Predictive Representations in Hippocampal and Prefrontal Hierarchies

Abstract

As we navigate the world we learn about associations among events, extract relational structures, and store them in memory. This relational knowledge, in turn, enables generalization, inference, and hierarchical planning. Here we investigated relational knowledge during spatial navigation as multiscale predictive representations in the brain. We hypothesized that these representations are organized at multiple scales along posterior-anterior hierarchies in prefrontal and hippocampal regions. To test this, we conducted model-based representational similarity analyses of neuroimaging data measured during virtual reality navigation of familiar and unfamiliar paths with realistically long distances. We tested the pattern similarity of each point-along each navigational path-to a weighted sum of its successor points within different temporal horizons. Predictive similarity was significantly higher for familiar paths. Overall, anterior PFC showed predictive horizons at the largest scales (~875m) and posterior hippocampus at the lowest (~25m), while the anterior hippocampus (~175m), pre-polar PFC, and orbitofrontal regions (~350m) were in between. These findings support the idea that predictive representations are maintained at higher scales of abstraction in the anterior PFC, and unfolded at lower scales by pre-polar PFC and hippocampal regions. This representational hierarchy can support generalization, hierarchical planning, and subgoals at multiple scales.

Predictive Maps in Rats and Humans for Spatial Navigation

“…Since the SR encodes the environment's transition structure, it is itself a model that can be leveraged for intuitive planning (Baram et al, 2018) or more explicit planning procedures such as a tree search -which may also provide a function for hippocampal replay (Mattar & Daw, 2018;Ida Momennejad, 2020;Ida Momennejad et al, 2018;Ólafsdóttir et al, 2017;Pfeiffer & Foster, 2013). Our work also adds to findings that the SR provides an account of hippocampal representations observed in rats and humans (Brunec & Momennejad, 2019;de Cothi & Barry, 2020;Garvert et al, 2017;Stachenfeld et al, 2017) by showing it also fits their spatial navigation behaviour in dynamic environments. We found less evidence for model-based agent behaviour in the rodent trajectories relative to the human behaviour, which may reflect less successful planning mechanisms or less propensity to use them.…”

“…The SR has been able to provide a good account of behaviour and hippocampal representations in humans (Bellmund et al, 2019;Brunec & Momennejad, 2019;Garvert et al, 2017; and rodents (de Cothi & Barry, 2020;Stachenfeld et al, 2017). Additionally, since the SR can be learnt using TD learning rules, it attributes a role for dopamine in forming sensory prediction errors (Gardner et al, 2018) -similar to what has been observed experimentally (Menegas et al, 2017;Sharpe et al, 2017;Takahashi et al, 2017).…”

Functionally distinct high and low theta oscillations in the human hippocampus

“…Many previous studies of hippocampal oscillations in rodents have focused on signals in the dorsal region, which is analogous to the posterior hippocampus in humans 11 , or on signals that are consistent across the length of the hippocampus 17 . However, a different line of work in humans [18][19][20] and animals [21][22][23] emphasized that there are substantial variations in function for neural activity recorded at different positions along the length of the hippocampus. This suggested to us that human hippocampal oscillations at different A-P positions could have distinct spectral and functional properties.…”

Learning Structures: Predictive Representations, Replay, and Generalization

“…Consistent with predictions, SR could account for fMRI pattern similarity in statistical learning [35] and learning non-spatial relational concepts [37] . A recent fMRI study has shown that during planned hierarchical navigation, predictive horizons of small to medium lengths are represented along the long axis of the hippocampus (with longer horizons in anterior hippocampus of humans, corresponding to ventral hippocampus in rodents) and the prefrontal cortex (PFC) hierarchy [40,41] . The largest horizons were represented in gradually more anterior regions of the rostral and orbital prefrontal cortex, corresponding to Brodmann areas 10 and 11 [40,41] (Figure 2).…”