An Indiana capacitor-manufacturing cohort (n=3,569) was exposed to polychlorinated biphenyls (PCBs) from 1957 to 1977. The original study of mortality through 1984 found excess melanoma and brain cancer; other studies of PCB-exposed individuals have found excess non-Hodgkin lymphoma and rectal, liver, biliary tract, and gallbladder cancer. Mortality was updated through 1998. Analyses have included standardized mortality ratios (SMRs) and 95% confidence intervals (CIs) using rates for Indiana and the United States, standardized rate ratios (SRRs), and Poisson regression rate ratios (RRs). Estimated cumulative exposure calculations used a new job-exposure matrix. Mortality overall was reduced (547 deaths; SMR, 0.81; 95% CI, 0.7-0.9). Non-Hodgkin lymphoma mortality was elevated (9 deaths; SMR, 1.23; 95% CI, 0.6-2.3). Melanoma remained in excess (9 deaths; SMR, 2.43; 95% CI, 1.1-4.6), especially in the lowest tertile of estimated cumulative exposure (5 deaths; SMR, 3.72; 95% CI, 1.2-8.7). Seven of the 12 brain cancer deaths (SMR, 1.91; 95% CI, 1.0-3.3) occurred after the original study. Brain cancer mortality increased with exposure (in the highest tertile, 5 deaths; SMR, 2.71; 95% CI, 0.9-6.3); the SRR dose-response trend was significant (p=0.016). Among those working >or= 90 days, both melanoma (8 deaths; SMR, 2.66; 95% CI, 1.1-5.2) and brain cancer (11 deaths; SMR, 2.12; 95% CI, 1.1-3.8) were elevated, especially for women: melanoma, 3 deaths (SMR, 5.99; 95% CI, 1.2-17.5); brain cancer, 3 deaths (SMR, 2.87; 95% CI, 0.6-8.4). These findings of excess melanoma and brain cancer mortality confirm results of the original study. Melanoma mortality was not associated with estimated cumulative exposure. Brain cancer mortality did not demonstrate a clear dose-response relationship with estimated cumulative exposure.