Twenty genotypes of common beans were evaluated in trial fields, in 1998/99 and 1999/2000 seasons, in Santa Maria, RS, Brazil, in order to verify the necessity of the utilization of border rows to improve the precision of regional bean trial. A completely randomized design with split plot with three replications was used. Plots contained four rows with 4m and the row spacing was 0.5m, with density of 15 seeds per meter. Useful plot size was two central rows discarting 0.5m in border. It was concluded that border utilization, with the useful plot size, does not improve the precision of the regional bean trial, because the error means squars of yield are homogeneous, common mean of the yield, coefficient of variation and the least signifficant differences are similar and the linear correlation coefficient is significant, in the two crop seasons, when yield in the useful area and the total area were consideraded.
experimental precision; common bean; experimental unity
