Departing from the historical background of scientific interest in soil fertility and sustainable agricultural production in sub-Saharan Africa, a review is conducted of nutrient budget studies carried out in semi-arid West Africa at scales ranging from individual fields to the sub-continent. For both, nitrogen and phosphorus, the comparison discloses largely diverging balances calculated for similar agro-ecosystems. In a first step, the modes of calculation of the nutrient budgets are examined. It is demonstrated that the calculations used in the different studies differ by the variables and biophysical processes taken into account, and by the choice of spatial scales as well as related time scales. One important discrepancy between approaches is whether and to which extent nutrient flows are internalized when upscaling. The extent to which the impact of individual and communal management, especially of pastoral and forestry resources, on nutrient flows is accounted for is a second cause of divergence. Moreover, it was observed that nutrient budgets tend to be increasingly negative as the spatial scale of the study increases from farm to sub-continent. This unexpected trend is traced back to the lack of internalization of nutrient flows when upscaling. The complexity of the scale patterns of nutrient flows and that of the interactions and the tradeoffs in the effects of management calls for the use of models to calculate nutrient budgets. Therefore, in a second step, examples of a static model, a multiple-goal linear programming model and a decision rules model were reviewed, all of which include the calculation of nutrient flows and balances and which were applied to West-African farming systems. The models are analyzed for their way of dealing with the critical issues of spatial and temporal scales and the impact of resource management on nutrient flows, taking into account that they have different objectives and were designed for different spatio-temporal scales. To conclude, suggestions were made for strengthening the use of models as tools enabling ex-ante testing of alternative agricultural technologies and policies that could improve soil nutrient balances in semi-arid sub-Saharan Africa.