Testing Hypotheses on Risk Factors for Scientific Misconduct via Matched-Control Analysis of Papers Containing Problematic Image Duplications.
Level 4 - case-series / case-control
Matched case-control study of published papers
PubMed 29460082 · doi:10.1007/s11948-018-0023-7
What was done
A pre-registered matched-control study evaluating four hypotheses regarding risk factors for scientific misconduct (pressure to publish, lack of social/peer control, absence of national misconduct policies, and male gender). Cases were papers identified in a prior systematic screening of PLoS ONE as containing problematic image duplications, classified into three categories of increasing complexity (Category 1 being most likely unintentional error to Category 3 being most likely intentional fabrication). Each case was matched to two control papers, and univariable and multivariable analyses were conducted on parameters linked to the hypotheses.
What was found
The abstract reports directional associations without quantitative statistics (no exact effect sizes, odds ratios, or p-values): - Category 1 duplications showed mostly no significant associations with tested parameters. - Categories 2 and 3 showed numerous statistically significant associations supporting the roles of academic culture, peer control, cash-based publication incentives, and national misconduct policies. - No clear support was found for the "pressure to publish" hypothesis. - Female and male authors were equally likely to publish duplicated images. - Country-level parameters exhibited stronger effects than individual-level parameters, with developing countries significantly more likely to produce problematic image duplications.
Why it matters
This study provides rare controlled, pre-registered empirical evidence assessing common hypotheses about scientific fraud, indicating that structural factors like national policy and cash incentives are stronger determinants of integrity issues than gender or general publication pressure.
Limits
Exact sample sizes (n), effect sizes, and test statistics are not reported in the abstract. The analysis was limited to visible image duplications from a single mega-journal (PLoS ONE), which may not generalize to other fields, non-image misconduct (such as text plagiarism or numeric data falsification), or journals with different review standards. Intentionality was inferred from duplication complexity rather than direct investigation findings.
Cited by
- contradicts The pressure on researchers to publish landmark papers in top-tier journals is one of the primary drivers of scientific fraud.