Resumo. Methods of geostatistics were used in the identification of size and structure of space variability of some physiochemical attributes of soils under seasonal deciduous forest areas, which were called mature forest, secondary forest and ¿capoeirão¿. The areas, located in Santa Tereza, RS, were sampled during the period of 2002 and 2003, comprising the soil classes: Argiluvic Chernosol, Cambisol Ta and Litholic Neosol. Systematic sampling was performed with regular spacing grid of points varying of 30 m for "capoeirão" and 40 m for mature and secondary forest, in an amount of 80 sample points in a total area of 7.34 ha. For each point, measurements "in situ" were made of the soil bulk density, clay content, pH, potassium, organic matter and phosphorus. The structure of space variability was identified through the variograms, being defined the needed parameters for the Kriging. The most variable attribute was the organic matter, being not adjusted to any model of variogram, while the pH was the least. In the analysis of the space variability structure through the geostatistics, the clay and bulk density, both in area of secondary forest, and pH in the experimental areas, had showed the biggest space dependence. The Gauss model for the clay attribute obtained the largest reach (a), being 400 m and 388 m in the areas of secondary and mature forest, respectively. The crossed validation showed accuracy in the variographic models adjustment, being the pH variable the one which presented the optimum performance, 0.87; 0.94 and 0.92 in the areas of mature forest, secondary forest and ¿capoeirão¿, respectively.