Abstract structural rules in novel meaning composition: Divergent learning strategies revealed by implicit and explicit measures

The ability to generalize previously learned knowledge to novel situations is fundamental for adaptive behavior, including interpreting unfamiliar compositional words. To investigate how people construct novel meanings based on structural rules, we developed a semi-artificial language paradigm that combines explicit meaning judgments with implicit semantic priming measures, allowing us to assess rule-based composition across multiple behavioral levels. Participants learned pseudo-words composed of known stems and unknown affixes, and were subsequently tested on novel combinations that required generalizing abstract structural rules. Across three behavioral experiments, we found reliable evidence for generalization on both implicit and explicit measures, yet these measures diverged in sensitivity. Critically, individuals differed in the degree to which they relied on the sequential order of word parts: Some employed a “building” strategy that integrated parts positionally, while others adopted a “mixing” strategy that ignored order. We further explored whether this variability could be explained by individual differences in stable trait factors. Together, this work offers a methodologically integrated approach for studying compositional meaning construction and provides an experimentally controlled paradigm suitable for future neuroimaging research on rule-based inference.