Background. Microwave ablation has been used to replace the traditional incisions used in the surgical treatment of atrial fibrillation. However, dose-response curves have not been established in surgically relevant models. The purpose of this study was to develop dose-response curves for the Flex 10 (Guidant, Inc) microwave device in both the acute cardioplegia-arrested heart and on the beating heart. Methods. Twelve domestic pigs (40 to 45 kg) were subjected to microwave ablation in either the arrested (n = 6) or beating heart (n = 6). The cardioplegia-arrested heart was maintained at 10°to 15°C while six atrial endocardial and seven right ventricular epicardial lesions were created in each animal. On the beating heart, six right atrial and seven ventricular epicardial lesions were created. Ablations were performed for 15, 30, 45, 60, 90, 120, and 150 seconds (65 W, 2.45 GHz). The tissue was stained with 2,3,5-triphenyl-tetrazolium chloride, and sectioned at 5-mm intervals. Lesion depth and width were determined from digital micrographs. Results. Mean atrial wall thickness was 2.8 mm (range, 1 to 8 mm). In the arrested heart, 94% of atrial lesions were transmural at 45 seconds and 100% were transmural at 90 seconds. In the beating heart, only 20% of atrial lesions were transmural despite prolonged ablation times (90 seconds). Ventricular lesion width and depth increased with duration of application, and were similar on the arrested and beating hearts. Conclusions. Microwave ablation produces linear dose-response curves. Transmural lesions can be reliably produced on the arrested heart, but not consistently on the beating heart.