1.1 Introduction
Chinese Korean refers to the varieties of Korean spoken by descendants of immigrants who came to China between the mid nineteenth century and the end of the Second World War (Jin, Reference Jin2008). As with many heritage languages, Chinese Korean has multiple sources of linguistic influence. First, the different varieties of Chinese Korean can be traced back to the different ancestral dialects spoken by the original immigrants who came from different regions in Korea. In addition, Chinese Korean has been influenced by the standard varieties of Korean in North and South Korea. The North Korean standard (i.e., Munhwaeo) was used as the model for Chinese Korean standardization (i.e., Joseoneo) in the mid twentieth century (Tai, Reference Tai, Zhou and Sun2004), but there has been an increase in exposure to Seoul Korean through media and travel since the establishment of diplomatic ties between China and South Korea in 1992. At the same time, the influence of Mandarin, the local contact language, has increased in recent years; the majority of Chinese Korean speakers are Korean–Mandarin bilinguals (Jin, Reference Jin2008), and there has been a shift from Korean to Mandarin as the dominant language in some ethnic Korean communities (Han, Reference Han2011, Reference Han2014; Schertz et al., Reference Schertz, Kang and Han2017).
The goal of the current study is to document the realization of two front rounded vowels /y/ and /ø/ in two Chinese Korean dialects spoken in the Chinese cities of Hunchun and Dandong, stemming from two different ancestral dialects (i.e., Northern Hamgyeong and Northern Pyeongan; see Figure 1.1), to explore the influence of contact languages (i.e., Mandarin and Seoul Korean) and ancestral dialects on these heritage languages. These vowels have diphthongized in many dialects of Korean, including Seoul Korean. Impressionistic descriptions of vowels in Hunchun and Dandong suggest diphthongal realizations, but quantitative acoustic studies are not available, as these vowels are omitted in most acoustic studies of the vowel systems of these dialects (Kim, Reference Kim2009; Li, Reference Li2016; Liu, Reference Liu2021; Xu, Reference Xu2019). In this study, we examine the influence of the following on the realization of /y/ and /ø/: (1) Mandarin, the local contact language, (2) Seoul Korean, the supralocal norm, and (3) the North Korean dialect serving as the contemporary descendant of the ancestral dialect.Footnote 1
Figure 1.1 Map of Korea and Northern China.
1.2 Background
In this section, we provide a brief introduction to Korean vowels, previous studies on front rounded vowels in Chinese Korean, and background on Hunchun and Dandong Korean.
1.2.1 Korean Vowels
Table 1.1 shows the ten monophthongal vowels of Standard Korean. Most dialects of Korean do not retain the ten-vowel system of monophthongs due to various mergers and shifts. The vowels /y/ and /ø/ have been diphthongized to [wi] and [we] in many dialects of Korean and Seoul Korean, in particular (Choi, Reference Choi2002; Shin, Reference Shin, Brown and Yeon2015; Shin et al., Reference Shin, Kiaer and Cha2012; Sohn, Reference Sohn1999). In addition, the contrast between /e/ and /ɛ/ has been neutralized to /e/ in Seoul Korean. The South Korean Standard (i.e., Pyojuneo) pronunciation guidelines specifically indicate that /y/ and /ø/, although listed as monophthongs, may be pronounced as diphthongs, thus acknowledging the prevalence of diphthongal pronunciations (Ministry of Culture Sports and Tourism, 2017).
Table 1.1 Monophthongs of Standard Korean – South Korean (i.e., Pyojuneo), North Korean (i.e., Munhwaeo), and Chinese Korean (i.e., Joseoneo).
| Front | Back | |||
|---|---|---|---|---|
| Unrounded | Rounded | Unrounded | Rounded | |
| high | i | y | ɨ | u |
| mid | e | ø | ʌ | o |
| low | ɛ | a | ||
Standard North Korean (i.e., Munhwaeo) pronunciation guidelines, on the other hand, explicitly state that these vowels should be pronounced as monophthongs (North Korea National Language Assessment Committee, 2010). However, studies on contemporary North Korean vowels indicate age- and dialect-based variation, with most reporting some form of diphthongal realizations of the vowels or a loss of rounding (Jung, Reference Jung2013; Kang, Reference Kang1997; Kwak, Reference Kwak2003, Reference Kwak2019; Lee et al., Reference Lee, Park and Jeong2009). Available instrumental studies on monophthongal vowels exclude the front rounded vowels from acoustic analyses on account of the fact that they are diphthongized or pronounced as front unrounded vowels for both Pyeongan (So, Reference So2010) and Yukjin and Hamgyeong dialects (Chung, Reference Chung2011; Kang, Reference Kang1997). As will be discussed in Section 1.5, we have very limited knowledge of the status of front rounded vowels in the North Korean dialects from the early twentieth century, and both Pyeongan and Hamgyeong dialects likely had variable realizations of these vowels.
While Standard Chinese Korean (i.e., Joseoneo) pronunciation guidelines similarly dictate that the front rounded vowels should be pronounced as monophthongs (China’s Joseon Language Assessment Committee, 2016), phonological descriptions often do not include the front rounded vowels in the inventory of monophthongs (Han, Reference Han2006; Joen, Reference Joen1996; Kang, Reference Kang2021; Li, Reference Li2016),Footnote 2 and instrumental studies on monophthongs do not include these vowels (Kim, Reference Kim2009; Liu, Reference Liu2021; Xu, Reference Xu2019), with the exception of Jin (Reference Jin2012).
1.2.2 Previous Literature on Heritage Language Vowels and Chinese Korean Front Rounded Vowels
Vowel production has been intensely researched in the literature on heritage language phonetics and phonology in recent years (Chang, Reference Chang, Montrul and Polinsky2021; Polinsky, Reference Polinsky2018). Studies have found that the production of heritage language vowels is influenced by those of the dominant languages. For example, heritage speakers of Western Armenian living in the US produced the Armenian vowels /i, ɛ, a/ more similarly to their English counterparts compared to homeland speakers or late L2 learners (Godson, Reference Godson2003, Reference Godson2004). Furthermore, heritage Korean speakers produced English and Korean vowels with acoustic differences from English and Korean monolinguals, respectively (Baker & Trofimovich, Reference Baker and Trofimovich2005).
For Chinese Korean front rounded vowels, Jin (Reference Jin2012) conducted a large-scale instrumental study on Korean /y/ produced by Chinese Korean speakers living in Shenyang, home to around 80,000 Chinese Korean speakers. Shenyang is well known for its successful maintenance of Korean ethnicity, and the ancestral dialect of the Shenyang variety is Northern Pyeongan. Based on a visual inspection of F2 and F3 formant movements and auditory categorization, she found that the monophthongal realization [y] was the dominant variant (81%), attributing this to the influence of the Mandarin monophthong /y/, because /ø/, which has no counterpart in Mandarin, was realized as diphthongal. She also found that the diphthongal variant [yi] was attested around 11 percent of the time and was produced more by female speakers, who were assumed to be more status-conscious and likely to favor prestigious forms (i.e., Seoul Korean /y/ [wi] in this case) than male speakers. Similarly, Liu (Reference Liu2021) conjectured that the use of diphthongal pronunciations of /y/ and /ø/ in younger Yanbian Korean speakers developed from recent contact with South Korea, although no evidence of age-based variation was provided.
1.2.3 Present Study
We build on Jin (Reference Jin2012) and examine /y/ and /ø/ in two other Chinese Korean communities, Hunchun and Dandong. The most commonly spoken Korean varieties in these cities stem from two distinct North Korean dialects, Northern Hamgyeong and Northern Pyeongan, respectively. Hunchun is part of the Yanbian Korean Autonomous Prefecture, where Korean is an official language alongside Mandarin, and a third of the population is ethnic Korean. Hunchun proper has a population of around 200,000, of whom 36 percent are ethnic Koreans (China Data Center, 2006). On the other hand, in Dandong, which is a large city of around 800,000 inhabitants, Korean is a minority language spoken within a smaller ethnic Korean community of around 20,000 individuals (Cui, Reference Cui2011). These sociolinguistic contexts make the two communities good case studies to probe the role of ancestral dialects and contact languages on heritage language variation. The different status of Korean and Mandarin in these two Chinese Korean communities is also reflected in the self-reported proficiency and language usage data of our participants, which will be discussed in Section 1.5. Korean is the dominant language for older speakers in both communities, as well as younger speakers in Hunchun, while Mandarin is overtaking Korean as the dominant language in younger Dandong speakers.
1.3 Methodology
1.3.1 Participants
A breakdown of Korean speaker participants by dialect, gender, and age is provided in Table 1.2. Dandong and Hunchun are our main target heritage speaker groups, and Northern Hamgyeong and Seoul serve as homeland comparison groups. Northern Hamgyeong is the ancestral dialect of Hunchun. Note that the number of male speakers in the Northern Hamgyeong group is much smaller than that of the other groups. This is because 72 percent of North Korean refugees in South Korea are female (as of December 2018, according to the Ministry of Unification) and many male speakers work and have limited availability. We include the male data for completeness, but any gender-based analysis for Northern Hamgyeong speakers will be interpreted with appropriate caution. For socioeconomic and geographical reasons, the majority of North Korean refugees originate from the Northern Hamgyeong region. Consequently, we lack a comparison group from the Northern Pyeongan homeland.
Table 1.2 Distribution of Korean speakers.
Dandong (n=66) | Hunchun (n=61) | Northern Hamgyeong (n=38) | Seoul (n=57) | |
|---|---|---|---|---|
| Older (year of birth <1970) | 20F,Footnote a 14M | 17F, 13M | 10F, 1M | 15F, 17M |
| Younger (year of birth ≥1970) | 15F, 17M | 15F, 16M | 22F, 5M | 11F, 14M |
a. F = female, M = male.
In addition, we collected data from sixteen native monolingual Mandarin speakers (eight from Dandong and eight from Hunchun) to compare their production of front rounded /y/ against the heritage Korean speakers’ production of /y/ in both Korean and, for a subset of 51 Hunchun and 57 Dandong speakers, Mandarin. Dandong and Hunchun speakers also completed a background questionnaire about their language proficiency in Korean and Mandarin and their exposure to the Korean language and its associated culture. Dandong and Hunchun data were collected in the summer of 2015, Seoul data in the summer of 2011, and Northern Hamgyeong data from recent North Korean refugees residing in Seoul in the winter of 2016–2017 (Kang & Yun, Reference Kang and Yun2018; Yun & Kang, Reference Yun and Kang2019). We note that the North Korean speakers differ widely in their degree of accommodation to Seoul Korean, and they are not a homogeneous group linguistically. Their length of residence in Korea ranged from three to 164 months, with a mean duration of 57 months at the time of data collection (Yun & Kang, Reference Yun and Kang2019). Nevertheless, their speech, especially their vowel production as a group (Kang & Yun, Reference Kang and Yun2018), differs significantly from Seoul Korean and is indicative of contemporary Northern Hamgyeong Korean.
Table 1.3 Distribution of Mandarin speakers.
| Dandong (n=8) | Hunchun (n=8) | |
|---|---|---|
| Older (year of birth <1970) | 1F, 2M | 1F, 2M |
| Younger (year of birth ≥1970) | 3F, 2M | 2F, 3M |
1.3.2 Stimuli
Stimuli were produced as part of a wordlist and included one target word for each vowel: /ysin/ (‘prestige’) for /y/, and for /ø/, /øsonca/ (‘grandson (daughter’s son)’) for northern dialects and /øsaŋ/ (‘credit purchase’) for Seoul speakers.Footnote 3 The Northern Hamgyeong speakers produced only one repetition, while all other speakers produced two repetitions. These words were chosen to represent the two front rounded vowels in word-initial position with a following coronal consonant. For normalization purposes, words containing each of the other monophthongs in word-initial position were also included and analyzed (see Table 1.4). A total of 3,946 tokens of produced words were analyzed after omitting 114 tokens that were skipped, mispronounced, or produced with excessive background noise. Mandarin speakers and a subset of Chinese Korean speakers also produced a Mandarin target word, yu2 /ý/ (‘fish’), as part of a larger list of Mandarin words. Two repetitions were included for most speakers, but there is only one repetition for some due to omissions, errors, or noise, for a total of 243 Mandarin tokens.
Table 1.4 List of stimulus words.
| /ysin/ | 위신 | ‘prestige’ |
| /øsonca/ (northern dialects) | 외손자 | ‘grandson (daughter’s son)’ |
| /øsaŋ/ (Seoul) | 외상 | ‘credit purchase’ |
| /atɨl/ | 아들 | ‘son’ |
| /ɛpi/ | 애비 | ‘father’ |
| /ʌlɨn/ | 어른 | ‘adult’ |
| /emi/ | 에미 | ‘mother’ |
| /olɛ/ | 오래 | ‘long time’ |
| /uli/ | 우리 | ‘we’ |
| /ɨn-i/ | 은이 | ‘silver-nom’ |
| /ilɨm/ | 이름 | ‘name’ |
The participants were fitted with a high-quality lapel condenser microphone (Audio-Technica AT831B), and their speech was recorded in a mono setting at 44,100 hertz (Hz) on a Zoom H4n recorder. The material was presented using PsychoPy (Peirce, Reference Peirce2007) on a laptop or a tablet, and the participants read the material at their own pace. The wordlist was randomized for each participant.
1.3.3 Acoustic Measurements
The first three formants, F1, F2, and F3, were measured at ten equally spaced time points across the duration of each vowel using Praat (Boersma & Weenink, Reference Boersma and Weenink2021). F1 is a major correlate of tongue body height, and all other things being equal, a higher F1 indicates a lower tongue body. F2 correlates with tongue body frontness; that is, all other things being equal, a higher F2 indicates a more fronted tongue body. F3 is associated with lip rounding, and all other things being equal, a lower F3 indicates more lip rounding (Jin, Reference Jin2012; Ladefoged, Reference Ladefoged1996). For consistency and replicability, we used Praat’s FormantPath function to identify the optimal formant track instead of manually setting or correcting speaker- or vowel-specific formant measurement settings. To remove obvious tracking errors, we identified outliers for each formant of each vowel for each speaker subgroup of the same gender, age group, and dialect. Any formant measurement away from its relevant mean by 2 standard deviations or more (i.e., 4.7% of the Korean vowel measurements and 4.1% of the Mandarin vowel measurements) was removed.
To allow for a direct comparison of formant values across speakers of different ages and genders, Lobanov transformations (i.e., a speaker- and formant-specific z-score transformation; Lobanov, Reference Lobanov1971) were conducted. These normalized values were converted back to the Hz scale using the overall mean and standard deviation of all speakers’ vowel measurements for each vowel. This conversion was done for ease of interpretation of formant values and did not affect the statistical analysis.
To quantify the degree of change across the course of vowels, we calculated the difference between the mean of the first three and the last three measurement points for each formant. This method was chosen, rather than the difference between individual time points, to reduce data omission due to missing measurements and minimize the influence of a single errant measurement. These distances were compared across speaker groups using a linear regression model in R (R Development Core Team, 2021).
1.4 Results
Figure 1.2 summarizes the distribution of ten vowels in the four dialect groups and their formant trajectories in the F2–F1 space (panel a) and the F2–F3 space (panel b).
Figure 1.2 Average formant trajectories of monophthongs by dialect in the F2–F1 (panel a) and F2–F3 space (panel b). Vowel symbols represent starting points, while arrows represent the full trajectory of the formants of each vowel. The target vowels, /y/ and /ø/, are shown in black.
Before we examine the front rounded vowels, we note a couple of crucial differences between Seoul Korean and the northern dialects. First, Seoul Korean does not contrast /e/ and /ɛ/, while the contrast is intact in the northern dialects. The vowel /o/ is raised to a position close to /u/, with /ʌ/ staying low and back in Seoul Korean, while in the northern dialects, /ʌ/ is produced higher than /o/. In addition, /ɨ/ is fronted to a central position in Seoul Korean, with younger speakers producing it more fronted than older speakers (age difference not shown), while /ɨ/ stays in a back vowel position, overlapping with /u/ for the northern dialects. These are all expected based on previous studies (Kang, Reference Kang and Kong2016; Kang & Kong, Reference Kang and Kong2016; Kang & Yun, Reference Kang and Yun2018) and provide an assurance that our formant measurement method successfully captures dialect-level vowel differences even without any manual error correction.
Turning to the focus of our study, Seoul /y/ and /ø/ travel the farthest distance, especially along the F2 and F3 dimensions, compared to the other dialects, indicating a large degree of diphthongization. Among the northern dialects, Hunchun and Northern Hamgyeong vowels show larger formant movements than what we observe in Dandong vowels. The four panels of Figure 1.3 provide examples of formant tracks of diphthongal Seoul pronunciations and monophthongal Dandong pronunciations.
Figure 1.3 Sample spectrograms of /y/ and /ø/, as produced by Seoul and Dandong speakers. Panel (a) /y/ (Seoul, female, born in 1990); (b) /y/ (Dandong, female, born in 1956); (c) /ø/ (Seoul, female, born in 1960); (d) /ø/ (Dandong, female, born in 1995).
Figure 1.4 shows average formant tracks by dialect. For both vowels, F3 increases across the vowel, indicating a decrease in lip rounding, while the extent of F2 increase across the vowel reflects a change from a more posterior to a more anterior articulation. For /ø/, F1 increases, which corresponds to the tongue body lowering across the vowel.
Figure 1.4 Average formant tracks of /y/ and /ø/ by dialect. The lines show locally estimated scatterplot smoothing, and gray shading indicates 95% confidence intervals.
We can observe a generally steeper rise of F2 and F3 in Seoul vowels compared to those of other varieties, while Dandong vowels show the least steep formant movements. This is in line with the observation that in Seoul, these vowels are realized as diphthongs, changing from a central or back rounded to front unrounded articulation; that is, [wi]~[ʉ̯i] for /y/ and [we]~[ʉ̯e] for /ø/. The F2 onset of Seoul vowels is often much higher than expected for a back rounded articulation, and hence the Seoul onglide may be more properly transcribed as [ʉ̯] rather than [w] in many cases (see Jun & Bishop, Reference Jun and Bishop2009 for a similar observation for Seoul /y/). Dandong has minimal F1 and F2 movements, which is an indication of little tongue body movement, vertical or horizontal, while the rise of F3 for some speakers indicates a change in lip rounding; that is, [y]~[yj]~[ɥi] for /y/ and [ø]~[øe̯]~[ø̯e] for /ø/. Hunchun pronunciations are similar to those of Dandong in showing relatively more stable average F1 and F2 values than those of Seoul. Northern Hamgyeong /y/ is intermediate between the Chinese dialects and Seoul, while Northern Hamgyeong /ø/ patterns with the Chinese dialects.
To quantify and compare the degree of diphthongization statistically, we calculated the difference between the mean formant values of the first three and the last three measurements of each vowel for each formant (i.e., FormantRise, Hz). The boxplots in Figure 1.5 provide a summary, showing a greater rise in F2 and F3 in Seoul than in the northern dialects across corresponding age and gender groups.
Figure 1.5 Formant rises for F1, F2, and F3 for /y/ and /ø/ by age, gender, and dialect. Boxplots show the distribution of vowel tokens.
The scatterplots in Figure 1.6 present another view of the data, with the x-axis representing F2 rises and the y-axis showing F3 rises. We see that the ellipses for Seoul Korean are in the upper right quadrant in both cases, while those of the other dialects are positioned more centrally, indicating less steep F2 and F3 movements.
Figure 1.6 Scatterplot of F2 and F3 rises by dialect group. Each dot represents a vowel token, and the ellipses represent 95% confidence intervals.
For our statistical analysis, a linear mixed-effects model was built for each formant for each vowel. The response variable was FormantRise (normalized Hz), and the predictors were Dialect (i.e., Dandong, Hunchun, and Seoul), Gender (i.e., female and male), Age (i.e., older and younger), and all interactions. A by-participant random intercept was included. Sum coding was used for all predictors such that the model intercept estimated overall formant rises across all speaker groups. For each model, overall effects of the predictors were tested by the Wald chi-square test, using the Anova() function of the car package (Fox et al., Reference Fox and Weisberg2013), and significant effects were probed further by the Wald chi-square test, using the testInteractions() function of the phia package with Holm correction for p-value adjustment (De Rosario-Martinez, Reference De Rosario-Martinez2015). Table 1.5 summarizes the significance of fixed-effect predictors, and all significant effects based on post-hoc tests are summarized in Table 1.6.
Table 1.5 Summary of p-values corresponding to the effects of dialect, gender, and age of speaker on the diphthongization of /y/ and /ø/ (not including Northern Hamgyeong).
| /y/ | /ø/ | |||||
|---|---|---|---|---|---|---|
| F1 | F2 | F3 | F1 | F2 | F3 | |
| Dialect | 0.540 | <0.001 *** | <0.001 *** | <0.001 *** | <0.001 *** | 0.003 ** |
| Gender | 0.143 | 0.004 ** | 0.745 | <0.001 *** | 0.079 † | 0.261 |
| Age | 0.951 | 0.134 | 0.491 | 0.045 * | 0.007 ** | 0.890 |
| Dialect * Gender | 0.029 * | 0.504 | 0.965 | 0.493 | 0.015 * | 0.136 |
| Dialect * Age | 0.747 | 0.235 | 0.588 | <0.001 *** | 0.148 | 0.318 |
| Gender * Age | 0.041 * | 0.437 | 0.889 | 0.601 | 0.947 | 0.227 |
| Dialect * Gender * Age | 0.124 | 0.778 | 0.344 | 0.927 | 0.246 | 0.717 |
*** p < 0.001, ** p < 0.01, * p < 0.05, † p < 0.1; gray shading means p > 0.1.
Table 1.6 Summary of significant effects based on post-hoc tests.
| /y/ | /ø/ | |||||
|---|---|---|---|---|---|---|
| F1 | F2 | F3 | F1 | F2 | F3 | |
| DialectFootnote a | None | D<H<S | D<H<S | D~H<S (Y) | D~H<S | D~H<S |
| D<NH<S | NH<S | NH<S (Y) | NH<S | |||
| (OF,YM,YF) | ||||||
| Gender | F<M | M<F | None | F<M | M<F (H) | None |
| (S,O) | ||||||
| Age | None | None | None | Y<O (D,H) | O<Y | None |
| O<Y (S) | ||||||
a. The parentheses include the subgroups for which the comparison holds (D=Dandong, H=Hunchun, NH=Northern Hamgyeong, S=Seoul, O = Older, Y = Younger, F = Female, M = Male).
Note that the Northern Hamgyeong group was not included in the main statistical models due to the previously mentioned gender imbalance in this group; that is, the number of male speakers in the Northern Hamgyeong group is small (i.e., one older and five younger). Another set of models was run with the Northern Hamgyeong data to compare this dialect to others, but age or gender effects internal to Northern Hamgyeong were not probed.
To anticipate the overall findings, both vowels are more diphthongal in Seoul Korean than in both Chinese Korean dialects and Northern Hamgyeong Korean. Amongst non-Seoul dialects, /y/ was more monophthongal in Dandong than Hunchun and Northern Hamgyeong, but /ø/ did not differ significantly.
For /y/, F1 changed only minimally overall (estimated overall mean rise = 17.7 Hz), and the main effect of Dialect was not significant. A significant interaction of Gender with Dialect and Age was found. Since the tongue body should remain high throughout the vowel /y/ and F1 relatively stable under either monophthongal or diphthongal realization, we treat these interactions as spurious and do not interpret them.
F2 and F3 rose substantially for all dialects and the dialects differ in the extent of the rise, as shown by the significant main effect of Dialect. F2 changes the most for Seoul (estimated mean, here and henceforth = 425.5), least for Dandong (140.9), and intermediate for Hunchun (250.6). A significant main effect of Gender was found for F2, with female speakers showing a steeper rise than male speakers in the normalized Hz scale across all three dialects. This is consistent with the female speakers favoring the newer variant in the case of Seoul and the more prestigious Seoul-like variant in Dandong and Hunchun (Jin, Reference Jin2008). For F3, Seoul showed a significantly greater rise (646.0) than Hunchun (581.2), which in turn had a steeper rise than Dandong (414.9). A test including the Northern Hamgyeong data showed that Northern Hamgyeong patterned with Hunchun, showing a steeper rise than Dandong and a less steep rise than Seoul for F2 and F3, although the difference between Dandong and Northern Hamgyeong did not reach significance for F3. These results show that /y/ is most diphthongal in Seoul, least in Dandong, and intermediate in Hunchun and Northern Hamgyeong.
For the mid front rounded vowel /ø/, F1 rises significantly overall (86.7), which means that the tongue body lowers through the vowel. There was a significant main effect of Dialect, Gender, and Age, and a significant interaction of Dialect and Age. Follow-up tests show that F1 rises significantly more for Seoul (96.1) than Dandong (63.6) and Hunchun (77.4) for younger groups only. The difference between Dandong and Hunchun was not significant. For the Chinese dialects, male speakers and older speakers had a steeper rise than females and younger speakers, which is somewhat unexpected if diphthongization of /ø/ is an innovation, which tends to be led by younger and female speakers. For Seoul speakers, younger speakers had a steeper rise than older speakers, which is in line with a change toward diphthongization.
As for F2 for /ø/, there was a significant effect of Dialect and its interaction with Gender. Follow-up tests showed a consistent pattern across both genders of a greater rise for Seoul speakers than for the other two dialects (Seoul: 297; Hunchun: −33.5; Dandong: −10.3). We found a significant effect of AGE across dialects and a significant effect of Gender for Hunchun, where younger speakers and female speakers produced more Seoul-like (i.e., higher) values than older speakers and male speakers, respectively, which is consistent with a change toward diphthongization led by younger speakers who are female. While F3 significantly rose across vowels for all three dialects, here again, Seoul showed a steeper rise (329.6) than both Dandong (240.6) and Hunchun (270), with the difference between the Chinese dialects not reaching significance.
Northern Hamgyeong did not differ significantly from the Chinese dialects along any of the three formants for /ø/. With respect to Seoul, Northern Hamgyeong speakers had a flatter F1 trajectory than the Seoul speakers for the younger group and a flatter F2 than Seoul for all speaker groups except for the older male group (with a single Northern Hamgyeong speaker).
1.5 Discussion and Conclusions
We found clear differences between Seoul and the northern dialects for both vowels, with more diphthongal pronunciations in Seoul. We also found that the northern dialects differed in their realizations of /y/, with more monophthongal productions in Dandong than in Hunchun or Northern Hamgyeong, while the realizations of /ø/ did not differ across the northern dialects. Jin (Reference Jin2012) suggested that the dominant monophthongal realization of /y/ in Shenyang Korean is due to the influence of Mandarin. As we will see below, our findings support the idea that Mandarin phonology plays a role in the variation of /y/ pronunciation across dialects. Unlike Jin (Reference Jin2012), we found that /ø/, which has no Mandarin counterpart, is also more monophthongal in the northern dialects than in Seoul. Interestingly, this difference cannot be straightforwardly attributed to contact with Mandarin since it is found not only in Chinese Korean dialects but also in Northern Hamgyeong.Footnote 4 Instead, the more monophthongal realizations of these front vowels appears to be a general property associated with the northern dialects. At the same time, we did see evidence of the role of Mandarin in the different degrees of diphthongization in Dandong versus Hunchun. Specifically, Dandong, which is a community where Mandarin is more dominant, showed more monophthongal pronunciations of /y/ than Hunchun; on the other hand, the two communities did not differ in their realizations of /ø/. Since /y/, but not /ø/, has a monophthongal Mandarin counterpart, this discrepancy is plausibly attributable to increased contact with Mandarin in Dandong.
Prior to examining the Mandarin contact effect in more detail, we explore three alternative possibilities that could account for the difference between /y/ in Dandong and Hunchun: (1) different ancestral dialects, (2) different degrees of contact with Seoul Korean, and (3) differences tied to local Mandarin /y/. The first possibility is that there is a difference in the ancestral dialect from which the heritage dialects are derived. As we only have data from contemporary Northern Hamgyeong Korean, the descendant of the ancestral dialect of Hunchun, and no comparable data from Northern Pyeongan, the ancestral dialect of Dandong, we cannot rule out the possibility that a comparable difference exists between the North Korean ancestral dialects or their contemporary descendants. However, there is no indication of any such difference between contemporary Hamgyeong and Pyeongan Korean (Han, Reference Han2006; Li, Reference Li2016; So, Reference So2010).
If anything, Hamgyeong Korean likely had front rounded monophthongs in the early twentieth century, but Pyeongan Korean did not. What we know about the realization of front rounded vowels in ancestral dialects in the late nineteenth to the early twentieth century is based on Korean textbooks (Ross, Reference Ross1877, based on Northern Pyeongan; Orthodox Missionary Society, 1904, based on Yukjin) and fieldwork descriptions (Ogura, Reference Ogura1927, Reference Ogura1944). Similar to contemporary descriptions, these earlier works point to instability and variation of these vowels, but one proposal (Ogura, Reference Ogura1927, Reference Ogura1944, as cited in Kwak, Reference Kwak2003 and Jung, Reference Jung2013) includes the two front rounded vowels in the monophthongal inventory of Hamgyeong but not Pyeongan. If this contrast persisted in our Chinese heritage varieties, we would expect a more diphthongal realization of the vowels in question in Dandong than in Hunchun, which runs contrary to our findings.
The second possibility is that different /y/ realizations are due to different degrees of contact with homeland Korean varieties, and Seoul Korean in particular. Figure 1.7 summarizes self-reported data on the consumption of Korean language media (1–5, with 1 indicating “no consumption” and 5 indicating “everyday consumption”) by our participants, as well as the number of months they have spent visiting South Korea. As is apparent in panel (a) of Figure 1.7, Hunchun speakers consume significantly more Chinese Korean media than those in Dandong, indicating a high level of local Korean language dominance and exposure (two-sample t-test: t = −8.7072, df = 128, p < 0.001).Footnote 5 The two communities did not differ, however, in terms of their exposure to South Korean or their North Korean ancestral varieties. Chinese Koreans consume practically no North Korean media (t = 0.7984, df = 128, p = 0.4261) and both communities report an equally high level of South Korean media consumption (t = −1.3784, df = 129, p = 0.1705), which is largely dominated by Seoul Korean. As can be seen in panel (b) of Figure 1.7, the two communities did not differ significantly in the average duration of visits to South Korea (t = −0.2264, df = 132, p = 0.8212).
Figure 1.7 Self-reported language exposure data by dialect and age. Panel (a) shows consumption of Korean language media (1–5 scale), and panel (b) shows duration of visits to South Korea (months).
In other words, the community-level difference in /y/ pronunciation between Dandong and Hunchun cannot be attributed to exposure to South Korean dialects, and in particular to Seoul Korean.Footnote 6 This is also consistent with our finding of no community-level difference in /ø/ diphthongization; that is, if Seoul Korean influence were the cause of different levels of /y/ diphthongization in Dandong versus Hunchun, with all other things being equal, we would expect a similar difference for /ø/, which is more diphthongized in Seoul Korean. Our results contrast with those of Jin (Reference Jin2012), who attributed the diphthongal variant of [yi] in Shenyang Korean to the influence of Seoul Korean as a supralocal prestige form.
The third possibility is that there may be different realizations of Mandarin /y/ in Dandong and Hunchun. To examine this hypothesis, we also collected Mandarin speech data from local native monolingual Mandarin speakers as well as heritage Korean speakers. If the difference in the diphthongization of Korean /y/ between the two communities shown above were to stem from a difference in Mandarin /y/, we would expect to see more monophthongal /y/ in the Mandarin spoken in Dandong than in that of Hunchun. Panel (a) of Figure 1.8 shows average formant tracks of the word yu2 /ý/ (‘fish’), as produced by sixteen native monolingual Mandarin speakers (eight from Dandong and eight from Hunchun), and panel (b) of Figure 1.8 shows formant tracks associated with productions of the same Mandarin word by a subset of our Chinese Korean speakers (57 from Dandong and 51 from Hunchun). The formants are shown in raw frequencies (Hz), without normalization. All three formants do not rise and remain stable across vowels for both Mandarin speaker groups. The heritage Korean groups show more movement, especially the older speakers and the Hunchun speakers.
Figure 1.8 Average formant tracks of Mandarin /y/ by (a) Mandarin and (b) heritage Korean speakers in Dandong and Hunchun. The lines show locally estimated scatterplot smoothing, and gray shading indicates 95% confidence intervals.
Linear mixed-effects models with Dialect as the predictor and Speaker as a random effect showed that Hunchun Korean speakers show a significantly larger F2 and F3 FormantRise than speakers of the other three groups, while the other three groups did not differ from one another. Since local native Mandarin productions did not differ in degree of diphthongization across the two communities, the difference between Hunchun and Dandong speakers’ Korean /y/ production cannot be attributed to the difference in their local Mandarin /y/. The two cities differed, however, in how heritage Korean speakers produced Mandarin /y/; that is, Dandong speakers produced a more monophthongal Mandarin /y/ than Hunchun speakers, mirroring the same community-based difference found in the production of Korean /y/, as reported above. In other words, Dandong speakers’ more monophthongal productions of /y/ in both languages align more closely with the Mandarin norm, while Hunchun speakers’ productions are more diphthongal, aligning more closely with their pronunciation of Korean vowels. We can connect this pattern to the overall difference in the status of Mandarin and Korean in the two communities. Korean is one of the official languages of Yanbian, where Hunchun is located, while in Dandong, Korean is a minority language in a large city, and Mandarin is more present and increasingly dominant.
This different status of Korean and Mandarin is also reflected in the self-reported proficiency and language usage data that we collected. Panel (a) of Figure 1.9 summarizes participants’ proficiency rating (1–5) in Korean and Mandarin. Older speakers are more dominant in Korean, but younger speakers show high proficiency in Mandarin. Younger Hunchun speakers retain high proficiency in Korean, but younger speakers of Dandong Korean show signs of decline in heritage language maintenance. Similarly, panel (b) of Figure 1.9, which presents self-reported percentages of the time that speakers use Korean and Mandarin at home and in other contexts, shows that Korean is the dominant language of use in Hunchun, while in Dandong, Mandarin is overtaking Korean among younger speakers. This provides evidence of how phonetic norms and language change can be driven not only by the presence of contact, but also by level of engagement with another community language.
Figure 1.9 Self-reported language background by dialect, gender, and year of birth.
To conclude, our study examined realizations of front rounded vowels in two communities of heritage speakers of Chinese Korean, and acoustically and statistically compared them to those of homeland Korean speakers from two dialects and native monolingual Mandarin speakers from relevant local communities, in order to examine the potential influences of language dominance and contact on dialectal variation. We found that the two communities differed in their production of /y/ but not /ø/, the latter of which has no Mandarin monophthongal counterpart and therefore does not encounter resistance to change towards diphthongization. By taking into account productions of multiple vowels, speaker populations, and languages, along with demographic information about our speakers, we were able to triangulate the source of this community-level difference to the relative dominance of Mandarin in our two communities of interest.
