The accuracy of the 23Na nuclear magnetic resonance (NMR) method for measuring the sodium concentration in erythrocytes was tested by comparing the NMR results to those obtained by emission-flame photometry. Comparisons were made on aqueous solutions, hemolysates, gels, ghosts, and intact erythrocytes. The intra- and extracellular 23Na NMR signals were distinguished by addition of the dysprosium tripolyphosphate [Dy(PPP)7-2] shift reagent to the extracellular fluid. The intra- and extracellular volumes of ghosts and cells were determined by the isotope dilution method. Our results indicate that greater than 20% of the intracellular signal remains undetected by NMR in ghosts and cells. When the cells are hemolyzed, the amount of NMR-detectable sodium varies depending on the importance of gel formation. In hemolysates prepared by water addition, the NMR and flame photometry results are identical. The loss of signal in ghosts, cells, and undiluted hemolysates is attributed to partial binding of the Na+ ion to intracellular components, this binding being operative only when these components exist in a gel state. In a second part, 31P NMR was used to monitor the penetration of the shift reagent into the cells during incubation. Our data demonstrate that free Dy3+ can slowly accumulate inside the red cell.