here, we examined participants' self‐reported use of cognitive map‐based navigational strategies in relation to their pHPC/aHPC hippocampal volume ratio. We find that greater reported cognitive map use was related to significantly greater posterior, relative to anterior, hippocampal volume in two separate samples of young adults. Further, greater reported cognitive map usage correlated with better performance on a self‐initiated navigation task. Together, these data help to advance our understanding of differences between aHPC and pHPC and the greater role of pHPC in spatial mapping.

Converging evidence has also been reported in the mnemonic domain. Greater pHPC/aHPC volume ratios were found to relate to better memory across diverse paradigms, including source memory judgments for scenes and recollection responses for pairs of words and pictures, suggesting a trade-off between the contributions of anterior and posterior hippocampal segments (Poppenk & Moscovitch, 2011). Specifically, the right pHPC/aHPC volume ratio showed a stronger relationship with memory than raw aHPC and pHPC volumes alone. These results suggest that individual differences in complex spatial and mnemonic abilities requiring a richly detailed representation may rely on a larger pHPC, which may entail a smaller aHPC.
Based on these previous findings, we chose to focus on the ratio of volumes as our target measure of interest. Given that differences in aHPC and pHPC pathology are found in Alzheimer's disease (AD) and healthy aging (Llado et al., 2018;Ta et al., 2012), a better understanding of the relationship between hippocampal long axis structure and navigational ability may also inform our understanding of pathology progression and protective factors.
Here, we examined the volume ratios of posterior relative to anterior hippocampal segments in two studies of younger adults who completed a navigational strategies questionnaire (NSQ) assessing their reliance on cognitive map strategies (NSQ published in Brunec, Bellana et al., 2018;see Appendix A). This questionnaire was designed to quantify the degree to which individuals rely on map-based strategies and spatial memories when navigating in the real world. Questions include self-reported strategies when navigating (i.e., "When planning a route, do you picture a map of your route or do you picture scenes of what you will see along the way?") and ratings of navigational ability (i.e., "Do you find it easy to read and use maps?"). In the present analyses, we sought to determine, in two independent datasets, if individuals who reported greater use of mental maps (higher mapping scores) had larger pHPC/aHPC volume ratios. We predicted that higher pHPC/aHPC volume ratios should relate to greater reliance on map-based navigational strategies, consistent with predictions based on previous studies and theories of specialization along the hippocampal long-axis (Poppenk et al., 2013;Strange, Witter, Lein, & Moser, 2014). This prediction is based on evidence of a trade-off between aHPC and pHPC function, observed in a range of episodic memory tasks (Poppenk & Moscovitch, 2011) and spatial abilities (Maguire et al., 2000), as well as a strong link between pHPC function and spatial behavior (Fanselow & Dong, 2010;Ryan, Lin, Ketcham, & Nadel, 2010  . The second study (i.e., the London dataset) included 25 participants (mean age 23.1 years, SD = 3.04; 13 female). One additional participant was excluded due to below chance performance on the inscan task. High-resolution T1-weighted structural scans were acquired using a 1.5 T Siemens Avanto MRI scanner at the Birkbeck-UCL Centre for Neuroimaging (TR = 12 ms, TE = 5.6 ms, 1 × 1 × 1 mm 3 resolution). The study was approved by the UCL research ethics committee and the Birkbeck-UCL Centre for Neuroimaging (BUCNI) ethics committee. Functional data from this experiment have previously been reported .
In the Toronto study, participants navigated freely by choosing their route between specified start and end points, in contrast to the London study, in which participants made navigational judgments at decision points but could not navigate off-course (Appendix B). In the Toronto study, participants were required to navigate using arrow keys, such that each keypress advanced their position by one step in the direction of their choice. Therefore, we were able to calculate navigational efficiency, defined as the difference in Euclidean distance to goal with each step. We found a significant relationship between mapping scores and navigational efficiency (r = .486, p = .004; Figure 1), supporting the notion that higher mapping scores relate to more efficient navigation. Navigational efficiency significantly correlated with mapping scores in the Toronto dataset. Mapping scores above 0 indicate a preference for map-based navigation, and scores below 0 indicate a preference for scene-based navigation. Navigational efficiency was calculated as the change in distance to goal with each step (key press). This measure was converted to z-scores to enable us to combine data across two separate experiments. The shaded area represents 95% confidence intervals around the fitted linear trendline Participants' hippocampi were extracted using FSL FIRST (Patenaude, Smith, Kennedy, & Jenkinson, 2011), after which they were manually segmented into anterior and posterior portions based on the location of the uncal apex (Poppenk & Moscovitch, 2011). Further following the method presented by Poppenk and Moscovitch (2011), hippocampal volume ratios were calculated by dividing posterior segment volumes by anterior segment volumes. Ratios above 1, therefore, indicate greater pHPC, relative to aHPC, and ratios below 1 indicate greater aHPC, relative to pHPC. There was no significant difference in volume ratios across the two datasets in either the left hemisphere (t(56) = −.075, p = .940) or the right hemisphere (t (56) = .369, p = .713). The mean volume ratio in the left hemisphere across both datasets was .984 (SD = 0.171), and the mean volume ratio in the right hemisphere was .930 (SD = 0.155). The mapping scores in the London study (M NSQ = 5.52, SD NSQ = 3.66) were significantly higher than those in the Toronto study (M NSQ = 3.09, SD NSQ = 4.53); t(56) = 2.19, p = 0.033. The left and right hippocampal volume ratios were then correlated to participants' mapping scores measured by the NSQ across both datasets (M NSQ = 4.14, SD NSQ = 4.32). In the combined dataset across both studies, there was a significant relationship between right hippocampal volume ratio and mapping (r = .397, p = .002; Figure 2c), but not between left hippocampal volume ratio and mapping (r = .180, p = .176; Figure 2b). For illustrative purposes, we also calculated the mean volume ratio (left and right hemispheres combined) and correlated it to mapping scores (Figure 2a). These results suggest higher pHPC/aHPC volume ratios relate to higher mapping scores, providing evidence that posterior hippocampal function relates to navigational strategy.
The same pattern of results broadly held when the data were split by study. In the Toronto dataset, there was a significant relationship between both right volume ratio and mapping (r = .352, p = .044), and left volume ratio and mapping (r = .410, p = .018). In the London dataset, there was a significant relationship between right volume ratio and mapping (r = .482, p = .015), but not between left volume ratio and mapping (r = −.049, p = .815; Figure 2). The relationships observed in the combined sample, therefore, broadly hold up in each of the individual samples with minor variations, though it is important to note that the sample sizes in each of the individual studies may be too small to draw strong conclusions about the relative differences between them.
To control for possible confounds, we ran a series of control analyses. We found no difference between male and female participants in mapping scores (t(56) = −1.366, p = .177) or volume ratios (right hippocampus: t(56) = 0.765, p = .447; left hippocampus: t (56) = 0.962, p = .340). There was no relationship between hippocampal volume ratio and Santa Barbara Sense of Direction Scale (Hegarty, Richardson, Montello, Lovelace, & Subbiah, 2002), designed to measure spatial ability (right hippocampus: r = −.097, p = .471; left hippocampus: r = −.034, p = .801), suggesting that our results are not related to general navigational ability but to map use specifically. In the Toronto dataset, we also found no significant relationship between hippocampal volume ratio and navigation efficiency (right hippocampus: r = .188, p = .294, left hippocampus: r = .251, p = .160), again supporting the specificity of the link between pHPC/aHPC volume ratios and navigational strategy, but not ability.
Last, to control for whole brain volume, we calculated a partial correlation predicting mapping scores from pHPC/aHPC volume ratio while controlling for whole-brain cerebrospinal fluid, white matter, and grey matter volume estimates. The partial correlation was significant for the right volume ratio (r = .447, p < .001), but not left volume ratio (r = .177, p = .196).
Together, these analyses suggest that pHPC/aHPC volume ratios, particularly in the right hemisphere, are related to greater reliance on cognitive maps. This relationship holds up even after controlling for grey matter, white matter, and CSF volumes, and appears to be specific to navigational strategy, but does not extend to in-task navigational ability. Individuals with larger posterior, relative to anterior, hippocampal volumes in the right hemisphere tended to rate their use of map-based navigational strategies more highly. Map-based spatial navigation requires an integrated, fine-grained spatial representation  and the use of flexible behavioral strategies when planning novel goal-directed routes (Wolbers & FIGURE 2 Correlations between (a) mean (left/right average) pHPC/aHPC, (b) left pHPC/aHPC, and (c) right pHPC/aHPC volume ratios and mapping scores. A volume ratio above 1 indicates a larger pHPC, relative to aHPC, and a volume ratio below 1 indicates a larger aHPC, relative to pHPC. The trendlines are plotted for the combined sample (both London and Toronto studies), but individual participants are represented by shapes corresponding to each of the two studies. The shaded areas represent 95% confidence intervals around the fitted linear trendlines Hegarty, 2010). Constructing a novel route within a learned environment shares similarities with episodic reconstruction, in that both involve the reinstatement of a broad episodic context and retrieval of individual details (Brunec, Moscovitch, & Barense, 2018). In line with existing evidence that recollective ability relates to larger hippocampal volume ratios (Poppenk & Moscovitch, 2011), the present results indicate convergence across mnemonic and spatial domains. This interpretation is consistent with recent theoretical views proposing that the pHPC supports fine-grained representations while the aHPC supports more coarse-grained representations Howard et al., 2014;Milivojevic & Doeller, 2013;Poppenk et al., 2013;Robin & Moscovitch, 2017;Sheldon & Levine, 2016). In recent work from our teams, we find functional neuroimaging evidence for the distinction between map-based and scene-based navigation, such that higher mapping scores relate to more variable voxelwise dynamics in pHPC , and more pronounced goal-distancecoding responses .
While our effects replicate across two independent samples, the magnitude of the correlations in both studies was moderate. This finding likely signifies that other factors mediate the relationship between hippocampal volume ratios and self-reported navigational strategies.
These other factors might include variations in the ability of participants to reflect accurately on their navigational styles and variance in navigational tendencies depending on the experience and familiarity with an environment. The cities where the two samples of participants resided also have very different configurations: Toronto has a highly regular grid-like structure and London does not. This difference in the environments experienced by participants over their lifetimes may also relate to a difference in navigational styles (Spiers & Maguire, 2007) and, therefore, the difference in mean mapping scores across the two samples. Future work is needed to relate individual differences in navigational abilities to differences in environmental configurations, especially since differences have been observed between different measures of space syntax and aHPC and pHPC activity . While we cannot infer causation based on these correlational data, evidence that pHPC/aHPC ratios increase with experience in London taxi drivers implies that as these specialized populations develop extremely proficient mapping abilities, their hippocampal volumes may change accordingly, although evidence suggests that change in pHPC may occur on a more rapid timescale than in aHPC (Maguire et al., 2000(Maguire et al., , 2006Woollett & Maguire, 2011;Woollett, Spiers, & Maguire, 2009). Whether more extensive training would lead to a trade-off between pHPC and aHPC volumes, and whether a similar mechanism might operate in the general population should be explored in future longitudinal studies of mapping abilities and changes in aHPC and pHPC volumes. Existing evidence suggests that recently, but not remotely, learned environments and routes necessarily require or activate the hippocampus (Hirshhorn, Grady, Rosenbaum, Winocur, & Moscovitch, 2012;Moscovitch et al., 2005;Patai et al., 2017;Rosenbaum et al., 2000;Rosenbaum, Ziegler, Winocur, Grady, & Moscovitch, 2004). Although the present data suggest that a larger pHPC/aHPC ratio is associated with implementing a map-based strategy, it is not clear whether it is a necessary condition for using cognitive maps effectively in remotely learned environments. Some evidence suggests that right hippocampal volume is predictive of navigational abilities (Nedelska et al., 2012;Schinazi et al., 2013), though a study has also reported a significant relationship between right aHPC volume and topographical memory (Hartley & Harlow, 2012). In a subset of the data reported here, we found that navigational efficiency was related to self-reported use of cognitive maps, but not directly to hippocampal volume ratios. This observation is consistent with prior evidence showing no link between navigational abilities and hippocampal volume in the general population (Maguire et al., 2003;Weisberg, Newcombe, & Chatterjee, 2018). The latter finding raises the possibility that increased pHPC volumes in taxi drivers reflect their spatial navigation strategy rather than ability alone. The inconsistencies in these results may stem from the differences in the metrics of navigational abilities being studied, which warrants further investigation. These results could have implications for understanding AD and mild cognitive impairment (MCI). As spatial disorientation is an early and common symptom of AD, the relationship between navigational strategy and the detection of pathological aging patterns needs to be explored in future work (Coughlan, Laczó, Hort, Minihane, & Hornberger, 2018). Recent research has found that atrophy of the pHPC in cases of MCI and AD is associated with tau-pathology, Aβ-pathology and declines in verbal and spatial memory (Lindberg et al., 2017;Llado et al., 2018), whereas nonpathological aging has generally been associated with mid-or anterior, but not posterior, volume reductions (Malykhin, Huang, Hrybouski, & Olsen, 2017;Rajah, Kromas, Han, & Pruessner, 2010;Ta et al., 2012). Thus, changes to pHPC/aHPC volume ratios could potentially serve as indicators of MCI or AD vulnerability, and accompany changes in spatial memory and navigation strategy.   . In this experiment, participants actively navigated along 12 long routes (2-10 km) with different degrees of familiarity and 3-4 turns. The participants constructed the routes with the experimenter ahead of the experiment. They were allowed to create routes anywhere within a 42 × 27 km region of Toronto. In the previously unpublished Toronto experiment, participants navigated in a much smaller area, constrained to the downtown University of Toronto campus (430 × 340 m). In this experiment, participants were required to complete a large number of short routes across campus, each containing at least two turns. In both experiments, because participants were navigating actively, we were able to calculate the decrease in Euclidean distance to goal with each step (each key press the participants made to move in the environment) as a measure of navigational efficiency. However, because of the structure of the routes and because the navigated area was rectangular in the latter experiment, the decrease in Euclidean distance to goal per step was higher than in the first experiment. To be able to combine data across the two experiments, we z-scored the values within each experiment and correlated the resulting z-scores to mapping scores.
The functional data from the London study were previously reported, along with more detail about the task . In this study, participants navigated along routes constructed from Google Street View images, using the same software as in the Toronto study.
Participants completed 8 familiar and 8 unfamiliar routes. Navigation in this task was not activeinstead, participants were passively guided between decision points and were required to make direction judgments when they reached junction points or when new goals were presented.
As the participants did not navigate actively and their responses did not affect the actual trajectory of the path, the calculation of a similar navigational efficiency measure as in the Toronto study was not possible.