Bi-plurilingual formulation practices in conversational writing interactions

The present article approaches the subject of oral formulation practices in multilingual and conversational writing interactions. Within the framework of pluriliteracy in the context of foreign language practice at university, special emphasis is placed on the manifestations of linguistic repairs as a mechanism, integrated in sequences of interactive formulation. By analyzing extracts of an interaction between three students according to the method of conversation analysis (CA), linguistic and interactional particulars of these repairs are identified, allowing for a first typology of formulation procedures with embedded repair operation. A second aim is to demonstrate in what way different languages are being used during the formulation process. By doing so, the question of how students draw on their plurilingual resources is discussed. The overall issue is addressed through a theoretical framework of functional language distribution in conversational writing.


Introduction
In the field of plurilingualism, qualitative studies on the plurilingual academic literacy practices of university students are still lacking. The fact that plurilingualism occurs not only in verbal but also in written communication has been largely neglected in academic undertakings. While it might be tempting to subsume one form of communication under the other, multiple authors have pointed out that writing in several languages requires different structural resources from speaking (see Oesterreicher 1986, 2011;Maas 2008Maas , 2010Erfurt 2017). Furthermore, little attention has been paid to phenomena in relation to plurilingual and academic writing processes and the possible recourse to plurilingual repertoires as opposed to cross-cultural linguistic differences on the surface of academic writing products (Donahue 2019). In contrast, Dengscherz (2019b) pro-vided an extensive qualitative study and monograph on professional writing in various languages. She raises the question, inter alia, of how plurilingual students and scientists draw on their linguistic repertoires when drafting academic textsa question of high relevance for the present study as well.
This article focuses on formulation practices of plurilingual students in conversational writing interaction 1 within the scope of academic literacy development and learning through the use of plurilingual repertoires. For this purpose, I present preliminary results from a qualitative case study involving students of Romance languages at Goethe University Frankfurt. The dataset of conversational writing interactions (n = 21) was gathered in a class I taught from 2017 to 2018, 2 and includes audio recordings of the writing interactions, screen recordings of the electronic writing processes and text products (for more information about setting and text genre, see 3.1). The methodological framework consists of the principles of conversation analysis (CA), with the qualitative case study focusing on linguistic repair operations within formulation sequences. By analyzing passages of one conversational writing interaction, a preliminary typology of formulation passages with embedded repairs will be developed. A higher-order objective, building on the conversation analytical results, consists of an examination of the usage of different languages during the formulation sequences. In this context, the question of pragmatic-functional language use will be addressed.
In section 2.1, I will refer to pertinent research on conversational writing interactions in anglophone, francophone and germanophone scientific communities in order to draw upon the starting points for my own research. In addition, in section 2.2, I will elucidate the characteristics of oral procedures for formulations within the format of conversational writing interactions and, in doing so, show how repair operations can be conceived as embedded elements of interactive formulation practices. In section 2.3, I embed the use of different languages for formulation within the spectrum of plurilingual academic writing. The analysis of selected passages of a conversational writing interaction, with three student participants -Amna, Carola and Maike 3is at the core of section 3. After analyzing the material, social and interpretative framework of the interaction, I provide an analytical differentiation into three variants of repair operations within intro-1 Conversational writing interaction derives from the German expression konversationelle Schreibinteraktion (see Hinrichs 2014) and refers to a writing format where two or more co-authors, sharing a physical space, discuss and draft a text together.
2 For the didactical considerations on the seminar concept, see Stierwald (2019). 3 All names are pseudonyms. ductory formulation sequences. Finally, the analytical results will be discussed in section 4, against the background of plurilingual writing practice.

Considerations on conversational writing interactions 2.1 State of research
Collaborative writing represents a highly researched subject in the anglophone scientific community. One of the first major monographs, devoted to the question of processes, effects and implications of collaborative writing in the workplace and classroom, is Singular texts, plural authors (1990) by Lisa Ede and Andrea Lunsford. The authors pose the fundamental question of why people should write together andbuilding on theoretical and historical deliberations on collaborative and group writing, as well as on their own questionnaire-interview studythey conclude with the injunction that writers must be able to collaborate in the twenty-first century. A very interesting result of their survey is that people perceive improved content and greater sensitivity to audiences because of the enlarged base of knowledge and experience as major advantages of this writing mode. Shifting attention to collaborative and group writing against a plurilingual background, one can detect a whole body of research pertaining to collaborative writing in L2 contexts. Storch (2011), for example, highlights that carefully designed and guided collaborative writing tasks represent an optimal format for L2 language learning. Different criteria such as task type, L2 proficiency and the social relations between the interacting writers influence the efficiency of this writing format and, accordingly, of the learning effect.
Looking to the francophone scientific field, the Lyon research group, led by Robert Bouchard, Interactions, Acquisition et Apprentissage, engages with conversational writing interactions in the anthology Les processus de la redaction collaborative (Bouchard and Mondada 2005). Here, Apothéloz (2005) addresses the question of how writers progress during the interactive writing process by examining techniques of reformulation in the collaborative fabrication of text. In the German scientific landscape, the research group led by the conversation analysts Ulrich Dausendschön-Gay, Ulrich Krafft and Elisabeth Gülich also focuses on conversational writing interactions, already investigating in 1992 how exolingual writing situations between German-and French-speaking interlocutors foster the acquisition of a foreign language. Building on that, in 1996, Dausendschön-Gay and Krafft analyzed formulation processes in the acquisition of a foreign language by posing the question of how knowledge of text genre and formulation solutions are related. Based on their observations, they categorized writing systems of interactive formulation processes Dausendschön-Gay 1999, 2000). Moreover, Faistauer (1997) examined conversational writing processes (and products) of students in her own class of German as a foreign language at university, in order to look for factors which contribute to more creativity and motivation in writing, on the one hand, and an improved quality of a text product, on the other hand. Overall, although this research has made headway in understanding how interlocutors (of different linguistic backgrounds) organize their writing interactions, it left the role of linguistic resources used in formulations unexamined.
Considering studies on conversational writing interactions in higher education dating from this century, one can notice that they primarily take place in a monolingual context, conceiving writing tasks and participants as being monolingual (see Lehnen 2000;Hinrichs 2014). 4 Moreover, when research into foreign languages is undertaken, it usually takes place in an L2-context, which, by focusing on the development of L2 literacy, has a clear orientation towards one target language. Questions of how students' plurilingual repertoires interact with their academic writing within a collaborative scenario are therefore not explicitly approached in such studies. In addressing this research gap, the present study takes a broader view on linguistic repertoires by critically discussing the actual use (or non-use) of plurilingual resources within oral procedures of formulation. Lehnen and Gülich (1997) have defined the following characteristics of oral procedures regarding formulation within conversational writing interactions: -Pauses or hesitations in the transition from conceptually oral to conceptually literate formulations -Change of speech rate and intonation (rhythmization) -Meta-discursive expressions (e. g. comments on speech) -Orientation towards norms and explicit conditions of constitution for the common text -Cooperative reference to the co-author(s) as a procedure of reassurance and a sign of relationship work Their focus is on L1-writing, hence, an examination of whether a multilingual context reveals different or additional characteristics for formulation practices remains. Also adding to our understanding of oral procedures of formulation, Beißwenger (2017) distinguished between oral and written resources in formulation processes within conversational writing interactions. Oral or interactional resources can be used in order to verbalize ad hoc alternative formulations or supplements to an already written texthere referring to the present status of a written draft at a specific point in the production process. As a result, writers perform ad hoc repairs, which are embedded in oral formulation sequences and, therefore, is comparable to repair in spoken language (see Mroczynski 2014). The written resource of the already existing text helps to conceptually differentiate between the activities of planning and formulating on the one hand, and the activities of reading and evaluating on the other hand. Consequently, and in contrast to the first type of repair, participants in a writing interaction can also carry out repairs at a later stage of their writing processes while revising an already completed text product. The present article focuses on the first type of ad hoc repair.

Plurilingual writing practices
According to the Austrian linguist and didactician Sabine Dengscherz (2019a), plurilingualism in the context of writing can be addressed on three different levels: on an individual level, writers have different language biographies and respective writing experiences in the languages they have acquired and learned throughout their lifetime. On a situative level, writers can work in various languages by, for instance, reading scientific literature in one language and writing an academic text (based on that literature) in another language. Here, the writing task represents a relevant factor for the use of different languages. On a productorientated and textual level, this plurilingual working mode can be (temporarily) visible in the emerging text product (see Canagarajah 2011 for a study on codemeshing in final versions of academic essays).
For the study and analysis of formulation practices with embedded repairs within conversational writing interactions presented herein, these levels shall be considered as intertwined. Although I primarily look at effective plurilingual practices in writing, the students' linguistic backgrounds, as well as the didactic format of the seminarspecifically designed to foster plurilingual practices in speaking and writingmust be considered. From this, I classify the conversational writing interactions of the corpus as multilingual, even if the formulation passages might primarily be marked by bilingual practices. 5 The analysis of bi-plurilingual formulation practices is based on the notion of functional language distribution. As Krafft and Dausendschön-Gay (1999) observed in their study of students writing dyads, when sharing the same L1 (French or German) and writing a text in the same L2 (German or French, respectively), there is a functional language distribution 6 between the two languages. While the L2 is used to propose, correct, dictate or read through formulations, the L1 is used as a common working language in which the participants coordinate their activities, discuss content, comment on propositions, control morphological, syntactical and semantical aspects of expressions or constructions, and debate advantages and disadvantages of a procedure. This kind of functional language distribution can also be observed in my data, on a macro-level. However, on a microlevel, the picture seems to get more complex due to heterogeneous linguistic repertoires, as demonstrated in section 3.3.

Frameworks and pre-conditions
The method of linguistic conversation analysis (CA) pursues conversation theoretical aims by determining "the general in the specific" (Brinker and Sager 2006: 180). Regarding the present subject of examination, the intention is to detect general principles of interactive, dialogical formulation practices. Before carrying out the analysis of exemplary formulation passages, the frameworks and pre-conditions of the conversational writing interactionincluding the material base, the social conditions of the conversation, the linguistic competences of the participants, as well as the interpretative stance of the researcherwill be defined, following the suggestions of Brinker and Sager (2006). These preliminary considerations are meaningful for further analysis and interpretation because they clar-ify under which circumstances a conversational writing interaction has taken place.
The material base consists of an audio recording, a screen recording and a text product. The audio recording is considered central to the analysis, while the screen video serves as a support function and is used selectively in order to provide an impression of the emerging text in addition to the verbal interactions of the conversation. Concerning the social conditions, the conversation can be classified as a writing interaction between three students which took place during a bloc session 7 of a research seminar. 8 The interaction partners, here referred to by the pseudonyms Amna, Carola and Maike, had never written together, and consequently, did not build upon a shared experience of cooperation. Amna took on the role of the writer, using her own computer. The institutional frame of the seminar favored specific interaction norms between the interlocutors, whose social roles were that of learners in an academic, university class. As fellow students, they acted on an eye-to-eye level and were driven by the intention to formulate an academic text, according to their discipline's scientific writing norms. It should be noted that the conversational writing interaction took place in a semi-experimental setting of data collection, namely, the research seminar. This format could have had an impact on the students' attitudes towards the writing task; their awareness of being recorded might have prompted a disposition to putting more effort into fulfilling the assignment than one would typically encounter under natural conditions.
The students had to write a research proposal in French for a fictional study on the topic of 'ways to work with language biographies in a school class'. In preparation for this assignment, they were required to master two material-based writing tasks beforehand (principle of scaffolding 9 ), drafting two question-guided excerpts in French, based on scientific texts. Students were to make use of German guidelines (as formulated by the Faculty of Educational Science at Bielefeld University 10 ) for their research proposals and the writing task presented at the seminar. As a result of their conversational writing interaction, lasting approxi-mately 90 minutes, the students drafted a research proposal of about 18 sentences (see Appendix 1). Regarding the visual text organization, they structured their research proposal into five sections, with the second and fifth sections consisting of just one sentence each. This structure was informed by notes they drafted, in an initial brainstorming phase, for each of the five sections.
In order to fulfill the writing task, the students had to demonstrate a certain ability of academic text composition in French. The following table provides a general overview of the participants and of their linguistic repertoires, at the time of data collection (winter semester 2017/2018). Although Carola and Maike were both enrolled as first-semester students in the Master's program 'Moving Cultures', they had different focal languages: Carola's was French (with English as an additional language), whereas Maike's was English (with French as an additional language). Amna, however, was a bachelor's student 12 in her 7 th semester with the subject combination of Romance studies and 11 The diatopic language varieties spoken by the studentswhich were gathered through reflexive texts about language biographies and linguistic repertoiresare specified in the table.
12 As the seminar's priority was to maximize the degree of linguistic diversity of its participants, it was open to students of different study programs (B. A., M. A., teacher training). political science. The three students shared the competence of English as a first foreign language. Apart from this common resource, they were heterogenous in terms of their first languages and in the order in which they learned additional languages. Both Carola and Maike acquired German as a first language, whereby Maike had an additional first-language proficiency, in Polish. Concerning the target language of French, they showed a similar school career, where French was introduced as a second foreign language. Amna, contrary to her interlocutors, did not grow up in Germany, but in the border region of Switzerland and France. Consequently, she had both Swiss French and, due to her family background, Moroccan Arabic as first languages. Altogether, the interaction partners shared the competence of academic writing in French, however, this might have been developed at different stages in their academic careers, and to different degrees. 13 With respect to my interpretative stance, I had assumed the roles of the teacher and researcher and, consequently, scaffolded and observed the writing interactions without actually participating in them. However, a certain experience with the writing format employed in my research, outside Goethe University Frankfurt (in the form of self-experiments), allowed me to gain an inside view as to the processes of production, although I had not participated in a setting similar to the one employed in this particular research. Having been both an undergraduate and a graduate student at Goethe University Frankfurt, I had a similar academic socialisation to the participants of my study and understood that the format of conversational writing interaction was unfamiliar to them. Nevertheless, the asymmetric relationship between teacher and students on the one hand, and between researcher and researched subjects on the other hand, had to be kept in mind while preparing and carrying out the seminar. 14 Regarding the macro-structure 15 of the conversational writing interaction of Amna, Carola and Maike recorded for my analysis, six higher-level phases of interactive writing and negotiation can be distinguished: 13 There is reason to assume that Amna's formal register of French is more developed than could possibly be the case with the other students (see Maas 2010 for a differentiation of the linguistic repertoire in intimate, informal and formal registers) and that she performs dominantly during formulation processes. This assumption can be confirmed in view of the fact that she takes the role of the writer at the computer, suggesting and controlling formulations throughout the whole writing interaction. 14 I discuss these power relations through the perspectives of role plurality and collaboration in a paper on collaborative action research (see Stierwald, planned publication in October 2020). 15 The macro-structure results from the inventory of the conversation that was created before transcription (see Deppermann 2008 for the steps of data preparation in CA research). 1) Preparation and preliminary talk (0:00-4:29), approx. 4.5min 2) Brainstorming and note-taking (4:30-18:49), approx. 14min 3) Exchange on own school experiences (18:50-25:54), approx. 7min 4) Transition to writing process (25:55/first recording-2:39/second recording), approx. 5min 5) Intense process of negotiation and formulation (2:40-53:29), approx. 51min 6) Short closing (53:30-54:59), approx. 1.5min On the whole, the interactive writing process is structured into two parts: after a stage of experience exchange, knowledge sharing and note-taking (phases 1 to 3), there is a short transition (phase 4), followed by an intense process of negotiation and formulation (phase 5), where all formulation sequences can be situated. Due to organizational and technical circumstances, there were two recordings made (see point 4 in the list), as the group had to change its writing space after 30 minutes of conceptual workthere was confusion regarding room reservations with another group, thus, the audio recording and the screened video were interrupted. Lasting approximately 51 minutes, the processes of negotiation and formulation covered about two thirds of the total conversational writing interaction. The method of detailed sequence analysis, as explained below, will be applied to a collection of passages from phase 5.

Detailed sequence analysis
To begin, my analysis takes a microscopic approach that starts with a detailed sequence analysis of one conversation passage (Deppermann 2008). 16 This passage was defined as a prototypical example based on a collection of formulation procedures with embedded repair(s) 17 for the conversational writing interaction of the three student participants. 18 The example can be considered prototypical because it represents the most frequent way my participants formulated a sentence segment for their final text, with Amna as the dominant formulator and Carola 16 In contrast, a macroscopic approach consists of analyzing the whole conversation on the basis of its structural components. 17 There can be one or multiple repairs inside one passage. 18 The whole corpus consists of 21 conversational writing interactions (see the explanations in the introduction). The interaction of interest lasts for 83 minutes and comprises 35 formulation passages, 17 of which pertain to those with embedded repair(s). and Maike acting as supporters. The prototypical example, as well as the other examples selected for analysis were chosen based on their naturalness. This means that the immediately preceding and subsequent sequences to the focused conversation passages were considered in order to discern context. Then, these passages were analyzed according to different aspects of detailed sequence analysis, including the following (Deppermann 2008): I. Paraphrase and description of action II. Expression design and verbalization dynamic III. Timing IV. Context analysis V. Consequence expectations VI. Interactive consequences VII. Sequence patterns and macro-processes Heuristic questions are linked to each of the seven analytical sections. These are, according to Deppermann (2008), open questions that do not prescribe theorydriven classification categories, rather, they stimulate and systematize the material-based search for features and coherences within conversations. These heuristic questions, with which a conversation passage is analyzed according to the principle of sequentiality, foster a certain analytic sensitivity to the description and do not explicitly result in standardized procedural rules. It is important to note that, only for the prototypical example, all seven heuristics were used. The analysis of all other examples was more selective and synthetic; only three heuristics were used, namely, I, II and IV. 19 After an analysis of the prototypical example, a hypothesis concerning the conversation practice was deduced. This hypothesis consists of assuming that there is a clear functional language distribution between interlocutors' dominant language of verbal communication inside the classroom (German) and a language being studied and trained (French), with the first having the function of a common working language and the second playing the role of a language of formulation (see language usage observed by Krafft and Dausendschön-Gay 1999, Section 2.3). I endeavored to systematically verify the hypothesis throughout the course of my analysis (constitution of subject) and selected additional examples based on the collection of formulations (sampling). Strategic example selection was guided by the criterion of variance of components and forms. 20 Selected examples were once again systematically analyzed according to the principles of detailed sequence analysis (analysis of subject). 21 An examination of the prototypical example is presented in section 3.3.1, followed by an analysis of two deviating examples in sections 3.3.2 and 3.3.3.

Different variants of embedded repair within formulation sequences
The analysis results in three sub-collections of formulations with embedded repair (for the categories of repair referred to hereinafter, see Schegloff et al. 1977): self-initiated and self-executed repair (8 cases), other-initiated and other-executed repair (7 cases), as well as self-initiated and other-executed repair (1 case). In the following sections, the structures, as well as linguistic and interactional particularities of the repair operations, will be explained by means of an exemplary case that illustrates this practice clearly. 22

Self-initiated and self-executed repair
In the first presented passage of the conversational writing interaction, the students began to formulate the question of the research proposal based on the premise that students' and teachers' reflections about individual language biographies reveal advantages for the learning and teaching of languages in a school context. 20 Deppermann (2008) defines the search for variance of components and forms as one of seven strategies of cross-case analysis. 21 Following Deppermann (2008), a fourth step of cross-case analysis consists of a repetition of these three steps until reaching a point of theoretical saturation. Throughout the course of analysis of the conversational writing interaction, these steps will be repeated with the aim of generating a conceptually dense theory (Strauss et al. 1998).
22 It should nevertheless be stressed that theoretical saturation is not yet reached and that, consequently, the presented structures cannot be considered as generalizable practices (see Deppermann 2008 for generalizability as one quality criterion for conversation analysis). Status of certitude of the results has to be checked throughout the course of a deeper analysis of this writing interaction and using it in other writing interactions.

Am [nv]
de la

Am [tr]
exactly there is a phrase ((laughs)) <<laughing> for that that is the problem >

Am [tr]
<<pp> okay what are the advantages of the > (1.5) okay Amna assumes the dominant role of the formulator by initiating and guiding the whole formulation process in this passage. At first, she introduces her proposal with the meta-discursive comment also haben wir gesagt ʻwell, we have saidʼ (segment 71), assigning by the pronoun wir ʻweʼ a certain responsibility for the ensuing formulation draft, to the writing group. She then proceeds in a self-dictating mode, marked by a slow rhythmization and a constant, sometimes slightly increased intonation (segments 72-76). Carola supports Amna in the formulation process by offering the lexeme avantages ʻadvantagesʼ (segment 75), which she had already proposed in the preceding conversation passage as an alternative to the two earlier suggested lexemes bénéfices ʻbenefitsʼ and profit. Amna takes up this proposition, but also adds the lexeme bénéfices, marking it with the aid of the forward slash as a further option (segments 76-85). After a short break, she starts a repair out of self-initiative (segments 86-89). This repair operation is introduced by the German colloquial negation nee, and executed by the repetition of the expression bénéfices, considered as inappropriate in this context, and, lastly, by its electronic deletion accompanied by an incomplete justification is zu wirtschaft ʻis too economicʼ. As Amna practically does not offer any alternative for bénéfices, the repair consists of deleting rather than correcting the lexeme. One might argue that due to the already existing alternative in the form of avantages, the interaction partners see no need to think of another option. Carola approves this decision immediately (segment 89) and Amna repeats the beginning of the sentence thus far redacted (segment 91). After one and a half seconds, Carola offers a completely different formulation for the question of the research proposal, changing the type of the construction from nominal to verbal (segments 94-96). However, Amna does not react to this suggestion and proceeds with formulation, becoming quieter with her voice. Then, Maike speaks for the first time in this passage, posing the metalinguistic question about whether there is a possible variant of Selbstreflexion ʻself-reflectionʼ in French (segments 102-106). By doing so, she returns to Carola's oral pretext réflexion de soi-même ʻreflection of oneselfʼ, uttered at an earlier point in the conversation. Amna interrupts her on the last syllable, replying to this question with a negation and providing the further information that this expression would have to be paraphrased in French. She laughs when denoting this fact as problematic (segments 106-110). After this, she takes up the thread of formulation at an increased speech rate, proceeds with self-dictation and reads, after a pause of 11 seconds, silently through the beginning of the formulation, ending the passage with okay (segments 111-124).
With respect to the languages used in this passage of the conversational writing interaction, one can detect a rather functional distribution between German and French. At the beginning, Amna makes a meta-discursive comment in German. This follows the preceding passage of negotiation, held in German, where the students have already problematized the lexeme bénéfices and looked for alternatives (avantages, profit). Moreover, German can be considered here as a language of metalinguistic consideration, as the semantic and structural characteristics of French are evoked as relevant context, but not further discussed. In contrast to this, French is clearly the language of formulation, identifiable through Amna's mode of self-dictation and Carola's suggestions (which are sometimes framed by German particles like oder ʻorʼ and vielleicht ʻmaybeʼ, confirming the meta-function of this language). With regard to the constitution of subject, as described above in the analytical procedure, this hypothesis of functional language distribution needed to be tested in a further case analysis.

Other-initiated and other-executed repair
After working on the question for the research proposal, the interlocutors turned their attention to the objective of their fictional study and generated the idea that the investigation could help to establish a new pedagogical method of language learning.

Am [tr]
pedagogical method of teaching >

Ma [v]
(.) hmhm, (-) hmhm- In this example, Carola introduces the formulation, recognizable by her audible act of inhaling and her meta-discursive comment peut-être on peut dire ʻmaybe one/we could sayʼ (segment 460). Her presentation of the formulation is marked by pauses and signals of hesitation ähm, which indicates that she is thinking aloud while formulating (segments 460-466). Before finishing her proposition, Amna anticipates Carola's idea and cuts her off (segment 466). While Carola was willing to keep her right to speak by holding her pitch course constant or by slightly raising her voice at the end of an intonation phrase, she shows herself ready to hand over this right to Amna, recognizable by her falling intonation and abrupt closing oder sowas ʻor something like thatʼ (segment 466). Amna intervenes in an increased speech rate, initiating and executing a repair (segments 466-468). This operation is neither introduced by a previous check-back, nor does it have any follow-up elements. The repair is instead uttered immediately with a light hesitation, noticeable by the switch from nouveau 'new (m)' to nouvelle 'new (f)' and the repetition of the indefinite article. It further consists of a specification: the new system of language education becomes a new pedagogical method 24 of teaching, so that the statement is shifted from a macro-to a meso-level. Even before Amna finishes her alternative draft, Carola and Maike interrupt her by expressing their consent to it (segment 468-469).
Here, a functional distribution between the languages is not as evident as in the first example. Apart from some hesitation markers ähm and the particle of consent ja, Carola solely draws upon French: she introduces her formulation with a French meta-discursive comment (in contrast to Amna's German meta-comment in the first passage), expresses the formulation in the target language and also expresses her consent in French exactement, ouais 'exactly, yeah'. Consequently, French can be considered to be the language of formulation and process organization. As a counterargument to this supposed organizational and meta-discursive function of French, one might argue that social and cooperative aspects could play a decisive role in the use of this language. Nevertheless, French keeps its function as language of the repair.

Self-initiated and other-executed repair
At an even later stage in the writing process, the three interaction partners worked on formulating some methodological considerations concerning the temporal frame of the fictional study.

Ca [tr]
and uh then in order to evaluate the (success) (-) 25 This spelling is intentional in order to reflect the specific pronunciation of the speaker. After Amna's invitation to Carola to continue with formulation (segment 1125), the latter resumes the role of the formulator and repeats her idea of the processual character (segments 1126-1132). While doing so, she hesitates in search of the appropriate wording in French so that her way of speech shifts from a conceptually literate point to a rather conceptually oral point within the continuum (see Koch and Oesterreicher 2011), recognizable by colloquial turns such as man braucht_n ʻone needs aʼ or on peut pas f\ le faire dans une semaine ʻone can't do it in one weekʼ. In addition to that, she clicks her tongue and makes a meta-discursive comment comment on dit ça ähm while quarrying for a suitable expression or paraphrase for the large time requirement of the study. The fact that she becomes quieter could indicate her readiness to hand over the right of speech. In direct reaction to this, Amna reads aloud the beginning of the sentence redacted thus far (segments 1133-1137). Her short pauses can thereby be related to her act of correcting the accenting (textually) of the word méthode. Her high-pitched voice at the end of the intonation phrase could be interpreted as another prompt for Carola to proceed with formulation. According to this expectation, Carola elaborates on the sentence, ending the intonation phrase with a high-rising pitch and, in this way, requesting Amna to control or repair her formulation proposal (segments 1140-1141). Amna acts correspondingly and executes a repair by transforming the verbal construction pour évalouer ʻin order to evaluateʼ into the nominal construction un temps d'évaluation ʻan evaluation timeʼ and, thus, changes the meaning of the sentence through the integration of the new information: cette méthode a besoin de temps ʻthis method needs timeʼ becomes cette méthode a besoin d'un temps d'évaluation ʻthis method needs an evaluation periodʼ. Nevertheless, she undoes this wording again in response to Carola's further elaboration of the sentence, as can be perceived in the subsequent segments (1149)(1150)(1151)(1152)(1153). The fact that the repair does not occur or affect formulation in the final text product, unlike in the first two examples, limits its scope for the ensuing writing process. Regarding language use, the functional borders between German and French become blurred again. Carola begins her conceptual considerations in German but switches quickly after the structural particle also ʻwellʼ over to French in order to negotiate the content. Strictly speaking, her expressions do not represent a conceptually literate formulation, but rather a conceptually oral pre-text. Likewise, she makes the meta-comment comment on dit ça ähm to indicate her reflection process in French. However, French keeps its clear function as the language of formulation and repair in this passage of the conversational writing interaction, identifiable by Amna's and Carola's turns. The closing considerations (segments 1149-1167), which ultimately lead to the annulment of the repair, serve in content negotiation and are uttered (by Carola and Amna) in French, however.

Summary of findings
For a preliminary typology of formulation procedures with embedded repair(s), one can differentiate a repair operation according to: a) involved actors; b) functions for the ongoing writing process and/or the writing product; as well as c) their occurrence in the final text. While in the first two examples, repair is initiated and executed by Amna, who controls writing on the computer, in the last example, it is initiated by the co-author Carola, but is nevertheless still executed by Amna. This fosters the dominant role of the person who is actually in charge of writing down the formulations. Furthermore, one can detect the following proce-dures and related functions in the repair operations (see Schegloff 1977 for seven procedures of repair): a withdrawal with justification implying exclusion of a lexeme for the further writing process (3.3.1), a replacement leading to increased precision of formulation in the target text (3.3.2), and a search that indicates an ongoing process for finding an appropriate formulation in the interaction (3.3.3). With regards to occurrence of repair in the final text, there are different repercussions on the writing product. For the first two examples, one can claim that the repair operation leaves a visible trace in the final text, though rather indirectlyin the first example (3.3.1), through lexeme deletion. On the contrary, the repair does not appear in the final text in the third example (3.3.3) and thus, represents an intermediate status of the sentence within the emergent writing product.
Moreover, the hypothesis of functional language distribution between German as a common working language and French as language of formulation, as proposed in the last paragraph of section 3.2, can be refuted. The heterogenous linguistic repertoires of the participants in this particular study generate more complex developments. Indeed, German is used for a range of organizational activities, expressed by discourse, consent or formulation markers, by meta-discursive comments and metalinguistic, as well as conceptual considerations (pretexts) while French is used for suggestions of formulation, repair and self-dictation. However, the usage of French expands to more activities. The participants (mostly Carola) use French for conceptual considerations and meta-discursive comments (that resemble thinking aloud processes), as well as for consent markers. In this regard, there is an interesting pattern in Carola's propositions in Transcript 2, segment 468 and in Transcript 3, segment 1127 where she opens her turn in German and just then switches (back) to French. Here, it would be important to look for similar language patterns in this (and other) writing interaction(s of the corpus), as this could offer further insight into language use on a micro-level. Overall, the findings obtained throughout the iterative course of analysis (constitution of subject, sampling, analysis of subject) in this single interaction, highlight the differentiation of the functional uses of language. In a future analytical step, I will examine these insights within the context of the corpus to see if any further discernments can be perceived.

Concluding remarks
The first aim of this paper was to create a preliminary typology of repair operations within the setting of multilingual and conversational writing interaction. The differentiation examined in three variants reveals revealing insights into formulation and repair procedures undertaken in a writing group of three students.
Nevertheless, this typology must be further developed with a view to other writing interactions of the corpus, which are similar in their basic frameworks and preconditions, but slightly different concerning the number of participants and their respective linguistic repertoires. A typology is normally developed on the basis of a larger sample and already contains a few illustrative types for the examined subject. However, I decided to create a typology for the relatively small sample of formulation passages within one conversational writing interaction because it would establish an analytical point of departure (not to be confused with an analytical template) in order to structure the data material and to define analytically relevant axes. A second aim consisted of analyzing the actual language use of participants and the functions of this usage. In this regard, a bilingual mode could be detected on the data surface, with German and French revealing different functions for the working and formulation process. Unlike the findings of Krafft and Dausendschön-Gay (1999), the functional language distribution seems to be more complex insofar as French, the language being studied and trained, is not exclusively the language of formulation, but also of content negotiation, organization and meta-commentary. Nevertheless, the formulation practice per se, including the different variants of repair, was never undertaken in German nor in any other language of my participants' language repertoires. In this sense, French kept its predominant function as the language of formulation and repair in this exercise of writing interaction.
Given the didactic format of the seminar, which was specifically designed in order to foster an active usage of languages from one's own linguistic repertoire, as well as a reflection upon usage, there emerged some significant questions for further consideration. Why did the students not make use of any other (shared) linguistic resources while negotiating and writing together? Alternatively, how would it have been beneficial to use additional languages that are neither part of the normal communication mode between the participants nor a component of the writing task? Could a greater plurilingual behavior during the writing interaction have increased metalinguistic awareness and language transfer (for the concept of language transfer, see Cummins 2008;Odlin 2013;Haim 2015) 26 and, consequently, have enhanced an elaboration of formal registers across languages?
Regarding the functional distribution of the languages, it appears obvious that the students did not draw on additional linguistic resources because of their lack of specific functionality. If they had discerned an added value in the usage of any additional language, they would have likely been more inclined to apply the resources connected to that language. Based on these considerations and taking into account the idea that functions of language use are not predefined but are rather emergent and dynamic in the context of a conversational writing interaction, bi-plurilingual formulation practices have to be further analyzed and differentiated. Moreover, it would be impactful to examine the transition from oral negotiations of formulations to their written realizations in order to understand the micro-procedures of formulation at a textual level. Which elements of oral communication do interlocutors incorporate into their text, and what kinds of reduction do they undertake? What criteria contribute to the success or acceptance of an oral repair operation? Considering seminar formats within the field of higher education, there is also incentive to reflect upon the potential of the studied formulation procedures for writing strategies in philological disciplines.