This paper presents a robust procedure for implementing glottal inverse filtering that requires no manual intervention. Speech from a number of individuals, male and female, has been analysed using this procedure and a glottal model has been fitted to the inverse filtered waveforms. Using the excitation waveform defined by the model, the vocal tract filter has been re-estimated using whole-cycle larynx synchronous analysis. The paper discusses the resultant formant tracks and also the intra-speaker and inter-speaker variability of the glottal model parameters for three selected vowels.