PlagiatScanner.de
Integrity · reporting

How to report negative results responsibly

A negative or inconclusive result is not a failed report. Present the question, design, data quality, estimate, uncertainty and limitations so readers understand what the study did not show, without turning that absence into proof that no effect or relationship exists.

INT-10 · Checked 4 September 2026

“Not significant” does not mean “no effect”

A statistical test may provide insufficient evidence against a null hypothesis under its assumptions. That does not establish the null as true. Small samples, imprecise measurement, high variability or unsuitable design can leave broad uncertainty. Report effect or association estimates, intervals, data quantity and methodological limits rather than a binary label alone.

A result pointing in the expected direction may still be inconclusive. Conversely, a precise estimate can exclude effects of practical importance even when described as negative. Interpretation begins with the question the design can answer. An observational analysis does not automatically establish causality, and an absent difference in one sample does not govern all contexts.

Reporting scheme without outcome cherry-picking

  1. Prior question: state the hypothesis or research question and whether it existed before analysis.
  2. Design: describe population, collection, comparison, measurement and analysis sufficiently to establish scope.
  3. Data flow: report approached, included, excluded and analysed cases with reasons.
  4. Primary finding: give estimate, suitable uncertainty and method regardless of direction or appeal.
  5. Data quality: explain missingness, measurement problems, deviations and material assumptions.
  6. Additional analysis: separate planned, justifiably changed and exploratory variants.
  7. Contrary results: show findings inconsistent with the preferred explanation or limited to subgroups.
  8. Interpretive boundary: state what conclusion is supported and what remains open.
  9. Literature comparison: examine differences in sample, method and setting, not only supportive studies.
  10. Next test: propose a specific better measure or study without recasting the negative result as provisional confirmation.

Use precise sentences rather than proof claims

Overstated and more defensible wording
Too strong More defensible Reason
“There is no relationship.” “The estimate was imprecise and compatible with zero in this sample.” Separates evidence from a universal claim.
“The hypothesis was disproved.” “The predefined analysis did not support the expected direction.” Links conclusion to test and design.
“The method does not work.” “Under these conditions, no measurable advantage was observed.” Defines setting and measurement.
“Only group A responded.” “The exploratory group difference requires independent testing.” Avoids overclaiming a post-hoc subgroup.

Do not use “no evidence for” to disguise a design that provided little information. Report precision and relevant effect sizes. Equivalence or non-inferiority questions require suitable pre-specified approaches; an ordinary non-significant result does not answer them automatically.

Keep all relevant outcomes visible

Outcome switching occurs when planned outcomes are replaced, reframed or omitted after results are known. Compare the report with proposal, registration or analysis plan. Changes may be methodologically necessary, but timing, reason and effect should be explicit. Report the original outcome too or explain why it could not be analysed.

Avoid spin. An unfavourable primary result does not become positive because one secondary analysis, unplanned subgroup or selective phrase looks promising. Such observations can motivate further work. Label them exploratory, list material variants and keep the principal question at the centre of the results.

Negative cases in qualitative research

Qualitative projects do not use the same significance logic but still encounter disconfirming evidence. A negative case does not fit an emerging account. Ask whether category, case boundary or theory needs revision. Record when it appeared and how previous coding was revisited. One counterexample does not mechanically invalidate every interpretation, but it cannot disappear because it complicates the narrative.

Report variation, minority perspectives and conditions in which a pattern was absent. Do not choose quotations purely for vividness. A case matrix can connect main pattern, negative case, context and effect on interpretation while avoiding details that enable re-identification.

Failure, interruption and unusable data

A technical failure or inadequate measure differs from a negative substantive result. If the instrument cannot assess the question, the conclusion is “not reliably testable under these conditions”, not “no effect”. Explain failure, diagnosis, affected data and attempted repair. Preserve originals and do not silently delete failed runs.

Stopping early may be ethically or operationally appropriate. State the rule, decision and consequence. Do not invent a success story after the event. A well-documented failure can still expose a methodological boundary without supporting broader effectiveness claims.

Final audit for a balanced account

  1. Can every pre-specified primary question be found?
  2. Do estimates and uncertainty accompany test decisions?
  3. Are exclusions and missing data fully explained?
  4. Are exploratory analyses unmistakably labelled?
  5. Are contrary findings and negative cases interpreted?
  6. Is the conclusion no broader than design and evidence?
  7. Does the literature discussion include credible disagreement?

Ask a knowledgeable reader to examine only the abstract and result tables, then state the main finding. If their account conflicts with the discussion, narrative emphasis may be distorted. Correct structure and wording, not the result.

Check the title and headings too: definitive labels can overrule cautious paragraphs. Add the population, period or condition where it materially bounds the finding.

Create an outcome inventory: Before drafting, list every pre-specified outcome and material analysis as completed, not analysable, changed or added exploratorily. Link each status to its place in the report or the reason analysis was impossible. Compare tables, abstract and conclusion with the inventory so findings do not vanish merely because they are awkward to narrate. The inventory belongs to the audit trail and can be more compact than a public registry, but it should show when a decision occurred and whether the result was already known. A change after results is not automatically improper; it demands especially clear labelling and cautious interpretation.

Method boundary: This guide cannot replace statistical or disciplinary advice. The suitable analysis depends on design and question.

Sources

  1. American Statistical Association: Statement on Statistical Significance and P-Values.
  2. EQUATOR Network: Reporting Guidelines.
  3. ALLEA European Code of Conduct for Research Integrity.