A study was performed to verify the relationship between the least significative difference in mean percentage (LSD%) of the variable grain yield and the mean and experimental error of the morphological characteristics observed throughout the experiment cultivar cicle, and the number of treatments in corn. Using data from a net of corn cultivars trials in the Rio Grande do Sul State, a pathy analysis was carried out, with LSD% as the main variable and as explaining variables some other variables measured in the trials. The results showed that variables such as grain yield, plant height, heigth of the ear and the final number of plants in corn, with more homogeneous values inside each trial, lead to a smaller mean square error, which reproved the experimental precision.