Even where there is no outright fraud involved, simple statistical errors, “publication bias” and perverse incentives can render “breakthroughs” less noteworthy when the studies reporting them are looked at more closely.