In cardiac arrest due to prolonged ventricular fibrillation (VF), defibrillation is more likely to result in a perfusing rhythm if chest compressions are performed first. Furthermore, the VF waveform can predict the shockability of VF and thus automated external defibrillators (AEDs) are being designed to analyze the VF waveform to direct therapies. However, it is unknown whether the VF waveform is dependent on recording direction, which could be altered by incorrect placement of AED patches. VF was induced in 26 patients with ischemic cardiomyopathy and 19 patients with dilated cardiomyopathy and recorded in six limb leads. Frequency characteristics (mean, median, dominant frequency, and bandwidth) were computed as well as amplitude-based measures: amplitude spectral area (AMSA), slope, signal amplitude, and slope divide by signal amplitude (slope-amp). Frequency characteristics were similar in all leads. However, AMSA, slope, and signal amplitude were significantly affected (P<0.001) by lead. In particular, for ischemic cardiomyopathy patients, between leads I and II, AMSA varied from 29.4+/-3.2 to 49.3+/-4.6 mV Hz (mean+/-SEM, P<0.001) and slope varied from 1.5+/-0.2 to 2.4+/-0.3 mV/s (P<0.001). Slope-amp was similar in all leads. There were no significant differences between ischemic and dilated cardiomyopathy patients. Amplitude measures of VF are significantly affected by limb lead ECG recording direction. This work suggests that AED patches must be correctly and consistently placed if amplitude-based measures are used to decide whether to deliver a defibrillatory shock.