<![CDATA[Research on academic writing has long described the rhetorical structure of research article introductions through Swales’ Create-a-Research-Space (CaRS) model. However, the question of whether rhetorical segmentation is systematically realised in surface linguistic form remains underexamined. The present study provides empirical evidence that rhetorical segmentation in academic introductions is reflected in the distribution of sentence-initial expressions. The analysis is based on the CaRS-50 corpus, comprising fifty research article introductions annotated at sentence level for Moves and Steps. Sentence-initial spans up to the first comma were extracted automatically and examined in relation to sequential Move transitions. Across 1,297 sentences, 36.4% begin with an identifiable opener, while 21.5% of adjacent sentence pairs involve a Move change, establishing a corpus-level baseline transition probability. Recurrent openers were evaluated relative to this baseline. The results show that sentence-initial lexical items vary in their transition behaviour relative to the corpus baseline: some occur above the baseline at points of Move change, whereas others predominantly mark within-Move continuation. Sentence-initial marking is therefore not confined to rhetorical boundaries but participates in both continuation and segmentation contexts within the annotated Move sequence. Without modifying the three-Move CaRS framework, the study introduces a probabilistic perspective on Move organisation. The findings extend genre-analytic descriptions of Move structure into a measurable account of rhetorical progression through a probabilistic analysis of sequential Move transitions and sentence-initial lexical distribution.]]>
