Figure 15:

Simulation 3c: Second-order conditioning. Simulation results contrasting canonical second-order conditioning (top; 50% maintenance trials) with a variant in which CS2 activity endures until the time of the omitted US (bottom; also 50% maintenance trials). The latter converts the relation between CS2 and US nonoccurrence from a trace-like to a delay-like conditioning relation and converts a positive dopamine response to the CS2 (top) into a negative one (bottom), i.e., a conditioned inhibitor (simulation 3b).