Miller and Cohen (2001) compare the PFC to a railroad switchman: it doesn't carry the trains, it sets which track they take. In their Stroop example (after Cohen, Dunbar and McClelland, 1990), color and word inputs travel along two pathways to shared responses. The word pathway is stronger, so naming the ink color only wins when a PFC task unit biases the color pathway.
Choose a task and a stimulus, then run a trial. Each run adds a little random noise, so trials near the edge can go either way.
Run many noisy trials of each stimulus type with the current task, PFC strength and pathway strengths. Reaction time is in model cycles and is averaged over correct trials only.