Dear Shelly, Dear Bruce Thanks for both your answers. Going through Shelly's slides was a nice refresher course. Bruce, ok, I see why normalization is not affected. However.... The data were already correctly aligned. I've now explicitly tied the data to the standard, using the Data menu, and the auto-align shift was only 0.02 eV. So this had no effect on the spike. Setting the lower bound of the spline range to 0 mainly shifts the position of the spike to near k=0, and so hides it somewhat when we multiply chi by a power of k. But it's still visible in E space (a large negative peak in the spline). Setting a spine clamp at the low end changes things slightly but not qualitatively. Changing E0 would seem to be a reasonable step. After all, there is no reason why k=0 should coincide exactly with the first inflection point on the edge!! But again, all attempts, for Eshift varying from -5 to 10eV, make no qualitative difference. All other tweaks I've tried also maintain the spike, or else distort the spline so much as to introduce other errors. For comparison, I took the first curve (labeled 0.12) in project AuChloride_LCF_example.prj from Scott's talk. There is something of a sharp departure between the spline and the data visible in E space, as in my case. But the plot in k space -- even of chi itself without multiplying a power of k -- is nice and smooth. All this is in http://www.msri.org/~levy/files/example.prj which contains only one group from me and one from Scott. Would you be willing to take a closer look? I'd not be asking except that I'm completely stumped. My data seem perfectly normal. Thanks again, Silvio