In the current discussion about fraud (see Correspondence, Nature 439, 782–784; 200610.1038/439782b), the important issue of what actually makes people cheat in the first place has not been addressed. Whether an individual cheats and lies is dependent on many factors, but instant personal advantage is the central one.
Unlike nanotechnology and genetics research, where lucrative patents can beckon, my field of Earth sciences has little to offer but basic research in public institutions. The pressure on the relatively few permanent jobs available to a growing number of scientists in this field is therefore high, and hence the rewards of publishing in the highest-ranked journals are extremely significant for one's career. The temptation to fumble with the data a bit for the sake of the story, or to include a big-shot author as a showcase, is understandable.
In the most notorious recent cases (the Korean stem-cell work and Jan Hendrik Schön's nanotechnology work), the peer-review system must be said to have failed, as the frauds were unveiled by people from outside the immediate process. Were the referees the weakest link, and were both the editors and the referees blinded by the aura of the authorships?
Despite some disadvantages, anonymous peer-review remains the fairest way to prevent publication bias (see “Three cheers for peers” Nature 439, 118; 2006). But who ensures the quality of the referees and the refereeing process? Clearly, it is the editor's responsibility to overlook lobbying by authors and referees alike and to improve procedures if necessary to ensure fair assessment of all submissions, not favouring those from ‘star’ authors.
I believe it is best for the publisher not to reveal authors' names or affiliations to the referee until the submitted manuscript has been accepted or finally rejected. Withholding the authors' identities in this way will not only ensure high quality, but will stop rejected authors from believing that their submission had been treated unfairly compared with those of ‘star’ authors.