Can be identified if they appear together with a META FRAME
5. INSERT: IMPLICIT ELICIT
Occurred as eliciting signals ‘how do you say this?’
6. INSERT: NON- ELICIT
Can be identified in cases of missing vocabulary, occasional access blockings, the nature of the topic or context and the attitudes of the speaker
7. WIPP (Without Identified Pragmatic Purpose)
Constitutes short elements, in most cases grammatical function words, such as pronouns, prepositions, connective adverbs and conjunctions, rather than content words.
Table 6.4 Types of code-switches according to William and Hammarberg (2001:26)
In addition, William and Hammarberg (1998) did not establish categories for the roles of background languages, but explained their functions during the conversations. The language that supplies material for the learner’s expressions in L3 has been identified as supplier language. The authors, go on to explain that the primary supplier language is L3 itself and was referred as primary supplier language, while the background languages were referred as external supplier languages. L1, as a dominant language, was identified to have an instrumental role because it dominated in various pragmatically functional language shifts and supported the interaction, therefore it functioned as external instrumental language. On the other hand, the role of the L2 was found to have a prominent supplier role in the learner’s construction of new words and was identified as an external supplier language. The differences between the role of two languages is that the external instrumental language is supplementary to the utterances in L3, while the external
supplier language contributes to and influences these utterances in the formulation process.
The analysis that will be presented in this study is based on Hammabrg’s model of code-switching categories and the role of background knowledge in L3 production.
6.3.1 Methods
The code switching analysis of bilingual Albanian learners of English is based on the recording spoken data in a classroom. First, students were told that the purpose of the conversation was to have a refaxed informal conversation. Second, they were instructed to discuss the text they had learned in the previous class. As the discussions are usually led by the teacher, in this case they were asked to lead the discussion by themselves, which appeared to be difficult for them as they were not ready to communicate in English only. Having this argument, the students were told to communicate freely with each other by using their background knowledge or teacher’s support. Finally, I did the recording while the students were talking about the text “Formation of Gender Roles” which they had already read before. The recording was done twice, once with the group of Low Bilinguals and next with the High Bilinguals. The recording lasted fifteen minutes and the interaction among the students and the teacher were recorded. Finally, the data were analyzed taking into account the situation that triggered the code-switching.
6.3.2 Resutls
The conversations were transcribed and then analyzed quantitatively and qualitatively. The two fifteen minutes recorded conversations resulted in 197 word tokens in total of which 58 make up code switches from L1 and L2. There were also some words, which could not be transcribed due to background noise on the tape.
Considering the code-switching categories, the transferred items from L1 and L2 were analyzed to find out the differences between L1 and L2 in L3 production. Table 6.5 shows the number of influence from L1 and L2 over the seven categories identified by Hammarberg (2001).
Table 6.5 Quantitative overview of code-switches in L3 production
The text corpus of recorded conversation comprises of 197 word tokens, of which 58 make up word types (code-switches) or an influence index of 0.50 from both bilingual groups of English learners.
The results obtained from this analysis indicate that L1 influence index of code-switches in LB is 0.39, while L2 influence is 0.19. In the other hand, influence index of code-switches in high Bilinguals indicate L1> 0.21 and L2 > 0.24. These findings show that the influence index of code switches from L1 in Low Bilinguals is twice higher than L2, while the influence index in high Bilinguals is almost the same.
Adding the point of influence index across the groups, the general results of code switching category display the difference between two bilingual groups.
The influence index of code switching in LB is 0.58 and HB is 0.45 or an approximate ratio of 5 to 4 which represents the difference between the influences of backround languages in L3 production. As a whole, considering the different number of participants in the study, LB=48 and HB=67 or an approximate ratio of 4 to 6, and the difference of influence index ratio (5:4) per group, it can be suggested that High Bilinguals code-switched less than their peers which leads them in a slight advantage during the third language learning.
Pursuing the general results further to frequency of categories and their distribution over L1 and L2, the results show that both groups rely more on their first
Low Bilinguals High Bilinguals
Category n L1 L2 L1 L2 EDIT 12 6 - 2 4 META COMMENT 5 3 - 2 - META FRAME 6 2 - 2 - INSERT EXPLICIT ELICIT 13 2 3 4 5 INSERT IMPLICIT ELICIT 10 2 2 4 3 INSERT NON-ELICIT 9 3 3 - 3 WIPP 3 1 1 - 1 Total number of CS 58 19 9 14 16 Total number of CS influence index across the groups
0.50
0.39 0.19 0.21 0.24
language. To begin with the EDIT category, the influence index in LB from L1 is 0.12 and no switches from L2 were found. In comparison to HB, the influence index from L1 is 0.02 while from L2 is 0.05.
In the second and third category, the meta categories, switches came from Albanian only. The influence index in LB is 0.10 while in HB is 0.05. By contrast, the insert categories appeared to be more frequent from L2 in both groups. The influence index of LB from L1 is 0.14 while from L2 was 0.16. In comparison to the influence index of HB, L1 is 0.11, while L2 is 0.16. These results represent higher influence from L2 in both groups, and at the same time they indicate that L2 (0.16) influence in HB is stronger than L1 (0.11).
The last categories, WIPP switches, were the least frequently used, and the influence index in LB is L1 > 0.02 and L2 >0.02 which shows the minimal use of both languages. In comparison to HB, the influence index shows no use of L1 and minimal use of L2 > 0.01.
6.3.3 Findings and discussion
In the following discussion the occurrences of code-switching will be discussed, based on Hammarberg’s model of code-switching categories such as EDIT, META with two subcategories: COMMENT and FRAME; INSERT with three subcategories: EXPLICIT ELICIT, IMPLICIT ELICIT, NON-ELICIT, and WIPP (Without Identified Pragmatic Purpose). The EDIT constitutes code-switches with self-repair elements, META categories constitute metalinguistic elements, and INSERT categories have been identified as primary contents of conversation (not editing or metalinguistic elements). These categories have been interpreted as having a pragmatic purpose and the speaker has no attempt to use them in the L3, while WIPP (Without Pragmatic Purpose) category has been identified as part of utterance in formulation in L3.
EDIT
This category comprises switches constituted from editing elements in self repair or in managing the interaction. Such elements of code-switching were found by bilingual Albanian students in moments where they could not recall for the exact word in English, or they were not sure if the English word was equal to the L1 Albanian. The following excerpt illustrates such speaker’s self-repair elements.
Excerpt 1
This example shows sequences of self-repair, the student used a very common L2 Macedonian term such as виц /vic/, (line 1) for the L3 English words ‘joke’. The student code switched to L2 because of the lack of or unfamiliarity with terms in L1 Albanian. The L1 Albanian term amvise (line 3) occurred as access blocking for the original L3 English word ‘housewife’.