Grammatical generalisation in statistical learning: The conribution of vocabulary, explicit and implicit knowledge

The learning and generalisation of grammatical regularities is fundamental to successful language acquisition and use. Research into statistical learning has started to consider how this process occurs through the implicit encoding and integration of grammatical regularities. Drawing on lexicalist and multi-componential theories, this thesis uses a statistical learning framework to explore two possible contributors to grammatical generalisation: 1) vocabulary, and 2) explicit and implicit knowledge. Across four studies adults and children learnt an artificial language containing two semantic categories denoted by a co-occurring determiner and suffix. The first two studies examined the role of vocabulary. Unlike child learners, in adult learners increasing the variability of vocabulary items in the learning context enhanced grammatical generalisation. When adults’ vocabulary knowledge was reduced to that of the children, the adults’ grammatical generalisation was likewise reduced to the same levels as the children’s. In the first two studies, the explicit knowledge that emerged in the course of the study influenced adults’ generalisation, but not the children’s. Using more sensitive measures of explicit knowledge, the final two studies also found a contribution of explicit knowledge in adult learners, but only for regularities that showed the highest level of learning. In addition, the contribution of implicit knowledge was examined using two measures: a confidence rating measure revealed evidence of implicit knowledge of all the regularities tested, but this was not detected when using a novel, reaction time-based measure. To conclude, vocabulary is important for grammatical generalisation, with both explicit and implicit knowledge contributing to adult grammatical generalisation. The theoretical implications of these findings are discussed along with future developmental considerations.