The Role of Previously Learned Languages in the Thought

420 Research in the Teaching of English Volume 47 May 2013

The Role of Previously Learned Languages in the Thought Processes of Multilingual Writers at the Deutsche Schule Barcelona

Brandon D. Tullock

Marta Fernández-Villanueva

University of Barcelona, Spain

In recent years, scholars have voiced the need for research which focuses on the ability of multilinguals to write across multiple languages rather than on the limitations that they face when composing in a non-native language. In order to better understand multilingual writers as resourceful and creative problem-solvers, the current study aims to investigate how German/Spanish/Catalan multilinguals draw on the full extent of their linguistic repertoires to solve lexical problems while writing in their fourth language, English. Think-aloud data were collected from 10 informants

(8 female, 2 male; ages 16-17) in a German immersion secondary school in Barcelona, Spain.

Analysis of the participants’ protocols revealed that the activation of lexical items across several languages was a common approach to tackling lexical problems. The writers’ resourcefulness and creativity were apparent in the activation of cognate forms and their willingness to experiment with language. In their metacomments, they expressed awareness of their strategic behavior as well as their degree of satisfaction with their solutions. It is argued that more research into the strategic behavior of multilingual writers is necessary in order to inform multilingual pedagogy.

Introduction

Multilingual writers are a rapidly growing demographic in today’s modern, globalized society, as more and more individuals are finding it necessary to engage in written communication in multiple languages on a regular basis. To date, however, most research on multilingual writing has been conducted with the monolingual native speaker as a model and has viewed phenomena such as code-switching and the appearance of non-standard forms in written texts as representative of a deficit.

This gives cause for concern to a number of scholars, who work at what Ortega and Carson (2010) call “the interface between L2 (second language) writing and

SLA (second language acquisition)” (p. 48). Grosjean’s (1989) statement that “a bilingual is not two monolinguals in one person,” along with Cook’s (e.g., 2010) notion of “multi-competence,” that is, the unique set of skills and metacognitive knowledge developed as a result of knowing multiple languages, have challenged

420 Research in the Teaching of English Volume 47, Number 4, May 2013

Copyright © 2013 by the National Council of Teachers of English. All rights reserved. g420-441-May13-RTE.indd 420 4/19/13 11:24 AM

T ullock and

F ernández

-V illanueVa

Thought Processes of Multilingual Writers 421 the idea of a monolingual norm and led certain researchers to call for a reorientation of writing research. According to Ortega and Carson, this different approach will allow us to focus on “what multicompetent writers can do, as opposed to only understanding what they cannot or wish not to do in their L2” (p. 65). Canagarajah

(2006), taking a similar stance, argues that multilingual writing research needs to challenge a monolingual bias in order to understand multilinguals better as resourceful problem-solvers, “shuttling creatively through discourses in order to achieve their communicative objectives” (p. 591). Textual differences, he says, should be considered as the result of strategic choices rather than unconscious errors and that by studying the processes behind these choices we can come to understand the resources multilinguals bring to their texts.

Perhaps one of the most important resources multilingual writers possess is their ability to refer to their full linguistic repertoire while composing. In a recent review, Manchón, Murphy, and Roca de Larios (2007) note that writers have been shown to deploy their L1 (first language) for a variety of purposes in all stages of composition. This behavior has been observed in multilingual writers at all proficiency levels and seems to facilitate rather than inhibit text production by reducing the load on working memory and providing access to key lexical items (see also

Manchón, Roca de Larios, & Murphy, 2009). L1 use during L2 writing is a widely documented phenomenon, and researchers generally agree that writers use all the languages in their linguistic repertoire while writing. This is evidenced by studies of cross-linguistic influence, which have found evidence of transfer from multiple languages in multilinguals’ written texts (e.g., Cenoz & Gorter, 2011). However, few studies of the multilingual writing process have gone beyond two languages, and the role played by non-target languages besides the L1 during composition remains unexplored. Meanwhile, in the field of multilingualism, the role of the

L2 in L3 (third-language) acquisition and production has proved to be a fruitful area of study, as L3 learners have been shown to rely heavily on their L2, especially in early stages of learning (e.g., Hammarberg, 2001; Williams & Hammarberg,

1998; see also Falk & Bardel, 2010). In contexts with favorable attitudes toward bilingualism, being bilingual has also been linked to advantages in the acquisition of additional languages due to increased metalinguistic awareness, knowledge of personal learning strategies, enhanced communication strategies, and a wider linguistic repertoire that can be exploited as a base for transfer (Cenoz, 2003). Thus, we feel that more research is warranted on how multilinguals’ repertoires at the resource level (Hayes, 2012) dynamically interact while writing in environments where multilingualism is socially promoted.

The aim of the present study is to go beyond two languages and examine the role that previously learned languages play in the lexical searches of German/

Spanish/Catalan multilinguals as they write in their L4 (fourth language), English.

It was inspired by “The Tyrol Study,” where Jessner (2006) used concurrent thinkg420-441-May13-RTE.indd 421 4/23/13 12:00 PM

422 Research in the Teaching of English Volume 47 May 2013 aloud protocols to gain insight into the cross-lexical problem-solving behavior of

German/Italian bilingual students at the University of Innsbruck as they wrote in English. The research presented here was carried out in a German immersion secondary school in Barcelona, Spain, where students have daily contact with two Germanic and up to three Romance languages. The analysis will focus on the writers’ strategic use of all the languages in their repertoire as they search for lexical items, and the discussion will highlight the strategic awareness of these multicompetent individuals in order to identify what they are actually doing as creative and resourceful problem-solvers.

Literature Review

Multilingual Writing

Despite its age, Flower and Hayes’ (1981) Planning-Translating-Reviewing model contains features that are still current in modern representations of writing. This is especially true in Hayes’ (2012) latest revised model, which includes a control, a process, and a resource level—the control level including motivation, goal setting and writing schemas; the process level including the writing process and the task environment; and the resource level containing working memory, attention and long-term memory (along with all learners’ linguistic repertoires). The three macroprocesses involved in writing—planning, translating, and reviewing—remain in the Hayes (2012) model as proposing, translation, transcription, and evaluation. (Henceforth, “translation/translator” will be referred to as “formulation/ formulator” in order to avoid confusion in a foreign language context.) In the latest model, writers proceed through these stages in a recursive fashion, shifting their focus to one process or the other as they attend to both high-level and low-level goals. At the process level, the “proposer” is responsible for generating a preverbal message. The “formulator” receives this input and converts it into a pre-text, or candidate text, which is stored in an articulatory buffer in working memory where it is evaluated. If deemed acceptable, it is added to the written “text thus far,” and if not, the process may begin again with the generation of new ideas or alternative pre-texts. The “evaluator” is supposed to be involved in all of these stages, making judgments related to the intended written product (Hayes, 2012).

The main area where foreign language (FL) writing distinguishes itself from

L1 writing is in formulation. For formulation to be fluent, writers must have easy, if not automatic, access to a wide range of target language knowledge. In foreign language contexts, however, this knowledge is often limited, making it necessary for writers to divert their attention away from the text as a whole in order to attend to linguistic issues. Writers have been found to devote most of their time to formulation (between 60% to 80%) when writing in either the L1 or the FL (Roca de Larios, Marín, & Murphy, 2001; Wang & Wen, 2002; see also Manchón et al., g420-441-May13-RTE.indd 422 4/19/13 11:24 AM

T ullock and

F ernández

-V illanueVa

Thought Processes of Multilingual Writers 423

2009). However, FL formulation entails a greater amount of problem-solving, where writers take steps to bridge the gap between an initial state and their communicative goals. Roca de Larios et al. (2001) found that the ratio of fluent formulation to problem-solving formulation was 5:1 in L1 writing and 2:1 in FL writing.

Lexical issues are common in FL writing, which involves a wider range of lexical problems than monolingual writing. Roca de Larios, Manchón, and Murphy (1996) put into four categories the types of problems L1 and L2 writers face in trying to transform preverbal messages into language that accurately reflects the intended meaning and conforms to the appropriate rhetorical style and register of the task.

P1 problems occur when writers struggle to retrieve the words needed to express a concept from long-term memory. P2 problems are unique to bi/multilingual writing; they occur when the appropriate lexical item is available in a non-target language and must be translated into the language of the text. P3 problems involve a writer’s desire to improve upon an available item in the target language, and P4 problems indicate insecurity about the accuracy or the appropriateness of a selected lexical item. P4 problems are theoretically common to both L1 and L2 writers but were only found in L2 data in the above-mentioned study.

When FL writers encounter lexical problems, they use all the resources available to them, including prompts and the already written text, dictionaries when available and linguistic resources stored in long-term memory, along with the writers’ L1 and other known languages (Manchón et al., 2007). One way of bridging the gap between a concept and the language needed to express it is by generating pre-texts, or tentative formulations, in the L1 or another known language. This behavior can be accompanied by a wide range of strategies, such as generating synonyms until an appropriate match is chosen and translated, repetition, paraphrasing, or segmenting one’s meaning (see Manchón et al., 2007). This tendency for writers to activate the L1 while writing led Wang and Wen (2002, p. 239) to state that “the

L2 writing process is a bilingual event.”

Another way learners use other known languages to tackle lexical problems is by backtranslating target language pre-texts into the L1, using it as a yardstick to measure the appropriateness of the item they have chosen. This is a variety of backtracking, the process of moving one’s attention back and forth between already written text and the text that is currently being produced. Backtracking helps learners focus on the main task objectives, review the main points of what they have written, and stimulate text production by freeing up cognitive resources.

It may involve a complete rereading, a summary, or paraphrasing of the text or writing prompt in any language (Manchón, Roca de Larios, & Murphy, 2000).

Much research has been conducted in order to determine how variables such as proficiency, task complexity, and general writing expertise affect the extent to which writers use other languages while composing. However, mixed results have been found, and these variables seem to interact in complex ways. Nevertheless, g420-441-May13-RTE.indd 423 4/23/13 12:00 PM

424 Research in the Teaching of English Volume 47 May 2013

L1 use seems to be common to FL writers at all proficiency levels (Van Weijen,

Van den Bergh, Rijlaarsdam, & Sanders, 2009; Wang, 2003; Wang & Wen, 2002;

Woodall, 2002). In a recent study, Murphy and Roca de Larios (2010) identified six different purposes for which advanced learners of English as a foreign language switched to the L1 during lexical searches. Aside from generating pre-texts and backtracking, learners also used their mother tongue to access metalinguistic knowledge encoded in the L1 and to monitor their writing process by engaging in self-talk to focus attention on problems, evaluate solutions, and comment on their strategic approach.

Having seen the important role the L1 plays in bilingual composition, one may ask the question, “What is the role of the L2 in L3 composition?” The following section allows us to view this issue in light of the broader discussion surrounding the role of the L2 in L3 acquisition and use.

Cross-linguistic Interaction in Multilinguals

Traditionally, research on cross-linguistic influence, or transfer, has only focused on the effects of the L1 on a language currently being acquired. However, as researchers have begun to pay more attention to other languages known to multilingual subjects, they have seen that cross-linguistic influence is not limited to transfer from the L1 but can involve all of a multilingual’s languages. Multilingual learners have been shown to prefer relying on their L2 to support the acquisition of an L3, especially at the early stages of acquisition (e.g., Hammarberg, 2001; Wang, 2003;

Williams & Hammarberg, 1998).

The literature on cross-linguistic influence has identified an extensive list of factors that contribute to the activation of one language over another (see Hall &

Ecke, 2003). However, some factors play a more important role than others, and the three that appear most frequently are (psycho)typology or language distance, proficiency, and L2 status. Learners tend to activate languages that are perceived as being more closely related to the target language during production—or having less “language distance.” They also show a preference for languages in which they are more proficient, but L2 status may override this, as learners may choose their

L2 over their L1 on the basis of its status as a foreign language, especially if this language is typologically close.

Studies of cross-linguistic influence on writing have shown the importance of transfer strategies to multilingual writers. Navés, Miralpeix, and Celaya (2005) studied the effects of grade level on the strategies of borrowing and lexical inventions. The researchers found a decrease in the use of both of these strategies as grade level increased, but this difference was only significant for borrowing. In the trilingual context of the Basque Country, Cenoz and Gorter (2011) examined the texts of adolescent writers across all their languages and found evidence of transfer among all three languages, even influence of the L2 and L3 on the L1. g420-441-May13-RTE.indd 424 4/19/13 11:24 AM

T ullock and

F ernández

-V illanueVa

Thought Processes of Multilingual Writers 425

Currently, one of the most influential multilingual models is the Dynamic

Model of Multilingualism (Herdina & Jessner, 2002; Jessner, 2006, 2008), in which languages are not conceived of as separate entities but rather as dynamic and interdependent systems, with the acquisition of new language systems being supported by those already present in the multilingual mind. The language systems known to a learner are in constant flux and interact in complex ways. This concept of “cross-linguistic interaction” goes a step further than cross-linguistic influence and includes phenomena such as code-switching and cross-linguistic comparison.

According to the Dynamic Model of Multilingualism, the presence and interaction of multiple language systems in the multilingual mind produces the so-called

Multilingualism-factor, or M-factor, a phenomenon that may be responsible for advantages in further language acquisition.

Under the Dynamic Model of Multilingualism paradigm, Jessner (2006) conducted one of the few studies devoted to the writing process of multilinguals. She analyzed the strategic behavior of German/Italian multilinguals as they searched for words while writing in English. Although the participants had grown up in German/

Italian speaking households, German turned out to be the dominant language used by most of the writers, who were studying at the University of Innsbruck in Austria.

It was the preferred language for starting a search for lexical items and the source of most borrowings, while Italian was used to backtranslate and evaluate items about which learners had doubts, especially in the case of Latin-based cognates.

Jessner also analyzed metalanguage as the most explicit form of linguistic awareness. She found that language switches preceded by metalanguage were more frequent in trilingual than in bilingual strategies. Metalanguage also seemed to be related to language dominance, as most of the Italian metalanguage was produced by only two subjects, who were Italian dominant and had daily contact with Italian through university classes and Italian-speaking friends. Those students who were

German dominant showed a clear preference for German in both their compensatory strategies and their metalanguage. Jessner identified two types of awareness exhibited by her informants: cross-linguistic awareness, defined as the tacit or explicit awareness of the other active non-target languages during production, and metalinguistic awareness, which made this comparison possible. (See also

Jessner, 2005, 2008.)

Research Questions

The present study aims to contribute to what is known about multilingual writing by observing the interplay between the languages that make up the repertoires of multilingual writers as they engage in lexical problem-solving in a context in which multilingualism is socially promoted. The following research questions will be addressed: g420-441-May13-RTE.indd 425 4/23/13 12:00 PM

426 Research in the Teaching of English Volume 47 May 2013

1. To what extent do multilingual writers think in previously learned languages while composing in English?

2. To what extent do writers rely on previously learned languages when searching for lexical items in English?

3. What role is played by previously learned languages in lexical searches?

Method

Participants

The participants (8 females, 2 males; ages 16-17) were selected from the 11th class, or grade, of Gymnasium, the university preparatory path, at the Deutsche Schule

Barcelona (henceforth the Schule) on the basis of their linguistic configuration.

At the Schule, German is the language of instruction, but students also attend Spanish class every day for all twelve years. As a result, students at the Schule develop near-native-like levels of competence in both German and at least one community language, regardless of their L1, which may be German, Spanish, or

Catalan. Catalan is also taught for the first nine years, but then it becomes an optional subject along with French. English is taught as a foreign language for three hours a week from the 5th year onward. In the 11th year, students are supposed to have approximately a B2 (upper-intermediate) level of English.

The linguistic configuration of these students is particularly interesting because of the typology of the languages involved. Spanish, Catalan and French are Romance languages, whereas English and German are both West Germanic.

Despite its classification as a Germanic language, English derives a large portion of its vocabulary from Romance languages.

Since students at the Schule come from varied linguistic backgrounds, the status of each language in the mind of the participant depends on his or her L1.

The sociolinguistic status of each language taught at the Schule also varies:

• German—language of instruction

• Spanish—community language with instructional support

• Catalan—community language with fewer hours of instructional support

• English—first foreign language

• French—second foreign language

All 26 students who volunteered to take part in the study were asked to fill out a sociolinguistic background questionnaire that included questions about the acquisition of their languages as well as their patterns of use at home, at school, and at activities outside school. As proficiency in both German and at least one of the community languages was considered to be a crucial factor, only those who were born in Spain and had received all of their formal education at the Schule g420-441-May13-RTE.indd 426 4/19/13 11:24 AM

T ullock and

F ernández

-V illanueVa

Thought Processes of Multilingual Writers 427 were considered as potential candidates. As a further control for L2 proficiency, the eligible students completed the DIALANG word-recognition task 1 in their L2

(German or Spanish), and only those who scored above 600 (advanced) were included. The L1 was determined by using the students’ self-reported first language and crosschecking them with the language background and use questionnaires.

Three participants had Spanish as an L1, 3 had Catalan, and 4 had German. English proficiency was also measured using the DIALANG word recognition task.

After these measures were taken, eleven willing participants qualified to take part in the study. However, one participant, who failed to perform the task, was later eliminated.

Writing Task

The writing task for this study was a response to the prompt: “How is growing up

nowadays different from growing up in your parents’ generation?” This topic was chosen because it was related to what the students were presently studying in their

English classes, the assumption being that the ideational content necessary for composing the essay would be easy to access and more working memory would be available to deal with lexical problems. The students were given 30 minutes to write approximately 200 words. These are roughly the conditions given to the participants in Wang’s (2003) study and are similar to the constraints faced by writers during English language examinations.

Data Collection Procedure

For the purposes of the present study, data were collected using think-aloud protocols, also referred to as verbal or concurrent protocols. This method was selected because it gives immediate, direct access to writers’ thought processes (Cohen, 1996).

The authors are aware of the criticism that has been raised against this method and took steps to reduce threats to validity. According to Manchón, Murphy, and

Roca de Larios (2005), the main threats associated with this method are reactivity to the task and interaction effects between the writer and researcher. In order to combat reactivity, the informants were given clear and simple instructions about the task they were expected to perform, and they had several minutes to practice thinking aloud on their own before carrying out the task. Meanwhile, we avoided creating expectations about the kind of information we were after. Modeling thinking aloud was avoided, as this might influence their behavior (Wang, 2003).

In order to minimize interaction effects, the task instructions were identical for all the participants, and interaction between the writer and researcher was also kept to a minimum in order to reduce variability in the way the informants performed the task.

The data collection procedure was carried out over three days. On the first day, the students were familiarized with thinking aloud and were given a mock composition to practice on their own. On the second day, two interlocutors (an g420-441-May13-RTE.indd 427 4/23/13 12:00 PM

428 Research in the Teaching of English Volume 47 May 2013

English native speaker along with a Spanish native speaker) conducted two recording sessions in which four students were recorded individually in separate rooms at the Schule at the same time as their regular English lessons. All instructions were given to the participants in English so as not to influence their language-switching behavior. Once again, they were given a mock composition to practice thinking aloud for about ten minutes as the interlocutors selectively listened in to make sure that they performed the technique properly. Then the students were given the actual task and began writing their compositions. The interlocutors sporadically listened in as the students were writing and prompted them to keep talking or to speak up when necessary. In order to keep interaction to a minimum, interlocutors only intervened if the informant was quiet for more than eight seconds. Once the end of the time limit was reached, the subjects were told to finish their compositions, which were then collected so that they could be referred to when segmenting the transcribed data. Students were allowed to finish early or were granted a couple of extra minutes to finish up their compositions when necessary. No dictionaries were provided to the writers, as the study aimed to find out how they used the resources available to them in their current stage of linguistic development. On the third day, the remaining three students were recorded following the same procedure.

This time an additional interlocutor, a native speaker of German, was present.

Data Preparation

The first step in preparing the data for analysis was transcribing the recorded think-aloud data. The first author, along with another trained linguist who was a fluent speaker in all the languages involved, transcribed word for word what the subjects said, along with phenomena such as long pauses, hesitations, yawning and coughing. The data were then segmented and matched to sentences in the students’ actual written compositions. Once this was done, the data were prepared in three phases.

First, words in the compositions and words in the protocols were counted.

Composition time was measured from the instant the subjects started talking to the moment they finished. The next step was to isolate formulation from planning, taken to mean the planning of ideas, and from revision, which includes revisions made to the already written text. Following Roca de Larios et al. (2001), formulation included “both the verbalization of written material and those other utterances that, because of their strict linear nature (lexical units, syntactic structures, etc.) could be considered clear candidates for becoming part of the text” (pp. 510-511). Changes made to the sentence currently being written were also categorized as formulation.

Then, lexical searches during formulation were isolated. In order to be considered a lexical search, two elements were required: evidence of a lexical gap between what the learner wanted to say and what he or she was able to say, as well as steps taken toward solving the problem. Sometimes a problem was revealed g420-441-May13-RTE.indd 428 4/19/13 11:24 AM

T ullock and

F ernández

-V illanueVa

Thought Processes of Multilingual Writers 429 through explicit problem indicators such as “How do you say…?” At other times the perception of a problem was related to implicit clues such as long pauses or repetition of a pre-text with a rising intonation. Those instances where the writer expressed a word in a non-target language and then effortlessly translated the word into English without any intervening steps were coded as binary searches and were not included in lexical searches (see Appendix 1 for transcription conventions):

Example of a Binary Search

“…because they lived in another societat, (C: society) society.” CR

Lexical searches were also distinguished from restructuring problems, which involve the search for an alternate syntactic structure after a problem had been encountered with the initial structure:

Example of Restructuring

“This is a a topic which…[cough] which… ¿cómo se dice con el que (S: how do you say

‘with which’)? With…No. No para empezar (S: no, not to start out) [cough]” This is a topic which with that I CP

Orthographical problems were also excluded from the analysis. Once lexical searches had been identified they were counted and categorized according to the languages involved.

In the final phase, the lexical searches were broken down into the steps taken, starting with the occurrence of the problem and going up to the point where the subject either decided on a solution, even if it was only a tentative one, or gave up and moved on. Once the lexical searches had been broken down into their components, those problem-solving steps that involved the use of a non-target language were identified and put into the following six categories taken from Murphy and

Roca de Larios (2010): generating lexical units or pre-texts; backtracking or back-

translating; self-questioning, problem-indicating and problem-focusing; evaluation;

metalinguistic appeal; and metacomments.

In order to establish consistency in coding the data, the first author and a trained colleague coded two of the compositions separately. These decisions were then compared and any disagreements were resolved. Then the same author coded the rest of the compositions.

Results

Before interpreting the results, it is important to mention some important differences that were observed regarding participants’ English proficiency and the number of words in the protocols and the compositions. Half of the participants scored at an advanced learner level (601-900) while the other half scored at the intermediate level (401-600). The means for the Spanish and Catalan L1 students were similar (569 and 594, respectively), while the mean proficiency of the German g420-441-May13-RTE.indd 429 4/23/13 12:00 PM

430 Research in the Teaching of English Volume 47 May 2013

L1 students turned out to be much higher (706). Additionally, composition length varied, ranging from 138 words to 371. Finally, some students were better verbalizers than others, and the number of words in the protocols varied widely. The descriptive statistics of the informants’ protocols and compositions are reported in Table 1 along with their proficiency scores in English and their L2.

RQ1: To what extent do multilingual writers think in previously learned languages while composing in English?

The first research question was asked in order to obtain a general overview of the languages that were active during composition. In order to respond to this question, the percentage of words in each language were calculated following the procedure outlined in Manchón, Roca de Larios, and Murphy (2000). The words in the protocol were totaled, and then the words in each language were counted.

Then the number of words in the students’ compositions was subtracted from both the total number of words in the protocol and the total number of English words. Finally, a percentage was calculated in each language. The percentages as well as the raw number of words produced in each language are reported in Table 2.

In accordance with previous writing studies, languages other than the target language were active during the production. These were not limited to the L1, however. All students, regardless of L1, thought in German, although three used fewer than 10 German words. Eight of the informants, including 2 German-dominant and all Catalan-dominant writers, thought in Spanish, while use of Catalan was confined to the Catalan L1 writers. CR used every language in her linguistic rep-

T able

1. Descriptive Statistics

Writers*

English

Proficiency

L2 Proficiency

(G=German,

S=Spanish)

Words in

Protocol

Words in

Composition

SN

SB

SJ

CA

CP

CR

GC

GJ

GS

GR

457(Int.)

532(Int.)

720(Adv.)

690(Adv.)

531(Int.)

561(Int.)

590(Int.)

727(Adv.)

730(Adv.)

778(Adv.)

G730

G940

G802

G649

G807

G652

S730

S703

S730

S778

2,789

2,681

1,709

1,163

1,652

1,442

1,676

1,040

1,067

1,103

304

184

223

275

285

327

371

263

217

138

*Note: The first letter of the abbreviation indicates the writer’s L1 (Spanish, Catalan, or German). The second letter is part of the writer’s pseudonym. g420-441-May13-RTE.indd 430 4/19/13 11:24 AM

T ullock and

F ernández

-V illanueVa

Thought Processes of Multilingual Writers 431

CP

CR

GC

GJ

GS

GR

SN

SB

SJ

CA

T able

2. Overall Language Use in Protocols

Writers

English

% / words

German

% / words

Spanish

% / words

35.7 / 888 0.5 / 12 63.8 / 1,585

88.4 / 2,392 0.6 / 15 11.0 / 274

49.1 / 953

62.5 / 830

0.6 / 9

0.2 / 2

50.3 / 747

36.8 / 327

88.0 / 1,487

14.3 / 486

0.1 / 1

1.6 / 18

5.3 / 72

4.9 / 54

76.0 / 1,363 9.8 / 128 14.2 / 185

97.3 / 1,019 2.7 / 21

77.1 / 872 22.9 / 195

28.0 / 408 71.4 / 689 0.6 / 6

Catalan

% / words

French

% / words

0.5 / 4

6.7 / 92

79.1 / 882 0.2 / 2

Words in

Protocol

(Adjusted)

2,485

2,497

1,486

888

1,367

1,115

1,305

777

850

965 ertoire, including French, a language she had been studying for under than three years, lending support to the notion that writers draw upon their full linguistic repertoire while composing.

The majority of the writers in this study preferred their L1 except for CA and

GC, who both used Spanish more than their L1 to support their composition. According to the information gathered in the background questionnaire, CA comes from a Spanish-Catalan bilingual household and speaks only Spanish with one of her parents, and her patterns of use reveal that she mostly uses Spanish with friends at school. At the beginning of her composition, she starts to express herself in Catalan before switching to Spanish:

Primer, primer les diferencias de (C: first, first the differences from) ayer . . . una introducción . . . (S: yesterday . . . an introduction)

After these first four words, she never goes back to formulating her thoughts in

Catalan. GC, on the other hand, maintained a balance between German and Spanish while composing. Her linguistic background questionnaire shows that, unlike the other German L1 speakers, she mainly uses Catalan at home and only sometimes uses German to communicate with her family. Meanwhile, with friends at school she always uses either German or Spanish and never Catalan. The predominance of Spanish and German over Catalan at the Schule could have influenced these individuals to restrict their Catalan use to non-academic contexts.

g420-441-May13-RTE.indd 431 4/23/13 12:00 PM

432 Research in the Teaching of English Volume 47 May 2013

RQ2: To what extent do writers rely on previously learned languages when searching for lexical items in English?

The learners in this study carried out 111 lexical searches while writing their compositions, ranging from 1 to 36 lexical searches per individual. As some participants wrote much longer compositions than others, the frequency of lexical searches was calculated by dividing the number of lexical searches by number of words in the composition. To make proficiency-related differences more visible, the informants are organized according to proficiency level in Table 3.

The mean number of lexical searches per 100 words was 4.13 (sd = 3.14).

There was a great deal of variation in the frequency with which learners tackled lexical problems, ranging from 0.38 to 11.84. A high Spearman rank correlation was found between English proficiency and the frequency of lexical searches (r =

-0.8788, p <.001). This correlation is not surprising, given that we would expect lower proficiency learners to have more difficulty accessing lexical items, which would in turn lead to more frequent lexical problems.

On average, students with a Romance language as an L1 encountered more lexical problems than those with German as their L1. While the overall higher

English proficiency of the German L1 group is likely to be partially responsible for their low number of lexical searches, this interpretation should be regarded with caution. Closer inspection suggests that it may also be attributed to one participant’s lack of verbalization, as evidenced by frequent long pauses as well as a low number of words in her protocol. In addition, another German-dominant participant was beset with ideational problems from the outset, and spent most

T able

3. Lexical Searches by English Proficiency Level

Writers

CA

SJ

GJ

GS

GR

Total

SN

CP

SB

CR

GC

English

Proficiency

690

720

727

730

778

457

531

532

561

590

Lexical

Searches

3

3

111

7

8

1

36

15

13

10

15

Lexical Search

Frequency/100 words

2.55

3.59

0.38

1.38

2.17

11.84

5.26

7.07

3.06

4.04

Lexical Searches in Non-target

Languages

35/36

15/15

11/13

10/10

13/15

7/7

6/8

1/1

3/3

3/3

104/111 g420-441-May13-RTE.indd 432 4/19/13 11:24 AM

T ullock and

F ernández

-V illanueVa

Thought Processes of Multilingual Writers 433 of his time thinking about what to say in response to the prompt rather than how to say it. As a result, he spent less time formulating and wrote a much shorter composition than the others.

Of the 111 lexical searches in our data, 104 of them involved non-target languages. Even GJ, who carried out 97.3% of her composition in English, used her L1 to help her access a lexical item. A closer look at the searches involving non-target languages revealed which ones were used. Of these 104 searches, 81 were bilingual,

22 were trilingual, and 1 was quadrilingual. These results are similar to Jessner’s

(2006), who also found more bilingual than trilingual searches.

Overall, Spanish was clearly dominant in the bilingual searches (62 of 81, see

Table 4). This dominance can be partially accounted for by the fact that over half of these were produced by Spanish speakers, who overwhelmingly preferred their

L1. This preference for the L1 was also shared by 3 of the 4 German speakers and

1 Catalan speaker.

Seven of the 10 participants used up to 3 or 4 languages in their lexical searches

(see Table 5). The three German students who tackled fewer lexical problems than the other participants only used German. The most frequent combination of languages used in lexical searches was German/Spanish/English (13), followed by

German/Catalan/English (4) and Spanish/Catalan/English (4), while the combinations Catalan/French/English and German/Spanish/Catalan/English each occurred once. Again, a clear preference for the L1 can be observed, as 19 of the 23 tri- or quadrilingual lexical searches involved the participant’s L1.

T able

4.

Languages Involved in Bilingual Searches

Writers

SN

SB

SJ

CA

CP

CR

GC

GJ

GS

GR

Total

G/E

2

3

3

11

2

1

5

5

10

S/E

29

7

9

62

C/E

3

2

5

Total

29

9

5

5

13

2

11

1

3

3

81 g420-441-May13-RTE.indd 433 4/23/13 12:00 PM

434 Research in the Teaching of English

T able

5.

Languages Involved in Multilingual Searches

G/C/E

Trilingual

S/C/E C/F/E Writers

SN

SB

SJ

CA

CP

CR

GC

Total

G/S/E

6

2

1

2

2

13

4

4

1

3

4

1

1

Volume 47 May 2013

Total

6

2

1

2

2

22

1

8

Quadrilingual

G/S/C/E

1

1

RQ3: What role is played by previously learned languages in lexical searches?

The third research question was asked in order to shed light on writers’ purposes for switching to languages besides the target language and the L1. As the main interest of this study was to go beyond L1 use, those three subjects who only used their L1

German were excluded from this part of the analysis. The following examples were selected for the dual purpose of exemplifying the categories that were used (from

Murphy & Roca de Larios, 2010) and for illustrating the idiosyncratic approach these multicompetent learners took to tackling lexical problems.

Generating Lexical Units or Pre-texts

One of the most common ways learners use previously learned languages during lexical searches is to produce pre-texts, or candidate texts. The generation of pretexts indicates the selection of a preverbal concept for which the target language item is not immediately accessible, so lexical units consisting of one or more words were produced in one or more languages in order to bridge the gap between the writer’s intended meaning and the target item. In the protocols, this was often accompanied by repetition and the activation of synonyms across languages. In the present study, this process seemed to be almost effortless, as when both Spanish and German were activated before producing the equivalent term in English:

Aunque sea muy innecessario (S: although it’s very unnecessary), obwohl (G:

although), aunque (S: although), although it, it’s a bit innecessary…” SA

However, this process could also entail a struggle to activate the target linguistic form, as in the following example, where SJ switched back and forth between German and Spanish before finally settling on “time”: g420-441-May13-RTE.indd 434 4/19/13 11:24 AM

T ullock and

F ernández

-V illanueVa

Thought Processes of Multilingual Writers 435

“. . . that each Epoche (G: epoch) , Epoche, die Epoche, época (S: epoch) each, that each, each, // época, each // Epoche, época, each Epoche, época, // that each . . . time,” SJ

Backtracking and Backtranslating

Backtracking, or rescanning already produced text, is a strategy that can be deployed in a target or non-target language. Backtracking can involve reading over what has already been written in order to stimulate text production (P1 problems, to use

Roca de Larios, Manchón, and Murphy’s, 1996, terminology). Additionally, when a learner has doubts about the correctness or appropriateness of a lexical item (P4 problems), the writer may backtranslate the item in order to evaluate it (Manchón et al., 2000). Both kinds of backtracking appear in the excerpt below. CR reiterates the first part of her clause, repeating the stem, “sense” (C: without) in order to stimulate the generation of ideas. Once she expresses her idea in Catalan, she generates pre-texts in Catalan and French. Finally, she comes up with the English form “take” and, after backtranslating it into Catalan and finding it acceptable, she includes it in her text:

“. . . sense . . . ehm, sense vigilar, (C: without watching out) sense, sense, sense que

la policia te (C: without the police), with the, with the policia, sense que la policia ehm

. . . et detengui . . . (C: detaining you) arrête, arrête, (F: arrest) ehm, eh the police . . . take, t’emporti, (C: take you away) ehm sense que la policia t’emporti (C: without the

police taking you away).” CR

Self-Questioning, Problem-Indicating and Problem-Focusing

Writers monitor their problem-solving activity as they become aware of their lexical problems and attempt to focus on them though questions or comments that lead to strategic behavior. Here, SC’s self-questioning reveals the activation of both Spanish and German:

“You have the, you have the, you have ehm easier options, easier opti- no, opciones (S: options) you have easier, ¿cómo se dice opciones en inglés? (S: how do you say ‘options’

in English?) You have easier ways to get in connection, connect, wie heißt denn..? (G:

how do you say..?) Para conectarse con alguien, para poder (S: to connect with someone, to be able to) connect you have easier ways to . . .” SC

Sometimes the question frame did not necessarily match the language of the pretext, as in the following examples, both from CR:

Com dir (C: how to say) ‘vor allem’ (G: above all)?” CR

Com es diu (C: how do you say) ‘Einkommen’ (G: income)?” CR

Following Murphy and Roca de Larios (2010), we also included fillers such as “a

ver (S: let’s see)” in this category, as they served a problem-focusing purpose and often preceded re-readings or other strategic behavior.

g420-441-May13-RTE.indd 435 4/23/13 12:00 PM

436 Research in the Teaching of English Volume 47 May 2013

Evaluation

Evaluation goes hand in hand with problem-focusing. Learners also engage in self-talk to evaluate the correctness or appropriateness of their solutions. Here, the facility with which the participants switched between languages is evident. In this example, CP uses Spanish for evaluation, Catalan for problem-focusing, and

German for generating a pre-text:

“. . . that folgs (G: leads) to that, that eh, cause, no, causes otra vez no, (S: not ‘causes’ again) ehm that leads to..aviam ja hi he posat leads (C: let’s see have I put down ‘leads’ yet)? Hmm..which leads, si. Bueno, es igual (S: Yes, well it’s okay) that leads to ehm a different, a different, com es diu? A veure (C: how do you say it? Let’s see), and that leads to a different growing up . . .” CP

Metacomments

In lexical problem solving, writers made metacomments, which provided insight into the writers’ awareness of their problem-solving approach as well as shed light on affective factors such as engagement with the task. An example is in the following excerpt, where SN has decided to include the term “actuality” despite her observation that it “sounds Spanish.“

“I think to live in the, in the actuality, so bueno una esp-, una españolada que le estoy

metiendo (S: well I’m making it sound very Spanish),. . .” SN

On the other hand, note CR’s appropriation of the English language as she shows no reservations about coining the word “dictature” in the example below:

“They lived in, dictadura en anglès (dictatorship in English)? Pff, Bueno me lo saco de

la manga y pim pam (S: well, I’ll make it up [literally: I’ll pull it out of my sleeve] and

‘pim pam’), they, // they lived in a dictature.” CR

Metalinguistic Appeal

Metalinguistic appeal is another strategy where learners focus on language as an object using metalinguistic terms, which can be standard or explained in the subject’s own words. Two examples are the following:

sinónimo (S: synonym), ehm allow, ehm… zulassen (G: allow)?SB

otra palabra para conectar (S: another word for connect).SC

The top three purposes for which writers used previously learned languages were for generating lexical units; evaluation; and self-questioning, problem-indicating and problem-focusing; whereas backtracking and backtranslating ; metacomments; and metalinguistic appeal were found to a lesser extent. Spanish was g420-441-May13-RTE.indd 436 4/19/13 11:24 AM

T ullock and

F ernández

-V illanueVa

Thought Processes of Multilingual Writers 437 the dominant language in all the categories, and it was also the most widely used, which is related to the overall preference for Spanish by most of the informants in this sample. Catalan was the second most frequently used language, followed by German. French was only used once. The widest array of languages was used for generating pre-texts, and this purpose was the main one for which German was used. In Figure 1, the overall frequencies of each purpose are displayed and broken down by the language used.

While Spanish was the dominant language used for generating pre-texts, every single writer in the sample also used German for this purpose, as is illustrated in

Figure 2. Further qualitative analysis of the subjects’ language switching behavior in generating pre-texts revealed several episodes where writers seemed to be at least tacitly aware of language distance at the word level. CR, who did most of her thinking in Catalan, generated lexical units in the widest variety of languages, including the cognate forms: arrête (F: arrest), Einkommen (G: income), and Haushalt (G: household). Although in all three of these cases she opted for a simplified target form, it is possible that she was tacitly aware of and was attempting to activate the

English cognate forms. Her use of cognates also appears in the following excerpt, when she backtranslates “attend” and expresses doubts about its appropriateness but decides to include it anyway:

F igure

1.

Purposes for Using Non-Target Language, by Language g420-441-May13-RTE.indd 437 4/23/13 12:00 PM

438 Research in the Teaching of English Volume 47 May 2013

F igure

2.

Generating Pre-texts: Individual Frequencies

“women had to stay at home and, ehm.. ehm fer coses a casa (C: do things at home),

Haushalt (G: Household), ehm Haushalt, ehm ehm Haushalt, ehm com es diu (C: how do you say it)? At home and do and and clean la casa o cuinar (C: the house or cook) or attend attender, prestar attenció,(C: tend to, pay attention) o ayudar, o (S: or help or) hmm hmm hmm attend… bueno jo diria que no es pot dir però bueno (C: well, I don’t think you can say that but oh well) or attend the kids.” CR

Her behavior is similar to that exhibited in SN and SB’s metacomments above, where the writers balance their strategic use with awareness of what does and does not conform to target norms.

Discussion

The aim of the present study was to examine how multilinguals’ linguistic repertoires interact at the resource level in order to solve lexical problems during foreign language writing. The use of think-aloud protocols to gather our data was crucial in order to gain access to the cognitive processes that underlie strategic behavior during writing. Although this method was not entirely problem-free (some students experienced difficulty thinking aloud; long pauses in the protocols indicated that not all thoughts were verbalized), we were able to gather rich information that sheds light on the multilingual writer as a creative problem-solver.

From our results, we can conclude that multilingual writing is indeed a multilingual event. As expected, learners drew on all the resources in their linguistic repertoires, activating multiple languages while writing. Not only daily contact but g420-441-May13-RTE.indd 438 4/19/13 11:24 AM

T ullock and

F ernández

-V illanueVa

Thought Processes of Multilingual Writers 439 also daily use seem to have a great influence on the likelihood that a given language will be activated during writing, which is exemplified by the predominance of

Spanish and German in our participants’ protocols. Nearly all the lexical searches carried out by our participants involved language-switching, which corresponds to a growing body of research that has found this to be a common strategy among foreign language writers of all proficiency levels. In our study, this practice was not avoided but was embraced and employed strategically, suggesting that, rather than conflicting with one another, multilinguals’ languages cooperate during writing, aiding in text production.

The most common reason for activating previously learned languages during lexical problem-solving was for generating pre-texts, bridging the gap between proposed concepts and the target language needed to express them, while backtracking was found to a lesser extent. Both generating pre-texts and backtracking were often carried out with the help of cognates, which can be seen as evidence of cross-linguistic awareness, also found in Jessner (2006).

The participants’ linguistic flexibility is exemplified in other aspects of their behavior, including code-switching, idiosyncratic borrowings, and mixed word formations. This behavior, supported by the information contained in metacomments, demonstrates that the appearance of nonstandard forms such as “live in the actuality” and “dictature” in the written texts were not the result of ignorance but rather conscious decisions made despite an awareness of their deviation from the normative standard. Here, the sociolinguistic context most likely plays an important role. These writers, operating in a context of socially valued multilingualism, develop and support their own norms which deviate from the monolingual standard. This has been described as a characteristic of English as a Lingua Franca in Europe (Seidlhofer, Breiteneder, & Pitzl, 2006).

In conclusion, while this study has shed light on how multilinguals’ languages interact at the resource level, it also indicates that there is still work to be done on how researchers can jointly study and model the interaction of cognitive processes with motivation, task engagement and, especially in the case of multilinguals, socialaffective factors in multilingual writing. Future research entailing a wide variety of tasks across all participants’ languages and carried out in different social contexts would be helpful in creating a more robust picture of the multilingual writer as a creative problem-solver, a perspective that is increasingly necessary in the modern foreign language classroom, where more and more students are entering school with knowledge of more than one language. g420-441-May13-RTE.indd 439 4/23/13 12:00 PM

440 Research in the Teaching of English Volume 47 May 2013

A ppendix

1: T rAnscripTion

c onvenTions

Bold

Plain text

Italics

Example of featured category

Words in English, the target language

Words in a non-target language (German, Spanish, Catalan, or French)

.

,

Underline Written text

Strikethrough Correction made to written text

..

Short hesitation

Short pause

Medium pause

//

Long pause

Denotes where text was abbreviated for space-saving purposes

XX (e.g., CP) At the end of each excerpt the speaker’s participant code is given in italics, with the first letter representing the participant’s L1 and the second letter representing part of the writer’s pseudonym.

(text in parentheses)

Translation of non-English words. The source language is indicated as G:

(German), S: (Spanish) or C: (Catalan). Clarifications regarding the translations are provided in square brackets within the parentheses.

NoTe

1. Available at http://www.lancs.ac.uk/researchenterprise/dialang/about.

ReFeReNces

C anagarajah

, S. (2006). Toward a writing pedagogy of shuttling between languages:

Learning from multilingual writers. College

English, 68(6), 589-604.

C enoz

, J. (2003). The additive effect of bilingualism on third language acquisition: A review. International Journal of Bilingualism,

7(1), 71-87.

C enoz

, j., & g orter

, D. (2011). Focus on multilingualism: A study of trilingual writing. The Modern Language Journal, 95(3),

356–369.

C ohen

, A. (1996). Verbal reports as a source of insight into second language learner strategies. Applied Language Learning, 7(1), 5-24.

Cook, V. (2010). Multi-competence. Retrieved from http://homepage.ntlworld.com/ vivian.c/Writings/Papers/MCentry.htm

F alk

, Y., & B ardel

, C. (2010). The study of the role of the background languages in third language acquisition. The state of the art.

International Review of Applied Linguistics in

Language Teaching, 48(2-3), 185–219.

F lower

, l., & h aYes

, J. R. (1981). A cognitive process theory of writing. College Composi-

tion and Communication, 32, 365–387.

g rosjean

, F. (1989). Neurolinguists, beware!

The bilingual is not two monolinguals in one person. Brain and Language, 36(1), 3-15.

H all

, C. j., & e

Cke

, P. (2003). Parasitism as a default mechanism in L3 vocabulary acquisition. In J. Cenoz, B. Hufeisen, & U. Jessner

(Eds.), The multilingual lexicon (pp. 71-86).

New York: Kluwer Academic Publishers.

h ammarBerg

, B. (2001). Roles of L1 and L2 in

L3 production and acquisition. In J. Cenoz,

B. Hufeisen, & U. Jessenr (Eds.), Cross- linguistic influence in third language acquisi-

tion: Psycholinguistic perspectives (pp. 21-41).

Clevedon, UK: Multilingual Matters.

h aYes

, J. R. (2012). Modeling and remodeling writing. Written Communication, 29,

369-388.

g420-441-May13-RTE.indd 440 4/19/13 11:24 AM

T ullock and

F ernández

-V illanueVa

Thought Processes of Multilingual Writers 441

H erdina

, P., & J essner

, U. (2002). A dynamic model of multilingualism: Perspectives of

change in psycholinguistics. Clevedon: Multilingual Matters.

J essner

, U. (2005). Multilingual metalanguage, or the way multilinguals talk about their languages. Language Awareness, 14(1), 56-68.

J essner

, U. (2006). Linguistic awareness in

multilinguals: English as a third language.

Edinburgh: Edinburgh University Press.

J essner

, U. (2008). A DST model of multilingualism and the role of metalinguistic awareness. The Modern Language Journal,

92(2), 270-283.

M anchón

, r. M., M urPhy

, L., & r oca de

L arios

, J. (2005). Using concurrent protocols to explore L2 writing processes: Methodological issues in the collection and analysis of data. In P. K. Matsuda & T. Silva (Eds.),

Second language writing research: Perspectives

on the process of knowledge construction (pp.

187–202). Mahwah, NJ: Lawrence Erlbaum.

M anchón

, r. M., M urPhy

, L., & r oca de

L arios

, J. (2007). Lexical retrieval processes and strategies in second language writing: A synthesis of empirical research. International

Journal of English Studies, 7(2), 149-174.

M anchón

, r. M., r oca de

L arios

, J., &

M urPhy

, L. (2000). An approximation to the study of backtracking in L2 writing. Learning

and Instruction, 10(1), 13-35.

M anchón

, r. M., r oca de

L arios

, J., &

M urPhy

, L. (2009). The temporal dimension and problem-solving nature of foreign language composing processes: Implications for theory. In R. M. Manchón (Ed.), Writing in foreign language contexts: Learning, teaching

and research (pp. 102-129). Bristol: Multilingual Matters.

M urPhy

, L., & r oca de

L arios

, J. (2010).

Searching for words: One strategic use of the mother tongue by advanced Spanish EFL writers. Journal of Second Language Writing,

19(2), 61-81.

N avés

, T., M iraLPeix

, i, & c eLaya

, M. L. (2005).

Who transfers more . . . and what? Crosslinguistic influence in relation to school grade and language dominance in EFL. Internation-

al Journal of Multilingualism, 2(2), 113-134.

O rTega

, L., & c arson

, J. (2010). Multicompetence, social context, and L2 writing research praxis. In T. Silva and P. K. Matsuda (Eds.),

Practicing theory in second language writing

(pp. 48-71). West Lafayette, IN: Parlor Press.

R oca de

L arios

, J., M anchón

, r. M., &

M urPhy

, L. (1996). Strategic knowledge in

L1 and L2 composing: A cross-sectional study. Proceedings of the European Writing

Conference, SIG Writing (pp. 1-23). Barcelona

Autonomous University.

R oca de

L arios

, J., M arín

, J., & M urPhy

, L.

(2001). A temporal analysis of formulation processes in L1 and L2 writing. Language

Learning, 51, 497-538.

S eidLhofer

, B., B reiTeneder

, a., & P iTzL

, M. L.

(2006). English as a Lingua Franca in Europe:

Challenges for applied linguistics. Annual

Review of Applied Linguistics, 26, 3-34.

V an

W eiJen

, d., van den

B ergh

, h., r iJLaars

daM

, g. & s anders

, T. (2009). L1 use during

L2 writing: An empirical study of a complex phenomenon. Journal of Second Language

Writing, 18(4), 235-250.

W ang

, L. (2003). Switching to first language among writers with differing second-language proficiency. Journal of Second Language

Writing, 12(4), 347-375.

W ang

, W., & W en

, Q. (2002). L1 use in the

L2 composing process: An exploratory study of 16 Chinese EFL writers. Journal of Second

Language Writing, 11(3), 225-246.

W iLLiaMs

, s., & h aMMarBerg

, B. (1998). Language switches in L3 production: Implications for a polyglot speaking model. Applied

Linguistics, 19(3), 295-333.

W oodaLL

, B. R. (2002). Language-switching:

Using the first language while writing in a second language. Journal of Second Language

Writing, 11(1), 7-28. g420-441-May13-RTE.indd 441 4/23/13 12:00 PM