The cost would be that the results would not be as generalizeable. A second
way to reduce variance is by using a within-subjects
design. In these designs, the overall level of performance of each
person is subtracted out. This usually reduces the variance substantially.