Second language (L2) attrition refers to the decline of acquired L2 skills, which is typically caused by the insufficient use of the L2 (De Bot and Weltens, 1991), or the change of language environment which changes the dominant language used in daily life, thus results in the decrease of L2 use (Olshtain, 1989: 151). The person who suffers from language attrition is called an “attriter” (Nakagawea, 2013). Here are some predictions on L2 attrition which are relevant to the present study. Firstly, younger language learners are predicted to be more vulnerable to language attrition than older ones (Yoshida, 1988; Cohen, 1989; Olshtain, 1989). Secondly, lexical skills are suggested to be relatively less immune to atrition than other linguistic skills (Weltens & Grendel, 1993; Kuhberg, 1992; Moorcroft & Gardner, 1987). Based on these predictions, the present study investigateda child Chinese returnee’s L2-English attrition on productive lexicon and oral fluency over a period of 12 months.

Interference hypothesis (IH) serves to explain the attrition of one language in the language environment of another dominant language, which indicates that attrition is resulted from the increasing influence of the “newly dominant, competing language” (Köpke & Schmid, 2004; Bardovi-Harlig & Stringer, 2010). Although its validity may compromise considering about L2 users’ knowledge of multicompetence as proposed by Cook (1991), IH is indirectly supported by the findings in Seliger (1991), in which the subjects were found unconsciously process L2 input after a period of time without L1 input. The subjects were speculated to have replaced the complex L1 rules with simpler L2 rules on semantics which are similar to each other (Bardovi-Harlig& Stringer, 2010).

The process of language attrition, according to Berko-Gleason (1982), is similar to its acquisition, which exhibits the features of continuity, gradualness, and directionality. Although it is sometimes viewed as a procedure of “forgetting” about the langue skills, the curve of language attrition does not follow the traditional Ebbinghaus curve (Ebbinghaus, 1913; “Forgetting Curve,” 2014). Loftus & Loftus’ (1976) retrieval-failure theoryis widely applied to the explanation of language attrition. That is, due to the difficulty in the retrieval of the target information from memory, the attritional language skills are temporarily unavailable rather yet not disappear permanently (Loftus & Loftus, 1976; also see Weltens& Grendel, 1993). Bahrick (1984) agrees with this view and indicates that instead of being lost permanently from memory, an anttriter’s language skills are retained in a “permastore”.

Early studies claim that a fixed amount of knowledge will be lost irrespective of an attriter’s total amount of knowledge and level of language proficiency (Bahrick, 1984). The process of language attrition was predicted to follow a fixed. For example, regression hypothesis (RH) claims that the process of language attrition follows the reverse order of its acquisition. That is, the first learned language skills(s) attrite(s) last, whereas the last acquired language skill(s) attribute(s) first (Jakobson, 1968; Cohen, 1975; Hansen, 1999a, b; Hayashi, 1999; Hedgcock, 1991). RH is supported by findings in Cohen (1975), Hansen (1999), Hayashi (1999), Hedgcock (1991) and Taura (2008). However, counterevidence is available in other studies. For instance, Jordens, de Bot &Trapman (1989) examined German casemaking, which “meets the conditions of gradualness and a more or less fixed order of acquisition”. Yet, no evidence was found regarding the reverse sequence between L1 acquisition and attrition. More similar findings areavailable from Håkansson’s (1995), Tomiyama (2009) and Hedgecock (1991). Similar to RH, Myers-Scotton’s (2002) 4-M1 model also suggests that the process of language attrition follows a fixed order. However, contrary to RH, 4-M model notes that language attrition pattern follows the same process as its acquisition path. That is, the first acquired skills suffer first from attrition, such as the findings presented in Taura (2008).

The Activation hypothesis (TH), however, holds totally different views on this issue. According to TH, language attrition is irrelevant to the sequence of its acquisition. Instead, language proficiency is the key factor regardinglanguage attrition: the best learned skills are the most immune to attrition (Paradis, 2007; also see Bardovi-Harlig& Stringer, 2010; Neisser, 1984). In some studies, learners of higher language proficiency levels are found to have better retention than those of lower proficiency levels (Kaufman, 1995; Reetz-Kurashige, 1999; Nagasawa, 1999b). Hansen (1999) agrees with TH and indicates, “the more you know, the less you loss” (Hansen, 1999). However, in some studies, such as Weltens& Grendel (1993), Bahrick (1984), and Weltens (1989), the subjects’ language proficiency level was not found to be a decisive factor for their language attrition.

Moreover, it is revealed that there is a period of time before attrition is noticeable, despite that during the period of which the target language is disused or not dominantly used. It is viewed as a period of time before the “initial plateau”, or before the start of noticeable drop of language skills (Russell, 1999; Weltens& van Els, 1986). In Tomiyama (1999), for instance, the subject was observed to be able to speak L2-English spontaneously by the 19^th month of returning to Japan. The subject’s “initial plateau”, therefore, was concluded at this point. Meara (2004) depicts this period of time as a “silent buildup period during which loss is laying a foundation”. Individual subject may have different durations of time before the start of attrition (Meara, 2004). The subject in Kaufman and Aronoff (1991) showed rapid attrition by the 12^th month of observation. In contrast, the subject observed by Kuhberg (1992) displayed server attrition by the 15^th month. Bahrick (1984) points out that the language attritional curve follows the sequence of “initial-decline-then-plateau”. However, the two subjects’ “initial plateau”/ the end of the “silent builduppeirod” in Kuhberg (1992) was reported to be at the 6^th month, before which their fluency remained constant. After 2-months’ sign of initial attrition, however, a drastic decline of various language skills displayed. A similar pattern was revealed by Kaufman and Aronoff (1991), despite that the “initial plateau”/ the end of the “silent buildup period” of fluency was found to be at the 8^th month.

Oral fluency is frequently defined as the ease and flow of the student’s speech in comparison with native speakers(Harmer, 2006). It can be gauged with the criteria: (1) the degree of smooth, which displays in the number and durational time of pauses (halting); (2) the speed of speech; (3) breaks between words; (4) whether seems require great effort (Day &Shapson, 1987). Oral fluency can also be measured by (1) speech rate; (2) length of runs; (3) rate of articulation; and (4) hesitation devices (i e. stalls, repairs, parenthetical remarks) (Hieke, 1985; also see Ascione, 1993).

Oral fluency attrition is typically viewed as the implication of the start of language attrition as a whole (de Bot &Weltens, 1991). It can be observed in terms of pauses (include silent pauses, filled pauses—voiced fillers which do not contribute lexical information, such as uh, um, well in English), decline in the rate of speech, the amount of talk (verbosensess), selfrepairs, and repetition (Yukawa, 1998; Kuhberg, 1992; Nagasawa, 1999a). The L1-Turkish subjects in Kuhbert (1992), for instance, displayed oral fluency attrition in the form of gradually decline of speed in the speaking of Germany. Moreover, Beardsmore (1972) suggests that the strategies for the elaboration of sentence structures, the selection and insertion of lexical items are all significant aspects that show a language speaker’s degree of oral fluency.

The use of complementary strategies2 is widely viewed as another implication of linguistic attrition. It is widely applied to the study on lexical attrition. Nonetheless, considering that complementary strategies are the result of the increase in the length of time needed for the retrieval of certain language items, it may be at least in part caused by lexical attrition. Therefore, it would be necessary to examine an attriter’s complementary strategies, if any, in the study of oral fluency. Complementary strategies can be observed in different ways. For instance, Cohen (1989) investigated two children’s attritional performance after 1, 3, and 9 months of discontinued contact with L3-Portuguese. At least 6 lexical production strategies were observed:L1-based strategies, intralingual strategies and code-switching. Similarly, in Kuhberg (1992),two L1-Turkish participants switched some German words/phrases to Turkish by the end of the 15^th and 20^th month respectively.

Previous studies on language attrition obtained different results. The participants in some studies displayed more severe attrition in oral fluency than others (ie., Cohen, 1989; Tomiyama, 2008). Both linguistic (i e. linguistic influence from the dominant language) and extralinguistic factors (i e. age of the learner) are reported may display significant impact on an attriter’s attritional performance (Bardovi-Harlig & Stringer, 2010).

Language leaners’ age is widely accepted to be one of the most significant factors. Younger language learners are predicted to be less immune to attrition than older ones under the same or similar circumstances (Berman &Olshtain, 1983; Hansen, 1983; Yoshida, 1988; Cohen, 1989; Olshtain, 1989; Pillier, 2007). For instance, Berman & Olshtain (1983) found a greater quantitative and qualitative loss of L2-English among 5-to-8-year-old L1-Hebrew speakers than among the older children, despite they experienced similar termination of contact with English. Similarly, Fujita (2000) revealed that L1-Japanese returnees who were younger than 9 years old were less immune to L2-English attrition than those who were older than 9 years old. Moreover, in Cohen (1989), although the two subjects of different ages both displayed a rapid decline in the type and number of words in the production of L2-Portugues, the older subject displayed less dramatic drop. However, the validity of this study may compromise concerning it is lack of valid evidence in support of the two subjects’ comparable proficiency level of Portuguese (Tomiyama, 2008).

Nonetheless, the interaction between age and other variable(s) was revealed to be able to lead to different results. Hansen-Strain (1990) examined the attrition of Japanese skills by 4 L1-English children of different ages. According to the age-related attritional hypothesis discussed above, the younger subjects were expected to suffer from the most severe and rapid loss of language skills. Thus it would not be surprising to find that the oldest subject’s (9 years old) communicative ability remained constant over a period of 10 months after leaving Japan. However, the youngest subject (3 years old) was observedto have better retention than the other two subjects of older ages (4 and 7 years old). It was attributed to the younger subjects’ higher English proficiency level. Similar findings are available from Hansen &Chantrill (1999), in which the subjects’ achieved levels of literacy were revealed to be a significant predictor for their retention of oral skills.

Additionally, other non-linguistic factors, such as individual subjects’ attitude, motivation and personality were also found to be significant for their maintenance of a target language knowledge(Moorcraft & Gardner, 1987; Snow, Padilla, & Campbell, 1988; Nakagawa, 2013; Yoshitomi, 1999). For example, the subjects in Yoshitomi (1999) of different personalities showed variance in the use of communication strategies, as well as the choice of type and token of vocabulary items. Paraphrasing and progressive retrieval were found among the more talkative subjects. Moreover, subjects offavorable attitude and motivation were reported to have displayed less language attrition than those with negative attitude and motivation (Nakagawa, 2013).

14-M model divided morphemes into 4 types: (1) the content morpheme, such as nouns, verbs, adjectives; (2) the early system morpheme—the definite article “the”;  (3) the bridge late system morpheme—the copula “be”; and (4) the outsider late system morpheme—irregular past tense. According to 4-M, language learners are predicted to both acquire and lose their morphemes in the order of content, early system, bridge late system and outsider late system morpheme (Myers-Scotton, 2002).  2Complementary strategies in the research of language attrition refers to an attriter’s communication strategies used to compensate for their insufficient linguistic knowledge (Nakagawa, 2013).

The participant and the method employed in the present study are the same as that in Li (in press). However, the present study examined the participa nts’attritionalperformance in oral fluency, whereas Li (in press) focuses on the investigation of the participant’s attritional performance in productive lexicon and morphology.

The participant (Lin, female) had lived in Ottawa with her parents for 3 years (from 2 to 5 years old) before returning to China. Both her parents were college-educated. When she was in Canada, she spoke Mandarin at home but English in the kindergarten. Most kids in the kindergarten were native English speakers. Lin did not receive any formal instruction in English. After returned to China, the dominant language was Mandarin both at home and in the kindergarten, except 3 English lessons per week in the kindergarten (30 minutes per lesson). The content of the lessons was mostly lexical acquisition (i e., numbers, colours, memorized songs, etc.). Most of the time, the teacher used Mandarin for instruction. At the end of the 5^th month after returned to China, Lin went back to Ottawa for a 20-day summer camp with native English speakers. After that, she returned to China again.

Lin’s English proficiency could be native-like when she first returned to China. For instance, the informal report card provided by the kindergarten in which Lin studied in in Ottawa also indicated that she could listen, speak, read, and writelike native English speakers. Her parents noted taht Lin did not have any difficulty in the daily communication with her peers in the Kindergarten. She was also an active girl with favorable attitudes toward study.

This longitudinal case study was designed to investigate whether the change of dominant language environment led to Lin’s L2-English attrition in oral fluency.

Lin was followed for 12 months since 2 weeks after returned to China. Her L2-English proficiency was tested 6 times by the end of the 12^th month (see Table 2.1 below).

The study was conducted at Lin’s home. The investigator was a bilingual English teacher who worked at a University in China. She could speak English with native-like proficiency. The investigator only spoke English to Lin and her parent since the first time they met. Therefore Lin did not know the investigator could speak Mandarin. It was intended to elicit her best English production in the study. According to the investigator, Lin answered the questions of the investigator spontaneously and fluently in English.

Language attrition is predicted can be measured by the negative change of the attriter’s linguistic performance in quality and/or quantity, which includes “from presence to absence, complex to simple, easy to difficulty, native-use to non-native use, avoidance, and overproduction” (Tomiyama, 2000). Therefore, Lin’s performance in the 6 tests was compared to detect her L2-English attrition in oral fluency.

Storytelling tasks were reported to be suitable for the assessment of children’s oral fluency (Bardovi-Harlig& Stringer, 2010; Cohen, 1989). It was employed in the present study. The picture book Frog, Where Are You? (Meyer, 1969. see Appendix), which is widely used in child language attrition research was adopted in the present study. This book has various settings in 24 pages of detailed pictures without any writing text. Lin was asked to depict a story according to the pictures as well as she could. She was given as much time as she needed for preparation. There was no time or length limitation in this task. Each time, the investigator played a little bit with her before the start of the task, so that to ensure the test could be carried out in a friendly atmosphere. Each session lasted about 30 minutes, which was auditorily recorded and transcribed.

For the measurement of the collected data, similar to that inNakagawa (2013), the dysfluency rate was calculated: the duration of inter-sentential pauses (pauses occur between sentences) and intro-sentential pauses (pauses occur in the middle of a sentence) were counted. Moreover, complementary strategies, such as code-switching were noted.

As shown in Fiture 3.1 above, Lin showed the lowest dysfluency rate in test1, and similar dysfluency level in test4, while relatively higher dysfluency rate in test 2, test 3, test 5 and test 6. It seems the longer Lin left Canada, the higher the dysfluency rate was observed. Repeated-measures ANOVA illustrated this speculation (F=3.787; p=0.019).

Lin was also observed to employ complementary strategies in the storytelling task. For instance, self-repairing was typically viewed as another symbol of dysfluency in oral English. No self-repairing was found in the first test. In the second test, however, Lin repaired herself in 2 different sentences (There are (pause) there were some frogs. There’s a frog behind (pause) um (pause) behind a tree (pause) behind a log.). There was one self-repairing sentence in the test 4. More self-repairing sentences occurred in test 5 (n=5) and test 6 (n=7). Moreover, code-switching from English to Mandarin in test5 and test6 companied by the occurrence of pauses in the middle of a sentence. The first code-switched word was found in test 5 (English antler was code-switched to Chinese Lingyang). As a noun, antler itself is not frequently used in daily life, particularly in a Mandarindominant speaking environment. An interesting finding was that Lin generalized antler with animal in the following narration of the study in test5 and test6. As a complimentary strategy, this finding is identical with that in Cohen (1989). Additionally, some of the words used in test1, 2 and 3 were not found in test 4, 5 and 6, such as log and forest. Most of the “missing words” were nouns.

The present study examined a young child returnee’s attritional characteristics of oral fluency. As evident in the baseline data (test1), Lin did not have any difficulty in expressing herself, despite there were few grammar mistakes. Although it was hypothesized that an attriter with favourable attitude and motivation may suffer less from attrition (Nakagawa, 2013), unfortunately, Lin displayed noticeable attrition very quickly after returned to China for 3 months. Her positive attitude towards English learning and cheerful personality seem did not help her retention of these skills as predicted by Yoshitomi (1999) andNakagawa (2013). Noticeable decline was observed in terms of the increased dysfluency rate from test1 to test3, with slightly decrease in test4, whereas followed by more severe and continuous decline in test 5 and test 6. The length of time that she left Canada was found displayed significant effect on her attritional performance in oral fluency.

These findings may have confirmed the hypothesis of IH (see the discussions in Bardovi-Harlig& Stringer, 2010). That is, due to the influence of the “newly dominant competing language”, which was Mandarin, Lin suffered from L2-English attrition, at least in the aspect of oral fluency. Findings on her code-switching from English to Mandarin words may have illustrated it. The trend of the attrition also seems to exhibit the characteristics of gradualness and directionality as predicted by Berko-Gleason (1982). It may be because after returned to China, Lin lost the change to use English on a daily basis. She could hardly maintain the acquired English knowledge from the English classes in China, which were at beginners’ level, and most of the time, the English teacher used Mandarin as the introduction language to teach. However, there were some words which were supposed to be frequently used, yet were code-switched to Chinese, such as he and then. Nevertheless, these mistakes only occurred once in the test. Lin correctly produced these words in the preceding and following content. Therefore, these code-switched words can either be viewed as the slip of tongue, or the embodiment of the increasing influence of L1-Mandarin on her production of L2-English on sentence level.

In addition, Lin’s “initial plateau” (Russell, 1999; Weltens & van Els, 1986) or “silent buildup period” (Meara, 2004) seems to be at the end of the 3^rd month after returned to China, which was much quicker than those found in previous studies (e g., Kaufman and Aronoff, 1991; Kuhberg, 1992; Bahrick,1984). For instance, in Tomiyama (1999), the subject’s “initial plateau” of L2-English was found at the end of the 19^th month after returned to Japan. This finding is consistent with the prediction that indiivusl subjects may have different time of durations before the start of language (Meara, 2004). Considering that Lin

Another significant finding of this study is that in test3, before which Lin went back to Canada for 20 days, her oral fluency displayed improvement. During the 20 days in Canada, Lin interacted with native English speakers, thus had the chance to use English on a daily basis. As predicted by the retrieval-failure theory, instead of losing the acquired language skills permanently, language attrition is in fact the temporary difficulty in the retrieval of the acquired language skills (Loftus & Loftus, 1976; Weltens & Grendel, 1993). Moreover,the interaction between age and other factors, such as an attriter’s proficiency level of a target language, his/her attitude, motivation, and even personality, were all revealed may display significant effect on his/her attritional performance. Lin’s native-like proficiency level, positive attitude towards English study, as well as cheerful personality may all have contributed to her improvement in the storytelling task after returned to China from the 20-day summer camp in Canada.

Additionally, the age factor may have played essential effect in her attritional performance. It is predicted that learners are both vulnerable to language attrition (Berman & Olshtain, 1983; Hansen, 1983; Yoshida, 1988; Cohen, 1989; Olshtain, 1989) and have advantages in second language learning (see Lenneberg’s hypothesis of CPH). Given that the subject Lin was a young child (5 years old), her rapid attrition as well as fast reactivation of the language skills displayed in test3 confirmed the two predictions. Nonetheless, due to lack of comparison between Lin and older participants, it was not sure whether subjects of older age perform differently from Lin.

One of the limitations of this study is that the same testing material was repeatedly used in the 6 tests, Lin’s performance in the latter tests may have been benefited from the repeated testing experience. Nonetheless, it was unclear whether and/or to what extent Lin has been benefited from the repeated testing experience. Moreover, the disadvantages are outweighed by the advantages of using the same testing material, because changing stories introduces many uncontrolled variables.

This study examined a young Chinese returnee’s (Lin) L2-English attritional performance in the aspect of oral fluency over a period of 12 months. The results indicate, Lin’s proficiency in productive lexicon and oral fluency declined at the end of the 3^rd month after returned to China. However, after a 20-day stay in Canada in the 6^th month, some improvement was observed in the aspect of oral fluency. Nonetheless, with the time increased after left Ottawa, more severe decline was occurred in the following tests. The duration of time that she left Canada was found to be statistically significant for her attritional performance.