High-resolution measurements of distortion product otoacoustic emissions (DPOAEs) from three different experimental paradigms are shown to be in agreement with the implications of a realistic 'two-source' cochlear model of DPOAE fine structure. The measurements of DPOAE amplitude and phase imply an interference phenomenon involving one source in the region of Strong nonlinear interaction of the primary waves (the strong 'overlap' or generation region), and the other source region around the DPOAE tonotopic place. The component from the DPOAE place can be larger than the one from the generator region. These findings are supported by the analysis of the onset and offset of the DPOAE when the higher-frequency primary is pulsed on and off. The two-source hypothesis was further tested by adding a third tone closer in frequency to the DPOAE which modifies the amplitude of the component from the DPOAE place and leaves the one from the generator region unchanged. The results agree well with the model prediction that the variation with frequency; and implied latency, of the phase of the DPOAE tonotopic-place component are greater than the corresponding quantities for the component from the generation region.