A modern edition of Troilus and Criseyde belongs to an editorial tradition that dates back to 1933 (Robinson) or 1894 (Skeat) or 1881 (Furnivall) or 1866 (Morris) or 1854 (Bell) or 1775 (Tyrwhitt--whose edition didn't include this poem but who was the first modern editor to understand Chaucer's language and versification) or 1721 (Urry) or 1561 (Stowe) or 1532 (Thynne) or 1483 (Caxton), or even to the early fifteenth century, when scribes often juggled manuscripts to patch gaps in the text or (intentionally or not) substituted familiar words or spellings for obscurities in the text they were copying from. A new edition is a part of history and owes much to the past.
At the same time, any edition is an interpretation of the poem for the present, focused through the lenses of the present. It will, and should, reflect its times. Ninety years ago, this statement would have seemed misguided: the object of editing was to reach through the obvious gaps and biases of history toward the original poem as the poet wrote it. In the early twenty-first century, one objection to this goal is that the author's text is unrecoverable. Chaucer's older French contemporary Machaut left autograph manuscripts, but the earliest extant manuscripts of Troilus and Criseyde are from about fifteen years after Chaucer's death, and it has been impossible to establish a manuscript genealogy or a reliable record of the poet's process of composition.
It is still possible to do what we can with what we have--the manuscript and early print record of the poem, and our knowledge of Chaucer and of late medieval manuscript transmission--and to arrive at a readable text that is likely to approach the poem as it existed in the 1380s. In this way, an editor's job is the same now as in the early twentieth century. In another way, the job has changed because the audience and the medium have changed. Online texts are replacing print texts. Current readers, especially but not only beginning student readers, are less likely to have experience of earlier periods of the language; the text presents new problems.
The present edition responds to the need for a responsible electronic text of the poem which is accessible and useful to student readers as well as scholars, and which makes use of some of the resources available to online publication: free and open access, mouseover notes to offer instant navigation within the site, and easy revision as new sources and issues arise. In the explanatory and textual notes, I have tried to make the discussion both intelligible to student readers and useful to Chaucerians. I have avoided technical terms when possible, and when that is not possible, I have tried at least to define the terms in place.
There are two very different ways to look at electronic publication. For several years now, we have had high-tech digital editions that can see more in a medieval manuscript than is accessible to the naked eye. These, for the modern editor, are a shining city on the hill. The present edition is more like the mimeograph machine in the church basement, but it is functional.
This introduction to the text of the edition will focus on the editing history of Troilus and Criseyde and on my editorial method, but there is a wider point to make as well: for this poem at least, and perhaps to some degree for other Middle English texts, some of the key theoretical and methodological questions have been not so much answered as exhausted. The real issue for the coming decades is to provide responsible and useful editions--for both students and scholars--that make the transition from the print world to the online one, and from current price tags to free and open access.
Troilus and Criseyde survives in sixteen manuscripts, none earlier than about 1415, another handful of manuscript fragments, and three early print editions (Caxton, Thynne, and de Worde) often given manuscript standing by editors. Four manuscripts are superior, in terms of completeness, coherence, and plausibility. Three of these are from the early fifteenth century: Corpus Christi College, Cambridge number 61 (the one with the frontispiece); Campsall (now Pierpont Morgan Library 817); and St. Johns College, Cambridge ms. L.1. The fourth, Harleian 2280, from the mid-fifteenth century, closely resembles Corpus.
William McCormick (the Globe editor) argued for three stages of composition and three manuscript types, α, β, and Γ, with Γ the finished product. (Corpus, Campsall, and Harleian 2280 are type Γ; St. John's is β to line 4.430 and α from there to the end.) R. K. Root, who was McCormick's student, completed McCormick's manuscript study, and his own edition made β the finished product, because β manuscripts looked to him like Γ manuscripts with another layer of revision, consisting mainly of several transposed passages. For both, the stages-of-composition hypothesis satisfied the need for genealogy and made it possible to edit by recension. Root used twin base texts, St. John's and Corpus, correcting the α section of St. John's by Corpus and making Corpus his first resort for problems in St. John's in books one through three.
Robinson (1933, 1957) worked from Corpus, rejecting Root's β-as-end-product theory but without a principled explanation. Root's influence is clear in Robinson's text, which frequently emends Corpus by St. John's, mainly on metrical grounds. Baugh is similar to Robinson. Donaldson and Howard worked from Corpus. Fisher used Campsall, which by then was in the U.S. To that point, no one had refuted Root's genealogy, but no one appeared to like its practical consequences for the text of the poem. In 1982, Windeatt offered a principles objection to the β theory, pointing out that even a clearly demonstrable manuscript tradition is not necessarily a stage of authorial composition. Barney (the Riverside editor) and Windeatt followed Corpus. (Barney's readings do sometimes disagree with Robinson's, and when they do, they are usually closer to the manuscript.)
Then in 1992, Ralph Hanna carried Windeatt's work farther, offering a plausible scenario for the β tradition. On circumstantial but compelling evidence, Hanna argues that it is misleading to speak of three manuscript types and impossible to draw a stemma: only Γ is a real tradition of the poem. Type α consists of two extant manuscripts and parts of others, pointing to a hypothetical early archetype probably two removes from Chaucer; and type β is historically real but poetically a mirage, an accident of late medieval commercial book production. β scribes received the text piecemeal from a clearing house, in bound or loose quires, and the gaps, transpositions, and shifts of allegiance in the manuscripts occur in units of four, five, or eight stanzas or their multiples: quarto or folio page or quire. A leaf is missing or reversed and there is a gap or transposition in the copy. The wrong quire arrives and the copy changes allegiance. A delivery is late and the scribe resorts to an inferior backup. By the late fifteenth century, a series of such events became a hodgepodge exemplar, and Caxton's edition set it in type.
Hanna's work ought to have been a methodological breakthrough, but it mainly justified the instincts of the editors, who, again, had simply ignored Root's theory rather than argue it down. And his work ought to have had wider implications for editing theory, but it was published while medieval studies was busy absorbing poststructuralism. So his article neither inspired nor inconvenienced most people, because by then not many people were particularly moved to recover Chaucer's intention or particularly troubled by the loss.
What remains for the present-day editor is either George Kane's monster eclecticism or best-text editing. Troilus and Criseyde is not a good candidate for Kane's method, for one important reason. Heavily eclectic editing is risky, but it is easy to understand why it would appeal to an editor of Piers Plowman. But with Troilus and Criseyde, the aggregate text is a much different thing. The manuscript tradition sounds like a thicket of brambles if we listen to the editors, but the manuscripts are really remarkably consistent, considering the means of transmission and the scale of language change in the fourteenth and fifteenth centuries. So the issue is less the manuscript record itself than what the editor does with and to the manuscript record. There are hundreds of decisions to make about individual lines, and these are all consequential, in other words worth the effort of constructing an edition. But most of them are not essential to most readers of the poem. In the whole poem--if we set aside omissions, transpositions, dialect spellings, metrical hiccups, plausible synonyms, and other well-attested scribal tics--there are sixteen consequential variants: sixteen places where the choice of manuscript can radically affect the reading of a line. But any critical interpretation resting on the reading of one of these lines would be unbelievable, and any interpretation resting on any combination of them would be nonsensical; there is just no pattern there.
The notes linked to the text of the poem will comment on significant variants, but a few illustrations here will be useful. For line 3.391, Campsall and three other manuscripts, including the β type Rawlinson, have "Right as thy knave, whider so thou wende." In Campsall, knave is corrected from sclave. The other manuscripts and all other modern editions--including Root--have "Right as thy sclave." Chaucer's poem has departed from Boccaccio's at this point, so the Filostrato is no help. Lyric tradition in the 1580s would prefer sclave, but not in the 1380s. It is hard to say, as it often is, which word the principle of the durior lectio would prefer. The MED has only one citation for slave earlier than Chaucer's line (Southern Legendary, Beckett, c. 1300); and the next citation after Chaucer's line is from 1440. The Concordance has one entry for slave, this line, so if the word is Chaucer's, it is very much the harder reading. I have adopted knave, with a note calling attention to the issue. But it is difficult to see the line as a crux of the poem.
In three places, manuscripts diverge over pite and piete(e). Corpus has the following:
- His eyen two for pite of herte (4.246)
- For evere in on his herte pietous (5.451)
- Conceyved hath myn hertes pietee (5.1598)
For 4.246, Harleian 2280 and Cambridge G.g.4.27 share the Corpus reading; St. John's has "piete"; Campsall has "pite of his herte." For 5.451, St. John's has "pietus," essentially the same as the Corpus reading; Campsall and Cambridge G.g.4.27 have "pitous." For 5.1589, Corpus and St. John's agree; Campsall and Harleian 2280 have "pite(e)." In all three lines, the issue is syllable count, and piete / pietous seems to make sense. But MED senses for the two words current in the fourteenth and fifteenth centuries overlap considerably; each is defined by the other. Even if modern editing can unravel the textual problem, it seems unlikely that it can clarify the semantic situation.
Most of the sixteen variants are similarly benign, but three are more consequential or just more interesting.
The first of these is 1. 234-35, for which Corpus and three minor manuscripts have "To serven love which that so soone kanne / The fredom of youre hertes to him thralle." Other principal manuscripts, and all other modern editors, have "To skornen love . . ." But the couplet at the end of the stanza--"For love is he that alle thing may bynde / For may no man fordo the lawe of kynde"--clearly justifies the Corpus reading. Fisher, working from Campsall, has a right to skornen, but the Corpus editors who run to Campsall here are taking a premature peek at the answer book.
The second, line 5.382 ("As ravenes qualyn or schrichyng of thise owles"), sheds light on earlier editorial practice. This is the reading of Corpus Christi, a Γ manuscript. St. John's, a β manuscript, similarly has qualin. Rawlinson, which Root identifies as the only manuscript to belong to group B throughout, has qualine. Other principal manuscripts have qualm, and all modern editions except the present one adopt that reading. Robinson cites the MED, which defines qualm as the cry of a raven but cites only this line, a circular argument. Editors have preferred the monosyllable on metrical grounds, and the cadence (an extrametrical syllable at the caesura) is rare in Chaucer but not inconsistent with his metrical system. And qualyn is both grammatically appropriate and onomatopoetic.
The third is line 5.1809. Corpus, Campsall, and all but two other manuscripts have Troilus rising to the seventhe spere after death. St. John's and Rawlinson (both β) have viij spere and viijthe spere in Roman numerals. All modern editors emend to eighthe. The source here is Boccaccio's Teseida (Book 11, stanza 1), which has urtava (ottava). A great deal has been written on the line, much of it unnecessary. Editors justify eighthe on several grounds: the stanza closely resembles the parallel passage in Boccaccio; the scribal process viijthe > vijthe > seventhe is understandable; and eighthe puts Troilus in the sphere of the fixed stars, which makes more sense than the sphere of Saturn. And counting outward from earth in this way is consistent with Book 3, line 2, which places Venus at the third sphere.
But the editors also have reservations. There is the unanimous testimony of Γ manuscripts for a start. And medieval reckoning is inconsistent: if we count inward from the primum mobile, depending on where we start, Troilus could pause at the sphere of Venus or the moon, either of which could make sense--Venus given the plot of the poem, and the moon as the first vantage point above the sublunary world. There is, however, a strangely ignored piece of evidence within this very passage. From his temporary vantage, Troilus sees "with ful avysëment / The erratik sterrës herkenyng armonye" (5.1811-12). In other words, he looks down and sees the planets beneath him--all of them, apparently--and then he looks down farther to "This litel spot of erthe." And the only place from which he could do all this is the sphere of the fixed stars.
Given the manuscript record, the logical course for an editor is to choose a Γ manuscript and stick with it as far as possible. (The other modern editors have done the first but not always the second.) I chose Corpus because the Parkes-Salter photographic facsimile was available when I started to work with the poem. (A facsimile of Campsall is now available as well.) I used the facsimile to correct the Chaucer society transcription of Corpus (Root found 20 errors in it; I found 114) and worked from that, consulting the other six-text manuscripts, a facsimile of Thynne's edition, Root's collations, and the other modern editions, starting with Skeat.
The first question is when to desert the base text. There are 128.5 lines missing from Corpus; I supply these mainly from Campsall, with St. John's as next resort. Apart from these missing lines, most departures from the base manuscript in previous editions of the poem are of two kinds: the correction of apparent scribal error or eccentricity, and the correction of lines that are metrically defective or anomalous or just annoying to the editor. Robinson in particular has a Victorian ear for metrical regularity and little tolerance for metrical resolution or headless lines. But every modern editor sometimes deserts his base text on grammatical and metrical grounds.
In a very few places, I do likewise, where Corpus is clearly outside Chaucer's metrical system because of an obvious omission or interpolation, but in most places I leave the line as in Corpus, even where the anomaly is surely scribal, in order to put the metrical system in perspective by letting the reader hear the consequence of its occasional breakdown. Overall, my edition is more conservative--closer to the base manuscript--than other modern editions.
The second question is how and when to modernize the text. Modernize is still a fightin' word among Chaucerians, but every editor does it. If you want an unmodernized text, read the Parkes-Salter facsimile. If you want a relatively unmodernized print text, read the Chaucer Society transcriptions. Modern editors change u to v and vice versa, add phrase and sentence punctuation, capitalization, and quotation marks, and selectively normalize or modernize spelling, all of which alter the text. Editors sometimes disagree on where a sentence ends (there is no pointing in Corpus), and this too can affect the meaning of a passage. Donaldson mentions Chaucer's tendency to leave major phrases suspended ambiguously between sentences (a technique that postmodern poets tend to think they invented). When this happens, editors punctuate to disambiguate. So, again, the question is not whether or not to modernize but where to stop. Where is the line between best-text editing and eclectic editing? (It has been suggested to me that any semantically based intrusion on a base manuscript exposed the soul of an eclectic. I see the point, but if we grant it, the label is redundant; any edition, including the Chaucer Society transcriptions, is eclectic.)
My object has been to produce an edition that will be useful to scholars but also accessible to twenty-first century undergraduates, including those with no prior experience of Middle English, which, these days, means almost everyone. Unfamiliar nouns and verbs are a problem, and even the most familiar function words may be unrecognizable. Michael Murphy's online text, which is unattributed and full of puzzling constructions, thoroughly modernizes the spelling. Donaldson's says he tries to avoid spellings that would be unrecognizable to Chaucer. Donald Howard normalizes, apparently to the values of Middle English generally. Even Fisher, whose edition nicely catches the flavor of Campsall, now and then corrects "ungrammatical" spellings. A parenthetical note of caution: actual medieval grammar is never as clean as our textbooks make it look, any more than modern grammar is.
My first inclination was to modernize, but it quickly became clear that this is unnecessary and that even Donaldson's level of normalizing is unnecessary. It is possible to produce an accessible text by normalizing the more obscure spellings of a few function words, and of another few dozen frequently used nouns and verbs, to values that appear elsewhere in Corpus, and the spellings of another few dozen words to values attested in the other principal manuscripts. Altogether, the process involves about 5% of the poem's vocabulary. In the case of one word, the second person pronoun the, this edition substitutes a modern spelling not authorized by the four principal manucripts, in order to avoid confusion with the definite article. (Thee does appear in Harleian 1239.)
Normal editorial practice in the past has been to gloss major class words and leave students to fend for themselves with the grammatical connective tissues. But sentence meaning depends on function words and syntax, and to a new reader, the density of function words and basic verb forms can make a page of text look like an optometrist's chart. It is impractical to gloss these often enough to fix the antique spellings in the mind of a Middle English language learner. And even if it were practical, the tactic would not help the student find a rhythm as a reader. It is impractical, particularly in a survey course, to spend enough time on language at the beginning of the term to solve the problem completely. But it is practical to substitute eye for ye, see for se, alle for al, such for swich, lieth for lith, thenne for than, and pitee for pite (which looks like /pait/ to a modern eye). Or to use hir for the feminine singular pronouns and hire for the genitive plural form. (The spelling her does appear occasionally in Campsall, but hir is intelligible, and it keeps the edited text closer to Corpus and to normal early fifteenth century spelling.)
This normalizing does slightly oversimplify late Middle English grammar and phonetics, but (a) these are well-attested manuscript spellings; and (b) someone who can't follow the poem won't catch the finer grammatical points anyway. The apparatus to the edition includes a list of normalized spellings with their manuscript attestations. For the rest of the vocabulary, I have let Middle English be Middle English. I have not aimed at spelling consistency, as Donaldson and Howard systematically did and as other editors have done, intentionally or not, to greater or lesser degrees.
My normalizing is heaviest at the beginning of Book One; it gradually thins out over the course of the poem, until, by the last half of Book Five, even a beginning student is reading something close to a transcription of Corpus. In other words, normalizing should be both a reading aid and a teaching tool.
Glossing is relatively heavy at first, and glosses are in the right margin, where they are easy to absorb into the flow of reading. Whenever possible, a gloss connects a Middle English word to a present-day descendant. Glosses also clarify syntax when space permits, and notes extend the discussion in some places. The edition marks reduced final and medial e by the diaresis, the same mark used by Tyrwhitt. Otherwise, there is no phonetic marking other than spelling.
The process described to this point would apply equally to a print edition. The advantages of electronic publication will be clear; the most obvious of these are open access and cost. A physical advantage is that explanatory notes can be done as mouseovers, keeping the reader close to the text and giving Middle English learners an added incentive to make use of the notes. A third advantage is ease and speed of revision, as new textual evidence or new issues and sources enter the discussion or as responses to the text suggest the need for additional notes or glosses.