Stealing into Print.

One of the few pleasures that followed the sustained gangbang of the American economy in the eighties by a club of smug Wall Street stock boys was seeing some of those white-collar rapists end up in prison. Oh, sweet justice, that even powerful and once-romanticized high financiers could be made to pay for their crimes and at last be seen as the oily swindlers they are.

Quite another reaction follows the uncloaking of a fraudulent scientist. Here, there can be no glee, no triumph, no sense of gotcha! Instead, the idea of researchers breaching our fundamental trust is frankly revolting. Scientists are supposed to be the closest thing to secular divinity, our wise divulgers of nature's mysteries. They have no other purpose but to tell the truth. So when they are caught marking the backs of mice with black ink and calling the splotches skin grafts or smoothing out a doubtful graph with a few fabricated data points, what else can one feel but profound disgust? And the fact that ever more cases of scientific fraud are coming to light only heightens the impression that nothing is sacred, everybody is corrupt, and that hell isn't large enough to hold all the sinners.

Given the emotions that surround the issue of fraud in science, Stealing into Print* is surprisingly cool, cerebral, and, on occasion, dry. It is nevertheless an important work, for it explores in scholarly detail how new discoveries are disseminated, how the exalted peer review system really works, and how difficult it is to stop fraud from permanently sullying the historical record once it has been published. Marcel LaFollette, a professor of science and technology policy at Washington University in St. Louis, is not interested in merely recounting the most harrowing fraud stories, nor does she try to address every aspect of scientific chicanery. For example, she makes no attempt to explain who is likely to cheat or why.

Instead, LaFollette takes an astringent look at one piece of the intricate business of fraud: how fraudulent papers get published and what can be done to remedy, or at least lessen, the incidence of printed fraud. By considering the process of scientific publication, she opens up wider questions of responsibility and accountability. In particular, she questions whether scientists are capable of policing theft own ranks. "What does the discovery of unethical conduct say about current systems for evaluating and disseminating research-based knowledge?" she asks. "And how may those systems be changed by efforts to detect, investigate, mitigate, and prevent wrongdoing ?" We may be unable to count on scientists to keep theft house clean, she claims, because most of them refuse to see the dirt for what it is. "[The] discovery of unethical or illegal conduct contradicts scientists' selfimages, theft beliefs about how 'real' scientists act," she writes. "As a consequence, the community often engages in wholesale denial of the problem's significance." Indeed, until quite recently, scientists have treated with hostility anybody who even mentioned the word fraud, viewing the issue as a creature of media hype blown out of proportion to titillate a publie that otherwise finds science dull--which sells newspapers and feeds a growing strain of anti'intellectualism.

The author finds fault with nearly everybody in and around the scientific arena. She is critical of scientists who lobby to get their names on a paper but who cry ignorance or innocence when one of their coauthors tums out to be a fraud. She lambastes the editors of scientific journals who sit passively on the sidelines while others fight over whether a paper that appeared in their publication is genuine or a hoax. She complains that most science reporters, far from being the "hyperactive pit bulls" most scientists be- lieve, are in fact timid invertebrates who rarely pursue evidence of potential infractions. And while she generally applauds whistleblowers for having the courage to defy their colleagues and expose malfeasance, she worries that, in their zeal to stamp out fraud, the watchdogs may trample the rights of the unfairly accused. The way to combat fraud, she says, is to formalize our handling of it, admit with some detachment that a certain percentage of scientists is likely to cheat, and design a systematic method for investigating claims and punishing proven perpetrators.

Cell block

LaFollette, alas, has not managed to discover what so many of us would like to know: exactly how many scientists commit fraud. But she doubts that dishonesty is as rare as most scientists believe; Daniel E. Koshland Jr., editor of the journal Science, recently declared that "scientists are ethically 99.9999 percent pure," although he offered no evidence to support his sweeping contention. LaFollette does not, however, unequivocally side with those cynics who believe that fraud is epidemic. Whether it is about to become so is another matter. She proposes that there is a greater commercial incentive for researchers to cheat today than 25 years ago. "Scientific knowledge is a major new commodity," she writes. She also suggests that the widespread use of technology such as computer networks, which can spread information swiffly and without any cross check, makes cheating easier than ever.

One of the book's most compelling chapters derails the different types of possible research and publication fraud, a sort of guided tour of skulduggery. Scientists may out-and-out fabricate an experiment, as William Summerlin did at Memorial Sloan-Kettering Hospital in New York when he decorated those mice with a felt-tip pen to make it seem that their skin grafts had taken. Or scientists may refer in print to data that don't exist--what Thereza Imanishi-Kari is said to have done in the widely publicized case of fraud that resulted in the resignation last year of her famous collaborator, David Baltimore, from his post as president of Rockefeller University. Scientists may seriously misrepresent authentic data, manipulating and massaging the resuits until everything fits perfectly into a preordained hypothesis. They may steal or plagiarize others' ideas, materials, and writings; put somebody's name on a paper without that person having contributed to the work; or claim credit for discovering something that others have found before.

Nor is deceit limited to the authors of a paper. Those who handle a manuscript after its submission to a journal also have the opportunity to breach the trust others have placed in them. LaFollette goes into some depth in describing the peer review system that is supposed to help keep research clean. Upon receiving a scientific manuscript, a journal editor normally sends it out to two or three researchers who work in a similar field for their comments and criticisms. On occasion, a reviewer may be a direct competitor of the paper's authors, and there have been cases in which reviewers appear to have stolen data from a submitted manuscript to expedite their own experiments.

By and large, however, LaFollette praises reviewers for theft diligence and commitment. Sometimes they may spot cases of apparent fraud simply in the normal course of checking over the submitted figures to see if they all make sense. I know of one case in which a Columbia University researcher who was reviewing a paper that had been submitted to the journal, Cell, calculated that if the author had done what he claimed to have done, he would have required as many laboratory petri dishes for one brief experiment as an entire scientific team would use in 10 years. The paper in question was never published, and the accused scientist was thrown out of his lab.

But LaFollette doubts that peer review on its own can do much to uncover fraud. The best way to detect chicanery is to repeat the experiment, which no reviewer has the luxury or inclination to do. Furthermore, scientists' definitions of what it means to review a manuscript vary. Some may scan for gross violations of logic, while others may recalculate some or all data tables.

Another enormous problem in rooting out fakery is distinguishing between fraud and an honest mis-' take. In rare instances, evidence of wrongdoing may be unequivocal. That happened recently at the National Institutes of Health, when Mitchell Rosner, a young graduate student, reported in Cell that he had done critical controls for experiments designed to explore the molecular details of embryonic development. But when others subsequently questioned the experiments and performed several checks on Rosner's work, proof that he had faked the controls was so overwhelming that Rosner had no choice but to sign a confession. The paper was speedily retracted in the bluntest possible language.

More often, the question of fraud is far murkier, and accusations and counteraccusations may fly for years. The most publicized example of this frustrating ambiguity is the aforementioned case of Thereza Imanishi-Kari and her disputed paper on the subject of immunology that also appeared in Cell. When her student, Margot O'Toole, had difficulty reproducing the results, she took her qualms to authorities at the Massachusetts Institute of Technology, where they both worked. Although David Baltimore, the senior author on the paper, dismissed Dr. O'Toole as a "disgruntled post-doc," rumors of a possible misdeed would not disappear. Eventually, the case was investigated by a congressional committee, before which Imanishi-Kari claimed that she had done everything that had been published, while O'Toole insisted that Imanishi-Kari's laboratory notebooks did not jibe with the published data. Baltimore thundered against political busybodying in the basilica of science and declared that Imanishi-Kari was merely a bit slapdash in her bookkeeping. Eventually, the Secret Service got involved. By forensically analyzing the paper Imanishi-Kari had used for counting radiation signals, the agents concluded fraud had occurred. Devastated and seemingly humbled, Baltimore retracted the paper.

The story still isn't over. Imanishi-Kan continues to claim her innocence and as of late July had reportedly hired her own forensic team to reanalyze the Secret Service's analysis of her data. That follow-up research, she said, trashed the initial forensic work. Baltimore, ever reluctant to admit to any error in judgment, told The New York Times that he was planning to retract his retraction. He argued that in light of the new forensic examination, the entire paper must be considered valid once again. Many of his colleagues fiercely disagreed, saying that nobody has been able to reproduce the results of the Cell paper and that Baltimore would be worse than foolish to cancel the retraction before repeating the entire experiment. As it looks now, Imanishi-Kari's guilt or innocence may never be known.

Salient solutions

Given the difficulty of proving fraud, how can its incidence be reduced? LaFollette does not pretend to have a tidy formula, but she believes that a good place to start is in accreditation--deciding whose names go on papers submitted for publication. Too often, researchers are listed as authors when they have done nothing more than provide advice over the telephone. They know little about the most important steps of the experiment and therefore cannot be depended on to detect suspicious results. They want all the benefits of another item for their resumes, but none of the responsibility for assuring that the entire paper is accurate. LaFollette suggests that every name listed on a publication should be held accountable for every part of the work; only when such considerable obligations are met, she argues, will science truly be able to serve as its own sentry. If nothing else, that policy would cut down on the list of 10, 20, even 100 names that now clutter the top of the bigger scientific reports.

Stealing Into Print is by no means a completely satisfying book. For instance, LaFollette relies too heavily on examples taken from nonscientific disciplines like history and art to make her points. If scientists are going to argue that science is more pristine than nearly every other trade, the skeptic's best ammunition is a blizzard of specific evidence to the contrary. LaFollette also quotes extensively from novels about scientists, as though the way a fictional character in a Dorothy Sayers mystery responds to fraud is how it actually happens. This is a particularly puzzling approach, because LaFollette has taken such pains not to indulge us with the psychodramas and operatics of genuine scientific frauds.

Still, for anybody who cares about the meaning and texture of science and who wants to see it soar once again above the current miasma of publicized fraud, Stealing into Print helps illuminate the darkness.

Natalie Angier is a science reporter for The New York Times.

* Stealing into Print: Fraud, Plagiarism, and Misconduct in Scientific Publishing. Marcel C. LaFollette. University of California Press, $30.