Second language processing of derived words in English : an ERP investigation

Abstract

What happens when second language speakers read derived words? Do their brains process and store these words as their whole-word forms, or are these words broken down into their constituent morphemes and stored individually? Behavioral literature has come to mixed conclusions, with some studies finding evidence of L2 morphological decomposition, and others finding evidence of whole-word storage. In this thesis, we present two studies that utilized electroencephalogram (EEG) to investigate the time course of L2 derived word processing in English. In our first study, we recorded EEG signals from advanced native Cantonese speakers of English and native English speakers, testing priming to three different types of prime-target pairs: morphologically related (swiftly-swift), only semantically related (explode-burst), and only orthographically related (surgeon-surge) pairs. For native English speakers, morphological priming occurred as attenuated early (350–430 ms) and late (430–500 ms) N400 amplitudes. Orthographic priming occurred in both N400 windows (350–500 ms), and orthographic inhibition occurred in the form of a late negativity (500–600 ms). Semantic priming occurred as an attenuated N400 in both time windows as well in the late negativity window (350–600 ms). L2 speakers replicated the L1 morphological and orthographic priming effects, but did not show orthographic inhibition during the late negativity window. Furthermore, L2 semantic priming was only significant from the late N400 window onwards (430–600 ms), and showed inhibition during the early N400 window. Importantly, we observed morphological priming that was stronger than the effects of orthographic and semantic priming combined in both the L1 and L2 groups, revealing that advanced L2 speakers decompose morphologically complex words during visual word processing as L1 speakers do. In our second study, we recorded EEG signals to two groups of L2 learners with even higher proficiency than the advanced L2 speakers in study 1, representing exceptional L2 learners, to investigate whether L2 morphological processing becomes more native-like as proficiency increases. We found evidence of morphological decomposition in both L2 groups, and that as proficiency increases, L2 morphological processing became more native-like but still showed qualitative differences from L1 processing. Semantic priming was not significant in the early N400 window for either of our exceptional L2 groups, and was longer lived in our higher proficiency group than our lower proficiency group. Additionally, morphological priming that was greater in the lower proficiency group remained significant in the late negativity window for both L2 groups, suggesting that L2 morphological priming was slower than L1 processing but became faster as proficiency increased. Together, the results of the first and second study show that advanced L2 speakers decompose morphologically complex words like L1 speakers do, but that early morphological processing is not facilitated by semantic relatedness, and that orthographic overlap does not cause inhibition until exceptional levels of proficiency are reached. Our results contribute to the current body of literature on L2 morphological processing that consists only of behavioral studies, providing neurophysiological evidence of native-like processing in exceptional L2 speakers that still contained qualitative differences in its duration.