Role of oscillations and plasticity in generating a temporal code from a rate code

Mayank R. Mehta and Matthew A. Wilson

Center for Learning and Memory,
Department of Brain and Cognitive Sciences,
Massachusetts Institute of Technology

Recent studies have shown that the hippocampal rate code for space is dynamic, such that size, specificity and shape of place fields change dramatically with experience (Mehta et al., 1997, 2000). As a result, the place fields become negatively skewed, such that the firing rate is low as a rat enters the place field but high at the end. It was shown that these results could arise due to spike timing dependent plasticity (STDP) of the feed-forward synapses from CA3 to CA1. In addition to this asymmetric rate modulation, hippocampal neuronal firing is also modulated by an ~8 Hz theta rhythm. Further, the phase of the theta rhythm at which a neuron fires precesses to smaller values as the rat advances through a place field. This phase precession (O'Keefe and Recce, 1993) thus results in a unique phase, or spike time with respect to the local field potential, for each spatial location. The origin of this temporal code and its relationship to the rate code are unclear.

We show three novel results that address these issues. First, the mean phase of spikes at a given location, i.e. the temporal code, is highly correlated with the mean firing rate of the neuron at that location, i.e. the rate code. Second, the phase precession is nonlinear such that the high phase (360-180 degrees) is more correlated with position than the low phase (180-0 degrees). Third, hippocampal inhibitory neurons also show phase precession, such that the firing rate is correlated with the phase.

These results are explained using a simple feed-forward model. This suggests that a robust temporal code can arise from an interaction between oscillations, delayed inhibition, and an asymmetric rate code that arises from temporally asymmetric, NMDA dependent LTP.

Similar phenomena have been observed in the directional, spatio-temporal receptive fields in V1 (Movshon et al., 1978; Reid et al., 1987), and our model suggests that V1 direction selectivity could arise due to STDP during development.