Results of experimentation are never conclusive as variability at various levels affect the results. In a clinical trial the variation of the results from one patient to the other is well known and the success of a treatment is measured against the background of this variability (using a statistical test). Less often assessed, though often very important, is the variability of the success from one center to the other. When important, the variability between centers can be incorporated as part of the yardstick for statistical significance. 

Similarly, in brain research when a subject repeatedly performs a task and the brain activity is measured, differences are evaluated relative to variation between repetitions. Yet equally important is the variability between one subject to the other, which is sometimes addressed but often is neglected. In a simple experiment of animals behavior, for example, when comparing one treatment to the other, there is variability between the results of one mouse to another, which is always addressed. But there is also variability between laboratories, which is evident when the experiment is replicated in another laboratory, which is hardly ever addressed; there might also be variability due to the treatment being tested on different strains of mice.

The choice of the appropriate level of variability is thus crucial in order to assure replicability over centers, in clinical research, over people in brain research, over laboratories and/or over strains in mice behavior experimentation.