You are on page 1of 19

Methodological questions about translation research

A model to underpin research into the mental processes of translation*


Stuart Campbell and Berta Wakim
University of Western Sydney

Research on the mental processes of written translation has beneted less from the ndings of psycholinguistic and cognitive research than interpreting research has. This has left translation research relatively impoverished in the theoretical grounding of research methodology. While the performance speed of interpreting seems to make interpreting research closer to bilingual speech research, the shared features of interpreting and translation suggest that they are points on a continuum rather than discrete production modes. A Translation-Interpreting Continuum is proposed, which allows various production modes to be described in terms of psycholinguistic notions, and which includes a language development dimension to account for second language production. The Continuum allows some very basic questions about translation research methodology to be answered, and opens up the opportunity for a better integration of translation and interpreting research. Keywords: translation, interpreting, research methodology, psycholinguistics, language acquisition

Introduction

. A knowledge gap in translation research While empirical investigations of written translation (e.g. Krings 1986, 1987, Dechert 1987, Gerlo 1987) have made inroads into our understanding of the mental processes underpinning this complex phenomenon, written translation has attracted relatively little interest from researchers in the psycholinguistic and cognitive domains of bilingualism. On the other hand research into oral
Target 19:1 (2007), 9. issn 09241884 / e-issn 15699986 John Benjamins Publishing Company

Stuart Campbell and Berta Wakim

translation, or interpreting, has interacted much more with research into these elds to the extent that leading authorities on bilingualism are now beginning to discuss simultaneous interpreting (e.g. de Bot 2000, Paradis 2000, MacWhinney 2005) as a topic of interest, and there are even neurolinguistic accounts (e.g. Paradis 1994). Research into interpreting utilises and adapts the ndings of psycholinguistic and cognitive studies (e.g. Gile 1995, de Groot 1997, 2000, Christoels and de Groot 2005, Christoels el al. 2006) most likely because of the anity between interpreting and bilingual speech production, particularly in speed of delivery; authorities such as Gile (1995, 1997) and Christoels and de Groot (2005) cite the relative slowness of written translation as the reason to focus on interpreting rather than translation: [I]nterpreters work at speech delivery speed, whereas [t]ranslators generally have hours, days, or even weeks to complete the operations (Gile 1995: 111112). The prospect for collaboration between interpreting and translation research is however beginning to be recognised, for example in Schner (2004) where a number of (mainly interpreting) researchers chart some of the common ground. Despite a recognition in Translation Studies of the need to frame empirical translation research rigorously (see, for example, Toury 1995: 222240), research into the mental processes of written translation is hampered methodologically because of insucient theoretical underpinning, especially in research dealing with professional translation tasks. This deciency makes it dicult to frame research questions and to design robust studies. Interpreting research on the other hand has healthy debates about experimental design (e.g. Shlesinger 2000: 10). In this paper we aim to describe this methodological problem and to pose some simple but as yet unanswered methodological questions. We then present a novel approach to dening data types based on ideas drawn from psycholinguistic and cognitive research into bilingualism and second language acquisition, and conclude by attempting to answer the methodological questions posed earlier. .2 Methodological problems in translation research The methodological problem mentioned above arises in experiments dealing with the translation of whole texts. These experiments try to emulate professional contexts, in an attempt to throw light on the development of professional competence. The authors research team encountered a quandary in the design of an experiment into the way translators negotiate information prominence between source and target texts: The authors wanted to study variation among translators in their capacity to retrieve and use information that would be needed to produce well formulated text as it unfolded. For example if three subjects translated the same sentence as It is politics that fascinates me, Politics fascinates me, and I am

Methodological questions about translation research

fascinated by politics we might reason that they dier in the way they were able to retrieve the relative prominence of the counterparts of I, fascinate and politics from the source text and use that information to construct the target text. In the design stage of the experiment we recognised that the way such information is retrieved must vary according to the production mode, a fact not lost on Second Language Acquisition researchers:
It is not surprising that tasks which are based on dierent skill components such as dictation, translation or sentence repetition would yield dierences in the performance of language learners because the dierent skill components utilise different components of the language production or comprehension system. (Pienemann 1999: 276)

Lambrechts idea of allosentences resonates strongly with the interpreters or translators task in solving information structure problems, or indeed any discourse level problems. Essentially the dilemma is one of selecting among alternative structures generated by the grammar expressing the same propositional meaning (Lambrecht 1995: 340) in the target language on the basis of the discourse value of the source language string. But while translators might have the potential to generate these allosentences, it may be that the cognitive demands of the production mode aect their ability to deploy them. In this respect, Gile (1995) relates dierences in coping tactics between interpreting and translation to processing capacity and time constraints (Gile 1995: 207) present in one mode and not the other. For example, when a translator has unlimited time to review and check a translation, deciding among It is politics that fascinates me, Politics fascinates me, and I am fascinated by politics could be the result of unhurried and reective reasoning, with unlimited access to the source text and previous drafts of the target text. On the other hand if a translator is working against the clock, we could hypothesise that they retrieve just as much information as they need from a constantly refreshing mental representation of the translation task with recourse to slower reective reasoning when diculties occur. We could further hypothesise that this representation includes both source and target language elements as attention is switched between the two. If there is indeed such a dierence between the processing of slow and fast translation, is it not implausible to suggest that fast translation may have more in common in some respects with some types of interpreting? Could it be for instance that decisions about information structure are made very rapidly, and (to use the term very cautiously) automatically in both fast translation and simultaneous interpreting? And is it not implausible to suggest that phenomena like information structure are handled dierently between dierent modes of interpreting? In simultaneous interpreting the source text is

Stuart Campbell and Berta Wakim

present at the time of the delivery of the target text; in long consecutive interpretation, the source and target texts are separated in time, so that the choice between It is politics that fascinates me, Politics fascinates me, and I am fascinated by politics may depend on retrieving information from a mental representation of the unfolding text that is inuenced by the content but not the formal features of the source text. It may be then, that interpreting research and translation research are not after all categorically dierent, but are part of a broad suite of interrelated research contexts. To return to the practical problem of methodology, a number of questions in experimental design arise related to mental processing of the translation task. Here are some of the questions relevant to experiments on written translation: a. Should subjects translate against the clock, and if so how fast should they translate? This question impinges on the kind of mental processes that the experiment is trying to model. For example, speeded translation would more likely elicit examples of horizontal processing or transcoding (Gile 1995, Campbell 2000) than unspeeded translation. But how fast is fast? b. Should subjects be allowed to edit as they work? This question is especially pertinent given that Campbell (1998: 138) shows that student translators vary in the eectiveness of their on-line editing, so that editing does not necessarily improve the quality of the target text; in that case the relevance of editing to the experiments aims must be considered. Bear in mind too that editing might obliterate the output of what we might call unconscious processing and replace it with the output of conscious processing. c. What aids should be permitted, e.g. translation memory software, dictionaries, etc.? Livbjerg and Mees (2003: 130) report on experiments where dictionary look-up had no eect on quality. But the use of translation aids might introduce an additional variable that needs to be controlled, e.g. the relative rate of dictionary look-up. d. Should special texts be written for the experiments? For example, where an experiment is trying to elicit information about processing at the text level, is it legitimate to downplay the complexity at sentence or phrase level, e.g. by reducing the amount of tricky lexis, with the risk of a loss of authenticity to the professional task? For example, Dillinger (1994: 158) constructed a multidimensional prole of some thirty lexical, syntactic, and semantic properties of text for his experimental study of dierences in comprehension processing between experienced and novice interpreters. Shlesinger (2000: 11) advocates sometimes extract[ing] and combin[ing] segments from dierent texts, or to create them from scratch where a rare feature cannot be found in sucient numbers in a natural text.

Methodological questions about translation research

e. Should subjects handwrite or type? Typewriting ability introduces an additional variable that could aect mental processing, e.g. by introducing an additional focus of attention for slow typists. On the other hand, authenticity is sacriced in an experiment involving handwriting, given that almost all professional translation is done on a computer. Each of these questions needs to be satisfactorily answered in order for the experiment to generate the best quality data. Conversely, the questions provoke further questions that help focus research questions: For example, can on-line editing be made more ecient through training, and what would the cognitive and psycholinguistic rationale for such training be? How does varying the speed of translation aect performance, and what aspects of mental processing are responsible for the eect? And can we share insights from interpreting by asking how the interpreter divides his/her attention to perception and or production, and how this is carried out in real time and to what extent interpreters can really shift their attention consciously (de Bot 2000: 66). Broadly speaking, the design of whole-text experiments in translation research requires theoretical underpinning. We take the view that such an underpinning should embrace a range of bilingual production modes, e.g. translation, sight translation and interpreting in its various modes, and that these production modes form a continuum. This continuum labelled for convenience in the following pages as the Translation-Interpreting Continuum oers a badly needed theoretical foundation for research design in written translation. It should be mentioned that the continuum metaphor is not infrequently employed in building linguistic models; Grosjeans Language Mode Continuum (2001), Tourys normative continuum (1995), and Tarones (1989) capability continuum all resonate with the ideas developed here. .3 Two principal elements of the Translation-Interpreting Continuum The rst element is a model of the mental representation of the translation task, with these characteristics: (a) The model must include the switching of attention, e.g. between comprehension and production since in some production modes the two occur more or less simultaneously; simultaneous interpreters, for example, require split conceptual attention (Christoels and de Groot 2005: 456). (b) It must make reference to automaticity in some way in order to account for the difference between low speed and high speed tasks. Later in this study we will invoke transcoding as one of the surrogates for the theoretically complex notion of automaticity. (c) It must deal with the issue of word order, i.e. the variation across production modes in their potential to present target text information in a

Stuart Campbell and Berta Wakim

dierent order from the source text. (d) It must allow for retention or decay of the formal structure of the source text, e.g. in fast translation and consecutive interpreting respectively. The second element is a developmental dimension to account for production in the second language. This dimension is required as a logical consequence of the fact that for some production modes, e.g. triadic consecutive interpreting, there will be output in a second language.1 From an empirical standpoint, the dimension also acknowledges unidirectional production modes into the second language; the Translation Studies discipline no longer ignores the existence of such practices (see Campbell 1998, Anderman and Rogers 2005). This dimension is also theoretically justied in line with Campbells (1998) view that translators into L2 are by denition learners of L2.2 The developmental dimension of the TranslationInterpreting Continuum proposed here has some anity with Tarones capability continuum, which assumes:
[t]hat variability in the regular language behaviour of second language learners which is associated with the use of dierent elicitation tasks is caused by style shifting along the IL continuum, which in turn is caused by variable shifts in the degree of attention which the learner pays to language form (Tarone 1989: 15)

2. Prerequisites for a model of the mental representation of the translation task 2. Mental representation and discourse We conceptualise the mental representation of the translation task as a continuously refreshing representation, not unlike discourse processing during reading with its interplay between input, long term memory and the discourse representation in working memory. The additional elements involved in the TranslationInterpreting Continuum are output in various modes and the variability of input (i.e. the absence or presence of the source text). In fact, studies in the area of reading research (e.g. Sanford and Garrod 1981, Sanford 1990) may give us important clues to the way translators mentally process input in modes like written translation and sight translation. For example, Sanford (1990: 527) describes the reading process as [utilising] a fast pattern-matching facility only when failures of matching occur are more detailed analytical procedures assumed. Myers et al. (1998) also argue that reading is a dumb autonomous process in which concepts and propositions in the discourse representation and in the readers knowledge base resonate as a function of the degree of match to the input (Myers et al. 1998: 132). This is astonishingly close to a description of the transcoding versus

Methodological questions about translation research

meaning-based strategies. Klin et al. (2004), on the other hand, try to show how salience in the information to be processed aects its representation in memory; readers continually adjust the amount of processing devoted to each text input on the basis of its perceived importance, or salience (Klin et al. 2004: 512). There is a clear analogy here with our view that the handling of information structure varies according to production mode. In Translation Studies, attempts have been made at using discourse models in the analysis of whole texts: Tirkkonen-Condit (1985) studied translated text while Mackintosh (1985) applied Kintsch and Van Dijks (1978) model to interpreting. What interpreting and translation studies seem to lack, however, is the connection between discourse and mental processing. A similar lack is raised by Tokowicz and Perfetti in relation to studies of bilingualism, suggesting that the paucity of psycholinguistic studies of text representation could be the result of the natural focus on word and, to a lesser extent, syntactic-level processes that are the building blocks of comprehension (Tokowicz and Perfetti 2005: 175). In fact our research into information structure requires that we deal with both the mental representation of discourse and the syntactic-level processes; this is a case par excellence where it is necessary to be both looking at sentences from above (Lambrecht 1995: 340), and hypothesising about how translators and interpreters look at sentences from above or in other words how they make a mental representation of the task. Given that metaphors are frequently employed to model complex phenomena, we conclude with a simple analogy based on Chafes suggestion that we might think of what is going on in a discourse as if it described states and events unfolding on a stage. We could then say that at any particular point in the discourse there are certain things on stage (Chafe 1972: 50). The mental task of the translator is not unlike watching Romeo and Juliet in English on one side of the stage and attempting to play various parts in Arabic on the other side. 2.2 Attention switching We discuss here four aspects of attention switching that need to be incorporated into the mental representation: higher versus lower order processes, macro and microplanning, decoding and recoding, and the issue of the simultaneity of comprehension and production. De Bots study of limited attentional resources deals with higher versus lower order processes, and is based on Levelts model in which these resources are assigned to higher order processes while subskills at the syntactic and phonological levels are carried out without paying attention unless an error is detected

Stuart Campbell and Berta Wakim

(de Bot 2000: 78). Experimental work in this area appears to be conned to interpreting research, e.g. that reported in Daro et al. (1996) which explores the distribution of attention in various modes of interpreting, with the aim of validating teaching strategies; a complex interaction is described between mistakes and left/right ear superiority and focalisation on input versus output. In relation to the translation end of the continuum, a question worth asking is to what degree translators use anticipation when in fast mode; we might anticipate for example that foreknowledge of the source text topic and a degree of background knowledge would increase the capacity of second language translators to focus attention away from higher order processes to monitoring syntactic output. To our knowledge, there has been no experimental work in this area. In tasks where comprehension and planning for production occur, i.e., in the comprehension and planning phases involved in processing of monological discourse, [r]etrieving information to be expressed can involve substantial memory search, inference, and planning (Levelt 1989: 126). Evidence from studies by Henderson, Goldman-Eisler and Skarbek (1966) and Beattie (1983) suggest that speakers split attentional resources to varying degrees between the phases of macroplanning and microplanning of monologues, the length of the text being an important factor in dictating the degree of switching. We would expect that points along the Translation-Interpreting Continuum would be characterised by dierent degrees of attention allotted to macroplanning and microplanning. For example, slow translation might permit a larger allocation of eort to macroplanning than simultaneous interpreting. Additionally, in a bilingual situation, more specically in translation tasks, not only do translators have to split attentional resources between the macroplanning and microplanning of the text, but they also have to do so between the decoding of input and the recoding and production of the output: Interpreters must self-activate the target language (TL) system to encode what they have just decoded from the source text (ST) while at the same time they must keep a dierent SL utterance (or part of an utterance) in short term memory (Paradis 1994: 322). The interpreters use of attentional resources during decoding and recoding may vary according to the translation strategy used: either the conceptually mediated route which relies on implicit linguistic competence or the structural route which relies on metalinguistic knowledge (Paradis 1994: 319). Presently there appear to be no experimental studies that contrast the ecacy of the model at distant points along the Continuum. Finally, simultaneous interpreting entails switching between comprehension and production simultaneously (Christoels and de Groot 2005: 456). However, there has been little or no research on the nature and extent of such switching at other points on the Continuum such as in fast translation.

Methodological questions about translation research

2.3 Automaticity and control The extreme ends of the Translation-Interpreting Continuum seem to be characterised by a dierence of automaticity in performance: Contrast a translator who has a whole day set aside to work on a tricky 1000-word text, with a simultaneous interpreter working at 95120 words per minute; at a production rate approximately 50 times faster, the interpreters performance must in some way be more automatic. Automaticity in language processing is, however, characterised in a number of ways, such as rapidity, eortlessness, unconscious nature, and ballistic nature, and does not refer to a unitary construct (Segalowitz 2005: 372). The idea of using speed alone as an indicator of automaticity seems inadequate here; simultaneous interpreting is so dramatically fast compared with slow translation that the automaticity of interpreting must have more than just a quantitative explanation. The notions of cognitive load and eort as characteristics of automaticity have proponents in research on interpreting and translation: Giles Eort Models use a processing capacity framework to demonstrate that simultaneous interpreting relies on three controlled (i.e. non-automatic) Eorts or processing components (Gile 1995: 162). Automaticity as unconscious processing is acknowledged in Think-Aloud Protocol studies on translation, but is not of course accessible to study, being a proceduralized text processing which is hidden from introspective observation indicated by a complete lack of concurrent verbalization (Dechert 1987: 102103). The ballistic characterisation of automaticity has not been explored in interpreting and translation studies; experiments by Favreau and Segalowitz (1983) showing dierences between weaker and stronger bilinguals in their capacity for ballistic processing are based on primed lexical decision tasks, which are very remote from professional tasks. Indeed the most useful way of thinking about automaticity in interpreting and translation research may be in relation to explicit knowledge. A key notion here is that Only what is noticeable can be noticed (Paradis 2004: 34); we cannot observe implicit computational procedures, but we can have explicit knowledge of controlled processes, even though they may be speeded up to the extent that they appear to be automatic (Paradis 2004). Automatic processes in this sense are, then, those that we cannot explicitly know. To operationalise this denition for interpreting and translation research we need to ask what it is that interpreters and translators explicitly know of their mental processing: One part of the answer is to model the mental representation of the interpreting/translation task, i.e. the element of the task that can be noticed. What this approach is likely to suggest is qualitative dierences along the Continuum rather than just dierences in processing speed.

Stuart Campbell and Berta Wakim

2.4 Transcoding One interpretation of automaticity is that it does not entail attention to the task being performed, a notion that provides the foundation for the distinction between meaning-based and transcoding strategies in interpreting. The meaning based strategy entails attention, while the transcoding strategy entails automaticity as small translation units are transposed into the other language without necessarily rst being fully comprehended and integrated into the discourse representation (Christoels and de Groot 2005: 459). According to Paradis (1994: 329), [d]irect transcoding by automatic application of rules may occur at various levels of the linguistic system, including phonology, morphology, syntax and lexical semantics. While transcoding is generally seen as a feature of interpreting, Campbell suggests that transcoding be seen as the special version of automatic lexical access that translators learn to do (Campbell 2000). In fact one could argue that in both interpreting and translation, a vast amount of automatic production must occur without being comprehended and integrated in the discourse representation (Christoels and de Groot 2005: 459) in such areas as inectional morphology and high frequency lexis; whether this is transcoding or not seems to be purely a question of denition. One might expect too that the extent of transcoding will vary by task and by the competence level of subjects. When a simultaneous interpreter hears the phrase to the best of my knowledge for the ve thousandth time in their career, the likelihood is that they will transcode. A literary translator will chew over the same phrase checking for cultural equivalence, intertextual references, ironic avour and all the other things that dont happen automatically. On the other hand, an experienced freelance translator faced with a two-hour deadline for an adavit will do something very similar to the simultaneous interpreter. Competence level is likely to interact with transcoding in very many ways. Interpreters working into the second language will vary in their capacity to automatically execute grammatical rules. Variations of a quite dierent order are likely in such areas as lexical collocations; an experienced professional interpreter or translator will be able to render a collocation like nuclear family with apparent automaticity, while a learner will resort to the meaning-based strategy. On the other hand transcoding may not necessarily equate with competence; it is important to know when to have recourse to transcoding and when to avoid it, for example in the translation of information structure, where the pragmatic value of the message could be lost as a result of transcoding (Gile 1995: 7576). To return to the issue of the mental representation of the interpreting/translation task, transcoding is clearly a critical factor because of the extent to which the source text is present in the representation. In at least two places on the Translation-Interpreting Continuum, transcoding is less likely to occur, i.e. in long consecutive interpreting

Methodological questions about translation research

where the source text will have decayed to some extent by the time that production of the target commences, and in slow translation when the translator revises parts of the target text without reference to the source. 2.5 Word order Perhaps the most striking aspect of interpreted and translated output is the deviations of word order from the source text. It is well known that basic sentence order varies in accordance with the grammatical systems specic to each language. Besides coping with these basic problems, interpreting and translation involves two further issues of word order. One is the pragmatic phenomena that aect word order beyond basic sentence order. For example, the experiments that motivated this paper were particularly interested in exploring word order in relation to information structure. Along the Translation-Interpreting Continuum, the capacity to choose among allosentences like It is politics that fascinates me, Politics fascinates me, and I am fascinated by politics will depend on the priority that each mode assigns to pragmatics, and the competence of the practitioner to make pragmatic choices; slow translation will assign a very high priority to pragmatics, and simultaneous interpreting perhaps a lower priority, depending on the skill of the practitioner. The nature of the not inconsiderable skills required to achieve pragmatic equivalence in interpreting is detailed in Hale (2004). The other word order issue is related to the linearisation problem (Levelt 1981, 1989), i.e. that the temporal order of propositional content does not necessarily correspond with the linear arrangement of words dictated by syntax, so that language production necessarily entails storing information for later use in the sentence. As an issue of monolingual production alone, the word order should be of no special concern to us. However, the issue is a very worrisome one in simultaneous interpreting, where the practitioner does not merely map propositional content onto syntax, but has to formulate from the source text a representation of the propositional content, propose a plausible temporal ordering, and then map that ordering onto the target syntax. Numerous earvoice studies (de Bot 2000: 6768) have explored how simultaneous interpreters accomplish this task, especially between languages with very dierent basic word orders, e.g. from SOV into SVO. It seems plausible that at various points on a Translation-Interpreting Continuum, there may be a dierent balance of attention to pragmatics (e.g. information structure) and word order. In sight translation, for example, the need to deal with the document in the original order may require less attention to the pragmatic issue and more to word order. As we have suggested, in simultaneous interpreting, the processing required to solve word order problems especially in less competent practitioners may overwhelm the capacity to pay attention to pragmatics. With sight translation and

Stuart Campbell and Berta Wakim

simultaneous interpreting the added burdens of the co-presence of source and target texts and extemporaneous delivery accentuates the linearization problem. To shift production modes, what might be the relative balance of pragmatics and linearization in slow and fast translation? In fact a question like this is at the heart of this paper, which is trying to provide the theoretical tools to describe data types, to formulate research questions and to design methods to answer them. 2.6 Retention and decay of source text We have suggested earlier that the decay of the source text, especially in long consecutive interpreting, is a factor in distinguishing between the types of processing along the Translation-Interpreting Continuum. Decay of the source text is known as deverbalisation in the context of simultaneous interpreting research, and has been posited as a distinguishing factor between meaning-based interpreting and transcoding; in meaning-based simultaneous interpreting it is suggested that there is complete deverbalisation of the source language. The research on this issue is summarised by Christoels and de Groot (2005: 461), who conclude that complete deverbalisation seems unlikely even though it seems plausible that both transcoding and meaning-based interpreting occur . We are not aware of empirical studies that explore the extent of decay of the source text in other modes along the Translation-Interpreting Continuum (cf. the experimental study in Isham (1994) that relates to retention of sentence form in simultaneous interpreting), although logic suggests that with long consecutive interpreting at least, the constant refreshing of working memory would have ushed the source text from any kind of buffer storage by the time delivery of the target begins. De Groot (2000: 55) suggests that the opportunities for recovery [of source material] in speech perception and simultaneous interpreting are much less favourable than in the case of reading and written translation. Retention and decay of the source text is a strong candidate as a distinguishing factor along the Continuum; in fact it serves to shape important questions about, e.g. fast translation processing.

3. The developmental dimension 3. Specialised expertise It has already been suggested that, for example, capacity to match pragmatics with grammatical structure varies with the practitioners competence. In the eld of interpreting, a good deal of work has been done to demonstrate what constitutes expertise, for example chunking of information, reasoning, speed of

Methodological questions about translation research

processing, individuals knowledge base and its organisation (Moser-Mercer et al. 2000: 108). Variation in all or some of these may aect performance along the Translation-Interpreting Continuum, not necessarily in interpreting alone. 3.2 Second language competence Cutting across these skill components is competence in second language production. Given, as we have mentioned, that the translation studies discipline no longer ignores the existence of translation into the second language (and that of course triadic interpreting always involves working into a second language), L2 competence must be accepted as a criterion at all points along the Continuum. In fast translation, for example, Campbell (1998) has identied three levels of textual competence, and has described aspects of a developmental pathway between the levels. In accordance with the notion that learning to translate into L2 is a special kind of second language acquisition (Campbell 1998) we feel justied in borrowing from second language research to esh out this developmental dimension; we know for example, that cognitive complexity of tasks (Gopher 1993, Long 1996, Shebilske et al. 1999) and mental control, e.g. stress (Wegner 1994), all aect L2 performance, and it seems reasonable that the same factors aect interpreting and translation performance into L2. While the developmental dimension is the least developed aspect of the Continuum it acts as a generator of research questions about, for example, the transition from controlled to automatic processing, the acquisition of transcoding skills and the acquisition of pragmatic competence, across a range of production modes.

4. The Translation-Interpreting Continuum in overview The elements of the Translation-Interpreting Continuum discussed above can be summarised graphically (if crudely) in Figure 1.
Production modes Slow Fast Sight Simultaneous Consecutive translation translation translation interpreting interpreting Attention switching slow fast Automaticity as low high transcoding Potential for word high low order variation Retention/decay of high low ST formal structure

Elements of mental representation

Figure 1. Translation-Interpreting Continuum in Overview

Stuart Campbell and Berta Wakim

The Continuum as presented here includes only ve modes. There is no reason why it could not be expanded to include, for example, sight interpreting, liaison interpreting and computer-aided translation. This Continuum is essentially a set of hypotheses out of which (a) research questions can be generated and (b) appropriate data sets and research methods developed. A glance at Figure 1, suggests for example that in sight translation, subjects who can achieve a high degree of transcoding could allocate more processing resources to word order variation in order to produce more stylistically authentic output; a testable hypothesis can be developed, along with suitable data and research procedures.

5. Methodological questions revisited To return to the questions we posed about translation research methodology at the beginning of this paper, we conclude by giving some brief answers to show how the Translation-Interpreting Continuum may shed some light on why and how we should research translation: Q Should subjects translate against the clock, and if so how fast should they translate? A It depends on what we want to nd out. If we are interested in the extent of automaticity of processing in translation subjects, then we could design experiments in which they translate progressively faster so that we can observe changes in output. At a more applied level, we know virtually nothing about how fast translation examinations should be; but we may be able to design experiments that show at what speed certain skills are optimally elicited, e.g. the ability to manipulate word order to achieve textual equivalence. Q Should subjects be allowed to edit as they work? A Again, this depends on what we are interested in discovering. Editing in fast translation may give us insight into the richness or eectiveness of the mental representation of the translation task, e.g. by showing the size of the chunks being processed. If on the other hand we were interested in the underlying ability to produce well-formed sentences in L2 regardless of the discourse features of the translation, we might require fast work with no editing. Q What aids should be permitted, e.g. translation memory software, dictionaries, etc.? A The intuitive answer is that if the use of such aids disturbs the even ow of processing, then experiments should not permit them. On the other hand, in professional work, translation aids are routinely used and represent just one more

Methodological questions about translation research

attractor of attention. One could, then, imagine an experiment that analyses the allocation of attention resources in fast translation. Q Should special texts be written for the experiments? A Unequivocally yes, if the special text will help to provide a more eective data set: For example, if we were interested in the level at which subjects transcode, e.g. single words, formulaic expressions, grammatical constructions, then we might write experimental texts rich in the categories that interest us. Dillinger argues that:
there seem to be no objective criteria to judge the articiality or naturalness of the texts used, and more importantly, there is no evidence that the articiality of the texts makes any dierence in how interpreters interpret. (Dillinger 1994: 158159)

Q Should subjects handwrite or type? A Despite the fact that professional translators almost always type, handwriting is an easier variable than typing skill to control in fast translation experiments. If subjects are to type in fast translation, then they need to be trained to the same level. In slow translation, there is no need to control for typing ability, and in fact keystroke-recording software, e.g. Translog (Jakobsen and Schou 1998) can provide very rich data about mental processes. To conclude, in our view translation research in the domain of mental processing can proceed only haltingly without a well-grounded model of this kind. Our hope is that we will see a better integration of research across the translation-interpreting boundary, and encourage in Translation Studies the kind of high quality experimental research that is commonplace in psycholinguistics.

Notes
* The research reported here was supported by the University of Western Sydney Research Grant scheme. We acknowledge comments on earlier drafts of this paper by Raymond Chakhachiro (University of Western Sydney), Ayla Al-Rubai (Al-Mustansariyya University, Iraq) and Zhang Delu (Ocean University of China). . We acknowledge the enduring debate on the existence of true bilinguals. 2. Second Language Acquisition research is marginally interested in translation per se, but only as an elicitation tool. It is worth mentioning that Second Language Acquisition research has on occasion used translation tasks in order to elicit certain structures in learner language, but that most of these tasks were based on word or sentence level. For example, Ravem (1968) used sentence translation into the second language; the procedure was meant to produce data similar to spontaneous speech. Hkansson (2001) studied word order variation in L1 transfer into L2 using sentence translation.

Stuart Campbell and Berta Wakim

References
Anderman, Gunilla M. and Margaret Rogers, eds. 2005. In and out of English: For better, for worse? Bualo: Multilingual Matters. Beattie, Georey. 1983. Talk: An analysis of speech and non-verbal behaviour in conversation. Milton Keynes: Open University Press. Campbell, Stuart. 1998. Translation into the second language. London: Longman. Campbell, Stuart. 2000. Critical structures in the evaluation of translations from Arabic into English as a second language. The translator 6:2. 211229. Chafe, Wallace L. 1972. Discourse structure and human knowledge. Roy O. Freedle and John B. Carroll, eds. Language comprehension and the acquisition of knowledge. Washington: V.H. Winston, 1972. 4169. Christoels, Ingrid K. and Annette M.B. de Groot. 2005. Simultaneous interpreting: A cognitive perspective. De Groot et al. 2005. 454479. Christoels, Ingrid K., Annette M.B. de Groot and Judith F. Kroll. 2006. Memory and language skills in simultaneous interpreters: The role of expertise and language prociency. Journal of memory and language 54:3. 324345. Danks, Joseph H, Stephen B. Fountain and Michael K. McBeath, eds. 1997. Cognitive processes in translation and interpreting. Thousand Oaks, California: Sage Publications. Daro, Valerie, Sylvie Lambert and Franco Fabbro. 1996. Conscious monitoring of attention during simultaneous interpretation. Interpreting 1:1. 101124. de Bot, Kees. 2000. Simultaneous interpreting as language production. Hyltenstam and Englund, 2000. 6588. de Groot, Annette M.B. 1997. The cognitive study of translation and interpretation: Three approaches. Danks et al. 1997. 2556. de Groot, Annette M.B. 2000. A complex-skill approach to translation and interpreting. Tirkkonen-Condit and Jskelinen 2000. 5368. de Groot, Annette M.B. and Judith F. Kroll, eds. 2005. Handbook of bilingualism: Psycholinguistic approaches. New York: Oxford University Press. Dechert, Hans. W. 1987. Analysing language processing through verbal protocol. Frch and Kasper, 1987. 96112. Dillinger, Mike. 1994. Comprehension during interpreting: What do interpreters know that bilinguals dont? Lambert and Moser-Mercer 1994. 155189. Frch, Claus and Gabriele Kasper, eds. 1987. Introspection in second language research. Clevedon, Avon: Multilingual Matters. Favreau, Micheline and Norman S. Segalowitz. 1983. Automatic and controlled processes in the rst- and second-language reading of uent bilinguals. Memory & cognition 11:6. 565 574. Gerlo, Pamela. 1987. Identifying the unit of analysis in translation: Some uses of Think-aloud Protocol data. Frch and Kasper, 1987. 135158. Gile, Daniel. 1995. Basic concepts and models for interpreter and translator training. AmsterdamPhiladelphia: John Benjamins. Gile, Daniel. 1997. Conference interpreting as a cognitive management problem. Danks et al. 1997. 196214. Gopher, Daniel. 1993. The skill of attention control: Acquisition and execution of attention strategies. David E. Meyer and Sylvan Kornblum, eds. Attention and performance XIV:

Methodological questions about translation research Synergies in experimental psychology, articial intelligence, and cognition. Cambridge, Massachusetts: The Massachusetts Institute of Technology Press, 1993. 299322. Grosjean, Franois. 2001. The bilinguals language modes. Janet Nicol, ed. One mind, two languages: Bilingual language processing. Malden, Massachusetts: Blackwell, 2001. 122. Hkansson, Gisela. 2001. Against full transfer evidence from Swedish learners of German. Lundt University, Department of Linguistics Working Papers 48. 6886. Hale, Sandra Beatriz. 2004. The discourse of court interpreting: Discourse practices of the Law, the Witness, and the Interpreter. AmsterdamPhiladelphia: John Benjamins. Henderson, Alan I., Frieda Goldman-Eisler and Andrew Skarbek. 1966. Sequential temporal patterns in spontaneous speech. Language and speech 9. 207216. Hyltenstam, Kenneth and Birgitta Englund, eds. 2000. Language processing and simultaneous interpreting: Interdisciplinary perspectives. Amsterdam: John Benjamins. Isham, William P. 1994. Memory for sentence form after simultaneous interpretation: Evidence both for and against deverbalization. Lambert and Moser-Mercer1994. 191211. Jakobsen, Arnt Lykke and Lasse Schou. 1999. Translog documentation. Gyde Hansen, ed. Probing the process in translation: Methods and results. Copenhagen: Samfundslitteratur, 1999. 920. Kintsch,Walter and Teun A. Van Dijk. 1978. Toward a model of text comprehension and production. Psychological review 85:5. 363394. Klin, Celia M, Kristin M. Weingartner, Alexandria E. Guzman and William H. Levine. 2004. Readers sensitivity to linguistic cues in narratives: How salience inuences anaphor resolution. Memory & cognition 32:3. 511522. Krings, Hans P. 1986. The translation strategies of advanced German learners of French. Juliane House and Shoshana Blum-Kulka, eds. Interlingual and intercultural communication: Discourse and cognition in translation and second language acquisition studies. Tbingen: Gunter Narr, 1986. 263276. Krings, Hans P. 1987. The use of introspective data in translation. Frch and Kasper 1987. 159176. Lambert, Sylvie and Barbara Moser-Mercer, eds. 1994. Bridging the gap: Empirical research in simultaneous interpretation. AmsterdamPhiladelphia: John Benjamins. Lambrecht, Knud. 1995. Information structure and sentence form: Topic, focus, and the mental representations of discourse referents. Cambridge; New YorkUSA: Cambridge University Press. Levelt, Willem J.M. 1981. Dj vu? Cognition 10:13. 187192. Levelt, Willem J.M. 1989. Speaking: From intention to articulation. CambridgeMassachusets: The Massachusetts Institute of Technology Press. Livbjerg, Inge and Inger M. Mees. 2003. Patterns of dictionary use in non-domain-specic translation. Fbio Alves dos Santos, ed. Triangulating translation: Perspectives in process oriented research. AmsterdamPhiladelphia: John Benjamins, 2003. 123136. Long, Michael H. 1996. The role of the linguistic environment in second language acquisition. William C. Ritchie and Tej K. Bhatia, eds. Handbook of second language acquisition. San Diego: Academic Press, 1996. 413468. Mackintosh, Jennifer. 1985. The Kintsch and van Dijk model of discourse comprehension and production applied to the interpretation process. Meta 30:1. 3743. MacWhinney, Brian. 2005. A unied model of language acquisition. de Groot and Kroll, 2005. 4967.

Stuart Campbell and Berta Wakim Moser-Mercer, Barbara, Uli. H. Frauenfelder, Beatriz Casado and Alexander Knzli. 2000. Searching to dene expertise in interpreting. Hyltenstam and Englund 2000. 107131. Myers, Jerome L. and Edward J. OBrien. 1998. Accessing the discourse representation during reading. Discourse processes 26:2/3. 131157. Paradis, Michel. 1994. Toward a neurolinguistic theory of simultaneous translation: The framework. International journal of psycholinguistics 10:3. 319335. Paradis, Michel. 2000. Prerequisites to a study of neurolinguistic processes involved in simultaneous interpreting. Hyltenstam and Englund, 2000. 1724. Paradis, Michel. 2004. A neurolinguistic theory of bilingualism. AmsterdamPhiladelphia: John Benjamins. Pienemann, Manfred. 1999. Language processing and second language development: Processability theory. AmsterdamPhiladelphia: John Benjamins. Ravem, Roar. 1968. Language acquisition in a second language environment. International review of applied linguistics 6. 165185. Sanford, Anthony J. 1990. On the nature of text driven inference. David. A. Balota, Giovanni B. Flores DArcais and Keith Rayner, eds. Comprehension processes in reading. Hillsdale, New Jersey: Laurence Erlbaum, 1990. 515535. Sanford, Anthony J. and Simon C. Garrod. 1981. Understanding written language: Explorations of comprehension beyond the sentence. ChichesterNew York: Wiley. Schner, Christina, ed. 2004. Translation research and interpreting research: Traditions, gaps and synergies. Clevedon, EnglandBualo: Multilingual Matters. Segalowitz, Norman. 2005. Automaticity in bilingualism and second language learning. de Groot and Kroll, 2005. 371388. Shebilske, Wayne L., Barry P. Goettl and Wesley J. Regian. 1999. Executive control and automatic processes as complex skills develop in laboratory and applied settings. Daniel Gopher and Asher Koriat, eds. Attention and performance XVII: Cognitive regulation of performance: Interaction of theory and application. Cambridge, Massachusets: The Massachusetts Institute of Technology Press, 1999. 401432. Shlesinger, Miriam. 2000. Interpreting as a cognitive process: How can we know what really happens? Tirkkonen-Condit and Jskelinen 2000. 315. Tarone, Elaine. 1989. On the variability of interlanguage systems. Fred R. Eckman, Lawrence H. Bell and Diane Nelson, eds. Universals in second language acquisition. Rowley, Massachusetts: Newbury House, 1989. 923. Tirkkonen-Condit, Sonja. 1985. Argumentative text structure and translation. Jyvskyl: University of Jyvskyl. Tirkkonen-Condit, Sonja and Riitta Jskelinen, eds. 2000. Tapping and mapping the processes of translation and interpreting: Outlooks on empirical research. Amsterdam: John Benjamins. Tokowicz, Natasha and Charles A. Prfetti. 2005. De Groot and Kroll 2005. 173177. Toury, Gideon. 1995. Descriptive Translation Studies and beyond. AmsterdamPhiladelphia: John Benjamins. Wegner, Daniel M. 1994. Ironic processes of mental control. Psychological review 101:1. 34 52.

Methodological questions about translation research

Rsum
Ltude des processus mentaux en traduction crite a, moins que la recherche en interprtation, pu bncier des rsultats de recherches psycholinguistiques et cognitives. En consquence, les mthodes de recherche appliques la traduction nont pas dassises thoriques comparables. Alors que la vitesse de performance de linterprtation semble rapprocher la recherche en interprtation de la recherche en parole bilingue, les traits partags par linterprtation et la traduction semblent des points sur un continuum plutt que des modes de production dirents. Larticle propose un continuum traduction-interprtation, qui permet de dcrire dirents modes de production en termes de notions psycholinguistiques ; ce continuum comprend une dimension de dveloppement linguistique qui permet de rendre compte de la production dune seconde langue. Le continuum permet de poser des questions fondamentales la mthodologie de la recherche en traduction ; il cre galement des possibilits pour une meilleure intgration de la recherche en traduction et en interprtation.

Authors address
Stuart Campbell and Berta Wakim School of Humanities and Languages University of Western Sydney PENRITH SOUTH DC 1797 Australia e-mail: S.Campbell@uws.edu.au

You might also like