Is there a tipping point in neuronal ensembles during learning?

Abstract

Learning is important for humans and can be disrupted by disease. However, the essence of how learning may be represented within a neuronal network is still elusive. Spike trains generated by neurons have been demonstrated to carry information which is relevant for learning. The present study uses well-established mutual information (MI) analysis techniques to better understand learning within neuronal ensembles. Spike trains in tetrode recordings from the dorso–lateral striatum were used for computing MI as rats learnt a T-maze procedural task. We demonstrate that in in-vivo recordings the growth of MI is reflected in the behavioral response as learning proceeds. These changes in MI are seen to correspond to three phases, a low MI value, namely early learning, a rapid increase in MI value, task-acquisition and stabilization of MI, over-training. Over multiple training sessions, small changes in MI within the neuronal network suddenly produce a big change in ensemble MI during the task acquisition phase. This phase represents the “tipping point” in the neuronal network where the MI growth builds habits during motor learning in the striatum.