From semantic structure to conventionalization

Abstract The present study investigates potential differences among 40 quantifier scales regarding measures that may indicate variations in conventionalization among different scales, i.e., the extent to which their terms are stored as scalemates in the mental lexicon. We derive these measures from two experiments and corpus data. In addition to exploring the relationships between these measures, we also investigate their correlation with scalar inference ( SI ) rates. We compare the quantifier scales to scales from two categories: highly and moderately conventionalized. Our findings indicate that quantifier scales pattern with moderately conventionalized scales in the experimental tasks. Additionally, we identify two significant predictors of SI rates that are not directly derived from the semantics of the quantifiers: confidence in an elicitation task and the observed/expected ratio of sentences in a corpus. Finally, we propose that the semantic and non-semantic features generally align, working together to reinforce the connection between the scalemates.