STOCHASTIC MODELS AND CHOICE BEHAVIOUR.

A general stochastic representation of the processes of choice was studied, and several current models were examined as special cases. Alternative models were specified by the pattern of accumulator states that were supposed to lead to an overt response. Three cases were emphasized 1 a simple accumulator or recruitment scheme, 2 accumulator with inhibition in which excitation coming into one accumulator returns all others to zero, and 3 random walk, a simplified version of the optimal statistical decision device, in which response occurs when the excitation in one accumulator exceeds that in any other by a threshold value. The distribution of decision times, the probability of alternative responses, and indices of vacillation and confidence are derived as formulas in terms of model parameters for models operating in discrete and continuous time. A method was developed for obtaining these properties for any model capable of being expressed as a finite Markov chain. Attention was given to properties that take distinctly different predicted forms for different models. The most important of these is the relation between the latency and relative frequency of a response. Author