Beyond speed and beyond prediction: The multi-faceted nature of speech processing in preschoolers and how it relates to language development

Efficient language processing requires both the ability to predict as well as flexibly revise. Numerous studies have demonstrated that young children can make predictions during online language processing tasks, but less is known about their ability to quickly revise interpretations when initial predictions are infelicitous. Adults readily revise initial interpretations during language processing across linguistic domains—recovering from garden paths, adjusting to acoustic variation, and resolving lexical ambiguities. However, it is relatively unknown how revision mechanisms develop in early childhood, and, even more critically, how revision ability relates to the acquisition of language. We examined four- and five-year-olds’ (N=60) phonological revision abilities using an eye-tracking paradigm that captured children’s initial phonological interpretation of a word and subsequent revision and recovery. We characterized children’s processing along 11 distinct dimensions, and employed unsupervised clustering analyses to identify three distinct speech processing types among the children. Results showed substantial individual differences in processing strategies with different patterns of initial interpretation, as well as subsequent revision, underlying similar language outcomes. Processing type membership significantly predicted children’s vocabulary and pre-literacy skills, suggesting that phonological revision mechanisms may underpin multiple domains of language development. Overall, our findings align with both Bayesian inference and error-driven learning accounts where the probabilistic prediction and resolution of mismatches may drive language development across multiple linguistic domains.