Simulation, in the time domain, of a complete source‐vocal tract model including the frequency‐dependent losses is quite impossible. But, in order to study the volume flow at the level of the glottis in the best conditions, we can use a two‐mass model loaded by a circuit representing the impedance of the vocal tract. The spectrum of the volume flow has been calculated when the model is loaded by the representative impedance of the vocal tract and when court circuiting. The mean value of the glottal impedance has been determined from comparisons between both spectrums. This value has been measured for different frequencies corresponding to several vowel formants. Formant frequency variations due to the glottal impedance and synchronization effects (pointed out by amplitude variations around formant frequency} can be observed on the source spectrums. Some of these results are in agreement with the real data. Some others have to be verified. In conclusion, and with other results elsewhere reported, it seems that a two‐mass model of the vocal cords is, until now, a good approximation of the real behavior of the glottal source.