An indirect enzyme-linked immunosorbent assay (ELISA) for Brucella abortus antibodies detection in bovine milk and serum samples was validated. The assay use B. abortus smooth lipopolysaccharide as antigen, immobilized on a polystyrene matrix; milk diluted 1:2 or serum diluted 1:50, in a buffer containing divalent cation chelating agents EDTA and EGTA (ethyleneglycol-bis-aminoether-N,N,N',N'-tetraacetic acid) to reduce non-specific reactions; and a mouse monoclonal antibody specific for an epitope of bovine IgG1, conjugated with horseradish peroxidase. A total of 2646 sera and 2119 milk samples from cows older than 24 months were obtained from 12 brucellosis-free herds for at least the previous 5 years. Milk samples were obtained in parallel with serum samples. The remaining 527 serum samples were from dry cows. All cattle were vaccinated with B. abortus strain 19 between 3-10 months of age. Five hundred and fifty-two milk samples and 562 serum samples were obtained from 6 infected herds with abortions where B. abortus was isolated at least once no more than 6 months before sampling. The complement-fixation test (CFT) on serum samples was considered the gold standard. Serum samples were also tested with the official screening test: the buffered plate antigen (BPA) test. The cut-off point was determined using receiver-operating characteristic (ROC) analysis. For milk samples, it was fixed at 36 percent positivity (PP) giving a sensitivity of 99.6% with a 95% confidence interval (CI) of 98.6-99.9%. The specificity was 99.1% (CI 98.9-99.4%). For serum samples, the cut-off was fixed at 53 PP giving a sensitivity of 99.6% (CI 98.6-99.9%) and a specificity 98.6% (CI 98-99%). The BPA test showed a relative sensitivity of 99.6% (CI 98.6-99.9%) and a relative specificity of 98.6% (CI 98.1-99%). Our results indicate that the indirect ELISA is a highly sensitive and specific test and can be adapted to process a large number of samples.