Hysteresis loops taken at a surface are often found to change even when the probing depth of the magnetic measurement varies by only a few lattice spacings. We report an interpretation of these changes in the framework of a simple model which assumes that only one sheet of thickness d at the surface is magnetically different from the bulk. The model reproduces the experimental hysteresis loops with a variety of materials. One obtains the magnetic anisotropy Ks of the surface and the exchange stiffness A1,2 along a path perpendicular to the surface. For Fe(100), Ks and A1,2 are found to be very small; furthermore, the average easy direction of the surface magnetization is at approximately 75°to the one in the bulk. © 1988 The American Physical Society.