For many years, scholars have thought that the manuscript source used by William Caxton to correct his first edition of the Canterbury Tales was a manuscript probably of the very best quality. In 1939, Thomas Dunn wrote a doctoral thesis on the subject, and for his research he used the Manly and Rickert collation cards. Technological advances made in the last decade of the twentieth century have made it possible to collate the witnesses of the Tales using computerised tools.

This work presents an analysis of the stemmatically significant variants found in Cx2 and attempts to offer a plausible hypothesis concerning the position of the manuscript source of Cx2 in the textual tradition of the Canterbury Tales. This thesis is organised in eight chapters. Chapter 1 focuses on the scholarly work surrounding Caxton's second edition and his editorial practices; chapter 2 contains the bibliographical description of one of the copies of Cx2; chapter 3 studies the question of the order of the tales; chapter 4 offers a synthesis of what, for the purposes of this particular research, is understood as a textual variant; in chapters 5, 6 and 7 the analyses of the data and some partial conclusions can be found. The findings of this work appear in the conclusions (chapter 8). There is an electronic appendix to this work in which data that were not deemed essential to its understanding can be found. The electronic appendix includes the complete collation of Cx2 against Cx1, collations of all the available witnesses and variants which were considered repetitive or uninformative.

This work shows that witnesses of the text which have remained unclassified up to this point might be genetically related. Especially evident is the relationship between Ad3 Ch Ha4 and the manuscript source of Cx2. It also appears that Cx2 shares with El and Gg variants which originated below the archetype. This thesis suggests that more work is required in order to clarify the stemmatic relations in the textual tradition of the Canterbury Tales.