The aim of this randomized study was to compare the reliability of the treadmill test at constant-load (C-test, 3 km/hr; fixed grade of 12%) recommended in Germany with that of the graded-exercise test (G-test, 3 km/hr; increase in grade of 3.5% every 3 minutes) propagated in the United States. In 50 patients with an absolute claudication distance (ACD) in the C-test of between 50 and 400 m, the two treadmill tests were carried out in randomized order on one and the same day, and repeated on 3 days within 1 week. For the initial claudication distance (ICD), the intraclass correlation was 0.88 in the C-test and 0.87 in the G-test. For the ACD the coefficients were identical at 0.91. The within-subject variation (CVwithin) in the C-test and G-test was 25% and 27% for the ICD and for the ACD 24% and 21%, respectively. The between-subject variation was very similar with 72% and 73% (ICD) and with 78% and 68% (ACD). However, in ACDs below 100 m and between 100 to 150 m, the C-test showed significantly smaller coefficients of variation than the G-test: 13% vs 81% and 14% vs 50%, respectively. In conclusion, the results showed that both C-test and G-test are equally well reproducible.