Scientific images are measurements presented through acquisition settings, processing and layout choices. A published panel can be informative while showing only part of the original record. Reading image integrity begins with understanding that record and the figure’s stated purpose, not with making an accusation from an unusual-looking feature.
Identify what the panel represents
Read whether an image is a representative field, a quantified measurement, a composite, a projection or an illustration. These roles differ. A representative panel does not reveal the full distribution of observations used in an analysis.
Nature Portfolio’s image guidance requires faithful representation of original data, disclosure of important processing and clear demarcation of juxtaposed images. It also explains why unprocessed files and metadata may be needed to assess a concern.Nature Portfolio — Image integrity and standards (opens in a new tab)
| Detail | Question |
|---|---|
| Panel identity | Which sample, condition and time are shown? |
| Scale | What size or intensity does the display represent? |
| Processing | What adjustments or projections were applied? |
| Source data | Can the published panel be linked to an original file? |
A missing scale bar or an unclear colour key may limit interpretation without implying that the underlying image is false. State the practical limitation precisely.
Distinguish transparent processing from a changed impression
Images often need processing for display, but the operation can affect what becomes visible. Contrast, cropping and colour mapping should be understood in relation to the comparison the figure invites.
In an original hypothetical figure, two microscopy panels use different intensity ranges, making both appear equally bright. Without the ranges or a common quantitative scale, their visual brightness cannot establish equal signal intensity.
Similarly, a crop may make a relevant region easier to see while removing contextual features. The reader needs enough information to know what was omitted and whether the displayed region represents the quantified data.
Describe the observable issue before interpreting it
Potentially repeated features, abrupt boundaries or inconsistent labels can justify closer inspection. Begin by identifying the panel, location and apparent relationship. Avoid claiming a cause that the visible figure cannot establish.
A shared control image might be legitimate if the same experiment supports several comparisons and the reuse is disclosed. An undisclosed reuse or a mismatch between the caption and source file raises a different question. The surrounding record determines the interpretation.
Nature’s guidance specifically discusses original gel and blot images, disclosed lane rearrangements and matching unprocessed scans to published figures. These checks connect the presentation to its source rather than relying only on the final layout.Nature Portfolio — Image integrity and standards (opens in a new tab)
An image similarity tool can flag a candidate issue, but its output is not a finding of misconduct. Resolution may require original files, acquisition metadata and an explanation from the authors or journal.
Assess the consequence for the reported result
Ask whether the issue affects a decorative illustration, a representative panel, the underlying quantitative measurement or a central comparison. Those cases can have different consequences for the paper’s claim.
If an original file resolves a labeling ambiguity, update the appraisal accordingly. If the necessary records are unavailable, state what remains uncheckable and how central that uncertainty is to the result.
Look for a formal correction or other publication notice before repeating an old concern as unresolved. Equally, a polished replacement image does not explain an underlying data problem unless the correction addresses it.
A useful evidence note identifies the observable discrepancy, the records inspected and the remaining interpretive limit. This keeps image scrutiny rigorous and fair, without treating every presentation flaw as fraud or every attractive figure as verified evidence.
Sources and further detail
- Nature Portfolio — Image integrity and standards (opens in a new tab)
Processing, original-data, juxtaposition and gel/blot guidance read at the journal policy page. No real image is accused of manipulation. The two-panel intensity example is original; this is a reading guide, not an image-forensics audit.
Sources checked 19 September 2026. Worked examples are illustrative unless a supplied report is explicitly identified. This article has not undergone independent scientific peer review.