In his discussion of “The Early Printed Editions of the Canterbury Tales” Dr. Greg endeavored to discover the affinities of the manuscript which was used by Caxton in correcting his second edition, printed in 1484. “We have already seen,” he notes, “that C8 was printed from C7, but we have Caxton's own word for it that the text was carefully revised by comparison with a new manuscript, and this claim is amply borne out by the table of collations.” In other words, whatever manuscript copy Caxton may have had in hand for his second edition, he seems to have used it only for corrections where he felt his earlier source to be at fault. From a thorough examination of the variant readings in the first 116 lines, Dr. Greg concluded that it “does not appear possible to determine the affinities of Caxton's second manuscript for the opening of the Knight's Tale.” According to Koch, Caxton's second manuscript is to be assigned to the “Eλ-Dd group,” but Dr. Greg finds that none of the MSS of this group is consistently more successful than others in the Petworth or Corpus groups: “It cannot, then, honestly be said that there is anything to choose between these three groups, and we seem bound to suppose either that Koch's conclusion was invalid or that the relations of the manuscripts are not constant.” Nevertheless, Dr. Greg's investigation is not without results. The manuscript used by Caxton, he concludes, must belong either to the Eλ-Dd, the Petworth, or the Corpus group. Among these he narrows the selection by his table to a comparatively small number of MSS, and he points out that of those which he has examined “A [BM Addit. 35286] does best.” However, the variants shown in lines 70 and 74 of the Knight's Tale “seemingly bar this manuscript.”