In order to compare different methods of fracture registration, we sought all nonvertebral fractures suffered during 8 years (1988-95) among 21,441 persons invited to a survey in 1979/80. We registered a total of 54 hip fracture cases through three separate sources (self-report, computer linkage to the local radiographic archives, discharge register), whereas forearm fractures (a total of 291 cases) were registered through two separate sources (self-report, computer linkage to the radiographic archives). The registration of fractures at other sites (a total of 1321 cases) were from one source (computer linkage to the local radiographic archives), and we have compared three ways of obtaining data from this single source (no ascertainment, ascertainment of records coded as fracture, ascertainment of all records). Ninety-three percent of all hip fractures and 97% of all wrist fractures in the entire study population were found by computer linkage to the radiographic archives, whereas the discharge register detected 87% of all the hip fractures. Computer linkage with ascertainment gave no overreporting of fractures. Among the 11,626 persons who answered a follow-up questionnaire in 1994/95, 97% (CI 84-100%) of all hip fractures and 72% (CI 66-78%) of all wrist fractures were self-reported. We conclude that a computerized search of radiographic archives is a viable method of fracture registration.