We examined 76 patients suffering from Staphylococcus aureus septicemia, from 1976 to 1979, none of whom were drug addicts. In contrast to other studies, we found a significant incidence of S aureus endocarditis in patients with a demonstrable portal of entry, a relatively high incidence of metastatic foci in patients without endocarditis, and endocarditis possibly as frequent in patients with hospital-acquired infection as in those with community-acquired infection. These findings suggest caution in determining the length of therapy for S aureus septicemia based on clinical grounds alone.