University of Virginia Library

Chaucer's Revision of Troilus and Criseyde
by
Kevin K. Cureton

The theory, once axiomatic, that the manuscripts of Chaucer's Troilus and Criseyde preserve revision states of the text has been brought into question in recent years. In his new edition of the poem,[1] Barry Windeatt dismisses the view of the text developed by William McCormick and Robert Root. In Root's formulation, the bewildering manuscript relationships, which had proved impossible to classify in a fashion suitable for editing by recension, would be explained by recourse to the hypothesis of a single manuscript, a fair copy of the first draft in Chaucer's possession, successively corrected and revised by the poet.[2] Early copies made from this manuscript would preserve the unrevised form of the poem, what McCormick and Root called the alpha text; copies made after Chaucer had performed some or all of his revisions would preserve beta or gamma texts. McCormick, who made the original classifications, held that beta represented the intermediate and gamma the final forms of the text. In Root's hypothesis the beta manuscripts would derive from the authorial copy in its finally revised form and the gamma manuscripts from a single copy made from Chaucer's working text at a time when most but not all of his revisions had been incorporated. His hypothesis not only provided Root with the means to explain the genesis of three authentic, authorial versions of the poem; it would also account for the confusion of agreements between unrelated manuscripts and the shifting of individual manuscript affiliation from one textual tradition to another, without recourse to coincident variation and contamination. All could be explained by the confusing Chaucerian exemplar, with its additions and cancellations.

Root's account of the textual tradition of the Troilus has not met with complete acceptance. In particular, the privileged status he accords the beta variants has largely been repudiated, and most subsequent scholars have preferred to edit and cite from gamma texts.[3] But the evidence for the earlier, unrevised form of the poem had until recently gone unchallenged. In discussing the text, Windeatt minimizes the importance of the alpha variants, treating the vast majority of them as scribal in origin and the rest as chance preservations of alternative readings present in Chaucer's draft. Windeatt goes to great lengths to demonstrate that certain passages absent from or added later to various alpha manuscripts—Troilus's Hymn to Love in Book 3 (1744-71), his predestination soliloquy in Book 4 (953-1085), and his ascent to the spheres in Book 5 (1807-27)—are not the evidence for revision Root claimed them to be but are, from context, an integral part of Chaucer's original

conception of the poem.[4] Indeed, Windeatt's discussion of these passages is central to his attack on Root's theory of revision, and his conclusions are accepted without further demonstration in Ralph Hanna's highly critical evaluation of Root's textual work.[5] Hanna goes far beyond Windeatt in attacking Root's conception of the poem as inherently flawed. In his view, Root's depreciation of the extent of scribal responsibility for variation allowed him to recognize all variation as possibly authorial, and his selective analysis in The Textual Tradition of Chaucer's Troilus provided him with what he sought, a means of identifying authorial readings "scientifically" and so of generating a text free of editorial interference.

Many of the criticisms levelled by Windeatt and Hanna are well-founded. Certainly coincident variation and contamination play a larger role than Root would allow, and his fanciful hypothesis explaining the confusion of manuscript relationships is more convenient than credible. Much of his enthusiasm for detailing revision states rather than identifying and eliminating scribal error may be explained by the fact that his was the first full account of the new textual theory to be published.[6] But if Root's enthusiasm for revision went too far in one direction, the skepticism of Windeatt and Hanna goes too far in the other, and there is a real danger that a new orthodoxy will simply replace the old. Since the appearance of Windeatt's edition most writers who mention the question of revision in the Troilus seem to regard Windeatt's as the final word, and only Charles Owen has come to the defense of Root's theories.[7] In the new Riverside Chaucer, Windeatt and Hanna are said to have "persuasively undermined Root's argument,"[8] and if this new edition of the Robinson text has anything like the influence of its predecessors, this is the sum total of what many future Chaucerians will come to know and believe about the textual tradition of the Troilus. With this much at stake, a more thorough reexamination of the textual situation is clearly warranted.

The three extended passages mentioned previously, those which Root took to be late additions to the text, have been discussed more extensively by textual critics than any other part of the poem, and because Windeatt's reinterpretation of this evidence underlies his rejection of revision, these passages may serve as a convenient starting point. Troilus's Hymn to Love in Book 3 (1744-71), adapted from Boethius's Consolation of Philosophy Book 2, metrum 8, was originally omitted from Ph and H2, two of the four manuscripts attesting alpha for this part of the poem,[9] although the scribe of Ph has supplied the passage on an inset leaf. The other alpha manuscripts, Gg and H5, give the song in its normal position. As Root comments, the omission in PhH2 might seem to have been accidental, particularly since these manuscripts are related genetically by a long series of common errors and so in agreement have the authority of only a single witness.[10] But there is nothing about the passage that might have led a scribe to omit it; the opening and closing lines of the song contain no verbal similarities to the textual environment that could induce an eyeskip error, and the passage is too short for its omission to have been caused by the loss of a leaf in an antecedent copy.

Nothing about its subject matter would seem likely to provoke a scribe to suppress it. There is evidence, moreover, to support the view that Chaucer's composition of the song may have been independent of his writing of the narrative here. Throughout the latter part of Book 3 (1310ff) Chaucer has been following Filostrato 3, but the song Boccaccio gives his Troiolo had already been used by Chaucer in the Proem to Book 3. In its place Chaucer supplies a different song adapted from the Boethian meter generally taken to have been the source of Boccaccio's inspiration. It is not inherently unlikely that such passages as Chaucer added to the narrative of Boccaccio were composed separately, especially when, as here, the textual evidence suggests independent composition.

This much Windeatt will concede: "What may survive at this point is a randomly preserved trace of the process of Ch's composition of the poem. . . . Such an expansion perhaps existed originally in the form of a physical addition to the draft which has been confused by certain scribes" (p. 38). Windeatt thus agrees with Root concerning the separate composition of the Hymn; he takes issue only with Root's theory that the poem circulated in an early version without the passage. Chaucer always intended to include a song here, he says, for "the context is, and must always have been, utter nonsense without it" (p. 38). To be sure, the context is better served by the inclusion of the song, but its omission does not make "utter nonsense" of the passage:

It is at least possible to take "this manere" to refer to the "wordes" Troilus exchanges with Pandarus about Criseyde and her beauty and womanhood (Owen, p. 158, makes a similar point). Certainly Chaucer does not introduce the song with words as unambiguous as those of Boccaccio, who says that Troiolo sang "qual qui sanz'alcun mezzo si divisa" (3.73,8).[11] It is also instructive to compare Chaucer's handling of the other songs and letters whose words he reports. Troilus's first song, a lament adapted from Petrarch's Sonnet 132, is introduced by the comment that "who-so list it here, / Loo, next this vers he may it fynden here" (1.398-399). The song of Antigone is followed by the line "And of hir song right with that word she stente" (2.876). Troilus "gan synge as ʒe may heere" (5.637) his third and final song. And in their letters to one another in Book 5, both Troilus and Criseyde "seyde as ʒe may here" (5.1316 and 1589). In these cases the reference to the verbatim report of the song or letter is unambiguous. The same cannot be said about the Hymn to Love.

The weight of the evidence, then, supports the independent composition of the Hymn. The only circumstance indicating that its omission might be scribal in origin is the inclusion of the passage in the alpha manuscripts Gg and H5, and even this discrepancy may be easily explained. Like Ph and H2, Gg and H5 are closely related,[12] and we therefore have only two independent witnesses to alpha here, one including and the other omitting the passage in question. There is every possibility that the scribe of some antecedent copy from which both Gg and H5 descend supplied the missing stanzas from another source, just as the scribe of Ph has done. In the end Windeatt must admit that the textual situation here reflects "the processes of Chaucer's composition"; he differs with Root only over Chaucer's intentions. Root held that "Chaucer's text existed for a time without the Boethian hymn to love" (p. 157), and he would attribute its inclusion in the majority of manuscripts to late copies made from Chaucer's working text after its substantial revision. Windeatt argues that Chaucer always intended to include the Hymn and that its absence from alpha may have resulted from its being "in the form of a physical addition to the draft which has been confused by certain scribes" (p. 38). On the independent composition of the Hymn, however, Windeatt and Root are in agreement.

For the predestination soliloquy of Book 4 (953-1085), Chaucer again abandons the narrative of the Filostrato and looks to Boethius, and again the textual evidence suggests that the interpolation was composed independently, although the nature of that evidence is extremely complex. Of the four manuscripts attesting alpha here, only H3 omits the passage in its entirety and without indication that the scribe noticed any irregularity in the text; line 1086 follows immediately after line 952. Gg omits all but the final, "transitional" stanza of the soliloquy, so that line 952 is followed by 1079, without any gap or mark in the manuscript. Ph originally omitted the whole passage, but the scribe has added it on two inset leaves and has marked its point of insertion, just as he had done in the case of the Hymn to Love. In J the passage follows normally, but after line 1078, at the bottom of folio 83r, a note has been appended: "her faileth thing yt / is nat yt made."[13] The back of the folio is blank; the next leaf, last of the quire, has been cut out, and the text resumes with line 1079 after a blank space the length of one stanza, at the top of folio 84r. Because in J there are normally five stanzas to a side, the space left empty corresponds to either six or sixteen stanzas, depending on when and for what reason the missing leaf was removed, but in any case no part of the text has been omitted. To further complicate matters, the entire passage is missing in H4, which is regularly a beta manuscript in this part of the poem.

The possibility that the omission is scribal in origin must be admitted, particularly since manuscripts representing two distinct traditions of the text, one presumably revised and the other unrevised, share in this textual feature. Unlike the Hymn to Love, the predestination soliloquy is long enough (nineteen stanzas) for its omission to have been caused by the loss of leaves in an antecedent copy. But the context of the passage and Chaucer's

treatment of his sources here make scribal omission highly unlikely. How did it happen that a scribe, or scribes, accidentally passed over precisely those lines in which Chaucer is departing from the narrative of Boccaccio? The context is even more significant, for Chaucer's narrative makes perfect, arguably superior, sense without the soliloquy. Pandarus goes from Criseyde and finds Troilus in the temple; if we omit the suspected interpolation, the narrative runs thus: Pandarus makes reference only to Troilus's wish to die; about the long disputation concerning free will and foreknowledge he says nothing. The suspicion that the passage was inserted by Chaucer after his composition of the main part of the narrative is strengthened by the wording of the "transitional" stanza, lines 1079-1085. Troilus has concluded his argument that "the bifallyng / Of thynges that ben wist bifore the tyde, / They mowe nat ben eschued on no syde"; there is a break in his speech at 1079, after which he reasserts his desire to die, and Pandarus, of whom we have heard nothing for over 130 lines, abruptly reenters. Thus the narrative returns to the situation preceding Troilus's soliloquy, and the transition is less than successful from a number of standpoints. For one, the action it relates is not consistent with what has gone before, for Pandarus has already "fond hym al allone" in the temple at line 947, and so his coming in at 1085 repeats an action already described. This may have been an oversight of Chaucer's, but it seems rather to have been deliberately contrived so that Pandarus's response might apply only to Troilus's suicidal lament and not to the soliloquy, which it would seem he should have overheard. Furthermore, line 1079 of the transitional stanza and line 1086 contain a maladroit verbal repetition: Troilus invokes "almyghty Iove in trone," and Pandarus begins his reply "O myghty god . . . in trone." If Chaucer had written these two stanzas successively it seems unlikely that he would have given them such similar initial lines, unless Pandarus's speech is meant to mock that of Troilus.

The context of the soliloquy, then, tends to refute Windeatt's assertion that "the poem is unlikely to have been regarded as authentically completed without the predestination passage" (p. 40). Comparison with the Italian also fails to provide the corroboration he claims. Windeatt maintains that Chaucer's Pandarus "expresses himself much more urgently" than Boccaccio's Pandaro, that this "implies that Troilus is in a much more desperate state than in Fil," and that "without the presence of the soliloquy there is nothing to suggest that this is so or to draw such a response from Pandarus" (p. 42).

As we have seen, however, Pandarus comments only on Troilus's desire to die, something that could easily provoke the response he makes. The difference in emphasis, however real it may be, exists in spite of and not because of the soliloquy.

To return to the textual situation: the evidence of Ph and H3 is unproblematic. There are some variations in their reading of 4.950-952, which immediately precedes the soliloquy where present, but these differences do not indicate that "at any stage of composition represented by the extant manuscripts, Ch intended the predestination soliloquy to be present in his poem," as Windeatt argues (pp. 41-42). The two versions are these:

   

PhH3	REST
he fast made his compleynt and his mone Bysekyng hem to sende hym oþir grace Or from þis world to done hym sone to pace	fful tendrely he preyde and made his mone To doon hym sone out of this world to pace ffor wel he thoughte ther was non other grace

The differences here are minor and, likely as not, scribal. But Windeatt seizes upon the word compleynt in PhH3 and argues that, paradoxically, the manuscripts that omit the soliloquy imply its presence more emphatically than those which contain the passage. His reasoning is specious, however; compleynt and mone are nearly synonymous and neither need refer to anything more than Troilus's suicidal lament. The reading of PhH3 may represent scribal amplification or, if authorial, may have been altered by Chaucer to eliminate the redundancy, but in any case the variations do not affect the question of revision. It remains clear that the soliloquy was absent from the immediate copies used by each scribe, although here as elsewhere the Ph scribe has noted the omission and supplied the passage later from another source. Gg, which contains only the transitional stanza, presents something of an enigma. As both Root and Windeatt note,[14] the stanza of transition implies the presence of the soliloquy, for Troilus is described as "Disputyng with hym self in this matere" (4.1084). The text as it stands in Gg makes no sense, and Root's proposal, that Chaucer "began, not uncharacteristically, at the end, with the stanza of transition"[15] and that copies were made of his working draft in this state, is hard to credit. Windeatt's judgment, that the situation in Gg is scribal, is undoubtedly correct, but he provides no hypothesis to explain how this stanza came to be preserved independently of the soliloquy to which it refers so explicitly.

A partial explanation may be suggested by the even more confusing situation in J. Root (p. 212) notes a generally close agreement between Gg and J in this part of the poem, so it is significant that here again the transitional stanza is divorced from the soliloquy, this time by a blank space left in the manuscript. Much has been written about this feature of the text and about the scribal note concerning the "thing yt is nat yt made." In Root's formulation both note and lacuna go back to an ancestor of J whose scribe was awaiting "matter not yet composed" (p. 217). This hypothesis fits his overall conception of the way the poem was revised, but it is neither necessary nor particularly compelling. As Windeatt points out, the note follows the

soliloquy instead of preceding it, as we would expect if the scribe or one of his predecessors had in fact been awaiting new matter from the poet.[16] A more likely sequence of events would be one in which the scribe of some ancestor of J had as his exemplar a manuscript like Gg in which the transitional stanza appeared without the soliloquy.[17] This scribe either recognized the inconsistency presented by the stanza or, through familiarity with copying the text, realized that some substantial part of the narrative was missing; in any event he left space in the manuscript and supplied the soliloquy later from another source. It would appear from the situation in J that he left too much space and that the resulting gap in the manuscript was taken over by subsequent scribes. The note may have found its way into the manuscript line at any point. It may have originated with the interpolator himself, in which case Root is correct in suggesting that a subsequent scribe must be responsible for moving the note to its present position; or it may reflect the attempt of a later copyist or reader to account for the gap.[18] But to ascribe the condition of the text in J to the circumstances immediately surrounding Chaucer's composition of the poem contributes nothing substantial to the theory of revision.

We are still left with two puzzling situations: the independent preservation of the transitional stanza in Gg and J, and the omission of the entire passage from H4, which attests beta throughout the last three books. Any number of hypotheses might be advanced to explain the first of these, though Root's solution, as noted above, seems more far-fetched than necessary. All that can be said for certain is that the state of Chaucer's production copy made it possible for scribes to pass over either the entire passage or the stanzas of the soliloquy alone. Windeatt suggests that "It would not be unlikely if the long soliloquy were actually put together as an exercise separately and then married to the draft, always with the consequent possibility of scribal misinterpretation" (p. 41); it may be that the stanza of transition was inserted in a different fashion, perhaps in the margin rather than on the separate leaves containing the soliloquy, so that it might be copied independently. This state of affairs would also allow for the omission of the passage from a manuscript like H4 representing a different tradition of the text. Whatever the true situation may have been, the context of the soliloquy and its relationship to Chaucer's sources reinforce the textual evidence suggesting that the passage was composed independently, as even Windeatt will concede, and that it was inserted into the preexisting and logically completed narrative.

The evidence for revision is less clear in the case of Troilus's ascent to the spheres, 5.1807-1827. Although here again Chaucer abandons the narrative of the Filostrato, this time turning to the stanzas of the Teseida (11.1-3) that describe the ascent of Arcita's soul, and although certain manuscripts omit the passage, the witnesses to the text do not divide along the usual lines. Of the manuscripts attesting alpha in Book 5, only Ph originally omitted the three stanzas, and here again the scribe has supplied them on an inset leaf. H3 and J include the passage normally, as presumably would the closely related Gg and H5 had the five leaves containing 5.1702-1869 in Gg and the entire conclusion of the poem in H5, from 4.687 on, not been lost. On the

other hand, H2 and H4, which were alpha manuscripts in the early part of the poem but here attest beta, omit the stanzas.

Again Windeatt is concerned to argue from context that "these stanzas were always to be included where they occur" (p. 39). He bases his assertion on two short phrases included in all manuscripts: "false worldes brotelnesse" (5.1832), words which, he says, "implicitly contrast this world and the next one where Troilus has gone," and "Swich fyn hath his estat real aboue" (5.1830) which "can have little relation to anything that has gone before, except to the hero's ascent to the spheres." This reasoning is prejudiced by knowledge of the content of those stanzas, however. Nothing incongruous would be apparent to a reader whose text lacked the description of Troilus's ascent, for the transition between lines 1806 and 1828 is perfectly smooth and logical:

To Windeatt's arguments it must be objected that the phrase "in this wise he deyde" refers explicitly to the description of Troilus's death at the hands of Achilles rather than to anything related in the three stanzas of his ascent. The connection is obscured by the interposition of twenty-one lines. More importantly, Windeatt's interpretation of "estat real aboue" is questionable; he takes "aboue" in the very narrow sense of physical superiority and would seem to take the line to mean "His royal state had such an end above [i.e., in heaven]." But "aboue" can also connote superiority in rank or power, and the line can be read "His exalted royal state had such an end,"[19] which is closer in meaning to the lines of Boccaccio being paraphrased ("cotal fine ebbe il lucido splendore / che lui servava al solio reale," 8.28,5-6) and which parallels the other lines describing the end of worldly vanities. Taken on its own, the "Swich fyn" stanza in no way assumes the presence of Troilus's ascent.

If the context allows, even suggests, the possibility that the passage was added by Chaucer, the textual evidence does not support the omission as a feature of alpha. In the case of the Hymn to Love, the two alpha manuscripts that include the passage are closely related and can be supposed to have been subjected to scribal interpolation at some point in their shared ancestry.

The one beta manuscript that shares in alpha's omission or textual confusion of the predestination soliloquy may have lost the passage as the result of some disruption in the production copy stemming from the insertion of the passage. In these cases there is good reason to hold that the omissions are genuine features of the alpha tradition, whether or not one believes that alpha represents the poet's first intentions. But the omission of the ascent to the spheres cannot be similarly classified. In this instance Windeatt is probably right to ascribe the textual situation to a "muddle by early scribes over a passage which was perhaps available to them separately because it did not form part of the author's main source" (p. 39). On the other hand, he is not justified in concluding further that "the passage was certainly added in the composition but not necessarily, indeed improbably, in a revision" (p. 40), a judgment based solely on his dubious arguments from context. We have no way of knowing when Chaucer interpolated the three stanzas, although the coincidence of his departure from the Filostrato with a disruption in the text again suggests independent composition. Root may be guilty of "denigrating as scribal the peculiarities of H3 and H2H4" in order to maintain the omission as a feature of alpha, as Windeatt argues,[20] but Windeatt is himself guilty of distorting the contextual evidence in accordance with his skepticism about revision. We again find context, treatment of source, and variation in the manuscripts all pointing to the insertion of a passage into the preexisting narrative, even though in this instance no particular tradition of the text can be associated with one state of the poem or the other.

There are, in addition to these large-scale variations between the manuscript groups, several instances of a more limited nature in which, nevertheless, the differences between the textual traditions may be attributed to the hand of the poet. A single stanza in Book 1, lines 890-896, is preserved only in PhH2H4, the manuscripts attesting alpha here, and in the early printed edition of Thynne, which presents a conflated text agreeing with alpha on occasion in Book 1.[21] The stanza comes soon after Troilus's revelation that Criseyde is the object of his love, whereupon Pandarus exhorts Troilus with these words:

The authenticity of the stanza is not disputed, although the narrative of Boccaccio does not provide a test in this case. Hanna argues that the omission in the majority of manuscripts may be scribal, the result of eyeskip error encouraged by the appearance of "And" as the first word in this stanza and the one following,[23] but the similarity seems rather small to have diverted a number of scribal eyes. If the variation is due to Chaucer, it is possible that the stanza was either added or deleted after the entire passage had been written. There would seem to be no good reason for the poet to have taken the trouble to insert this one stanza, and it is assumed by both Root and Windeatt that he cancelled it as digressive.[24] In this instance, at least, they are agreed on a feature of alpha that appears to represent an early form of the text.

In two other cases the evidence for revision involves not the presence or absence of material but the order of stanzas in one tradition of the text as opposed to another. In each instance both textual orders are possible, although one is demonstrably closer to the reading of the Filostrato. In the central love scene of Book 3, the narrator introduces two stanzas in his own voice in which he subjects his poem to the correction of lovers:

The voice here is distinctively that of Chaucer's narrator, the "sorwful instrument / That helpeth loueres, as I kan, to pleyne," but the ideas are developed by Chaucer from part of a single stanza of Boccaccio's, Filostrato 3.33: In alpha and gamma, Chaucer's adaptation of these sentiments appears in the same position as in Boccaccio, between a stanza beginning "O blisful nyght of hem so long i-sought" and another describing the lovers' wonder at finding themselves together at last, a close translation of Filostrato 3.34. But in H3JRCx, four of the five witnesses attesting beta here, the two stanzas are shifted to a position after 3.1414, and in the fifth beta manuscript, H4, the stanzas appear in both places. In the later arrangement the narrator's comments do not interrupt the flow of the story, for they follow the description of the night's activities and are succeeded by the intrusion of the dawn.

The position of the stanzas in beta may be scribal, but if so the new

arrangement could hardly have come about through unconscious scribal activity; nothing in the nature of the text would seem likely to induce the shift of so short a passage so far into the narrative. There are, moreover, certain variations in individual readings that seem designed to accommodate the shift, and these are taken by Root to be "important evidence to the deliberate nature of the shift" (p. 157). In alpha and gamma, line 1323 immediately precedes the stanzas and describes the lovers' bliss "That is so heigh that al ne kan I telle." In beta, the line becomes "That is so heygh þat no man kan it telle," presumably because the narrator's protestations of his inability do not here follow. The first line of the shifted stanzas is also altered in beta, so that it begins "But how al thogh" rather than "But sooth is, though."[25] Finally, whereas in most manuscripts line 1415 reads "But whan the cok, commune astrologer," in beta the line begins "Whan that the cok," which avoids the juxtaposition of two lines beginning "But" in these manuscripts.

Windeatt admits that the order in beta is just as possible as that in alpha, but he is more inclined to explain the differences as scribal. The thrust of his argument is that the stanzas fit the earlier context better:

The stanzas may be more effective in their alpha/gamma position, but this proves nothing more than that they were originally composed for this context; it does not mean that they were not or could not have been moved by Chaucer. As to the "number of bizarre scribal errors," the only example cited by Windeatt involves five stanzas in R, 3.1212-1246, that are copied twice, once in their normal position and again after 3.1099. In this instance, however, the stanzas are shifted to an earlier position, which is just what we would expect if a scribe were to mistake the point at which he was to resume copying but later discover his error and begin again where he had in fact left off. There are only a very few genuine examples of a shift of stanzas to a later position in the text, and in none of these is there any evidence of the new context being prepared for the transplanted material.[26]

The repositioning of the narrator's apostrophe to lovers, far from being a "bizarre scribal error," represents a deliberate alteration of what must have been Chaucer's original plan for the poem. The responsibility for the shift may lie with the scribes; they were certainly not averse to introducing changes into the texts they copied, even those which, like the Troilus, do not invite and in some ways discourage their active intervention.[27] A motivation for the change is easy to imagine and would presumably be the same as that attributed to Chaucer, the desire to remove the passage to a less intrusive location. But if we are to accept this explanation of the textual situation, we must attribute to the scribe a conscious alteration of the text, involving the rewriting of lines to fit the stanzas to their new context, which at the same

time changes the overall structure of the scene but little. It is difficult to imagine scribes going to so much trouble to so little effect, and there is no evidence that they undertook such drastic revision elsewhere in the text of the Troilus. On the other hand, such a change can easily be ascribed to the workings of the poet, who would be more likely to take trouble over subtle arrangements of his text.[28] If Chaucer was indeed responsible for the shift, as the weight of the evidence seems to indicate, we have in this instance one of the very few examples in which beta alone preserves what appears to be a revised form of the text.[29]

The second case of revision involving the shift of a stanza comes in Book 4, in a passage describing the visit paid Criseyde by her friends upon learning that she is to be traded to the Greeks in exchange for Antenor. In Boccaccio the women take their leave, whereupon Criseida breaks down in grief:

Although in translating Chaucer renders Boccaccio's single stanza in something more than two, his Criseyde goes through the same motions as does Criseida. But the sequence of events Boccaccio relates is observed among the Troilus manuscripts only in alpha, here attested by PhGgH3J. In these manuscripts, stanza 108 of Book 4, lines 750-756, which describes Criseyde's weeping, beating her breast, and calling on death, follows the lines in which she falls on her bed (733-735) and is followed in turn by a stanza in which she tears her hair and again invokes death before beginning a long complaint (736-742). In all other witnesses to the text, stanza 108 follows the first stanza of this complaint, interrupting Criseyde's speech with a description of her actions taken from the central portion of Boccaccio's stanza. There is also significant variation in the manuscripts over the reading of 750-752, the opening lines of stanza 108, and of 757, which begins the second stanza of Criseyde's complaint and so in beta and gamma immediately follows the shifted passage. The two versions of 750-752 are these:[30]    

PhGgH3J	REST
The salt teris from her eyen tweyne Out ran as shour in Aprill ful swithe; Her white brest she bet & for þe peyne	Ther-with the teris from hire eyen two Down fille as shoure in Aperil ful swithe; Hire white brest she bet and for the wo

The variations between "tweyne . . . peyne" and "two . . . wo" as well as between "Out ran" and "Down fille" may be scribal and in any case do not affect the question of revision. The important difference lies in the initial wording of each version. Neither "The salt teris" nor "Ther-with the teris" is closer to the Italian, but each is appropriate to the stanza in its particular context. In alpha the stanza comes in the course of a descriptive passage and so requires no transition from what precedes. In beta and gamma, "Therwith" signals a break in direct speech, and although the change is not strictly necessary, it does demonstrate the deliberate nature of the shift. Even more significant are the variations in line 757, which in beta and gamma marks a return to direct discourse: "She seyde, 'how shal he don and ich also?'" In alpha the speech prefix is unnecessary, and the line reads "What shal he don, what shal I do also?"[31]

It seems clear that, as with the two stanzas in Book 3, the variations between alpha and beta/gamma represent a deliberate alteration of the text, and that alpha must approximate what Chaucer originally wrote. The alternative, that Chaucer composed the passage with the stanza in the beta/gamma position but that either he or some scribe later moved it, is hardly likely. Chaucer could have done so, but for the poet to revise his text to conform to his source after having written the stanzas in a deliberately original order would defy all reasonable expectation; for a scribe to restore the sequence of the Italian would require a degree of luck or intuition difficult to credit. The converse possibility, that it is the beta/gamma arrangement that is scribal, must be admitted, but the same arguments against a scribal explanation for the shift in Book 3 apply here, and to deny these would be to indict every editor of the Troilus, from Caxton to Windeatt, for printing the non-Chaucerian order in the present passage. Windeatt cannot get around these facts, although he again attempts to denigrate the authority of alpha:

Windeatt is forced to admit in this instance what he has taken such pains to deny elsewhere, that alpha preserves an early form of the text revised by Chaucer at some point after the major task of composition.

In nearly all of the larger variations between the textual traditions, Windeatt's arguments from context fail to confute the evidence for revision. Only in the case of the Hymn to Love does he present a credible argument that the form of the received text must always have been that intended by Chaucer, and even here he will admit that the composition of the passage was independent of Chaucer's writing of the narrative. Only in the case of Troilus's ascent to the spheres does a suspected revision cut across the familiar lines of the textual traditions. The six examples thus far considered provide more than sufficient evidence that Chaucer revised his text, however locally and at whatever point in the process of composition. This evidence justifies a careful examination of the manuscript variants to determine whether any of these may be due to changes made by the poet.

In evaluating the textual work of Manly and Rickert, George Kane considers the question of revision in the text of the Canterbury Tales:

External evidence of the kind that allows detailed analysis of a poet's method of composition, such as exists in the holograph manuscripts of some of Petrarch's poems, does not survive for any of Chaucer's works,[33] and in the great majority of the Troilus variants the editorial judgment Kane speaks of is the only tool available to decide the question of revision. There are a number of instances, however, in which comparison with Chaucer's source can provide a kind of external information against which to test the priority and originality of readings. Variants closer to the source in detail or in phrasing must be considered the work of Chaucer; to believe that any particular scribe had both a sufficient knowledge of Italian and the access to a manuscript of the Filostrato that would allow him to make such alterations is difficult enough, without further crediting him with taking the time and trouble to compare the poems line by line.[34] When these Italianate readings accompany others whose authenticity is accepted, either because of overwhelming manuscript support or because of their "Chaucerian" qualities, authorial revision may be supposed to be behind the variation, and the direction of the revision can be presumed to have been away from the Italian.[35] The logical alternative would be to discard as scribal the readings furthest from the Italian, even where these form part of the received text, as they so often do. No editor has so far been willing to take this step.[36]

Root's argument for the primacy of alpha rested in large part on some striking variants confined to certain portions of the poem, most notably 1.1-1.300, 2.701-2.1113, and 3.400-4.1450. Windeatt accepts some of these as offering "an intriguing glimpse into the process of the poem's composition" (p. 44), but both he and Hanna are quick to point out that, in Hanna's words, "at many points by his [Root's] own account β or γ resembles the Italian more nearly than does a."[37] This state of affairs should hardly be surprising, however, considering the relatively extensive corruptness of the alpha manuscripts in general and Ph in particular, compared with the manuscripts of beta or gamma. The fact remains that there are a number of alpha variants that are remarkably closer to the Italian at points in the text where equally striking and apparently authorial readings are also to be found, while the converse situation is extraordinarily rare, although one could not infer this from the very general comments of Windeatt and Hanna. Clearly a more detailed analysis would reveal whether there is any justification for their objections to what Hanna sees as Root's belief in "a's unique similarity to the Italian" (p. 199).

In selecting the evidence for analysis, some criteria must first be established.

We are concerned here only with variations meeting the following conditions: 1) Chaucer is following his source, usually the Filostrato; 2) the traditions of the text are at variance; 3) the alternative readings are not manifestly corrupt; 4) one reading is closer to the source than the others. A list of some 103 examples was compiled by McCormick and revised by Root for publication by the Chaucer Society,[38] and these together with a few additional cases noted by Windeatt in his Commentary may serve as a basis for analysis. To begin with the more striking instances in which alpha appears closer to the Italian, we find a number of significant variants in two stanzas of Book 1, lines 78-91:[39]

• 1.78 a+Th: Wherfor to departe all softely
• Rest: ffor which forto departen softely Per che segretamente di partirsi (1.9,1)
• 1.83 a: Hopyng in hym kunnyng hem to rede (kunnyng hem] them k. W)
• Rest: In trust that he hath konnynge hem to rede da lui sperando sommo e buon consiglio (1.9,7)
• 1.85 a+Th: Grete rumour was whan hit was ferst aspyed (rumour] noyse W. was] gan H2H4; began W; rose Th)
• Rest: The noise vp rose whan it was first aspied Fu'l romor grande quando fu sentito (1.10,1)

In the first instance, despite Windeatt's comment that "Phetc is further from Fil," both Wherfor and ffor which are equally plausible and idiomatic translations of the Italian, and in fact the former is used by Chaucer to translate per che on three occasions in the Troilus (1.430, 1.988, 3.281). In the other examples, however, alpha is much closer to the Italian in translating sperando as hopyng and 'l romor grande as Grete rumour,[40] and while Root may be correct in calling the first of these "a rather bald translation," Windeatt admits in both instances that alpha "perhaps retains a trace of Ch's translating process."

• 1.110-111 a: Byfor Ector on knees she fyll adoun (+CxTh) Wiþ chier & voys ful pytous and wepyng (chier &] clere, W)
• Rest: On knees she fil biforn Ector adown (b.E.s.f. GgH5) With pitous vois, and tendrely wepynge ginocchion si gittò a piè d' Ettore, e con voce e con vista assai pietosa (1.12,5-6)

Alpha retains the Italian vista, which is not translated in the other manuscripts. Windeatt concedes the point, but he quibbles about the variant word order in 110, where alpha is "further from [Filostrato] 12/5," when in fact the more natural word order in alpha marks the variant as typically scribal and therefore irrelevant to the discussion of proximity to the source.

• 1.124 PhH2H4: And she hym þonkyd oft in humble chere
• Rest: And she hym thonked with ful humble chere Ella di questo il ringraziò assai (1.14,5)

The variation here is slight; alpha's oft (a reading not shared by W) may retain the original assai, as Windeatt notes.

• 1.169 a: Among þe which was Criseida (was] was this H2; was than W)
• Rest: Among thise othere folk was Criseyda Tra li qua' fu di Calcàs la figliuola Criseida . . . . (1.19,1-2)

Root and Windeatt note alpha's more literal translation of Tra li qua', and although its form of the line is not metrically regular, it is unlikely to have arisen by scribal substitution for the more explicit thise othere folk.

• 2.734-735 a+Th: Men love wymmen al this towne about; (al] þour al, Gg) Be they þe wors? [why] nay, without dout (why] GgH5Th; om PhH2)
• Rest: Men louen wommen al biside hire leue, And whan hem leste namore, lat hem byleue Io non conosco in questa terra ancora niuna sanza amante . . . . . . . e come gli altri far non è peccato, né ne può esser alcun biasimato. (2.70,3-4 and 7-8)

Alpha is closer to the Italian in including al this towne about, suggested by in questa terra, and in Criseyde's rhetorical question, which retains the rationalization of Boccaccio's heroine that it is no sin to do as others do.

• 3.1482 aβ: Seth þat desire right now so streynith me
• Rest: Syn that desire right now so biteth me (biteth] brenneth Cl; bitleth H1) sì mi stringe il disio del ritornarci (3.46,2)

The beta manuscripts join alpha in the majority reading streynith, which translates stringe much more literally than biteth, the reading of only five manuscripts and a possible scribal substitution. But Windeatt joins other editors in regarding biteth as "authentic . . . and stronger," which leads him inevitably to the conclusion that because alpha "must stem from Ch's translating," the readings reflect authorial revision.[41]

• 3.1617 aβ: Til al was seyd, and þan he þs answerde
• Rest: Tyl al was seyde, and than he hym answerde (hym] thus him H3) . . . e poi così rispose lieto a' detti suoi (3.59,7-8)

Beta again joins alpha in a minor variation that does, nevertheless, reflect the Italian (così) more closely than the alternative reading. Windeatt rejects hym as scribal, reasoning that the word has been attracted from the preceding line ("And Pandarus ful sobrely hym herde"), but modern editors are divided as to whether to emend the line.[42]

• 4.37 PhJ: Þis purpos & þat day þei [issen] mente (issen] J; issu Ph)
• Rest: This purpos and that day they fighten mente (they fighten mente] t. fouhten m. H4; þe þus m. H2; of assignement H5) incontro a' Greci uscì ne' campi piani (4.1,6)

Issen, "to go out," translates uscì literally, and there can be little doubt that, as Windeatt puts it, the reading "retains here a trace of Ch's translating," even though Chaucer is not following Boccaccio line by line at this point and, in fact, 4.39 may be a closer paraphrase of the Italian: "Ector and many a worthi wight out wente." Of the manuscripts reading issen, only Ph attests alpha here, although the confusion in H2 and H5 may well stem from the scribes' attempt to make sense of a difficult term, and the remaining alpha manuscript, Gg, is out at this point, having lost two leaves. It is thus possible to regard issen as a genuine feature of alpha, although J does not normally give alpha readings until 4.431.[43]

• 4.246-247 α: His eyen too for pyte of his hert So wepyn þat þei semyn wellis twey
• Rest: His eyen two, for piete of herte, Out stremeden as swifte welles tweye; Li miseri occhi per pietà del core forte piangean, e parean due fontane (4.28,1-2) (Piangono si che paion due fontane, 5.24 in the Baroni ed., Paris 1789)

As both Root and Windeatt note, alpha is clearly closer to the Italian, particularly in the form of the line presented by Baroni in the Paris edition. At the same time, the reading Out stremeden as swifte is, in Windeatt's words, "superior and obviously authentic" and must therefore have come about as the result of authorial revision.

• 4.258 α: But þo bygan his teris out more to breste Þat wele vnnethe þe body may suffise
• Rest: But tho bygonne his teeris more out breste, That wonder is the body may suffise che 'l capo e 'l petto appena gli bastava (4.29,7)

Alpha's wele vnnethe translates the Italian appena, "scarcely," while "the reading wonder is also seems authentic in a context where Ch has doubled the extent of T's suffering" (Windeatt).

• 4.290 α: How shal my sorowful lyf don in þis cas? (sorowful] reuful Gg)
• Rest: What shal my sorwful lif don in this cas? come farà la mia vita dolente (4.33,3)

A minor variation, but alpha's How does translate come with greater precision. Root does not attach much significance to this example, but Windeatt will admit that alpha "possibly reflects Ch's translation."

• 4.318 α+ADH2: Who shal now yeue comfort to my peyne? (my] your H3)
• Rest: Who shal now ʒeuen comfort to the peyne? (the] thy S1Th) chi darà più conforto alle mie pene? (4.36,3)

Considerable variation in the reading of this one word may mark the example as scribal, despite the coincidence of the Italian mie. In particular, the agreement of the nonallied ADH2 and the coincident variation of S1 Th and H3 in a third reading make difficult any claim for authorial revision here. Root and Windeatt do not comment.

• 4.590 α: Ne preciously, but help thi self anone
• Rest: Ne corteisly, but help thi selue anon (corteisly] preciently R; curyously Cx) Non guarda amor cotanto sottilmente (4.72,1)

Windeatt remarks that "preciously is perhaps closer to sottilmente," and in fact, as Root notes, corteisly does not fit the context of the English or the sense of the Italian, and it may be a corruption of curyously. Most editors nonetheless accept corteisly as Chaucer's.[44]

• 4.596-597 α: Hit is no [rape] in my dome ne no vyse Hir to witholdyn þat þe louith moost (rape] iape PhGg. no(2)] om GgH5)
• Rest: It is no shame vn-to ʒow ne no vice Hire to withholden that ʒe loue most Tu non hai a rapir donna che sia dal tuo voler lontana . . . (4.73,1-2)

Though Chaucer's paraphrase is not exact in either case, alpha retains the Italian rapir,[45] while the other reading is also apparently authentic. Windeatt admits that the variation "may reflect Ch's translation process here."

• 4.762-763 α: And cursid be þat day which þat Argyue Me of her body bare to bene a lyve
• Rest: O moder myn, that cleped were Argyue, Wo worth that day that thow me bere on lyue! Oh, trista me, che'n mal punto fui nata, dove ti lascio, dolce l' amor mio? Deh, or foss' io nel nascere affogata (4.88,3-5) Mal' abbia il giorno, che al mondo fui nata, E che di me mia madre ebbe desio! (5.73 in the Paris ed.)

Windeatt comments that "The Phetc [alpha] variant perhaps reflects the reference to birth in Fil 88/5," but Chaucer's paraphrase is even closer to the corresponding lines in the Paris edition, which Root claims "correspond more nearly to α than βγ." In fact, both versions seem to owe something different to the variant readings of the Paris edition, and it is likely that both are authorial.

• 4.820 α: Whan she hym sawe she gan for shame anone
• Rest: Whan she hym saugh she gan for sorwe anon La qual come lui vide, fra la braccia per vergogna nascose la sua faccia. (4.96,7-8)

Alpha's shame translates vergogna precisely, and Windeatt sees "a trace here of Ch's translating process."[46]

• 4.882 αβ: As he þat [shortly shapith hym] to dey (shapith hym shortly, Ph; shortly he þat, H2H4)
• Rest: ffor verray wo his wit is al aweye il qual del tutto in duol ne vuol morire (4.102,8) Che cerca disperato di morire (5.84 in the Paris ed.)

The reading of alpha and beta is clearly closer to the Italian, while the other version "offers a somewhat simpler sentence structure" in Root's view and, according to Windeatt, improves on Chaucer's first attempt, which "seems clumsy coming before 883-884."[47]

• 4.906 α: To se hym in þat woo þat he is ynne
• Rest: To sen that sorwe which that he is inne ma più m' è di veder Troiolo afflitto (4.105,2)

The syntax of alpha (Pandarus sees Troilus in woe) is somewhat closer to the Italian than that of the other version (he sees the sorrow Troilus is in), as Windeatt and Root note.

• 4.1124α+RH2H4: But whan þat hit was tyme for to go
• Rest: And whan that it was tyme for to go . . . ma quando tempo gli parve di dovere andare (4.113,3-4)

All but one of the witnesses to beta join alpha in this minor variation, which is, nonetheless, closer to the Italian (ma). But is actually more appropriate in context, for Troilus has "seyde vn-to it more," but now he will, "withouten mo, / Unto hire com." The reading And, while generally preferred, may have been attracted from the preceding line. Windeatt accepts the reading of his copy-text without comment.

• 4.1214 α: And he answerid, "hert myn, Criseyde, . . ."
• Rest: And he answerde, "lady myn, Criseyde, . . ." A cui el disse:—Dolce mio disiro (4.124,3)

Although Windeatt does not comment, the alpha reading may retain something of the Italian not found in the other manuscripts. Root notes that "All MSS. read herte myn in 1216, which may explain the revision," but by the same token it might be concluded that alpha preserves a scribal error of anticipation.

• 4.1218 α: And he bygan comforte her as he might
• Rest: And he bigan to glad hire as he myghte . . . e 'l suo martiro, come potea, con parole alleggiando, la confortò . . . (4.124,5-7)

Alpha's comforte is clearly closer to the Italian, as Root and Windeatt note, but to glad has generally been preferred and may be the result of revision.[48] It should be said, however, that all manuscripts read And hire to glade at 1220, and it might be argued that the majority reading at 1218 represents another scribal error of anticipation.

• 5.1028 α: The causes why, the soth for to telle (causes] cause H3)
• Rest: The cause whi, the sothe forto telle e da queste cagion sommossa . . . (6.34,7)

A minor variation, one in which H3 does not share, but the plural causes is closer to the Italian (queste cagion), as Windeatt and Root note.

These twenty-four examples represent only a portion of those readings in which alpha may be thought to preserve an early stage in Chaucer's composition of the poem; they do not by any means include all of the most interesting or significant of the alpha variants, but have been selected because in these cases the Filostrato provides a test of authenticity. Windeatt (p. 43) may be justified in describing them as being "of very varying quality," but they are all arguably closer to the Italian, and in at least half of the examples the resemblance is so striking as to rule out coincidental similarity. Windeatt relegates discussion of this evidence largely to his Commentary, and his typical remark is that alpha "perhaps retains a trace of Ch's translating process." In his argument dismissing alpha's claim to priority, his examination of specific resemblances to the Italian is so brief that it may be quoted in its entirety:

To this he adds in a note: Having thus dispensed with alpha, Windeatt notes "the many instances throughout the poem where the other MSS are closer to the Italian than Phetc" (p. 43), and with that comment rests his case. Hanna remarks that "Root attempted to handle this problem by defining as scribal errors all α passages which do not resemble the Italian more nearly than βγ" (p. 287, n. 17), but neither he nor Windeatt defends these assertions. Certainly there are many instances throughout the poem where an individual manuscript or a manuscript group can be shown to be further from the Italian than the other manuscripts; the question is whether such variants have any claim to authority. In many of the examples analyzed above, the readings of the received text are widely regarded to be the work of Chaucer, despite the existence of variant readings closer to the Italian. In most cases, however, the readings further from the Italian will likely prove to be scribal, and in fact most are manifestly corrupt. As a result, it may be fair to ask whether the remarks of Windeatt and Hanna regarding the "many instances . . . where the other MSS are closer to the Italian than Phetc" are germane to the question of revision. Are there, in fact, "many instances" in which alpha preserves a reading further from the Italian that is not a clear case of error? The examples noted by Windeatt—Hanna provides none of his own—should resolve the matter.

• 1.57-58PhH4W: Knowyn þing is how þat þe grekys strong (þing] well it W) Wiþ armes in a m. shippes went
• Rest: Yt is wel wist how that the Grekes stronge In armes with a thousand shippes wente Erano a Troia li greci re d' intorno nell' armi forti, . . . . (1.7,1-2)

Windeatt remarks that "in 58 Phetc With for In is further from nell' armi," but in fact not all of the alpha manuscripts give this reading, and in any case the change is merely one of word order.

• 1.117-118 α: And seyd, "lete ʒour fadris tresoun gone To sory hap, & ʒe ʒour self in ioy . . . ."
• Rest: And seyde, "lat ʒoure fadres treson gon fforth with meschaunce, and ʒe ʒoure self in ioie . . . ." dicendo:—Lascia con la ria ventura tuo padre andar che n' ha offeso tanto (1.13,5-6)

Windeatt is correct in noting that "Phetc's reading To sory hap is further from con in Fil 13/4 [sic]," although it could be argued that in translating ria ventura as sory hap, alpha is somewhat closer to the Italian. But meschaunce is undoubtedly Chaucer's, and in any case To sory hap would be a familiar expression (cf. Romaunt 7582) and therefore an easy scribal substitution.

• 1.159 α: In meny wise shewed, as y rede
• Rest: In sondry wises shewed, as I rede e 'n diversi atti mostra suoi amori (1.18,4)

Windeatt remarks that the "Phetc reading meny is further from Fil here," and sondry is in fact a more exact translation of diversi. Meny is easier and is for that reason a likely scribal substitution.

• 1.272 PhH2H4: His eye perceyvid, and so depe hit went
• Rest: His eye percede, and so depe it wente (percede] procede CpCl; preceded D; persedyn Gg; departed R; perceded S2) l' occhio suo vago giunse penetrando (1.26,5)

The variations here are so many and so obviously scribal that Windeatt's comment ("Phetc [H2H4Ph] is corrupted away from Fil") is entirely gratuitous, and in fact one witness to alpha, W, reads persed. A difficult expression and confusion over abbreviation evidently compounded the errors here.

• 1.407 PhH2H4: And if y in myn owne lust brenne (y] om H4. brenne] I b. H4)
• Rest: And if that at myn owen lust I brenne S' a mia voglia ardo, onde 'l pianto e lamento? (Petrarch, Sonnet 132)

As Windeatt notes, the majority reading is "closer to Petrarch (S' a) than Phetc reading in," but the simple substitution of an easier expression marks the variant as scribal. Note that W again fails to attest a supposed alpha reading further from Chaucer's source.

• 1.483 α+R: That all þe grekys as the dethe hym dredde (all þe] alle R; all H4)
• Rest: That the Grekes as the deth him dredde che li Greci il temien come la morte (1.46,8)

A beta manuscript joins alpha in adding all, which is not in the Italian. The addition of a word that improves the meter might be charged to the poet or the scribes; the variation here is of no importance.

• 2.1093α+JRH4: This Pandare vp þer with, & þat betyme
• Rest: This Pandare tok the lettre, and that by-tyme (the lettre] om H3) Pandaro, presa la lettera pia (2.108,1)

Alpha and three of four beta manuscripts omit Pandarus's taking the letter and so are further from the Italian. No scribal explanation for the change suggests itself, but the omission of the lettre in H3 would seem to indicate some confusion in the manuscripts having nothing to do with revision.

• 3.349 α: And al þe rehetyng of his sikes sore
• Rest: And al the richesse of hise sikes sore (richesse] tresour H4; thoughtis Cx) Li sospir, ch' egli aveva a gran divizia (3.11,5)

The majority reading richesse reflects the Italian divizia; the variant readings of alpha, here attested by only two manuscripts, and of H4 and Cx look very much like scribal substitutions, although rehetyng, "assailing, attacking," is in many ways the harder reading.[49]

• 3.1360α+others: And wel a C tymes gan he syke
• β+SI: And wel a thousand tymes gan he syke . . . e nessuna ora sanza mille sospiri valicava (3.37,3-4)

Alpha and a majority of the other manuscripts give a reading further from the Italian, but either may be the original; the error is a common one perhaps occasioned by a confusion over the interpretation of Roman numerals.

• 4.238 α: In his distresse, as y shall ʒow devise (distresse] distreste Gg)
• Rest: In his woodnesse, as I shal ʒow deuyse ch' uom non parea, ma arrabbiata fera (4.26,8)

Windeatt claims that woodnesse is the "harder reading, and closer to Fil, than distresse in Phetc," but although Chaucer is following Boccaccio closely in the preceding lines, the line in question bears only a slight resemblance to the Italian, and neither reading can thus be said to be "closer to Fil." Alpha's distresse is easier, however, and can for that reason be dismissed as scribal.

• 4.295-296 α: What shal y do? y shal whil y may dure In woo, in turment, and yn cruel peyne
• Rest: What shal I don? I shal, while I may dure On lyue in torment and in cruwel peyne Io piangerò e sempre doloroso starò dove ch' io sia, mentre la vita mi durerà 'n questo corpo angoscioso! (4.34,1-3)

Alpha is further from the Italian than the other manuscripts, which translate mentre la vita / mi durerà as while I may dure / On lyue. Alpha's In woo is easier, as Windeatt notes, and probably represents scribal amplification upon the torments elaborated in the remainder of the line.

• 4.306 PhGgH5: ffle forth a none and do myn hert to brest (to] om GgH5)
• Rest: ffle forth out of myn herte and lat it breste esci del core e Criseida segui (4.34,7)

H3, which joins alpha at 4.300 and generally adheres to its readings throughout the rest of the poem, does not share in this minor variation. Windeatt is correct in saying that "Phetc's line is further from Fil 34/7," but alpha does not "omit" del core, as he claims, but instead employs hert in a syntactic construction different from that in the Italian. The change is typical of scribal activity.

• 4.347 α: ffor Antenore to chaungyn [so] Criseyde (so] GgH3H5; om Ph)
• Rest: ffor Antenor to ʒelden so Criseyde e come aveano ancora per partito preso di render Criseida i signori (4.43,3-4)

The majority reading ʒelden reflects render more accurately than alpha's chaungyn, an easy scribal substitution in this context (cf. 4.665, where chaunge is used by Chaucer in a similar construction).

• 4.506-507 α: Or deþe me slowe, y wold have ʒevyn hire But now his comyng is to me so swete (his]om H3J. comyng is] is c. J)
• Rest: Er thow me slowe I wolde haue yeuen hire; But now thi comynge is to me so swete Morte, tu mi sarai tanto soave, quant' è la vita a chi lieta la mena (4.61,1-2)

The direct address of death in the majority of the manuscripts follows Boccaccio's practice in the corresponding stanza of the Filostrato, though Chaucer is not translating the Italian verbatim here. Alpha exhibits the scribal tendency toward greater explicitness.

• 4.537 α: Rys vp anone and lete þis sorow be
• Rest: Ris vp anon and lat this wepyng be caccia via il dolor, caccial via, caccia l'angoscia tua e li dolenti guai, rasciuga il tristo pianto della faccia (4.65,4-6)

The Italian provides support for both sorow and wepyng (dolor and tristo pianto); the majority reading is probably Chaucer's, though either would be an easy scribal substitution for the other.

• 4.602 α+Cx: And fleith from wrecchis for hir cowardise
• Rest: And weyueth wrecches for hire cowardise . . . La Fortuna aiuta chiunque ardisce e' timidi rifiuta. (4.73,7-8)

Alpha's fleith from is probably further from rifiuta, "spurns," than the majority reading weyueth, "neglects, abandons." The substitution of an easier expression may be suspected here; compare 2.284, where the "good" manuscripts Cp and Cl have corrupted weyuen, here in rhyme position.

• 4.732 αβ+Si: In to þe chambre went vp out of þe hall (went vp] Ph; w. GgH2H4; rest vp w.)
• Rest: In-to hire chambre vp went out of the halle nella camera sua piangendo piano se n' entrò dentro . . . (4.86,5-6)

A minor variation in which half the manuscripts correctly reflect the Italian (camera sua) while the others make the easy substitution of the definite article for the personal pronoun.

• 4.1134 PhGg: As y seyd arst, for sorow and [for] wepyng (for(2)] Gg; om Ph)
• Rest: As I seyde erst, for wo and for sobbyng (wo] sorwe H3) sì gl'impedivan gli angosciosi pianti e' singhiozzi e' sospiri, . . . (4.115,4-5)

Windeatt claims that "Phetc omits singhiozzi" (p. 52, n. 10), but in fact only two of four alpha manuscripts actually read wepyng for sobbyng, which are in any case interchangeable in scribal practice.

These eighteen examples include all of the "many instances" noted by Windeatt where alpha is further from the Italian. Most of these he treats only in his Commentary, but in a note supporting his contention that alpha's similarities to the Italian are not unique, he singles out a number of them, among which are several of the least compelling.[50] If the Italianate readings in alpha are "of very varying quality," as Windeatt suggests, these are not; with hardly an exception, they are uniformly scribal. Only alpha's rehetyng (3.349) can be

defended as difficult enough to merit attention. Furthermore, in five instances the variant further from the Italian is not attested by all alpha manuscripts, and in another five cases other manuscripts join alpha in the divergent reading. The same is true in nearly all of the remaining examples in which variation may be tested against Chaucer's sources. These need not be examined in detail and have instead been summarized in an appendix. In most cases at least some of the alpha manuscripts are among those witnesses that are close to the Italian, although the resemblance is often trifling and may be coincidental. The same is true where alpha does not reflect the original, and in such cases other manuscripts join alpha in what look like easier readings. The evidence suggests that Root was correct in ascribing such variation to scribal error, despite Hanna's implication that he thus justified an unwarranted belief in "α's unique similarity to the Italian." To say this is not to disagree with Windeatt's view that "The evidence for a sustainedly distinct version or state of Ch's text, earlier and closer line by line to the Italian source, does not exist" (p. 43), an idea never advanced by Root and not defended here. It is to say that where remarkable variants undeniably closer to the Italian exist side by side with other, equally authentic readings, the Italianate form is invariably preserved by alpha. This state of affairs reinforces the evidence from the large-scale variations suggesting that in at least some of its particulars alpha reflects an early stage in Chaucer's composition of the poem.

What then may be said of Windeatt's view of the text of Troilus and Criseyde as compared with the theory advanced by Root? Although the case made here places greater emphasis on alpha's priority than does Windeatt and disputes a number of his major conclusions, it is not intended as a defense of Root's overall conception of the textual tradition. In fact, the view taken here may in some ways be more compatible with Windeatt's position than with Root's. Root was concerned to show that Chaucer "published" his poem in an early version, and that he later made changes in the publication copy supplied by his scribe Adam. Quite apart from the lack of any evidence to support it, this hypothesis involved Root in some logical inconsistencies, as Hanna notes, the most important being that "if production of fair copy was a prerequisite for medieval publication (a view invoked in discussing α), Chaucer should have had a second fair copy produced in order to publish the β version."[51] In contrast, Windeatt offers no new and elaborate theory of the text, preferring not to speculate as to just how the traditions of the text arose and how authorial variation came to be preserved in the manuscripts. We should probably prefer Windeatt's agnosticism to Root's certitude, but on one point we can do more than speculate. It is clear from the manuscript evidence that the Troilus must have been copied more than once from some manuscript or manuscripts containing authorial variation. A single publication copy cannot account for the kind of authentic revision detailed above, unless the authorized version of the text was later conflated with the poet's earlier drafts. Whether these copies were made before Chaucer had completed his alterations, perhaps by associates of the poet enlisted "ther nede is, to correcte" what he had written,[52] or

whether his foul papers continued to be available after he had finished with the poem, allowing alternative readings to enter the textual tradition, we have no way of knowing. For this reason theories intended to establish a sequence in the process of revision are speculative at best; priority may be determinable in an individual instance of variation, but the manuscript evidence does not justify the formulation of larger patterns.

More may also be said about the character of alpha. Root used the designation α to refer to two separate entities, the corrected fair copy of Chaucer's first draft and the extant manuscripts supposed to represent it.[53] In his theory of the text it was also a version of the poem, distinct in time, and in the mind of the poet, from the later forms. As established in the foregoing argument, however, Root's theory presupposes a process of composition, publication, and textual transmission difficult to credit. Because no sequence can be established in the process of revision, it is impossible to regard the alpha readings in aggregate as forming a version of the poem, in any ordinary sense; and, in any case, the differences between alpha and beta or gamma are neither as extensive nor as pervasive as those between the versions of Langland's Piers Plowman or, for that matter, between the F and G Prologues to Chaucer's own Legend of Good Women.[54] Alpha is first and foremost a manuscript group or, more accurately, a tradition of the text represented throughout the poem by Ph and locally by a shifting alliance of manuscripts, and as such must be seen as a scribal product, as both Windeatt and Hanna remark. "What the Phetc MSS largely present," says Windeatt, "is the character of a manuscript grouping, not a 'version' of the text. To all intents and purposes the Phetc family has no identity which is anything other than scribal, except in certain parts of Books I, III and IV" (p. 45). Hanna says much the same, although in terms considerably more forceful and more critical of Root's treatment of the evidence: "Thus Root almost forgets to identify α as a genetic group (or scribal version, which it is . . .); only belatedly and offhandedly in The Textual Tradition, pp. 254-255, does he note this fact" (p. 286, n. 11). But to recognize the scribal character of alpha is not to denigrate the authority of many of its readings. Every text that has its origin in a manuscript culture survives as a scribal construct,[55] and textual traditions are defined in large measure by the common errors of copying shared by the individual manuscripts that make up the group; these observations are as true of gamma or beta as of alpha. The distinctive features that identify alpha as a group may not correspond to any authorized "version" of the poem as envisioned by Chaucer, but a substantial number of them, large and small, may be traced to the hand of the poet.

Finally, it may be asked in what precise sense the Troilus may be said to have been revised. Root used the term "revision" in its most formal sense, to refer to the changes made by the author on his completed and scribally published text, and in the portion of his Conclusion on "The Method of Revision" (pp. 256-260) he spells out in detail his theory concerning the fair copy produced by "Adam scriveyn" and the corrections and revisions performed on it by Chaucer. Windeatt justifiably rejects this hypothesis, but despite his efforts to minimize the importance of authorial variation, he finds himself referring

time and again to "a 'rough' early draft" and to "traces of the process of composition." He declines to term these examples of "revision," but what else do we call the writer's process of drafting, with its additions, alterations, and deletions?[56] Windeatt distinguishes authorial variation in the Troilus from what he agrees to call revision in the texts of Piers Plowman and the Confessio Amantis by saying that, by contrast, "most of the distinctions between the TC MSS give the impression of being curiously on top of one another, in both the temporal and spatial sense of the expression" (p. 50), but he has no more proof of the synchronism of variants than Root had of their sequence in an orderly process of revision.

Windeatt summarizes his view of revision in the text of the Troilus with an intriguing analogy: "To say that TC existed for a while without its philosophical passages is comparable to saying that St Paul's Cathedral existed for a while without its dome: that is, until the plan implied by the rest of the structure was completed" (p. 51). Apart from exaggerating the prominence of the Boethian interpolations, the comparison endows them with an inevitability that the contextual evidence will not support. In point of fact, however, the analogy is apt in ways Windeatt does not intend, for the familiar aspect of Wren's masterpiece reflects only the final stage of a plan that underwent at least two major revisions and was altered by the architect even after construction had begun. In addressing the question of revision in the Troilus, Windeatt betrays an almost teleological approach to the text: because the finished poem evinces a certain plan, he seems to say, that plan must always have been the one intended by Chaucer. To be fair, there are certainly features of the text that belong to any state of the poem represented by the extant manuscripts, but the variations under discussion are not of this sort. Unlike Root, we cannot claim to know when the revisions were made or how they came to be preserved in the manuscript tradition. But we must recognize that in certain circumstances the evidence reveals Chaucer's continuing engagement with his text, whether during the initial process of composition or some time after he had completed the main part of the narrative, and that the revisions made by Chaucer altered the structure and meaning of the poem in significant ways and were not simply the inevitable final touches to a preconceived plan, as Windeatt would have it. In the absence of holograph manuscripts and explicit statements of authorial intent, the variations preserved in the manuscript tradition provide invaluable evidence for the method of composition of the greatest English poet of his time.

Notes

Appendix

• I. Alpha variants split, some closer to and some further from source.
• 1.63: H2H4W omit By Paris don; cf. da Paris fatta (1.7,8).
• 1.271: PhH2 read þat for a; cf. la (1.26,4).
• 1.405: H4W and others read me so for to me; cf. sì (Petrarch, Sonnet 132, line 4).
• 1.456: H4 and W read Nor on and Neuer of for Ek of; cf. Né (1.43,1).
• 1.457: H4 reads hundred for thousand; cf. mille (1.43,2), and see 1.531 and 1.546.
• 1.465: PhH2H4 and others read his for him; cf. suo (1.44,3).
• 1.471: PhH2W and others omit other; cf. altri (1.45,2).
• 1.531: H4 reads hundred for thousand; cf. mille (1.54,7), and see 1.457 and 1.546.
• 1.537: Ph reads lyf for self; cf. vita (1.55,4), but note that Chaucer has already used lyf in the previous line.
• 1.546: H4 reads hundred for thousand, this time in agreement with Fil; cf. cento (1.57,8), and see 1.457 and 1.531 above.
• 1.614: H4 and others read fallen for folwen; cf. seguire (2.8,6).
• 1.623: PhH2 and others include than; cf. dunque (2.9,8).
• 2.516: PhH2 and others read yn a fere, a corruption of afer, which reflects Io non gli era vicin (2.57,1) more closely than after.
• 2.1196: PhH2 read he wrote for ʒe woot; cf. lo scriver (2.118,4).
• 3.1348: PhGgH5 and others read we for ʒe; cf. deh, può egli esser ch'io con teco stea (3.35,4).
• 3.1748: GgH5 and others read endytyþ for knetteth; cf. Boece, Book II, metrum 8; both words appear in this context.
• 3.1779: PhH2 read Out of Troy for In tyme of trewe; cf. tempi delle triegue (3.91,1).
• 4.248: GgH5 read þerwith þe sobbis for The heighe sobbes; cf. gli alti singhiozzi (4.28,4).
• 4.493: Gg and H5 read hauede and haue had, respectively, for leuede; cf. vivea (4.58,3).
• 4.519: PhH3 and others read Thus for This; cf. Così (4.63,1).
• 4.799: H3J read al redde for y-red; cf. a pieno (4.95,1).
• 4.1363: H3 includes the grekys; cf. Greci (4.135,1).
• 5.211: Gg and others read walwith, which translates volgendo (5.19,1) more exactly than the other variants.
• 5.541: H3 reads ylyght for i-hight; cf. luminoso (5.53,3), but note that all MSS refer to light in 5.543.
• 5.645: PhJ and most other MSS retain thus, which others omit; cf. così (5.67,1).
• 5.1261: Gg and R read hire for you; cf. ti (7.30,1).
• II. Alpha and others closer to Chaucer's source.
• 1.324: α and others read the for his; cf. al palagio (1.32,2).
• 1.530: α and others read be for by; cf. si scuopre (1.54,6).
• 2.977: α and others read Troyes for Troyens; cf. Troie (2.81,4).
• 2.1091: JR read Iwis for lettre; cf. Lettera mia (2.107,7).
• 3.1415: β reads Whan that for But whan; cf. Ma poi (3.42,1), but note that β's reading is probably a revision occasioned by a shift of stanzas; see the discussion of the apostrophe to lovers above.
• 3.1804: H3 reads This for Thus; cf. Così (3.93,5).
• 4.78: α and others read and for or; cf. e (4.6,1).
• 4.262: α and others include thus; cf. sì (4.30,4).
• 4.295: α and others read What shal I don for What I may don; cf. che farò io (4.33,7).
• 4.594: α and most other MSS read a lite i-founde for a litell stounde; cf. alquanto (4.72,7).
• 5.84: H2H4 read lete for do; cf. far (5.12,8).

---

184




• 5.412: αβ read seyn for wene; cf. diria (5.35,6).
• 5.632: αγ read Thenchesoun for The entencioun; cf. cagione (5.61,2).
• 5.924: αβ read kyng for lord; cf. re (6.22,8).
• III. Alpha and others further from Chaucer's source.
• 2.761: H4JR read Vnwist to him for Upon this knight; cf. non sarà saputo (2.69,7).
• 3.1346: H3 reads euer his eye for neuere his look; cf. gli occhi (3.35,2).
• 3.1477: H5R read Ioye for lif; cf. ben (3.44,8).
• 3.1802: R reads all for ech; cf. tutti (3.93,3).
• 4.51-55: H3 omits Maugre in 51, reads and for or in 53, and mentions Hector in 55; cf. Fil 4.3, especially e in line 2 and d' Ettore in line 5.
• 4.57: H3 makes Priam the originator of the truce; cf. Chiese Priamo triegua e fugli data (4.4,1).
• 4.87: α and others corrupt lefte into le(e)ste or loste; cf. lasciai (4.7,8).
• 4.94: Some MSS read and for the second O; cf. e (4.8,4).
• 4.127: HR reads this for his; cf. Questo (4.12,1).
• 4.138: H3H5 omit reference to Toas, who is not mentioned in Fil 4.13, but Chaucer is following Guido and Benoit here.
• 4.261: ADS1S2 read the for thus; cf. t' ho (4.30,3).
• 4.511: α and most other MSS read herte for hete; cf. foco (4.61,5), but also note cor (4.61,8).
• 4.733: H2H4 read wo for ded; cf. la dolente (4.87,1).
• 5.64: H2Th read guise for wise; cf. guisa (5.10,1).
• 5.67: R reads wallys for valeye; cf. vallo (5.10,4), although the majority reading is probably Chaucer's own mistranslation.
• 5.1674: αβ read bright for myn; cf. mia (8.12,2).
• 5.1809: Only JR read eighthe for seuenthe et al.; cf. ottava (Teseida 11.1,4).
• IV. Significance of variation with respect to source not clear.
• 1.158: McCormick and Root claim smellyng is closer to the Italian than smelle(n); cf. riveste i prati d' erbette e di fiori (1.18,2).
• 1.373: Windeatt says "Ph corrected," referring presumably to the reading lorne; cf. perduto (1.35,2), although it is not clear why he cites this reading as an example of α's being further from the Italian. Only H2 actually reads borne.
• 1.442: αβ read day fro day, which McCormick and Root take to be closer to di giorno in giorno (1.41.1) than day by day, but the resemblance is not clear.
• 2.968: H4 reads ben stoupyng for stoupen, and H4 and others read stalk(e) for stalkes; cf. chinati and stelo (2.80,2 and 3), though the variation is trifling.
• 2.1005: PhH2 and others omit second right, which is not in Fil 2.91.
• 4.89: JH3Cx read this for that; cf. Di ciò (4.8,1).
• 4.495: H2H4 read it for first that; cf. come 'l potrò da me così cacciare, / come ragioni (4.58, 4-5), though proximity to the Italian is difficult to see.