Gomez, R. L.
University of Arizona
Why is it that infants are such rapid statistical learners in laboratory studies but it takes months and years to learn a language? Our work tests retention over time to begin addressing this question. Infants who stay awake after statistical learning retain verbatim information 4 hours later but nothing the next day, whereas those who nap generalize a rule 4- and 24-hours later. How does sleep preserve learning? Here we test two theories of sleep-dependent consolidation. According to one view, sleep neural replay strengthens memories during sleep, in which case infants may generalize by analogy to a remembered instance. By another view, sleep sweeps the brain of weaker memories, retaining only the stronger ones in a process of cortical synaptic downscaling. In this case infants forget the less frequent details of the stimulus with sleep (the words themselves) but retain the more frequent details that serve to reactivate the rule. The mechanism, reflected in the experimental outcome, has implications for whether statistical learning accrues rapidly or slowly over developmental time.