In 1948, Edward C. Tolman first proposed the idea of a cognitive map, a mental representation of the spatial environment (or, put more simply, the ability to visualise an area within the mind’s eye).
Many scientists at the time believed that behaviour was caused by an individual responding to stimuli. This was commonly known as the stimulus response theory. Tolman disagreed with this idea – he believed that the cause of behaviour was more complex. His experimental work involved rodents, which he placed in a maze with food inside. The food was always placed in the same part of the maze but the route to reach the food was altered. If he was correct, then the rodents should be able to gather information about their environments and use this to build a cognitive map. Their cognitive map would therefore enable them to find the food (even if the layout of the maze changed), by using the knowledge of the position of the food relative to their starting position in their cognitive map.
The maze consisted of a main chamber connected to multiple corridors or ‘arms’. In an initial training phase the rodent was placed in a maze which only had one ‘arm’, into which the animal had to turn right in order to enter the food chamber (see diagram 1 above). In the food chamber was a food reward. This conditioned the animals to always turn right at an intersection in order to obtain food. Tolman predicted that this training period should allow the animals to learn where the food was in relation to the main chamber. However subscribers to the stimulus response theory might have predicted that the training period would only teach the animal to turn right at an intersection to find food. After training, the rodents entered the experimental phase: they were placed in a maze (diagram 2) with a range of arms, and the original arm no longer led to food (oh rats!).
Despite the change to the maze, the rats managed to choose the shortest way to the reward. Therefore although the rodent had learnt to turn right towards food, it still successfully navigated the new maze to find its reward. Tolman interpreted this as evidence for the rat knowing the general direction it should aim for, so theorised the rat possessed a mental map of the maze. This research was criticised because some scientists believe that this experiment failed to rule out simpler explanations for these results. Despite this, people did not give up on the concept of a cognitive map. Since then, many ideas of positioning systems in the brain have been explored. In 2014 the Nobel Prize in Physiology and Medicine was awarded to John O’Keefe, May-Britt Moser and Edvard I. Moser for the discovery of the cells that constitute this positioning system; these aid in building a mental map that is then stored in our memory.
Before explaining their fascinating discoveries, I shall take a moment to explain the reason for this inaugural guest article. Being a Master’s student at Uppsala University I had the pleasure of attending a lecture by both May-Britt and Edvard Moser. Coming to stand in the queue at 8am on a Saturday (obviously after a night out) for an hour and a half was definitely worth the experience of listening to these highly inspiring people, even if I took a couple of hours to realise they were actually Nobel laureates. (I tried taking a selfie with May-Britt Moser but I didn’t get to the front in time).
Getting back on-topic, how do animals perceive space? The hippocampus is part of the brain located in the medial temporal lobe. The hippocampus contains place cells (pyramidal neurons of the hippocampus), which provide the animal with a continuously updated and dynamic representation of space and its location in that space (sounds a bit Doctor Who, I know). O’Keefe (1976) showed these neurons fired whenever an animal was in a certain location (known as the place field). By manipulating the environment – for example turning the lights on and off – O’Keefe determined that this firing was determined purely by the position of the animal relative to the environment, rather than sensory stimulus, motor behaviour or incentive factors. Different place cells will have different place fields, and there is no apparent arrangement of these cells in relation to each other. An animation of different place cells firing as an animal moves through different place fields can be seen here.
More distinct firing regions were found in another region, also located in the medial temporal lobe in the enthorhinal cortex (separating the hippocampus and the neocortex). Experimental evidence showed that cells in this area show finely-tuned spatial firing, but each cell has multiple firing locations. In simpler terms this means that when an animal enters a specific part of its environment only a very specific set of cells will fire; however, each cell may fire in multiple locations. When all this information is integrated, each small part of its environment will have a specific fingerprint, or grid of firing cells. This indicates that they might be involved in spatial navigation. Each grid formed by the firing of the grid cells can be characterised by spatial separation between points in the enclosure at which the grid cells fire, as well as relative orientation towards the external reference axis – for example the walls of the enclosure. This pinpoints a very precise navigation system.
When an animal is released into an enclosure it wanders away from the starting position and performs path integration: this is akin to a very complex mathematical calculation, which uses the distance and direction moved to estimate where the start point was. In doing this the animal manages to roughly keep track of the changing position. Research suggests that even ants can carry out path integration – who knew ants were expert mathematicians? Path integration initially determines the firing activity of the grid and place cells. Fuhs & Touretzky (2006) proposed that the MEC (medial enthorhinal cortex) is organised so that the activity of the neighbouring cells is similar. This means that the activation of the grid cells is similar for the cells close to each other, and the more physically separated the cells are, the more different the pattern of their activation is (the cells fire at different places in the area). The firing of each cell is unaffected by ‘noise’ (interference) present far away in the tissue. During the lecture Edvard Moser called it a network of Mexican hat connectivity, because cells close to each other have high activity. Further away from these cells the firing activity decreases (at a particular time). Thus, on a graph of cell activity, the pattern formed has a structure of a perfect sombrero!
McNaughton et al (2006) argued with this statement, stating that cells in the MEC are randomly distributed, and there is no pattern of connection between them. Instead each cell has the same representation of space – the activity of different cells is random and independent of each other. Even though the two models suggest a different organisation, they both propose that the system is influenced by speed of motion, i.e. when an animal moves more quickly there is a higher cell firing rate.
As the animal starts exploring a new environment it takes several minutes for the system to reach a so-called stable state, meaning it takes a few minutes for the cognitive map to be organised. What is more, the system is plastic , so the firing pattern may change over time as the area is explored. Thus, the natural question at that point would be: can a mental map be remembered? When does the information about the area fade away?
Tsao (2013) found that in the rat LEC (lateral enthorhinal cortex) there are cells that fire when an object is placed in the enclosure, but there are also cells that fire even when the object is removed (called trace cells). These memories can last for up to 11 days. During her lecture May-Britt Moser talked about HM, a patient who had a damaged hippocampus, which impaired his ability to form memories. HM was therefore incapable of spatial navigation. Every one of us has experience of using space as a framework for storing memories – we remember things better when we associate them with a specific location. May-Britt Moser informally tested this on her daughter, who used the location of different rooms in the house to remember the map of South America. Consolidation of hippocampal memory traces occurs over a period of inactivity, such as sleeping or eating. Experimental evidence shows that coactivation of grid cells in the hippocampus continues while a rodent is sleeping, so enables information consolidation. Therefore this suggests how a cognitive map can be remembered.
What is equally interesting is that the hippocampus also receives olfactory information from the LEC – odours are known to be very important cues for information retrieval. What scents remind you of particular events? Some evidence for this was supplied by a mouse called Emma, who was trained to go to one place in an enclosure when smelling chocolate and another place when smelling banana. (Sweet). Researchers found that it was in fact the cells in the hippocampus that responded to Emma sniffing chocolate. When Emma had been trained to go to a particular area after smelling chocolate, her hippocampal cells fired whenever she was in that particular location, whether chocolate was present or not. The same results could be produced when smelling banana. Hence, spatial information about odours changed the wiring of the neurons in the hippocampus. Many different external cues from the environment appear therefore to be intertwined.
Not much is known about the origin of the spatial representation system during development, nor how this system is able to filter out errors. A lot remains to be done. If I got you interested in this research follow May-Britt Moser and Edvard Moser, both at NTNU’s Kavli Institute for Systems Neuroscience.