Abstract: The continuous phase angular demodulation method disclosed digitally processes the signal in baseband after having over-sampled it with reference to the bit period. The processing consists in routinely demodulating sub-sets of differential phases at the bit period T.sub.b shifted with respect to one another by fractions T.sub.b /q of this period, in correcting them, a priori, by the phase deviations associated with a set of pre-defined d frequency drifts and in computing, for the d.q sets thus obtained, a noise criterion. The set of demodulated bits chosen is the one that reduces this noise criterion to the minimum. The set of bits then enables computation of the phase variation emitted and, using this variation and the measured variation, the real frequency drift. Another demodulation taking this drift into account is done for the q sets of initial differential stages and the set minimizing the noise criterion is then chosen and fixes the synchronization bit by the sampling instants associated with it.