Critics fault flaws in some scientific research studies
SCIENTISTS generally agree that fraud in research is intolerable. But what of other faults that weaken reports of scientific work - careless writing, inaccurate data, or the common practice of ``honorary'' authorship in which coauthors of research reports may have had little to do with the work? Dr. Walter Stewart and Ned Feder of the US National Institutes of Health think this problem may be a serious threat to the integrity of scientific literature. They urge an in-depth study to see if their concern is justified. Meanwhile, a preliminary study they have done has recently been published in Nature after a three-year delay caused by threats of libel suits by some coauthors of some of the papers cited.
A number of journals, including Nature, had earlier refused to publish the study. Now, as an accompanying editorial in Nature explains, the paper has been edited so that it no longer crosses the bound between allowable criticism and libelous accusations.
Stewart and Feder analyzed papers associated with a celebrated fraud that came to light in May 1981. Medical research scientist John C. Darsee had fabricated many of the data used in his reported work while at Emory University and the Harvard Medical School. There are 109 papers by Darsee with 47 coauthors in the Stewart and Feder study. The two scientists did not look at the fraud itself, which was already thoroughly investigated. They wondered, instead, if there were other signs of carelessness in those papers, whose coauthors were not alert enough to detect their colleagues' fraud.
They explain in their Nature paper that they found what they consider ample evidence of a general laxness. There were a variety of careless errors, such as discrepancies between captions of illustrations and the body of the text. Data reported in previous studies were reused without warning readers that they were not new data. And there was, in Stewart and Feder's judgment, considerable ``honorary authorship'' - a term they coined for scientists who co-sign papers reporting work with which they have had little actual contact.
Nature's editorial points out that the extent to which this kind of study reflects a general problem is debatable. Many of the errors cited seem trivial - a point that Eugene Braunwald of the Harvard Medical School, a coauthor of some of the Darsee papers, makes in an accompanying critique. Nevertheless, one is left with the feeling that there is indeed a general problem. As Nature notes: ``At the very least, [Stewart and Feder] have shown that a group of published papers do not provide a full explanation of the data reported in them.''
Nature adds that, to judge from general experience, the problem of honorary authorship is widespread. This is hardly a new complaint. It has surfaced whenever cases of fraud have been investigated. Nor is it novel to note, as Nature does, that the source of such corruption lies in the ``publish or perish'' environment within which scientists pursue their careers. Nature suggests ``decoupling the literature from promotion prospects.'' That may seem a reasonable goal. But since the productivity of scientists is equated with publication, achieving it would require a cultural change throughout the world scientific community. Perhaps journals, such as Nature, should insist on accountablity of coauthors as a first step.
There is, however, a more fundamental problem. Arthur Neufeld, director of research at the Eye Research Institute in Boston, Mass., points out that the nature of research itself has changed in a way that threatens scientific integrity. One of the main strengths of scientific research has been the ability of workers to reproduce each other's results. This has served as a check on error (and fraud) and has been a principal means of advancing knowledge. Writing in New Scientist, Neufeld points out that it often is no longer practical to try to reproduce another's experiments.
Expensive, specialized equipment and long, specialized training may be involved, so that only the original research team has the means of carrying out the experiment. Also, with so much to be explored in most fields today, scientists may feel they have no time to test another's work.
The result, Neufeld says, is that scientists tend to leave details of their methods out of research papers. The assumption is that their work won't be independently checked. More and more, Neufeld notes, scientists are having to take each other's reported results on faith, even when those results are crucial to a scientist's own work. And if the pressure to publish leads to sloppily written and inaccurate papers, that faith can be badly misplaced.
A Tuesday column. Robert C. Cowen is the Monitor's natural science editor.