自1980年代以來，魯曼就將自我指涉/異已指涉當成社會系統 的必要成分。本文首先從魯曼的理論發展來追蹤自我指涉/異已指涉 這組區別的出現和變化，並從其中推導出一個隱含的弔詭：與訊息區 別開來的告知是系統自我觀察的成果，但自我觀察卻是系統取得關於 自身的訊息，由此訊息既和告知區別開來，也不和告知區別開來。直到後期，魯曼才試圖解決這個弔詭，但他的解決方法引發了更多問 題：首先，他沒有解釋異已概念為何需要擴充；其次，過程性自我指 涉和反省現在既被他視為自我指涉，也被他視為異已指涉；最後，他 也沒有解釋自我觀察為何有些涉及訊息的製造，有些則否。有鑑於此，本文建議將這組差異視為偶然的成分，它在歷史中既非必然出現，也非不可能出現。如此一來，研究者便能追問它的可能性條件。

Since 1980s Niklas Luhmann has treated the distinction between self-reference and hetero-reference as the necessary component of social systems. This article examines how this distinction emerges and changes throughout the development of Luhmann's system theory, and a paradox hidden behind the distinction. The utterance, which differs from information, is the result of self-observation of the system. But the selfobservation refers to the information about the system itself. That is, an utterance is an information and not an information at the same time. Luhmann had not tried to solve this paradox until the end of 1990s. His solution, however, leads to more problems. First, he cannot explain why the concept of “hetero-” must be extended. Secondly, procedural selfreference and reflection are also treated as hetero-reference; Lastly, he cannot explain why some of the self-observations are involved in information and some not. In order to overcome these dilemmas, this article suggests that the distinction between self-reference and heteroreference is a contingent component of social systems; it is neither necessary nor impossible. As a result, researchers should investigate its conditions of possibility.