This is a wider issue, but how do they prevent knowledge of the experiment adversely affecting the results?