In rodents, spatial memory and navigation involves at least four different types of neuron, located in and around the hippocampus, whose properties are well characterised. Place cells, for example, fire selectively when the animal is in a specific location; grid cells fire periodically as it traverses a space; head direction cells increase their activity when it faces a particular way; and border cells, whose existence was reported just a few months ago, fire when it is close to anything that resembles a border. Yet, despite the relative wealth of cellular data, the neural mechanisms underlying spatial navigation are still unclear.

We do not yet know what specific features of the environment the place cells are sensitive to, or how ensembles of neurons in the hippocampus encode spatial memories. In the visual system, the topographical features of an image are mapped directly, so that adjacent parts of the image are encoded by neighbouring cells in the retina and early visual processing areas of the brain. In the hippocampus, however, there is no obvious relationship between place cell activity and an organism's precise location within its environment. Location is therefore thought to be encoded randomly in large populations of cells distributed sparsely and widely throughout the hippocampus.

For the new study, Demis Hassabis of the Wellcome Trust Centre for Neuroimaging in London and his colleagues constructed a virtual reality environment, consisting of two separate rooms with distinct layouts and landmarks (above), using a modified version of the graphics engine used for the role-playing video game Fable. Four participants were asked to navigate this virtual space, and to find specified locations within it, while their brains were scanned with high spatial resolution functional magnetic resonance imaging (fMRI). 

The researchers analysed their data with an algorithm similar to the one which was recently used to reconstruct visual images from the activity of the visual cortex. They focused on the activity of several dozen clusters of neurons in the posterior part of the hippocampus, which is known to be involved in spatial memory. Each of these clusters comprises some 10,000 cells, so that the activity of several million of the approximately 40 million hippocampal neurons was recorded.

Thus the researchers identifed the activation patterns associated with a number locations in the virtual environment. They found that they could distnguish between these patterns, and were therefore able to predict the precise location of their participants in the virtual space. The researchers were effectively reading spatial memories as they were being encoded and consolidated in the hippocampus.

This study confirms that spatial memories are encoded by the co-ordinated activity of large numbers of neurons distributed widely throughout the posterior portion of the hippocampus. But by decoding the activity associated with the memories of specific locations, the researchers have demonstrated that the brain's representations of space are encoded in an orderly, structured manner, and not randomly, as previously thought. The study therefore marks a big step toward understanding how spatial memories are encoded, and how this process breaks down in conditions such as Alzheimer's Disease.

Related:

• Developmental topographagnosia
• Rats know their limits with border cells
• Reading the contents of working memory
  
• Visual images reconstructed from brain activity 

Hassabis, D. et al (2009). Decoding Neuronal Ensembles in the Human Hippocampus Curr. Biol. DOI: 10.1016/j.cub.2009.02.033.