On May 25, 2016 the peer-reviewed journal Nature published the results of an in-house survey on reproducibility of scientific experiments. The survey included more than 1,500 Nature readers representing researchers from the following disciplines: Biology (703), chemistry (106), earth and environmental (95), medicine (203), physics and engineering (236), and other (233). 52% of participants agreed that there is a “significant crisis” of reproducibility in science; however, less than 31% think that failing to reproduce published results means that the results are wrong. 73% of participants think that at least half of the publications in their field are reproducible. Most of the researchers indicated to have failed reproducing an experiment, either someone else’s or their own; however, less than 20% of respondents reported to have been contacted by another researcher unable to reproduce their work. Only a minority of participants reported to have attempted publishing a replication study. 33% of the researchers said their laboratory or institution has taken concrete steps to improve reproducibility within the past five years. Redoing experimental work or asking someone else to repeat the work, as well as improving documentation and standardization of experimental methods seem to be the most common practices therefor, according to the survey. In terms of factors contributing to irreproducible research, most participants named “selective reporting” and “pressure to publish.” As factors contributing to improved reproducibility, a majority of participants named “better understanding of statistics,” “better monitoring/supervision,” and “more robust design.”
Read more
Nature (May 25, 2016). “Reality check on reproducibility.”
Monya Baker (May 25, 2016). “1,500 scientists lift the lid on reproducibility.” Nature
