This study employs text mining to quantitatively analyze peer review reports, aiming to show that modifying review modes can mitigate publication bias. Results-Blind Review (RBR) aligns with Registered Reports (RRs) in homogeneity, while Transparent Peer Review (TPR) represents a transparent version of traditional peer review. Using BMC Psychology as a case study, we applied LDA topic modeling to identify reviewers’ focal points, sentiment analysis to assess emotional tones, and content analysis to evaluate outcome polarity. Results reveal that RBR (functionally equivalent to RRs) emphasizes research quality and methodological rigor more prominently than TPR, exhibits more stable reviewer sentiment, and shows no significant preference for positive results. The study confirms the presence of publication bias in peer review and demonstrates that pre-registration and RRs can effectively reduce it. Publishing stakeholders should critically reflect on their role in perpetuating bias and actively promote innovative review models to counteract it.
He et al. (Wed,) studied this question.