Experiencing two odors as a mixture can later increase their perceived similarity when presented separately. Such an increase in similarity can be used as an implicit measure of how well participants remember the mixture. Three experiments tested the resistance to interference of this effect by first giving participants exposure to two 2-odor mixtures and then presenting each odor from one mixture (interfered pair) separately many times. After the latter interference phase, participants rated the similarity of the two odors in each pair. The experiments varied in terms of the number of initial exposures to the mixtures, the number of odor presentations in the interference phase, and the control conditions employed. No difference in similarity ratings for the interfered and non-interfered pair was found in any experiment. In contrast to the lack of interference with this implicit measure, an explicit measure of memory based on participants' ratings of odor frequency revealed that they could recall to some degree that the interfered odors had been presented alone in the second phase. These results suggest that implicit memory for odor mixtures is highly resistant to interference and is consistent with a configural encoding account of this effect.