The People Who Look at Pictures for a Living

The Western blot is an unglamorous piece of scientific furniture: a dark rectangle of film or a digital scan showing bands of protein, arranged in lanes, compared across experimental conditions. It is also, according to a growing community of forensic reviewers, one of the most frequently fabricated objects in biomedical literature.

The manipulations are often surprisingly crude. The same band, with the same texture and background noise, appearing in two different lanes representing different samples. A gel that has been spliced together from multiple runs. A figure that has been rotated, flipped, or subtly stretched — detectable because the background speckle of the image betrays the duplication. Once you know what to look for, the evidence can be visible to anyone willing to look closely enough.

A small number of people have made looking their vocation.

An Informal Watchtower

There is no central agency charged with auditing the integrity of the roughly two million papers published each year. Instead, the task has fallen to a scattered coalition: retired researchers with time and pattern recognition, image forensics specialists, academic misconduct scholars, and volunteer collectives who comb the literature for anomalies.

Their findings are documented on forums, preprint comment servers, and institutional blogs — including the anonymous PubPeer database, where post-publication critiques accumulate in public view. Some sleuths have built reputations rivaling those of the journals themselves. Their work has triggered thousands of retractions and, in a growing number of cases, federal investigations into researchers whose laboratories were funded by the National Institutes of Health.

The scale of the problem they have uncovered is difficult to overstate. Analyses of the retraction literature suggest that a substantial fraction of retractions cite image manipulation or duplication rather than honest error. And because retractions represent only the misconduct that gets caught, the true prevalence of manipulated figures in the literature remains unknown — a gap that hangs over every downstream field that builds on published results.

Why Retraction Moves So Slowly

When a credible allegation lands on a journal's desk, the expected sequence is straightforward: the editors investigate, the authors respond, the paper is corrected or retracted. In practice, the process can take years — sometimes a decade or more.

The reasons are structural. Journal publishers often lack the staff and expertise to adjudicate complex forensic claims. Universities, which bear responsibility for investigating their own faculty, face conflicts of interest that are hard to overstate: the institution's reputation, its grant funding, and its patent portfolio may all be implicated by an adverse finding. Whistleblowers who report misconduct internally describe retaliation that makes external publication look safer by comparison.

Even when a retraction finally arrives, the damage lingers. Retracted papers continue to be cited for years after their withdrawal, their findings absorbed into reviews, meta-analyses, and the working assumptions of junior scientists who never learned the original result was compromised.

What It Means for Drug Development

For fields translating basic biology into therapies — gene therapy among them — the integrity of the preclinical record is not an abstraction. A gene therapy program that advances toward the clinic on the strength of animal studies containing fabricated data will discover the truth eventually, but at the cost of years and tens of millions of dollars. The clinical trial becomes, in effect, the first honest replication attempt.

The pattern is familiar from the broader reproducibility literature. When biotech and pharmaceutical companies attempt to replicate academic findings before licensing them, a large share fail to confirm — and image manipulation, where present, is among the most damning explanations.

Toward Systemic Solutions

The detective model, however effective, is a symptom rather than a cure. A sustainable fix looks different: journals deploying automated image-screening software at submission; funders requiring raw data deposits alongside publication; institutions with genuinely independent misconduct offices; and career incentives that reward replication and rigor rather than volume of high-impact publications alone.

Some progress is visible. Major publishers now routinely run manuscript images through forensic screening tools. The NIH has expanded training in responsible research conduct. Data-sharing mandates are slowly becoming the norm rather than the exception.

But the underlying truth the sleuths have exposed is uncomfortable: the scientific literature has been policed largely by volunteers, in their spare time, without institutional support. Until integrity checks are built into the publishing pipeline itself, the health of biomedical research — and the therapies built upon it — will depend on the continued vigilance of people who look at pictures for a living.