web analytics

It Is Difficult to Have No Bias in Scientific Research

oVW9uLoLZbCUpGEet75QBgXPG1MzbAPJWLTUWgqh5fBKAQAA6wAAAEpQ_260x196The path to seek truth is not smooth. Bias in scientific results is always a great obstacle to impede scientific development. In a recent research published on PNAS, English scientists found that exaggeration in scientific research is more likely to happen in behavioristics research, especially for American researchers.

Daniele Fanelli, an evolutionary biologist from University of Edinburgh, and co-workers collected 82 meta-analysis results in genetics and pathergasiology fields. “The reason to choose these two subjects is because they used to apply meta-analysis approach, which helps us obtain data easily.” Fanelli said that they extracted 1174 copies of original data from the meta-analysis and found that pure behavioristics research is more likely to report extreme effects; American research, especially those that don’t involve biological parameters, is highly probable to be bias to experimental hypothesis.


Above figure shows the ratio between effect size and summary effect size from each meta-analysis (in log scale). The size of circles depend on the standard deviation of research data. “0” at Y-axis means the effect size corresponds completely with the summary effect size from the meta-analysis. Researches are divided into purely biological/non-behavioral, purely behavioral and bio-behavioral. US: United States; AS: Asian countries; EU15: 15 European countries; O: other countries.*

*Image source:  Daniele Fanelli and John P. A. Ioannidis. (2013). US studies may overestimate effect sizes in softer research. PNAS.

“Probably the cheating and bias in scientific research become more in recent years, however to some extent, these behaviors always perplex science.” Fanelli said that the phenomenon might come from the relatively low consistency in methodology of behavioristics research. In addition, the thought of “publish-or-perish” is popular in America, which results in more positive results published by U.S. researchers.

In China, there also exists similar evaluation system that is closely related with publication amount and impact of articles. “Some researchers from Chinese universities can even gain direct reward if they get published on certain journals. Some scholars and decision makers believe that such reward system might lead to more severe problems compared with western countries.”Fanelli said, “If we really want to do good scientific research, the key is to make sure that scientists are rewarded because of their research but not directly due to the results. In China, some universities’ policy is to offer money reward for those who publish on Nature and Science, and clearly this is not going to help promote science development.”

“I believe the whole scientific system should make some changes.” Fanelli supplemented:” For example, journal side can process peer review and accepting papers based on the introduction and experimental parts of articles, without even know the accomplished results. The same fashion can also be employed by funding agency.”

Fanelli also suggested readers: “Don’t take it as truth for any new results. Only those have been screened and revised thesis should be trusted.” He said, “Scientists are also human, and all human have bias and might not be honest some time. This is an unavoidable fact in any fields of scientific researches, especially for areas that are immature in theory and methodology.”

Fanelli hopes their research method could be applied into other fields and he said:” Only when scientists can repeat and criticize each other’s research, science can advance. We hope all other scientists could do these things using our findings.”

 Image sourceShutterstock