The measurement of ammonia in biological fluids is the only way to diagnose and evaluate hepatic encephalopathy, but samples for ammonia measurement cannot be stored or sent by post. Two analysers for use in veterinary practice have recently become available, the VetTest and the Blood Ammonia Checker II; the reliability of ammonia measurements in canine blood with these two analysers has been evaluated by comparing the results with a standard automated enzymatic assay. Blood samples from 39 dogs, with a range of ammonia concentrations from 5 to 589 microM, were used simultaneously in the three assays. The blood samples were placed immediately on ice, and the measurements were made in duplicate. The intra-assay coefficients of variation were 13.7 per cent for the VetTest, 4.7 per cent for the Blood Ammonia Checker, and 2.8 per cent for the enzymatic assay. The correlation coefficients over the entire range of concentrations were 0.79 between the VetTest and the enzymatic assay, and 0.98 between the Ammonia Checker and the enzymatic assay. The ammonia concentrations recorded in the enzymatic assay were divided into 12 samples within the normal range (0 to 50 microM), 18 samples with moderately increased concentrations (51 to 150 microM), and nine samples with concentrations above 150 microM. No correlation or a poor correlation was found between the results from the VetTest and those from the enzymatic assay from 0 to 50 microM (R = 0.27) and from 50 to 150 microM (R = 0.51; P = 0.05). The results from the VetTest were only reliable in samples with the highest concentrations (R = 0.93; P < 0.05). In contrast, the results from the Ammonia Checker correlated well with the results from the enzymatic assay over all the ranges: R = 0.79 (P < 0.05) from 0 to 50 microM, R = 0.86 (P < 0.05) from 50 to 150 microM, and R = 1.00 (P < 0.05) in samples exceeding 150 microM.