03 · What You Need to Know
Selection Is Unavoidable; Misrepresentation Is Not
No Research Paper Can Display Everything
A microscopy study may produce thousands of images. An interview project may generate hundreds of pages of transcripts. A case study may accumulate years of records. A quantitative experiment may contain far more observations than can appear individually in a paper.
Researchers therefore select.
They choose representative images, illustrative quotations, exemplar cases, excerpts, tables, figures, and particular observations to discuss. Selection is not a methodological defect by itself. It is part of scholarly communication.
The integrity question is what principle governs that selection.
An Illustration and Evidence for Prevalence Are Different Things
An example can be used simply to show what a phenomenon looks like. It can also be presented as evidence that the phenomenon is common, typical, consistent, or dominant.
Those are different claims.
If one participant gives an unusually vivid quotation, researchers may use it to illustrate a theme if the methodology supports that use. But the quotation should not create the impression that all or most participants expressed the same view when the broader dataset shows substantial disagreement.
Illustrative selection
A genuine example is chosen to make a documented phenomenon concrete, without implying that the example alone establishes its frequency or representativeness.
Misleading selective presentation
Examples are chosen or contradictory evidence omitted so that readers receive a materially distorted impression of what the overall research showed.
“Representative” Is a Scientific Claim
Researchers often label microscopy images, histological fields, cases, quotations, or examples as “representative.” That word should not be treated as decorative figure language.
A representative example is being presented as reasonably characteristic of some relevant body of evidence.
If a treatment produces the expected cellular pattern in 5 of 100 fields while the remaining 95 look substantially different, choosing one of those five and labeling it “representative” would require a very different justification from selecting an image near the center of the observed pattern.
Community-developed microscopy guidance specifically addresses the need to select representative images appropriately and to communicate relevant variation rather than relying on an unexamined best-looking field.
The Most Dramatic Image Is Not Necessarily the Most Representative Image
Researchers naturally gravitate toward clear images. A figure should be interpretable, and technical quality matters.
But “clearest” can quietly become “strongest effect.”
If researchers repeatedly select the field with the largest difference between groups, readers may receive a visual impression much stronger than the quantitative analysis supports.
Image selection should therefore be connected to the underlying experimental design and analysis rather than conducted as an informal beauty contest after the results are known.
Selective Omission Can Matter Even When Nothing Is Digitally Edited
A figure may pass every pixel-level image-integrity check and still mislead.
Suppose 20 images were collected under the same condition. Four contain the predicted feature and 16 do not. Showing only the four positive images without explaining the broader distribution can distort the apparent consistency of the finding.
This is why research image integrity extends beyond Photoshop operations. What researchers choose not to show can matter as much as how the displayed pixels were processed.
Qualitative Quotations Must Remain Faithful to Their Context
Qualitative researchers often use participant quotations to support interpretations and allow readers some access to the underlying material. Quotations are necessarily selective because a manuscript cannot reproduce every transcript.
A genuine quotation can nevertheless be misleading if its surrounding context reverses or substantially qualifies its apparent meaning.
For example, quoting “the platform made everything easier” would be misleading if the participant's full statement was essentially that it made one minor task easier while making the overall work considerably more difficult.
Ellipses, excerpting, and editing for readability should therefore preserve meaning rather than manufacture a cleaner statement.
Do Not Choose Quotes Solely Because They Sound Perfect
A particularly eloquent participant may provide the quotation every qualitative researcher secretly wishes all participants had delivered in complete publication-ready sentences.
That does not make the quotation unusable. It does mean the researcher should ask whether the quote genuinely supports the analysis rather than merely expressing the researcher's interpretation unusually well.
Quotes should illuminate the evidence, not function as dialogue written by the dataset on behalf of the author.
Contradictory and Negative Cases Can Be Analytically Important
Some research methods explicitly attend to deviant, negative, or contradictory cases. Even where a formal negative-case procedure is not used, evidence that challenges an emerging interpretation may be scientifically important.
Researchers do not necessarily need to give every minority observation equal space. But systematically suppressing contradictory material can make an interpretation appear more uniform or certain than the evidence supports.
The appropriate treatment depends on the methodology. A qualitative study, case series, ethnography, survey, experiment, and diagnostic study will not use the same representativeness criteria.
Case Selection Should Match the Claim Being Made
Case-based research often involves deliberate rather than statistically representative sampling. Researchers may select an extreme case, critical case, typical case, information-rich case, or theoretically important case.
Such selection can be entirely legitimate if the case is represented according to the logic under which it was chosen.
An extreme case should not quietly become evidence of what usually happens. A purposively selected success story should not be presented as though it estimates the typical effect of an intervention.
The integrity issue lies partly in matching the inferential claim to the sampling logic.
Cherry-Picking Numerical Results Raises the Same Basic Problem
The principle extends beyond images and qualitative material.
If a study measures ten outcomes and researchers discuss only the two favorable ones while creating the impression that the intervention consistently succeeded, selective reporting may distort the overall evidence.
Not every unreported analysis or observation is falsification. Researchers make legitimate editorial and analytical choices. But the Federal Research Misconduct Policy explicitly recognizes that omission of data can constitute falsification when it misleads readers about the research results.
Selection Criteria Chosen Before Seeing Results Can Reduce Bias
When feasible, researchers can establish rules for selecting representative images, cases, time points, fields, quotations, or examples before knowing which ones best support the desired conclusion.
For microscopy, for example, sampling fields systematically or linking image selection to quantitative analysis can reduce discretionary selection of unusually favorable fields.
For qualitative work, a transparent account of how excerpts were chosen and how contradictory material was handled can make the interpretive process easier to evaluate.
Not every selection criterion can or should be prespecified. Exploratory and interpretive research often develops iteratively. The broader objective is to prevent the desired conclusion from becoming the hidden selection rule.
Quantification Can Help Put Representative Images in Context
A representative image is often most informative when accompanied by systematic quantification across the relevant sample rather than asked to carry the inferential burden alone.
If a figure shows one microscopy field per condition, quantitative results across biological replicates can help readers understand whether the displayed field reflects the broader evidence.
This does not mean every visual observation must be quantified. It means the strength of the claim should match the evidence used to support it.
A Genuine Figure Can Still Misrepresent the Underlying Data
Selective presentation is closely related to a broader problem: factual accuracy at the component level does not guarantee accuracy at the level of interpretation.
A graph can contain correct numbers while using a misleading scale. A photograph can be genuine while being an extreme case. A quotation can be verbatim while being stripped of decisive context.
The related question of whether a technically accurate figure can still misrepresent the underlying data therefore goes beyond selective examples and concerns the entire relationship between representation and evidence.
When Can Selective Presentation Become Falsification?
Under the PHS definition, falsification includes changing or omitting data or results such that the research is not accurately represented in the research record.
The Federal Research Misconduct Policy clarifies that accepted practices may appropriately omit data, but omission is considered falsification when it misleads readers about the results of the research.
This does not mean every editorial omission or imperfectly chosen example constitutes misconduct. A formal finding requires a significant departure from accepted practices of the relevant research community, intentional, knowing, or reckless conduct, and proof by a preponderance of the evidence.
The applicable research method matters greatly. Selection should therefore be evaluated against the practices of the relevant research community rather than one universal rule about how many images, cases, or quotations must appear.
Watch Out
Ask what a reader would conclude from the selected evidence without seeing everything you omitted. If that impression differs materially from what the full body of evidence supports, the problem may be selection rather than the authenticity of the individual examples.
