Familiar words iconicity shapes novel words meaning

Iconicity, the perceived resemblance between a word’s form and meaning, is systematically related to the semantic organization of familiar vocabulary. This raises the possibility that iconicity-related regularities also constrain how people interpret unfamiliar words. We tested this proposal by learning mappings from (sub)lexical representations to large-scale English iconicity ratings and examining whether predicted iconicity for pseudowords was associated with the semantic quality and specificity of human-generated definitions. Higher iconicity predicted from semantic representations was associated with definitions that aligned more closely with their target pseudowords and distinguished them more selectively from alternative forms. In contrast, orthographic features predicted iconicity in familiar words but showed no corresponding associations with pseudoword definitions. The positive associations for semantic representations generalized across multiple encoders, including analyses in which different models were used to predict iconicity and evaluate definitions. These findings suggest that iconicity-related semantic structure in familiar vocabulary is associated with the construction of meanings for unfamiliar forms. They are consistent with discriminative and complementary learning accounts in which previously acquired form-meaning relations provide graded constraints on novel interpretations and, more broadly, they connect theories of iconicity grounded in perceptual and experiential resemblance with distributional accounts of lexical knowledge, suggesting that linguistic experience organizes and supports the generalization of resemblance-related semantic information to novel forms.