Recent accounts of selective adaptation in speech perceptin have proposed that either one or two leves of processing are adapted. Most of the previous experimental results can, however, be accounted for by either type of model. In the present experiments, two aspects of the selective adaptation paradigm were manipulated. The spectral (frequency) overlap between adapting and test syllables was manipulated along with differences in interaural presentation (adapting in one ear, testing in the other). The results indicated that the adapting syllables drawn from the test series and adaptors with no spectral overlap with the test series both produced significant changes in subjects ratings of the test stimuli. However, the identical adaptors caused significantly more adaptation than the nonoverlapping adaptors. Moreover, the nonoverlapping adaptors produced 100% interaural transfer of adaptation, indicating a central locus of this effect. The identical adaptors drawn from the test series showed approximately 50% interaural transfer. Taken together, these results strongly suggest that two levels of processing are involved in selective adaptation to place of articulation in stop consonants. One is a peripheral level that is relatively frequency specific and the other is a central level that integrates information over a wider frequency (spectral) range.