Summation tests were used to assess whether participants in a human predictive judgement task learned that some cues (foods) prevented the occurrence of the outcome (migraine). In three experiments the preventative effect of a conditioned inhibitor, I, trained in the design, P+, PI-, I-, was stronger than that of a negative control cue, N, that had been presented alone with the same frequency and simply predicted no outcome. This control cue, N, also passed the summation test when compared to a novel control cue. Experiments 2 and 3 examined whether the latter negative control effect was attributable to differential inhibition. Manipulations designed to alter context learning and the magnitude of the negative contingency did not affect the properties of the negative control cue. Thus, these experiments did not support the possibility that a cue given simple negative training acquires differential inhibition. Regardless of the mechanism underlying the negative control effect, the consistent finding that the conditioned inhibition cue, I, reduced prediction of the outcome more than did the negative control cue, N, provides evidence for true conditioned inhibition whose acquisition requires co-occurrence of the preventative cue with a positive training cue.