Fat embolism was assessed at necropsy and correlated with clinical findings in the patients who died among 854 with fractured hips admitted to hospital between 1967 and August 1971. Sixteen cases of clinical importance were found, eight of which were judged to have been fatal or to have seriously contributed to death. Frequencies were as follows: 2.4 to 3.3% among 424 patients with subcapital fractures; 0.7 to 0.8% in the 405 with trochanteric fractures; 4.1 to 7% among subjects treated without operation, representing 30% of those who died within seven days; and 0.9 to 1.1% among patients treated by pinning, nailing, or nail-plating. The higher frequency in the conservatively treated group is probably related to selection of poor-risk subjects. Fat embolism was found in 6.8 to 8.0% of those with subcapital fractures treated by primary Thompson's arthroplasty which utilizes acrylic cement, and in none of those given Moore's prostheses for which cement is not used. Study of a larger group after Moore's prosthesis is required to establish its lack of special risk. Fat embolism accounted for all the deaths within seven days of Thompson's arthroplasty and for most within 14 days; it was clearly related to surgery in some cases.A possible explanation of the hazard of Thompson's arthroplasty is that fat globule entry is enhanced by a rise of intramedullary pressure due to proximal occlusion of the reamed marrow cavity. A controlled trial of the effect of venting the marrow cavity on the frequency of fat embolism is warranted. It is possible that the acrylic monomer may also contribute to venous entry of medullary fat. The higher-age group of those with subcapital fractures and associated chronic cardiac and pulmonary disease might make them more susceptible to fat embolization than those in whom arthroplasty is also carried out for chronic hip disease.