Signatures of Adaptive Memory Search: How Early Linguistic Input Shapes Strategic Use of Lexical Information

Producing words during the semantic fluency task (e.g., “name all the animals you can in a minute”) involves not only activating the relevant semantic representations, but also being able to flexibly utilize phonological information to facilitate retrieval. However, the mechanisms by which individuals effectively use phonology as a latent cue during memory search remain understudied. In this study, we used a computational approach to examine the use of different lexical sources during memory search in two categorical domains (Animals and Foods) within the semantic fluency task. We captured the relative saliency of semantic and acoustic information by creating blended lexical representations obtained from two language models trained on text (word2vec) and speech (speech2vec). To examine process-level mechanisms, we implemented computational models based on optimal foraging principles to characterize transitions within and across clusters of related items. These models were evaluated within a sample of normal hearing children and a naturally occurring control group with limited early auditory and language experience – prelingually deaf children with cochlear implants. Overall, in domains where semantics and phonology were more highly correlated (i.e., Foods), greater use of phonology in within cluster transitions was associated with better fluency performance for those with normal hearing. In contrast, though children with cochlear implants amplified acoustic structural information in these domains, they were less able to flexibly leverage phonological information to retrieve more items within clusters. Together, this work highlights how early linguistic experience shapes the ability to use semantic and phonological cues to navigate lexical space in a controlled, adaptive manner.