Suppose that cocoa does not have direct effects on health. Recruit 60 people, give 30 of them cocoa and 30 of them placebo. Due to random chance, you'll someti

Suppose that cocoa does not have direct effects on health. Recruit 60 people, give 30 of them cocoa and 30 of them placebo. Due to random chance, you'll sometimes get positive results, and sometimes get negative results. If all of those studies are published, then there isn't a huge problem -- a metastudy would show that on average cocoa doesn't affect health. But instead, support that only the positive studies are published. A third party reviewing the relevant literature will assume that all studies conducted showed a positive result, when in fact they were only the studies that were published. This is the problem with having sponsored research: the research itself can be done properly, but a bias in publishing can give an inaccurate view of the true science at hand.