A GOES M1.5 solar flare was observed in NOAA AR 10652 on 2004 July 27 around 20:00 UT with the Coronal Diagnostic Spectrometer (CDS) aboard the Solar and Heliospheric Observatory (SOHO) spacecraft. Images obtained with SOHO's Extreme-ultraviolet Imaging Telescope and with the Transition Region And Coronal Explorer satellite show that the CDS slit was positioned within the flare, whose emission extended 1 arcmin along the slit. Rapid cadence (9.8 s) stare spectra obtained with CDS include emission from the upper chromosphere (He I at 584.3 Å), transition region (O V at 629.7 Å), corona (Si XII at 520.7 Å), and hot flare plasma (Fe XIX at 592.2 Å), and reveal that (1) the flare brightened in its southern parts before it did so in the north; (2) chromospheric evaporation was "explosive" during the first rapid intensity increase observed in Fe XIX, but converted to "gentle" during the second; (3) chromospheric evaporation did not occur in the northern portion of the flare observed by CDS: the brightening observed there was due to flare material moving into that location from elsewhere. We speculate that the initial slow, steady increase of Fe XIX intensity that was observed to start several minutes before its rapid increase was due to direct coronal heating. The change from explosive to gentle evaporation was likely due to either an increased absorption of beam energy during the gentle event because the beam passed through an atmosphere modified by the earlier explosive event, or to a weakening of the coronal magnetic field's ability to accelerate nonthermal particle beams (via reconnection) as the flare progressed, or both.