How to write a research paper

Before you write: the main finding and the target journal

Settle two things before the first sentence.

First is the main finding of the paper, written in two sentences only. Many scholars write the topic here. “We studied the adoption of UPI among small retailers in Coimbatore” is only a topic. Compare it with a real finding. “Among 412 small retailers in Coimbatore, perceived transaction cost predicted UPI adoption more strongly than perceived security, reversing the pattern reported in urban studies.” Suppose only the first kind is coming to you, then hold the draft for now. Mostly the analysis is incomplete, or two papers are squeezed into one.

A PhD thesis usually gives two to four papers, and every paper needs a main finding of its own. Many first drafts run to 12,000 words and read like a project report. Mostly a full chapter was pasted into the paper.

Second, choose the journal before you draft. The journal decides the word limit and the reference style. It also decides whether the abstract is structured, how long the introduction can go and who will read the paper. A paper written for “any Scopus journal” suits none of them. For making a shortlist, see our guide on how to choose a journal. After you pick one, download two or three recent articles from it close to your topic. Keep these articles open beside the draft, they show what that journal really accepts, better than the guidelines.

The IMRaD structure and what each section does

Some scholars treat IMRaD as a formality. Actually each section answers one question. Readers go directly to the section that has their answer. A reviewer who wants to check your sample size will open Methods. Someone deciding whether to cite your work will read the abstract and the first paragraph of the Discussion.

These word counts take a 6,000-word limit, which is common in management and social science. Engineering and life-science journals many times allow 3,500 to 5,000 words, and letters or short communications allow far less. When the journal limit is smaller, you trim each section a little, but the content stays complete.

A few journals combine Results and Discussion into a single section. Some keep Methods at the end of the paper. Others add a Literature Review or Theoretical Framework section between the Introduction and Methods. You see this mostly in management and education. Go with the journal. The logic underneath stays the same, from problem to method to evidence to meaning.

The order in which to write a research paper

A reader goes through a paper from title to references. Don’t write it in that order. Started first, the introduction is the hardest part, because you don’t yet know exactly what the paper will claim. Most experienced authors start from the evidence and work outwards, in this order.

1. 1

   Tables and figures

   Pick the three to six displays that carry your findings. The whole paper gets built around these.

2. 2

   Methods

   This one is the easiest, because you already know what you did, and the small details are still fresh in your mind now.

3. 3

   Results

   Go through the tables and figures in a sensible order and describe what they show, but the meaning of it you can leave for the Discussion.

4. 4

   Discussion

   Here you interpret the main findings, compare them with earlier studies and write the limitations.

5. 5

   Introduction

   By now you know exactly what the paper shows. So setting up the gap it fills becomes much easier.

6. 6

   Abstract and title

   Keep these for the end and write them from the finished paper, so they match what is really inside.

Tables and figures force you to decide early which comparisons really matter and which are only supporting detail for a supplementary file. If 14 tables are coming, that is actually two papers or a thesis chapter.

Do the first draft fast and rough. A complete rough draft is much easier to fix. Scholars who polish the introduction for one month many times never reach the discussion.

What to put in each section of a research paper

Title

Keep the title at 10 to 20 words. Name the main variables and the population or material, and where useful, the design also. “Effect of fly ash replacement on the compressive strength of M30 concrete: an experimental study” is a good title. “A study on concrete” is not. Keep abbreviations out of the title, and those openings like “A study of” or “An investigation into” are just wasting words there.

Introduction

The introduction works like a funnel of three or four paragraphs. (The CARS pattern is explained with an annotated example in our introduction guide.) Begin with the broad problem and why it is important, supported by a citation or a figure. Then come down to what is already known on your own question. Next show what is still not known, and why that missing piece is a problem. In the last paragraph give your aim or research question, and in many fields a line on how you went about the study. Keep the literature focused on the gap. A two-page history of the field belongs in your thesis. If the gap is not coming in one sentence, get it down to one line in your notes before drafting. The same line will help you later when writing the abstract.

Methods

Give enough detail for another person to repeat the study. Usually that covers the study design, the setting and dates, and the sample with how it was selected. It covers the materials, instruments or questionnaires and their source. After that comes the procedure, and then the analysis with the software and version (SPSS 29, R 4.4, MATLAB R2024b and so on). Wherever a reviewer may question a choice, give the reason, for example why you took a convenience sample or used a particular statistical test. Our guide on choosing a statistical test is there for scholars confused about the test. Where your study had people or animals, write the ethics approval and consent details here itself.

In many fields there are reporting guidelines which journals expect authors to follow. CONSORT is for randomised trials, STROBE for observational studies and PRISMA for systematic reviews, and there are others too. The full list is kept by the EQUATOR Network. The checklist of whichever one fits your design should be open while you are writing the Methods itself.

Results

Give the findings in the same order as your objectives or hypotheses, with the most important result first. Point to the patterns and key numbers in the text, without repeating every table cell. Along with effect sizes put the confidence intervals or standard deviations, since a reviewer gets very little from p-values alone. Interpretation of the numbers you hold back for the Discussion. “Group A scored higher than Group B (mean difference 4.2, 95% CI 1.8 to 6.6)” is a Results sentence. “This suggests that training improves confidence” goes in the Discussion.

Discussion

Start the discussion with a short paragraph giving the main finding again in simple words. Then put it next to earlier studies and show where your result agrees with earlier researches. Where it differs, discuss about a reasonable explanation for the difference. After this write the implications for theory or practice, the limitations, and at last some ideas for future work. Reviewers in every field look for overclaiming first, so go slow with your claims. Cross-sectional survey data can’t show causation, whatever you are hoping.

Conclusion

Many journals ask for a short conclusion as a separate section. Here just a few sentences on what the study found and why it is important will do. Don’t bring in new results or new citations here.

References

Only cite papers you have read yourself, never one copied from another paper’s reference list. Match each reference with the original for authors, year, volume, pages and DOI. Start using a reference manager from the first draft itself. After a rejection, changing 60 references by hand to a new journal style takes one full day.

Language and style for scholars writing in English as a second language

Reviewers don’t expect beautiful English from you. They want English they can understand without effort, and a few habits take care of most of that.

• One idea per sentence. Grammar mistakes and confusing meaning mostly come in when a sentence carries many clauses. Break such sentences into two or three.
• Use the same term for the same thing throughout. If you wrote “customer loyalty” in the introduction, keep the same words in the results also and don’t change to “consumer retention”. Readers will think a new word means a new idea.
• Tense. Use past tense for what you did and what you found (“we measured”, “scores increased”). For established knowledge and for what your tables show, use present tense (“Table 2 shows”).
• Drop the long formal phrases. Instead of “It is observed that there is an increase in”, simply write “X increased”. Indian academic English many times carries this formal padding from report-writing habits, and journal editors will remove it anyway.
• Active voice is acceptable in most journals now, and so is “we”. See how recent articles in your target journal handle it.

Read the draft aloud once. Where you stumble, a reviewer will stumble too.

On AI writing tools

Most major publishers now allow some limited use of AI tools for language editing, provided you disclose it. No publisher accepts an AI tool as an author. The policies are different from one publisher to another and still changing, so go through the current policy of your target journal. No policy permits text or references which you didn’t check yourself. AI tools many times make up citations that look completely genuine, and even one fake reference can sink the paper.

Revising and checking before you submit

Keep the finished draft aside for a few days. Then read it like a reviewer. It is even better if you give it to colleague who works outside your topic. After reading only the abstract and introduction, can they tell you the main finding? If not, it isn’t clear yet.

• The research question or objective comes in the last paragraph of the introduction, as one clear sentence.
• Each table and figure is cited in the text in number order, and each one can be understood without reading the text.
• The numbers are the same in the abstract, the results text, the tables and the figure labels.
• Each claim in the discussion can be traced back to one specific result.
• Limitations are explained plainly, and for each one you say how it affects your conclusions.
• All in-text citations are there in the reference list and all references are cited in the text, in the journal’s style.
• Word count, abstract length, keywords and heading levels are as per the author guidelines of the journal you picked.
• Declarations are ready for ethics approval, funding, conflicts of interest, data availability and author contributions.

Before submitting, run a similarity check and see what it is flagging. Text you reused from your own thesis or from an earlier conference paper is also counted. For that, most journals want you to rewrite the text or else tell them about the earlier version. The technical faults that get a paper sent back without review are covered in our guide to why papers get desk rejected. For the letter you send with the paper, see the cover letter guide.

Papers that do not follow IMRaD

Review articles are arranged by theme, and a systematic review will also have a methods section describing the search. Case studies and case reports go along the story of the case. Humanities papers are usually argued essays, and their headings follow the argument. All of them still do the same basic work of stating the problem, showing the evidence and explaining what it means.

If you want a subject specialist to check the structure and argument of your draft before submission, our research paper review does that kind of review, while the paper and the findings in it are still yours only.