Wilcox, Allen J.
There is a widespread notion that scientific journals do not welcome null findings. This preference for the positive is thought to contribute to a publication bias toward “false-positive” results. There may be some truth to this. But there is also a bigger picture here: not all null results are created equal.
Some findings are not worth finding in the first place. This can be especially true for negative results. For example, investigators may be tempted to launch studies in search of a specific exciting result (“high-risk, high-impact”). Such studies hold the investigator hostage to fate—especially if the hypothesis is highly speculative. When speculative studies come up null (as they so often do), the authors face an array of bad alternatives: shelve the whole project, try to persuade a journal to publish an inconsequential finding, or (the worst) manufacture a “positive” result out of meaningless fragments. Authors can avoid such problems by asking themselves before the study whether the results will be publishable regardless of their direction. If not, the study may not be worth doing.
Other null findings may remain unpublished because authors are disconcerted by the message. Results showing benefits of smoking, for example, or harms of “good” dietary practices may be slow to reach publication because authors think the complexity of the real world will undermine public health messages.
Editors cannot control authors’ decisions about what to submit. However, at Epidemiology, the direction of findings is not our concern. Our chief criterion in selecting articles for publication is the contribution that a manuscript makes to the literature. We are unlikely to give priority to positive results (however well-done the study might be) that confirm well-established associations. Conversely, we welcome null results that challenge conventional wisdom or prior expectations. It is notoriously hard to disprove any hypothesis, of course, and so negative studies must have the precision and strength of design to be reasonably persuasive. However, when these strengths are present, a null result can be as scientifically valuable as any other. “Negative” papers in Epidemiology are easy to find.1–5
In response to growing concerns about publication bias, some journals are making special provisions to attract negative papers—even to the point of providing incentives.6 We trust that our prospective authors do not need such fanfare, but perhaps it’s worth going on record: identifying blind alleys can be as important as unlocking new doors.
1. Stein CR, Savitz DA, Bellinger DC. Perfluorooctanoate and neuropsychological outcomes in children. Epidemiology. 2013; 24:590–599
2. Cea-Soriano L, Johansson S, García Rodríguez LA. Risk factors for falls with use of acid-suppressive drugs. Epidemiology. 2013; 24:600–607
3. Elliott P, Shaddick G, Douglass M, de Hoogh K, Briggs DJ, Toledano MB. Adult cancers near high-voltage overhead power lines. Epidemiology. 2013; 24:184–190
4. Collins JJ, Bodner KM, Bus JS. Cancer mortality of workers exposed to styrene in the U.S. Reinforced plastics and composite industry. Epidemiology. 2013; 24:195–203
5. Stokholm ZA, Bonde JP, Christensen KL, Hansen AM, Kolstad HA. Occupational noise exposure and the risk of hypertension. Epidemiology. 2013; 24:135–142