Poster #: 34
Topic: Development (infancy, childhood, adolescence, and aging)
Wednesday, Sep 9, 2015
Prediction errors in word recognition and learning in young children
1Cognitive Brain Research Unit, Institute of Behavioral Sciences, University of Helsinki, Helsinki, Finland
2Institute of Behavioural Sciences, University of Helsinki, University of Helsinki, Finland
Previous studies have suggested that in adults, word recognition is driven by predictions. Here we show that inferring predictions is crucial to language learning in children. We recorded auditory event-related brain potentials (ERP) to the same target syllables in isolation (e.g., [ka]) and within context (e.g., [ku][ka], together forming the word [kuk:a] ‘flower’). As compared to no context condition, the context condition was hypothesized to allow more efficient processing of the target syllables due to word-level predictive coding, which was expected to be enhance and accelerate ERP responses. In the context condition, the target syllables completed either a familiar ([kuk:a] ‘flower’) or an unfamiliar ([kuk:e] with no meaning) word form. Results showed that 12- and 24-month olds’ electrophysiological brain responses to heard syllables are faster and more robust when the preceding word context predicts the ending of a familiar word. For unfamiliar, novel word forms, however, word-expectancy violation generates a prediction error response, the strength of which significantly correlates with children’s vocabulary scores at 12 months. This suggests that predictive coding serves as the neural mechanism for both word recognition and early learning of novel word forms. We argue that language development is guided by the same prediction-based learning mechanism reported across domains in human and non-human animals.