I ran into some mysterious crashes of R and when stepping through the code it appeared to me that a call to GScale(0.0,1.0,1,dd) (as e.g. in do_plot_new(), plot.c:458 ) possibly with min=-0.04 runs into (graphics.c:1911) Rf_gpptr(dd)->logusr[0] = Rf_dpptr(dd)->logusr[0] = log10(min); which should end up with a domain error. As I cannot reproduce this fully, I asked myself if there is some mysterious handling of this situation. Can anybody explain this to me? Thomas.