Blocking

Blocking pulls a known source of extraneous variation out of the treatment comparison instead of letting it sit in the noise. For example, if you expect men and women to respond differently to a training program, block by sex and randomly assign the programs within each group so every program is tried on both. Blocking is the experimental analogue of stratifying in sampling, but the two are not the same move: you stratify a population before selecting a sample, and you block experimental units before assigning treatments. Both build groups that are similar inside, and only blocking involves random assignment.