Changes in soil organic carbon (SOC) stocks and its distribution in depth due to land use change have been reported worldwide. The objective is to establish the variation of the SOC pattern, in surface and deep soil layers, of the Pampas of Argentina as affected by agriculture during the last four decades. For the estimation of past carbon stocks, soil data from more than 2000 soil profiles were obtained from surveys (1960-1980). Soil variables were reported from the soil surface to the bottom of the profiles or to the petrocalcic horizon. Soil bulk density and organic carbon content were determined. For the carbon stock at present, a soil sampling was performed in 2008 at eighty two farms widespread over the region and at each farm paired treatments were sampled accounting for common vegetation types and land uses. Bulk density and SOC were determined up to 1 m depth. Rainfall and temperature were obtained from climatic records. An artificial neural network model was developed that allowed the estimation of the SOC stock (R²= 0.64) based on climate, soil properties, vegetation type and land use at county level. This regional model, linked to remote sensing information, estimated SOC stock at present of 4.11 Gt compared to the estimated stock from soil surveys of 4.16 Gt for an area of 48.5 Mha. Agriculture determined a reduction of 16 % of SOC to 50 cm in sampled sites. The stratification pattern of SOC in depth was not affected by the treatments; so that vegetation and land use impacted the SOC sequestered in soil, but not its allocation in depth. At regional scale only a small decrease of total SOC stock was produced while at county scale soil with SOC content higher than 100 t ha^-1 to 1 m depth loose carbon. Sequestration prevailed below this threshold.