See the previous posting http://stoney.phys.ualberta.ca/~e614/Projects/E614MEETINGS/00094/ for explanation of the procedure and MC parameters. To test Art's suggestion that angular smearing is important, the job was rerun without multiple scattering (MULS 0), the rest of parameters as before. Energy scale is reproduced correctly now, as you can see from the attached picture. Some speculations: theta_true which I use in analysis instead of a reconstructed angle comes from positron momentum just after muon decay. Since phase space volume is proportional to sin(theta), it's more probable that theta will increase due to multiple scattering. Energy loss is proportional to 1/|cos(theta_track)| which is systematically bigger than 1/|cos(theta_true)| There is also a second-order in delta(theta) effect: path length increase for a positive delta(theta) is bigger than corresponding decrease for a negative delta(theta). This also gives systematic bias towards bigger energy loss comparing to what is expected from 1/|cos(theta_true)| These two effects perhaps can explain, why the answer was wrong. However, shift of spectrum edge x_max in each angular bin can give either negative or positive bias to the energy scale depending on how the shift varies with theta. The observed bias was negative. Is there an explanation for this? I hope that use of reconstructed angle will give better energy calibration than use of MC theta_true, because reconstructed angle should be directly related to the track geometry and thus to the amount of material passed. (Of course, biases in reconstruction can badly affect the calibration...) Andrei