A major problem in defining the chronology of geomagnetic reversals is linked to the detection of short (<30 kyr) time intervals between reversals (TIBR). Published polarity timescales do not usually include the shortest TIBR; therefore the timescales are inherently incomplete. The purpose of this paper is to investigate the effects of this incompleteness on the analysis of the geomagnetic polarity timescale; the ultimate goal is to provide constraints on the Earth's core processes. The effects of inclusion/exclusion of the shortest TIBR are verified by comparing the results obtained by statistical analysis of real and synthetic series of events; for the real sequence, two basic cases are considered in which the tiny wiggles are attributed either to short TIBR or to paleointensity fluctuations. Particular attention is paid to the influence of measurement errors estimated for the most recently published Cenozoic timescale. By following the minimalist philosophy of Occam's razor, which is particularly suitable for studying poorly known processes, the reliability of the simplest model, i.e., the Poisson process which is symmetric in polarity, can be checked. The results indicate the plausibility of a generalized renewal process; the only regularity is relative to the long-term trend, which is probably linked to core-mantle coupling. In detail, a uniform exponential trend in the last 80 Myr is found for the timescale; it is not presently possible to estimate the influence of the inclusion of tiny wiggles because they are well-resolved only in the last 30 Myr, a period in which both series are stationary. The sequences, with and without tiny wiggles, are symmetric in polarity, with no evidence of low-dimensional chaos and memory of past configurations. The empirical statistical distribution of the TIBR departs slightly from a theoretical exponential distribution, i.e., from a Poisson process, which can be explained by a lack of short anomalies, and/or by a generating process with wear-out properties (a more general renewal process). A real exponential distribution is sustainable only if the number of missing short TIBR in the last 30 Myr is larger than the number of tiny wiggles observed in the same period.¿ 1997 American Geophysical Union