The Journal of Specialised Translation
Issue 05 - January 2006
Retrospection methods in translator training and translation research
This article presents the results of the Copenhagen Retrospection Project which was carried out at CBS in 2004 in connection with research in translation processes. It is a systematic comparison of several methods for introspection, the Integrated Problem and Decision Report (IPDR) by Gile (2004), Retrospection with Replay with Translog (R+Rp) and Retrospection with Replay combined with cognitive clarification via an immediate dialogue(ID) between the subject and the observer (R+Rp+ID). The methods were tested and compared in terms of their applicability, their influence on the number of problems and decisions mentioned by the subjects, the amount of information each of them provides, their long-term effect on trainees, as well as their general value for translator training and translation research. The most important results are that they all contribute to raising awareness of the translation process and that they - though they differ in many respects, especially as to source consultation - are suited to be used complementarily, both together and combined with other methods.
Copenhagen Retrospection Project, IPDR, R+Rp, R+Rp+ID, introspection methods, cognitive clarification, immediate dialogue, translation processes, source consultation, complementarity.
Over the last few years I have carried out a research project in which I have investigated Störquellen, which are sources of disturbance in translation processes(SDs), i.e. sources of translation problems that are either overlooked in translation teaching or that cannot be coped with properly because of lack of time. The project consisted of five series of experiments and control experiments. Altogether about eighty subjects, all final year students from CBS, were involved (Hansen 2005a).
More than twenty different SDs were discovered (see also section 2). The conclusion of the study was that much trouble and time could be saved if students, at various stages during their translator training, were submitted to individual tests. What I am thinking of here is not the usual kind of test where errors are marked in target texts and discussed, but individual translation competence profile tests that allow students to become aware of some of the causes of their errors. In my project, I had used Translog - with and without time pressure - and retrospection, which I combined with Translog's replay function and an immediate retrospective dialogue, R+Rp+ID.
The dialogue is a process of negotiation with the purpose of reaching subjective and inter-subjective identification and clarification of a phenomenon of interest, which in translation processes for example can be: a problem, a strategy, a decision, a revision, a cursor movement, an attitude, an error, aspects of the source text or the translation brief. A dialogue creates clarity about where the discussed phenomenon belongs in the conceptual structures of the subject. On the one hand, the dialogue is oriented towards the describing subject him/herself who is going through the process of identification, clarification and verbalization, and who, as a result of it, develops and increases his/her own understanding of the phenomenon. On the other hand, it is oriented towards the observer who tries to understand the message and to promote the act of clarification, and who gives feed back. As mentioned in Hansen (2003a: 35) the condition for such a dialogue in an experiment with translation processes is that the subject is able to verbalize his/her thoughts about phenomena, problems, actions and decisions. This presupposes that he/she already has reached some degree of clarity and has the expressive means available. Because of the immediate retrospective dialogue the method combination R+Rp+ID requires a great deal of effort from the observer and the challenge is now to find other methods or method combinations that can make it easier to detect and eradicate SD’s, or - at best - to prevent them.
Two other retrospection methods seemed to be useful: IPDR, a retrospection method described by Gile (2004), and retrospection with replay, but without the immediate dialogue: R+Rp.
In this article I will discuss the application of IPDR and R+Rp in translator training and translation research. I will show the results of a comparison of processes and products where the two methods were applied, the kind and amount of information each of them provides about individual processes and the influence they might have on the processes and products, especially with respect to errors. I will also incorporate comments from the subjects concerning the application of the methods in translator training and of their advantages and disadvantages. Finally, I will discuss method combinations and evaluate their usefulness and effectiveness for detecting, eradicating and preventing SD’s in translation processes. In order to clearly distinguish this study with IPDR from the experience described by Gile (2004), I call this study "the Copenhagen Retrospection Project".
In many studies of translation research the objective is an investigation of general issues regarding the complex phenomenon of 'translation', without much consideration of translator training. In this study, where it is students' processes that are investigated, research and training are closely connected: results from research have an immediate impact on training and vice versa.
2. SDs in translation processes
Often we realize that something is wrong in a translation product and we mark errors - again and again - but we do not exactly know the causes of the errors. I have defined 'SDs' as "causes of problems and errors which cannot be explained by deficient knowledge or skills be they social, communicative or linguistic" (Hansen 2005a: 17/303f.) . They are not errors in themselves and neither can they be explained away by the lack of competence in a certain area, as for example lack of social, communicative or linguistic competence. Analyses of the complex relations between the individual profile, process and product are necessary in order to detect them. SDs are for example unfortunate attitudes, habits or strategies, different kinds of compulsive behaviour, and fundamental misunderstandings. They include an over-attention to detail, laziness of thought, presumptuousness or, conversely, lack of self-confidence, strategies that are not appropriate to the translation process at hand, although they have been shown to be useful in other situations, fear of interference and various reasons for lack of control.
3. Retrospection methods
3.1. IPDR (Integrated Problem and Decision Reporting)
Gile (2001: p2) asserts that "IPDR has shown its didactic usefulness in a process-oriented training approach over the years". The advantages of this method are that it is a convenient form of retrospection which is not very time consuming for the trainer because the students' comments are directly readable. It provides "information about students' problems, both individual and collective, and information about their translation strategies". It is "a means to raise their awareness of various components of the translation process and promotes best efforts towards maximum quality" (Gile 2004: p2). As it is not bound to special experimental situations or to a particular environment, IPDR can be widely used in the classroom. The method can be used without further cues or reminders from the teacher except for the initial instruction. The comments allow a precise analysis of general and specific problems, and can be used for synopsis and feedback.
Gile describes several advantages of the method for translator training. With respect to IPDR for research purposes he is, however, more guarded. (section 2.4.3), especially because reporting takes time and effort, and students cannot always be relied on to do it thoroughly.
In the following comparison and discussion of the methods, some of Gile's observations with IPDR are integrated and discussed.
3.1.1. Translating with IPDR
When translating with IPDR, the subjects are assigned a realistic translation task with a translation brief and are asked to comment on every problem they meet during the translation process. The brief for the retrospection is as follows:
Report every problem that you encountered, how you attempted to solve it, and why you decided on the solution that you adopted.
For every sentence, term or sentence structure for which you used an outside source indicate specifically the full references of that source (including the bibliographical references or web site address, and the qualifications of a human source), and preferably give the context (sentence or paragraph) where you found the information.
Make sure this is done individually, for every problem, using no general statements such as "I used this or that web site".
3.2. Translation processes with Translog's replay function: R+Rp and R+Rp+ID
When using the retrospection methods with Translog and replay, subjects are given a translation brief and are asked to work on a realistic translation task. During the translation they are alone in the room. As soon as they have finished, the process and product are saved and the writing process is replayed. The subjects see all their cursor movements, pauses and revisions on the screen and describe, in retrospect, what thoughts went through their minds during the process. They comment on the problems they encountered and the strategies and aids they used in order to solve them. These comments are recorded and transcribed.
The only differences between R+Rp and R+Rp+ID are the observer's role and the combination of the replay with an immediate retrospective dialogue.
Retrospection with replay means that the subjects work alone throughout the whole test or experiment. The observer only disturbs them after the translation process is finished in order to save the target text and to establish and start the replay function. After that, the subjects observe the replay of their writing process on the screen and comment on the translation process, problems and problem solving. The reports are recorded and transcribed.
For this method-combination, which I used in experiments from 1997 until 2002 (Hansen 2005a), the observer is present during the replay and listens to the retrospection. Immediately after the subject stops commenting on his/her translation process, the observer initiates a retrospective dialogue with the subject about phenomena like the subject's behaviour during the process, individual problems, problem solving, errors and any other issues that might seem to be relevant (see section 1).
4. Theoretical background
This comparative study of methods is based on interdisciplinarity. On the one hand, it combines translation inherent disciplines, such as translation theory, communication, linguistics, text linguistics, pragmatics and stylistics. The choice of the source texts and the translation briefs are based on insights from these disciplines, as are the descriptions and categorizations of the problems mentioned during the experiments, and the classification of errors from the assessment of the translation products (Hansen 2005a). The assessment was carried out in accordance with the ideas of functional translation combined with the ethical norms for translation and interpreting in Danish society (Hansen 1997). As to the assessment of products, see "research design" in section 5.
Interdisciplinarity in this study also means intermethodology. Disciplines and research patterns from psychology, phenomenology, natural sciences and social sciences provide empirical translation research with useful tools, methods and techniques. I used introspection methods (Ericsson & Simon 1993), complementary approach (Bohr 1959/1964), description methods from Copenhagen phenomenology (Tranekjær Rasmussen 1955), communication and interview techniques (Moustgaard 1981, 1990), combination and triangulation of methods and data (Denzin & Lincoln 1994), and coding procedures from Grounded theory (Strauss & Corbin 1998).
4.1. Retrospection in TS - reminders and retrieval cues - observers' effect
The different kinds of retrospection described in this article are introspection methods in line with the more often used "think aloud" (TA). In comparison with TA, which takes place simultaneously with the translation process, retrospection has until now been regarded as less reliable and more error-prone. As it is carried out after the process and as subjects easily forget what they have done, they are believed to tend to distort their observations (Ericsson & Simon 1993: 18, Krings 1986: 68).
When using TA during a translation, many cognitive processes are carried out automatically, with the result that subjects tend to stop talking, especially in situations of high cognitive load (Jääskeläinen 1999: 101). In order to avoid this and to enhance production of TA, Ericsson & Simon (1993: 83/256) proposed the use of reminders to make the subject speak when he/she falls silent. However, every time subjects are asked to continue talking, they are inevitably also reminded of the fact that they are participating in an experiment and that there is an observer present.
When using retrospection, the central issue is always the act of "recalling". Memory, recognition and retrieval are the most important aspects. Ericsson & Simon (1993: 117) proposed that recognition is easiest using different reminders and retrieval cues which also can have uncontrolled observers' effects.
Disadvantages of that kind can be reduced to a minimum if methods are combined and/or triangulated. Retrospection combined with a replay of the writing process makes this method more reliable. Furthermore, it opens up for new possibilities, because the replay makes subjects recall some of their thoughts during the process automatically - without resorting to any reminders or retrieval cues. The ST, as well as pauses and revisions, which can be seen on the screen, take over the function of reminders and cues in stimulating the subject’s memory.
With IPDR, subjects choose themselves when to write the reports of their problems and decisions - parallel to every act of problem solving, after having finished the first draft, or after having written the final target text. In this method, each of the recognized problems and decisions functions as a reminder or retrieval cue.
The advantage with IPDR and R+Rp in relation to TA is that the observer does not need to interfere in the process. Bias from observers’ effects can be minimized because reminders and cues are unnecessary. A further advantage is that with both methods the different modes of expression writing and talking are not used simultaneously, so that there is no impact on the translation process from talking.
R+Rp differs from IPDR as retrospection with replay always takes place after the whole process is finished and, thus, should not have any direct influence on the translation process itself (see section 6.2). With IPDR the reports can be written during the process and this can have an impact on the process (see section 6.2).
4.2. Combination and triangulation
Experiments with R+Rp provide the researcher with three different sources of observation and qualitative and quantitative data: the recorded retrospection, the log files from the translation process and the evaluation of the translation product. With IPDR the researcher has two sources of observation and qualitative and quantitative data: the written report and the evaluation of the translation product.
In this study, I combine data from processes with data from products and individual retrospective reports. For the comparison, results from the experiments with IPDR are triangulated with results from R+Rp with respect to certain aspects. Finally, the result of the comparison is triangulated with reports from the subjects on how they felt about using the two methods, and with observations from my experiments with R+Rp+ID. (For further discussion of combination and triangulation see Hansen 2005a: 61).
4.3. Some definitions
Terms like "problem" and "strategy" are in my study used in the common meaning of the terms, as they are found in Duden (1996) and Leo (2005). A problem is a "difficult (unsolved) task", or a "difficult question proposed for solution". The problems that subjects usually mention during the retrospection can be broken down into different categories such as lexical, structural or pragmatic problems (see section 6.1). A strategy is "a plan of action intended to accomplish a specific goal" (Hansen 1999: 44 and 2005a: 312).
5. Research design
In two series of experiments, I tested IPDR and R+Rp with six subjects, who were final year students at CBS with Danish as their mother tongue. Five of the students were female and the sixth was male - all about the same age. They had followed different courses in translation theory, textual analysis, revision of translated texts and translation for special purposes. That is why they were used to analysing and discussing translation problems and strategies. They had already written IPDRs with two translation tasks into both directions. The Copenhagen Retrospection Project was part of a special course about translation processes.
For the experiments - translating from Danish into German and from German into Danish - the subjects were divided into two groups. According to results from pilot experiments, the groups seemed to be quite equal with respect to translation competence. In one series, they were asked to translate two short original Danish tourist flyer texts from Lyset over Skagen, text 1a and text 1b, for German tourists. For the translation into Danish, two texts, one from Stern 51/2003, text 2a, and one from the German popular science journal Psychologie Heute from April 2004, text 2b, were translated for a Danish popular science journal Psykolog Nyt. The text pairs were the same text type, had the same function and target group and they were about the same length. In pilot experiments and earlier experiments with other subjects (Hansen 2003b:325f), I had checked that the texts were of roughly the same degree of difficulty and contained the same number of potential problems. I used neutral texts, which did not require special knowledge, because I wanted to avoid the situation where special traits in the texts dominate the experiments and the results. For the source texts and translation briefs, see the appendix. All together 24 experiments were conducted.
I had used texts 1a and 1b in earlier experiments, where I investigated the connection between time management and the quality of translation products when translating different text types. Those experiments were also carried out with a group of six subjects, who then translated with R+Rp+ID (Hansen 2002, 2003a, 2003b). This means that I now should be able to compare the use of R+Rp and IPDR with R+Rp+ID, but the comparison suffers from some bias because the individual translator profiles in 2002 differed considerably from those in 2004. In 2002, two of the subjects were bilinguals with very different backgrounds, whereas in 2004 all subjects had a more similar linguistic background, as they all had learned German at Danish schools and university only. As to the impact of bilingualism on translation processes between Danish and German (Hansen 2005a).
In the present project, the experiments with R+Rp were carried out at my office. Every subject had done a pilot experiment so that they felt comfortable with the situation and had got used to the computer. The experiments with IPDR were carried out at home. The subjects did not talk to each other about the experiments. They all received exactly the same instructions for the retrospection tasks. I used the wording of IPDR (see section 3.1.1) for all experiments. As already mentioned, all aids and reference books usually used by the subjects were available, including the internet.
First I asked group 1 to translate text a using Translog and R+Rp while group 2 translated text a with IPDR. Then I switched over, so that group 2 translated text b with R+Rp while group 1 worked on text b with IPDR. After having transcribed the tapes from the R+Rp, the reports from IPDR and R+Rp and the translation products were analysed anonymously. The products were evaluated in terms of errors which were marked by two potential recipients of the target texts and myself. Only errors we spontaneously agreed on were registered. All decisions not considered errors by the three evaluators where classified as 'good'.
These experiments with IPDR and R+Rp were followed by a delayed retrospective dialogue (DD) with the subjects, individual feedback, and a discussion in class. The dialogue had to be 'delayed', because none of this could take place until all the translation products had been evaluated and the reports and tapes had been analysed, i.e. about two weeks after the experiments.
6. Results of the comparison
The analysis was carried out on all four translations, the two processes and products into Danish and the two processes and products into German. I concentrated on the following issues:
6.1. Writing versus talking: number of problems and source consultations
With IPDR subjects have to write and with R+Rp they talk. As Gile (2004: p2 and section 2.4.3) points out, with IPDR "writing about all the problems and decisions takes time and effort". In comparison, R+Rp does not take much of the subjects' time. It is carried out at double or even triple writing speed and because of the pauses and revisions during the process there is always enough time for reporting.
Thus an important difference between the two methods is that R+Rp is very easy for the subjects to carry out. This may be one of the reasons why, with R+Rp, the number of problems reported is larger than with IPDR. The Fig. 1 and Fig. 2 show this:
Fig. 1: Problems, dictionaries and the internet reported
Danish - German:
Fig. 2: Problems, dictionaries and the internet reported
The difference is most marked with respect to the number of problems mentioned; in the use of dictionaries and the internet there is not much difference. In the translation process from Danish into German, with IPDR there are more comments on dictionary use and internet searches than with R+Rp.
There may be another reason why one could expect more problems reported with R+Rp than with IPDR. As Gile (2004: section 2.4.3) says, "students cannot be relied on to do it thoroughly". As mentioned in section 3.2 and 4.1, with R+Rp the pauses, cursor movements and revisions on the screen animate the subjects to recall and to comment on their thoughts during the process. When they see on the screen that they change the structure of a sentence several times, they almost automatically comment on their structural problems. In addition, the subjects know that after the experiments the observer can repeat the replay and check which of the problems they had (these can be seen on pauses and revisions) but did not comment on.
The problems mentioned by the subjects with R+Rp can be broken down into lexical (problems with finding a word or term), structural (the question of how to construct of a sentence, word order), idiomatic (the question of whether a term would be used in the situational context) or pragmatic problems (how to adapt the TT to the TT-receiver). In addition, there may be problems with the rationale, which I called "semantic logical" problems (sem.log). They talked about reception or production problems (i.e. that they did not understand the source text or did not know how to express the meaning in the target text), spelling problems or problems with respect to the appropriate style in the context. Sometimes subjects mentioned word flexion in relation to the syntax of the sentence (morphological problems or msyn).
Though the subjects using IPDR in the Copenhagen Retrospection Project had been trained in applying the method before the experiments and though they were asked to report all problems, some of the problems seem not to have been reported.
As Gile (2004: section 2.2.1.a) says, "students report any problem they consider significant". It is their choice and perhaps they considered some of their problems less significant. This can, for example, be illustrated with spelling problems, which emerge with the R+Rp, when the subjects see their orthographic revisions during the replay - in both directions. Spelling difficulties are not mentioned at all with IPDR. See the following figures that give a breakdown of the kinds of problems mentioned with IPDR and R+Rp:
Fig. 3: Kinds of problems mentioned (German-Danish)
Problems mentioned when translating from Danish into German with IPDR and R+Rp:
Fig. 4: Kinds of problems mentioned (Danish-German)
As these diagrams show, apart from the spelling problems nearly the same kinds of problems are reported with both methods.
6.2. Influence of the methods on the translation process
The experiments show similar results as to the number of appropriate translation solutions, i.e. good decisions, taken in relation to the problems mentioned. In both directions about 80% of the problems are followed by a satisfactory decision. See the following figures:
Problems and good decisions when translating from German into Danish:
Fig. 5: Problems and good decisions
Problems and decisions when translating form Danish into German:
Fig. 6: Problems and good decisions
In terms of their influence on the process, the two methods also differ in many respects: As Gile (2004) points out, IPDR is not an online task and the reports are "an integral part of the translation assignments". The act of reporting can have a direct influence on the process. R+Rp, in contrast, is an online task and the translation process should not be influenced by it because the retrospection takes place separately after the translation process has finished.
In my experiments from Danish into German, I realized that the methods had different kinds of positive and negative impacts on the translation processes:
6.3 Influence of the methods on the products - with respect to errors
The influence of IPDR and R+Rp on the process is partly reflected in the kinds of errors in the product. The products of the translations from German into Danish showed more errors with IPDR than with R+Rp whereas the products from Danish into German showed the reverse. This can be due to certain reasons that cannot easily be controlled in experiments where tasks and methods are switched; for example, the fact that texts or parts of texts in relation to subjects always differ with respect to kinds and number of difficulties. According to their experience and precognition, subjects/groups of subjects may react differently when confronted with them. There could also be a retest-effect.
Errors with IPDR and R+Rp in both directions:
Fig. 7: Number of errors with IPDR and R+Rp in both directions
The potential influence from IPDR and R+Rp can be identified through a rough classification of the main errors in the translation products.
In this classification, all kinds of interferences, such as lexical or structural influence from the source language on the language of the target text, are gathered under int. The choice of a wrong collocation from dictionaries or the internet is marked lex-sem. With semantic logical errors, sem.log, the meaning of larger parts of the text is distorted (also in relation to the source text), and this without any special pragmatic reason or explanation. An idiomatic error, idiom, is not a semantically incorrect expression, however a native speaker would not use it in the actual context. Morphological errors are marked as msyn. Under others different kinds of errors are gathered, for example errors as to pragmatics, style or reference.
As to the number and distribution of errors with both methods - see the following Figures 8 and 9:
Fig. 8: Types of errors (German-Danish)
Fig. 9: Types of errors (Danish - German)
Looking at the translations from German into Danish, the most significant difference between the two methods concerns semantic logical errors. Apart from potential reception problems, a reason could be that when translating into their mother tongue, some of the subjects are used to begin by writing the whole translation in one stretch. As also can be seen with R+Rp - they write in one go and just leave out anything they cannot translate at once - and then go back later in the process and revise the draft. The subjects did not use this common practice in the experiments with IPDR. As I said earlier, five of the six subjects chose to write footnotes immediately when they encountered a problem. This may have disturbed their usual flow.
With IPDR, in the translation products from Danish into German, I noted that the subjects, on several occasions, produced morphological and idiomatic errors immediately after the reporting. The reason for this can be that some of the subjects had problems when they switched back from parallel written reporting to translating. One subject, S3, mentions this problem - and she made this kind of error. The fact that subjects report in their mother tongue, Danish, while translating into the foreign language, in this case German, may also have some influence. However, this observation about error-creating influence from IPDR needs further investigation. Maybe more training with IPDR would have minimized this impact of the method.
In the products of the translations from Danish into German, the predominant difference can be noticed with respect to morphological errors (msyn) and incorrect choices of expressions from dictionaries and the internet (lex-sem). I assume that these different results can be explained, at least in part, by the differences between the experimental conditions, already mentioned in section 6.2. Whereas morphological problems and some of the lexical problems, which are regarded as elementary and perhaps embarrassing, with IPDR can be solved quietly and covertly at home - with Translog they cannot be hidden, and the R+Rp may have made the subjects feel stressed, so that they did not take the time to solve these problems properly.
6.4. Additional information provided by the methods
As the term IPDR "Integrated Problem and Decision Reporting" indicates, the method is an approach concerning "problems" and "decisions", and as Gile (2004: p2 and section 2.1.1a) points out, it is also a useful method for awareness raising and information collection about "translation strategies" and "various components of the translation process". In the experiments, using both methods, the subjects felt an urge to give explanations for various components of their processes. In order to categorize these comments, I used the coding procedures from Grounded theory (Hansen 2005a: 315f). The comments could be categorized as: intentions, attitudes, strategies, behaviour and control, in short 'iasbc-comments' (see Tables 1-4 in the appendix). I also added some of my spontaneous observations.
The application of R+Rp in both directions created a larger variety of such additional comments than IPDR. I did not count how many times subjects made the same kind of comment. This difference in variety is caused by the different experimental conditions. As can be seen in Table 2 and Table 4, some of the comments - they are marked with an asterisk * - seem to be animated by the dynamic process on the screen. These are, for example, explanations like "*Writes XXX for unknown words/larger units - inserts them in the revision phase" or "*Reformulates sentences several times". They are typically provoked by the replay.
Translog is a computer software program, developed by Jakobsen and Schou (1999). It provides the possibility to log keystrokes, revisions and pauses during the writing process. The process can then be replayed on the screen and reviewed on a log file.
6.4.1 Analogy and differences as to the comments
As mentioned, the translation tasks, the briefs for the translations and also for the retrospection were identical. The number and kinds of problems encountered generally depends largely on the character and difficulties of the source text in relation to the translation competence of the subject. That is why the kinds of problems, decisions and errors were identical with both methods used and why some overlap as to the 'iasbc-comments' could be expected. On Table 1+2 and on Table 3+4 (see appendix) the comments from IPDR and R+Rp of each of the subjects are shown. The overlapping comments are written in bold/italic.
The results show that - depending on the method - the subjects' focus was directed towards different aspects of the translation process: Whereas with R+Rp the subjects focused intensely on the process of producing and structuring the TT, with IPDR they focused more on information acquisition, especially when they comment on their use of dictionaries and the internet. Thus IPDR provided me with important information about source consultation that I did not obtain from R+Rp, whereas R+Rp showed a larger number of structural problems and decisions and a greater variety of information about other aspects of 'iasbc' (see Figures 1-4 and in the appendix tables 1-4). The investigation showed that it is not the absolute quantity of commented look-ups in dictionaries or on the internet that is important, but more the quality of the descriptions of the source consultation. The better quality of the IPDR comments on source consultation may also be due to the fact that with R+Rp subjects comment after the process is finished. Then they sometimes have forgotten what sources they have used or cannot recall exactly. With IPDR, if the comments are written immediately after a problem or task is solved, this kind of introspection according to Ericsson and Simon (1993: 19) should be even more precise than TA and give "the closest approximation to the actual memory structures". This is perhaps supported by these experiments: with IPDR the number of Google searches is mentioned 20 times (in total in both directions), whereas with R+Rp the exact number of Google hits is only remembered and mentioned once. An example of a typical comment with IPDR:
I was in doubt as to sauber and rein, because they give different associations. I tried both in relation to Umwelt on google.de and found that both can be used. There were many more hits with sauber (about 6000) than with rein (about 100). I chose the latter because it sounds best in a flyer and because I also found it in this connection (S5, my translation).
As to translator training, my conclusion is that best results about processes can be obtained if both methods are applied complementarily. This was confirmed when I had the delayed dialogue and the evaluation of the methods at class after the experiments (see section 7).
6.5. IPDR and R+Rp: Applicability, advantages and disadvantages for training and research
Both methods provide a means to raise students' awareness of what they are thinking and doing when they translate. Their application gives students and trainers valuable additional insights into the various components of the complex, individual translation processes, which complements the traditional preliminary textual analysis übersetzungsrelevante Textanalyse of source texts and reflections about appropriate translation strategies (e.g. Hönig 1995).
As to applicability, IPDR obviously has advantages. Gile (2004) characterises the method as a convenient type of retrospection. This may be the case for the observer/trainer as IPDR costs little in terms of information collection, but the method requires some effort from the subjects/students. R+Rp is easier for the subjects/students, but it is very time consuming for the observer/trainer, because the reports have to be transcribed or - for teaching - at least to be listened to carefully, and the log files and results have to be analysed.
A second advantage of IPDR is that the sample size is not restricted. As Gile (2004: section 184.108.40.206) points out, the method can be used with whole classes and without any special experiments. IPDR is also very flexible because the reports can be written anywhere and at any time - conversely, with R+Rp the replay always presupposes the establishment of a special experiment. Though Translog easily can be used at home, the replay function of the software cannot be established without the observer/trainer present.
A third advantage of IPDR is that the method provides the possibility to re-think and sometimes re-translate passages of the first draft. As Gile (2004: 2.4.3) points out, the writing about a segment often generates further reflection and results in an improvement of the TT. With R+Rp, it is not possible to re-translate because the translation process is finished before the replay starts.
It can be concluded that IPDR has many advantages over R+Rp. This is especially true of the everyday teaching situation, where it is impossible to do R+Rp with large groups because it requires too much time and effort from the trainer.
6.5.1. Naturalness of the translation processes
As mentioned, an advantage of both methods is that they can be used without reminders and cues. Nevertheless, the use of either method distorts translation processes when compared to natural processes without any kind of introspection.
On one hand, with IPDR the translation process is more natural than with R+Rp because many translators like to interrupt their translation process and to do something else - sometimes because they have a problem they have to think about, and sometimes just to let their translation rest for a while. For Danish students, it is quite usual that they visit libraries or contact experts (Hönig/Hansen 2000: 336). IPDR invites subjects to do this, whereas with R+Rp subjects are tied to a certain place and longer interruptions are inconvenient or even impossible because Translog, in principle, is based on the logging of time, keystrokes and cursor movements. On the other hand, it is quite unusual for 'real' translators to give written reports about their problems and decisions during or after a translation process - as they are asked to do with IPDR.
With R+Rp, on the one hand the translation process is more natural than with IPDR because it is not interrupted by the act of reporting - on the other hand, some of the subjects obviously become stressed in the experimental situation. This has also been shown in the control experiments in translations with and without R+Rp that I carried out in 2003 (Hansen: 2005a: 106).
When considering how natural the translation process is using the two methods, the conclusion must be that both methods are useful tools for gathering information about problems, decisions, intentions, attitudes, behaviour, strategies, and control during translation processes - but what is investigated is not the natural translation processes. The experimental situation and the application of introspection methods must have some impact on the subjects' thoughts and actions (Hansen 2005b).
6.5.2.IPDR and R+Rp as research tools in empirical research
The fact that it is not the natural process that is investigated with these retrospection methods is an important factor in empirical research. Gile (2004: p2) asserts that IPDR "holds some promise for local and multi-centre empirical studies into translation expertise acquisition", but he also points out (2.4.3) that its usefulness as a research tool may depend on the research project. He adds that IPDR does not provide comprehensive information because students cannot be relied on to do it thoroughly. As experiments with IPDR are usually carried out at home, they are not controlled. One could imagine that students or subjects tend to select what they want to mention and what they prefer to hide. As can be seen from the comparison with R+Rp with respect to reported problems and source consultation (Figs. 1-4) and also from the number of different comments as to 'iasbc' (Tables 1-4), on the whole IPDR seems to provide less information, or at least less varied comments, than R+Rp. The importance of this must be weighed against the fact that the method can be used on a larger scale and be combined with other methods (see section 7).
As Gile (2004: section 2.4.3) points out, "one aspect of the translation process which is particularly well covered with IPDR is ad hoc information acquisition". With respect to source consultations, IPDR provides more data - and also more precise and relevant data - than R+Rp. Here IPDR, if it is carried out during the process, is maybe even more precise than TA (see section 6.4.1).
As IPDR can be used on large samples without much effort from the observer, it is easy to gather a large, directly accessible corpus to work on. This may compensate for lack of comprehensiveness in the individual reports with respect to some aspects of the processes. Gile (2004: section 2.4.3) proposes that it might be an idea to introduce specific questions and/or instructions. It is also possible in the brief for the retrospection to focus on students' reactions to selected problems or aspects of the process.
In contrast to IPDR, the strength of R+Rp is that it is carried out in a controlled setting and that the process is double-checked by use of the log-file. With R+Rp, subjects automatically feel animated to take a stand on their problems - or better cursor movements, revisions and pauses - because they can be seen on the screen. During the retrospection, subjects cannot really ensure that their comments accurately reflect what they did during the process. This can be seen on the log-files. My observation is that serious problems and SD’s in translation processes can often be discovered via discrepancies between what subjects say during the retrospection - and their real actions, which are logged. The log files also complement and supplement the observations from the retrospection and the translation products by providing data about time management, pauses, segmentation and revisions. For the qualitative interpretation of the data from retrospection by the observer, the combination with the objective, quantitative data from a log file is an important means to reduce bias.
7. Method combinations with ID and DD
As mentioned, in my earlier experiments (Hansen 2003a, 2003b, 2005a), I always used Translog with R+Rp+ID, which means that the retrospection was carried out under my observation and was followed by an immediate dialogue with the subject and individual feedback.
The experiments in this study with IPDR and R+Rp were combined with 1) a delayed dialogue with the subjects, 2) an individual feedback and 3) a discussion of the methods and results in class. None of this could take place until all the translation products had been evaluated and the reports and tapes had been analysed, which means that at least two weeks elapsed before I could meet the subjects again.
The dialogue is a very useful part of experiments in empirical research - especially if the objective is to create awareness and to improve translation processes. With the Copenhagen Retrospection Project, I now can compare the combination of R+Rp with the two kinds of dialogue, R+Rp+DD in this study and R+Rp+ID from my earlier project. Additionally, I can look at the method combination IPDR+DD, where the IPDR is the starting point of the dialogue, because I have tried to use exactly the same procedure and kind of dialogue after the processes with IPDR as after those with R+Rp.
With the immediate dialogue, as I have used it, the observer follows the translation process on the screen - from the subject's first look at the translation brief until the final product is completed - and listens simultaneously to the retrospection. Combined with previous knowledge from questionnaires providing personal profiles on the subjects' linguistic background, the observer can gather a large amount of relevant data and signals. The combination of data from profiles, processes and products then makes it possible, in the dialogue, to peel off one layer of a problem after the other until the observer and the subject are able to encapsulate the fundamental reason for a problem, or until the subject discovers and can explain the causes of a problem him/herself. A precondition for this kind of "here and now research" is that the observer, already during the process, analyses and combines data and results from the personal profiles, subject's comments - in interaction with cursor movements, revisions and the final products - with a view to planning the immediate dialogue.
In the actual study with IPDR and R+Rp, I tried to transfer the same procedure on the application of the delayed dialogue. I also gathered all available data about the personal profiles and combined them with the reported problems (kinds of problems), decisions, comments as to use of dictionaries, internet and other sources of information, errors in the products (kinds of errors), and any further comments the subjects had made as to their 'iasbc' (see Tables 1-4).
It showed that both, processes with IPDR and processes with R+Rp can be equally suitable for a starting point for a dialogue. However, the delayed dialogue showed to be laborious with both retrospection methods. The problem was that some weeks after having carried out the experiments, the subjects did not remember their translation processes, products, or comments on problems and decisions. I had to work hard to help them remember. The conclusion must be that irrespective of the retrospection method it is easiest to analyse and use the results as quickly as possible.
Though the results of a delayed dialogue seem to be poorer than the results when I used the immediate dialogue, I will give some examples to show that the delayed dialogue nevertheless can be a useful tool in order to gain additional information when using either of the retrospection methods, IPDR or R+Rp.
My attitude when gathering data from retrospection was that every single report and comment from both methods could prove to be important. Some comments or combinations of comments proved to be signals that something could be wrong or needed explanation. In my experiments, such signals were for example:
In accordance with the advantages of each of the methods (see section 6.5), in some cases IPDR and in other cases R+Rp displayed advantages as the basis for the delayed dialogue. Sometimes both were equally suitable as the starting point for the dialogue.
Example 1 (advantage IPDR)
Having analysed the IPDR of S2 for the translations into both directions, I was concerned primarily by her heavy use of the internet. On the one hand she obviously based many of her decisions solely upon the quantity of google hits, and on the other hand, she used the internet in a well-considered manner, as can be seen from the following example, where she is able to abstract from the google hits:
I chose to mention the names of the artists. There are not many hits on google.de (72 hits) concerning P. S. Krøyer, but I think as Skagen is famous also because of them, they must be mentioned. (S2, my translation)
The delayed dialogue with S2 showed that she used the internet in order to get ideas. She said that she always feels trapped by the expressions of the source text, that she has no alternative ideas in the target language, and that she has been trained to observe strict fidelity to the source text. She admitted that she uses the internet as a super-dictionary, an observation Gile (2004: section 3) also made with some of his students.
The dialogue showed that she was an expert with respect to the internet and that there was no reason to be concerned about her heavy use of that tool. However, it also emerged that she has a general problem, i.e. that she suffers from a lack of ideas in the target language when she reads the source text.
Example 2 (both equal)
Inconsistent comments are nearly always an important signal that something may be wrong. That is why I had a closer look at the processes and products of S3. Several of my observations from both IPDR and R+Rp were contradictory (Tables 1-4). Sometimes S3 shows a great deal of self-confidence - then again she is in doubt. She seems to be good at taking decisions but, on the other hand, she is insecure. The delayed dialogue showed that she was quite competent but that she suffered from lack of self-confidence.
With IPDR I also found an attitude in the process of S3 which seems to be a SD. Some translators develop preferences for special idioms or lexical units, and these preferences are so strong, that they totally ignore all signals that the words may not fit in the context. They just use them whenever there is the opportunity. She says for example: "I have chosen to translate gelten with holde vand, an expression I like very much.” This expression does not fit stylistically. She also says: "I have to mention that "zwar" is one of my favourite words. It also always gives me much pleasure to use the expression vel at mærke.
I had observed this phenomenon several times in my earlier experiments with R+Rp+ID, in connection with reverbalization, where it also proved to be a cause of errors (Hansen 2005a: 334).
Example 3 (both equal)
The reports of S6 with IPDR and with R+Rp showed heavy use of dictionaries. This can be noticed with both methods. The subject is quite competent without the heavy use of dictionaries - a phenomenon I observed in earlier experiments (Hansen 2003a/2005a). This excessive use of dictionaries can have different causes. Sometimes these causes can be discovered with experiments where subjects also are asked to translate under time pressure, and where their spontaneous translations without aids can be compared with their translations with aids (Hansen 2005a).
Example 4 (advantage R+Rp)
S5 used a lot of time for her translations with R+Rp. Data from the log files of all subjects show that for the translation into German they, on average, spent 47 minutes (some spent 31 minutes - S5 spent 75 minutes). The keystroke average was 45 per minute (the fastest subjects had 65 - S5 had 27).
Like S6, S5 showed heavy use of dictionaries - in both directions and with both methods. In her IPDR-reports she writes that she finds inspiration in dictionaries and that she needs security checks. With IPDR (Danish-German), she writes that she looks up in several dictionaries to solve one problem.
With R+Rp (German-Danish) she says that she guesses. With R+Rp (Danish-German) the log file shows that she took many pauses (49) longer than 20 seconds. She used them for at least 19 series of look-ups in several dictionaries. She used the internet additionally for checking collocations.
With R+Rp, on the log file, it can be seen that she reformulates many times and she comments on her many reformulations of the same correct sentence into several new correct versions. She obviously cannot decide on one of them and, finally she goes back to her first solution. My first impression was that she suffered from lack of linguistic competence. With R+Rp, I observed that she often was forced to look up quite simple everyday words – words she actually should have known - and that she even with respect to these could not decide.
During the delayed dialogue she explained that she spends lots of time for her translation tasks and that she is generally insecure and cannot take any decisions. As her translation products were quite good - translation tasks with time pressure perhaps could also help her.
8. Evaluation from the subjects
So far, I have done experiments with IPDR in translation teaching with about 50 Danish students. They said that they regarded the method as very useful. In 2004, I used IPDR with equal success in intercultural communication, a kind of technical writing or free composition of texts in the foreign language. A few of the subjects said that they did not like their translation process being disturbed by writing the reports. Some also mentioned that it took them too much time, and that they forgot what they were thinking when they did the reporting parallel to problem solving during the process.
I had never tried R+Rp without my being there during the replay before these experiments and I have not used it since. The six subjects who participated in these experiments with both methods said that they preferred R+Rp because it was easier for them to use it. They expressed that they liked to see their processes on the screen. Some of them said that they had felt stressed and that talking had been a bit awkward.
IPDR and R+Rp are two complementary retrospection methods - where one is particularly useful for showing information acquisition and the other particularly efficient with respect to time management, structural problems and revisions. Both can be combined with each other and with other methods, for example with questionnaires, clarifying dialogues and/or experiments with time pressure where subjects are asked to translate spontaneously and without the use of aids.
Depending on the focus of interest or research issue, many different kinds of designs can be created for both teaching and research. If necessary, it is also possible to create special designs to fit specific individual needs.
9.1. Conclusion as to detecting, eradicating and preventing SDs
The results corroborate the assumption that IPDR and R+Rp provide different kind of information. There is however some overlap, see appendix, Tables 1-4.
In order to be effective, both methods have to be combined with each other and/ or with other methods.
IPDR seems to have advantages in terms of finding some kinds of SD’s, especially in connection with source consultation which is covered more exhaustively with IPDR than with R+Rp.
R+Rp has advantages with respect to detecting SD’s that can only be discovered via the combination of replay and log-file, i.e. in relation to structural problems, reception problems, inappropriate changes and revisions. The log-files provide an excellent tool for revealing discrepancies between what subjects think they do and what they really do.
As to detecting SD’s, the methods complement each other. They seem to be most effective if they are combined with an immediate dialogue and individual feedback.
Following one’s own process with replay on the screen and discussing it immediately after the replay, i.e. as long as many thoughts with respect to problems and decisions are still present, tends to produce a strong effect on subjects. Therefore, R+Rp proved to be very effective for eradicating SD’s when the method was combined with the immediate dialogue (Hansen 2005a).
If IPDR is used in combination with such an immediate dialogue, i.e. shortly after the process, it may have a similar eradicating effect. This would have to be tried in practice.
Both methods help to raise subjects' awareness of what they do when they translate, and both seem to have a comparable influence on keeping the translator on track - as Gile (2004) expresses it. He assumes that "a set of norms and strategies that have not only been practised, but also thought and written about, are more resistant to attrition over time", and that "the awareness of the norms gained by the students may make it easier for them to remain on the track". As the application of IPDR is much easier for the teacher and as it can be used on a large scale, this method may be better at preventing SD’s.
The influence of R+Rp is due to an immediate effect of impression and fascination. All subjects who have done retrospection in my experiments liked following their own processes on the screen, talking about it and receiving individual feedback.
For preventing SD’s and for guaranteeing a long-term effect, the dialogue and feedback seem to be crucial with both methods.
Bohr, N. (1959/64). Atomfysik og menneskelig erkendelse I und II. København: J. H. Schultz.
Denzin, N.K. & Lincoln, Y.S. (eds) (1994). Handbook of Qualitative Research . Thousand
Oaks, CA: Sage.
Duden (1996). Duden Deutsches Universalwörterbuch. Mannheim: Dudenverlag.
Ericsson, K.A. & Simon, H.A. (1984, 1993 2). Protocol Analysis: Verbal Reports as Data.
Cambridge, Mass.: MIT Press.
Gile, D. (2004). Integrated problem and decision reporting as a translator
training tool. The Journal of Specialised Translation 2: 2-20. < http://www.jo s t r ans.org >
Hansen, G. (1995a/2001 3). Einführung in das Übersetzen. Copenhagen: Copenhagen
Business School Press. Munksgaard.
Hansen, G. (1997). Success in translation. Perspectives. Studies in Translatology 5.2:
Hansen, G. (1999). Das kritische Bewusstsein beim Übersetzen. Copenhagen Studies in
Language 24. 43-67.
Hansen, G. (2002). Zeit und Qualität im Übersetzungsprozess. Copenhagen Studies in
Language 27. 29-54.
Hansen, G. (2003a). Controlling the process. Theoretical and methodological reflections
on research in translation processes. In Alves, F. (ed.) Triangulating Translation.
Amsterdam: Benjamins. 25-42.
Hansen, G. (2003b). Interferenz bei Referenz im Übersetzungsprozess. In van Vaerenbergh, L. (ed.) Linguistics and Translation Studies. Translation Studies and Linguistics. Linguistica Antverpiensa 1/2002. 303-326.
Hansen, G. (2005a). Störquellen in Übersetzungsprozessen. Copenhagen: CBS, Samfundslitteratur.
Hansen, G. (2005b). Experience and emotion in empirical translation research with think-aloud and retrospection. In Lee-Jahnke, H. (ed.) Processes and Pathways in Translation and Interpretation . META 50/ 2. 511-521.
Hansen, G. & Hönig, H.G. (2000). Kabine oder Bibliothek? Überlegungen zur Entwicklung eines interinstitutionell anwendbaren Versuchsdesigns zur Erforschung der mentalen Prozesse beim Übersetzen. In Kadric, M., Kaindl, K. & Pöchhacker, F. (eds) Translationswissenschaft. Festschrift für Mary Snell-Hornby zum 60. Geburtstag. Tübingen: Stauffenburg. 319-338.
Hönig, H.G. (1995/1997). Konstruktives Übersetzen. Tübingen: Stauffenburg.
Jääskeläinen, R. (1999). Tapping the Process: An Explorative Study of the Cognitive and
Affective Factors Involved in Translating. Joensuu: University of Joensuu.
Jakobsen, A.L. & Schou, L. (1999). Translog documentation. Copenhagen Studies in
Language 24. 149-184.
Krings, H. P. (1986). Was in den Köpfen von Übersetzern vorgeht. Tübingen: Narr.
LEO. (2005). Link Everything Online. English-German Dictionary . Online-Service der
Informatik der Technischen Universität München.
Moustgaard, I.K. (1981). Beskrivelse og kommunikation. Kapitler af beskrivelsens
psykologi. København: Munksgaard.
Moustgaard, I.K. (1990). Psychological Observation and Description. Bergen: Sigma.
Strauss, A. L. & Corbin, J. (1998). Basics of Qualitative Research: Techniques and
Procedures for Developing Grounded Theory . London: Sage.
Tranekjær Rasmussen. E. (1955/1967 2). Virkelighed og Beskrivelse. In Nordisk Sommeruniversitet . København: Munksgaard und Psykologiske Tekster 6. København: Akademisk Forlag. 14-60.
Translation briefs, source texts and some proposals for target texts
Auftrag für 1a und 1b:
Das Touristenbüro von Skagen möchte die Broschüre Lyset over Skagen auf Deutsch neu drucken. Sie werden gebeten, den folgenden Abschnitt zu übersetzen:
Auftrag für 2a:
Die dänische populärwissenschaftliche Zeitschrift Psykolog Nyt gibt eine Artikelserie über Arbeitslosigkeit in Europa heraus. In diesem Zusammenhang werden Sie gebeten, den folgenden Artikel aus Stern 51/2003 zu übersetzen, den man als Hintergrundmaterial benutzen will. Übersetzen Sie folgenden Abschnitt:
Auftrag für 2b:
Sie werden gebeten, im Rahmen der Artikelserie auch einen Artikel aus der Zeitschrift Psychologie Heute, März 2004 , für die Leser von Psykolog Nyt zu übersetzen. Bitte übersetzen Sie folgenden Ausschnitt des Textes:
Subjects' comments as to 'iabsc' with IPDR and R+Rp in both directions and some observations
Table 1: IPDR German - Danish
Table 2: R+Rp German - Danish
Table 3: IPDR Danish-German
Table 4: R+Rp Danish - German
i. Translog is a computer software program, developed by Jakobsen and Schou (1999). It provides the possibility to log keystrokes, revisions and pauses during the writing process. The process can then be replayed on the screen and reviewed on a log file.The subjects only used these two sources of information from outside.
© 2006 JoSTrans