False perspectives on human language: Why statistics needs linguistics