Second-order conditioning sometimes depends on mediation by first-order cues. The competing assumption, that there is a direct second-order and outcome association, is challenged by experiments that have shown that responding to the second-order cue is attenuated after the first-order cue is extinguished. Alternatively, we propose that extinction of the first order cue could result in mediated extinction to the second-order cue. We will discuss experiments designed to differentiate between these two accounts.