Immunoglobulin reactions to Salmonella dublin in serum and milk from 4 groups of lactating cows were measured by an indirect ELISA. The groups consisted of (1) cows that were natural carriers of S dublin in the mammary gland, (2) experimentally infected cows that did not become carriers, (3) cows inoculated with a commercial S dublin bacterin, and (4) cows used as S dublin-negative controls. Milk and serum samples were obtained at monthly intervals. Models for predicting carrier status were developed by use of stepwise logistic regression. Independent variables consisted of serum and milk IgG and IgM titers to S dublin lipopolysaccharide and a ratio of IgG to IgM. The utility of a single sample vs multiple samples obtained at 1-month or 2-month intervals was tested by comparison of goodness-of-fit chi 2 P values for 8 models predicting carrier status. Immunoglobulin reactions specific to S dublin were a significant predictor of carrier status (P less than 0.001). Serum IgG titers specific for S dublin were the most important variable for predicting carrier status. Two serum IgG titers to S dublin obtained 2 months apart was a better predictor of carrier status than measurement of the IgG:IgM ratio from a single serum sample. Immunoglobulin recognizing S dublin epitopes also were detected in milk samples. In milk, performing 2 ELISA 60 days apart to determine IgG and IgM reactions to S dublin appeared to be useful for the prediction of carrier status, but was not as accurate as models for serum immunoglobulin reactions.