Plastic Mandarin tones: regional identity in prosody

1.1 The city of Changsha and its accents

Changsha is the capital of Hunan province, in south central China. Historically, Changsha has been a regional centre for administration, transportation, commerce, and culture. Nowadays, it is one of the core cities in central China with a total population up to 10 million (Hunan Provincial Bureau of Statistics 2021), covered by extensive train and bus networks, waterways, and air travel.

Changsha Xiang has been the main variety spoken in the city of Changsha, especially in the downtown area. It is largely mutually unintelligible with Standard Mandarin. Other Xiang varieties such as Ningxiang and Liuyang are also spoken in the counties under the jurisdiction of Changsha. The integration of Changsha and two neighbouring cities, Zhuzhou and Xiangtan, into the Greater Changsha Metropolitan Region (also known as the Chang-Zhu-Tan Megapolis) has also brought speakers of different Xiang varieties into Changsha. Changsha is the only city in Hunan Province that has positive net migration rate over the years (Hunan Provincial Bureau of Statistics 2017), and the great majority of migrants are from other regions within the province, thanks to its impressive economic growth in recent years. The surge of migrants in the labour market yields increasing complexity of the local linguistic repertoire, since there are numerous linguistic varieties within Hunan, many of which are mutually unintelligible.

While Standard Mandarin is required and omnipresent in television, films, radio, lectures, and formal meetings, Changsha has been dominating everyday communication and the local opera Huāgǔxì (also known as Flower Drum Opera). Contact between Standard Mandarin and Changsha in this multilingual society has resulted in a spectrum of non-standard varieties blending elements of both languages in varying degrees and manners. It is thus possible to identify two broad categories of non-standard Mandarin produced by Changsha speakers: Hunan-accented Mandarin and Plastic Mandarin. Hunan-accented (or Xiang-accented) Mandarin is characterised by perceptual and production differences in segmental phonology between Xiang vernaculars and Standard Mandarin, while largely retaining the tone patterns of Standard Mandarin. For example, some Changsha speakers are not perceptually sensitive to the distinction between /ʂ/ and /s/ in Mandarin, a non-native contrast in many Xiang vernaculars including Changsha. This is illustrated in two Hunan-accented Mandarin examples [1] shī shēng and [2] dú shū by a Changsha speaker taken from a transcript in Wu (2005: 29). The corresponding Standard Mandarin, Changsha, and Plastic Mandarin pronunciations are provided for comparison. [1] and [2] demonstrate that the Changsha speaker employs either the alveolar fricative [s] or the alveolo-palatal fricative [ɕ], acoustically similar segments available in the vernacular Changsha, as alternatives to the retroflex fricative [ʂ] in Standard Mandarin, while showing no difficulty in producing Standard Mandarin tones. Such proficiency-based variation also includes the realisation of syllable onsets /n/ as [l], /tʂ/ as [ts], / tʂʰ/ as [tsʰ], /x/ as [f], and the syllable coda /ŋ/ as [n], patterns that are commonly observed among adult Xiang learners of Mandarin. These segmental features are often stigmatised or regarded as production errors, while compliance with the Standard Mandarin segmental features is associated with ‘educatedness’ and oral competence in contexts where Mandarin is expected.

[1]

师 生 teacher(s) and student(s)
	shī shēng
Standard Mandarin	[ʂ ʂəŋ]
Changsha	[sz̩ sən]
Hunan-accented Mandarin	[sz̩ sən]
Plastic Mandarin	[ʂ ʂəŋ]

[2]

读 书 read books
	dú shū
Standard Mandarin	[tu ʂu]
Changsha	[təʊ ɕy]
Hunan-accented Mandarin	[tu ɕy]
Plastic Mandarin	[tu ʂu]

In contrast, we can see that in [1] and [2] the segments in Plastic Mandarin align with those in Standard Mandarin, but the tones differ. The phoneme inventory of Plastic Mandarin is largely identical to that of Standard Mandarin (Jing and Niu 2010), though there may be individual variation. The distinctive tonal patterns serve as the primary perceptual and social cues for Plastic Mandarin. Interestingly, prosodic deviations from Standard Mandarin (as in Plastic Mandarin) appear to be considerably less stigmatised than segmental deviations (as in Hunan-accented Mandarin).

1.2 The crystallisation of Plastic Mandarin

One of the earliest mentions of this term Plastic Mandarin in academic work appears in Wu (2005: 22), referring to “a variety that combines the features of the local Changsha dialect with features of Standard Mandarin”. Translated literally from the term 塑料普通话 used by local speakers, it was originally an umbrella term covering various Mandarin-vernacular hybrid varieties, including L2 Mandarin accents spoken by adult Xiang learners of Mandarin. Hence, the boundary between Plastic Mandarin and Hunan-accented Mandarin is not always clearly delineated.

In recent decades, the referent of Plastic Mandarin has become crystallised and increasingly identified as a particular non-standard Mandarin with distinct and consistent tonal patterns spoken by young people. This was validated in Xu et al. (2012)’s study: about 90 % of their 104 survey respondents in Changsha correctly identified the recording of Plastic Mandarin among recordings of four major local varieties, including Standard Mandarin, Plastic Mandarin, Changsha-accented L2 Mandarin, and Changsha; more than half (55.8 %) of them differentiated Plastic Mandarin from Changsha-accented L2 Mandarin. The gap between identification and differentiation is largely attributable to the fact that some participants considered Changsha-accented L2 Mandarin as a variant of Plastic Mandarin. Despite the absence of authoritative definition or standard, most respondents could recognise and were highly aware of Plastic Mandarin. Furthermore, Jing and Niu (2010) observed that the most speakers of Plastic Mandarin were below 30 years old, and those with high academic qualifications were capable of and inclined to code-switching between Standard Mandarin, Plastic Mandarin, and Changsha.

Plastic Mandarin is spreading to broader geographic and social space and to a wider range of age groups, which has come to be linked increasingly to Changsha. Considering that older generations in Changsha do not speak Plastic Mandarin, it was likely formed in the past two or three decades, during a period when Standard Mandarin was extensively promoted through education and media under national language legislation and strong language policy. This push toward linguistic standardisation can be seen in the designation of the third week of every September, since 1999, as the “Putonghua (Standard Mandarin) Promotion Week” across the country, an annual national event jointly guided by nine central government bureaus including the Ministry of Education. Activities and competitions for the annual event are organised through the public school system and by various state and cultural institutions.

1.3 Contact and acquisition: the formation of a new Mandarin dialect

The scholarly discourse on the nature of Plastic Mandarin revolved around two theoretical strands: (1) Plastic Mandarin as a fossilised interlanguage (Jing and Niu 2010; Mao 2009) and (2) Plastic Mandarin as a contact-induced new variety (Fu 2010). The former concept of interlanguage in the field of second language acquisition is specific to individual non-native speakers, while the latter concept in the field of sociolinguistics places the focus on collective linguistic behaviour arising in multilingual communities. Both contact and acquisition are essential factors in the formation of Plastic Mandarin.

Interlanguage refers to the language system that learners develop in the process of acquiring a target language and has a structurally intermediate status between the native language and the target language (Selinker 1972). The accented Mandarin produced by older adults in Changsha tends to exhibit characteristics of interlanguage, i.e. permeability, dynamism, and systematicity. There are some uniform patterns alongside considerable individual variability in the accented Mandarin, as evident in the classification of heavy, medium, and light accents in Fu (2010)’s examination of accented Mandarin produced by 200 native Changsha speakers aged from 16 to 55. The homogeneous linguistic forms deviant from the norm of Standard Mandarin in Plastic Mandarin, nevertheless, should not be analysed as persistent errors or failed approximations as in the fossilised interlanguage account, since most Plastic Mandarin speakers in this study were aware of the mismatch between Plastic Mandarin and Standard Mandarin and capable of producing more standard forms.

The sociolinguistic perspective of contact and new dialect formation is highly relevant to the case of Plastic Mandarin. For the millennial generation, the norms of Plastic Mandarin were not introduced by their parents nor grandparents. The lack of continuity between generations serves as a marker of a new dialect (Kerswill 2020). The contact between Changsha and Standard Mandarin in the city of Changsha may not be as apparent, given that there has not been large-scale migration of Standard Mandarin speakers to Changsha in recent years. The contact accompanied the process of adopting Standard Mandarin under the language policy and was sustained and accelerated in the digital space including mass media and digital communication. Considering Changsha and Standard Mandarin, one might regard Changsha as a case of extended diglossia (Fishman 1967) with bilingualism, where two varieties coexist in the society but are used in different domains: Standard Mandarin with a highly codified system is the High variety used in written form and formal domains and Changsha is the Low variety marginalised in formal education and thriving in ordinary conversations at home and wet market. The compartmentalisation by usage domains of the two varieties in a diglossic situation is often not clear-cut or stable in reality (Sayahi 2020), especially when coupled with non-native competence. Changsha presents a fluid diglossic situation, or diaglossia in Auer (2005)’s term, as the local linguistic landscape is also filled up with intermediate variants.

The rise of homogeneous Plastic Mandarin during the contact, akin to the focusing stage in Trudgill’s three-stage theory of new-dialect formation (Trudgill 2004), complicates the functional distribution of varieties. Plastic Mandarin tends to encroach upon the High variety in contexts such as education and corporate meetings as well as infiltrate the Low variety in some informal contexts. Zhang et al. (2024) have observed an ongoing change in language use in Changsha, identifying vernacular-preserving and vernacular-lost groups of speakers.

1.4 The citation tones of Standard Mandarin and Changsha

Given that tone plays a central role in distinguishing Plastic Mandarin from other varieties, it is essential to first review the citation tones of Standard Mandarin and Changsha, as displayed in Tables 1 and 2, where each tone category is illustrated with at least one [ma] syllable, using both tone letters and numerals (Chao 1930). While the tones of Standard Mandarin are well-established (e.g. Lee and Zee 2003), the description of Changsha tones is based on the seminal work of Firth and Rogers (1937), Beijing University (1995), Li (1998), Wu (2005), and Bao and Chen (2005). Recent work by Wu (2023) suggests slightly different Changsha tone values (e.g. T1 is 34 instead of 33; T2 is 223 instead of 13; T3 is 43 instead of 41; T5 is 31 instead of 21), though the findings are based on data from a single speaker. In the ‘Tone’ column, the Arabic tone numbers are arbitrary but conventional, and their historical tone category is also labelled. The tone categories in Changsha and Standard Mandarin that evolved from the same Middle Chinese category are mostly labelled with the same number to indicate their connection. For instance, Tone 2 in Standard Mandarin and Tone 2 in Changsha are both evolved from the yáng píng category of Middle Chinese.

Table 1:

Citation tones of Standard Mandarin.

Tone	Description		Example
1 yīn píng	˥ 55	high – level	[ma] ‘mother’ 妈 ‘wipe’ 抹
2 yáng píng	35	mid – rising	[ma] ‘hemp’ 麻
3 shǎng	214	low mid – dipping	[ma] ‘horse’ 马
	21	low – falling
4 qù	51	high – falling	[ma] ‘abuse’ 骂 [pʰa] ‘fear’ 怕

Table 2:

Citation tones of Changsha.

Tone	Description		Example
1 yīn píng	33	mid – level	[ma˧] ‘mother’ 妈
2 yáng píng	13	low – rising	[ma] ‘hemp’ 麻
3 yīn shǎng	41	mid high – falling	[ma] ‘horse’ 马
4 yīn qù	45	mid high – rising	[pʰa] ‘fear’ 怕
5 yáng qù	21	low – falling	[ma] ‘abuse’ 骂
6 rù	24	low mid – rising	[ma] ‘wipe’ 抹

In Standard Mandarin (Table 1), Tone 3 has two common variants, excluding the sandhi form: the full form and the reduced form , although the former is the prescriptive form in textbooks. Tone 3 is typically reduced to a simple low fall tone [ 21] without the rising tail in non-utterance final positions (Chao 1968). Changsha has a more complicated tone system than Standard Mandarin, with 6 distinct citation tones. The tonotactic gap in the [ma] set in Table 2 reflects that Changsha evolved from the ‘register distinction’ of qù tone (Norman 1988) in Middle Chinese. In other words, Tone 4 in Changsha usually co-occur with syllables with a voiceless obstruent onset. Therefore, [pʰa] with a voiceless bilabial stop onset is employed instead to illustrate the yīn qù tone (Tone 4) in Changsha. Changsha is not standardised and sometimes the /a/ vowel may be realised in a more retracted and raised manner as [ɐ] or [ɔ] by some speakers for instance.

1.5 Research questions

Tone has been identified as one of the prosodic aspects of language that is highly diffusible through language contact (Ratliff 2015). Plastic Mandarin manifests Mandarin tonal variation and change in a multilingual community.

The primary objective of this study is to establish the tone system of Plastic Mandarin through phonetic-acoustic analysis. A few prior studies, such as Jing and Niu (2010), Xu et al. (2012), Wang (2018), and Shen (2019), have reported the four citation tones of Plastic Mandarin, but these were either based on impressionistic descriptions or fundamental frequency (f0) measurements of data from a single speaker. Building upon these studies, this empirical study examines the prototypical f0 contours of citation tones in Plastic Mandarin employing speech data from a group of high school students, the primary social group associated with this variety and for whom Plastic Mandarin constitutes a predominant part of their everyday linguistic repertoire.

The second objective of this study is to explore the potential mechanisms underlying the formation of Plastic Mandarin tones. During language contact, tonal features from the two source varieties – Standard Mandarin and Changsha – coexist. Adolescents may have selectively adopted certain Changsha features associated with local identity into their Mandarin, contributing to the emergence of a distinct tonal system. In order to attest the hypothesis that Plastic Mandarin has borrowed certain tone patterns from Changsha, one has to at least demonstrate that similarities in pitch contours between Plastic Mandarin and Changsha tones are strongly present. A parallel dataset of Plastic Mandarin and Changsha citation tones, produced by the same group of multilingual participants, serves as a perfectly controlled research ground where we can track both group-level patterns and speaker-specific variation. This dataset enables us to evaluate the potential correspondence between each individual’s Changsha and Plastic Mandarin tone production.

Specifically, the present study addresses the following specific questions:

1. What is the prototypical pitch contour of each tone in Plastic Mandarin?

2. How similar the pitch contours are between the tones of Plastic Mandarin and Changsha?

2.1 Participants

Twenty-one young adults contributed to the elicitation of Plastic Mandarin tones (sixteen females and five males; mean age = 17.24 years, SD = 0.70 years). Their demographic profiles are presented in Table 3. They were all students at the Changsha Nanya Middle School and recruited from pre-existing friendship groups. Changsha Nanya Middle School is a boarding school where students are connected not only through classes and academic work, but also through extracurricular, social, and daily activities. This environment fosters a speech community characterised by dense and multiplex social networks and high frequencies of Plastic Mandarin interaction.

Fourteen participants were born and raised in Changsha; the others were born in nearby cities in Hunan province and moved to Changsha at an early age. On average, the participants had spent 15.71 years in Changsha at the time of elicitation. At home, many boarding school students also speak Changsha or other Xiang vernaculars, particularly with their grandparents; 16 out of 21 participants (76 %) reported always or often doing so. Thirteen of them who grew up in downtown Changsha, reported Changsha as their only Xiang vernacular, and self-rated their proficiency as at least a moderately high (a score of 4 or 5 on a 5-point Likert scale; highlighted in grey in Table 3) contributed to the elicitation of Changsha tones. This subgroup consisted of ten females and three males (mean age = 17.08 years, SD = 0.28 years). No history of speech or hearing disorders was reported. Written informed consent was acquired from subjects prior to recording.

2.2 Materials

The materials for Plastic Mandarin recordings contain four sets of Mandarin monosyllables composed of different segments; within each set these syllables differ only in tone (see Table 4). These sets of Mandarin syllables were selected because (1) their Changsha counterparts (cognates) share a similar or matched segmental composition, facilitating tonal comparison; (2) they are high-frequency words; and (3) their syllable structures are relatively simple – these syllables are either vowel-only or consist of a consonant-monophthong structure, featuring three vowel contexts /i/, /u/, /a/. The majority of them are fully voiced to minimise potential f0 perturbation introduced by voiceless consonants. The corresponding Changsha sets in Table 5 were used for Changsha recordings.

2.3 Procedure

The elicitation of Plastic Mandarin and Changsha was mostly conducted in a sound-attenuated music room in the Changsha Nanya middle school, China. Two additional Plastic Mandarin speakers were recorded in a sound studio in the Phonetics Laboratory, University of Oxford. Each participant had a Rode Lavalier microphone clipped to their collar. The recording was made on one channel at a sampling rate of 44.1 kHz and 16-bit quantization using a Marantz (PMD661 MK.II) digital audio recorder.

Participants were instructed to articulate the monosyllables and repeat each six times, with pauses in between. The instructions and materials were presented on a piece of paper and on a computer screen in written Chinese. Participants were self-paced without interruptions and all utterances of a participant were recorded in one WAV audio file.

The participants were not trained to be aware of the citation tones of Plastic Mandarin since the tones have not been explicitly established. Instead, they had overt knowledge of the four tones of Standard Mandarin due to years of formal learning and training in schools. To prevent participants habitually switching to Standard Mandarin, they were instructed to think of or think aloud words, phrases, or any utterances that contain the target syllable in Plastic Mandarin before the monosyllabic articulation. Prior to the monosyllabic elicitation, participants engaged in other tasks conducted in conversational Plastic Mandarin with their friend (e.g. Xu 2024). The researcher, a speaker of both Plastic Mandarin and Changsha, conducted the sessions in the language corresponding to the elicitation task.

2.4 Acoustic measurement

The recordings were time-aligned to Chinese characters at the syllable level using the Penn Phonetics Lab Forced Aligner (P2FA) for Chinese (Yuan and Liberman 2008, 2015). The output TextGrid files containing the temporal boundaries of each syllable were checked against the corresponding audio files and misalignment errors were manually corrected with the aid of spectrograms and waveforms. The interval of each target syllable, padded with 50 ms silence before and after, were extracted and saved as individual WAV files. Such zero-padding prevented information loss due to windowing in subsequent acoustic analysis.

The f0 contour of the entire voiced portion of each syllable was initially extracted in the first pass, revealing that f0 estimates at the syllable edges were often inaccurate, primarily due to creaky voice associated with speech initiation and termination. As discussed in Xu and Zhang (2024), f0 was extracted over the entire voiced portion of each syllable in citation form, without excluding any segment a priori based on certain phonological assumptions. This study then adopted an intensity-dependent semi-automatic approach to trim the edges of f0 tracks of monosyllables, whereby the intensity was measured along with each f0 point. For monosyllables, the intensity curve typically resembles an inverse U-shape: the beginning and end exhibit a steep rise and a plunge respectively, while the middle portion reaches a plateau. The algorithm filters out f0 measurements in edge intervals of low intensity, by excluding edge f0 estimates whose intensity falls more than one standard deviation below the mean intensity of the token. This process was achieved by a customised Python script that employs the Parselmouth library (Jadoul et al. 2018) to implement Praat measurements (Boersma 1993). As demonstrated in Figure 1, f0 measurements outside the specified interval were excluded from further analysis. Empirically tested, this method was conducive to excluding inaccurate f0 estimates at edges as well as attenuating effects attributed to a neighbouring voiceless consonant. The plateau intervals were visually checked and revised if it excluded a significant portion of a token, which would result in information loss. Within each f0 track interval, f0 estimates were obtained at 21 equidistant points, corresponding to intervals of every 5 % of the duration. f0 measurements were further checked and manually corrected for artifacts such as octave errors. Utterances with a creaky portion in the middle of the vowel were discarded due to missing f0 measurements. In total, there are 2,618 f0 contours of Plastic Mandarin tones and 1,848 f0 contours of Changsha tones.

Figure 1:

Trimmed f0 track (black solid) and intensity (pink dashed) of the syllable 雾 wu in Plastic Mandarin. f0 track segments outside the marked interval were not considered.

2.5 Statistical analysis

4.1 Changsha tones

Figure 4 visualises the distribution of normalised f0 data at every 5 % intervals across the voicing portion of syllables grouped by tone category in Changsha. It shows that the tonal realisations among participants are largely internally consistent. For Tones 1, 2, 3, and 6, the distributions are unimodal, with the ridge representing the most likely f0 contour. The f0 values of Tone 4 are primarily distributed above 0, although there are some much thinner bands below 0 in Figure 4(b), representing deviations. Upon check, two additional elicited variants of the Tone 4 morpheme [uː] 务 ‘task’ were identified. However, these variants were not consistent across the group, with each one produced by only a single participant. The f0 distribution of Tone 5 bifurcates and there are two different f0 patterns: one resembles the high rising Tone 4 and the other largely occupies a lower f0 region below 0. The bimodality exhibited in their production of Tone 5 is mostly attributed to the alternative pronunciation of the Tone 5 morpheme [uː] 雾 ‘fog’. Both pronunciations are acceptable, and this is related to a more ubiquitous phenomenon among Chinese varieties that some characters have differentiated literary and vernacular readings (Ho 2015). The doublets of readings result from language contact and reflect lexical stratification under the influence of the standard language of education and literary instruction at various periods (Norman 1979), especially for non-Mandarin regions such as Changsha. The two patterns in Figure 4(b) align with Li (1998)’s observations: the vernacular reading of many Tone 5 syllables is a low falling tone (the majority of Tone 5, historical yáng qù tone), while their literary reading is a high rising tone (the majority of Tone 4, historical yīn qù tone).

Figure 4:

Distribution of f0 (semitone) of citation tones in Changsha across time. (a) Superimposed f0 distributions of all six tones of Changsha, for comparison; (b) f0 distribution for each tone separately, for clarity.

The variations in Changsha tones shown in Figure 4 suggest that Changsha is not as strictly prescribed as Standard Mandarin and that Changsha may be undergoing tonal change. In many modern Mandarin varieties, the historical qù Tone has been maintained as a single tone category (Mandarin Tone 4). In contrast, it was split into two registers that became modern Tones 4 and 5 in Changsha. Peng (2006) thus predicted a trend toward the merging Tone 4 and Tone 5 in Changsha, reflecting the levelling effect of Changsha becoming more similar to Mandarin.

Figure 5 illustrates the prototypical contour of the six tones, excluding the less common variants. Changsha tones produced by these young Changsha participants can be represented as 33, 23, 43, 45, 21, 24. It is worth noting that a 5-level system may not precisely depict the exact f0 height but provides a simplified and stylized representation that preserves the phonological contrasts. Although both Tone 3 and Tone 4 begin with the numeral “4”, the actual starting pitch for Tone 4 is much higher than that of Tone 3. Tone 3 is represented as beginning with “4” to maintain a clear contrast with Tone 1. A comparison between the current Changsha data and tone values documented in the literature in Table 2 suggests that Tone 3 in Changsha may have shifted from /41/ to /43/.

Figure 5:

Observed data (error bars indicate ± standard error) and growth curve model fits (lines) for f0 contour of each tone in Changsha.

4.2 Cross-varietal comparison

Visually inspecting the prototypical tone patterns in Figures 3 and 5, several similar tone pairs of the same (historical) tone category can be identified. For instance, Plastic Mandarin and Changsha have comparable Tone 4s, Tone 2s, and Tone 1s. In addition, Plastic Mandarin Tone 3 and Changsha Tone 5 have a similar contour.

4.2.1 The high rising Tone 4 and level Tone 1

Figure 6 illustrates the model predictions for the two groups of Tone 4 contours with 95 % pointwise confidence intervals. Overall, Changsha Tone 4 is higher than Plastic Mandarin Tone 4. The parametric term of language variety from the GAMM is statistically different from zero (t = −3.45, p < 0.001), indicating that they have different f0 heights. By removing such effect of language variety on overall f0 contour height through centring the predicted f0 (see Figure 6b), we can find that the contour shape between Tone 4 of the two varieties does not statistically significant differ between the two varieties over the course of the tone.

Figure 6:

Estimated f0 trajectories of high rising Tone 4s and comparison. (a) Estimated f0 trajectories (GAMM fits) of Plastic Mandarin and Changsha Tone 4, with 95 % pointwise confidence intervals. (b) Same trajectories, centred to facilitate shape comparison.

There is some variation in the individual realisation of Tone 4 in Plastic Mandarin. Figure 7 shows f0 contours of Tone 4 in Plastic Mandarin and Changsha by each speaker in an overlaying manner. Speaker d104 produced mostly high falling Tone 4 in Plastic Mandarin, distinct from the corresponding high rising Tone 4 in Changsha. For speakers d103, d105, and d106, some of their Plastic Mandarin Tone 4 contours also feature a falling tail. In contrast, the remaining speakers, in particular speakers d102, d110 and d112, produced Plastic Mandarin and Changsha Tone 4 contours that were highly similar and often overlapping.

Figure 7:

f0 contours of Tone 4 in Plastic Mandarin (purple) and Changsha (orange) by speaker.

For Tone 1, we are mainly interested in f0 heights since these are level tones. Tone 1s in both varieties sit mid-range in the tonal space. Plastic Mandarin Tone 1 is generally a bit higher than the Changsha counterpart, illustrated by the model predictions of the GAMM in Figure 8. The estimated height difference based on the parametric term in the GAMM is 0.68 semitone (t = 3.00, p = 0.0027), which is statistically significant.

Figure 8:

Estimated f0 trajectories (GAMM fits) of Plastic Mandarin and Changsha Tone 1, with 95 % pointwise confidence intervals (duration: 0.27 s; vowel: /i/).

4.2.2 The low rising and low falling tones

Figure 9 shows the comparison of the two groups of low rising tones (Plastic Mandarin and Changsha Tone 2) and low falling tones (Plastic Mandarin Tone 3 and Changsha Tone 5). Plastic Mandarin Tone 2 is closely resembles its Changsha counterpart throughout the course of the tone and the estimated curve difference between them in Figure 9(a) is non-significant for the first 89 % of the tone duration. Plastic Mandarin Tone 3 and Changsha Tone 5 tend to reach similar pitch heights in the last 37 % of their tone contours.

Figure 9:

Estimated f0 trajectories of low rising and low falling tones and comparison. (a) Estimated f0 trajectories (GAMM fits) of Plastic Mandarin and Changsha Tone 2, with 95 % pointwise confidence intervals. (b) Estimated difference between the two groups in (a) with the associated 95 % pointwise confidence interval. (c) Estimated f0 trajectories (GAMM fits) of Plastic Mandarin Tone 3 and Changsha Tone 5, with 95 % pointwise confidence intervals. (d) Estimated difference between the two groups in (c) with the associated 95 % pointwise confidence interval. The highlighted red area in (b) and (d) indicates where the confidence interval excludes zero.

5.1 The phonetic bias account: recurrent tone contour changes

In his analysis of Bangkok Thai tone changes, Pittayaporn (2018) suggested that most phonetic variation is not random but biased and that tone variants arising from the production and perception of lexical tones tended to be crystallised and phonologised. For instance, f0 peaks tend to be delayed rather than early (e.g. Hao et al. 2017; Xu and Wang 2001), introducing a bias toward the rightward shift of tone peaks. This tendency may be attributed to common articulatory mechanisms such as carryover effects and truncation of tone production due to time pressure (Yang and Xu 2019).

Table 7 summarises the Standard Mandarin-Plastic Mandarin tonal variation by tone categories, by comparing the tone patterns of Plastic Mandarin and Standard Mandarin. When tone patterns are conceptualised as probabilistic approximation of high or low pitch target(s) that anchored to specific time points relative to the onset of a syllable, a varied tone pattern can be generated when the pitch target moves in the temporal domain (leftward or rightward) and/or the pitch domain (upward or downward). These tone variation patterns summarised in Table 7 are in fact among the most common and recurrent patterns of tonal change observed cross-linguistically, as identified in Yang and Xu (2019)’s review of 52 studies on tone change in East and Southeast Asia. For example, that the mid-rising tone becomes the low-rising tone (Tone 2) is also observed in Dong’gan, Gaotian Zhuang, Tai Phuan, and Lalo; the contour reduction of the dipping tone, low fall-rise to low falling tone (Tone 3), is also observed in Taiwan Mandarin, Ma’ya, Libo, Taipusi. Interestingly, the rightward sliding and flattening of Tone 4 is also reminiscent of the changes observed in Thai Tone 4 between the early 20th century and 1962. Many unrelated languages have parallel development of certain types of tone patterns. The fact that Standard–Plastic Mandarin tonal variation largely aligns with the crosslinguistic tendencies strengthens the phonetically-based account of tonal contour changes.

5.2 The borrowing account: contour similarity between Plastic Mandarin and Changsha tones

The comparison of the tone patterns of Plastic Mandarin and Changsha leads to an alternative promising account of the mechanism of tone change, the contact-induced borrowing. There is a striking similarity in the shape of the f0 patterns between the Plastic Mandarin Tones 1, 2, and 4 and their corresponding Changsha tones. This suggests that Plastic Mandarin may have borrowed certain tone patterns from the corresponding Changsha tones, in which the historical tone category continue to play a mediating role. The average f0 heights can differ slightly between these similar pairs of tones, which may imply that speakers borrow the tone shapes and adjust the tone heights when creating a new tonal system instead of direct copying. The high rising tone tends to be higher in Changsha than in Plastic Mandarin, which may be attributed to Changsha’s more complex tone system, potentially requiring a larger tonal space for distinction.

In Standard Mandarin, syllables with Tone 4, the high falling tone, tend to be judged the most prominent among other tones in both prosodically strong and weak positions (Deng 2010). Tone 4 is also the most frequent tone in Standard Mandarin (Wu et al. 2020). The change in the Tone 4 tone pattern is, therefore, highly perceptually salient, and it puts the borrowed Changsha Tone 4 with a distinct high rising pattern in the spotlight. Furthermore, the high rising tone pattern can serve as a signature of Xiang varieties. The majority of Xiang vernaculars retain a tone category evolved from the high (yīn) register of qù tone (Bao and Chen 2005), typically manifesting as a high rising or mid-to-high rising tone, as seen in varieties such as Changsha, Xiangtan, Xiangxiang, Yueyang, Shaoyang, Shuangfeng, Loudi, and others. Rather rarely we find such high rising qù tone in other most spoken Chinese varieties. For instance, there is no such high rising tone pattern in the tone inventories of Guangzhou Cantonese, Xiamen Min, Suzhou Wu, and Meixian Hakka.

The adaptation of Changsha tone patterns as new canonical tone patterns has allowed Plastic Mandarin to express a shared local and cultural identity and social solidarity. While specialised lexicon associated with local food, festivals, activities, and so on is often the site or marker of cultural identity in a local Mandarin variety, the case of Plastic Mandarin shows that identity can be marked prosodically through pitch.

5.3 The hybrid account: Mandarin lexical tone variation

While the phonetic bias account does not fully account for the crystallisation of the Plastic Mandarin tone patterns, the contact-induced borrowing account does not explain why Tone 3 in Plastic Mandarin does not borrow from Changsha Tone 3. It is likely that other biases or constraints also exert some influence in the process. These accounts may not be mutually exclusive. In addition to the phonetic biases, Pittayaporn (2018) identified three systemic biases in tonal contour changes: contour maximisation, contour accentuation, and avoidance of similar tones, which favoured variants of dynamic tones with a greater and more dramatic f0 excursion, and tone variants dispersed in phonetic space. Such biases were supported by perception experiments. For example, dynamic pitch excursions improve speech recognition in noise, and speech with dynamic pitch is consistently better perceived than monotone speech (Shen and Souza 2018). The Tone 3 in Changsha [] lies in the mid-high space of the pitch range, which would not be optimal in maximising perceptual contrast if adopted by Plastic Mandarin, given another two mid-range tones, Tone 1 [] and Tone 2 [], and a high-range Tone 4 [] in Plastic Mandarin. The low falling Tone 3 [] in Plastic Mandarin, instead, magnify the perceptual distance between Tone 3 and other tones.

The phonetic and systemic biases, as well as the contact-induced borrowing effects, altogether can offer a coherent explanation for the synchronic Plastic Mandarin variation. In addition, I propose that social biases play a crucial role in the formation of Plastic Mandarin, with social identity formation guiding this change. Figure 10 schematises the potential phonologisation of the high rising Tone 4 contour, following Pittayaporn’s (2018) approach. A pool of structured and random variants motivated by phonetic biases, by borrowing, or by chance is first generated. The variants illustrated in Figure 10 are not exhaustive but are theoretically plausible. Most of them were chosen based on common cross-linguistic patterns identified in Yang and Xu (2019, Table 8). Specifically, a high falling tone /51/ may transform into a falling tone with a rising offset [53] (e.g. Old Beijing) , a high level tone [55] (e.g. Dalian, Lanzhou, Bangkok Thai) , a mid rise-fall tone [341] (e.g. Chongqing) , or mid rising tone [45] (e.g. Penang Hokkien) . Some of these variants are reflected in individual Tone 4 productions (see participants d103 – d106 in Figure 7 for examples). All of these variants are characterised by the clockwise tone shift pattern (Yang and Xu 2019; Zhu 2018), a well-attested cross-linguistic tendency in tone change observed across East and Southeast Asia. The clockwise pattern, essentially a temporal realignment of pitch movements, can be considered as a phonetic bias that influences the directionality of tonal variation. We may also consider [41], with a lowering onset, as a random phonetically similar variant, and [45] as a borrowed variant. Among all these variants, [45] was preferred and prevailed as the primary variant, likely because it closely resembles or embodies the Changsha Tone 4 (yīn qù), the ‘signature’ Xiang tone, which, in turn, serves as a socially marked variant associated with regional identity. This high pitch tone is uniquely distinct from the other Plastic Mandarin tones, which primarily occupy the low-to-mid pitch range, thereby maximising perceptual contrast and revealing a systemic bias towards avoiding similar tones. The bottom part of Figure 10 shows that [45] is selected as the new canonical tone value in Plastic Mandarin, reflecting selectional biases that involve both systemic and social factors.

Figure 10:

Schematic illustration of the phonologisation of the high rising contour of Tone 4 in Plastic Mandarin. Variants [53], [55], [341], and [45] reflect the clockwise bias in tonal variation. [45] may also be a borrowed variant and [41] may be a phonetically similar random variant.

Table 8:

Summary of Plastic Mandarin lexical tones.

Tone	Phonetic description	IPA representation
1	a mid level tone	[] 33
2	a low to mid rising tone	[] 23
3	a low falling tone	[] 21
4	a high rising tone	[] 45