Asymmetric Mismatch Negativity and Underlying Representations of Lexical Tones in Cantonese

Introduction The mismatch negativity (MMN) has been used extensively to examine the sensitivity of individuals to speech sound contrasts. Nevertheless, it is relatively rare to find reports contrasting sounds with each sound serving as deviant in one condition and standard in another in a passive oddball paradigm. A few studies that have done so have obtained asymmetry in MMN amplitude and in some cases latency for some pairs of sounds but not others (German vowels: Eulitz & Lahiri, 2004; Japanese vowels: Ikeda et al., 2002; Hindi consonants: Shafer et al., 2004). Eulitz and Lahiri (2004) explained MMN asymmetry in terms of presence or absence of conflicting phonological features between the surface form representing the deviant stimulus and the underlying form with underspecified feature accessed by the standard stimulus. This study employed a passive oddball paradigm to examine whether asymmetric MMN occurred in two tonal contrasts, T1 [+Upper register, h(igh) tone] vs. T6 [-Upper, l(ow)] and T4 [-Upper, hl] vs. T6, with T6 serving as deviant and standard stimuli in different blocks. Method Twenty native Cantonese speakers who could discriminate and produce all six distinctive tones in the language with 100% accuracy from Law et al. (2013) participated in this study. Electrocephalography (EEG) was recorded from 64 Ag/AgCl electrode sites arranged according to the International 10-20 system. Statistical differences between true and dummy difference waves were assessed by a non-parametric cluster-based permutation test. Paired t-tests were conducted on each tonal contrast comparing MMN amplitudes and latencies of T6 as deviant with standard at FCz. Results and Discussion Table 1 shows statistically reliable clusters in T6/T1 (deviant/standard) and T1/T6 conditions, and significant but unreliable clusters in T6/T4 and T4/T6. There was a tendency of greater difference in MMN amplitude between the two standard/deviant allocations in T1 vs. T6 than T4 vs. T6. This was verified by paired t-tests revealing higher mean MMN amplitude for T6/T1 but comparable amplitudes in T4 vs. T6 at FCz. No significant differences in peak MMN latencies were found. Our results are compatible with a proposal of underspecification of the register feature for low register tones underlyingly, i.e. T4 [ , hl], T5 [ , lh], and T6 [ , h]. This hypothesis makes the prediction of asymmetric MMN for T2 [+Upper, lh] vs. T5. In particular, enhanced MMN is expected for T5/T2, which can be confirmed in future study. References Eulitz, C., & A. Lahiri (2004). Neurobiological evidence for abstract phonological representations in the mental lexicon during speech recognition. Journal of Cognitive Neuroscience, 16(4), 577-583. Ikeda, K., Hayashi, A., Hashimoto, S., Otomo, K., & Kanno, A. (2002). Asymmetrical mismatch negativity in humans as determined by phonetic but not physical difference. Neuroscience Letters, 321, 133-136. Law, S.-P., Fung, R., & Kung, C. (2013). An ERP study of good production vis-à-vis poor perception of tones in Cantonese: Implications for top-down speech processing. PLoS ONE, 8, e54396. Shafer, V. L., Schwartz, R. G., & Kurtzberg, D. (2004). Language-specific memory traces of consonants in the brain. Cognitive Brain Research, 18, 242-254.link_to_OA_fulltex