Decoding memories

The ensemble activity of hippocampal neurons carries information about the location of an animal in its environment (see ‘A code for space‘). For example, one may find that cell 1 is only active when a rat is in location A; and cell 2 is only active when the rat is in location B. Conversely, if at some time in the future we observe that cell 2 is active, then we would predict that the rat is most likely in location B. This technique is called neural decoding. More generally, once the spatial correlates of population of hippocampal neurons has been established, then at any time in the future the animal’s location can be decoded based solely on the observed activity of these neurons.

Hippocampal place cells do not signal an animal’s current physical location at all times. The ensemble activity pattern switches to a distinct state when the animal is sleeping or when the animal is awake but sits quietly. During these ‘offline’ states, hippocampal neurons are mainly silent, except for short lasting bursts that occur about one or two times per second. In a recent paper we applied neural decoding to these population bursts to analyze their content. We showed that during such bursts spatial sequences (i.e. trajectories) are replayed at high speed. An example of this phenomenon is shown in the video below. Here a rat is stopped in the middle of a long track. Overlaid onto the video is a set of arrows that represent our estimate of the animal’s (virtual) location solely based on the spiking pattern of hippocampal neurons. It is as if the animal is running along the track in his mind. We are presently studying the role of this hippocampal memory replay in planning behavior, evaluating past performance and memory consolidation.