The yield and reproducibility of various methods of ventricular premature beat (VPB) detection are examined in otherwise healthy middle-aged men first found to have VPB in a 2-min lead I ECG rhythm strip. With a combination of an isometric and treadmill exercise test, VPB were repeatedly detected in 83%. Test-retest reliability in classifying the subjects by frequency of VPB was 67%. The reliability of detecting complex VPB was 47% for multiform VPB, 35% for pairs of VPB, 17% for runs and 36% for VPB showing the R-on-T phenomenon. These data suggest that simple and complex VPB detected by the described methods among normal men are poorly reproducible in the individual. Taking the VPB reproducibility for the group as a whole, the proportion of subjects having different frequency or kinds of ectopic beats is reasonably stable for the different test occasions.