Measurements of the intensity ratio of the 589.0/589.6 nm sodium doublet in the terrestrial nightglow over an 8-year period, involving >300 separate determinations, have established that it is variable, the value RD = I(D2)/I(D1) lying between 1.2 and 1.8. Sky spectra from the Keck I telescope with the High-Resolution ¿chelle Spectrometer (HIRES) ¿chelle spectrograph and the Keck II telescope with the ¿chellette Spectrograph and Imager (ESI) ¿chelle spectrograph were used in this analysis. The result contrasts with the accepted view, from earlier measurements at midlatitude, that the ratio is 2.0, as expected on statistical grounds. The lack of dependence of the ratio on viewing elevation angle, and hence Na slant column, allows self-absorption to be ruled out as a cause of the variability. The data suggest a semiannual oscillation in the ratio, maximum at the equinoxes and minimum at the solstices. Airborne measurements over the North Atlantic (40¿--50¿N) in 2002 show an even larger range in the nightglow ratio and no correlation with the upper mesospheric temperature determined from the OH 6--2 bands. A laboratory study confirms that the ratio does not depend on temperature; however, it is shown to be sensitive to the /2> ratio. It is therefore postulated that the variable ratio arises from a competition between O reacting with NaO(A3Σ+), produced from the reaction of Na with O3, to yield D-line emission with a D2/D1 ratio greater than about 2.0, and quenching by O2 to produce NaO(X2Π), possibly with vibrational excitation, which then reacts with O to produce emission with a ratio of less than 1.3.