Why does the single neuron activity change from trial to trial during sensory-motor task?

Single neuron activities from cortical areas of a monkey were recorded while performing a sensory-motor task (a choice reaction time task). Quantitative trial-by-trial analysis revealed that the timing of peak activity exhibited large variation from trial to trial, compared to the variation in the behavioral reaction time of the task. Therefore, we developed a multi-unit dynamic neural network model to investigate the effects of structure of neural connections on the variation of the timing of peak activity. Computer simulation of the model showed that, even though the units are connected in a cascade fashion, a wide variation exists in the timing of peak activity of neurons because of parallel organization of neural network within each unit.