Some Un-Revolutionary Aspects of Computer Editing


Hoyt N. Duggan, University of Virginia

Forthcoming in The Literary Text in the Digital Age, ed. Richard Finneran (Ann Arbor: University of Michigan Press). An earlier version of this essay was delivered at the Annual Convention of the Modern Languages Association, December 27th, 1994, San Diego CA.

Unless otherwise noted, items published by the Institute for Advanced Technology in the Humanities are copyrighted by the authors and may be shared in accordance with the Fair Use provisions of U.S. copyright law. Redistribution or republication on other terms, in any medium, requires express written consent from the author(s) and advance notification of the publisher.

My title makes sense, of course, only because we share a conviction that electronic technology has revolutionized our discipline. Many of us who work with computers in editing are convinced that we stand on the verge of a major reconceptualization of the nature of text, that like printers in the age of the incunabula, we are initiators of a vast cultural project, the ultimate ends of which we cannot see. At the same time, any survey of electronic texts presently available will suggest that this has not yet become a golden age of textual editing. All too often in the preparation and publication of electronic texts, the chief criterion for choosing a text to be "input" for the electronic edition is expiration of copyright.1 Computerized literary scholars have so far demonstrated a tendency to exploit the computers for quick-and-dirty textual solutions, to expend their efforts upon compiling massive amounts of textually uneven materials rather than upon producing reliable scholarly editions. Nevertheless, there is no intrinsic reason why the electronic archive should be the home of out-of-date and inferior texts. Computer technology has enabled Thomas Cable and me to discover the underlying formal constraints governing composition of late Middle English alliterative, metrical constraints undiscovered and largely unsuspected in a century and a half of scholarship.2 Larry Benson's sophisticated program for producing lemmatized concordances using the head words of the Middle English Dictionary offers extraordinary opportunities in stylistics and lexicography.3 As the present volume alone makes abundantly clear, a number of important scholarly editorial projects are underway. Norman Blake's and Peter Robinson's Canterbury Tales Project, already well begun, is being joined by other large projects, such as the Kevin Kiernan, Andrew Prescott, and Paul Szarmach color digital facsimile edition of Beowulf.4 Ralph Hanna, Thorlac Turville-Petre, and I are planning with members of the British Library staff a similar color digital facsimile edition of Cotton Nero A.x. Electronic editions of the Old English penitentials and psalter glosses are in progress or nearing completion. Yuki Matsumoto has recently completed an electronic text of The Destruction of Troy; Vincent McCarren is at work on the late Middle English/Latin Medulla; and a group of scholars headed by Greg Waite in New Zealand is producing a large computerized textbase of early Tudor verse, prose, and drama. For the past several years I have worked with a team of scholars on an electronic archive designed to present in digital form the entire medieval and renaissance textual tradition of the fourteenth-century dream vision Piers Plowman.5 My interest in the text of the poem began in my attempts to determine whether Langland's metrical system and practice were consonant with those of other poets of the "Alliterative Revival." Though my comparisons of the Athlone A and B texts with the manuscripts increased my respect for the extraordinary care and intelligence of the editorial work in both editions, they also convinced me that George Kane and E. Talbot Donaldson--like pre-Tyrwhitt editors of Chaucer--had worked in essential ignorance of the poet's meter, mistaking both his alliterative practice and the rhythmic constraints that governed composition at half-line level.6 In editing both the A and B texts, they occasionally rejected authorial manuscript readings and replaced them with unmetrical, and thus unauthentic, lections and conjectures. Moreover, in the B text, their very full list of variants did not always record metrically substantive variants. Other scholars at the same time pointed to yet other difficulties with the theory and execution of Kane's and Donaldson's text.7

In 1987, Robert Adams and I began to consider collaborating to construct a new critical edition of the B text. Adams created a relational database of Piers Plowman manuscripts, designed to permit elaborate comparisons of the lections within the B tradition with those of the A, C, and Z versions.8 Though his program at that time did not make machine collation feasible, Adams's preliminary efforts convinced us that the B archetype itself could be restored. Moreover, we found the reconstructed B archetype with notable frequency revealed a rationale for Langland's revision that had eluded Kane and Donaldson or had, perhaps, not been considered in their enthusiasm for using the agreement of the A and C versions to edit B. In focusing too narrowly on differences and agreements among the A, B, and C versions at the level of the half-line, Kane and Donaldson from time to time failed to account for possibly authorial changes at the level of the sentence or verse paragraph.9

In 1991 Adams and I were joined by Eric Eliason, Ralph Hanna, and Thorlac Turville-Petre, who also had an interest in the text of Piers Plowman. By then, we were coming to think that only an electronic edition could provide an adequate basis for a radically new approach to text editing, one that puts aside the traditional controversies on whether to create conservative "best text" editions of the poems or to exercise judgment to create more reliable critical editions. As we have continued to design and construct the Archive, we realized the degree to which the centuries-long dispute between conservative and interventionist editors has been technologically and economically constructed. Indeed, we now believe that the historical antagonism between textual conservatives and interventionists is inherent neither to editorial principle nor practice, but rather has been motivated by the limitations of print technology, limitations that need not constrain editors working with electronic media. The economics of publishing printed codices have sustained the often acrid conflict between proponents of two partial editorial ideas, each in itself answering well to practical necessities, but neither in itself fully adequate. That is, the costs of setting, printing, storing, and merchandising books have tended to force editors and their publishers to decide among three practical alternatives: "best" text editions with full or selective variants recorded from other witnesses; parallel text editions (or entire series of "best" texts); or reconstructed critical texts. With any of these alternatives, the characteristically protean medieval text becomes fixed in print either as an editor's reconstruction of an author's putative "original" or as one or more of the "least bad" scribal copies. With each, the exigencies of print deprive the reader, who receives either a privileged editorial critical reconstruction or a privileged scribal version. The editor of the electronic edition need not privilege either kind of text.

The vast expanses of cyberspace permit editors to represent fully even very complicated textual traditions. Editors of works with complex textual histories, like that of Piers Plowman, can produce as many documentary or "best text" editions of individual manuscripts or early printed witnesses as their judgment (or patience) dictates. They can, moreover, produce digital facsimile editions of the original texts in 24 bit color with capacities for image manipulation that sometimes make electronic facsimiles superior to seeing the original in a good light. Moreover, each facsimile edition can be hypertextually linked to its documentary edition. In the case of the Piers Plowman Electronic Archive, we have decided to provide documentary texts of all fifty-four manuscripts and of the three authority-bearing sixteenth-century printed texts, each hypertextually linked with a color digital facsimile, each, moreover, accompanied by codicological, paleographical and linguistic description as well as with textual notes. The scholar who wishes to do so may consult fifty-four "best" texts of the three versions of Piers Plowman, of the six A and C combinations, of the single A and B splice, and of the three combinations of AC with "a long B continuation."10 Within a single large, hypertextually-linked archive, scholars may interrogate the entire textual tradition of the poems. Some scholars may conclude that the editorial project is completed with the production of accurate documentary and facsimile editions of all the witnesses. We take it only to have been well begun, for though the surviving manuscripts constitute the major evidential base from which an editor might attempt to reconstitute authorial texts, not one of the Piers Plowman manuscripts by itself reflects accurately what Langland wrote. Though some manuscripts are in general better witnesses to the authorial text than others, no one manuscript bears more a priori authority than any other. Therefore, if we are to read a text more nearly authorial than is represented in any one manuscript witness, it is incumbent upon editors to construct archetypal and critical editions of the authorial versions from this corpus of variants. Scholars have reasonably complained of the losses and distortions of textual information in printed critical editions. The wealth of textual information conveyed in many Piers Plowman manuscripts by changes in hand, script, and color is ordinarily only hinted at in printed texts. That information will easily be represented in an electronic text both in relatively unmediated fashion in high quality color facsimile editions and in machine-searchable form by sophisticated SGML markup in the documentary editions. Consider, for example, Corpus Christi College, Oxford, MS 201 (henceforth F), in which the scribe typically marked changes between the English alliterative text and the Latin quotations by changing the script from anglicana bastarda to a rubricated textura fere, marking the hierarchy both by color and by change of script. In some instances, the scribe further highlighted the Latin text by beginning a line of text with a Lombard capital. Larger structural divisions within the passus are set off with four-line ornamental capitals, and the smaller strophic divisions are indicated with alternating red and green or red and blue paraph signs. Important words and phrases within the English text are emphasized by red touches on initial letters or in some cases with red touches on every character and with red underlining. A color facsimile will put this textual information directly before those readers who cannot readily consult the manuscript itself. In addition, SGML markup within the transcribed text makes the distribution of such features machine searchable. In short, the customary loss of information from the hand-produced manuscript text by reduction to the printed text need not occur in the electronic text. Readers an ocean away from the originals may experience in a relatively unmediated form most of the physical qualities of the manuscript itself.11

Transcribing the manuscripts inevitably presents interpretive problems. Any transcription, no matter how literal an editor intends it to be, is an interpretive act, an abstraction from the brown and red and blue inks on vellum or paper, an attempt to represent the textual intentions of long dead scribes. Every leaf presents many occasions for the exercise of editorial judgment. Late medieval Piers Plowman manuscripts differ in the number and density of suspensions and abbreviations the scribes utilized, the relative care with which corrections were noted and made, the care or lack of care with which letter forms were made. Most scribes use a variety of allographic forms. Should the editor distinguish sigma "s" from the long form? Should two-lobed "a" be distinguished from the single lobed form? How should occasionally erratic word and morph divisions be represented? Should roman numerals be spelled out or left as is? Should abbreviations and suspensions be resolved or be represented graphically or by entity references?

Peter Robinson and Elizabeth Solopova are right to argue that editors should provide graphemic rather than graphetic transcriptions.12 That is to say, editors will wish to use markup to represent just those graphic features that carry textual meaning and ignore other graphic features that do not convey such meaning. Distinguishing marks that carry meaning from meaningless flourishes is not always simple. In initial transcriptions editors must record all of the ambiguous graphic features so that they may be subjected to graphemic analysis. For instance, scribal flourishes, such as those frequently appearing on word-terminal <g> or <k>, must be recorded until it can be determined whether the scribe intended them to represent a final <e> or not.13

SGML markup permits editors of documentary texts to record discrepancies between scribal intention and performance. Lapses of the pen can be easily entered in dual form, one with the text as the scribe wrote it and another with his probable intention. For example, the F scribe refers to Christ as a "kaue child" in a context concerned with Christ's masculinity. This lection we have entered with a <sic> tag so that when the text is later collated with other manuscripts its true reading is reflected, but the same tag will include the intended reading "knaue child," so that a reader searching for "knaue" will be able to locate this occurrence. Neither version is privileged save by the interest of the user. Nor is either suppressed editorially. Both readings, what the scribe actually wrote and what the editor thinks he must have intended to write, are represented in the edited text.

Simple accuracy is perhaps the hardest thing to achieve in an edition, and electronic technology scarcely affects the labor of transcribing and proofreading. Transcription at a keyboard, like writing on animal skins with a quill, still takes place character by character. Optical character recognition software does not presently exist to shorten the task. Proofreading still requires serial re-readings to compensate for eye-skip, arrhythmia, dittography, homoeoteleuton, or for the manifold other failures of concentration that have marred scribal efforts since literacy began. Our original plan provided for five separate readings of each transcription against the photocopies and a final proofreading against the original manuscript, but our practical experience with F suggests that in the case of a difficult or complex manuscript more readings will be necessary.14

The apparatus criticus of the electronic edition will differ inevitably from that of printed editions. For instance, traditional printed editions of Middle English texts include somewhere in the introduction a description of the paleographic features of the scribal hand, usually accompanied with a list, sometimes with illustrations, of the common abbreviations and suspensions used by the scribe. Frequently editors provide black-and-white photographic facsimile pages as part of the front matter of printed texts so that readers may see a representative page or two. In the electronic edition, the editor can supply, as I did in my IATH research report on the World Wide Web, digital images of a full set of abbreviations and suspensions. The one in my research report is crude, even by the standards of present technology, taken as it is by cut-and-paste methods from xerox flow copy of the manuscript and then scanned on a flatbed scanner. In future, such paleographic descriptions can be constructed hypertextually with color facsimiles and with sufficient examples to produce for the user something approaching full analysis of the scribe's practice. Moreover, such descriptions, unlike their counterpart in printed texts, will in the completed Archive constitute fifty-four hypertextually-linked data bases of late medieval scribal practice.

When all fifty-four manuscripts have been transcribed with SGML markup, readers will have fifty-four "best text" documentary editions, each accompanied by and linked to a color facsimile edition. How many scholars, including those most insistent upon privileging scribal versions of the poem, will read fifty-four editions of very similar poems? Even in the case of a culturally important work such as Piers Plowman, it is not at all clear that so many marginally different "poems" will find readers. As it stands, few have read all three published versions with the same attention they have given B. How many scholars will wish to read several dozen manuscript witnesses differing in quite minor ways one from another? This profusion of textual material, even with meticulous hypertextual linking, is at once too much and too little. Too much, certainly, if one thinks of the text only as an aesthetic object. Such a view of text is, of course, more than a little parochial, since literary texts serve a variety of other functions in modern attempts to recreate and understand our past. Less parochially, fifty-four electronic transcriptions and facsimiles--perhaps none of them ever serving as a traditional reading text--will offer scholars not only new ways to study the text and the textual tradition of the poem but also possibilities for gaining fresh insights into other aspects of late medieval literary culture. Students of text reception may readily access formerly inaccessible marginal and interlinear annotations or significant scribal changes to the text itself. The Archive will enable study of the changes both in language and literary focus wrought by the revising sixteenth-century scribe who created Toshiyuki Takamiya, MS 23, or by Robert Crowley's protestant rebaptism of the poem in his three 1551 printed editions of the B text, each converting Langland's Middle English into something more appropriate for its Tudor audience. Since Piers Plowman was copied in virtually every late Middle English dialect, historical linguists will be able to study patterns of regional variation in lexicon, phonology, and orthography. The facsimiles will be useful to students who once lacked access to large collections of primary manuscript materials. Moreover, students of form and style and meter can add their own markup for other, as yet unimagined, kinds of study. It matters little that no one is ever likely to want to read all fifty-four documents. Many will want to use them. Even so, despite the proclaimed "death of the author," many readers still want to read the poems that Langland wrote, as he wrote them and without scribal lapses or accretions. Medievalists, of course, have never had occasion to celebrate the kind of author whose demise has been announced.15 We know, for example, virtually nothing about the historical poet who composed the three versions of Piers Plowman. The autobiography of Long Will, the dreamer-protagonist, is constructed as much by thematic considerations as by knowable events in the poet's life and may well be completely fictional. Whether "Langland" means anything to us other than a name to attach to the three poems, we recognize in those poems the work of a poet of extraordinary power. We want to read his words. Without wishing for one moment to deny the intrinsic interest of scribal practice to bibliographers, students of reception, linguists, metrists, or cultural historians, we recognize that scribal accretions are of secondary interest to most readers. In any case, the import and direction of scribal changes are recognizable only in relation to a concept of the authorial text. That is, we cannot study the reception history of texts like Piers Plowman as it is reflected in the manuscript witnesses until editorial work distinguishes what each scribe or editor did from what was inherited from an exemplar. Without doing collations, without establishing archetypal or critical texts, without an attempt to construct a stemma, textual variation is difference without significance. Though the fifty-four documentary editions constitute necessary groundwork for critical editing, each witness with its own set of differences will not be fully meaningful until its relationship to the other texts in the tradition can be established.16 That requires critical editing.

This reference to constructing archetypal and critical texts moves us into the "unrevolutionary aspects of textual editing," even into what some will consider hopelessly out-dated notions of editing. It is currently fashionable to think claims to literary knowledge vestigial remains of an earlier, more naive and optimistic age.17 A fashionably automatic rejection of archetypes and critical texts prevails among post-modernists and some romantically inclined traditional scholars. One traditional scholar has complained of "the sterile operating room (or terminal intensive care unit) of the modern critical edition" of medieval poems, complaining that they have functioned to suppress and distort "the realities of medieval texts and manuscript production."18 Scholars influenced by Jerome McGann's work with nineteenth-century authors and his sociological critique of the Greg-Bowers school of critical editing also reject the possibility of rescuing medieval texts from the depredations of scribes. Still others, inspired by post-structuralist "theory," proclaim their indifference to authorial texts and reject critical editing as a viable scholarly project.19

Though readers understandably want to read authorial texts, editors of Piers Plowman have always known that the authorial text has disappeared, lost to the accidents of time and hand-copying. Even if, miraculously, we were to reconstitute from the maze of scribal witnesses Langland's final texts precisely as he intended to leave them, we would never be confident that we had done so. This situation, of course, does not distinguish the texts of Piers Plowman from those of Chaucer or most other medieval authors. Lacking autograph manuscripts, editors of medieval texts inevitably engage in an attempt to understand the relationship of the extant witnesses to a lost text. Inevitably they will create a new text, something of their moment in history, using all of the available documentary evidence or some portion of it to attempt to recreate what we have lost to time. In this important respect, editors of medieval texts differ from editors of more recent texts such as my colleague Jerry McGann deals with in The Rossetti Archive.20 Unlike the editor of a medieval work, he enjoys a surfeit of authority. He works with texts either handwritten by the author or printed under his supervision, where differences between and among versions are, for the most part, differences between authorial texts. In the case of Piers Plowman, we have not separate versions of the poet's work but flawed documentary witnesses to the work.

The attempt to recreate a perfect authorial text is, under the circumstances, quixotic. Textual scholars--both conservative traditional scholars and progressive post-structuralist theorists--agree at least on that. Typically, the conservative critique consists of the following points: the empirical observation that modern critical editors not infrequently reach diametrically opposed conclusions from the same set of basic data, a reminder that medieval scribes, whatever their deficiencies of intellect or attention, nevertheless knew the language in ways that modern scholars can never know it, and the claim that a conservatively edited, well chosen "best text" at the very least provides a genuinely medieval document that some medieval readers could have read, whereas the critical edition is a hybrid modern monster, one that never existed before the cultural project of modern scholarship.21 Postmodern theorists have added the further objection that the critical edition fetishizes the author. Tim Machan, for instance, though he does not deny that someone called William Langland "may have been the efficient cause of the text, or that it is possible to recover the text he was responsible for, or that his own efforts may be superior to others' according to certain aesthetic standards," argues that "the valorization of his text is necessarily an interpretive imposition on the manuscript evidence."22 Machan characterizes the goal of traditional textual criticism, the establishment of an authorial text, as "only one of the things, and perhaps the most historically problematic, which one can do with a medieval work" (13). The traditional privileging of the authorial is vicious, from Machan's point of view, largely because it "precludes--pronounces 'wrong' or 'other'--other kinds of evidence and analysis which may in face be less historically problematic and are, in any case, equally valid as aesthetic activities"(14). Machan's objections, occasioned by printed critical editions are, of course, made irrelevant by the electronic archive. I think they are irrelevant to good critical editing in printed texts to the degree that they rely upon Machan's creation of an unnecessary and false disjunction between an editor's respect for the material objects in which texts survive and his desire to reconstitute the authorial text. Good critical editing requires that a competent editor pay minute attention to the manuscript as a physical object, recording and analyzing its physical features, its ordinatio, its provenance, its original audience, its punctuation--in brief, its thingness. An editor's preliminary attention to codicology and paleography is not exclusively preliminary. Ralph Hanna wrote a description of the casting-off procedures of the scribe who copied MS Lambeth Palace 491, but it was only after he and David Lawton had completed their critical edition of the poem that they were able to account for Hanna's observations of the physical text.23

In the electronic archive, the reconstituted authorial text is privileged only at the level where such privileging is appropriate. At such a level, the Piers Plowman team will attempt to recuperate the authorial texts of the three versions. At a different level of the archive, we will attempt to represent the scribal versions of the text--their ordinatio of the page, their passus divisions and rubrications, their verse paragraph markers, their hierarchy of scripts, their use of color and line in emphasizing and presenting text, their choice of dialect terms, spellings, and so forth. An electronic archive can and should be as faithful at this level to scribal intentions as--on a different level of abstraction--it is to the author's text. In the electronic archive, the user/reader will be able to move easily between levels, but the editors must first do the traditional editorial work and supply the linkages.24

In the case of many medieval texts, attempts to publish a reconstructed authorial text are misguided. When, for example, the evidence shows that the extant copies are not derived from a common written original--as is frequently the case with lyric poems, some popular romances, tropes, learned commentaries, saints lives, to name only a few genres in which textual mouvance represents the ordinary state of affairs--there is no "original" to be derived from the mass of variants.25 However, until the editor has analyzed the evidence, until he has undertaken the fundamental activities of critical editing, he cannot know whether--or to what extent--the textual tradition is monogenous or not. Critical editing does not always mean that a single authorial text will be reconstituted, only that the editor has thought in a systematic and disciplined way about the documents and other textual evidence.

Under some circumstances, reliable authorial texts cannot be reconstructed by critical methods from scribal copies, but in many cases, something nearly approaching the authorial text can be restored. Thorlac Turville-Petre and I edited one such text, The Wars of Alexander, in which the accidental survival of two good manuscripts, the relative fidelity of the poet's translation to a known source which itself survives in many witnesses, and our knowledge of the metrical constraints under which the poet wrote all conspired together to make a critical edition possible.26 We know, of course, that we did not reconstruct the original poem, only a modern version of it, but that modern version is closer to that original poem than either of the extant witnesses. Ralph Hanna and David Lawton have similarly just completed a critical edition of The Siege of Jerusalem, a poem with known sources and nine surviving witnesses.27 Their text, like Turville-Petre's and my text of The Wars of Alexander, which is not the original, has exactly as much authority as its editors can persuade readers that it carries. That is, the reconstructed text, like interpretive scholarship in literature--or for that matter in physics or history--is a rhetorical construct within the framework of an intellectual discipline.

Our preliminary work with Piers Plowman supports Skeat's theory that the poem survives in three different versions, but we are willing to forsake that theory if the evidence should require that we do so. Our work so far with the B text supports the Kane-Donaldson view that the extant witnesses are all descended from a single archetype, an early text already textually at some remove from Langland's final text. We are skeptical of various claims that any large number of textual variants represent authorial revisions, though again that is a position that can be reversed when all the evidence is available and subjected to analysis.28 We are inclined to think that the Athlone editors and A.V.C. Schmidt were correct in finding two major families of manuscripts in the textual tradition of the B version, though again we will cheerfully change our minds if the evidence should require it.29 Once all the manuscripts have been transcribed with SGML markup, we will have immediate access to all the relevant textual evidence, more accessible and useable than it ever has been before because the computer puts at the editor's immediate disposal a vast array of materials minutely searchable. I will use a single example to indicate the range of evidence electronic editors will be able to bring to bear upon textual problems. It should be remembered that the archival base for our editorial decisions is still in preparation. We have, so far, only a tiny fraction of the textual information we plan to accumulate, especially from the textual tradition of C. The following example, therefore, is a provisional hypothesis, and can only suggest how we will distinguish authorial from scribal lections.30

In Passus 13.140, the B witnesses are about evenly divided on the reading of the a-verse:

With wordes and with werkes, quod she ....

and with] WHmCrLMRF; and YGOC2CB

Kane and Donaldson, probably on aesthetic grounds (they almost invariably select the more laconic or syllabically shorter variant), select the reading of YGOC2CB.31 A similar set of variants occurs at B 3.2:

Wiþ Bedeles and with baillies .... WHmYCLR

With bedellus and baillifs .... MHm2CrGOC2BH

In this instance, both A and C versions agree with the more laconic reading, and Kane-Donaldson again select the shorter lection. If our final analysis--including use of cladistic analysis--continues to show but two main textual families, the archetypal reading here cannot be determined by recension.32 The logic of attestation, however, tends to support the first reading with its agreement of four of the best beta manuscripts, CHmLW, with the better alpha witness R. F's characteristic eccentricity complicates the situation with its reversed a- and b-verses. F's b-verse "with bedel and bayly" offers weak support to the group without the second "with." The poets of the Alliterative Revival composed in phrasal units of metrically regular half-lines.33 Their wonderfully flexible grammar of composition represents a metrically defined subset of the syntactic possibilities of Middle English. These poets, Langland included, tended not only to say the same things in the same way in lexical formulae but also used the same syntactic-rhythmic frames over and over in the course of a long poem. The poems manifest multiple rhythmic realizations of the same grammatical collocations. The relevant grammetrical pattern for the preceding example can be described as follows:34


prep + NPobj + { cj + prep } + (art) NPobj


With presently available software, it is a tedious but relatively simple matter to assemble all instances of this pattern in all three versions of Piers Plowman.35 Both Langland's usus scribendi and the rhythmical rationale for the usage become instantly clear. Though he composed a number of a-verses in which only one strong dip appears, he generally preferred to provide three or four syllables in the one strong dip.36 The preposition is repeated where it is needed to provide a second or third unstressed syllable in a single strong medial dip:

     A 5.68 B 5.86   With werkes or with wordes ...
     A 10.4 B9.4     With wynde and with water ...
     C.8.321         With craym and with croddes .... 
     B 13.289        With inwit and with outwitt ... /V/.
     B 17.69         Wyth wyn and with oyle ....
It is also repeated when it provides a syllable to make a strong medial dip:

     B 5.600         With no lede but with Loue ...
The second with tends to be omitted when the noun head of the prepositional phrase is trisyllabic or when other unstressed syllables intervene, e.g.:
     B 5.89          With bakbitynge and bismer  ....
     B 7.12          With patriarkes and prophetes ....
     C.2.91          With vsurye and auaryce .... 
The second preposition is also omitted generally when the first noun phrase contains both a stressed adjective and noun:

     AB P.16         Wiþ depe dikes & derke ...

A 7.285 Wiþ good ale & glotonye ...

B 6.300 With grene poret and pesen ...

B 13.359 With false mesures and mette ....

C.18.154 With fyue fisches and loues ....

However, it appears to be required when the second noun phrase has two stresses:

     A 7.205         Wiþ fuyr or wiþ false men ... 

B 18.305 With glorie and with grete li3te ...

C 19.243 With wyles and with luyther wittes ....

Since the language permits that either the conjunction and preposition may be present or omitted, depending upon the rhythm of the verse, it should not surprise us that scribal variation is common. The set of variants that began this investigation appear again in the poem:

   (1)    A 5.68     Wiþ werkis [or wiþ] wordis ...

or wiþ] UVHEKWM; or DLN; and with R

B. 5.86 With werkes or with wordes ...

with (2)] om. GH

(2) A 7.210 Wiþ mete or [wiþ] mone ...

wiþ (2)] RUDChVHJLKNMH3; in E; om. TW

(3) A 10.4 Wiþ wynd & wiþ watir ...

wiþ (2)] om. DJW

B 9.4 With wynde and with water ....

No variants.

C 10.130 With wynd and with water ... (Pearsall)

C.11.130 With wynd & water .... (Skeat)37

In the case of B 13.140, the agreement of both alpha and beta families shows the lection rejected in the Athlone text to have been archetypal, and the grammetrical evidence suggests Langland himself is responsible for the repetition of the preposition.38 In the case of B 3.2, agreement of versions A and C, added to the grammetrical evidence, indicates that the archetypal reading is in error and that the repeated with is scribal.

The arguments advanced here are, I repeat, provisional. Only a portion of the evidence is now available in machine-searchable form, and a search of other monosyllabic prepositions, for instance, may complicate or controvert this hypothesis. Consideration of with alone reveals a few anomalies. An apparent exception appears in Skeat's text at B 20.124:

          With glosynges and with gabbynges ....

In that verse, about half the witnesses also omit the metrically and syntactically otiose second with: C2GYOCBR lack it; WHmCrLMF have it. It is worth remarking that the variation occurs in both alpha and beta families as it did at B 3.2. In this verse, Kane's and Donaldson's preference for the more laconic reading led them to the correct reading, rejecting the second with. More problematic verses appear in which all of the witnesses agree in a reading inconsistent with the patterns observed above:
          B 15.520   With londes and ledes .... 

B 19.230 With sellyng and buggynge ... (also C.21.235)

Whether such verses reflect corruption of the archetype or constitute important evidence against the hypothesis advanced here must await analysis of the larger corpus we are assembling.

I chose this particular a-verse pattern, partly to demonstrate how swiftly and reliably the computer permits us to order a substantial base of evidence, and partly because it raises the critical issue of the recovery of authorial intention. Did Langland intend to write "With x and with y" or "With x and y"? Indeed, can he be said to have had any particular intention at all in regard to these syntactically and semantically redundant monosyllables? How would Langland have explained his choice? Would he have known consciously what he intended? Or would he have had to speak the verse to hear what "sounded" right? I suspect it would have been the latter, that he and other fourteenth-century alliterative poets would not have made conscious decisions to include or elide prepositions in such half-lines any more than we do in ordinary speech. Nevertheless, one may argue that Langland intended to compose alliterative verse and that his conscious intention to do so immediately involved him in thousands of unconscious intentions entailed by the metrical grammar of composition.

Generative metrists in recent years have shown that very powerful unconscious linguistic rules underlie and govern poetic composition, just as similar rules determine the shape of our every utterance, rules inaccessible to most of us, though no less powerful for being so.39 Such rules are rarely broken because they do not rise to a level of conscious control. The full set of such rules for alliterative poetry have not been discovered, though we have recently made some steps in the direction of understanding them. The hypothesis proposed here is not such a rule, though it may be preliminary to discovery of such a rule. Within the limitations imposed by the surviving evidence, we can recover and describe some of the linguistic and metrical rules governing composition. In the instances we have just examined, if the tentative hypothesis proposed to account for Langland's practice with respect to this grammetrical frame should hold up when all the evidence has been collected and considered, we will have another basis for determining what Langland "intended" to write.

Electronic technology offers powerful new tools to scholars who want to distinguish what Langland and other poets wrote from scribal changes. Degrees of accuracy and consistency formerly unattainable by even the most scrupulous editors are enabled by the computer. Machine collations, unmarred by the inevitable human failures of eye, attention, and memory, will replace hand-produced collations. Deliberations on archetypal and critical texts will be speeded by virtually instantaneous access to all the other manuscript versions, to concordances of each manuscript version as well as to machine readable texts of other modern editions of all three versions. In addition, we and other scholars will have access to electronic texts of every other major Middle English alliterative poem and to electronic concordances to each of those texts. Presented with a set of variant readings, we can in a matter of moments examine all of the contexts in which the alternative lections appear, either elsewhere in Piers Plowman or even within the whole tradition of alliterative verse. It would be a pity if the postmodern distrust of empirical evidence and distaste for the methods of the harder sciences should cause us to turn our backs on critical editing just as the tools become available that will help us do it better than ever before.40


  • 1 The editor of The European English Messenger 3 (1994): 86-87, describes his experience with an inaccurate electronic edition of Virginia Woolf's The Waves and the absence of any response to a "set of letters to literary critics and media experts at other universities, asking what on earth one is to do with all this electronically stored textery." He concludes with an appeal to readers "who have discovered what to do with or to a text on the computer [to] advise us as to what pleasures those of us miss who are puzzled by all these oysters that refuse to yield the pearls they had promised. Surely the emperour of CD does have some clothes on?"Back

  • 2 Thomas Cable, The English Alliterative Tradition (Philadelphia: University of Pennsylvania Press, 1991); Hoyt N. Duggan, "The Shape of the B-Verse in Middle English Alliterative Poetry," Speculum 61 (1986): 565-67; "Alliterative Patterning as a Basis for Emendation in Middle English Alliterative Poetry," SAC 8 (1986): 73-105; and "Final -e and the Rhythmic Structure of the B-Verse in Middle English Alliterative Poetry," MP 86 (1988): 119-45. "Notes toward a Theory of Langland's Meter," YLS 1 (1987): 41-70; "Langland's Dialect and Final -e," SAC 12 (1990): 157-91; "Stress Assignment in Middle English Alliterative Poetry," JEGP 89 (1990): 309-329.Back

  • 3 For a description of the program, see Larry Benson, "A Lemmatized Concordance of Chaucer," in Computer-Based Chaucer Studies, ed. Ian Lancashire, CCH Working Papers 3 (Toronto: University of Toronto Centre for Computing in the Humanities, 1993), pp. 141-160.Back

  • 4 For the Beowulf facsimile project, see Kevin Kiernan's research report "Digital Preservation, Restoration, and Dissemination of Medieval Manuscripts," published on the World Wide Web at the following URL: For the Canterbury Tales Project, see Norman Blake and Peter Robinson, eds. The Canterbury Tales Project: Occasional Papers Volume I, Office for Humanities Communication Publications, 5 (Oxford: Oxford University Computing Services, 1993).Back

  • 5 For a description of the project, see my "The Electronic Piers Plowman: A New Diplomatic-Critical Edition," Æstel 1 (1993): 1-21, and my hypertextual research report "Creating an Electronic Archive of Piers Plowman," Publications of the Institute for Advanced Technology in the Humanities, Research Reports, Second Series (1994) on the World Wide Web at the following URL:

  • 6 George Kane, ed., Piers Plowman: The A Version (London: Athlone Press, 1960); George Kane and E. Talbot Donaldson, eds., Piers Plowman: The B Version (London: Athlone Press, 1975).Back

  • 7 David C. Fowler, "A New Edition of the B Text of Piers Plowman," Yearbook of English Studies (1977): 23-42; Charlotte Brewer, "The Textual Principles of Kane's A-Text," Yearbook of Langland Studies 3 (1989): 67-90; "Authorial Vs. Scribal Writing in Piers Plowman," in Medieval Literature: Texts and Interpretation, ed. Tim William Machan, MRTS Texts & Studies, 79 (Binghamton, NY: MRTS, 1991), pp. 59-89; and "George Kane's Processes of Revision," in Crux and Controversy in Middle English Textual Criticism, ed. A. J. Minnis and Charlotte Brewer (Cambridge, 1992), pp. 71-96; Robert Adams, "Editing Piers Plowman B: The Imperative of an Intermittently Critical Edition," SB 45 (1992): 31-68; Thorlac Turville-Petre, "Review of Kane-Donaldson, Piers Plowman: The B Version," Studia Neophilologica 49 (1977): 153-155.Back

  • 8 For the controversy on the so-called Z-version, see William Langland, Piers Plowman: The Z Version, eds. A. G. Rigg and Charlotte Brewer (Toronto: Pontifical Institute of Mediaeval Studies, 1983), and George Kane, "The 'Z Version' of Piers Plowman," Speculum 60 (1985): 910-930. Kane makes the more compelling case.Back

  • 9 "Editing Piers Plowman B: The Imperative of an Intermittently Critical Edition," SB 45 (1992): 31-68.Back

  • 10 George Kane, "The Text," in A Companion to Piers Plowman, ed. John A. Alford (Berkeley and Los Angeles: University of California Press, 1988), p. 178.Back

  • 11 Murray McGillivray, "Electronic Representation of Chaucer Manuscripts: Possibilities and Limitations," in Computer-Based Chaucer Studies, pp. 1-15, predicts that electronic technology will bring virtual reality to manuscript studies so that readers can be given the smell and feel of turning the pages of the virtual manuscript. At the rate of technological change, that prediction cannot be dismissed as hopelessly utopian.

    We considered for a time producing our digital facsimiles from black-and-white microfilm. There are some obvious reasons for doing that: in many cases, microfilm copies already exist and thus do not require further handling of the originals. They are, furthermore, relatively inexpensive to purchase initially and to reproduce in quantity. For many documents, the images produced are clear and entirely readable. Readers of microfilm lose, of course, all of the textual information encoded in color. However, editors have used microfilm copies of manuscripts for editorial work for well over a century, and digitized microfilm is likely to continue to provide a basis for many editions. For instance, the editors of The Canterbury Tales Project are preparing some of their digital facsimiles from black-and-white microfilm. However, we have come to believe, partly on the basis of having begun our experiments with an unusually difficult manuscript (F), that digitizing black-and-white microfilm is a secondbest, make-do solution and that editors will wish, where it is at all possible, to produce 24 bit color facsimiles.Back

  • 12 "Guidelines for Transcription of the Manuscripts of the Wife of Bath's Prologue," The Canterbury Tales Project, Occasional Papers Volume I, Office for Humanities Communication Publications, 5 (Oxford: Oxford University Computing Services, 1993), pp. 19-52.Back

  • 13 Scribal behavior is frequently inconsistent. The scribe who copied F from time to time adds a flourish to words ending in <g, k, t>. When the end of the word coincides with a phrasal boundary or the caesura, the scribe not infrequently uses that flourish to serve in place of his usual virgule, but in other instances, as at the top of fol. 24v, the flourish appears to be utterly meaningless, and in some other instances, he writes both the flourish and a virgule at a phrasal boundary.Back

  • 14 Since graduate student members of my team of assistants were learning both how to read medieval handwriting and to insert SGML markup while they were transcribing and proofreading, it appears likely that with more experience fewer passes will be necessary.Back

  • 15 For thoughtful comment on this formalist gambit, see Seán Burke, The Death and Return of the Author: Criticism and Subjectivity in Barthes, Foucault and Derrida (Edinburgh: Edinburgh University Press, 1992).Back

  • 16 Robert Adams decries "the recent penchant for ... treating scribal errors as instances of medieval literary criticism" ("Editing Piers Plowman B," p. 33). Cf. Barry Windeatt, "The Scribes as Chaucer's Early Critics," SAC 1 (1979): 119-141, and Derek Pearsall, "Editing Medieval Texts: Some Developments and Some Problems," in Textual Criticism and Literary Interpretation, ed. Jerome J. McGann (Chicago: University of Chicago Press, 1985), p. 103, with George Kane, "The Text," p. 194.Back

  • 17 For critiques of the hermeneutic suspicion characteristic of the post-modern establishment, see Paisley Livingston, Literary knowledge : humanistic inquiry and the philosophy of science (Ithaca, N.Y. : Cornell University Press, 1988); Humberto Eco, The Limits of Interpretation (Chicago: University of Chicago Press, 1990); Interpretation and overinterpretation, with Richard Rorty, Jonathan Culler, Christine Brooke-Rose (Cambridge: Cambridge University Press, 1992); and Paul R. Gross and Norman Levitt, Higher superstition : the academic left and its quarrels with science (Baltimore: Johns Hopkins University Press, 1994). Back

  • 18 Pearsall, "Editing Middle English Texts," pp. 104, 102. Back

  • 19 Some students of medieval texts have misread McGann, who distinguishes between the textual situation in pre-print societies and that after printing houses had been established. McGann is aware that whereas the editor of a nineteenth-century poet must select his text from a variety of authorial texts the editor of the classical or medieval text may lack any authorial document. Though medieval poets may have expected scribes to make mistakes and change their texts--and we know what Petrarch and Chaucer thought of meddling or careless scribes--such an expectation hardly constitutes an analogy with a publishing house's style sheet. As John Bowers has recently shown in "Hoccleve's Two Copiers of Lerne to Dye: Implications for Textual Critics," in The Papers of the Bibliographical Society of America 83 (1989): 437-472, even authorially produced texts can often lack textual authority. I believe that Bowers has drawn the wrong implications from Hoccleve's scribal practice and that the poet, acting as his own scribe, worked from a scribal psychology. Others who have less difficulty than I do in proofreading my own work may be more receptive to Bower's take on the evidence.Back

  • 20 An elegant hypertextual demonstration of the textual principles at issue in constructing the Rossetti Archive is available on the World Wide Web at the following URL: Click on "Publications of the Institute."Back

  • 21 For a cogent critique of critical editing from a conservative position, see E. G. Stanley, "Unideal Principles of Editing Old English Verse," PBA 70 (1987): 231-273.Back

  • 22 Tim William Machan, "Late Middle English Texts and the Higher and Lower Criticism," in Medieval Literature: Texts and Interpretation, ed. Tim William Machan, Medieval and Renaissance Texts and Studies, 79 (Binghamton: MRTS, 1991), p. 11. Machan's fuller development of his theoretical stance, Textual Criticism and Middle English Texts (Charlottesville, VA: University of Virginia Press, 1994), arrived too late to be considered here. Similar textual positions, informed by recent critical theory, have been staked by Robert S. Sturges, "Textual Scholarship: Ideologies of Literary Production," Exemplaria 3 (1991): 109-131, and John Bowers, "Hoccleve's Two Copies."Back

  • 23 Ralph Hanna, III, "Booklets in Medieval Manuscript: Further Considerations", SB 39 (1986): 100-111. The Hanna-Lawton edition of The Siege of Jerusalem is forthcoming (Colleagues Press, 1995).Back

  • 24 Bernard Cerquiglini, Eloge de la variante; Histoire critique de la philologie (Paris: Seuil, 1989), imagines a new kind of electronic text not unlike that proposed here, though he appears to be content with the archival element of compiling vast collections of linked data but without considering it necessary to move beyond accumulation of data. As Mary B. Speer trenchantly remarks, "What Cerquiglini envisions is a decentered, pluralistic collection of transcriptions, accompanied by the mechanisms for comparative searches: medieval texts defined, in effect, as the raw material for linguistic enquiry. Criteria of authorial or redactional authenticity, philological truth, and scribal error or idiosyncrasy would be replaced by mere usefulness" ("Editing Old French Texts in the Eighties: Theory and Practice," RPh 45 (1991): 22. Back

  • 25 Rupert Pickens, "Jaufré Rudel et la poétique de la mouvance," CCM 20 (1977): 323-337; (ed.) The Songs of Jaufré Rudel, Studies and Texts 41 (Toronto: Pontifical Institute of Mediaeval Studies, 1978); A. C. Baugh, "Improvisation in the Middle English Romance," Proceedings of the American Philosophical Society 103 (1959): 418-454; "The Middle English Romance: Some Questions of Creation, Presentation, and Preservation," Speculum 42 (1967): 1-31; Murray McGillivray, Memorization in the Transmission of the Middle English Romances, The Albert Bates Lord Studies in Oral Tradition, 5 (New York: Garland, 1990); Sherry L. Reames, "Mouvance and Interpretation in Late-Medieval Latin: The Legend of Saint Cecilia in British Breviaries," in Medieval Literature: Texts and Interpretation, 159-189; Julia Boffey, "Middle English Lyrics: Texts and Interpretation," in Medieval Literature: Texts and Interpretation, 121-138; Gunilla Iversen, "Problems in the Editing of Tropes," TEXT 1 (1981): 95-132; Marjorie Curry Woods, "Editing Medieval Commentaries: Problems and a Proposed Solution," TEXT 1 (1981): 133-45.Back

  • 26 The Wars of Alexander, EETS SS 10 (Oxford: Oxford University Press, 1989).Back

  • 27 See David Lawton's discussion of their editorial use of metrical theory in "The Idea of Alliterative Poetry: Alliterative Meter and Piers Plowman," in Suche Werkis to Werche: Essays on Piers Plowman In Honor of David C. Fowler, ed. Míceál F. Vaughan (East Lansing: Colleagues Press, 1993), pp. 147-168.Back

  • 28 Charlotte Brewer has argued for extensive authorial revisions in the texts of Piers Plowman in "The Textual Principles of Kane's A-Text," YLS 3 (1989): 67-90, and in "Authorial Vs. Scribal Writing in Piers Plowman," in Medieval Literature: Texts and Interpretations, pp. 59-89. For a critique of Brewer's argument, see Adams, "Editing Piers Plowman B," 64-65, Ralph Hanna, "Producing Manuscripts and Editions" in Crux and Controversy in Middle English Textual Criticism, eds. A. J. Minnis and Charlotte Brewer (Woodbridge, Suffolk: D. S. Brewer, 1992), pp. 109-130; and my review of Machan's Medieval Literature: Texts and Interpretations in SAC 15 (1993): 234-235.Back

  • 29 A. V. C. Schmidt, ed. William Langland. The Vision of Piers Plowman (London: Dent, 1984), pp. xxxv-xxxviii; Kane-Donaldson, Piers Plowman: The B Version, pp. 16-69.Back

  • 30 Citations from A are taken from Kane's edition. Those from B are from my computerized copy of W. W. Skeat's edition The Vision of William Concerning Piers the Plowman in Three Parallel Texts ...., 2 vols. (Oxford: Oxford University Press, 1886), while quotations from the C version are from Derek Pearsall's edition, Piers Plowman by William Langland. An Edition of the C-Text (London: Edward Arnold, 1978). The variant lections are supplied from Kane's A text and the Kane-Donaldson text of B.Back

  • 31 Both Schmidt and Kane-Donaldson tend to prefer the syllabically shorter reading when other evidence cannot be adduced. In general, that preference is shaped by both classical poetry and by post-Renaissance verse traditions, traditions perhaps not fully relevant to Middle English alliterative poetry.Back

  • 32 Cladistic analysis is a methodology developed over the past three decades by evolutionary biologists to reconstruct the descent of related species. Both evolutionary biologists and textual critics seek to explain the existence of a varied population as the product of branching descents over time from a common ancestor. For cladistic theory, see Elliot Sober, Reconstructing the Past (Cambridge, MA: MIT Press, 1988), and N. I. Platnick and H. D. Cameron, "Cladistic Methods in Textual Linguistic and Phylogenetic Analysis," Systematic Zoology 26 (1977): 380-385. For successful attempts to use cladistic analysis to determine lines of manuscript descent, see Peter M. W. Robinson and R. J. O'Hara, "Cladistic Analysis of an Old Norse Manuscript Tradition," Research in Humanities Computing 4, ed. Nancy Ide and Susan Hockey (forthcoming), and the same authors' "Computer-Assisted Methods of Stemmatic Analysis," in The Canterbury Tales Project: Occasional Papers Volume I, pp. 53-74.Back

  • 33 For the notion of grammetrical frames, see R. F. Lawrence, "The Formulaic Theory and its Application to English Alliterative Poetry," in Essays on Style and Language, ed. Roger Fowler (London: Routledge & Kegan Paul, 1966), pp. 166-183; "Formula and Rhythm in The Wars of Alexander," ES 51 (1970): 97-112; and my "The Rôle of Formulae in the Dissemination of a Middle English Alliterative Romance," SB 19 (1976): 265-288.Back

  • 34 Optional elements are in parentheses. The central stacked elements indicate that the conjunction or preposition may be omitted, but not both. In this frame, both the conjunction and second preposition are elidable because they are strongly implied. NP=noun phrase; art=article; cj=conjunction; prep=preposition; unstressed elements are indicated in lower case. One syllable in the noun head of NP carries both stress and alliteration.Back

  • 35 Over three dozen a-verses of this form with the preposition "with" occur in the three versions of Piers Plowman, counting verses repeated in two or three versions only once. Shortly after I had laboriously assembled the data presented here using a word processing program and a set of macros, Thornton Staples and Susan Gants at The Institute for Advanced Technology in the Humanities began to plan with present and former Fellows to develop a search and display program to facilitate precisely this kind of textual search.Back

  • 36 A weak dip consists of a single unstressed syllable. A strong dip in the a-verse consists of two to six or (rarely) seven metrically unstressed syllables. The preponderance of a-verses display two strong dips in Langland's verse as well as in the a-verses of other Middle English alliterative poets, but some verses of the form x/xx/x appear in all three versions of Piers Plowman.Back

  • 37 Similar variants appear in A 3.2, B 3.2; A 6.78, B 5.600, C 7.238; B 5.89; B 13.140; B 13.289; B 20.124; C 6.116.Back

  • 38 Langland uses "quod he/she/I" in this position in ways uncharacteristic of other alliterative poets. My sense, based on uncompleted research, is that such fillers are extra-metrical and do not affect the argument here.Back

  • 3 The literature on generative metrics is extensive. A good place to begin is Gilbert Youman's introductory essay "Introduction: Rhythm and Meter," in Phonetics and Phonology, Volume 1: Rhythm and Meter, ed. Paul Kiparsky and Gilbert Youmans (San Diego: Harcourt Brace Jovanovich, 1989), pp. 1-14. Back

  • 40 For a similar view of philology in the present academic environment, see Siegfried Wenzel, "Reflections on (New) Philology," Speculum 65 (1990): 11-18. The entire issue, entitled The New Philology, was edited by Stephen G. Nichols. See also Kevin Brownlee, Marina Brownlee, and Stephen G. Nichols, eds. The New Medievalism (Baltimore and London: Johns Hopkins University Press, 1991), and Keith Busby, ed. Towards a Synthesis? Essays on the New Philology (Amsterdam and Atlanta, GA: Rodopi, 1993).Back

    IATH WWW Server
    Last Modified: Wednesday, 11 Jan 95 13:02:22 EDT