A sequential model of two-choice intensity identification

Abstract

A model of perceptual decision-making in two-choice intensity identification tasks is advanced. The model assumes that sensory pathways encode the physical intensity of the stimulus in the firing rates of sensory afferents, characterized by exponentially distributed interarrival times. The decision-making process entails a sequential comparison of each interarrival time with memory traces from prior stimulus exposure. This yields a random walk process reminiscent of the two-choice RT model by Stone (1960), but with an additional stochastic element introduced by variable sampling times. The model provides a reasonable account of data garnered in a brightness identification task (Experiment 1), aligning with distributional RT statistics and intensity effects on mean RTs. Several post hoc assumptions, such as variability and bias in the starting point of the random walk, are required to accurately predict error RT distributions, however, which introduces problematic asymmetries in predicted error probabilities. Applying the model to a loudness identification task (Experiment 2) necessitated the additional assumption of variability in transduction rates to overcome challenges in accommodating longer RTs for errors compared to correct responses in this task.