The paper investigates the cellular and circuit mechanisms underlying time integration in perceptual decisions, using a simplified two-variable version of a biophysically realistic cortical network model. The authors found that slow time integration can be robustly achieved if excitatory reverberation is primarily mediated by NMDA receptors, while fast AMPA receptors alone do not produce comparable decision times. The model exhibits two distinct modes of behavior: winner-take-all competition for decision computation and the presence or absence of attractor states for working memory. The decision process is linked to the local dynamics in the "decision space" near an unstable saddle steady state, which separates the basins of attraction for the two alternative choices. This framework provides a quantitative explanation for the dependence of performance and response time on task difficulty and the longer reaction times in error trials compared to correct trials observed in monkey experiments. The simplified two-variable model offers a biophysically plausible framework for studying perceptual decision-making in general.The paper investigates the cellular and circuit mechanisms underlying time integration in perceptual decisions, using a simplified two-variable version of a biophysically realistic cortical network model. The authors found that slow time integration can be robustly achieved if excitatory reverberation is primarily mediated by NMDA receptors, while fast AMPA receptors alone do not produce comparable decision times. The model exhibits two distinct modes of behavior: winner-take-all competition for decision computation and the presence or absence of attractor states for working memory. The decision process is linked to the local dynamics in the "decision space" near an unstable saddle steady state, which separates the basins of attraction for the two alternative choices. This framework provides a quantitative explanation for the dependence of performance and response time on task difficulty and the longer reaction times in error trials compared to correct trials observed in monkey experiments. The simplified two-variable model offers a biophysically plausible framework for studying perceptual decision-making in general.