The determination of the density of the horizons of a soil profile allows to evaluate certain properties, such as: porosity, hydraulic conductivity and water storage. However, measured data is not always available or easy to obtain. Thus, it was objective with the work to build and evaluate models of pedotransference function based on multiparameters of the soil as alternatives for the estimation of soil density in areas with agricultural potential of the state of Piauí. Thus, soil samples were collected from 42 soil profiles in the state of Piauí at depths of 0.0-0.20; 0.20-0.40; and 0.40-0.60 m e, characterized in terms of chemical and physical attributes. Through multiple regression analysis, two pedotransfer function models were generated: i) including all determined attributes and, ii) only the particle size analysis and total organic carbon. For both models, all classes (generalized character) were considered, as well as each individual soil class Latossolo (Ferralsol), Argissolo (Acrisol), Plintossolo (Plinthosol) and Neossolo (Arenosol). The generated pedotransference functions are alternative to estimate bulk density of soil, obtaining "excellent" performance in model I for the "Latossolos" (Ferralsols) and "Argissolos" (Acrisols), which included a greater number of variables in the construction of the predictive model of bulk density.