A massive seven-year project exploring 3,900 social-science papers has ended with a disturbing finding: researchers could replicate the results of only half of the studies that they tested 1. The ...