Encoding the Edge: Manuscript Marginalia and the TEI

Laura Estill (Texas A&M University)

<additions>, <label>, <head>, <note>: with over five hundred elements, why can’t I find the one I need?[1] By tracing the different TEI (Text Encoding Initiative) elements and attributes available to describe and record marginalia, this paper shows how the placement and purpose of early modern marginalia require careful consideration by digital editors. Encoding marginalia encourages scholars to think about the purpose of the added text in a different way from the editors of pre-digital editions (and even digital facsimile editions). Encoding marginalia is an editorial act because encoders must consider, in concrete, quantifiable ways, how the marginalia relates to the rest of the text. Furthermore, TEI-encoded texts can be used as a corpus or in a database that allows scholars more flexibility in searching the text.  Peter Robinson claims that “there is no such thing as a non-interpretive transcript.” If the act of transcribing alone is interpretive, encoding takes things a step further: as Susan Schreibman elegantly explains, “Text encoding, like any other area of textual scholarship, is not theory-free. It is subjective, theoretical, and interpretative.”[2]  When considering marginalia, encoding raises questions about the definition of text, body, and marginalia itself. Ultimately, encoding marginalia is a data-modeling question that asks us to rethink the nature and function of our texts and paratexts, questioning where one ends and the other begins.[3]

H. J. Jackson begins her book, Marginalia: Readers Writing in Books, with the claim that “We all know the reader-annotated book of the present-day, and we prefer not to think about it” (1).  Jackson outlines the basic practices of marginalia: ownership marks or introductory material often appear in the front of books and notes are scribbled in margins.  For centuries, handwritten marginalia in printed books was generally overlooked (and, indeed, sometimes literally washed away). Recent scholarship has shown the importance of marginalia: William Sherman’s Used Books opens by explaining how schools taught early readers to write in their books. Copies of books that were well-used would be less likely to survive.  Heidi Brayman Hackel and Sonia Massai both show that marginal notations can help us to access what early readers thought of printed works. However, margins were not only for handwritten marginalia. In early modern books, printed text often appeared in margins as textual glosses or religious explications. Indeed, some texts with variorum-style commentary were printed with more annotations than text proper. William Slights highlights the multiple functions of printed marginalia in the early modern period, including amplification, simplification, and correction. Evelyn Tribble argues that “reading the margin shows that the page can be seen as a territory of contestation upon which issues of political, religious, social, and literary authority are fought” (2). The importance of text that appears in the margins of books cannot be overstated, be it handwritten or printed.

This article begins by assessing the options for encoding marginalia using TEI, showing that no single solution covers all cases.  One the benefits of the TEI is that it fosters interoperability between projects because it offers standardization; still, a survey of existing practices shows that when it comes to encoding marginalia, researchers share little consensus of best practices. And while it might be easy to advocate that the TEI adopt a new element for transcribing marginalia, the multifunctional and heterogeneous nature of marginalia could preclude the possibility of such a gesture towards standardization, as the final part of this article demonstrates. I conclude by turning to the manuscripts written by Archbishop William Sancroft (1617-1693) in order to analyze the varied deployments of marginalia. Ultimately, I argue that no one-size-fits all solution can exist for encoding marginalia; each editorial team must make decisions based on the specifics of their project. It is precisely because of the range of approaches that writers and readers take to marginalia that encoders and editors must weigh these different options carefully and document their decisions explicitly.

Encoding Marginalia: Existing Options and Practices

The challenge of transcribing marginalia with TEI is perhaps surprising, considering the length and thoroughness of the “Manuscript Description” chapter in the P5 Guidelines for Electronic Text Encoding and Interchange. Marjorie Burghart and Malte Rehbein note that the community of “‘manuscript encoders’ can be regarded as the largest sub-community within the TEI” (10). In 2011, the TEI incorporated the suggestions from the Workgroup on Genetic Editions, chaired by Fotis Jannidis, which added a great deal of functionality to manuscript encoding, including the ability to encode the location of text on a page, the direction of the text, and for editors to suggest the order in which a document was written (Jannidis, Burnard, Pierazzo, and Rehbein). The Proust Prototype (Pierazzo and André) shows the potential value of genetic encoding—that is, rendering text in blocks written in different zones on the plane of a page.

The genetic encoding additions to the TEI make encoding marginalia easier when the manuscript is treated as a document, but there are still improvements to be made when encoding the contents of a manuscript as a work (in the Tansellian sense).[4] Indeed, the genetic edition guidelines allow editors to rely closely on digital facsimiles and to undertake documentary editions and diplomatic editions. As the lingua franca of digital editions, however, TEI still needs to cater to those without large grants or digital humanities centers: that is, those who might not be able to publish page images and those who are more interested in the work-as-text than the document.  Even those who are working from a single document (as, for instance, a manuscript) might choose not to use the genetic encoding model of the TEI, which requires giving the attention to reconstructed revision timelines and the layout on the piece of paper as a geometric plane. This is the case for those creating TEI documents that would be incorporated into a database, for editors who are following the traditional copytext-variant method, and for those using TEI to encode multiple witnesses (documents) of a work.

While the TEI offers valuable ways to describe marginalia as metadata in the Manuscript Description element (<msDesc>), particularly in the Physical Description (<physDesc>), encoding marginal additions in the body of the XML document can be a fraught undertaking. When encoding a document in TEI, there are two main sections: the header (<teiHeader>), which is the paratextual material that describes the work, document, and the encoding process, and the text (<text>), which contains the text itself—generally as copied from one document, but sometimes with other witnesses incorporated. In the case of manuscript studies, the <teiHeader> describes the manuscript (provenance, ink, watermarks, size), and the <text> furnishes transcriptions of what appears on the physical codex, scroll, or leaf. If you search “marginalia” in the TEI guidelines, the first result is the element <additions>, which as the guidelines note, “contains a description of any significant additions found within a manuscript, such as marginalia or other annotations.”[5] <additions>, however, is found in the <physDesc> (physical description) part of the manuscript header, which only describes the manuscript and does not present its text. The <additions> element is ideal, for instance, for noting that a seventeenth-century manuscript includes nineteenth-century commentary, as is the case with Bodleian MS Eng. poet. e. 97. In the <additions> element, users can describe the marginalia in a given manuscript, and, if desired, transcribe marginalia. Relying too heavily on <additions> can lead to multiple shortcomings: first, although the narrative description of the marginalia would be suitable for a catalogue entry or introduction to a manuscript, such a description is by nature paratextual. Second, if the marginalia is transcribed only in the <additions> element, it literally separates the marginalia from the text (and the <text>), which suggests that the marginalia is not, in fact, the text worth studying. Another unsatisfactory solution is to redundantly transcribe all marginalia in two places, both the header and the body. While describing marginalia in the header is can be a valuable undertaking, I contend that it is necessary to position a transcription of it in the <text> of the TEI document, as often the marginalia is the most fascinating part of a manuscript.

The inadequacy of <additions> prompted Laura Mandell to ask on the TEI listserv: “Can anyone direct me to a tag that allows for actually transcribing marginalia?” The range of responses shows that the community does not agree on any one element for transcribing marginalia in the text of the document.  Without one specified for this use, projects lose potential interoperability; furthermore, large data mining analysis would be unable to compare the marginalia in different digital corpora because it is encoded differently.  (Comparatively, a data mining project could easily find all epigraphs in TEI editions, as epigraphs are encoded with the very intuitive element <epigraph>.) In the handful of responses to the TEI listserv, we hear about projects using <note>, <label>, <add>, <seg>, <milestone>, and even <argument>. The listserv discussion pointed out that each of these tags would only be appropriate for certain types of marginalia (added by hand into a print book, printed in the margin, written at the time of composition into a manuscript, etc.). Some projects use more than one of these elements for marginalia; others use the same element for both marginalia and other items, such as editorial footnotes. In these cases where a single element is not used to define marginalia and only marginalia, to search only marginalia would involve detailed drilling into the xml in order to create a list of elements used and attributes that distinguish them--it is not a recipe for interoperability or intuitive searchability.

Each element that an encoder might consider using for transcribing marginalia as part of the text carries a different semantic weight.[6] Semantically, the definition of a <note> element (“a note or annotation,” as defined by the TEI Guidelines) could apply to many different kinds of marginalia: authorial, editorial, readerly… or even those absent-minded doodles we find in school textbooks. The <note> element lends itself particularly well to footnotes and other notes that are anchored at one spot in a line of text.[7] The Deutsches Textsarchiv, which focuses on marginalia printed in books, uses only @place to differentiate between notes in the margin and footnotes. Indeed, their marginalia functions in the same way as footnotes, being notes included by the author, early editor, or printer. However, many marginal additions refer to longer sections of texts (which, for instance, could be indicated by curly braces or marginal lines) or to multiple and disparate sections of texts (indicated, for instance, by arrows). The use of <note> for such disparate uses requires the use of attributes to clarify what type of note is being encoded, as recommended by the TEI Lite customization.[8] The University of Saskatchewan’s Textual Communities exemplifies this practice by encoding marginalia with a @place attribute and editorial notes with @resp (responsible party) and @type attributes.[9]  Leaving the @resp attribute off marginalia makes sense (despite TEI Lite’s decree), as in the case of a great deal of marginalia, we have no way to know who wrote it. As Massai points out, there is also much marginalia that we cannot date with any certainty (153). Textual Communities is designed to serve a range of scholars, from those working on Beowulf to those working on postcards, and as such, has a broad understanding of marginalia. As a multipurpose element, <note> can enclose a range of transcribed marginalia, that is, “any additional comment found in a text” (TEI Guidelines 3.8.1).

The definition of the <label> element, like the <note> element, could apply to manuscript marginalia: “any label or heading used to identify part of a text, typically but not exclusively in a list or glossary.” The TEI guidelines provide many examples of the use of <label>, primarily focused on lists and graphs, but offer only one example of <label> as an example of “Notes, Annotation, and Indexing.” The guidelines suggest that <label> would be appropriate to use for marginal interventions that “provid[e] a heading or descriptive label for the passage concerned.” Claire Carlin and Martin Holmes’s Le Mariage sous L’Ancien Régime is an example of a project that has adopted the use of <label> to describe the marginal explanations present in early modern French printed texts.
To determine which tag to use for marginalia, Paul Schaffner asks, “is this thing in the margin best regarded as intrinsic or extrinsic to the main text?” Both <note> and <label> suggest that the marginalia is extrinsic to the text.  Schaffner suggests that in the right circumstances (such as marginal numbering of the verses of a Psalm), <milestone> could be appropriate for this extrinsic text.[10] Similarly, Schaffner shows how <argument> could be appropriate for text found in the margin, such as “2 God's spirit shewes it selfe an holy Spirit, in begeting good motions” (Jenison 223, Schaffner’s example). Schaffner admits, however, that using <argument> can be “violen[t] to the source” and can lead to tag abuse, as <argument> elements have to come at the front of a section (<div>, or textual division). The EEBO-TCP (Early English Books Online-Text Creation Partnership) suggests that <label> might be appropriate for some marginalia, but that <stage> (stage direction) would also be appropriate when encoding plays with marginal stage directions. With narrowly-defined tags holding marginalia, projects will need to use more than one tag. If a project uses @place consistently and accurately, a user could still be able to search with relative ease to find all text in the margin.  The countless attribute values that are possible on all elements describing marginalia (LHMargin, margin, left_margin, etc.) would make large-scale analysis of multiple manuscript corpora almost impossible without painstakingly creating a list of each project’s idiosyncratic choices.

If, at one end of the spectrum, we have <argument> and <milestone> as granular elements that can only contain very particular kinds of marginalia, at the other end of the spectrum we have <seg> (arbitrary segment), which offers an even more capacious understanding of marginalia than <note>. The Guidelines explain that <seg> “represents any segmentation of text below the ‘chunk’ level.”[11] Roughly speaking, <seg> tags enclose a portion of text that is less than a paragraph long (although not always). Sites such as the International Greek New Testament Project and the Adams Digital Edition use <seg type="margin"> (Leblanc).  And while <seg> seems to have a definition broad enough to do the job, its attribute constrictions add another challenge: @place is not allowed on <seg>, which is why these projects use @type. This only adds to the proliferating and confusing options for encoding marginalia.

If <seg> and <note> are too big, and <stage> and <argument> are too small, perhaps <add> is the Goldilocks element that is just right.  <add> “contains letters, words, or phrases inserted in the source text by an author, scribe, or a previous annotator or corrector” (TEI Guidelines 3.4.3). Some projects use <add> only to describe material supplied by the author (such as words that have been crossed out and then added after).  For instance, the Walt Whitman Archive uses <add> for notes that Whitman pasted into his books. Digitizing Whitman’s Annotations and Marginalia, an ancillary project to the Walt Whitman Archive, focuses entirely on annotations that Whitman made to other texts.  Matt Cohen, a contributing editor to the Whitman Archive, explains that Whitman’s textual interactions are not easily categorized (from scribbling over printed text, to pasting in printed sheets). Digitizing Whitman’s Annotations and Marginalia uses <add> for Whitman’s clippings, but nests three additional tags within before covering the text of the clipping. Based on their work with the Whitman’s Annotation project, Erica Fretwell and Cohen weigh the pros and cons of even using TEI at all to encode digital editions in general, and marginalia in particular.

All of the projects that I have described so far encode marginalia in a way that makes sense for their project and follows the TEI guidelines--but each primary investigator had to determine the best way to encode the marginalia for their documents and works, as there is no one-size-fits-all solution from the TEI. In the next section of this essay, I show how conceptualizing manuscript marginalia for encoding challenges our understanding of the nature of marginalia.

Heterogenous Marginalia: The Case of Archbishop Sancroft’s Manuscripts

Slights acknowledges that “Just what is ‘text’ and what ‘margin’ in a particular book is not always easy to say” (61). He argues that “The difference between primary text (remembering that primacy is not always easy to determine) and annotation is semiotically crucial in the production of meaning” (62). That is to say, there is more to texts in documents than can be defined with the binary of “intrinsic” and “extrinsic.” Perhaps there are easier distinctions to be made when marginalia is handwritten in printed books—although even then, in the case of authorial revisions, stop-press corrections, or (say) Whitman’s notes in another book, there is no easy answer as to what is “marginal.” The following examples are taken from already challenging potential cases: handwritten manuscripts with handwritten notes in the margin and elsewhere.

Archbishop William Sancroft was a prolific manuscript compiler who wrote dozens of notebooks for his personal reference and scholarship.[12]  Here I consider the parts from plays that Sancroft copied. In one of his manuscript miscellanies, Sancroft transcribed selections from “Yorke, & Lancaster,” referring to Shakespeare’s 2 Henry 6, which was originally published as The Contention of York and Lancaster (1594).[13] Figure 1 shows that Sancroft reconsidered his original title, “Yorke, & Lancaster. 1[st] part,” blotted the indication that this was the first part, and added it to the left-hand margin. Sancroft differentiated between “Yorke, & Lancaster” and “1[st] part,” perhaps considering the former a heading and the latter a subheading. In this case, <head> would be an appropriate element for both, with @place and @type attributes to make the difference clear.

Figure 1: Bodleian Libraries, University of Oxford, MS Sancroft 97, p. 80.

In other cases, however, Sancroft did not identify his source in a way that can be captured with the <head> element, as <head> refers to the opening of a section of a document.[14] Figure 2 shows another site where Sancroft has copied an extract from a play, delineated this extract from surrounding material, and provided the play’s title. Here, Sancroft offers a quote from John Webster’s The White Devil, “There are not Jewes enough; for why else do so many Christians turn Userers?...” Sancroft labels the play by its alternative title, Vittoria Corombona, and gives an act/scene reference. Sancroft’s added title should be treated in the same way as his identification of 2 Henry 6, but the <head> element cannot be used as it is currently defined in the TEI Guidelines. Moreover, even if in the future, the <head> element could be placed at the middle or end of a textual division, Sancroft seems here to be providing more of a reference than a heading. Sancroft’s identification of the Henry 6 plays and Vittoria Corombona function in the same way, and therefore should be treated in the same way; thus <label> seems a clearer way of encoding Sancroft’s text.

Figure 2: Bodleian Libraries, University of Oxford, MS Sancroft 18, p. 5.

These two examples from Sancroft’s manuscripts raise another question: does marginalia have to be in the margin? Figure 1 and Figure 2 show how Sancroft’s additions should not be taken as separate from the text itself: Sancroft placed labels that function similarly as centered titles, in the margin, and in-line with the text itself. The TEI element <label> allows for these two play titles to be semantically the same, and their placement on the page can be encoded using the @place attribute.[15]

Sancroft’s in-line commentary on the extracts from plays is not limited to providing a reference to his source-text such as a title or act and scene reference. As Figure 3 illustrates, Sancroft copied a series of extracts from The Merry Wives of Windsor, including a paraphrase of Falstaff’s lines describing “sack” (that is, fortified wine) “brewed with pullet sperm.” This is one of Falstaff’s more colourfully imaginative phrases whose meaning might not be clear at first glance. Sancroft added the note “Eg[g]s,” marked as an aside with a bracket, to explain what “pullet [chicken] sperm” means. In this case, the TEI element <gloss>, “a phrase or word used to provide a gloss or definition for some other word or phrase,” would perfectly capture the Archbishop’s addition.

Figure 3: Bodleian Libraries, University of Oxford, MS Sancroft 29, p. 86.

A comparable example to the “Eg[g]s” gloss would be Sancroft’s note in Figure 4 which says “Richard 3d”: both are similarly demarcated from the surrounding text, and both offer context for the extract that they are about. However, the <gloss> element does not suit the latter. Although “Richard 3d” might seem more akin to the other cases where Sancroft referred to play titles (The Contention betwixt the two famous Houses of Yorke and Lancaster and Vittoria Corombona), Sancroft’s note is not a play title, but rather an explanation of the passage’s content. The speech in Figure 4, adapted from 3 Henry 6 (not Richard III), is about Richard III’s birth. Sancroft anachronistically identifies the infant Richard with his future title as king. And while <gloss> is the obvious choice to encode the explanation of “pullet sperm” as eggs, <gloss> does not capture Sancroft’s use of “Richard 3rd” to identify the subject matter of King Henry’s speech. An encoder could use two different elements in this situation, but this would suggest that the two textual interventions differ. Yet they are substantially similar: with both “Eg[g]s” and “Richard 3d,” he is providing explanatory notes that provide a context for understanding his extracts from plays. In order to encode both “Eg[g]s” and “Richard 3d” in the same way, a user is faced with the same choice regarding Sancroft’s titling of extracts: <note> or <label>. Furthermore, the encoder would have to decide if this gloss and explanatory note function in the same way as the titles that Sancroft provides for his sources. In this case, it would be reasonable to include the @type attribute on <label> if an encoder wanted to distinguish between Sancroft’s references and his explanations.

Figure 4: Bodleian Libraries, University of Oxford, MS Sancroft 29, p. 103.

Sancroft not only transcribed and paraphrased individual lines and speeches from early modern plays; he also copied dialogue. In his selections from Webster’s The Devil’s Law Case, Sancroft copied a short exchange: “A. I have a suit to you. B. Value it not the less, if I say, Tis granted already. A. You are all Bounty” (see Figure 5). Sancroft’s marginal notation next to these extracts is the play’s title (which, as we have seen, he also writes centered above his text or sometimes run-in to the text itself.) In this example of Websterian dialogue, Sancroft uses A and B to indicate the speakers. A full-text version of this play, either in print or manuscript, would give the characters’ names, Leonora and Contarino, or an abbreviation of them. An encoder would then use the element <speaker>, which was designed for exactly that situation. There is possibly an argument for using <speaker> to represent A and B because they are stand-ins for the characters’ names (or at least abstractions of each interlocutor).[16] In Figure 1, Sancroft indicated that the first lines he copied from 2 Henry VI were spoken by an ambassador. As with “Richard 3d” (in Figure 4), Sancroft’s “Embass[ador]” could easily be mistaken for a speaker name. The print version of Shakespeare’s play from which Sancroft was copying clearly names the Duke of Suffolk, though he is indeed acting as an ambassador when he entreats King Henry, “Your gracious excellence that are the substance of that great Shaddow I doe represent.”[17] Sancroft’s label, ambassador, not only explains the role of the speaker; it also describes the content of Suffolk’s words.

Figure 5: Bodleian Libraries, University of Oxford, MS Sancroft 29, p. 69.

Sancroft is not the only early modern reader who copied selections from plays, nor is he the only reader who copied dialogue. One anonymous mid-seventeenth century compiler sometimes chose to mark his dialogue with A and B, as Sancroft did, to represent the speakers. But at other instances in the same manuscript (Folger MS V.a.87), this manuscript compiler instead used “An[swer]” or “Reply” to indicate a change in speaker. For “An[swer]” and “Reply,” it would be inappropriate to use the <speaker> tag. Again, we return to <label> as a good fit. And, while <label> can fit all of these examples, <note> does not seem to apply to, for instance, the marking of speakers as A or B.

As I touched on above, applying attributes to the <label> element can clarify how this element is used. The @place attribute is a natural fit for the <label> element and can be used to indicate the placement of the label on the page, with attribute values such as “LHMargin” or “inline.” The @type attribute could be used to distinguish between Sancroft’s references, descriptions, and indication of speakers, though an encoder could also conceptualize all of these as “labels” that Sancroft applied to the text in order to increase his understanding when he turned back to his manuscript miscellany. Currently, however, the responsibility attribute class (att.responsibility) is not permitted on the element <label>, which means that the attribute used to indicate parties responsible, @resp, cannot be applied.[18] In the case of Sancroft’s manuscripts, this would not be an issue, as his hand is generally the only one present. In other manuscripts, however, where multiple people have included marginal notes, additions, explanations, or glosses, it would be important to indicate the change in hand, and, when known, the different writers—as in the example of the manuscript, which I mentioned in the introduction, with marginal notes added in the seventeenth and nineteenth centuries, Bodleian MS Eng. poet. e. 97.

In preparing an edition of a manuscript, an editor could simply reproduce the marginalia as it stands and allow a reader to determine what it refers to, by offering a diplomatic edition or a page transcription. The TEI allows for more than editions: it can be used in XML databases. A scholar who considers the manuscript extracts from early modern plays not as documents or complete texts, but rather as a corpus or dataset, could want search results that bring up not an entire manuscript, but instead, a list of the extracts from a particular play or lines spoken by a specific character across a variety of manuscripts. Scholars interested in the reception of early modern drama would undoubtedly want to know if an early reader had added marginal comments or changed the text. As the examples from Sancroft’s manuscripts show, early readers had no compunction about changing what they copied. The TEI offers a simple mechanism for allowing both the original and the paraphrase to be encoded, as both options can be nested in a <choice> element and displayed alongside each other in a search result. Marginal notations and in-text notes can also appear in search results, but this requires the encoder to clearly mark the label’s subject and referent. Ideally, the search results returned would allow users to click through to a page facsimile or diplomatic transcription of the complete manuscript page, which would encourage database users to draw their own conclusions.

In the Sancroft examples, both his in-text and marginal labels apply to fairly easily identifiable passages of text. “Eg[g]s” and “Richard 3d” have direct applications—to “pullet sperm” and Henry’s speech, respectively. His centered title, “Yorke, & Lancaster,” extends to both part one and part two (see Figure 1). His marginal label, “1st part,” extends only partway down the page, until he provides another marginal label, “2d part,” and begins copying from the next play. Multiple labels can refer to the same point in the text, as “Yorke, & Lancaster,” “1st part,” and “Embass[ador]” all have overlapping referents: the Ambassador’s speech is part of the first part of 2 Henry 6. Sancroft’s manuscripts illustrate the range of engagement that one early modern reader could have with dramatic texts--and he was certainly not the only reader who copied notes into a personal notebook, which is just one of many kinds of document wherein we find marginalia.

British Library Lansdowne MS 1185 is another late seventeenth-century miscellany that contains copious selections from early modern drama. The anonymous Lansdowne compiler, like Sancroft and other seventeenth-century readers, included multiple labels that perform functions as diverse as offering cross-references, indicating source-text or speaker, or distilling a passage’s theme. And while some marginal notations clearly refer to one extract or a handful of extracts, the encoder must still decide what the label refers to in cases where a comment is scrawled in a margin, or begun and hastily deleted. As an example, Figure 6 shows a page of the Lansdowne manuscript where the left-hand margin includes the marks “John,” “Of the,” “The Pope’s Authority ridiculd,” “John” (a second time), “p. 9.” Another speaker name, “Lewis,” floats between text and margin. An encoder would have to decide where, in the text, each label starts and each label ends. Does “The Pope’s Authority ridiculd” refer to the whole page? Lewis’s interjection, “Brother of England you blaspheme in This,” performs exactly the opposite: he is arguing against John’s anti-Catholicism. An encoder would also have to decide if “Of the” should be linked to any text and if the glyph next to “f. 9” is a misshapen “F” or an unknown symbol—and, furthermore, if it is a symbol, if and how it should be encoded. Moreover, the encoder would have to judge whether the marginal <label> should be encoded as “Of the The Pope’s Authority ridiculd p. 9,” or should instead be divided at any point. The text itself might even support a reading that separates “The Pope’s Authority” and “ridiculd.”

Figure 6: (c) The British Library Board, Lansdowne MS 1185, f.9v.

Although facsimiles, documentary encoding, and diplomatic editions might more accurately capture the complexities of manuscript documents, textual encoding is necessary if these manuscripts are to be searchable in a larger database. Facsimiles are only “searchable” by those who can read a particular handwriting: as of now, early modern handwriting cannot be digitally searched by, for instance, transforming it through Optical Character Recognition (OCR). Rather, page images, like the manuscript itself, can only be combed through by individual scholars, which is a time-consuming task and does not allow for the application of textual tools. Diplomatic editions and documentary encoding, which represent the text as it appears on the page as accurately as possible, might be keyword searchable, but without editorial markup they cannot be searched by category (for instance, you cannot search simply the incipits of poems or only the marginalia). Both documentary and textual encoding are valuable scholarly contributions that require an encoder to be an editor; both can function together and complement each other (Brüning, Henzel, Pravida 19-22). The TEI guidelines allow encoders to determine which elements of their documents are most important and encode accordingly; in the case of many early modern miscellanies, some of the most important things to include are transcriptions of the text and its marginalia, notes, and jottings.

Manuscript miscellanies are collections of smaller parts, mostly taken from other sources: they can be considered as a unified document, but they can also be considered as textual snippets taken from multiple works that need to be able to be searched and displayed as such.  The benefits of putting these snippets into a database, in this case, are that it enables users to search across manuscripts and display many results. A corpus of manuscripts encoded in a standard way (that, for instance, use the same schema) can serve as the basis for a database that allows for complex queries.

Encoding marginalia is an editorial act not least because the encoder decides what the marginalia refers to and how it functions in relation to the rest of the text. The TEI encourages editors and bibliographers to carefully consider their texts: in the TEI, there will always be multiple ways of approaching a text, each of which will carry inherently different conceptualizations of that text. This essay has argued for a consideration of marginalia and additions as text (in the <text>) rather than paratext (relegated to the <teiHeader>) and has explored the multiple ways in which marginalia and textual additions can be presented in TEI.

Even if the TEI decided to add a new tag to signal marginal text, this would not solve the challenges inherent in considering what text is and what counts as paratext in a given document.  As the examples from Sancroft’s manuscripts show, text in the margin can serve the same function as text placed anywhere else on the page. Turning our focus to the text on the edge of the page changes our understanding of the text on the center of a page and upends traditional conceptualizations of text and paratext.
As encoders grapple with marginalia and textual additions in their documents, they gain a better understanding of the texts they work with and how textuality functions writ large. It is only with strong documentation (of both the XML and the project) that their particular choices can be made clear. And with well-documented and consistently-encoded TEI projects, perhaps we will find ways to make the margins more searchable using Linked Open Data. In the meantime, scholars are already considering how we can add virtual margins to the digital projects we produce (Schofield and Weber; Wittern).  Ultimately, the encoding choices we make affect not only how we value and display the dialogues that have been taking place in margins for centuries, but also how we interact with these ideas and join their conversation.

Works Cited

Boyd, Jason and John Bradley, with Carolyn Black, Sally-Beth MacLean, and Pauel Vetch for the Fortune Theatre Records (FTR) Team. “The Fortune Theatre Records: A Prototype Digital Edition, Records of Early English Drama (REED).” White Paper. University of Toronto, 2013. Online. <http://reed.utoronto.ca/wp-content/uploads/2013/12/FortuneWhitePaper.pdf>
Burghart, Marjorie, and Malte Rehbein. “The Present and Future of the TEI Community for Manuscript Encoding.” Journal of the Text Encoding Initiative 2 (2012).  Online. <http://jtei.revues.org/372>
Brüning, Gerrit, Katrit Henzel, and Dietmar Pravida. “Multiple Encoding in Genetic Editions: The Case of ‘Faust.’” Journal of the Text Encoding Initiative 4 (2013). http://jtei.revues.org/697.
Carlin, Claire, Martin Holmes, et al. Le Mariage Sous L’Ancien Régime. Online. <mariage.uvic.ca>
Cohen, Matt. “Editing Walt Whitman’s Marginalia Today: Digital Humanties Method at the Edge.” Hall Humanities Center Lecture, University of Kansas. 1 May 2014. Online. <https://mcohenlab.wordpress.com/2015/05/09/digital-humanities-method-at-the-edge/>
Deutsches Textarchiv (DTA). “DTA-Basisformat – Inhaltliche Erschließung des Volltextes.” Online. <http://www.deutschestextarchiv.de/doku/basisformat_texterschliessung_inhaltlich>.
Digitizing Whitman’s Annotations and Marginalia. “Encoding Guidelines for Annotated Documents.” Online. <http://asmodeus.ws/cohenlab/Annotations_Guidelines_Draft.pdf>
Early English Books Online--Text Creation Partnership (EEBO-TCP). “EEBO tagging cheat sheet (1).” Online. http://www.textcreationpartnership.org/docs/dox/cheat.html
Flanders, Julia, and Fotis Jannidis. “Data Modeling.” A New Companion to Digital Humanities. Ed. Susan Schreibman, Ray Siemens, and John Unsworth.  Wiley-Blackwell, 2015. Ebook. DOI: 10.1111/b.9781118680643.2016.00018.x.
Fretwell, Erica, and Matt Cohen. “Interface Design for Static Multimedia Documents.” Society for Textual Scholarship, New York. March 2009. Online. http://asmodeus.ws/cohenlab/STS_09_CohenFretwell.pdf
Hackel, Heidi Brayman. Reading Material in Early Modern England: Print, Gender, and Literacy.  Cambridge: Cambridge UP, 2005. Print.
International Greek New Testament Project. “IGNTP guidelines for XML transcription of New Testament manuscripts using the TEI P5.” 2011. Online. <http://epapers.bham.ac.uk/738/1/IGNTP_XML_guidelines_1-1.pdf>
Jackson, H. J. Marginalia: Readers Writing in Books. New Haven: Yale UP, 2001. Print.
Jannidis, Fotis, Lou Burnard, Elena Pierazzo, and Malte Rehbein. An Encoding Model for Genetic Editions. Revised Edition. Text Encoding Initiative, 2010. Online. <http://www.tei-c.org/Activities/Council/Working/tcw19.html>
Jenison, Robert. The Christians apparelling by Christ. London: 1625. Online (Early English Books Online).
Leblanc, Ondine. “Re: Marginalia.” TEI Listserv email. <http://markmail.org/message/hxqvtcagcad6cfw2#query:+page:1+mid:wr2yjzoumds2lcmn+state:results>
Mandell, Laura. “Marginalia.” TEI Listserv email. <http://markmail.org/message/hxqvtcagcad6cfw2#query:+page:1+mid:4rkkhdkogie3lbta+state:results>
Massai, Sonia. “Early Readers.” The Oxford Handbook of Shakespeare. Ed. Arthur F. Kinney.  Oxford: Oxford UP, 2011. 143-61. Print.
Niles, Rebecca, and Mike Poston. “Re-Modelling the Edition: Creating the Corpus of Folger Digital Texts.” Early Modern Studies and the Digital Turn. Ed. Laura Estill, Diane K. Jakacki, and Michael Ullyot.  Toronto: Iter, 2016.  119-46. Forthcoming.
Pierazzo, Elena. “Digital Documentary Editions and the Others.” Scholarly Editing: The Annual of the Association for Documentary Editing 35 (2014). Online. <http://www.scholarlyediting.org/2014/pdf/essay.pierazzo.html>.
Pierazzo, Elena, and Julie André, with Rafaele Viglianti. “Autour d’une séquence et des notes du Cahier 46: enjeu du codage dans les brouillons de Proust. Around a sequence and some notes of Notebook 46: encoding issues about Proust's drafts.” [Proust Prototype.] King’s College London. Online. http://research.cch.kcl.ac.uk/proust_prototype
Robinson, Peter. “Some Principles for the Making of Collaborative Scholarly Editions in Digital Form.” Digital Humanities Quarterly, Forthcoming. Draft online. <https://www.academia.edu/12297061/Some_principles_for_the_making_of_collaborative_scholarly_editions_in_digital_form>.
-----. “Towards a Theory of Digital Editions.” Variants 10 (2013): 105-32. Online. <https://www.academia.edu/3233227/Towards_a_Theory_of_Digital_Editions>
Schaffner, Paul. “Re: Marginalia.” TEI Listserv email. <http://markmail.org/message/hxqvtcagcad6cfw2#query:+page:1+mid:cia4v536xsqyflgs+state:results>
Schofield, Scott, and Jennette Weber. “Opening the Early Modern Toolbox: The Digital Interleaf and Digital Commonplace Book.” Scholarly and Research Communication 4.3 (2013). http://src-online.ca/index.php/src/article/view/127/324
Schreibman, Susan. “Digital Scholarly Editing.” Literary Studies in the Digital Age: An Evolving Anthology.  Ed. Kenneth M. Price and Ray Siemens. New York: Modern Language Association, 2013. Online. <https://dlsanthology.commons.mla.org/digital-scholarly-editing/>
Sherman, William. Used Books: Marking Readers in Renaissance England. Philadelphia: U of Pennsylvania P, 2007. Ebook (Project MUSE).
Slights, W. E. Managing Readers: Printed Marginalia in English Renaissance Books. Ann Arbor, MI: U of Michigan P, 2001. Print.
Tanselle, G. Thomas. A Rationale of Textual Criticism. Philadelphia: U of Pennsylvania P, 2011. Ebook (Project MUSE).
Tribble, Evelyn B. Margins and Marginality: The Printed Page in Early Modern England. Charlottesville, VA: U of Virginia P, 1993.
Text Encoding Initiative (TEI). “Feature Requests.” http://sourceforge.net/p/tei/feature-requests/
-----. P5: Guidelines for Electronic Text Encoding and Interchange. Version 2.8.0.  2015. Online. <http://www.tei-c.org/release/doc/tei-p5-doc/en/html/index.html>.
-----. “TEI Lite.” Online. <http://www.tei-c.org/Guidelines/Customization/Lite/>.
Textual Communities. “Default Transcription Guidelines.” University of Saskatchewan.  Online. <http://www.textualcommunities.usask.ca/web/textual-community/wiki/-/wiki/Main/Default+transcription+guidelines>
The Walt Whitman Archive. “Encoding Guide for Poetry Manuscripts.” Online. <http://www.whitmanarchive.org/about/guidelines/index.html>
Wittern, Chrisian. “Beyond TEI: Returning the Text to the Reader.” Journal of the Text Encoding Initiative 4 (2013). <http://jtei.revues.org/691>
Women Writers Project. 2007. “What is Text Encoding?” Women Writers Project Guide to Scholarly Text Encoding. Northeastern University. Online. <http://www.wwp.northeastern.edu/research/publications/guide/html/encoding.html>

[1] I would like to thank Krista May, Constance Crompton, and Kailin Wright for their thoughtful feedback on this paper.  A version of this essay was originally presented at the 2013 Canadian Society for Digital Humanities meeting at the University of Victoria.  This essay is part of the work towards DEx: A Database of Dramatic Extracts, which has been generously funded by the Social Sciences and Humanities Research Council of Canada (SSHRC) and Texas A&M University, and is forthcoming from Iter: Gateway to the Middle Ages and Renaissance at dex.itercommunity.org.

[2] For more on encoding as an interpretive and scholarly act, see the Women Writers Project and Rebecca Niles and Mike Poston

[3] For a discussion of data modeling, including how that can apply to text encoding and XML, see Julia Flanders and Fotis Jannidis and Niles and Poston.

[4] One of the valuable additions for manuscript encoding that came out of the Workgroup on Genetic Editions is the <metamark> element, which can be used to encode marginal crosses, manicules, and the like that indicate how a reader should approach the text. The <metamark> element exemplifies how encoders must consider the function of marginalia. This essay follows G. Thomas Tanselle’s definitions of text, work, and document: the document is one physical copy of a work; a work can appear in multiple documents (sometimes called “witnesses”).  The text is the words that make up a work and that are written in a document. Peter Robinson (“Towards a Theory”) and Elena Pierazzo similarly follow this terminology: see their essays for more of the kinds of editions made possible online (and with the TEI).

[5] Unless otherwise noted, all quotations about the function of TEI elements and attributes are taken from the P5: Guidelines for Electronic Text Encoding and Interchange version 2.5.0, last updated 26 July 2013, available at www.tei-c.org.

[6] Similarly, the customization of a schema that proscribes the use of elements can alter their semantic value. See the TEI Guidelines 23.3, “Personalization and Customization.”

[7] For an example of a project that uses anchored <note> elements for marginalia, see Jason Boyd and John Bradley’s discussion of the Records of Early Drama, particularly 2.1 and 5.1.

[8] In TEI Lite (a “specific customization of the TEI tagset, designed to meet ‘90% of the needs of 90% of the TEI user community’”), the instructions are clear for how to encode marginalia in the body of a document: “All notes, whether printed as footnotes, endnotes, marginalia, or elsewhere, should be marked using” <note>. Encoding marginalia in a <note> element means that the @resp (responsible party) attribute has to be applied to “distinguish between authorial and editorial notes, if the work has both kinds.”

[9] The Textual Communities Default Transcription Guidelines offers the following examples: <note place="margin-left"> encodes marginalia and <note place="ed" resp="PRM"> encodes editorial notes.

[10] While all of the examples in the TEI Guidelines for milestone do not enclose text (they are <milestones/>, it seems reasonable that, like its counterpart <p> (page), it could be used to represent numbers that are written in the document.

[11] The TEI Guidelines explain the different levels elements can exist on in “Informal Element Classifications” (

[12] For a more detailed discussion of Sancroft’s manuscript collection and his transcribing habits, see Author 2015, especially chapter five.

[13] Bodleian MS Sancroft 97. For manuscript transcriptions, conventional scribal abbreviations have been silently expanded, such as ‘ye’ for ‘the’ and ‘Xn’ for ‘Christian.’ Other expansions are marked with square brackets. Both u/v and i/j have been normalized.

[14] There has been discussion in the TEI community about whether <head> elements should be allowed in more places than openings. See, for instance, Syd Bauman’s and Brett Barney’s SourceForge feature requests, http://sourceforge.net/p/tei/feature-requests/387/ and http://sourceforge.net/p/tei/feature-requests/390/, respectively.

[15] <note> can also be differentiated with the @place attribute, but as discussed above, could be semantically tied to editorial notes as well.

[16] If <speaker> were used for “A.” and “B.”, the @who attribute could be applied to refer to the names of the characters who spoke the text in the complete play, even though the manuscript compiler omitted them.

[17] All print versions of 2 Henry VI name Suffolk as the speaker; none have him described as “Ambassador.”

[18] Holmes has proposed to have @resp be a member of att.global, which would make it available to be used on any TEI element: http://sourceforge.net/p/tei/feature-requests/443/.