Bats, brains and memory palaces: shared computations behind navigation and memory?
Humans are capable of remarkable feats of cognition. Some of us are able to remember extraordinary amounts of information by drawing on ‘memory palaces’, mental visualisations of where memories are stored. These abilities may seem uniquely human. However, recent research has highlighted how advanced memory techniques may draw upon the same neural computations used for spatial navigation across many species.
In a now-famous philosophy paper, Thomas Nagel invites us to ask what it is like to be a bat. He argues that the bat has experiences we, as humans, can never imagine: what is it like to fly, to echolocate? For Nagel, we are trapped in our own little sensory bubble, unable to truly imagine what it is like to be outside of our own physical frame.
While this may be true, we may have more in common with bats than one might think at a first glance. Many systems of the human brain are similar to those of other species via common ancestry: a bat may look very different to us, but it still has a visual cortex, an auditory cortex, and so forth.
One brain region which is well-preserved across species is the hippocampal-entorhinal system. The hippocampus is a strange seahorse-shaped horn of grey matter, in humans buried below other ‘neocortical’ systems. Since the mid-1900s, we have known that the hippocampus is instrumental in forming new memories: a famous case study from Brenda Milner documents how Patient HM was rendered unable to form new memories after bilateral damage to the hippocampal system.
The hippocampal formation is also involved in navigation. Navigation problems can be solved by many methods: the way we move around our home kitchen, for example, is very different to how we would explore the open seas. Nonetheless, cell populations in the hippocampus and adjacent entorhinal cortex ‘track us’, releasing bursts of energy (‘action potentials’) when we move across specific locations. This is true both in bats and humans: when you move across your kitchen floor, a place cell might fire only when you are next to the door – for a bat, a place cell might fire when it nears the cave entrance. This is solid evidence that something navigation-related is being tracked by your hippocampal formation.
In the age of modern neural networks, it is easy to imagine the brain as a big mess of cells which, with enough training, learn to handle the problems we throw at it. Some people, indeed, argue that navigation and memory can be solved this way, and we over-interpret single cell results (Luo et al., 2024).
Despite this complexity, the brains of different species appear to converge on the same solutions within the hippocampal formation. Primates, bats and rodents all appear to use grid cells, neurons which typically fire in a hexagonal pattern, to navigate. Across species, this extends beyond spatial navigation into more complicated mnemonic functions: specialised cells map out specific pitches in rodent brains, while human neural patterns track the relationship between conceptual information.
So, to recap: across species, the hippocampal formation responds selectively to information related to navigation and memory. Across species, cell populations appear to respond in similar ways. This suggests that there is something in common between the neural mechanisms used for navigation and memory.
To bring all this under one umbrella, researchers have theorised that the hippocampal-entorhinal system carries out the same computations, across domains and species. One way to approach this has been to build up mathematical models which can explain both neural firing patterns and animal behaviour. These models produce predictions for future studies, which leads to a virtuous cycle of science. For example, models of grid cells based on attractor networks led to the remarkable prediction that grid cells can be described as a torus, a donut-like projection of neural data into a lower-dimensional space. This was duly discovered by a team led by Nobel prize-winners May-Britt and Edvard Moser.
As mentioned above, the computations that lead to grid cells appeared to be shared both across species and across domains. When we navigate a social network or other ‘non-spatial’ memory spaces, we recruit the hippocampal formation in a similar way to that when we navigate real-world spaces. This raises an intriguing question: can we use models to explain both space and memory?
In a Nature paper from last year (Chandra et al., 2025), researchers attempted to do just that. A team led by Ila Fiete (MIT) imagined the grid cell attractor network as a spatial ‘scaffold’ to which we can attach memories. The same processes, we might think, could describe the distance between rooms in a house or move between your memories of different subjects. In extreme cases, they argue, super-learners could exploit the spatial specialisation of these systems and anchor their memories to specific locations – just as we do when using memory palaces.
This is, of course, theoretical work which leaves as many questions open as it answers. We know a lot about how the brain handles spatial navigation, but it is not clear if these processes can always be easily recycled into navigation of memory. In a physical space, movement is continuous and well-defined; is this really true for memory? What would it mean to move between my knowledge of algebra and my knowledge of literature? Nonetheless, mathematical models provide a computational bedrock for further study which will allow us to probe if, and how, well-understood features of spatial navigation can be ‘borrowed’ in service of memory. My own PhD work moves in that direction, showing that the human entorhinal cortex processes non-spatial ‘action’ in the form of mathematical operations and eye movements.
All in all, these computational models provide an exciting glimpse at how navigational machinery may be recycled for memory. From an evolutionary perspective, this may help us understand how early humans evolved complex memory systems which enable the advanced communication characteristic of our species. Moreover, it may suggest that high-performing memorisers who use memory palaces are in fact drawing on an archaic link between space and memory.
I will finish by returning to bats, who have similar hippocampal machinery to us. We might see and feel the world differently from a bat – certainly very few of us use echolocation – but the same gears and cranks which allow a bat to fly around a cave may help us to build societies and remember, well, quite a lot of things.
Chandra, S., Sharma, S., Chaudhuri, R., & Fiete, I. (2025). Episodic and associative memory from spatial scaffolds in the hippocampus. Nature, 638(8051), 739–751. https://doi.org/10.1038/s41586-024-08392-y
Gardner, R. J., Hermansen, E., Pachitariu, M., Burak, Y., Baas, N. A., Dunn, B. A., Moser, M.-B., & Moser, E. I. (2022). Toroidal topology of population activity in grid cells. Nature, 602(7895), 123–128. https://doi.org/10.1038/s41586-021-04268-7
Luo, X., Mok, R. M., & Love, B. C. (2024). The inevitability and superfluousness of cell types in spatial cognition. eLife, 13. https://doi.org/10.7554/eLife.99047.1
Sounds like something mankind’s always known but science had to figure out for itself… Well, when everyone has already jumped to a conclusion and you follow in tiny, painstaking steps, you learn more about the road there. So as much as I’m often annoyed by watching scientists waste time in the dark, digesting data like blind, deaf, primitive bacteria with no perception except through their guts, with no knowledge about the world around them except what they can conclude from their food, I put up with it as a necessary evil on our way to precise, detailed knowledge enabling real technological progress which drowns my lazy butt in luxury. Even if opening your eyes and looking out of the window from time to time may accelerate the process and save you decades of research.
Spiritual journeys, spiritual worlds where ghosts of our ancestors live, sharing their wisdom with shamans with the exceptionally rare magic power of not being total morons, fairy worlds, fantasy worlds, haunted houses with an architecture vaguely reflecting the architecture of the brain, magic woods, magic cities, labyrinths the Matrix or spaceships filled with people sleeping in hibernation pods like neurons in a network… The echoes from our inner sonar howl and thunder all around us, providing you with data like a natural MRI.
The more you focus on a subject, the more resources your brain mobilizes to deal with it: You put energy on your inner sonar, call out loud to get answers. Then you shut up and listen, so you get your best ideas under the shower, after hearing the whispering echoes that you couldn’t hear while your inner king was roaring. The brain is a state. That’s how it works. Because that’s how most, if not all, networks work.
And now my inner VR generator shows me a group of neurologists studying a brain scan, till one of them shouts: “I found him! Here’s Frodo! Right over there, between Narnia and Neverland!” It’s not like that, it rather seems there are a lot of worlds in there mixed like lego bricks, a certain frequency and volume of your inner sonar activates the red bricks, another the yellow bricks, they close their circuits and form the VR world they were made for, but they still use all the other bricks and the network they form – your mind is a state, and it doesn’t matter if you are ruled by a coalition of red, yellow and green bricks or black and red bricks, if the neural nation below stays a mess and is incapable of supporting it, because it’s strangulated by semi-visible subconscious networks too stable and powerful to allow necessary change.
But if the hierarchy inside our brain continues outside as a social hierarchy, it might continue downside. Which means, just calming and listening to the voices within lets you travel to new worlds, just like shrinking would let you travel to the universe of cockroaches, then that of bacteria, more strange to us than Mars or Venus. How deep does it go? If you were really, really silent, if you had enough patience, could you understand the whispers of your atoms, instead of just drawing energy from their screams?
A star is far away, a grain of dust is small, because they are silent. Your neighbor is farther from you than your grandma on the phone, because there is a wall separating you from him and a wormhole enabling closeness to her. Time and space are shaped by communication, interaction, data travel, and they form intertwined networks, parallel universes embedded inside each other, threads, webs, labyrinths that bend and distort each other even if they are not aware of each other, and the only communication that links them all is the domino way – make them all move at the same time, make them pass on a signal till it reaches someone, something, that reacts to it. That’s Gotham City, Batman. Listen to it, look at it, taste it, smell it. It’s everywhere, inside and outside of you, and you are a part of it.
The world is a wave computer trying to solve an impossible equation: 1=0. It’s a bubble creating smaller bubbles testing different calculations, varying them, experimenting with possible solutions of the unsolvable, and any bubble that stays stable for some time is considered a success and is tested in a bigger, more complex bubble, and it all goes on till the bubble pops. Mathematics as usual. As our brain is just another bubble like any other, you kind of live inside a much larger brain around you. Details differ, and that’s where neurology steps in. But it may help to understand the general rules, and the only people with the superpower of not perceiving the rules at work are the dead and the unconscious.
If one takes this connection between the sense of place and memory seriously, one would also have to attribute a kind of sensory quality to spatial intelligence. The inner sensorium for place and movement anchors in this image the remembered in the real world, i.e. in our inner image of the real world.