This paper discusses the strengths and weaknesses of three different methods for quantifying prehistoric soil erosion. Method A estimates erosion by determining the amount of colluvium stored downhill. Method B involves reconstructing a former erosion surface using truncated soil horizons. Method C compares the elevation of a paleosol beneath an earthwork with the modern surface in the surrounding area. Each method was applied to a Celtic earthwork (Viereckschanze) at Poign (near Regensburg) in Southern Germany in order to cross-check the different results. For an erosion area of 3.6 ha and during 300 years of agricultural usage, Method A calculates a minimum erosion rate of 20.8 t/ha/a. Method B computes 10 t/ha/a of soil loss. Method C yields the highest rate of erosion with 24.2 t/ha/a. We have confidence in Method C, which implies an underassesment of soil loss in using methods A or B. © 2007 Wiley Periodicals, Inc.