Prediction and error in early infant speech learning: A speech acquisition model

In the last two decades, statistical clustering models have emerged as a dominant model of how infants learn sounds their language. However, recent empirical and computational evidence suggests that purely methods may not be sufficient to explain speech sound acquisition. To early development perception, present study used two-layer network trained with Rescorla-Wagner learning equations, an implementation discriminative, error-driven learning. The contained no priori linguistic units, such phonemes or phonetic features. Instead, expectations about upcoming acoustic signal were learned from surrounding signal, spectral components extracted audio recording child-directed both inputs outputs model. evaluate performance, we simulated infant responses in high-amplitude sucking paradigm using vowel fricative pairs continua. simulations able discriminate consonant predicted perception data. also showed greatest amount discrimination expected frequencies. These results suggest discriminative provide viable approach modelling