A multilayer drift‐diffusion broadcast model: Converting temporal evidence into response timing

Abstract

The peak interval procedure, in which subjects learn to cluster their anticipatory responses around delays to biologically critical outcomes, provides benchmarks for interval timing models. These include (a) the bistable response regime demarcated by an abrupt increase (start times) and a decrease (stop times) in response rates around the target time in individual trials; (b) bell‐curve‐shaped average response curves that superpose on a relative time scale; (c) lower coefficient of variation of stop times than start times; and (d) specific correlational patterns between these parameters, their mean, and difference. This article presents a multilayer computational model that predicts these key markers of timing behavior in the peak interval procedure. The first layer constitutes the clock stage, comprising uncoupled diffusion processes approaching a single boundary. The first crossing of the threshold for each diffusion process is broadcast to the downstream motor layer. The motor layer integrates the broadcast signals with a leak term scaled by the drift rate to avoid saturation and reset of motor system activity. The model is extended to account for real‐time temporal discrimination. This family of models departs from the single‐clock approach to modeling and meets the core empirical benchmarks of timing behavior.