The Influence of Aspects of Social Identity on the Development of L2 Phonology

Over time, second language (L2) speech production changes as the learner gains more experience with the language. Factors such as interaction with native speakers of the learner’s L2 are known to play an important role. It is less clear to what extent, if at all, aspects of social identity influence the development of the L2 (Hansen Edwards 2008:372–373). This longitudinal study examines the development of the actor Jackie Chan’s L2 (English) phonology. His speech production in two time periods 9 years apart is contrasted: in 1998, before he gained success in the English-speaking world, and in 2007, after he had released multiple Hollywood blockbusters. To check that factors such as age of acquisition were not the only reason for a lack of alignment over time towards English native-speaker norms, another context was taken from 2007, namely, an interview with a French native speaker. In foreigner-directed speech, there exists a pressure to produce more standard variants (Zuengler 1991:234). If Chan uses fewer non-standard variants with a non-native English speaker than with a native English speaker, it would imply that there is some degree of intent in his usage of non-standard forms, and that his development, or lack thereof, is not only due to uncontrollable factors of second language acquisition. Two variables are examined: his production of stops in word-final codas containing a single stop and in word-final consonant clusters containing a stop as the final consonant. Native speakers of English generally pronounce these stops, whereas native speakers of Chinese often simplify them by deleting or glottalising them (Setter et al. 2010:15, Hansen 2001:340). In 2007, Chan is found to use a greater rate of the standard non-simplified variant than previously; however, he also simplifies his pronunciation by deleting the stop in the codas more often than in 1998. He uses standard forms that align with English native speakers to a greater extent when talking to non-native speakers. After 9 years of working in Hollywood, he would have gained more experience with English due to his social network consisting of more English native speakers, resulting in the expected increased alignment with native-speaker norms. However, his English has developed so that the non-standard variant of deletion is also used


Introduction
There exists a basic assumption that as a second language (L2) learner gains more experience with the L2 and interacts more with native speakers of that language, their speech production aligns further with that of the native speaker. However, research on sociolinguistic aspects of bilingualism has shown that a speaker may choose to use phonological features from their first language (L1) in order to express their social identity. The question then arises of how great a role the social identity of the speaker plays in the development of their skills in a new language. In this paper, a longitudinal study was conducted on the L2 speech of Jackie Chan, a famous Chinese actor, who has starred in multiple famous American films over the last two decades. Chan's L1s are Cantonese and Mandarin, and he is an adult L2 learner of English. The study examines the change in his production of final stops in single codas and in consonant clusters in interviews taking place 9 years apart, before and after he gained success in the Englishspeaking world. Overall, the development of Chan's English does appear to have been influenced by his social identity. Although his phonology has in general become closer to that of a native speaker over time, the usage of certain variants influenced by his L1s has, in fact, increased as pressures from the linguistic marketplace become less prevalent.

Literature Review
The field of bilingualism is still fairly young, and thus there are still numerous theoretical issues to be resolved. In particular, there is no clear-cut answer to how important the factor of identity is to the acquisition of an L2 (Hansen Edwards 2008:272-273). There exists a general assumption that an increase in usage of an L2 results in the general amelioration of L2 skills, both for children and adults (Beardsmore 1982:19). More specifically, it appears that the pronunciation of L2 learners who have greater contact with native speakers of the L2 is more closely aligned to that of a native speaker (Bayley 1996:111). However, numerous studies have demonstrated that bilingual speakers may continue to use certain phonological features of their L1 in order to emphasise their social identity as a bilingual speaker and as a native speaker of their L1 (Hansen Edwards 2008:262). Unfortunately, most research conducted around this issue only focuses on either the social identity of different speakers at various levels of proficiency or the change in a speaker's phonology over time (Hansen Edwards 2006:1). Very rarely does a longitudinal study also concentrate on the social factors affecting a person's L2 development. Thus, it seems opportune to analyse the change in a person's speech production over time in specific relation to aspects of their social identity.
As mentioned above, empirical research has so far demonstrated a definite difference between the speech production of less and more experienced L2 speakers. This appears to be caused by increased usage of the language (although this remains controversial, cf. Edwards 2008:255); much more importantly, it is also influenced by acquiring a social network which includes native speakers of the L2. A greater amount of interaction with native speakers allows the L2 speaker to receive the quantity and quality of input necessary to learn the sociolinguistic norms governing the production of the target variable (Bayley 1996). In addition to the second language acquisition (SLA) processes involved, this change also appears to be influenced by a very important social factor, namely, the linguistic marketplace. A bilingual speaker may find it necessary to maximise their alignment with a monolingual native speaker of the L2 in order to reduce communication difficulties and increase their linguistic capital (Beardsmore 1982:40). Further individual differences within SLA are dependent on a person's motivation (Skehan 1989:51). Individuals who find it necessary to improve their L2 for economic reasons (i.e., instrumental motivation) or due to the importance they place on assimilation (i.e., integrative motivation) will desire to change their L2 pronunciation in order to sound more similar to native speakers. Aligning to a greater extent with native-speaker norms can aid a person in their career or in their ability to be an active participant in the local community.
Conversely, a bilingual speaker may also actively choose to retain certain phonological features of their L1 as variants within the L2 (Hansen Edwards 2008:257). Phonology is a "domain within which one's identity is expressed" (Zuengler 1988:34), so speakers who continue to feel a strong connection to the community and culture of their L1 may be less willing to align this aspect of their speech to that of a native speaker of the L2. Retention of * Special thanks are due to Emma Fong for helping to collect and check the data and assisting me with the Results section. L1 phonological features can emphasise a speaker's pride in their heritage or strong connections to their L1 community (Hansen Edwards 2008:257-8). As the bilingual's usage of L1 phonological features is influenced by a desire to express their social identity, this variation can perhaps be explained by speaker design theory. Speaker design theory posits that speakers construct a social identity through their speech, converging or diverging from their interlocutors at will as well as selecting which variant to use to better express their identity (Schilling-Estes 2004:388-9). A person who is under no specific pressure to accommodate to their interlocutor may converge to a much lesser extent and instead use features of their native variety (Shapp et al. 2014:156). It seems likely, then, that an L2 speaker who has over time lost the necessity to gain greater linguistic capital would similarly find it unnecessary to attempt to align with native-speaker norms. As of yet, however, the scope of possible variation is unclear, as there has been little research done on SLA in regard to speaker design theory.
Unfortunately, it is uncertain how much this study can further current knowledge on these issues. For one, only one individual is analysed, and not a particular community; moreover, the specific interaction between Chan's career as a Chinese actor in the American market and his language skills make him dissimilar to most other L2 learners.
The results from such analyses are always difficult to extrapolate to the general public (Shapp et al. 2014:150-1). Confounding the issue further, SLA can be influenced by so many unique factors that it is generally more difficult to draw generalisations from single individuals (Skehan 1989:148). Nevertheless, there is still some value to this case: it helps define the scope of variation and whether factors of social identity can influence the development of the learner's L2 to any notable extent.

Speaker Background
Jackie Chan was born in Hong Kong in 1954 and at the age of 7 joined the Peking Opera School. His L1 is Cantonese, but he is also fluent in Mandarin; for the purposes of this study, they are both described as being his L1s, as Mandarin influences his English similarly to Cantonese. The first variety of English as an L2 that Chan encountered is likely to have been Hong Kong English; however, according to him, his first real experience with the language came as an adult: in his 20s, he briefly did construction work in Australia. Soon after, he returned to Hong Kong to work in movies and had a long and successful career there. He attempted several times to break through to the American market but had a lukewarm reception until the 1995 hit Rumble in the Bronx. He starred in the movie Rush Hour 3 years later, which proved to be his first giant Hollywood blockbuster. Over the next decade, he became a household name, with movies such as Shanghai Noon and The Tuxedo (Chapman n.d.).

Data Collection
For this study, the necessary data were collected from online video clips, three of which were hosted on YouTube and two on AlloCiné. The sound from the files was ripped using Soundflower 2ch (Ingalls 2012) and recorded with Audacity (Audacity Team 2012). The variables were then coded manually using Praat. The author and a colleague transcribed, coded, and checked Chan's speech together. Originally, each person coded different clips. However, an interrater reliability check of 10% of the data found a too low agreement rate of 74%, so we re-examined the data, choosing to rely primarily on acoustic analysis. We used formant transitions only as secondary confirmation. We then checked a different 15% of the data, which gave an agreement rate of 85%. The person checking changed any token whose value they did not agree with to the preferred value so that any discrepancy in coding would not be split completely by clip and thus by the independent variable. The data collected were then analysed and the graphs constructed using R (R Core Team 2013).

Hypothesis and Social Variables
For this longitudinal study, the main independent variable examined was time period. Chan's interviews from 2 different years were examined: 1998 and 2007. In 1998, the first Rush Hour movie came out-this was Chan's first giant blockbuster success in America. In 2007, 9 years later, after filming multiple movies in English and achieving both commercial and critical success, the movie Rush Hour 3 was released (Chapman n.d.). While filming in America, Chan's social network of English native speakers increased; however, as he had already proven himself to belong to the Hollywood A list, the pressures of the linguistic marketplace were less intense in 2007 than they had been in 1998. We anticipated that comparing these two periods would reveal how factors of social identity influenced the development of Chan's English. Our hypothesis was thus two-tailed: we predicted that these contrasting factors would cause Chan's English to become either more or less aligned with native-speaker norms over time. As the interviews were about, and in relation to, promoting the same movie franchise, the effect of the topic on variation could be kept relatively stable. Although Bell (1984:177) proposes that in mass media the target audience (i.e., English-speaking fans of Chan) is perhaps equally or even more influential than the direct interlocutor, we ensured that both interviewers were also similar to each other: they were white, middle-aged men, who speak General American English.
A possible problem in the research design lay in its generally being more difficult for the bilingual to acquire native-speaker-like L2 phonology if the L2 is learnt post-pubescently, regardless of any other factors (Ortega 2009:23). As Chan began speaking English frequently only as an adult, this meant that a lack of change towards closer alignment with native speakers over time might be due to age of acquisition rather than social identity factors. Therefore, to check that it was not age of acquisition that was solely responsible for a lack of difference between the two time periods, a second context for the 2007 period was chosen: one in which Chan was interviewed by a fellow non-native English speaker from a different linguistic background to his own. In foreigner-directed speech, there exists a pressure to accommodate to the interlocutor by means of greater articulation and use of standard variants, for example (Zuengler 1991:234). Although there has been little research on foreigner-directed speech by non-native speakers, it has been previously shown to exist (Milk 1990). Thus, if Chan's speech when talking to a non-native speaker contains fewer non-standard features caused by the influence of his L1s, then the lack of alignment to native speakers over time could not solely be due to general SLA factors. Rather, it would mean that this lack of change is caused by some intent to emphasise aspects of his social identity. For the non-native interlocutors, French native speakers were chosen; unfortunately, it was not possible to find enough data of Chan talking to just one interviewer. However, we ensured that the interviewers were French native speakers and the intended audience was French fans in all three interviews. 1

Linguistic Variables
The Whether a variant is standard or non-standard depends on the segment following the final cluster: simplifying (i.e., reducing or deleting) the cluster when the following segment is a vowel or a pause is non-standard. His glottalisation of the stop in both a single coda and in consonant clusters is influenced by the phonology of Cantonese and is thus present in his speech, which can be classified as Hong Kong English (Setter et al. 2010:15). Deletion of stops in single consonant codas, on the other hand, is found in speakers who have Mandarin as their L1 (Hansen 2001:340). The deletion of the stop in consonant clusters is a possible variant found in both Mandarin and Cantonese L1 speakers. It is also found in numerous varieties of English among native speakers; however, there are stricter phonological rules as to when the cluster can be simplified. In particular, native English speakers generally simplify the cluster only when it is followed by a consonant, 2 whereas both Mandarin and Cantonese speakers can simplify clusters containing any stop, also when the word-final coda precedes a vowel or a pause (Bayley 1996:97-9, Deterding et al. 2008. These two variables were chosen because they can both be argued to be relatively salient: the non-standard variants of both are used when imitating stereotypical East Asian speakers (Chun 2004:272-3).
Other specific linguistic factors were taken into account when coding. In Hong Kong English, the variety Chan speaks, the voicing of final stops is neutralised, and both voiced and voiceless stops are pronounced as voiceless (Setter et al. 2010:15). Therefore, in this study, the categories were collapsed, and we coded for the stops without separating them in relation to voicing. The environment where the variable occurred was also coded for: we noted whether the stop was followed by a consonant, a vowel, or a pause. This was done in particular because of the consonant cluster, whose phonological environment affects whether an English native speaker would simplify it.
Another linguistic factor we considered was assimilation. As simplifying the coda due to assimilation is a phonological process and not necessarily dependent on a speaker's linguistic background, all tokens where the coda was followed by the same consonant were removed from the data. Finally, we ensured that no words which were inflected using the -ed suffix were coded for. Previous research has shown that the rate of coda simplification of Mandarin L1 speakers is significantly greater within regularly inflected verbs (Bayley 1996:109). As it is phonological simplification and not that related to a grammatical process which is of interest in this paper, leaving these tokens in would have unnecessarily skewed the results. Altogether, the data from 1998 contained 19 tokens within a cluster and 33 single stops; the 2007 American interview had 51 consonant cluster codas and 105 single stop coda tokens; the French interviews had 23 consonant cluster coda tokens and 44 single stops.

Results
Overall, this longitudinal study shows that Chan's English underwent a noticeable change between 1998 and 2007. In regard to the single coda simplification (see Figure 1), Chan initially deleted 21% of the stops, glottalised 67%, and pronounced 12%. In the later American interview, he deleted 26% of the stops but pronounced the final consonant without simplification 30% of the time. He also glottalised 44% of the consonants. When being interviewed by non-native speakers, Chan deleted only 16% of the stops, of which 45% were glottalised and 36% pronounced. In 2007, when speaking to an English native speaker, he deleted 5% more than in 1998, although deletion is a variant not found in native-speaker varieties. However, he also pronounced consonants without any simplification 18% more than in 1998, which shows developed adherence to native-speaker norms. As native English speaker cluster simplification is constrained by the phonological environment, Chan's cluster simplification before vowels and pauses, something not possible in native varieties, was considered separately from simplification before consonants. As seen in Figure 2, Chan did not pronounce the stop in the coda at all in this environment in 1998, instead deleting and glottalising it equally. By the time he was filming Rush Hour 3, he deleted 71% of the final stops in clusters, only glottalised 12% of them, and used a branching coda 18% of the time. In the French interviews, Chan deleted 90% of his consonants; he showed no signs of glottalisation in these interviews. As in the case of the single coda, with native speakers he deleted and pronounced the final stop at a greater rate in 2007 than in 1998. Chan's cluster simplification before consonants is seen in Figure 3. Similarly to cluster simplification before pauses and vowels, Chan simplified the cluster 100% of the time in 1998; he deleted 64% and glottalised 36% of the final stops. Nine years later, Chan deleted 91% of the stops, glottalised 6%, and pronounced them 3% of the time when interviewed by an English native speaker. With the French interviewers, he deleted the stops 92% of the time and pronounced the stop without simplification 8% of the time. Here too, there is evidence of Chan using a more complex coda at a later period, whereby he exhibits a more native-speaker-like style of language production. Moreover, simplification before consonants has been found to be common among numerous native-speaker varieties also (cf. Wolfram 1997:115, Green 2002.

Discussion
The results show that Chan's English has changed between 1998 and 2007 across all variables. In general, he uses simplified forms less in the latter period and adheres more broadly to native-speaker norms. However, the interpretation of the results is complicated by the possible differing salience of the variables, most clearly visible in the results of the French interviews. The increases and decreases across the variables between the two periods are generally similar and may thus be caused by the same factors; on the other hand, the results of the French interviews are very different. For example, the deletion of the stop in the cluster occurs at a greater rate in the French interview than in the American one, whereas the deletion of the single coda stop occurs more rarely. This might be due to the relative salience of the variables, with the single coda simplification being more salient than the cluster simplification. If single coda simplification is more salient, it would result in Chan receiving a greater amount of explicit negative feedback (i.e., being corrected), leading to him taking greater care over its pronunciation in foreigner-directed speech. As cluster simplification happens in some varieties of English as well, it would draw less negative attention, and Chan may thus not concentrate on it while speaking with French interviewers. This possible difference in salience could help explain the results of the French interviews: issues of social identity are not considered as important when issues of mutual comprehension are at the forefront.
Nevertheless, despite the different rates, an overall development of Chan's further alignment with native speakers can be seen. This is most noticeable in the fact that the rate of simplification of the final stop has decreased by Rush Hour 3. In regard to the single coda, it is evident in the increase of 18% in his pronunciation of the stop between the two interviews. Similarly, a jump from 0% to 18% can be observed in the pronunciation of all consonants within a cluster before vowels and pauses. This reveals two insights into Chan's L2 development: the phonotactics of his interlanguage now allow a non-simplified branching coda; additionally, as he simplifies less before vowels and pauses, there is some evidence of an acquisition of native-like constraints concerning the environment. Although these jumps may not be extremely large, they nevertheless clearly show that after Chan's social network increased to include more English native speakers, his English developed overall to become more like that of a native speaker. Moreover, in the American 2007 interview, Chan directly mentions attempting to develop his English and voices his frustration over his failures: "Until now I try to get in American market, try to learn everything American, learn from Chris Tucker [his co-star in the Rush Hour trilogy]. I just cannot bring in. They speak so fast on the set. I just follow whatever they tell me to do." Although he has not improved as much as he had wished, he very clearly shows his desire for his English to be more like that of a native speaker and his efforts in this direction. Such motivation may have led to the change in pronunciation between 1998 and 2007.
However, Chan does continue to use non-standard variants influenced by his social identity; in fact, the amount of deletion across both variables increased between the interviews. In addition, when comparing the single coda variable in Chan's American interview to the interviews with non-native speakers, it becomes clear that the rate of Chan's alignment with native speakers cannot only be due to an imperfect acquisition of English. He is clearly not accommodating to and aligning with the English native speakers to the greatest extent possible. These results can perhaps best be explained by the decreased pressures of the linguistic marketplace. Chan no longer has to accommodate as much as his proficiency allows in order to break through in Hollywood; rather, he has already succeeded. He recognizes as much: "You know, I remember 20 years ago when I in Eiffel Tower. I have to wait on a line, squeeze in the small elevator. But now I have a private elevator! Wherever I go, people recognize me this way" (AlloCiné 2007). By 2007, Chan can afford to speak without monitoring the interference of his L1s. On the other hand, however, his success in American movies involves playing a very specific character-a Chinese martial artist. Perhaps, then, it is not that Chan no longer needs to draw on linguistic capital, but rather that he very much does so, emphasizing his Chinese identity by using certain non-standard variants to indirectly index his appeal as a foreigner, or a Chinese native speaker, in movies. Nonetheless, this interpretation is still related to his general success, as within these 9 years, he has been able to "flip" the linguistic market to his advantage so that his non-standard language is seen to have more worth. Either way, then, this development in his English depends on his social identity and his success.
A further potential identity-related development in Chan's English is the natural result of his increase in both the deletion and pronunciation of stops, namely, the decrease in glottalising. In 2007, his glottalisation across all variables and in all contexts was far lower than in 1998; the sole exception is the single stop variable in Chan's French interviews, which could perhaps be explained by him attempting to accommodate as much as possible by at least partially pronouncing the coda. A possible explanation for this choice to delete rather than to reduce could be a sort of emphasis of extremes. Some stops he would previously have glottalised are instead pronounced due to alignment with native-speaker norms; others may have been deleted as the greater lenition is an even more marked way to express his Chinese identity. Another, more tentative, explanation could be a shift in Chan's L1s. In the timespan between the interviews, he opened a film studio in mainland China. Additionally, he has spoken several times about associating Mandarin with Chinese ideology and Chinese people, and has begun emphasizing his identity as being Chinese, i.e., not specifically from Hong Kong, by frequently wearing Chinese clothing (JC News 2003). The shift in his social identity to emphasise the wider Chinese culture may have resulted in Mandarin becoming both more frequently used and more important to him than Cantonese. This would explain why the variant that is caused by interference from Mandarin, i.e., deletion, is in 2007 more prevalent than the one from Cantonese, i.e., glottalising.
Overall, Chan's L2 development displays the expected increase in alignment with native speakers but is also influenced by the desire to reinforce his social identity as a native speaker of his L1s, Mandarin and Cantonese. Certain features have become more aligned with those of a native speaker, but some non-standard variants are used to a greater extent, possibly due to the difference in competing pressures from the linguistic marketplace. Although there are several conceivable explanations as to the exact nature of the cause of change, and it is difficult to tease apart to what extent factors of social identity have a significant effect, it is nonetheless clear that they are important in L2 development. Thus, this study shows that it is necessary for any longitudinal study of bilingual speakers to take into account not only factors such as age of acquisition and social network but also the social identity of the people being studied.

Conclusion
To conclude, this longitudinal study examined the factors influencing a bilingual's acquisition of an L2. It aimed to find out to what extent the subject, Jackie Chan, further aligned with English native-speaker norms over 9 years due to changes in his social network and a desire to achieve success in the American market, and to what extent his identity as a Chinese national caused him to choose to continue using variants influenced by Cantonese and Mandarin. Both factors proved to play an important role: Chan produced stops in both single stop codas and consonant clusters to a greater extent at a later period, thereby showing his closer alignment with native speakers. However, he also deleted stops in both variables at a greater rate in 2007, and an analysis of the single stop codas in his foreigner-directed speech demonstrated that he was not, in fact, aligning to the greatest extent possible with American speakers; these two traits appear to be an indexing of his identity as a successful Chinese national. A further confounding factor is that by 2007, he appears to be emphasizing his identity as a person from China, and not specifically from Hong Kong, more than in 1998, leading him to use variants influenced by Mandarin more frequently than before. This brings up the interesting question of how talking in English to another person from China would influence Chan's pronunciation, and in particular, how this might change depending on whether the interlocutor's L1 was Cantonese or Mandarin. Another direction which builds on this study could be considering the effect of topic change: if Chan uses fewer non-standard variants when not talking about films in which he uses Chinese martial arts, this could strengthen the explanation that his usage of non-standard forms is to emphasise his foreignness in films. Overall, it cannot be claimed that factors of social identity influenced the development of Chan's L2 to a greater extent than other factors related to SLA. What is clear, however, is that the social identity of a person must be taken into account when constructing longitudinal studies on the acquisition of an L2.