Humans are sensitive to temporal redundancies in their environment. When the identity of a target stimulus is correlated with the duration of the preceding interval, performance is better for frequent than for infrequent combinations of target and interval. This effect has been demonstrated several times in current timing research. However, it can be accounted for by 2 starkly contrasting explanations. The standard account has explained it in terms of learning associations between intervals and stimulus-response events. But, alternatively the effect might be due to partial trial transition costs, because infrequent time-event combinations are proportionally more often partial transitions (i.e., transitions of either interval, or target). We conducted 3 choice response time experiments to distinguish between both explanations. The results clearly show that the time-event correlation effect is due to learning, not to partial transition costs.