On the Shape of Mathematical Arguments