This paper describes the statistical analysis of an experiment with 22 soybean varieties, carried out at the Campinas Central Experimental Station, in order to test their resistance to root knot nematodes. The same check variety (Abura) was planted in each plot in order to estimate the infestation of the soil by nematodes. Subjective scores have been ascribed to each plot and to the check variety from 0 to 5, a score 5 to the varieties with highest incidence of nematodes. Two observations were made at intervals of 60 and 90 days after germination. The averages of these scores were used for the statistical analysis. Considering y as the variety score and x as the check variety score, it was possible by means of covariance analysis to eliminate from the experimental data the variation correlated with the variation of the check variety. With this analysis, the error variance was reduced by 13%.