Abstract
“Go back to the sentence that changed the argument.” The instruction is simple until a reader has to carry it out. On a long roll, the reader must traverse a continuous surface and regain a place through visible column, remembered sequence, or an external cue. In a codex, an opening, a leaf, a running head, a page number, or a mark in the margin may help. In a printed reference work, an index or contents structure may name a route before the reader begins turning pages. In a structured digital publication, a table of contents, landmark, heading, link, page location, search query, and remembered reading position may all compete or cooperate.
This paper calls the resulting arrangement a reading wayfinding contract: the relation among a text’s form, its cues for orientation, its permitted routes, its reference affordances, its return mechanisms, and the work a reader must do to recover a relation. The contract is not a rating of media. It is a way to replace the blunt question “Is a screen better than a book?” with more exact questions. Does the reader need to follow sequence, locate a prior claim, compare distant passages, enter through a topic, resume after interruption, or share a stable reference? Which cue supports that action, and what does the cue cost or conceal?
The paper traces selected elements of the contract across roll, codex, printed reference apparatus, and structured digital text. It then offers the Route–Return Matrix, a practical instrument for documenting the reader task, orientation cue, route, reference affordance, return mechanism, friction, and access question in any publication. The object is not to crown the page or celebrate search. It is to learn that navigation has always been designed work—and that the route which makes a passage easy to find may not be the route that best preserves a reader’s sense of the whole.
1. The sentence that moved away
A reader closes a book, leaves a screen, or is interrupted halfway through an argument. Later, one sentence remains with unusual force: it revised the reader’s understanding of the subject, but the reader cannot quote it. The memory is not a bibliographic record. It is a rough relation—somewhere before the example, after the author’s rejection of a familiar explanation, perhaps near a short heading. The reader must now do three different things: return to an earlier location, identify the sentence, and recover the path from that sentence to the argument now underway.
Calling all three operations “navigation” hides the difficulty. A reader may be able to reach the beginning of a work without finding a particular passage. A reader may search a phrase without understanding its place in the sequence. A reader may know the page number but lose the route that led from one claim to the next. A good reading environment therefore does more than expose text. It gives a reader ways to enter, orient, traverse, reference, and return.
The practical stakes become clearer when the same request is placed in different reading forms. A scroll offers continuity: the text runs onward, and the reader moves through a material surface. A codex makes leaves and openings available as discrete physical fields. A printed scholarly book may supply a contents list, running heads, page numbers, an index, cross references, and reader-made marginal marks. A digital publication can combine a default sequence with a global table of contents, headings, landmarks, links, progress displays, page-location systems, and search. None of these arrangements is simply “more navigable.” Each makes particular reader work easier and relocates other work elsewhere.
This paper’s central question is therefore not which medium wins. It is this:
What must a reader be able to recover, and which material and informational cues make that recovery possible?
Answering requires a vocabulary more precise than old versus new, print versus screen, or scrolling versus paging. The proposed vocabulary is the reading wayfinding contract.
2. The reading wayfinding contract
A wayfinding contract describes the practical agreement a reading form makes with a reader. The form may not announce the agreement. A page number does not say, “I will help you cite and return.” A contents list does not say, “I will offer a map before you enter.” A search box does not say, “I may retrieve a word while leaving the shape of the argument to you.” But each feature assigns some work to a cue, a system, or the reader’s memory.
The contract has six parts.
| Contract element | Reader question | Example cues |
|---|---|---|
| Form | What physical or computational object is being encountered? | Roll, bound leaves, printed volume, reflowable publication, linked document. |
| Orientation cue | How do I know where I am? | Visible column, opening, heading, page number, progress display, landmark. |
| Permitted route | How can I move? | Continuous traversal, turning leaves, contents route, index, link, query, next/previous control. |
| Reference affordance | How can I name or share a location? | Quoted incipit, leaf or page locator, section label, index entry, stable link. |
| Return mechanism | After a detour, how do I get back? | Physical mark, page memory, running head, history control, bookmark, landmark, retained query. |
| Friction | What labor or loss accompanies the route? | Traversal time, ambiguous location, broken sequence, visual overload, inaccessible cue, loss of larger shape. |
The distinction between sequence and global navigation matters especially in a digital publication. EPUB 3.3 defines a package document whose spine supplies a default reading order, while the EPUB navigation document contains human- and machine-readable global navigation information such as a table of contents.[1] A work can therefore have an order without supplying an adequate map, or it can provide a map whose entries do not faithfully express that order. The W3C’s EPUB accessibility guidance emphasizes that a table of contents conveys structure and ordering as well as links; it helps readers locate where they are, go where they intend, and return to earlier locations. It advises that entries generally follow the publication’s linear order unless there is a clear reason not to.[2]
This is a contemporary standards observation, not a claim about all reading systems. The important teaching point is older and broader: a sequence is not the same thing as a route through a sequence. A reader may be carried forward in an orderly way and still be poorly supported when asking, “Where was that?” or “How does this section relate to the last one?”
3. The roll: continuity without a page address
A roll presents text as continuity. The reader encounters a surface that must be unrolled and rerolled; location is inseparable from traversal. This does not make the roll primitive, nor does it mean that a reader of rolls lacked devices for orientation. It means that the wayfinding contract is organized differently from a book made of bound leaves.
Book historian Gary Frost urges caution precisely where casual stories become too smooth. His account of early codex adoption notes that discussions often credit codex form with better access to intermittent passages, but also says that this advantage “might not be apparent” if we could observe a skilled reader consulting scrolls.[3] He further notes that early books—both scroll and codex—lacked many later text-management devices, including word spacing, pagination, and punctuation.[3] The lesson is not that form is irrelevant. It is that a form cannot be credited with every later cue that readers now associate with it.
The history of page numbers makes the same point from another direction. A University of Pennsylvania institutional book-history illustration explains that a scroll’s continuous text lacks page breaks in the modern sense, while codex leaves introduced a different organizational possibility.[4] Yet page numbering did not appear automatically with the codex. Binding signatures helped artisans assemble gatherings, but they were not necessarily reader-facing locators; readers and producers developed headings, colored initials, chapter divisions, running aids, marginal signs, foliation, and pagination over time.[4] The relevant contrast is therefore not “scroll equals no navigation” and “codex equals navigation.” It is continuous traversal versus a changing set of page-based and paratextual reference possibilities.
A reader using a roll may rely heavily on local visual features, remembered sequence, opening and closing actions, a known portion of the surface, or external descriptive language. These routes can be quite effective for continued reading. They become harder to describe with a modern page address because the reader’s relation to the text is not stabilized by numbered leaves. That is a difference in reference affordance, not a proof that one form makes better readers.
The surviving record also limits the story we can tell. The British Library’s account of Papyrus 745, a small Latin parchment fragment dated to the late first or early second century, treats it as important evidence for early codex history while stressing that the available evidence is regionally uneven and open to revision.[5] The account notes the strong presence of early Christian codices alongside the continued use of rolls for much Greek literary writing, cautions against generalizing from Egyptian survival patterns, and presents travel portability as a hypothesis rather than a settled cause.[5]
That restraint is part of the instrument. A reader using the Route–Return Matrix should write “uncertain historical route” rather than filling a historical gap with a confident improvement story. The purpose of a navigation history is not to flatten past readers into defective versions of present users.
4. The codex and the emergence of reference work
The codex changes the reader’s field of action. Bound leaves create openings, allow a reader to hold several locations in relation, and make physical cues such as thickness, leaf position, and page spread meaningful. Later reference apparatus can build on that field: chapter heads, running heads, tables, indices, cross references, foliation, and pagination can become increasingly powerful ways to locate a claim or name a passage.
But a page is not a locator merely because it has edges. The page must enter a reference system. Page numbering, in the strict sense, numbers both recto and verso pages; foliation numbers leaves; signatures serve an assembly function; and a running head or decorative division can establish local orientation without being a stable citation address.[4] These are not minor technical distinctions. They separate a device that helps a binder construct a book from a device that helps a reader find a thought again.
The printed index makes the work of reference especially visible. The University of Georgia Press describes an index not as a mechanical word list but as analyzed and arranged entries with locators and cross references.[6] Its formulation matters because an index does not merely expose every occurrence of a word. It selects, groups, disambiguates, subordinates, and sends a reader toward related terms. A locator turns a general topic into a route; a cross reference prevents a reader from becoming trapped by the wrong label.
That practice exposes a broader principle: navigation is editorial interpretation. The contents list proposes a hierarchy. The running head declares which division is currently in view. The page number turns a physical opening into a communicable coordinate. The index decides what counts as a meaningful entry and what relation deserves a cross reference. Readers can, of course, resist or supplement these routes with bookmarks, dog ears, marginalia, inserted slips, memory, and conversation. The contract remains distributed among form, apparatus, and reader practice.
The codex is therefore not best understood as a static rival to the digital screen. It is a platform on which differing navigation systems can be constructed. A thin novel with no contents list and no index offers a very different contract from a legal commentary, a dictionary, a scholarly edition, or an illustrated atlas, even when all are bound books.
5. The screen: default order, global routes, and query
A structured digital publication can preserve a default reading sequence while offering routes that do not resemble turning leaves. EPUB 3.3 describes the publication as structured and semantically enhanced web content packaged for distribution. Its package document specifies a spine—the default order in which content documents are rendered—while its navigation document can provide a table of contents and other global navigation information for both human and machine use.[1]
This arrangement makes several things possible. A reader can begin with a sequence, move by heading, jump through contents entries, use a page-list or location mechanism where available, follow a link, search a word or phrase, and return through a reader system’s own history or bookmark behavior. The features should not be treated as redundant. A search query retrieves language; a contents list displays a designed hierarchy; a landmark points to major structural territory; a page location can help synchronize discussion across versions; a heading may convey rank within an argument.
The distinctions have access consequences. W3C’s EPUB accessibility techniques say that a contents list helps readers understand both structure and order, and warn that users can have difficulty locating themselves, moving where they intend, or returning when contents entries diverge from linear reading order.[2] The same guidance distinguishes EPUB landmarks from ARIA landmarks: they aid navigation differently, and one does not simply replace the other.[2] The paper does not turn that guidance into an accessibility certification. It instead treats the distinction as a wayfinding fact: a route available to one reading mode or system may be partial, absent, or differently exposed in another.
Search supplies an especially tempting false solution. It can make a remembered word immediately reachable even when a reader has no clear sense of page position. Yet a remembered word is not always available; the reader may remember an idea, a contrast, an image, or an argument’s function rather than its vocabulary. Search can also return a string without returning the reader to the argumentative path that made the string meaningful. The answer is not to reject search. It is to record what job search actually performs and what additional cue restores the larger shape.
6. One request, four routes
The following comparison is a constructed teaching scene. It does not claim that a late-antique reader, a Renaissance index user, and a contemporary phone reader face identical institutional conditions. It holds the task steady only long enough to make the contract visible.
A reader needs to recover an earlier claim, compare it with a later qualification, and then return to the argument in progress.
| Form | Likely orientation cue | Route to prior claim | Reference / comparison resource | Return problem |
|---|---|---|---|---|
| Roll | Visible portion, column, remembered sequence, external description. | Traverse the continuous surface toward the remembered region. | A known opening phrase or externally supplied description may matter more than a page address. | The reader must reconstruct the former stopping point after movement through the surface. |
| Codex without developed reference apparatus | Opening, leaf position, visual division, reader mark, remembered physical relation. | Turn or thumb through leaves; hold multiple openings in relation where the object permits. | Material opening and local visual cues can support comparison, but stable naming may remain limited. | A reader may retain one opening physically, but detour and resumption still require a cue. |
| Printed reference work | Page number, running head, contents entry, index term, cross reference. | Use a numbered or indexed route rather than traverse every intervening page. | An index can analyze and group a topic, while page locators permit citation and comparison. | The reader can return by page, but page coordinates alone do not explain the argumentative relation. |
| Structured digital text | Heading, contents item, landmark, progress/location indicator, visible search result. | Follow a contents path, navigate by heading/landmark, invoke search, or use a link/history control. | A query can retrieve wording; semantic structure can provide global orientation; a link can encode a selected relation. | The system may retain history, but return behavior and cue exposure vary by reading environment. |
The table’s purpose is not to make the last row look advanced. A reader may need a stable shared page reference in one setting and a broad semantic entry route in another. A page opening can support side-by-side comparison that a narrow screen makes awkward; a landmark can provide nonvisual global access that a visible page spread alone does not. A query can retrieve quickly while dispersing the reader into isolated occurrences. The right question is never merely, “What is faster?” It is, “What route preserves the relation the reader needs next?”
7. Teaching instrument: the Route–Return Matrix
The Route–Return Matrix turns that question into a practical observation exercise. Choose a real publication—a catalogue, novel, research report, textbook, manual, digital edition, or long web document—and choose one actual reader task. Do not start with a preference for an interface. Start with the task.
| Field | Prompt | Example observation |
|---|---|---|
| Navigation task | What is the reader trying to do? | “Find the claim that qualifies the opening argument, compare it with the later case, then return to the case.” |
| Form / condition | Which particular representation is being used? | “Reflowable EPUB viewed on a phone with enlarged type.” |
| Orientation cue | What indicates location? | “Section heading and a reading-system progress display; no stable print-page cue.” |
| Permitted route | How can the reader move? | “Contents entry, heading navigation, search, sequential swipes, bookmark.” |
| Reference affordance | What can the reader name, share, or revisit? | “Section title and quoted phrase; page location changes with reader settings.” |
| Return mechanism | What restores the earlier route after detour? | “Bookmark plus the heading hierarchy; search result alone is insufficient.” |
| Friction | What costs or ambiguities appear? | “Search finds the phrase, but the reader loses the prior sequence and must rebuild it.” |
| Access / test question | What needs direct checking? | “Are the hierarchy and return route usable in the intended reading systems and access modes?” |
The matrix is deliberately not a scorecard. A filled entry should not produce a number that declares a publication good or bad. It should reveal a decision. If a reader needs to compare two distant diagrams, the matrix may expose the value of stable page relations or multiple visible panels. If a reader needs to enter a long work by concept, it may expose the need for an index, meaningful headings, or semantic search. If a reader is likely to be interrupted, it may reveal that a progress display is less useful than a recoverable heading path and bookmark. If a reader cannot rely on visual page position, it may reveal an untested dependency on location alone.
The instrument’s final field is intentionally an access / test question, not an accessibility verdict. A matrix can identify a cue that may disappear, a route that may not be exposed, or a return mechanism that requires direct testing. It cannot certify the experience of every reader or replace standards-based assessment and reader research.
8. Why a universal verdict fails
The temptation to turn every navigation comparison into a verdict is strong. Pages seem to promise stable location. Scrolling seems to promise continuity. Search seems to promise retrieval. Yet the empirical record is conditional enough to make verdict language irresponsible.
An abstract-only 2022 study of Norwegian university students reading informational text on smartphones or tablets reported tendencies toward better integrated understanding and more strategic backtracking in paging conditions, while also finding that scrolling on a tablet could produce a more positive reading experience than some alternatives.[7] Its abstract does not authorize a broad claim: the readers, devices, informational task, and outcome measures matter.
A 2025 CHI Extended Abstract approached the question differently. Lee and Vogel compared scrolling and pagination on a smartphone with 100 valid crowdsourced responses, eighth-grade-level stories, and multiple-choice comprehension tests. In that experiment, they found no statistically significant differences in measured comprehension, reading duration, or workload between the scroll and page conditions.[8] Their discussion names limitations that the reader of any interface study should notice: genre, question format, working-memory variation, prior practice, and disagreement among earlier results.[8]
These sources do not cancel one another. They teach the same methodological lesson: a medium label is too coarse. The route is embedded in a task, an interface mechanic, a text type, a reader population, a device, a measure, and a history of learned use. The Route–Return Matrix is valuable because it stops an editor or designer from carrying a claim about one condition into every other condition.
9. A task-based navigation rule
The navigation rule is simple enough to use before a redesign, a publication specification, or a reader study:
Name the reader’s next act before judging the navigation cue. Then ask how the form lets that reader enter, orient, traverse, reference, and return.
The rule places a demand on any claim that a feature is better. Better for what task? Better for whom, under which representation? Better at initial sequence, at subsequent recovery, at non-linear comparison, at concept entry, at shared citation, or at a particular access route? A page number, heading, index, link, landmark, search field, bookmark, progress indicator, and running head may all be good answers to different questions.
A practical first exercise requires no specialized software. Take one short but structured work and give two people different tasks: one must resume after interruption; the other must locate and compare two distant claims. Ask each person to describe the cue they used, the route they took, the point at which they became uncertain, and how they returned. Fill two Matrix rows. The result will not tell a publisher which medium to choose. It will make the publication’s actual wayfinding contract available for judgment.
Boundary statement
This paper does not prescribe an interface, rank the scroll, codex, printed page, and screen on a single scale, determine whether one medium produces better comprehension, reconstruct a universal history of reading, or certify accessible navigation. It teaches a bounded historical and comparative vocabulary for route, return, reference, and recovery. A navigation feature becomes meaningful only when a reader’s task, condition, and limits are named.
References
[1] World Wide Web Consortium, EPUB 3.3 (W3C Recommendation, 2026)
[2] World Wide Web Consortium, EPUB Accessibility Techniques 1.2 (Group Note, 2026)
[3] Gary Frost, “Adoption of the Codex Book: Parable of a New Reading Mode,” Book and Paper Group Annual 17 (1998)
[4] “Page Numbers,” Cultures of the Book at Penn
[5] Cillian O’Hogan, “The Beginnings of the Codex,” British Library Medieval Manuscripts blog (2015)
[6] University of Georgia Press, “Indexing Guidelines”
[7] Rogier A. C. E. K. H. et al., “Movement, or Both? Investigating Effects of Screen Size and Text Movement on Processing, Understanding, and Motivation when Students Read Informational Text,” Reading and Writing (2022)
[8] Austin Lee and Daniel Vogel, “Is Pagination Better than Scrolling when Reading on a Phone?,” CHI EA ’25 (2025)