We evaluated the performance of the Du Pont aca ammonia procedure with regard to (a) linearity, (b) precision, (c) interferences, and (d) effect of anticoagulants. Linearity extends to 2,000 mumol/L. The SD of the method was essentially constant (2 to 3 mumol/L) and independent of the NH3 concentration. Hemoglobin, bilirubin, and lipemia do not interfere. Either EDTA or heparin is suitable as anticoagulant. Recovery of NH3 added to plasma samples averaged 102% (range: 97--107%). We established normal values by measuring NH3 in 188 plasma samples from apparently healthy individuals. The 95% confidence range is from 10 to 35 mumol/L. The aca ammonia method compares very well with that of Kingsley and Tager but correlates less strongly with that of Reinhold and Chung. We describe a protein-based solution with stable NH3 concentration that is suitable as a control material.