On the Arbitrariness of Linguistic Signs