Where groups within a population are likely to respond differently to an experiment, the variation between subjects may be so great that simple randomized experiments will not reveal differences in treatments which actually exist. The randomized blocks design is analogous to stratified sampling for estimation purposes in the same situation, where the randomization is done within groupings of similar subjects—these groupings are the ‘blocks’ of the name. Each treatment is then applied to equal numbers of subjects in each block, but the choice of which subject gets which treatment is randomized within each block. For example, if two methods of improving memory are to be compared it might be considered likely that age and gender may be factors, in which case the blocks might consist of males or females of different age-groupings.