A method of optimizing signal averaged recordings is described, in which the signal average remains unbiased. This is in contrast to techniques such as weighted averaging, where beats are given a weighting in the average that is proportional to their noise content. In a study of 22 ambulatory subjects, noise was reduced via signal averaging to a mean RMS level of 0.57+or-0.24 mu V. The number of beats averaged range from 198 to 3354 with a mean of 1410. The level of noise after averaging the ambulatory recording was greater than but still comparable to that of conventional signal averages performed on the same day. These results suggest that it is feasible to perform quality controlled signal averaging from ambulatory ECG recordings. >