It is normal to regard the loss in a ferromagnetic material undergoing magnetization as the sum of a hysteresis loss, which might be measured by a ballistic method, and an eddy-current loss, calculable from a knowledge of the shape and resistivity of the material and the flux changes involved. For some sheet steels the measured loss is considerably in excess of this sum, and it is with this discrepancy that the paper is concerned. A pulse method of excitation is used, a constant voltage being applied suddenly to the specimen for times ranging from 0.2 to 2.5 microsec. It has been found that, in most instances,B/Hcurves are either linear or can be closely approximated by functions of the formB=B̂(1−ε−αH) orB=B̂(ε−αH−1), whereB̂and α are constants. By adopting these exponential expressions, accurate calculation of eddy-current losses in the non-linear regions has been made possible. In the paper, several theories of the cause of the discrepancy are suggested and examined in the light of experimental evidence in which the rate of change of mean flux density, the polarizing field, the grain orientation and the lamination thickness are varied. It is concluded that the discrepancy is at least partly caused by a variation of permeability and resistivity across the lamination thickness; but it is also shown that a reduction of the effective permeability, due to the inability of the elementary irreversible changes of magnetization (Barkhausen jumps) to be completed in the time available, could also contribute to a “discrepancy loss” which would vary with the factors above in a manner consistent with the experimental results.