The role of prediction in language learning: evaluating error-based theories of language acquisition

Fazekas, Judit
(2020) The role of prediction in language learning: evaluating error-based theories of language acquisition. PhD thesis, University of Liverpool.

[img] Text
201204456_Jun2020.pdf - Unspecified

Download (2MB) | Preview


This project examines prediction’s role in language acquisition by assessing the projections of error-based language acquisition theories, specifically the Dual-path model. It seeks to expand on results obtained by the prime surprisal paradigm and to develop novel studies building on this method. In Chapter 2, we examined whether prime surprisal appears with passive as well as dative structures in adults. While we found a numerical, but non-significant prime surprisal effect in the dative study, no such effects appeared with passives. In Chapter 3, in a four-stage prime surprisal-based intervention study we examined whether surprising sentences lead to more learning than predictable ones. Here we found preliminary evidence for enhanced learning rates for abstract structures after surprising input with 5-6 year-old child participants. These long-term effects were stronger in younger children, although they were dissociated from immediate prime surprisal. We found no significant evidence for verb-dependent error-based learning effects in either group or for an abstract learning effect in the adult group alone. In Chapter 4 we compared the ERPs associated with predictable and surprising dative sentences in adult participants and found an enhanced N400 effect in surprising (as opposed to predictable) DOD (double object dative) sentences but we found no such effect in prepositional (PD) sentences. We also found no significant differences in the P600 region. While the N400 effect shows predictability-related processing differences, based on these results alone we cannot determine whether these differences are the result of active prediction or differences in the ease of integration. Some of our key results (such as the replication of immediate prime surprisal effects in two of our studies, the preliminary evidence for enhanced learning rate with unexpected structures or the N400 effects) support error-based learning theories. However, other results (such as the dissociation of immediate prime surprisal and longer-term learning effects or the lack of P600 effect in surprising sentences) raise questions about the exact nature of the mechanism leading to these effects and warrant further examination.

Item Type: Thesis (PhD)
Divisions: Faculty of Health and Life Sciences
Depositing User: Symplectic Admin
Date Deposited: 14 Aug 2020 10:39
Last Modified: 09 Apr 2022 07:12
DOI: 10.17638/03090904