University of Virginia Library

Modern scholarship has perhaps been less conscious than it should of the fragile bases for some of our assumptions about the Chaucer canon. Were it not for Chaucer's own testimony in the Prologue to the Legend of Good Women and the Retractions to the Canterbury Tales [1] Chaucer's oeuvre would appear rather slimmer than it does. For example, neither the Book of the Duchess nor the House of Fame is attributed to Chaucer in any of the surviving manuscripts. And of the twelve manuscripts of the Legend of Good Women, only one ascribes it to Chaucer, Bodleian Arch Selden B. 24, a manuscript so unreliable textually, that its claims might, in isolation, make one ponder the claims of another witness, Bodleian Rawlinson C. 86, which credits one of the Legends to Lydgate.[2]

Even Chaucer's own accounts pose problems which cannot be resolved. We cannot now identify a number of works he claims.[3] And we continue to wrestle with problems of canonicity unresolved by Chaucer's testimony, in shorter passages like lines 31-96 of the Book of the Duchess,[4] or the concluding roundel of the Parliament of Fowls,[5] or of various links in the Canterbury Tales.[6] And on a larger scale, recent scholarship has raised questions about the authenticity of the Canon's Yeoman's Tale [7] and the Tale of Gamelyn.[8]

Then there are works not specifically mentioned by Chaucer at all but which have been accepted into the canon on various grounds. Into this category fall many of Chaucer's lyrics, and at least one more substantial work, Anelida and Arcite. I have recently re-examined the manuscripts of this poem and come to the conclusion that the received views of its authorship and unity merit some reconsideration.

Anelida and Arcite is invariably printed by modern editors as a single, incomplete, poem of 357 lines, by Chaucer, and divided into two related parts, the first, The Invocation and Story (lines 1-210) and the second, the Complaint of Anelida (211-350) followed by a brief unfinished continuation (351-357). An examination of the manuscripts casts doubt on all these editorial assumptions.

The authorship of Anelida and Arcite has never been seriously questioned.[9] Skeat speaks of the "internal evidence" as "completely satisfactory" linking it with other poems, like the Book of the Duchess and the House of Fame about which "we need say no more."[10] And he does not.

Robinson simply announces categorically that the poem "is of undoubted authenticity."[11]

It will, I trust, become clear why it is simpler to respond to the evidence by categorical assertion than reasoned analysis. But it is worth considering the factors that have led to its being so unhesitatingly accepted into the canon. Some of these are clear. It is on a subject that links it to a known work of Chaucer's, his Knight's Tale. It has the incremental weight of every printed edition from Caxton onwards behind it. And the metrical virtuosity of the Complaint itself has consistently excited admiration.[12]

In addition, the work was also a relatively popular one during the fifteenth century. It survives in twelve manuscripts.[13] Of Chaucer's shorter poems only the Balade de Bon Conseil (twenty-two manuscripts), Lak of Stedfastnesse (fourteen manuscripts) and the ABC (thirteen manuscripts) occur in more. There may be adventitious reasons for such survival in these other cases. The brevity of the first two makes them good "fillers" in fascicular manuscripts, which assemble collections of Chauceriana. And the appeal of the ABC seems somewhat inflated by its appearance in five manuscripts of the Middle English prose translation of Deguileville's Pelerinage de la Vie Humaine.[14] Anelida and Arcite is thus fairly distinctive among the shorter works ascribed to Chaucer in its appeal to fifteenth-century readers, a circumstance which, together with its intrinsic merits and its general collocation with genuine works of Chaucer, has perhaps created a predisposition not to enquire too searchingly into its antecedents.

Of the twelve manuscripts of Anelida and Arcite, only three include any ascription to Chaucer:[15] British Library Additional 16165, Trinity College Cambridge R.3.20 and British Library Harley 372. At first glance, such testimony might seem impressive compared to the contemporary attributions for other works included in his canon. And the fact some important manuscripts, such as Bodleian Library Fairfax 16, do not attribute it to Chaucer, is not of overmuch weight, since Fairfax, although of considerable textual importance, only ascribes a few lyric texts specifically to him.

It must be stressed, however, that none of the manuscripts is very early, and most are quite late. The earliest one is possibly British Library Additional 16165, which cannot, however, be before 1422-23.[16] Trinity R.3.20 cannot, on internal evidence, be much before 1430 but was probably compiled before 1437.[17] Fairfax 16 and Tanner 346 have been dated from the second quarter of the fifteenth century.[18] Harley 372 has been dated at about 1450 (Brusendorff, p. 190n). Harley 7333 was probably begun c. 1450-60 but was "continued for many years."[19] (Anelida and Arcite comes about half way through.) Longleat 258 was probably done about 1460-70 (Seeton, p. 92). Huntington HM 140 is somewhere between 1450-80 (Manly/Rickert, I, 434). Most of the others are very late in the fifteenth century: Digby 181 has been dated c. 1483-85;[20] and Pepys 2006, Bodley 638 and Cambridge University Ff.i.6 are probably later than that.[21] It may be worth pointing out that Anelida and Arcite does not appear in the first early Chaucerian anthology,

Cambridge University Gg.4.27, made in the first quarter of the fifteenth century.

Possibly the two earliest manuscripts are those written by the mid-fifteenth century scribe, John Shirley: British Library Add. 16165 and Trinity, Cambridge R.3.20. In addition, a further text, Harley 7333, seems to derive from a lost Shirley original; this manuscript was probably written in a house of Austin canons in Leicester (Manly/Rickert, I, 214-218).

Shirley himself remains one of the more enigmatic figures in fifteenth-century publishing history. We possess a relatively large number of manuscripts which he copied as well as several that clearly derive from originals which he wrote. But the purposes of his manuscript production have resisted confident analysis. He has been termed "one of England's earliest publishers,"[22] and conversely, an earnest antiquarian, whose transcriptions were made for private pleasure.[23]

But whatever his motives, his Chaucer attributions have proved of considerable significance in establishing the canon. For example, he provides either unique or corroborative authority for attributing Chaucer's Complaint Unto Pity, Complaint to his Lady, The Complaint of Mars, The Complaint of Venus, Womanly Noblesse Truth and others.[24]

The Shirley connection is of particular importance because it gives the greatest weight in support of the attribution of Anelida and Arcite to Chaucer: Shirley attributes it to him in both the Additional and Trinity manuscripts. In conjunction with the attribution in Harley 372 this might seem more than sufficient evidence to leave the poem secure within the canon. But, in fact, the implications of all these attributions are less clear than might appear initially. To clarify why this is so, it is necessary to briefly consider how the poem appears in the manuscripts, and, in particular, what light is shed by a consideration of the content, order and title of the poem as evidenced in them.

It should perhaps be stressed that neither the content nor the order of Anelida and Arcite in modern editions is that of the majority of the manuscripts, particularly the early ones. This is strikingly demonstrated by the manuscripts Fairfax 16 and Bodley 638. These manuscripts are closely related textually. They both present the main divisions of the text in reverse order. That is, the Complaint (lines 211-350) appears before lines 1-210. Both these manuscripts agree in describing lines 211-350 as "The complaynt of feyre Anelida on fals Arcyte." Fairfax simply repeats this title for lines 1-210. But Bodley characterizes these lines as "The boke of feyre Anelida and fals Arcite."

Such an anomalous ordering is not easy to account for, except by an assumption that the poem as we now conceive it was only available to these copyists as two separate parts, not necessarily related or sequential. There is in both these manuscripts a concluding rubric that suggests some link between these two parts: after line 210 (here the actual last line of the poem), line 211 is used as an explicit referring back to the antecedent Complaint (see

Appendix, nos. 3 and 7). But what this explicit signifies is unclear. If it were some instruction about the sequentiality of the portions of the text, it is puzzling that the scribes did not implement it in their copying. It seems that their exemplar consisted of two distinct components that they were disposed to believe were non-sequential: Complaint and the narrative portion of the text.

The validity of this unambitious hypothesis can be demonstrated in other manuscript contexts. Thus in what is likely the earliest surviving version of Anelida and Arcite, Additional 16165, Complaint and narrative portions once again appear in reverse order, but this time completely separated. On folios 241v-243v occurs "The compleynt of Anelyda," more or less lines 211-350. Later, on folios 256v-258v occurs a defective version of lines 1-210. These lines are titled "Balade of Anelyda Quene of Cartage made by Geffrey Chaucer."

It is, moreover, clear that the Complaint portion came to enjoy a distinct existence separate from the rest of the poem. For example, in four of the twelve manuscripts lines 211-350 (or 357) appear as a separate poem: Trinity, Cambridge, R.3.20, Pepys 2006, Cambridge University Ff.i.6 and Huntington Hm 140. In all these manuscripts these lines are designated as "The compleynt of anelyda," and this designation is also given to these lines in manuscripts that contain a consecutive text of lines 1-350, British Library Harley 372 and 7333 as does Longleat 258 (lines 1-357). Bodleian Digby 181, which also has a consecutive text, describes lines 211-357 as "litera Anelida regine." And other manuscripts, while not differentiating the second half of the text by heading, do so in other ways. Thus Tanner 346 leaves a substantial space between lines 210 and 211.

Several points emerge. In the first place, the evidence of rubrics emphasizes the scribes' apparent sense of the two parts of the poem as distinct entities. And the evidence of the manuscripts themselves suggests the initial separateness of these entities in the forms in which they were available to scribes. This is clearly the evidence of the earliest surviving copies, Shirley's Additional 16165 and his Trinity R.3.20 (which contains only the Complaint), and the Fairfax/Bodley ones, as well as of the other copies which only include the Complaint. In fact, only five of the twelve manuscripts transcribe the complete poem consecutively, more or less in the order we now read it in modern editions: Harley 372, Harley 7333, Digby 181, Tanner 346 and Longleat 258. And, as I have noted, all these differentiate between the two parts either by rubrics or in other ways.

These points have an obvious bearing on the matter of attribution. For there is no manuscript that unambiguously ascribes the entire poem to Chaucer. Of the two Shirley attributions, one, in Additional 16165, credits Chaucer with the first part of it; the other, in Trinity R.3.20 gives him the second half (the only part that occurs there). And Harley 372 has opposite the last line of the poem (line 350), the name "Chaucer" in the hand of the scribe. The evidence I have presented suggests that this last attribution cannot be

confidently extended to encompass the entire poem. (There are other reasons for some scepticism about this Harley attribution to which I will return.)

The extent to which any attribution can be reasonably held to cover the entire poem also bears on the last stanza (lines 351-357), and, in particular, the question of the poem's supposedly unfinished state. Doubts have been expressed from time to time about the likely authenticity of these lines,[25] which are, of course, the basis for any belief that the poem is unfinished. There are obvious grounds for suspicion. This stanza obscures the parallelism of lines 211 and 350 ("So thirleth with the poynt of remembraunce . . . Hath thirled with the poynt of remembraunce"), the symmetry of which might seem to mark a formal ending. More compellingly, lines 351-357 appear in only four manuscripts: Longleat 258, Digby 181, Cambridge Ff.i.6 and Tanner 346. None of these attributes the poem to Chaucer. Only Tanner 346 is relatively early. And Tanner has a cryptic note opposite the beginning of this stanza—"Belliger." What this note means is unclear. But it does seem to represent an attempt to make a discrimination of some kind between this stanza and what precedes it. In sum, the evidence for the authenticity of these lines is very insubstantial.

But perhaps the most puzzling aspect of lines 351-357 appears to have escaped consideration. That is, not the question of their authenticity, but the reason why they are there at all. If any effort was made by someone other than Chaucer to continue the poem beyond line 350, why was it done in such a desultory, undeveloped way? Fifteenth-century copyists were evidently capable of writing links or conclusions to Chaucer's poem when they felt it necessary. How, then, is such a clumsy and aborted effort to be explained?

A possible explanation may lie in the ordering of the earliest forms of Anelida and Arcite. As I have noted, both Fairfax and Bodley transpose the main portions of the poem, and Additional 16165 also separates them. It may be possible that lines 351-357 were originally a clumsy attempt to link the second part of the poem (211-350) to the following first part (1-210). There are grounds that could lead a copyist to believe this was a reasonable thing to do, if confronted with two apparently separate but topically related poems. The reference in line 355 to Anelida's sacrifice to Mars provides a link to the opening line of the poem ("Thou ferse god of armes, Mars the rede"). And this opening stanza, with its prayer that Mars "my song contynue and gye" (6), could be read as implying the precedent nature of the Complaint (lines 211-350). A copyist might have felt it appropriate to insert a transitional passage to make clearer this relationship. Such a passage would presumably have been added to an exemplar on a separate sheet, and could, in the course of transmission, easily become misplaced or overlooked. Hence its survival at the end of Anelida and Arcite in some manuscripts and its omission in others.

Such an argument rests on a sense of the original fragmented state of the transmission of the early exemplars of Anelida and Arcite. The strange ordering of the text of the Complaint in Additional 16165 appears to support such a view.[26] It seems that it took some time for it to achieve a stable transmitted

text, reflecting the order and content as these appear in modern editions. The uncertainty of the earliest surviving manuscripts with respect particularly to order leaves few grounds for believing that the modern ordering of the text corresponds to Chaucer's intentions. What we appear to have are two originally separate texts which fifteenth century scribes came gradually to amalgamate to form the poem we now call Anelida and Arcite.[27]

There remains the question of authorship itself. I have already pointed to the ambiguity of the attribution in Harley 372, which cannot be confidently held to extend to the whole poem but which may only be intended to cover the Complaint. The other two attributions to Chaucer appear in the Shirley manuscripts, Additional 16165 and Trinity R.3.20. Looking at these attributions together, certain things are immediately puzzling.

In Additional, where the two parts appear separately and reversed, only the first (lines 1-210 in fragmentary form) is ascribed to Chaucer: "Balade of Anelyda Qwene of Cartage made by Geffrey Chaucyer" (f. 256v). This is the only clear attribution of these lines to Chaucer in any of the manuscripts. The second part (211-350) appears alone in Trinity where it is described as "þis compleynt of Anelyda Qweene of Cartage . . . englysshed by Geffrey Chaucier." But in the Shirley-derived Harley 7333, both parts appear consecutively; neither is ascribed to Chaucer. Thus, while at different times he credits each part of the poem to Chaucer, at no point does Shirley ascribe all of it to him. And, in what was probably his last transcriptional effort involving it, the preparation of the exemplar from which Harley 7333 derives, he ascribes none of it to Chaucer.

It is clear, as I have shown, that in the two manuscripts Shirley actually copied, Additional 16165 and Trinity R.3.20, Shirley believed that the separate parts of Anelida and Arcite were wholly separate poems. It seems that his information about both of them was less than reliable. Thus his rubric for the narrative portion of the poem (1-210) in Additional, "Balade of Anelyda Qwene of Cartage made by Geffrey Chaucyer" (f. 256v), contains two inaccuracies (the poem is not a balade and Anelida was not queen of Carthage). It also, as a statement, stands in marked contrast to the prolixity of Shirley's rubrics in the Trinity and Harley manuscripts and elsewhere. The rubric therefore can be placed in contexts of inaccuracy and possible uncertainty which do not serve to strengthen the credibility of Shirley's unique attribution.

The rubrics for the Complaint itself in Additional are also curious. Here Arcite is characterized (uniquely) as "chiualrous."[28] And it again (together with Trinity) agrees against all other witnesses (including the Shirley-derived Harley 7333) in describing Anelida as "Qweene of Cartage" (rather than "Hermonye" as she is represented in four other manuscripts—including Harley 7333).

Such contradictions between Shirley's own manuscripts and the one derived from his activities, Harley 7333, indicate that he had access to conflicting sources of information for his various copies, and that by the time

he was preparing the exemplar for Harley 7333 he seems to have rejected information he had utilized when he was copying Additional.

It is certainly the case that Shirley appears to have rejected any textual authority he may have once felt to inhere in Additional 16165. In the case of lines 1-210 he appears simply to have rejected Additional completely by the time he came to prepare Harley 7333's source text. There are (very roughly) some eighty substantive variants between the two for the 131 lines they have in common.[29] Harley includes one stanza omitted in Additional (141-147). And the degree of variation often amounts to virtual rewriting; for example, in Additional lines 188-9 read:

In Harley 7333 these lines become The readings of Harley 7333 seem generally superior to those of Additional for this part of the text.

In the case of the Complaint (211-350), matters are somewhat more complex. Three Shirley-related manuscripts survive. In the probable earliest, Additional, the text appears in a unique, wholly anomalous order. Apart from the question of order, there are about thirty-five substantive variants between this text and the one in Trinity. Again, these are often striking in their extent. Thus line 318 in Additional reads

which becomes in Trinity And line 332 as it appears in Additional is while in Trinity it is

Harley 7333 has a high degree of agreement with Trinity against Additional (it agrees with Trinity, for example, in lines 318 and 332 above). It does, however, include lines 265-268, which do not appear in either Trinity or Additional. And it does have about twenty-five substantive variants of its own from either of these manuscripts. It seems probable that the Harley text is a conflation of Trinity with some other source that also, of course, included lines 1-210.

The range of witnesses Shirley seems to have employed suggests a dissatisfaction with the forms of the text to which he had access at particular

times. On the evidence of its use in Harley 7333, he appears to have felt the later Trinity text preferable to the Additional one; and in the exemplar for Harley 7333 he rejected Additional for the text of lines 1-210 in preference to the unidentified further source.

This repudiation of Additional in Shirley's subsequent scribal activities with Anelida and Arcite does raise substantial questions about the canonicity of the poem in general, and lines 1-210 in particular. As I have noted, Shirley provides the sole unambiguous authority for these lines in Additional. The curious form of this attribution, its evident incorrectness in some particulars, and its subsequent omission in Harley must leave the authenticity of these lines open to substantial doubt.

It is also noteworthy Shirley did not credit Chaucer with the authorship of the Complaint in Harley 7333 after he had done so in the Trinity manuscript. If this indicates some subsequent denial of a belief in Chaucer's authorship it is worth recalling for a moment the authority to be attached to the only other manuscript ascribing any part of the poem to Chaucer, Harley 372. As I have noted earlier, the name "Chaucer" appears opposite line 350 in this copy. If it is uncertain how much of the poem this ascription was intended to cover, it is also uncertain whether the ascription of any of it can be held to be inherently reliable. Harley 372 is a fascicular or "booklet" manuscript.[30] The fascicle that contains Anelida and Arcite also contains a copy of the Middle English version of La Belle Dame Sans Merci that is unique among the extant witnesses in ascribing this poem to Sir Richard Roos.[31] The conjunction of this unique ascription with the ascription of some (unclear) portion of Anelida to Chaucer suggests that either the copyist has access to some unusually authoritative source or to one that was very unreliable. It is unclear which is the case.

In Harley 7333 Shirley also established for the first time in his dealings with the poem a consecutive text of lines 1-350. In the light of Shirley's other efforts with the text, such a decision must be seen once again as part of his continuing effort to make sense of the manuscript evidence as it became available to him. It seems probable that by the time he came to prepare the exemplar for Harley 7333 he had become aware of other, conflated orders linking the distinct poems that now form Anelida and Arcite, such as those represented in Tanner 346 and Harley 372. But it is clear that he was not in a final way convinced that the whole work was by Chaucer, and he may possibly have come to believe that none of it was by him.

External testimony serves to confirm at least his suspicions about lines 1-210. The earliest attribution of Anelida and Arcite to Chaucer apart from Shirley's is that in the Prologue to Lydgate's Fall of Princes, probably begun c. 1432; here Lydgate offers a lengthy discussion of the Chaucer canon, which adds several works to Chaucer's own accounts. Among these is

The characterization of the poem as a "compleynt" is suggestive since this characterization is reserved almost exclusively for the second part of the poem.[33]

To sum up: an examination of the manuscript evidence suggests that the incomplete work we call Anelida and Arcite and credit to Chaucer is, in fact, not incomplete, and is not one poem but two, only one of which was possibly written by Chaucer. It seems likely that the part of the poem which is now lines 1-210 was grafted on to a different poem which may have been, but is not certainly Chaucerian.[34] The evolution of the organization of these texts and the evident uncertainty about both organization and attribution can be traced through the sequence of surviving witnesses, particularly those copies connected with Shirley and the Fairfax/Bodley ones.

More generally, what these arguments suggest is some of the ways in which manuscript evidence, evidence of compilation, textual transmission and presentation (especially, in this instance, the rubrics and divisions of a text) can be employed to call into question editorial and critical assumptions about the nature of the received text. Anelida and Arcite provides a clear demonstration that the text itself often cannot be considered profitably in isolation from the manuscript contexts in which it appears. And a consideration of such manuscript contexts may help, in the present instance, to achieve a clearer understanding of how this poem came to achieve a status during the fifteenth century that can be shown to be non-authoritative and may even be non-canonical.