Sandwich ELISAs have become a widely used method for the quantitative detection of serum proteins. However, they can be biased by a variety of interfering substances. As reported recently, we observed false-positive levels of interferon (IFN)-alpha and -beta in up to 27% of sera from healthy blood donors using commercial ELISAs. We now demonstrate that two different groups of naturally occurring heterophilic antibodies (IgG-type) are responsible for these titers. Group I (representing 85% of positive samples) binds to the Fab region of IgG from goat, mouse, rat, horse, and bovidae (but not rabbit). Group II (15%) recognizes an epitope in the Fc region of mouse, horse, bovine, and rabbit (but not goat or rat) immunoglobulins. The antibodies did not crossreact with human IgG subclasses but contributed to false-positive IgG rheumatoid factor levels obtained using a commercially available ELISA. To investigate the susceptibility of assays to these artifacts, various combinations of capture and detection antibodies have been tested. On this basis, we defined the relative risks that standard ELISAs might be influenced by heterophilic anti-immunoglobulin antibodies. In general, assays that use monoclonal antibodies for both capture and detection are less susceptible than others which include at least one polyclonal antiserum. However, only systems utilizing rabbit F(ab')(2) fragments have been found to be immune to this interference.