Mismatch responses to lexical tone, initial consonant, and vowel in Mandarin-speaking preschoolers

Abstract

The present study investigates how age, phonological saliency, and deviance size affect the presence of mismatch negativity (MMN) and positive mismatch response (P-MMR). This work measured the auditory mismatch responses to Mandarin lexical tones, initial consonants, and vowels in 4- to 6-year-old preschoolers using the multiple-deviant oddball paradigm. The data showed the coexistence of MMN and P-MMR in the same age group when responding to the three types of syllabic features in Mandarin. The transition from a predominantly positive response to a predominantly negative response supported the multiple MMN mechanisms. Congruent with the phonological saliency hypothesis and the phonetic acquisition order of Mandarin in behavioral studies, for the compulsory elements of Mandarin syllables, lexical tones, and vowels, the larger deviants elicited adult-like MMNs, whereas the smaller deviants elicited P-MMRs. The optional elements of the Mandarin syllables, the initial consonant, only elicited P-MMR in preschoolers. These findings suggest that MMN and P-MMR index different functional characteristics and may provide information on when and how children's speech perception becomes automatic at different developmental stages.

Introduction

Humans achieve efficient speech communication based on precise mapping between sound and meaning. An efficient listener segments continuous speech into elementary bricks (such as phonemes, syllables, and pitch) that can combine in various ways to convey different meanings to decipher information. Although speech production does not become significant in the first year of life, studies have demonstrated that infants are able to discriminate many phonetic contrasts before 6–8 months of age and begin to show increased discrimination sensitivity to native phonemes with a decline in perceptual sensitivity to non-native phonemes by 12 months (Kuhl et al., 2006, Kuhl et al., 1992). Most important, studies have demonstrated that the early phonetic perception of infants with a family risk of dyslexia significantly differs from that of infants without such a risk and a strong relationship exists between the speech perception of early infancy and later language abilities (Kuhl et al., 2005, Richardson et al., 2003, Tsao et al., 2004).

The behavioral measurements for speech perception, such as discrimination or identification tasks, often require one's overt responses and motivation. Researchers have recently shown an increased interest in using mismatch negativity (MMN), a component of auditory event-related potentials (ERPs), as a relatively objective measure to reflect the brain's ability to discriminate speech elements. Näätänen, Gaillard, and Mäntysalo (1978) first described MMN as a frontocentral distributed negativity peaking between 100 and 250 ms from the onset of a stimulus when the average ERP to a frequent (standard) stimulus is subtracted from that obtained from a rare (deviant) stimulus (Näätänen et al., 1978). The MMN amplitude increases, whereas peak latency decreases, as the discriminability between the standard and the deviant increases (Näätänen, Paavilainen, Rinne, & Alho, 2007). MMN can be elicited even when the participant does not attend to the stimuli (for example, while they are reading a book or watching a silent movie). It suggests that MMN involves a mechanism of automatic and pre-attentive auditory change detection and thus may serve as an excellent tool for assessing auditory discrimination, particularly for infants and children with limited attention or motivation. Several studies have demonstrated that children with or without inherited risk of reading difficulties showed different MMN characteristics to certain speech features (Leppänen et al., 2002, Lyytinen et al., 2001, Lyytinen et al., 2003, Schulte-Korne et al., 2001, van Leeuwen et al., 2006, van Leeuwen et al., 2007). Most importantly, the brain responses to speech sounds successfully predict their later language development (Guttorm et al., 2005, Lyytinen et al., 2004, Molfese, 2000). Thus, MMN may provide a new diagnostic method for early identification of developmental language and literacy disorders.

Although MMN is well established in adults, the developmental trajectories of MMN in infants and young children remain unclear. Negative mismatch responses have been observed in newborns and infants. However, such mismatch negativities usually persist for a longer time interval or occur in a later time window than those typically seen in adult data (Alho et al., 1990, Cheour-Luhtanen et al., 1995, Cheour et al., 1997, Cheour et al., 2002, Martynova et al., 2003, Morr et al., 2002). For example, Alho et al. (1990) reported an adult-like MMN in sleeping newborns while listening to pure tone changes (1000 Hz standard vs. 1200 Hz deviant). The mean peak latencies were seen at 296 ms at Fz and 270 ms at Cz. Cheour and colleagues found that the adult-like MMN peaked at approximately 380–400 ms in newborns and infants when responding to changes in pure tone and vowel (Cheour et al., 1998, Cheour et al., 1999). Other studies reported that the mismatch responses in infants are often positive rather than negative between 200 and 450 msec (Dehaene-Lambertz, 2000, Friederici et al., 2002, Friedrich et al., 2004, He et al., 2008, Leppänen et al., 2004b, Leppänen and Lyytinen, 1997, Morr et al., 2002, Shafer et al., 2011). For example, Leppänen & Lyytinen (1997) observed a positive mismatch response (P-MMR) in response to a pure tone change between 250 and 350 ms in newborns (Leppänen & Lyytinen, 1997). Friederici et al. (2002) examined the mismatch responses of 2-month-old infants to syllables varying in vowel duration (short/ba/ vs. long/ba:/) and found that the P-MMR peaked at approximately 400 ms, particularly for the long syllable (Friederici et al., 2002).

There are several hypotheses to account for why and when the positive or negative mismatch responses could be observed in early infancy. The first account is mainly based on the physiological characteristics and the neural maturation of the infants. Trainor, Samuel, Desjardins, and Sonnadara (2001) and Trainor et al. (2003) identified P-MMR in 2-month-old infants, but MMN in 6-month-old infants in detecting temporal changes (Trainor et al., 2001, Trainor et al., 2003). He, Hotson, and Trainor (2007) observed a similar developmental shift from the discriminative positive slow wave to the adult-like MMN by recording the brain responses to infrequent pitch changes of piano tones in infants from 2 to 4 months old. Their data showed an increase in the left lateralized positive slow wave in 2- and 3-month-old infants, compared to a faster adult-like MMN in 3- to 4-month-old infants. Both slow P-MMR and fast adult-like MMN coexist in 3-month-old infants, suggesting a possible developmental transition of mismatch responses from a slow positive wave to an MMN-like fast negativity (He et al., 2007, He et al., 2008, He et al., 2009). In summary, the adult-like MMN response gains in prominence while P-MMR diminishes as infants grow. These observations suggest that neural maturation may play a role in the developmental changes of mismatch responses.

Recently, a few studies reported the existence of P-MMR in preschoolers and older children (Ahmmed et al., 2008, Maurer et al., 2003a, Maurer et al., 2003b, Shafer et al., 2000), and suggest that the developmental shift from P-MMR to MMN is not a specific phenomenon for infants. These studies argue that certain stimulus-related factors, such as the short inter-stimulus interval and smaller deviance magnitude that decrease the discriminability between standard and deviant, might be a prerequisite for the observed positive mismatch response. For example, Maurer et al. (2003a) examined the mismatch responses in adults and in 6- to 7-year-old children using an oddball sequence with smaller frequency, phoneme deviances, and shorter intervals, compared to most previous studies. A typical frontocentral MMN was evident in adults, whereas children showed a consistent frontal P-MMR. The data from source localization indicated that both P-MMR in children and MMN in adults originated from the superior temporal plane, although with opposite polarity. Furthermore, the data showed that P-MMR in children increased with the degree of phoneme deviance (but not with the degree of tone deviance). They thus claimed that the P-MMR observed in children must possess the same functional aspect as MMN in adults, which reflects additional or increased (rather than decreased) neural activation to deviants compared to standards. Maurer et al. (2003b) applied the same paradigm to 6- to 7-year-old children with or without familial dyslexia risk and found that children at risk had a larger P-MMR to phoneme deviance recorded at mid-frontal electrode sites than the control group. Ahmmed et al. (2008) examined the degree of frequency separation (standard: 1000 Hz, deviants, 2%: 1020 Hz, 5%: 1050 Hz, and 10%: 1100 Hz) between pure tones required to generate MMN in 7- to 9-year-old children with or without specific language impairment (SLI). Both control and SLI groups showed P-MMR for 2% contrast at 400 ms of inter-stimuli interval (ISI). For the control group, the P-MMR changed to MMN with a change in tone contrast from 2% to 5% at 400 ms ISI. However, children with SLI required a frequency separation of more than 10% to show MMN as strongly as was observed in the control group. These findings suggest that both stimulus-related and biological criteria are needed for P-MMR to switch to MMN (Ahmmed et al., 2008). However, researchers disagree as to whether P-MMR is a precursor of MMN (Kushnerenko et al., 2002, Leppänen et al., 2004b).

The present study aims to investigate how age, phonological saliency and deviance size of various speech features affect the presence of P-MMR by tracing the developmental trajectories of mismatch responses to Mandarin lexical tones, initial consonants, and vowels in 4- to 6-year-old preschoolers. Mandarin is a tonal language with a relatively simple syllable structure. There are four possible syllable structures in Mandarin: V, CV, VC, and CVC. A Mandarin syllable possesses four possible elements: tone, syllable-initial consonant (onset), vowel, and syllable-final consonants. Tone and vowel are compulsory units for Mandarin syllables, whereas initial and final consonants are optional. Studies in speech errors made by Mandarin-speaking children have suggested that these four elements are acquired with different developmental sequences. In general, lexical tones first, followed by final consonants and vowels; and finally, initial consonants. Hua and Dodd (2000) proposed a phonological saliency hypothesis to account for phonetic acquisition order of Putonghau (Mandarin Chinese), in which the following factors determine the phonological saliency of the possible elements in a syllable (Hua & Dodd, 2000). First, the status of a component in the syllable structure, particularly whether it is compulsory or optional (a compulsory component is more salient than an optional one). Second, the capacity of a component in differentiating lexical meanings of a syllable; a component is more salient if it is more capable of differentiating lexical meanings. Third, the number of permissible choices with a component in the syllable structure; the more permissible choices it has, the less salient it is. However, phonological saliency is language specific. The saliency of a particular phonological feature is determined by its role within the phonological system of a particular language, not by its reference to other languages. For Mandarin, tone has the highest saliency because it is compulsory for every syllable and has only four permissible choices, high level (T1), high rising (T2), low dipping (T3), and high falling (T4). Changes in lexical tones will change the lexical meanings, such as T1: yi1 “clothes”; T2: yi2 “aunt”; T3: yi3 “chair”; and T4: yi4 “easy”. The vowel is also a compulsory syllabic component. However, its saliency is lower than the lexical tone, due to the relatively large number of options it has (there are 22 vowels in Mandarin, including monophthongs, diphthongs, and triphthongs). The initial consonant has the lowest saliency because its presence in the syllable structure is optional and it has 21 permissible choices.

Studies addressing the mismatch responses to speech features of Mandarin syllables are scarce, and most focus on lexical tone. A larger MMN was found for the T1/T3 contrast compared to the T2/T3 contrast in native Chinese speakers, but not in native English speakers (Chandrasekaran, Krishnan, & Gandour, 2007). This accords with findings from the multidimensional scaling (MDS) analysis with paired comparison judgments of tonal dissimilarity for the perceptual dimensions of lexical tones (Gandour & Harshman, 1978). Pitch contour for T2 is acoustically more similar to T3 than to T1. Data from tonal discrimination and identification confirmed that T2 and T3 are acoustically similar and are more often confused with one another, compared to other tonal pairs. Luo et al. (2006) reported that lexical tone contrast evoked a stronger MMN than initial consonant contrast, particularly in the right hemispheres (Luo et al., 2006). Xi, Zhang, Shu, Zhang, and Li (2010) used the MMN paradigm to investigate the categorical perception of Mandarin lexical tone in native Chinese adults and showed that both within- and between-category deviants elicited MMNs. Congruent with Luo et al. (2006), the MMN record from F4, an electrode positioned at the right frontal site, was marginally larger than that record from F3, an electrode positioned at the left frontal site. However, the categorical effect was only found at F3 in the left frontal site, in which the across-category deviant elicited a larger MMN than the within-category deviant (Xi et al., 2010). Zhang et al. (2012) applied the set stimuli in children with and without dyslexia. They found that the across-category deviant elicited larger mismatch negativity than the within-category deviant recorded from F3 in the left hemisphere in age-matched controls, but not for dyslexic children (Zhang, et al., 2012). Meanwhile, it seems that both within- and between-category deviants elicited positive mismatch responses around 200 ms, although this was not included in their data analysis. In short, the laterality characteristic of MMN to lexical tone is still unclear. Furthermore, it is worth investigating when the adult-like MMNs in response to various phonological features of Mandarin syllables, such as lexical tones, vowels, and consonants, would appear in Mandarin speaking preschoolers.

To the best of our knowledge, this is the first MMN study to investigate the developmental trajectories of mismatch responses to lexical tone, initial consonant, and vowels in preschoolers from 4 to 6 years old. To examine how the deviance size affects the transition from positive to negative mismatch responses, we adapted the multiple-deviant paradigm, which contains two deviants (large deviant and small deviant) and one standard. We predicted that, the presence of P-MMR and MMN would depend on language experience and phonological saliency. It is more likely that P-MMRs would be observed in younger children, especially when responding to less salient phonological units, such as the initial consonant, and to contrasts that are difficult to discriminate (such as the small deviant). On the other hand, it is more likely that adult-like MMN would be found in older children because they may have developed robust representations to support an automatic, pre-attentive discrimination.