Asymmetries in the perception of Mandarin tones: Evidence from mismatch negativity

Abstract

The mismatch negativity (MMN) component of the electroencephalogram reflects phonological asymmetries: specifically, greater MMN is elicited by infrequent (deviant) tokens when the features of the frequent (standard) token are phonologically underspecified than when those features are phonologically fully specified. Such asymmetries have not been studied in non-Indo-European languages or in suprasegmental contrasts, which might not be represented in the same way as segments. Therefore, the present study investigated the neural representation of Mandarin contour tones. Mandarin third tone (T3) is realized as second tone (T2) in certain contexts, whereas it has no alternation relationship with fourth tone (T4). The present study examined MMNs elicited by T3~T2, T3~T4, and T2~T4 contrasts (in both directions). Asymmetrical MMN effects were elicited in both T3~T2 and T3~T4 contrasts: MMN was smallest when the standard was T3. The results suggest that T3 has a less specified neural representation than both T2, which it alternates with, and T4, which it does not. This may be due to the featural representation of T3 (in standard Mandarin it is a Low tone, which is typologically less marked) or due to its behavior (i.e., it may be the case that features which alternate always trigger reduced MMN effects).

Neural signals to violations of abstract rules using speech-like stimuli

“…Syllables in the 3rd position (XXY deviant) differ in their consonant voicing and place of articulation, and in their vowel height (Mioni, 1993;Kramer, 2009;Paoli, 2016). While it should be noted that whether all these phonological features have a neural representation is on itself an open debate (Hestvik and Durvasula, 2016;Politzer-Ahles et al, 2016;Schluter et al, 2016Schluter et al, , 2017, in the case of our stimuli set, the number of phonological features that change for each deviant condition is the same.…”

“…While previous work in the literature has shown that the generation of predictions can serve word processing, phonemes in these experiments were either omitted (Bendixen et al, 2014), or replaced either by other phonemes (Cornell et al, 2013;Politzer-Ahles et al, 2016;Schluter et al, 2017) or by a non-linguistic sound (Kashino, 2006;Groppe et al, 2010). Because of this, changes in low level auditory features might have contributed to the recorded signals.…”

Encoding category-level and context-specific phonological information at different stages: An EEG study of Mandarin third-tone sandhi word production

“…The significant effect of the T3 prime using the half-third sandhi variant supports the presence of a shared underlying representation for all the T3 tonal variants (Chien et al, 2016(Chien et al, , 2021, and cannot be explained by the surface representation view assuming the only storage of the surface form. However, the nature of the underlying representation is still debated, which has been argued to be the canonical form /214/ (as in the canonical representation view mentioned in Zhou and Marslen-Wilson, 1997), an abstract Low tone (e.g., Cheng, 1968), or an underspecified representation (e.g., Politzer-Ahles et al, 2016). Based on recent priming studies, it is probable that the underlying tone of T3 is represented as an abstract Low tone (Chien et al, 2020(Chien et al, , 2021Meng et al, 2021), but other studies have reported inconsistent findings (Politzer-Ahles et al, 2016;Zhou and Marslen-Wilson, 1997).…”

“…However, the nature of the underlying representation is still debated, which has been argued to be the canonical form /214/ (as in the canonical representation view mentioned in Zhou and Marslen-Wilson, 1997), an abstract Low tone (e.g., Cheng, 1968), or an underspecified representation (e.g., Politzer-Ahles et al, 2016). Based on recent priming studies, it is probable that the underlying tone of T3 is represented as an abstract Low tone (Chien et al, 2020(Chien et al, , 2021Meng et al, 2021), but other studies have reported inconsistent findings (Politzer-Ahles et al, 2016;Zhou and Marslen-Wilson, 1997). As the current study was not designed to test the nature of the underlying representation, further research is still required.…”

“…Zhang et al, 2022; hereafter, we will label T3 as the underlying form and T2 as the surface form for convenience). Although researchers disputed over the nature of the underlying representation (see discussion in Chien et al, 2016Chien et al, , 2020Chien et al, , 2021Meng et al, 2020;Politzer-Ahles et al, 2016;Zhou and Marslen-Wilson, 1997) and computation (e.g., Politzer-Ahles and Zhang, in press), it has been suggested that the generation of T2-like sandhi variant could be operated as an online tonal substitution rule that changes the underlying T3 to a T2 (Chang and Kuo, 2020). Under this account, the T3 sandhi production involves a serial encoding process, with the underlying T3 firstly specified and the surface T2-like sandhi tone generated at a later time point before phonetic spellout or articulation (e.g., J.…”