The naturalness of formulaic language is a key variable in the acceptability of multilingual academic writing. Such naturalness is best reflected by lexical collocations between linguistic units. However, no well-established framework has systematically studied collocational acceptability in multilingual academic contexts. This study fills the gap by proposing a triangulated model to assess academic lexical collocations produced by Indonesian authors in English Language Teaching (ELT) research publications.The model integrates three components: corpus data, statistical metrics, and cross-group speaker judgments based on Kachru's (1985) three circles of English.Findings show that many recurrent collocations, such as specific Adverb-Verb and Verb-Noun structures, are semantically transparent and moderately acceptable, even when they diverge from prototypical native-speaker norms. By applying a decision logic of MI Score > 3.0, a negative T-score, and I-CVI > 0.78, the proposed model contributes to research on corpus linguistics in education by linking phraseology competence and multilingual academic communication.
Copyrights © 2026