Studies intended to determine the effect of "Cambridge" roller working parameters such as specific weight and working speed on the soil density and density distribution in soil layers 0-10, 10-20, 2-30 em deep. The experiments were conducted on turf-podzolic soils of 16.7, 15.9, 15.2% moisture content and initial density 1.12, 1.12, 1.36 g/cm³ measured in respective soil layers using as the variable parameters working speed 4-12 km/h and specific roller weight 420-580 kg/m. The tests of crushing-pressing roller showed that its specific weight and working speed affected significantly the soil density and thus - the crop yielding. The optimum soil density at growing cereals on turf-podzolic soils was 1.2-1.4 g/cm³. On the basis of experimental results the regression equations were derived which describe the effect of such parameters as the specific weight and working speed of the roller on soil density in the layers 0-10, 10-20, 20-30 cm deep. At using working parameters determined according to regression equations (4 km/h working speed, 460 kg/m specific weight) in experiments, the yield of oats and peas mixture increased by 28% in comparison to the yields obtained at roller working parameters generally recommended for this region. Derived regression equations enable the optimization of roller working parameters for given soil-climate conditions.