Pexels
Longtime Retraction Watch readers know the scientists on our Leaderboard have changed over the years. But one characteristic has remained relatively constant: There are few women on that list – in fact, never more than one at a time.
So when a recent paper dove into whether retraction rates vary by the gender of the authors, we were curious what the authors found.
The team, from Sorbonne Study Group on Methods of Sociological Analysis (GEMASS) in Paris, sampled 1 million articles from the OpenAlex database, then referenced the Retraction Watch database to compare against their sample.
After identifying gender based on the first names of the authors of the articles, the researchers looked at single-author studies by both men and women and teams of researchers of both mixed and single-gender researchers. Regardless of gender, single-author papers were less likely to be retracted than studies with multiple authors. Mixed teams were more likely to have retractions than single-gender teams. Mixed author groups led by women were slightly less likely to face retractions than mixed groups led by men.
We asked the researchers to expand on the methods and results of their study, which appeared March 6 in Quantitative Science Studies, the journal of the International Society for Scientometrics and Informetrics. The authors, Abdelghani Maddi, Emmanuel Monneau, Catherine Guaspare-Cartron, Floriana Gargiulo, and Michel Dubois, responded collectively via email. The questions and responses have been edited for brevity and clarity.
Retraction Watch: What are the main takeaways from the analysis?
GEMASS Researchers: The study provides a nuanced examination of how gender influences the dynamics of scientific retractions. It highlights that the gender composition of author teams—whether single-authored or collaborative, including mixed-gender groups—can correlate with distinct patterns in article retractions.
The analysis suggests that gender might affect both the likelihood of an article being retracted and the underlying processes that lead to such outcomes. This unexpected influence calls into question common assumptions about scientific risk-taking and error correction, indicating that the interplay between individual traits and team dynamics is more complex than previously understood.
Overall, beyond gender-related aspects, our findings highlight several factors that increase the likelihood of retraction:
- Publications in Open Access
- Publications in high-impact journals
- Medium-sized research teams (3 to 10 authors)
- Conversely, publications with project-based funding were associated with a reduced probability of retraction
- Research in the health and life sciences exhibited higher retraction probability of retraction compared to the social and physical sciences.
Retraction Watch: What are some of the other takeaways beyond article authorship?
GEMASS Researchers: The research points toward significant broader implications for the scientific community. Mixed-gender teams, in particular, may benefit from a more balanced approach to oversight and self-correction, potentially mitigating errors and reducing the incidence of retractions. The findings underscore that retraction is not solely a consequence of isolated mistakes or misconduct but is also intertwined with social and cultural dimensions, including potential biases in peer review and the public perception of scientific credibility.
These findings challenge common assumptions about scientific rigor and integrity across genders. The evidence implies that while male-dominated research environments might be more prone to issues related to misconduct, the relatively lower incidence of retractions among women could reflect either a more cautious approach or differences in the opportunities and pressures encountered in their careers. Ultimately, the study calls for a deeper examination of how gender dynamics intersect with research practices and accountability, highlighting the need for nuanced policies that promote both equity and excellence in science.
Retraction Watch: You categorized gender by looking at authors’ first names, but were only able to determine gender for 70% of your original sample. Why did you pick this method? What are its shortcomings?
GEMASS Researchers: We chose the first-name method because it is the most commonly used approach to infer gender in large datasets, with tools like Genderize.io, Gender API, NamSor, and Wiki-Gendersort. Manual inference would be challenging in our case, given that our sample includes over 2.6 million authors.
However, we recognize the limitations of automated gender detection methods, particularly in terms of accuracy. To mitigate this, we applied a validation process, both technical and functional, to ensure that the observed global patterns in our data aligned with findings from existing literature.
Retraction Watch: While single-author studies are less likely to be retracted than multiple author studies, your study found the risk stops increasing, or even decreases, when a study has more than 10 researchers. How do you interpret this threshold?
GEMASS Researchers: While this result may seem counterintuitive, it aligns with prior research. Specifically, Rathmann and Rauhut (2019) suggested that larger research groups may benefit from enhanced social control and oversight. Similarly, Sharma (2021) observed that smaller teams tend to be more vulnerable to retractions, underscoring the importance of team size in maintaining research integrity.
Our findings introduce additional nuance: medium-sized teams– which we define as three to 10 authors–are at higher risk of retraction compared to larger teams. This suggests that while collaboration can mitigate misconduct risks, medium-sized teams may face challenges related to unclear roles and responsibilities, potentially increasing oversight gaps.
Retraction Watch: How do you interpret the result that mixed collaborations led by women are less likely to be retracted than mixed collaborations led by men?
GEMASS Researchers: Both the odds ratios and the crude ratios indicate that mixed-gender publications with a male first author have a higher likelihood of retraction compared to those led by a female first author. This difference remains statistically significant in our regression analysis, with odds ratios of 2.27 versus 1.65, even after controlling for other variables.
Without qualitative data, we can only hypothesize. One potential explanation, suggested by existing literature, is that gender-based social norms may influence both behavior and how misconduct is judged. Women, for example, may be less frequently less exposed to retraction due to tendencies to adopt more cautious, apologetic behavior, while men may engage in riskier behaviors. It is possible that women commit misconduct at similar rates but are less often detected or sanctioned. It is also possible that women exercise their responsibilities as first author differently from men.
Retraction Watch: Why do you think studies led by men are more likely to be retracted for misconduct, while studies led by women are more likely to be retracted for errors or disguised text plagiarism?
GEMASS Researchers: While male- and female-led publications share certain breaches of integrity, such as plagiarism, fabrication, image or text manipulation, the retraction profile of male-led publications is over-represented in characterized departures from integrity, i.e., misconduct, and breaches of ethics, such as lack of ethical authorization, whereas the retraction profile of female-led publications is over-represented in errors. Beyond this observation, our study, based on two databases, does not go any further in understanding these disparities. It would be necessary to add a qualitative approach to explore correction practices in greater detail.
Retraction Watch: Are there sufficient studies with non-binary authors to be included in this kind of analysis?
GEMASS Researchers: The practice of scientific signatures remains to our knowledge indifferent to the question of gender. When we sign an article, we are not supposed to declare ourselves as “male” or “female”. However, it is possible to identify the relative share of men and women by analyzing first names, often in conjunction with country-specific conventions – e.g., recognizing that “Andrea” is typically a woman’s name in the United States but a man’s name in Italy. In this study, we used the Genderize.io algorithm to infer the gender of the authors. Current bibliometric datasets lack metadata on non-binary authorship, which makes this analysis impossible. To our knowledge, no studies, on this topic, have yet integrated this dimension.
Retraction Watch: Do you believe journals should ask researchers for their gender or pronouns before publishing?
Collecting this information could indeed improve the depth and accuracy of gender-based analyses. If journals were able to collect this type of data by giving researchers the opportunity to declare their gender, they would enable researchers to improve the accuracy of detection, given that automated tools fail to classify around 30% of cases due to the presence of initials only or unisex first names. This would also enable us to broaden research questions surrounding gender disparities in research collaborations.
Like Retraction Watch? You can make a tax-deductible contribution to support our work, follow us on X or Bluesky, like us on Facebook, follow us on LinkedIn, add us to your RSS reader, or subscribe to our daily digest. If you find a retraction that’s not in our database, you can let us know here. For comments or feedback, email us at [email protected].