Assessing scalar implicatures in second language learners: A systematic review

• Synthesizes 25 years of L2 scalar implicature research using PRISMA methods. • Reveals major methodological sources of variability across studies. • Shows L2 SI performance depends on scale type, exposure, and proficiency. • Calls for multi-scale, cross-linguistic, and online methods in future research. This systematic review examines studies on how adult second language (L2) speakers derive scalar implicatures – inferences such as interpreting ‘some’ as ‘some but not all’ – and how methodological, linguistic, and individual factors modulate their pragmatic performance. Following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, a systematic search was conducted across Scopus, Web of Science, and Google Scholar, covering studies published between 2000 and 2025. After screening and applying the inclusion criteria, a total of 19 experimental studies were retained for analysis. These studies collectively test over 2500 participants from diverse linguistic backgrounds and L2 proficiency using a wide range of tasks (truth-value judgment, covered box, self-paced reading, sentence evaluation). Results from the 19 reviewed studies reveal substantial methodological heterogeneity but highlight robust qualitative patterns. L2 learners generally succeed in deriving SIs, yet outcomes vary with (i) exposure type (classroom vs. immersion), (ii) proficiency assessment, (iii) task demands, and (iv) lexical diversity of the tested scales. Evidence from recent studies suggests that differences often reflect processing constraints and lexical-semantic transfer rather than a genuine pragmatic difficulty. Moreover, factors such as response format (binary vs. graded) and stimulus informativeness significantly affect judgments, challenging long-standing assumptions about L2 pragmatic insensitivity. We conclude that L2 pragmatic competence in scalar inference is modulated by input experience, individual factors, cognitive resources, and task design. The review calls for methodological standardization and for expanding beyond the canonical < some, all > scale to encompass scalar diversity, cross-linguistic transfer, and online processing measures.