For many years physicists lagged way behind biologists in the perpetration of scientific fraud. But they have caught up in spectacular style with the ambitious opus of Jan Henrik Schon of Bell Labs, who placed seven of his fictive works in Nature and nine in Science. All those ad hoc explanations for biomedicine's leadership role in fraud--that entrance to medical school selected for corner-cutters, or that the mathematical structure of physics leaves little slack for fudging figures--must be laid aside.
Fraud in science is a minor irritant from one perspective, a serious problem from another. Most instances of fraud concern work of little importance and are quickly forgotten. Some practitioners forsake the safety of mundane fabrication and concoct spectacular experiments about matters at the cutting edge of their fields. But one can argue that the more ambitious the fraud, the more quickly it will be discovered.
The Schon case does not strongly support this contention. His fraud remained undetected for two years, during which time he produced work of such seeming distinction that he was talked of for the Nobel Prize. He was detected because of an insider's tip, not by the usual checking mechanisms of science; refereeing and replication. Had he had the good sense to stop in time, his oeuvre might have gained him a professorship from which he could have directed the work of an army of honest PhDs and laid a firmer basis for his scientific standing. Perish the horrid thought that undiscovered Schons throng the halls of academe.
Science is a cumulative process, however, and in the long run each brick must bear the load of those placed above it. So there is much force to the argument that incorrect results of any kind--whether obtained by fraud, self-deception, or other regrettable human frailties--cannot last indefinitely.
But they can last a long time, breezing past the conventional checkpoints of scientific quality without the slightest difficulty. This is the sense in which fraud is a serious problem, both of methodology and of public relations. Scientists point to the refereeing system as a guarantor of quality, but in the next breath will assert that referees cannot be expected to detect fraud. In fact, a referee, who after all is just doing an unpaid paper review, cannot test for much more than plausibility. That's a useful function, but it's not very effective as a screen against fraud.
Replication is central to scientific methodology, but in practice it's almost never an exact duplication of the kind necessary to support an accusation of fraud. There are plenty of honest reasons why two researchers may get different results from the same experiment. A claim that cannot be replicated is generally ignored, not publicly repudiated. Like refereeing, replication plays a useful purpose in science, but it is not designed to detect fraud and rarely does so.
Many, perhaps most, cases of fraud come to light because someone in the perpetrator's laboratory, someone in a position to observe his behavior and see the raw data, gets uncomfortable enough to blow the whistle. The front line of defense against fraud is not methodological but personal. The lab chief is in the best position to detect fraud. Only he can demand to see the lab notebooks, evidence that is beyond the reach of outsiders.
Science, by this analysis, is institutionally vulnerable to fraud. Its quality control mechanisms do not prevent fraud, yet as each new case bursts into public view, scientists find themselves put in the generally false position of declaring that there is no need to worry, because the quality control mechanisms of science infallibly detect fraud.
A more direct answer would be that research is not a process that can be made efficient. There is an inevitable degree of waste in the system, and fraud is generally not a serious enough problem to justify any measure that would cost significant time or money. However, it has not proved to be a popular response to go before Congress or the news cameras and declare, "Fraud happens--forget about it."
There's a strong case for viewing the prevention of fraud as the direct responsibility of the lab chief. If the people he or she has hired are disturbed enough to cook data, the lab chief should get to know about it. If the lab chief puts his name on the concoction, intending to draw credit for it, he deserves a big share of the blowback. But at present every fraud case seems to end the same way. The perpetrator disappears from view, slinking off to become a pathologist in a Midwestern hospital. And the lab chief receives the commiseration of his pals for the unfortunate occurrence that fate visited on him.
Nicholas Wade is a science writer for The New York Times.