I TAKE MY TITLE from The Book of Margery Kempe , that autobiographical account of travels and visions that survives in a single copy from the middle of the fifteenth century. In an early passage the Book presents Kempe as having been urged by ‘worthy and worshipful clerkes’ to ‘don hem wryten and makyn a book of hyr felyngys and hir revalacyons’ (I. 76-80). 1 This language (‘don hem wryten’) suggests that Kempe will have the book written for her, but she then says that ‘sum proferyd hir to wrytyn hyr feelings wyth her owen handys’ (I. 80-1). ‘Her’ in this formulation is ambiguous since in Kempe’s English ‘her’ could well mean ‘their’, as in ‘some offered to write this book with their own hands’; but Kempe might well have meant ‘her’ because she then goes on to say that she refused these offers because ‘sche was comandyd in hir sowle that sche schuld not wrytyn so soone’ (I. 82-3). It is odd that Kempe should imagine herself writing at all since she then goes on to explain how the Book was written, first, by a man ‘dwellyng in Dewchlond’ and then, again, by a ‘prest’ when it turned out that the Dutchman wrote English so poorly (I. 89-90, 97). The Book says that Kempe was ‘not lettryd’ (I. 4290), but there are indications that ‘unlettered’ does not mean wholly illiterate in this case. As a number of commentators have noticed, the Book describes any number of instances in which Kempe seems able to read, as, for example when she is hit on the head by masonry in a church while she has ‘hir boke in hir hand’ (I. 659), 2 or when Christ says that he will be pleased with her ‘whethyr thu redist or herist redyng’ (I. 7342).

One way around these contradictions is to presume that Kempe was much more competent than she admits. Lynn Staley has argued that Kempe’s account of her scribe was itself a ‘trope’, a necessary attribute of the ‘conventions of spiritual writings by or about women’ that the Book absorbs so fully. 3 Moreover, when Kempe says that she is ‘not lettryd’ she may only mean that she was illiteratus or unable to read Latin, and so she still might have been able to understand spoken Latin phrases just as she almost certainly could read the English written for her. 4 Or she may have thought she was ‘not lettryd’ because she was unable to form letters on parchment or paper. Reading and writing were separable skills in the Middle Ages because wielding a quill or reed required specialised training. 5 Not all men who could read and understand Latin found it easy to put words on a page. Perhaps, then, Kempe understood writing in her own hands to be writing in their own hands (a task that, literate though she was, she needed someone to do for her).

This last possibility is given support from slightly later in the fifteenth century in the extraordinary store of letters that survive from the Paston family. A letter of (about) 1481 from Margery Paston to her husband, John Paston III, for example, is written in a very neat hand, but it is also very clearly the hand of someone other than Margery, a secretary who has been identified as Richard Calle, the family bailiff. 6 Only the letter’s closing, ‘Be yowre servaunt (and bedewoman)’ and the letter’s signature, ‘Margery Paston’, seem to have been written by Margery in a ‘distinctively halting and uncontrolled hand, as of someone beginning to learn to write’. 7 Margery’s literacy was clearly marginal, and all of the other letters from her in the collection seem to have been written by male secretaries, as indeed are all the other letters in the collection ‘written’ by the other Paston women. 8 But the use of secretaries was hardly confined to the women of the family. A letter from John Paston I to his wife Margaret in 1460, for example, is also in the hand of Richard Calle. 9 Once again, only the signature – in what Norman Davis describes as a ‘coarse ill-formed hand’ – belongs to the person we might call the writer of this letter. 10 But John Paston was also very clearly ‘lettered’ not only because he could clearly read but because, when he wanted to, he could also clearly write. A lengthy letter from him written in 1465 is not in the hand of any secretary but, rather, in John’s own coarse but clearly competent script. The Paston letters show, in other words, that so far from it being a particularly female mode of literacy in the fifteenth century, men also often dictated what they wanted written to secretaries, and they clearly did so even when they could have written the document themselves. Such practices make perfectly clear how Kempe could have understood writing in the hand of another to be writing ‘with her own hands’. They also make clear just how extensive the role of secretaries in this period might have been.

I have suggested in my title that Kempe’s Book might teach us something about Chaucer and Langland not least because very little has interested scholars of Middle English more than the way that their poems were first written down, and particularly how those first texts relate, or can be related, to the extant copies of the texts from which we now edit their work. In that context Margery Kempe’s mode of secretarial production must seem highly unusual, since, when textual scholars have imagined the reproduction of Chaucer’s and Langland’s writings, they have always imagined copying , a process whereby a scribe turned ‘foul papers or wax tablets on which the poet wrote first drafts into the first fair copies of his work’. 11 Much about such copying is well documented: we know it tended to involve ‘independent craftsmen … employed on a particular commission’, using ‘exemplars’ or copy texts that they exchanged with one another, often piecemeal. 12 And most of the textual complexities in extant manuscripts have seemed well explained by the complexities necessarily involved in transforming an author’s drafts into a polished (and saleable) text, and then from the errors of transcription that arise, and are then layered, when a text is copied repeatedly. The B-text of Piers Plowman offers a useful example of such explanations here because it has proved particularly challenging to explain how the extant texts of the poem actually derive from its ‘archetype’ (the single text from which, it is always assumed, all these extant texts must have been copied). The foundational problem is usually thought to be that this archetype was so messy, a ‘scribal fair copy (already replete with small errors)’ which ‘was then subjected to extensive, direct authorial revision’, with the result that ‘it had blocks of recently composed or revised texts attached at various points as extra leaves or scraps of membrane’. 13 To explain the relationships between the fifteen extant copies of the B-text of Piers Plowman on this basis it has also been necessary to hypothesise the existence of two manuscripts that we lack, ‘alpha’ and ‘beta’ (sub-archetypes or direct copies of the B-archetype) even though the relationship between these lost texts has itself proved very difficult to determine or understand. 14 So reliant is this set of explanations on the presumption that texts can only relate to one another by the process of copying, however, that it rests on the presumption that Langland only had access to corrupt, error-filled copies of his own text. This is inherent in the view that B was revised from a ‘scribal fair copy’ (rather than a text in Langland’s own hand) and it was foundational to the view, advanced by George Kane and Talbot Donaldson, that when Langland made the extensive revisions to B that produced the C-text he still only had a ‘corrupt scribal copy’ of B available to him. 15 In what must be a highly surprising moment in the light of such views, Kane also followed H. J. Chaytor in the introduction to his edition of the A-text of Piers Plowman in entertaining the possibility that ‘a more likely source of variants’ of the kind found in Piers Plowman manuscripts might have been ‘oral transmission’ not errors introduced by copying. 16 Kane, predictably enough, rejected the possibility because ‘the character of the variants’ does not support it; 17 yet, as I will suggest, it is exactly the kinds of variation that led to the hypotheses I have just traced that actually constitute some of the best evidence we have of transmission of this kind in Middle English.

Before turning to that evidence it is probably useful to establish that the mode of literacy employed by Kempe and the men and women of the Paston family was not a fifteenth century development. Oral transmission and the use of secretaries has a very long history in the production of texts in the West. Cicero says, for example, in one of his letters to Atticus, that he dictated it ( dictavi ) syllable by syllable ( syllabatim ) to a secretary, whom he names. 18 The Codex Sinaiticus, an important fourth century copy of the Bible in Greek, seems to have been produced by dictation since it displays a series of errors best explained as a consequence of mishearing. 19 Possidius says, in his biography of Augustine of Hippo, that notarii or secretaries were employed to record Augustine’s sermons as he preached. 20 Michael Clanchy provides striking visual evidence of the use of secretaries in the early thirteenth century from a copy of Peter of Poitiers’ Compendium or Tabula Historiae . 21 This roll offers a visual primer to biblical history, diagramming its progress as well as briefly narrating its key events, and images right at its beginning also offer an account of the roll’s own production: a roundel on the left side shows a figure who must be Peter of Poitiers with a codex in front of him (its written contents are indicated by vertical lines), gesturing towards a figure in a roundel at the opposite side of the roll who has his head turned towards Peter, the better to hear what he is saying. This second figure is dipping his pen in ink with his right hand while, to his left, there is a roll of vellum (exactly like the one on which the Compendium has been written) on which he is clearly preparing to write. Sermons and academic lectures were so frequently taken down by secretaries in the Middle Ages that Malcolm Parkes has traced the development of systems of ‘tachygraphy’ (rapid writing) in which the formation of letters, simplification of words, and abbreviation of content made it easier to take down oral performances as what were then called reportationes . 22 Parkes also describes a number of manuscripts ‘from at least the late eleventh century onwards’ which contain written versions of lectures delivered in the Oxford schools. 23 Much in the manner of Margery Kempe, Anselm of Canterbury and Bernard of Clairvaux corrected the reportationes that secretaries made of the sermons they delivered. 24 It is no accident that both the Pastons and Cicero dictated letters because, historically, secretaries have been used with particular frequency in this genre, and the prevalence of secretarial work in the Middle Ages, particularly in later medieval England, owes a great deal to the wide remit of the letter itself: as Malcolm Richardson has pointed out, all the writs, wills, petitions, and charters produced so copiously in medieval courts and households were, in form and by genre, letters. 25 Scriveners arose as a significant class in the thirteenth century in England, and formed themselves into a craft towards the end of the fourteenth century, to produce the flood of documents England’s bureaucratic culture required, and it seems right to presume that many of these letters were produced, like the Paston letters, by one literate person’s oral dictation recorded by another. 26 Among the exigencies such a mode of production addressed was the need for speed, since several written copies can be made from one act of dictation. It has been argued, for example, that many copies of the Divine Comedy show evidence of production in this way because its instant popularity required that a large number of copies be made quickly. 27 Nor did secretarial production of this kind die out after the Middle Ages. A manuscript of Nicholas Harpsfield’s Life and Death of Sir Thomas Moore seems to display numerous substitutions derived from mishearing rather than miscopying. 28 Stenography was developed in the late sixteenth century to ensure that accurate versions of sermons could be taken down in full at the speed of their delivery. 29 In the twentieth century, Henry James developed a rheumatic cramp in his right wrist while writing What Maisie Knew and slowly moved to producing all of his novels by a process he characterised as ‘fierce legibility’. 30 Mary Weld, who was employed as one of his secretaries, reported in her diary that she spent 194 days writing out The Wings of the Dove to James’s dictation. 31 The age in which the figure of the secretary is a person may have ended in the twenty-first century, but dictation may be said to have a growing role for those willing to take the risk of producing documents with the aid of Apple’s ‘Speech Interpretation and Recognition Interface’ or what an iPhone calls ‘Siri’.

As the roundels in Peter of Poitiers’ Historia make clear, dictation did not necessitate memorisation, but some of the difficulty we now have in imagining scenarios in which secretaries worked, or oral transmission occurred, for poem’s as substantial as Piers Plowman may derive from our own memorial limitations. In a culture where an iPhone can not only take dictation but distribute our cognition (where we can effectively ‘remember’ the whole of any text for which we can type four consecutive words into Google) we may find it impossible to imagine that Margery Kempe held the whole of her Book in her head, although this is exactly what the Book ’s account of its own creation implies. And whoever designed the frontispiece of one of the better manuscripts of Chaucer’s Troilus and Criseyde , Cambridge, Corpus Christi College MS 61, had no trouble imagining that Chaucer had the whole of Troilus in his head. This famous image depicts a figure who is clearly Chaucer before an audience who seem to be listening to him perform the poem this book contains. 32 The image is usually described in modern reproductions as ‘The Poet Reading to an Audience’, 33 but as Derek Pearsall has observed, ‘the book from which “the Poet” is supposed to be “reading” is not at all obvious in the picture’ because ‘it is, in fact, not there and has only been supplied by modern interpretation’. 34 Pearsall aligns the representation of Chaucer in the frontispiece with a tradition of portraiture depicting the delivery of sermons, and he illustrates this iconography with a portrait of Petrus de Aureolis, archbishop of Aix, shown preaching to an audience in a fourteenth century copy of his Compendium super Bibliam (London, British Library, MS Royal 20.C.vii) and two portraits of the archbishop of Rouen ‘preaching a crusade’ in two fourteenth century manuscripts of the Chroniques de France (London, British Library, MS Royal 20.vii and MS Royal 20.C.vii). 35 Portraits of authors reading to an audience in the manner of the Corpus Christi frontispiece are, according to Pearsall, a later development. 36 But one could also say that the pulpit and the iconography of preaching in this frontispiece do not so much associate Chaucer with a different model of reading as a different model of writing , in which an author like Chaucer (or Margery Kempe) could hold the whole of what they had composed in mind so readily that it was not only easy (and common) to perform that text for an audience but just as easy to dictate it to some scribe who would then be recording rather than copying it from the author’s original, producing fair copies not from foul papers, but as so many reportationes .

Such ‘writing’ would have been enabled by memories that were more generally trained, and practised, in remembering very long texts. We might well believe that Chaucer could hold the whole of Troilus and Criseyde in his head if we can also believe Augustine’s claim that his school friend, Simplicius, could not only recite all of Virgil from memory but could recite it backwards. 37 Thomas Aquinas seems to have been praised less for the books he wrote than for the prodigious memory that was thought to have made the writing of those books possible, 38 and he seems to have composed the Summa Theologiae mentally ‘calling in a secretary when he was finished to take down his dictation in a fair hand’. 39 Mary Carruthers has argued that ‘the terms “oral” and “written” are inadequate categories’ for describing what she prefers to call ‘memorative composition’, a practice not just possible but common in both pedagogy and the practices of individual writers throughout the Middle Ages. 40 Such composition was understood and identified technically as cogitatio in rhetorical traditions as far back as the first century. In the Institutio Oratoria , for example, Quintilian said that the orator’s training should advance from a procedure where he revises with stylus and tablets to cogitatio proper where he revises mentally, so that what ‘he conceives by cogitation’ ( quae cogitarit ) will be just as useful to him ‘as what he has written and learned by heart’ ( quam quae scripserit atque edidicerit ). 41 In the fourth century, the rhetor Julius Victor distinguished the kind of composition in which ‘we write’ ( scribimus ) from the composition in which ‘we think’ ( cogitamus ). 42

To be sure, such methods and definitions of writing, like much of the long tradition of secretarial work I have traced above, lie at a considerable remove from late fourteenth century England. My other work on literacy in this period suggests that these are precisely the years when transformative change occurred in England in a matter of decades (in 1390 most tuition in grammar schools was probably in Latin, but by 1410 that same tuition seems to have occurred in English). 43 Moreover, there is a different tradition of Chaucer portraiture which insists that Chaucer was the kind of writer who wielded a pen. The most famous instance of this iconography is the portrait in the left-hand margin of folio 183 of San Marino, Huntington Library, Ellesmere 26 C 9, in which Chaucer is depicted on horseback pointing to the beginning of the Tale of Melibee. The poem this portrait accompanies is of course haunted by the spectre of oral transmission (since it is part of the general fiction of the Canterbury Tales that each tale is recited rather than read out) as well as dictation (since the narrator not only represents himself as a fellow pilgrim but as their secretary, faithfully recording all these oral performances). So it also makes sense that the programme of illustrations in this manuscript, which tend to depict each pilgrim with instruments or clothing that give some indication of his or her profession, identify Chaucer’s profession by means of a penner, the case for a writing instrument that is shown hanging around his neck. This detail is repeated through a long tradition of Chaucer portraiture which seems to have been based on an original by an artist who knew what Chaucer looked like. 44 It is also possible to see Chaucer at work with his pen in the lyric that survives in Cambridge, Trinity College MS R.3.20, usually understood to be by Chaucer, not least because it is called there ‘Chaucer’s Wordes Unto Adam, His Owne Scriveyn’:

Adam scriveyn, if ever it thee bifalle

Boece or Troylus for to wryten newe,

Under thy long lokkes thou most have the scalle,

But after my makyng thow wryte more trewe;

So ofte adaye I mot thy werk renewe,

It to correcte and eke to rubbe and scrape,

And al is thorugh thy negligence and rape. 45

The poem defines ‘for to wryten’ as the process whereby Adam leaves the traces on a page that Chaucer must then rub away and then ‘correcte’, presumably by writing in his own hand. And yet, as Mary Carruthers has noticed, this lyric can also be seen to make a careful distinction between cogitatio and scriptio , between writing that occurs in the first instance mentally, or what Chaucer calls ‘my makyng’, and then physically as Adam writes it, ‘after’ – that is, as a faithful reproduction of – that prior ‘makyng’. 46 In this reading, this lyric understands ‘Adam scriveyn’ as a secretary who writes to Chaucer’s dictation, and Chaucer’s corrections as similar to those Kempe has to make when her Book is so ill written the first time. The lyric is then much more about Kempe’s mode of literacy than the more standard models of copying, a literacy distributed between two persons (one who composes and the other who makes the marks on the page), two distinct practices that only become a single act of ‘writing’ insofar as they are joined together by the processes of dictation.

If Adam scriveyn was a secretary and such writers were in use in the production of literary manuscripts in the fourteenth and fifteenth centuries, it is odd that we hear so little about them. But, as Malcolm Richardson has also put it ‘the scriveners of London are a shadowy bunch’ and while it seems clear from their number and the sorts of documents that survive in their hands that these writers ‘fulfilled many of the functions of notaries’ the details of their professional lives ‘seem just beyond the grasp of history’. 47 Professional writers, separate from the clerisy, were certainly growing in number from the thirteenth century onward in England, and London in particular. By 1357 they had assembled themselves as a craft that was eventually called the Scriveners Company, but, at its inception and in Chaucer’s day, was called the Craft Writers of Court Letter. 48 Such incorporation was itself a measure of a more widespread process since, like so many craftsmen in the period, the scriveners banded together to protect themselves from the many non-citizens of London who were stealing their custom by taking up their trade. As ever too, it is a craft’s prohibitions that give us the richest portrait of its activities, and so it is a set of accusations against scriveners who work on Sundays that shows us best what they probably did every day:

That he keeps his stall without a workshop open on Sundays and important feasts [literally festis duplicibus ‘double feasts’] and writes and exercises his craft openly on the aforesaid days in the sight of the people, and hangs outside many documents and various writings to the great disgrace of all the honest men of his craft and as a wicked example to other workmen, and in clear contempt of dear mother church and against the command of the lord bishop. 49

This portrait does not say that these scriveners wrote to the dictation of passers-by but it evokes exactly the scene that remains common enough in Mumbai or Istanbul, where writers even today can be seen sitting ready with pen and paper or a typewriter to produce documents to order, where what a ‘writer’ is is someone prepared to hire out his literacy to the illiterate and where dictation is the only method available to the illiterate person to produce a text. If such labours were the stock in trade of the members of the Scriveners Company it may be relevant too that the Adam of Chaucer’s lyric has been identified with a member of that company, the Adam Pinkhurst thought by some to have written the earliest manuscripts of the Canterbury Tales as well as at least one copy (which survives in fragment) of Troilus and Criseyde . 50

It is inherent to the mode of literacy I have been trying to describe that it occupies a borderland just beyond the precincts of surviving evidence: it is, of course, only the writing that is the product of dictation that could survive. But the problem is also in part conceptual. What I have so far treated as a category of literacy could also be understood as a problem for what has recently been called the ‘history of mediation’, 51 and one of the complexities that attends any such history, as John Guillory has pointed out, is that the concepts required to identify or describe a medium may be ‘absent but wanted’ for several centuries after a medium’s initial appearance. 52 Langland may not mention what Chaucer only implies or the Book of Margery Kempe describes in contradictory ways because none of these writers makes a distinction between secretarial work and their own ‘making’, between writing made by their own or another’s hands. Kempe’s Book may give us a little more guidance here, however, while also helping me scale the sort of claim I want to make about the role of dictation in the production of late medieval texts. For in its detailed account of its own composition the Book also details the method of Kempe’s corrections: as her friend the priest reads it out to her she follows along ‘helpyng where ony difficulte was’, collating the written form the Book now has with the ‘book’ she has in her mind or memory as she dictates (I. 133). Kempe makes clear, in other words, that a writer need not have remembered her (or his) whole text in order to dictate a new one, but, also , that the process of dictation using a textual prompt leaves ample room for changing that text in the process of transmission, repeating but also revising or re making, introducing errors to be sure, but doubtless also improving, augmenting, elaborating the text being used as a prompt as it is being retransmitted orally to be written anew. Kempe’s practice here seems to me not just a middle way between one version of the Book and another, but between the model of foul papers made fair by copying and the proposition that Chaucer held the whole of Troilus in his head. It also puts before us a more concrete way of tracking the role of a secretary since he would sit, not behind all written versions, but between two of them. As I have said, George Kane rejected the role that oral transmission of this kind might have played in the production of the A-text of Piers Plowman , but in order to reject it, Kane also detailed the characteristics he would expect such transmission to produce. What we do not have in the textual history of the A-text, he says, is ‘compression at one point and expansion at another, dislocation of matter other than palaeographic, large omissions evidently made for abridgment or from defect of memory, and marked unevenness in the accuracy of reproduction’. 53 What is most perplexing about the list of course is that, even if it does not describe the ‘character of the variants’ in the textual history of the A-text, it most certainly describes the textual tradition of Piers Plowman as a whole where it is precisely instances of the first three of Kane’s criteria – compression and expansion, dislocation of material, and large omissions – that characterise the relationships between the A-, B-, and C-texts. The list also calls attention to the similar sorts of relationships that can be found in the textual traditions of a number of other Middle English texts, including Chaucer’s Troilus and Criseyde . So I would like to turn now to an exploration of the possibilities of oral transmission, and the evidence of secretarial work, in such examples. To my mind the revelatory traces are exactly those Kane identifies, but the signature of the practice is nothing more than a text that survives in versions , a set of witnesses that are each authentic but are equally impossible to relate to one another because it is the oral medium rather than a sequence of errors in copying that stands between them. In such cases, where we have come to assume that some intermediary or original version has been lost, we might now imagine that it never existed.

If we believe that Chaucer could have known Troilus and Criseyde well enough to recite the whole of it without a text in front of him, we might also imagine that he left out a passage or two as he recited it, or moved around stanzas, or added new material, revising the text as he dictated it. Three key omissions from some of the most important copies of the poem have exactly the shape this sort of rolling revision would have produced. The most important of these copies is San Marino, Huntington Library MS HM 114, an otherwise authoritative manuscript that lacks important passages that most other authoritative manuscripts include. The gaps are: (1) the song Troilus sings about love in Book III (1744-71); (2) Troilus’s meditation on predestination in Book IV (953-1085); and (3) the description of Troilus’s ascent to the spheres in Book V (1807-27). 54 R. K. Root argued in his edition that these passages were evidence that Chaucer had put an early version of the poem into circulation. 55 Barry Windeatt has argued very persuasively that the poem requires all three of these passages, although, to be sure, the gaps are not problematic at the level of the stanza (since it is possible to read along in a version of the poem in which these stanzas are lacking without stumbling). Moreover, all three of these passages have what might be described as a compositional relationship since all of them depart from Chaucer’s primary source for Troilus and Criseyde , Boccaccio’s Filostrato . For Troilus’s song Chaucer turns to Book II, metrum 8 of Boethius’s Consolation of Philosophy . For Troilus’s meditation on predestination in Book IV Chaucer turns to Book V, prose 3 of the Consolation . For the description of Troilus’s ascent to the spheres at the end of Book V, Chaucer translates Book XI.1-3 of Boccaccio’s Teseida . In Windeatt’s view, the omission of these three passages from MS HM 114 is the result of errors in transcription: they ‘perhaps existed originally in the form of a physical addition to the draft which has been confused by certain scribes’, 56 or ‘were perhaps available to them separately’ because they ‘did not form part of the author’s main source’. 57 But, if the scribe of this book was confused, he also had an infallible bias against all those passages in which Chaucer had changed his compositional technique, as if he had some way not only of reading the text but of understanding which texts were behind it. Another way to put the point is that the scribe of the Huntington manuscript had the preternatural ability to make distinctions that were, at root, Chaucer’s distinctions, invisible in that sense on the surface of any copy, but absolutely decisive for the making that had produced the original text. Imagining that Chaucer was responsible for these gaps does not require us to imagine Chaucer dictating the whole of Troilus from memory but, rather, producing it exactly as Kempe produced the final copy of her book, working from a text of something like MS HM 114 which lacks these passages, dictating a copy like that of Corpus Christi MS 61, which has them, adding these passages as he dictates, perhaps even using the passages from Boethius and the Teseida as prompts or reminders as he does so. This model may resurrect the long-rejected proposition that Chaucer released an early version of Troilus into circulation, but there are verifiable instances of Chaucer releasing early versions of his poems, 58 and it is not necessary to embrace the whole of Root’s theory, which annexed many more differences between the manuscripts to its claims, to accept that these three omissions conform better to the discrepancies produced by dictation than to the kinds of error that are usually produced by copying.

A less dramatic (but to my mind even more compelling) set of examples can be found in the manuscripts of the B-texts of Piers Plowman and those hypothesised sub-archetypes ‘alpha’ and ‘beta’, ‘the lost ancestors of the “two main families” of the surviving B manuscripts’ that I have already mentioned. 59 Alpha and beta are stages that accounts of the textual tradition of the B-text have required because there are eighty-one passages in which these two traditions significantly disagree, with alpha containing some passages beta lacks and beta containing some passages that alpha lacks. In their edition of the B-text, Kane and Donaldson explain every one of these differences as matters of eye-skip, but as John Burrow and Thorlac Turville-Petre have pointed out, case by case, these differences are often thoughtful. Let me take just one example from Passus 8, four lines that the beta tradition has but the alpha tradition lacks:

For thei be men of this moolde that moost wide walken,
And knowen countrees and courts and many kynnes places –
Bothe princes paleises and povre mennes cotes,
And Dowel and Do-yvele, wher thei dwelle bothe. 60

This is not a crucial addition but it is generally agreed to be wholly Langlandian. Since Will has just met two friars whom he is about to ask for directions to Dowell, the passage explains his hope – which is a hope Langland generally has – that friars, as a profession, will be of some help in this search. These lines are missing from the A-text of Piers Plowman but they are in the C-text so the presumption must be that beta represents an intermediary revision of B on its way to C. There are, however, passages omitted from the beta tradition that can be found in alpha and these are also understood by Burrow and Turville-Petre (and most editors), as authentic revisions of the B-text. How could two versions of the same text both contain and omit so much authorial material? Of the many models that have been proposed to explain these discrepancies, Burrow and Turville-Petre’s is brilliantly simple but it also relies on the scribal mechanics typical of Middle English textual scholarship. 61 They imagine the B-archetype in the usual way as a ‘scribal fair copy that Langland used as his working text’ with ‘additions on loose leaves, in the margins and interlinearly’, and they suggest that the alpha and beta readings are authentic but discrepant because the scribes who produced these texts made ‘heavy weather of the disorder that faced them’ not least because they ‘were copying without Langland’s supervision and perhaps without his authority’. 62 And yet this heavy weather consists of exactly the compression and expansion Kane said we should expect of oral transmission, whereas the scribal model of copying must presume that the scribes of both alpha and beta were generally producing the text in front of them faithfully while also nodding with great frequency, and also always nodding at different times. The distinctions between these hypothesised cases are again, moreover, not so much scribal as compositional, with alpha and beta looking for all the world like different Langlandian performances of Piers Plowman , and these compressions and expansions are exactly what we would expect should a poet have made minor revisions as he dictated one version of his poem while looking at another, or simply forgot some lines, or added a few, when dictating from memory.

I could conclude here with a set of examples from a third Ricardian poet, since Gower’s Confessio Amantis also survives in two distinct versions (with variant prologues and conclusions) as well as what appears to be an intermediate version in which passages are inserted into Books 5 and 7 and passages are rearranged in Book 6. 63 I mention this last possibility not to explore it, however, but to make clear that the kinds of compression, expansion, and, in this case, dislocation that imply secretarial work are routine rather than exceptional in fourteenth century English literature. But I would like to turn, for my last example, to a slightly different sort of evidence, which is more minor in every single case, but so pervasive that it can be found in every poem Chaucer wrote, in every manuscript copy. The issue brings me fittingly and helpfully too to the work of F. W. Bateson, who published an article in Essays in Criticism in 1975 whose title asked the question, ‘Could Chaucer Spell?’ This is to frame humorously a problem that has vexed every editor of Chaucer not so much because the spelling of so many words differs from manuscript to manuscript but because the final - e in so many words is so variable and so many of these variations confound Chaucer’s metre: that is, final - e often appears in manuscripts at the end of words where the metre suggests we should not pronounce it, but it also disappears in some manuscripts where the metre absolutely requires it. The standard theory is that these final - e ’s (most of which were remnants of the more elaborate series of declensions and conjugations in Old English) were fragile enough by the end of the fourteenth century for Chaucer to have exploited the ambiguity, sometimes requiring them to be pronounced, sometimes assuming they would not be, and that Chaucer’s scribes, many of them writing at a later date, either misunderstood Chaucer’s intentions (because they now pronounced the words differently) or were careless about these small points. The issue may be illustrated with three examples from the manuscripts of the Book of the Duchess . In line 236, all four of the extant witnesses of the poem have ‘Men to slepe ne for to wake’ which has an extra syllable in ‘slepe’ (or two dips between ‘slepe’ and ‘for’), if we assume (as the bulk of the readings in the witnesses help us to) that Chaucer wrote lines in which stressed syllables regularly alternated with unstressed syllables. 64 Editors of Chaucer have generally refused to emend for metre so this line is usually printed in the rough form it has in all the witnesses, but it is easy to imagine that what Chaucer wrote was a line that omitted the - e at the end of ‘slepe’ and therefore was perfectly regular: ‘Men to slep ne for to wake’. 65 Similarly, lines 644 and 645 have two problems in them in almost all of the witnesses, a dip lacking at the end of ‘fals’, but an extra dip at the end of ‘aboute’: 66

So turneth she her fals whele
Aboute for it is nothing stable.

That is a very awkward rhythm, but the lines scan perfectly if we assume that there was a final - e on ‘fals’ and no final - e on ‘aboute’:

So turneth she her false whele
About for it is nothing stable.

The presumption that Chaucer’s - e ’s were somehow progressively lost or confused in the copying of his texts easily explains why ‘slep’ and ‘about’ acquired an - e in these lines but ‘false’ lost one, but Bateson was led to ask his question by the larger problem: why, if these - e ’s matter so fundamentally to metre, do the scribes get them wrong so often? Why is Chaucer’s poetry as it survives in all its witnesses so frequently spelled in a way that does not scan? Bateson answers his question by insisting that we should not rely on ‘the syllables [that] can be counted on the page’ because ‘the primary function’ of those syllables in Chaucer’s poetry ‘was to be heard’. 67 To put this claim in the form I have been advancing it, Chaucer could doubtless spell in a way that would give each of his lines regular metre, but his scribes could not. If we insist on the standard model of the foul papers for the production of extant copies of Chaucer’s texts that answer must lead to another question: if Chaucer’s scribe (say Adam scriveyn) was looking at Chaucer’s foul papers – at his good spelling – why did he so regularly fail to get the - e ’s right? The answer here, I think, is not just that Chaucer’s syllables were meant to be heard but that they were heard. Adam, like the other scribes who wrote the first copies of Chaucer’s poems, tended to get the final - e ’s wrong so often because they never saw them. They spelled ‘slepe’ and ‘fals’ and ‘aboute’ however they pleased because they were writing down what Chaucer said, not what he wrote, using whatever spelling they were accustomed to use for each word, in no immediate contact with Chaucer’s own regular final - e ’s. To put that theory in its broadest form: the final - e ’s in all the manuscripts of Chaucer’s poems look exactly as we would expect them to if Chaucer provided his scribes not with foul papers, but with a performance like the one depicted in the Corpus Christi MS 61 frontispiece (or in the Canterbury Tales for that matter) where the task at hand was not to copy but to record, producing not copies but reportationes .

In describing the implications of literacy, Brian Stock says that ‘there is in fact no clear point of transition from a nonliterate to a literate society’. 68 What I am really suggesting here, then, is that there is no clear point of transition in literacy between the writing that occurs in the mind and the writing that occurs on a surface. Secretaries clearly existed in all walks and precincts of textual production from Rome until as recently as yesterday, and they certainly existed in fourteenth century England in abundance. If we lack firm evidence of secretarial production for the poems of Chaucer and Langland – or for the Book of Margery Kempe for that matter – neither, in most cases, do we have the foul papers that have so far seemed to explain so much to textual scholarship. The nature of women’s literacy in the Middle Ages is notoriously difficult to specify or even determine since every word a woman composed might well have been mediated through the agency of some man or men. So, all details aside, my argument here might suggest most of all that it is also the nature of men’s literacy in this period that may be harder to specify than we have thought because all of the writing we have from the period may have been just as heavily mediated through the agency of secretaries. The final question, then, is not only Bateson’s (‘could Chaucer spell?’) but the question that incisive query should lead us to: how – that is, in what manner – did Chaucer and Langland ‘write’ when they wrote in their own hands?

Notes

I would like to thank the electors for the Bateson Lecture, Richard Carwardine, president of Corpus Christi College, Seamus Perry, Hermione Lee, Colin Burrow, Helen Small, and Val Cunningham for their kind invitation. I am also grateful to Richard Beadle, Vincent Gillespie, Jill Mann, Derek Pearsall, and Daniel Wakelin for bibliographical guidance and advice while researching this topic.
1 References in the text are to The Book of Margery Kempe , ed. Barry Windeatt (Harlow, 2000).
2 See also Book , p. 9, and Karma Lochrie, Margery Kempe and Translations of the Flesh (Philadelphia, 1991), pp. 101-4 and 114-27.
3 Lynn Staley, Margery Kempe’s Dissenting Fictions (University Park, Pa., 1994), p. 33.
4 On the term illiteratus see M. T. Clanchy, From Memory to Written Record: England, 1066-1307 , 2nd edn. (Oxford, 1993), pp. 229-30 and 266.
5 See ibid., pp. 47-8.
6Paston Letters and Papers of the Fifteenth Century , 3 vols., Early English Text Society ss 20-22 (Oxford, 2004-5), vols. i and ii ed. Norman Davis; vol. iii ed. Richard Beadle and Colin Richmond. See vol. i, p. lxxvi (on Calle), pp. 664-5 (for the letter).
7 Ibid., vol. i, p. xxxvii.
8 Ibid., pp. xxv-lxxix.
9 Ibid., pp. 92-3.
10 Ibid., p. xxxvi.
11 Linne R. Mooney, ‘Chaucer’s Scribe’, Speculum , 81 (2006), 121.
12 A. I. Doyle and M. B. Parkes, ‘The Production of Copies of the Canterbury Tales and the Confessio Amantis in the Early Fifteenth Century’, in M. B. Parkes and Andrew G. Watson (eds.), Medieval Scribes, Manuscripts and Libraries: Essays Presented to N. R. Ker (1978), pp. 163-210: 167. See also Daniel Wakelin, Scribal Correction and Literary Craft: English Manuscripts, 1375-1510 (Cambridge, 2014), pp. 87-94.
13 Robert Adams, ‘The R/F MSS of Piers Plowman and the Pattern of Alpha/Beta Complementary Omissions: Implications for Critical Editing’, TEXT , 14 (2002), 117.
14 Robert Adams, ‘Evidence for the Stemma of the “Piers Plowman” B Manuscripts’, Studies in Bibliography , 53 (2000), 173-94.
15 George Kane and E. Talbot Donaldson, introduction to Piers Plowman: The B Version , rev. edn. (1988), pp. 1-220: 98.
16 George Kane, introduction to Piers Plowman: The A Version (1960), pp. 1-172: 144. See also H. J. Chaytor, From Script to Print: An Introduction to Medieval Literature (Cambridge, 1945), p. 129.
17 Kane, introduction to Piers Plowman: The A Version , p. 144.
18 Cicero, Epistolae ad Atticum , ed. D. R. Shackleton Bailey (Teubner, 1987), 13.25, cited in Paul Saenger, Space Between Words: The Origins of Silent Reading (Stanford, 1997), p. 48.
19 T. C. Skeat, ‘The Use of Dictation in Ancient Book-Production’, Proceedings of the British Academy , 42 (1956), 193.
20 M. B. Parkes, ‘Tachygraphy in the Middle Ages: Writing Techniques Employed for Reportationes of Lectures and Sermons’, in id., Scribes, Scripts and Readers: Studies in the Communication, Presentation and Dissemination of Medieval Texts (1991), pp. 19-33: 19.
21 Cambridge, Mass., Harvard University, Houghton Library fMS Typ 216, helpfully reproduced in Clanchy, From Memory to Written Record , as plate XII (pp. 468-9).
22 Parkes, ‘Tachygraphy in the Middle Ages’, p. 32.
23 Ibid., p. 20.
24 Mary Carruthers, The Book of Memory: A Study of Memory in Medieval Culture , 2nd edn. (2008), pp. 242-3.
25 Malcolm Richardson, Middle-Class Writing in Late Medieval London (2011), p. 58.
26 On the increasing need for scriveners see Nigel Ramsay, ‘Forgery and the Rise of the London Scrivener’s Company’, in Robin Myers and Michael Harris (eds.), Fakes and Frauds: Varieties of Deception in Print & Manuscript (Detroit, 1989), pp. 99-108: 102-4.
27 Skeat, ‘The Use of Dictation’, p. 203.
28 Ibid., p. 206.
29 Parkes, ‘Tachygraphy in the Middle Ages’, p. 33.
30 Leon Edel, Henry James: The Treacherous Years, 1895-1901 (Philadelphia, 1969), p. 175.
31 Leah Price and Pamela Thurschwell, ‘Introduction: Invisible Hands’, in Price and Thurschwell (eds.), Literary Secretaries/Secretarial Culture (Aldershot, 2005), pp. 1-12: 1.
32 For a clear reproduction see Derek Pearsall, ‘The Troilus Frontispiece and Chaucer’s Audience’, Yearbook of English Studies , 7 (1977), plate I.
33 This is the caption given the image as the frontispiece to both volumes of P. M. Kean, Chaucer and the Making of English Poetry (1972). See also vol. i, p. 25 for a reference to the ‘reading’ depicted in this image.
34 Pearsall, ‘The Troilus Frontispiece’, p. 70.
35 See ibid., plates II and III and p. 71 n. 3.
36 Ibid., p. 71 n. 2.
37 Carruthers, The Book of Memory , p. 21.
38 Ibid., p. 14.
39 Ibid., p. 251.
40 Ibid., pp. 240-1.
41 Quintilian, The Orator’s Education , ed. and trans. D. A. Russell, 5 vols . (Cambridge, Mass., 2001), iv. 370-1 (10.6). Translation slightly altered.
42 Carruthers, The Book of Memory , pp. 250-1.
43 See Christopher Cannon, ‘From Literacy to Literature: Elementary Learning and the Middle English Poet’, PMLA , 129 (2014), 350-2.
44 Derek Pearsall, The Life of Geoffrey Chaucer (Oxford, 1992), pp. 287-8.
45The Riverside Chaucer , ed. Larry D. Benson (Boston, Mass., 1987), p. 650.
46 Carruthers, The Book of Memory , pp. 242-3.
47 Richardson, Middle-Class Writing , pp. 46 (‘scriveners’) and 47 (‘seem’).
48 Richard Firth Green, ‘The Early History of the Scriveners’ Company Common Paper and its So-Called “Oaths”’, in Simon Horobin and Linne Mooney (eds.), Middle English Texts in Transition: A Festschrift Dedicated to Toshiyuki Takamiya on his 70th Birthday (York, 2014), pp. 1-20: 2.
49 Ibid., p. 8. See also Nigel Ramsay, ‘Scriveners and Notaries as Legal Intermediaries in Later Medieval England’, in Jennifer Kermode (ed.), Enterprise and Individuals in Fifteenth-Century England (Stroud, 1991), pp. 118-31: 123.
50 Mooney, ‘Chaucer’s Scribe’, p. 98.
51 Paula McDowell, ‘Defoe’s Essay upon Literature and Eighteenth-Century Histories of Mediation’, PMLA , 130 (2015), esp. 575.
52 John Guillory, ‘Genesis of the Media Concept’, Critical Inquiry , 36 (2010), 321.
53 Kane, introduction to Piers Plowman: The A Version , p. 144.
54 The other manuscript lacking Troilus’s song is London, British Library, MS Harley 3943; those lacking Troilus’s meditation on predestination are Cambridge, University Library, MS Gg.4.27 (although this manuscript contains the last stanza) and BL, MS Harley 2392; the manuscripts lacking the description of Troilus’s ascent to the spheres are MS Harley 3943 and MS Harley 2392. See Troilus and Criseyde , ed. B. A. Windeatt (Harlow, 1984), pp. 37-41.
55 On Root’s theories and his edition see Ralph Hanna, Pursuing History: Middle English Manuscripts and Their Texts (Cambridge, 1996), pp. 97-114, and Troilus and Criseyde , ed. Windeatt, pp. 36-7.
56Troilus and Criseyde , ed. Windeatt, p. 38.
57 Ibid., p. 39.
58 I am thinking here of the distinctly different Prologues (‘F’ and ‘G’) we have for The Legend of Good Women , but also the theory that Chaucer added some passages to the Wife of Bath’s Prologue after he had ‘finished the text’: John Matthews Manly and Edith M. Rickert, The Text of the Canterbury Tales, Studied on the Basis of All Known Manuscripts , 8 vols. (Chicago, 1940), ii. 191. Windeatt also traces a large number of lines and phrases in MS HM 114 which seem to ‘go back to Chaucer’s first translation of the Italian’ and that therefore do suggest that this version of the text is ‘the survival, possibly corrupted, of a “rough” early draft’ (p. 44).
59 J. A. Burrow and Thorlac Turville-Petre, ‘Editing the B Archetype of Piers Plowman and the Relationship Between Alpha and Beta’, Yearbook of Langland Studies , 26 (2012), 134.
60The Vision of Piers Plowman , ed. A. V. C. Schmidt, 2nd edn. (1995), 8. 14-17.
61 Among the models that Burrow and Turville-Petre work to simplify is Ralph Hanna’s ( Pursuing History , pp. 203-43), but Hanna also helpfully summarises and corrects a number of the other theories that have been advanced.
62 Burrow and Turville-Petre, ‘Editing the B Archetype of Piers Plowman ’, p. 160.
63The English Works of John Gower , ed. G. C. Macaulay, 2 vols., EETS es 81-82 (1900-1), vol. i, p. cxxviii.
64 The four witnesses are Oxford, Bodleian Library, MS Fairfax 16, MS Bodley 638, MS Tanner 346, and Thynne’s edition (1532).
65 The Riverside Chaucer has ‘Men to slepe ne for to wake’ (p. 333), while Skeat adds a ‘for’ to smooth the rhythm at the beginning of the line, while leaving ‘slepe’ intact (so ‘Men for to slepe, ne for to wake’). See The Complete Works of Geoffrey Chaucer , ed. W. W. Skeat, 7 vols. (Oxford, 1894), i. 285.
66 Thynne departs from the other witnesses here and has ‘false’ in line 644.
67 F. W. Bateson, ‘Could Chaucer Spell?’, E in C , 24 (1974), 9. Bateson is thinking less of final - e here than of the way a performance of a poem might depart from its written form by the ‘elision’ of syllables (p. 14) or of the way, when read, an unstressed syllable might ‘hover between full articulation and the merest whisper’ (p. 15).
68 Brian Stock, The Implications of Literacy: Written Language and Models of Interpretation in the Eleventh and Twelfth Centuries (Princeton, NJ, 1983), p. 9.