Pearl millet plays an important role for food security in arid regions of Africa and India. Nevertheless, it lags far behind other cereals in terms of genetic improvement. Improving its root system could improve pearl millet tolerance to abiotic constraints (drought and low nutrient availability) and lead to a significant increase in production. The objective of this work is to characterize pearl system root system development in order to produce knowledge for breeding, mainly targeted on tolerance to drought stress occurring at the early growth stages. First, we described the dynamics of early pearl millet root system development and the anatomy of the different root types. This work revealed the existence of three anatomically distinct types for lateral roots. We also showed the existence of variability in primary root growth and lateral root density in a diversity panel derived from cultivated varieties. Our study also revealed a large variability among the growth profiles of lateral roots. To further analyze this diversity, the growth rates of a large number of lateral roots were measured daily and a statistical model was developed to classify these lateral roots into three main trends, according to their growth profiles. These three categories distinguish roots according to their growth rate and their growth duration. These different lateral root types are randomly distributed along the primary root and there seem to be no influence of root types on the intervals between successive lateral roots. The three growth types correspond, though imperfectly, to the three anatomical types evidenced in the first chapter. A similar work has been performed on maize, which was used to compare these two phylogenetically close cereals. Finally, we initiated the search for genetic markers associated to primary root growth, a trait potentially involved in early drought stress tolerance. A large panel of genetically fixed pearl millet inbred lines was phenotyped, confirming the