The 2f1-f2 distortion product otoacoustic emission (DPOAE) response, measured as a function of the primary frequency ratio r shows a bell-shaped curve, which depends, in principle, on the interplay among three factors: The dependence on r of the overlap between the basilar membrane (BM) responses to the primary tones, linear interference between the wavelets coming from spatially distributed sources, and nonlinear suppression phenomena. This issue has been investigated using a numerical nonlinear cochlear model, and an analytical linear cochlear model in which DPOAEs were generated as nonlinear perturbations, to assess the reliability of a method for estimating cochlear tuning from the width of the experimental DPOAE level vs. ratio function. A fixed-f2 acquisition paradigm was used, varying r continuously in the range 1-1.5, for two f2 values (1250 and 5000 Hz). The resulting experimental "spectra" have been time-frequency analyzed to select the distortion component, and compared with the models' simulations. The experimental results are well matched by both models, suggesting that the main features of the DPOAE vs. ratio curve are quite independent of nonlinear suppression phenomena. From the same model simulations, the basilar membrane BM tuning can be directly estimated. A strong linear correlation was found in the model between the width of the BM response and that of the DPOAE level vs. ratio relation, supporting the reliability of estimates of cochlear tuning obtained from the evaluation of the DP versus ratio experimental curves.