Subjects are divided into blocks based on a common characteristic or variable, and then randomly assigned to different treatments within each block. This design reduces the variability due to the blocking variable that is not of interest and focuses solely on the treatment variable.