Research students need to be taught how to handle experimental data and when to select and reject results, the American Association for the Advancement of Science in Philadelphia heard last week, writes Julia Hinde.
Without this, less experienced scientists may find themselves inadvertently crossing the line between good research and unacceptable practice - especially as the pressure to produce results and publish papers increases.
Diane Hoffman-Kim, a postdoctorate fellow in neuroscience at Harvard University and the Bunting Institute of Radcliffe, said that education about which experimental data to include in analysis does not occur efficiently by example alone. "If you can't work out what data is noise and what is not, you won't be able to publish anything. But we just don't learn by osmosis. Very often these ideas are not discussed - they are seen as intuitive," she said. "I would like to challenge the scientific community to set up courses for undergraduates and research students to make explicit such intuitive processes."
She added that it was not always easy for junior researchers to question supervisors if they did not understand or agree with data selection, or if they wanted further clarification as to why a decision was made. "People don't ask because they don't want to be perceived as ignorant. There is also a climate in which questions about these issues are taboo. If you question why a supervisor has disregarded a result, then maybe he is going to think you may suggest misconduct."
Massachusetts Institute of Technology scientist Stephanie Bird, who is leading the way in establishing such courses, said: "We believe the situation is very similar in the UK. You don't really know what conclusions to draw unless someone tells you what the policy is underlying the practice. What we are trying to do is get the scientific community to be open about what the practices are."