The capacity of LLMs to create coherent mappings between representational fields -- visual, linguistic, sonic -- by chasing connotative rather than denotative chains.