Imprimir Republish


Data scientist reveals that papers whose results are not confirmed continue to receive citations

Data scientist Paul von Hippel of the University of Texas at Austin, USA, evaluated the extent to which scientific papers in the field of psychology lose credibility when their results are not confirmed by later studies. In a survey published in June’s issue of the journal Perspectives on Psychological Science, he analyzed the number of citations—a measure of an academic article’s influence—of 98 papers from a psychology journal. All were published in 2008 with attempts made to replicate the results in 2015.

Von Hippel found that among studies that were not replicated, the number of citations only dropped by about 5–9% compared with those for which the findings were successfully reproduced. He also noted that less than 3% of papers in which the original articles were cited mentioned the replication study. The researcher offered a series of recommendations on how to prevent flawed results from continuing to serve as a reference. One suggestion is that authors should have to cite not only original studies but also any replication studies that have been carried out, and proposes that search tools and scientific databases be improved to increase the visibility of replication studies. Von Hippel did not choose psychology as the field of study at random: over the past decade, the discipline has been shaken by a series of cases of articles falling into disrepute because their results were not repeated in subsequent experiments—some due to errors and others due to fraud.