Hello Everyone Thanks for all the responses. Many were about twinning detection. IMO, the best tip was toward the cctbx utilities (I think these come from the phenix developers??) for twinning detection and detwinning. The output from mmtbx.xtriage and mmtbx.twin_map_utils is particularly clear. (Thanks to Peter Zwart and Bill Scott) In the particular case of the (not so) hypothetical dataset I was concerned about - that dataset is not twinned, rather the struck refinement was due to severe data anisotropy [not noticed during late night synchrotron data reduction but detected with sfcheck - we hadn't yet figured that out when I posted the message.] A run through the anisotropy server at UCLA http://www.doe-mbi.ucla.edu/~sawaya/anisoscale/ has unstuck the refinement and made the maps much more pleasing. Sue Sue Roberts Biochemistry & Biopphysics University of Arizona [log in to unmask]