In very weak fields, the eddy-current losses in a lamination may be calculated exactly but the observed loss is always somewhat greater. Most of this extra loss varies with frequency and field in the same way as eddy-current loss, and if a quantity η equal to the ratio of the observed to calculated loss is defined it is found that it increases rapidly with decreasing thickness of sheet. It is suggested that this effect arises from neglect of the true mechanism of magnetization, and it is shown that the classical calculation of eddy-current loss is invalid when the distance between domain walls becomes comparable with the sheet thickness. Theoretical predictions seem to be in good agreement with observations made by Richards, Walker and Lynch.