We investigate the red supergiant (RSG) content of the SMC and LMC using multi-object spectroscopy on a sample of red stars previously identified by BVR CCD photometry. We obtained high accuracy (< 1 km s −1 ) radial velocities for 118 red stars seen towards the SMC and 167 red stars seen towards the LMC, confirming most of these (89% and 95%, respectively) as red supergiants (RSGs).Spectral types were also determined for most of these RSGs. We find that the distribution of spectral types is skewed towards earlier type at lower metallicities: the average (median) spectral type is K5-7 I in the SMC, M1 I in the LMC, and M2 I in the Milky Way. Our examination of the Kurucz Atlas 9 model atmospheres suggests that the effect that metallicity has on the appearance on the TiO lines is probably sufficient to account for this effect, and we argue that RSGs in the Magellanic Clouds are 100 • K (LMC) and 300 • K (SMC) cooler than Galactic stars of the same spectral types. The colors of the Kurucz models are not consistent with this interpretation for the SMC, although other models (e.g., Bessel et al.) show good agreement. A finer grid of higher-resolution synthetic spectra appropriate to cool supergiants is needed to better determine the effective temperature scale. We compare the distribution of RSGs in the H-R diagram to that of various stellar evolutionary models; we find that none of the models produce RSGs as cool and luminous as what is actually observed. This result is much larger than any uncertainty in the effective temperature scale. We note that were we to simply adopt the uncorrected Galactic effective scale for RSGs and apply this to our sample, then the SMC's RSGs would be under luminous compared to the LMC's, contrary to what we expect from stellar evolution considerations.In all of our H-R diagrams, however, there is an elegant sequence of decreasing effective temperatures with increasing luminosities; explaining this will be an -3important test of future stellar evolutionary models. Finally, we compute the blue-to-red supergiant ratio in the SMC and LMC, finding that the values are indistinguishable (∼ 15) for the two Clouds. We emphasize that "observed" B/R values must be carefully determined if a comparison to that predicted by stellar models is to be meaningful. The non-rotation Geneva models overestimate the number of blue-to-red supergiants for the SMC, but underestimate it for the LMC; however, given the inability to produce high luminosity RSGs in the models that match what is observed in the H-R diagram, such a disagreement is not surprising.