Electrophysiological dynamics of Chinese phonology during visual word recognition in Chinese-English bilinguals

Silent word reading leads to the activation of orthographic (spelling), semantic (meaning), as well as phonological (sound) information. For bilinguals, native language information can also be activated automatically when they read words in their second language. For example, when Chinese-English bilinguals read words in their second language (English), the phonology of the Chinese translations is automatically activated. Chinese phonology, however, consists of consonants and vowels (segmental) and tonal information. To what extent these two aspects of Chinese phonology are activated is yet unclear. Here, we used behavioural measures, event-related potentials and oscillatory EEG to investigate Chinese segmental and tonal activation during word recognition. Evidence of Chinese segmental activation was found when bilinguals read English words (faster responses, reduced N400, gamma-band power reduction) and when they read Chinese words (increased LPC, gamma-band power reduction). In contrast, evidence for Chinese tonal activation was only found when bilinguals read Chinese words (gamma-band power increase). Together, our converging behavioural and electrophysiological evidence indicates that Chinese segmental information is activated during English word reading, whereas both segmental and tonal information are activated during Chinese word reading. Importantly, gamma-band oscillations are modulated differently by tonal and segmental activation, suggesting independent processing of Chinese tones and segments.


Behavioural Results
Table S1 Mean error rates and reaction times (with SD in brackets) of English and Chinese Experiments +Segment +Tone +Segment -Tone -Segment +Tone -Segment -Tone

Materials and Design
Concreteness score is based on Brysbaert, Warriner, and Kuperman (2014). Semantic Scores of English word pairs and their Chinese equivalent pairs were obtained using a 5-point semantic rating study with a different group of 20 Chinese-English bilinguals to make sure that all the critical word pairs are not related in meaning.

ERP Analysis
Previous studies have consistently reported an N400 reduction when English word pairs contained Chinese phonological repetition (Thierry & Wu, 2007;Wu & Thierry, 2010, 2012a). Thus, we had a-priori knowledge of the time-window. Furthermore, we selected electrodes which showed maximal effects in previous studies (Thierry & Wu, 2007;Wu & Thierry, 2010, 2012a. As indicated in Maris and Oostenveld (2007) and discussed in FieldTrip's on-line tutorial, cluster-based permutation tests can be conducted with apriori selected channels and latency to increase the sensitivity. Therefore, for the English experiment, we conducted cluster-based permutation tests on a-priori selected channels and time window.
We also conducted traditional AVOVA analyses with the data of the English experiment.
We first used the mean global field power across all electrodes in all semantically unrelated conditions (+Segment +Tone, +Segment -Tone, -Segment +Tone, -Segment -Tone) to detect the latency of the peak amplitude in the N400 window from 350 ms to 500 ms. The peak latency was 392 ms. Next, the mean amplitude of the time window which extended ±25 ms surrounding the peak (367 ms to 417 ms) was calculated. The mean amplitude of the next 50-ms time window (417 ms to 467 ms) was also calculated. For the Chinese experiment, we tried to identify the critical time-window for the P200 and N400 components using the mean global field power (Lehmann, 1987;Lehmann & Skrandies, 1980;Picton et al., 2000). However, using the mean global field power measured across all electrodes and all unrelated conditions (+Segment +Tone, +Segment -Tone, -Segment +Tone, -Segment -Tone), the peaks for 100-250 ms and 350-500 were detected at 148 ms and 459 ms respectively. The latencies of the peaks were unexpected because the 148-ms peak was relatively early for the P200 component and the 459-ms peak was relatively late for the N400 component. Based on the latencies of these peaks, it is very likely that the P200 component temporally overlapped with an early component and that the N400 component temporally overlapped with a later component.
Although the mean global field power is an unbiased measure of electrical signal amplitude across the scalp, it may fail to distinguish peaks of temporally overlapping components (Dien, 2012). To further investigate the issue of overlapping components in our data, the ERP waveforms were averaged across all unrelated conditions (+Segment +Tone, +Segment -Tone, -Segment +Tone, -Segment -Tone). As can be seen in Figure S7, the P200 component is preceded by the N170 component and the N400 component is followed by the LPC component. Because the patterns in the ERP data were more complex than expected and different from previous studies, the findings of previous studies were not very helpful in guiding our analysis. Therefore, we decided to include the whole time window (200-800 ms) and all electrodes in the cluster-based permutation analysis of the Chinese experiment.