This is the accepted version of the article: Vargas-Urpi, Mireia. «When non-renditions are not the exception. A corpus- based study of court interpreting». Babel, Vol. 65, núm. 4 (2019). DOI 10.1075/babel.00103.var This version is avaible at https://ddd.uab.cat/record/211416 under the terms of the license This is a post-print version of the following article: Vargas-Urpi, Mireia (2019). When non-renditions are not the exception. A corpus-based study of court interpreting. Babel, vol. 65, n. 4. DOI: 10.1075/babel.00103.var When non-renditions are not the exception A corpus-based study of court interpreting Mireia Vargas-Urpi Universitat Autònoma de Barcelona This article seeks explore the nature and function of non-renditions in a corpus of transcriptions of 55 authentic interpreted court proceedings celebrated in Barcelona (the TIPp corpus). By doing so, it establishes a dialogue with Cheung’s (2017) contribution about non-renditions in court interpreting in Hong Kong. The transcriptions of the TIPp corpus were annotated using the software EXAMARALDA following Wadenjsö’s (1998) distinction between “talk as text” and “talk as activity”. Non-renditions were considered a part of “talk as activity”. A distinction was made between justified non-renditions, i.e. those that were used to ask for a pause to interpret, to ask for clarification, to confirm possibly misheard information and to retrieve parts of the original message in case of a lapsus, and unjustified non-renditions, e.g. when interpreters give advice to the defendants or warn them, when they answer on behalf of defendants, or when they supply information not provided in the original utterances. The findings reveal alarming averages of non-renditions in the bilingual parts of the trial (58.3 per bilingual hour), with a higher ratio of unjustified non-renditions. These findings have a clear correlation with the poor working conditions of court interpreters in Spain and reveal an urgent need of professionalization of this practice in this country. Keywords: court interpreting, non-renditions, impartiality, TIPp corpus 1. Introduction ‘Non-renditions’ were first introduced by Wadensjö (1992, 1998) as part of a taxonomy that includes various kinds of renditions produced by dialogue interpreters. In Wadensjö’s words, the taxonomy was developed to go beyond the traditional units of interest in translation studies – source text and target text – and better account for the complexity of naturally occurring interpreter-mediated face-to-face interaction, in which both the progression and the content of talk are continuously shaped by what participants add and what they take up. (2015: 352) ‘Non-renditions’ are interpreters’ utterances or responses that do not correspond to any prior original utterance. Wadensjö analyses non-renditions as part of interpreters’ discourse management or coordinating tasks, which include ‘explicit coordinating moves’, as when interpreters ask for clarification or for a pause to interpret (Wadensjö 2015: 353). According to Cheung (2017: 174), this categorization of interpreters’ utterances helps determine if the interpreters “are playing additional roles” beyond that of translating what has been said. Cheung (2017, based on Wadesnjö 1998) explains that non-renditions in court interpreting can be textual or interactional. Textual non-renditions are used to ask for clarification in situations of misunderstanding or communication breakdown (Mikkelson 2008), while interactional non- renditions are more related to the coordination of turn-taking. There seems to exist the assumption that non-renditions are less frequent in court interpreting than in other contexts (e.g. healthcare interpreting), because of the “more restrictive” nature of the court setting. For instance, in Spain, the use of the term ‘intercultural mediation’ to refer to healthcare and educational interpreting practices seems to reflect this broader flexibility towards non-renditions in these contexts. Contrastingly, the use of the third person by court interpreters to refer to themselves in countries such as the USA (Angermeyer 2015) is a good example of the restrictions the court poses to interpreters when providing utterances of their own. Nevertheless, previous research has shown that non-renditions are also common in court interpreting and worth exploring to better understand court interpreters’ behaviour (Angermeyer 2015, Berk-Seligson 1990, Cheung 2017, Hale 2004, among others). This article analyses non-renditions in a corpus of 55 court proceedings collected in Barcelona: the TIPp corpus.1 By doing so, it establishes a dialogue with Cheung’s (2017) study of non- renditions in court interpreting in a corpus collected in Hong Kong. It also aims to further explore how the analysis of non-renditions can help us explain court interpreters’ behaviour and roles. The article contributes to a better understanding of the reality and quality of court interpreting in Spain’s criminal courts, which, before the existence of the TIPp corpus, “had never been tackled in a systematic and rigorous descriptive study using a representative oral corpus” (Orozco-Jutorán, 2019). 2. Non-renditions, visibility and impartiality Cheung’s (2017) article provides a thorough review of the evolution of non-renditions in previous research in dialogue interpreting in general and court interpreting more specifically, including the debate of whether certain kinds of non-renditions should be avoided in court interpreting settings because they can compromise “the perceived impartiality of interpreters”, among other reasons (Cheung 2017: 178). Non-renditions certainly reflect interpreters’ visibility (Angelelli 2004), which has been extensively documented and accounted for in dialogue interpreting research. In fact, Ozolins (2016) uncovers the myth of invisibility when pointing to the absence of any identifiable literature or codes of ethics advocating invisibility. The question is thus whether or to what extent these ‘visible’ utterances affect impartiality, a key principle in both public service and court interpreting codes of ethics. 1 The acronym stands for “Traducción e Interpretación en los Procesos penales” (Translation and Interpreting in criminal Proceedings). Cambridge’s impartial model about public service interpreting in general is clear about the kind of situations where an interpreter may use a non-rendition: Interpreters using the impartial model relay on messages accurately, completely and in as closely as possible the same style as the original. They do not give personal advice or opinions; do not add or omit parts of the message; do make every effort to foster the full, accurate transfer of information; do maintain strict confidentiality. They will intervene only when they need clarification of part of a message; they cannot hear what is being said; they believe a cultural inference has been missed; they believe there is a misunderstanding. (2002:123) The codes of ethics of the NAJIT and the APTIJ (Spanish Professional Association of Court and Sworn Interpreters and Translators) also mention that “[c]ourt interpreters who do not hear or understand what a speaker has said should seek clarification” (NAJIT 2002). However, asking for a clarification can be controversial, because it may impinge on the speaker’s strategy of ambiguity (Niska 1995, cited in Cheung, 2017: 178). Codes of ethics for court interpreting do not usually foresee non-renditions to explain cultural differences or to warn in case of misunderstanding. Instead, the code of ethics of the APTIJ suggests that “[c]ultural terms that have no direct equivalent in Spanish [the language of the court] or that may have more than one meaning must remain unchanged”. Another problem concerning non-renditions is that participants who do not know the language in which the non-rendition is uttered are excluded from that monolingual exchange (Cheung 2017:178). Non-rendition to ask for clarification may certainly result in a monolingual conversation over the meaning of the unclear part, letting other participants aside. One of the few codes of ethics that covers this kind of situations is the guide proposed by García-Beyaert et al. (2015), which includes the principle of ‘transparency’: “the interpreted encounter must be transparent so that everyone knows what is happening at any time. (…). The interpreter should also interpret his or her own utterances whenever he or she has to intervene and speak as the interpreter” (2015:9). Note that this code of ethics is not exclusive for court interpreting but for community interpreting in general, thus one may wonder to what extent this ‘transparency’ principle would be feasible to apply in a court, where the interpreter has limited scope for turn- taking or conversation management. 3. Interpreters in Spanish courts This section focuses on the description of the context where the data collection took place (i.e. Spanish courts), which may help understand and contextualise the findings presented in the following sections. The leitmotiv in any contribution about court interpreting in Spain is usually the poor working conditions for interpreters in Spanish courts, which directly affect quality. Before the transposition of the Directive 2010/64/EU, which took place in April 2015, the only requirement for a person to work as an interpreter in a court of law was to acknowledge a certain level of competency in a foreign language (Ortega Herráez 2006: 54). The transposition of Directive 2010/64/EU turns translation and interpreting into an essential element in the rights to effective legal protection in the exercise of lawful rights and interests before the courts in order to avoid any state of defencelessness, namely the right to be informed of the accusation against them, the right to a public process with all procedural guarantees and the right to self-defence, as enshrined in Article 24 of the Spanish Constitution. However, the directive is not explicit in terms of interpreters’ qualification to work in a court and, for the time being, court interpreters in Spain are still not required to pass any kind of accreditation process (either by means of an exam or specialised formal education) to interpret in a court of law. The provision of the service is outsourced in most regions in Spain, as Del Pozo Triviño and Blasco Mayor explain in the following quote: [A]dministrations publish a tender to which private companies present bids. The fact that the company acts as an intermediary means that the rates received by interpreters are reduced significantly, which in turn means that many professionals do not accept the terms and conditions imposed by the companies awarded the bid and the companies in turn seek to recruit people with little or no training or experience (2015: 47). The lack of specialised training often results in poor quality in the interpretation provided, as had already been pointed out by Onos (2014). In this respect, the first results of the TIPp project (see Arumí et al 2017; or Orozco-Jutorán, 2019) are alarming: on average, a 54% of the total duration of the court proceedings is not interpreted to the person with limited competency in the language of the court. Not only this, but the frequencies of interpreting mistakes (omissions, additions, distorted meaning) in the parts that are actually interpreted are also high, with an average of 21 serious mistakes per bilingual hour (see Orozco-Jutorán, 2019, for more information about this). Specialised education would help improve the quality of court interpreting, but the interpreter is not the sole responsible for quality: it should be a shared responsibility with other participants in the interpreted event. In tune with this, the TIPp corpus has also revealed high speech rates in judges and lawyers’ turns, overlapping speech and interruptions, which are also important obstacles to provide accurate interpretation (Vargas-Urpi 2017). 4. Method and corpus For the purpose of the project “The quality in translation as an element to safeguard procedural guarantees in criminal proceedings: development of resources to help court interpreters of Spanish - Romanian, Arabic, Chinese, French and English”, 2 the MIRAS research group 2 For more information about the research project see: http://pagines.uab.cat/tipp/en obtained access to the video recordings of criminal proceedings where interpreting took place between 2010 and 2015 in almost half of the criminal courts in the city of Barcelona. Due to limited time and resources, the first stage of the project focused on the court proceedings celebrated during the first semester of 2015 (January-June). Three language combinations were selected for this first stage: English-Spanish, French-Spanish and Romanian-Spanish. This resulted in a corpus of 55 court hearings which lasts 1,116 minutes in total (18.6 hours) and amounts 190,000 tokens.3 Table 1 presents detailed information of the corpus. The total duration corresponds to the total minutes of the court proceedings, whereas the bilingual duration (339.2 minutes) refers to the minutes in which the interpreter provides interpreting and is audible in the recording (i.e. the bilingual minutes do not include whispering). Interpreting was Total duration Bilingual Languages Court proceedings requested for the (minutes) minutes defendant EN/ES 19 18 371 123 FR/ES 9 7 190 92 RO/ES 27 26 555 124.2 Total 55 51 1116 339.2 Table 1. The TIPp corpus of authentic court proceedings This corpus is substantially smaller than Cheung’s (2017) 100-hour corpus in terms of length, but it is larger if we consider the number of hearings included: 55 in the TIPp corpus as opposed to Cheung’s 5 cases. Nevertheless, in the study of specialized discourse, the quality or content of the data may be more relevant than the size of the corpus (see Orozco Jutorán, 2019). 3 For more information about the corpus, see Orozco Jutorán (2017; 2018; 2019) and Arumí & Vargas- Urpi (in press/2018). Furthermore, we believe the TIPp corpus is representative mostly because its random sampling, “where every subject of the population being studied has an equal chance of being selected as a participant in the research” (Orozco Jutorán, 2019), and thus can be considered generalizable. In addition, the corpus consists of authentic, naturally-occurring data, where researchers have been non-participant observers. The court proceedings were transcribed verbatim. The EXMARaLDA4 software was used both as a transcription and annotation tool. The EXMARaLDA interface resembles a musical score, with each speaker’s interventions annotated on a different tier (see figure 1).5 Tiers can be added below the transcription for annotations. The software was especially useful because it also allows the creation of ad hoc categories for annotation and the extraction of quantifiable data. The annotation categories created followed Wadensjö’s distinction between ‘talk as activity’ and ‘talk as text’ (Wadensjö 1998: 21) and had already been applied in previous research by the MIRAS group concerning PSI in social services and education (Arumí and Vargas-Urpi 2017). “Talk as activity” covered three broad categories: conversation management problems, non- renditions, and direct vs. reported speech (see Arumí & Vargas-Urpi, 2018). 4 See: http://www.exmaralda.org/en/ We would like to express our gratitude to Bernd Meyer, who suggested the use of this software, and Thomas Schmidt, developer of the software package, who so kindly helped us with this research. 5 For more information about the transcription using EXMARaLDA, see Arumí and Vargas-Urpi (2017) and Orozco Jutorán 2018. Figure 1. Screenshot of EXMARaLDA’s interface Two tiers were used to annotate the information concerning non-renditions. The first tier for annotating non-renditions considers whether they are justified (J), unjustified (NJ) or reactive tokens (RT).6 The second tier specifies the type of non-rendition: - Pause (J): when the interpreter asks for a pause to interpret; - Clarification (J): when the interpreter asks for clarification or explains something that was expressed ambiguously; - Confirmation (J): when the interpreter seeks to confirm that he or she has understood or heard the information clearly; - Retrieval (J): when the interpreter is aware that he or she is missing some information and asks to retrieve it; 6 Reactive tokens refer to non-renditions where the interpreter merely acknowledges that she or he received the information in the original utterance; for instance, when she or he simply says “vale” (okay) before providing the rendition. This article focuses only on justified and unjustified non-renditions. - Advice, instructions, warning (NJ): when the interpreter warns the defendant on how to behave, gives instructions on how to answer a question, or gives him or her advice of any kind; - Answer (NJ): when the interpreter answers on behalf of the defendant, probably because they have already discussed the situation with the defendant or with his or her lawyer prior to the trial; - Extra information (NJ): when the interpreter gives information to any of the participants or asks them questions not posed in the original utterances. The annotations introduced in EXMARaLDA were exported into Microsoft Excel workbooks created specifically for the present project. Each workbook contains results for a single language combination, with a separate spreadsheet for each trial, plus an additional summary sheet for all the trials for that language combination. One of the limitations of this study is that we did not receive any kind of background information about the specific interpreters that appear in the recordings of the corpus (for instance, about their previous education). Thus, we can only refer to the general situation of court interpreting in Spain (see section 3 above) to account for interpreters’ behaviour and use of non-renditions. 5. Findings In order to contextualise the use of non-renditions by the interpreters in our corpus, it must be reminded that, on average, only 46% of the court proceeding is interpreted to the defendant, either in short consecutive (30%) or whispered (16%). In other words: more than a half of what is said by judges, lawyers or witnesses during a trial (54%) is not interpreted to the defendant. 5.1. Relative frequency of non-renditions Table 2 presents the average frequency of non-renditions in the TIPp corpus. The figures correspond to the number of non-renditions per bilingual hour. Average per Average Range Languages bilingual hour deviation EN/ES 94.1 64.91 240 FR/ES 28.0 32.44 126 RO/ES 92.8 77.6 600 Average 75.7 58.32 -- Table 2. Average number of non-renditions per bilingual hour. The table shows that the language pair with most non-renditions was English-Spanish, with an average of 94.1 non-renditions per bilingual hour (this corresponds to 1.6 non-renditions per bilingual minute, on average). This proportion is followed closely by the Romanian-Spanish language pair, with 92.8 non-renditions per bilingual hour. These averages triple that of the French-Spanish language pair, with 28 non-renditions per bilingual hour. The average deviation and the range values reflect the high dispersion of the data analysed. It must be noted that in the subcorpus of each language combination there are trials with zero non- renditions. In other trials, very short bilingual parts (i.e. the interpreter only speaks for some seconds) contain various non-renditions, resulting in a higher average number of non-renditions per bilingual hour for that specific trial. For example, in a court hearing with Romanian-Spanish interpreting, the interpreter intervened only once, lasting less than 10 seconds. However, her intervention included two non-renditions, as may be noted in extract 1. Extract 1. Judge: Pregúntele si tiene algo más que alegar en su defensa. Ask him if he wants to allege anything else in his defense. INT: Dacă tu doreşti să spui ceva în plus în apărarea ta ce nu ai spuis până acum. Adaugi sau nu? Whether you want to say anything else in your defense, something you still haven’t said. ¿Do you add something or not? When these two non-renditions are considered in terms of ‘number of non-renditions per bilingual hour’, the result in an average of 600. It is the only example of this kind in the corpus. Nevertheless, the relatively high dispersion of the data analysed also accounts for the differences between the final results presented here and the results of the pilot project presented in Arumí and Vargas-Urpi (2018). 5.2. Justified vs. unjustified non-renditions For the purpose of this project, a distinction was made between ‘justified’ and ‘unjustified’ non- renditions, as already mentioned before. This distinction seemed important to us, because one of the final goals of our project was to evaluate the quality of court interpreting in the corpus analysed and to make recommendations on how to improve it. In tune with this, some non- renditions were considered justified because they were used as coordination strategies (when interpreters ask for a pause or to retrieve possibly misheard information), or as strategies to better understand the original utterance (when they ask to confirm or clarify information). This is coherent with Wadensjö’s (1998) and Cheung’s (2017) interactional-oriented vs. text-oriented non-renditions. Unjustified non-renditions do not respond to those textual or interactional orientations but reflect that the interpreter sometimes trespasses the principle of impartiality by giving advice, warning, answering or providing further information. It must be noted, though, that Cheung (2017: 181) includes these kinds of non-renditions in the interactional orientation. Table 3 presents the average frequency of each kind of non-renditions in the TIPp corpus. The figures correspond to the number of justified and unjustified non-renditions per bilingual hour. Language Justified Unjustified EN/ES 26.3 50.2 FR/ES 14.4 11.9 RO/ES 25.6 65.7 Average 22.8 45.5 Table 3. Average number of justified or unjustified non-renditions per bilingual hour. The total average (including the three language combinations) shows that unjustified non- renditions nearly double those that are considered justified. In other words, interpreters give advice, instruct, warn, ask questions or give information on their own three times every four minutes during the bilingual parts of the trials. There are also important differences among the three language pairs. At one end of the spectrum, the French-Spanish combination is in the best position, with more justified than unjustified non-renditions, and relatively lower averages. At the other end of the spectrum, the Romanian-Spanish corpus presents an alarmingly high average of unjustified non-renditions per bilingual hour. 5.2.1. Justified non-renditions Regarding the purpose of justified non-renditions, table 4 presents the average frequency per bilingual hour of each kind. Justified non-renditions per bilingual hour Language TOTAL Pause Clarification Confirmation Retrieval EN/ES 26.4 4.4 13.7 6.3 2.4 FR/ES 14.4 2.0 9.8 2.6 0 RO/ES 25.6 0 18.4 6.7 1.5 Average 22.8 2.1 14.3 5.5 1.4 Table 4. Average number of each kind of justified non-renditions per bilingual hour. Table 4 shows that justified non-renditions are most frequently used to ask for clarification (14.3 times per bilingual hour on average) or confirmation about the information received (5.5 times per bilingual hour on average), which are two closely related categories. Non-renditions that ask for a pause or to retrieve possibly forgotten information are scarcer. In this respect, what may be interesting is that the textual orientation of justified non-renditions is much more frequently used than the interactional orientation, which is also plausible if we consider the unequal power distribution of the court. In Cheung’s (2017: 181) study, interactional non- renditions were also more frequently directed to Cantonese-speaking defendants and witnesses (75% of total non-renditions in Cantonese) than to English-speaking members of the judiciary (11% of total non-renditions in English). In fact, justified non-renditions to ask for a pause only occur in 4 of the 55 court hearings analysed, one corresponding to the French-Spanish subcorpus and three corresponding to the English-Spanish subcorpus, accounting for 11 non-renditions. Of these 11 non-renditions to ask for a pause, 10 were directed to the defendant and only 1 to a member of the judiciary (a defence counsel). This example is presented in extract 2. Extract 2. DC: (…) y que después ustedes fueron de fiesta en una discoteca, ¿no? ¿Me pueden decir cuál? (…) and, aftewards, you went party to a disco, right? Could you tell me which one? INT: And afterwards, you went to party D: [Sí. [Yes. INT: [to which disco?= D: = [((inaudible)) DC: [A ver, ¿usted bebió aquella noche? [Let’s see, did you drink that night? INT: Perdone, que… si me puede responder… u::: u::: una pregunta, antes de::: Excuse me… could he answer… a::: a::: a question, before:::= DC: =Perdón… eh, perdón. Perdón, perdón. =Sorry… eh, sorry. Sorry, sorry. INT: So, eh, in which disco were you, if you can remember? Extract 2 shows a particularly confusing extract, mostly due to overlaps and interruptions. The defence counsel asked two questions and when the defendant was answering the second question, the defence counsel overlapped with him, preventing the interpreter from hearing the defendant’s response, which was also inaudible in the recording of the court proceeding. The interpreter politely asked for a pause before continuing with the next question, thus clearly performing his role as coordinator. The low frequency of non-renditions to ask for a pause may be related to another strategy identified in the corpus. Sometimes, interpreters “impose” a pause to interpret by starting to interpret aloud (not in a whispering mode) while the defendant is still talking: the interpreter interrupts the defendant or witness to interpret. This is also possible because the defendant and the interpreter share the same microphone. This strategy is much more frequent than non- renditions to ask for a pause, as it occurs, on average, 32.19 times per bilingual hour in the TIPp corpus. As for non-renditions to ask for clarification, they occur in 12 of the English-Spanish court proceedings (63% of this subcorpus), 5 of the French-Spanish subcorpus (55.5%) and 8 of the Romanian-Spanish subcorpus (29.6%). Contrastingly, the highest average frequency per bilingual hour is found in the latter. Again, this is because various non-renditions to ask for clarification are found in trials with very few bilingual minutes. For instance, in one of these court proceedings there are three non-renditions to ask for clarification in a court proceeding with only two bilingual minutes. Extract 3 shows various non-renditions to ask for clarification in a relatively brief exchange. Extract 3. A: ¿Qué pasó? What happened? INT: What happened? D: I was in LPH eating. INT: You were…? D: I was in LPH. INT: LPH? D: In front of Bojangles = INT: = Okay. Estaba eh... en el... L... P... H... Los Pollos Hermanos, que está en frente del, del Bojangles, estaba sola, de pie =Okay. She was eh… at the… L… P… H… Los Pollos Hermanos, which is in front of the… of the Bojangles… she was standing there, alone D: So I was stand there, eating = INT: = comiendo = = eating = D: yeah, when they put on uniform, with eh, their machine, so they're standing, traffic, waiting traffic, there, there I was coming there back = INT: = police officers, you’re talking about? D: Yes, they stop me the traffic = INT: = Dos agentes de, de policía estaban controlando el tráfico, de, eh, estaban de uniforme, e iban con las motos = Two police officers, they, they were controlling the traffic, they, eh, were wearing uniform and they were on their motorbikes The interpreter first asks for repetition after possibly mishearing what the defendant says and then asks again when she seems to have trouble in understanding what the defendant means. In the third non-rendition she reacts to the defendant’s ambiguous turn when referring to “them”. 7 This last example, though, can also contradict the principle of accuracy, where interpreters should maintain the content and style of the original discourse. If the defendant had been ambiguous on purpose, the interpreter’s question would have challenged this “strategy of ambiguity” (Niska 1995). Similarly, in the following example (extract 4), the interpreter seeks to obtain more accurate information of the object mentioned by the defendant suggesting a simile (“an arch”), which she repeats twice in the following excerpt. Extract 4. 7 Note that there are additions and omissions by the interpreter in the previous excerpt. Interpreting mistakes in the TIPp corpus are studied by Arumí et al. (2017) and Orozco-Jutorán (2019). DC: Eh, usted cuando fue a hacer el vis-à-vis, ¿antes de esto los funcionarios le hicieron algún tipo de registro? When you went to the vis-à-vis, before that, did the officers do any kind of search on you? INT: Când ai mers să faci vizaví, vizita la el funcționarii de la pușcărie te- au înreg...te-au căutat. Ți-au făcut percheziție? When you went to the vis-à-vis, to visit her, the prison officers did not sea::: did not examine you. Did they frisk you? D: Da, mi -au trecut prin aparatul ăla. Yes, they made me pass that device. INT: Prin arc? Through an arch? D: Da. Yes. INT: Ah, estuve pasando por un aparato, por un arco. Eh, I was passing through a device, an arch. D: Dar nu s-a întâmplat nimic. But nothing happened. INT: Un arc? An arch? D: Da, ceva așa ((indica con las manos la forma del aparato)) Yes, something like this ((signals with his hands the shape of the device)) INT: Es un aparato que = It’s a device that = DC: = Sí, sí, la detección… =Yes, yes, for the detection. The interpreter, again, oversteps her role by seeking to obtain more detailed information. While it is true that the more concrete the information is, the easier it is to interpret accurately, these examples also reflect that non-renditions to ask for clarification or confirmation challenge the principles of accuracy and impartiality. 5.2.2. Unjustified non-renditions The annotation system included three categories of unjustified non-renditions, as already described in the previous section. Our idea of unjustified non-renditions is similar to Cheung’s (2017:181), “non-translational types of interactional non-renditions”: Non-translational types of interactional non-renditions were made when the interpreters assumed the roles of court clerks to carry out such court procedures as administrating oaths/confirmations, reminding witnesses/defendants that during recess they were not allowed to discuss the court case with anyone, or reminding the witnesses/defendants that they were still under oath/confirmation after recess. The use of the term “unjustified” seeks to broaden the scope of this category, which can also include answers on behalf of the defendant or adding questions on behalf of the members of the judiciary. Table 5 shows the average frequency of each kind of unjustified non-rendition. Unjustified non-renditions per bilingual hour Language TOTAL AIW Answer Extra-info EN/ES 50.2 7.3 6.3 36.1 FR/ES 11.9 1.3 1.3 8.5 RO/ES 65.7 7.3 12.1 45.4 Average 45.5 5.7 7.1 32.0 Table 5. Average number of each kind of unjustified non-renditions per bilingual hour. (AIW: advice, instructions and warning to the defendant). The results reflect that non-renditions to supply extra information are the most frequent, as they account for a 70% of unjustified non-renditions. Some interpreters use their own knowledge or experience in the court system to talk on behalf of the members of the judiciary (the judge, attorney or defence counsel). The following examples (extract 5) may help understand how interpreters overstep their role when providing this extra information. This excerpt is from a court hearing that lasted 3 minutes. The audible bilingual part accounted for 1 minute approximately (a third part of the hearing). In one minute of interpreting, the interpreter produced 7 unjustified non-renditions: 6 to add extra information and 1 to give advice to the defendant. Extract 5 shows the context of the first two non-renditions of this hearing. Extract 5. [1:30:09 – 1:50:09] J: Dígale al acusado que se le pide como autor de un robo con fuerza en tentativa. Tell the defendant he is requested as the author of an attempt of burglary. INT: Ah, deci te acuză de un furt cu spargere în grad de tentativă. Oh, so, you are accused of an attempt of burglary. D: Nu imi acord deloc. Să îmi iasă creierii pe ochi. I don’t remember anything. Let my brains get out through my eyes. INT: Ești de acord sau nu ești de acord? Do you agree or not? D: Nu îmi acordo nada, deloc. I don’t remember anything. INT: Ascultă, tu ești de acord? Listen, do you agree? D: Nu că nu sunt de acord, nu știu deloc... Not that I don’t agree, I don’t know anything… INT: Vale, dice que no se acuerda de nada de este. Ok, he says he doesn’t remember anything about this. As can be noted in the previous extract, the judge simply produces an informative statement, which the interpreter renders in her first turn. However, after receiving the answer from the defendant, instead of relaying it to the judge, the interpreter adds a question that is not present in the judge’s original utterance, and she even repeats that question once more. The interpreter takes on the role of the judge twice in just 20 seconds and maintains a brief monolingual exchange with the defendant. In the following extract, from another court proceeding, the interpreter uses information he has known from the preparation of the trial with the lawyer to add extra information to what the judge has just said. Extract 6. J: Vamos a ver. Mire. Dígale que tenemos dos opciones: o celebramos el juicio o podemos, o puede aceptar una pena que sería de ahhh, 8 meses de multa con cuota diaria de 3 euros. Tiene que decidir si quiere hacer el juicio o quiere aceptar esta pena. Let’s see. Look. Tell him we have two options: we can either celebrate the trial or we can, or he can accept the punishment, which would be of eeehh 8 months of fine with a daily fee of 3 euros. He has to decide whether he wants to do the trial or accept this punishment. ((There is some whispering during the judge’s turn)) INT: Acum asculți. Atent. Și îți zic. Sau facem procesul și îți dă altă pedeapsă sau facem o înțelegerea pe care ai vorbit-o cu domnul avocat și rămâne în 8 luni de amendă a câte 3 euro pe zi care sunt 720 de euro. Ce vrei să facem? Now, listen. Carefully. I’ll tell you. Either we do the trial, or [and] they will give you another punishment, or we make the deal you talked about with mister lawyer and you will have a fine for 8 months with a daily fee of 3 euros, which will be 720 euros. What do you want to do? D: Și astea 8 luni să le fac în pușcărie? Will I go to prison during these 8 months? INT: Nu. Plătești amendă, ți- am zis. No, you only pay the fine. I told you. In the previous extract, the interpreter oversteps her role in various parts of her rendition. She relays the basic information of the judge’s original utterance, but she adds her own paternalistic tone (“Now, listen. Carefully”) and information from the preparation of the hearing with the defence counsel. In her second turn, the interpreter does not translate the defendant’s question to the judge but answers it herself adding a final cue (“I told you”). According to Hale’s (2008) classification of court interpreters’ role, the interpreter of this excerpt would be acting as the advocate for the powerless participant. Other examples of unjustified non-renditions to add information include adding comments on what has just been interpreted (e.g. Bueno, supuestamente, “Well, allegedly”) to detach from what has been said, possibly as a face-saving strategy. In other occasions, non-renditions expand a question to make it easier to answer, as in extract 7. Extract 7. A: ¿El tren entonces estaba en marcha? Was the train moving then? INT: Și atunci trenul mergea? Trenul mergea în timpul ăsta sau era o stație? And then was the train moving? Was the train moving at that time, or was it at a station? Unjustified non-renditions to answer on behalf of the defendant are less frequent, but there are also some examples in the corpus. For instance, in extract 8, the interpreter answers the attorney’s question based on the information provided by the defendant in a previous answer. Extract 8. A: Con posterioridad a estos hechos, esto fue en febrero de 2012 la denuncia, ¿ha tenido algún otro incidente con este señor, con el agente, con este agente de ferrocarriles? After the incident, which was in February 2012 according to the report, have you had any other incident with this man, with the officer, with this rail officer? INT: Reclamația s-a făcut în februarie de 2012. După această dată ai mai avut cu el vreo întâlnire, o discuție? The report was filed in February 2012. After that, have you met him or argued with him? D: M-am întâlnit dar nu am vorbit cu el, l-am văzut doar. I met him, but I didn’t talk to him. I only saw him. INT: Solamente le vio, pero no tuvieron ningún altercado. She only saw him, but there was no incident. A: Después me refiero, desde 2012 hasta ahora. I mean afterwards, from 2012 until now. INT: Sí, de 2012 solamente, solamente lo vio, pero... Yes, after 2012 she only, she only saw him but… This may seem a rather innocuous example of unjustified non-rendition, because the information provided by the interpreter is already in a previous turn. However, by directly answering the attorney’s question, the interpreter neglects the defendants’ right to confirm, reformulate, expand or modify a previous answer. Finally, the least frequent unjustified non-renditions are those aimed at giving advice or instructions to the defendants or warning them. For example, in a court hearing, while the members of the judiciary are discussing the preliminary questions, the interpreter and the defendant engage in a whispered side conversation. Certain words from this side conversation are audible in the video recording and the interpreter is heard saying “be careful”, “don’t do any crime”, “don’t go in”. In another court proceeding, the interpreter told the defendant “there’s no problem, you can say it”, to encourage her to tell the truth about her working as a prostitute (see Arumí and Vargas-Urpi, 2018). In other examples, the interpreters use this kind of non-renditions to tell the defendants or witnesses how to behave during the hearing: they tell them when to stand up, when to approach the microphone, when to sit, or, for example when not to talk, as in extract 9. Extract 9. Court proceeding 81. PP: Sí. Le puede preguntar, por favor, cuántos eran en total aquel día. Yes. Can you ask him, please, how many of them were in total that day? INT: Câţi eraţi în total? How many were you in total? D2: Patru. Four. D1: Ah, eram patru. Ah, we were four. INT: Tu nu răspunzi, el răspunde. Cuatro personas. ((to the defendant 1)) You are not answering, he is answering. ((to the private prosecutor)) Four people. In extract 9, there are two Romanian-speaking defendants. The private prosecutor addresses a question to defendant 2, but defendant 1 also answers and the interpreter warns him that it is not his turn to answer. 5.4. Self-initiated vs. other prompted Cheung (2017) analyses whether non-renditions were self-initiated or other-prompted, because this parameter “can be used to measure an interpreter’s active coordination”. This parameter was not included in the initial annotation of the TIPp corpus but has been explored specifically for the purpose of this article. In Cheung’s study (2017: 182), there were much more self-initiated non-renditions (1267 out of 1524 non-renditions) than other-prompted (257). Self-initiated non-renditions were more frequently addressed to the Cantonese-speaking persons (mostly witnesses and defendants) (1031 out of 1524), while other-prompted non-renditions were more frequently addressed to the English-speaking persons (the court language) (250). In the TIPp project, the various categories used to classify justified and unjustified non- renditions mainly reflect self-initiated non-renditions. Other-prompted non-renditions were not the object of our study because the pilot project already revealed a very low presence of this kind of non-renditions. Among the few examples of other-prompted non-renditions in the TIPp corpus, there are the interpreters’ responses after the judges introduce them. Table 6 presents information about the number of court proceedings where the interpreter is presented, and whether this presentation prompts any kind of response from the interpreter. The presentation Court proceedings The interpreter is Languages prompts a non- (total subcorpus) presented rendition EN/ES 19 6 6 FR/ES 9 0 0 RO/ES 27 8 5 Total 55 14 (25.4%) 11 Table 6. Presentation of the interpreter during the court proceeding As may be noted in the previous table, the interpreter is not regularly presented. Instead, in most court proceedings, he or she starts to interpret without any kind of introduction. In 11 examples, the presentation of the interpreter results in other-prompted non-renditions. In 6 of these cases, he or she is asked to swear an oath, as shown in excerpt 10. In these cases, the interpreters’ other-prompted non-rendition is the answer to the judge’s request. It is interesting that there are only two judges using this formula in the TIPp corpus, who are thus responsible for the six examples of this kind. Extract 10. Court proceeding 121. J: Muy bien. A ver señor intérprete, en primer lugar, ¿jura o promete desempeñar fiel y correctamente su cargo? Good. Let’s see, Mr interpreter, firs of all, do you swear or promise to faithfully and duly fulfil your task? INT: Juro, señoría. I swear, your Honour. J: Muy bien. Pues, bajo apercibimiento que en caso contrario incurrirá en falso testimonio. Good. Otherwise you are under penalty of perjury. In 4 court proceedings, the judges ask the interpreters to identify themselves by saying their name, again prompting a non-rendition. In one case, the presentation is more informal (e.g. “The defendant is assisted by an English-interpreter, Mr. A. That’s you, right?”) and prompts a ‘yes’ to confirm the information. In only one example, the interpreter whispers at the defendant after his presentation. As the whispering is not audible in the recordings, it is impossible to know whether he interprets his own presentation to the defendant. In the other examples, the interpreter’s presentation and/or oath are not interpreted to the defendant, who is excluded from this monolingual exchange. 6. Conclusions Non-renditions are not exceptional in the TIPp corpus. Instead, this article has shown that, on average, non-renditions are frequently used by court interpreters in the TIPp corpus, a fact that may be generalizable to court interpreting in Spain. The dispersion of the results, with certain court proceedings concentrating high frequencies of non-renditions in very short bilingual parts, suggests that there may be important differences among court interpreters. The differences among the various language combinations are also worth mentioning: the French-Spanish subcorpus seems to be in a better position than the other two. A possible hypothesis to consider is the linguistic proximity between French and Spanish, which might prevent interpreters from uttering certain non-renditions that would be understandable to the other party. We might also hypothesize whether French-Spanish interpreters are better trained than Romanian-Spanish or English-Spanish interpreters, but without access to interpreters, this question is difficult to answer. The fact that the Romanian-Spanish subcorpus has the highest average of unjustified non-renditions may also be related to a desire to act as advocates for their compatriots in the case of Romanian interpreters. Note that all the Romanian interpreters shared the same origin with the defendants they interpreted for, while in the English and French samples interpreters do not always the same origin with the defendant, as these languages are sometimes used as lingua franca (Vargas Urpi, forthcoming). Further research is needed to explain these differences and confirm or refute these hypotheses. Concerning the functions of non-renditions, the study bases on the premise that not all non- renditions are undesirable and thus seeks to acknowledge the interpreter’s task in discourse management. This is the rationale behind the distinction between justified and unjustified non- renditions: justified non-renditions respect the principles of impartiality and accuracy and can be considered as part of the interpreters’ coordination tasks; while unjustified non-renditions are those that contravene these principles, especially impartiality. Even so, the findings reflect higher rates of unjustified non-renditions, which are particularly recurrent when used to provide extra information. In general, unjustified non-renditions show that interpreters sometimes act as advocates for the powerless participant, while in other occasions they serve as institutional assistants for the institutions (e.g. when they answer directly, based on information they already know), thus acting as advocates for the powerful participant (Hale 2008). As for the justified non-renditions, the analysis reflects that questions to ask for clarification sometimes contradict the principles of accuracy and impartiality, especially when there might be a strategy of ambiguity, as already pointed out by Cheung (2017). Further analysis is needed to ascertain to what extent clarification requests can affect the intention of original utterances and the development of the interaction. Non-renditions to ask for a pause to interpret are relatively scarce in the corpus. It is interesting because these are the only non-renditions that purely reflect a conversation-management strategy. The low frequency of non-renditions of this kind contrasts with the high frequency of interruptions by the interpreter, which seems to be a more recurring conversation-management strategy. The fact that pause requests are more often addressed to defendants (the powerless participant) than to other participants in the interaction (judges or lawyers, the powerful participants) reflects court hierarchies and emphasizes the already existing power imbalance, as also happened in Cheung’s (2017) study. The power imbalance is also evident when the defendant is excluded from the monolingual exchange involving the presentation of the interpreter by the judge or the court assistant. The lack of transparency in these cases results in the defendant not being aware that the interpreter has sworn an oath before starting to interpret. While it is true that the tempos of court proceedings make it difficult to relay to the defendant everything that is said in monolingual exchanges, more transparency would contribute to balance power. All in all, the detailed results presented in this article contribute to the description of the reality of court interpreting in Spain. The abuse of non-renditions reflects poor quality of court interpreting, especially in the case unjustified non-renditions. The examples also demonstrate scarce awareness of ethical principles and urgent need to improve court interpreting practices in Spain. Finally, concerning the method, while it is true that some categories could overlap somehow (e.g. non-renditions to ask for confirmation or clarification) and that the taxonomy should have included ‘other-prompted non-renditions’ as a separate category, it is also a first attempt at describing and classifying the various kinds of non-renditions in court interpreting according to their function. Future research might base on these categories or adapt them to further provide more accurate descriptions of the purpose of non-renditions and to explore their influence in the development of interpreted court proceedings. Acknowledgements This paper is part of the project “Translation quality as a guarantee of criminal proceedings. Development of technological resources for court interpreters in Spanish-Romanian, Arabic, Chinese, French and English language pairs (TIPp project), funded by the Spanish Ministry of Economy and Competitiveness (FFI2014-55029-R). References Angermeyer, Philipp Sebastian. 2015. Speak English Or What?: Codeswitching and Interpreter Use in New York City. New York: Oxford University Press. APTIJ. 2010. Código deontológico para intérpretes y traductores judiciales y jurados. Code of ethics for court and sworn interpreters and translators. http://www.aptij.es/img/doc/CD%20APTIJ.pdf Arumí, Marta; Bestué, Carmen; Gil-Bardají, Anna; Orozco-Jutorán, Mariana; Vargas-Urpi, Mireia, and Vigier, Francisco. 2017. La qualitat de la traducció com a factor de garantia del procés penal: desenvolupament de recursos per a intèrprets judicials (el projecte TIPp). Report. URL: https://ddd.uab.cat/record/176075 Arumí, Marta, and Vargas-Urpi, Mireia. 2017. “Strategies in public service interpreting. A roleplay study of Chinese–Spanish/Catalan interactions”. Interpreting, 19 (1): 118-141. doi: 10.1075/intp.19.1.06aru Arumí, Marta, and Vargas-Urpi, Mireia. 2018. “Annotation of interpreters’ conversation management problems and strategies in a corpus of criminal trials in Spain: the case of non- renditions”. Translation and Interpreting, 13 (3): 421-441. doi: 10.1075/tis.00023.aru Berk-Seligson, Susan. 1990. The bilingual courtroom: Court interpreters in the judicial process. Chicago: University of Chicago Press. Cambridge, Jan. 2002. “Interlocutor roles and the pressures on interpreters”. In Traducción e Interpretación en los Servicios Públicos. Nuevas necesidades para nuevas realidades, ed. by C. Valero Garcés and G. Mancho Barés, 119-124. Madrid, Servicio de Publicaciones de la Universidad de Alcalá. Cheung, Andrew K. F. 2017. “Non-renditions in court interpreting. A corpus-based study”. Babel, 63 (2): 174-199. doi: 10.1075/babel.63.2.02che Del Pozo Triviño, Maribel and Blasco Mayor, M. Jesús. 2015. “Legal Interpreting in Spain at a turning point”. MonTI, 7: 41-71. doi: 10.6035/MonTI.2015.7.1 García-Beyaert, Sofía; Bancroft, Marjory A.; Allen, Katharine; Carriero-Contreras, Giovanna and Socarrás-Estrada, Denis. 2015. “Ethics and Standards for The Community Interpreter. An International Training Tool”. In The Community Interpreter, ed. by M. Bancroft et al., 4- 30. Columbia: Culture & Language Press. Hale, Sandra. 2004. The discourse of court interpreting: Discourse practices of the law, the witness, and the interpreter. Amsterdam – Philadelphia: John Benjamins. doi: 10.1075/btl.52 Hale, Sandra. 2008. “Controversies over the role of the court interpreter”. In Crossing Borders in Community Interpreting: Definitions and dilemmas, ed. by C. Valero Garcés and A. Martin, 99-121. Amsterdam – Philadelphia: John Benjamins. doi: 10.1075/btl.76.06hal Mikkelson, Holly. 2008. “Evolving views of the court interpreter’s role”. In Crossing borders in community interpreting, ed. by C. Valero-Garcés and A. Martin, 81-97. Amsterdam – Philadelphia: John Benjamins. doi: 10.1075/btl.76.05mik NAJIT. 2002. Code of Ethics and Professional Responsibilities. URL: https://najit.org/wp- content/uploads/2016/09/NAJITCodeofEthicsFINAL.pdf Niska, Helge. 1995. “Just interpreting: Role conflicts and discourse types in court interpreting”. In Translation and the law, ed. by M. Morris, 293-316. Amsterdam – Philadelphia: John Benjamins. doi: 10.1075/ata.viii.21nis Onos, Liudmila. 2014. La interpretación en el ámbito judicial: el caso del rumano en los tribunales de Barcelona. PhD dissertation. Universitat Autònoma de Barcelona. URL: http://hdl.handle.net/10803/285160 Orozco-Jutorán, Mariana. 2017. “Anotación textual de un corpus multilingüe de interpretación judicial a partir de grabaciones de procesos penales reales”. Revista de Llengua i Dret, 68: 33-56. doi: 10.2436/rld.i68.2017.3034 Orozco-Jutorán, Mariana. 2018. “The TIPp Project: Developing Technological Resources Based on the Exploitation of Oral Corpora to Improve Court Interpreting.” Intralinea. URL: http://www.intralinea.org/specials/article/2316 Orozco-Jutorán, Mariana. 2019. “A Mixed-Methods Approach to Research in Corpus-Based Interpreting Studies. Translation and Interpreting in Criminal Procedures in Spain”. In Qualitative and quantitative research methods in legal translation/interpreting studies, ed. by L. Biel, V. Sosoni, J. Engberg and R. Martín Ruano. Ortega Herráez, Juan Miguel. 2006. Análisis de la práctica de la interpretación judicial en España: el intérprete frente a su papel profesional. PhD dissertation. Universidad de Granada. URL: http://hdl.handle.net/10481/977 Ozolins, Uldis. 2016. “The myth of the myth of invisibility?”. Interpreting, 18 (2): 273-284. doi: 10.1075/intp.18.2.06ozo Vargas-Urpi, Mireia. 2017. “Court interpreting as a shared responsibility: judges and lawyers in a corpus of interpreted criminal proceedings”. Revista Canaria de Estudios Ingleses, 75: 139-154. Vargas-Urpi, Mireia. Forthcoming. “Juicios multilingües en Barcelona desde la perspectiva de la sociología del lenguaje: lenguas dominantes, lenguas minorizadas y lenguas invisibles”. Onomazéin, 47. DOI: 10.7764/onomazein.47.08 Wadensjö, Cecilia. 1992. Interpreting as Interaction: On Dialogue Interpreting in Immigration Hearings and Medical Encounters. Linköping: Linköping University, Department of Communication Studies. Wadensjö, Cecilia. 1998. Interpreting as Interaction. London – New York: Longman. Wadensjö, Cecilia. 2015. “Non-renditions”. In Routledge Encyclopedia of Interpreting Studies, ed. by Franz Pöchhacker, 352-353. London – New York: Routledge.