The phenomenology of spontaneous otoacoustic emissions (OAEs) is compared to theoretical predictions given by models in which they are produced by active nonlinear oscillators. Along with the well-known Van der Pol oscillator, a new active oscillator model is proposed and analyzed here. Numerical simulations and multi-scale analytical computation results are compared to experimental data of neonatal spontaneous and evoked OAEs. A simple analysis technique is proposed, in which the time evolution after a click stimulus of the amplitude of each spectral line corresponding to a spontaneous OAE is studied. Apart from a few stationary lines, an approximately exponential decay law, with characteristic damping coefficients in the 20-200 Hz range, was found to fit the data. These results are not compatible with a Van der Pol oscillator model, and show that some important aspects of the OAE phenomenology can be better explained by the proposed oscillator. Other interesting features of the spontaneous end evoked OAE phenomenology, such as spontaneous OAE suppression by external tones and the following recovery, as well as stimulus/response curves in the linear and nonlinear mode of acquisition, are also well reproduced by the proposed model.