In France, sunflower variety trials for new varieties registration on the National List are planned in the fields as randomized complete block designs with four replicates. In this paper 119 such trials (from 1993 to 1996) are analysed as if they had been planned with a resolvable incomplete block designs (IBD) to investigate whether a different system of blocking would have been better than the one actually used. A postblocking technique (Patterson and Hunter, 1983; Ainsley et al., 1987) is used to establish a relationship between error mean squares and block size in order to predict the efficiency of IBD . It is shown that when recovering inter-block information the IBD average efficiency would have been 1.33 for five plots per block; with intra-block information only, the IBD average efficiency reaches 1.24. Block size has an effect on efficiency. The results are in accordance with those of Patterson and Hunter (1983) who recommended that the optimum block size should be roughly equal to the square root of the number of varieties in the experiment.