As the popular saying goes, “Don’t believe everything you read on the internet.” A disturbing trend in science suggests you should add, “and don’t believe everything you read in scientific journals either.”
The gold standard in science is a carefully designed experiment that yields reproducible results. Scientific theories from Galileo, Isaac Newton, Albert Einstein and other brilliant scientists remained theoretical until their proofs could be replicated
Financial incentives and career promotion often push today’s scientists to produce eye-catching results for high-profile publications. Unfortunately, these incentives can spur corner-cutting and loss of scientific objectivity. This corner-cutting has huge costs. Government and businesses invest billions of dollars based on research findings. Sick and vulnerable people put their hopes on breakthroughs in medical science. When published science proves to be wrong, however, potentially irreparable damage is done to public trust.
For instance, a few years ago, drug-makers Bayer and Amgen reported dismal results from attempts to reproduce published cancer treatment research. Only 25% of the results reviewed by Bayer could be reproduced in an independent laboratory. Amgen could replicate only six out of 53 studies for which it commissioned analysis.
More recently, 270 volunteer scientists affiliated with the Center for Open Science tried to reproduce results of 98 original research papers published in three psychology journals. As reported subsequently in Nature magazine, only 39 of the replication attempts were successful.
A subsequent survey by Nature found that more than 70% of researchers have tried and failed to reproduce another scientist’s experiments, and more than half have been unable to reproduce their own experiments! More than one-half of those surveyed by Nature described all of this as a “significant crisis” for science.
A 2012 study aimed at linking genetics, political ideology and mental illness illustrates several serious problems with the lack of research reproducibility. Two respected scientists undertook to analyze a data set and determine if political orientation and a particular personality disorder were linked. Their conclusion, published in The American Journal of Political Science, was that politically conservative individuals are much more likely than political liberals to manifest a personality pattern typified by aggressiveness and interpersonal hostility.
When two other social scientists attempted to reproduce these results, they discovered a grade school mathematical error in the original analysis. Correcting the math mistake produced the opposite result: liberals, not conservatives, seemed much disposed toward aggressive behavior and interpersonal hostility.
Unfortunately, the two researchers who made the elementary math error resisted correcting or retracting their original findings for three years, during which their initial findings were cited as established science in dozens of other scholarly articles. Not until publication of an article which detailed their mistake was imminent did they correct the record.
Untrustworthy research results make Americans more skeptical about science. This should worry every scientist and advocate for science. Research that furthers our national interest and is of the highest intellectual merit is crucial to creating and maintaining public trust in our federally funded science. The American public should demand scientific integrity and a healthy, transparent and vibrant research and technology environment.
One step my committee has taken in achieving this goal is passing the American Innovation and Competitiveness Act, or AICA, the last bill from the 114th Congress signed into law. The AICA included several provisions to maximize the nation’s investment in research, including a first step toward addressing research reproducibility. The AICA requires the National Research Council (NRC) to study and recommend ways for Congress and the NSF to address failed research reproducibility.
Pending completion of the NSR report, here are steps NSF can take now:
- Set aside three percent of research funds for reproducibility studies.
- Enforce the requirement that researchers provide timely public access to all data generated by federally funded research.
- Take the lead for establishing an electronic database to enable public access to research data.
- Encourage scientific journals to publish credible reproducibility studies and establish consistent retraction policies.
Congress and the White House should support and encourage the work of scientific agencies like the US Office of Research Integrity in the Department of Health and Human Services and the NSF Inspector General. Transparency and public accountability are the best antidotes to sloppy science, research bias, misrepresentation and misuse of public funds. They are also essential to maintaining public trust in science.