We have a stated energy resolution in the BEMC NIM paper (1.5% + 14%/sqrt(E)), but I've never seen a study to verify it, so I decided to check it using electrons from the 2006 data (all productions). I broke the electrons into bins based on their track momentum with bin edges at {1.5, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0}.  Additionally, the electrons had to enter and exit the same tower and fall within a fiducial area 0.02 x 0.02 in eta-phi space on the face of the tower (out of 0.05 x 0.05).

For each bin in electron momentum, I plotted E/p and fit the distribution to a gaussian (see attached pdf). The sigma of the gaussian should be the energy resolution, as long as the momentum resolution is small. We can see this assumption fails in the higher bins from the plot below. I fit the resulting points to a function r^2 = a^2 + b^2/E over the momentum range [1.5,6.0]. We find that a = 11.0% and b = 15.4%. These numbers do not agree with the values stated in the NIM paper, but the momentum resolution begins to have a significant effect in the last couple fit bins, which cause the resolution to fall more slowly.