The calculation of the anion gap is widely used in the diagnosis of metabolic acidosis. It is often taught that the increment in the anion gap will exactly match the fall in serum bicarbonate during a simple metabolic acidosis of the high anion gap type; if the changes in the anion gap and bicarbonate level are not equivalent, a second acid-base disorder should be suspected. The assumptions upon which this formulation is based are largely unsubstantiated. This review critically examines these assumptions and their clinical implications. Discrepancies between the increment in the anion gap and the reduction in serum bicarbonate must be interpreted cautiously.