Absolute 31P-NMR measurements of ATP, ADP and 2,3-diphosphoglycerate (2,3-DPG) in oxygenated and partly deoxygenated human erythrocytes, compared to measurements by standard assays after acid extraction, show that ATP is only 65% NMR visible, ADP measured by NMR is unexpectedly 400% higher than the enzymatic measurement and 2,3-DPG is fully NMR visible, regardless of the degree of oxygenation. These results show that binding to hemoglobin is unlikely to cause the decreased visibility of ATP in human erythrocytes as deoxyhemoglobin binds the phosphorylated metabolites more tightly than oxyhemoglobin. The high ADP visibility is unexplained. The levels of free Mg2+ [( Mg2+]free) in human erythrocytes are 225 mumol/l at an oxygen saturation of 98.6% and instead of the expected increase, the level decreased to 196 mumol/l at an oxygen saturation of 38.1% based on the separation between the alpha- and beta-ATP peaks. [Mg2+]free in the erythrocytes decreased to 104 mumol/l at a high 2,3-DPG concentration of 25.4 mmol/l red blood cells (RBC) and a normal ATP concentration of 2.05 mmol/l RBC. By increasing the ATP concentration to 3.57 mmol/l RBC, and with a high 2,3-DPG concentration of 24.7 mmol/l RBC, the 31P-NMR measured [Mg2+]free decreased to 61 mumol/l. These results indicate, that the 31P-NMR determined [Mg2+]free in human erythrocytes, based solely on the separation of the alpha- and beta-ATP peaks, does not give a true measure of intracellular free Mg2+ changes with different oxygen saturation levels. Furthermore the measurement is influenced by the concentration of the Mg2+ binding metabolites ATP and 2,3-DPG. Failure to take these factors into account when interpreting 31P-NMR data from human erythrocytes may explain some discrepancies in the literature regarding [Mg2+]free.