Low-frequency masking is a recent clinical procedure for the differential diagnosis of sensory hearing loss. Currently this requires the recording of the phase-dependent masked subjective threshold, which is time consuming and not always accurate. As an objective method, the recording of modulated distortion product otoacoustic emissions (DPOAEs) can be performed continuously, and with better frequency specificity. Results of measurements of the low-frequency modulated two-tone DPOAE 2f1-f2 in the human ear, and its dependence on various acoustic parameters, are presented here for the first time. Similar to the masked hearing threshold, the pattern of the phase-dependent modulated DPOAEs displayed two minima, at the phases of maximal rarefaction and condensation, respectively, with a latency of about 4 ms (suppressor frequency 32.8 Hz). The smaller dip, at maximal condensation, appeared only for a high suppressor level, and for a low level of the primary tone f2. The modulating effect measured for the primary frequencies f1 = 2.5 kHz and f2 = 3 kHz, decreased for 4 and 4.8 kHz, and vanished for 5 and 6 kHz. The results are discussed using a cubic distortion model based on the Boltzmann function for mechano-electrical transduction of the hair cells. The saturation behavior of the increase of the DPOAE level at different phases is compared with the growth rates of the DPOAE level in normal hearing and in sensory hearing loss.