Existing findings on the truth effect could be explained by recollection of the statements presented in the exposure phase. In order to examine a pure fluency account of this effect, we tested a unique prediction that could not be derived from recollection of a statement. In one experiment, participants judged the truth of a statement that had the same surface appearance as a statement presented earlier but contradicted it, for example “crocodiles sleep with their eyes open” one week after having heard “crocodiles sleep with their eyes closed”. We predicted and found that participants judged contradictory statements as being more false than new statements after a delay of only a few minutes, but judged them as more likely to be true after one week. In contrast to earlier findings, this result cannot be explained by accounts relying on recollection of the previously presented statements.