The semantics behind the inference: How first language quantifier systems shape scalar reasoning in second language learners

This study investigates how L1-Japanese learners of English pragmatically enrich quantifiers in the L2, focusing on the role of L1 semantics, lexical competition, and contextual informativeness. Our central aim was to chart cross-linguistic and interlanguage differences in the underlying quantifier systems of English and Japanese, and to assess how these shape the lower and upper semantic thresholds of scalar terms like some. Using the gumball paradigm developed within a Constraint-Based framework, we elicited gradient naturalness judgments across two experiments to map how some is distributed over fine-grained numerosity continua. In unpressured contexts with limited competitors (Experiment 1), L2 learners patterned closely with native English speakers in their interpretations of some, suggesting emerging target-like intuitions. However, when the lexical alternatives a few and most were introduced alongside some (Experiment 2), systematic L1-based transfer effects emerged: learners interpreted a few and most in line with their Japanese counterparts (sukoshi, hotondo), displaying narrower or delayed acceptability ranges. These transfer patterns disrupted scalar coordination, resulting in an expanded acceptability window for some and a unique scalar reasoning profile. Together, the results map how cross-linguistic differences in quantifier systems affect L2 learners’ scalar representations and highlight the need for models of inference development to incorporate language-specific constraints on quantifier meaning alongside consideration of pragmatic mechanisms.