By means of a selfadjusting experimental protocol based on the evaluation of the cardiac impedance, peak current thresholds were determined to defibrillate canine hearts with transventricular capacitor discharges. From 346 successful defibrillations performed on 20 mongrel dogs, an average value of 0·82 A/kg (s.d.=0·3) was found. The energy showed considerably more variability (mean of 0·22 J/kg, s.d.=0·156), suggesting that it is not the most adequate electrical parameter to describe the dose for defibrillation. From these results, a recommended dose equal to the average value plus two standard deviations (i.e. 1·42 A/kg) would appear as acceptable at an essentially 100% success level, assuming it is the first trial and that no other deterioration exists. The procedure applied in these determinations made use of a constant current 12 kHz impedance meter to estimate the load imposed on the defibrillator by the fibrillating heart. The overall average value was 25·6 Ω (s.d.=4·9) ranging from 17 to 44 Ω. The average time constant RC of the exponential discharge was 1·63 ms (s.d.=0·31) using in all cases a 63 μF capacitor. There is, of course, an error in taking the modulus of Z as equal to R and also due to the low current used in the measurements, as opposed to the high current that flows during actual defibrillation. However, it does not seem to be significant from a practical point of view. Defibrillators with a built-in Z meter and a microprocessor to automatically set the value of the discharge are, as a consequence, quite conceivable. The results given here confirm, and, in a sense, improve on those already reported in the literature.