Can journal reviewers dependably assess rigour, significance, and originality in theoretical papers? Evidence from physics

Peer review is a key gatekeeper for academic journals, attempting to block inadequate submissions or correcting them to a publishable standard, as well as improving those that are already satisfactory. The three key aspects of research quality are rigour, significance, and originality but no prior study has assessed whether journal reviewers are ever able to judge these effectively. In response, this article compares reviewer scores for these aspects for theoretical articles in the SciPost Physics journal. It also compares them with Italian research assessment exercise physics reviewer agreement scores. SciPost Physics theoretical articles give a nearly ideal case: a theoretical aspect of a mature science, for which suitable reviewers might comprehend the entire paper. Nevertheless, intraclass correlations between the first two reviewers for the three core quality scores were similar and moderate, 0.36 (originality), 0.39 (significance), and 0.40 (rigour), so there is no aspect that different reviewers are consistent about. Differences tended to be small, with 86% of scores agreeing or differing by 1 on a 6-point scale. Individual reviewers were most likely to give similar scores for significance and originality (Spearman 0.63), and least likely to for originality and validity (Spearman 0.38). Whilst a lack of norm referencing is probably the biggest reason for differences between reviewers, others include differing background knowledge, understanding, and beliefs about valid assumptions. The moderate agreement between reviewers on the core aspects of scientific quality, including rigour, in a nearly ideal case is concerning for the security of the wider academic record.