ABSTRACT We derive empirical constraints on the nucleosynthetic yields of nitrogen by incorporating N enrichment into our previously developed and empirically tuned multizone galactic chemical evolution model. We adopt a metallicity-independent (‘primary’) N yield from massive stars and a metallicity-dependent (‘secondary’) N yield from AGB stars. In our model, galactic radial zones do not evolve along the observed [N/O]–[O/H] relation, but first increase in [O/H] at roughly constant [N/O], then move upward in [N/O] via secondary N production. By t ≈ 5 Gyr, the model approaches an equilibrium [N/O]–[O/H] relation, which traces the radial oxygen gradient. Reproducing the [N/O]–[O/H] trend observed in extragalactic systems constrains the ratio of IMF-averaged N yields to the IMF-averaged O yield of core-collapse supernovae. We find good agreement if we adopt $y_\text{N}^\text{CC}/y_\text{O}^\text{CC}=0.024$ and $y_\text{N}^\text{AGB}/y_\text{O}^\text{CC} = 0.062(Z/Z_⊙)$. For the theoretical AGB yields we consider, simple stellar populations release half their N after only ∼250 Myr. Our model reproduces the [N/O]–[O/H] relation found for Milky Way stars in the APOGEE survey, and it reproduces (though imperfectly) the trends of stellar [N/O] with age and [O/Fe]. The metallicity-dependent yield plays the dominant role in shaping the gas-phase [N/O]–[O/H] relation, but the AGB time-delay is required to match the stellar age and [O/Fe] trends. If we add ∼40 per cent oscillations to the star formation rate, the model reproduces the scatter in the gas phase [N/O]–[O/H] relation observed in external galaxies by MaNGA. We discuss implications of our results for theoretical models of N production by massive stars and AGB stars.