3.1 Introspective Data Sources

Methodologically speaking, introspection generally means examining one’s thoughts and perceptions. An introspective data source enables us to look at research participants’ reports of inner thoughts and emotions. As will be noted in the works cited in the following two subsections, translation process research has made use of two introspective sources: the think-aloud method conducted in individual sessions and the dialogue think-aloud method conducted synchronously with more than one research participant.

3.2 Observational Data Sources

Observational data sources have been increasingly used in translation process research published since the beginning of the millennium. Increasing use is particularly notable in employing technology-based observation data sources. Conversely, only a few translation process studies have relied upon the researcher’s observation.

3.3 Retrospective Data Sources

Retrospective data sources are employed for assessing research participants’ cognitive processes. Their use is based on the assumption that participants can retrieve such processes from the information stored in long-term memory (Ericsson & Simon Reference Ericsson and Simon1993). One main advantage of using retrospective sources in collecting data about translators’ processes is that they do not interfere with the translation process itself; however, the retrospective data retrieved by research participants may be incomplete or inaccurate (Englund Dimitrova & Tiselius Reference Englund Dimitrova, Tiselius, Mees, Alves and Göpferich2009). Interviews and questionnaires are the two main retrospective data sources that have been used in previous translation process research.

3.4 Which Data Source(s) to Use?

The issues discussed in subsections 3.1, 3.2, and 3.3 clearly indicate that each data source has its strengths and weaknesses. Therefore, there is no best single source of data about the translation process. Given that each source can reveal different dimensions about translation cognition, it is important to use those appropriate for the research purpose and context. For example, think-aloud protocols may be a more reliable source for modelling translation processes, whereas computer keystroke logging will be a better option for collecting data about temporal aspects (e.g., pause duration and frequency) and revising behaviours. Meanwhile, semi-structured interviews and researcher’s observation may fit well within the ethnographic approach which involves researching the long-term changes and challenges in translation processes in workplace environments (see, e.g., Asare (Reference Asare, Angelelli and Baer2016); Risku (Reference Risku2014; Reference Risku, Schwieter and Ferreira2017)).

Triangulation – that is, collecting data using more than one data source – has become a common methodological feature in translation process research. It was Jakobsen (Reference Jakobsen and Hansen1999) who first introduced triangulation into translation process research. It is generally noted that translation process studies have increasingly implemented data source combination; for instance, by combining think-aloud protocols with interviews, keystroke logging or screen recording, triangulating keystroke logged data with eye-tracking, screen recording or retrospective interviews, and combining interviews with screen recording, questionnaires or researcher’s observation. Matsumoto (Reference Matsumoto1994) recommends using one of four data triangulation types by combining: (a) introspective with retrospective data sources; (b) different retrospective data sources; (c) retrospective data sources with product data; or (d) retrospective with observational sources. While some translation process studies used two sources only (e.g., Angelone Reference Angelone, Shreve and Angelone2010b; Englund Dimitrova Reference Englund Dimitrova2005), others relied upon three sources (Faber & Hjort-Pedersen Reference Faber, Hjort-Pedersen, Jakobsen, Mees and Göpferich2009) or more (Göpferich Reference Göpferich, Jakobsen, Mees and Göpferich2009). When deciding upon data triangulation, it is important to select the triangulation approach that helps in collecting the data covering the translation process dimensions addressed by the study.

4.1 Translation Problem-Solving Processes

The translation process by its nature is a problem-solving activity. As implied in Section 2, some researchers view translation strategies as procedures for solving problems (e.g., Krings Reference Krings, House and Blum‐Kulka1986; Lörscher Reference Lörscher1992). Due to the nature of the translation process, the most common problem type translators are expected to encounter is lexical problems. Mondahl and Jensen (Reference Mondahl and Jensen1996) note that translators solve the lexical problems they encounter using one of two strategies: (a) retaining the communicative goal of the source text part by finding a lexical alternative close to it (i.e., achievement strategy); and (b) simplifying the meaning of the source text part due to inability to find an acceptable equivalent to it (reduction strategy).

Some published works have focused particularly on conceptualizing or modelling translators’ problem-solving processes. Such works, for instance, include the ones reported by Krings (Reference Krings, House and Blum‐Kulka1986), Lörscher (Reference Lörscher1991), Wilss (Reference Wilss1996), and Shih (Reference Shih2015). In all these works, the researchers have tried to depict the potential steps or strategies for solving translation problems (see detailed modelling examples in the next section). Regarding decision-making in translation problem-solving, Shih (Reference Shih2015) states that:

With the increasing use of keystroke logging and eye-tracking in translation process studies, researchers have started to investigate the association between translators’ pausing and their problem-solving. In most studies, these two technology-aided observation data sources are combined with translators’ stimulated retrospective accounts. Translators’ long pauses are normally interpreted as signals of their problem-solving processes (Muñoz Martín & Olalla-Soler Reference Muñoz Martín and Olalla-Soler2022). For reviews of research on this issue, see Kumpulainen (Reference Kumpulainen2015) and Muñoz Martín and Cardona Guerra (Reference Muñoz Martín and Cardona Guerra2019).

Additionally, other studies have looked at the correlates of translators’ problem-solving strategies. Araghian, Ghonsooly, and Ghanizadeh (Reference Araghian, Ghonsooly and Ghanizadeh2018), for instance, combined think-aloud protocols and keystroke logging in exploring the correlation between translators’ problem-solving and their self-efficacy levels. Núñez and Bolaños-Medina (Reference Núñez and Bolaños-Medina2018) also examined how translators’ intrinsic motivation and competence predict their self-perceived problem-solving efficacy.

4.2 Translators’ Reading Behaviours

While completing translation tasks, translators spend considerable time not only reading target- and source-text parts but also reading related language and information sources. Each reading type or behaviour translators perform has a strategic purpose. As Hvelplund (Reference Hvelplund, Jakobsen and Mesa‐Lao2017) explains, translation reading is different from normal reading comprehension; it involves reading the source text intensely and thoroughly prior to finding its target text equivalent.

In the two early decades of translation process research, translators’ reading was generally studied as a part of the whole translation process. This can be noted, for instance, in Krings (Reference Krings, House and Blum‐Kulka1986) and Lörscher’s (Reference Lörscher1991) works in which reading represents a component or strategy of their proposed translation process frameworks. An early study focusing solely on reading was reported by Shreve et al. (Reference Shreve, Schäffner, Danks and Griffin1993) who examined translators’ source text reading processing. To address this issue, they used quantitative measures through calculating clause reading time and perceived translation difficulty. They asked 10 translation students to read a text on a computer screen and to translate it. After reading the text, the students were asked to circle the text they perceived would be difficult to translate, and to indicate the cause of the difficulty. Shreve and colleagues found a diversity in their participant translators’ reading comprehension processing. These translators were also interested in lower linguistic levels while reading the source text. They conclude that ‘the translator’s reading of a text may be to some extent more thorough and deliberate than that of an ordinary reader’ (p. 36).

The introduction of eye-tracking in translation process research has opened a novel window for studying translators’ reading behaviours. Jakobsen and Jensen’s (Reference Jakobsen, Jensen, Gopferich, Jakobsen and Mees2008) work is perhaps the first published in-depth eye-tracking study of translators’ reading. They compared professional and student translators’ reading of four similar text types with different purposes: comprehension, preparation for translation, immediate oral translation (i.e., sight translation), and immediate written translation. Their eye-tracking data revealed that while the student translators paid more visual attention to the source text than the target one, the professional translators paid more visual attention to the target text. It was also found that the translators spent more time reading during text translation as compared to other reading tasks.

Other eye-tracking studies have investigated translators’ text reading from different angles. For example, Schaeffer et al. (Reference Schaeffer, Paterson, McGowan, White, Malmkjær, Jakobsen and Mesa-Lao2017) collected eye-tracking data to compare reading for comprehension and reading for translation. They also manipulated the number of target words potentially to be included in the translation of a source word. They found a significant effect for translation task manipulation on total reading time and reading frequency. In another study, Hvelplund (Reference Hvelplund, Jakobsen and Mesa‐Lao2017) used eye-tracking data in studying the cognitive activities underlying the following four translation reading types: source text reading, source text reading during typing, current target text reading, and emerging target text reading. He found significant differences among the four reading types. Hvelplund concludes that ‘the traditional [source text-target text] dichotomy in translation research should be complemented with further itemisation of the types of reading’ (Hvelplund Reference Hvelplund, Jakobsen and Mesa‐Lao2017, p. 74). In a recent eye-tracking study, Mizowaki, Ogawa and Yamada (Reference Mizowaki, Ogawa and Yamada2023) explored the reading behaviours involved in two different translation revision conditions: linear translation where the word order of the source text and target text is similar; and non-linear translation where the word order of the source text and target text is dissimilar. Their study showed that in the non-linear translation condition translators expended greater monolingual and bilingual reading efforts.

4.3 Translators’ Information Source Use

Resourcing or looking for information in linguistic and non-linguistic sources is part and parcel of translators’ tasks. It is generally regarded an instrumental sub-competence in translators’ work (Kuznik & Olalla-Soler Reference Kuznik and Olalla-Soler2018). Translators’ workspace has excitingly changed in the last twenty decades as they now have access not only to traditional printed sources but also to various electronic and online ones (Sycz-Opoń Reference Sycz-Opoń2019). Such changes in translators’ workspace have been accompanied by a considerable amount of research on their use of information sources.

Some studies have addressed translators’ source use from a survey angle. For example, Hirci (Reference Hirci2012) explored Slovene–English translation difficulties and how they are influenced by the availability versus unavailability of bilingual and monolingual printed and electronic references. Hirci’s study drew upon pre- and post-experiment questionnaires to collect data about the student participants’ task-specific source use and their evaluation of the translation tools and information sources they use in translation. Kuznik and Olalla-Soler (Reference Kuznik and Olalla-Soler2018) also conducted a longitudinal study to examine translation students’ source use process skills through measuring the total numbers and types of references consulted, and the stage and duration of source use. The students’ source use profiles were correlated with their translation process and product aspects. In a later study, Sycz-Opoń (Reference Sycz-Opoń2019) used think-aloud, observation and computer screen data to look at student translators’ information-seeking behaviours while converting legal texts from English into Polish. The data of this study was analysed in terms of ‘the information most often looked-up in sources, the sources most often consulted, the level of satisfaction with source consultation, the reasons for non-satisfaction, and the problems commonly encountered during the search for information’ (Sycz-Opoń Reference Sycz-Opoń2019, p. 152).

Other studies have focused only on translators’ real-time use of digital sources. In such studies, we can note how technological advances have influenced researchers’ methodological choice as they have mainly depended on computer-aided observation tools, such as keystroke logging and eye-tracking. For example, Bundgaard and Christensen (Reference Bundgaard and Christensen2019) combined keystroke logs with interviews in their investigation of translators’ source consultations in the post-editing tasks of English–Danish machine-translated technical texts. The digital sources their participants consulted include: concordances, termbases, reference texts, Webpages, and online and offline dictionaries. In three studies, Hvelplund (Reference Hvelplund, Jakobsen and Mesa‐Lao2017, Reference Hvelplund2019, Reference Hvelplund, Svoboda, Biel and Sosoni2023) used eye-tracking and/or screen recording data to look at different issues associated with translators’ use of digital sources. The digital sources translators accessed in the studies varied slightly but included: bilingual and monolingual dictionaries, search engines, reference works and websites, and terminology and conversion tools. In the first study, Hvelplund (Reference Hvelplund, Jakobsen and Mesa‐Lao2017) compared translators’ use of digital resources while drafting and revising their texts. In the second study, Hvelplund (Reference Hvelplund2019) explored the patterns of translators’ source use and how these patterns relate to the flow of the translation process. This study showed that digital sources are used during several translation stages, in one of the following patterns: source text – digital sources – source text, source text – digital sources – target text, target text – digital resource – source text, and target text – digital sources – target text. In a more recent study, Hvelplund (Reference Hvelplund, Svoboda, Biel and Sosoni2023) examined the relationship between European Union (i.e., institutional) translators’ source use and their cognitive attention distribution. The translators were found to spend about a quarter of translation task time interacting with digital sources.

4.4 Translators’ Revisions

In the last two decades, growing attention has been given to researching the changes made to the translated texts. Prior to the 2000s, researchers examined revision as a sub-component of the translation process. With the emergence of machine-translation and translation memory systems, researchers have explored the cognitive acts involved in changing different types of translated outputs. Jakobsen (Reference Jakobsen, Dam, Brøgger and Zethsen2018) tried to draw a distinction among the cognitive acts of these text types as follows:

Research focusing solely on human translation revisions seems to have been published only during the 2000s. The studies reported by Breedveld (Reference Breedveld2002) and Shih (Reference Shih2006) are perhaps two of the earliest focusing on translators’ revision. Two main types of translation revisions can be distinguished: (a) online revisions made during the drafting phase; and (b) end revisions made during the final phase of the translation task (Carl, Dragsted, & Jakobsen Reference Jakobsen, Alvstad, Tiselius and Hild2011; Jakobsen Reference Jakobsen, Dam, Brøgger and Zethsen2018). Like all the areas of translation research, methodological developments can be noted in revision studies. For example, the two early translation revision studies Breedveld (Reference Breedveld2002) and Shih (Reference Shih2006) reported depended on think-aloud protocols and retrospective interviews, respectively, whereas later studies (e.g., Alves & Vale Reference Alves, Vale, Hansen-Schirra, Neumann and Čulo2011; Schaeffer et al. Reference Schaeffer, Nitzke and Tardel2019) used keystroke logging and eye-tracking. Data analysis of translators’ revisions also varied depending on the methodology used. For example, Shih (Reference Shih2006) focused on identifying what revision means to translators, the numbers of revisions made, and the aspects checked in revision. On the other hand, Englund Dimitrova (Reference Englund Dimitrova2005) analysed the syntactic, lexical, morphological, content and orthographic revisions made by translators.

It is worth mentioning that most previous translation revision studies have been concerned with translators’ self-revision rather than other-revision; the latter type aims mainly at optimizing translated text accuracy and fluency and making sure it conforms to the norms of the target language (Jakobsen Reference Jakobsen, Dam, Brøgger and Zethsen2018); that is, it means revising the text translated by another translator (Jakobsen Reference Jakobsen, Dam, Brøgger and Zethsen2018; Robert & Brunette Reference Robert and Brunette2016). Very little attention has been paid to examining the cognitive processes associated with other-revision. Among the few relevant studies published are those reported by Robert (Reference Robert, Way, Vandepitte, Meylaerts and Bartłomiejczyk2013, Reference Robert2014) and Robert and van Waes (Reference Robert and van Waes2014). A main issue investigated in these studies is comparing bilingual and monolingual other-revision processes and products.

As translation revision has become a more technology-mediated process, an increasing number of studies have focused on examining translators’ cognitive processes while editing technology-assisted translated texts. This increase has mainly been associated with the improvements in the quality of machine-translated outputs. The significance of (post-)editing research is that it may reveal the translator’s cognitive capacity and the potential human efforts needed when interacting with machine-translated products, and the relationship of such efforts with machine translation errors and source-text features (Jakobsen Reference Jakobsen, Dam, Brøgger and Zethsen2018; Nunes Vieira Reference Nunes Vieira2017). Some post-editing translation process studies have adopted a comparative approach by examining the cognitive behaviours used in editing machine-translated texts versus human-translated ones (e.g., Daems et al. Reference Daems, Vandepitte, Hartsuiker and Macken2017; Jia, Carl, & Wang Reference Jia, Carl and Wang2019), or investigating professional versus non-professional translators’ machine-translated output post-editing strategies (e.g., de Lima Fonseca Reference de Lima Fonseca2019) or the post-editing processes of outputs translated by different machine translation systems (e.g., Koglin & Cunha Reference Koglin and Cunha2019); for a review of the issues addressed in post-editing research, see Sun (Reference Sun, Lei and He2019). Compared to its machine-translated post-editing counterpart, relatively less research has dealt with translators’ cognitive acts when editing the outputs produced with the aid of translation memory systems. These studies include those reported by Dragsted (Reference Dragsted2004, Reference Dragsted2006), O’Brien (Reference O’Brien2007), O’Brien, O’Hagan, and Flanagan (Reference O’Brien, O’Hagan and Flanagan2010), and Christensen and Schjoldager (Reference Christensen, Schjoldager, Sharp, Zock, Carl and Jakobsen2011). These studies have mostly depended on technology-based observation tools in collecting data.

4.5 Translators’ Cognitive Effort Allocation

A considerable number of studies have addressed the cognitive efforts translators allocate to translation processes; this effort allocation is sometimes labelled cognitive attention distribution. According to Gile and Lei (Reference Gile, Lei, Alves and Jakobsen2020), the main reason for studying translators’ cognitive effort allocation is to understand ‘the link between the effort invested by the translator and his/her performance. … The correlation is assumed to be strong for low effort investment; the translation done carelessly will most probably exhibit weaknesses’ (Gile & Lei Reference Gile, Lei, Alves and Jakobsen2020, p. 265).

Methodologically, previous research has investigated translators’ effort allocation from different angles such as examining their problem-solving and information processing, the time allocated to different translation phases/processes, and pausing and gaze fixation time. Jääskeläinen (Reference Jääskeläinen1999), for instance, depended on think-aloud protocols in identifying the verbalization instances signalling professional versus non-professional translators’ attention to the following translation process categories: (a) procedural comments and global translation strategies; (b) source text or comprehension; (c) target text processing; and (d) other issues. Jääskeläinen found that her participant translators paid the largest amount of attention to target text processing. Using computer keystroke logging, Jakobsen (Reference Jakobsen and Hansen2002) probed the time translators allocate to the following three phases: (a) the orientation phase during which the translator reads the whole source text or part of it; (b) the drafting phase in which the translator produces a partial or complete translation draft; and (c) the end revision phase during which the translator revises the translated text draft and tackles newly discovered problems in it. The keystroke logging data in Jakobsen’s study showed that the orientation phase made up 3% of the translators’ processing time, whereas the drafting and revision phase accounted for 77% and 20%, respectively.

The introduction of eye-tracking in translation process research has offered researchers a different option for studying translators’ effort allocation. In their dependence on eye-tracking, translation process researchers have followed the implications drawn from Just and Carpenter’s (Reference Just and Carpenter1980) seminal work which implies the potential of using eye fixations for identifying instances of effortful cognitive processing. For example, the studies reported by Jakobsen and Jensen (Reference Jakobsen, Jensen, Gopferich, Jakobsen and Mees2008), Sharmin et al. (Reference Sharmin, Špakov, Räihä, Jakobsen, Göpferich, Jakobsen and Mees2008), and Pavlović and Jensen (Reference Pavlović and Jensen2009) compared the differences in eye fixation during source text versus target text reading. These studies found that the latter type of text processing is more cognitively effortful than the former. In a more in-depth study, Hvelplund (Reference Hvelplund2011) combined eye-tracking with keystroke logging data to investigate student and professional translators’ cognitive effort allocation to source text processing, target text processing, and parallel source and target text processing. In his data analysis, Hvelplund focused on (a) attention or time measurement units with identifiable boundaries of cognitive processing (or implied problem-solving), and these were marked by eye movement data (fixations and saccades) and typing activities; and (b) attention or effort allocation shifts indicated by shifts in visual attention. Hvelplund’s study revealed that the translators paid significantly more attention to target text reformulation than to source text comprehension, and that the professional translators undertook more automatic processing than the student translators whose heavier cognitive load was indicated by their attention switching between cognitive processes.

An important issue related to studying translators’ cognitive effort allocation is whether translation is a parallel process or not. Parallel processing in translation occurs when its processes overlap, for example when translators perform two processing tasks synchronously with one task at the centre of attention and another undertaken subconsciously; conversely, sequential processing occurs if there is no overlap (Balling, Hvelplund, & Sjørup Reference Balling, Hvelplund and Sjørup2014). In general, parallel processing in translation is viewed as a dimension in skilled translators’ performance as it indicates its automaticity (Hvelplund Reference Hvelplund2011). Some eye-tracking studies (e.g., Balling, Hvelplund, & Sjørup Reference Balling, Hvelplund and Sjørup2014; Yang, Wang, & Fan Reference Yang, Wang and Fan2022) provide evidence for parallel processing in translation.

Despite the increase in published research on translator effort allocation, some concerns have been raised about the methodological approaches used in it. For example, Hvelplund (Reference Hvelplund2011) suggests that since think-aloud protocols data can only reflect a limited portion of translators’ processing, they may not show the focus of their attention reliably. On the other hand, Gile and Lei (Reference Gile, Lei, Alves and Jakobsen2020) consider that while eye-tracking data indicators such as gaze duration and pausing may inform us about translators’ cognitive effort intensity, they cannot accurately indicate the specific sub-processes within each translation component or the causes of translators’ behaviours. It is believed, therefore, that triangulation can help in data validation and enrichment in studies of translators’ effort allocation.

4.6 Translation Process Styles

A translation process area yet to be given due attention is translators’ cognitive styles. According to Mossop (Reference Mossop, Chesterman, Salvador and Gambier2000), translation process style is the translator’ general translation production procedures or habits, the order in which translators perform cognitive processes while translating, and how they distribute them over the phases of the translation task. It can also be defined as the particular recurrent tendencies and individual behavioural characteristics or profiles noted in the translator’s processing (Carl, Dragsted, & Jakobsen Reference Jakobsen, Alvstad, Tiselius and Hild2011; Dragsted & Carl Reference Dragsted and Carl2013). What is meant by translators’ cognitive styles is explained from Borg’s (Reference Borg2018) description of translators’ varied approaches to starting the translation task:

The importance of studying translators’ styles is that it provides information about the most effective ways of completing translation tasks (Mossop Reference Mossop, Chesterman, Salvador and Gambier2000).

Very few studies have focused on translators’ cognitive styles. Mossop (Reference Mossop, Chesterman, Salvador and Gambier2000) is perhaps the first researcher to call for investigating translators’ styles. He refers to the possibility of categorizing translators’ styles drawing upon Chandler’s (Reference Chandler1993) taxonomy of writing process styles. He also provides two imaginative illustrative examples of two translators’ search for information during the three phases of the translation task (i.e., pre-drafting, drafting, and post-drafting).

The first published empirical attempts to study translators’ processing styles are likely the ones reported by Carl, Dragsted, and Jakobsen (Reference Jakobsen, Alvstad, Tiselius and Hild2011) and Dragsted and Carl (Reference Dragsted and Carl2013). Their research on translators’ styles was influenced by the earlier writing process work by Flower and Hayes (Reference Flower and Hayes1980), Boehm (Reference Boehm1993), Chandler (Reference Chandler1993), and van Waes and Schellens (Reference van Waes and Schellens2003). These two studies made use of eye-tracking and keystroke logging data. In the first study, Carl, Dragsted, and Jakobsen (Reference Jakobsen, Alvstad, Tiselius and Hild2011) classified translators in terms of their revision as: (a) online revisers: translators making revisions while drafting the target text; (b) end-revisers: translators spending more than 20 per cent of the translation time on revising their first draft; and (b) constant revisers: translators combining both behavioural patterns of revisions. In the second study, Dragsted and Carl (Reference Dragsted and Carl2013) identified other categories of translation styles during the pre-drafting and drafting phases, along with the revision styles mentioned above. In Dragsted and Carl’s taxonomy, translators’ cognitive styles during the pre-drafting phase include: (a) head-starters: translators who transcribe the target text immediately without any planning; (b) quick planners: translators who read a few words or sentences before transcribing the first target text part; and (c) scanners: the translators who scan the text rapidly. Online planning styles during the drafting phase include: (a) broad-context planners; translators orienting themselves in a broad context, and attending to a sentence or a group of sentences further down in the text part being read or translated; (b) narrow-context planners: translators focusing on a small text part, for example a few words ahead of the text part being translated; and (c) sentence planners: translators preferring to plan one sentence after reading it.

Meanwhile, Alves and Vale (Reference Alves, Vale, Hansen-Schirra, Neumann and Čulo2011) classify their participant translators’ cognitive styles drawing on keystroke logging data. Their taxonomy of translators’ revision styles include the following categories: (a) drafters: translators who revise the target text in the drafting phase more than in the revision phase; (b) revisers: translators who revise the target text in the revision phase more than in the drafting phase; (c) recursive drafters/revisers: translators who revise the same parts of the source text during the drafting and the revision phases; and (d) non-recursive drafters/revisers: translators who do not revise the same parts of the source text during the drafting and the revision phases.

The exemplary research reviewed in this subsection shows the interesting insights that can be gained from studies on translators’ cognitive styles. However, these studies are generally scarce. Ascribing such paucity to the time-consuming nature of and the difficulties involved in this specific translation process research area, Mossop (Reference Mossop, Chesterman, Salvador and Gambier2000) suggests researchers should start with small-scale studies addressing it. Given that this area is still in its infancy, future relevant studies may investigate the explanatory variables of translators’ styles such as task type, and translator’s experience.

4.7 Translation Process Training

Though translation process research is based on the assumption that it can enrich translator training by enhancing our understanding of trainees’ cognitive behaviours and familiarizing them with effective strategies, not much attention has been given to researching translation process training. Arguably, this low level of attention is associated with a lack of emphasis on integrating process or strategy instruction into translator training. As Pym (Reference Pym, Malmkjær and Windle2011) notes, ‘process-based research … picks up several aspects that are rarely fore-grounded in the pedagogical models based on products. These include speed, the capacity to distribute effort in terms of risk, the use of external resources … and … reviewing’ (Pym Reference Pym, Malmkjær and Windle2011, p. 321). This negligence of process training research stands is in sharp contrast to its importance, which lies in raising translator trainees’ awareness of their task performance weaknesses and strengths and of effective translation strategies, and helping translation teachers understand trainees’ problems (Gile Reference Gile2009; Massey & Ehrensberger-Dow Reference Massey, Ehrensberger-Dow, Kiraly, Hansen and Maksymski2013; Pym Reference Pym, Mees, Alves and Göpferich2009).

Some process training literature is mainly prescriptive, that is, confined to providing suggestions for training translators in translation processes. For example, Gile (Reference Gile2009) used his sequential model of the translation process to show how to raise trainee translators’ awareness of the effective implementation of source text comprehension and target text reformulation processes. Specifically, he refers to drawing trainees’ attention to checking the outcome of their source text comprehension and target text reformulation processes against their linguistic and background knowledge and available resources.

The few available translation process training studies have used different approaches. For example, Alves (Reference Alves2005) reported a pilot study in which the translation processes of 18 student translators were recorded using Translog. The students’ retrospective accounts were collected immediately after the translation task during the replay of Translog data. In the following phases, the students listened to their comments, and cross-analysed and discussed them while watching their Translog pausing and segmentation patterns, compared their retrospective comments to the translated texts, and noted the differences between their performance in the drafting and revision phases. The positive outcomes Alves found in this study led him to conclude that process training is a way to bridge the gap between declarative and procedural aspects in translator training. Pym (Reference Pym, Mees, Alves and Göpferich2009), who drew upon Mossop’s (Reference Mossop, Chesterman, Salvador and Gambier2000) taxonomy of translator styles, also reported using translation task screen-recorded data as a tool for translator training. After collecting this data, Pym asked his participant students to observe and evaluate their translation processes in the following way:

Unlike the previous research, Massey and Ehrensberger-Dow (Reference Massey and Ehrensberger-Dow2011) conducted a translation process training study in which they used more than one observational data source and also collected their data from both students and teachers. Specifically, they gathered a corpus of keystroke logging, eye-tracking, and screen movement data from eight students at four different phases (during their translation degree programme, after completing it, and after gaining professional experience). The eight students were invited back in individual sessions in which they watched anonymously selected recorded sessions of their peers’ translation processes, and were asked to provide commentaries on them. Immediately afterwards, the students were interviewed about their own translation processes, and were asked to compare them to those of their peers. Meanwhile, the translation teachers of these students also attended individual interview sessions in which they watched some recorded data of the translation processes and commented on students’ translation strategies. The results of Massey and Ehrensberger-Dow’s study imply that engaging students and teachers in exploring translation processes in this way resulted in heightening students’ awareness of key aspects of the translation process at different levels, fostering students’ peer-to-peer learning through allowing them to gain comparative insights into translation strategies, and helping teachers understand students’ translation behaviours and realize the potential of using observational data about the translation process as a diagnostic tool in investigating trainees’ translation performance.

It may be concluded that translation process training is still in its infancy. On the one hand, the research reported in this regard is scarce. On the other hand, it can best be described as ‘pilot’ studies, as some of studies were described by the researchers reporting them (e.g., Massey & Ehrensberger-Dow Reference Massey and Ehrensberger-Dow2011; Pym Reference Pym, Mees, Alves and Göpferich2009). It is likely that progress in this research area requires identifying the teachable and non-teachable translation process components, and the processing features translators can use consciously or subconsciously.

5.1 Components of the Translation Process

As indicated in the above paragraph, global models give descriptions of the whole translation process and the main components it encompasses. Two different approaches can be identified in researchers’ conceptualizations of translation process components. First, some researchers have identified the components of the translation process in terms of its phases. An example of this classification type is Jakobsen’s (Reference Jakobsen and Hansen2002) above-mentioned taxonomy (see Section 4.5) in which the translation process is divided into the orientation, drafting, and end-revision phases. Mossop (Reference Mossop, Chesterman, Salvador and Gambier2000) also used an earlier similar translation process taxonomy with the following three phases: (a) pre-drafting phase: exploring the source text and getting acquainted with it; (b) drafting: composing the target text; and (c) post-drafting: reviewing and making any necessary changes in the target text. While these phase-based taxonomies have their rationale, they do not consider the recursive nature of the translation process. In other words, they suggest that translators implement these phases in a linear way, an indication contradicting the well-acknowledged assumption about the recursiveness of the translation process.

The second but much more common way to classify the translation process components is to categorize them in terms of their processing nature. According to Hurtado Albir and Alves (Reference Hurtado Albir, Alves and Munday2009), translation is conceptualized in light of the interpretive theory as a process of three components: understanding or interpreting the source text, deverbalization or target text synthesis, and re-expression or linguistic formulation. Translation process researchers have further elaborated on these three components. For example, Gerloff (Reference Gerloff, House and Blum‐Kulka1986) identifies the following translation process components: problem identification, linguistic analysis, information retrieval, information search and selection, text reasoning and contextualisation, and task monitoring. Krings (Reference Krings, House and Blum‐Kulka1986) also suggests that the translation process encompasses four components: comprehending the source text message and using sources, retrieving equivalent text parts, monitoring the equivalent generated, and decision-making and reduction. For Shreve and Koby (Reference Shreve, Koby, Danks, Fountain and Mcbeath1997), there are four main components in the translation process; these are: comprehending and interpreting the source text message, reformulating the source text message into the target language, and expressing the transposed message in the target language. On the other hand, Hvelplund (Reference Hvelplund2011) believes that the translation process can generally be conceptualized as encompassing two main types of processing: source text processing and target text processing. Source text processing involves reading and comprehending a word/sentence part through fixating it, identifying physical properties of its letters, and lexically analysing it. As for target text processing, it entails reformulating and transcribing.

Some models have also provided global conceptualizations of the translation process components. For example, in Bell’s (Reference Bell1991) model, the translation process is composed of two phases: source text analysis and target text synthesis. The two phases depend on short-term and long-term memories. As noted in Figure 2, source text analysis starts with the visual recognition of a word/sentence part which is then analysed syntactically, semantically, and pragmatically. Through these analyses, which are combined with lexical searches, the translator tries to generate a semantic representation of the source text part drawing upon some kind of planning. Once the semantic representation is approved by the monitor, the translator starts synthesizing the target text part pragmatically, semantically, and syntactically; all the three text synthesis types are also mediated by lexical searches and end with transcribing the target text part. Bell points out that processing the source or target text at syntactic, semantic, and pragmatic levels has no fixed order as the translator’s regular online revision is associated with processing regression.

Figure 2 Bell’s (Reference Bell1991) translation process model.

On the other hand, Danks and Griffin’s (Reference Danks, Griffin, Danks, Shreve, Fountain and McBeath1997) model suggests that the translation process encompasses source text comprehension, target text reformulation, evaluation, and target text production. Their model provides a detailed account of source text reading in particular. According to this model, for the mental representation of the source text message to be formed, the translator draws upon background knowledge in processing its orthographic and lexical features, and analysing its phrasal, propositional, and sentential context. The model also proposes that the translator evaluates the reformulated text in terms of its meaning, and how it meets their intent, the source text writer’s message, and user’s expectation.

Gile (Reference Gile2009) also proposed a sequential model in which he views translation as a process with two-phase processing operations as set out in Figure 3. According to this model, each translation unit – which could be a single word, a sentence part, a sentence, or more – is processed through two phases: comprehending the source text and reformulating the target text. After reading the source text part, the translator temporarily assigns a meaning to it; this meaning is either approved or rejected by the monitor. Once accepted, the translator starts formulating the target text, and the reformulated text is also to be approved by the monitor. The two phases of comprehension and reformulation are mediated and optimized by decision-making which normally depends on knowledge available to the translator from two sources: (a) their knowledge base comprising linguistic knowledge of the source and target texts and background extra-linguistic knowledge; and (b) their task-specific or ad hoc knowledge.

Figure 3 Gile’s (Reference Gile2009) sequential model of the translation process.

More recently, Oster (Reference Oster, Schirra, Czulo and Hofmann2017) provided a translation process model adapted from Levelt’s (Reference Levelt, Brown and Hagoort1999) theory of lexical access in speech production. Oster’s translation process model includes the following components: source text reading, comprehension and conceptualization, and target text formulation, monitoring, and articulating the message. Oster explains the roles of these processes as follows:

Despite the important insights gained from the global models reviewed in this subsection, they have not explicated some roles of particular translation sub-processes or components. For example, the roles of monitoring and information search are only partially explained. More importantly, the models do not show clearly how translation processes/strategies are organized and coordinated. Meanwhile, the processes of target text reviewing and revision are rarely included. Overall, the translation process involves more components and more strategic details than those described in the above-mentioned global models. More details are given in the following two subsections.

5.2 Modelling Translation Problem-Solving Processes

Since translation by nature is a problem-solving activity, many attempts have been made to depict the processes involved in solving translation problems. The importance of these models is that they reveal more detailed information about the complexities of the translation process. Krings (Reference Krings, House and Blum‐Kulka1986) developed an early translation problem-solving model, shown in Figure 4. In Krings’s model, the translation process is viewed as a problem-solving operation triggered by encountering a problem with no satisfactory solution. In order to solve the translation problem, the translator makes use of the following types of strategies:

• Problem identification or analysis strategies: reflecting upon the problem and analysing it using reasoning, inferencing, and available references.

• Retrieval strategies: drawing upon inter-lingual association, the semantic analysis of the source-language text item, and references to produce the target translation equivalent.

• Monitoring strategies: evaluating the translation equivalent(s) generated and deciding whether to accept them or not; if the translator does not accept the proposed solution, they may look for a reduction in the target text (such as dispensing with a particular semantic feature).

Figure 4 Krings’s (Reference Krings, House and Blum‐Kulka1986) translation problem-solving model.

Another early conceptualization of translation problem-solving was proposed by Lörscher (Reference Lörscher1992) who argues that translation strategies have their starting-point when the translator realizes a problem, and ending-point when the translator finds a solution or realizes that the proposed solution is inappropriate. Between realizing the translation problem and finding its solution or discovering its insolubility, further cognitive operations may occur. According to Lörscher (Reference Lörscher1992), the translator may produce multiple translation versions due to solving the translation problem unsuccessfully at the first attempt, trying to optimize the target language text through finding a better alternative version. Lörscher differentiates between two categories of translation strategy elements: the original elements which are part of strategic or problem-solving phases of the translation process and the potential elements which are part of non-strategic phases of the process. Drawing upon think-aloud data, Lörscher found the following 22 translation strategies in the two categories:

Original elements of translation strategies

Potential elements of translation strategies

In Kiraly’s (Reference Kiraly1995) model of translation problem-solving, translation is viewed as a process relying upon three elements: (a) information sources stored in the long-term memory such as translation-related and non-translation-related schemata, and lexico-semantic and syntactic knowledge; (b) the relatively uncontrolled processing centre in which translation operations are performed intuitively and less consciously; and (c) the relatively controlled processing centre in which these operations are performed strategically and more consciously. Kiraly states that the translator’s mind is the information-processing system in which the translation product is formed as a result of the interaction between uncontrolled and controlled operations depending on the information sources stored in the long-term memory. He also explains that a translation problem occurs in the uncontrolled processing centre when the translator is unable to find a tentative translation output; in this case, the problem is attended to in the controlled processing centre. In case the translation problem remains unsolved, it goes back to the intuitive processing centre where long-term memory information interacts with source text input subconsciously.

In conceptualizing the problem-solving process in translation, Wilss (Reference Wilss1996) draws a distinction between two knowledge types: declarative knowledge or knowing what, and procedural knowledge or knowing how. According to Wilss (Reference Wilss1994), decision-making in the translation process is an information-processing feature through which the translator’s cognitive system interacts with their linguistic and non-linguistic knowledge bases, translation task specifications, and text-specific problem space. The interaction among these factors enables the translator to form an internal representation of the translation problem and a solution for it. For Wilss (Reference Wilss1996), decision-making in the translation process involves the following six steps: identifying a problem, clarifying it, searching for and retrieving relevant information, choosing a problem-solving strategy, selecting a solution, and evaluating the solution. Wilss explains that with each of these six steps some further problem(s) may occur, thus causing a delay or an interruption in solving the translation problem.

Robert (Reference Robert2014) developed a conceptualization of problem representation in translation revision. She differentiates between ill-defined and well-defined problem representations. The former representation type occurs if the translator has a vague detection of the problem, or a rejection of the translation. The latter occurs in the form of an intentional, maxim-based or rule-based diagnosis of the problem.

Finally, Shih (Reference Shih2015) proposed a translation end-revision problem-solving process model which is adapted from Wilss’s (Reference Wilss1996) model. This adapted model fills a modelling gap related to translation problem-solving and decision-making during end-revisions. As Figure 5 shows, the steps of the translation end-revision problem-solving process in Shih’s (Reference Shih2015) model are: identifying a problem, defining it, generating a solution, testing or evaluating the solution, accepting the solution or rejecting it, and confirming the decision made through bolstering or de-emphasizing.

Figure 5 Shih’s (Reference Shih2015) model of the end-revision decision-making and problem-solving process.

The models reviewed in this subsection highlight the central role of monitoring in the translator’s decision-making when solving translation problems (Kiraly Reference Kiraly1995; Krings Reference Krings, House and Blum‐Kulka1986; Lörscher Reference Lörscher1992; Shih Reference Shih2015; Wilss Reference Wilss1996). This role is also indicated in some global conceptualizations of the translation problem (e.g., Bell Reference Bell1991; Gerloff Reference Gerloff, House and Blum‐Kulka1986; Gile Reference Gile2009; Oster Reference Oster, Schirra, Czulo and Hofmann2017). Other works have discussed in detail how translators monitor their cognitive processes (e.g., Carl & Dragsted Reference Carl and Dragsted2012; Hansen Reference Hansen and Alves2003; Oster Reference Oster, Schirra, Czulo and Hofmann2017; Schaeffer & Carl Reference Schaeffer and Carl2013; Tirkkonen-Condit Reference Tirkkonen-Condit2005). Collectively, these works associate monitoring use with translators’ evaluation of proposed or transcribed target text alternatives and taking decisions upon them. Tirkkonen-Condit (Reference Tirkkonen-Condit2005), for instance, states that ‘the monitor’s function is to trigger off conscious decision-making to solve the problem’ (Tirkkonen-Condit Reference Tirkkonen-Condit2005, p. 408). Bell (Reference Bell1991), Gile (Reference Gile2009), and Oster (Reference Oster, Schirra, Czulo and Hofmann2017) also view that translators use monitoring for evaluating the proposed target text alternatives (or target text representation) before transcribing them. Arguably, these conceptualizations of monitoring use in the translation process are limited. The next subsection highlights this issue.

5.3 Towards a Remodelling of the Translation Process

While the global and problem-solving models reviewed in subsections 5.1 and 5.2 have significantly widened our understanding of the translation process, some gaps remain unaddressed in translation process modelling. Particular translation processing features have neither been included nor fully explained in these models. For translation process modelling to be complete, we need to explicate the roles of the monitoring, information search, target text reviewing and revision components in translators’ processes.

Monitoring plays a far greater role in the translation process than indicated in previous models and relevant literature. Its role is not only limited to evaluating translation alternatives and problems and taking decisions upon them. In writing process theories, monitoring is generally viewed as the strategist and coordinating component which allows the writer to move from one composing process to another (Flower & Hayes Reference Flower and Hayes1981). It is also defined as checking on the information being processed and how it is being processed (Ruiz-Funes Reference Ruiz-Funes1999). Moreover, writing process literature indicates that monitoring is the component responsible for organizing text composing processes and managing them, evaluating the proposed or written idea/text, and also regulating one’s motivation during writing (see Abdel Latif Reference Abdel Latif2021). Likewise, it is assumed that translators not only monitor their source text understanding, and proposed or transcribed target text production, but they also monitor their task performance procedures, and how they are implemented. Accordingly, there is a need for expanding the definition of monitoring in translation process research.

Similarly, the role of resourcing or the search for information is not fully described in previous translation process models. Some but not all of the conceptualizations reviewed above include an information search element (e.g., Bell Reference Bell1991; Gerloff Reference Gerloff, House and Blum‐Kulka1986; Krings Reference Krings, House and Blum‐Kulka1986; Oster Reference Oster, Schirra, Czulo and Hofmann2017; Wilss Reference Wilss1996). These conceptualizations generally tackle translators’ linguistic (mostly lexical) information retrieval, but do not refer to their use of sources. Thus, these conceptualizations are inconsistent with the research findings indicating the large amount of time translators spend on using information sources (see Section 4.3). During translation tasks, translators look for lexical or grammatical items, and they may also have to find information about the meaning of some unfamiliar source text words/phrases prior to finding target text equivalents. In searching for these different information types, translators depend first on retrieval of the information stored in their long-term memory, and – if their memory retrieval attempts fail – they use external information sources. We can conclude, therefore, that translators’ search for information either in the form of memory retrieval or source use constitutes a central part which takes much of translation task time (e.g., Hvelplund Reference Hvelplund, Jakobsen and Mesa‐Lao2017, Reference Hvelplund2019, Reference Hvelplund, Svoboda, Biel and Sosoni2023). As a result, we need to demystify translators’ information search, and characterize it as a separate component in translation process modelling.

Additionally, we cannot ignore reviewing and revision when modelling translators’ processing. While performing their tasks, translators review the proposed target text (i.e., check how it matches the intended meaning in the source text prior to transcribing it), and also the transcribed text (through verifying its meaning, and scanning and reading it). In their think-aloud protocol data, Mondahl and Jensen (Reference Mondahl and Jensen1996) noted translators’ use of the former review type (i.e., verifying the meaning of the proposed target text) through back translation. Specifically, they found that translators try to identify ‘differences between the source text and a potential translation equivalent. … [And they] may also test adequacy by “translating back” to the source text in order to check whether the two expressions are similar’ (Mondahl & Jensen Reference Mondahl and Jensen1996, p. 103). While this view of the role of reviewing in the translation process concurs with some writing models (e.g., Abdel Latif Reference Abdel Latif2021; Chenoweth & Hayes Reference Chenoweth and Hayes2001), translation modelling literature generally implies that reviewing is an embedded process in translators’ text production (Danks & Griffin Reference Danks, Griffin, Danks, Shreve, Fountain and McBeath1997; Gile Reference Gile2009; Oster Reference Oster, Schirra, Czulo and Hofmann2017) or revision (e.g., Shih Reference Shih2015; Wilss Reference Wilss1996). The gradual disappearance of the think-aloud method in translation process studies (see Sun (Reference Sun2011)) and researchers’ over-dependence on keystroke logging and eye-tracking have not helped in distinguishing translations reviewing from text production or revision; arguably, the distinction between these processes can be made through gathering concurrent verbalizations rather than keystroke logging and/or eye-movement data only. For a better modelling of the translation process, we need to conceptualize or characterize the three processes (monitoring, target text reviewing, and revision) as separate components due to the following reasons: (a) monitoring is responsible for supervising all translation processes and procedures by initiating, and approving or rejecting them; therefore, it interferes in all translation processes rather than in reviewing and revision only; (b) reviewing the target text through reading or scanning may not necessarily result in changing it; (c) translators may change the target text without reading or scanning it (for example, in case of realizing an error made while transcribing); and d) revision entails many complexities as translators make changes in the target text at the syntactic, lexical, morphological, content and orthographic levels through additions, omission or substitution (see Englund Dimitrova Reference Englund Dimitrova2005).

Overall, the issues reviewed and discussed in this section indicate a need for a different translation process model which provides a more detailed and clear-cut description of its components and shows how they are managed and coordinated. In this work, a translation process model is proposed to meet these requirements. The proposed model is adapted from Abdel Latif’s (Reference Abdel Latif2021) writing process model given in Figure 6. This adaptation is based on the assumption that writing and translation processes are similar and share a number of common phases and characteristics (Dam-Jensen & Heine Reference Dam-Jensen and Heine2013; Göpferich & Nelezen Reference Göpferich and Nelezen2014; Immonen Reference Immonen2006; Kobayashi & Rinnert Reference Kobayashi and Rinnert1992; Risku, Milosevic & Pein-Weber Reference Risku, Milosevic and Pein-Weber2016; Robert & Brunette Reference Robert and Brunette2016; Uzawa Reference Uzawa1996). Risku, Milosevic, and Pein-Weber (Reference Risku, Milosevic and Pein-Weber2016), for instance, point out that:

Likewise, Immonen (Reference Immonen2006) found that both writing and translation require approximately the same time proportion in the task orientation phase, though translation needs more revision and monitoring behaviours. Noting such common characteristics in writing and translation cognition, Dam-Jensen and Heine (Reference Dam-Jensen and Heine2013) suggest that these forms of cognitive research could benefit from each other (p. 90). With regard to process modelling in the two fields, they specifically state that:

As noted in Figure 6, Abdel Latif’s (Reference Abdel Latif2021) writing process model describes text composing as a process comprising seven sub-processes or components. In this model, monitoring is the central process responsible for managing and organizing the process of text composition, and evaluating the proposed or written idea/text and writing procedures. With these multiple functions, monitoring plays the role of working memory in the writing process. It encompasses the following types of writing process strategies: (a) task-management: setting goals for writing procedures and observing task time and reminders; (b) ideational and metalinguistic reasoning; (c) evaluation: identifying a problem, approving or rejecting proposed writing alternatives and procedures; and (d) motivation or emotion regulation. This expanded definition suggests that the following six writing process components are monitored:

• Searching for content: activating memory retrieval and using external sources for finding what to write (searching for ideas) or for how to write it (searching for the target linguistic alternative(s)).

• Ideational planning: proposing ideas for writing at the essay or whole text, paragraph or sentence level.

• Linguistic rehearsing: proposing the linguistic form of the sentence or sentence part to be written prior to deciding whether or not to include it in the text.

• Reviewing: verifying the meaning of the proposed or written text and scanning and reviewing the text written.

• Transcribing: translating the proposed text into written language for the first time using a pen/pencil and paper, or typing tools.

• Text revising: changing the written text; writers’ revisions can be classified in terms of their operation type (addition, deletion, and substitution), their timing (online and post-writing), and the linguistic unit level changed (revising a sentence, a phrase or a word; editing grammar, spelling, or punctuation).

For detailed definitions of these writing process components and their think-aloud protocol examples, see Abdel Latif (Reference Abdel Latif2021).

Figure 6 Abdel Latif’s (Reference Abdel Latif2021) writing process model.

Based on Abdel Latif’s (Reference Abdel Latif2021) writing process model and the insights gained from the translation process literature reviewed in this work, the model proposed in this work provides a more detailed description of the translation process, and shows how its components interact with each other. As Figure 7 illustrates, the model depicts translation as a process with the following seven components or sub-processes: monitoring, source text representation, searching for information, target text rehearsing, target text reviewing, transcribing, and revising. Monitoring in this proposed model assumes the role of working memory as it is the organizer and coordinator of the whole translation process and its other components. It initiates the goals of the translation task and its subtasks or parts, allows the translator to move from one sub-process or step to another, forces the translator to accept or reject the proposed or transcribed target text, identifies any potential problem(s) in it, and alerts them to check task time and notes. In the proposed translation process model, monitoring manages and supervises the following six components:

• Source text representation. This process encompasses reading a source text part (e.g., a word, phrase, and a sentence) and interpreting its meaning in the source language. Source text interpretation is a prerequisite for the translator to be able to start searching for its target text equivalent. This representation process may occur subconsciously and synchronously during source text reading if the translator is familiar with its meaning in the source language; but it may take some time if the translator is unfamiliar with the meaning of the source text part being read, for example when it is jargon or has two potential source language meanings. In this latter case, the translator will reflect upon the meaning of the source text part, or – when necessary – use an online or printed reference to reach its optimal representation. Here, the monitor will interfere to approve or reject the translator’s representation of the source text part. If rejected, another source text representation attempt will be made.

• Searching for information. According to the explanation given at the beginning of this subsection, the translator searches for different types of information during the translation task. It is expected that the translator will be preoccupied with searching for appropriate word/phrase equivalents. The translator normally starts with memory retrieval of the information stored in the mental lexicon; thus, memory retrieval draws upon long-term memory. If memory retrieval attempts have not succeeded, the translator searches for the target word/phrase equivalent in online or printed dictionaries or other lexical sources. The translator can also access machine translation and/or translation memory systems while performing tasks to facilitate their work. Meanwhile, they may follow the same memory retrieval/source use cycle in accessing other types of linguistic information; for example, when trying to phrase a particular sentence in an appropriate grammatical form, or when making sure of the meaning of a given word/phrase in the source language to represent it accurately. In each information search process, the monitor interferes again to approve or reject the information retrieved from the long-term memory or found in the sources consulted.

• Target text rehearsing. This process involves reformulating the target text before transcribing it. After retrieving the equivalent potentially matching the source text part, the translator mentally rehearses how to write it in a meaningful target text unit (e.g., a phrase, sentence part or a sentence). Target text rehearsing might be also made in a written form. The monitor normally supervises this process through accepting or rejecting the proposed target text alternative(s).

• Target text reviewing. As indicated in the beginning of this subsection, reviewing does not only include scanning or reading the transcribed target text, but it also encompasses checking the suitability of the proposed target text (e.g., through back translation or repeating the rehearsed text). After reviewing either the proposed target text or the transcribed one, the monitor decides upon the suitability of each alternative.

• Target text transcribing. This process involves transcribing the target text part for the first time through handwriting or typing tools. Once the target text part is transcribed, the monitor allows the translator to move to another process.

• Target text revising. Revision entails changing the transcribed target text as a result of identifying a problem in it. The revision process is triggered by the monitor and also ends once the revised target text part is approved by it. Translators’ revisions can be classified into different categories such as the syntactic, lexical, morphological, content, and orthographic changes made (Englund Dimitrova Reference Englund Dimitrova2005), the operation type (addition, deletion, and substitution), or the linguistic unit level changed (a sentence, a phrase or a word; grammar, spelling, or punctuation) (Abdel Latif Reference Abdel Latif2021).

Figure 7 The proposed translation process model.

As the proposed translation process conceptualization shows, monitoring attends to each of the six remaining translation processes and the strategies representing them. It supervises these processes, evaluates the suitability of their outcomes, and based on this evaluation it allows the translator to move from one process to another. The monitor’s supervisory role also interacts with the information retrieved from the long-term and short-term memories, the information found in the sources consulted, and with the source text translated so far and the target text produced. The translator performs a large number of cognitive activities; each activity relates to completing a part in the task. A particular cognitive activity in the translation task is initiated by the monitor either through setting a goal for this activity or through noticing a problem. After the initiation of the translation activity by the monitor, the translator may engage in any other cognitive process such as source text representation, target text rehearsing or reviewing, or revision. The translator’s cognitive processes occur recursively in no fixed order, and the one translation activity does not necessarily include all these processes; the occurrence of a particular set of translation processes in the one activity depends on activity purpose, time constraints, cognitive overload, or the automaticity resulting from the ease of the subtask at hand. For example, the cognitive overload associated with limited target language knowledge may cause the translator to be unable to monitor all the translation operations and to allocate more efforts to information search during the task (for an example in writing process research, see Abdel Latif (Reference Abdel Latif, Bailey and Damerow2014a)). The translator’s cognitive style may also play an influential role in the implementation and regular use of particular translation processes. Likewise, a high degree of familiarity with the source text genre may lead the translator to rehearse target text parts less regularly or monitor some processes, or to undertake minimal information search during the task. The impact of source text genre and task familiarity on the automaticity of the translator’s cognitive processes can be compared to the automaticity levels in car driving of novice drivers (who normally concentrate fully on their driving) versus expert drivers (who drive more automatically due to their driving familiarity and experience). Compare also this conscious–subconscious perspective to Oxford’s (Reference Oxford1990) view that language learners use cognitive strategies consciously at the beginning, but after some time they use them automatically. Despite the variance in using translation processes in different cognitive activities, the translator will use all of them while completing the whole translation task. In other words, varied attentional efforts will be allocated to different translation components and the strategies representing them depending primarily on the translator’s target language proficiency, familiarity with source text genre, task complexity, and available information sources. These issues can be regarded as testable hypotheses in future research.