This study aims to determine the appropriate size of field plots for field experimentation with sunflower. An experiment was conducted in randomized complete blocks design with 14 cultivars of sunflower and 10 replications. The field plots consisted of four rows of six-meter long rows, spaced 0.7 m and 0.3 m between plants, with a total area of 16.8 m2. The useful area of the plot (7.56 m2), consisting of the two central rows, was divided into 12 basic units, each one consisting of three plants in the row (0.63 m2). The production of sunflower grains obtained in basic units was grouped in order to form portions of seven kinds of five different predefined sizes. The appropriate size of the experimental plot was estimated by the following methods: a) Intraclass correlation coefficient b) Maximum modified curvature c) Segmented linear model with plateau and d) Hatheway (1961). Were also estimated the soil heterogeneity coefficient (b) and the detectable difference among treatments (d). There was a reduction in the coefficient of variation with increasing the size of the plot. The soil of the experiment showed high heterogeneity ( = 1.0585). They were estimated by the methods of the intraclass correlation coefficient, maximum modified curvature and segmented linear model with plateau, respectively, the optimal plot sizes corresponding to 2.52, 3.74 and 2.48 m2. The maximum modified curvature method presented estimate of the optimum plot size more appropriate, together with the detectable difference between means of cultivars to accurately assess the yield of sunflower grain. The plot of 3.74 m2 of useful area was considered appropriate to assess the yield of sunflower grains and it was smaller than the size generally used in researches with sunflower. Though the Hatheway method (1961), they were estimated several very aplicable plot sizes. Considering all the cultivars for the same difference to be detected among means of cultivars, the use of a portion of smaller size with the largest number of replicates required less experimental area than the larger plots with a fewer number of replications