Thanks to everyone who has responded so far. I'd like to provide a few more details since I'm also interested in the practical ramifications of the theoretical arguments presented thus far.
1. density map at 3.2 A resolution was used to refine a model that had already been built and refined at 2.0 A. 2. bound small molecule ligand was fit into the difference density using RSR within Coot (with tight geometric restraints as generated by ProDRG). 3. ligand was not refined any further using standard reciprocal space methods. 4. to my eyes, the ligand fit didn't look very good (even factoring in the resolution of the maps). 5. I would have started with RSR in Coot and then followed with something like Refmac, but was told the results wouldn't be any different. 6. debate ensues, starting with this quote I pulled from one of Michael Chapman's papers in *Acta Cryst.* (1999). D*55*, 464-468: "...real-space methods have been largely superseded by reciprocal-space methods owing to the implicit dependence of real-space methods upon the phases used to calculate the electron-density map, which are often much less accurate than the diffraction amplitudes." Thanks again for the responses -- this "gray-haired" crystallographer has enjoyed the commentary on a subject that sometimes gets taken for granted. Regards, Mike
