On what grounds? Justifications of student translators for their translation solutions

ABSTRACT This paper reports on an empirical experiment in which a group of MA student translators justified their translation solutions. The aim of the work was to determine to what extent students rely on their theoretical knowledge of translation in their justifications and use the metalanguage of the field. The data consists of transcribed retrospections which were carried out immediately after the students had completed the translation task, with screen recordings of translation processes as the cue. In the retrospection, students commented on their translation processes. The results indicate that approximately one third of all justifications are based on a gut feeling, and the use of metalanguage is scarce. The results point to the need to practice the skill of justifying one’s solutions orally during translator training.


Introduction
The ability to justify one's translation solutions can be regarded as one aspect of translation competence (TC). According to the European Master's in Translation network's framework for translator and translation competence (EMT (European Master's in Translation) 2017,8), MA students in translation should know how to 'analyse and justify their translation solutions and choices, using the appropriate metalanguage and applying appropriate theoretical approaches'. In other words, students should know how to apply theoretical knowledge and concepts assimilated during training in their working practice. The ability to provide a solid justification for translation solutions is needed especially when a translator makes changes to a target text (TT) in comparison to the source text (ST), for example when a ST cannot be translated literally due to interlingual or cultural differences.
This paper provides empirical data on the way this aspect of TC developed in a group of MA students majoring in English and translation or Russian and translation. The paper explores how students justify their translation solutions: to what extent they use the metalanguage specific to the field of translation (studies) or refer to theoretical knowledge. The students involved in the present study had all completed a BA in translation, including both translation classes and compulsory translation studies (15 ECTS). Therefore, it can be assumed that they already possessed practical and theoretical knowledge about translation and could exploit it in justifying their decisions.
First, we look into the concepts of 'metalanguage' in translation and 'translation theory' and define them for the purpose of this paper. Then we introduce the educational setting of the study, looking into the contents of the training that the participating students have been exposed to. Third, we explain the research material and methods in detail. Finally, we present the results of the data-driven analysis. In our concluding remarks, we discuss the relevance of the results from the perspective of translator training, bringing out limitations and restrictions of the study and proposing ideas for future research.

On the relation between theory, metalanguage, and translation competence
Some professional translators and translation students appear to regard translation theory as irrelevant and useless for the practice of translation Vottonen and Jääskeläinen 2018, 94). According to a survey by , university trained translators rate theory very low on their list of ideal university training. In some educational settings, such as Brazil, translation as an academic area and professional translation have been regarded almost as two separate domains (Milton 2004). Indeed, the notion of 'translation theory' appears to be largely missing from the accounts of TC, apart from the recently published EMT model. Is 'translation theory' really a mere abstraction to be discussed in academic ivory towers, with no relevance in the working life of translators?

On the concept of 'theory'
Our answer to the question above is obviously 'no'. We firmly believe that the negative attitudes towards 'theory' can at least partially be explained by different understandings of the term 'theory' itself. In Katan's survey (2009), for example, translators were asked how important they regarded a module titled 'Translation Theory' as a part of translator training, without any specification of the course contents or definition of a 'theory'. As discussed by Chesterman (2007), the concept of 'theory' can indeed be understood in different ways. First, a 'theory' can refer to an all-encompassing scientific explanation of a phenomenon. 'Translation theory', in the singular, may be taken to refer to a single theory that can explain all translation. Such a theory is unlikely to exist: translation as a multifaceted phenomenon can hardly be harnessed by any single theory. On the other hand, our experiences as teachers of translation are that any academic research conducted in the field of translation can be loosely referred to as 'theory' by students, and that students are confused about the number of 'theories' in the field. Among students, a 'theory' may be conceptualised as 'building blocks on which a discipline is built' and 'the conceptual tools that make it possible for one to understand a discipline and what it is about' (Hanna 2009, 148-149, our emphasis), or even as having 'no clarity about the relevance for real-life situations' (149). These quotations reflect the attitudes of MA translation students in the UK, implying that theory may indeed be conceptualised as an abstract way of approaching a phenomenon without any clear link to practice; theory is needed to understand the discipline, not the practice. When 'theory' is understood in any of the ways described above, the negative attitudes towards it seem understandable: it may not be relevant for a practicing translator to know by heart the various theories that are offered to explain translation, or it may not make sense to students to read an esoteric academic article discussing a specialised topic the context of which they are not familiar with.
What is relevant, however, is to acknowledge that any practice, translation included, is based on a set of principles, a system of ideas that explain the phenomenon. This is the very basic definition of a 'theory', provided for example by the Oxford living dictionary. As Chesterman (2007) notes, a (better) understanding of a phenomenon is the general goal of any theory, continuing that, 'a theory of translation is thus a view of translationor some part or aspect of it -which helps us to understand it better.' A theory, then, is an instrument or a tool with the help of which translators know what they are doing. Theory in this sense is simply knowledge about translation.
In this instrumental sense, 'theory' does not have to be based on academic research, but it can stem from any written information on the topic in question -or even from experience. Hence, a theory of translation can also be personal or empirical knowledge about translation (Vottonen and Jääskeläinen 2018, 85). Presas and Martin de León (2014, 275) refer to this as 'implicit theories'. Implicit theories are based on the assumption that translators apply both operational and conceptual or theoretical knowledge to understand and translate texts (280). Implicit theories can cover several aspects, such as the translator's aim or role, or the relationship between source and target language and texts, and they are built through generalisation from particular cases, lacking systematic or deliberate reflection (280). From the professional perspective, implicit theories of beginners in translation may appear narrow and restricted, reflecting a layperson's idea of translation as a mechanical word transfer process. In other words, views of translation may vary from very rudimentary ones to a thorough understanding of the phenomenon of translation in different contexts, and each view in itself is a kind of 'theory' of translation, forming the guiding principles for the actual practice.

Knowledge about translation as a component of TC
In this paper, 'translation theory' is understood in its instrumental sense. We feel that emphasising the instrumental value of 'a theory' also clarifies its importance as a part of translator training and its relevance to practitioners of translation. In its instrumental sense, translation theory is also incorporated into the current TC models, such as that of PACTE (e.g. 2003). Perhaps due to ambiguity of the term 'theory', PACTE prefers 'knowledge about translation' to 'translation theory'. We adopt the same practice in this paper. In PACTE's TC model, knowledge about translation is one of the subcompetences that is specific to translation, distinguishing translators from other language professionals (e.g. PACTE 2003). Following the dichotomy proposed by Anderson (1983), the concept of 'knowledge' is divided into declarative and procedural knowledge. The first refers to theoretical knowledge (in the sense described above) which enhances the understanding of translation, consisting for example of the facts, concepts, principles, ideas and theories, whereas the latter refers to knowing how to do something (Anderson 1976).
Knowledge about translation sub-competence is defined as declarative knowledge (implicit or explicit) about how translation functions, i.e. types of translation units, processes required, methods and procedures used, types of problems (PACTE 2003, 59). In addition, knowledge about translation includes knowledge related to professional translation practice (work market, different types of briefs, clients and audiences) (59). 'Knowledge about translation', then, covers issues which are commonly discussed within 'translation theory', as well as practices in the field. A key element of this sub-competence is the concept of translation, which according to PACTE (2014) can be either dynamic or static. In the first case, translation is regarded as a textual, interpretative, communicative and functional practice, whereas in the latter case it is considered as a linguistic and literal, mechanical language change process (98).
In Göpferich's TC model, knowledge about translation or theoretical knowledge is not explicitly listed as a sub-competence. However, 'topics covered and methods employed in theoretical & practical translator training' are considered to set the frame for the emergence of TC as a whole (Göpferich 2009, 20-23). Training influences a translator's self-concept, i.e. the way they see their role and the role of translation in society (22). This, in turn, partly determines the employment of various sub-competences in different translation situations. In other words, translator training enhances students' ability to evaluate translation situations from various perspectives, in order to see what is essential and important in different situations, and to employ their competence accordingly. As far as theoretical training is concerned, it could be assumed that the more approaches to translation are presented during training, the more solid understanding of the practice students are likely to gain. Moreover, topics covered during training are likely to influence the type of metalanguage students learn to use.

Metalanguage in translation (studies)
Perhaps surprisingly, few TC models explicitly mention the ability to use field-specific metalanguage as a part of TC. However, in the EMT competence framework (EMT (European Master's in Translation) 2017,4,8) it is specified that students should 'know how to justify their solutions and decisions by applying the metalanguage and theoretical approaches'. This TC model is designed as a set of learning outcomes for MA level translator students in European universities, and it aims at reflecting the skills and knowledge required from translators in the current translation industry. The ability to provide sound justification for one's solutions is certainly one way of showing one's expertise.
In the EMT framework, 'knowledge' is understood as 'the outcome of assimilation of information through learning', and it refers to 'the body of facts, principles, theories and practices that is related to a field of work or study' (EMT (European Master's in Translation) 2017, 3). The need to equip students with such knowledge, or to aim at 'a deep understanding of the processes' involved in translation is mentioned as one of the general goals of MA level translator training (4). Although the five main areas of competence defined in the framework reflect procedural knowledge (skills) rather than theoretical knowledge, it is explicitly mentioned that each sub-competence is a combination of skills and knowledge (e.g. EMT (European Master's in Translation) 2017, 9); a skill does not emerge without knowledge about the task in question.
What, then, is exactly the metalanguage of translation? If this question is approached from the perspective of multicomponent TC models, it can be deduced that there is no single metalanguage which a translator should master. First of all, there is the metalanguage of translation studies, i.e. concepts arising from different theoretical approaches to translation. Due to the multi-or interdisciplinary roots of translation studies, the scholarly meta discourse is influenced by metalanguages from other fields (Gambier and van Doorslaer 2009, 1), such as linguistics, psychology and sociology. Secondly, there is the metalanguage of translation practice, i.e. the language of the profession. This obviously has parallels with the academic discourse, but is not the same 1 . Professionals' way of talking about translation reflects the business aspects, incorporating the terminology of service provision and business management. In this paper, we are interested in all the metalanguage which students use to justify their decisions, since all metalanguage is considered to reflect knowledge about translation.
Students already possess translation-related knowledge when entering a training programme (Jääskeläinen 2004, 28;Kumpulainen 2016, 132). This knowledge may be based, for example, on grammar translation at school, or their personal observations about the way translations have been produced. Possibly their understanding may have been influenced and somewhat refined by the theoretical books they have studied for the university entrance examination (Kumpulainen 2016, 132). Students can already translate when beginning their translator training; as Pym (2009) notes, 'we train people not just to translate, which they can already do, but to translate well, perhaps for a specific purpose, market or technological environment.' Hence, students have implicit theories, but there tends to be a mismatch between the academic (and professional) view and the beginners' conceptualisations. In order to train students to translate well, up to professional standards, their understanding of the practice of translation needs to become more diverse. Therefore, one of the most important tasks of translator training is to redefine the concept of 'translation' (Jääskeläinen 2004, 28), to make students aware of the multifaceted nature of translation in general and of professional translation in particular. The goal of training is not to prove students' implicit theories to be wrong, but to modify the conditions under which the initial understanding holds true (Bereiter 2002, 366). Training aims at complementing initial knowledge about translation, focusing on professional aspects in particular. Through deliberate practice (Ericsson 2006, 692;Shreve 2002, 158), training also enforces the proceduralisation of knowledge so that students can exploit the translation-related knowledge they have acquired in order to justify their translation solutions when need be, using the metalanguage of the field.

Participants, material and methods
At the time of data collection, translator training at the University of Eastern Finland (UEF) covered both BA and MA studies: students completed a 3-year BA programme in translation first, after which they moved on to the MA programme. The BA programme included a compulsory module of translation studies (15 ECTS), and compulsory translation classes into L1 and L2 (28 ECTS in total in the English major, 21 ECTS in the Russian major). In addition, the BA programme included compulsory studies of Finnish for Translators and Interpreters (25 ECTS). One of the learning outcomes in BA degrees was that students recognised the role of theoretical knowledge in translators' and interpreters' problem solving and in work in general (Curricula of foreign languages and translation 2014-2017, 45, 148). In this educational setting, students were exposed to both the theory and practice of translation prior to their MA studies.
Theory has also been incorporated to some extent into all translation courses (Vottonen and Jääskeläinen 2018, 95). At the very least, a translation task is always accompanied by a brief, i.e. the purpose of translation. Students also learn to justify their decisions in written commentaries, which they are required to submit with each translation task. In them, they comment on any difficulties they have encountered during translation and justify the solutions provided. The metalanguage of commentary instructions focuses on the function of the translation (vs. the function of the TT), and on global and local translation strategies. In the theoretically-oriented courses, a functional approach to translation has been put forward as the generally accepted guideline in the practical work, although other approaches are also introduced and discussed during training (linguistic approaches, manipulation, etc.). Hence, the training provides students with the instruments to justify their translation solutions using the metalanguage of the field. Whether or not this is the case in practice is the focal point of this paper.
The data analysed for this paper consists of translation process recordings and retrospections of eight MA students, four of whom were majoring in English language and translation, and four in Russian language and translation 2 . The participants had completed all the theoretical and practical translation classes of the BA studies. Six of the participants were in the first year of MA studies and two were in the second year.
The participants were asked to perform a translation task from either English or Russian into their first language (Finnish) with different briefs, which were expected to reveal different types of justifications and decision-making 3 . Three participants translated the text for the travel magazine Mondo, two for the travel section of the daily newspaper Helsingin Sanomat, and three for the children's news section of Helsingin Sanomat. The English text described Tower of London (324 words) and the Russian text described New Holland, a travel attraction in Saint Petersburg (259 words). The students could use all the time they needed to finish the translation task.
The translation process was recorded with the screen recording software Camtasia Studio 8. Immediately after the task, participants performed retrospection in which they described their translation process to the researcher with the screen recording as a cue. Cued retrospection has been criticised as a data collection method due to the longer duration of experiments, potentially causing tiring and lack of motivation and resulting in incomplete reports by the subjects (Göpferich and Jääskeläinen 2009, 181). Moreover, despite the cues, retrospection relies on memory, which may be erroneous due to the time gap between translation and reporting (Saldanha and O'Brien 2013, 125). Despite these weaknesses, cued retrospection was chosen as the method for the study due to its potential to open a window into the participants' long-term memory which, when triggered by cues, may help to elucidate how problematic situations were solved during translation (Englund Dimitrova and Tiselius 2014, 280). During the retrospections, students could describe their translation process and decision-making freely. During silent moments or clear problems during the translation process, the researcher asked 'what is going on here' or 'what were you thinking at this point', without prompting the use of theoretical concepts. Retrospections were recorded and transcribed.
The data was analysed using content-based analysis. We scrutinised the transcriptions simultaneously with the screen recordings, picking out points at which students refer to decision-making and provide some justification for their solutions (rather than, for example, just explaining what they are doing). These instances are epitomised in the verbal protocols by words and phrases such as therefore (I decided), because, on the other hand, I decided that, I was wondering if, and so forth. The use of metalanguage was also considered as a potential signal of verbalised decision-making. The analysis was datadriven: we did not reflect students' justification on any existing categorisation, but different categories arose from the data. A similar approach was taken by Pavlović (2010) in her study on novice students' arguments in their decision-making in L1 and L2 translation.
In section 4, a summary of the overall use of metalanguage is first provided and discussed. Following this, the categorisation emerging from the data is presented. Each category is also discussed from the perspective of TC, as an indicator of translationrelated knowledge. In addition, the use of metalanguage in each category is commented on. Where applicable, a comparison is made between the present study and the results of Pavlović (2010). However, some categories emerging from our data have no parallel category in Pavlović (2010).
Our hypothesis was that students would rarely refer to any specific theory or approach to translation (such as the skopos theory), but rather, express concepts and principles that reflect their knowledge of translation. Some metalanguage, for example referring to the 'target text purpose' or to the 'needs of a target reader' when justifying a strategy, can be regarded as an explicit manifestation of a certain concept of translation, or certain theoretical approaches to translation. However, not all metalanguage of translation can be directly linked to a specific type of approach to translation: 'parallel texts' might serve as an example of this. A student may justify a solution because 'a similar expression was found in a parallel text'. It could also be assumed that in our data, knowledge of translation can be manifested implicitly: a student may, for example, look for a solution to a problem on various sites through Google, and then finally decide to go for an option provided by an online dictionary (this process can be detected in screen recording data), without uttering any further justification.

Analysis and results
In this section we present a summary of the use of metalanguage in the verbal protocols, introduce the categories that could be identified as justifications for translation solutions, and explain the typical features of each category.
In the analysis, students majoring in English language and translation are referred to by the code 'E' followed by a number 1, 2, 3 or 4, whereas the code 'R' followed by a number indicates students majoring in Russian language and translation. The letters after the student codes indicate the translation brief: M for Mondo travel magazine, HS for Helsingin Sanomat newspaper, and HSC for Helsingin Sanomat for children.

Use of metalanguage
As discussed in section 2, the metalanguage of translation borrows from various neighbouring fields in addition to its field-specific discourse. In this study, metalanguage used by the students could be characterised as a combination of basic vocabulary in translation (studies) and linguistics. The total number of instances considered as the use of metalanguage was 113 or, excluding repetition, 56 (see Appendix 1). Metalanguage of translation covered references to different translation strategies (often used as verbs such as replace, omit, add, provide explanation), the verb to translate, nouns or noun phrases such as the reader, source text, Finnish equivalent, context, source, translation, term, translation solutions (that affect the whole text), brief, parallel text, and definition. In addition, basic vocabulary of linguistics, related to grammar and the use of language, was also identified. This covered words such as sentence, clause, word, word order, passive, agent structure, filler word and euphemism.
As Appendix 2 indicates, the use of metalanguage varied considerably between individual students, E4HSC showing the most frequent use with 45 references and R4HSC the least use with only four references. Generally, the students translating from Russian used metalanguage less frequently than those translating from English; approximately 78% of all metalanguage use appeared in the verbalisations of the latter group. However, variation within this group was considerable: of 88 instances, E4HSC's share was 51%. As was expected, no explicit references were made to any specific translation theory, and a considerable part of the instances -37 of the 113, i.e. approximately 33% -was attributed to the reference of sentence, words, clause, term, translation, and translate, which is very basic linguistic or translation-related vocabulary. Concepts such as parallel text, context or brief were rarely mentioned explicitly: the first twice and the other two once each. The use of metalanguage in different justification categories is discussed in the next subsection, in connection with the presentation of each category (see Appendix 3 for the distribution of metalanguage in different justification categories).

Justification categories
How, then, did students justify their decision-making, if the use of metalanguage was infrequent and linguistically basic? Eleven categories were identified from the cued retrospection data, which were labelled as follows: gut feeling/'sounds better', parallel texts, clarifying the target text (TT), translation brief, dictionary, context, Google search results, hurry, Google Translate, maintaining the style of the source text (ST), and differences between source and target language and culture. The distribution of justifications into different categories used by different students is presented in Table 1 and discussed in more detail in the following subsections. In total, we identified 367 justifications in the data.

Gut feeling
With the exception of participant E4HSC, students justified their solutions most commonly by what we labelled gut feeling (33.5% of all justifications), often indicated by expressions such as 'this sounds better' (see example 1) or 'looks better'. Simple statements of 'ending up' with a specific solution (see example 2) were also classified into this category. The many instances in which a student explained a solution without using any type of translation-related metalanguage or referring to any external source were categorised as being justified by a 'gut feeling'.  Example 1: 'Suddenly, it came to my mind that it could be "area", and then it started to look really good.' (R1M) 4 Example 2: 'But then it sounded somehow funny, so I ended up with that "famous" again. ' (E3HS) This finding is similar to that of Pavlović (2010, 68-69, 78): in her study, one third of students' arguments fell into the category of what Pavlović labelled 'sounds better'. This category seems to be very close to our 'gut feeling' since it includes all arguments in which no other reason than 'a feeling' is given for a solution. This finding may suggest that students' TC with regard to their ability to justify translation solutions is not very solid at this stage of their MA studies.
In this category, the use of metalanguage was rather vague. Students made some metalinguistic references that described their actions or grammar and language use (see Appendix 3). However, non-specific use of metalanguage does not necessarily correspond to a lack of knowledge about translation, but can have several explanations. First, in this experiment, the students justified their solutions to their teacher, not to a client. This may have influenced their motivation to carefully consider the reasons behind specific solutions. On the other hand, one might also expect metalanguage to be used in the academic training context in particular, since translation tasks are frequently discussed and commented on in translation classes by teachers using the field-specific metalanguage -and the same is expected of the students.
Secondly, not all the knowledge on which our decision-making is based can easily be retrieved from the long-term memory: the nature of knowledge can change, and what has once been learnt from books and lectures (declarative knowledge) may turn into tacit knowledge when blended with knowledge gained from experience, for example (e.g. Bereiter 2002, 148). As Polanyi (1966, 4) argues, 'we know more than we can tell'. Hence, what surfaces as 'gut feeling' or 'intuition' may in fact be based on learned principles and theories, which have turned into tacit, implicit knowledge that is difficult to verbalise, implying a role of non-conscious and intuitive decision-making in the participants' translation processes (see Pavlović 2010, 80;Hubscher-Davidson 2013). Student translators may even favour intuition over deliberate analysis when they are faced with particularly complex translation problems (Hubscher-Davidson 2013).
Thirdly, students may possess knowledge about translation, but are yet unable to apply that knowledge in their translation practice -or in justifying their practice. Example 3 might illustrate such a case: the student appeared to be thinking about the target readers and their needs, but failed to express this explicitly, simply stating that it 'does not make this sound any worse' if an explanation is added to the translation.
Example 3: 'Well yeah, okay this "beefeater" is quite a well-known word, but still it does not make this sound any worse if I quickly explain it.' (E2M)

Parallel texts
The second most frequent justification category was parallel texts with 63 (17.2%) justifications. Individual students justified their solutions with parallel texts to a varying degree. This category includes instances in which students justified their translation by referring to Finnish-language websites in which they found solutions to their problems (see example 4). Parallel texts were most often sought to solve lexical issues.
The use of metalanguage in this category was scarce: only one student referred to the term 'parallel text' explicitly, whereas most explained how they found a certain site, found a suitable expression there and therefore decided to use it. This, however, may be partly due to the nature of the data collection method: watching a video of one's actions and explaining them simultaneously may prompt narrative explanations rather than analytical reflection. It might be more natural to say that 'here I open a website which seems to be about the same things as the source text' than 'now I turn to a parallel text'. Students also made singular references to metalanguage relating to information searching, such as source, information search, provide explanation, and term. No critical comments were expressed with regard to the trustworthiness of the sites used. Parallel texts were commonly sought from Wikipedia: students sought a Finnish-language Wikipedia article on the topic of their ST to find out how 'Wikipedia calls' this and that. Some students mentioned explicitly that Wikipedia is always the first source of information for them (see example 5). Despite the scarce use of metalanguage, the very fact that students searched for parallel texts suggests that they regarded translation as something other than simply a word replacement procedure, and therefore, translation solutions may be best sought elsewhere than in dictionaries.

Clarification of the target text
Clarification of the TT (13.4%) was the third most frequently used justification, appearing to a varying degree in all students' data. In these cases, students expressed a willingness to 'do something' to the TT in order to make it more fluent, for example by splitting long ST sentences to make the TT easier to read (see example 6). In addition, students wanted to add some information to the TT in order to make it more understandable. However, reasons for changes were not always clarified but remained rather vague, as in example 6, in which the student argued that 'it might be easier to use a full stop in Finnish'. Here, too, the use of metalanguage was scarce: in total 15 words considered as metalanguage were used. Few references were made to translation strategies (clarify, add, omit, simplify) or to grammar (word order, clause, structure). However, the mere consideration of the clarity of the TT implies knowledge of translation: students appeared to be aware of the need to pay attention to interlingual differences in translation, searching for solutions conforming to the norms of the target language rather than following the ST structure (see example 7).
Example 6: 'Well, I thought there are so many commas in these sentences that it might be easier to use a full stop in Finnish every now and then.' (R3HSC) Example 7: 'Then I decided to put it in this way, because I think that in Finnish texts it is more usual.' (R4HSC) Again, our results resemble those of Pavlović (2010, 71, 78), whose category of 'pragmatic/textual reasons', consisting of arguments that have to do with text-linguistic or pragmatic considerations, accounted for 12% of all arguments in students' verbal protocols.

The translation brief
One of the most apparent indications of a specific approach to translation was justifications based on the translation brief (11.4%). In comparison to Pavlović's (2010) corresponding category of 'Target text reader', this justification appears to be more common in our data. Six out of eight students justified a part of their solutions on this ground; for E4HSC this was the most frequent type of justification. All justifications that either explicitly stated or implicitly hinted at the needs of a target reader or specific communicative needs of the text were categorised into this class. Most commonly, the justification of being 'the brief' must be read between the lines, as in example 8, in which the student explained why place names such as 'The Tower of London' should be transferred in their original language into the Finnish translation.
Example 8: 'Then I thought that it would be good to be in English in here too, because in principle you could put some kind of translation in the end for those who don't know English, but it would be useful to be in the original language as well, because they go to that place. And all signs are in English.' (E1M) Singular references were made to brief, reader, and strategy and the use of translation strategies such as translate directly, provide explanation, simplify, omit, and replace by superordinate. Although the use of metalanguage was most frequent in this justification category (see Appendix 3), it is attributed largely to E4HSC, whose verbalisation covers 64.3% of the metalanguage in this category.
This justification category also suggests a dynamic approach to translation, implying that translation is regarded as a communicative (rather than a merely linguistic) activity in which solutions are to be considered against the needs of the target reader and the function of the text.

Dictionaries and context
The justification category of dictionaries (10.4%) includes instances in which students looked up a word in a dictionary and accepted it without questioning or any further reasoning (example 9). In these instances, the students did not comment on the suitability of the word in the context in any way, but readily accepted the word suggested by the dictionary (example 10). This justification emerged in all students' data, being the dominant one in R2s retrospection. Typically, no specific metalanguage was used in these justifications -only singular references to definition, reader, word, and clause.
We separated the simple use of dictionaries, without any further reasoning or references to the context, from the cases in which students consulted a dictionary but based their decision on the context. Such justifications were classified into the category of context (example 11). Context was also used as justification when omitting or adding text (example 12). All in all, however, only 11 solutions (4.4%) were justified by the context. Only one student explicitly referred to the term context. In other instances, the reasoning was deduced from the verbalisations, such as 'I usually prefer Finnish (equivalents) but I thought this might be the most suitable here'. Other references to metalanguage were original, translate, and word.
These two justifications imply a somewhat different type of knowledge about translation: although the blind reliance on dictionaries suggests a rather mechanical understanding of translation as a simple language-transfer process, the category of context implies a more critical approach, showing appreciation of translation as communication rather than language transfer and also an awareness of how languages function.
Example 9: 'If this says that it is "contemporary", then it is "contemporary"'. (R4HSC) Example 10: 'Because I didn't have any previous experience of this word, I just figured I trust this [dictionary].' (E1M) Example 11: 'Okay, "scaffold", yes, it sounds logical in this context.' (E1M) Example 12: I thought I would put 'St. John's chapel' in brackets, but in the end I deleted the explanation in the brackets, because later in the text it turns out that it's a church.' (E3HS)

Other categories
Some solutions were also justified by Google search results (3.0%), hurry (2.5%), Google Translate (1.9%), maintaining the style of the ST (1.6%), and differences between the source and target language and culture (0.8%).
In the first justification category, students performed a Google search and made their decision based on the search results, without opening any links (example 13).
Example 13: 'I ended up with "navy jail" (direct translation from Finnish laivastovankila), because Google gave more results for that.' (R4HSC) In the second justification category, students felt they were simply running out of time and therefore had to settle for the solution they had been offered (example 14).
Example 14: 'In the end I was in too much of a hurry so I didn't have time to check it and I left it like that.' (E1M) Screen recordings show that some students also exploited the Google Translate tool to translate problematic words or phrases, trusting its suggestions completely without any apparent criticism (example 15). Such an uncritical attitude towards machine translation may suggest a somewhat mechanical concept not only of translation as an activity that can be carried out by a machine with no human input, but also of languages in general.
Example 15: 'But then it stubbornly wanted to be "innovative" in Google (translate), so I figured so be it.' (R1M) The last two justification categories, 'maintaining the style of the ST', and 'differences between source and target language and culture', by contrast, appear to imply more sophisticated knowledge about translation. An attempt to conform to the style of the ST in translation implies loyalty to the ST writer and their stylistic choices, hence reflecting the idea of dynamic equivalence (example 16).
Example 16: 'Here I tried to pursue the same kind of lightness that was present in the source text.' (E1M) The most infrequently used justification, differences between the source and target language and cultures, refers to instances in which students explicitly stated that the languages involved in the task differ with regard to a specific aspect and therefore some changes are necessary in translation (example 17). This justification category implies contrastive knowledge about languages and cultures and an ability to exploit it in translation (example 18).
Example 17: 'Because "ground floor" is like "first floor" in Finnish.' (E2M) Example 18: 'I was wondering how to start this sentence, because I always notice that the word order is different in Russian and in Finnish, so I just have to figure out what is the core of the sentence before starting to write it.' (R4HSC) All in all, there were only a few explicit references to metalanguage in these last five categories (see Appendix 3).

Discussion
Initially, we assumed that students might provide different types of justification for their decision-making in different briefs. Translating for children was expected to prompt more changes to the TT on the basis of the needs of the TT readers. This assumption proved to be partially correct (see Appendix 4 for distribution of justification categories based on different briefs). Clarification of the TT and the brief were indeed used more often by translators translating for children. However, 'gut feeling' was the commonest justification in all tasks, including when translating for children, when it accounted for 29.7% of all justifications. The corresponding figure for Helsingin Sanomat was 25.3%, whereas among students translating for Mondo it was 42%. This considerable difference was unexpected and deserves more attention in further studies.
Despite the slight differences, gut feeling was the commonest justification type in all briefs. This may imply a lack of translation-related knowledge, but it may also imply that decision-making has become automatised as a result of training or experience, and is therefore unconscious (see also Pavlović 2010, 80). As Pavlović (2010) points out, further research is needed in order to shed light on the way decision-making by gut feeling correlates with the level of TC. While this justification category may actually be based on solid knowledge about translation, the knowledge that is not put into words may go amiss in client-translator interaction. Justifying one's decisions by a gut feeling or on the grounds of 'sounding better' may not convince the client of the translator's expertise. The study points to the need to pay attention to this in translator training so that students learn to make their knowledge about translation explicit if necessary.
Some justification categories do, however, clearly reflect knowledge about translation. Several categories imply that students regard translation as a dynamic communication process in which solutions are based on what works best in the given situation. The influence of translator training is evident in the use of parallel texts as well as in taking the brief into account in translation solutions. Taking the brief into consideration clearly indicates that students apply a dynamic approach to the translation. The fact that students use parallel texts rather than only bilingual dictionaries as trusted sources of translation solutions also implies a more sophisticated conception of translation: translation is more than a simple word replacement process. The same applies to the justification category of 'context'. In addition, referring to the clarity of the TT implies a sensitivity towards the target reader's needs. Some justification categories imply a different, more mechanical understanding of translation: the use of Google Translate and bilingual dictionaries as the 'authority' are cases in point. However, further studies are needed to establish whether students' uncritical reliance on Google Translate or dictionaries is indeed related to mechanical understanding of translation or rather, can be explained by (a lack of) motivation, which is argued to be an essential factor that influences students' translation processes (see e.g. Hubscher-Davidson 2007, 55;Chesterman 2009, 17). According to Núñez and Bolaños-Medina (2018, 292), intrinsic motivation plays an important role in students' decisionmaking, making them persistent problem-solvers.
The use of translation-specific metalanguage was rather vague in students' retrospections. Metalanguage consisted mostly of the very basic vocabulary of translation and language use. In future studies, different settings could be tested to encourage the use of metalanguage. For example, a teacher could take the role of a client and ask questions explicitly about specific solutions. In this way, students might be prompted to justify their solutions in a more focused manner. This might also decrease the narrative nature of retrospections: instead of explaining everything that is happening on the screen, the students could focus on the questions brought up by the 'client'.
One could also argue that metalanguage of translation (studies) may not even be appropriate when justifying solutions to a client, since the client does not necessarily understand it. In this sense, the use of metalanguage as a part of TC in the EMT framework is rather vague and could be fine-tuned. In addition to mastering the metalanguage, a translator must speak the same language as the client in order to be understood -or at least to be sensitive about how a solution can best be justified to different clients. During training, therefore, students should learn not only the metalanguage but also a client-friendly, convincing and professional manner of using it in different situations.