Thirty kidneys containing renal carcinoma became available for complete histological examination 2-8 days after therapeutic embolisation with various combinations of gel-foam, dura mater, steel coils and thrombin. There was extensive tumour infarction (greater than 50%) in 25 kidneys and in the other five kidneys there was no tumour infarction detected histologically. In the 24 kidneys with a single renal artery, the number with tumour infarction exceeding 50% was significantly higher in cases where the post-embolisation arteriogram had shown complete occlusion of the renal artery close to its origin than in cases where the renal artery had been incompletely occluded (P = 0.0016). Fragments of gel-foam or dura mater were significantly less likely to produce complete arterial occlusion on the post-embolisation arteriogram than mixtures of embolic material containing coils (P = 0.016) or thrombin (P = 0.042). In the six kidneys with two renal arteries, tumour infarction exceeded 50% in the five kidneys where the post-embolisation arteriogram had shown complete occlusion of both renal arteries, and there was no tumour infarction in the single case where an accessory renal artery had been left patent. It is concluded that complete occlusion of the renal artery (or arteries) is essential if tumour infarction in excess of 50% is to be guaranteed, and that this is more readily achieved with mixtures of embolic materials containing fragments of gel-foam or dura mater plus coils or thrombin than with fragments of gel-foam or dura mater on their own.