Magnetocardiography (MCG) is a method complementary to electrocardiography (ECG). We examined recording and reproducibility of atrial depolarization signal by MCG. Multichannel MCG over anterior chest and orthogonal 3-lead ECG were recorded in 9 patients who had paroxysmal lone atrial fibrillation and in 10 healthy subjects in duplicate at least 1 week apart. Data were averaged using atrial wave template and high-pass filtered at 25, 40, and 60 Hz. Atrial signal duration with automatic detection of onset and offset and root mean square amplitudes of the last portion of atrial signal were determined. Coefficient of variation of atrial signal duration by MCG at 40 Hz was 3.3% and difference between the measurements was 3.5 milliseconds on average. The corresponding figures obtained by signal-averaged ECG (SAECG) were 6.1% and 6.9 milliseconds. Coefficient of variation for root mean square of the last 40 milliseconds of atrial signal were 16% in MCG and 17% in SAECG. Reproducibility was best at 40-Hz filter and similar in patients and healthy subjects. In conclusion, the reproducibility of atrial signal variables in MCG is adequate and somewhat better than in SAECG and equal in patients with lone atrial fibrillation and healthy subjects. Magnetocardiography seems to be a potentially valuable method to evaluate features of atrial depolarization in patient studies.