Abstract Language production includes multiple subsystems such as the lexical, syntactic, and semantic domains. These subsystems do not operate in isolation but instead interact during production. Previous research has documented complexity trade-offs, cases in which an increase in complexity in one linguistic domain tends to be accompanied by a reduction in others. However, existing studies have primarily focused on trade-offs between syntax and lexicon using small datasets. Our study is the first to extend research on trade-offs into the semantic domain. Two innovative semantic complexity indices (semantic specificity and semantic richness) are incorporated to offer insights into the interaction with semantic complexity. Additionally, we evaluate the effect of sentence length. We conduct Spearman correlation tests and fit generalized additive mixed models to assess the trade-offs. Our results provide converging evidence of complexity trade-offs in natural language production across lexical, syntactic, and semantic domains. We then confirm that longer sentences make complexity trade-offs more prominent. Complexity trade-offs may be attributed to the constraints imposed by the limitations of cognitive resources and the need for efficient communication. Our findings may represent a universal feature of languages, offering insights into how efficiency and human cognitive resources shape language production.
Zhao et al. (Fri,) studied this question.