Results vs discussion: what goes where

Results versus discussion is one of the most common structural problems supervisors flag, and one of the easiest to fix once you hold the rule firmly: the results chapter reports what you found; the discussion chapter argues what it means. Evidence lives in one place, inference in the other, and every sentence in either chapter can be sorted by asking which of the two it is doing.

Simple to state, harder to apply. Real sentences blur the line constantly: a comparison to a prior study, an adjective like 'surprisingly', a robustness check you ran only after seeing the data. This guide gives you the clean division, the gray areas that trip students up, a one-question test that sorts any sentence, and the cases where a combined chapter is genuinely the right call.

Why the results and discussion split exists

The separation is not bureaucratic tradition; it is what lets an examiner audit your work. When evidence is reported neutrally in one chapter, a reader can check it, re-derive it, and, crucially, disagree with your interpretation while still trusting your data. Mixing the two forces the reader to untangle observation from opinion on every page, and readers made to work that hard become suspicious readers.

The split also protects you. In the viva, an examiner who challenges your interpretation is challenging the discussion; your results stand untouched behind it. A thesis that entangles the two puts the data itself inside every argument, which is a far worse position to defend from. Keep the evidence clean and you can concede an interpretation gracefully without losing anything else.

There is a drafting benefit as well. Writing the results chapter first, flat and complete, gives you a stable evidence base to argue from; the discussion then becomes a series of deliberate moves over known ground rather than a scramble to justify numbers that keep shifting as the analysis settles. Each chapter having exactly one job also makes both of them easier to start.

What belongs in the results chapter

The results chapter is a neutral report of what the study produced, in the order the analysis plan promised:

• Sample characteristics and descriptive statistics.
• The output of each planned analysis, reported without evaluation: effect, direction, uncertainty.
• Tables and figures with captions complete enough to stand alone.
• Signposting prose that walks the reader through the tables, stating what each shows.
• Deviations from the analysis plan, stated as fact with the reason.
• Null and negative results, given the same prominence as the results you hoped for.

What belongs in the discussion chapter

The tone of the results chapter is deliberately flat: 'Group A scored higher than Group B' belongs, while 'encouragingly, Group A scored higher' does not, and every evaluative word you keep out of that chapter strengthens it. The discussion is where judgment lives. It takes the reported findings as given and builds the argument:

• An explicit answer to each research question.
• Interpretation: mechanisms, explanations, what the findings mean.
• Comparison with prior literature: agreement, extension, contradiction, and why.
• Alternative explanations, and your reasons for weighing them as you do.
• Limitations, tied to their effect on each inference.
• Implications and recommendations, scoped to the evidence.

The gray areas that trip students up

Notice what the two lists share: nothing. Any sentence that would sit comfortably in both chapters is usually doing neither job well and should be split into a reporting half and an interpreting half. The gray areas are where that splitting takes actual thought.

Comparisons to prior literature are the biggest one. 'This is consistent with earlier findings' is interpretation and belongs in the discussion, even though it feels like reporting. In most thesis formats the results chapter cites no external literature at all; some disciplines allow brief methodological citations, but the default is none. If your results chapter has grown a reference list of its own, something has leaked.

Evaluative adverbs are the quiet one. 'Surprisingly', 'importantly', 'only', 'as expected': each smuggles a judgment into a report. Surprise is relative to expectations, and expectations are an argument. Move the surprise to the discussion, where you can actually discuss it. Qualitative work has its own version: presenting a theme with supporting quotes is results; explaining why the theme emerged and what it implies is discussion. Robustness and sensitivity analyses have their own rule, covered below, and mixed-methods theses hit every one of these edges at once, which is why the sentence-level test that follows earns its keep.

A one-question test for any sentence

When you cannot decide where a sentence goes, ask: could a careful reader verify this sentence from my data alone? If yes, it is results. If verifying it requires judgment, prior literature, or a theory, it is discussion. The question is not whether the sentence is true; it is what kind of warrant the sentence needs.

'Mean response time fell after the intervention' is verifiable from the data: results. 'The fall in response time suggests the intervention reduced cognitive load' requires a theory linking time to load: discussion. 'This mirrors what earlier studies observed in adult samples' requires the literature: discussion. The test is mechanical on purpose; it works precisely when your intuition is tired.

Run it over any paragraph you are unsure about, sentence by sentence. Most misplaced paragraphs are not wholly misplaced; they are reporting sentences with one inferential clause attached, and only the clause has to move.

Where robustness checks and post hoc analyses go

Analyses you ran because of what you saw, post hoc tests, sensitivity analyses, robustness checks, are still reported in the results chapter, clearly labeled as post hoc or exploratory. The label is the honesty; the location does not change. What moves to the discussion is the reasoning: why you ran them, and what they rescue or undermine in the main findings.

The same holds for assumption checks. The check and its outcome are results; the decision it justified ('a non-parametric test was therefore used') can be stated as fact in the results, but any argument about what the violation implies for the findings belongs in the discussion. Report the fork you took; argue about it later.

Tense and voice: the quiet signals of each chapter

Results prose is written almost entirely in the past tense: you report what the data showed. The discussion mixes tenses deliberately: past for your findings, present for claims you now advance as true, present or future for implications. If you find present-tense truth claims sitting inside the results chapter, that is often interpretation leaking in through the grammar rather than the content.

Voice differs too. Results prose is procedural and impersonal even where first person is permitted. Discussion prose argues, and arguments have an author: 'we interpret this as' and 'a more plausible explanation is' belong there and only there. Reading a page aloud usually reveals which chapter it actually belongs to, whatever heading currently sits above it. It is a cheap and surprisingly reliable diagnostic, and it doubles as a proofread.

When a combined results and discussion chapter works

Some fields, and much qualitative research, use a combined 'Findings and Discussion' chapter, presenting each theme or study and interpreting it immediately. Where interpretation is inseparable from presentation, as in much interpretive qualitative work, this structure serves the reader better than an artificial split that forces constant page-flipping.

Two cautions. First, this is a structural choice your program and supervisor must endorse; check the handbook and ask before committing a hundred pages to it. Second, combining chapters does not suspend the sentence-level rule: within each section, report first, then interpret, and keep the two visibly distinct. A combined chapter is a different arrangement of the same discipline, not an exemption from it.

If you do combine, borrow the signposting habit from the split format: open each themed section with the finding, close it with the interpretation, and let subheadings mark the turn. Readers, and examiners in particular, should always know which mode the prose is currently in.

Discipline differences you should respect

Lab and quantitative sciences usually hold the strict split, often with a short, dense results chapter and a longer discussion. Social sciences vary by method: quantitative theses mirror the sciences, while qualitative and mixed-methods theses more often combine. Humanities theses frequently have no results chapter at all; the argument itself is the structure.

Multi-study theses add one more pattern: results and discussion per study, plus a general discussion that integrates across studies. If that is your shape, the general discussion is where the thesis-level argument lives, and it must do more than summarize the per-study discussions in sequence.

In every case the authority is local: your handbook, your supervisor, and recently passed theses from your own department outrank any general guide, this one included. Borrow the structure of a thesis your examiners have already accepted.

Keeping the split honest in CiteDash

The workflow maps cleanly onto the two chapters. For results, CiteDash's Data Analysis takes a CSV upload, runs any of 13 named statistical tests, and produces charts, which gives you neutral, reportable output to write up; the full walk from raw file to written results section is covered in dissertation statistics: CSV to Results chapter.

For the discussion, drafting in the Thesis Editor is grounded: the comparison-with-literature paragraphs cite database-backed papers, never free text, and the Fact Checker verifies claims against held full-text PDFs before they reach you. The chapter that leans hardest on prior work is the one where citation integrity matters most, and it is the one place you should never rely on an assistant's memory of a paper.

When late revisions move sentences between the two chapters, run your sorting pass again before submission. Misplacement usually re-enters a thesis during editing, not during drafting, because a sentence trimmed for length so often keeps its number and loses its neutrality.

A sorting checklist before you submit

Before the full draft goes to your supervisor or the examinations office, run this pass:

• Every table and figure is referenced in the results prose.
• No citations to prior literature inside the results chapter, unless your discipline explicitly expects them.
• Every research question has one explicit answer sentence in the discussion.
• No number appears in the discussion that was not first reported in the results.
• Evaluative adverbs ('surprisingly', 'importantly') are gone from results prose.
• Limitations appear once, in the discussion, not scattered through the results.