Writing an Empirical Paper: Structure, Appendix, Example

Two papers, the same survey, the same 180 responses, two very different grades. Students who write an empirical paper rarely fail because of the data, and often because of how the data get written up.

The break happens in the same place almost every time: finding and meaning get mixed together. You write an empirical paper by building the theory and the research question first, then disclosing in a methods section how you collected and analyzed the data, reporting the findings without judgment, and placing them in context only in the discussion, where evaluation is finally allowed. By the end you will have an outline you can hand to your advisor.

📌 The short version

• Rackham calls the doctoral text a dissertation, the master’s a thesis.
• Many US master’s programs require a capstone instead of a thesis.
• Purdue OWL puts interpretation in the discussion, not the results.
• 45 CFR 46.104 exempts many surveys, but not from IRB rules.
• No federal US law governs student survey data like the GDPR.

Empirical Paper vs. Literature Review: What Actually Differs

An empirical paper is a piece of academic writing that tests its research question against data instead of answering it from the literature alone, and that discloses how those data were produced and by what rules they were analyzed.

Empirical does not have to mean that you send out your own questionnaire. Federal regulation treats the reuse of existing data as research in its own right: 45 CFR 46.104(d)(4) sets out a category for “secondary research uses of identifiable private information,” and the Common Rule counts you in, since 45 CFR 46.102(e)(1) speaks of “an investigator (whether professional or student).” Reanalyzing an archived dataset is empirical work.

Both forms carry the same weight, they simply answer different kinds of questions. A question about the state of a debate needs no new data collection, a question about how one specific group behaves does. If the choice is still open, the overview of empirical research helps you decide.

How to Structure an Empirical Paper: A Sample Outline

The structure of an empirical paper follows the same two-part split almost everywhere in the US: a theoretical part that defines terms and leads into the research question, then an empirical part with methods, results, and discussion.

The names come first, because they differ from British usage. In the United States the doctoral text is the dissertation, the master’s text the thesis. The master’s policy of the University of Michigan’s Rackham Graduate School shows it from the other side: coursework “with ‘doctoral,’ ‘dissertation,’ or ‘preliminary’ in the title” does not count toward a master’s degree. The University of Cincinnati adds a wrinkle: “A master’s thesis is required by some programs, and a master’s capstone project/experience is required in others.”

No single outline is prescribed nationally. Rackham leaves the arrangement to the field and asks one thing only: however the content is organized, the dissertation “should include an introduction and conclusion in which the student integrates the entire scope of the research that has been undertaken” (Rackham Dissertation Handbook, 2023).

1. Introduction. Why the topic matters, what you want to find out.
2. Literature review and theory. Define terms, present models, order the field.
3. Research question and hypotheses. Derived from theory, not fitted to data.
4. Methods. Design, instrument, sample, field period, analysis procedures.
5. Results. Findings ordered by hypothesis or subquestion, without evaluation.
6. Discussion. Placement in the literature, limitations of your own study.
7. Conclusion. The answer to the research question, and what stays open.
8. References and appendices. Sources, figures, instrument, dataset.

An outline is a template, not a statute. If your topic calls for it, results and discussion can be cut differently, after a check with your advisor. The guide to empirical research in final papers shows how to get there.

How Much Theory an Empirical Paper Needs

An empirical paper needs enough theory to justify every measurement decision it makes. American graduate schools publish no fixed ratio between theory and empirical work, and Rackham deliberately leaves the arrangement open. The practical test is simpler than a page count: if a reader cannot see from your theory chapter why you measured these variables and tested these hypotheses, the theory part is too thin, however long it runs.

IMRaD, and Why It Is Not a Rule in the United States

IMRaD stands for Introduction, Methods, Results, and Discussion, and it orders journal articles rather than degree work. No US graduate school requires the scheme. Rackham allows a dissertation to take “a variety of scholarly forms, including a single monograph, or an ensemble of papers, essays, manuscripts or articles.” A student paper usually adds a separate theory chapter that a journal article does not need, because the case for the design has to be made somewhere.

Our reading tip

Kate L. Turabian: A Manual for Writers of Research Papers, Theses, and Dissertations. Chicago: University of Chicago Press. Rackham lists Turabian in the bibliography of its own handbook, next to the MLA Handbook and The Chicago Manual of Style, but in the eighth revised edition of 2013 and still under the older title “A Manual for Writers of Term Papers, Theses, and Dissertations.” The University of Chicago Press dates the tenth edition, revised by Joseph Bizup, William T. FitzGerald, and Juli Parrish, to November 2026, so check which edition your program expects.

The IMRAD shape with theory, methods, results and discussion is only one pattern among several. Which others exist, what order your graduate school requires and how far down to number your headings is covered in the guide to thesis structure.

The Methods Section: What Has to Be in It, Without Gaps

A methods section has one job: a stranger should be able to repeat your data collection from it without asking you a single question. Everything needed for that belongs in it, and nothing else does.

The most common failure here is not incompleteness but the wrong genre. Many papers spend two pages explaining what an online survey is and not one sentence justifying why it suited this question. Purdue OWL is blunt about the stakes: “your study methods form a large part of your credibility as a researcher and writer.”

What belongs in the methods section

• Design, and the reason it fits the research question.
• The instrument with its origin: self-developed or adapted.
• Operationalization, meaning which item measures which construct.
• Sample, recruitment channel, and the number of cases reached.
• Field period with concrete dates.
• Pretest: how many people, and what changed because of it.
• Analysis procedures and the software you used.
• IRB determination or approval, consent, and anonymization.

One item gets skipped again and again: the pretest. Letting two or three people from your target group fill in the questionnaire costs an afternoon and catches the items nobody reads the way you meant them. Which decisions come before that is sorted out in the guide to research design.

💡 Tip

Clear the IRB before you collect anything. Rackham states the rule without an exception: “No dissertation based on the use of humans as subjects can be accepted without prior review and approval by the appropriate IRB.” Treat the review date as a hard deadline in your schedule.

Reporting Results Without Interpreting Them Yet

The results section reports what you measured, not what you think about it. Keeping those two apart is the sharpest fault line in student papers and the easiest one to avoid.

Purdue OWL describes the results section as the place where findings appear “both narrated for the readers in plain English and accompanied by statistics,” and reserves the next section for meaning: “Your discussion section is where you talk about what your results mean.” Explaining a coefficient in plain words is part of reporting. Judging it is not.

What the line looks like in practice shows up sentence by sentence.

Two orders work for the section: hypothesis by hypothesis, or along the subquestions of your research question. Either one is defensible, mixing them is not. Which procedure fits your data is covered in the guide to data analysis methods.

The Discussion: Interpret, Compare, Name the Limits

The discussion answers what your findings mean and how much weight they carry. Evaluation is allowed here and nowhere else, which is why this is the section where numbers turn into a claim.

Four moves carry a good discussion, usually in this order. Working through them one at a time gets you through the hardest section of the paper without detours.

1. Summarize. The central findings in a few sentences, without the tables.
2. Connect. Does your result confirm, contradict, or refine prior work?
3. Limit. The limitations of your study, weighed rather than listed.
4. Look ahead. The question your finding raises for a follow-up study.

Limitations often decide the impression a paper leaves. A small, non-random sample is no flaw as long as you name it and say what your claim is therefore limited to. Purdue OWL sets the dose: “Briefly consider your study’s limitations, but do not dwell on its flaws.” The standards behind that judgment are laid out in the guide to quality criteria.

⚠️ Heads-up

Limitations are a quality signal, not a confession. The mistake is not the small sample, the mistake is the sentence “The results are representative of college students in the United States” underneath it. Write instead who your finding holds for and who it does not.

What Goes in the Appendix and What Does Not

The appendix of an empirical paper holds everything a reader needs in order to judge your data collection but that would break the flow of the main text. First in line is the instrument itself, your questionnaire or your interview guide.

No nationwide rule governs the contents, and program handbooks differ in the details. Purdue OWL names the typical items, among them “the questionnaire used in the research, a detailed description of an apparatus used in the research.” The same page sets the labeling convention: a single appendix stays “Appendix,” several become “Appendix A,” “Appendix B,” and so on.

The row on licensed instruments surprises many. Rackham requires written permission from the copyright owner for anything in a dissertation that falls outside fair use, and tells students to “keep full documentation of every instance for which they have received permission.” Anyone using an established scale therefore cites it and reproduces at most a sample item. Check your own program handbook before you submit.

What Happens to Your Data After You Submit

After you submit, what happens to your raw data is decided by your IRB approval and by your institution, not by a national privacy law. The United States has no federal counterpart to the GDPR for student surveys, and that absence is the answer rather than a hole in the research.

Protection runs through the Common Rule instead. 45 CFR 46.102(e)(5) defines “identifiable private information” as private information for which the identity of the subject “is or may readily be ascertained by the investigator.” Once the identifiers are gone, the information stops being identifiable private information under that definition.

“Exempt” does not mean “no IRB.” 45 CFR 46.104(a) exempts such activities from the rest of the policy “except that such activities must comply with the requirements of this section.” An anonymous survey falls under 46.104(d)(2)(i). Collect answers that can be traced back to a person and you land in 46.104(d)(2)(iii), which requires “a limited IRB review.”

• Surveys of minors cannot use that exemption, says 46.104(b)(3).
• Class projects stay unsettled: 46.102(l) lists four exceptions, none coursework.
• Students are covered either way under 46.102(e)(1).

Retention periods come from your IRB approval and your institution, and no federal code sets a single number for degree work. Anyone who never stores names or email addresses ends up with a dataset that may be kept, because it identifies nobody. Which settings that takes is covered in the guide to data privacy in surveys. Your program’s rules govern, and this article is not legal advice.

Common Mistakes When Writing an Empirical Paper

Five mistakes turn up in empirical degree papers over and over, and all five are writing mistakes rather than research mistakes. The data collection was fine, the write-up was not.

Writing mistakes have one advantage: all five can be collected in a single revision pass, and fixing them needs neither new data nor new literature.

Interpretation Slips Into the Results Section

The most common mistake comes out of helpfulness. Writers want to make it easy for the reader and put the meaning right next to the number. The discussion then has nothing left to say and repeats itself, so both sections look half finished. The test is quick: if a sentence in the results carries a judging word such as “encouraging,” “surprising,” or “unfortunately,” the sentence belongs one section further on.

The Methods Section Explains the Method Instead of Justifying It

A definition of the online survey sits in every textbook and interests nobody in a methods section. What is wanted is the reason this form of data collection fit the research question better than an alternative you weighed. One sentence often does the whole job, along these lines: interviews would have gone deeper, but would have put the target sample size out of reach.

Raw Statistical Output Lands in the Text

Output windows in statistical software are built for the person operating the program, not for a reader. Build your own table with the statistics that bear on your hypothesis, name the units, and keep the full output for a supplementary file. Purdue OWL also sets a threshold for using a table at all: a few numbers belong in the sentence, not in a table.

The Instrument Is Missing From the Appendix

Without the questionnaire or the interview guide, not a single claim in the paper can be checked. The instrument still goes missing surprisingly often, usually because the survey tool exports no printable version. Export it as a PDF before you send it out, including skip logic and answer options. Months later the survey may be long since closed.

The Limitations Are Boilerplate

“The sample was small” is an observation, not a limitation. A limitation names which claim no longer holds because of it: with 84 participants from a single degree program, no comparison between fields is possible, while a tendency inside that program can still be described. Exactly that distinction is what shows methodological understanding.

Conclusion

An empirical paper does not win because the data are spectacular. A paper wins because every decision is documented well enough for someone else to follow it. Keeping methods, results, and interpretation apart makes your path easy to see, and that is the requirement behind all the formal rules.

In the results you report what happened. In the discussion you say what it means. That line is half your grade.

What to Do Next

• Still need to build your survey? How to create a questionnaire
• Unsure about your hypotheses? How to write a hypothesis
• Want the sequence from idea to analysis? The empirical research process

Still need to run the survey for your empirical paper?

With empirio.ai, an online survey tool from Germany, you build the questionnaire, collect the responses anonymously, and export the dataset for your appendix when you are done.

Create a survey for your paper