Portland
Neuroscience Book Club

Book archive / Book report

Dark and Magical Places: The Neuroscience of Navigation

Christopher Kemp · Report shared July 26, 2023

About this bookDownload text

Dark and Magical Places. The Neuroscience of How we Navigate

By Christopher Kemp

The author states that this book is an attempt to understand the shortcomings of his own brain. He is a self-professed poor navigator, and is fascinated with those who aren’t. In this book, he delves into the neuroscience of how our brains navigate the world.

Julius Caesar Arantius first described and named the hippocampus in the 1560’s, but we only began to understand its role in navigation in the 1950’s, with patient HM, who had both hippocampi and some surrounding cortex removed to treat his severe epilepsy. As a result, while his short term memory was intact (because of an intact prefrontal cortex), his long term memory was severely affected. His episodic memory (personal memories of his life) was erased for several years before the surgery, though he retained some semantic memory (factual memories). In addition, he was unable to form new explicit memories, but could still learn motor tasks, though without conscious recollection of learning them before. All this hinted at the function of the hippocampus and surrounding areas. The hippocampus does seem to be involved in storing memories, though over time, they seem to be transferred elsewhere, but the hippocampus does seem to be involved in both storing and retrieving highly detailed memories. It is also involved in binding together different aspects of an experience so it can be recalled as a whole cohesive memory. It accounts for only 0.5% of the brain’s weight, but is a critical component of memory. It is also central for spatial navigation. There seems to be some bilateral differentiation as well, with the right hippocampus playing a larger role in navigation, and the left playing a larger role in episodic memory.

Note: some researchers think that the hippocampus is a general sequence generator and does not actually represent space or time as such1. The brain has no sensors for directly detecting distance or duration, and so neither space nor time can cause a change in the brain. They point out that neuronal assemblies that fire in a way that correlates with the passage of time also tend to correlate with other frames like egocentric actions, objects, and learned routes. They propose that the hippocampal system may simply point to items stored in the neocortex in the same order they were experienced during learning.

Other areas also play a role in different aspects of navigation including the parahippocampal place area, the occipital place area, and the retrosplenial region.

The author details many stories of hikers who have gotten lost in the woods after stepping only briefly off a trail. Amanda Eller stepped off a trail in a dense forest in Hawaii and was lost for 17 days. She was found only 7 miles from the parking lot where her car was parked, and only a mile from the coast. She faced a lot of public ridicule and criticism.

Edward Tolman pioneered the use of mazes in the 1930’s to study not only spatial navigation in animal models, but learning and memory generally. He discovered that when presented with a new obstacle in a familiar maze, rats will find alternative routes, showing that they weren’t just memorizing a series of turns as behaviorists through at the time.

Nora Newcombe at Temple University in Philadelphia is studying spatial navigation in humans using VR environments. She has found that people fall into three roughly equal sized categories: good navigators who form mental maps, ok navigators that can learn routes but don’t integrate them together, and people who can’t seem to build a mental map at all.

There are two general strategies for learning to navigate an environment, a hippocampal-dependent one, and another one that relies on the caudate nucleus. Veronique Bohbot at McGill University in Montreal has imaged human subjects in an fMRI while they navigate a virtual radial arm maze surrounded by landmarks. She has found that when asked to navigate to a spot in the maze after subjects have familiarized themselves with it, some subjects used a spatial strategy using the landmarks (increase in hippocampal activity), and some used a non-spatial strategy, like counting the arms of the maze from the starting point (hippocampus doesn’t increase in activity, but caudate nucleus does). The caudate nucleus is part of the striatum and is involved in associative learning and a wide range of other functions and is sometimes called part of the brain’s “autopilot” network. As the subjects got more familiar with the maze, more and more of them started using the caudate nucleus to do the task to unburden the hippocampus. This is normal and is seen in animal models as well. As we learn an environment, we stop using a costly spatial strategy, and relegate to autopilot when we’ve become familiar enough with it. The problem is, if you only use the autopilot strategy, you don’t develop a flexible cognitive map, if you deviate from the path at all, you’re lost.

In the 1970’s, John O’Keefe at University College London discovered place cells. Using electrode recordings from a rat navigating a maze, O’Keefe found that some cells fired only when the rat was in a particular location in the maze. They fired at rest at a rate of roughly once every ten seconds, but when activated, fired at up to twenty times per second, acting like a locating beacon. They’ve been found primarily in the hippocampus, and are a type of pyramidal cell. Later, O’Keefe also discovered misplace cells that fire when an animal travels to a location where an expected object is suddenly absent. Place cells may be part of a broader neuronal network dealing with the sequences of actions an animal is taking and what to expect once there, they may be general knowledge representers. But how do place cells work? There are relatively few of them, and yet they can encode large stretches of space and lots of different locations. There are approximately 200,000 place cells, and about 10% of them are active at a given time. As you move around an environment, a different 10% becomes active, forming a continuous representation informed by which cells are active together at a given location. However, we don’t yet know the pattern to predict which place cells will fire together. There’s no topographical relationship between two place cells (two adjacent place cells don’t represent adjacent spatial locations). Additionally, they may fire as part of the same ensemble, or not. Note: each place field ranges from about one to 15 body lengths in rodents but are thought to enlarge when learning a larger space. Place cells also display sharp-wave ripples. These are sharp spikes lasting only 100 ms or so, but they involve up to 100,000 neurons tightly synchronized together. They have historically been studied most during sleep and are thought to be involved in long-term learning because disrupting them is as detrimental to memory as removing the hippocampus in rats. Place cells produce sharp wave ripples, which is thought to represent a reexamination of recent spatial memories, and about 70% of sampled prefrontal cortex neurons changed activity in response to the ripples, suggesting they have brain-wide impacts. As we navigate, place cells seem to rapidly switch between representing our current location and an array of possible futures at a rate of about 8 times per second. There is also evidence of place cells in human epilepsy patients fitted with electrodes navigating a VR environment.

How are place cells informed? Edvard and May-Britt Moser discovered grid cells in 2005. These cells have a strictly hexagonal firing pattern that repeat across an entire area where the animal is walking around. They are in the entorhinal cortex, the hub through which information passes both into and out of the hippocampus. Unlike place cells, their firing fields sit in a predictable pattern. They can fire in multiple locations, but if two grid cells fire together in one environment, they’re likely to fire together in others. As an animal learns an environment, its entire suite of grid cells is eventually recruited. They seem to encode location, distance, and direction. It is a general measurement system that feeds into the hippocampus and can then be applied onto a map.

In 1984, James B. Ranck Jr. at SUNY in Brooklyn discovered head direction cells. Using recording electrodes in rats, he found cells that preferentially fired when the rat was facing a particular direction. Again like a beacon, they begin to fire as the animal turns in it’s preferred firing direction, and increase in firing rate until animal is facing directly that way. They seem to be scattered throughout the brain and are reliant on the vestibular system in order to function normally.

One of the largest existing datasets of human navigation is players of the game Sea Hero Quest. It was designed as an early diagnostic marker of dementia in 2016, but over 4 million diverse people from 195 countries have played the game. The following main findings have emerged. We continuously get worse at spatial navigation as we age, starting at about age 20. The distribution of spatial skills is not normal, but skewed towards poor navigators. The most important determinant of navigation ability is GDP of the player’s country. Finally, the often reported sex difference between men and women doesn’t exist in countries where men and women are treated equally. Interestingly, participants who live in high-entropy (non-grid) cities score better as well, indicating that orderly grid cities don’t challenge us spatially and may result in less spatial practice and thus skill.

Sex differences in spatial navigation: there is a well replicated sex difference between men and women with men scoring better on mental rotation tasks and women scoring better on object location. Studies have also shown that men use more spatial strategies to navigate an environment than women. These differences have often been attributed to division of labor in our evolutionary history, despite evidence for this theory being very weak. New evidence is showing these differences may simply be a result of gender differences in activity patterns, which are influenced by cultural expectation and opportunities, which we know have differed dramatically between men and women. In a 2007 study, women who were made to play an action video game for ten hours scored more similarly to men on a mental rotation task. We know from Sea Hero Quest that women who are not given an opportunity to travel freely or drive tend to have poor navigation skills, but women in other more gender equal countries don’t show this deficit.

Lukas Kunz at the University of Freiburg noticed that place cells, grid cells, and head direction cells are encode allocentric (external space) but how does the brain represent egocentric space? Kunz recorded from human epilepsy patients and found that these is a population of cells that fire according to your position relative to an anchor, like your home. Particular cells fire when your home is on your left a few meters, others when it is to the right, and others at different distances.

Evolutionary history of navigation: During our evolution into modern humans, our parietal lobes expanded, which enhanced our visual and spatial skills, but this may have been exapted into more complex functions like numerosity and navigation. It is involved in tracking the location of the body and its parts. The author describes an interesting case study of a man with an overgrowth in a parieto-occipital region, and when asked the location of the hospital he was in, the location was different each time, but stated with confidence.

Dead reckoning is another factor in our navigation. The desert ant, for example, which zig zags away from its home and shoots a straight path back to it does this by keeping an internal record of its accumulating step count. Bees use optic flow data for the same function. Importantly, this must be integrated with outside information, like the position of the sun to navigate well. Dead reckoning can be very error prone when used alone, but is an important part of path integration (continual position updating as you move through it) for many animals. In a 2009 study, Jan Souman at the Max Planck Institute for Cybernetics, tested the ability of people to walk in a straight line through a forest. People were good at this on sunny days, but on cloudy days were disastrously bad. After 4 hours of walking, most of them did not end up more than a mile from their starting location. Note that his sample size was only 6 people. Other people can do this task, like the Tsimane, detailed below.

Neuroscientist Giuseppe Iaria is studying people who are unable to make cognitive maps. He described a developmental disorder in which people are so severely spatially challenged that they get lost within their own neighborhoods, but have no obvious brain damage or any other cognitive complaint. Iaria estimates that 1-2% of the normal population might have Developmental Topographical Disorientation. These people may have different patterns of activation in brain areas critical for navigation. One participant was found to have a retrosplenial cortex (an information aggregating region in the parietal lobe that is densely connected with the hippocampus) that doesn’t adapt to familiar scenes like control subjects. The structural connections seem normal, but functionally, the region is almost totally disconnected. There seems to be a genetic component, with one third of DTD participants reporting a first degree family member who has similar spatial problems. When Iaria recruited family members, even those that didn’t report spatial difficulty were likely to fail the spatial tests in the lab. We just live in a world where you can live a pretty normal life without navigating spatially.

Ben Trumble studies the Tsimane, a group of isolated indigenous people living in the Amazon Basin. They navigate a large forest area in low light, without many navigational cues. The river they live along is serpentine. Yet they can point in the directions of far away locations, like a town many miles away, with pretty good accuracy. Men and women, young and old, all perform equally well.

Kali Rimfield at King’s College London has studied 1300 twin pairs and found that spatial tests including mental rotation, map reading, route memory, and navigation have a heritable component. About 84% of the variance in our spatial abilities is explained by genetics. Furthermore, the different components of spatial ability cluster together, indicating that they are served by a common ability.

In the final chapter, the author illustrates our reliance on GPS and how that can get us into trouble. Noel Santillan was in Iceland and misspelled a town name by one letter, and followed his GPS for 4 hours until he ended up in an isolated fishing village. People have also died in GPS related incidents, for example driving off the end of a bridge that had become defunct years ago, or became stranded. Human imaging shows that when people are given a set of directions to use to navigate an environment, the hippocampus is not engaged. We could perhaps redesign GPS to give us directions that are more landmark based and conceptualized, perhaps even integrated with our personal data (e.g. turn right at that bookstore you visited a few weeks ago). In a study testing this, subjects remembered landmarks better and navigated more accurately than subjects who got traditional GPS directions.

Appendix:

What to do if you’re lost:

Note: not sure how practical these are?

How to become a better navigator:

References:

1. Buzsáki, G. & Tingley, D. Space and Time: The Hippocampus as a Sequence Generator. Trends Cogn Sci 22, 853-869 (2018).

Club notes from the original shared report. Formatting has been standardized; the substantive notes are preserved. Dates refer to when the report was shared.