Gated stochastic accumulator model. (a) Neural activity recorded from FEF visual neurons provided salience input. VT, activity when a target fell in the neuron's RF, VD, activity when distractors fell in the RF. Salience input was gated, g such that stochastic accumulation could not begin until a critical level of input was achieved. Once reached, movement units corresponding to choices into (MT) or away from (MD) the RF began accumulating evidence. Movement accumulators were subject to leakage, k and lateral inhibition, β. (b) A decision was committed once accumulation reached a fixed threshold, θ. The model not only accounts for overt behaviour but also accounts for the form of FEF movement neuron activity.