4 Findings and Discussion
4.1 Findings by Research Instruments
4.1.2 Findings from the Think-Aloud Protocols
Some of the data presented in this section have been previously published in the International Journal of Bilingual Education and Bilingualism as a partial report (Boksay Pap, 2016) while the data analysis process was still in progress. Certain parts of the already published content are presented here unchanged. Other parts, for instance some of the core codes, have undergone some reformulations and modifications during the process of data interpretation—a common phenomenon of the coding process (Charmaz, 2003; Kelle, 2007;
Strauss & Corbin, 1998). As a result, the present work contains these changes along with verbal protocol excerpts that have not been published elsewhere. The transcription of the think-aloud verbal protocols follows the transcription guidelines set by Van Someren, Barnard and Sandberg (1994) in which no capital letters are used so as to speed up the transcription process.
As I have indicated in the Methods Section, the premise behind the think-aloud procedure was that it is one of the research strategies belonging to the hypothetico-deductivist arsenal, which makes possible the study of some important observable manifestations of a series of mental activities which cannot be directly observed (Camps, 2003b; Ericsson & Simon, 1984; Leighton, 2017). My assumption is that in the light of the current state of scientific knowledge the verbalizations which surface during a think-aloud process, without offering a complete picture, are the best available windows into individuals’
thoughts during a problem solving task (Libben & Jarema, 2002). As is seen in what follows, the analysis of data collected during the think-aloud process revealed relevant patterns of reliance on different languages, language alternation, and strategies used to create the L3 narrative composition.
Table 9. The number of Instances the Three Languages L1 Hungarian, L2 Romanian, and L3 English Were Relied on
This section has two main parts. The first part presents the findings connected to the languages on which the participants relied while composing the L3 narrative. In this part, the findings regarding language reliance are presented with their connection to the text-composing strategies employed by the participants in the process of creating the L3 narrative based on wordless pictures. In the second part, the findings regarding language reliance are presented with their relation to alternations between L1 Hungarian, L2 Romanian and L3 English. The presentation of the findings is accompanied by a discussion focusing on the relevance of the findings to the study from the point of view of the research questions and from the standpoint of theory construction.
4.1.2.1 Reliance on Languages and the Strategies Employed in the Composition of the L3 Narrative Text
In the data analysis, reliance on different languages while composing the L3 text emerged as a category with three main dimensions: reliance on L1 Hungarian, reliance on L2 Romanian, and reliance on L3 English. Table 9 displays the number of instances the languages L1 Hungarian, L2 Romanian and L3 English were relied on during the composition of the L3 English narrative along with the strategies adopted by the participants in the text-composing process.
Because the primary focus of the research is reliance on languages and access of non-native language lexical units, the text-composition strategies identified in the think-aloud protocols were interpreted as characteristics of each of the three dimensions of the category reliance on languages while composing the L3 text. The resulting axis, shown in Figure 20, is a hierarchical one in which, in a top-down arrangement, the category reliance on different languages while composing the L3 text occupies the uppermost position and it is described by the three dimensions: reliance on L1 Hungarian, reliance on L2 Romanian, and reliance on L3 English. Each dimension has several characteristics represented by the text-building strategies employed in the composition process. These strategies appear enumerated in the first columns of Table 9 and will be discussed under the relevant dimension.
4.1.2.1.1 Reliance on L1 Hungarian
The first dimension of the category language reliance while composing the L3 text is reliance on the L1 Hungarian. Its characteristics are the following:
- 1. Identifying the task, - 2. Setting goals,
- 3. Justification of paths taken, - 4. Linguistic assistance, and - 5. Creating meaning.
These characteristics are detailed below, and their discussion is supported with excerpts from the participants’ verbal protocols.
As shown in Table 9, participants relied on their L1 Hungarian primarily in metacognition, that is, in the process of planning and monitoring the creative processes of crafting out the linguistic text (Flavell, 1976; Olsen Stevens, 2017; Quigley et al., 2011). We know from research in the areas of development of reasoning ability, language learning, and inner speech that metacognition is intimately linked to inner speech, the latter fulfilling essential roles in “mental orientation, conscious understanding, (…) overcoming difficulties”
(Vygotsky, 1934/2012, p. 242). We also know that both metacognition and internalised Figure 20. Hierarchical Axis of the Category LANGUAGE RELIANCE, its Dimensions and Characteristics
speech, as higher psychological processes, have their roots in social interaction, in the social, collective discourse (Proust, 2013; Vygotsky, 1978; Wertsch, 1979). In light of these premises, it comes as no surprise that the language in which we learn to think, in the overwhelming majority of cases our L1, is the language we rely on in our metacognitive activities. In the case of the Transylvanian-Hungarian L1 participants, L1 Hungarian fulfilled several roles in cognitive control and monitoring.
4.1.2.1.1.1 Identifying the task
Identifying the task took place, in the case of all 12 participants, exclusively in L1 Hungarian. Excerpt (17) shows such a case. English translations are given in square brackets and follow the Hungarian transcription. The numerals at the left hand margin represent the number of the lines within the larger transcription unit. The initials of the participant are given at the end of the excerpt in round brackets:
(17)
3 a képek alapján ez egy mese [based on the pictures this is a story]
4 egy kutya…és egy macska [a dog …and a cat]
5 aki együtt laknak a tulajdonosukkal [who live together with their owner]
6 kellene egy bekezdés [an introduction would be necessary]
7 valami…[something…]
8 once upon the time ha jól emlékszem [once upon the time if I remember well]
(Verbal protocol excerpt; participant: B.A.)
As we can see in Excerpt (17), the participant identified the genre of the narrative he had to create, briefly took stock of the characters in the pictures, reflected on the L3 English target expression usually used to introduce stories, and made a comment on his ability to retrieve the right expression.
Excerpt (18), an excerpt from the verbal protocol of participant S.B., who was attending education in L2 Romanian at the time of the data collection, illustrates similar movements. The participant identified the type of the narrative he had to compose, he evaluated the length of the series of pictures, and the scenes they showed, expressed his attitude towards the difficulty of the task and retrieved an L3 English expression with which to start the story:
(18)
4 na ne ez komoly [is this a joke?]
5 most mesét kell írnom? [do i have to write a story?]
6 elég sok kép van [there are quite a lot of pictures]
7 egy öreg asszony és az állatai [an old lady and her animals]
8 jó ezzel gyorsan megleszek [ok I will finish this in no time]
9 szóval ez egy mese [so this is a story]
…
12 úgy kezdődik hogy there was [it starts with there was]
(Verbal protocol excerpt; participant: S.B.) 4.1.2.1.1.2 Setting goals
L1 Hungarian was also relied when setting goals to carry out the narrative task. Setting goals appeared as a recurrent metacognitive movement throughout the whole composition process, and it took the form of procedural comments, recapitulation of what had been already performed with the aim of planning of what had to be still carried out. Excerpts (19), (20), (21) and (22) illustrate different instances of such metacognitive movements. In Excerpt (19), participant K.J. planned what he had to include in the following section of his composition:
(19)
34 most arról kell írjak [now i have to write
35 hogy összeszedi a recepthez valókat [that she gathers the ingredients]
(Verbal protocol excerpt; participant: K.J.)
In Excerpt (20) participant P.L., having realized that he did not know a certain word in L3 English planned to turn to the dictionary:
(20)
67 ezt nem tudom angolul [i don’t know this in English]
68 meg kell keressem a szótárban [i have to look it up in the dictionary]
(Verbal protocol excerpt; participant: P.L.)
In Excerpt (21), participant F.L. planned how she would undertake the narrative composition task and decided on the length of the introductory part:
(21)
16 először egy bekezdést írok [first i write an introduction]
17 utána majd meglátom [then i’ll see]
18 legyen három kép a bekezdés [let’s have three pictures for the introduction]
(Verbal protocol excerpt; participant: F.L)
In Excerpt (22), participant N.A. evaluated the amount of text created for a certain picture in the series, was not content with how much he had written, and then remarked that what he had written had to be connected to the next section, and started re-examining the pictures:
(22)
49 ehhez a képhez még nem írtam eleget [i haven’t written enough about this picture]
50 össze kellene kötni a következővel [it should be linked to the following one]
51 na, lássuk csak mi van itten [well, let’s see what we have here]
(Verbal protocol excerpt; participant: N.A)
Previous research studies have already shown that along with the metacognitive strategy of task-identification, other metacognitive strategies, such as procedural comments, evaluation of already performed work and planning of further steps in problem solving are best served by the L1 of participants who know and use more than one language. Bilinguals’
and second language learners’ reliance on the L1 in metacognition related to solving mathematical word problems—numerical thinking—has been documented by Celedón-Pattichis (1999) and Vaid and Menon (2000). Reliance on the L1 in the case of highly fluent bilinguals’ and professional translators’ metacognitive processes has been researched by Rudd and Fraser (2000) in relation to linguistic problems. A consistent finding in these studies was that metacognitive processes seem to be supported best by the multiple-language-user participants’ L1. However, Vaid and Menon (2000) identified that L2 was also relied on in metacognition by those participants whose language of instruction was L2 from an early age. Vaid and Menon’s (2000) findings, although based only on self-reported data, are noteworthy as one of the few available empirical results on using a foreign language in numerical thinking and because they raise the issue of the language of early instruction as a potential factor in influencing multilinguals’ language choice in higher level mental activities.
Rudd and Fraser’s (2000) findings related to the use of L1 in metacognition during translation activity are also noteworthy as they indicate that the task’s relative difficulty is another factor which has the potential to affect the language choice in higher level mental activities. The researchers found that in the case of the L2 linguistic tasks which were experienced as being relatively difficult by the participants the amount of metacognitive talk increased and was carried out in the professional translator participants’ L1.
In the present study, the findings related to the group of Transylvanian-Hungarian L1 participants are in line with the aforementioned studies regarding the factors of language of early education and of relative task difficulty. The language of early schooling in the case of all 12 participants was L1 Hungarian, with six participants enrolling in mainstream L2
Romanian education at the age of 14, and, as it has been already discussed, all participants used their L1 in the metacognitive processes. In addition, when the participants faced challenging sub-tasks, such as naming in L3 English an object or activity with which they were less familiar or unfamiliar, the amount of L1 metacognitive talk increased as they commented on what step to follow, recapitulated what they have done and planned what to do next. Illustration of such metacognitive talk in L1 is given in Excerpt (20), where participant P. L. decided to use the dictionary for looking up an L3 English word. An example in which L1 Hungarian was employed as a result of the perceived difficulty of the task to be carried out is Excerpt (23). Participant N.A. was unfamiliar with the process of turning milk into butter (see pictures in Appendix A). As a consequence, he had to interpret the situation shown in the pictures without having the possibility to access previous knowledge and experience (Cavanagh, 2011; Hashtroudi et al., 1984). His attempts to make meaning of what he saw were accompanied by L1 Hungarian metacognitive talk while he evaluated the already performed sub-task (line 54), tried to retrieve an L1 Hungarian word without success (lines 57, 58, 59, 60, 64, 65, 66), decided to turn to the Hungarian-English dictionary (line 63), abandoned the L1 Hungarian-L3 English route and turned to L2 Romanian (line 67), and came up with the L2 Romanian word. Then he seemingly gave up his undertaking to find the L3 English word (lines 69, 70), to finally return to his search (line 71), and decided to follow the L2 Romanian-L3 English route to find the target word:
(23)
54 most már megvan a teje [now she has got her milk]
55 nem tudom mit akar vele [i don’t know what her intention is with it]
56 kiönti [she pours it]
57 de mibe? [but into what?]
58 nem is tudom mi ez [i don’t know what this is]
59 mi a neve [what its name is]
60 mi a neve? [what is its name?]
61 she puts the milk in the 62 in the
63 hol az a szótár? [where is the dictionary?]
OPENS THE HUNGARIAN-ENGLISH DICTIONARY 64 na mi a neve magyarul? [well, what is its hungarian name?]
65 mit keressek? [what should i look for?]
ADDRESSES THE RESEARCHER
66 nem tudom mi ennek a dolognak a neve magyarul [i don’t know what this thing is called in hungarian]
/…/
67 románul [in romanian]
68 ez egy castron [ROM] [this is a CASTRON]
69 kész ezt kihagyom [enough i leave this out]
70 nem írom bele [i don’t write about it]
71 de mit csinál? [but what is she doing]
72 hol a román szótár? [where is the romanian dictionary?]
(Verbal protocol excerpt; participant: N.A)
Excerpt (23) also captures one of the several instances of translanguaging, defined as “the deployment of a speaker’s full linguistic repertoire” (Otheguy, García, & Reid, 2015, p. 281) and the presence of more than two languages within a larger stretch of discourse.
Translanguaging phenomena contain valuable information with respect to the organizational principles of the MML and search processes at the word-meaning level. The issue of translanguaging is readressed in the present section in the discussion of Creating Meaning in Hungarian.
4.1.2.1.1.3 Justifying the adoption of different paths
A last group of metacognitive strategies carried out in L1 Hungarian was that of justifying why one or another path had to be taken during composition. This type of metacognitive talk materialized in the form of explanations directed seemingly to oneself while in fact it was targeted towards the researcher. From a research-logistics point of view, it is worth mentioning that these explanatory comments appear to have been triggered by the fact that the researcher was present—as a silent-observer—in the room where the think aloud activity was carried out. I interpreted this kind of metacognitive talk as one possible result of the “locational constraint” (Otheguy et al., 2015, p. 259), that is, a behaviour which was determined by the context of the think-aloud activity (and the whole research). As a result of the presence of the educational context, most participants indirectly positioned themselves towards the researcher in a talking-listening space characteristic to the intersubjective situations of learning activities (Lin, Chen, & Chen, 2012) in which teachers are providers of information. One such example of positioning is found in Excerpt (23), where the participant could not come up with the L1 Hungarian word to name a thing in the picture and addressed the researcher in an attempt to get assistance. Another example of positioning is captured in Excerpt (24). Participant S.B. explained why he would omit certain details from the composition and addressed the researcher to ask for her consent.
(24)
76 itt ez egy hosszú dolog [this one here is a long thing]
77 hogy elgondolja mit fog csinálni [where she thinks about what she is going to do]
78 de nem fogok mindegyikről írni [but I won’t write about each one]
79 mert az túl hosszúnak tűnik [because that seems too long]
80 csak összefoglalom [I only summarize it]
ADDRESSING THE RESEARCHER 81 jó? [ok ?]
(Verbal protocol excerpt; participant: S.B.)
Providing justification for taking one step over another in the process of composition was clearly the result of the researcher’s presence, and it did not contribute directly to the creation of the L3 English text. Nonetheless, the Hungarian utterances were coded and included in Table 9, and justification was taken to be a distinctive trait of a problem-solving process in the presence of a silent observer (Holsánová, 2008; Leighton, 2017).
4.1.2.1.1.4 L1 Hungarian as a linguistic assistant
In addition to metacognitive talk, L1 Hungarian was also employed by the participants in the process of retrieving L3 target lexical units. The instances in which a participant attempted to access an L3 English lexical unit starting from an L1 Hungarian one were coded L1 Hungarian linguistic assistant. These instances comprised all those situations in which a lexical unit of the L1 Hungarian was present as a starting point for the L3 English lexical unit search.
Attempting to create content in L3 English was the typical situation for the participants, and only in those cases in which the participants experienced difficulties in producing the L3 target lexical units did they employ the support of either L1 Hungarian or L2 Romanian. The role of L1 Hungarian as a linguistic assistant, or support language, was meagre as it can be seen in Table 9. There were only a few instances when the participants deployed their L1 Hungarian to prompt the retrieval of an L3 English target lexical unit. This finding runs counter to the conventional wisdom that it is the native language that acts as a primary support to the other languages acquired subsequently. Studies with multilingual learners concerned with the use of multiple languages in the process of learning an L3 have documented that there are three potential factors which influence the learners’ deployment of particular linguistic resources as support to additional languages, namely:
- the typological proximity between the language(s) known and the language to be learnt
- the proficiency level in the language(s) known and the language to be learnt, and - the method of learning.
The discussion of the findings relative to these three factors starts here, and it is taken forward in the following sections of the present work.
Typological proximity, known also as linguistic distance, is the degree of relatedness between languages. It has been shown that language learners, in the process of learning a foreign language, tend to rely on a previously known language if this is closely related to the language to-be-learnt (Ahukanna et al., 1981; Ecke, 2001; Fouser, 2001). Based on the influence of the typological-proximity factor, the participants’ reduced reliance on their L1 Hungarian as a support language in supporting the retrieval of L3 English target lexical units could be explained by the considerable language distance between L1 Hungarian, a Finno-Ugric language, and L3 English, an Indo-European language.
A second factor influencing the deployment of a linguistic resource as a support language in the learning of another is the learner’s proficiency level in his languages.
Research into bilingualism and SLA has shown that in the early phases of acquisition of a target language, when language command in that language is still incipient, reliance on another language as a support language is strong. Both the native language and non-native languages in which the participants were fluent have been documented to be relied on as support languages in the early phases of acquisition of an additional foreign language (Aronin & Toubkin, 2002; H. C. Chen & Leung, 1989). Interestingly, even non-native languages in which the participants were less fluent have been documented to be present as support languages to a non-native language in which the participants had a reduced level of proficiency (De Angelis, 2005).
From the proficiency level factor standpoint, the participants’ reliance on support languages can be explained by their levels of proficiency in English, as the participants’
command of L3 English was situated between low-intermediate and intermediate on the proficiency continuum. However, the different degrees of reliance on the two languages—the L1 Hungarian and the L2 Romanian—do not find their explanation in the tenets of the proficiency-level factor, as the participants were highly fluent in both these languages. As a consequence, their reliance on L1 Hungarian and L2 Romanian as support languages should have produced a more balanced pattern. However, this was not the case as the participants seemed to have relied more on their L2 Romanian in the process of retrieving L3 English words. The problem of the unbalanced reliance is further discussed in the section Reliance on L2 Romanian, and some tentative explanations are presented.
The method of learning the vocabulary of a second language has been acknowledged to play a role in whether and how the native language of the learner is relied on in this
process. The effects of different methods of teaching and learning on how the lexis of particular languages are relied on during the process of learning and later on have received some attention in SLA and bilingualism studies. To my knowledge, the effects different language teaching methods and learning strategies have on the reliance on different languages in multilingual contexts have not yet been explored. In terms of the methods of learning a foreign language, Singleton (1999) pointed out the important role the previous language acquisitional experience plays in the learning process of an L2, both in naturalistic and formal instruction environments. Obviously, in this situation, the learner can make use of his native-language acquisition experience; whereas in the situation of any additional language beyond an L2, the learner has a richer language acquisition experience to make use of. Singleton (1999) holds that acquiring the lexis of an L2 takes place quite similarly in the naturalistic and the formal instructional environment, and that the learners benefit from the streamlined language input they get from their interlocutors. In Singleton’s (1999) view, the only aspect in which the process of learning a new language differs in the naturalistic environment and the formal instructional environment is the way words become encoded in memory. In naturalistic learning environments, learners have to isolate the L2 lexical units on their own from the input they get. These lexical units then have the chance to become connected to the relevant lexical units of the L1.
By contrast, in formal instructional contexts of learning, the learners are usually presented with the lexical items of an L2 in an “atomistic” (Singleton, 1999, p. 50) fashion, that is, discrete L2 words appear along with their L1 translation equivalents. This way of dealing with vocabulary, Singleton (1999) remarks, makes the L2 words easily connected to their L1 counterparts. Singleton’s (1999) observation on how different methods of acquisition of an L2 can result in varying degrees of lexical connectivity between the L2 lexis and the L1 lexis is a valuable starting point in the present discussion of the role played by L1 Hungarian as a linguistic assistant.
As mentioned earlier, the language of education for six participants in the study was L2 Romanian. These participants were learning L3 English in a formal educational context in which the L3 English and L2 Romanian lexical units were being presented in the atomistic way mentioned by Singleton (1999, p. 50). Additionally, L2 Romanian was employed as a scaffolding language during L3 English lessons. Thus, these participants’ diminished reliance on L1 Hungarian as a linguistic assistant in accessing and retrieving L3 English lexical units during the building of the L3 composition could find its explanation in the method-of-acquisition factor. Conversely, in the case of the other six participants, who had L1 Hungarian as a language of education and the scaffolding language during the L3 English language lessons, the limited presence of L1 Hungarian as a linguistic assistant cannot be explained based on the method of acquisition factor.
The ways in which these three factors shape reliance on support languages in the retrieval process of a learnt but difficult-to-retrieve L3 target word are re-addressed in Sections 4.1.2.1.2 Reliance on L2 Romanian and 4.1.2.1.3 Reliance on L3 English.
4.1.2.1.1.5 Creating meaning in L1 Hungarian
Another characteristic of the dimension reliance on L1 Hungarian was creating meaning, or generating ideas. Creating meaning in the present study is interpreted as the process of “turning ideas into (…) sentences, and larger units of discourse within working memory” (McCutchen, 2006, p. 121). In other words, this means the use of verbal language for heuristic purposes in solving a picture-based problem to create a logical and coherent L3 English text content. The think-aloud protocols captured the participants’ verbalisations in the process of coming up with ideas triggered by the pictorial information in the different stages of intersemiotic translation. In the process of transposing pictorial information into verbal language, the L1 Transylvanian-Hungarian participants employed L1 Hungarian to generate ideas. However, the L1 was used to a lesser extent than L3 English and L2 Romanian.
One such example of creating meaning and relying on L1 Hungarian is found in Excerpt (25). Lines 29 and 37 contain L3 English utterances, which mark the beginning and the end of the L1 Hungarian inlay.
(25)
29 she wanted to make pancakes 30 na, várjunk [well, let’s see]
31 ehhez szüksége van a recept könyvre [for this she needs the cookery book]
32 szóval felköti a kötényét [so she puts on her apron]
33 leveszi a polcról a recept könyvet [takes the cookery book from the shelf]
34 kinyitja és megtalálja a receptet [opens it and founds the recipe]
35 elolvassa és látja mennyi minden kell [she reads it and sees how many things are needed]
36 szóval /…/ [well]
37 she needed to find the /…/
38 to find the /…/
39 huh mi az hogy recept könyv
(Verbal protocol excerpt; participant: Sz.E.)
Excerpt (25) illustrates an instance in which the participant, having already created text in the target L3, reverts to his L1 Hungarian to generate further content. Such switching between languages in the creation of a non-native text has been identified by previous SLA writing research involving participants who were learning a second or a foreign language