Abstract
A folder can make a persuasive story before anyone has read a line. One file is called draft, another copyedited, another proof, another final, and another sits in a repository history with a later timestamp and a reassuringly exact identifier. The sequence seems to say that raw writing became approved writing and then became publication. Sometimes it does. Often the available evidence is thinner. A filename is an assertion. A marked PDF may be a page proof, an exported working copy, or a reader’s annotation. A commit message saying final is a recorded message, not proof that the represented text was authorized for release. A clean document can be later than a marked document and still be only one unverified copy.
This paper calls such a representation a textual state: a text-bearing object, file, export, or record whose evidential meaning depends on the practice that produced it, its intended use, its documented relation to other representations, its authority signals, and the context through which it survives. The term is deliberately more modest than version when the relationship between objects is unknown. A textual state may be useful evidence without becoming the definitive text, the author’s intention, or the complete history of a work.
Drawing on bibliographical scholarship, a university special-collections finding aid, archival-description principles, university-press production guides, official Git documentation, and archival-authenticity writing, this paper introduces the Textual-State Evidence Map. The map teaches readers to separate what can be observed from what is claimed, what is documented from what is inferred, and what additional corroboration would be required before using words such as final, authorized, earliest, definitive, or authorial. It does not create a critical edition or prescribe archival retention. It makes a more immediate skill available: reading textual states without mistaking a sequence for a story. [1] [2] [3]
Working principle: A state can show that a representation exists and bears particular features. Its order, authority, purpose, and relation to other states must be evidenced rather than assumed.
1. The misleading sequence
Consider an explicitly constructed teaching packet for an imagined long-form essay. It contains five representations:
| File or object label | What a reader might too quickly conclude |
|---|---|
author-draft.docx |
This is the author’s earliest working text. |
copyedited_manuscript.docx |
Editorial work has been completed and accepted. |
pageproof.pdf |
This is a production proof ready for final correction. |
revised-proof.pdf |
This is the last approved state before publication. |
Repository commit message: final |
The text in this commit is the final authoritative version. |
The labels make a plausible line: authorial work, editorial work, typeset proof, correction, final release. But the packet alone does not establish that line. The first document could be a renamed copy created after editorial work began. The second could contain tracked changes that were never accepted. The PDF could have been generated for an internal layout check rather than formal proofreading. The revised proof could be a derivative saved by a freelancer outside the publisher’s system. The commit might record that one contributor called a state final for one branch, while a different state was subsequently transmitted to production.
This does not mean the packet is worthless. It means that each object must be read at the resolution the evidence permits. We can observe a filename, markup, page geometry, a timestamp displayed by an interface, a digital identifier, or a relationship recorded by a system. We may also have a finding aid, a production guide, correspondence, a repository policy, or a documented transmittal that helps interpret those observations. What we do not have automatically is an authorial narrative.
The difference is practical. A researcher who cites the pageproof.pdf as the “final text” may be overstating the evidence. An editor who receives a clean file named approved may need to know whether it was approved for copyediting, typesetting, proofreading, printing, or public release. An archive user who sees three closely similar copies may need to describe them as distinct representations before deciding whether one is an edition, issue, state, variant, transcript, scan, or post-publication copy.
For that work, this paper uses the phrase textual state. It means a representation in hand—not merely abstract text—that has a possible relation to other representations. The relation may be well documented, partly documented, disputed, or unknown. The phrase does not say that every file is a formal version, that every earlier state leads to a later state, or that all changes are authorial. It creates room for better questions.
2. State is not story
A textual state first gives us a thing to describe. It may be a physical copy, a handwritten manuscript, an annotated typescript, a galley, a bound volume, a scan, a transcription, a word-processing file, a PDF export, a repository object, or a commit record. It can bear visible textual, material, and technical features. It may carry a label that claims a position in a workflow. But its place in a history is a separate proposition.
| What can be observed in or with a representation | What may require external evidence | What cannot be concluded automatically |
|---|---|---|
| Text, visible markup, typography, paper or binding features, filename, identifier, message field, recorded location, or visible metadata. | Who generated it; for whom; when it entered a workflow; whether it derives from another state; whether an actor approved it; or whether a reported date concerns creation, copying, export, or accession. | Authorial intention, definitive authority, complete provenance, a stable “original,” or the reason every alteration was made. |
Bibliographical practice supplies a useful restraint. The Society of American Archivists defines descriptive bibliography as study of physical characteristics and bookmaking processes, and as close description of a work’s material features, including format, pagination, typography, illustration, paper, and binding.[1] The definition does not say that a single copy reveals an entire edition. Its attention to the object’s material form makes a more cautious claim possible: this is the copy, object, or representation being described.
Fredson Bowers makes the distinction sharper. Writing about analytical and descriptive bibliography, he explains that books can exist in editions, impressions, issues, and states, and that comparative examination matters to claims about their relations.[2] A cataloger may record a copy at hand; a descriptive bibliographer attempts to analyze material forms across more than one example and to distinguish the individual object from an edition-level account.[2] Historical terminology is contested and context-specific, especially outside print. Yet the methodological lesson travels well: an available representation is not automatically a complete account of the group to which it may belong.
The Bibliographical Society of America similarly separates enumerative listing from analytical, textual, and descriptive bibliography. Its educational resource presents bibliography as attention to recorded forms, material production, transmission, reception, and the relationships among texts that carry the same work.[3] That wider account is useful only if it does not dissolve differences. A list of files, a precise physical description, a comparative account of variants, and an argument about the intended text are related tasks, not interchangeable ones.
The Textual-State Evidence Map begins with this distinction. It does not ask, “Which one is the real version?” It asks three earlier questions: What representation is in hand? What claim am I about to make about it? What specific evidence would make that claim responsible?
3. A production packet is not one manuscript in five costumes
A finding aid can show why labels need their surrounding practices. The University of Iowa’s Alembic Press Archive describes materials connected to the design, production, and publication of printed books, ranging from holograph manuscripts through galley proofs, printing, binding, and distribution.[4] Across the archive’s item-level descriptions appear research notes, original and word-processed typescripts, autograph corrections, galley proofs, marked page proofs, dummies and prototypes, layout designs, printer’s notes, correspondence, final sheets, sample bindings, artwork, and finished copies.[4]
The finding aid gives a productive contrast, not a universal sequence. For Delittle—An English Wood Letter Manufacturer, it records a typescript and a word-processed typescript with autograph corrections, title-page designs, galley proofs, incomplete page proofs marked with typographic corrections, illustrations, a sample binding, and correspondence.[4] For another project, it describes a word-processed typescript with holograph corrections, a dummy/mock-up, page proofs, and a prototype binding.[4] Elsewhere the archive lists a printer’s copy of loose sheets alongside design notes, amended typescripts, and correspondence.[4]
These descriptions demonstrate that terms such as typescripts, galley proof, page proof, dummy, prototype, and final sheets name different material and operational situations. They do not tell us, without further examination, every decision that preceded the objects, all the directions in which influence moved, or the complete chronology of the work. A finding aid represents an archive’s description of aggregated materials. It is already interpretive and necessarily selective. It offers context; it does not eliminate the need to inspect specific objects or related correspondence.
University-press production guides illuminate the operational side of state labels. The University of Minnesota Press describes a sequence in which a final manuscript is transmitted to production, copyedited electronically, returned to the author with visible changes and queries, cleaned up for typesetting, and turned into typeset pages for proofreading and indexing.[5] It states that authors’ proof-stage alterations are limited to correction of errors and that wording or style changes are not allowed at that stage.[5] Cornell University Press likewise distinguishes a complete manuscript submitted to production, copyedited files with tracked changes, reviewed copyedited files, composition, PDF page proofs, corrected proof, revised proof, and the bound book.[6]
Neither guide provides a timeless definition of publishing. Both are institutional accounts of their own workflows. Their value is narrower and more useful: a representation’s meaning partly depends on the action it is prepared to support. A copyedited manuscript is not merely text after a draft. It is a file positioned for a particular form of review. A page proof is not merely a later-looking PDF. In these workflows, it is a typeset representation sent for proofreading and indexing under constraints on the kind of alteration expected. A corrected proof records a different action than a clean manuscript, and a bound book records a different production outcome again.
The historical and contemporary cases together show why the phrase version of the manuscript can conceal too much. A production packet is not one stable work wearing several costumes. It is a collection of representations tied to different operations: writing, responding, composing, correcting, comparing, manufacturing, describing, and distributing. To identify one state is not to narrate every relation among them.
4. Description is not decoration
If a file or object’s contents do not settle its status, where does context come from? Partly from description. The current Describing Archives: A Content Standard principles say that archival description exists to help people understand the past through traces in records.[7] The standard emphasizes that archival description should document lifecycle and administrative information not necessarily present in the records themselves: appraisal, custodial history, reformatting, restrictions, and other activities that might change a user’s understanding of the record’s content or context.[7]
That point matters because a textual state is never only a body of words. A document may appear in a folder because a donor copied it there, a repository exported it from a system, an archivist preserved it during transfer, or a publisher grouped it with production materials. Each circumstance can matter to a claim about relation or authority. DACS therefore treats records, agents, activities, and relationships among them as fundamental elements of archival description.[7] The relationship is not decorative metadata appended after the real work. It can be the evidence that makes a description useful.
DACS also complicates the wish for a simple original order. It notes that the order in which records arrive at a repository may not match their creator’s last-used order and that electronic records are commonly moved and maintained in file systems over their life cycles.[7] “Original order,” in this account, is not a privileged sequence that a reader can always recover by looking at folders. It is an intellectual construct that may illuminate activities and relationships if documented carefully.[7]
The Textual-State Evidence Map therefore separates three fields that are often collapsed:
| Field | Question it asks | A restrained example |
|---|---|---|
| State relation | What relation between representations is actually recorded? | “The finding aid groups this marked page proof with production files.” |
| Authority signal | What evidence says the representation was prepared, approved, or used for a specified purpose? | “A production guide identifies page proofs as material sent for proofreading.” |
| Uncertainty note | What is missing, contested, or only inferred? | “The filename says final, but no approval record or transmittal has been found.” |
The fields let a reader say something useful without pretending that a description is total provenance. “Grouped with” is not “created by.” “Sent for review” is not “approved for release.” “Later in a digital interface” is not “later in the author’s intellectual development.” Such phrases may sound cautious, but they are more informative than a falsely smooth story.
5. A digital snapshot has structure—but not omniscience
Digital histories can seem unusually authoritative because they expose identifiers, timestamps, messages, contributors, diffs, and graphs. Git’s official documentation shows why these fields are valuable and why their meanings must be kept technical. A basic git log entry can display a commit checksum, author name and email, date, and commit message.[8] Patch output can display textual differences associated with an entry. Graph output can display recorded branch and merge relationships.[8]
The documentation also distinguishes author from committer. The author is the person who originally wrote the work represented by a change; the committer is the person who last applied it to the repository.[8] That distinction is more than a software detail. It teaches the kind of question the Evidence Map asks. An author field, a committer field, a message, and a commit identifier are different claims with different meanings inside the system. They may record important technical and collaborative relations. They do not supply a complete account of authorship in the literary, editorial, contractual, or institutional sense.
Put a Git commit beside a marked page proof. The proof can visibly show typeset pages and corrections. The commit can visibly record an identifier, fields, an order within a repository history, and perhaps a patch. Both may be strong evidence for some claims. Neither one, in isolation, settles whether a text was authorized for publication, whether every relevant state survives, whether a contributor’s role is fully described, or what an editor intended.
The Council on Library and Information Resources’ Authenticity in a Digital Environment provides the useful counterweight. Its contributors note that authenticity carries different meanings across contexts, including meanings connected to originality, fidelity, integrity, provenance, context, and the intended use of information.[9] The report’s historical examples show that labels and initial descriptions can be challenged by internal or physical evidence; its discussion of digital objects warns that apparently available attributes and descriptive systems do not make authentication automatic.[9]
The report was published in 2000 and should not be treated as current technical forensics guidance. Its value here is methodological. A repository history, catalog record, file timestamp, or digital copy can be evidence. It can increase confidence when it is well documented and corroborated. But it does not make the map unnecessary. A technically structured graph tells us that particular records are linked in a particular recorded system. The map asks a further question: What does that relation establish for the claim we actually want to make?
6. The Textual-State Evidence Map
The Textual-State Evidence Map is a reader’s diagnostic sheet for a consequential representation. It is not a cataloging standard, a preservation schedule, a forensic examination, or a method for choosing a definitive text. It helps readers make claims at the scale the evidence supports.
| Map field | Reader records | Why it matters |
|---|---|---|
| 1. Representation in hand | Physical copy, file, export, transcript, scan, proof, or repository object; and the form actually accessible. | It prevents the reader from describing an unavailable original when only a derivative is present. |
| 2. Observed features | Text, markup, typography, filename, identifier, material features, visible metadata, or rendering. | Observation must be kept separate from interpretation. |
| 3. Claimed state label | Draft, proof, page proof, copyedited file, edition, snapshot, final, or another designation. | A label is evidence to evaluate, not an outcome. |
| 4. Generating practice | Writing, copyediting, typesetting, exporting, scanning, committing, binding, transcription, or unknown. | The practice that generated a representation gives the label practical meaning. |
| 5. Intended use and audience | Author review, copyediting, proofing, indexing, print manufacture, repository development, reader access, or unknown. | A state is partly defined by what action it was prepared to support. |
| 6. Relation to other states | Precedes, follows, derived from, compared with, contains a correction to, transmitted with, or relation unknown. | It prevents chronology from becoming guesswork. |
| 7. Authority signal | Signed approval, transmittal record, institutional instruction, documented role, publication designation, commit policy, or none found. | It separates a name on an item from evidence of authorized status. |
| 8. Context and provenance evidence | Finding aid, collection location, custodial history, correspondence, repository location, or descriptive record. | It makes surrounding evidence visible without mistaking it for certainty. |
| 9. Uncertainty or alternative reading | Missing states, unverified label, inconsistent date, possible copy/export, undocumented intervention, or incomplete record. | It turns limitation into an explicit part of responsible description. |
| 10. Corroboration needed | Compare copies, inspect material features, consult correspondence, check workflow policy, examine a repository record, or refrain from a stronger claim. | It identifies the next responsible step rather than forcing a conclusion. |
Apply the map to the constructed pageproof.pdf. The representation in hand is a PDF. Its observed features may include page numbers, running heads, typesetting, and a filename that says pageproof. The label suggests a state but does not establish one alone. The generating practice may be composition, export, or an undocumented later rendering; it remains unknown until a production record or related state supports the claim. The intended use might be proofreading and indexing if the PDF is accompanied by an institutional transmittal or fits a documented workflow. Without that evidence, the reader should describe it as a “PDF labeled pageproof” rather than a verified page proof.
If a second file called revised-proof.pdf contains fewer marks, the map records a possible relation. It does not automatically record that it superseded the earlier proof, was approved by the author, or became the printed book. A bound copy may show that publication occurred, but it still does not identify every decision in the path from proof to print. A comparison can reveal textual difference. It cannot, by itself, reveal why an alteration was made or who authorized it.
This is why the last two fields are not an apology for uncertainty. They are the instrument’s practical center. A researcher might need a publisher’s production correspondence. An editor might need an approval email or a current production-system record. An archivist might need a donor’s documentation, a transfer history, or material comparison. A person without access to corroborating evidence may still describe the representation honestly and stop short of a larger claim.
7. The restraint test
Use the Evidence Map only when a claim about status, relation, authority, provenance, publication, or textual history depends on a representation. It is not a requirement for every private save, temporary copy, or informal note. Its proportionality is part of its design. A writer need not turn ordinary drafting into an archival inquiry. The map becomes useful when the state will support a consequential handoff, publication decision, scholarly citation, preservation transfer, dispute, or historical account.
Before using a strong label, ask four questions:
- What status claim am I about to make? Is this a claim about a draft, proof, final state, edition, authorial copy, or published text?
- What in the representation itself supports the claim, and what comes from surrounding description? A visible feature and an archival relationship should not be treated as the same kind of evidence.
- Which relation is recorded, and which relation am I merely narrating? “Later in this folder” and “superseded by” are not equivalent statements.
- What would I need to see before using final, authorized, earliest, definitive, or authorial? The answer may be an approval record, a comparative bibliography, a publisher designation, a repository policy, material inspection, or a refusal to make the stronger claim.
Return to the constructed commit message, final. The responsible reader does not need to dismiss the state. It may be a well-identified, useful point in a recorded repository history. But its label is not its authority. The reader can say what the commit records, what relation it displays, what surrounding evidence supports its use, and where a broader story remains unproven. That is not a failure to know the text. It is the beginning of knowing what the text can honestly be made to say.
References
[1] Society of American Archivists. “Descriptive Bibliography.” Dictionary of Archives Terminology. https://dictionary.archivists.org/entry/descriptive-bibliography.html
[2] Bowers, Fredson. “The Function of Bibliography.” Library Trends 4, no. 4 (1956): 497–511. University of Illinois IDEALS. https://www.ideals.illinois.edu/items/5802/bitstreams/32025/data.pdf
[3] Bibliographical Society of America. “Bibliography Defined: Further Reading.” https://bibsocamer.org/about/what-is-bibliography/further-reading
[4] University of Iowa Libraries, Special Collections and University Archives. “Alembic Press Archive.” Finding aid. https://www.lib.uiowa.edu/scua/msc/tomsc900/msc882/alembicpressarchive.html
[5] University of Minnesota Press. “7. The Production Process.” Manuscript Preparation Guide and Production Overview. https://manifold.umn.edu/read/manuscript-preparation-guide-and-production-overview/section/9b98cb00-3860-479d-9c1f-fe7a48c23470
[6] Cornell University Press. “The Cornell Process from Manuscript to Bound Book.” https://cornellpress.cornell.edu/overview-of-the-publishing-process/
[7] Society of American Archivists, Technical Subcommittee on Describing Archives. “Statement of Principles.” Describing Archives: A Content Standard. https://saa-ts-dacs.github.io/dacs/04_statement_of_principles.html
[8] Pro Git. “2.3 Git Basics — Viewing the Commit History.” Git project documentation. https://git-scm.com/book/en/v2/Git-Basics-Viewing-the-Commit-History
[9] Council on Library and Information Resources. Authenticity in a Digital Environment. Washington, DC: CLIR, 2000. https://www.clir.org/wp-content/uploads/sites/6/pub92.pdf