The objective was to assess the extent and intensity of the spatial dependence of the fertility of a Red-Yellow Latosol, cultivated under a non-tillage system, by using both geo-statistical techniques and multivariate statistics. The experiment was conducted in an area planted with soybeans under a non-tillage system, in the city of Rio Verde, Goiás, Brazil. The soil was sampled at depths of 0-0.05m and 0.05-0.15m, in a sampling grid with 44 points. The attributes evaluated were the pH of CaCl2, Ca, Mg, Al, P, K, H+Al, CEC, BS, V and MO. The data were analyzed using descriptive statistical analysis, followed by principal component analysis, in order to generate components which would explain most of the variations in the attributes. Geo-statistics was used to check and quantify the degree of spatial dependence, using the principal components. Multivariate analysis, associated with the geo-statistical analysis, showed satisfactory results for the spatial distribution of fertility. Organic matter showed greater spatial continuity due to the characteristics of the no-tillage system. In general, the area presented low fertility for the cultivation of soybeans, as well as different patterns of spatial distribution between layers.