Academic writing at C2 — article-length papers, lit reviews, methodology

Academic writing is the genre most Russian-speaking advanced learners think they already command, and it is also the genre in which their writing most often fails to land with American editors and reviewers. The difficulty is rarely grammatical. Russian-trained writers at C2 control the vocabulary, the passive voice, the citation form. What they often miss is the rhetorical architecture of the modern American research article: the way an introduction must frame a gap, the way a literature review must do argumentative work rather than catalogue prior research, the way a methodology must be reproducible, and the way a discussion section must distinguish what the data show from what the author concludes.

The dominant structure of empirical research in American academia is IMRaD — Introduction, Methods, Results, and Discussion — used across the natural sciences, the social sciences, and increasingly in the humanities under different names. The structure is not a stylistic convention; it is the cognitive contract between writer and reader. The reader of a research article reads in the order abstract → introduction → results → methods → discussion, skipping and returning. The writer who organizes the paper to support this reading pattern is read; the writer who buries the contribution in a long literature review is not.

This lesson takes the article-length research paper as its object. The four sections of an IMRaD paper each have their own rhetorical demands, and at C2 you should be able to write each one to the standard of a peer-reviewed journal. The model passage below is a discussion section, which is where academic writing is most often weakest.

Structure — the IMRaD architecture

Article totals: 6000-12000 words for a typical journal piece.

Step-by-step craft

1. Write the introduction as a CARS move

The Create-A-Research-Space (CARS) model, formalized by John Swales, describes the standard moves of an academic introduction in three steps. Move 1: Establish the territory — say what is known and why it matters. Move 2: Establish the niche — show what is missing, contested, or unresolved in the existing work. Move 3: Occupy the niche — name what this paper contributes. The American convention is to compress these three moves into a brisk introduction, often only 700-1200 words, in which the gap is named clearly and the contribution stated directly. Russian-trained writers often spend too long on Move 1 (territory) at the expense of Moves 2 and 3.

2. Write a literature review that argues

The literature review is not a list. It is an argument about the state of the field, organized around the writer’s claim. A weak literature review proceeds chronologically (Smith (2010) found X. Jones (2012) found Y. Patel (2015) found Z.); a strong literature review organizes by question, debate, or methodological camp (Researchers studying X have divided along two lines. The first, exemplified by Smith and Jones, treats X as…; the second, exemplified by Patel, has argued that…). The review’s purpose is to set up the contribution; every cited work either supports the writer’s frame or is the position the paper is responding to.

3. Write methodology for reproducibility

A methods section is judged by whether another researcher could replicate the study from the description alone. This demands specificity that beginning academic writers often resist: the model number of the instrument, the version of the software, the exact wording of the interview question, the number of coders and the inter-coder reliability statistic, the IRB protocol number. Russian-speaking writers often write methodology in a higher register than the genre requires; the register is technical and plain, not literary.

4. Separate results from interpretation

The results section reports what was found, in the order in which the analyses were conducted. It does not interpret. Participants in the treatment condition scored 3.4 points higher on the post-test, t(58) = 2.31, p = .024. The interpretation — what the difference means, why it might have arisen, what its theoretical implications are — belongs in the discussion. Mixing the two is one of the most common reasons papers are returned for revision.

5. Write the discussion as a controlled argument

The discussion section is where the paper earns its publication. It restates the central finding, situates it in the literature, considers alternative explanations, names limitations, and suggests implications. The C2 challenge is to commit to a claim while acknowledging its limits. American academic prose tolerates strong claims if they are bounded; it does not tolerate either bland summary or unhedged overreach.

6. Manage passive and active voice deliberately

The cliché that academic writing is uniformly passive is false. Modern American practice mixes voices. The active voice is preferred when the agent matters (We tested four models); the passive is preferred when the agent does not matter or is conventionally absent (The samples were stored at -80°C). Russian-speaking writers often default to passive throughout, producing a prose that reads as bureaucratic. The rule of thumb: passive when the action is what matters; active when the actor is what matters.

7. Hedge with calibrated confidence

Academic claims are bounded. Our results show is a strong claim; our results suggest is a weaker claim; our results are consistent with is the weakest. Match the verb to the strength of the evidence. The C2 instinct to use prove (and the Russian instinct to translate доказывает as proves) overreaches in most contexts; demonstrate, establish, indicate, and support offer graded alternatives.

8. Calibrate the role of theory

American academic writing varies in its use of theory. In the natural sciences, theoretical apparatus is minimal and is introduced only to frame predictions. In the social sciences, theoretical framing is conventional but bounded — usually one to three paragraphs at most. In the humanities, theory can be more central but must still serve the argument rather than dominate it. Russian-trained writers often deploy more theoretical machinery than American journals expect; calibrate to the field.

9. Address limitations honestly

The limitations section is the writer’s chance to demonstrate critical self-awareness. Three to five concrete limitations, each tied to a feature of the study design, signal a careful researcher. Limitations that are merely formal (the sample could have been larger) signal a writer going through motions. The strongest limitations identify what the present study cannot answer and what future work would need to do to answer it.

Full model text — 750-word annotated discussion section

The model below is the discussion section of an empirical paper in a social-science journal. The rhetorical moves are marked in brackets.

Discussion

[Restate the central finding] The present study tested whether moderate exposure to political disagreement on social media reduces affective polarization, as predicted by contact theory, or amplifies it, as predicted by social-identity accounts. The data support the second prediction. Participants randomly assigned to a feed containing cross-partisan posts reported higher affective polarization at two-week follow-up than control participants assigned to a same-partisan feed (M_diff = 0.42 on a 7-point scale, 95% CI [0.18, 0.66], d = 0.31). The effect persisted across both Democratic and Republican subsamples and was robust to controls for baseline polarization, news consumption, and demographic covariates.

[Situate in the literature] These findings extend a small but consistent line of recent work that has called the contact hypothesis as applied to online disagreement into question. Bail and colleagues (2018) reported similar effects in a field experiment with Twitter users; Levendusky and Malhotra (2016) found that exposure to mixed political content increased rather than decreased perceived ideological distance. The present study replicates these findings under tighter experimental conditions and with a more representative sample. It does not, however, settle the broader question of whether all forms of cross-partisan contact produce these effects. The contact theory literature in face-to-face contexts (Pettigrew & Tropp, 2006) remains robust, and the divergence between online and offline findings is itself a significant open question.

[Consider alternative explanations] Three alternative explanations warrant consideration. First, the effect may be specific to the platform tested (Twitter/X); other platforms with different affordances — particularly those that surface content from acquaintances rather than strangers — might produce different effects. Second, the two-week follow-up window may capture only short-term reactance; longer exposure may attenuate or reverse the effect, a possibility consistent with habituation models. Third, the effect may be mediated by the quality rather than the fact of disagreement; if cross-partisan posts in the experimental condition were unusually contentious, the manipulation may have been confounded with hostility. Our content analysis (see Supplementary Material C) found that cross-partisan posts were rated as 18% more negative in tone than same-partisan posts, lending some support to this concern.

[Name limitations] Several limitations should be acknowledged. The sample, while reasonably representative of US social-media users, underrepresents users over 65 and rural users; generalization to those populations is unwarranted. The manipulation was administered for two weeks, a short interval relative to ordinary social-media use; effects of longer-term exposure are not addressed by the present design. Affective polarization was measured by self-report on a single scale; behavioral measures (donation patterns, willingness to interact across partisan lines) were not collected and would represent a valuable extension. Finally, the study was conducted during the 2025 midterm cycle, a period of unusually high political salience; the effect size may overstate what would obtain in lower-salience periods.

[Implications] If the effect we observed is real and generalizable, the policy implications are non-trivial. Platform interventions designed to “burst the filter bubble” by surfacing cross-partisan content may, contrary to the intent of their designers, increase rather than decrease political division. This conclusion should be drawn cautiously: a single study, however well-designed, cannot adjudicate a policy question of this importance. But the convergence of our findings with prior work suggests that the burden of evidence has shifted. Researchers and platform designers proposing cross-partisan exposure as a polarization remedy should now demonstrate, rather than assume, that their interventions produce the predicted effect.

[Forward look] Three lines of further work follow from these findings. First, dosage-response studies could establish whether more sustained exposure attenuates or amplifies the short-term effect. Second, mediator analyses could test whether tone of disagreement, rather than ideological distance, drives the polarization effect. Third, field trials with platform partners could test interventions designed to surface high-quality cross-partisan content while filtering hostility. We are pursuing the second of these in ongoing work and would welcome collaboration on the others.

A working sequence for the empirical article

For the C2 writer drafting an empirical research article, the following sequence is reliable.

1. Conduct the study; collect and clean the data.
2. Draft the methods section in detail while procedures are fresh.
3. Run the analyses; produce figures and tables.
4. Draft the results section, reporting in narrative order.
5. Draft the introduction, working from a one-sentence statement of contribution.
6. Draft the literature review, organized by debate or question.
7. Draft the discussion, with limitations and implications.
8. Draft the conclusion.
9. Draft the abstract last.
10. Revise the whole for voice consistency.
11. Format citations to the journal’s style.
12. Have at least one colleague read the draft before submission.

A 7000-word empirical article typically takes between three and twelve months from data collection to submission, with most of the time spent in writing and revision.

Writing the abstract

The abstract is the most-read paragraph of any academic paper, and the only paragraph that many readers ever see. It must compress the entire article into 150 to 300 words. Conventional structure includes background (one sentence), gap (one sentence), method (one to two sentences), results (one to two sentences), and implications (one sentence). A strong abstract is informative — the reader can decide whether to read the full paper based on the abstract alone.

Russian-speaking writers often produce abstracts that are too long, too vague, or too descriptive of the paper rather than of the findings. The abstract should read as a compressed version of the paper, not as a description of what the paper does. The paper examines the relationship between X and Y is descriptive; The relationship between X and Y is moderated by Z is substantive.

Writing the literature review — three working methods

The literature review is the section that requires the most rhetorical organization. Three working methods help structure it.

Method 1 — by debate

The most powerful organization treats the literature as a conversation. Researchers studying X have divided into two camps. The first, exemplified by…, argues that…; the second, exemplified by…, has countered that… The review positions the paper as an intervention in this conversation.

Method 2 — by evidence type

When the literature is methodologically diverse, organizing by evidence type can clarify. Experimental studies have produced one set of findings; observational studies have produced another; the discrepancy is partly explained by… This organization sets up methodological contribution.

Method 3 — by question

Organize by the question each cluster of studies has addressed. On the question of whether X causes Y, the literature has reached consensus. On the question of how X causes Y, agreement has been more elusive. The unresolved question becomes the paper’s territory.

Each method produces a coherent review; the wrong method (chronological by publication date) produces a list.

Writing the methods section

The methods section is where reproducibility is established. The reader should be able to replicate the study from the description.

Participants

Number, recruitment method, demographics, exclusion criteria. Two hundred and forty-eight US adults (52 percent women, mean age 34.6) were recruited through Prolific between February and April 2025… Specificity wins.

Materials

Every instrument, every survey, every stimulus. Version numbers, item counts, response scales. If a published scale is used, cite the original.

Procedure

Step by step, in the order the study unfolded. Participants first completed a baseline measure. After a two-week interval, they were randomly assigned to one of three conditions… The procedure section is where reproducibility is concretely operationalized.

Analysis plan

The statistical or analytical procedures used. Pre-registration, when relevant. Software versions. All analyses were conducted in R version 4.3.0 using the lme4 package…

A methods section that is too short almost guarantees the paper will be returned for revision. A methods section that is too long is rare but possible — supplementary materials can absorb detail that interrupts the main text.

Disciplinary variation in academic prose

The IMRaD structure dominates the empirical sciences and most social sciences, but academic prose varies significantly across disciplines. Three patterns are worth noting for the C2 writer.

Humanities

In literary studies, history, and philosophy, the standard article is a long argumentative essay rather than an IMRaD report. Sections are conceptual rather than procedural; the literature review is integrated into the argument rather than separated; theoretical engagement is more extensive. The American conventions are looser, but the demand for original argument is more pronounced.

Law

Legal scholarship (covered more fully in the previous lesson) follows neither IMRaD nor the humanities essay. Articles are organized around legal questions, with extensive footnoting and a distinct rhetorical structure. Legal academic writing is the closest American academic genre to European traditions.

Mathematics and theoretical computer science

Mathematical articles follow a theorem-proof structure that is even more compressed than IMRaD. The introduction sets the problem; the body proves results; the discussion is minimal. American conventions here are highly formal.

The C2 writer aiming at academic publication should identify the conventions of the target field early and apply them rigorously.

Common pitfalls

Literature review as catalogue

A list of Smith (X) found… Jones (Y) found… Patel (Z) found… organizes by source rather than by argument. Reorganize by question or debate.

Burying the contribution

The contribution should appear in the abstract, the end of the introduction, and the discussion. If a reviewer cannot find the sentence that says what this paper adds, the paper has failed.

Mixing results with discussion

The treatment group scored higher, which suggests that the intervention works. The first clause is results; the second is discussion. Keep them in separate sections.

Unhedged claims

This study proves that X causes Y. Empirical research almost never proves; it supports, indicates, demonstrates within bounds. Calibrate the verb to the evidence.

Methodology too thin

A methodology that another researcher cannot replicate from the description has failed its purpose. Specificity beats elegance in this section.

Discussion as restated results

The discussion’s job is interpretation. Restating what was found in different words is not interpretation. Compare to prior work, name limitations, draw implications.

Citation styles in American academia

American academic writing uses several citation styles, each dominant in particular fields. A C2 writer should recognize at least three.

APA

The American Psychological Association style governs psychology, education, and much of the social sciences. Citations are author-date in the text (Smith, 2020); the reference list is at the end. APA emphasizes the recency of evidence; the year is foregrounded.

MLA

The Modern Language Association style governs literature, modern languages, and humanities. Citations are author-page in the text (Smith 234); the works-cited list is at the end. MLA emphasizes the source’s pages and physical form.

Chicago

The Chicago Manual of Style governs history, theology, and parts of the humanities. Two variants exist: notes-bibliography (with footnotes) and author-date (similar to APA). Chicago is the most flexible of the three.

Verifying journal style

Every journal specifies a citation style in its submission guidelines. The C2 writer should verify the style before drafting and apply it consistently throughout. Citation style errors signal that the writer has not engaged with the journal’s conventions.

Writing the introduction — paragraph by paragraph

The introduction of a standard empirical article runs five to seven paragraphs. Each has a function.

Paragraph 1 — broad opening

The opening paragraph establishes the topic’s importance with a specific, sourced claim. Affective polarization in the United States has, by most measures, doubled since 2000 (Iyengar et al., 2019). Not generic.

Paragraph 2-3 — narrowing

The next paragraphs narrow from the broad topic to the specific question the paper addresses. Of particular concern has been the role of social media in this process…

Paragraph 4 — gap

The paragraph that names what is missing or contested in the existing literature. What has been less well examined is whether…

Paragraph 5 — contribution

The paragraph that states what the paper does. The present study tests this question directly, using…

Paragraph 6-7 — roadmap

A brief paragraph that previews the structure of the paper. The remainder of this article is organized as follows…

This five-to-seven paragraph structure is the standard model in American social-science journals; humanities and natural-sciences journals vary slightly.

Connectors and phrases bank

• Establishing territory: A substantial body of work has examined…; The literature on X has converged on the view that…; Researchers in this tradition have…
• Establishing niche / gap: Less well understood is…; What remains contested is whether…; Despite this consensus, three questions remain unresolved…
• Occupying niche / contribution: The present study addresses this gap by…; We extend this line of work in three respects…; Our contribution is to test…
• Hedging finding strength (graded): prove → demonstrate → establish → indicate → suggest → are consistent with → tentatively support.
• Discussion moves: These findings extend…; The present results converge with…; A plausible interpretation is…; An alternative explanation, which we cannot rule out, is…
• Limitations: Several limitations should be acknowledged; The generalizability of these findings is bounded by…; The cross-sectional design does not permit causal inference…
• Forward look: Three lines of further work follow from these findings…; A natural extension would test…; Future research should address…

Writing the results section

Results are reported, not interpreted. Five conventions govern the section.

Convention 1 — narrative order

Report results in the order in which they were analyzed, or in the order that supports the paper’s argument. The order should follow a logic the reader can track; do not present results as a random list.

Convention 2 — descriptive then inferential

Descriptive statistics (means, standard deviations, proportions) typically precede inferential statistics (tests, models, effect sizes). The reader gets the lay of the data before the test of the hypothesis.

Convention 3 — effect size, not just significance

Modern American journals require effect sizes alongside p-values. t(58) = 2.31, p = .024, d = 0.31 is the convention; t(58) = 2.31, p = .024 alone is now considered incomplete reporting.

Convention 4 — visual support

Figures and tables present complex findings more efficiently than prose. Refer to each figure and table in the text; do not let figures float without textual reference.

Convention 5 — no interpretation

The treatment group scored higher is a result. The treatment group scored higher, supporting the contact hypothesis mixes result with interpretation; move the second clause to the discussion.

Writing the conclusion

The conclusion synthesizes across the paper and looks forward. It is shorter than the discussion (200 to 500 words for most articles) and tighter in focus. The reliable structure includes a one-paragraph summary of the central finding, a paragraph on implications, and a paragraph on next steps. The conclusion should not introduce new evidence or new analysis; the work is done.

Common Russian-speaker writing mistakes

1. Generic territory openings — В современном мире… translated to In the modern world… is a calque that American journals reject. The territory must be specific to the literature the paper is joining.
2. The phrase more research is needed — Russian academic tradition tolerates this as a niche statement; American journals require the gap to be specific (what research is needed, why, what would resolve which question).
3. Uniform passive voice — Russian-trained writers often use the passive across all four sections. American practice mixes voices; methods can use the passive (The samples were prepared) but results, discussion, and introduction often benefit from the active (We tested, The data show).
4. Calque on доказать — prove used loosely. American academic writing reserves prove for mathematical proof and one or two formal contexts; everywhere else, use demonstrate, establish, indicate, or support.
5. Citation in-text style errors — Russian academic citation conventions differ from APA, MLA, and Chicago. Verify the journal’s style and apply it consistently; the in-text format (author-date vs footnote, page numbers vs no page numbers) varies by field.
6. Long literature reviews that catalogue — Russian academic tradition often produces 30-40 page reviews that comprehensively cover the field. American journals want focused reviews that set up the contribution; 5-10 pages of well-argued review beats 30 pages of catalogue.
7. Discussion as summary — In this paper, we have shown… followed by a paragraph restating the abstract. The discussion’s job is to do something with the findings — situate them, compare them to prior work, name their limits, draw implications. Restating is not enough.

Responding to peer review

The peer review process is where most academic papers spend most of their time. Three categories of revision recur.

Major revisions

The reviewer asks for substantial changes — additional analyses, different framing, more extensive engagement with prior work. Major revisions can take weeks. The C2 writer should treat them as opportunities to strengthen the paper rather than as obstacles to publication.

Minor revisions

The reviewer asks for clarifications, additional citations, or sentence-level adjustments. Minor revisions are usually completed within a few days.

Rejections

A rejection is a verdict on the paper at this journal. It is rarely a verdict on the paper itself. The standard move is to revise based on the reviews and submit elsewhere; most successful papers are rejected once or twice before publication.

The response letter

Every revision is accompanied by a response letter that addresses each reviewer comment. The convention is to quote the comment, describe what was done in response, and indicate where in the revised manuscript the change appears. A carefully prepared response letter is often what tips the balance toward acceptance.

The article-length argument — three reliable structures

Beyond the standard IMRaD article, three argumentative structures recur in academic writing across fields.

Structure 1 — the contrast

Two views or two cases are compared in detail; the contrast supports a larger claim. Account A predicts X; account B predicts Y. The evidence supports Y.

Structure 2 — the synthesis

Two or more apparently incompatible positions are reconciled through a third frame. The conflict between A and B has been treated as a binary; in fact, the two are compatible under C.

Structure 3 — the critique

A widely-accepted position is examined and limited. The dominant view holds X. This paper argues that X is true only under conditions Y and Z; outside those conditions, the position requires revision.

Each structure corresponds to a recognized rhetorical genre in academic writing. The C2 writer should be able to identify which structure a paper requires before drafting.

Summary

• IMRaD is the dominant architecture: Introduction, Methods, Results, Discussion.
• The introduction follows CARS: establish territory, establish niche (gap), occupy niche (contribution).
• The literature review is argumentative, not catalogue; organize by question or debate.
• Methods is technical and plain; reproducibility is the standard.
• Results report; discussion interprets. Keep them separate.
• Hedge claims with calibrated verbs: prove → demonstrate → establish → indicate → suggest → consistent with.
• Russian-speaking writers should especially watch generic territory openings and the uniform passive.

Next lesson: Creative nonfiction at C2 — memoir, personal essay, narrative nonfiction.