On the unit of description in phenomenography