To orient ourselves, we mainly need two pieces of information: where am I and in which direction am I heading?
Experiments in the rat have shown that these types of information are directly accessible and independently coded in the brain.
When the rat explores a new territory, so-called place cells and head direction cells form within only a few minutes. Place cells are active when the rat visits a particular area, no matter which direction it is facing. In contrast, head direction cells code the direction the rat is heading, independent of where it is.
Also humans presumably have these and other types of cells which specifically instruct its sense of orientation. Scientists around Mathias Franzius and Laurenz Wiskott from the Humboldt-University and Bernstein Center for Computational Neuroscience Berlin (Germany) have now developed a theoretical model that can explain the emergence of all orientation-specific cells that are known in rats and primates to date.
The model analyzes realistic image data that reflect the visual input a rat would perceive when exploring a new environment. The core of their model is a mathematical algorithm called "slow feature analysis", which extracts information relevant to orientation from the image data. On the basis of this algorithm, the model generates place cells and head direction cells -- without this being an a priori requirement.
Every receptor in the eye only captures a very small section of the perceived image. When shifting the gaze just a little, the information that every single receptor transmits will be quite different than before. While sensors deliver constantly changing data, the information important for orientation varies far more slowly -- the overall impression in the example above remains almost constant. Features that vary slowly can be obtained from image data by slow feature analysis.
With their model, the scientists could show that slow feature analysis allows the emergence of what one could call a "cognitive map" from the linear sequence of visual data a rat receives when moving through a new environment. In this map, positions are coded by place cells, whereas a directional reference frame is given by head direction cells. It is only after this learning process that disparate visual impressions can activate the same set of place- or head direction cells. If, for example, the rat is located in the northern corner of its cage, the same place cells will show activity, no matter if it is heading east or west.
The work will be published on August 31, 2007, in the scientific journal PLoS Computational Biology.
CITATION: Franzius M, Sprekeler H, Wiskott L (2007) Slowness and sparseness lead to place, head-direction, and spatial-view cells. PLoS Comput Biol 3(8): e166. doi:10.1371/journal.pcbi.0030166
Cite This Page: