Learning from mistakes

We re-examine the commonly held view that learning and memory necessarily r equire potentiation of synapses. A simple neuronal model of self-organized learning with no positive reinforcement is presented. The strongest synapse s are selected for propagation of activity. Active synaptic connections are temporarily "tagged" and subsequently depressed if the resulting output tu rns out to be unsuccessful. Thus, all learning occurs by mistakes. The mode l operates at a highly adaptive state with low activity. Previously stored patterns may be swiftly retrieved when the environment and the demands of t he brain change. The combined process of: (i) activity selection by extrema l "winner-take-all" dynamics; and (ii) the subsequent weeding out of synaps es may be viewed as synaptic Darwinism. We argue that all the features of the model are biologically plausible and discuss our results in light of recent experiments by Fitzsimonds et al. on back-propagation of long-term depression, by Xu ei nl. on facilitation of long-term depression in the hippocampus by behavioural stress, and by Prey and Morris on synaptic tagging. (C) 1999 IBRO. Published by Elsevier Scienc e Ltd.