Much research has been undertaken that seeks to understand the crop productivity response to soil erosion. Reported effects appear to be inconsistent with respect to both the magnitude of the response and shape of the response curve. This study was conducted to examine whether general patterns emerge when the results of experimental studies on soil loss are combined and compared. Results from a number of studies that relate crop productivity to erosion were collected and quantified. Important variables of a methodological or physical nature were identified. Both the magnitude and shape of the response curves were related to these variables. It appears that the experimental methodology has an overwhelming effect on the magnitude of the crop productivity response to soil erosion. The comparative-plot method showed an average reduction in crop productivity of 4.3% per 10 cm of soil loss, whereas the reduction averaged 10.9% for studies based on the transect method and 26.6% for desurfacing experiments. Physical variables affected the shape of the response curve: water deficit and physical root hindrance produced convex curves, whereas nutrient deficit resulted in linear to concave curves. The available data did not allow identification of any significant effect of the physical variables on the magnitude of the response, nor an effect of the research-method on the shape of the response curve. It is assumed that the desurfacing and transect methods overestimate the effect of soil erosion because (a) desurfacing experiments result in much stronger changes in soil properties than soil erosion that takes place gradually, and (b) transect methods often ¿include¿ effects of other processes that are related to topography. If this assumption is correct, then yield reductions of approximately 4% per 10 cm of soil loss should be considered realistic. Where nutrient deficits are avoided by fertilization, response curves are generally convex, implying that reductions will become increasingly severe with further erosion.