The greatest bias in the scientific literature comes from small studies that report large effect sizes, according to Stanford University's John Ioannidis and his colleagues in the Proceedings of the National Academy of Sciences this week.