What happens when a scientist investigates results that are "too beautiful to be true"?
Inside Tsuyoshi Miyakawa's attempts to improve reproducibility in science
Photo by David Clode on Unsplash
Some fields of science are currently experiencing a reproducibility crisis. This means that a large number of scientific studies, that have been peer reviewed and published, cannot be repeated with the same results. One of the causes for this is scientific misconduct. This may include p-hacking, the practice of adding in data until the results are significant, selective reporting, where only the desired results are shared, or outright falsification of data.
The Editor-In-Chief of the journal Molecular Brain, Tsuyoshi Miyakawa, recently wrote an editorial in which he examined the manuscripts he had handled over the last two years. As an editor, his job is to judge whether a manuscript is good enough to send to reviewers. In some cases he might decide to ask the authors to provide the raw data before sending it out for review, especially if the results look ‘too beautiful to be true.’
Over the two years of manuscripts he analyzed, Miyakawa sent out requests for the raw data 41 times. He only received data in 20 cases. In 19 out of those 20 cases, the data was either incomplete, or did not match with the results in the manuscript. The reason why in so many cases the raw data was not available or incomplete is not clear, but it might suggest that this data never existed in the first place.
Miyakawa calls for increased availability of raw data. His journal, Molecular Brain, now requires the raw data for every manuscript submitted to be made publicly available. This makes it easier to determine whether results are genuine, and it will increase transparency. Although the reproducibility crisis is complicated and will not be solved completely with just this requirement, it is definitely a step in the right direction.