Background: The magnetic resonance technique of arterial spin labeling (ASL) allows myocardial perfusion to be quantified without the use of a contrast agent. This study aimed to use a modified ASL technique and T1 regression algorithm, previously validated in canine models, to calculate myocardial blood flow (MBF) in normal human subjects and to compare the accuracy and repeatability of this calculation at 1.5 T and 3.0 T. A computer simulation was performed and compared with experimental findings. Results: Eight subjects were imaged, with scans at 3.0 T showing significantly higher T1 values (P > 0.001) and signal-to-noise ratios (SNR) (P < 0.002) than scans at 1.5 T. The average MBF was found to be 0.990 ± 0.302 mL/g/min at 1.5 T and 1.058 ± 0.187 mL/g/min at 3.0 T. The repeatability at 3.0 T was improved 43% over that at 1.5 T, although no statistically significant difference was found between the two field strengths. In the simulation, the accuracy and the repeatability of the MBF calculations were 61% and 38% higher, respectively, at 3.0 T than at 1.5 T, but no statistically significant differences were observed. There were no significant differences between the myocardial perfusion data sets obtained from the two independent observers. Additionally, there was a trend toward less variation in the perfusion data from the two observers at 3.0 T as compared to 1.5 T. Conclusion: This suggests that this ASL technique can be used, preferably at 3.0 T, to quantify myocardial perfusion in humans and with further development could be useful in the clinical setting s an alternative method of perfusion analysis.