Month: August 2022

In this research project we sought to compare the translation abilities of Spanish-English and Korean-English early childhood bilinguals (ECB) to second language learners (L2 learners) of Spanish and Korean. Early childhood bilinguals are those who acquire and develop their second language (L2) during early childhood (birth to 8 years old). Second language learners are those who learn their L2 at a more mature age, after already acquiring their first language (L1). We chose Spanish-English and Korean-English bilinguals due to the high population of Korean and Spanish speakers in Los Angeles, where this project was conducted (Data USA, 2019). For the purpose of this study we centered the context of translation/interpretation on a medical setting, due to the high-risk environment and necessity for accurate translations. Our participants included 7 ECBs of Spanish-English, 7 L2 learners of Spanish, 7 ECBs of Korean-English, and 3 L2 learners of Korean all of whom were 19-23 years old.

That being said, we attempted to answer the following questions:

• Do early childhood bilinguals have an advantage over second language learners when it comes to maintaining meaning when translating/interpreting?
• How does each group’s linguistic background affect their interpretation/translation skills?

Background

Previous research in Second Language Acquisition has often categorized L2 Learners as being more aware of their L2’s structure and linguistic properties, mainly due to their formal instruction and relatively late learning onset (McLaughlin, 1981). While this type of linguistic awareness might prove useful in certain situations, it does not necessarily equate to a greater ability to express meaning and communicate effectively (Colina, 2002). Because of this and the fact that L2 Learners are likely to have less exposure to their L2 than ECBs, we imagine that L2 Learners are less skilled at accurately expressing meaning in translations and interpretations, because they might be more focused on correctly translating the structure of the language.

Similarly, literature on the area of Language Brokering differentiates child L2 Learners and adult L2 Learners by their priorities when producing in the L2. For example, one study found that children prioritize accurate meaning when producing in the L2, whereas adults are more likely to prioritize grammatical correctness and structural form (Fathman, 1975 as cited in McLaughlin, 1981). In another study, it was found that children with prior language brokering experience had higher levels of metalinguistic abilities and enhanced problem-solving skills (Lopez et al., 2019). Given these findings, we decided to have our study focus on bilinguals with differing language acquisition backgrounds to better understand their priorities and tendencies when translating and interpreting.

When analyzing translations, it is important to understand what must occur in order to produce an accurate translation. Schwieter and Ferreira (2018) define translation and interpretation as processes that involve the comprehension of the input in the source language, its deverbalization or mental decomposition, and, lastly, its re-expression as output into the target language. Because preserving meaning while attempting to construct a message with a sound structure is difficult, we believe that our study can contribute to the body of knowledge in the area and potentially lead to new approaches for interpretation and translation training. Ultimately, it is crucial that all patients, regardless of whether they can speak English, have access to professional interpreters and translators as something as important as their health and well-being is at stake (Tonkin, 2017).

Method

We had participants reply to a 3-part questionnaire that included sociodemographic questions, a translation task and an interpretation task. The translation task asked participants to translate three medical terms from English to Spanish/Korean and three medical terms from Spanish/Korean to English. The interpretation task asked participants to interpret 4 sets of dialogue between a doctor and patient. The first two asked them to interpret the doctor’s English lines and the last two asked them to interpret the patient’s Spanish/Korean lines. This was to emulate an appropriate context for using interpretive skills. For the translation task, we looked at the percentage of participants that had correct translations, and for the interpretation task we identified patterns and tendencies for each group’s translations.

Results

For our Spanish speakers, we found that the ECBs performed better on five of the six single-word translations that are displayed in Figures 1 and 2. The ECBs were more likely to come up with the correct translation, except for the Spanish term “infección por estreptococos.”

We cannot fully understand why this happened, but we predict that the L2 Learners were better at making educated guesses for this term because they often used context clues when they did not know the exact translations. For example, when translating dolor de garganta into English, many of them said throat pain or pain of throat. However, this term actually means sore throat in Spanish. As individual words, dolor means pain and garganta means throat. Therefore, we predict that they made educated guesses based on their knowledge of these terms, but overall did not know the target translation. This distinction is important because throat pain and sore throats are very different medical ailments, which proves why specific medical translations are essential.

For our interpretation task, we knew that there would be more variability in responses, so we decided to analyze them by seeing if there were any patterns in how each group decided to translate. The responses from each group are displayed in Figures 3 and 4. When translating Is it difficult to breathe? we noticed that all of the L2 Learners had identical responses: Es difícil respirar?

 

 

 

 

 

 

They provided a very literal translation that coincides word-for-word with the English sentence. In contrast, the ECBs had more variety in their responses, using a broader vocabulary and more complex sentence structure. In general, the ECBs produced more complex interpretations that took the concept from English and conveyed it in a natural way in Spanish. Overall, the L2 Learners tended to maintain word choice and sentence structure across translation, while the ECBs demonstrated more flexibility in their responses.

Next, we noticed another difference in translation patterns for this sentence: Please point to the exact spot where it hurts. The ECBs provided very precise translations, whereas the L2 Learners were vaguer. As shown in Figure 3, five of the seven ECBs selected the Spanish verb apuntar as the translation equivalent of point. In contrast, none of the L2 Learners selected this word. In fact, all of them had unique responses. As shown in Figure 4, the L2 Learners use phrases like digame (‘tell me’), usar tu mano (‘use your hand’), señale con el dedo (‘point with your finger’) and toca el lugar (‘touch the place’) that have the intention of communicating pointing. It seems that they don’t know how to exactly express point, so they use these roundabout phrases instead. This is likely because the L2 Learners have weaker vocabularies.

 

 

 

 

 

 

Overall, the ECBs are mostly able to come up with a singular word to mean point, but the L2 Learners all used different phrases to try and communicate the same meaning.

While we noticed these differences when translating from English to Spanish, we did not identify many interesting patterns when the groups translated in the opposite direction. It seems like translating from Spanish to English was easier for both groups, because we had many more responses. Also, there was less variation in translations among groups and between groups, which led to very similar data.

For the Korean results, the ECBs also performed better on five of the six single-word translations. However, we significantly lacked data from the L2 Learners with only three participants. Most ECBs successfully translated the terms from English to Korean, whereas L2 learners appeared to struggle, averaging a 33% success rate.

When translating from Korean to English, we noticed that there were some over-translations among the ECBs. For example, our target word was 패혈성 인두염 (‘strep throat’), but they answered septic pharyngitis which is a more scientific term for strep throat.

Another term that was over-translated is 인후염 (sore throat) where we received the term laryngopharyngitis. In this case, sore throat should be pharyngitis rather than laryngopharyngitis which means a sore throat and an inflamed voice box. Since these terms are actually different, it is important to note since they may require different medical treatments.

 

In the interpretation portion for the Korean data, the participants translated, “Please point to the exact spot where it hurts.” from English to Korean. In the Korean language, there is no exact equivalent for the term, point, so this term is usually related to the phrase 가르켜주세요, which is directly translated to teach me. The closest direct translation for point is 짚어주세요. For this task, we saw that all of the ECBs answered using either 가르켜주세요 (teach me) and 짚어주세요 (point) when they interpreted for point. On the other hand the L2 Learners did not have correct answers as they completely neglected the term point since it is a more difficult term to translate and instead answered with 어디가 아파요? (Where does it hurt?) making their answers more concise and even changing the statement into a question.

 

 

 

 

 

 

For the next interpretation task we analyzed 배 위에 난방 패드를 붙이면 기분이 좋아져요 (I feel better when I put a heating pad on my stomach). While all of the ECBs mentioned the phrase, feel better (기분이 좋아져요), the L2 learners simplified the phrase by saying helps instead and restructuring the sentence from the target answer. Overall, the answers provided by the ECBs were much more detailed and accurate whereas L2 learners provided more concise answers or none at all. While the ECB answers varied, they used words with the same meaning consistently throughout all the tasks.

Discussion

An important finding in the translation task was that ECBs were better at translating medical terminology because they performed better in five out of six translation questions. This finding supports our hypothesis that the linguistic background of ECBs is advantageous to their translation abilities. For example, ECBs were more likely to give the actual translation for dolor de garganta (‘sore throat’), whereas L2 learners tended to make educated guesses based on their knowledge of what dolor (‘pain’) and garganta (‘throat’) mean, showing that they did not know the actual translation. This same finding was also true for Korean-English participants.

One important finding in the interpretation task was that ECBs had better lexicons and understanding of the language because they used specific Spanish words like apuntar (‘to point at’), whereas L2 learners had to find different ways to convey the English word point with words that they know such as, usar (‘use’), señale (‘show’), and toque (‘touch’). In addition to the frequency of ECBs acting as language brokers, our questionnaires also show that they acquired Spanish at the age of five or younger, and had more frequent exposure to the language compared to L2 speakers, which likely leads to improved vocabularies and understanding of the language. This same finding was also true for Korean-English participants.

The last finding in the interpretation task was that ECBs were more likely to be flexible and advanced in translating by using different structures. For example, they used Spanish phrases such as tiene dificultad para respirar? (‘Have difficulty breathing?’), se le hace difícil respirar? (‘Is it difficult for you to breathe?’), and others. From our questionnaire data, we saw that ECBs have more translation experience than the L2 learners, which likely prepared them to focus on translating meaning rather than the exact structure, because they know how a concept might be conveyed differently in each language. In contrast, L2 learners were likely to mimic the same structure of the first language, such as their word-for-word translation, es difícil respirar? (‘Is it difficult to breathe?’). This same finding was also true for Korean-English participants.

Based on the results of our study, we confirmed that ECBs were more likely to prioritize meaning, whereas L2 learners prioritized speech form and structure as expected from the studies we drew upon. Therefore, this further supports the notion that the L2 Learners’ awareness of language properties does not equate to the ability to accurately express meaning.

Our research can be particularly useful for the medical and healthcare industry, where professional translators and interpreters are needed to facilitate communication with patients. The findings of our study suggest that companies might prefer to hire ECBs because they are likely to have more experience with translating and interpreting than L2 learners. However, the slight variations in which ECBs expressed meaning also indicate that being bilingual does not necessarily mean you will be a perfect translator or interpreter, which is why it is important to have professionally trained experts. Our findings do show that there are benefits of learning a second language at an early age. Therefore, encouraging parents to consider bilingual education for their children at an early age can help develop their translating and interpreting skills, expand their lexicon, and allow them to comfortably use both languages, which could be seen with some of our participants.

Conclusion

Researchers should consider replicating this study on a much larger scale to ensure that there is enough data from each group. This study could also be improved upon by having pre-tests, such as cloze tests, that can accurately measure the fluency levels of the participants to compare the translation abilities of those that have similar proficiency levels. We didn’t measure the proficiency levels of participants other than by asking them to rate their own abilities in Korean and Spanish.

While the ECBs seemed to perform better than the L2 learners on a few different measures, ECBs still experienced their own struggles when translating certain medical terminology. This demonstrates that bilinguals are not born translators and being bilingual is not necessarily a sufficient requirement for professional translators and interpreters, especially in a medical setting.

 

References

Colina, S. (2002) Second Language Acquisition, Language Teaching and Translation Studies. The Translator, 8:1, 1-24, DOI: 10.1080/13556509.2002.10799114

Data USA. (2019) “Non-English Speakers in Los Angeles, CA.” Data USA, https://datausa.io/profile/geo/los-angeles-ca/demographics/languages.

López, B. G., Lezama, E., & Dagoberto Heredia, Jr. (2019). Language Brokering Experience Affects Feelings Toward Bilingualism, Language Knowledge, Use, and Practices: A Qualitative Approach. Hispanic Journal of Behavioral Sciences, 41(4), 481–503. https://doi.org/10.1177/0739986319879641

McLaughlin, B. (1981). Differences and similarities between first- and second-language learning. Annals of the New York Academy of Sciences, 379, 23–32. https://doi.org/10.1111/j.1749-6632.1981.tb41994.x

Schwieter, J., & Ferreira, A. (2018). Bilingualism, translation, and interpreting (pp. 251–266).

Tonkin, E. (2017). The importance of medical interpreters. American Journal of Psychiatry Residents’ Journal, 12(8), 13–13. https://doi.org/10.1176/appi.ajp-rj.2017.120806

Introduction

Attrition and incomplete acquisition have been one of the most intriguing concepts in language development–many children grow up learning multiple languages at once, especially in densely populated areas of bilinguals, including Los Angeles. Our study goal was to discover the degree of attrition in reading and writing amongst young adults in Los Angeles. We wanted to investigate the relationship between attrition of L1 and formal education in L2. In Los Angeles, 42% of the population speaks Spanish vernacular due to it being their heritage language while being formally educated in English, which could either lead to attrition in the future or incomplete acquisition of the Spanish Language. Attrition is the loss of language over time despite fully acquiring it during youth due to multiple factors such as “emigration, extensive use of L2 in daily life” (Schmid & Köpke, 2007), while incomplete acquisition is when children deviate from the norm and were never able to fully acquire a language (Köpke, 2018). In addition, Keijzer (2008) discusses Roman Johnson’s hypothesis, stating that “forgetting a language is the reversal process of learning the same language,” which was taken into account here because we believe this depends on age – a child may acquire a language in a few years but take many more as an adult for attrition to occur.

Linguists have hypothesized in the past through the interdependent development hypothesis that learning a second language could be accelerated if one or the other is morphologically rich, which Spanish is morphologically rich. In other words, the evidence points toward children being able to acquire both English and Spanish properly (Liew, 1996). In addition, since the parents’ L1 is Spanish, it would allow the child to be exposed to Spanish properly for the first few years of life, before heavily integrating into the American society to acquire English as well. While we cannot be certain whether attrition or incomplete acquisition, previous evidence and patterns of communication of parents tend to suggest the complete acquisition of both languages, meaning that attrition is more appropriate to test here. However, there has been evidence to support that if there is somewhat of a lack of language input for the child, then incomplete acquisition may occur – in a study about bilingual English-Spanish speakers, mothers were strongly working toward maintaining their children’s Spanish abilities: “the experience of these Spanish-speaking mothers reveal that while they perceived bilingualism as important for work and family communication, they often struggled to sustain their children’s Spanish development using various home language policies and discourse strategies…several mothers reported that their child was starting to prefer English, not only with siblings and peers but also with Spanish-speaking adult caregivers” (Surrain, 2018, p. 20). This evidence suggests that formal education, as young as preschool, can play a role in L1 language attainment if parents are not providing enough input of the heritage language to the children. This could translate to attrition in college students due to the potential lack of constant input of their L1. This is why we hypothesize that due to the lack of writing and reading and strictly using L1 orally, heritage speakers will greatly suffer in competency scores since their formal education is focused on English, the dominant language. Our prediction is that test scores will be middle to low.

Methods

Participants included 13 Spanish bilingual young adults from ages 18-21. All of them were recruited in the Los Angeles area and speak LA vernacular Spanish. First, the participants filled out a survey, investigating their perceived proficiencies in speaking, listening, reading, and writing in Spanish, as well as frequencies and demographics in which they would use LA vernacular Spanish. Then, they were given a college-level written test from CLEP (College Level Examination Program). Lastly, they were offered an optional interview to explore discrepancies between their perceived proficiency and the actual test results. We analyzed their degrees of attrition and deficiency in competency in their L1, mostly in morphosyntactic structures by observing what areas of morphosyntax they tend to struggle with the most.

Results

Survey Results:

 

 

 

Test Results:

 

 

Interview Results:

(Linguistic Landscape)

“Living in LA I read Spanish in a lot of different places, like billboards and fliers. It’s not everywhere in LA but I see it.”

“My parents and I try to keep up on our Spanish. We agree that we will exclusively speak it at home. And when we go to the market and places like that, we see it all over the place.”

“I never really took a Spanish class, but I see it when I am out in town or when I go to Mexican grocery stores. It’s not complicated stuff but I see it written and I read it.”

 

( Attrition / Dominant L2)

“I took some Spanish in high school because it was required, but I don’t read it often and I don’t write it ever. It’s mostly just speaking it at home.”

“I only really speak it at home with my parents, but other than that I don’t use it often.”

 

(Dialect and Prescriptivism)

“This kind of Spanish isn’t exactly the type of Spanish I use, it seems like this is more like Spain Spanish.”

 

Analysis

When observing the figures in the proficiency portion of the survey, our participants were fairly confident in their abilities as well-rounded bilinguals. There is a common consensus that the more someone speaks a language, the more comfortable they will become speaking the language. The average proficiencies were well above our predicted levels, as we predicted a steeper drop in the confidence in reading and writing. This goes against our hypothesis that predicted that formal education would have a significant effect on proficiency levels in reading and writing. There was a significant drop in confidence in relation to the participant’s confidence in their writing ability which does go in the direction of what we expected, just not by the amounts that were projected. One should keep in mind that the group is a mix of heritage speakers with and without formal Spanish education. This may have an impact on the data.

In the aspect of frequency of use, the frequency of speech portion is relatively high but our expectation was for a middling average. The frequency is similar to the perceived confidence in language use which is interesting. The frequency numbers drop in a similar way to the proficiency. The lower confidence level in reading and writing seems to have a correlation to the frequency of use, thus it can be stated that frequency of use possibly bolsters confidence in an area of speech. One note is that the frequency of writing in Spanish is even lower than the confidence level of writing in Spanish.

When asked about the social environments in which our speakers use Spanish, the results are extremely stacked in one direction. The language of choice in most instances was overwhelmingly in favor of English. The only instance of the prominent use of Spanish is with older family members. This in a way is logical since the parents’ L1 is Spanish and they would feel more comfortable communicating in the Spanish vernacular of their heritage language rather than English. Even with younger family members, the numbers polled much stronger toward English, but a decent number still spoke to them in Spanish. This information tracks with our hypothesis of attrition, in that English, has the dominant use in their day-to-day language usage. However, this was the only portion of our data to support our claim.

The Spanish reading and writing test proved to predominantly correspond with the perceived proficiency stated in the survey. Eleven of the participants scored ten or higher on a fifteen-question test. The most common score was twelve out of fifteen for four participants. Scores thirteen and fourteen account for another four participants.

One strange observation is the two outlier scores of eight correct answers. Upon viewing these tests individually, some seemingly simple mistakes were made, and for some, the commonality was a bit of struggle on the correct past form to use. One explanation could be a rushed test taker not completely reading the question upon answering, but this is not confirmed. The case of the incorrect past forms is a small commonality of about two questions incorrect on it. Here the participants had a hard time distinguishing between the preterit or the imperfect forms. The reason we believe that the test was rushed is that the clues to use imperfect are usually with the text, and if the text is not fully read, then the reader may not be alerted to use the imperfect. These two participants also missed some easy questions that a beginner Spanish learner would get correct, let alone a heritage speaker. There is no real way to truly claim that a rushed test is the reason for these outliers, but from the evidence presented, this may be the likely case. This also calls to question our methods, as they may be too long for participants to where they may rush through the test and skew our results.

Ultimately though, our participants exceeded our expectations in terms of testing. The test results produced were higher than expected with only our two outliers being in the middle. Even the questions that were wrong could be argued to be fully grammatical depending on the variety of Spanish used as this is a prescriptivist Spanish test. They did however struggle with a particular question pertaining to the present progressive, and whether or not to add a preposition. This phrasing is more typical of formal written language, rather than spoken language. There is a lack of a present progressive trigger word like “estar”, making this even more marked.

The final portion of the test was an optional interview portion asking about participants’ scores, many of whom responded with comments about living in the Los Angeles area and seeing Spanish in their environment. Linguistic landscape has played a large role in the test results according to our participants. Many of them are subjected to many forms of written Spanish stimuli in their surroundings due to the large Spanish-speaking community in Los Angeles. This was not taken into account when planning our experiment but gives us a good idea as to why our data concluded the way it did.

 

Discussion

Foremost, our biggest flaw was not taking the linguistic landscape into account for our project as a whole. Our early results were strange when first testing, we could not predict just how well our participants projected and performed. Our interview portion really shed light on our shortcoming of neglecting such an influence on heritage speakers in this area. That being said, if we were to do another experiment and frame a comparison against heritage speakers in a less optimal linguistic landscape where English is dominant. We believe that the results would be a staggering comparison as to how much the linguistic landscape plays a role in the limiting of attrition.

We believe that our method format was extremely solid in terms of the quality of data produced, however, the length of a survey, a test, and an interview severely limited the amount of people willing to participate. To some degree, we believe that it even affected our test results. The outliers in our test data suggest that two of our participants may have been speeding through our test without fully reading the questions. Even leaving the interview portion optional, the sequence of google form after google form has given us some unintended consequences.

But to the credit of the long format, the interview portion was eye-opening and put our entire experiment into a better perspective. Three comments describing aspects of the linguistic landscape were pivotal in understanding the prior results. However, the interview also highlighted the flaw of overly prescriptivist language use in our test. The comment that our test was too “Spain Spanish” was something that we knew would come up, however, there is no known formal test that we know of which would test for specific dialects. I believe that we do not have enough credibility as linguists to create this, but for other linguists moving forward, dialect-specific tests may be a useful tool.

Another issue that we observed was the inconsistency of heritage speakers with prior formal education in Spanish. In hindsight, we should have included several questions in the survey asking specific questions about their formal training. As for finding exclusively non formally trained heritage speakers, it’s hard to control for, especially since most schools and colleges require a foreign language. Many heritage speakers opt for taking Spanish as their foreign language in school leaving a small number of young adult heritage speakers to pick from to keep the data consistent.

 

Conclusion

In the survey, our test group projected that their proficiency was fairly high for reading and writing in Spanish, and their reading and writing test results confirmed this. The only evidence in our favor was that English was their preferred language of choice in most social scenarios, but this did not translate into the test data. Ultimately our hypothesis was proven false, as our group exceeded our expectation of proficiency and did not exhibit distinct signs of attrition in their Spanish reading and writing. This also demonstrated that formal education does not have as much of a tremendous role in proficiency in the heritage language as we expected. Thus, it can be concluded that for our group of Los Angeles Spanish heritage speakers, attrition was not present and that our bilinguals have attained high proficiencies in their L1s and their L2s.   

 

References

Elizabeth M. Liew (1996) Developmental Interdependence Hypothesis Revisited in the Brunei Classroom, Journal of Multilingual and Multicultural Development, 17:2-4, 195-204, DOI: 10.1080/01434639608666271

Goral M, Libben G, Obler LK, Jarema G, Ohayon K. Lexical attrition in younger and older bilingual adults. Clin Linguist Phon. 2008 Jul;22(7):509-522. doi: 10.1080/02699200801912237. PMID: 18568793; PMCID: PMC3128922.

Keijzer, M. (2004). First language attrition: A cross-linguistic investigation of Jakobson’s regression hypothesis. International Journal of Bilingualism, 8(3), 389–393. https://doi.org/10.1177/13670069040080030201

Keijzer, M. (2010). The regression hypothesis as a framework for first language attrition. Bilingualism: Language and Cognition, 13(1), 9-18. doi:10.1017/S1366728909990356

Köpke, B. and Genevska-Hanke, D. (2018) First Language Attrition and Dominance: Same Same or Different?. Frontiers in psychology, 9, 1963. https://doi.org/10.3389/fpsyg.2018.01963

Montrul, S. (2002). Incomplete acquisition and attrition of Spanish tense/aspect distinctions in adult bilinguals. Bilingualism: Language and Cognition, 5(1), 39-68. doi:10.1017/S1366728902000135.

Schmid, M. S., & Köpke, B. (2007). Bilingualism and attrition. Language attrition: Theoretical perspectives, 33, 1-7.

Surrain, S. (2018). ‘Spanish at home, English at school’: How Perception of Bilingualism shape Family Language Policies among Spanish-speaking Parents of Preschoolers. International Journal of Bilingual Education and Bilingualism. 1-15.

 

Introduction and Background

Code-switching is a linguistic phenomenon where speakers who are fluent in more than one language quickly change from one language to another while they’re speaking. It’s very common and is often tied to sociocultural factors. For example, a person might code-switch when they need to use a specific word that is cultural and doesn’t have a good translation in another language. They might also code-switch to show others that they are fluent in multiple languages and to connect themselves to a certain identity or heritage (Myers-Scotton, 1993).

We define accent hold-over as the transmission of the phonological system of one language to speech in another language, tied to code-switching. For example, imagine that a person who speaks both Korean and English is talking to a parent in Korean. They might briefly code-switch to English to say a short word or phrase, such as a person’s name or a culturally-related phrase, before switching back to Korean. We would potentially see an accent hold-over as they are switching between the two languages. Because the phonological systems, or the sounds present in the language, are very different between Korean and English, a person’s speech might be more accented than their normal speech for a short period of time.

Grosjean (1988) studied this phenomenon, which he called the “base-language effect,” and found that it exists in language processing. When two bilingual people are speaking and one person code-switches to another language, the person listening to them experiences a small lag in processing what they’re hearing. Their brain expects to hear sounds from the previous language, and needs a brief period of time to adjust and switch to the other language. A later study by Grosjean & Miller (1994) didn’t find this in language production, but research has been mixed. Some, such as Olsen (2013), have found this hold-over effect in production, while others, such as Šimáčková & Podlipský (2018), have not. Overall, there hasn’t been much research in this area, and past research has not focused on the impact of a speaker’s linguistic background.

In our study, we will look at how Korean-English bilinguals interact with each other, compared to how a Korean-English bilingual interacts with an English monolingual speaker. Our goal is to investigate how vowel and consonant sounds change (or remain the same) as a person switches between Korean and English mid-sentence. We chose these two languages partially out of convenience, because there are many Korean-English bilinguals in this area, and two of our group members are also Korean-English bilinguals. But we also considered the phonological inventories of the two languages: Korean and English have very different sounds, which could amplify these effects. Going into this experiment, we hypothesized that Korean-English bilinguals would code-switch and show more of an accent holdover effect when talking to a Korean-English bilingual compared to a monolingual English speaker.

Similar to the hold-over effect, Maye et al. (2008) showed that when listeners heard a segment of Wizard of Oz where the vowels were shifted down in the vowel space, listeners were easily adapted to the accent change. The listeners were then able to find a plausible implementation of the shifted accent in other real words. Based on this case study, we can argue that bilingual speakers who hear the hold-over accent in other speakers may be able to use it in their own speech as well, once they are accustomed to the sound changes. However, because our speakers will be reading a prewritten script that is written in both English and Korean, it is possible that the orthography of the script may elicit some accent adaptation (Norris et al., 2003).

More generally, we are also looking at Korean phonetic code-switching when reading Korean words in English sentences. Additionally, other factors that could affect code-switching in this instance include prosodic flow, language dominance, community norms (Myers-Scotton, 1993), and the bilingual’s personal language practices and fluency (Poplack et al. 2020). Poplack et al. (2020), when looking at the phonology of English borrowings spoken by French-English bilinguals, also found that there was a large amount of variation between different groups of consonants, so we will be investigating both consonants and vowels in this experiment.

Methodology

To explore our research question, we ran a small experiment: we gave pairs of participants a script to read off of, and recorded and analyzed their speech. We ended up with two pairs of participants. One pair had two Korean-English bilinguals, both of whom have lived in Korea and the United States and considered themselves fluent in both languages. The other pair had one Korean-English bilingual, who had also lived in both Korea and the United States, and a monolingual English speaker who wasn’t fluent in any other language and had little to no exposure to Korean. We accepted the language background that our participants told us, but a more extensive study should include a language test to formally assess this.

Our script was written by two Korean-English bilinguals, Joan and Yoori. The script had two speakers. Speaker one’s part was entirely in English but had Korean food borrowings: Korean words referring to cultural foods, transliterated into English text because English equivalents don’t exist. For example, we included the words ‘kimbap’ and ‘hangook hotdog.’ Speaker one was designed so any English speaker, regardless of their knowledge of Korean, would be able to read it. On the other hand, speaker two spoke mainly in English but code-switched to Korean. This code-switching was indicated in the script using Hangul, the Korean alphabet, unlike the food borrowings for speaker one that were in the English alphabet. By writing the words in Hangul, we indicated that speaker two should be entirely switching from English to Korean. Then, we could see how these mid-sentence code-switches changed the speaker’s English phonology and pronunciation.

Analysis ended up being a little difficult for us. We originally wanted to look at the sounds in Praat, but that ended up being unrealistic given the nature and scope of our experiment. Our main method of analysis was listening to our recordings by ear, because we had both English speakers who didn’t have exposure to Korean and Korean-English bilinguals. We looked at the phonological inventories for both languages to make sure we had accurate intuitions and picked out notable segments.

Results and Analysis

After recording the participants, we listened to the speech by ear. We expected words before and after a Korean word to have an accent holdover. Interestingly, there were no accent holdovers in the words we expected them to be. For example, in line 2 (see Figure 1), the word tired is pronounced [ˈtaɪəɹd] instead of [ˈtaɪjəd]. We also predicted that proper names would contain accent holdovers, such as in line 4 when the participant is asked to read Powell Library, but with the word library in Korean. Instead of pronouncing it as [pa:ul], the word Powell was instead pronounced as [ˈpaʊəl] by both bilingual speakers reading speaker 2 lines. Our script also gave opportunities to pronounce Korean food names, as well as a different nationality’s dish to see if food in a different culture and language would also be pronounced with a Korean accent holdover. We provide this in lines 6 and 7. While both bilingual speakers reading speaker 2 lines pronounced the Korean word kimbap in the proper Korean pronunciation, the word tacos in the same sentence was pronounced in a standard American accent as [ˈtɑkoʊz] instead of [ˈtakoz] as predicted. Interestingly, when the bilingual speaker was reading the speaker 1 script in line 7, the word kimbap and tacos was spoken in an American accent, rather than in the Korean pronunciation. However, an explanation for this would be that in line 7, the word kimbap was written in English instead of in the Korean orthography, possibly causing the bilingual speaker reading that line to have chosen to not read it in Korean.

One thing to note is that when the Korean-English bilingual speaker was reading speaker 1’s line in the script, all Korean food names were read in the standard American accent. For instance, in lines 7 and 8, the words kimbap, hangook hotdog, and ramen were expected to read as [kimpap’], [‘hanguk hatogɨ’], and [ˈlamjʌn] respectively. Instead, the words were elicited as [kɪmbap], [hɑnguk hɑt dɔg], and [ˈɹɑmjʌn]. [ˈɹɑmjʌn] particularly stood to us during our analysis because it appeared to be a combination of a standard American accent and the Korean accent. More specifically, the first syllable is pronounced with the [ɹ] that is not found in the Korean phonemic inventory, but the second syllable is pronounced [mjʌn] as a Korean speaker would.

Another accent incident we noticed was that both bilingual speakers reading speaker 1’s line produced a club’s name in a Korean accent. The club, KASA, or Korean American Student Association, was expected to be pronounced as [ˈkʰɑsə], but was instead produced as [kʰasa] by both speakers.

Discussion

Overall, we did not find any evidence of an accent holdover when bilinguals switched between Korean and English. In every instance of code switching, we observed that the accent being produced changed immediately along with the language in which the corresponding word was meant to be read. The previously mentioned cases that we found (e.g., the case of “KASA”) seem to be instances of code-switching, as opposed to accent holdover. This would indicate that there is no phonological accent holdover effect for bilinguals, reflecting the findings of Grosjean & Miller (1994) and Šimáčková & Podlipský (2018). This suggests a high level of flexibility in language systems and instantaneous or near-instantaneous shifts between languages, as past research has also indicated.

We did, however, have other particularly interesting findings almost entirely unrelated to accent hold-overs.  For instance, the words “KASA” and “ramen”, while formed or written in the script in English, were pronounced at least in part with a Korean phonemic inventory. This might suggest that there is a code switch for words that relate to Korean culture or identity, even if they are not formed in the Korean language in the script or in general.

When it comes to accent holdovers themselves, because of a number of limitations, we are not confident that our results definitively demonstrate that accent holdovers do not exist. The small sample size of participants was one, as we only tested one pair of speakers per context (bilinguals speaking to each other vs. a bilingual and a monolingual speaking). The sample size is especially notable due to high variation among bilingual individuals, which means that we may be required to recruit a larger sample size to account for this variability and to be able to construct a meaningful average.

Perhaps our most significant limitation is the fact that the context in which participants had to elicit words was largely unnatural, as it was based entirely off of a prepared script. While it was difficult to find an efficient way to elicit the data we needed for analysis, the scripted nature of elicitation could have affected results in unforeseen ways. Chief among these is the fact that Korean and English have radically different orthographies, which, over the course of this study, often forced us to choose how to spell certain target words such as “ramen”. Therefore, the scripted nature of the experiment might also yield an alternative explanation as to why the word “ramen” was pronounced the way it was: a word originally formed in Korean, it was spelled in its English borrowed version in the script, yielding a potential confusion within the bilingual speaker as to how he ought to pronounce the word he was reading (and therefore an odd code-mix). There is also a possibility that the speakers were conscious of their pronunciation after they noticed that their voice was going to be recorded. This might have made them more focused on pronouncing English words in the English inventory and Korean words in the Korean inventory so that neither their English nor their Korean would sound “butchered” or their transition would seem “incomplete” to the recording and listening experimenters. Considering all of the above-mentioned limitations, our results become difficult to interpret.

Conclusion

Despite suggestions of differences in language processing when it comes to perception of accent change in bilingual speech and code-switching, accent holdovers have not yet been observed in speech of bilinguals to one another, or in speech of a bilingual that is directed to a monolingual. There are several ways the limitations of this particular study could be addressed, but ultimately it appears that code-switching is instantaneous and happens not just on a general word level, but on the phonemic level as well: the inventories switch as much as does the vocabulary. This may illustrate a broader trend in that the multiple languages are placed in the bilingual speaker’s mind as different systems that, even when somewhat mixed (like in code-switching), remain firmly separated.

 

References

Grosjean, F. (1988). Exploring the recognition of guest words in bilingual speech. Language and Cognitive Processes. 3, 233–274.

Grosjean, F., & Miller, J. L. (1994). Going in and out of languages: An example of bilingual flexibility. Psychological Science, 5(4), 201–206.

Myers-Scotton, C. (1993). Social Motivations for Code-Switching: Evidence from Africa. Oxford: Clarendon Press.

Poplack, S., Robillard, S., Dion, N., & Paolillo, J. C. (2020). Revisiting phonetic integration in bilingual borrowing. Linguistic Society of America, 96(1), 126–159.

Maye J., Aslin R. N., Tanenhaus M. K. (2008). The weckud wetch of the wast: lexical adaptation to a novel accent. Cogn. Sci., 32, 543–562.

Norris D., McQueen J. M., Cutler A. (2003). Perceptual learning in speech. Cogn. Psychol., 47, 204–238.

Olson, D. J. (2013). Bilingual language switching and selection at the phonetic level: Asymmetrical transfer in VOT production. Journal of Phonetics, 41, 407–420.

Šimáčková, Š., & Podlipský, V. (2018). Patterns of short-term phonetic interference in bilingual speech. Languages, 3(3), 34. MDPI AG.