We compared two methods of ascertaining mortality in a historical prospective mortality study. Computerized Record Linkage (CRL) with the centralized historical Canadian Mortality Data Base (CMDB) was carried out on 2469 men and an attempt was also made to trace the subjects by individual follow-up (IFU). All but 88 were traced and 60 were reported to be dead. CRL was able to locate the deaths of three men who had been untraced by IFU. Contradictory information on vital status was obtained on 5 subjects--in 4 of them, the discrepancy was resolved in favour of CRL. Overall, CRL using the CMDB performed very well. We also consider factors that affect the relative costs of the two methods, which should be balanced against the accuracy of information obtained.