Peer comment wording in pre-service teacher microteaching tracks rating severity but not the dimensions rated low

Abstract In microteaching peer evaluation, pre-service teachers often give each peer both numerical ratings and a short written comment. This study asks whether the explicit content of those comments corresponds to the evaluator’s own ratings, both overall and for specific rated items. We analyzed 702 peer evaluations from a single Japanese cohort (27 pre-service teachers, each rating 26 peers; self-evaluations excluded). Each evaluation paired nine forced-distribution ratings, treated as within-evaluator relative ranks, with one short Japanese comment. Using a fixed rule-based lexicon, we measured five explicit comment markers (generic praise, concrete instructional referent, learner-oriented mention, feed-forward, and criticism) and related them to the ratings with within-evaluator permutation tests. Generic praise was common (67.1%), whereas explicit feed-forward and criticism were rare (about 13–14% and 5–6% of comments). Comments paired with lower relative ratings carried corrective markers more often, and this rose with the number and severity of low ratings. However, the link was general rather than targeted: when a comment named a teaching dimension, it was no more likely than chance to be one the evaluator had actually rated low (36.7% vs 39.6%). In this exploratory, single-cohort study, explicit corrective wording covaried with the overall severity of an evaluator’s relative ratings, but dictionary-detected dimension mentions did not preferentially identify the dimensions rated low, a loose coupling between the numerical and textual channels. We report transparent, rule-detected markers of explicit comment content, not human judgments of feedback quality.