Modifiable neuronal connections: an overview for psychiatrists.

Synaptic plasticity is currently the target of much neurobiological research, because it is thought to play an important role in brain function (particularly memory formation). However, it has attracted little attention from psychiatrists to date despite accumulating evidence that links it to various clinical syndromes, including amnesia and possibly psychosis. The purpose of this article is to present an overview of the two major arms of synaptic plasticity research-theoretical (the field of neural network modeling) and neurobiological (long-term potentiation). Artificial neural networks are a class of theoretical model that has been developed with the aim of understanding how information could, in principle, be represented by large numbers of interconnected and relatively simple units. Over the past few decades, several theoretical accounts of information-processing mechanisms have been developed, and these are briefly reviewed. The principle common to representation formation in nearly all neural networks is that of "associability"-the idea that streams of information are combined by forming, strengthening, or pruning connections between them to form new representations that can later be retrieved. Associability also lies at the heart of psychological theories of information storage in the brain. Research into associability has directed the attention of many experimenters toward the possible biological correlates of such mechanisms. Of particular interest is the recent discovery that some neurons appear to possess connections of modifiable strength. The implications of this finding for psychiatry are discussed in relation to representational disorders such as delusions and amnesia.