While studied extensively since their discovery by Kemp in the late seventies, the cellular basis of the phenomenon of otoacoustic emission remains unknown. Data from experiments in humans, chinchillas and Mongolian gerbils was used to test the hypothesis that otoacoustic emissions originate in the hair cell transduction apparatus. Specifically, a double Boltzmann model of the transducer predicts that emissions generated by a single tone (stimulus frequency otoacoustic emissions - SFOAE) should be measurable at stimulus levels 20 or more dB below neural threshold, but sufficient to modulate the activity of enough transduction channels to produce a macroscopically observable result. On the other hand, for a fixed low-level probe tone that evokes SFOAE, it should only be possible to demonstrate the presence of emission by using a suppressor tone large enough to drive the transducer into its nonlinear range, approximately where the suppressor level reaches neural threshold. This result should be independent of suppressor frequency. Both predictions were confirmed experimentally in all three species. The threshold suppressor level was consistently near the threshold of the compound neural response monitored with an extracochlear electrode, even for suppressors more than an octave higher than the frequency of a low-level (30 dB SPL) probe tone. Cochlear microphonic responses were always detected at the lowest levels demonstrating SFOAE. The hair cell transducer appears to be the site of interaction between the probe and suppressor tones for all suppressor frequencies, consistent with a single suppression mechanism. Nonlinear interactions demonstrated in SFOAE and CM between widely separated tones do not appear to have a correlate in the basilar membrane, suggesting that, at least under some conditions, pressure waves can be initiated directly from forces produced by the hair bundle.