¹ Vols, v–viii of The Text of the “Canterbury Tales”, by John M. Manly and Edith Rickert (Univ. of Chicago Press, 1940), hereafter, Manly and Rickert.

² In discussing the theories jointly elaborated by Dr. Manly and Dr. Rickert and presented in their admirable work, I shall for the sake of brevity avoid the double reference and mention only Manly.

³ Manly and Rickert, ii, 42–43, 475–476, 483; in, 481–182. It is simpler to mention only Manly as holding these views, but Chaucerians will recall that, in regarding as editorial the d order and indeed the order of the tale blocks in all our MSS, Dr. Manly was in agreement with Dr. Tatlock—“The Canterbury Tales in 1400,” PMLA, l(1935), 100-139, esp. pp. 127–31—and that the spuriousness of the links peculiar to d and of the “Marriage group” order in d had been recognized by many scholars. For Manly's ideas on the manuscript history of the tales I may refer to my article, “Manly's Conception of the Early History of the Canterbury Tales”, PMLA, lxi (1946), 379–415.

⁴ “The Evolution of the Canterbury ‘Marriage Group’, ” PMLA, xlviii (1933), 1041–59; Review of Manly and Rickert, MLN, lv (1940), 606–621; “Three Notes on the Text of the Canterbury Tales”, MLN, lvi (1941), 163–175; and “Author's Revision in the Canterbury Tales”, PMLA, lvii (1942), 29-50. For a brief indication of the main points, see Dorothy Everett in “The Year's Work in English Studies”, xxiii (1942), 50–51. Additional observations in support of Dr. Brown's explanation of the formation of the d order are offered by Margaret Schlauch, pp. 427–130 of “The Marital Dilemma in the Wife of Bath's Tale”, PMLA, lxi (1946), 416–430.

⁵ The c MSS are Corpus Christi 198, Lansdowne 851, and Sloane 1686. The four subgroups of group d are named after MSS Petworth, Royal 18 C.II, Lichfield 2, and Egerton 2863; for their composition see Manly and Rickert, ii, 49; the three single MSS are Sloane 1685, Delamere, and Harley 1758; the less constant members are Barlow 20, the Bodleian subgroup, Laud 600, Fitzwilliam McClean 181, Selden Arch.B.14, and Northumberland.—Italicized sigils in this article correspond to Manly's underlined sigils, i.e., they denote, not single manuscripts, but groups or subgroups of manuscripts. To simplify presentation I shall use the sigils d and √d to designate the group and its ancestor whether in the piece under discussion the group consists only of “constant” members or includes temporary associates, in which case the sigils in Manly and Rickert are d∗ and √d∗.

⁶ Inseparable is used in the same sense, or we are told that c does not show apart from d, i.e., has no error not shared by d. That d does not show apart from c is seldom added but can be assumed to be Manly's opinion wherever we find no statement to the contrary.

⁷ The only possible exceptions seem to be McT and PsT, where √b might be derived from the latest common ancestor of c and d. This of course does not affect the relation between these two. For all other pieces included in our list the evidence for an exclusive ancestor of c and d is decisive. It is true that many of the variants adduced by Manly occur also outside of cd or are lacking in contaminated and edited cd MSS, but considered in the aggregate, and supplemented by scores of similarly distributed variants not listed by Manly, they prove the point beyond a shadow of a doubt.

⁸ By the sigil √cd I never mean more than copies of particular pieces or passages. The √cd copies of the portions of CT adding up to our two thirds of the work and the √cd copies postulated for most of the portions constituting the other third were certainly never combined as parts of one CT exemplar.

⁹ In the tales in which the three c's are inseparable from d, Manly says, “either the three MSS used singly the exemplar from which the subgroups and single MSS of d were derived”, (this seems to mean derived independently of one another, i.e., according to fig. 1 rather than fig. 3) “or their intermediary in its few errors is with d by accidental coincidence, which is very improbable.” Radiation from √cd is definitely Manly's explanation for Gen Pro (see his diagram, ii, 86); less definitely for ReT (ii, 163, “Summary”). That such radiation involves the integration of portions of the √c exemplar into the √fd exemplar does not seem to have occurred to Manly; see e.g. i, 513; ii, 317,428, where it is assumed that, as late as ca. 1480–90, √c itself, still containing over nine tenths of CT, could have been used by the scribe of Sl².

¹⁰ “In most of the tales √c used the exemplar from which somewhat later was derived √d.” (i, 96).

¹¹ “Before 1740” i.e., in the first half of KnT, “there is no evidence for d∗ without c, and c does not appear apart from d∗. This does not mean that there was no Group c but only that it was in this part a subgroup of cd∗. The three MSS constituting c here and elsewhere had an exclusive ancestor … The absence of exclusive c variants is accounted for by the accuracy with which the scribe of √c reproduced his exemplar, √cd.”

¹² I shall keep referring to the √d editor in the singular, assuming that most of the responsibility for the preparation of the √d exemplar fell on one man. What assistance he may have had in the execution of the project I am unable to say.—The word exemplar will be applied exclusively to MSS (of the whole CT or of parts) apparently intended for copyists rather than readers.

¹³ A 1134 and 1153, where all d's are correct while two c's have errors, are left out of my list because, in each case, the two c's are the edited La and SI² and their variants are different.

¹⁴ Each subgroup is counted as having the reading of its most conservative member.—Bw, after A 1250, is no longer in d.

¹⁵ Those listed in the corpus plus the accidentally omitted En² and Ld¹. In no other case do we find more than four d's sharing a variant not in any c; see A 947, 981, 990, 1638; the insignificance of all those is obvious at a glance.

¹⁸ The Original, as Manly has shown, had many errors; there is certainly no reason to think that Hg's thirty-one variants in A 859–1740, three of them unique variants (Manly and Rickert, ii, 123), were all in his exemplar; Mg is described as remarkably close to its model Lc (i, 373; ii, 65), but in 200 Unes (A 952–1151) picked at random in our passage, I note seventeen divergences from Lc; for the seeming accuracy of the √Pw scribe in some portions of CT (ii, 68), the most probable explanation, in my opinion, is that some members of the subgroup copied certain pieces from √d directly or through a derivative other than √Pw.

¹⁷ It is not on these or any other variants that Manly based his belief that, for A 859–1740, the c's and the d's radiated from distinct exemplars. His feeling that, here and else where, it had to be so was due merely to his failure to see that the two CT exemplars √c and √d did not have to be postulated as coexistent.

¹⁸ Except for the two prose pieces, for which scattered specimens seemed conclusive enough, all the pieces involved have been examined in their entirety. Should additional illustrations seem of interest, I shall give here, for ReT and FrT, the lists of the variants which, isolated, might suggest different readings in hypothetically distinct √c and √d exemplars: A 3863, 3986, 4170, 4181; D 1350, 1620 (but the testimony of Cp is more reliable than the joint testimony of La and Sl²) and 1602.

¹⁹ Insignificant variations are disregarded, as well as the dissent of notoriously contaminated members of d.

²⁰ Few of these points were noted by Manly, with the result that √c and √d, for some of the pieces now on hand, are said to be inseparable (ii, 211), or a reservation is made on a basis different from mine (ii, 435).

²¹ Manly and Rickert, iii, 461–162.

²² I include within cd the group B0¹-D1-Ha²-Fi whose exclusion does not seem warranted by Manly's evidence, ii, 435.

²³ In agreement with this explanation is the fact that some omissions are peculiar to c, but none is peculiar to d; further, that the d scribes disagreed as to the position of his disciple, a sign that, in their exemplar, those words appeared as a correction.

²⁴ Apart from the cases listed in the text, the c MSS, in FkT, WBT, and CYPT, are not found a single time in agreement on an error not present in any d, nor the d MSS on one not shared by any c. In SNPT, where c counts only two members, I find two cases, G 94 and 180.—Some variants present in many but not all d MSS should perhaps be added to the list of corrections by √d, e.g. G-1471.

²⁵ The only non-cd MSS that do not have it are members of b∗, which is here close to c. 26 See e.g. A 3958, B 1111, D 753, E1938,1957, F1184,1188,1540. In other cases the disagreement is about the inclusion of some words; see A 3624, D 1430, 1603, 1604.

²⁷ Some of the cases in which the variant is found also outside of cd are almost as significant; see e.g. A 1203, D 763, B 1803,1921, 2075, G 728,1275.

²⁸ In D 1159, For gentilesse nys but renotnee, Cp and La, instead of renomee, read reneute and SI² reuerente. Nothing like either of those two forms occurs in any MS outside of the cd group. Of the d MS subgroups and single MSS, about half have two words instead of one; the first is reuent, reuerent, reuerence; the second, in spite of spelling variations, clearly points to the correct form renomee in the common ancestor. The other d MSS have, in one form or another, either the first word (reneute, reuerente) or the second {renomee, renoune). It seems clear that √d was a corrected copy, our √d man or an assistant having written the correct word renomee without thoroughly (if at all) cancelling the word which this was intended to replace; further, that the not thoroughly cancelled word was liable to be variously construed owing to 1) its third letter, consisting of two minims which could be read as n or as v, 2) something suggestive of an abbreviation for er over the second syllable, and 3) another letter not clearly nor v—that is, exactly the features whose presence in √c would account for reneute in Cp and La and reuerente in SI². It would be hard to avoid the conclusion that the copy of WBT hastily corrected for use as part of √d was the √c copy of that piece.—Very similar conditions are found in the next line, Of thyn auncestres for hir hye bountee, where √c, instead of hye bountee, had merely beauté, which the corrector replaced by bountee but again without cancelling the word. (Incidentally, his failure to add hye suggests correction from memory). Most scribes treat this second problem in the same way as the first; in both lines Pw, Mm, Ld¹, To and Nl have both words, Lc and Ps only the corrector's word, Ph³ only the √c word. Cf. also D 106, where √c, for continence, had conscience; a corrector wrote continence, and Dl reads continent conscience.

²⁹ It is principally though by no means exclusively in this third of CT that Manly finds in d (or d plus other members of the Large Group) variants which he believes authorial and relics of early versions. My reasons for not sharing this view were indicated briefly in the article already referred to, note 32. For other facts which clearly invalidate some of the evidence offered by Manly, see J. Burke Severs, “Did Chaucer revise the Clerk's Tale?”, Spec, xxi (1946), 295–302.

³⁰ For NPT and all or part of MkT, I am inclined to believe that the √d copy, distinct from √c, was derived from it through an intermediary on which a few corrections had been made.

³¹ See Plates II and III in Miss Margaret Rickert's chapter on Illumination, Manly and Rickert, i, 561–605. The pagination, handwriting, treatment of headings, etc. are very similar; the decoration in the d MSS is mostly in a style characteristic of a slightly later period, which, however, alternates with the older style of Cp and La (p. 570).

³² No other explanation seems at all plausible. The √c copies, with the possible exception of only one or two passages, could hardly have been rejected as corrupt, for the d text is as a rule very obviously worse; nor as too badly worn out, since, after decades of service as portions of √d some portions of √c were still in use; see below, n. 35.

³³ The losses of √c were few, probably five, but extensive, each of them involving consecutive pieces; see below, n. 53.

³⁴ Had the non-√c copies been found already extant, there would inevitably have been cases of overlapping with the √c portions, particularly within CT pieces. In view of the carelessness of the √d editor and of most d scribes, we would almost certainly find traces of this here and there in our MSS. The presence of such traces in one case unique in many respects, the Ph-Pd Link (see below, nn. 79, 82), greatly increases the significance of their absence elsewhere.

³⁵ This repatching of √d seems to have taken place about the middle of the century. What was at that time its condition down to FkT is far from clear. As to the second half of CT, the rarity of variants peculiar to Bo¹-Dl etc. in PhT, PdPT, ShT to ca. B 1370, Th to ca. B 2042, NPPT, McT indicates that, for those pieces, the new editor did not use derivatives of √d but √d itself; this may be true also of PsPT. For ShT after ca. B 1370, SNPT, CYPT, and MkT, he provided copies derived from √d but much more corrupt; for Th after ca. B 2042, and for some links, he used copies derived from antecedents of √cd; for PrT, a copy not related to any known one; for Mel, a copy derived from an antecedent of El. One gap, McP, he did not fill at all.—These vicissitudes of √c and √d as seen by the present writer would account for apparent coincidences for which Manly had no explanation, for instance the fact that, in ShT, Bo¹-Dl-Fi-Ha²-Sl¹ should appear as a distinct group ca. B 1370, i.e., just at the point where c and d become textually indistinguishable. What happened is this: The √c pages covering the beginning of the ShT, B 1191-1370, had gone astray before √d was prepared; the substitute copy provided by the √d editor and the old √c copy of B 1370 ff. were not attached, or at least not securely attached, to one another; the former happened to be preserved for the √Bo¹-Dl-etc. editor to use, while the latter was not. For Manly's comments see ii, 350.

³⁶ The evidence, mostly from the quire make-up and the inks, is abundant (i, 266–275; ii, 477–479) and no alternative explanation seems conceivable that would not involve gratuitous assumptions or most unlikely coincidences.

³⁷ See ii, 298–299 and the corpus of variants.

³⁸ See i, 224; ii, 284–285, 286, and the corpus of variants.

³⁹ Ha⁴ has lost the quire which contained F 617–1223 (i, 219); that the Sq-Fk Link was present is certain from the quire make-up and page content of Ha⁴. Other MSS probably indebted to Hg for one or both of those links (and not through √d) are Ht, Ps, Ra³-Tc¹-Ln, Bw (though largely a d MS), and Py. The indications of descent from Hg are particularly strong in Ht, which has SqT, MeT, and FkT in this Hg order, and the texts of MeT and FkT derived from Hg. (The relationship between Hg and Ht is constant throughout the MeT; Manly, ii, 279, is mistaken on that point.) The case of Py, which has the Sq-Me Link (two copies of it; ii, 298), is hopelessly confused. It is genetically related to Hg in various pieces including SqT and MeT, but is so corrupt that it seems impossible to determine the nature of the relationship, which may well be variable; cf. Bw, from Hg in Sq-Me Link, from an antecedent of Hg in FkT.

⁴⁰ My text differs on several points from that given in Manly and Rickert, iii, 481–482, as the text of the original. In st. 1, lines 1 and 2, although and And yet in Manly's text are certainly mistakes, for all ten MSS read though and Yet. On three points (st. 1, line 4, And vs. For; st. 2, line 2, worthi man vs. gentilman; line 7, now omitted) I adopt as the most probable reading of the original that of Bw-Ry²; see below, p. 478 and n. 71. I keep Manly's reading, say vs a tale in st. 2, line 2, as the more idiomatic telle vs a tale of Bw-Ry² is certainly an emendation; cf. Hg.

⁴¹ In Hg the word was caught by accident from the preceding line.—Manly does not seem to have noted these agreements between Hg and the stanzas.

⁴² Ht, Ii, and PI (I disregard Fi, too confused to be significant). In PI the Hg Me-Fk sequence is probably due to the link form, inherited from Hg through √Pw.

⁴³ E 2427, that omitted; 2432, sore that for sore; to for unto.

⁴⁴ Ra³-Tc¹-Ln, Bw-Ht, and Py obtained the text independently of √d, Ii and Nl through √d.

⁴⁵ That the passage was already lacking in √bcd, though not necessarily from the first, is suggested by the shift of affiliation of √b and the fact that none of the c scribes was able to fill the gap. On the other hand Ha⁴, which was probably written in the cd shop, has the passage and its text of it is related to √d's; see ii, 282. The significance of the concurrence which Manly notes between this breaking off of the c text and a change of ink in Hg will be discussed in a short article to appear in MLN.

⁴⁶ Both passages seem to have been lacking in the copies' from which √c was derived, but may in each case have found in the cd atelier the copy from which the La editor had supplemented the c text; see the variants at B 3202, 3969, 3981.

⁴⁷ Manly takes this for granted (ii, 42–43) without giving his reasons, which doubtless were these: The textual closeness of √c and √d favors derivation of one of the orders from the other rather than likeness by accident; the earliest c MSS are older than the earliest d's; the numbers appended to tales in some d MSS correspond not to the dbut to the c tale order.

⁴⁸ See n. 120 of my previous article.—As expected, yet worth noting as confirmation, figures reflecting the tale number of √c are appended in d MSS only to pieces whose √c copy we believe on other grounds to have been incorporated into √d.

⁴⁹ Unless there was a table of contents. The texts of some tales—CkT, MkT, NPT—would in all likelihood be obtained with links indicating their position; not so with C1T or Gamelin.

⁶⁰ See above, n. 30.

⁵¹ On this question.see Tatlock, op. cit., pp. 110, 119–120. 52 Gamelin, WBP, MeT, FkT, SNP, PhT, and ShT.

⁵³ √c's other three losses, likewise filled by the √d editor, may be dismissed in this note by a mere indication of their extent: 1) from ca. D 1740 to ca. E 1640, i.e., most of SuT, entire C1PT and Envoy, first 395 lines of MeT; 2) from C 329 to ca. B 1371, i.e., PdPT and first half of ShT; and 3) from B 3181 to H 104, i.e., MkT, NPPT, and McP. (I do not include F 1545–1624 [end of FkT] among the losses of √c, for rejection of the √c text as exceptionally corrupt [Manly and Rickert, ii, 302] seems more likely than √c's loss of so short a passage.)—In MeT, the point at which √d becomes indistinguishable from √c is, according to Manly, ca. 1691. But it is at 1640 that the series of variants peculiar to bed suddenly starts (1640, 1642,1645, 1646–47, 1650, 1656, 1662, etc.) and a little before that line that the evidence for eδ-Ry¹ and that for d suddenly come to an end. This seems to me to indicate the place of the shift much more clearly than does the first cd variant of classificatory value (which, besides, is not at 1692 but at 1677).

⁵⁴ Manly's statement about this link on ii, 154 is a mistake; cf. ii, 137. 55 Manly and Rickert, ii, 153.

⁵⁶ A critical edition of Gamelin is being prepared by Miss Mabel Dean, one of Dr. Manly's main collaborators, and co-author of The Text of the “CT.” I have examined the photostats through the first hundred lines, and found again and again most d MSS in agreement on trivial points against c, Ha⁴ and Ch. The √d copy, then, was certainly not the √c copy.

⁵⁷ To indicate Manly's position on this question: On one occasion he remarks that “the d pattern is derivable from the c pattern by adding the links, modified and spurious, which characterize the d arrangement, and moving Me to follow Sq in consequence of the discovery of the Sq-Fk Link and of its adaptation to the Sq-Me connection” (ii, 485; italics mine), but elsewhere is categorical in his assertion that the √d editor “having already misplaced Sq and Me, … was obliged to adapt the two newly-found links (Sq-Fk and Me-Sq) to suit the new arrangement.” (ii, 489; again in i, 275).

⁵⁸ Disregarding, of course, incomplete MSS and accidental disarrangements.

⁵⁹ Manly suggests this as a somewhat remote possibility (i, 274–275), but no alternative seems half as plausible.—As the Hg scribe never had CYPT, his SNT was not linked to any other; he could have placed it after Block A, covered by the first eight quires.

⁶⁰ The sequence, we recall, is compatible with all time and place indications in the text, and indeed was adopted by Skeat in what is still considered the most logical of possible orders for the CT fragments as we have them (see Tatlock op. cit., pp. 131-132; Manly and Rickert, ii, 491-492); yet—and this is the significant point—it is certainly not on any such ground that it was favored by those early editors, who were very far indeed from working out the best of possible orders. One wonders whether an extra textual note by Chaucer may not have indicated both the Nun as teller—internal evidence, we remember, assigns the Legend to a man—and the position of the tale immediately after the Franklin's.

⁶¹ The Cl-Fk sequence would be due to an early editor's insertion of MeT-D-CIT, a block in Chaucer's early plan, between the previously written and already linked SqT and FkT; for refs. see above, n. 4.

⁶² For the titles given to CT pieces in the MSS see The Manuscripts of Chaucer's Canterbury Tales, by Sir William McCormick (Oxford, 1933).

⁶³ Ry² and Ld² are “constant members” of group d; Bw is textually in the group in more than half of CT, and, with regard to content, order, and links, is entirely a dMS. The link is not found outside of d.

⁶⁴ For the complete texts of the spurious links of the d MSS and those of La, see Manly and Rickert, v, 437; vi, 3; vii, 3–4 and 110.

⁶⁵ The presence of the link in Ry² and its derivative Ld² is as easy to explain in one alternative as in the other, for, since Ry² in MeT is in group d, the link, if already present in √d, would easily reach the Ry² scribe in unbroken transmission with the rest of MeT. In Bw, which is in group d neither in MeT (except down to ca. E 1640) nor in WBP, the link would be borrowed from an antecedent of Ry². The two MSS are closely related in WBP and other pieces.

⁶⁶ Except that the subgroup Ry ² is represented here by Ry² only. Ld² follows another source from E 298 to F 944 (Manly and Rickert, i, 316) and has the Cl-Fk Link in its two-stanza form.

⁶⁷ This worthy clerk whan ended was his tale

Oure hoost seyde and swoor by goddes bones

Me were leuere than a barel ale

My wyf at hom had herd this legend ones

This is a gentil tale for the nones

As to my purpos wiste ye my wille

But thyng that wol nat be lat it be stille 68 Dr. Manly, in his remarks on the composition of the spurious link and on its stan-zaic form, never mentions the Host Stanza. It was apparently his opinion, and is implied fairly clearly here and there (see ii, 483; iii, 482) that the link was written as a two-stanza link, to which the Host Stanza was prefixed only later, on a derivative of √d.

⁶⁹ This is the explanation suggested briefly by Dr. Tatlock, op. cit., p. 120. That the rimer did not choose the easier course of recasting the Host Stanza into couplets may be due to the immense vogue of the rime royal as illustrated in Hoccleve and Lydgate.

⁷⁰ The clause quoth our hoste which in Fi, Ha², Lc and Nl makes the opening line too long, was of course not in the original but was added after the Host Stanza had dropped off; in Ha² it was kept after the Host Stanza was reintroduced.

⁷¹ 1) she hath (line 2 of 1st spurious stanza) vs. hath she in the other eight copies; cf. E 2429; 2) now (last line) absent in Bw and Ry² as in the source line F 6, present in all other copies of the spurious stanzas except Fi; 3) And, vs For (line 4 of 1st spurious stanza) which was no doubt caught from the previous line by mistake.

⁷² The only explanation in keeping with Manly's view that the d text of C1T probably represents a pre-CT version would of course be contamination; see ii, 265.

⁷³ ii, 483.

⁷⁴ ii, 286, 483; iii, 481.

⁷⁵ The √Pw scribe may have found, on loose leaves left between the C1T and FkT, both the stanzaic link and its source, and have chosen the latter. Or, if the copy of the stanzaic link happened to be out at the time, he may, looking for a link to introduce the FkT, have found the Hg text bearing perhaps the title of Prologue of the Franklin.

⁷⁶ G 1–119 is called prologue in very nearly all MSS which are here members of d, as also in La, here the sole representative of c.

⁷⁷ See Plate II in Margaret Rickert's chapter on Illumination, Manly and Rickert, i, 567.

⁷⁸ That the scribe of La shows no such hope, i.e., leaves no blank space for any E-F link, is no objection, for nothing suggests that the spurious links were composed by him.

⁷⁹ √c had a combination of the early form, which was only a Ph Endlink (C 287–299, quoted by Manly, ii, 325) and of the passage added later, C 300–328. The version supplied by the √d editor differed from that of √c in having in the opening twelve lines the variants introduced by Chaucer when he added C 300–328.

⁸⁰ Manly and Rickert, ii, 411.

⁸¹ One might argue that the Me-WB Link was written on a derivative of √d by a man who did not have the opening portion of the Me-Sq Link, previously discarded when the spurious stanzas were written—an attractive explanation at first sight, but hardly compatible with the indications of common authorship of Me-WB, Ph-Pd, and CY-Sh Links that we have noted.

⁸² At least five d scribes had copied C 287–299 in its c form (early version; see above, n. 79) before suspecting that something was wrong. One of them, the scribe of Dl or of an antecedent, started over again, copying from its beginning the revised link as supplied by the √d editor; the others, Ha², Lc, Ry¹, Ry ², pick up the √d text at C 300.

⁸³ Though Bw, Ld¹, Mm and Se derive their text of SuT from the √c copy, it does not seem likely that that copy was ever part of the √d exemplar.

⁸⁴ That the episode covers 136 lines, i.e. 4×34, suggests to Dr. Manly the loss of two folios. But the writer of the spurious lines must have had at least the end of the passage since he echoes the last line. This fact speaks also against Manly's alternative suggestion, that the d form represents an early and unfinished version of the tale (ii, 229).

⁸⁵ On F 1545–1624, the last eighty lines of FkT, see above, n. 53.

⁸⁸ See above, pp. 461-2 and n. 28.

⁸⁷ The SNPT illustrates every point: G 71 and 73–74 are supplied in their authentic form; 156–159 and 390 are replaced by spurious lines; nothing is done about 326–337.

⁸⁸ Scores of √cd variants in those pieces make nonsense; see e.g. A 999, 1005, 1137–38, 1373–74.

⁸⁹ That the √d scribe is not scrupulously faithful to a lost exemplar can in no case be demonstrated, but the fact that hiscopyalways differs oncountless trivial points from its closest known relatives, preserved or postulated, is, I believe, conclusive, as are, in the aggregate, the types of variants peculiar to d. I incline to think that one scribe wrote most, perhaps all, of the one third of CT not provided by the dismembered √c

⁹⁰ Py, Se, and Ry have the CY-Ph Link and Ha⁵ part of it; Nl has the two spurious stanzas; Ii and Ht have Hg's Sq-Me-Fk sequence with its two adapted links, but have √d'sCk-Gam, CY-Ph, and Pd-Sh Links.

⁹¹ Everything suggests that the √b editor owed his tale order to √d. We have seen that what makes the appearance of that lame order intelligible at the hands of √d is his applying to the √c pattern the Sq-Me Link of Hg. The √b editor did not have that link, and, further, had the Merchant's Prologue with its “weping and wailing” first line definitely tying to the Clerk's Envoy (whose stanzas, in b, appear in correct order, i.e., the “weping and wailing” fine is the last one). That in such circumstances √b should, independently of √d, have produced the same absurd tale order (Manly's view, ii, 485–486) seems to me highly improbable. There is no difficulty in the alternative. We have proofs that information as to tale order was (as might have been expected) occasionally transmitted with few texts or none (the α-El order transmitted to En³, e.g.); and of course the √b editor's close association with the cd shop is abundantly demonstrated by Manly's classification for a number of pieces.

⁹² Endorsement of the d order on the part, for instanec, of Thynne, Stowe, or Speght, is the more inexcusable as all of them had the “weping and wailing” piece which, in spite of the disarrangement of the stanzas of the CI Envoy, should have been recognized as intended to follow it; further, they could surely have seen the Sq-Fk link in its correct form, if not in a MS, at least in Caxton's second edition or in Wynkyn de Worde.