The results of a study of the sealar electrocardiogram, utilizing a wide-band recorder, expanded time scale and a low-pass filter, indicate that a recorder flat to at least 500 c.p.s. is required for faithful reproduction. With a high-pass filter, measurable residual signals are present at a cut-off frequency of 1,000 c.p.s., or higher. Therefore, to record these, an adequate response at 1,000 c.p.s. or more is reqnired. The high frequency energy of the electrocardiographic spectrum arises from the fast deflections contained in the original waveform. These may occur in a single fast deflection, notching, or both. Whereas, in the normal individual, high frequency energy usually arises from a relatively smooth, fast deflection, in abnormals the fast events may occur in conjunction with notching, and other deformities. Judging from the technic used in this experiment, a variable band-pass filter is valuable as an aid for studying the high frequency components of the electrocardiogram and establishing equipment requirements. Measurement of peak-to-peak voltage of residual signals gives partial but lot clear-cut separation of normal and abnormal subjects and would not seem to add to the value of high-fidelity electrocardiography per se in routine clinical diagnosis. It is possible that root mean square readings, reflecting the total energy content, might give a better separation between normal and abnormal subjects than peak-to-peak amplitude alone.