With the use of no-till farming, the soil physical and chemical properties are constantly changed, directly reflecting on plant yield and composition. In 2006, in the municipality of Selviria, Mato Grosso do Sul, Brazil, the linear and spatial correlations between the plant attributes and the attributes of a Rhodic Hapludox were analyzed. The objective of this work was to select among the analyzed soil attributes the one that best explained the variability of agricultural yield. For this, the geostatistical grid was installed for the soil and plant collection in 124 sampling points in a 4,000 m2 area. In the soil samples, collected at three depths (0-0.10 m, 0.20-0.30 m and 0.10-0.20 m), the microporosity, macroporosity, total porosity, bulk density and organic matter content were determined. In the plant, the yield (fresh and dry matter) and the crude protein content were determined. From the spatial viewpoint, there was a direct correlation between the dry matter yield and the soil organic matter content, as well as an inverse correlation between the crude protein content and the macroporosity and total porosity of soil. Therefore, the surface organic matter content behaved as a soil chemical quality indicator to estimate the dry matter yield of Cajanus cajan L. under the conditions of this research.
