Accent as Cultural Artifact: Identity and Power in Taiwan’s Traditional Chinese

Rachel Koh
Vanderbilt University

Abstract

In Lawrence Abu Hamdan’s “The Freedom of Speech Itself,” an asylum seeker’s accent functions as a biography of migration, a proof of nationality, and a test of eligibility for promised safety. Many of these asylum seekers are undocumented, so their accent becomes their passport proving the truth of their identity. Immigration officers test the linguistic features of their voice against that of a speaker from their alleged country of origin by analyzing an audio tape of their voice from an interview. This expectation is deeply flawed—because asylum seekers are often minorities fleeing persecution, it is unlikely that they would speak the “standard language” of the country that they are fleeing—but the asylum seeker must ensure that their voice matches this expectation regardless or risk deportation.

This is a good example of how identity and power become entangled in an accent. Australian writer A. A. Phillips coined the term “cultural cringe” in 1950 to describe an individual’s disavowal of their own culture as a result of internalized shame. We see cultural cringe at play in an immigrant’s desire to eliminate their accent through expensive accent reduction services (Ramjattan 47) or to whitewash the foreignness of their identity entirely (Mohabir). Cultural cringe has guided language learning for centuries. Elite institutions teach their students to speak with a refined, educated voice like Received Pronunciation (RP) in the UK and Standard American English (SAE) in the US (Mugglestone 24-26). Language learners are encouraged to abandon their native tongues in favor of a standard language to gain social standing and employability, or to pass English proficiency tests required by academic or professional institutions (Ramjattan 38).

This leads to the correlation between accent and wrongness, a relationship that has become part of social oppression—where maintaining authority over some “standard language” becomes a channel for upholding sociolinguistic hegemonies and erasing minority culture. In teaching minorities that their speech is flawed and inappropriate, education systems around the world are erasing accents and the cultural heritage they carry under the guise of promoting literacy and pursuing a developed nation. The oral tradition contained in accents—items like folklore, songs, and stories—gets lost in this persecution as younger generations lose the interest or ability to inherit it from their parents and grandparents (Mohabir 18-19).

In the years since the internet, language learning—like most facets of human life—has become increasingly dependent on technology. The technological mediation and disembodiment of voice that we heard in Abu Hamdan’s immigration interviews saturate almost every aspect of urban life from navigation apps to phone calls to karaoke machines. Within our context, natural language processing (NLP) and speech recognition technology act as adjudicators within language learning apps, allowing disembodied voices to function as a metric and judge of correctness. However, this technology remains far from perfect. In any language, models trained on a database of “neutral” hegemonic voices and accents often struggle to understand people who speak with accents different from this constructed standard of neutrality (“Here’s Why”; Linge). Used in language learning technology, this technology pushes learners to adopt this “accentless” standard, signaling that any accent that fails to fit into this mold is deficient and in need of correction.

However, I want to stress that this understanding of accent does not necessarily limit it to a negative role. If accent functions as a cultural artifact, then making the conscious decision to speak with that accent—and refusing to conform to some correct “standard” language—becomes a way for accented speakers to invert the cultural power imbalance and reclaim the value of their cultural heritage (Mugglestone 21). By thinking with an accent, we acknowledge the real problem of accented inequality that punctuates our hegemonic society. Redefining the accent from a marker of verbal deficiency into a vessel of linguistic heritage to be preserved and passed on gives accented speakers the rightful space to ensure that their communities are remembered.

Ramjattan’s conservationist effort to archive his grandmother’s Bhojpuri is a great example of this, but thinking with an accent starts with us and the ways we choose to interact with accents in our everyday lives. What is the voice we speak with? How is it accented, in its falling and rising tones, in the way words ebb and flow over each other to form a chain of meaning, and in the way it morphs as we speak to different people? What differences exist in the speech of our friends and families, and how do we react to their voices? By thinking with an accent, we expose the falsehood of the “standard language” and reclaim the accent as a marker of heritage and pride.

The Accent of Hellochinese: Democratizing Learning, Erasing Culture

Language learning applications have become popular options for amateur learners looking to develop their proficiency in a target language. I found the study of Chinese through these apps to be of particular interest because of the language’s significance in the China-Taiwan sociopolitical dynamic. Chinese has two main writing systems: traditional Chinese and simplified Chinese, used in Taiwan and China respectively (Republic of China [Taiwan]). The simplified system was created by simplifying some—but not all—commonly used characters to promote literacy (see table 1). Beijing has used simplified characters in all official and social settings since the system was created in the 1950s, while Taiwan has intentionally kept the traditional system as part of its rebellion against Beijing’s influence on the island (Guo 103). As a result, traditional Chinese cannot be separated from Taiwanese Chinese and the culture of its people, nor can it be meaningfully associated with the mainland.

Table 1

Some differences between traditional and simplified Chinese

Meaning in English	Romanization (pinyin)	Traditional	Simplified
Love	ài	愛	爱
Correct	duì	對	对
Book	shū	書	书
Head	tóu	頭	头
Noodles	miàn	麵	面
Eat	chī	吃	吃

Source: Pleco Chinese Dictionary, Pleco Inc., Vers. 3.2.73, 2023. Apple App Store, itunes.com/apps/PlecoChineseDictionary. Accessed 9 December 2023.

It is important to note that characters are pronounced the same in both traditional and simplified Chinese; they are writing systems, so there is no way of speaking in either of them (see table 1). Any differences in speech here are formed by the sociocultural separation between the communities that use each of these writing systems rather than rules inherent to these systems. It might be useful to think of this using the analogy of British versus American English here, where some words may be written differently (e.g. color versus colour), but it is the difference in cultures—and not spelling—that dictates the difference in pronunciation.

HelloChinese is a language learning app that focuses on teaching Chinese through reading, writing, speaking, and listening exercises. Built by developers from mainland China, the app defines standard Chinese as synonymous with the Beijing dialect and uses it as a standard of correctness (“Learn Chinese Pronunciation: Erhua”). In this case study, I will analyze a YouTube video uploaded by Taiwanese-Australian immigrants Eddy Chen and Brett Yang on their TwoSetViolin channel, in which they take a test in Traditional Chinese via the HelloChinese app. Both Chen and Yang speak with a Taiwanese accent in Chinese throughout the test (“Can We Pass a Chinese Test?”).

EDITOR'S NOTE: A figure accompanying this piece appears at this point in the PDF version but could not be embedded in this HTML version.

My first focus for this case study is the speaking test, taken by Chen. Chen spoke the sentence displayed on the screen, and his Taiwanese accent differed from the app’s accent for several words. There does not seem to be a clear relationship between accent and correctness based on how the app marked his words as correct or wrong in the sentence—the app marked sēn as wrong for 生 (shēng) but did not flag làng for 讓 (ràng), and marked 打 (dǎ) as wrong—which supports Linge’s findings that AI voice recognition remains relatively unstable and unreliable (see fig. 1). Therefore, I found it more useful to focus on the heard differences between the Beijing accent that the app employs as the “correct” accent it teaches versus the Taiwanese accent, as spoken by Chen.

In the video, we see Chen reattempting the sentence to reproduce the app’s accent. Even ignoring the app’s explicit adjudication of correctness, its obvious use of the Beijing accent encourages Eddy to believe that his accent is wrong and needs to be corrected, despite the fact that his Taiwanese accent very closely approximates how the sentence would be read by those comfortable with traditional Chinese. This is one way in which the app uses the flawed rubric of the Beijing accent to assert that a user’s rightful accent is, in fact, wrong.

EDITOR'S NOTE: A figure accompanying this piece appears at this point in the PDF version but could not be embedded in this HTML version.

The second part of this case study was the reading comprehension exercise, taken by Yang. This was a fill-in-the-blank exercise, where Yang was asked to complete the sentence with an appropriate word, here being 讓 (romanized by the app as ràng) (see fig. 2). The app defines this word as “told,” which is a meaning of the word only used in parts of mainland China (“让 [ràng]”; “讓 [ràng]”). The literal translation included with the sentence seems to acknowledge this fact, as the word is defined as “let” here. This proves that the app also uses Beijing vocabulary rather than a Taiwanese one. In the video, we see Yang puzzling over this exact mismatch between his intuitive understanding of the word and the app’s definition of it.

Both of these situations illustrate why the app’s use of the Beijing accent is concerning because users using the app to learn a language are likely to trust its authority on the correctness of language and meaning above their intuitive definitions, even if their intuitions were in fact right, as in this situation. The app’s use of the Beijing accent teaches a misleading and inaccurate understanding of how traditional Chinese is used by its Taiwanese community.

These apps are double-edged swords that promote linguistic diversity by gamifying the language learning process but also destroy the accented diversity of these languages. Many lessons in these apps acknowledge the cultural-linguistic relationship by personalizing exercises to represent the festivals, cuisine, clothing, and landmarks relevant to speakers of that language, therefore encouraging language users to understand the culture of the people whose language they are studying as an intangible part of the language learning process. However, because most of the languages are only offered in a single “standard” accent, these apps also destroy accent diversity by teaching that accents are wrong and unintelligible, and pushing for their elimination from “proper” speech.

In conclusion, language-learning apps democratize the language-learning process for millions worldwide, but they also inherit many of the linguistic prejudices found in the traditional classroom setting, especially that of accent as a linguistic shortcoming. In using the Beijing accent as the standard marker of correctness in Traditional Chinese, the developers of HelloChinese implicitly claim that Taiwan falls under Beijing's authority while disregarding its unique culture and identity. Language learning apps abuse their authority to falsely represent languages and their native speakers by forcing a hegemonic vocabulary and meaning onto a language. This is unfortunately difficult to solve in the current world, as the available dataset of voices for training AI speech recognition and NLP models is understandably saturated with samples from the hegemonic demographic, regardless of language. However, there is hope that this database will continue to grow as AI technology proliferates further in the future, allowing for more representative models of language learning to enter the market.

Works Cited

“Can We Pass a Chinese Test?” YouTube, uploaded by TwoSetViolin, 7 October 2021. www.youtube.com/watch?v=0pd4vMu2vzA.

 

Guo, Yingjie. Cultural Nationalism in Contemporary China: The Search for National Identity Under Reform. Routledge, 2004.

 

“Here's Why Your Phone Can't Understand Your Accent.” CBC, 11 September 2015, www.cbc.ca/radio/spark/292-what-you-say-will-be-searched-why-recognition-systems-don-t-recognize-accents-and-more-1.3211777/here-s-why-your-phone-can-t-understand-your-accent-1.3222569. Accessed 8 December 2023.

 

“Learn Chinese Pronunciation: Erhua.” YouTube, uploaded by HelloChinese, 5 June 2017, www.youtube.com/watch?v=VbziFBH8m7w.

 

Linge, Olle. “How Good Is Voice Recognition for Learning Chinese Pronunciation?” Hacking Chinese, 22 April 2019, www.hackingchinese.com/using-speech-recognition-to-improve-chinese-pronunciation-part-2/. Accessed 28 November 2023.

 

Mohabir, Rajiv. Antiman: A Hybrid Memoir. Restless Books, 2021.

 

Mugglestone, Lynda. Talking Proper: The Rise of Accent as a Social Symbol. Oxford University Press, 2010.

 

Phillips, A. A. “The Cultural Cringe.” Meanjin, Summer 1950, meanjin.com.au/essays/the-cultural-cringe-by-a-a-phillips/. Accessed 17 November 2023.

 

Ramjattan, Vijay A. “Accent Reduction as Raciolinguistic Pedagogy.” Thinking with an Accent: Toward a New Object, Method, and Practice, edited by Pooja Rangan et al., University of California Press, 2023, pp. 37-53.

 

“让 [ràng].” 百度百科 [Bǎidù bǎikē], baike.baidu.com/item/%E8%AE%A9/3400894. Accessed 9 December 2023.

 

“讓 [ràng].” 教育百科 [Jiàoyù bǎikē], pedia.cloud.edu.tw/Entry/Detail/?title=%E8%AE%93&search=%E8%AE%93#. Accessed 9 December 2023.

 

Republic of China (Taiwan), Ministry of Education. “Standard Typefaces for Chinese Characters: Preparation and Promotion.” Ministry of Education, 15 September 2013, Internet Archive. web.archive.org/web/20140915094836/http://english.moe.gov.tw/ct.asp?xItem=14786&ctNode=11446&mp=1. Accessed 7 December 2023.

---

Scaffold: A Showcase of Vanderbilt First-Year Writing, Volume 6, Spring 2024

Scaffold is a digital collection of first-year writing curated by the Vanderbilt Writing Studio. To highlight the developing writing processes and learning experiences central to the growth of new college writers, the collection pairs each piece with a recorded reflection from its author. Visit Scaffold’s website to listen to the authors reflect, to learn more, and to tap into this collection as a resource for college students and instructors alike.

The copyright to this work rests with the author. Proper attribution required. Vanderbilt retains a non-exclusive right to distribute the work as part of this collection.