We have analyzed the time series of geomagnetic field reversals using a correlation function of the telegraphic signal obtained from the Heirtzler scale by assigning ¿1 to intervals of normal/reversed polarity. Unlike the correlation function used by Naidu, the autocorrelation function we define is only very slightly sensitive to undetected short polarity intervals. Thus the resluts we obtain are valid irrespective of failure to detect short polarity intervals and give information about the dynamics of the time series of polarity reversals. We show that, within statistical noise limits, the successive polarity intervals are statistically independent and are distributed in time according to Poisson process, as already suggested by Cox using other mathematical means.