1. The circadian pacemakers controlling activity rhythms in four species of rodents are compared, as freerunning systems in constant darkness. In analyzing their stability the distinction is made between (1) spontaneous day-to-day instability of frequency, and (2) a longer-term lability, some of which is traceable to identified causes. 2. Serial correlation analysis indicates that the precision (day-to-day stability) of the pacemaker's period is ca. twice as good (estimated s.d. = 0.6 % of ¯τ in Mus musculus) as the already remarkable precision of the activity rhythm it drives (average s.d. = 1.2 % of ¯τ). 3. Identifiable causes of long-term lability include age and several features of prior entrainment by light. The period and photoperiod of a light cycle have a predictable influence on the subsequent freerunning period (τ) of the pacemaker; they cause "after-effects". So do single light pulses causing a phase-shift in the freerunning system. Constant light also has an after-effect opposite in sign from the after-effect of long photoperiods. 4. After-effects of "skeleton" photoperiods support the hypothesis that the transitions of light to darkness vv. are involved in the entrainment process which leads to changes in τ. 5. Both day-to-day instability and long term lability are most pronounced in species (Peromyscus maniculatus, Mus musculus) whose ¯τ is considerably shorter than 24 h; they are least pronounced in hamsters whose ¯τ is indistinguishably close to 24 h. 6. The differences between the species in τ and its lability are paralleled by differences in pacemaker lability as measured in light-induced after-effects and in the extent of changes with age. The species evidently differ in the "tightness" with which τ is homeostatically conserved.