In order to learn the temporal order of events through plasticity, two neurons representing sequential events A and B should be repeatedly activated in the same order over short time scales, about 10 ms, of Hebbian synaptic plasticity. During behavior neurons A and B would be typically active over time scale of about 1000 ms. However, neurons are believed to fire in a rate-modulated Poisson fashion. Hence, the behavioral order of events A and B, occuring over 1000 ms, would be rarely reflected over the short time scales of synaptic plasticity, resulting in poor learning.
A novel mechanism that can solve this scaling problem was proposed rescently (Mehta et al 2000), inolving an interaction between periodic inhibition and slowly varying excitation. Consistent with the predictions of this model, we find that the hippocampal temporal code is asymmetric and it becomes more robust with experience (Mehta et al. 2002). In particular, the correlation between the phase of the theta rhythm (at which a neuron fires a spike) and the position of the rat improves two fold with experience. As a result, the hippocampal spatio-temporal receptive fields change their structure from random to inseparable within a few trials. This mechanism generalizes easily to other modalities such as the structure of spatio-temporal receptive fields of direction selective neurons in V1