Transient global amnesia: Model, mechanism, hypothesis.

Transient global amnesia (TGA) is a rare but well-described neurological disorder characterised by an acute self-limited episode of dense anterograde amnesia with retrograde amnesia of variable extent. Although the clinical phenomenology, neuropsychology, neuroimaging, prognosis, and precipitating and predisposing factors of TGA are well described, the pathogenesis is not currently understood. Existing proposals - epileptic seizure, stroke or transient ischaemic attack, migraine aura - all have shortcomings. Based on a computational neural network model of mnemonic function, a new model of TGA pathogenesis is suggested, along with potential mechanisms, thus generating a potentially testable hypothesis of TGA pathogenesis. Based on the intrinsic neuronal circuitry of the hippocampal formation, specifically the rich CA3 recurrent collateral projections which have previously been characterised as a global autoassociative attractor network, a model of TGA is developed in which loss of fault tolerance, a property of attractor networks, results in catastrophic rather than graceful degradation. Excessive positive feedback through the CA3 recurrent collaterals produces runaway neural firing, with infinite gain in the short CA3 feedback loops, resulting in a singularity or discontinuity manifest as failure of synaptic transmission. In terms of the energy landscape, the system flips to a shallower, low firing rate, basin of attraction. At the mechanistic level, spreading depolarization is suggested to underpin these changes. These considerations permit the construction of a hypothesis of TGA pathogenesis which may have testable, falsifiable, predictions.