The number of recognized occupational fatalities in the United states counted by the Burea of Labor Statistics (BLS) Annual Survey, the most widely known source, is grossly underreported. For 1984, the BLS reported 3.740 occupational deaths, the National Institute for Occupational Safety and Health reported 6,258, and the National Safety council estimated 11,500. We analyzed the 1984 BLS survey counts for 26 stages, by comparing them to death certificate data for trauma, and to reported deaths to workers compensation. The BLS survey, based on self-reporting by employers, showed underreporting in almost all states, which in some cases was severe. Fatalities in small firms employing < 11 workers were particularly underrepresented. The low BLS total is not explained by a smaller number of workers covered in the survey than in the other published figures. By using existing data sources, we estimate the minimum number of recognized occupational deaths in the United States in 1984 at 7,771. Adjusting the number from death certificates for underreporting gives an estimate of 8,234 from trauma. When corrected for underreporting, death certificate data should be a better basis for counting occupational deaths than the BLS survey.