Self-translations in multilingual workplace interaction

Example 1:

Lightly (Publication 3, 2:24–2:57)

After Katja’s longer turn (line 9), Jegor repeats the word “kevyesti” in a soft voice (line 12) with level intonation. Partial repeats, especially when accompanied by questioning prosody, typically implement repair initiation (Dingemanse et al. 2014; Schegloff et al. 1977; for Finnish conversation see Kurhila and Lilja 2017). Jegor’s repeat is indeed treated as a sign of language trouble by Katja: She first provides the adjective (“kevyt”, line 14), from which the adverb kevyesti is derived, and after that utters the Russian translation of the adjective. Jegor receives the translation with an acknowledgement token (line 16).

As Example 1 shows, using a translation can be an efficient way to achieve intersubjectivity. As the repair initiation points out the problematic item, the co-participants can exploit the resources of the other language they share; the repair operation can be accomplished through a single adjective. Explaining the word in the original language (Finnish) would take more time and effort. Using translation as a repair thus helps keep the repair sequence short and thereby concurs with the preference for the progressivity of the talk (Stivers and Robinson 2006).

Previous research has shown that in L1-L2 conversations, L2-speakers’ repeats are generally treated as indicating the speaker’s unfamiliarity with the repeated word (Lilja 2010). It is worth noting, however, that in this case the more proficient speaker A does not produce a straight translation, which would resolve the unfamiliarity problem, but instead provides the Finnish adjective (kevyt, ‘light’) that is the root for the target adverb (kevyesti, ‘lightly’). In other words, she does not act just as an interpreter, translating the repeated word verbatim. Instead, she offers speaker B linguistic resources to analyse the unfamiliar word linguistically by relating the adverb (that was used) to the adjective root morpheme. As Russian translation, she also provides the basic adjective (лёгкий), not the adverb (легко). Thus, while the more proficient speaker A resolves the trouble by giving a translation, she does so pedagogically by giving linguistic information to speaker B, encouraging him to engage in learning rather than simply equipping him with the verbatim translation.

The next example illustrates an instance of a more indeterminate other-initiation of repair. Instead of focusing on any specific linguistic item, speaker B produces an open-class repair initiation (Drew 1997). The extract comes from professional training in a metal workshop in Germany. The instructor (WS) is building a bicycle stand with the trainee (DE). They have just welded two tubes together and now have to check that the metal bars are straight.

Example 2:

Red waterlevel (Geschnitten_Perfplus_Metall_5.8._teil8; 00:49-01:12)

After welding two pipes, the instructor announces “here we have to be flush” (line 01), meaning that they have to make sure that the two pipes have the same height. To check this, they need a water level. The trainer stands up, scans the environment (lines 09–12) and asks the trainee if they have a water level (line 12). The trainee does not respond to the question, but continues to look at the welded pipes (Figure 1). Since the two interactants are standing very close to each other, it is not likely that the trainee did not hear the instructor. Rather, it seems that the trainee does not react because he did not understand the instructor’s question. In order to get the trainee’s attention, the instructor addresses him by name (line 14, Figure 2) and tells him that they now need the red water level (line 14). He points towards a domain of scrutiny in the trainee’s back, making clear that he requests the trainee to fetch the object. The trainee now indexes his lack of understanding with an open-class repair-initiation “uh” (line 15, Figure 3). The trainer’s continuing point in the direction of the trainee’s back makes the latter turn around (lines 12–15, Figure 4). Yet, only when the trainer resorts to a partial self-translation of line 14 into the lingua franca English, saying “the red (.) the red” (line 16), the trainee recognizes the targeted object and walks off in the right direction to fetch the water level lying on the workbench (lines 18–24). In this case, a minimal, partial self-translation, containing only the colour adjective specifying the desired object, is sufficient to lead the addressee to a proper understanding of a request that was not understood in two earlier versions in German. It may be noted that the choice of precisely only the colour term is a way of most economically securing reference concerning the requested item, as the action meaning as a request can be taken as understood, while the technical term water level to be used for object-identification is probably known neither by the instructor nor the apprentice in English.

In the examples in Section 4.1, self-translations are produced by Speaker A as a potential remedy to the interactional trouble that Speaker B has signalled by his other-initiation of repair. Speaker B’s repair initiation can be specific, pointing out the exact word that causes trouble (Example 1), or an open initiation (Example 2) that signals that Speaker B is at a loss as to how to respond to Speaker A’s previous talk. It is worth noting that the self-translations in our data are very minimal. In Example 1, Speaker A merely translates the root of the problematic word repeated in the repair initiation (kevyesti vs. kevyt); in Example 2, the translation repeats just the content word (red) of the original request, which most economically allows for identifying the problematic referent. In other words, self-translators aim to resolve the interactional problem as economically as possible, making use of their shared linguistic repertoire so as to provide just enough linguistic elements to secure intersubjectivity and the progressivity of the course of action. In our cases, notwithstanding the specificity of the repair initiation, Speaker A treats the trouble as stemming from restricted language skills, and attempts to repair the trouble by exploiting linguistic elements from other languages.

Next we will discuss instances where the self-translation is not preceded by a repair-initiation. Nevertheless, in these cases the translation still serves as a means to resolve problems of understanding.

Example 3:

have you worked (Gastro_2.Kurs_1.Phase_3.Tag_00035; 04:48–05:02)

The instructor first asks the apprentice an open wh-question about where he worked (line 01). Since they had talked about the apprentice’s wishes for a future job before, the instructor puts particular stress on the perfect auxiliary “HAST” in order to indicate the topic shift to the apprentice’s professional biography. As the latter does not answer, the instructor reformulates his question, now using a thematically narrower polar interrogative (see Svennevig 2013), which inquires if the apprentice worked in a restaurant or at a hotel (line 03). The apprentice confirms (line 04). The instructor initiates repair, however, again highlighting the auxiliary (line 05). The prosodic design of his partial self-repeat indexes that he assumes that the apprentice has not understood the grammatical/topical shift to the past and therefore has confirmed erroneously.^[2] After a short lapse, the instructor self-translates his question from German into English, again using the analeptic short variant “HAVE you worked” (line 07), thus presupposing that the apprentice has understood and retained the object reference (hotel). The apprentice now disconfirms (line 08). With a particle response, the instructor treats the apprentice’s self-correction as conforming with his expectations (line 09). After the apprentice’s reconfirmation (line 11), the instructor indexes that the communicative task has been successfully completed and closes the sequence (line 12).

The next instance of self-translation after an inadequate response represents a case where the L2 speaker B produces an answer that does not fit the formal expectations of the question by Speaker A. The example comes from the Finnish NGO data. The extract is from a remote meeting, where Risto, the head of the organization, asks the employees to describe their ongoing projects. Jegor has just reported in Finnish that their newsletter has received 330 new subscribers during the weekend due to a publicity campaign they launched earlier. The extract starts with Risto’s follow-up question (lines 01–05).

Example 4:

Three hundred now or during the weekend (IMM1805, 5:51–6:23)

After giving positive feedback (line 01), Risto asks a specifying question: did Jegor mean 300 new subscribers during the weekend or since the launching of the campaign (lines 02–05)? As a response, Jegor quietly produces a very minimal response particle (line 06), which could function as a continuer. However, since Risto has just asked an alternative question, the particle does not fit as a type-conforming answer (cf. Raymond 2003): It does not determine which one of Risto’s propositions was correct. Jegor’s insufficient response is followed by a lengthy pause (line 07), after which Risto asks whether his question was difficult (line 08). He thus makes explicit that Jegor’s answer was inadequate. After another pause, Risto begins translating the question into Russian. Doing so indicates that he has interpreted the problem as Jegor’s limited skills in Finnish. When Risto has reformulated the alternatives in Russian, Jegor begins answering the question (lines 20–21). The self-translation by Speaker A has thus resolved the understanding problem and allows the conversation to progress.

Responses that attest to a misunderstanding or non-understanding can also be embodied. This is particularly relevant in the case of requests and instructions that call for embodied compliance, as in Example 6. It comes from the German catering course. The instructor (FR) has started rolling dough for cookies. He asks the apprentice (AP) to continue his work, getting the dough even thinner (lines 03–04).

Example 5:

much thinner (Gastro_1.Kurs_2.Phase_1.Tag_00006; 00:12-00:50)

The instructor asks the apprentice to take over rolling the dough until it is much thinner (lines 03–04). He accompanies his directive with a distance gesture (Streeck 2008). The apprentice grasps the cookie cutter, presents it looking at the instructor (line 05, Figure 5). In this way, he produces an embodied candidate understanding (Antaki 2012) of the next action that the instructor’s request makes relevant (i.e., cutting the cookies). Obviously, this understanding does not depend on linguistic comprehension of the instructor’s turn, but on pragmatic inferencing. It seems to build on knowledge about the workflow of baking cookies, mistaking the instructor’s distance gesture as referring to the cookie cutter, expecting the apprentice to perform the next step of the procedure. Yet, this attempt at replacing a lack of linguistic knowledge by pragmatic (professional) competence fails. The instructor takes the cookie cutter and puts it aside (Figure 6). He treats the apprentice’s inadequate response as caused by an insufficient knowledge of German and switches to a lingua franca: He self-translates his request from lines 03–04 from German into Spanish “MAS fino” (line 06). This third-position self-repair is followed by the instructor continuing his code-switch into Spanish, instructing the apprentice how to roll the dough properly (lines 08–14). The instructor only switches back to German (lines 16/18) when AP executes the task properly. In this case, the apprentice’s embodied response attests a misunderstanding that obviously rests in not having understood the instruction. The self-translation resorts to a lingua franca (Spanish), the choice of which is based on specific personal knowledge about the other’s language proficiencies. In other cases in the German data, English is used as a default lingua franca in the absence of specific information about the addressee’s language proficiency (cf. Extract 2).

In Section 4.2, we have presented three examples in which Speaker A produces a self-translation after an inadequate response by Speaker B. We have referred to three different types of responses as “inadequate”: in Example 3, Speaker B produces a turn that is unlikely to be true; in Example 4, Speaker B’s answer is not type-conforming and does not deliver the projected information; in Example 5, Speaker B produces an embodied candidate understanding that makes a misunderstanding of the projected action available. In all these cases, Speaker B’s response contradicts either the known facts or the projections created by the previous turn. The examples in this section resemble those presented in Section 4.1, as self-translations are used to secure intersubjectivity. However, self-translations after inadequate responses seem to belong to a more restrictive environment; they only occur after a strong normative expectation of a response has not been fulfilled (e.g., after first pair-parts of adjacency pairs). In such contexts, understanding must be demonstrated by a suitable verbal or embodied response. If Speaker B does not respond in a projected way, an understanding problem becomes obvious and Speaker A responds with a self-translation.

In the final section of our analysis, we will examine cases where a self-translation by Speaker A follows Speaker B’s turn that somehow fails to show affiliation with the preceding talk.

Example 6:

go go go (Gastro_1.Kurs_2.Phase_2.Tag_00033; 05:44–06:13)

The instructor (FR) asks the apprentices to start serving the dishes (line 02), specifically addressing the apprentice Alim (AL) again in line 06. However, the instructor at the same time realises that some of the dishes are not yet ready to be served and need some adjustments, which he requests from the apprentices (lines 02–03 and 08). He hands a plate to Alim (Figure 7), but then adjusts the items on the plate himself (Figure 8). In this way, a complex, possibly contradictory expectation of immediately serving versus first improving and only then serving the dishes is created for the apprentices. In other words, it might not be clear to them if the dishes are already ready to be served or not. Alim confirms (line 09), but does not grasp the plate again, which he put down in the meantime (Figure 9). In line 11, FR responds to AL’s inaction with an upgraded, repeated request to dispatch the dishes. He first uses a repeated German imperative (“GEH GEH DU KOMM,”), then switches to an English self-translation in a fast and loud voice, again repeating the imperative (“>GO GO GO,<”; Figure 10).^[3] This turn-design conveys urgency of the requested action (Mondada 2017). Already in overlap with the instructor’s repeated request, Alim accounts for his inaction by asking “nicht zusammen” (‘not together’, line 12). He thus indexes that his inaction rests on a discrepant expectation concerning the serving procedure: While the instructor asks the apprentices to serve each dish as soon as its preparation is finished, Alim (who is an experienced servant) expects that several plates should be served at once. This implies that he waits until more plates are ready for serving. His account makes clear that his inaction was not caused by any problem of linguistic understanding concerning the instructor’s requests, but was motivated by the assumption of a different prodecure for distributing the food. In line 13, the instructor rejects Alim’s alternative logic, claiming that it will take too long. However, Alim still does not comply with the request and instead asks to wait for still another plate (line 17). The instructor responds with another upgraded request to start serving immediately, again using repetition, loud and fast prosodic delivery, and self-translation (line 18).

In this extract, self-translation does not solve the action coordination problem. While by using self-translation, Speaker A treats Speaker B’s lack of compliance as a linguistic problem, B’s responses make clear that there is no linguistic understanding problem, but a mismatch of expectations concerning the task-related workflow, which is at issue. However, even when Speaker A’s and B’s different conceptions have become obvious to both of them, A still uses self-translation as a way to insist on his directive, as B continues not to comply.

Our last example also illustrates a case in which the L2 speaker rejects the need for self-translation, claiming explicitly that she has understood the previous turn. Example 7 is from the Finnish NGO data, a bit later on in the same conversation as Example 1. The participants have talked (switching between Finnish and Russian) about different ways of collecting feedback at an event. Before the extract, Risto has proposed distributing a questionnaire to the participants, after which Jelena has posed a question (in Russian) about the relevance of filling out a form on the spot. In the extract, Risto explains in Finnish why he thinks it is a good solution.

Example 7:

Everybody participates (Museum planning 1, 234, 5:50–6:29)

Risto gives reasons for his opinion in lines 01–08, summarising his point of view (‘getting everybody to participate’) in lines 12–14. Katja acknowledges this summarising turn verbally (line 15), but Jelena does not say anything. She produces a series of slight head nods (line 14), similar to the ones she produced earlier, while Risto explained his reasoning (lines 05, 06, and 12). At the end of Risto’s turn, Jelena shifts her gaze from Risto, first towards Katja (line 14), then down (line 15) and then forward, looking somewhere between Risto and Katja (line 16).

Jelena’s participation is thus fairly minimal after Risto has summarised his point of view. Given that Risto has provided an explanation for why they should do the action Jelena has questioned earlier, it seems that Risto is waiting for a more substantial response from Jelena. When no response follows (line 16), Risto explicitly asks Jelena if she has understood what was said (line 17), now switching to Russian, which is Jelena’s strongest language. Both his question and his code-switch imply that he treats Jelena’s minimal participation as potentially signalling some trouble of understanding. Almost at the same time, Jelena responds with the particle “okei” (line 18). The particle indicates ‘no problem’ and could also be interpreted as accepting Risto’s suggestion (Koivisto and Sorjonen 2021). Right afterwards, Jelena glances at Risto and produces a minimal response accompanied by several nods (line 20). However, Risto does not treat Jelena’s actions as a sufficient response: He switches to Russian and starts translating his prior turn (lines 22–23). In response, Jelena makes explicit that she understands, using a prosodically prominent (high onset) verbalisation of understanding (line 25). Risto acknowledges Jelena’s turn by reduplicated confirming particles (line 27); he leans backwards, displaying nonverbally that he withdraws from his prior action, and does not continue his translation.

Jelena’s position as a second language-speaker of Finnish makes it possible for Risto to treat her minimal responses as a sign of an understanding trouble. It is worth noting, however, that Jelena and Risto have had different opinions of how to perform the action in question; therefore, Jelena’s minimal participation may also index reservations against Risto’s proposal. Jelena’s restricted language skills in Finnish offer Risto an additional opportunity to promote his idea; by switching to Russian, Risto manages to verbalise his opinion once more and thereby pursue a (positive) response from Jelena. However, Jelena treats Risto’s Russian turn as specifically targeting a trouble of understanding – she does not comment on the content of the turn, but verbalises that the translation is not necessary.

In Examples 6 and 7, Speaker A uses self-translation to deal with Speaker B’s lack of affiliation concerning A’s prior action, such as a request (Example 6) or a proposal (Example 7). As in the cases of inadequate responses (4.2), self-translation is used as a third-positioned resource to repair a problem of intersubjectivity. However, the cases in Section 4.3 do not concern problems of (linguistic) understanding, but rather problems of non-shared expectations concerning the organisation of an activity (Example 6) or a difference in opinion (Example 7). That is, it is not so much the notional-epistemic side of intersubjectivity that is at stake here, but the more practical-normative side concerning consensus about appropriate procedures and (joint) actions. Sometimes, the L2 speaker clearly indexes that their response to A’s action is not caused by a lack of understanding (as in Example 7), but, for example, by a different idea of how to perform a task in a joint activity (see Example 6). Nevertheless, A’s choice of self-translation as a resource for redoing a first action and eliciting a modified, more acceptable response from B exploits the possibility of treating B’s response as indexing a linguistic understanding problem. In other words, self-translation is used here as a resource for pursuing affiliation that specifically builds on affordances created by the particulars of a participation framework involving an L2 speaker. However, since proficiency-related understanding problems of the L2 speaker occur repeatedly (to a varying extent), there seems to be a continuum between:

1. cases in which self-translation is used in response to an action by B that A considers to be insufficient because of B’s linguistic non- or misunderstanding, while in fact B has understood correctly, but does not affiliate; and

2. cases in which A uses self-translation as a resource that strategically treats B’s dispreferred response as seemingly based on a linguistic problem.