In the Sahel, pearl millet yields are affected by the proportion of hybrid phenotype plants resulting from genetic mixing between domesticated and wild forms. Man counteracts this mixing by applying a production method, the efficiency of which is quantified in this study. Under experimental conditions, cultivated and wild pearl millet were hybridised in order to obtain cultivated pearl millet seeds including a known proportion of F1 hybrids tagged by two different allozymes. These seeds were sown in the field and the cultivation was conducted following practices common in the Sahel. The evolution of the survival rate of plants and the frequency of hybrids were followed over several stages during the season: sowing, germinating, emergence, thinning, flowering and maturing of the seeds. Owing to plant mortality in the experiment, the average tendency was a hybrid frequency that decreased steadily in the first part of the growing season from 42% during germination, to 37% at emergence. It then fell to 17% after the thinning of the plantlets by the farmer. At the end of the cycle, after thinning, only 11% of mature plants were hybrids. Thus, under the combined pressures of natural and human selection, the frequency of hybrids in the field declined drastically. In interaction with natural pressure, the farmer's practices of selection of seeds, sowing in pockets and thinning have the combined effect of heavily selecting the cultivated genotype and limiting without completely preventing the introgression of wild pearl millet genes into the cultivated genome.