Research misconduct may leave traces in the text itself, not only in how the research is conducted, suggests a new study from Chalmers University of Technology, Sweden. By analysing scientific articles later retracted for misconduct, the researchers identified five recurring rhetorical “warning signs” that can indicate when a study is designed to appear credible despite unreliable foundations.

The findings could make misconduct detection more tangible and support training for doctoral students and early-career researchers. The study suggests that working with authentic retracted articles can help develop critical reviewing skills and strengthen research integrity at a time when scientific publishing is expanding rapidly.

In the study, published in the journal Accountability in Research, twenty doctoral students analysed retracted research articles as part of their reviewer training. The results show that the approach increases doctoral students’ “rhetorical sensitivity” – their attentiveness to linguistic signals that may indicate misconduct or serious shortcomings.

"We used these articles as a form of living course material in research integrity – an “opening Pandora’s box” approach where one dares to learn from problems instead of ignoring them. By working with real cases, we hope to make future reviewers more attentive to early warning signals", says Baraa Khuder, Senior Lecturer at the Department of Communication and Learning in Science at Chalmers, and the researcher behind the study.

When the doctoral students analysed the retracted articles, five recurring rhetorical patterns emerged that characterise unreliable research:

1. False or fake references (Intertextual falsification)
   References and prior research are used in misleading or incorrect ways. Articles may give the impression of strong support from literature, even though sources are distorted, misrepresented, or in some cases entirely fabricated.

2. Unclear research process (Methodological opacity)
   The methodology appears sound and detailed at first glance, but crucial aspects of the research process are unclear or omitted, making it difficult to assess the study’s reliability.

3. Inconsistent content (Rhetorical inconsistency)
   Different parts of the text do not align. Introductions, results, and conclusions may point in different directions, complicating critical evaluation.

4. Grandiose conclusions (Rhetorical overstatement)
   Findings are presented using overly confident language, portraying claims as indisputable and leaving little room for uncertainty or alternative interpretations.

5. Inaccurate terminology (Terminological distortion)
   Key concepts and technical terms are used incorrectly or inconsistently, creating conceptual confusion and potentially misleading the reader about what the study shows.

"The fact that these patterns recur across multiple retracted articles suggests that research misconduct and serious errors are often accompanied by similar rhetorical strategies. By recognising these warning signals, reviewers and readers can identify potential problems in a study at an earlier stage", says Baraa Khuder.

The article “Opening Pandora’s box: Developing reviewer rhetorical sensitivity through retracted articles” is written by Baraa Khuder at Chalmers University of Technology, Sweden. The paper has been published in the journal Accountability in Research.