An 80-year soil archive, the 42-plot experimental design at the INRA in Versailles (France), is used here to study long-term contamination by 137Cs atmospheric deposition and the fate of this radioisotope when associated with various agricultural practices: fallow land, KCl, NH4(NO3), superphosphate fertilizers, horse manure and lime amendments. The pertinence of a simple box model, where radiocaesium is supposed to move downward by convectional mechanisms, is checked using samples from control plots which had been neither amended, nor cultivated since 1928. This simple model presents the advantage of depending on only two parameters: α, a proportional factor allowing the historical atmospheric 137Cs fluxes to be reconstructed locally, and k, an annual loss coefficient from the plow horizon. Another pseudo-unknown is however necessary to run the model: the shape of historical 137Cs deposition, but this function can be easily computed by merging several curves previously established by other surveys. A loss of not, vert, similar1.5% per year from the plow horizon, combined with appropriate fluxes, provides good concordance between simulated and measured values. In the 0-25 cm horizon, the residence half time is found to be not, vert, similar18 yr (including both migration and radioactive decay). Migration rate constants are also calculated for some plots receiving continuous long-term agricultural treatments. Comparison with the control plots reveals significant influence of amendments on 137Cs mobility in these soils developed from a unique genoform.< Réduire