03 · What You Need to Know
The Goal Is Compression Without Changing the Evidential Pattern
Scientific Reporting Already Requires Simplification
A paper is not supposed to reproduce an entire dataset in prose.
ICMJE's current recommendations explicitly advise authors to present results in logical sequence, give the main or most important findings first, and emphasize or summarize important observations rather than repeating all the data already available in tables and figures. Extra material and technical details can be placed in appendices or supplementary materials when appropriate.
This is an important point: simplification is not an unfortunate compromise with scientific rigor. Good reporting requires selection and organization.
The integrity question concerns what the simplification preserves.
Technical Complexity and Evidential Complexity Are Different
Some complexity can be compressed with little loss of meaning.
Suppose six sensitivity analyses produce almost identical effect estimates. You probably do not need six paragraphs describing them separately. A concise statement that the conclusion was robust across the specified sensitivity analyses, supported by a table or supplement, may communicate the evidence more effectively.
Now suppose three analyses support the conclusion and three equally defensible analyses do not.
Summarizing that as “sensitivity analyses confirmed the result” would not merely reduce technical detail. It would remove evidential instability.
Technical complexity
Detail that can often be condensed without materially changing how the evidence should be interpreted.
Evidential complexity
Variation, uncertainty, contradiction, dependence, or exceptions that materially affect the strength or scope of the conclusion.
Good simplification removes the first where useful while preserving the second.
Look for the Pattern Across Results, Not Just the Most Attractive Result
Suppose an intervention is evaluated using five outcomes. Two show clear improvements, one suggests a smaller possible improvement, and two show little evidence of change.
You do not necessarily need to recite all five results every time you mention the study.
But “the intervention improved outcomes” may be too broad. “The intervention improved some outcomes, particularly X and Y, while evidence for the others was weaker” preserves the important pattern with relatively few words.
That is compression rather than concealment.
The wording becomes even more important when outcomes have different prespecified roles. A favorable secondary outcome should not be allowed to make an unfavorable primary outcome disappear merely because the secondary result produces a simpler sentence.
Mixed Findings Can Often Be Summarized as Mixed
Researchers sometimes treat “mixed” as a frustrating non-conclusion. It can be the correct conclusion.
A set of findings may suggest that an effect depends on outcome, context, measurement, population, time point, or analytical specification. Those patterns can be more informative than forcing the evidence into a binary “works” or “does not work” judgment.
ICMJE recommends placing findings in the context of the totality of relevant evidence, stating limitations, and avoiding conclusions that are not adequately supported by the data.
Sometimes the most faithful simplification is therefore something like:
“The results were mixed: the intervention improved engagement but did not provide clear evidence of improved achievement or retention.”
That sentence is considerably simpler than a full results table. It still preserves the main structure of the evidence.
Uncertainty Is Not Disposable Detail
Researchers often simplify estimates into labels: positive, negative, significant, nonsignificant, effective, ineffective.
Those labels can erase important uncertainty.
Imagine an estimated effect of 4 points with a narrow confidence interval from 3 to 5. Now compare it with an estimated effect of 4 points with an interval from -2 to 10. The point estimates are identical, but the evidence is not equally precise.
A simplified account should preserve uncertainty when it materially affects interpretation. Depending on the context, that may require reporting a confidence interval, describing the estimate as imprecise, or avoiding a categorical claim altogether.
Exceptions Matter When They Define the Boundary of the Claim
Suppose a program improves outcomes in four participating institutions but shows the opposite pattern in a fifth.
Can you summarize the overall result as positive?
Possibly, depending on the design, estimates, and inferential analysis. But the exception may matter if it suggests that effectiveness depends on institutional context.
The key question is whether the exception is incidental noise or evidence about the boundary conditions of the conclusion.
If omitting it allows you to make a claim that would otherwise require qualification, it probably belongs in the simplified account.
Tables and Figures Can Carry Complexity That Prose Does Not Need to Repeat
One of the easiest ways to avoid the false choice between exhaustive prose and oversimplification is to distribute information across formats.
The narrative can state the central pattern. A table can provide estimates for individual outcomes. A figure can show heterogeneity. Supplementary materials can document additional robustness analyses.
ICMJE specifically recommends using tables and figures where they help explain the argument and assess supporting data, while avoiding unnecessary duplication between text, tables, and figures.
This allows the prose to remain readable without making the evidence inaccessible.
Statistical Significance Is Not a Good Simplification Rule
A particularly risky shortcut is to summarize significant findings and omit nonsignificant ones.
That converts a statistical threshold into an editorial selection mechanism.
If ten related analyses are performed and two cross p <.05, reporting only those two does not simplify the complete pattern. It changes it.
This is one route by which ordinary writing decisions can become selective reporting.
Do Not Replace “Complicated” With “Positive”
A common form of oversimplification occurs when results are partly favorable and partly unfavorable, but the final narrative keeps only the favorable direction.
Empirical research on spin illustrates the risk. Boutron and colleagues examined randomized trials whose primary outcomes were statistically nonsignificant and documented reporting strategies that emphasized beneficial secondary, subgroup, or within-group results or distracted readers from the unfavorable primary result.
The lesson is broader than clinical trials. A simplified conclusion should emerge from the overall evidential pattern, not from whichever subset makes the strongest claim.
The Abstract Requires the Most Aggressive Simplification and Therefore the Most Care
An abstract may compress thousands of words into a few hundred. It cannot preserve every nuance.
Yet it should preserve the nuance that changes the conclusion.
If the primary finding is uncertain but a secondary result is favorable, an abstract can mention both concisely. If an important limitation restricts generalizability, a few words may be enough to signal it.
The principle is the same as when deciding which results deserve emphasis in an abstract: prioritize without giving readers a materially different study from the one reported in full.
Simplification Should Remain Consistent Across the Manuscript
Another problem arises when each section becomes progressively more confident.
The Results say, “The estimate was positive but imprecise.”
The Discussion says, “The findings suggest a beneficial effect.”
The Conclusion says, “The intervention was beneficial.”
The abstract says, “The intervention improves outcomes.”
No individual step may look dramatic, but uncertainty has gradually evaporated.
Watch Out
When shortening a result, check which words disappear. If the first casualties are always “may,” “uncertain,” “mixed,” “secondary,” “exploratory,” or “context-dependent,” the shorter version may be changing more than the word count.
04 · A Practical Example
Turning Eight Results Into One Accurate Paragraph
Hypothetical Example
A Multifaceted Student-Support Program
A university evaluates a student-support program using eight outcomes. Examination scores and retention show moderate improvements. Engagement improves slightly. Attendance and satisfaction show little change. Academic confidence improves only in one subgroup, while two wellbeing measures are highly uncertain.
Too Much Detail
A paragraph could report every estimate, confidence interval, test statistic, and p-value sequentially. That may be appropriate in a table, but it can obscure the larger pattern in prose.
Too Much Simplification
“The student-support program improved student outcomes.”
This is concise, but it implies broader consistency than the results show.
Useful Simplification
“The program showed its clearest benefits for examination performance and retention, with smaller evidence of improved engagement. Other outcomes showed little clear change, while the apparent improvement in academic confidence was limited to one subgroup.”
This version does not reproduce every statistic. It nevertheless preserves the important structure: stronger findings, weaker findings, null findings, and a subgroup-dependent result.
The detailed estimates can remain available in a table or supplementary material. The prose has simplified the presentation without inventing consistency.
