To evaluate the mechanism of excimer laser recanalization and compare the results with those of laser-assisted thermal probe recanalization and mechanical recanalization, a total of 42 human atherosclerotic totally occluded arterial segments (2-15 cm long) were recanalized by excimer laser with a 400-800 micron quartz fiber pulsed at 20 Hz with 50 mJ/mm2 of energy (n = 21), an Argon heated thermal probe at 10-12 watts (n = 11), a guidewire directed through a 6 Fr multipurpose catheter, or an angioplasty balloon catheter (n = 10). On histologic examination, the excimer laster created a single round lumen or multiple lumens ("Swiss-cheese" like appearance) with no evidence of thermal injury at the perimeter of the lumen. The incidence of perforation in vitro was less with an excimer laser catherter (8/21 or 38%) than with the thermal prove (10/11 or 91%) (p less than 0.01). However, serial histologic cross-sectional examination showed that the pathway of the devices were essentially the same in all recanalization procedures. The pathway of the device was located outside the atheroma but proximal to the internal elastic membrane in 13 arteries with the excimer laser (62%), in 10 arteries with the thermal probe (91%), and 8 arteries with mechanical devices (80%). These results indicate that although the eximer laser could recanalize human atherosclerotic arteries without thermal injury, the fiber frequently deflected around firm atherosclerotic plaque and advanced in a dissection plane between the plaque and media. A similar course was noted for the thermal probe or during mechanical recanalization with a guidewire and catheter. To insure the safety of an excimer fiber or a thermal probe to reopen complete occlusions, better guidance systems must be developed.