Numerous cases of precursory seismic quiescence have been reported in recent years. Some investigators have interpreted these observations as evidence that seismic quiescence is a somewhat reliable precursor to moderate or large earthquakes. However, because failures of the pattern to predict earthquakes may not, in general, be reported, and because numerous earthquakes are not preceded by quiescence, the validity and reliability of the quiescence precursor have not been established. We have analyzed the seismicity rate prior to, and in the source region of, 37 shallow earthquakes (M 5.3-7.0) in central California and Japan for patterns of rate fluctuation, especially precursory quiescence. Nonuniformity in rate for these pre-mainshock sequences is relatively high, and numerous intervals with significant (p<0.10) extrema in rate are observed in some of the sequences. In other sequences, however, the rate remains within normal limits up to the time of the mainshock. Overall, in terms of an observational basis for intermediate-term earthquake prediction, no evidence is found in the cases studied for a systematic, widespread or reliable pattern of quiescence prior to the mainshocks. In earthquake sequences comprising full seismic cycles for 5 sets of (M 3.7-5.1) repeat earthquakes on the San Andreas fault near Bear Valley, California, the seismicity rates are found to be uniform. A composite of the estimated rate fluctuations for the sequences, normalized to the length of the seismic cycle, reveals a weak pattern of a low rate in the first third of the cycle, and a high rate in the last few months. While these observations are qualitative, they may represent weak expressions of physical processes occurring in the source region over the seismic cycle. Re-examination of seismicity rate fluctuations in volumes along the creeping section of the San Andreas fault specified by Wyss and Burford (1985) qualitatively confirms the existence of low-rate intervals in volumes 361, 386, 382, 372 and 401. However, only the quiescence in volume 386 is found by the present study to be statistically significant. ?? 1988 Birkha??user Verlag.