How to Write an Introduction for a Research Paper
A research-paper introduction moves the reader from a known conversation into the specific problem this paper will solve. It is not an abstract, not a literature review, and not a standalone thesis statement. The Harvard College Writing Center says the opening presents an analytical question or problem and then offers an answer (the thesis). This page is that writing how-to. It is not how to write a conclusion, which leaves the paper, and not how to write a research paper, which covers the whole draft.
By CitationEasy Editorial Team34 min read
On this page
Quick Answer: How to Write an Introduction
Name the conversation, mark what is still wrong or unknown, then state what this paper will do. Course essays follow Harvard's three pieces: orienting information, what is at stake, and a thesis. Journal articles follow John Swales's CARS model: establish a territory, establish a niche, occupy the niche. USC Libraries restates the same job as four reader questions: what you studied, why it mattered, what was already known, and how this study adds knowledge. University of Toronto Writing Advice adds the length rule: most papers need one paragraph covering half to three-quarters of the first page, longer only when the document is long enough to earn it.
Starting templates (fill from the paper you wrote; do not invent a gap)
Course essay / humanities paper:
Readers often treat [text or debate] as [the usual reading]. That account cannot explain [a tension already in the source]. This essay argues that [arguable thesis], which matters because [the misreading it corrects].
IMRaD / empirical paper:
Prior work on [territory] has established [what is settled]. It remains unclear whether [specific gap]. This study tests [purpose, question, or hypothesis] in [population or setting].
Design / engineering note:
Existing systems for [task] still [the constraint the field already accepts]. We present [the artifact] that [the claim a reader can check in the methods].
Randomised trial (CONSORT 2025 Items 6–7):
[Why a new trial is needed, citing a systematic review when one exists]. The objective is to evaluate whether [intervention] improves [primary outcome] compared with [comparator] in [participants] at [timepoint].
Use the thesis-statement generator or the research-question generator only after you can already say what would count as an answer. Then format the sources the opening actually cites with the CitationEasy generator. This page is not how to write a thesis statement, which is the claim inside the opening.
What a research paper introduction actually does
Readers decide in the first screen whether the paper is for them. The introduction has to answer four questions before any methods or close reading begins. USC Libraries writes those questions as a mental road map: what you studied, why the topic was important, what was known before this study, and how this study advances knowledge. If a paragraph does not help with one of those four, it belongs later or nowhere.
That is a different job from neighbouring sections. The abstract is a standalone summary that often reports results. A literature-review chapter synthesises a field. A thesis statement is usually one or two sentences inside the introduction, not a substitute for it. A conclusion leaves the paper. This page is only about the opening section.
The introduction has to finish these jobs
- Name the topic at the right scale (not "since the dawn of time," not a methods dump)
- Orient a non-expert just enough to follow the claim
- Show what prior work has already established
- Identify a gap, tension, puzzle, or practical problem
- State the research question, thesis, or hypothesis
- Say what is at stake if the answer is right
- Optionally preview the paper's structure (common in long essays and some social-science articles; rare in short science letters)
- In a social-science paper, name the delimitations: what the study will not cover, and why
Course essay vs IMRaD article vs trial report
Harvard College Writing Center puts the same work in three pieces that travel across disciplines: orienting information, an explanation of what is at stake, and a thesis. John Swales's CARS model (Create a Research Space), built from research-article introductions, uses three moves: establish a territory, establish a niche, occupy the niche. CONSORT 2025 adds two extra jobs when the paper reports a randomised trial: Item 6 (scientific background and why a new trial is needed) and Item 7 (specific objectives for benefits and harms, including participants, interventions, comparator, primary outcome, and timepoint). Mixing the patterns is how students write a dramatic "hook" that a methods paper does not want, or a gap statement that a close-reading essay does not need. Choose the pattern the assignment actually grades.
Course essay / humanities paper
- Opens a problem a thoughtful reader could actually have
- Orients (text, author, debate) without a book-report plot summary
- Ends the opening on an arguable thesis
- May preview the argument in Government, Economics, and some long papers
- Anecdotes and literary quotations can work if they are on-topic
- Usually one paragraph; longer papers may use two or three
IMRaD research article or trial
- Opens on the research problem and why it matters now
- Cites only the studies the gap depends on
- States a purpose, objective, or hypothesis
- Does not report the study's own results (ICMJE)
- Trial papers also name the comparator and primary outcome (CONSORT 2025 Item 7)
- Dictionary hooks and classroom anecdotes usually fail
- Often 300–800 words; letters and Nature-length papers are much shorter
What writing centers and manuals actually require
Advice pages disagree about hooks. They agree about function. An introduction that only "grabs attention" and never frames a problem is doing entertainment, not academic work. The shortest official statements are more useful than a list of clever first lines.
How the authorities define the opening
Harvard College Writing Center, “Introductions”
The introduction to an academic essay will generally present an analytical question or problem and then offer an answer to that question (the thesis).
Harvard College Writing Center, “What Do Introductions Across the Disciplines Have in Common?”
While the introductions focus on very different topics, each one contains the key elements of an introduction: orienting information, an explanation of what’s at stake, and a thesis.
UNC Writing Center, “Introductions”
Your introduction is an important road map for the rest of your paper. Your introduction conveys a lot of information to your readers. You can let them know what your topic is, why it is important, and how you plan to proceed with your discussion.
University of Toronto Writing Advice, “Introductions and Conclusions”
A good introduction should identify your topic, provide essential context, and indicate your particular focus in the essay. It also needs to engage your readers’ interest.
USC Libraries, “4. The Introduction”
The introduction leads the reader from a general subject area to a particular topic of inquiry.
APA Style (7th edition headings)
Because the first paragraphs of a paper are understood to be introductory, the heading “Introduction” is not needed. Do not begin a paper with an “Introduction” heading; the paper title at the top of the first page of text acts as a de facto Level 1 heading.
IEEE Author Center, “Structure Your Article”
The introduction section includes a review of the existing literature to position your research within the broader scientific field and to show the novelty of your work. The introduction should also describe the question you’re trying to answer with your research and why that question is important to the field.
ICMJE Recommendations (January 2026, manuscript introduction)
Provide a context or background for the study (that is, the nature of the problem and its significance). State the specific purpose or research objective of, or hypothesis tested by, the study or observation. Cite only directly pertinent references, and do not include data or conclusions from the work being reported.
Those sentences come from the Harvard introductions handout, Harvard's introductions across disciplines page, the UNC introductions handout, Toronto's introductions and conclusions advice, the USC introduction guide, the APA Style headings page, the IEEE Author Center structure page, and the ICMJE manuscript-preparation recommendations (updated January 2026). Harvard is describing an essay that argues. ICMJE is describing a biomedical article that tests something. IEEE wants the novelty claim visible. All still require a problem, a purpose, and a reason the reader should continue.
Harvard’s stake test, and the two openers it rejects
Stake, not topic justification
It’s not enough to explain why the topic you’re writing about matters; rather, you need to explain what your essay adds to that discussion.
No cosmic first sentence
Avoid writing a very general opening sentence. While it may be true that “Since the dawn of time, people have been telling love stories,” it won’t help you explain what’s interesting about your topic.
No empty funnel
Avoid writing a “funnel” introduction in which you begin with a very broad statement about a topic and move to a narrow statement about that topic. Broad generalizations about a topic will not add to your readers’ understanding of your specific essay topic.
"People care about climate change" is a topic justification. "Existing city-level studies measure adoption of rooftop solar but not the municipal incentive designs that predict it" is a paper justification. Harvard's Dobbs example on the same page makes the same cut: it is not enough to say people care about reproductive rights; readers need to know what this thesis adds.
Introduction vs abstract vs literature review
These three get swapped in drafts because they all mention prior work. They are indexed differently, read differently, and graded differently. Keep the jobs separate and the citations stay honest.
| Section | Stands alone? | Reports your results? | How much prior work? |
|---|---|---|---|
| Abstract | Yes; databases index it | Usually yes, in informative abstracts | Almost none; no undefined abbreviations |
| Introduction | No; it hands the reader to the body | No in IMRaD (ICMJE); a thesis is a claim, not a result | Only the studies the gap needs |
| Literature review | Sometimes (standalone review article or thesis chapter) | No | A synthesised field, organised by argument |
If your introduction is already walking through twenty studies one by one, you have started the literature review early. If your abstract withholds the finding "to create suspense," you have written a bad abstract, not a good introduction. Write the abstract last, from the finished paper; draft the introduction late enough that the purpose statement matches the paper you actually wrote.
The CARS model (the research-article pattern)
John Swales analysed research-article introductions across disciplines and described a recurring rhetorical pattern in Genre Analysis (Cambridge University Press, 1990). He and Christine B. Feak restated it for graduate writers in Academic Writing for Graduate Students (3rd ed., University of Michigan Press, 2012). University writing centres, including USC Libraries' CARS page, teach the same three moves. You do not have to name CARS in the paper. You do have to perform the work.
Move 1: Establish a territory
Show that a research area is worth a paper: claim centrality ("X is a major cause of…"), make a topic generalisation ("Recent models treat Y as…"), and review the items of previous research the gap depends on. This is not a comprehensive review. USC's CARS guide calls Move 1 a reflection of key studies, not a chapter. The full literature review comes later if the genre requires one.
Territory language at the right scale (compressed from Vaswani et al. 2017)
Recurrent encoder-decoder models with attention became the default for machine translation, but they generate hidden states one position at a time and therefore resist parallel training on long sequences.
Territory language that is too wide
Since the dawn of civilization, humans have wanted to translate languages.
Move 2: Establish a niche
Name the opening your paper will occupy. Swales listed four ways in: a counter-claim, a gap, a raised question, or a continuation of a tradition. Gap language is the most common in student IMRaD papers. It should be specific. "Little research exists on social media" is not a niche. "Prior work measures screen time, not active versus passive use, in first-year students" is a labelled scale example, not a published finding. For a real niche, use the Clements–Gleason contest in Harvard's Crane Beach sample below, or CONSORT 2025's demand that a new trial be justified against a systematic review.
Niche signals (from the CARS tradition)
- Gap: "However, few studies have isolated…" / "It remains unclear whether…"
- Counter-claim: "Those estimates assume X; that assumption fails when…"
- Question: "Despite repeated observations of A, we still do not know why B occurs."
- Continuation: "These results suggest a need to test the same mechanism in…"
- Neutral contrast (when you do not want to attack prior authors): "Although research has been devoted to A, less attention has been paid to B."
USC's CARS guide warns against a sneering niche. You can mark a limitation without calling earlier work incompetent. The same page's neutral contrast — "Prior research has focused primarily on A, rather than on B" — does the Move 2 job and keeps the citation relationship usable.
Move 3: Occupy the niche
State what this study or essay will do. Outline purposes, announce the present research, and, if the journal's convention allows it, indicate the article's structure. Announcing principal findings in the introduction is field-specific: some science and economics papers preview the result; ICMJE tells medical authors not to include data or conclusions from the work being reported. Follow the venue, not a generic blog rule.
Move 3 as a real hypothesis (Harvard sample: Talia Blatt ’23, Crane Beach dunes)
Using the sand dunes at Crane Beach in Ipswich, Massachusetts, this study uses the chronosequence approach to study primary succession, characterizing an ecological community’s change over time with the proxy of distance from the shore. I hypothesize that overall diversity will increase with time, as pioneer grasses colonize and facilitate shrubs and eventually trees.
Structure sentence (long paper only; labelled template)
Section 2 reviews [the studies the gap depends on]; Section 3 describes the sample and models; Section 4 reports results; Section 5 discusses implications.
| Move | Reader question it answers | Typical placement |
|---|---|---|
| 1 Territory | What conversation is this? | Opening sentences and selective citations |
| 2 Niche | What is still wrong, missing, or unstable? | The "however" paragraph |
| 3 Occupation | What will you do about it? | Last sentences; purpose, question, or thesis |
The funnel, and the "dawn of man" failure
Many handbooks describe an inverted pyramid: start broader than the study, end on the exact question. That pattern matches CARS when "broader" still means this research area. It fails when "broader" means human history. UNC's introductions handout uses a maps metaphor: zoom to the scale of the question. If the assignment is about education in Frederick Douglass, you do not open at the latitude of Rome. Harvard is blunter: do not write a very general opening sentence, and do not write a funnel that begins with a very broad statement about a topic. Toronto says the same thing in three words of advice: get to the point.
Toronto’s length and first-sentence rules
How long most essays need
The introductions for most papers can be effectively written in one paragraph occupying half to three-quarters of the first page.
Raise the topic in sentence one
Get to the point as soon as possible. Generally, you want to raise your topic in your very first sentences. A common error is to begin too broadly or too far off topic. Avoid sweeping generalizations.
Those two pieces of Harvard advice are not a contradiction once you separate topic-scale context from cosmic context. "Recurrent sequence models are hard to parallelise" is a funnel step that earns its keep. "Since the dawn of time, people have been telling love stories" is Harvard's own example of a useless opener. If deleting the first two sentences would not change what the reader knows about your problem, they were decoration.
Wrong scale (UNC / Harvard "dawn of man")
Throughout human history, communication has shaped society. Today, technology is more important than ever. This paper discusses social media.
Right scale (territory already inside the problem)
Coastal dunes are key sites for the study of ecological succession, and theories of that succession remain contested between Clements’ climax community and Gleason’s contingent polyclimax.
Write the introduction late, even if you draft a placeholder first
UNC's advice is the one that saves the most revision time: try writing your introduction last. You may not know the argument until the body exists. A purpose statement written in week one will describe the paper you hoped to write. The introduction the reader sees has to describe the paper you did write. Toronto agrees: some writers draft an opening first to think, then compress or rewrite it after the body exists; others write it only at the end. USC adds a late-process check: after the body exists, reread the opening's method, structure, and promised outcomes so they still match.
UNC Writing Center
Sometimes it’s easiest to just write up all of your evidence first and then write the introduction last—that way you can be sure that the introduction will match the body of the paper.
University of Toronto Writing Advice
It can be fine to leave the writing of the introduction for a later stage in the essay-writing process. Some people write their introduction only after they have completed the rest of the essay. Others write the introduction first but rewrite it significantly in light of what they end up saying in the body of their paper.
A placeholder opening is fine for getting words on the page. Treat it as scaffolding. When the thesis or hypothesis changes, the first page changes with it. The conclusion should then invert the same movement: from your specific finding back to the stake you named at the start, not a new topic.
A drafting sequence that matches the finished page
- Write a one-sentence purpose statement you could show a supervisor (question, thesis, or hypothesis).
- List the three to eight sources the gap actually depends on, not everything you read. Evaluate and organise those notes first.
- Draft Move 2 (the niche) next. If you cannot state the gap without hedging it into nothing, the project is still a topic.
- Write Move 1 around that niche so the citations all point at the same hole.
- Write Move 3 so the purpose is narrower than Move 1 and wider than a methods sentence.
- If the paper is a randomised trial, add CONSORT 2025 Item 6 (why a new trial, citing a systematic review when one exists) and Item 7 (participants, intervention, comparator, primary outcome, timepoint).
- Only then decide whether the first sentence needs a hook. Most IMRaD papers do not.
- After the body and conclusion exist, rewrite the opening so no promise is unkept.
State a purpose a reader can test
Vague purpose language ("this paper will explore various aspects of…") is how introductions hide an unchosen project. A research question names a population, a relationship, or a text and an angle. A hypothesis predicts a relationship in advance. A thesis statement is a claim a thoughtful reader could dispute. Use the tool that matches the assignment: question and hypothesis for empirical papers, thesis for interpretive essays. The generators for a research question, hypothesis, and thesis are useful only after you can already say what would count as an answer.
Too vague
This paper will discuss climate policy and its many important effects on society.
Real hypothesis (Harvard sample: Blatt, Crane Beach)
I hypothesize that overall diversity will increase with time, as pioneer grasses colonize and facilitate shrubs and eventually trees. This biotic shift will likely be accompanied by an abiotic shift in soil texture, from coarse to fine. Altogether, I hypothesize that the dune chronosequence will show a progression in line with Clements’ vision of a maximally biodiverse climax community.
Real trial objective (CONSORT 2025 Item 7 example, Hopewell et al.)
To evaluate whether a structured exercise programme improved functional and health related quality of life outcomes compared with usual care for women at high risk of upper limb disability after breast cancer surgery.
Real interpretive thesis (Harvard sample: Eliza Hirsch ’25, Hamlet)
In Hamlet, the motif of ears being physically attacked creates a connection between physical and verbal violence and ultimately suggests that the two are equal.
How to start: openings that work, and five that writing centres reject
UNC says the first sentence should "actually say something useful" in a polished way. Harvard says you do not have to hook readers with a dramatic promise or an exciting fact. Both can be true. A useful first sentence is already interesting because it is already inside the problem.
Openings that usually earn their space
- A tension in the literature. Two credible findings that cannot both be true as stated. Harvard's Blatt sample opens on Clements versus Gleason, not on "nature is important."
- A precise stake. Who is harmed, mismeasured, or misread if the current account stands.
- A primary-source moment, in humanities papers, that the thesis will interpret (UNC's Douglass examples and Harvard's Hirsch Hamlet ear-count both start from the assigned text).
- A constraint the field already accepts. Vaswani and colleagues open on sequential computation, not on "the wonder of language."
- *A definition you cannot assume*, taken from a field source rather than a desk dictionary.
Five introductions UNC tells you not to write
The UNC handout names five failure modes that still fill first-year papers. They waste the only page many graders read twice. Toronto's pitfalls list overlaps on dictionary definitions and restated prompts. USC repeats the dictionary ban and adds a sixth social-science failure: a long history of the topic that belongs in the literature review.
- The placeholder. Vague sentences that occupy the "introduction space" without a claim: "Slavery was a tragedy. It had many aspects. Each created problems."
- The restated prompt. Repeating the assignment back to the person who wrote it. Toronto: "Don't repeat the assignment specifications using the professor's wording."
- The Webster's Dictionary opening. Anyone can copy a definition; UNC notes that the dictionary is also a weak authority for course concepts. Harvard gives the same ban. Toronto: "Don't provide dictionary definitions, especially of words your audience already knows."
- The "dawn of man" opening. "Since the dawn of man, slavery has been a problem" (UNC's example). Sweeping history that never reaches the thesis.
- The book-report opening. Author, title, publisher, and plot for a text the instructor already assigned.
If you need a definition, use the one your course or field actually argues over, and put it where the argument needs it. A term that is complicated belongs in a later paragraph or a dedicated subsection, not in sentence one as a stall.
Do you label the section "Introduction"? APA, MLA, Chicago, IEEE
In APA Style 7th edition, the answer is no unless the instructor explicitly asks for the word. The student-paper setup guide (updated 8 October 2025) says to begin with background and context and not to use the label unless it is specifically requested. Repeat the paper title, centred and bold, at the top of the first text page. That title is the Level 1 heading. If you need subsections inside the opening, APA says to use Level 2 headings, then a Level 1 heading for the next main section (Method, for example). See also headings and subheadings and paper format.
APA 7: the opening is already the introduction
APA Style student-paper setup guide
Begin with an introduction to provide background and context. Do not use the label “Introduction” unless specifically requested; the beginning of the text is assumed to be the introduction.
| Style / venue | Label the opening "Introduction"? | What sits above the first paragraph |
|---|---|---|
| APA 7 student or professional paper | No, unless the assignment demands it | The paper title, bold and centred (de facto Level 1) |
| MLA 9 essay | Usually no; MLA does not require the heading | The paper title, centred, title case, not bold |
| Chicago / Turabian course paper | Optional; follow the instructor or department template | Title page or first-page title, then the opening prose |
| IMRaD journal (ICMJE, most STEM) | Yes; Introduction is a named section | Abstract (and keywords) before the Introduction heading |
| IEEE transactions and many CS proceedings | Yes; numbered 1. Introduction is standard | Title, authors, abstract, then the numbered heading |
Putting "Introduction" on an APA student paper is one of the most common format marks lost on an otherwise sound opening. Putting no heading on a journal submission that uses IMRaD is the opposite error. Match the template, not a single universal rule.
How long should an introduction be?
There is no official word count in APA, MLA, Chicago, IEEE, or ICMJE. Length follows genre and the amount of orientation the thesis or hypothesis requires. Toronto's half-to-three-quarters-of-the-first-page rule is the best published planning number for course essays. A twenty-page paper may call for a two-page introduction; a five-page paper will not. Harvard notes that some Government courses want a short opening with a preview, while other disciplines accept two paragraphs or more. ICMJE specifies content, not paragraphs. CONSORT 2025 specifies what a trial introduction must cover, not how many words it takes. Use the table as a planning range, then cut anything that does not serve the problem, the niche, and the purpose.
| Document | Typical opening | What "too long" usually means |
|---|---|---|
| Short undergraduate essay (about 3–7 pages) | One paragraph, often 120–200 words | A second paragraph of background the body will repeat |
| Course research paper (about 8–15 pages) | One or two paragraphs, sometimes up to a page | A hidden literature review before the thesis appears |
| IMRaD journal article | Often 3–8 paragraphs / roughly 300–800 words | A systematic review; or results leaking into Move 3 |
| Randomised trial (CONSORT 2025) | Long enough for Item 6 plus a PICO-style Item 7 | A methods dump, or no systematic-review justification |
| Letter / Nature-length article | A few hundred words or less | Textbook background a specialist reader does not need |
| Thesis or dissertation Chapter 1 | A chapter, often with subsections | Methods, results, or a second literature-review chapter |
A thesis chapter called "Introduction" is a different genre from the first page of a ten-page seminar paper. It can include a longer stake, a chapter map, and more context. It still should not swallow the literature review or the methodology.
What changes by discipline
Sciences, medicine, and engineering
Lead with the problem and its significance, then the specific objective or hypothesis. ICMJE's three-sentence rule is the cleanest checklist in medicine: context, purpose, pertinent citations only, no results. CONSORT 2025 (Hopewell et al., BMJ 388:e081124) tells trial authors to justify a new trial against available evidence — ideally a systematic review — and to state objectives in participant–intervention–comparator–outcome language. IEEE wants the literature review in the introduction to show novelty, plus the question the paper will answer. Engineering papers often add a short statement of the approach or prototype after the gap ("we present a…") because the contribution is the design. Keep methods short here; the methodology section is where procedures become reproducible. Citation style is often APA in psychology-adjacent work, Vancouver or AMA in clinical journals, ACS in chemistry, or IEEE in electrical engineering — the syllabus or author guidelines win.
Social and behavioural sciences
CARS is native here. Name the theory or construct early enough that the gap is intelligible. Review competing findings, not every paper in the database. State questions or hypotheses in operational language (who, what is measured, what relationship). Practical significance ("so what for schools, clinics, or policy") belongs in Move 3 or the end of Move 1, not as a slogan in sentence one. USC also wants a delimitation: say what the study will not cover so a reader does not ask why you ignored an obvious alternative. APA 7 is the default in psychology, education, and nursing unless told otherwise.
Humanities
The opening is usually a problem plus a thesis, not a hypothesis. Harvard's stake test matters more than a literature gap: what misreading does the essay correct? You may bring a primary source into the first paragraph if the thesis will interpret that moment. Harvard's Hirsch sample counts the word "ear" in Hamlet before stating the thesis; UNC's Douglass examples start from the assigned Narrative. You should not open with publication facts the instructor already knows. MLA and Chicago notes are the usual documentation systems; they affect footnotes and the works cited, not whether the first page needs a hook.
How to cite in the introduction without turning it into a review
Introductions are where citation density goes wrong in both directions. Under-citing makes a gap look like a hunch. Over-citing makes Move 1 look like a database dump. ICMJE's "directly pertinent" test is the right one even outside medicine: if deleting the citation would not change the niche, it is probably body-text material. CONSORT 2025 is stricter for trials: if a systematic review of previous trials exists, the introduction should cite it.
- Cite the studies you are departing from, not every study on the topic.
- Prefer paraphrase for findings; reserve quotation for wording you will interpret.
- Keep locators (page or paragraph) when you quote or point at a specific claim.
- Do not invent a source to make the gap look real. If you cannot cite the absence, narrow the claim ("in this sample of reviews," "in the journals this course assigned").
- Format every in-text citation in the paper's style from the first sentence. Switching from informal names to APA on page three reads as unfinished.
- A missing year, a hanging name with no reference-list entry, and a title in the wrong case are the errors a citation checker catches after you draft.
If you are still gathering sources, stop and find scholarly work before polishing sentences. An elegant opening attached to weak evidence is still a weak paper. When the sources are in hand, generate the references in the assigned style rather than guessing punctuation.
Tense, person, and voice in the opening
| What you are doing | Usual tense | Notes |
|---|---|---|
| Stating established knowledge or a theory's current form | Present | "Attention mechanisms are part of encoder-decoder models." |
| Reporting a specific prior study's procedure or finding | Past | "Hochreiter and Schmidhuber (1997) introduced LSTM." |
| Announcing the aim of this paper | Present or present perfect | "This paper tests…" / "We propose…" |
| Previewing results (only if the venue wants it) | Present or past, consistently | Do not do this in ICMJE-style Introductions |
| Close reading of a primary text | Present | "Douglass treats literacy as…" |
First person ("I argue," "we tested") is acceptable in many humanities essays and in a growing number of science journals. APA 7 allows first person for your own actions. Harvard's Blatt sample uses "I hypothesize." Some lab reports and some instructors still want third person. Passive voice is a tool for methods ("participants were assigned"), not a requirement for every opening sentence. If the active/passive choice hides who did the thinking, rewrite. Clarity in the first paragraph is worth more than a Latinate vocabulary.
Worked examples from publications and writing-center samples you can open
Do not learn introductions from invented "Smith (2024)" papers. The science examples below are real, citable papers. The humanities and course-science examples are official teaching samples from Harvard and UNC, using Douglass's, Shakespeare's, and Blatt's own wording — not fabricated journal articles.
Announce-first: Watson and Crick (1953)
James D. Watson and Francis H. C. Crick, "Molecular Structure of Nucleic Acids: A Structure for Deoxyribose Nucleic Acid," Nature 171, no. 4356 (25 April 1953): 737–738, https://doi.org/10.1038/171737a0. The paper is a letter, not a modern IMRaD article, and it occupies the niche in the first two sentences.
Opening (Watson and Crick 1953)
We wish to suggest a structure for the salt of deoxyribose nucleic acid (D.N.A.). This structure has novel features which are of considerable biological interest.
What to steal
Move 3 comes first because the contribution is the structure. The next paragraphs then review Pauling and Corey's competing three-chain model and a note by Fraser — Move 1 in service of a niche already claimed. This pattern fits letters and some engineering notes. It is a poor template for a student psychology report that still owes the reader a gap.
Classic CARS: Vaswani et al. (2017)
Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin, "Attention Is All You Need," in Advances in Neural Information Processing Systems 30 (2017), https://doi.org/10.48550/arXiv.1706.03762. The NeurIPS PDF reprints the same purpose sentence. The introduction is a textbook CARS sequence.
Territory, niche, occupation (compressed from the published Introduction)
Move 1 (territory)
Recurrent networks, especially LSTM and gated recurrent models, were the established approach to sequence modelling and translation. Those models compute hidden states one time step at a time, which blocks parallel training on long sequences. Attention was already widely used, but almost always with a recurrent network.
Move 2 (niche)
The sequential constraint remains even after efficiency tricks. Attention-only architectures for transduction had not replaced recurrence.
Move 3 (their wording)
In this work we propose the Transformer, a model architecture eschewing recurrence and instead relying entirely on an attention mechanism to draw global dependencies between input and output.
What to steal
Every citation in those paragraphs supports the niche. There is no history-of-computing preamble. The purpose sentence names the model and the claim ("entirely on an attention mechanism") so a reader can predict the rest of the paper.
Harvard course science: Blatt on Crane Beach dunes
Harvard's introductions across disciplines page annotates a biology introduction by Talia Blatt ’23 for "Ecology: Populations, Communities, and Ecosystems." The sample orients on coastal-dune succession and chronosequence methods, marks the Clements-versus-Gleason stake, then occupies the niche with a testable hypothesis at Crane Beach in Ipswich, Massachusetts. That is Harvard's three pieces inside a science assignment — not a fake "solar rebate" study.
Official Harvard teaching sample (keep English in every locale)
Orienting information plus stake
Coastal dunes are key sites for the study of ecological succession – the process of structural change in an ecological community after a disturbance (Martinez et al., 2018). … Theories of succession remain contested: Frederic Clements envisioned a steady progression towards a climax community, which has been the traditional understanding. Conversely, Henry Gleason imagined a more contingent, polyclimax theory of succession, in which different organisms are adapted to live together in different environments.
What to steal
The first sentence is already inside ecological succession. The stake is a named theoretical dispute, not "people care about beaches." The hypothesis (quoted above) is specific enough to be wrong. Harvard annotated this sample to show orienting information, stake, and thesis in one opening.
Harvard course essay: Hirsch on Hamlet
The same Harvard page annotates Eliza Hirsch ’25, written for "Why Shakespeare?" The opening counts how often Hamlet mentions ears (twenty-five times; eight in act one) against "mouth" (five) and "nose" (two), then asks why the play joins words to life-threatening physical violence. The thesis — the motif of attacked ears equates physical and verbal violence — is disputable and already inside the text.
Official Harvard teaching sample
Hirsch thesis sentence
In Hamlet, the motif of ears being physically attacked creates a connection between physical and verbal violence and ultimately suggests that the two are equal.
What to steal
The puzzle is already in the assigned play. No dictionary, no dawn-of-time claim, no book-report publication facts. A reader knows which scenes must appear in the body.
Course essay: a problem-and-thesis opening (Douglass)
UNC builds its introductions handout around an assignment on Narrative of the Life of Frederick Douglass, an American Slave (1845). The useful first move is not "Webster defines slavery." It is a tension already in the text. Douglass writes that "education and slavery were incompatible with each other." A student introduction can start from that incompatibility and end on a thesis about how the Narrative uses interrupted literacy lessons as political evidence, not merely autobiography. That is Harvard's pattern: problem, stake, thesis. It is not CARS, and it should not pretend to be a social-science gap statement.
Illustration (student-essay scale, not a published article)
Student-essay illustration
Douglass's claim that "education and slavery were incompatible with each other" is easy to praise as a defence of literacy and easy to misread as a story of individual uplift. The Narrative spends as much energy on the moment Hugh Auld forbids Sophia Auld to continue the lessons as it does on any later skill Douglass acquires. This essay argues that the interruption is the political centre of the education plot: the text treats a literate enslaved person as a public danger to the slaveholding order, which is why the lessons have to stop and why Douglass presents their stopping as evidence, not only as personal loss.
What to steal
The quotation is from the assigned primary source, not a motivational poster. The thesis is disputable. A reader knows which scenes must appear in the body. Nothing here cites a fake "Jones (2024)" study of social media.
Do / don't
Do
- Match essay-with-thesis, article-with-gap, or trial-with-PICO, not all at once
- Write or rewrite the opening after the body exists
- Cite only the work the niche needs (and a systematic review, in a trial)
- Put the purpose, question, thesis, or hypothesis where a skimmer will find it
- Keep the first sentence inside the problem's scale
- Follow APA's no-"Introduction"-heading rule on APA papers
- Explain what this paper adds, not only why the topic is important
- Name delimitations in social-science papers so ignored alternatives are a choice
Don't
- Open with a dictionary, a dawn-of-time claim, or a book report
- Paste the assignment prompt back to the instructor
- Turn Move 1 into the literature-review chapter
- Report IMRaD results in the introduction unless the journal wants a preview
- Apologise ("this humble paper cannot…") or overclaim ("this study will solve…")
- Switch citation styles mid-opening
- Leave a placeholder first page in the file you submit
- Invent a "Smith (2024)" paper to make the gap look real
Revision checklist
Read the opening as a stranger would
- A colleague can predict the paper's argument or study from the opening alone (UNC's test)
- The first sentence is already inside the problem, not in human history
- Orienting information is the minimum a non-expert needs for the thesis or hypothesis
- The niche or stake is specific enough to be wrong
- The purpose statement matches the body and the conclusion
- Prior work is synthesised, not listed as "Author (Year) found… Author (Year) found…"
- No results appear in an ICMJE-style Introduction
- A trial introduction cites a systematic review when one exists (CONSORT 2025 Item 6) and states a PICO-style objective (Item 7)
- Section labelling matches APA, MLA, Chicago, IEEE, or the journal template
- Every citation has a reference-list or bibliography entry in one style
- Length is justified by genre, not by anxiety
- Specialised terms are defined only when the reader cannot proceed without them
- The abstract, if any, does not merely duplicate this page
Frequently asked questions
- How do you start an introduction for a research paper?
- Start inside the research problem, not with a dictionary or a history of humanity. In an IMRaD article, name the territory and move quickly to the gap, then state the purpose, question, or hypothesis. In a course essay, frame an analytical problem a thoughtful reader could have and end the opening on a thesis. Harvard College Writing Center: the introduction presents a problem and then an answer to that problem.
- What are the three parts of an introduction?
- For essays, Harvard groups them as orienting information, what is at stake, and the thesis. For research articles, Swales's CARS model uses three moves: establish a territory, establish a niche, occupy the niche. USC adds four reader questions (what, why, what was known, what this study adds). Hooks, anecdotes, and method previews are optional extras, not a required third of the page.
- How long should a research paper introduction be?
- Long enough to frame the problem and state the purpose, short enough that it is not a literature review. Toronto says most papers can use one paragraph covering half to three-quarters of the first page. Short essays often run about 120–200 words. Course research papers often use one or two paragraphs, sometimes up to a page. Many journal Introductions run a few hundred words across several paragraphs. Thesis Chapter 1 is a chapter, not a paragraph. There is no official percentage in APA, MLA, Chicago, IEEE, or ICMJE.
- Should I write the introduction first or last?
- Draft a placeholder if you need momentum, but write the version you submit after the body exists. UNC's introductions handout recommends writing the introduction last so it matches the evidence. Toronto says you may write it first and then rewrite it in light of the finished body. USC tells you to reread the opening after the study is done so promised methods and outcomes still match. Purpose statements written before the research is finished describe a different paper.
- Do you put a heading on the introduction in APA?
- No, not in APA 7 unless the instructor requires the word. Repeat the paper title at the top of the first page of text; APA treats that title as the Level 1 heading. The October 2025 student-paper setup guide repeats the same rule. Use Level 2 headings only if the opening itself has subsections. Journal IMRaD templates and IEEE papers are the opposite: they expect a section titled Introduction (often numbered 1. Introduction).
- What is the CARS model of introductions?
- CARS (Create a Research Space) is John Swales's description of how research-article introductions work: Move 1 establishes a territory, Move 2 establishes a niche (gap, counter-claim, question, or continuation), and Move 3 occupies the niche by stating the present purpose. It comes from Genre Analysis (1990) and is taught in Academic Writing for Graduate Students (Swales and Feak, 3rd ed., 2012). USC Libraries restates the same three moves. It is a pattern for journal articles, not a law for every humanities essay.
- What is the difference between an introduction and an abstract?
- An abstract is a standalone summary, often 150–300 words, written so a database record makes sense without the paper. Informative abstracts include results. An introduction is the first section of the paper. It orients the reader, frames a gap or problem, and states a purpose. ICMJE tells authors not to put the study's own data or conclusions in the Introduction. Write the abstract from the finished paper; see How to write an abstract.
- Can I use "I" in a research paper introduction?
- Often yes. APA 7 permits first person for your own research actions ("I analysed," "we recruited"). Many humanities essays prefer "I argue." Harvard's official Blatt ecology sample uses "I hypothesize." Some lab instructors and some journals still want third person or passive methods sentences. Follow the assignment or author guidelines; do not avoid "I" by writing dangling passives that hide the actor.
- Should a research paper introduction include a hook?
- Only if the hook is already a piece of the problem. Harvard warns against dramatic promises and random startling facts, and against the "dawn of time" love-stories opener. UNC allows an intriguing example, a quotation from the text under discussion, or a pointed question, and rejects dictionary, dawn-of-man, and book-report openers. IMRaD articles usually start on the research problem, not on a classroom anecdote.
- What must a clinical-trial introduction include under CONSORT 2025?
- Item 6: scientific background and rationale, including why a new trial is needed and a citation to a systematic review of previous trials when one exists. Item 7: specific objectives related to benefits and harms, naming participants, intervention, comparator, primary outcome, and the timepoint of that outcome. Hopewell and colleagues (BMJ 2025;388:e081124) give the published example of evaluating a structured exercise programme versus usual care for women at high risk of upper limb disability after breast cancer surgery. Do not put the trial's own results in the Introduction; ICMJE still forbids that.
Cite the sources the opening actually uses
Once the purpose statement is stable, format every in-text citation and reference-list entry in one style. Guessing commas is slower than generating the record and checking it against the manual.