Negative polarity items occur within the immediate scope of operators that support decreasing inferences. It is natural to expect the presence of a licensed NPI to facilitate the processing of decreasing inferences based on the licensor. In joint work with L. Bott and B. McElree I conducted a series of experiments to test this. We found no facilitation in terms of either verification accuracy or processing speed; instead, the presence of an NPI slowed down the processing of valid inferences. The talk will report on the experiments and consider various theories of licensing as potential explanations.