Abstract Soil loss due to crop harvesting (SLCH) has been established as an important soil erosion process that has significantly contributed to soil degradation in highly mechanised agriculture. This has stimulated the need to investigate the importance of this process of erosion under low input agriculture where, until now, only water and tillage erosion are known as important phenomena causing soil degradation. This study was conducted in Eastern Uganda with the following objectives: (1) to assess the amount of soil lost due to the harvesting of cassava roots and sweet potato tubers under low input agriculture, (2) to look into the factors that influence variations in these soil losses, and (3) to estimate the amount of plant nutrients lost due to SLCH for cassava and sweet potato. Soil sticking to roots and tubers was washed and the soil suspension oven dried to estimate the amount of soil lost after harvesting. Mean annual soil loss for cassava was 3.4 tonnes ha −1 and for sweet potato was 0.2 tonnes ha −1. Ammonium acetate lactate extractable soil nutrient losses for cassava were N = 1.71 kg ha −1 harvest −1, P = 0.16 kg ha −1 harvest −1, K = 1.08 kg ha −1 harvest −1 and for sweet potato were N = 0.14, P = 0.01 kg ha −1 harvest −1, K = 0.15 kg ha −1 harvest −1. Difference in soil loss due to crop harvesting for cassava and sweet potato could be due to: (1) smaller yields of sweet potato leading to smaller soil losses on an area basis, (2) smoother skin and less kinked morphology of sweet potato that allowed less soil to adhere, and (3) the fact that sweet potato is planted in mounds which dry out faster compared to the soil under cassava. Soil moisture content at harvesting time and crop age were significant factors that explained the variations in the soil lost at cassava harvesting. Soil loss under cassava justifies the need to conduct further investigations on this process of soil erosion under low input agriculture.