A paragraph moved by hand
Constructed opening claim. “I want to move this paragraph by dragging it.”
The request is easy to picture. A paragraph is visible in an outline or page-like canvas; the writer points, drags, releases, and sees the new order. The action seems more direct than describing a location, typing a command, or opening a structural panel. It may be.
But the phrase “move this paragraph” can conceal a difficult editorial operation. Is the paragraph a spatial block, a heading’s child, a numbered list item, a quoted source, a shared revision with comments, a footnote-linked claim, or a part of a larger argument? Does moving it also move notes, citations, tracked changes, accessibility labels, and references? Is the desired operation “put it visually here” or “make it the second reason after the counterexample”? The first is spatial. The second is relational and symbolic.
Direct manipulation is powerful because it lets people operate on visible representations with immediate feedback. Shneiderman’s classic account emphasized visible objects, rapid incremental and reversible actions, and physical actions in place of complex command syntax. [1] This paper does not turn those features into a universal standard for manuscript editors. It uses them to ask a more precise question: when is a visible representation a good stand-in for the editorial relation a writer is trying to change?
Hutchins, Hollan, and Norman offered a cognitive account of direct manipulation that distinguished semantic and articulatory distance and considered limits as well as advantages. [2] Their analysis is especially helpful for manuscripts because the feeling of “handling text” can be direct in one respect and indirect in another. A drag can be close to a spatial intention while still be a poor way to express a complex global condition.
This paper introduces the Manuscript Operation Fit Grid, an original Wexley teaching instrument. It records one operation, its relevant representation, the intended relation, direct-spatial affordance, symbolic or abstract control, reversibility, accessibility condition, and evidence need. It cannot score interfaces, determine accessibility, prove efficiency, or dictate a tool. Its purpose is to preserve the difference between moving a visible object and editing a structured relation.
Directness has more than one distance
A spatial operation can reduce the distance between intention and action when the object of interest is visible and the target relation is shown faithfully. Rearranging cards in a chapter plan is an intuitive example: a writer sees sections, moves one, and immediately sees a new sequence. The representation is close to the intended question, “What follows what?”
The same action becomes less direct when the representation hides the relevant structure. Dragging one paragraph through a manuscript may appear to move a block, yet the writer may actually be deciding:
- whether a definition belongs before a claim or after an example;
- whether a quotation should travel with its citation and explanatory context;
- whether a heading should carry two subsections with it;
- whether a shared comment should remain attached to a specific sentence;
- whether an instruction applies to all instances of a term; or
- whether a new sequence preserves the argument rather than merely a visual order.
These are not reasons to ban dragging. They are reasons to ask what the operation is trying to express.
Hutchins, Hollan, and Norman describe directness through the relation between user intention and system representation, not merely through a mouse, hand, or touch surface. [2] A command can sometimes be more direct than a drag if it states the intended relation with less ambiguity: “move this subsection under this heading,” “replace this term across the document with review,” “show comments from this reviewer only,” or “keep this citation attached to the paragraph.” Conversely, a visible manipulation can be clearer than a command when the intended relation is essentially spatial: resize a figure, reposition a callout, or arrange visible sections in a narrative sequence.
The object is not to place direct manipulation and abstraction on opposite sides of a quality scale. They often serve different parts of the same editorial task.
A manuscript has several simultaneous representations
The phrase “digital text” can imply one surface. A manuscript is usually several representations at once:
| Representation | It makes available | It can hide |
|---|---|---|
| Page-like flow | Rhythm, local reading order, visual space, breaks, emphasis. | Larger structural and semantic relations. |
| Outline | Heading hierarchy, sequence, section scale, navigation. | Sentence-level detail, citation attachment, prose rhythm. |
| Selection | The current unit of operation. | Whether the unit is too small or too large for the intended relation. |
| Comment and revision layer | Review conversation, proposed alternatives, status. | The final reading path or cross-document context. |
| Search or query result | Repeated terms, patterns, names, style problems. | The local argument and exception cases. |
| Structured metadata or markup | Roles, references, identifiers, semantics, reusable fields. | Immediate visual meaning for a reader. |
No one representation should be expected to do all editorial work. A flat page can make a global replacement feel arbitrary; a query result can make a sentence’s rhetorical role disappear. A card-based outline can make section movement easy while detaching it from prose transitions. A document editor becomes more capable when it lets a writer move among representations without claiming that one is the manuscript itself.
Novick and colleagues’ text-editor prototype extended direct reference and conversational forms of reference through direct-manipulation interaction. [3] The work is a prototype discussion, not a general usability result. It is useful because it rejects a simplistic choice between pointing and language. Text work routinely involves reference: this phrase, that occurrence, the third argument, the paragraph after the example. A well-designed operation has to make the referent and intended result legible.
Spatial power and symbolic control are complementary
Chugh and colleagues describe direct manipulation and programmatic systems as offering complementary strengths in their work on SVG: immediate visible feedback on one side, complex reusable abstraction on the other. [4] Their domain is graphics and programming, not authoring. Still, the complementary relation is valuable for manuscript work.
Consider a heading hierarchy. A writer may benefit from dragging a visible section card to rethink sequence. The same writer may need a symbolic command or clear structural action to move a subsection and all its descendants, preserve cross-references, or apply a change across several locations. The first operation makes a spatial plan visible. The second expresses a relation that may not fit on a canvas.
Likewise, a writer can drag a quotation into a new location, but might need a separate control to preserve its source relation, update a note, or mark it for verification. The drag is not defective because it does not do everything. It becomes risky only when an interface represents it as though every hidden consequence has been understood or resolved.
The useful question is not “Can a text editor be direct?” It is:
Which parts of this operation are well represented by visible spatial movement, and which parts need an explicit relation, command, preview, confirmation, or reversible path?
The Manuscript Operation Fit Grid
The Grid is a non-validated Wexley teaching instrument. It records a single operation in a defined manuscript state. It is not a usability benchmark, a product review, a shortcut guide, or an accessibility audit.
| Field | Record this | Do not infer |
|---|---|---|
| 1. Operation boundary | The selected unit, intended change, document state, and user goal. | That a named gesture represents the entire task. |
| 2. Relevant representation | Page flow, outline, selection, comments, search, metadata, or another view. | That the visible representation preserves every hidden relation. |
| 3. Spatial relation | Position, size, adjacency, hierarchy, grouping, or order that a direct action can show. | That every editorial relation is spatial. |
| 4. Symbolic or abstract relation | Scope, condition, identity, reference, pattern, permission, or rule involved in the operation. | That symbolic control is inherently difficult or indirect. |
| 5. Direct-manipulation path | What the user can visibly point to, move, resize, group, or reorder and the feedback received. | That the action is safe, complete, or reversible without evidence. |
| 6. Explicit-control path | Commands, parameters, preview, confirmation, search, markup, or structured choices needed. | That a command is necessarily less intelligible. |
| 7. Reversibility and inspection | Undo, preview, change history, comparison, and visible consequences. | That undo repairs every collaboration or interpretation problem. |
| 8. Accessibility condition | Alternative inputs, target scale, focus, screen-reader, keyboard, cognitive, or assistive-technology questions. | A conformance or usability verdict. |
| 9. Next evidence and non-conclusion | A concrete observation or test and what the Grid cannot decide. | Interface quality, efficiency, or adoption choice. |
The Grid does not require every operation to have both a drag and a command. It asks whether the available representation matches the actual relation. A spatial task may need no abstraction. A global or semantic task may need no canvas. A complex task may need both.
A constructed manuscript move
The following example is fictional. It describes no product, system, or observed user.
Initial claim: “Dragging is the most natural way to reorganize this chapter.”
Bounded operation: A fictional author wants to move a two-paragraph counterexample from the end of a section to immediately after a definition. The counterexample has one attached comment and a citation marker that refers to a source note.
| Grid field | Constructed record | What remains open |
|---|---|---|
| Operation boundary | Move two paragraphs and preserve the rhetorical relation to a definition, comment, and source marker. | Whether the move improves the chapter argument. |
| Relevant representation | The author uses an outline to see section order and page flow to see local transitions. | How comments and citations will behave in the chosen system. |
| Spatial relation | The counterexample should occur immediately after the definition. | Whether immediate adjacency is the whole editorial goal. |
| Symbolic relation | Citation attachment, comment location, heading hierarchy, and argumentative reason for the move. | Whether all hidden relations travel automatically. |
| Direct path | The author can drag a visible block in an outline and see a changed order. | Whether the block includes the correct paragraphs and descendants. |
| Explicit path | The author may need a move-under-heading command or preview to inspect citation and comment treatment. | Which interaction form is preferable for every user. |
| Reversibility | The move can be undone; before/after reading can expose a broken transition. | Whether undo restores a collaborator’s interpretation of a change. |
| Accessibility condition | The operation needs a keyboard-accessible alternative and a visible focus path; drag alone is not sufficient. | Formal WCAG conformance or accessibility outcome. |
| Next evidence | Inspect the moved comment and citation in a representative test. The Grid cannot choose an editor. | A product ranking or performance claim. |
The example does not declare the drag wrong. It divides the task into the part that direct movement makes visible and the part that still needs explicit inspection.
Scope is an editorial relation, not a screen size
Selection seems like a simple matter: choose a word, sentence, paragraph, section, or document. But scope often changes what the action means. A single deletion is different from a global replacement; a paragraph move is different from moving all children of a heading; accepting one tracked change is different from accepting a class of changes. The more a result depends on a pattern, condition, or hierarchy, the less a spatial surface alone can describe the intended scope.
This does not mean that explicit controls must be obscure. An interface can make scope inspectable by showing a preview, naming the units affected, offering a comparison, or allowing a writer to narrow or broaden a selected relation before committing. It can also make a direct action safer by exposing what travels with the visible object: comments, citations, headings, figure anchors, cross-references, and revision markers.
The Grid’s separate symbolic-relation field helps a writer notice when they are acting on a visible block but trying to change a rule. “Replace this phrase wherever it refers to the protagonist” is not a pure spatial operation. “Move the opening scene after the title” may be spatial at one level, but becomes structural if the scene includes nested sections and associated notes. Naming this difference makes a system more legible, not more abstract for its own sake.
Preview is a form of directness when consequence is not visible
Immediate feedback is often cited as a strength of direct manipulation. In a manuscript, the immediate visible result may still be incomplete. A page can show that a paragraph moved, while hiding that an endnote broke, a comment detached, a link target changed, or a table of contents needs to update.
A preview, change summary, or reversible comparison can therefore be more direct to the author’s true goal than a visually instantaneous action alone. It brings a consequence into the representation before the writer has to discover it later. The purpose is not to add confirmation dialogs everywhere. It is to choose inspection proportional to the operation’s hidden relations.
The important comparison is not “gesture versus dialog.” It is whether the interface lets the person see enough of the consequence to decide responsibly. A drag that previews the destination and affected structure may support direct handling; a command that clearly lists all affected occurrences may do the same. Both can be opaque if they hide the relation that matters.
Use the Grid to design alternatives, not to police preference
One writer may prefer an outline drag, another a keyboard command, another an accessible tree view, and another a query-like structural operation. The Grid should not turn those differences into a contest of sophistication or an assumption about “natural” behavior. It can reveal that each person is reaching the same editorial relation through a different path.
For a shared authoring system, this leads to a constructive design question: can the same underlying operation be exposed through more than one representation, with the same result and a readable record of what happened? If the answer is no, the next question is why. The restriction may be real, but it should be explicit rather than hidden behind the claim that one interface is inherently direct.
When views disagree, the disagreement is evidence
A manuscript can look settled in one view and unresolved in another. An outline can show an elegant sequence while the page view reveals a broken transition. A page can look visually balanced while a structural view reveals a heading level that is wrong. A comment layer can make an issue feel active while the clean reading view shows that the reader will not encounter its context. This is not always a defect. It can be the work object answering back through more than one representation.
The Grid can record these disagreements rather than trying to erase them. If a direct manipulation makes one view clearer but another relation less visible, the next step may be a comparison rather than a different gesture. A writer can rearrange an outline, then read the affected passage in flow. A system can show a selected object in the page view and its hierarchy in a structure view. A production path can surface an explicit warning when a visual move affects a nonvisual reference.
The aim is not to force writers through every view before every change. It is to recognize that the “right” representation depends on the relation under edit. Direct handling can be immediate and still need a second view for confirmation. Symbolic control can be exact and still need a page-like preview for craft. Their relationship is iterative, not adversarial.
This is especially important for shared documents. Another editor may encounter the change through a different representation and need to understand what the operation did. A visible, inspectable record of the relation can make a spatial move legible without turning every action into a technical lecture.
Direct does not mean hand-only
Direct manipulation is sometimes equated with a hand on a screen. That shortcut can exclude people who use keyboards, switch devices, voice systems, screen readers, or alternative pointers. WCAG 2.2 includes guidance relevant to dragging movements, target size, focus, and alternate interaction paths. [5] It does not tell a native manuscript editor how to work, nor does it certify an interface that happens to offer a keyboard shortcut. It reinforces the basic point: a direct-manipulation path should not become the sole path to an editorial operation.
The issue is conceptual as well as accessible. “Handle the text” can mean point at an object, name a relation, select a pattern, navigate a hierarchy, or invoke a reversible structured action. People need more than one way to make a change because documents themselves contain more than one kind of relation.
Let the manuscript answer back
Return to the opening claim: “I want to move this paragraph by dragging it.” After the Grid, a better account is possible:
The author wants to change a visible sequence relation, for which an outline drag may be useful. The operation also includes citation, comment, hierarchy, and argument relations that require inspection or explicit control. The Grid does not decide which editor is best.
That account protects the pleasure of direct handling without asking a drag to carry every editorial meaning. It allows spatial interaction to be powerful where it is faithful, and it keeps symbolic controls available where abstraction is the honest representation of the task.
References
[1] Ben Shneiderman, “Direct Manipulation: A Step Beyond Programming Languages,” Computer 16, no. 8 (1983): 57–69, https://doi.org/10.1109/MC.1983.1654471.
[2] Edwin L. Hutchins, James D. Hollan, and Donald A. Norman, “Direct Manipulation Interfaces,” Human–Computer Interaction 1, no. 4 (1985): 311–338, https://doi.org/10.1207/S15327051HCI0104_2.
[3] David G. Novick, Francisco Romero, Edgar R. Saenz, and Armando Sandoval, “Extending Direct Manipulation in a Text Editor,” SIGDOC (2002): 127–132, https://doi.org/10.1145/584955.584975.
[4] Ravi Chugh, Brian Hempel, Mitchell Spradlin, and Jacob Albers, “Programmatic and Direct Manipulation, Together at Last,” PLDI (2016): 341–354, https://doi.org/10.1145/2908080.2908103.
[5] W3C, Web Content Accessibility Guidelines (WCAG) 2.2 (2024), https://www.w3.org/TR/WCAG22/.