劉向輯錄屈原等人的作品，編成十六卷《楚辭》，今已佚亡，後來王逸為《楚辭》作注釋，編成《楚辭章句》。過去學界認為王逸《楚辭章句》的底本必然出自劉向編集之《楚辭》，本文針對這個問題，指出宋以來學者僅據王逸序言考證兩書版本的關係並不可靠，一方面宋代以前文獻未有論及《楚辭章句》底本源出劉向；另一方面，王逸受朝廷徵召，擔任校書郎一職，編撰《楚辭章句》時勢必涉及文本校勘。從漢代校勘古書的方法來看，校勘典籍先是兼備異本，除其重覆，作成定本後再對勘文字，並非以某一底本再加以對校的形式，故絕不能用後世的版本觀念稱劉本為底本。今人主張王逸依據劉本為底本撰寫《楚辭章句》的觀點，是受到宋代雕版印刷將「劉向」之名設置於刊本書題的影響，至今仍成學界主流觀點。辨清《楚辭章句》與《楚辭》兩書的關係，將有助吾人釐清古人底本的觀念，對進一步了解古書注本的形成亦有很大的裨益。

A sixteen juan version of Chuci [the Songs of Chu] was compiled by Liu Xiang through the collection of works by Qu Yuan and others. While this corpus of verses is lost today, Wang Yi’s annotated work allegedly based on Liu’s anthology is preserved and transmitted till now. Most scholars take for granted that Wang Yi was annotating Liu Xiang’s compilation. This paper seeks to challenge this assumption. Not only it is unreliable to use Wang Yi’s preface to prove the linkage between the two texts, there is also no textual evidence that could suggest that the base text of Chuci zhangju [the Annotated Songs of Chu] was Liu Xiang’s work before the Song dynasty. Wang Yi was summoned to court to serve as an official collator, thus one would surmise he would have compared different editions of the works and would attempt to edit the text with the resources he gathered in the imperial library. Using the methodologies scholars adopted to edit the text during the Han dynasty as reference, the common practice of collation would involve a collection of different editions and elimination of repeated texts. This process enables the compiler to formulate a base text for further revisions. In this case, it is impossible to refer to Liu Xiang’s compilation anachronistically as a base text for Wang Yi. This paper seeks to account for the reason behind why most scholars assert that Chuci zhangju was annotated based on Liu’s compilation. I would argue this assertion was made mainly because publishers often placed Liu Xiang’s name under the title of Chuci in the bibliographies compiled in the Song dynasty. To use the relationship between Chuci zhangju and Chuci as a point of departure, this paper seeks to understand the concept of base text in early China. This will contribute to the study of the process of compilations and annotations.