Weed control for the West African Sahel rainfed crops is done mainly manually, resulting in high labor requirements. Because of the seasonality of rainfed farming, weed control is often late and incomplete, resulting in considerable losses in crop yield. We examined the case of weed control in continuously cultivated pearl millet (Pennisetum glaucum (l.) R. Br.) on a sandy soil. During 3 years we evaluated the effects of pre-sowing ridging and combinations of hand and mechanical weeding powered by donkeys on: seasonal weed growth (separating within-row and between-row weeds), pearl millet yield, and labor requirements for weeding. Four weeding methods resulted in levels of weed control decreasing in the order: full field hand weeding (Wf); between-row weeding by animal traction plus additional within-row hand weeding (Wa h); between-row weeding by animal traction (Wa); and no weeding (W0). Weeding method was the single most important factor that affected weed growth and, hence crop yields. Depending on the year, average weed dry matter at harvest ranged from 140 to 270 kg ha−1 for method Wf to 3000 to 3520 kg ha−1 for method W0. Correspondingly, the highest millet yields ranging from 279 to 1012 kg ha−1 were obtained with Wf with a total weeding labor requirement of 70.1 h ha−1. The Wa h method required 51.8 h ha−1 and resulted in yields, depending on the year, that were 78 to 100% of the Wf method. The Wa method resulted in unchecked within-row weed growth that exceeded those in weedy check plots, and reduced crop yield to 28 to 34% of Wf. Finally, unchecked weeds reduced yield to 2 to 10% of Wf. Pre-sowing tillage reduced seasonal weed growth, increased crop yields, and was particularly effective in reducing the amount of crop weeds for the Wa weed control method. The continuous growing of millet did not change the amount of seasonal weed dry matter during the experiment.