The time constant (T) of left ventricular pressure fall is widely used as an index of ventricular "relaxation". It is not known whether its rate limiting step is deactivation, an enzymic energy consuming reaction whose rate is therefore sensitive to temperature, or elastic recoil. To distinguish between these possibilities, the time constant was measured by simple logarithmic (Tlog) and exponential (Texp) methods in 12 patients during cooling before coronary artery grafting. Ventricular loading conditions were altered by transfusion from bypass to maintain arterial and left atrial pressures constant in individual patients, though heart rate fell from 86 (8.4) to 68 (10) beats/min. Tlog increased from 49 (10) ms mean (SD), at 37 degrees C to 86 (15) ms at 31 degrees C, and Texp from 63(14) at 37 degrees C to 112 (23) ms at 31 degrees C with intermediate values at 34 degrees C. Texp proved sensitive to "noise" at low temperatures, but the overall change in Tlog with temperature was 9% per degree C--considerably less than that observed experimentally for the rate of tension decline of isolated myocardium, and possibly itself an overestimate because of the concomitant fall in heart rate. The relatively small effect of temperature on Tlog in humans, associated with a considerable load sensitivity appearing under hypothermic conditions, does not favour simple dependence on deactivation as the rate limiting step of left ventricular pressure fall, but suggests that its determinants may be complex.