| Summary: | 碩士 === 國立臺灣大學 === 語言學研究所 === 106 === Increasing number of studies have suggested that statistical learning may play a more fundamental role in supporting language acquisition than previously thought. However, the issue of how the left hemisphere (LH) and the right hemisphere (RH) learn to master the statistical regularities in language is still poorly understood. In view of this, the present study aims to take a critical first step toward understanding how the two hemispheres track statistical regularities. In this study, we targeted Mandarin Chinese native speakers to investigate: (1) are both hemispheres capable of picking up non-adjacent statistical regularities from language-like input? (2) If yes, do the learning trajectories differ between the two hemispheres? And critically, (3) do the two hemispheres rely on mechanisms of different nature to track statistical regularities in language, akin to how syntactic information is processed in the two hemispheres?
An ERP experiment with alternate monaural listening mechanism in presenting the materials was conducted in this thesis. In total, forty-seven right-handed young adults without familial sinistrality background (FS-) participated. With the classic artificial language learning paradigm, non-adjacent dependencies were chosen as the learning materials in this experiment. In order to systematically manipulate the transitional probabilities (TP) between test tokens, different numbers of the intervening items were assigned to create three different variability conditions (low, mid, and high).
The behavioral results failed to replicate past findings when all participants were considered; nevertheless, if focused on successful learners only, high variability condition indeed best facilitated the learning of non-adjacent dependencies. Alternating training and test phases might be determinant factors contributing to these conflicting findings since the TPs do not exactly meet the goal of the experimental design. The ERP results displayed symmetric brain responses for low and mid variability conditions, with P600 for low and N400 for mid. Only in high variability condition did LH and RH show asymmetric results—left-lateralized P600 effects and right-lateralized N400 effects. These brain responses have proven that both left and right hemispheres have the ability to undergo statistical learning processes; however, only in high variability condition can we see each hemisphere using different processing approach (LH: syntactic processing approach/ RH: lexical association based approach). Finally, together with behavioral results, we think the lateralization of P600 grammaticality effects may be closely linked to the syntactic proficiency of non-adjacent dependencies. These data patterns are based on a relatively small ERP data set, and more data need to be obtained before firmer conclusions can be drawn.
|
|---|
