Distortion product otoacoustic emissions (DPOAEs) provide a non-invasive and relatively direct method for evaluating cochlear travel time in humans. In the present study, the 2f1-f2 DPOAE latency was deduced from DPOAE phase shift according to f2 frequency shift, with f1 being fixed, using the ILO92 system. Latencies of 2f1-f2 DPOAEs were recorded at various primary frequencies and levels. Results were not wholly consistent with previous latency estimates. This study showed that the functions describing both 2f1-f2 DPOAE latency decrease with primary frequency increase, at several primary-tone levels, and 2f1-f2 DPOAE latency decrease with primary-level increase, at various frequencies, are best fit by exponential equations. The results of the evaluation of the latency-frequency-level relationship indicated the possibility of linking these three variables by the following expression: Latency = 68.30 exp (-0.027I -0.000434 f)+1.13.