The radioisotope 137Cs has been extensively used to provide information about soil erosion. The technique relies upon a number of assumptions that this review evaluates in order to establish what can be learned about soil erosion from the use of 137Cs. The assumption of locally spatially uniform fallout is broken down into three components: (1) atmospheric fallout is spatially uniform; (2) transfer to the soil is spatially uniform; and (3) no redistribution occurs during the transfer process. Evidence is presented to show that none of these assumptions is valid. Similarly, it is shown that the literature that continues to examine both the processes of adsorption of 137Cs onto soil particles, the conditions under which it may be desorbed and the uptake of 137Cs by plants, makes the assumption that 137Cs is rapidly and irreversibly adsorbed onto soil particles contestable. Evidence is presented for the movement of 137Cs across the landscape other than by movement of soil particles across the surface. Methods used to convert 137Cs inventories into rates of erosion are reviewed, as are errors associated with obtaining 137Cs inventories. Finally, the relationship of 137Cs inventories to current understanding of soil-erosion processes is assessed. It is concluded that that no current rates of soil erosion that are based upon the use of this technique are reliable, and that 137Cs cannot be used to provide reliable information about rates of soil erosion.