Batterink, L.
Western University
Statistical learning (SL) is the ability to detect structure in the environment. SL is typically studied using a direct memory test, in which learners are asked to discriminate between ?words? from a previously presented continuous speech stream and foil items. This offline measure has several limitations: it is influenced by peripheral cognitive processes such as memory retrieval, does not reveal the time course of learning, and assesses the outcome rather than the process of learning itself. To begin to bridge this gap, we have been tracking SL directly by using an EEG-based neural entrainment measure, in which learning is indexed by neural entrainment to hidden embedded words in continuous speech. Neural entrainment to embedded words increases as a function of exposure to reveal the progression of learning and predicts individual performance on later learning tests. Using this approach, we have also found evidence that SL (1) can occur outside of focused attention, (2) can occur remarkably rapidly, and (3) relies on overlapping mechanisms in pre-lingual infants and adults. These findings highlight the powerful and ubiquitous nature of SL. Future studies using this measure are continuing to explore the optimal conditions, underlying mechanisms, and age-related changes associated with SL.