Keywords

1 Literal Translation as a Translator’s Basic Procedure

Many findings suggest that replacing words of one language with those of another without more complex text analysis is the predominant strategy of individuals with little experience in translation, as I already argued in my previous paper (Płońska 2014). For instance, Lörscher notes that “most of the foreign language students … produce translations mainly by an exchange of language signs” (Lörscher 2005, p. 605). Königs and Kauffmann observe that translation procedures of foreign language students are vocabulary-centered and their mental activity is focused mostly on the vocabulary to the detriment of the grammar (Königs and Kauffmann 1996, pp. 18–19). Tirkkonen-Condit remarks that “novices tend to approach a translation task as a series of lexical or phrasal problems that are to be solved in the order in which they appear in the text. In novices’ performance, translation tends to proceed word by word, phrase by phrase, sentence by sentence” (Tirkkonen-Condit 2005, p. 408).

On the other hand, research carried out by Tirkkonen-Condit (2005) showed that a tendency to translate literally occurred in both beginner and experienced translators. This was visible not only in the translation process but also in the finished translations. The author claims that literal translation is the default procedure used by a person translating a text until that person notices a problem with the text of the translation. This finding is in line with the theoretical considerations of Newmark, who believes literal translation to be the basic translation procedure (Newmark 1988, p. 70). The tendency to apply literal translation as a default procedure was also noted by Mandelblit (1996). In her psycholinguistic experiment bilinguals were asked to translate idioms from French into English and vice versa, each participant translating into their mother tongue. According to the researcher’s hypothesis, the idioms with a different cognitive mapping in the target language would be more difficult and thus take more time to translate. For instance, the French expression “trouver le temps long” [lit. “to find the time long”; this and further English translations and annotations in square brackets are my own], which can be translated to English as “time is passing slowly”, would be more difficult to translate than the expression “perdre du temps”, which is a literal equivalent of the English idiom “to waste time”. In the first case, French uses the “time as space” metaphor, while English uses the “time is a moving object” metaphor. In the second case both French and English make use of the “time is a valuable object” metaphor. The results confirmed the author’s hypothesis but also showed that “when translating DMC [different mapping condition] sentences, subjects tended to first suggest a word-to-word (and “same mapping”) translation for the source sentence and only later propose the better translation” (Mandelblit 1996, p. 493).

At the same time, comprehension strategies of professional and non-professional translators seem to differ. Tirkkonen-Condit (2005) notes that beginner translators and amateurs focus on lexical units and seek information in external translation aids, while experts concentrate on the text itself, its semantic, pragmatic and inter-textual aspects, trying to extract as much information as possible. In other words, the comprehension strategies of amateurs have a local orientation, while those of experts are global.

These findings are in consonance with those of Jääskeläinen (1996). The researcher discovered that the authors of mediocre and poor translations rely more on linguistic knowledge, while the authors of good translations tend to apply world knowledge. According to her, “in the good processes most of the attention is directed at text comprehension, at relating the text to the extra-textual world. The less successful processes seem to remain more exclusively at the linguistic surface level” (Jääskeläinen 1996, p. 69). Similarly, analyzing translation processes of foreign language students, Königs and Kauffmann note “l’énorme restriction de l’activité de contextualisation qui ne s’effectue qu’au niveau de la phrase, voire même du syntagme” [“a huge restriction of contextualization activity which occurs only on the level of the phrase, or even of the syntagm”] (Königs and Kauffmann 1996, p. 19).

Accordingly, the mental representation of the text being translated seems to have significant importance for the translation process. In this chapter I investigate whether forming a full mental representation of the source text before taking up the task will influence participants’ translation behavior. In particular, I want to find out whether non-professional translators and translation students would produce less literal translations if they had a mental representation of the source text prior to translating. My research is based on the text comprehension model proposed by van Dijk and Kintsch (1983). This model distinguishes three main levels of text representation: the superficial level of words and syntax, the text base level consisting of propositions, and the situational model level presenting the situation described in the text. The main objective of the comprehension process is to develop an accurate situational model.

In contrast with the studies mentioned above, based on TAPs, my research employs keystroke log data from Translog (Carl 2012).

2 Literal Translation: Problems of Operationalization

One of the main challenges in this field of research is the lack of a single commonly accepted definition of literal translation (see Carl and Schaeffer n.d.; Chesterman 2011). For the purposes of translation process research, Carl and Schaeffer (n.d.) propose three definitional criteria for an ideal literal translation. According to them, a translation is literal if the word order is identical in source and target texts, if source and target text items correspond one-to-one and if each source text word has only one possible translated form in a given context. This last criterion is operationalised in terms of translation entropy (see Chap. 2, Sect. 2.4.7). Using these criteria it is possible to measure how literal a translation is.

As the present study is a part of a larger research project concerning other more complex translation strategies too, I needed a definition which would allow me to identify precisely the passages translated literally. Carl’s and Schaeffer’s criteria for an ideal literal translation seemed too narrow to cover all the instances of what I intuitively identified as literal translation. That is the reason why I used a different definition of literal translation, formulated by Kielar (2013, p. 51). Kielar’s definition is the one I refer to further in the text every time I talk about literal translation. According to this definition, in literal translation, the rules of the syntax of the target language are used to combine the words calqued from the source language as separate lexical units. This definition does not presuppose that the word order should be preserved in translation. In fact, French and Polish differ substantially in terms of word order. As noted by Gniadek (1979, p. 131–132), “en français l’ordre des éléments est fixé depuis la disparition de la flexion nominale, tandis qu’en polonais l’ordre des éléments est plus libre, parce que la forme du nom indique sa fonction dans la phrase” [“in French the order of items is fixed since the disappearance of the nominal inflexion, while in Polish the order of items is freer because the form of the noun indicates its function in the sentence”]. Given these differences, in the present study I decided not to apply the identical word order as one of the criteria of literality of translation. However, I wanted to verify if the notion of literality operationalized according to Kielar’s definition correlates with the notion of entropy proposed by Carl and Schaeffer. The value of entropy indicates how many different translations a given source text word has. If a word has only one possible translation, it has an entropy value of 0. I admit the possibility of a source text word having more than one literal equivalent in the target language. This intuition is based on my previous experience. At the same time I suppose that the number of literal equivalents of a given word is limited. Therefore, it seems to me that there should be a strong relationship between the literality of translation of a given word as defined by Kielar (2013), and translation entropy as defined by Carl and Schaeffer (n.d.). The entropy values should be significantly smaller for the words translated literally according to Kielar’s definition.

The study was aimed at ascertaining whether professional translators, non-professional translators and translation students differ in terms of the literality of the translations they produce. Firstly, I expected that the tendency to translate literally would decrease with experience, i.e. professionals would produce less literal translations than students and students less literal ones than non-professionals. Secondly, I investigated the impact of constructing an initial mental representation of the source text on the literality of produced translations. I hypothesized that participants who did form a mental representation of the text prior to translating would translate less text literally than those who did not. Thirdly, I analyzed the relationship between literality as defined by Kielar (2013) and entropy as defined by Carl and Schaeffer (n.d.). As I already stated above, my assumption was that the entropy values would be significantly smaller for the words translated literally according to Kielar’s definition.

3 Method

3.1 Participants

The study involved 19 professional French to Polish translators aged 28 to 61, 20 students of applied linguistics with French language aged 22 to 34 and 20 persons with advanced-level French language skills and without a background in translation aged 25 to 54. Further in the text I refer to these groups by the terms “professionals”, “students” and “non-professionals”. The professionals’ work experience ranged from 5 to 38 years at the time of the study. Among the non-professionals, 9 persons had DALF certificate (Diplôme approfondi de langue française) confirming their advanced knowledge of French. The other 10 persons were teachers of French in upper secondary schools (Polish: liceum) in Warsaw and one person was a teacher of French in a primary school (Polish: szkoła podstawowa) at the time of the participation in this research.

3.2 Materials and Procedure

The study was conducted on an individual basis. The task was to translate from French to Polish. The participants were assigned to translate an article for a magazine covering European issues, and were asked to prepare a text ready for publication without any need for further corrections. The task was preceded by brief technical instructions regarding the software. The process of translation was recorded using Translog. The participants had access to hard copies of a French-Polish dictionary and a monolingual French dictionary and their behavior during translation was filmed. No online dictionaries were put to use. After the task was finished I interviewed the participants about the completed task.

The article had been written for the purpose of the study by a French journalist having good command of Polish. In view of the study’s objective, it was deemed important for the text to have a narrative structure and be easy to understand but nonetheless present some problems in translation: idioms, metaphorical expressions, “false friends” etc. The length of 365 words was specified so that the text was long enough to reveal some regularities in participants’ behavior but not too long due to time constraints.

To investigate the role of forming an initial representation of the text, the participants were randomly assigned to two groups. In the experimental group, before taking up the translation task, the participants were given the following instruction (in Polish): “Please read the following text very carefully. In a moment you will be asked to answer some questions about its content and form”. After having read the French text, without being able to refer back to it, the participants filled in a questionnaire with a sentence recognition test and instructions to write a summary in Polish. The sentence recognition test contained four types of samples: literal samples from the text, paraphrases, correct conclusions, i.e. sentences which were not in the text but which are consistent with the text meaning, and the incorrect conclusions, i.e. sentences which are not consistent with the text meaning. In the control group the participants did not read the text and filled in a shorter version of the questionnaire regarding their personal information only.

The study was conducted on the premises of the University of Social Sciences and Humanities, in several upper secondary schools and at the participants’ homes. The time for each task was not limited.

3.3 Data Analysis

The dataset was added to the CRITT Translation Process Research Database (see Chap. 2). The translations were manually aligned using the YAWAT tool (Germann 2008). For the purposes of the analysis, the data were subsequently processed into a set of tables.

In line with Kielar’s definition of literal translation, for every passage of the French text I tried to imagine all the possible literal translations by using dictionary equivalents of French words and by connecting them according to Polish syntax rules. I used “The Great French-Polish Dictionary” (Dobrzyński et al. 1996) and “The Great Polish-French Dictionary” (Frosztęga 1995–2008) as a reference material. I took into consideration all the possibilities of word order available in Polish syntax. It does not mean that I physically made an exhaustive list of the possible literal translations for every sentence of the French text. Such a list would be very long for two reasons. Firstly, because according to the dictionary most of the French words used in the text have more than one Polish equivalent. Secondly, because in Polish the word order is freer than in French. I don’t think having a list of all the possible literal translations of all the sentences of the text would be necessary for the purpose of the subsequent analysis. Therefore, I looked for all the dictionary equivalents of the French words used in a given passage, and imagined how they could be connected according to Polish syntax rules to envisage how the word order could be changed in the sentences thus generated.

Afterwards, I compared the actual translations with the imaginary literal ones. I marked all the passages that matched literal translations in terms of word form as translated literally. It means that for every translation all of the source text words were labeled as translated literally or non-literally. An example of the labeling is provided in Table 13.1. The abbreviation “lit.” stands for “literal” while the abbreviation “non-lit.” stands for “non-literal”. The first column presents the original passage as well as its English literal translation (based on the dictionary entries). The next three columns present three of the possible literal translations I imagined: one with the same word order as in the original text (Imaginary literal translation 1) and two with different word order (Imaginary literal translations 2 and 3). The next two columns present two translations provided by participants. These translations are only partially literal. In the translation by Participant 1, lexical changes in rows 4 and 6 result in changes in form of the words in rows 5, 7 and 8. Accordingly, all the words in rows 4–8 are labeled as translated non-literally. In the translation by Participant 2, lexical change in row 5 results in changes in form of the words in row 7.

Table 13.1 Labeling of different versions of Polish translation of a French passage
Full size table

Articles, subject pronouns and possessive adjectives omitted in translation in accordance with the rules of Polish grammar, as well as the French prepositions “de” and “à” in the phrases translated as nominal inflections or adjectives, were considered together with the following words and labeled accordingly. The annotation was blind, i.e. I did not know who had produced the translations. Initially, I also adopted the procedure of back-translation described by Ivir (1997) in order to ensure that the target text words identified as such were indeed literal equivalents of the source text words. However, originally, this method was used to check the semantic content of translation segments of at least two words. When I tried to apply the method to separate words it turned out to be unavailing, because all the words appearing in the French-Polish dictionary as the equivalents of a given French word could be translated back by the means of the same French word using Polish-French dictionary.

3.4 Results

The total number of source text words translated literally was used as a measure of literality of translation. This variable was examined with a 3 × 2 (Experience [non-professionals, students, professionals] × Initial text representation [yes, no]) analysis of variance (ANOVA). The results showed a statistically significant effect of the main variable experience, F(2, 53) = 6.32, p < 0.01, η2 = 0.19 (see Fig. 13.1). The students translated significantly less text literally (M = 162.55; SD = 27.46) than non-professionals (M = 193.55; SD = 33.40).

Fig. 13.1
figure 1

Average total number of words translated literally depending on the experience

Full size image

Furthermore, the effect of the interaction between the variables Experience and Initial text representation was statistically significant, F(2, 53) = 5.78, p < 0.01, η2 = 0.18 (see Fig. 13.2). Simple effects analysis showed that students translated significantly less text literally than both professionals and non-professionals but only when the participants did not form a mental representation of the text prior to commencing work. Creating an initial representation of the text significantly reduced the amount of text translated literally among professionals and significantly increased this amount among students.

Fig. 13.2
figure 2

Average total number of words translated literally depending on the experience and depending on whether an initial representation of the source text had been formed

Full size image

I used the total number of source text words translated literally as a measure of literality of the whole translation. However, to investigate the relationship between the literality of translation of a given word and the entropy of translation alternatives I used a nominal variable literality concerning separate words. As I already stated above, for every translation, all of the source text words were labeled as translated literally or non-literally. The entropy values were also calculated for every source text word. In order to verify whether the nominal variable literality concerning separate words can be a good predictor variable of the entropy of translation alternatives, a one-way ANOVA was performed with literality as a factor. The effect of this variable was statistically significant, F (1, 22772) = 7470.62, p < 0.001, η2 = 0.25. The entropy values were smaller in the case of source text words labeled as translated literally (see Fig. 13.3).

Fig. 13.3
figure 3

Entropy of translation alternatives depending on whether the word was translated literally

Full size image

Another illustration of this relationship is provided by Fig. 13.4 presenting the number of occurrences of literal and non-literal translation depending on the entropy values.

Fig. 13.4
figure 4

Number of occurrences of literal and non-literal translation depending on the entropy of translation alternatives

Full size image

3.5 Discussion

The results of the study are unanticipated. As it turns out, the students translated less text literally than professional translators. Moreover, there are no significant differences in the amount of text translated literally between professionals and non-professionals. The results also show that an initial mental representation of the source text has a substantial impact on the subsequent translation process in terms of the frequency of words translated literally. According to expectation, having an initial representation of the text made the differences between groups less substantial. However, the influence of this variable is different for the three groups. The students who did form an initial mental representation of the source text translated literally more text than those who did not. In contrast, the professionals who did construct a representation of the source text before taking up the task made less frequent use of literal translation than those who did not.

The results concerning the entropy of translation alternatives conform to my preliminary expectations. For the words labeled as translated literally according to Kielar’s definition the entropy values are significantly lower. It means that the number of translations proposed by participants is significantly smaller in the case of words translated literally. This finding confirms the assumption that a word of one language has a limited number of literal equivalents in another language. It also shows that, to a certain extent, my operationalization of literal translation is compatible with Carl and Schaeffer’s conception described in Chaps. 2 and 9. The entropy, one of the criteria used by Carl and Schaeffer to measure the literality of translations, is a continuous variable and the literality operationalized according to Kielar’s definition is a nominal one. Nevertheless, considering the entropy criterion alone, the passages labeled as translated literally were significantly closer to an ideal literal translation as defined by Carl and Schaeffer than the passages marked as translated non-literally.

In my view students’ reluctance to translate literally might have been the result of translators training. Students learn at a very early stage that the use of literal translation often results in translation errors and can be regarded as a sign of incompetence. This is the probable reason why they perceive this procedure as their last resort and try to avoid it by all means. As their experience grows, they learn to recognize the situations permitting the safe use of literal translation. In contrast, experienced translators can consciously use literal translation allowing them to provide translations that are both acceptable to target norms and adequate to the source. It is even easier to apprehend given that the use of this procedure requires less time and effort (cf. Schaeffer and Carl 2014). As the differences in the use of literal translation between professional and non-professional translators weren’t statistically significant, it would be compelling to assess the quality of the translations and to examine whether the use of this procedure was associated with specific translation errors in any of these groups.

The study confirms the importance of forming a mental representation of text before commencing translation. In light of the results, I believe that having a mental representation of the source text before taking up the task allows the translator to be freer in their choice of available translation procedures, including literal translation. This would help in explaining why, contrary to my preliminary expectation, the students who did form an initial representation of the text translated literally more text than those who did not.

The effect of experimental condition could also be interpreted in terms of a possible pre-translation during the initial reading and a probable priming effect (see Chaps. 9 and 10, this volume). Schaeffer et al. argue that reading for translation is substantially different from monolingual reading. However, when the participants were presented the original text for the first time, they were told the objective of the reading was to be able to answer the questions about the text form and content. The participants knew they were going to translate a text during the study. They might have supposed the text they were reading to be the one they would translate later. Nevertheless, they weren’t explicitly told so. Certainly, some of the words of the original text were translated during writing the summary. On the other hand, the text was too long to be memorized, so at this stage the participants were writing a new text based on their recollection of the original text content rather than translating.

As far as the priming effect is concerned, the participants were presented not only the original text but also the sentence recognition test with different kind of samples, including paraphrases, correct conclusions and incorrect conclusions. They also wrote a Polish summary of the original text. As noted by Schaeffer et al. (see Chap. 9, this volume), in translation priming studies the priming from L1 to L2 was observed more often than priming from L2 to L1. Thus, it is legitimate to suppose that Polish words and syntactic constructions the participants used in their own summaries of the original text were more accessible to them during the subsequent translation. It might be an interesting concept for the future to examine the summaries written by the three groups of participants and to see how close they are both to the original text and to the translation text. An analysis of the time data from Translog could also shed a light on the role of a pre-translation and of a possible priming effect in the ulterior translation process.

A more complete picture of the translation process may be gained by researching the ways the three groups of participants apply more complex translation strategies. As a part of the current project, I also plan to take a closer look at the participants’ errors by analyzing the entire process of making corrections.