Accurate measurements have been made of the 23Na nuclear magnetic resonance spectrum of sodium nitrate single crystals from -190 °C to 300 °C in an applied magnetic field of 6000 Oe. The nuclear electric quadrupole coupling constant has a magnitude which is convenient for the separation of first- and second-order shifts of the resonance lines. The separation of the two satellite lines, to which second-order perturbations do not contribute, follows the theoretical first-order variation closely. The frequency of the central line, and the mean frequency of the satellite lines to which first-order perturbations do not contribute, accurately follow the theoretical behaviour using the parameters determined from the first-order examination. The magnitude and angular variation of the second moment of the central line at room temperature is in excellent agreement with the theory of Kambe and Ollom and provides the first experimental test of their theory. At all orientations of the crystal the second moment is approximately 10% smaller than would be expected if the quadrupole coupling constant were zero, when Van Vleck's expression would be applicable. The width of the line decreased over the range 200° to 280 °C to a small value; this reduction is ascribed to diffusion of the sodium ions. The quadrupole coupling constant was found to decrease by a factor of ten over the whole range of temperature, with a discontinuity in the temperature dependence at the second-order phase transition, 276 °C. This unusually large decrease is much greater than can be explained by the effects of lattice expansion or the tendency of the crystal structure to become somewhat more nearly cubic. The behaviour must clearly be associated with the changes in the crystal which lead to the second-order phase transition. Experimental information bearing on the nature of the phase transition is reviewed. It is shown that the evidence can be reconciled by a dynamic-disorder model in which the nitrate groups reorient about their threefold axes with increasing rapidity as the temperature rises. Although this motion does not become free it encourages the disordering of orientation of the nitrate groups leading to a phase-transition of the order-disorder type. This motion does not however directly account for the reduced electric field gradient experienced by the sodium nuclei. It is proposed that this reduction arises from the oscillation of the sodium ions predominantly in directions perpendicular to the triad axis of the unit cell, and that this motion is a secondary consequence of the mechanism of the phase transition. Calculations have been made of the electric field gradient, and these support the proposed mechanism of dynamic disorder.