03 · What You Need to Know
Scientific Images Should Be Treated as Data, Not Decoration
An Image Can Be Part of the Research Evidence
A microscopy image, electrophoretic gel, Western blot, radiological image, photograph, astronomical image, histological section, or other research image may contain information from which researchers and readers draw scientific conclusions.
That makes image integrity conceptually similar to numerical data integrity. ORI explicitly notes that images are frequently data in science and that manipulation of an image may therefore raise a question of falsification depending on the particular circumstances.
The important issue is not whether Photoshop, ImageJ, Fiji, proprietary instrument software, or another application was used. Software does not determine integrity. What the processing does to the evidence does.
Image Processing and Image Falsification Are Not Synonyms
Digital image data often require processing to make them interpretable or suitable for publication. Community-developed microscopy guidance, for example, recognizes operations such as cropping, rotation, resizing, brightness and contrast adjustment, channel display, annotation, and preparation of figures.
A journal may permit some processing while restricting other operations. Standards also differ among imaging techniques and disciplines.
Legitimate image processing
Processing improves visualization, formatting, measurement, or communication while preserving the scientific meaning of the underlying image and following applicable standards.
Misleading image manipulation
Processing creates, removes, conceals, duplicates, relocates, selectively emphasizes, or otherwise alters information in a way that changes what the evidence appears to show.
Adding Something That Was Not There Is a Major Problem
One of the clearest concerns is digitally introducing a feature that the original image did not contain.
Examples include copying a band into a gel lane, cloning a cell into a microscopy field, duplicating a signal to make a pattern appear more frequent, or copying an object from another image and presenting it as part of the original scene.
Depending on what was created and how it was represented, such conduct may implicate fabrication, falsification, or both concepts. The formal classification depends on the facts and applicable misconduct standard.
Removing Something Can Be Equally Misleading
Researchers may be tempted to remove visual features they regard as irrelevant: an unexpected band, a background spot, a cell that looks unusual, a blemish, or a signal that complicates interpretation.
Digital cloning, healing, smudging, blurring, or local background replacement can make such removal almost invisible.
Yet a feature that looks aesthetically inconvenient may be scientifically meaningful. Removing it can create a false impression about the specimen or experiment.
Nature Portfolio image-integrity policies, for example, prohibit touch-up tools such as cloning and healing tools and caution against processing that causes data to disappear.
Selective Local Processing Is More Problematic Than Uniform Processing
A global adjustment changes the entire image according to the same operation. A local adjustment targets only a selected region.
This distinction matters because local processing can selectively strengthen one feature, suppress another, or make experimental and control regions look more different than they actually are.
Scientific imaging guidance has therefore generally treated simple adjustments applied uniformly across an image much more permissively than selective alteration of particular regions.
Even global processing can become misleading if it is excessive. The operation being applied to the whole image does not justify clipping weak signals, saturating strong signals, or otherwise destroying relevant information.
Duplicating an Image or Part of an Image Can Misrepresent Independent Evidence
Duplication does not always mean that the duplicated content was digitally altered. Sometimes the same image is reused and labeled as though it represented a different experiment, condition, sample, participant, or time point.
That can mislead readers about the amount or independence of the underlying evidence.
Within a single image, duplicated regions can create the appearance of repeated biological structures or experimental signals that did not exist in the original field.
The integrity problem therefore concerns representation, not merely whether individual pixels were changed.
Splicing Images Can Be Legitimate Only When the Assembly Is Represented Honestly
Researchers sometimes need to juxtapose material from different images. For example, nonadjacent lanes from a gel may be presented together to focus attention on relevant samples.
Whether this is permitted depends on the applicable journal and disciplinary standards. Where such assembly is allowed, boundaries generally need to be clear and the figure should not imply that nonadjacent material was originally contiguous.
Nature Portfolio guidance, for example, requires rearranged nonadjacent gel lanes to be clearly delineated and the rearrangement stated in the figure legend.
The more specific problem of combining images from different experiments into one figure deserves separate scrutiny because a technically neat composite can create a false impression about experimental context.
Cropping Can Change Meaning Without Changing a Single Remaining Pixel
Cropping removes context from an image. That is often harmless and useful. Researchers may crop empty borders, focus attention on a region of interest, or make a figure fit a publication layout.
But a crop can also conceal contradictory evidence, remove relevant controls, hide an unexpected band, or make a selected region appear representative of a much larger field.
Community-developed microscopy guidance allows cropping provided that it does not change the meaning conveyed by the image. The critical issue is therefore not whether cropping occurred but what information disappeared with the crop.
Brightness and Contrast Adjustments Can Either Clarify or Distort
Brightness and contrast adjustments are common because raw images are not always displayed optimally for human viewing.
Policies such as Nature Portfolio's generally permit such processing when it is applied appropriately across the entire image and consistently to controls. Adjustments should not make data disappear or selectively emphasize experimental evidence relative to controls.
The boundary around brightness, contrast, and cropping adjustments therefore depends on how those operations affect the underlying information and comparison.
Representative Images Must Actually Be Representative
A different form of manipulation can occur before any pixel is edited: selective image choice.
Suppose an experiment produces 100 microscopy fields with substantial variation. A researcher chooses the single field showing the strongest expected effect and presents it as “representative,” despite knowing that most fields look quite different.
The selected image may be completely unedited. The presentation can still mislead.
Modern microscopy guidance emphasizes representative image selection and recommends showing sufficient images to communicate relevant variation when needed. The related question of selectively showing images, cases, quotes, or examples therefore extends beyond digital editing.
Image Manipulation and a Finding of Research Misconduct Are Not the Same Thing
ORI cautions against conflating an image discrepancy with a finding of falsification or research misconduct. Detecting duplicated pixels, an unexplained splice, or another anomaly establishes a question that requires investigation.
Original image data are particularly important for determining what occurred.
Under the current PHS framework, a formal research misconduct finding requires more than an image discrepancy. It 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. Honest error and differences of opinion are excluded.
Watch Out
An image can look “better” after processing while becoming scientifically worse. Do not use aesthetic improvement as the criterion for editing research images. Ask whether every consequential operation preserves what the underlying data actually show.
