Using 40-year experiment data from a mono-modal rainfall area of northern Togo, we analyzed soil fertility dynamics when 2 and 3-year fallows were alternated with 3-year rotation of groundnut, cotton and sorghum. The control treatment consisted to continuous cultivate the soil in a rotation of groundnut/cotton/sorghum without fallow periods. For each rotation, two fertilisation rates were applied: no fertilisation and mineral fertiliser application during the cropping and/or the fallow periods. Yields of unfertilised crops, which averaged 1 t ha?1 during the first years of cultivation, were often nil in the long-term. In the long-term, yields of fertilised cotton and sorghum decreased by 32 and 50 %, respectively compared to the average of 2.4 and 1.6 t ha?1 obtained during the first decade of cultivation. The long-term decline in crop productivity was mitigated when fallow periods were alternated with cropping periods, and consequently there was partial compensation in terms of production for the unproductive fallowed plots. Long-term yields of fertilised cotton and sorghum in the periodically fallowed plots were 40 and 50 % higher than those in continuously cropped plots, respectively; they were 90 and 60 % higher than those in continuously cropped plots without fertilisation. Like for crop productivity, soil C, N and exchangeable Ca and Mg decreased less in periodically fallowed plots than in continuously cropped plots. The limited soil C decline when fallows were alternated with crops appears to be the consequence of no-tillage period rather than the effect of the highest C inputs to the soil.