Tuesday, August 8, 2006

Corruption and Fraud in Science

Water, Air & Soil Pollution (2006)
DOI 10.1007/s11270-006-9209-8

J. T. Trevors & M. H. Saier, Jr.

Science is conducted by people, not all of whom are honest and credible, and some of whom unfortunately do not place the interests of humanity and our common biosphere ahead of their own selfish agendas.Honest mistakes are sometimes made because of our human foibles, but in this editorial we address the problem of deliberate corruption.The importance of this issue is emphasized by the fact that intentional errors and over-generalizations arriving at misleading conclusions for the purpose of justifying unwarranted actions can be extremely destructive and cause the less perceptive reader to be confused, when confusion may not be warranted.

Corruption in science can manifest itself in many forms.For example, it may take the form of plagiarism,falsified academic credentials,or the use of deliberately misleading statistical approaches.Even more serious are acts involving deliberate manipulation of data, fabrication of data sets or theft followed by publication of the stolen data.Also worthy of concern are more subtle acts such as peer reviews that are not objective, providing a competitor with a time advantage for publication.For this last reason, honest, credible reviewers must be used in the review process.This means that chief editors and senior editors must be knowledgeable about the scientific credentials and ethical values of potential reviewers.

Editors, editorial boards and manuscript reviewers have an arduous task when reviewing submitted articles.It is a significant challenge to determine if any of the above-mentioned activities have occurred with regards to the article.For example, parts of plagiarized articles from journals and books are difficult to recognize as reviewers cannot possibly have read all relevant research articles.The infraction goes unnoticed until the article is in circulation, and even then it may escape being identified as corrupt.While plagiarism is certainly not as serious a crime as deliberately misleading an audience,it is relatively straightforward to acknowledge the source of information, and therefore plagiarism should and can easily be avoided by conscientious authors.

Fabricated and stolen data sets that are slightly altered are difficult to identify.Raw data sets are not submitted to journals. Inaccurate data can be made to appear statistically significant by selectively altering or removing data.Evidence has been presented, for example, that Gregor Mendell must have selected his data to get results reflecting the statistically improbable high degree of accuracy reported. Fortunately in this instance, the conclusions were correct, but this is by no means always the case. Once again it is the difficult task of the reviewer to determine or guess if data selection has occurred from an examination of the submitted data sets.

The teaching of integrity in science should be part of academic scientific training programs. Training in integrity, just as training in the scientific method, must be an enforced policy of the institutes where research is conducted.Ethical practice also has to be learned by example; consequently, proper ethics must be taught by the supervising scientists.This becomes a difficult task when it is recognized that some supervising scientists are themselves not fully credible.Greed and the need for funding, recognition and promotion all provide unfortunate incentives for capable scientists to follow a path that does not maximally benefit humanity and the Earth_s biosphere.

The peer review process is the standard that is used worldwide for publication of scientific data and concepts. While not perfect, most scientists agree that this is the best method to guarantee publication of quality science. The question is how to identify problems of ethical misconduct in submitted articles at the peer review stage. Some suggestions are:

1. Articles that are poorly prepared, with misspellings and unacceptable grammar, but with apparently excellent data sets and conclusions, may trigger an alert as to the origin of the data. Articles with honestly submitted data sets are generally well written at the time of submission.Exceptions might be non-native English users who have not been able to obtain assistance in text preparation.

2. Articles that do not contain explanations of how experiments were replicated, that is, do not provide a sufficiently detailed description of the methods, allowing another scientist to repeat the observations, should not be accepted.Experimental methods without proper positive and negative controls are also suspect.The same can be true when detection limits are not reported.Sometimes articles contain representative data or representative images. This should be an alert to reviewers as to what the other images and data that are described but not shown may actually be.

3. Improper use of statistics or inadequate descriptions of statistical methods are problems that should lead to rejection if not corrected following the editorial review. The reviewer should be trained to recognize these problems.

4. Articles that contribute no new knowledge but are simply reaffirming what was previously published are often not acceptable because they represent minimal contributions to our knowledge.Such papers can also be generated by falsification of data sets with relatively little risk of being caught, unless, of course, the originally published data and conclusions were wrong.

5. Articles where no testable hypothesis was proposed in advance of data generation can be problematic.However, it should be noted that large data sets of information generated over extended periods of time with no obvious purpose in mind are sometimes useful and advance our knowledge of a field, subsequently allowing postulation of a coherent model.

6. Articles that are of only local interest, or that have used outdated methods, may be limited as to their scientific value.Conversely, if a particular area contains, for example, endangered species, unique medicinal plants or has not been the subject of previous research, articles of only local relevance can bring useful information to light.

7. Articles that don_t contain up-to-date references or don't cite relevant articles in the same discipline are sometimes suspect.They indicate a lack of familiarity of the authors with the background information. Reviewers must also be on the alert for copyright violations.This may also be a difficult task since no one is fully aware of the extensive scientific literature.

8. Articles where the conclusions are not supported by the data sets presented should not be acceptable. It is easy for exuberant authors to over-interpret their results.But such conclusions can also indicate carelessness, or even an attempt to prove a case without the requisite data.

9. Articles that report values that differ significantly from the existing literature should be scrutinized with care to ensure that errors have not been made. On the other hand, it is precisely these papers, when correct, that can lead to major advances in science.

10. It is sometimes necessary to ask the submitting authors for in press and previously published articles to ensure there is little or no overlap between articles.When these are obtained, the referees must read them carefully to ensure that there are no internal inconsistencies or suspect results.

Even when all of the above-mentioned precautions are taken, articles that should not be published may elude the review system. Fortunately, the scientific process will correct such errors with time.It is impossible, perhaps not even advisable, to attempt to create a repressive error-free system of scientific inquiry. Honest mistakes have at times led to major advances. And every credible scientist should have the opportunity to make his or her opinions heard,even when they go against the consensus opinion.