Abstract
A margin is not empty space waiting for a reader’s personality to arrive. It is a working zone in which a text may be corrected, indexed, claimed, disputed, rehearsed, decorated, redirected, or made available to another reader. Yet the visual resemblance among these marks invites a persistent error: an underline, a proofreader’s caret, a medieval nota, and a shared digital comment are all called “annotation,” as if the common name settled their social purpose.
This paper offers a more disciplined unit of analysis: the marginal transaction. A marginal transaction joins a mark to a surface, a passage, an actor, a candidate purpose, an audience, and a time condition. The point is not to eliminate interpretation. The point is to keep an observed mark from becoming an invented biography of the person who made it. Drawing on manuscript and rare-book cases, archival principles, a contemporary Web-annotation standard, and annotation research, the paper introduces the Margin Function Atlas: a reusable instrument for describing what is visible, stating what is inferable, and labeling what remains unknown.[1] [2] [3]
Working principle: Describe the mark before assigning it a motive. Record the uncertainty before telling the story.
The crowded edge of a sentence
Consider one plain sentence: “The committee shall publish the revised notice after review.” It appears four times on a teaching sheet.
The first copy bears a pencil question mark beside after review, followed by the words “who reviews?” The second has a proofreader’s caret moving the same phrase. The third has a small printed pointing hand directing attention toward review. The fourth is a screen capture: committee is highlighted, and a visible comment asks, “Name the approving body.”
At first glance, all four additions live at the edge of the sentence. That fact does not make them the same act. The pencil note may be a reader’s comprehension question, a reminder to the writer, an anticipated discussion prompt, or a query that no one else was meant to see. The caret may be a proposed correction, a house-style instruction, or a step in an editorial exchange whose acceptance cannot be determined from the sign alone. The pointing hand may help retrieval, but it may also announce that a passage deserves attention. The shared comment has a precisely selected target and a visible text body; depending on the service, it may also have an account identity, notification route, revision history, audience setting, and export behavior.
The temptation is to flatten these differences into a pleasant generalization: people have always “written in the margins.” That statement is not false, but it is too shallow to teach with. It joins together actions that carry different authority, different audiences, and different consequences for the text. A reader who learns only to identify a mark will overread it. A reader who learns to ask what transaction the mark performs can move more carefully between a manuscript page, a marked paperback, a copyedited proof, and a web annotation.
The question, then, is not whether a mark counts as annotation. The more useful question is: what relationship does this mark establish, under what conditions, and with what evidence?
The marginal transaction
A marginal transaction is an observed or reconstructed relation among seven conditions: a mark, a surface, a base passage, an actor, a candidate purpose, an audience, and a time condition. The word transaction is intentional. It directs attention away from a mark as a private object and toward an act that connects a text with someone, somewhere, and sometime. It also prevents the analyst from treating the mark’s graphic form as a complete explanation.
| Condition | Question for the reader | Why it cannot be skipped |
|---|---|---|
| Mark | What is physically or digitally present: a word, line, sign, color, drawing, reply, or correction? | Similar forms can serve different functions, and different forms can serve one function. |
| Surface | Is the act in a side margin, gutter, flyleaf, interlinear space, printed apparatus, screen layer, or separate pane? | Location affects access, direction of attention, durability, and the relation to the base text. |
| Base passage | What word, region, image, or document state is addressed? | A general page note is not the same kind of relation as a claim-level correction. |
| Actor | What evidence identifies the maker: signature, hand, account, institutional role, or none? | Identity confidence is not the same as authority, ownership, or motive. |
| Candidate purpose | Is the likely action correction, navigation, citation, ownership, critique, instruction, play, or something else? | Purpose is an inference that should remain revisable. |
| Audience | Is the mark private, directed to a collaborator, meant for a future holder, visible to a class, or public? | The same words perform differently once an audience changes. |
| Time condition | Was it made during production, later reading, revision, or an unknown interval? Is it tied to a particular version? | A surviving trace does not automatically record the original publication or reading event. |
The contemporary Web Annotation Data Model makes a related analytical separation explicit. It treats an annotation as a structured relationship among resources, typically including a body and a target, and gives importance to the annotation’s motivation. Its model can identify a target as a specific segment or state of a resource rather than merely a whole page.[1] This does not turn a medieval border or a pencil mark into a JSON object. It does, however, remind us that a contemporary system must distinguish relations that human readers also need to distinguish: what is being addressed, what is being added, who or what created the relation, and which state of a text is at issue.
A standard cannot determine why a historical person made a mark. Nor can a mark, by itself, establish which interpretation a later observer prefers. The analytical value of the marginal transaction is that it places observation and inference in different boxes. “An unsigned word is written in brown ink beside a paragraph” is an observation. “The book’s owner rejected the paragraph” is an inference. It may be a plausible inference; it is not the same kind of statement.
The margin has never had one job
A border that asks the reader to look again
Margins are often imagined as supplementary territory, separate from the text’s “real” content. A Bodleian case shows why that division can mislead. Charlotte Ross describes two fifteenth-century manuscripts in which decorative borders and initials carry possible reading instructions or devotional messages. In MS Douce 322, a border’s vine-like forms resemble repeated nota abbreviations, a widespread sign directing a reader to take notice. In MS Rawl Liturg. e.3, wording is embedded within an illuminated initial at the opening of a psalm.[2]
The historical interest of these pages is not that they prove every decorative border is secretly a note. Ross offers alternative interpretations for the embedded devotional phrase, including a possible personal motto, and closes with a direct warning against treating every material detail as symbolic.[2] That restraint is as valuable as the example itself. The page teaches at least three things at once. First, spatial placement can direct a reader’s attention before the body text begins. Second, a visual element can be decorative and functional without becoming transparent evidence of an individual’s intention. Third, an edge-of-text feature may be part of the production design rather than a later reader’s intervention.
The working question for such a case is not, “What did the margin mean?” It is: Which functions does the evidence make plausible, and which social story would exceed that evidence? In the Margin Function Atlas, the mark form, surface, target, and time condition would be recorded before its possible devotional or navigational function is proposed.
A margin that contradicts, corrects, and plays
The Kenneth Spencer Research Library’s Marvelous Medieval Marginalia exhibition makes another correction to the single-purpose account. Its curatorial essay reports a copy of Lactantius’s Divine Institutes in which an early reader wrote falsum (“false”) and struck through portions of the text. The curator carefully leaves the mark’s force unsettled: it may express disagreement, or it may assert that material was spurious or misattributed.[3] This is not a technicality. A margin can contest a proposition, contest a textual attribution, or perform a gesture whose authority is impossible to recover from the page alone.
The same exhibition describes additional functions: scribal faces and decorative play in one manuscript, as well as copious notes and images in a 1540 printed book. The point is not that doodling is merely a charming interruption to serious reading. A doodle can be unrelated to the base text, visibly responsive to it, or evidence of a scribal working condition that a later scholar cannot fully reconstruct. The exhibition also provides a useful continuity claim: print altered the conditions of a page but did not end readers’ and owners’ interventions in it.[3]
The distinction matters because words such as commentary can launder uncertainty. To call falsum “commentary” is permissible only if the word is understood broadly enough to include competing interpretations of what is being contested. To call it evidence that a particular reader rejected the author’s worldview would be a leap. The Atlas therefore requires a separate authority signal field: a mark may look corrective or forceful without proving that it was an authorized correction, a final decision, or a settled belief.
A book as a surface of possession, play, and use
A recent peer-reviewed study of children’s marginalia in rare-books collections broadens the range again. Kuecker and colleagues describe an uneasy special-collections history in which marks may be treated either as defacement or as evidence of past use. Their study does not identify one natural category called “marginalia.” It discusses ownership marks, notes, lists, drawings, imaginative play, coloring, short original writing, homework notes, and other traces that are differently related to a book’s printed text.[4]
This research has a methodological consequence. The researchers identify provenance as the most helpful measure for determining a likely creator, while also recognizing that the maker of a mark is sometimes impossible to know. They used indicators such as collection context and internal description to locate potential examples, while acknowledging that common cataloguing practices often make such traces difficult to discover.[4] A margin’s silence can be a problem of survival, cataloguing, acquisition history, or prior restoration—not proof that no reader acted there.
The point is not to treat all uncredited marks as equal cultural evidence. Rather, the paper asks a reader to avoid a false choice between dismissal and romance. An unremarkable shopping list in a book may tell us little about the argument on the facing page, but it may still be evidence of possession, domestic use, or the book’s availability as blank paper. A named inscription may tell us more about a chain of custody than about reading comprehension. A correction may tell us more about a working editorial relation than about the future printed text. These are not failures of the marks. They are differences in what kind of question the marks can answer.
| Reduction to avoid | What the evidence instead supports |
|---|---|
| “A mark in a book is a reader’s response to the text.” | A mark may be an ownership claim, a production feature, a navigation sign, a correction, play, a non-textual use of paper, or an uncertain combination. |
| “A forceful mark proves an author’s or reader’s belief.” | A forceful mark can signal disagreement or authority, but its maker, target, timing, and institutional standing may remain unknown. |
| “Print ended the active, alterable page.” | Print constrained some page conditions while readers and owners continued to add marks in individual copies.[3] |
| “A digitally precise target removes interpretive difficulty.” | A precise target can clarify what was selected, while purpose, audience, identity, and later circulation still require separate analysis.[1] |
The same passage across surfaces
The four copies of the committee sentence can now be read as four marginal transactions rather than four instances of a generic habit. The following comparison is deliberately constructed for teaching. It should not be confused with a claim about a particular historical object.
| Version | Most defensible observation | Candidate function | What remains unknown without more context |
|---|---|---|---|
| Pencil question: “who reviews?” | A short pencil note sits beside the phrase after review. | Comprehension query, revision prompt, or discussion preparation. | Whether it was private, addressed to an author, written before or after publication, or intended for another reader. |
| Editorial caret | A conventional correction sign relocates a phrase. | Proposed syntactic revision. | Whether the revision was accepted; whether the mark came from an authorized editor; what house style or version it belongs to. |
| Printed pointing hand | A printed sign directs attention to review. | Navigation or emphasis. | Whether it was placed by author, printer, editor, or designer; whether it signals a legal, rhetorical, or instructional priority. |
| Shared digital comment | A highlight targets committee and a comment requests a name. | Collaborative query or editorial instruction. | The audience setting, reviewer role, notice path, permanence, or whether the author acted on it, unless platform information is also preserved. |
The digital example has some affordances that paper may not visibly retain. The W3C model can record a relationship among a body, target, motivation, agent, and state; it also aims to allow annotations to be shared or moved across systems, including in private use.[1] But formal recordability is not the same as social clarity. A data model can allow a field for an agent; it cannot guarantee that the account belongs to the person the analyst assumes. A saved comment can preserve a target; it cannot restore an audience setting that changed, a notification that was never seen, or the social stakes of a conversation at the moment it occurred.
Contemporary reading research reinforces the need to distinguish function from medium. A 2024 Association of College & Research Libraries synthesis describes studies of social annotation in educational settings in which student comments included text-to-text connections, text-to-practice connections, and monitoring for understanding; in the reported case, comment patterns changed during the course from repetition or agreement toward more additive or challenging contributions.[5] That is evidence of differentiated activity in a particular learning context. It is not evidence that every digital note becomes dialogue, nor that all forms of shared annotation improve reading. The same synthesis reports that format, access, task, device, and the difficulty of organizing notes shape digital-reading experiences.[5]
The lesson is not “paper is private and digital is social.” A printed book can carry a public apparatus, an institutional proof mark, or accumulated use by several holders. A digital note can remain private, unread, or disconnected from its intended conversation. The teaching question is more exact: What does this surface make visible, durable, addressable, or shareable, and which of those conditions are actually evidenced in this case?
The Margin Function Atlas
The Margin Function Atlas is a field instrument for editors, students, researchers, and tool designers. It is not a machine for discovering motive. It is a guardrail against treating a provisional function as an observed fact.
The instrument begins with description. Only then does it request a ranked function hypothesis. Its final column requires the analyst to name the inference that would be unsafe and the contextual evidence that would reduce the risk.
| Field | Record | Example of a restrained entry |
|---|---|---|
| 1. Mark form | The visible or digitally rendered sign. | “Pencil question mark plus six words.” |
| 2. Surface and location | The material or interface position and relation to the base text. | “Outer margin beside line 12; no line crosses the text.” |
| 3. Base passage / target | The exact or bounded object addressed. | “Phrase ‘after review.’” |
| 4. Actor evidence | Signature, account, hand comparison, role, or unknown status. | “Unsigned; hand differs from body text; actor unknown.” |
| 5. Time condition | Date, version, layer, sequence, or uncertainty. | “Date unknown; ink appears later than the base text, but no dating evidence is available.” |
| 6. Candidate function | One or more hypotheses, not a single declared truth. | “Possible comprehension query; competing hypothesis: editorial query.” |
| 7. Audience and visibility | Private, collaborative, public, future holder, or unknown. | “No explicit audience evidence; future-reader use is possible but unconfirmed.” |
| 8. Persistence and transfer | Erasable, copied, version-bound, exportable, catalogue-described, or otherwise preserved. | “Pencil mark survives in present copy; prior erasure cannot be assessed.” |
| 9. Authority signal | Whether the mark appears to request, propose, direct, or finalize an action. | “Caret signals a proposed correction; acceptance status unavailable.” |
| 10. Interpretive risk / needed context | The unsafe inference and the evidence required to test it. | “Do not infer ownership; seek inscription, acquisition record, or hand comparison.” |
Applied to the Spencer Library’s reported falsum case, the Atlas yields a disciplined result. The mark form is a word accompanied by struck-through lines. Its location is adjacent to a passage in MS C61 as described by the exhibition. The actor is an “early reader” only in a broad curatorial category. Its candidate function may be disagreement or a claim about spurious attribution. Its authority signal is strong in appearance but indeterminate in institutional effect. The interpretive risk is clear: it would be unwarranted to say that the maker rejected the whole work, held an authorized editorial position, or intended the annotation as a final textual correction.[3]
That result may feel less dramatic than a story about a defiant medieval reader. It is better evidence. The instrument does not silence interpretation; it makes the cost of interpretation visible.
What the trace cannot tell you
The strongest counterweight to marginal enthusiasm comes from archival thinking. Anne Gilliland-Swetland describes archival evidence as inseparable from context: provenance, original order, custody, aggregation, preservation activity, and the relation among materials all affect what a document can reliably show. She emphasizes that evidential integrity depends on tracing contextual relationships rather than retrieving isolated information alone.[6]
A bound book is not automatically an archival fonds, and marginalia should not be forced into archive terminology as though that solves attribution. The analogy is narrower and useful: do not sever a mark from the conditions that give it meaning. If a page was trimmed, rebound, digitized selectively, described minimally, inherited without acquisition records, or viewed apart from nearby marks, the analyst has lost information that may matter. If a digital comment is exported without its target version, audience setting, identity context, or moderation history, an equally serious contextual loss can occur.
This is also a privacy principle. A marginal trace can become more public as it travels: a schoolbook enters a special collection, a private comment is included in an archive, a screenshot leaves a limited discussion space. The existence of a trace is not a blanket permission to narrate its maker’s inner life. When an analyst can identify a person, the ethical question grows sharper, not smaller. What does the collection record establish? What information was likely intended for a particular audience? What would publication add to the original exposure?
The practical consequence is modest but demanding. Treat provenance and audience as evidence categories, not as afterthoughts. If the answer is unknown, write unknown. A mature reading of margins does not turn every gap into an invitation for confident explanation.
Interpretive restraint test
Before drawing a conclusion from any mark at the edge of a text, apply this test.
| Test question | A responsible answer looks like |
|---|---|
| What can I literally describe? | Form, location, target, visible relation to the base text, and material or technical state. |
| What function is plausible? | A ranked, conditional hypothesis with at least one alternative where the evidence allows it. |
| Whose act do I think this is? | A statement of evidence level: named, probable, possible, or unknown. |
| What authority does the mark carry? | A distinction between visual force, institutional role, and proof of final textual effect. |
| What context would change my interpretation? | Handwriting evidence, cataloguing data, version history, audience setting, custody record, or neighboring marks. |
| What should I decline to claim? | Any motive, biography, final decision, or social meaning that the surviving trace does not support. |
Margins are working space because they make room for more than one operation around a text. They are not simply the place where the reader speaks. Sometimes they carry a reader’s question; sometimes they direct attention, assert possession, host a correction, display a production decision, hold a private trace, or reveal a later system of description. The task is neither to celebrate every mark nor to dismiss it as noise. The task is to identify the transaction, preserve the uncertainty, and let the evidence determine how far interpretation may travel.
References
[1]: https://www.w3.org/TR/annotation-model/ “W3C Web Annotation Data Model,” W3C Recommendation, 23 February 2017.
[2]: https://historyofthebook.mml.ox.ac.uk/2020/12/15/hidden-in-plain-sight-secret-messages-in-manuscript-marginalia/ Charlotte Ross, “Hidden in Plain Sight: Secret Messages in Manuscript Marginalia,” History of the Book, University of Oxford, 2020.
[3]: https://blogs.lib.ku.edu/spencer/marvelous-medieval-marginalia/ Eve Wolynes, “Marvelous Medieval Marginalia,” Kenneth Spencer Research Library Blog, University of Kansas, 2025.
[4]: https://rbm.acrl.org/index.php/rbm/article/view/27062/34940 Elliott Kuecker, Katie Grotewiel, and Zoe Thomas, “If This Book Should Chance to Roam: The Importance of Children’s Marginalia in Rare Books Collections,” RBM: A Journal of Rare Books, Manuscripts, and Cultural Heritage, vol. 26, no. 2, 2025.
[5]: https://acrl.ala.org/dols/2024/07/top-5-articles-on-digital-reading-and-annotation/ Brittni Ballard, “Top 5 Articles on Digital Reading and Annotation,” Digital Scholarship Section Blog, Association of College & Research Libraries, 2024.
[6]: https://www.clir.org/pubs/reports/pub89/archival/ Anne J. Gilliland-Swetland, “The Genesis and Rationales of Archival Principles and Practices,” in Setting the Stage: Introduction to the Archival Paradigm, Council on Library and Information Resources, 2000.