EssayArcEssayArc / Research & pedagogy
Writing as Thinking: The Research Foundations of EssayArc
This is the web version of the PDF report, with the same content in a form that screen readers, magnifiers and small screens can use.
Download the PDF (12 pages; not tagged for screen readers)
An undergraduate writing methodology informed by leading university programs and published faculty scholarship.
The purpose of writing instruction is to develop a writer who can make, examine and defend choices.
EssayArc brings together Kwame J. Granderson’s Write to the Top methodology, a comparative review of university writing pedagogy, and a documented analysis of scholarly prose. Its central commitment is student authorship: AI supports inquiry and revision while students develop their own interpretations and work with human peer learning facilitators.
- 6 university programsPublished teaching principles
- 146 scholarly worksSource-audited benchmark corpus
- 4 disciplinary profilesAuthor-balanced comparisons
Prepared for Kwame J. Granderson, JD · September 11, 2026 · Research synthesis and methodology report • EssayArc implementation reviewed at September 11, 2026.
Research foundations / 01
A research-informed approach to developing writers
EssayArc’s strongest academic foundation is the convergence of three sources of knowledge: the teaching practices of leading undergraduate writing programs, Granderson’s explicit instruction in the craft of prose, and the observable patterns of published scholarly writing. Each contributes something different. Together, they support a coherent approach to helping students think critically and write deliberately. [1, 2, 3, 4, 6, 10, 11]
University pedagogy supplies the learning principles
The comparative review covers Brown, Duke, Swarthmore, Princeton, Carleton and Harvard. Their published materials emphasize purposeful inquiry, defensible claims, evidence and analysis, attention to readers, revision, and reflection. Brown’s peer-fellow model also offers a clear foundation for making the student-PLF relationship central to EssayArc. [1, 2, 3, 4, 6, 7]
Published scholarship supplies examples of accomplished practice
The StyleBot study examines 146 works across literary criticism, history, social research and legal scholarship. The collection includes identifiable works by faculty at Brown, Harvard, Duke, MIT, Stanford, Northwestern, Chicago, Princeton and UCLA. It provides a concrete reference for discussing sentence structure and disciplinary variation, rather than relying on an invented universal ideal. [11, 12, 17, 18, 19, 20, 21, 22, 23, 24, 25]
Write to the Top supplies an explicit craft vocabulary
Granderson connects thesis development, relationships among ideas, supporting arguments, sentence balance, transitions, vocabulary, rhetoric and conclusions. EssayArc translates these concepts into focused exercises and feedback that leave the intellectual work with the student. [10, 14, 15]
The academic case for EssayArc rests on its methods: evidence-based interpretation, deliberate craft, revision and conversation.
How this report was prepared
This is a purposive documentary review and implementation synthesis. It compares official program guidance, examines the supplied book and the documented corpus study, verifies representative faculty affiliations through university profiles, and maps those findings to EssayArc’s current tools. The six institutions form the comparison set supplied for the earlier review; this report does not treat their numerical ranking as independently established. “Leading” identifies the institutional comparison set, not a measured national league table.
Research foundations / 02
What the university programs contribute
The sources include course outcomes, writing-center guidance, a peer-fellow program, grading criteria and portfolio requirements. They are complementary teaching resources, not six versions of a single national rubric.
| Program | Published emphasis | Application in EssayArc |
|---|---|---|
| Brown | Peer fellows respond to drafts and confer with writers before revision. They support writers rather than assign grades. [1] | A concise conversation brief helps a PLF discuss the student’s choices and unresolved question. |
| Duke | Academic writing involves inquiry, evidence, audience, genre, ethical source use and an iterative process that includes reflection. [2] | Focus feedback on a writer’s purpose and reasoning; connect revision decisions to future writing. |
| Swarthmore | Writers connect claims, reasons and evidence, explain those connections, and reconsider claims when support is inadequate. [3] | Ask whether each supporting paragraph makes and explains a defensible contribution to the thesis. |
| Princeton | Motive, analysis, thesis and structure are linked. Organization follows the argument’s needs; grading criteria reward substantive reasoning. [4, 5] | Treat geometric maps as exploratory aids. Use intellectual criteria for formative feedback without assigning a paper grade. |
| Carleton | A portfolio brings writing from several settings together with reflection on the writer’s development. [6] | Record what changed and why. Use discussion and later writing to examine whether a strategy transfers. |
| Harvard | An arguable thesis answers an analytical problem. Body paragraphs explain evidence; relevant counterarguments test an interpretation. [7, 8, 9] | Examine the problem behind the thesis and the explanation and competing interpretations within the argument. |
The shared principle is that writing improves through decisions a student can explain. A polished sentence, a diagram or an AI comment matters educationally when it helps the writer understand and revise those decisions.
Research foundations / 03
From Write to the Top to a learning method
Write to the Top predates this comparative program review. Its opening account describes Granderson learning from the published writing of his UCLA professors and identifying teachable patterns in organization, diction, transitions and rhythm. EssayArc’s later research review adds institutional pedagogy and a broader empirical sentence-reference collection to that foundation. [10, 11]
Investigate relationships before settling on a claim
Geometric reasoning makes relationships visible: a line can help investigate change, a circle a return, and a triangle a relation among positions. These are prompts for discovery. The revised Thesis Forge lets students use a map, revise it when evidence does not fit, or proceed through open exploration. A thesis is assessed through its problem, evidence, explanation, scope and defensibility, rather than completion of a shape. [10, 15]
Teach sentence balance as rhythm through structure
The book’s sentence-balancing technique concerns variation among simple, compound, complex and compound-complex sentences. Length alone cannot describe these relationships among clauses. EssayArc’s useful comparison is therefore a structural pattern the student can hear and interpret in context, not a word-count contest. The book offers an illustrative ratio while expressly allowing variation; the newer discipline profiles are a separate descriptive study, not proof of that illustrative ratio. [10, 11]
Make the craft exercises serve a real piece of writing
The Writing Studio brings the craft vocabulary to a student’s own paper. Its seven choices focus attention on an identifiable writing problem. Students inspect a passage, consider the rubric, decide what to change and write the revision themselves. This connects practice to authorship rather than treating successful exercise completion as the final outcome. [14]
| Learning movement | Student’s intellectual work |
|---|---|
| Investigate | Identify a problem and test relationships against evidence. |
| Practice craft | Try a structural or rhetorical choice and explain its effect. |
| Draft and revise | Develop an interpretation; reconsider it as the writing reveals gaps. |
| Discuss and reuse | Work with a PLF, choose a next step and apply the learning elsewhere. |
These movements are recursive. Placing Craft before Draft in the interface gives students a route through the tools; it should not imply that craft must be perfected before drafting begins. [2, 4]
Research foundations / 04
A foundation in published faculty scholarship
The source ledger identifies the works used in the benchmark study. The examples below connect that ledger to official faculty biographies. They establish the presence of accomplished university scholars in the collection; they do not rank every author or imply institutional endorsement. [12]
| Faculty and institution | Representative work in the collection |
|---|---|
| Stuart BurrowsBrown University [17] | The Subject of StyleSource S083 |
| Glenda R. CarpioHarvard University [18] | Conjuring the Mysteries of Slavery: Voodoo, Fetishism and Stereotype in Ishmael Reed’s Flight to CanadaSource S016 |
| Thomas PfauDuke University [19] | Of Ends and Endings: Teleological and Variational Models of Romantic NarrativeSource S044 |
| Noel JacksonMIT [20] | The Time of BeautySource S084 |
| Ana Raquel MinianStanford University [21] | Offshoring Migration Control: Guatemalan Transmigrants and the Construction of Mexico as a Buffer ZoneSource S045 |
| Wendy Nelson EspelandNorthwestern University, emerita [22] | Rankings and Reactivity: How Public Measures Recreate Social WorldsWith Michael Sauder; source S049 |
| Lisa BernsteinUniversity of Chicago [23] | Copying and Context: Tying as a Solution to the Lack of Intellectual Property Protection of Contract TermsSource S018 |
| Kim Lane ScheppelePrinceton University [24] | Autocratic LegalismSource S010 |
| Devon W. CarbadoUCLA [25] | Strict Scrutiny & The Black BodySource S067 |
Affiliations describe the university profiles reviewed for this report, not necessarily the author’s affiliation when each work was published. Source identifiers point to the study ledger; titles are presented with normalized punctuation. The corpus includes multiple forms of scholarly prose, including articles, review essays and book-related material.
Research foundations / 05
How the scholarly benchmarks were constructed
The expanded study replaces the earlier 32-essay pilot with analysis of usable main text throughout the available collection and a documented external expansion. It seeks a fairer description of writing patterns by limiting the influence of both unusually long works and frequently represented authors. [11, 12]
| Selection and preparation | Documented accounting |
|---|---|
| Original collection | 101 PDFs examined; 100 usable PDFs yielded 102 separately attributed works. One PDF contained three distinct contributions. |
| Exclusion and replacement handling | One PDF was excluded because clipping made its text unusable. An alternate Word version of another work was treated as the same work, not an extra observation. |
| External expansion | 44 additional works acquired across the four fields; 146 works in the final collection. |
| Final text analyzed | 1,145,017 word tokens; 40,824 candidate sentence units. Of these, 37,392 were classified, 1,110 marked as fragments and 2,322 as uncertain. |
Keep the author’s main prose
Preparation removed non-main-text material such as bylines, abstracts, references, notes, recurring headers, captions, displayed quotations and reproduced verse. Quotations embedded within an author’s sentence were generally retained. Source-level corrections and exclusions are documented so that the boundaries of the analysis can be inspected. [11, 13]
Give works and authors a fair share
First, each work receives a sentence-type profile based on its classified sentences. Within each discipline, works connected by a shared author or coauthor are grouped into author networks. Each network receives equal weight; that weight is divided equally among its works. Consequently, a long essay contributes a profile, not thousands of extra votes. [11]
Work weight = 1 / (number of author networks × works in that network)
The discipline profile is the weighted mean of its work profiles. The all-disciplines reference then gives each of the four disciplines one-quarter of the weight. This prevents the largest disciplinary collection from defining the overall comparison. The expansion followed the original collection’s fields and genres; it was purposive, not a random national sample. [11]
Research foundations / 06
What the sentence profiles show
The reference collection shows substantial variation within and across disciplines. These profiles describe the grammatical structures used in the sampled works; they do not specify how a student should sound or how often a “good” writer must use a sentence type. [11]
| Field | Works | Author networks | Simple | Compound | Complex | Compound-complex |
|---|---|---|---|---|---|---|
| Literary criticism | 49 | 25 | 29.6% | 4.2% | 55.8% | 10.4% |
| Historical scholarship | 27 | 21 | 39.3% | 4.7% | 49.7% | 6.3% |
| Social research | 33 | 25 | 40.8% | 4.5% | 48.5% | 6.2% |
| Legal scholarship | 37 | 26 | 35.7% | 3.0% | 56.5% | 4.8% |
Figure and table: equal-author-network, equal-work discipline profiles. Percentages use classified sentences and may differ from 100% in total because of rounding. Source: EssayArc’s source-corrected benchmark dataset. [11, 12] In the PDF, these percentages are also drawn as stacked bars; the table above holds the same values.
Read a pattern, then investigate its effect
A complex sentence joins an independent clause with at least one dependent clause; that label does not measure intellectual sophistication. A writer might subordinate an idea to explain a condition, use a compound sentence to coordinate claims, or place a simple sentence after longer structures to create emphasis. The instructional question is what the choice does for this argument and this reader.
A comparison should lead back to the student’s prose
The student’s distribution should appear beside the selected discipline, followed by examples and the sequence of sentence types in the paper. The useful next move is to examine a passage, read it aloud and consider an alternative arrangement. Success is a reasoned revision, not convergence on a benchmark percentage. [10, 14]
Research foundations / 07
A focused rubric that keeps the student in charge
The Writing Studio organizes feedback around seven choices. This structure translates broad academic principles into manageable decisions. It does not offer whole-paper grading or substitute an AI judgment for the instructor’s assessment. The criteria below summarize the current focus definitions. [14]
| Student selects | What the analysis examines | What the student practices |
|---|---|---|
| 1. Thesis paragraph | The analytical problem, relevant context, arguable interpretation, relationships among the claim’s elements, significance and fit with the body. [7, 10] | Explain what problem the claim answers and why the interpretation matters. |
| 2. Supporting paragraphs | Topic sentences, contribution to the thesis, evidence, explanation, relevant competing interpretations and connections between paragraphs. [3, 8, 9] | Show how evidence warrants a claim; test an alternative explanation. |
| 3. Conclusion | Synthesis, return to the thesis, implications and the effectiveness of the ending. [10] | Explain what the argument now allows the reader to understand. |
| 4. Sentence balancing | Distribution and sequence of sentence types, local rhythm, emphasis and structural choices. [10, 11] | Revise a passage deliberately and explain the change in rhythm or meaning. |
| 5. Rhetorical usage | Devices actually present, their purpose, audience, effect and restraint. [10] | Choose a device because it serves an idea, rather than as decoration. |
| 6. Vocabulary usage | Precision, connotation, register and economy, with attention to the writer’s voice. [10, 14] | Select the word that expresses the intended relationship or distinction. |
| 7. Transitions | Logical relationships within and between paragraphs and whether connective language expresses them accurately. [10, 14] | Make the movement of the argument visible to a reader. |
Each focus should point to the student’s actual text, explain the reader’s difficulty or a successful choice, and offer a bounded next action. A question can deepen that response, but it should not replace a clear explanation of the relevant criterion.
Research foundations / 08
Make human conversation the center of growth
Brown’s Writing Fellows model offers a particularly useful design principle: feedback prepares a conversation that helps the writer revise. EssayArc adapts that principle through a brief view of the student’s work and choices. The PLF’s role is to listen, probe reasoning and help the student identify a next step. [1, 16]
What did you try? What changed? What remains unresolved?
Before the meeting: a student-owned record
A concise brief gives the PLF a relevant passage or change and the student’s account of the attempt. The student writes “Let’s discuss,” identifying the uncertainty they want to bring to the conversation. AI feedback can be part of the background, but the brief should foreground the student’s thinking rather than ask the PLF to administer a queue of AI comments. [16]
During the meeting: test the reasoning together
The PLF can ask why the writer changed a claim, how the evidence supports it, or what a reader would need to understand. The point is to help the student make a defensible choice. A meeting may lead to a narrower thesis, a different interpretation, more evidence, or a decision to retain a passage with a clearer rationale.
After the meeting: capture the agreed next action
“Our next step” records the outcome of the student-PLF conversation. It should be concrete enough to try and modest enough to complete: compare two interpretations of one source, explain the link between two paragraphs, or revise a passage’s sentence pattern and describe its effect. It is a human discussion outcome, not an AI assignment. [16]
| Evidence of development | Why it is useful |
|---|---|
| The earlier and revised passage | Makes a substantive change available for discussion. |
| The student’s reason for the change | Reveals understanding rather than simple acceptance of advice. |
| The unresolved question | Makes uncertainty a productive starting point. |
| The next attempt in another context | Helps examine transfer beyond this one paper. |
Carleton’s reflective portfolio practice and Duke’s emphasis on reflection and transfer support this longer view of development. A future portfolio can extend the current conversation record across assignments; it should not be confused with a feature already proven to measure learning. [2, 6]
Research foundations / 09
A credible foundation, with a clear research agenda
The evidence supports describing EssayArc as a research-informed writing environment grounded in published university pedagogy and a documented collection of scholarly prose. Its distinctive contribution is the combination: explicit craft instruction, focused analysis of a student’s own writing, and a human conversation about revision choices.
What the research establishes
The documentary review establishes identifiable connections between EssayArc’s criteria and established teaching guidance. The corpus audit establishes which works were analyzed, how main text was prepared and how disciplinary sentence profiles were weighted. Official biographies establish the affiliations of the faculty examples presented here. These are inspectable foundations for the methodology, rather than a claim based only on institutional prestige. [1, 2, 3, 4, 5, 6, 7, 8, 9, 11, 12]
What the research does not yet establish
Neither university guidance nor a faculty corpus demonstrates that using EssayArc causes learning gains. The benchmark study is descriptive and its sentence labels have measurement limits. It used a parser-based corpus pipeline, with AI-assisted review and sensitivity checks; it has not received an independent human-annotator validation study. Student analysis uses a different AI pipeline, so agreement between the two approaches still needs direct calibration. [11]
The field samples meet the study’s author-diversity floor, but each still has one or more unresolved coverage, stability or measurement criteria. The current percentages are therefore provisional descriptive references. More documents alone would not resolve classification error or selection bias. The collection is not a random sample of all faculty writing. [11]
The next evidence to collect
| Research priority | A useful test |
|---|---|
| Feedback quality | Have writing educators independently judge whether feedback accurately interprets the passage and helps a student revise. |
| Sentence measurement | Use independently annotated sentences from student and scholarly prose to test labels, disagreements and chart comparability. |
| Learning and transfer | Compare initial, revised and later independent writing with blinded human rubric assessment and a suitable comparison design. |
| Student-PLF development | Study whether the brief improves the substance of conferences, the student’s explanations and follow-through. |
Research foundations / 10
Sources & research records
Official program materials and faculty profiles were reviewed for this synthesis in September 2026. Numbered citations link to the original sources. The book was supplied by the author; corpus records and implementation files are EssayArc research and product records.
- 1. Brown University. Writing Fellows Program. Sheridan Center for Teaching and Learning. View source for source 1
- 2. Duke University. Writing 120: Academic Writing. Thompson Writing Program. View source for source 2
- 3. Swarthmore College. Using Evidence. Writing Program. View source for source 3
- 4. Princeton University. Writing Lexicon. Princeton Writing Program. View source for source 4
- 5. Princeton University. Writing Program Grading Standards: Revisions. Revised August 2021. View source for source 5
- 6. Carleton College. Writing Portfolio Requirements. Requirements effective Winter 2025. View source for source 6
- 7. Harvard College Writing Center. Thesis. View source for source 7
- 8. Harvard College Writing Center. Anatomy of a Body Paragraph. View source for source 8
- 9. Harvard College Writing Center. Counterargument. View source for source 9
- 10. Granderson, Kwame J. (2025). Write to the Top. First print edition. ISBN 979-8-89514-018-5. Supplied book; especially chapters 4-10 and 12.
- 11. EssayArc. StyleBot Full-Corpus Benchmark Review. September 2026. Internal research report published with the application. View source for source 11
- 12. EssayArc. StyleBot source audit summary. Work-level corpus accounting and benchmark provenance. View source for source 12
- 13. EssayArc. StyleBot source exclusions. Acquisition, extraction and text-boundary decisions. View source for source 13
Research foundations / 11
Sources & faculty verification
Implementation references identify the reviewed release rather than an external evaluation. Faculty biographies verify affiliations; the corpus ledger identifies the particular works analyzed.
- 14. EssayArc. Studio focus definitions and rubric. Implementation snapshot, September 11, 2026; repository access required. View source for source 14
- 15. EssayArc. Thesis Forge reasoning framework. Implementation snapshot, September 11, 2026; repository access required. View source for source 15
- 16. EssayArc. PLF conversation component. Implementation snapshot, September 11, 2026; repository access required. View source for source 16
- 17. Brown University. Stuart Burrows. Faculty profile. View source for source 17
- 18. Harvard University. Glenda Carpio. Department of English faculty profile. View source for source 18
- 19. Duke University. Thomas Pfau. Scholars at Duke faculty profile. View source for source 19
- 20. Massachusetts Institute of Technology. Noel Jackson. Literature faculty profile. View source for source 20
- 21. Stanford University. Ana Raquel Minian. Faculty profile. View source for source 21
- 22. Northwestern University. 2025 Emeritus Faculty. Wendy Nelson Espeland. View source for source 22
- 23. University of Chicago Law School. Lisa Bernstein. Faculty profile. View source for source 23
- 24. Princeton University. Kim Lane Scheppele. Sociology faculty profile. View source for source 24
- 25. UCLA School of Law. Devon W. Carbado. Faculty profile. View source for source 25
Scope note: this report uses “research-informed” to describe the documented basis for the methodology. It does not assert adoption or endorsement by the named institutions, a ranking of the individual scholars, or demonstrated causal effectiveness of the software.