We have obtained long-slit spectra of NGC 253 in the J, H, K, and N bands, broadband images in the J, H, and Ks bands, narrowband images centered at the wavelengths of BrGamma and H2(1,0)S(1), and imaging spectroscopy centered on [NeII](12.8um). We use these data and data from the literature in a comprehensive re-assessment of the starburst in this galaxy. We derive the supernova rate from the strength of the infrared [FeII] lines. We find that most of the H2 infrared luminosity is excited by fluorescence in low density gas. We derive a strong upper limit of ~37,000K for the stars exciting the emission lines. We use velocity-resolved infrared spectra to determine the mass in the starburst region. Most of this mass appears to be locked up in the old, pre-existing stellar population. Using these constraints and others to build an evolutionary synthesis model, we find that the IMF originally derived to fit the starburst in M 82 (similar to a Salpeter IMF) also accounts for the properties of NGC 253. The models indicate that rapid massive star formation has been ongoing for 20-30 million years in NGC 253---that is, it is in a late phase of its starburst. We model the optical emission line spectrum expected from a late phase starburst and demonstrate that it reproduces the observed HII/weak-[OI] LINER characteristics.