A total of 110 total knee replacements (TKRs) was randomised to receive either a cemented or an uncemented prosthesis. Postoperative venography at five to seven days was used to compare the prevalence, site and size of deep-vein thrombosis (DVT). We also compared the findings with those of postoperative venography in a group of patients with cemented total hip replacements (THRs). The total prevalence of DVT was significantly greater after uncemented (81%) than after cemented TKR (55%). Both knee groups had a significantly higher prevalence of DVT than in cemented hip replacements (32%). We found no difference in the proportion with proximal DVT in the three groups (14%, 15% and 16%). The median length of the thrombi was significantly greater after cemented (26.5 cm) than after uncemented TKR (11 cm) or after cemented THR (7 cm). This difference was mainly due to greater lengths of distal rather than proximal thrombi. We conclude that the use of cement may affect the formation of DVT after joint replacement, but does not appear to lead to an increased incidence