This study compared estimates of sinoatrial conduction time (SACT) obtained by constant atrial pacing (CAP) and premature atrial stimulation (PAS) with measured SACT in isolated rabbit right atrial preparations. Transmembrane potentials and surface electrograms were recorded from the sinus node and crista terminalis, respectively. The crista terminalis was paced 5, 10 and 15 beats/min faster than the spontaneous sinus rate with a train of eight pulses. Estimate of SACT by CAP was taken as the difference between the first atrial return cycle and the mean spontaneous cycle length. SACTs at 5, 10 and 15 beats/min faster were 76 ± 10, 86 i 10 and 96 + 10 msec (mean SEM; n = 12), respectively; correlation coefficients with the true SACT were 0.7, 0.54 and 0.4. Consecutive determinations of SACT by PAS and CAP in the same preparation (n = 6) at 10 beats/min faster gave SACTs of 86 + 13 and 79 ± 14 msec, respectively, compared with true SACTs of 79 10 msec. Shortening of sinus node action potential, depression of automaticity and shifts in the site of the primary pacemaker contributed to the errors in both techniques. Estimation of SACT by CAP may be further complicated by failure of sinus node capture. Principles to minimize some of these errors are also presented.