Professional Documents
Culture Documents
Special issues of established journals tend to circulate within the orbit of the
subscribers of those journals. For the Benjamins Current Topics series a number
of special issues of various journals have been selected containing salient topics of
research with the aim of finding new audiences for topically interesting material,
bringing such material to a wider readership in book format.
For an overview of all books published in this series, please see
http://benjamins.com/catalog/bct
Volume 77
Describing Cognitive Processes in Translation. Acts and events
Edited by Maureen Ehrensberger-Dow, Birgitta Englund Dimitrova,
Séverine Hubscher-Davidson and Ulf Norberg
These materials were previously published in Translation and Interpreting Studies
8:2 (2013)
Describing Cognitive Processes
in Translation
Acts and events
Edited by
Maureen Ehrensberger-Dow
Zurich University of Applied Sciences
Séverine Hubscher-Davidson
Aston University
Ulf Norberg
Stockholm University
doi 10.1075/bct.77
Cataloging-in-Publication Data available from Library of Congress:
lccn 2015019668 (print) / 2015026242 (e-book)
isbn 978 90 272 4265 5 (Hb)
isbn 978 90 272 6820 4 (e-book)
Introduction 1
Models of what processes? 7
Andrew Chesterman
Shared representations and the translation process: A recursive model 21
Moritz Schaeffer and Michael Carl
ELF speakers’ restricted power of expression: Implications for
interpreters’ processing 43
Michaela Albl-Mikasa
The role of intuition in the translation process: A case study 63
Séverine Hubscher-Davidson
The effect of interpreting experience on distance dynamics: Testing the
literal translation hypothesis 85
Nataša Pavlović and Goranka Antunović
The impact of process protocol self-analysis on errors in the translation
product 105
Erik Angelone
Opening eyes to opera: The process of translation for blind and partially-
sighted audiences 125
Sarah Eardley-Weaver
Index 149
Introduction
1. The chapters were originally published in a special issue of the journal Translation and
Interpreting Studies (8:2, 2013), bringing together papers presented at two international con-
ferences in 2011, Text-Process-Text, at Stockholm University, Sweden, and Translation Process
Research: Breaking New Ground, at Aston University, Birmingham, U.K.
doi 10.1075/bct.77.01int
2015 © John Benjamins Publishing Company
2 Introduction
past decade of process-oriented translation studies and its most prominent topics
and tendencies, see Muñoz (2014).
In parallel with the field of process-oriented studies, another important field
has evolved in Translation Studies, also having as one of its focal points the transla-
tor/interpreter, but seen from the perspective of being embedded in his/her social
environment: translation sociology or the sociology of translation (e.g., Wolf and
Fukari 2007; Kinnunen and Koskinen 2010; Angelelli 2012). This field could be
said to have originated in the core of Descriptive Translation Studies and in the
conceptualization of translation norms (see Toury 1995) but has since been fur-
ther developed by inclusion of theories and concepts from sociologists such as
Bourdieu and Latour.
Chesterman (2009) proposed a new subfield within Translation Studies, which
he called Translator Studies. He further suggested a tripartite division of this sub-
field, with cultural, cognitive, and sociological branches, the last two of which are
the focus of the present volume. The sociological branch would deal with “transla-
tors’/interpreters’ observable behaviour as individuals or groups or institutions,
their social networks, status and working processes, their relations with other
groups and with relevant technology”, and the cognitive branch would deal with
“mental processes, decision-making, the impact of emotions, attitudes to norms,
personality” (Chesterman 2009: 19). It is clear that there is overlap in some of the
objects of study in the cognitive and the sociological branches and hence a poten-
tial for a rapprochement and perhaps cooperation. Examples of studies which may
be seen as located at the interface between the two branches are Buzelin (2007) and
Risku and Windhager (2013/2015). The relevance of such potential convergence
seems especially clear in view of the rapid changes in the working conditions of
translators and interpreters. Increased use of technology, specialization of transla-
tion work processes in large enterprises as well as the outsourcing of translation
tasks are just some factors which characterize the development of modern transla-
tion professions. How those changes in the sociological translation event impact
the cognitive translation act is a fruitful avenue for future study.
The six chapters in this volume address various aspects of translators’ and
interpreters’ observable and non-observable processes, thus enabling readers to
reflect further on the concept of translator studies and a possible merging of cog-
nitive and sociological approaches to understanding the phenomenon of interest.
In the first chapter, Andrew Chesterman provides a theoretical framework for
the volume with Toury’s (1995) proposed distinction between different translation
processes: on the one hand, cognitive translation acts, i.e., the mental processes,
and on the other, sociological translation events, into which translation acts are
embedded. Chesterman here proposes a third term, translation practices, to desig-
nate the translation process at the historical and cultural level. The author further
Introduction 3
claims that translation acts can be studied from three different perspectives: as
virtual (potential) processes, as reverse-engineered (reconstructed) processes, or
as actual (observed) processes.
In terms of those distinctions, the other chapters of the volume can be charac-
terized as studies of the actual cognitive translation acts, of other processes related
to the translation acts, or of processes that are related to the sociological transla-
tion event. They share an interest in actual, observed processes but differ with
respect to the phenomena under investigation as well as the methods used to gain
an increased appreciation and understanding of those phenomena.
Focusing on bilingual processing, clearly part of the act of translation, Moritz
Schaeffer and Michael Carl re-evaluate the psycholinguistics of the bilingual lexi-
con and evidence from bilingual priming studies in order to gain insights into
automated translation processes. They argue that translation involves the activa-
tion of shared cognitive representations and report experimental evidence for a
facilitative effect of translation over simple reading on the recall of a source text.
In the light of their findings, they propose a recursive model of translation which
re-defines the literal translation hypothesis and the monitor model in terms of
horizontal and vertical bilingual processing during the production of a target text.
In her chapter, Michaela Albl-Mikasa considers the special challenges that
ELF (English as Lingua Franca) input presents in the act of mediated bilingual
processing in conference interpreting. She draws on a small case study of a student
interpreter’s rendering into German of short speeches given by ELF speakers. Her
analysis of the data suggests that interpreting ELF speech may be compromised
on two levels: non-standard input may affect the interpreter’s comprehension
processes; and unconventional expressions and structures prevent the interpreter
from being able to rely on previously established links in rendering the target text.
Especially interesting with respect to the act-event distinction is that the study also
includes the ELF speakers’ perspectives on their own discourse, as deliverers of the
source speech. In her conclusion, Albl-Mikasa discusses the didactic implications
of trying to prepare students for unpredictable ELF input.
Probing how non-conscious processes might influence the act, Séverine
Hubscher-Davidson examines the psychological construct of intuition and ex-
plores the role that it might play in translation. She reviews the literature that has
considered the importance of intuition in translation and explains the construct
in light of recent psychological research. Self-report data from a case study of a
student translating a literary text is analyzed in order to highlight the possible in-
fluence of intuition in decision-making behavior during translation. On the basis
of her analyses, she claims that the potential to enhance intuitive decision-making
should be exploited, since it might be an important part of the translation process.
4 Introduction
Similar to Schaeffer and Carl, Nataša Pavlović and Goranka Antunović focus
on the literal translation hypothesis. Using Translog recordings of interpreters’ and
translators’ processes, they test whether more literal renderings during the initial
act of translation become freer during self-revision. Hypothesizing that interpret-
ing experience will have an effect on the types of self-revisions made at various
stages of the process, they compare revisions in terms of distance to the source text
structure. Their hypothesis is only partly confirmed by their results. An important
finding is that many self-revisions are ‘neutral’ in terms of distance to the source
text, and the authors conclude that this has to be taken into account in future stud-
ies on self-revision.
By encouraging students to reflect on their own processes afterwards, Erik
Angelone moves them a step away from the immediate cognitive act. In his ex-
ploratory study of self-revision processes, he compares the efficacy of three types
of translation process protocols as prompts for post-drafting translation problem
recognition and revision. Students created a process protocol while translating
and then self-reflected about their problem-solving performance on the basis of
that protocol. They were given the opportunity to revise their translations follow-
ing their analyses, and then their final versions were marked for various types of
errors. Of the three protocol types, screen recordings were consistently associated
with the fewest errors in the final translations, suggesting that this type of prompt-
ing would be a useful revision technique in process-oriented translator training.
In the last chapter of this volume, Sarah Eardley-Weaver considers a very dif-
ferent type of translation process, one which is more closely linked to the concept
of a sociological translation event. She discusses the growing field of audiovisual
translation and examines audio description and touch tours for partially-sighted
and blind patrons of opera performances. Proceeding from the framework of actor
network theory also used in studies within sociology and translation (Abdallah
2010), she explores the roles of various actors involved in such translation process-
es and argues that the complex networks would be best described as iterative, with
the audience’s reception of the translation product feeding back into subsequent
processes. Audience reception studies, she points out, should be an integral part
of the investigation of any translation process and not just audiovisual translation.
With the chapters of this volume, we hope to have provided impetus for further
work, combining the act and event perspectives.
Birgitta Englund Dimitrova
Maureen Ehrensberger-Dow
Séverine Hubscher-Davidson
Ulf Norberg
Introduction 5
References
Andrew Chesterman
Most process research on translation has so far focused on the cognitive dimen-
sion; the investigation of sociological processes has not attracted as much atten-
tion, although the recent “sociological turn” in translation studies may influence
future tendencies (e.g., Wolf and Fukari 2007). The distinction between these two
basic dimensions was already implied by Gideon Toury (1995), in his discussion of
natural translation and sources of feedback. Feedback, he wrote, may come from
the translation recipients, and also from:
1. My sincere thanks are due to the two critical referees who made many helpful suggestions on
the initial version of this chapter. I should undoubtedly have taken more of them on board than
I have been able to.
doi 10.1075/bct.77.02che
2015 © John Benjamins Publishing Company
8 Andrew Chesterman
those who have commissioned the act of translating, and sometimes from the orig-
inator of the utterance to be translated as well. When realized by actual persons,
these roles (in the sociological sense) — all parts of the interactional makeup of a
translating event — may, of course, partially overlap. (1995: 249, emphasis original)
Reading somewhat between the lines here, many scholars (including myself) have
taken “act of translating” — or translation act — to refer to the cognitive process,
whereas the translation event is the observable sociological framework in which
the cognitive translation act takes place. In the revised version of the book (Toury
2012: 67–68), this distinction is drawn more clearly and developed at some length.
The locus of the translation act is said to be “the human brain.” Yet Toury insists
there that the cognitive cannot be completely separated from the environmental.
The relation between the mental act and the situational event is one of “comple-
mentarity and containment.” This is a position that has also been taken by other
scholars such as Risku (2010), who have explored the relevance of the notion of
situated cognition to Translation Studies. Toury calls for research on the interac-
tion between these two levels, the mental and the situational. A recent example of
how this interaction might be studied is Jones (2011), on poetry translation; this
study combines think-aloud protocols (TAPs), interviews, and agent networks.
An act is thus embedded in an event: the event is the sociological or situ-
ational context of the act. The event is directly observable: one can follow a trans-
lator’s overt behavior, observe phone calls, emails, use of the internet, physical
movements, and so on. But the act is not directly observable, one can only make
inferences from the behavior one can see. Even the insertion of electrodes in the
brain does not provide direct access to cognition itself, only to the electric pulses
and neuron activities etc. which manifest it. A translation event normally involves
other actors too, of course, who also perform relevant cognitive acts.
In principle, the distinction between the cognitive translation act and the so-
ciological translation event seems clear, although both terms refer to processes
taking place in time; moreover, some models appear to incorporate aspects of both
dimensions. In translation, these time scales are very different: one is measured in
seconds or microseconds, the other in hours or days, or even months. There does
not yet seem to be agreement about how to define precise starting and ending
points of a translation act or event, however, or how to conceptualize the overlap
between them. Aspects of the event, for instance, such as the details of the brief
and the definition of the intended addressees, presumably influence the mindset
of the translator and hence the cognitive translation act, perhaps even before he/
she even receives the source text. Let us nevertheless suggest that a translation
act begins when the translator begins to read the source text, and ends when the
translator decides to take no further action in revising the translation (although
Models of what processes? 9
there might still be further thought on the subject). The act may of course be in-
terrupted. A translation event, on the other hand, could be said to begin when the
translator accepts the job (or perhaps when the client begins to look for a transla-
tor?), and ends with, say, payment of the bill (or perhaps when the first recipient
reads the translation?). Translation events have been investigated in different ways,
via workplace studies, revision procedures, the analysis of translator teams and
networks, translator agency, and so on (e.g., Kinnunen and Koskinen 2010). Here
too, interruptions are common.
This chapter mainly concerns translation, but it is worth noting that, in inter-
preting, the distinction between act and event often seems less clear-cut. But the
interpreting act and event can still be studied separately, as indeed they are, using
different kinds of methods and models.
implicitly predicts that it will also apply to yet-to-be-studied instances of the phe-
nomenon in question, indeed to all possible instances belonging to the same set.
In this weak sense, a descriptive model could be said to be explanatory in that it
generalizes, by predicting applicability to unknown instances (on explaining via
generalization, see Chesterman 2008). However, explanations and predictions do
not inevitably go hand in hand, as is well known.
In a stronger sense, a model can be predictive of consequences that can them-
selves be tested empirically. A classic example from chemistry is Mendeleev’s
nineteenth century model of the elements, the periodic table, which arranged the
elements by group and by atomic weight. The first versions of this model had gaps
at certain points; the model predicted that these gaps would eventually be filled, as
new elements were discovered. And they were.
As we shall see, not many models of the translation process appear to be par-
ticularly predictive.
In what follows I will distinguish between different kinds of models of the
cognitive translation process according to the ontological status of what is being
modeled. These different kinds of models actually represent different senses of
‘translation process,’ although all are concerned with the cognitive translation act,
not the sociological event. My presentation of model types is based on Toury’s
discussion (2002; 2012: Chapter 2) of three different senses of the term ‘translation
problem.’ Toury does not give labels to these different senses, but refers to them
as problem1, problem2, and problem3. In brief, Toury’s distinction is as follows.
The first sense is the potential problem of the translatability of a given source-text
item into a given target language, under given conditions: how might this ST item
be translated here? A problem in the second sense is identified by starting with a
given target-text item that functions as a factual translation solution, and then at-
tempting in retrospect to reconstruct the translation problem for which it has been
selected as a solution, and also to reconstruct the thinking that led the translator
to this solution. And the third sense is the notion of a problem as it is observed to
be experienced by the translator, during the process of a given act of translation,
via traces left e.g., by interim solutions, by verbal reflection, or by pauses in the
process. In Toury’s revised version of his book (2012), these three senses of ‘trans-
lation problem’ are argued to correspond to different senses of the translation act.
With these senses in mind, consider what they imply for an understanding
of different models of the process of the translation act. I propose to distinguish
three kinds of models: models of virtual processes, corresponding to Toury’s first
sense of ‘problem’; those of reverse-engineered processes, corresponding to Toury’s
second sense; and those of actual processes, corresponding to his third sense. I now
look at these types in more detail, and give some examples.
Models of what processes? 11
First, consider a model of what I will call a virtual process: this would be the po-
tential path from one sense of ‘translation problem’ to a potential solution, show-
ing for instance the possible strategies for the translation of an allusion or a pun
(taken as translation problems, in Toury’s sense 1). Such a model is pedagogical or
advisory/prescriptive in nature, and starts with something in the source text that is
treated as a translation problem, or as we saw Toury put it, a problem of translat-
ability. It thus takes a prospective approach. Based on intuition, or experience, or
on the analysis of many translations, the model then outlines possible courses of
action leading to possible solutions, in theory. It is a simplified, idealized model, of
possible decision processes leading to acceptable solutions (although, clearly, real
translators might not behave optimally). Such a model might also purport to rep-
resent the translation process in general, not just the solution of a given problem.
Models of the virtual process are predictive in the weak sense that if translators
follow the advice illustrated by the model, it is presumably assumed that the results
will tend to be more acceptable than if the advice is not followed. In other words,
the model implies a prediction that its use will lead to beneficial effects: hence its
usefulness in the classroom. In principle, therefore, these models are testable: we
can test whether their use really does lead to better translations than cases where
they are not used. (But do we have any such tests?)
An early example of such a virtual model is Nida’s (e.g., 1969) well-known
three-stage model, comprising analysis, transfer, and restructuring, with obvious
pedagogical aims. Nord’s “looping model” (1991) takes Nida’s pedagogical ap-
proach a step further. She starts with the analysis of the skopos, then the analysis
of the source text, followed by the production of the translation. The model loops
back and forth between these three. Feedback comes from the emerging target text
itself, too, as later decisions affect earlier ones. Hönig (1995) proposes an “ideal”
flow-chart model, including the translator’s macro- and micro-strategies, moni-
toring, etc. Interestingly, he explicitly compares his virtual model with Krings’ ac-
tual one (see below).
Levý’s game-theoretical model (1967) takes a teleological point of view of the
virtual translation process, in which the translation act is represented as a series of
decisions, which are like moves in a game. Alternative solutions to a given ST seg-
ment are generated, assessed according to specified criteria (such as stylistic natu-
ralness vs. semantic closeness, or type of implied audience). Most of the examples
he analyzes are modeled in terms of binary decisions. Levý takes the process of
translating to be “a game with complete information”; this means a game where
every decision and move is influenced by the knowledge of previous decisions and
their consequences.
12 Andrew Chesterman
this case also reading. Like other TAP researchers, Shreve et al. assume in general
that (longer) pauses indicate problems or points where special efforts are needed.
The third kind of model aims to represent a process in real time. We could call it
a concurrent model, of an actual process. Such models make use of data derived
from observation of the process as it takes place, using such methods as think-
aloud protocols, keyboard logs, eye-tracking, and the like. They are thus different
from reverse-engineered models in that they observe the process as it unfolds,
rather than retrospectively inferring from the end-result how this result might
have arisen. (True, a model of an actual process also involves inferences concern-
ing the nature of the unobservable cognitive activities presumably taking place.)
Models of an actual process are different from virtual models in that they are de-
scriptive, not prescriptive (e.g., Kussmaul 2007: 91).
Krings’s model of the actual process (1986), based on TAP data, is a group-
ing of types of strategies (of comprehension, retrieval, decision-making, monitor-
ing, and reduction) to solve problems of different kinds. Lörscher’s model (1991)
aimed to represent the abstract kinds of “strategies” translators use when they
come to a segment of source text that they cannot translate routinely: they have
to stop and think. Lörscher’s strategies are thus problem-solving ones, and do not
represent the whole of the translation act. Examples of such strategies are: the
realization of a translation problem; its verbalization; formulation of a tentative
solution; assessing this solution. The classification is based on TAP data, with a few
non-professional translators. What is modeled is part of an observed process, at a
very general level. He concludes that the strategies he discusses are unlikely to be
specific to translation itself, but pertain to text-processing generally.
Jakobsen’s recent work with Translog and eye-tracking (e.g., 2011) is of con-
siderable interest as an approach to the observed cognitive process, i.e., what I
have called the actual process. Among his results he cites data (2011: 40) suggest-
ing that at least on some occasions, comprehension, formulation of a translation,
and actual typing could not possibly always be in sequence, linearly, but overlap;
this is because some recorded chunks are so long they could not possibly have
been held in short-term memory. In other words, there must have been some
parallel processing going on. An alternative interpretation of these results might
argue that they go against the assumption that there are separate “modules” for
memory, formulation, and physical execution (typing) in the first place.
14 Andrew Chesterman
On the basis of his eye-tracking and keystroke data, Jakobsen summarizes the
translation “micro-cycle” as a series of six steps, which he presents as a “small al-
gorithm” (2011: 48).
1. Moving the gaze to read the next chunk of new source text (and constructing
a translation of it)
2. Shifting the gaze to the target text to locate the input area and read the current
target-text anchor word(s)
3. Typing the translation of the source-text chunk
4. Monitoring the typing process and the screen outcome
5. Shifting the gaze to the source text to locate the relevant reading area
6. Reading the current source-text anchor word(s)
As thus formulated, these steps partly refer to the translation act and partly to the
event. Typing is an observable feature, and hence pertains to the translation event;
constructing a translation and monitoring, however, are cognitive acts. Jakobsen
adds that specifically steps 3 and 4 are not necessarily in linear sequence; that there
may be a good deal of recursion; and that steps can be skipped. Steps move some-
times from source to target and sometimes from target to source.
When models of actual processes are sufficiently formalized, computational
models of the translation act can be developed, with the goal of improving in-
teractive machine translation systems. Carl (2012), for example, reports research
of this kind, based on the actual behavior of expert translators. This behavior is
studied via keystroke logging and eye-tracking, but the derived “production rules”
include inferred mental operations such as shifting attention and actually translat-
ing, as well as the observable actions of shifting gaze and typing. Carl also outlines
a statistical model of the process studied, which appears to match actual human
behavior more closely. This model so far only deals with “unchallenged” (i.e.,
problem-free, routine) translation.
The relations between models of these three kinds of processes prompt a number
of questions. In the first place, whose processes are being studied? Much of the
research on actual processes has compared novice translators to professionals or
experts, on the assumption that something can thus be learned about the develop-
ment of translation competence, or about potential application in translator train-
ing (e.g., Jääskeläinen 1999; Englund Dimitrova 2005). There has thus been an
implicit or explicit quality variable involved: it is often the cognitive processes of
Models of what processes? 15
good translators that we are really interested in. Kussmaul (2007), for instance, is
particularly interested in successful, creative translation solutions.
There is a potential problem here, concerning the relation between virtual
processes and actual ones: our simplified models of virtual processes may be peda-
gogically useful, but they may not in fact correspond to what professionals actually
do. (For an early discussion of this issue, see Lörscher 1989.) Is there evidence, for
instance, that translators proceed through a logical series of binary decisions, as
suggested by some virtual models? (All translators? Or just some types?) Models
of virtual processes can thus generate predictions that can be tested.
Similarly, if reverse-engineered processes can be shown to lead, in theory, to a
given solution, to what extent might such a process correspond to an actual pro-
cess which did lead to this solution? What might be the significance of the real-
ization that two or more different processes might arrive at the same solution?
Might reverse-engineering even suggest more efficient ways of proceeding? Or is
the point of reverse-engineering models only to generate questions and then per-
haps hypotheses: could the process have been like this? And how might reverse-
engineered models relate to virtual ones?
Models of actual processes are likely to prove a good deal more complex, and
involve much more variation, than models of the first two process types, which
naturally tend to be idealized and simplified. Of course, any model may turn out
to be inaccurate, incomplete, or simply wrong, if it makes false predictions. Most
of the examples mentioned do not score highly on predictiveness, except in the
pedagogical sense of predicting their own usefulness in training, and in the weak
descriptive sense of purporting to be general models. None seem to be strongly
predictive, in the sense of a model exposing itself to the risk of being falsified by
predictions that are not realized.
In terms of testability, we could say that models of reverse-engineered process-
es could be checked to some extent by retrospective interviews, or tested against
other kinds of triangulated data elicited during the process itself, as indeed is done
(e.g., Englund Dimitrova 2005). However, there is no guarantee that models of a
reconstructed process necessarily represent the actual process which terminated in
a given translation.
Models of the sociological event have been rarer than those of the translation act;
in many cases they are only implicit. My first, simple example is from the profes-
sional translator Robert Bly (1984), who wrote a classic essay about his own way of
translating poetry (specifically, a Rilke poem). He summarized this as eight steps.
16 Andrew Chesterman
9. Concluding remarks
If models incorporate causality in some way, they can make strong predictions and
thus be tested. Some of the cognitive models we glanced at above do include causal
factors: Gile’s Effort Model, for instance; and to some extent Levý’s decision-mak-
ing model, in that decisions are seen to be partly based on (i.e., caused by) cost-
benefit calculations on the pros and cons of various alternatives. But most of the
ones we have looked at are non-causal, so far; at least, they are not explicitly causal.
Retrospective methods such as interviews or error analysis may suggest causal
factors, but these are inferred via the logic of reverse engineering. The challenge
is then to test a reverse-engineered model against real-life translation, against an
actual process.
A similar challenge faces scholars who draw pedagogical inferences from their
descriptive models, as these inferences also imply a cause-and-effect dimension.
But we need more empirical tests of the validity of such inferences: whether, for
instance, translation quality would be improved by the adoption of such-and-such
a model as a basis for translator training. We do not know enough about how pre-
dictive our models might be.
Indeed, some scholars may feel that we are not at all ready yet to build any-
thing more than very simple models, and that research should proceed by gener-
ating and testing specific hypotheses: if supported, these hypotheses may then be
synthesized into a complex (and causal) model. One recurring observation is that
many studies indicate a great deal of variation across subjects, both in translation
and in interpreting. This makes generalization risky; but without generalizations
we cannot build models.
If one does set out to construct a model, I conclude, it is important to be clear
about what kind of process one is seeking to model; i.e., about what kind of model
one is building, based on what facts and/or hypotheses, and what assumptions.
What testable hypotheses might the model generate? What kind of predictions is
Models of what processes? 19
the model capable of making, if any? If it can make explicit predictions, it can be
tested, and maybe progress can be made.
References
Bly, Robert. 1984. “The Eight Stages of Translation.” In Translation. Literary, Linguistic and
Philosophical Perspectives, ed. by William Frawley, 67–89. Newark, NJ: University of
Delaware Press.
Buzelin, Hélène. 2005. “Unexpected Allies: How Latour’s Network Theory could Complement
Bourdieusian Analyses in Translation Studies.” The Translator 11 (2): 193–218.
DOI: 10.1080/13556509.2005.10799198
Carl, Michael. 2012. “A Computational Cognitive Model of Human Translation Processes.” In
Emerging Applications of Natural Language Processing: Concepts and New Research, ed. by
Sivaji Bandyopadhyay, Sudip K. Naskar, and Asif Ekbal, 110–128. Hershey, PA: IGI Global.
Chesterman, Andrew. 2008. “On Explanation.” In Beyond Descriptive Studies. Investigations in
Homage to Gideon Toury, ed. by Anthony Pym, Miriam Schlesinger, and Daniel Simeoni,
363–379. Amsterdam: John Benjamins. DOI: 10.1075/btl.75.27che
Englund Dimitrova, Birgitta. 2005. Expertise and Explicitation in the Translation Process.
Amsterdam: John Benjamins. DOI: 10.1075/btl.64
Gile, Daniel. 2009. Basic Concepts and Models for Interpreter and Translator Training. Revised
edition. Amsterdam: John Benjamins. DOI: 10.1075/btl.8
Göpferich, Susanne. 2009. “Towards a Model of Translation Competence and its Acquisition:
The Longitudinal Study TransComp.” In Behind the Mind: Methods, Models and Results in
Translation Process Research, ed. by Susanne Göpferich, Arnt Lykke Jakobsen, and Inger M.
Mees, 11–37. Copenhagen: Samfundslitteratur.
Holz-Mänttäri, Justa. 1984. Translatorisches Handeln. Helsinki: Suomalainen Tiedeakatemia.
Hönig, Hans. 1995. Konstruktives Übersetzen. Tübingen: Stauffenberg.
Jääskeläinen, Riitta. 1999. Tapping the Process: An Explorative Study of the Cognitive and Affective
Factors Involved in Translating. Joensuu: University of Joensuu.
Jakobsen, Arnt Lykke. 2011. “Tracking Translators’ Keystrokes and Eye Movements with
Translog.” In Methods and Strategies of Process Research, ed. by Cecilia Alvstad, Adelina
Hild, and Elisabet Tiselius, 37–55. Amsterdam: John Benjamins. DOI: 10.1075/btl.94.06jak
Jones, Francis R. 2011. Poetry Translating as Expert Action: Processes, Priorities and Networks.
Amsterdam: John Benjamins. DOI: 10.1075/btl.93
Kinnunen, Tuija, and Kaisa Koskinen (eds). 2010. Translators’ Agency. Tampere: Tampere
University Press. http://urn.fi/urn:isbn:978-951-44-8082-9.
Koskinen, Kaisa. 2008. Translating Institutions. Manchester: St. Jerome.
Krings, Hans P. 1986. Was in den Köpfen von Übersetzern vorgeht: eine empirische Untersuchung
zur Struktur des Übersetzungsprozesses an fortgeschrittenen Französischlernern. Tübingen:
Gunter Narr.
Künzli, Alexander, and Maureen Ehrensberger-Dow. 2011. “Innovative Subtitling. A Reception
Study.” In Methods and Strategies of Process Research, ed. by Cecilia Alvstad, Adelina Hild,
and Elisabet Tiselius, 187–200. Amsterdam: John Benjamins. DOI: 10.1075/btl.94.14kun
20 Andrew Chesterman
1. Introduction
Guided by the question of what is a translation, Toury (1995: 144) put forward the
notion of “assumed translation” and postulated that a translation is a text for which
“there [is assumed to be] another text, in another culture/language, which has both
chronological and logical priority over it.” Toury (1995: 145) further suggested that
the latter postulate “entails the assumption that the process whereby [the translation]
came into being involved the transference from the assumed source text of certain
features that the two now share.” Toury (1995: 145) further argued that the latter two
postulates imply “that there are accountable relationships which tie [the translation]
to its assumed original, an obvious function of that which the two texts allegedly
1. The authors would like to thank the editors and the reviewers for their valuable and insightful
comments and suggestions.
doi 10.1075/bct.77.03sch
2015 © John Benjamins Publishing Company
22 Moritz Schaeffer and Michael Carl
share…” In other words, the relationship between source and target text can be de-
fined in terms of what both texts share — during the production of the target text.
In terms of the investigation of the process of translation, a central question
is therefore when and how the features which are shared between source and tar-
get text are identified, because identification of interlinguistic equivalence is a
distinctly translational process and one which could clearly distinguish reading
and writing for translation from reading and writing for comprehension or other
forms of rewriting.
Two models of the translation process, which have been adapted from the
models representing the bilingual lexicon have been referred to as “vertical ver-
sus horizontal translation” (De Groot 1997: 30). These two translation models dif-
fer in terms of the assumed cognitive processes. The vertical model is essentially
composed of two monolingual systems: one for decoding the source language and
one for encoding the captured meaning in the target language. The horizontal
translation model implies that items in the lexicons of the translator are linked via
shared memory representations and that cognitive processes during translation
are specific to the combination of both languages involved (see horizontal arrows
in Figure 1). In both translation models a source text is read and a target text is
produced, but they differ in terms of the processes which are specific to transla-
tion. The X in equation (1) represents these processes:
(1) Translation = Source Text Reading + Target Text Writing + X
On the one hand, vertical translation assumes that the source text is parsed and
abstracted into more or less language specific concepts or even non-linguistic
concepts and then re-expressed in the target language. Only after the transforma-
tion of input tokens into concepts does the translator proceed to translate these
concepts into lemmas which belong to the target language and finally into the
output. In vertical translation, there is no clear distinction between monolingual
and bilingual processing, because higher level conceptual representations or pos-
sibly non-linguistic conceptual representations are not necessarily specific to any
given language and may be involved in monolingual and bilingual processing
equally: whether a conceptual meaning, which is not specific to a given language,
is re-expressed in a different or the same language does not necessarily distinguish
translation from re-writing or a general problem solving activity. In other words,
the vertical translation method involves monolingual source language decoding
processes, conceptual, non-linguistic representations, monolingual target lan-
guage encoding and comparison between the propositional content of source and
target language utterances, i.e., problem solving (this process is represented by the
vertical arrows in Figure 1).
Shared representations and the translation process 23
Problem solving
Contextualised Contextualised
propositional propositional
content content
Shared representations
(limited context)
Monolingual
Monolingual
decoding
encoding
NP VP NP
In horizontal translation, items in the two languages are linked via shared rep-
resentations. In this case, a source text noun phrase — verb phrase — noun phrase
construction activates a cognitive representation which it shares with the target
text. In vertical translation, source text items are decoded once enough context
is available and the resulting propositional content is encoded into the target lan-
guage once translation problems are solved.
The data elicited by the use of Think Aloud Protocols (TAPs) in research on
translation processes has focused on the description of translation as the employ-
ment of translation strategies (e.g., Krings 1986; Lörscher 1991). While these strat-
egies may deal with the relationship between texts in different languages, this does
not necessarily imply that the cognitive processing during this kind of problem
solving is any different from the one employed when solving problems in rela-
tion to a (number of) text(s) in one language. Diamond and Shreve (2010) review
studies which search for neural and physiological correlates of translation and find
that the evidence does not seem to suggest that neural activity during translation
is different from neural activity during non-translational language use. The evi-
dence does however seem to suggest that translation is a complex problem solving
activity, but not a “primarily linguistic one” (Diamond and Shreve 2010: 309). The
variable X in equation 2 represents translation specific processes, if translation is
viewed as primarily a problem solving activity.
(2) X = problem solving
24 Moritz Schaeffer and Michael Carl
Horizontal translation, on the other hand, assumes that translation does not nec-
essarily occur via conceptual, but via shared semantic and syntactic representa-
tions: via “replacement of SL linguistic structures of various types (words, phrases,
clauses) by the corresponding TL” (De Groot 1997: 30). Paradis (1994: 329) sug-
gests that in the horizontal method, equivalence is established on different levels
by the application of automatic rules which have been learned during training. In
most cases, morphology, syntax, and semantics are the most likely locus of “trans-
ference” (Toury 1995: 145) in translation.
Horizontal translation does not necessarily require higher level represen-
tations, i.e., it is not necessary to fully comprehend either source or target text.
Paradis (1994: 332) argues that the horizontal translation method is subserved by
explicit metalinguistic knowledge, whereas the vertical translation method is sub-
served by implicit linguistic competence. The horizontal translation model implies
that the cognitive processes during translation are specific to the two languages in-
volved, because items in the lexicons of the translator are linked via shared memo-
ry representations. Source text reading activates, thus, shared representations and
the same shared representations serve as a basis for target language writing. This
process is represented in equation 3.
(3) X = activation of shared memory representations
In the next sections we reformulate the literal translation hypothesis and the
monitor model (Ivir 1981; Toury 1995; Tirkkonen-Condit 2005), and we provide
evidence for the horizontal translation model as a default mode in translation
processing. We look at a number of experiments from psycholinguistics which
indicate, on the one hand, that recall does not occur on the basis of verbatim
representations, i.e., that the target text is regenerated on the basis of heightened
activation of different aspects of the items to be recalled. On the other hand, prim-
ing experiments show that much of the translation activities can be explained by
shared syntactical and semantic representations. We further provide evidence that
shared representations play a central role during automatic translation. We will
then discuss a recursive model of translation which integrates strategic problem
solving processes with automated translation processes.
In other words, it is not possible to conclude whether the processes by which the
target text was produced were either horizontal or vertical if only the product is
taken into consideration. Whether a highly literal target text has been produced in
a very automatic manner or not cannot be determined on the basis of a compari-
son between source and target text alone.
3. Automatic translation
by controlled, conscious processing, but only on the basis of the output of the
automatic process, not on the basis of the aspects of the automatic process itself.
Only after the output has been received and evaluated by some kind of monitor
can these processes become consciously controlled. Therefore, we assume that the
monitor plays a prominent role in the translation process and that shared rep-
resentations are accessed very early during the process, i.e., that the horizontal
method is a very early process while the vertical method is a later process.
This understanding of the horizontal method differs from how Paradis (1994)
understands the horizontal method, because Paradis (1994: 332) assumes that ex-
plicit metalinguistic knowledge regarding translation equivalents forms the basis
for the horizontal method. Our view is that priming forms the basis for the hori-
zontal method: the influence of a previously processed item or structure on a sub-
sequently processed item or structure forms the basis for horizontal translation.
This is in line with Pickering and Ferreira (2008: 447, italics in the original) who
argue that priming “reflects the operation of an implicit learning mechanism,” i.e.,
that repeated exposure to primes creates long-lasting memories. So rather than
learning about translational equivalents, implicit mechanisms during repeated ex-
posure to source and target texts establishes shared representations in the transla-
tor’s long term memory.
When copying a text by re-typing it in the same language, “no lexical or structural
transfer is required” (Carl and Dragsted 2012: 129). Apart from corrections of ty-
pos and given adequate blind typing skills, reading and writing can thus occur
concurrently. In theory, it is possible to assume that during the process of copying
a text, the same representations which have been activated during reading are also
activated during writing. In other words, it is possible to assume that for copying,
lexical access is sufficient in order to allow for a reproduction of the text and that
higher levels of e.g., discourse integration are not necessary. But the data (eye-
tracking and keyboard logging) Carl and Dragsted (2012) present shows that par-
ticipants do reread the text to be copied even when there was no typo. In particular,
the data shows that regressions (re-reading of already fixated words) in the source
text are triggered when a particular item is literally easy to comprehend, but dif-
ficult to integrate contextually. Interestingly, re-reading is triggered by production
activity. Carl and Dragsted find re-reading occurred immediately before typing
corresponding items — both during copying and during translation. Regressions
during reading are normally associated with processes which integrate lexical
Shared representations and the translation process 27
representations into larger units of meaning (e.g., Reichle, Warren, and McConnel
2009). This suggests that even when source and target are literally the same (as
is the case when copying a text), non-literal or non-verbatim representations are
activated and serve as a basis for regeneration in the output.
Carl and Dragsted (2012) also confirmed the finding from Dragsted (2010)
that translators only look a few words ahead when reading and writing at the same
time. This means that production of a target text item normally depends on recall
of the source text item. When typing the translation of word N, the translator’s
eyes normally do not fixate word N, but a different word a few positions ahead in
the text. These findings from Carl and Dragsted (2012) and Dragsted (2010) are
in line with the Potter-Lombardi Regeneration hypothesis (Potter and Lombardi
1990; Lombardi and Potter 1992). Potter and Lombardi (1998: 265) argue that re-
call does not occur on the basis of a verbatim representation in memory, but on
heightened activation, i.e., primed representations, but maybe more importantly,
that rather than reproducing a verbatim representation of the sentence, partici-
pants regenerate the sentence on the basis of the primed representations. In sum,
even within the same language and when re-writing results in a verbatim copy of
the source text, literality is not a concept which adequately describes the processes
associated with this reading and writing behavior. Instead, we suggest describing
translation processes in terms of automaticity and shared representations.
In priming studies, participants are not aware of the effects of primes. Priming is
normally interpreted as evidence of implicit mechanisms and not explicit memory
retrieval. Potter and Lombardi (1998) presented native English participants first
with the target sentence and then with the prime sentence. The recall of the target
sentence was evaluated in terms of the influence from the prime sentence. For
example, when the target was a double-objective dative sentence, such as “The
prompt secretary wrote her boss a message every week,” the prime sentence was
a prepositional dative sentence, such as “The tycoon willed that mansion to his
young nephew grudgingly” (Potter and Lombardi 1998: 268). They found that par-
ticipants were significantly more likely to recall a sentence with the same syntactic
structure as that of the prime sentence than with an alternative syntactic structure.
28 Moritz Schaeffer and Michael Carl
6. Bilingual priming
The bulk of structural priming studies investigates language effects within one lan-
guage. Fewer studies have presented primes and targets in different languages. They
are, though, consistent in their findings: the results from a number of studies using
two languages (Heydel and Murray 2000; Meijer and Fox Tree 2003; Loebell and
Bock 2003; Desmet and Declercq 2006; Salamoura and Williams 2007; Schoonbaert,
Hartsuiker, and Pickering 2007; Kantola and van Gompel 2011) seem to suggest
that if particular features of syntax exist in both of the bilingual’s languages, then
they are likely to prime across languages. These studies used genetically and ty-
pologically close languages: English, French, German, Dutch, Spanish, and Greek.
These findings are in line with Hartsuiker, Pickering, and Veltkamp’s (2004) shared
syntax account. The authors propose that, where the syntax of the two languages of
the bilingual are sufficiently similar, they share a representation. In addition, there
is ample evidence for semantic representations which are shared across languages.
Hartsuiker et al.’s shared syntax account (2004) extends Pickering and Branigan’s
(1998) monolingual model of how lemmas are represented cognitively. Hartsuiker
et al. (2004: 481) assume that entries in the bilingual mental lexicon consist of
“combinatorial nodes” which connect lemmas, concepts, word forms, and syntac-
tic information. This means that these nodes are fairly abstract and general and
far from verbatim representations and this means also that many different words
from both languages share nodes at the respective levels of lemma, concept, word
form, and syntactic information: “…such combinatorial nodes are connected to
all words with the relevant properties, irrespective of language” (Hartsuiker et al.
2004: 481). However, aspects which are shared between the two languages in the
bilingual mind are specific to the language combination, because not all languages
always share the same aspects. This makes shared representations very different
from non-linguistic conceptual representations which can be shared by all lan-
guages. It is very likely that transference during horizontal translation occurs via
these shared representations on different levels (semantics, syntax, etc.): accessing
one relatively abstract and general shared representation or combinatorial node,
rather than two language specific verbatim representations is faster and more effi-
cient. Regeneration (Potter and Lombardi 1990; Lombardi and Potter 1992) on the
basis of combinatorial nodes with a maximum amount of overlap between source
and target text resembles copying rather than problem solving: the source text is
read, shared representations are primed and the same shared representations serve
as a basis for the regeneration in the target language.
Meijer and Fox Tree (2003) report from a study which used recall as a measure of
priming, but participants in this study were Spanish-English bilinguals and prime
and target sentences were presented in two different languages. The English target
sentences were constructed with verbs which allowed both double-object and prep-
ositional object constructions, such as “the waitress will bring the noisy customers
a tray of drinks” which can also be expressed as “the waitress will bring a tray of
drinks to the noisy customers” (Meijer and Fox Tree 2003: 187). The Spanish prime
sentences always had a prepositional object. And after the presentation of Spanish
30 Moritz Schaeffer and Michael Carl
the process of translation as long as the emerging target text is acceptable but also
that shared representations inhibit the process of translation when the emerging
target text does not conform to target norms. The stimuli in the study by Vandepitte
and Hartsuiker were tightly controlled in terms of the acceptability of a target
text with the same agent and predicate, i.e., using the same agent and predicate
in the translation of critical items was not acceptable, while this was acceptable
in the translation of control items. Vandepitte and Hartsuiker found significantly
longer translation (writing) times for critical items and interpreted these results
as evidence for cognitive effort during translation of items with cross-linguis-
tic differences as compared to items with cross-linguistic sameness. Vandepitte
and Hartsuiker instructed participants “to type the first translation they found”
(2011: 76, italics in the original). It is therefore possible that not only the nature of
the stimuli, but also the translation brief, i.e., the reading or translation purpose,
has an effect on how the stimuli are encoded. That how the source text is encoded
is a matter of choice, strategy or habit was already suggested by Isham (1994) on
the basis of a study which investigated recall in simultaneous interpreters.
Our study aimed to test whether translation primes semantic and syntacti-
cal representations, as described by Vandepitte and Hartsuiker (2011: 81), because
this could suggest that shared representations facilitate the process of translation.
We wanted to test whether reading a sentence in one language and translating it
into another language primes representations shared between the two languages
more than reading only for comprehension does. We used verbatim recall as a
measure of priming (Potter and Lombardi 1998; Meijer and Fox Tree 2003). The
source text was the target and the translation was the prime.
10.1 Participants
Six translators who habitually (at least 5 years’ experience) translated into German
from English and five translators who habitually translated into French from
English were selected from registers available at professional bodies for translators
(IOL, ITI, and AIIC). All participants had English as their L2. No other require-
ment was taken into account for the selection of participants. The results from one
French participant had to be excluded, because she recalled the target text instead
of the source text in the translation condition.
10.2 Procedure
Two excerpts (844 and 794 words) from The Old Man and the Sea (Hemingway
1975) were presented one sentence at a time using ePrime. The instruction for the
translation condition was to produce a rough written draft of each sentence, not
32 Moritz Schaeffer and Michael Carl
a polished translation. During the translation condition, the source text sentence
remained visible while translating. To indicate that they were happy with their
translation, the participants hit the escape key. A new sentence or a recall prompt
would appear and the previous sentence would no longer be visible. The procedure
during the reading for comprehension condition was exactly the same apart from
the fact that participants did not translate. Recall prompts appeared every 1–5 sen-
tences at irregular intervals during both conditions. The recall prompt instructed
participants to recall as much as possible of the sentences they had either just read
for comprehension or which they just translated, and to type as many words from
the last recall prompt up to and including the last word before the current recall
prompt. It was stressed that recall should be as accurate as possible. There were 24
recall prompts in each condition. The order of presentation of the two excerpts
and tasks were counterbalanced.
10.3 Scoring
Scoring was carried out in two different ways. The lexical score represents lexical
aspects of the recall and disregards syntax. The total number of possible correct
scores is the number of words in each of the two texts (844 and 794). For the lexi-
cal score, only words which were recalled exactly in the same form were counted
as correct, with the exception of singular/plural. Figure 2 is an example of how
the lexical score was calculated: for this sentence, the score was 6/7 (86%), be-
cause [the trousers] was recalled as [his trousers]. For the lexical score, words were
counted as correct irrespective of word order (see Figure 3). For the sentence in
Figure 3, the lexical score was 10/14 (72%).
Source text To wake the girl from her sleep he went over to their little house.
Source text To wake the girl from her sleep he went over to their little house.
For the syntactical score, the recall and the source text were parsed using the
Berkeley Parser (Klein and Petrov 2007) and each parsed sentence was aligned
starting with the last word. Only words which were recalled in the same order
(starting from the end) and with the same part of speech tag were counted as cor-
rect. This is in line with Jarvella (1971) who also employed two recall measures:
recalled sentences aligned from the end and recall of words irrespective of order.
The total number of possible correct syntactical matches for the two texts was 1553
and 1598, which correspond to the total number of part of speech and phrasal tags
for each text. Only recalled words in exactly the same position (from the end of the
sentence) and with exactly the same part of speech and phrasal tags were counted
as correct disregarding lexical differences — the fact that [slid] was recalled as
[pushed] did not affect the recall score (see Figure 4). The syntactical score for this
sentence was 19/19 (100%).
10.4 Results
The lexical and syntactical scores for each participant in each condition in Table 1
are reported as mean correct recall as proportions of 1.
Within participants comparison showed that participants recalled on aver-
age significantly more after translation than after reading for comprehension on
both scores: Lexical recall (t(9) = 3.38, p < .008) and syntactical recall (t(9) = 3.34,
p < .009). Figure 6 shows the mean recall score for both measures and tasks; error
bars indicate the confidence interval (95%). The magnitude of the effect accord-
ing to Pearson’s correlation coefficient for the syntax score was large (r(8) = .584,
p > .076), though not significant and the magnitude of the effect for the lexical
score was also large (r(8) = .599, p > .067), but not significant.
34 Moritz Schaeffer and Michael Carl
Table 1. Mean recall scores (lexical and syntactical) for each participant
Lexical Syntactical
Participant Translation Reading Translation Reading
German 1 0.81 0.49 0.57 0.33
German 2 0.61 0.65 0.25 0.29
German 3 0.74 0.64 0.31 0.32
German 4 0.71 0.75 0.46 0.45
German 5 0.76 0.51 0.42 0.25
German 6 0.80 0.69 0.42 0.37
French 1 0.84 0.78 0.55 0.45
French 2 0.77 0.65 0.38 0.21
French 3 0.48 0.26 0.26 0.07
French 4 0.80 0.56 0.49 0.27
1.00
0.80
0.60
Mean
Translation
Reading
0.40
0.20
0.00
Lexis Syntax
RecallMeasure
Error Bars: 95% CI
Figure 5. Lexical and syntax mean recall scores for reading for comprehension versus
translation
10.5 Discussion
The significantly higher recall during translation from English into French and
into German than during English monolingual comprehension may be explained
through increased activation of shared representations during translation. The re-
sults suggest that during reading for comprehension, a single presentation of the
text activates monolingual English representations. During translation, monolin-
gual English representations are activated, but, where these representations coin-
cide with French or German target language representations, these shared aspects
are activated twice and hence facilitate recall and increase the overall recall scores.
Shared representations and the translation process 35
Contextualised
propositional content
Monitoring
Decoding Decoding
SL Semantics TL Semantics
SL Syntax NP VP NP TL Syntax
SL Morphology TL Morphology
SL Phonology TL Phonology
SL Phonetics TL Phonetics
SL Orthography TL Orthography
Encoding Encoding
Pickering and Branigan’s (1998) model of the lemma stratum predicts a lexi-
cal boost when primes and targets have identical verbs. Hartsuiker et al.’s model
(2004) also predicts a translation equivalent boost via shared semantic represen-
tation. This lexical boost is normally found from L1 to L2. During translation,
many words in a sentence are lexical translation equivalents and in this study, the
target text was the prime and was in the L1. In this sense, the higher recall is best
explained by reference to repeated access of the same representations.
Our experiment was not controlled in terms of the stimuli and is ecologically
valid to a certain extent: both tasks were relatively continuous in that the indi-
vidually-presented sentences in both tasks were part of a longer text. Traditional
priming studies require control stimuli in order to make sure that the use of a
particular structure in the target sentence is an effect of the prime structure. The
control stimuli in, for example, Potter and Lombardi (1998) were sentences with a
structure which could not prime the target, i.e., the authors used intransitive sen-
tences as controls for transitive target sentences. In translation, it is more compli-
cated to find control stimuli because only syntactic structures which cannot have
a shared representation can be said not to prime. In other words, only sentences
which cannot be translated using the same cognitive representation can be said
not to prime and therefore act as control stimuli. But to decide whether something
36 Moritz Schaeffer and Michael Carl
Rather than describing the cognitive processes during translation in terms of the
literality of the target text in relation to the source text as in Tirkkonen-Condit,
Mäkisalo, and Immonen (2008), we propose to describe the literality of the pro-
cess in terms of shared representations and automaticity. Englund Dimitrova
(2005: 53) argues that “in the translation process there is an alternation between
automated and strategic problem-solving processing.” Automaticity is here in-
dexed by pauses in the typing activity and/or by problem indicators in the TAPs
(Englund Dimitrova 2005: 156). We propose that future studies could measure
automaticity in terms of the degree to which source text reading and target text
production occur at the same time. Hartsuiker et al. (2004: 487) predict that trans-
lation and simultaneous interpretation “will be facilitated when the two languages
have parallel syntactic structures; this is because processing the source structure
will act as a syntactic prime that leaves the target structure with residual activa-
tion.” In other words, the closer the cognitive alignment between source and target
text representations, the more efficient is the process. We propose that instances
of concurrent reading and writing during translation are indicative of automatic
processes and shared representations. The resulting target text may or may not be
considered a formally corresponding translation according to equivalence theo-
ries such as that proposed by Catford (1965).
In the following section, we will explain how and why both the horizontal and
the vertical model can be integrated into one model of translation which accounts
Shared representations and the translation process 37
The model we propose is very similar to the one proposed by Christoffels and de
Groot (2005: 474, Figure 22.2 b). Christoffels and de Groot (2005: 471, our brack-
ets) argue that “…there is some [indirect] evidence to suggest that, in addition to
meaning-based [vertical] interpreting, transcoding [horizontal] also takes place.”
De Groot (2011: 324) proposes separate input and output lexicons for source and
target language systems to allow for target text production monitoring at the same
time as source text comprehension. In our model, vertical processes act as the
monitor for target text production by the horizontal processes. In other words, we
do not assume separate input and output lexicons for source and target language
(i.e., 4 lexicons). During decoding, both horizontal and vertical processes are al-
ways active at the same time. We assume that the horizontal process is an early
process while the vertical processes depend on context which becomes available
later, as processing advances in the chunk or text. Early during source text reading,
shared representations are activated which then serve as a basis for regeneration
in the target language (circles in Figure 6). As long as the target text being pro-
duced conforms to the target norms and contextual considerations of the vertical
processes, regeneration on the basis of shared representations is not interrupted.
But when the target text is not acceptable, the interim translation, either kept in
working memory or already partially produced as target text, is adapted to target
norms by vertical encoding processes (vertical arrows in Figure 6). Adaptation
during encoding to target norms is an essentially monolingual process. In other
words, vertical processes, i.e., monolingual encoding and decoding processes act
as monitor. In this sense, De Groot (2011: 322) argues that for “…such monitoring
to take place no special mechanism needs to be postulated…” As represented in
Figure 6, horizontal processes access shared representations (NP/VP/NP in this
case) automatically and early during the process. Vertical processes monitor out-
put from the horizontal automatic processes.
The shared syntax account (Hartsuiker et al. 2004) predicts that syntax is not
separate for each language. The neurofunctional model proposed by Paradis also
predicts:
…each language is represented as a subsystem of the language system. When one
language is selected for expression, the activation threshold of the other language
38 Moritz Schaeffer and Michael Carl
While entries in each of the different aspects of the language system may belong
to a different language which can be inhibited selectively, there is only one lan-
guage system — with one subsystem per language of the bilingual. In line with
Paradis (1994: 321) our model predicts that the monolingual monitoring activity
and the automatic default procedure use one language system and balance the
activation threshold for source and target language since during concurrent read-
ing and writing the activation threshold for both source and target language is
similarly high. When the translator is not reading and writing at the same time,
the activation threshold for the non-active language is higher than for the active
language. In both these modes, both languages are never completely inhibited and
always active to a certain extent as described by the “monolingual-bilingual mode
continuum” (Grosjean 1997: 168).
The vertical processes control the acceptability of the target text, but also need
to monitor equivalence. These processes need to assess whether source and target
are the same. In order to do this, the monitoring processes need to have a tertium
comparationis: the monitoring processes assess whether source and target texts
are compatible in terms of propositional content and shared conceptual represen-
tations — when the activation threshold for each language is relatively balanced.
The model consists of a recursive cycle which integrates horizontal and verti-
cal source and target language processes: the monitor needs to compare whether
the source is the same as the target, but it is equally important to make sure that
the target is the same as the source. Vertical processes access the output from the
automatic default procedure recursively in both the source and the target language
and monitor consistency as the context during translation production increases.
13. Conclusion
In this chapter, on the basis of a translation priming study, we revised the lit-
eral translation hypothesis and the monitor model (Tirkkonen-Condit 2005) in
terms of translation processes: an automatic translation procedure produces de-
fault translations on the basis of shared representations (horizontal translation).
Vertical processes interrupt this automatic process if the resulting target text is not
acceptable. Malmkjær (2005) argues that the evidence put forward by Tirkkonen-
Condit (2005), Eskola (2004) and Gellerstam (1996) suggests that the literal trans-
lation hypothesis might be cognitively determined and that it is not restricted to
one language combination of closely related languages. The latter combination of
Shared representations and the translation process 39
factors makes it therefore a good candidate for a phenomenon which can be ob-
served across languages and contexts. We explained the literal translation hypoth-
esis in terms of the translator’s bilingual lexicon. This increases the likelihood that
automatic translation on the basis of shared representations is a process which
occurs across language combinations.
In sum, our model describes how the relationship between source and target
text can be defined in terms of what both texts share — during the production of
the target text.
References
Kantola, Leila, and Roger P.G. van Gompel. 2011. “Between- and Within-Language Priming is
the Same: Evidence for Shared Bilingual Syntactic Representations.” Memory and Cognition
39: 276–290. DOI: 10.3758/s13421-010-0016-5
Keatley, Catharine W., John A. Spinks, and Beatrice De Gelder. 1994. “Asymmetrical Cross-
language Priming Effects.” Memory and Cognition 22: 70–84. DOI: 10.3758/BF03202763
Klein, Dan, and Slav Petrov. 2007. “Improved Inference for Unlexicalized Parsing.” In Proceedings
of Human Language Technology Conference of the North American Chapter of the Association
of Computational Linguistics (HLT-NAACL), Rochester, New York, 22–27 April 2007, ed. by
Candace L. Sidner, Tanja Schultz, Matthew Stone, and Cheng Xiang Zhai. Rochester, NY:
The Association of Computational Linguistics.
Krings, Hans P., 1986. Was in den Köpfen von Übersetzern vorgeht:eine empirische Untersuchung
zur Struktur des Übersetzungsprozesses an fortgeschrittenen Französischlernern [What
Happens in the Heads of Translators: An Empirical Investigation of the Structure of the
Translation Process in Advanced Students of French]. Tübingen: Gunter Narr.
Loebell, Helga, and Kathryn Bock. 2003. “Structural Priming across Languages.” Linguistics 41:
791–824. DOI: 10.1515/ling.2003.026
Lombardi, Linda, and Mary C. Potter. 1992. “The Regeneration of Syntax in Short Term Memory.”
Journal of Memory and Language 6: 713–733. DOI: 10.1016/0749-596X(92)90036-W
Lörscher, Wolfgang. 1991. Translation Performance, Translation Process and Translation
Strategies. A Psycholinguistic Investigation. Tübingen: Gunter Narr.
Macizo, Pedro, and Maria T. Bajo. 2006. “Reading for Repetition and Reading for Translation:
Do They Involve the Same Processes?” Cognition 99: 1–34.
DOI: 10.1016/j.cognition.2004.09.012
Malmkjær, Kirsten. 2005. “Norms and Nature in Translation Studies.” Synaps 16: 13–19.
Meijer, Paul J.A., and Jean E. Fox Tree. 2003. “Building Syntactic Structures in Speaking: A
Bilingual Exploration.” Experimental Psychology 50: 184–195.
DOI: 10.1026//1617-3169.50.3.184
Paradis, Michel. 2004. A Neurolinguistic Theory of Bilingualism. Amsterdam: John Benjamins.
DOI: 10.1075/sibil.18
Paradis, Michel. 1994. “Toward a Neurolinguistic Theory of Simultaneous Translation: The
Framework.” International Journal of Psycholinguistics, 10 (3): 319–335.
Pickering, Martin J., and Holly P. Branigan. 1998. “The Representation of Verbs: Evidence from
Syntactic Priming in Language Production.” Journal of Memory and Language 39: 633–651.
DOI: 10.1006/jmla.1998.2592
Pickering, Martin J., and Victor S. Ferreira. 2008. “Structural Priming: a Critical Review.”
Psychological Bulletin 134: 427–459. DOI: 10.1037/0033-2909.134.3.427
Potter, Mary C., and Linda Lombardi. 1990. “Regeneration in the Short-Term Recall of Sentences.”
Journal of Memory and Language 29: 633–654. DOI: 10.1016/0749-596X(90)90042-X
Potter, Mary C., and Linda Lombardi. 1998. “Syntactic Priming in Immediate Recall of
Sentences.” Journal of Memory and Language 38: 265–282. DOI: 10.1006/jmla.1997.2546
Reichle, Erik D., Tessa Warren, and Kerry McConnell. 2009. “Using E-Z Reader to Model
the Effects of Higher Level Language Processing on Eye Movements during Reading.”
Psychonomic Bulletin and Review 16: 1–21. DOI: 10.3758/PBR.16.1.1
Ruiz, Carmen, Natalia Paredes, Pedro Macizo, and Maria T. Bajo. 2008. “Activation of Lexical
and Syntactic Target Language Properties in Translation.” Acta Psychologica 128: 490–500.
DOI: 10.1016/j.actpsy.2007.08.004
42 Moritz Schaeffer and Michael Carl
Salamoura, Angeliki, and John N. Williams. 2007. “Processing Verb Argument Structure
across Languages: Evidence for Shared Representations in the Bilingual Lexicon.” Applied
Psycholinguistics 28: 627–660. DOI: 10.1017/S0142716407070348
Schoonbaert, Sofie, Robert J. Hartsuiker, and Martin J. Pickering. 2007. “The Representation of
Lexical and Syntactic Information in Bilinguals: Evidence from Syntactic Priming.” Journal
of Memory and Language 56: 153–171. DOI: 10.1016/j.jml.2006.10.002
Schoonbaert, Sofie, Phillip J. Holcomb, Jonathan Grainger, and Robert J. Hartsuiker. 2011.
“Testing Asymmetries in Noncognate Translation Priming: Evidence from RTs and ERPs.”
Psychophysiology 48: 74–81. DOI: 10.1111/j.1469-8986.2010.01048.x
Tirkkonen-Condit, Sonja, Jukka Mäkisalo, and Sini Immonen. 2008. “The Translation Process
— Interplay between Literal Rendering and a Search for Sense.” Across Languages and
Cultures 9: 1–15. DOI: 10.1556/Acr.9.2008.1.1
Tirkkonen-Condit, Sonja. 2005. “The Monitor Model Revisited: Evidence from Process
Research.” Meta 50: 405–414. DOI: 10.7202/010990ar
Toury, Gideon. 1995. Descriptive Translation Studies and Beyond. Amsterdam: John Benjamins.
DOI: 10.1075/btl.4
Vandepitte, Sonia, and Robert J. Hartsuiker. 2011. “Metonymic Language Use as a Student
Translation Problem: Towards a Controlled Psycholinguistic Investigation.” In Methods and
Strategies of Process Research: Integrative Approaches in Translation Studies, ed. by Cecilia
Alvstad, Adelina Hild, and Elisabet Tiselius, 67–92. Amsterdam: John Benjamins.
DOI: 10.1075/btl.94.08van
ELF speakers’ restricted power of expression
Implications for interpreters’ processing
Michaela Albl-Mikasa
1. Introduction
The study of English as a lingua franca (ELF) is an even younger discipline than
that of interpreting studies. Although both belong under the wider umbrella of ap-
plied linguistics, it is only very recently that the two fields have begun to consider
the mutual relevance of each other (see Cook 2012). With studies so far having
concentrated on the effect of non-native accents on the interpreting task (espe-
cially McAllister 2000; Sabatini 2000; Cheung 2003; Kurz 2008) and the advantage
of having the non-native speaker’s first language as one of the interpreter’s work-
ing languages (see Taylor 1989; Basel 2002; Kurz and Basel 2009), consideration of
doi 10.1075/bct.77.04alb
2015 © John Benjamins Publishing Company
44 Michaela Albl-Mikasa
the wider impact of ELF on interpreting has only just started to take off (see Albl-
Mikasa 2010; 2012a; 2013b; Reithofer 2010; 2013). It has become clear from this
research that interpreters are disconcerted by the unprecedented spread of English
as a global lingua franca. However, it is not simply a potential existential threat to
the profession — ELF developments threaten the very multilingualism on which
it thrives, (see Cook 2012: 256) — that gives rise to major concerns on their part.
In fact, changing market conditions — such as more conferences in English only,
more ‘bilingual ELF conferences’ (namely, interpreting only between ELF and one
local language instead of multilingual conferences with a greater number of dif-
ferent language booths), less appreciation of interpreter services by customers, a
greater pressure to deliver higher quality to demonstrate the added value of those
services, and so forth (see Albl-Mikasa 2010) — are not the only disturbing factors.
It is also, and quite fundamentally, on the micro-level of working conditions or ac-
tual processing that interpreters face new challenges. In my questionnaire survey
among 32 professional interpreters on the implications of ELF on the interpret-
ing profession (Albl-Mikasa 2010), the respondents complained that their work
was increasingly strenuous and tiring due to additional cognitive load; faced with
ELF speakers, they had to expend extra capacity in source text comprehension (to
grasp foreign accents and recover unfamiliar expressions, to resolve unorthodox
syntactic structures and compensate for the lack of pragmatic fluency on the part
of the non-native speaker); faced with non-native English listeners, they had to do
so in the target text production (for accommodation to differing proficiency levels
of non-native addressees). As a result, in a number of qualitative statements to
open questions, the interpreters expressed that job satisfaction was on the decline
(2010: 141). Furthermore, the restricted power of expression on the part of ELF
speakers was identified as a major trouble spot (see Albl-Mikasa 2013a).
It is arguable that the perceived burden is simply a reaction to new circum-
stances, which deviate from well-trodden paths, internalized ways, and proven
experience, and that such perceptions are exacerbated by the (as yet) unavailability
of special training measures to help would-be interpreters build up the required
compensatory strategies geared to ELF conditions. It needs to be noted, however,
that the complaints voiced in the questionnaire survey (see Albl-Mikasa 2010)
and in the in-depth interview corpus of 10 professional conference interpreters
(see Albl-Mikasa 2012b) were made by very experienced professionals who have
been in the profession for decades rather than years; this means that they have had
the chance to become accustomed to a constantly growing number of non-native
English speakers and audiences in conference settings by gradually working their
way through the emerging challenges. The complaints about taxed resources, ex-
haustion, and declining job satisfaction due to the ‘ELF factor’ ran across the whole
spectrum of interpreters with a working experience of 10 to 40 years. Whichever
ELF speakers’ restricted power of expression 45
way one may look at the situation, it is a difficult one for interpreters and begs the
question as to what exactly it is that makes their work in settings with ELF speak-
ers and listeners especially challenging. The focus of this chapter is on exploring
the nature of processing difficulties arising from ELF.
To understand why interpreters keep complaining about ELF and the exact par-
ticularities typical of non-native English speakers that impinge upon the interpret-
ers’ processing, and thus to understand the consequences for the transfer process
and the resulting difficulties and possible new strategic requirements, one has to
bear in mind that interpreters work under conditions that are different from those
of monolingual communication settings in which ELF is usually used (and which
ELF speakers’ restricted power of expression 47
have been the basis of the specialist ELF literature so far). As Braun and Kohn
point out, under bilingual conditions people’s “routinely practiced comprehen-
sion and production skills are on unfamiliar territory” (2012: 190). The particular
and aggravating processing conditions in mediated multilingual communication
(as opposed to unmediated monolingual communication) are described in some
detail by various authors:
1. Kohn (1990: 109–112; 2004: 220–223) outlines the three sources of conflict
inherent to translation: (a) the continuing presence of the source text, (b) the
lack of semantic autonomy on the part of the translator, and (c) the need for
deeper reception and analysis as a basis for choosing adequate expressions in
target text production.
2. According to Anna Kucharska (2009: 14–18), interpreting typically takes place
in “deficient situations” due to a great number of mode-, time-, interpreting-
and situation-specific restrictions (see also Braun 2004).
3. Furthermore, what makes interpreting an especially complex task is the cogni-
tive need of interpreters to focus their attention — contrary to monolingual
processing (see Brown and Yule’s Sweden example in Section 2 above) — on
all levels of mental representation and processing at the same time. Focus on
the level of linguistic surface structures is necessary for the re-textualization of
linguistic expressions and equivalence effects in the target language; focus on
the level of the propositional text base for complete rendering and better mem-
orization of details; and focus on the level of the mental model for a grasp of the
overall meaning or overarching coherence and control of the communicative
situation (see Albl-Mikasa 2011: 56–57).
These particular bilingual conditions of mediated communication are the backdrop
against which we need to look at ELF as an additional aggravating factor. It is in this
context that it becomes clear why perceptions of the efficiency of ELF communica-
tion differ. While ELF researchers have collected ample evidence emphasizing the
effectiveness of ELF communication (see Seidlhofer 2011), findings in interpreting
studies are much less positive. On one hand, the factors facilitating and support-
ing monolingual ELF communication in largely interactional settings do not apply
to mediated communication when the presentation of monologic speeches leaves
no room for meaning negotiation and co-construction and when the requirement
of their full and detailed rendering does not allow for let-it-pass strategies (see
Reithofer 2010: 149). On the other hand, as Karin Reithofer (2013) finds in a larger-
scale study, understanding of source speech propositional content in conference
settings can be significantly higher among conference participants listening to the
interpreter than those listening to the non-native English original. As far as the
interpreters are concerned, respondents to my 2010 questionnaire survey seemed
48 Michaela Albl-Mikasa
to agree that non-native English speakers can be easier to interpret for than native
speakers, especially when they speak slowly and use simple structures. At the same
time, a clear majority (69%) of them favored native-speaker source text producers
who were felt to be not necessarily easier but better or more pleasant to interpret
for, as their speech flow, speech structure, and precision in expression facilitated
the interpreters’ task. In contrast, ELF speaker-induced pronunciation, structure,
and expression-based comprehension problems were perceived to impede the un-
derlying processes of anticipation and inferencing (see Albl-Mikasa 2010: 136).
To gain further insight into what might be behind interpreter complaints and
perceptions regarding input by ELF speakers, I should like to draw on a small-
scale case study. The study is too restricted in scope to be representative or to allow
for broader generalization of the results. Moreover, it involves a student interpreter
rather than a professional. The reason is that this study, carried out for a B.A. thesis
under my supervision (see Dschulnigg and Hansen 2011), was conceived as a kind
of pilot study for a larger study involving professionals.1 As such, it studied a stu-
dent in order to amplify a problem that the professional interpreters interviewed
had previously identified as restrictions in the power of expression on the part of
ELF speakers (see Albl-Mikasa 2013a). The idea was to investigate the basic dif-
ficulties posed by these restrictions and to identify different types of restriction
before testing their validity with professional interpreters.
In the small-scale B.A. study, three German native-speaker students in the B.A. in
Translation program who did not have English as one of their study languages but
had taken an ELP (English for Language Professionals) course were asked to give
a five-minute speech in English on one of the following randomly assigned topics:
the nuclear disaster in Japan, the Libyan crisis, and WikiLeaks. They were given a
sheet with topic-related content notes in German and had 20 minutes to prepare
and look up some of the necessary terminology on the Internet or in a diction-
ary. All three speeches were simultaneously translated into German by the same
1. I did not ask Nadine Dschulnigg and Leslie Hansen to co-author this chapter (which I nor-
mally do in the case of publication of M.A. thesis results), because the setup, study design, and
guiding ideas behind the thesis resulted from a joint discussion, in which my input was salient,
and because the classification of interpreter stumbling blocks I propose in this chapter does not
figure in the thesis. Full credit goes, of course, to the two students for collecting and discussing
highly interesting data and for rendering a clearly relevant B.A. account.
ELF speakers’ restricted power of expression 49
In the example in table 2, the speaker points out in the retrospective interview that the interpreter
50 Michaela Albl-Mikasa
In the following, I will discuss one or two examples for each category.
In the example in Table 2, on the nuclear disaster in Japan, the speaker points out in
the retrospective interview that the interpreter did not render the idea of a benevo-
lent gesture on the part of the emperor, which she had wanted to express by saying
“he was really kind with (eh) with the folk there, cause he has like a building.” The
interpreter, by contrast, reports that she was rather irritated by the phrase “the
emperior imperior like the king [clearing throat] the king of Japan.” Upon hearing
“king” as the last piece of information, she started off with “der König” (king), only
to rephrase it as “Kaiser” (emperor) when she realized it was the emperor that was
being referred to. All this, however, distracted her to such an extent that she did not
hear the rest of the proposition and omitted it. It should be added that the omitted
part of the proposition was not well expressed either and would hardly have given
away the meaning of the benevolent gesture intended by the speaker.
In the second part of the example in Table 2, the speaker points out that
“Badezimmer” (bathroom) as used by the interpreter does not convey what she
meant. At the same time, she admits to having been uncertain about what kind of
building or Badeanlage (bathhouse) it actually was that she had found described in
a newspaper some time ago. This is not the only concept, however, that the speaker
found difficult to convey. As seen from her use of “I don’t know” and “like,” she
Table 2. Example of incoherent input and resulting interpretation (Dschulnigg and Hansen
2011:87)
In the second part of the example in table 2, the speaker points out that “Badezimmer”
(bathroom) as used by the interpreter does not convey what she meant. At the same time, she
ELF speakers’ restricted power of expression 51
had general difficulty in expressing things. In fact, the interpreter emphasizes that
it was the very presentation of the content matter by the speaker that caused a
comprehension problem on her part, as she was unable to imagine what exactly
was meant and therefore summarized the source text, which may have led to the
omission of some details.
Looking at the example in Table 2, it becomes clear how difficult it must be for
the interpreter to figure out the exact meaning of an utterance on the basis of poor
source text input. It is almost impossible to infer the actual circumstances from the
very general term “bath” unless one has the corresponding background knowledge;
this is even more so if such a term is embedded in an overwhelmingly unclear pas-
sage (laughter, “take a half half bad,” etc.). In other words, the example highlights
the extent to which explicit input supports the interpreter’s task.
The example in Table 2 illustrates a case where the ELF speaker becomes con-
fused and is unable to complete her output production in an efficient way (evidence
that this is not a rare occurrence in ELF speech can be found in the TELF database).2
This can lead to serious comprehension and processing problems on the part of the
interpreter, who has to invest extra energy to clear up and reorganize the confused
and incoherent speaker output in order to reconstruct what the speaker meant or
would have said had she been able to express herself more adequately.
In the example in Table 3, concerning the crisis in Libya, the speaker criticizes in
the retrospective interview the interpreter’s use of “Versammlung” (assembly) in
the target text, which to her mind does not relate to the governmental body consti-
tuted by the opposition she intended to refer to. However, had the speaker known
that in English-speaking media this body is called ‘National Assembly’ and had
she used the phrase accordingly, the interpreter could have more easily come up
with the German equivalent Nationalrat (National Council). “General Assembly,”
by contrast, conjures up Generalversammlung (general meeting) in people’s minds.
From the fact that the interpreter hesitated about which word to use (saying “ein
eine Versammlung”), it seems likely that she must have figured out that the speaker
could not have been referring to a Generalversammlung; as a result, her strategy
was to use the more general term Versammlung.
2. Evidence can be found in TELF, the Tübingen English as a Lingua Franca corpus and data-
base (http://www. telf.uni-tuebingen.de).
up with the German equivalent Nationalrat (national council). “General Assembly,” by contrast,
conjures up Generalversammlung (general meeting) in people’s minds. From the fact that the
interpreter hesitated about which word to use (saying “ein eine Versammlung”), it seems likely
52 that
Michaela
she mustAlbl-Mikasa
have figured out that the speaker could not have been referring to a
Generalversammlung; as a result, her strategy was to use the more general term “Versammlung.”
Table 3. Example of imprecise input and resulting interpretation (Dschulnigg and
Hansen 2011: 89)
Table 3. Example of imprecise input and resulting interpretation (Dschulnigg and Hansen
What is also striking in the example in2011:89)
Table 3, but goes unnoticed by the speaker,
is her use of “foreign minister of justice.” The person in question is actually the
‘former’ minister of justice, which suggests that the words ‘former’ and ‘foreign’
53
were confused by the ELF speaker. For unknown reasons, this part is omitted by
the interpreter. It is possible, however, that the omission was caused by irritation
about the contradictory collocation and the fact that too much energy would have
been needed to draw on background knowledge and decide whether the former
minister of justice or of foreign affairs was meant by the speaker. By contrast, the
step from ‘former minister of justice’ to früherer Justizminister (former justice min-
ister) would have been straightforward, not taking up much capacity. Thus, while
omissions are often seen as ‘classical interpreter mistakes,’ some of them may not
be attributable to the interpreter, but to the (ELF) speaker.
This example demonstrates that the restricted ability of the speaker to express
precisely what she intends to convey can lead to general comprehension problems
or to the interpreter being puzzled and needing to invest additional resources to
draw on background knowledge and make decisions that are normally pre-empt-
ed by the speaker upon output production.
In the example in Table 4, concerning the Japanese nuclear disaster, the speaker
expresses her feeling in the retrospective interview that the target text expression
“schwimmen” (swim) is too positive a term for what she wanted to convey, namely,
“dass die Häuser ‘weggespült’ wurden” (that houses were washed away). The inter-
preter, however, upon hearing “flow away” (which conjures up the picture of cars
floating around), would have had to invest some effort to pin down the intended
meaning and come up with wegspülen. Had the speaker provided the standard
term ‘to wash away,’ this might have — at least in the given context — narrowed
down the meaning immediately and helped the interpreter retrieve the direct
equivalent wegspülen from her multilingual lexicon, without requiring any extra
inferential effort.
55
54 Michaela Albl-Mikasa
has to infer exactly what the speaker meant when using an imprecise term in order
to then activate the corresponding expression.
The examples given in the previous section point to two major processing con-
straints arising from ELF speaker source text production. First, incoherent or im-
precise input may hamper the interpreter’s comprehension processes and make it
difficult for her to grasp exactly what the speaker is trying to get at, thus under-
mining the interpreter’s decision-making directed at finding adequate target lan-
guage solutions. Second, the use of conventional (lexical or structural) expressions
by the speaker might make it easier for the interpreter immediately to retrieve
corresponding translation equivalents, while ELF speaker output can impede the
interpreter’s falling back on established transfer links. Even though they are from a
small study, these findings may be relevant when considered against the backdrop
of the cognitive discourse processing framework given in Section 2 above as well
as results from ELF research.
The two constraints are interesting in that they are closely related to the two
basic strategies of interpreting, namely, meaning-based and word-based interpret-
ing. The former is assumed to take place on the basis of the (mental) representa-
tion of a (more or less) fully comprehended source text meaning, whereas the
latter involves “the literal transposition of words or multiword units” (Christoffels
and de Groot 2005: 459). During their task, interpreters “inevitably shift from
one pattern to the other according to internal and external circumstances” (Gran
1989: 98). This is explained by the Heidelberg School cognitive discourse model
of interpreting (see Section 2 above) as follows: processing takes place in terms of
more global or more local coherence building, in other words, it is geared toward
higher (the mental model) or lower (the surface structures) levels of mental repre-
sentation (see Albl-Mikasa 2011). The first constraint refers to (global coherence-
oriented) meaning-based processing, which is dealt with in detail in the specialist
literature on interpreting where researchers describe how comprehension and
analysis of the source text are fundamental for target speech production (see Albl-
Mikasa 2007: 11–132) and serve as the basis for translational decisions (see the
third source of translational conflict specified by Kohn in Section 3 above). From
Gile’s Effort Models (2009), it can easily be inferred that successful interpreting is
undermined when additional resources have to be expended to grasp unfamiliar
source language accents and structures produced by ELF speakers. I should, there-
fore, like to concentrate on the second constraint in the following discussion.
ELF speakers’ restricted power of expression 55
The second constraint is situated at the level of surface structures because it fo-
cusses on (matching) linguistic items. Interpreters, not being experts in each and
every field, will, at times, have to forego deeper understanding (meaning-based in-
terpreting) and process at the level of the lexical and terminological input (word-
based interpreting). What is often called “transcoding” (see Kalina 1998: 118) is
not an emergency measure, but a full-fledged strategy in circumstances that do
not allow for meaning-based interpreting. Psycholinguistic research has produced
evidence that transcoding occurs and that it “implies the existence of direct links
between representations of the linguistic elements of one language and the cor-
responding representations in the other language” (Christoffels and de Groot
2005: 471). This is in line with reports by an experienced interpreter (I-1) saying
that, in many cases, interpreters choose to rely on activation of their “autopilot”
(see Albl-Mikasa 2012b: 75), that is, on surface-level transcoding and on transfer
trails carved out by high-frequency use rather than in-depth analysis. A similar
remark was made by another interpreter in the same interview corpus: “I do a lot
of patterns, technical collocate patterns. I don’t even try to understand them; I still
believe that I can do a good job” (I-2). Given time constraints and capacity limi-
tations, important strategies to relieve pressure include activation of translation
routines (see Göpferich 2008: 156) as well as the routine assembly of translation
equivalents, or “reflex-like equivalence relations” (see Kutz 2002: 190; my transla-
tion and emphasis). Such direct and easy links and paths between source text items
and target text solutions are either established through rehearsal and practice or
ad hoc. An example of how such routines emerge is provided in my interview
corpus (see Albl-Mikasa 2012b: 79). One of the interpreters (I-6) reported render-
ing the source language term “first pass rate” by “Erstläuferrate” even though the
English term rate is typically translated as Quote in German. The reasons given
for this decision were that this is the term the German experts in the audience
were familiar with and that it made interpreting easier because it took less effort
to cross over from rate to Rate rather than from rate to Quote. It greatly facilitates
the interpreter’s task to store such expressions, phrases, formulaic sequences, and
their corresponding target language expressions in a routine way, and to build up
a ‘mental translation memory’ to fall back on during interpreting. Based on the
following explanations, I hypothesize that this represents a major impediment for
interpreters when it comes to ELF speaker source text production.
Retrieval, in general, and from the ‘mental translation memory,’ in particular,
depends on conducive, activating input items, as explained by recourse to the “prin-
ciple of encoding specificity” (formulated by van Dijk and Kintsch 1983, who view
memory as consisting of the traces that result as a function of depth of processing):
56 Michaela Albl-Mikasa
In order to retrieve an item from memory, the retrieval cue must match, at least
partially, the encoded item. The cue must be appropriate for the encoding. The
efficiency of memory is determined not by either encoding or retrieval alone, but
by the encoding-retrieval interaction. This is the principle of encoding specificity
[…]. (van Dijk and Kintsch 1983: 334)
According to this principle, ease of retrieval, transfer processing, and source text
rendering are to some degree determined by the extent to which an input cue (as
provided by the source speech producer) matches the item encoded and stored by
the interpreter (see Albl-Mikasa 2007: 82–84). If ELF speaker output is not made
up of the idiomatic expressions the interpreter has learned and internalized, there
will be a mismatch between the activating (ELF speaker) input item and the item
encoded (by the interpreter).
Such a mismatch is likely to occur, if one considers the differences between na-
tive and non-native English speakers identified in the literature on ELF. Of native
speakers, it can typically be said that “a language user has available to him or her a
large number of semi-preconstructed phrases that constitute single choices, even
though they might appear to be analyzable into segments” (Sinclair 1991: 110). As
Barbara Seidlhofer (2009) points out, such prefabricated chunks serve to relieve
memory, ease processing, support memorization, and make for efficient retrieval.
If all interlocutors have access to conventionally prefabricated phrases, they do not
have to make sense of each linguistic component bit by bit. In the case of prover-
bial sayings, the very beginning of such formulaic patterns (e.g., When in Rome…)
will suffice for the listener to complete the phrase (… do as the Romans do). The
idiom principle, therefore, relates to the least effort principle, since idiomatic us-
age of language “reduces the language user’s online processing load and so facili-
tates communication” (Seidlhofer 2009: 197) in that “[o]ne linguistic item primes
expectation for the occurrence of others, with different degrees of predictability”
(2011: 130). Contrary to native or highly proficient speakers of a language (such as
interpreters), however:
ELF users will tend to construct what they have to say more analytically, in a
bottom-up fashion, drawing on what is semantically encoded in the grammar
and lexis of the language — in other words, by recourse to what Sinclair calls the
‘open-choice’ principle. (Seidlhofer 2009: 202)
phrases will find that many ELF speakers’ unidiomatic expressions do not match
the choices in their repertoire and, therefore, do not help to activate encoded
items, including the corresponding translation equivalents. This undermines re-
trieval and may put the interpreter off track. Such derailment may not only apply
to trainees. For, while one could argue that, in the example in Table 5, a profes-
sional would have had enough background knowledge to immediately come up
with Sondergipfel (extraordinary summit) upon hearing “special meeting,” the
same professional may not manage to do so in the less obvious cases that often
arise in interpreting situations.
Obviously, such retrieval constraints can have an adverse effect on funda-
mental strategic interpreting processes, such as anticipation and segmentation
(see Chernov’s 2004 probability-prediction model of inference and anticipation).
Conscious knowledge of preconstructed, conventional phrases as part of the in-
terpreters’ competence not only supports fast retrieval of established direct links,
but, in addition, facilitates anticipatory processes (see the example above, When
in Rome… do as the Romans do). Similarly, prefabricated chunks point the way
toward meaningful segmentation supportive of processing. Whereas the boundar-
ies of indecomposable holistic items are easily recognized (e.g., take a gap year), it
takes more time to reconstruct decomposable semantic chunks (e.g., work abroad
as a volunteer), which rely on sense creation processes. As a consequence, the ELF
speakers’ creative appropriation of resources to serve their communicative pur-
poses may be a major asset for unmediated interactional ELF communication, as
pointed out by ELF researchers (see Seidlhofer 2011: 97–111), but may become an
impediment to the interpreters’ processing.
6. Conclusion
References
Research. Festschrift für Kurt Kohn, ed. by Michaela Albl-Mikasa, Sabine Braun, and Sylvia
Kalina, 159–181. Tübingen: Narr Francke Attempto.
Kohn, Kurt. 1990. “Translation as Conflict.” In Confli(c)t. Proceedings of the International
Symposium‚ Contact + Confli(c)t, Brussels, 2–4 June 1988, ed. by Peter H. Nelde, 105–113.
Brussels: Association Belge de Linguistique Appliquée (ABLA papers 14).
Kohn, Kurt. 2004. “Zur Rolle des Übersetzens beim Sprachenlernen.” In Neue Perspektiven in
der Übersetzungs- und Dolmetschwissenschaft. Festschrift für Heidrun Gerzymisch-Arbogast
zum 60 Geburtstag, ed. by Juliane House, Werner Koller, and Klaus Schubert, 217–233.
Bochum: AKS-Verlag.
Kohn, Kurt. 2011. “English as a Lingua Franca and the Standard English Misunderstanding.”
In English in Europe Today. Sociocultural and Educational Perspectives, ed. by Annick De
Houwer, and Antje Wilton, 71–94. Amsterdam: John Benjamins. DOI: 10.1075/aals.8.06koh
Kohn, Kurt, and Michaela Albl-Mikasa. 2002. “Note-taking in Consecutive Interpreting.
On the Reconstruction of an Individualised Language.” In Linguistics and Translation
Studies. Translation Studies and Linguistics. Linguistica Antverpiensia1, ed. by Leona Van
Vaerenbergh, 257–272. Antwerpen: Hoger Instituut voor Vertalers & Tolken.
Kohn, Kurt, and Sylvia Kalina. 1996. “The Strategic Dimension of Interpreting.” Meta 41 (1):
118–138. DOI: 10.7202/003333ar
Kucharska, Anna. 2009. Simultandolmetschen in defizitären Situationen. Strategien der translato-
rischen Optimierung. Berlin: Frank & Timme.
Kurz, Ingrid. 2009. “The Impact of Non-Native English on Students’ Interpreting Performance.”
In Efforts and Models in Interpreting and Translation Research, ed. by Gyde Hansen, Andrew
Chesterman, and Heidrun Gerzymisch-Arbogast, 179–192. Amsterdam: John Benjamins.
DOI: 10.1075/btl.80.15kur
Kurz, Ingrid, and Elvira Basel. 2009. “The Impact of Non-native English on Information Transfer
in SI.” Forum 7 (2): 187–212.
Kutz, Wladimir. 2002. “Dolmetschkompetenz und ihre Vermittlung.” In Übersetzen und
Dolmetschen. Eine Orientierungshilfe, ed. by Johanna Best, and Sylvia Kalina, 184–195.
Tübingen: Francke.
Mauranen, Anna. 2009. “Chunking in ELF: Expressions for Managing Interaction.” Intercultural
Pragmatics 6 (2): 217–233. DOI: 10.1515/IPRG.2009.012
Mauranen, Anna. 2012. Exploring ELF. Academic English Shaped by Non-native Speakers.
Cambridge: Cambridge University Press.
McAllister, Robert. 2000. “Perceptual Foreign Accent and its Relevance for Simultaneous
Interpreting.” In Language Processing and Simultaneous Interpreting, ed. by Kenneth
Hyltenstam, and Birgitta Englund Dimitrova, 45–63. Amsterdam: John Benjamins.
DOI: 10.1075/btl.40.05mca
Pöchhacker, Franz. 2012. “Qualität, die man versteht: ein funktional-kognitiver Ansatz.” In
Dolmetschqualität in Praxis, Lehre und Forschung. Festschrift für Sylvia Kalina, ed. by
Barbara Ahrens, Michaela Albl-Mikasa, and Claudia Sasse, 19–31. Tübingen: Narr Francke
Attempto.
Reithofer, Karin. 2010. “English as a Lingua Franca vs. Interpreting — Battleground or Peaceful
Co-existence.” The Interpreters’ Newsletter 15: 143–157.
Reithofer, Karin. 2013. “Comparing Modes of Communication. The Effect of English as a Lingua
Franca vs. Interpreting.” Interpreting 15 (1): 48–73. DOI: 10.1075/intp.15.1.03rei
62 Michaela Albl-Mikasa
Séverine Hubscher-Davidson
This chapter explores the psychological construct of intuition and its influence
in decision-making behavior. Intuition was defined by Jung (1971) as a primary
mode of perception operating subconsciously. As opposed to sensing personal-
ity types who prefer concrete details, intuitive personality types prefer to acquire
information by imagining possibilities (Myers and Myers 1995). In this chapter,
an analysis of verbalization data from a translation process study is discussed
in order to demonstrate the influence of intuition on decision-making during
the translation process and to explore the implications of this influence. Recent
studies have found that intuition plays a role in learning and decision-making
tasks involving affect (Laborde, Dosseville, and Scelles 2010: 786). Intuition is
therefore here viewed as a potentially vital component of translator behavior
which could predict individuals’ translating effectiveness.
1. Introduction
Translation Studies (TS) scholars would probably like to think that the field has
moved on from a time when it was thought that translation was mostly an un-
conscious process that translators would have difficulty explaining (Weaver 1989).
Certainly, studies in the area of Translation Process Research have multiplied over
the last few years and provided useful insights into the workings of the black box.
doi 10.1075/bct.77.05hub
2015 © John Benjamins Publishing Company
64 Séverine Hubscher-Davidson
Nonetheless, much remains to be explained, and experts in the field still question
what exactly constitutes decision-making. In 1989, Weaver commented that, when
faced with a choice between two words, the translator hears the words in some
corner of his mind and likes the sound of one better than the other (1989: 117); in
1998, Newmark argued that intuition and imagination are as valuable in transla-
tion as awareness of what one is doing (1998: 77); in 2008, Munday postulated that
translation is not scientific and that the translation process always involves an ele-
ment of poetic taste (2008: 227). Over the years, it would seem that the idea of an
objective, conscious, straightforward decision-making process in translation has
not been universally accepted. Could there be a mysterious je ne sais quoi within
each translator that informs their choice of translation solutions? Could this inex-
plicable element be intuition? When weighing up different potential solutions to the
translation of a particular segment, can a first instinctive or unconscious thought
produce a more successful solution than one that is the result of careful analytical
deliberation? This chapter will attempt to provide an answer to these questions.
The concept of intuition can be said to have been rather neglected in academic
circles. It is often perceived to be an irrational and unreliable concept pertaining to
popular psychology rather than a scientific phenomenon worthy of investigation.
Despite this, the last two decades have witnessed a surge of interest in the topic. In
TS, process research has not yet fully explored the depths of nonconscious infor-
mation processing despite the will to go beyond traditional approaches to under-
stand the mechanisms underlying the translation act. Translators have often been
thought to deviate from rational patterns of behavior (Wilss 1996: 37), but I would
argue that what may appear irrational in the course of a translation act can actually
be the result of a highly effective nonconscious phenomenon.
In this chapter, an attempt is made to discuss the concept of intuition and
to demonstrate its influence on decision-making during the translation process.
First, relevant TS studies that have addressed intuition in a psychological context
are reviewed. Second, the construct is clarified and recent research in the field of
intuition is discussed. Third, an analysis of verbalization data from a translation
process study (henceforth referred to as ‘the case study’) is presented. The final
section of this chapter addresses some key issues in relation to the translator’s intu-
itive behavior. Intuition is viewed as a potentially vital component of translator be-
havior, and one which could be predictive of individuals’ translating effectiveness.
While it has been acknowledged for some time that cognitive and attitudinal fac-
tors can be very important during the translation process (e.g., Laukkanen 1996;
The role of intuition in the translation process 65
Jääskeläinen 1999; Dancette 2010), the concept of intuition has received little at-
tention in TS and, to my knowledge, only a handful of scholars have employed the
term. In order to understand what role intuition plays in the translator’s decision-
making process, this section presents a brief chronological review of studies in
translation which have explored the concept.
As early as 1981, Straight argued that translation decisions are a result of the
translator’s own reading of the original, and personal responses to the source text
(ST) and to the translation-in-progress (1981: 45). If that reading process matches
that of ST readers, then the translation is more likely to be successful, he claimed.
Although the idea that a personal reading of a ST might be helpful when translat-
ing is conceivable, Straight made other, less convincing, observations. He argued
that the translator’s decisions may be “influenced by an essentially ‘intuitive,’ right-
hemispheric mental apparatus” (1981: 49). In his view, the influence of the right
hemisphere is the reason why it is not always possible to provide logical explana-
tions for translation decisions. However, there is no evidence supporting the idea
that different thinking styles lie within each hemisphere and, in fact, the two brain
hemispheres are said to work together during cognitive effort (OECD 2002; Wager
et al. 2003; Pinel and Dahaene 2010). As a result, Straight’s theory does little to
enlighten us.
In 1988, Wilss made a distinction between objective and subjective transla-
tor behaviors, arguing that subjective intuition is used when a translator is faced
with a problem requiring a new approach in order to find a solution.1 In his view,
“the sphere where intuition is needed depends on the translator’s knowledge and
experience” (qtd. in Komissarov 1995: 351). Both Straight and Wilss conceived in-
tuition as a cognitive process influencing judgment and decision-making in trans-
lation, though they remained less clear as regards what activates and characterizes
intuitive processes.
Robinson (1991) argued that more work needed to be done on the ways in
which our bodies send us signals regarding what we know and how we should re-
act. Convinced that “we are guided much more powerfully […] by those autonomic
responses called “intuition” or “gut reactions”” (1991: x), Robinson calls transla-
tion a largely intuitive process, and postulates that we have somatic responses to
words on an unconscious level. The idea that translators’ body conditioning will
provoke a reaction, a feeling that a word may be ‘right’ before they are consciously
aware of what is happening, hints at the powerful influence of intuition.
Komissarov (1995) picks up Wilss’s distinction between objective and subjec-
tive behaviors during the translation process, and suggests:
1. The author’s comments in this section are based on Komissarov’s review of Wilss’s work, due
to the original only being available in German.
66 Séverine Hubscher-Davidson
Decisions can be taken either after an objective analysis of facts and situations or
on the strength of subjective intuition. […] In many cases it is the intuitive type of
decision-making that plays the decisive role in the translating process. (1995: 349)
3. Intuition in Psychology
Defining intuition is not an easy task. Epstein (2008: 23) highlights that intuition
has been given so many meanings that one might wonder whether the term has
any meaning at all. Nevertheless, recent work in cognitive psychology has placed
intuition firmly within the context of dual process models:
In the dual-process framework, intuition is […] automatic, holistic, affective, fast,
and associative, as contrasted with rational thought which is deliberate, analytical,
non-affective, slow, and rule-based. (Pretz and Totz 2007: 1249)
Dual process models make a clear distinction between intuitive and deliberate
thinking, qualifying the former as effortless and perception-like — the “low road”
— and the latter as analytic and controlled — the “high road” (Myers 2010: 371–
372). Our brains therefore are believed to process information using both conscious
analysis and nonconscious intuition. Whereas the former is said to lead to reasoned
decision-making, the latter occurs simultaneously and automatically, thus result-
ing in gut-feelings and implicit, spontaneous thoughts. Interestingly, as intuition
is extremely quick and works in parallel with conscious analysis, it is believed to
exceed the capacity of working memory, thus only making the outcome of intuition
available to us (Evans 2010: 314). This outcome is known as an intuitive judgment
(Dane and Pratt 2009: 3). Therefore, intuition processing is perceived to be mostly
nonconscious, and intuition outcomes are said to be experienced consciously.
Dual process models have been adopted in many different areas of psychology,
including the study of learning, memory, thinking, reasoning, and decision-making
(Evans 2010: 313). Though the categorical distinction between conscious analysis
and intuitive behavior is useful in terms of structuring our thinking about intuition,
Hammond (2010: 330) argues that judgments are not necessarily made at either
pole of the cognitive continuum, but could come from somewhere near the middle.
Betsch and Glöckner suggest that intuition and analysis share the cognitive work,
with intuitive processes working in the background while analytic processes are
2. In this chapter, it is understood that problem solving may consist of more than one instance
of decision-making.
68 Séverine Hubscher-Davidson
The mind
Automatic Deliberate
judgments resulting and
from acquired self-aware
expertise (inferential decision-
Nonconscious intuition) making
and intuitive
Complex,
non-sequential
judgments
(holistic intuition)
“match patterns between their environment and deeply held knowledge struc-
tures” (Dane and Pratt 2009: 6).
Creative intuition, on the other hand, is believed to be an input into the cre-
ative process and is defined as being “linked to a creative act of synthesis in which
disparate elements are fused together in novel combinations [to generate] some-
thing ‘truly new’ ” (Dane and Pratt 2009: 9–10). Creative intuitions are also said
to be preceded by an incubation period, i.e., “an antecedent to the rapid, holistic,
and associative operations that produce the intuition” (Dane and Pratt 2009: 11).
Table 1 was first published in Dane and Pratt (2009: 5), and I have adapted it so as
to illustrate the distinction between the two types of intuitions.
Despite employing different terminology, other scholars studying the concept
(e.g., Pretz and Totz 2007; Hogarth 2010; Sinclair 2011) also refer to two types
of intuitions. Sinclair (2011: 5–6) calls these intuitions ‘inferential’ and ‘holistic.’
While inferential processing, like problem-solving intuition, relies on matching
information with existing knowledge and experience, holistic processing, like cre-
ative intuition, relies less on existing knowledge.3
Pretz (2011: 24) argues that holistic intuition is likely to be most appropriate
when problems are perceived as highly complex, and inferential intuition is best
used when problems are perceived as relatively simple, either due to prior experi-
ence or the straightforward nature of the task. It could be argued that the use of
holistic intuition would be particularly well-suited to a literary translation task
where problems are likely to be complex in nature and to require the integration
of knowledge from different domains, whereas inferential intuition would be best
used in a technical or domain-specific translation task where pattern matching
may be more easily done.
Table 1. Two types of intuition (adapted from Dane and Pratt 2009: 5)
Intuition Type Description Nature of Affect Level of
Associations Incubation
Problem-Solving Automatic acts of Largely convergent/tight; Relatively Low to none
(inferential) recognition due to Based on highly specific low intensity
pattern matching domain knowledge
3. From this point onwards, the terms ‘inferential’ and ‘holistic’ will be consistently used to refer
to these two types of intuition.
70 Séverine Hubscher-Davidson
3.3 Measurement
4. Many authors have discussed the role of affect in intuitive judgments, and the impact of
emotional states on intuition and decision-making. Although an important avenue for further
research, dealing with the concept of ‘affectively charged’ intuitions (Dane and Pratt 2009) is
beyond the scope of this chapter.
The role of intuition in the translation process 71
In this section, I analyze three extracts from the Think Aloud Protocols (TAPs)
of a former student on the Master’s program in Translating and Interpreting at
the University of Bath. This data was collected as part of the doctoral work that I
conducted on the translation process, the aim of which was to better understand
the influence of students’ personality traits on their translations, as expressed
through their decision-making behaviors. In brief, the study took place in 2004;
the students were all English native speakers, with an average age of 27.4; and they
were asked to produce a translation of a literary text from French into English.
The text was 221 words in length and was originally published in a 1960s volume,
presented like a tourist guide of Paris and describing its mysterious aspects, with
stories of witches, plots and haunted places (Hubscher-Davidson 2009: 179). This
text was chosen due to its appealing content, which would, it was hoped, help keep
students interested and motivated. There was no specific brief, access to dictionar-
ies or time constraints.
The participants were required to think aloud and to complete a version of the
MBTI as well as other questionnaires. Once the participants finished, the target
5. Over 15 years ago, Robinson (1997: 73) hinted at the idea that certain types of learners pro-
cess information in a more intuitive manner than others. His early observations are consistent
with personality theory.
72 Séverine Hubscher-Davidson
texts were collected and assessed anonymously. Interestingly, out of the twenty stu-
dents who participated in the experiment, the five most successful were all classed
as intuitive types by the personality measure. Thus, results indicated a strong cor-
relation between success in the task and the presence of an intuitive trait among
participants. The following examples are extracts from the TAPs of one of these
intuitive students, T2, who performed particularly well in the task and scored very
highly on intuition. His protocols were selected as they seemed to present traces
of both intuitive and rational behaviors, thus demonstrating the interplay of both
systems during translation.
The intention here is not to reassess the correlation between the personality
trait and translation competence, nor to judge the quality of translation solutions
adopted. Rather, the aim of reviewing the protocols is to highlight the mechanisms
of intuitive behavior during the translation process and to gain a better under-
standing of its influence during decision-making in light of the literature reviewed
above.
Please refer to the appendix for the complete ST, the ST and TT extracts, and
the three excerpts from T2’s TAPs, which are analyzed below.
Excerpt 1
In the first excerpt (see Appendix), the student is wondering how to translate the
first part of the seventh sentence of the ST, i.e., la lune est pompeuse ou fugitive. The
student starts by verbalizing his understanding of the moon’s trajectory through
Paris and ‘behavior’ during this time. If intuition is, as Robinson (1991: 15) puts it
“a feeling that a word is or is not ‘right,’ ” it could be argued that T2 has an initial
intuition that using the word pompous to describe the moon is ‘not right’: ‘I don’t
know, I don’t really like pompous here.’ Having verbalized this subjective feeling,
T2 then proceeds to verbalize what can only be described as an attempt to reason,
to rationalize his initial reaction: ‘I don’t think it lends itself very well to the, well
I suppose it is metaphorical language […] there’s no reason to stay quite literal.’
Despite this effort to bring back decision-making to a conscious and analytical
process which is based on knowledge and experience (i.e., recognition of meta-
phorical language and the preferred norm for non-literal renderings), T2 cannot
seem to shake off his initial intuition that pompous is not the best translation solu-
tion: ‘but my instinct is that it’s not…I prefer imperious.’ It might seem surprising
that T2 does not opt for imperious in his final TT, but this decision is consistent
with findings from other process studies (see Kussmaul 1995). Although it could
be said that T2 has come full circle, i.e., that he did not act on his intuition but
opted for a predictable literal translation, two important features can be extracted
from this specific process.
The role of intuition in the translation process 73
Firstly, it would seem that the student experiences some awareness that his de-
cision-making process could be influenced by intuition or instinct: ‘I’m not really
sure whether [my] hesitation with the word pompous is to do with a well-informed
instinct or a lack of appreciation of [2 sec. pause] the meaning of the passage.’
This is consistent with outcome-oriented intuition research, which highlights that
it is possible to be aware that intuition processing is happening on a conscious
level (Sinclair 2010: 380). T2’s observation is evidence of a tension between the
underlying process and the point at which it is consciously experienced. This ‘in
between’ state is called fringe consciousness (Topolinski and Strack 2009). In this
case, the realization that there may be some processing mechanism that escapes
his immediate awareness leads T2 to postpone making a decision: ‘I’ll just put a
note over that one.’
Secondly, T2 subsequently makes the decision to personify the moon as being
coy. He expresses dissatisfaction with his first translation of fugitive as fleeting, and
settles on a solution which transfers the elusive and playful nature of the moon.
In this instance, T2 opts for a creative solution without much debate or, it seems,
without really knowing why: ‘coy I suppose, I can’t really think of another mean-
ing, word, right now.’ Clearly this is not true, since he previously came up with an
alternative, fleeting. Although the reasons underlying his choice may have seemed
unfathomable to the student, it could be argued that his previous intuition and
reflection on the personification of the moon as imperious might have somehow
triggered an appropriate subsequent creative solution. I would therefore venture
that T2’s prior experience with a potentially well-informed instinct may have sub-
consciously encouraged him to rely on inferential intuition and prevented him
from analyzing alternative solutions.
Excerpt 2
In the second excerpt, T2 exhibits continued doubt in relation to the first part
of the seventh sentence. He is aware that continuing to play around with la lune
est pompeuse will not help him find a solution to au hasard des arrondissements
(‘it still doesn’t resolve the issue of depending on’). Nonetheless, having chosen
a translation solution that went counter to his initial intuition, the translation he
settled on could still feel problematic to him and he may be attempting to find
an alternative that might better match his initial somatic response. The verbaliza-
tions also suggest that T2 is working on at least two ST segments simultaneously,
as he repeats the moon is pompous several times while trying to solve au hasards
des arrondissements, thus perhaps still processing one problem subconsciously
while verbalizing in relation to another. Betsch and Glöckner (2010: 281) claim
that extensive information integration can occur even if cognitive capacities are
74 Séverine Hubscher-Davidson
Excerpt 3
In the third excerpt, T2’s initial verbalization, frosty, immediately triggers a feel-
ing of uncertainty and an alternative TL word is offered (‘um […] don’t know
about frosty, icy I think’). This observed behavior recalls Epstein’s (2010) definition
of intuition as an almost instant sense of knowing seemingly resulting from un-
conscious information processing. Interestingly, T2 then approves of his solution
by expressing a positive feeling: ‘I think I’m quite happy with that.’ According to
Sinclair (2010: 381), a confirmatory feeling can accompany the outcome of intu-
ition processing as it emerges into consciousness, and it could be argued that T2
experiences this feeling. It could also be argued that T2 is experiencing the kind of
somatic response discussed by Robinson (1991) whereby his reasoning seems to
be emotionally directed (e.g., ‘I think I’m quite happy with that’) and his gut feel-
ing is intellectually monitored (e.g., ‘there might be a better way of saying that’).
T2 then tackles l’Esplanade des Invalides. The first (literal) translation that T2
verbalizes is Esplanade of the Invalides, a solution he initially seems content with
(‘think I’m just gonna put that’). However, he then expresses the need to ‘check
that,’ and verbalizes a ‘temptation’ to shorten the translation to the Invalides.
This could be an indication that T2 is experiencing a vague sense of knowing
that is emerging into consciousness, i.e., the outcome of an intuition. Not having
heard part of the verbalization related to this decision, I made the exception of
The role of intuition in the translation process 75
interrupting the verbalization in order to clarify what led to this second solution.
The student provided two justifications for his ‘temptation.’ The first, ‘[be]cause it’s
simpler,’ seemed to be an attempt at humor. The second justification seemed to be
an attempt to reason his choice: ‘because I think it’s something that is quite easily
recognizable to a lot of people.’ It could be argued that a conscious and systematic
consideration of relevant factors may not yet have taken place and that T2 is there-
fore struggling to explain the outcome of his intuition in terms that would not
amount to admitting the use of embarrassing guessing behavior.
T2’s second justification is guided by his assumption that target readers, like
him, will be educated, Francophile, and thus familiar with the green open space
that extends north of the Parisian Hôtel des Invalides. However, T2 quickly retracts
his assumption and decides that perhaps not all of his target readers would under-
stand what the Invalides are, berating himself for his subjectivity and insufficient
reflection: ‘that’s quite a subjective point of view […] when I actually think more
about it […] on second thought.’ T2 is clearly emphasizing the fact that his initial
intuitive solution had not been the result of careful deliberation, and that now he
‘knows better.’ It could be argued that the student is making a focused attempt to
bring back the decision-making process to a rational analysis of facts and to steer
clear of what Straight would call his personal, intuitive responses to the ST and to
the translation-in-progress (1981: 45).
Another salient feature of T2’s justification is the reference to how target read-
ers would behave upon reading the translation and to what knowledge they might
have as regards the Invalides. This verbalized concern for target readers is relevant
for at least two reasons. First, it could be argued that verbalized sensitivity to tar-
get reader behaviors is a feature of T2’s personality type and, more specifically, a
feature of his intuition. Second, this problem-solving behavior is consistent with
holistic intuition. Indeed, if innovation draws on experience from everyday life in
order to create something new (Sinclair 2010: 382), it is clear that T2 is drawing
on his life experience in terms of reader expectations to create an adequate TL so-
lution.6 This combination of experience and current stimulus results in disparate
patterns being associated in new and creative ways (Sinclair 2010: 383).
6. This is in conformance with Kussmaul and Tirkkonen-Condit’s observation that good trans-
lators are more prepared to use their world knowledge and their own inferences about text and
text type as a basis for their decisions (1995: 190).
76 Séverine Hubscher-Davidson
5. Discussion
intuitions were then followed by reflective processes, thus verifying the initial feel-
ings with explicit reasoning. This is what Evans (2010: 314) calls the default-inter-
ventionist structure, suggesting that dispositional thinking style determines the
likelihood of this intervention. Interestingly, the personality instrument identified
T2 as an INFJ or Introverted Intuitive Feeling Judging type, which suggests that
he is exceptionally intuitive, and that this intuition is responsible for his imagina-
tion, originality, and visions (Myers and Myers 1995: 110). However, INFJs are also
judging types, which implies that they do not usually exploit unplanned events,
and that they are rational and depend on reasoned judgments (Myers and Myers
1995: 75). T2’s preference for intuition is therefore tempered by his propensity to
judge. According to Myers and Myers (1995: 110) introverted intuitive types need
to cultivate this judging process to balance their intuition and to combine vision
and practicality successfully.
T2’s rational thinking could be perceived on several occasions (e.g., ‘when I
actually think more about it’), and it is interesting to note this link between per-
sonality traits and the handling of intuitive judgments, which is in line with the
observation made by Robinson (1997: 73) that learning styles influence informa-
tion processing. As alluded to in Section 3, successfully integrating intuition and
analysis can improve performance by drawing on both conscious and noncon-
scious resources (Kuhnle 2011: 228). As a result, people who may not have well-
balanced traits may be inclined to rely on their impulsive and effortless intuitions,
thus running into trouble. Promoting a reliance on ‘gut feelings’ would therefore
be ill-advised, as they can be misleading in the absence of an active reasoning
process.
One of the key findings from the analysis is that both holistic and inferential in-
tuitive judgments were present alongside conscious analysis in the TAPs, which
suggests that the translator had to deliberate on the appropriateness of both types
of intuitions.
Interestingly, the literature suggests that participants in reasoning experiments
tend to rely on intuition as it requires less effort than reasoning, particularly when
tasks become more complex (Frederick 2005; see also Stanovich 2009; Evans 2010;
Betsch and Glöckner 2010). The issue here is that an experienced translator will
probably have useful inferential intuition to rely on, but novices could instead rely
on biased judgments. According to Hogarth (2010: 343), inferential intuition can
be trusted when a person’s past experience is representative of the current situa-
tion but, if this is not the case, one should be careful.
78 Séverine Hubscher-Davidson
Despite the fact that intuition processing is not a conscious process, recent re-
search suggests that it is possible to train ourselves to pay attention to cues in
our environment that may help uncover solutions to problems. Interestingly, this
would suggest that translators may be trained to become more in tune to the out-
comes of their intuition. Sinclair (2010: 380) suggests that this requires a goal-
directed thought process aimed at a specific task, and that operating on a primed
nonconscious level might enable intuition processing to be activated. This ties in
with Robinson’s (1991) argument that translators should pay particular atten-
tion to their somatic responses, and with Bǎlǎcescu’s (2006) point that translators
should be made aware of valid intuitions during training. This study provides ad-
ditional evidence that thorough training in the handling of intuitive judgments
when translating would be beneficial.7
Dane and Pratt (2009: 31) suggest that different types of intuition may be use-
fully considered during different stages of the decision-making process, and that
individuals should return to their initial intuitions once they have undertaken an
analytical assessment of a problem in order to re-evaluate these. This would also
enable individuals to stay attuned to the potential emergence of holistic intuitions
which, we have noted, require a period of incubation and can become useful in the
final stages of complex decision-making.
Although consensus is yet to be reached on the best way to develop intuition,
the potential to enhance translation decision-making processes should be exploit-
ed. Dane (2011: 221) suggests that even attending to inaccurate intuitions gives
one the opportunity to compensate for them, and decide what role they should
play in decision-making.
7. For those wishing to apply formative techniques in the translation classroom, Robinson’s
1997 volume provides exercises designed to develop subliminal skills alongside conscious pro-
cessing.
The role of intuition in the translation process 79
6. Conclusion
The present study aimed to shed light on the role that intuition played in one
translator’s decision-making in order to enhance our understanding of the cir-
cumstances when trusting intuition is an appropriate strategy in translation. It
is clear that validation studies are needed to build on this subjective analysis, but
the study demonstrated the worthwhile nature of further investigating the mecha-
nisms underlying intuition.
The method used to collect information on the translation process in this
study remains the topic of a vigorous debate, since the validity of verbal reports
as a way to tap into the translator’s mental processes has been hotly contested.
Jääskeläinen (2011: 16) argued that the method is based on the idea that partici-
pants can report accurately on information that is consciously processed but not
on automated processes. This implies that intuitive outcomes may feature in the
protocols, but not intuitive processes, which are not consciously experienced. In
addition, among other issues, Göpferich and Jääskeläinen (2009: 179) report that
think-aloud data do not necessarily reflect the order in which mental processes
occur. It must therefore be borne in mind that the instances of intuitive behavior
gauged from T2’s protocols are just a partial representation of the complex and
(mostly) unconscious processes actually taking place.
However, I would argue that TAPs are still useful for exploring intuition. One
area where the design of intuition studies can be improved is by experimenting
in multimodal situations (Betsch and Glöckner 2010: 290), and thinking aloud
could be one of the multimodal methods with which intuitive thinking processes
are investigated. Wilss (1998: 59) highlights that understanding translators’ pre-
choice behaviors must be a priority in future studies, particularly as the develop-
ment of internalized processes is not well understood. The impact of intuition in
translation processing may be far more important than we imagine, and the topic
should no longer be a marginal area of enquiry. For example, it has been found
that speeded tasks encourage a greater reliance on intuition, to the detriment of
logical reasoning (Evans 2010: 318). This is clearly relevant for those studying the
translation process of professionals working under real-world time constraints.
This study demonstrated that it is time to raise awareness among the transla-
tion community not only of the benefits of taking ‘the high road’ but also the im-
portance of intuitive processing. Long gone is the time when translators’ intuitions
were considered to be imponderable and inaccessible to rational explanation.
80 Séverine Hubscher-Davidson
References
Jääskeläinen, Riitta. 1999. Tapping the Process: An Explorative Study of the Cognitive and Affective
Factors Involved in Translating. Joensuu: University of Joensuu.
Jääskeläinen, Riitta. 2011. “Back to Basics: Designing a Study to Determine the Validity and
Reliability of Verbal Report Data on Translation Processes.” In Cognitive Explorations of
Translation, ed. by Sharon O’Brien, 15–29. London: Continuum Studies in Translation.
Jung, Carl Gustav. 1971. Psychological Types. Princeton, NJ: Princeton University Press.
Kiraly, Don. 1995. Pathways to Translation: Pedagogy and Process. Kent, OH: Kent State
University Press.
Komissarov, Vilen Naumovich. 1995. “Intuition in Translation.” Target 7 (2): 347–354.
DOI: 10.1075/target.7.2.10kom
Kuhnle, Claudia. 2011. “The Benefit of Intuition in Learning Situations.” In Handbook of
Intuition Research, ed. by Marta Sinclair, 227–236. Cheltenham: Edward Elgar Publishing.
Kussmaul, Paul. 1995. Training the Translator. Amsterdam: John Benjamins.
DOI: 10.1075/btl.10
Kussmaul, Paul, and Sonja Tirkkonen-Condit. 1995. “Think-Aloud Protocol Analysis in
Translation Studies.” TTR: Traduction, Terminologie, Rédaction 8 (1): 177–199.
DOI: 10.7202/037201ar
Laborde, Sylvain, Fabrice Dosseville, and Nicolas Scelles. 2010. “Trait Emotional Intelligence and
Preference for Intuition and Deliberation: Respective Influence on Academic Performance.”
Personality and Individual Differences 49: 784–788. DOI: 10.1016/j.paid.2010.06.031
Laukkanen, Johanna. 1996. “Affective and Attitudinal Factors in Translation Processes.” Target 8
(2): 257–274. DOI: 10.1075/target.8.2.04lau
Lörscher, Wolfgang. 1991. Translation Performance, Translation Process, and Translation
Strategies. Tübingen: Gunter Narr.
Munday, Jeremy. 2008. Style and Ideology in Translation. London: Routledge.
Myers, David G. 2010. “Intuition’s Powers and Perils.” Psychological Inquiry: An International
Journal for the Advancement of Psychological Theory 21: 371–377.
Myers, Isabel Briggs, and Peter B. Myers. 1995. Gifts Differing: Understanding Personality Type.
Mountain View, CA: Davies-Black Publishing.
Newmark, Peter. 1998. “The Customer as King: A Response to Hans G. Honig.” In Translation
and Quality, ed. by Christina Schäffner, 75–78. Clevedon: Multilingual Matters.
OECD. 2002. Understanding the Brain — Towards a New Learning Science. Paris: OECD
Publishing.
Pinel, Philippe, and Stanislas Dehaene. 2010. “Beyond Hemispheric Dominance: Brain Regions
Underlying the Joint Lateralization of Language and Arithmetic to the Left Hemisphere.”
Journal of Cognitive Neuroscience 22 (1): 48–66. DOI: 10.1162/jocn.2009.21184
Pretz, Jean E. 2011. “Types of Intuition: Inferential and Holistic.” In Handbook of Intuition
Research, ed. by Marta Sinclair, 17–27. Cheltenham: Edward Elgar Publishing.
Pretz, Jean E., and Kathryn Sentman Totz. 2007. “Measuring Individual Differences in Affective,
Heuristic, and Holistic Intuition.” Personality and Individual Differences 43: 1247–1257.
DOI: 10.1016/j.paid.2007.03.015
Robinson, Douglas. 1991. The Translator’s Turn. Baltimore: John Hopkins University Press.
Robinson, Douglas. 1997. Becoming a Translator. London: Routledge.
Sadler-Smith, Eugene, and Erella Shefy. 2004. “The Intuitive Executive: Understanding and
Applying ‘Gut Feel’ in Decision Making.” Academy of Management Executive 18 (4): 76–91.
DOI: 10.5465/AME.2004.15268692
82 Séverine Hubscher-Davidson
Appendices
(Caradec, François and Jean-Robert Masson (eds). 1985. Guide de Paris Mystérieux. Collection
Les Guides Noirs. Nouvelle Édition. Paris: Éditions Tchou)
whether that’s going to be accurate enough (4s) if I take the one minute option then perhaps that
might be accurate enough (1s)
The aim of this process-oriented study is to test the literal translation hypothesis,
i.e., the frequently held view that, in their search for the most suitable translation
solution, translators move from more literal towards freer renderings. The study
uses keystroke-logging to investigate translation processes involving self-revision
in terms of ‘distance dynamics.’ The latter term refers to the fact that each new
solution to a particular translation problem can be closer or more distant from
the source item than the previous one. A more specific hypothesis is also tested,
namely that experience in interpreting has an effect on distance dynamics. To
this end, processes of two groups of subjects — translators without substantial
interpreting experience, and interpreters — are compared. The findings chal-
lenge the literal translation hypothesis, and highlight some methodological
issues that should be tackled in future studies.
1. Introduction
In this chapter we present the findings of a study in which we tested the literal trans-
lation hypothesis, more aptly named ‘deliteralization hypothesis’ (see Chesterman
2011). This hypothesis reflects the frequently held view (see Ivir 1981; Toury 1995;
Tirkkonen-Condit 2005; Englund Dimitrova 2005; Shlesinger and Malkiel 2005;
Jakobsen, Jensen, and Mees 2007) that translators first consider cognate solutions
and that their self-revision involves a move from such literal solutions towards
freer ones. For example, Tirkkonen-Condit (2005) posits that “literal translation is
a default rendering procedure, which goes on until it is interrupted by a monitor
doi 10.1075/bct.77.06pav
2015 © John Benjamins Publishing Company
86 Nataša Pavlović and Goranka Antunović
that alerts about a problem in the outcome,” building on Ivir (1981; 1988), who
claimed:
[t]he translator begins his search for translation equivalence from formal corre-
spondence, and it is only when the identical-meaning formal correspondent is
either not available or not able to ensure equivalence that he resorts to formal cor-
respondents with not-quite-identical meaning or to structural and semantic shifts
which destroy formal correspondence altogether. (Ivir 1981: 58)
more distant from the source item than the previous solution)? This brings us to
the final question that needs to be addressed: does the claim apply to all translation
processes, regardless of who translates, in which direction, what type of source text
is involved, whether there is time pressure or not, etc.?
In this research we set out to test the literal translation hypothesis, restricting its
scope to actual text production rather than applying it to cognitive processing in
general. We assumed, as does Chesterman (2011: 28), that keystroke logging is
a suitable method to study such text production. The general hypothesis of our
research was therefore formulated as follows: in their search for the most suitable
translation solution, translators move from more literal towards freer renderings.
In an earlier study (Antunović and Pavlović 2012), which involved students
translating from their second and third languages (L2 and L3), we had already
tested this hypothesis but the evidence supporting it was not convincing enough.
In the present study we decided to test professional subjects and, more specifically,
to compare translation processes of translators (without substantial interpreting
experience) with interpreters working on the same translation task. Interpreters
were chosen because of their experience in immediately producing acceptable TL
solutions, enhanced by deverbalization / conceptualization habit (the relevance
of the concept in interpreter training is confirmed in e.g., Gile 1990: 33 or, in par-
ticular, Dejean Le Féal 1993). We assumed that the ability to immediately produce
acceptable TL solutions would be activated also in the written translation tasks,
and that the first solution suggested by interpreters would be less literal than that
proposed by translators. Instances of deliteralization were therefore expected to be
less frequent in the interpreter group.
The above assumptions led to our more specific hypothesis, namely that expe-
rience in interpreting has an effect on ‘distance dynamics.’ The effect is expected to
be evident in that translation processes of interpreters, when compared to those of
translators, involve: a. fewer self-revisions on the whole; b. fewer self-revisions that
involve deliteralization; c. a relatively smaller share of deliteralizing self-revisions
in the overall number.
Subhypotheses (b) and (c) build on the findings of Jakobsen et al. (2007). In
their study inspired by Shlesinger and Malkiel (2005), they compared the process-
ing of idioms by interpreters and translators. They had expected both groups of
subjects to start off by looking for solutions based on idiom-to-idiom translation,
which, however, proved to be true only of the translator group. Contrary to the au-
thors’ expectations, occurrences of ‘direct transfer’ (literal translation) were found
88 Nataša Pavlović and Goranka Antunović
3. Method
In this study we tested six professional translators and six interpreters (four of
whom were ESIT graduates) with Croatian as their first language (L1 or A) and
English as L2 or B-language. At the time of the experiments, all of the subjects
had five or more years of professional experience in translation and interpreting
respectively, and were working for EU institutions.
The two groups — the translators and the interpreters — were asked to trans-
late the same text, 168 words long, from the EU portal at www.europa.eu, en-
titled Documents you will need. The subjects were informed about the origin of
the source text (ST) and were given a realistic translation brief. The translation
sessions were recorded with the help of the keystroke-logging program Translog
2006,1 based on Jakobsen and Schou (1999) and further developed by Jakobsen
(2006). Having been introduced to the program, the subjects were asked to do
a warm-up task in order to familiarize themselves with the application. No ex-
ternal resources were available, and the time was limited in order for the logging
program to capture a maximum number of solutions. We believed that with time
constraints the subjects would be less likely to formulate solutions in their mind
before writing them down and thereby make them inaccessible for analysis.
Following the experiments, all problems and self-revisions were identified
and listed. For the purposes of this research, ‘problem’ was defined as every seg-
ment of the ST that involved at least one self-revision. Self-revisions encompassed
all the changes introduced into the target text (TT) by the translator during the
process of translation. As in our previous research on self-revision (Antunović
and Pavlović 2011; 2012), we excluded typos and followed the method used by
Englund Dimitrova (2005: 116) and Brenda Malkiel (2009: 156) in deciding what
counted as an instance of self-revision. They both observe that several distinct,
1. http://translog.dk/
The effect of interpreting experience on distance dynamics 89
Example 1:
The ST segment a common visa policy was initially translated as ‘zajednička poli-
tika za vize’ [common policy for visas]. At a later point in the drafting process, this
initial solution was changed into ‘zajednička politika u području viza’ [common
policy in the area of visas]. While still in the process of drafting the translation
(i.e., before the whole ST was translated), the translator was presumably not happy
with the solution resulting from the first revision and changed it into ‘zajednička
politika viznog režima’ [common visa regime policy]. In our analysis this solu-
tion counted as the second revision. It produced the solution that was kept in the
final version of the TT, which means that this particular example involved no self-
revisions in the post-drafting phase.
Example 2:
The ST contained the segment the Schengen rules, which produced ‘pravilima o
Schengenu’ [rules about Schengen] in one of the transcripts. The translator appar-
ently felt this initial solution should be revised, and substituted it with ‘pravilima
o schengenskom sustavu’ [rules about the Schengen system]. Deciding to apply
a different orthographic rule, the translator then changed the spelling and used
a capital initial S- in the adjective ‘schengenskom.’ The second revision thus pro-
duced ‘pravilima o Schengenskom sustavu’ [rules about the Schengen system],
which was kept until the end of the drafting process. Once the whole ST was trans-
lated, the subject started reading it again, introducing some new revisions. Among
those self-revisions in the post-drafting phase were two further changes of the
segment taken up here, i.e., the third and the fourth revision of the initial solu-
tion. While the third revision involved a return to the initial solution (‘pravilima o
Schengenu’), the fourth read ‘schengenskim pravilima’ [the Schengen rules] (note
that this phrasing is closer to the ST segment than any earlier solution). As can be
gathered from this description, Example 2 provided solutions relevant for all three
stages of our analysis reported here: (a) four self-revisions with regard to the whole
process, (b) a first revision of an initial solution, and (c) two self-revisions in the
post-drafting phase.
4. Findings
The specific hypothesis formulated in this study states that substantial experi-
ence in interpreting affects interpreters’ self-revision when they undertake writ-
ten translation. The first subhypothesis focuses on the number of self-revisions:
The effect of interpreting experience on distance dynamics 91
As it was pointed out earlier (see Section 1), the various formulations of literal
translation hypothesis in the literature do not state clearly to what part of the
self-revision process the hypothesis applies. With no specific phases mentioned,
it seems justified to take it as applying to the process as a whole. We therefore
counted the overall number of the three different types of self-revisions in the
interpreters’ and in the translators’ process. In addition, allowing for different in-
terpretations, we also analyzed first self-revisions in their own right, as well as
self-revisions done in the post-drafting phase.
45%
39.47%
40%
35% 33.77%
30%
26.75%
25% T-moves
S-moves
20% Neutral revisions
15%
10%
5%
0%
Figure 1. Three types of self-revision, whole process, all subjects
The ordering of the three types of self-revisions changed when the two groups
of subjects were studied separately. In the translator group, T-moves (40.87%)
were still predominant, but S-moves proved more common than neutral revisions
(34.13% vs. 25%). In the interpreter group, however, T-moves were less common
than distance neutral revisions (37.74% T-moves, 44.61% neutral changes, while
S-moves accounted for only 17.65% of all revisions) (Figure 2).
50%
44.61%
45%
40.87%
40% 37.74%
35% 34.13%
30%
25% T-moves
25% S-moves
17.65% Neutral revisions
20%
15%
10%
5%
0%
Interpreters Translators
The fairly high values for neutral self-revisions indicate that this type must not
be neglected in studies of revision processes. However, since such revisions result
in TL solutions which, compared to the solutions they have replaced, are neither
The effect of interpreting experience on distance dynamics 93
closer nor more distant from their respective source elements, they provide no
information on the direction of distance change. What is more, they can make the
comparison between the two groups of distance changing revisions less clear. We
therefore decided to also look at the number of T- and S-moves alone, excluding
neutral revisions (Figure 3).
38.5(68%) 18(32%)
Interpreters
T-moves
S-moves
51.5(54.50%) 43(45.50%)
Translators
0 20 40 60 80 100
Figure 3. Number of T- and S- moves only, whole process, interpreters vs. translators
With regard to our general hypothesis, it can be noted that T-moves were indeed
more numerous than S-moves with both groups of subjects. However, with the
45.5% share of S- moves in the translator group, the difference between the two
types is not very convincing. The interpreter group is more in line with the general
hypothesis, even if it also has a fair share of S-moves (almost one third of all dis-
tance changing revisions).
Figures 2 and 3 point at another interesting finding at this stage of our study.
In contrast to our specific subhypothesis (c) (stating that interpreters’ translation
processes, when compared to those of translators, involve a relatively smaller share
of changes from a more literal to a freer solution), the proportion of T-moves
turned out to be larger in the interpreter group (68.14% interpreters vs. 54.5%
translators).
initial solutions that they replace. If literal translations are default solutions, the
first self-revision will then have to involve a T-move, which means that the share
of such moves among first revisions would have to be very high.
As can be seen in Figures 4–6, the T-move values are indeed somewhat higher
for first revisions than they were for the whole process: for the combined group
of all the 12 subjects (Figure 4), the share has risen from 39.47% to 42.45%; in
the interpreter group from 37.74% to 43.21% (Figure 5) and from 68% to 71%
with neutral revisions excluded (Figure 6); in the translator group from 40.87%
to 41.67% (Figure 5) and from 54.5% to 59% with neutral revisions excluded
(Figure 6). However, it is just as obvious that the increase is not particularly high
in any of these comparisons.
45% 42.45%
40%
35% 33.96%
30%
25% 23.58% T-moves
S-moves
20% Neutral revisions
15%
10%
5%
0%
Figure 4. Three types of self-revisions, first revisions only, all subjects
50%
45% 43.21%
41.67%
40% 38.89%
35%
30% 29.49% 28.85%
T-moves
25% S-moves
20% 17.90% Neutral revisions
15%
10%
5%
0%
Interpreters Translators
Figure 5. Three types of self-revisions, first revisions only, interpreters vs. translators
The picture for first revisions differs from that for the whole process in two re-
spects. One is the order of the three types of changes in the interpreter group:
The effect of interpreting experience on distance dynamics 95
neutral revisions are here less numerous than T-moves (although they still account
for 38.89%, i.e., more than a third of all revisions). The other interesting difference
has to do with the comparison of the interpreter and the translator group with re-
gard to the number and the share of T-moves among first revisions: contrary to our
specific subhypotheses, interpreters made a larger number of deliteralizing first
revisions than translators (both in absolute figures and in terms of their share).
35(71%) 14.5(29%)
Interpreters
T-moves
S-moves
32.5(59%) 23(41%)
Translators
0 10 20 30 40 50 60
Figure 6. Number of T- and S- moves only, first revisions, interpreters vs. translators
With regard to our general hypothesis, the predominance of T-moves over S-moves
is manifest in all three groups (interpreters and translators combined, and the two
groups taken separately). However, the question still remains how predominant
they would have to be to consider the description of the self-revision process as
moving from more literal towards freer renderings unequivocally substantiated.
For, even among the first revisions, T-moves never account for more than half of
the revisions done if all three types of changes are considered (the values range
from 41.67% to 43.21%). The predominance seems more convincing if only dis-
tance changing moves are considered, but even then the relatively high share of
S-moves would have to be acknowledged (approximately one third of all distance
changing moves: 29% and 41% for the interpreter and the translator group respec-
tively (Figure 6), or 35.71% for the combined group of all twelve subjects).
translation process, and thereby check whether the post-drafting phase would cor-
roborate our general hypothesis more conclusively than the former two situations
(the whole process and first revisions only).
As it turned out, just the opposite was true in our study: the values for T-moves
in the post-drafting phase were lower than in the two situations described ear-
lier (4.2.1 and 4.2.2). For example, for the combined translator-and-interpreter
group, the T-move value was 24.32% in the post-drafting phase, which compares
to 39.47% for the whole process and 42.45% for first self-revisions. As can be seen
in Figures 7 and 8, T-moves accounted for less than one third of all revisions in the
combined and the interpreter group (24.32% and 15.91%) and only slightly more
in the translator group (36.67%).
60% 56.76%
50%
40%
T-moves
30% S-moves
24.32% Neutral revisions
20% 18.92%
10%
0%
Figure 7. Three types of self-revision, post-drafting phase, all subjects
80%
72.73%
70%
60%
50%
T-moves
40% 36.67% S-moves
33.33%
30.00% Neutral revisions
30%
20% 15.91%
11.36%
10%
0%
Interpreters Translators
The revision type whose share rose most noticeably in the post-drafting phase,
in parallel to the pronounced drop in the share of T-moves, were neutral revi-
sions. The rise was quite dramatic in the interpreter group (72.73%, or 16 out of
The effect of interpreting experience on distance dynamics 97
the post-drafting total of 22) but there was a certain rise in the translator group
as well (33.33% vs. 25% for the whole process and 28.85% for first revisions). This
has once again confirmed the significance of distance neutral moves for any self-
revision typology.
As regards the ratio between the two distance changing types of revisions in
the post-drafting phase, T-moves were once again more numerous than S-moves,
but their dominance could hardly be considered substantial: 58% : 42% in the
interpreter group and 55% : 45% in the translator group (Figure 9). It is worth not-
ing, though, that the absolute figures in this case are very low (6 and 10 distance
changing revisions in the interpreter and the translator group respectively), which
challenges the relevance of the percentages quoted here.
3.5(58%) 2.5(42%)
Interpreters
T-moves
S-moves
5.5(55%) 4.5(45%)
Translators
0 2 4 6 8 10 12
Figure 9. T- and S- moves only, post-drafting phase, interpreters vs. translators
Our general hypothesis applies to the translation process as such, and it emphasiz-
es the significance of deliteralizing changes (‘T-moves’ in our study) in the process
of self-revision, claiming that they are essential to the process. As we pointed out
in the introductory section of this chapter, there are a number of unclear points in
various formulations of the literal translation hypothesis in literature. The fact that
it can be interpreted in different ways has led us to refrain from making only one
conclusion with regard to its truthfulness.
We base our conclusions on the values established in our study for three differ-
ent types of self-revisions undertaken by our subjects. Since we analyzed different
parts of the process, we obtained several values for each revision type. To reflect
that finding, the share of each of the three types in the overall number of self-
revisions is expressed as a range, with the lowest and the highest share established
as its endpoints (Table 2). The values included here refer to the largest group in our
analysis (the combined group, interpreters and translators taken together), which
we deem most appropriate in the context of the general hypothesis. As has been
done throughout this chapter, two sets of values are presented: the values in the
first row show the shares with all three self-revision types taken into account; the
values in the second row indicate the shares between T-moves and S-moves only,
with distance neutral revisions excluded from the calculation:
As Table 2 indicates, moves from a more literal towards a less literal translation
solution (i.e., T-moves) always account for more than half of all self-revisions
(56–64%) when only two revision types are studied, and for between 24% and
42% when neutral changes are also included (the arithmetic mean of the range
being exactly 33%). Given that fact, we feel that the figures substantiate the view
that cognitive processing in translation involves a fair number of literal solutions,
as well as the expectation that a considerable portion of self-revisions involve del-
iteralization.
On the other hand, the other two revision types also appear regularly in our
study, in the processes of all subjects and regardless of the part of the process ana-
lyzed. The values for S-moves (36–44% when neutral revisions are excluded, and
19–27% when they are included) and for neutral changes (34–57%) are too high to
The effect of interpreting experience on distance dynamics 99
warrant the interpretation of our general hypothesis as implying that literal trans-
lations are default solutions in actual text production. The high share of neutral
changes (higher than that of T-moves!) prompts us to draw yet another conclu-
sion, namely that descriptions of distance dynamics cannot be limited to T- and
S-moves only but they have to take into account changes that are neutral with
respect to the distance between the original segment and a particular translation
solution.
It was also pointed out in the Introduction that formulations of the literal
translation hypothesis in the literature do not normally limit the group of transla-
tors to which it applies. The hypothesis is therefore taken to apply ‘in general,’ and
an implication for empirical studies is that it should be tested with different groups
of subjects, and that it is not essential to keep subjects with different characteristics
strictly apart. We therefore deem it appropriate to refer at this point to the results
of an earlier study (Antunović and Pavlović 2012). At that stage we tested the same
general hypothesis by analyzing twenty logs produced by ten language and transla-
tion students, each of them working from two different source languages.
As can be seen in Figure 10, the shares of T- and S-moves in the students’ self-
revisions (excluding neutral changes) correspond fully to the values established
in this study, and corroborate the above conclusions regarding the two different
interpretations of the literal translation hypothesis just mentioned.
80%
70% 68.14%
65.49%
60% 54.50%
50% 45.50%
40% T-moves
34.51% S-moves
31.86%
30%
20%
10%
0%
Interpreters Translators Students
Figure 10. Shares of T- and S- moves over the whole process (combined data from two
studies)
45%
40.80%
40%
34.33%
35%
30%
24.88%
25% T-moves
S-moves
20% Neutral revisions
15%
10%
5%
0%
Figure 11. Shares of the three self-revision types, all subjects (combined data from two
studies)
As we saw in 4.2.1, the interpreter group proved to have made a strikingly large
number of distance neutral revisions (44.6% vs. 25% in the translator group),
which may be interpreted as suggesting a lack of routine in solving some problems
typical of written translation (e.g., spelling or punctuation). Hence our first meth-
odological concern: to what extent is the assumed activation of the interpreters’
ability to immediately produce acceptable TL solutions hampered by their lack
of routine in the written translation mode? The challenges of comparing modali-
ties are well known (e.g., Shlesinger and Malkiel 2005); in this case a solution for
future studies might be to use a group of subjects who are equally skilled in both
modalities, and to compare their written tasks with those of the control group
(translators without any interpreting experience).
The large number of neutral moves in our study could also be interpreted as
being TL specific, considering the existence of several orthography manuals in
Croatia, with some mutually exclusive spelling and punctuation rules that often
make native speakers feel insecure about their writing. It seems plausible that in-
terpreters, who are not likely to write as much on a daily basis as translators, are
more insecure in this respect.
Another limitation of this study is the small sample (12 subjects), a characteristic
it shares with most process oriented translation research. One or two subjects who
get stuck on a translation problem and make an exceptionally large number of
revisions can skew the results, as can be seen from Table 3. The first row shows
the average distribution of T-moves and S-moves for the whole translator group
(6 subjects), and the second the average figures without the two subjects who were
markedly different from the rest. It can be seen that the figures for the latter group
102 Nataša Pavlović and Goranka Antunović
are much closer to those obtained for the interpreters (and, incidentally, to those
obtained for the students, see 5.1). Statistical measures such as the trimmed mean
are not worth performing on small samples, but that situation would change if the
experiments were replicated to involve more subjects. Future studies should thus
present a more convincing picture of distance dynamics, in particular if the find-
ings corroborate those obtained so far.
In the existing literature dealing with what we have referred to as the literal trans-
lation hypothesis, a plethora of terms are used (literal solution, formal corre-
spondent, contrastive correspondent, default solution, cognate, reflex translation,
unthinking translation, direct transfer, etc.), and the implications of differences
among those terms — and, more importantly, among the concepts behind them
— are insufficiently discussed. In our research, we adopted a very broad definition
and considered as ‘literal solutions’ those solutions that are obviously source-in-
spired. Future studies should pay closer attention to this issue and perhaps differ-
entiate among the concepts more clearly.
In the present study all subjects translated from their L2 (B-language) into their L1
(A-language). The students from our previous research, mentioned above in 5.1,
worked from their L2 and L3 into L1, and yielded similar results regarding dis-
tance dynamics. Future studies could further test directionality as a variable, e.g.,
by introducing translation from L1 into L2, which is a regular practice in many
settings. Language pairs might prove a relevant variable, as formal closeness of the
two languages could play an important role in the deliteralization process (see also
the comment in 6.1 regarding the Croatian language, indicating how specificities
of the TL may influence self-revision and distance dynamics). Another variable
that deserves further attention is the relationship between distance dynamics and
individual translating style. Finally, future studies could use different text types or
translation tasks, as the requirement of more literal vs. freer translation can also be
expected to influence distance dynamics.
The effect of interpreting experience on distance dynamics 103
References
Antunović, Goranka, and Nataša Pavlović. 2011. “Moving on, Moving back or Changing it here
and now: Self-Revision in Student Translation Processes from L2 and L3.” Across Languages
and Cultures 12 (2): 213–234. DOI: 10.1556/Acr.12.2011.2.5
Antunović, Goranka, and Nataša Pavlović. 2012. “Dinamika otklona od izvornika tijekom iz-
rade prijevoda [Distance dynamics in the translation process].” In Aktualna istraživanja u
primijenjenoj lingvistici. Zbornik radova s 25. međunarodnog skupa HDPL-a održanog 12.-
14. svibnja 2011. u Osijeku [Current research in applied linguistics. Proceedings of the 25th
International Conference of the Croatian Applied Linguistics Society], ed. by Leonard Pon,
Vladimir Karabalić, and Sanja Cimer, 199–214. Osijek: HDPL.
Chesterman, Andrew. 2011. “Reflections on the Literal Translation Hypothesis.” In Methods and
Strategies of Process Research: Integrative Approaches in Translation Studies, ed. by Cecilia
Alvstad, Adelina Hild, and Elisabet Tiselius, 23–35. Amsterdam: John Benjamins.
DOI: 10.1075/btl.94.05che
Dejean Le Féal, Karla. 1993. “Pédagogie raisonnée de la traduction.” Meta 38 (2): 155–197.
DOI: 10.7202/003451ar
Englund Dimitrova, Birgitta. 2005. Expertise and Explicitation in the Translation Process.
Amsterdam: John Benjamins. DOI: 10.1075/btl.64
Gile, Daniel. 1990. “Scientific Research vs Personal Theories in the Investigation of Interpretation.”
In Aspects of Applied and Experimental Research on Conference Interpretation, ed. by Laura
Gran, and Christopher Taylor, 28–41. Udine: Campanotto Editore.
Ivir, Vladimir. 1981. “Formal Correspondence vs. Translation Equivalence Revisited.” Poetics
Today 2 (4): 51–59. DOI: 10.2307/1772485
Ivir, Vladimir. 1988. “Translation and Backtranslation.” In Yugoslav General Linguistics
(Linguistic and Literary Studies in Eastern Europe), ed. by Milorad Radovanović, 131–143.
Amsterdam: John Benjamins. DOI: 10.1075/llsee.26.09ivi
Jakobsen, Arnt Lykke. 2006. “Research methods in translation: Translog.” In Computer
Keystroke Logging and Writing: Methods and Applications [Studies in Writing, 18], ed. by
Eva Lindgren, and Kirk P.H. Sullivan, 95–105. Oxford : Pergamon Press, 2006.
Jakobsen, Arnt Lykke, and Lasse Schou. 1999. “Translog Documentation.” In Probing the Process
in Translation: Methods and Results [Copenhagen Studies in Language, 24], ed. by Gyde
Hansen, 151–186. Copenhagen: Samfundslitteratur.
Jakobsen, Arnt Lykke, Kristian T.H. Jensen, and Inger M. Mees. 2007. “Comparing Modalities:
Idioms as a Case in Point.” In Interpreting Studies and Beyond, ed. by Franz Pöchhacker,
Arnt L. Jakobsen, and Inger M. Mees, 217–249. Copenhagen: Samfundslitteratur.
Malkiel, Brenda. 2009. “From Ántonia to My Ántonia: Tracking Self-Corrections with
Translog.” In Behind the Mind. Methods, Models and Results in Translation Process Research.
Copenhagen Studies in Language, ed. by Susanne Göpferich, Arnt L. Jakobsen, and Inger M.
Mees, vol. 37, 149–166. Copenhagen: Samfundslitteratur.
Mossop, Brian. 2001/2007. Revising and Editing for Translators. Manchester: St Jerome.
Shlesinger, Miriam, and Brenda Malkiel. 2005. “Comparing Modalities: Cognates as a Case in
point.” Across Languages and Cultures 6 (2): 173–193. DOI: 10.1556/Acr.6.2005.2.2
Tirkkonen-Condit, Sonja. 2005. “Evidence from Process Research.” Meta 50 (2): 405–414.
DOI: 10.7202/010990ar
Toury, Gideon. 1995. Descriptive Translation Studies and Beyond. Amsterdam: John Benjamins.
DOI: 10.1075/btl.4
The impact of process protocol self-analysis
on errors in the translation product
Erik Angelone
This chapter presents the initial results of an exploratory study examining the
efficacy of Integrated Problem and Decision Reporting (IPDR) logs, recorded
verbalizations, and screen recordings when used by translation students as revi-
sion tools for purposes of recognizing problems and mitigating errors in their
translations. Students were given the task of creating one of these three process
protocol types in conjunction with nine translations. They were then asked to
self-reflect on their problem-solving performance while paying particularly close
attention to a series of concrete problem indicator types found in the respective
protocols during a post-task retrospective session. The students had the chance
to make any desired revisions to their translations as a result of analyses before
submitting a final version. The revised versions were marked up for punctuation,
spelling, lexical, syntactic, stylistic, and mistranslation errors. The frequency of
errors for each of these types was calculated for each student in conjunction with
the process protocol utilized. The total number of errors in conjunction with
each process protocol type used, regardless of textual level, was also calculated.
Screen recording consistently proved to be the most efficacious process protocol
type in mitigating errors. The chapter concludes by positing several possible ex-
planations for the greater success rate of screen recording as a revision tool and
calls for screen recording to be implemented as a core component of a process-
oriented translator training curriculum.
1. Introduction
doi 10.1075/bct.77.07ang
2015 © John Benjamins Publishing Company
106 Erik Angelone
increasingly oriented toward fostering the trainee’s cognizance of how one trans-
lates, as shaped by the behaviors and strategies that tend to occur during the stages
of comprehension, transfer, and production. This cognizance, in large part, im-
plies that learners have a conscious awareness of the problems they encounter
while translating as well as of viable problem-solving strategies that can be ex-
ecuted in their presence. Large-scale longitudinal research projects, such as the re-
cently-completed Capturing Translation Processes project at the Zurich University
of Applied Sciences (2009–2012), the TransComp project at the University of
Graz (2008–2011), and the Copenhagen Retrospection Project at the Copenhagen
Business School (2004), have not only shed light on problems and problem-solv-
ing, but have also established some preliminary frameworks as to how learner
awareness of both can be trained through a variety of analyzable protocols, such as
keystroke logs (Jakobsen 1999; Hansen 2006) and screen recordings (Pym 2009;
Massey and Ehrensberger-Dow 2011; Angelone 2013). These protocols have con-
tinued to make their way from the research lab to the translator-training environ-
ment, and students now have the opportunity to engage in critical self-analysis of
their own problem-solving performance to an unprecedented extent.
Although translator training still tends to be biased in the direction of the
translation end product in and of itself (Dam-Jensen and Heine 2009: 1), particu-
larly when it comes to addressing errors, process-oriented translator training holds
great potential for both students and trainers when attempting to discern what po-
tentially triggered these errors in the first place. Process protocols can be utilized
as diagnostic tools, with problem indicators contained therein marking areas of
interest for purposes of problem-solving analysis. While a plethora of research has
shed light on how such protocols document various cognitive processes, such as
the utilization of internal and external support (PACTE 2005; Alves and Liparini
Campos 2009) and the allocation of cognitive effort during planning, drafting,
and revision (Pym 2009), relatively little research has been done on how translator
trainees can utilize protocols documenting their own performance for purposes of
problem awareness training and as a potential means for mitigating errors.
This chapter reports on the initial findings of an exploratory study aimed at
examining potential correlations between students’ utilization of different pro-
cess protocol types for purposes of problem recognition and error detection,
namely Integrated Problem and Decision Reporting (IPDR) logs (Gile 1994;
2004), recorded verbalizations of their thought processes, and screen recordings
documenting all on-screen activity, and the types and frequencies of errors ap-
pearing in corresponding translation products post retrospection and revision.
On a broader scale, this study explores efficacy in problem recognition and error
mitigation in a comparative manner when students use these three protocol types
as revision tools. It is hoped that the results will ultimately have an impact on
The impact of process protocol self-analysis 107
IPDR logs require the student to temporarily break away from the task at hand
for documentation purposes. My students reported occasionally losing their train
of thought as a result of this necessary disruption in natural translation flow, yet
were happy to have a concrete, personalized point of departure when reflecting on
their problems and problem-solving post task completion. Ultimately, the prom-
ise IPDR logs hold as a method for training problem awareness is dependent on
the student’s efficacy in recognizing and documenting problems in the first place.
Research has shown that problems in translation often go unnoticed by students
(Angelone 2010), implying that the logs they construct potentially run the risk of
being incomplete (Göpferich and Jääskeläinen 2009: 172) in terms of both overall
scope and the way in which problems and corresponding problem-solving strate-
gies are described. Interestingly, while incompleteness was in fact occasionally an
issue in the logs produced by my students, many of them expressed uncertainty
as to how much was too much when it came to documenting content in a log.
They mentioned feeling bogged down by the need to document “each and every
little problem.” Nevertheless, when it came time for them to self-reflect on the
110 Erik Angelone
problems they encountered, the data found on their IPDR logs were personalized,
learner-centered, and relatively straightforward.
degree of articulation slowly, but gradually drops off over the course of the transla-
tion task. In anticipation of this trend, the text lengths for all translation activities
in this study were deliberately kept relatively short (not exceeding 250 words) in
an attempt to mitigate fatigue.
Screen recording, one of the most recent methods used in translation process re-
search, is rapidly attracting interest within the field of process-oriented transla-
tor training as a user-friendly, ecologically valid tool for fostering the learner’s
self-awareness. For example, Pym (2009) had his students self-reflect on screen
recordings of their performance to foster greater awareness of planning, drafting,
and revision tendencies. Kujamäki (2010) had his students create unguided ret-
rospective commentaries to document their own performance as well as that of a
fellow student while watching respective screen recordings. The popularity of this
method as a revision tool is particularly evident at the Zurich Institute of Applied
Sciences, where all undergraduate and graduate students of translation have been
required to record some of their translation assignments using screen recording
(Massey and Ehrensberger-Dow 2011).
During a screen recording session, a software application unobtrusively re-
cords all on-screen activity that transpires over the course of the translation task.
Practically everything the translator does on the computer that is rendered on-
screen, from the CAT tools and online resources utilized for information retriev-
al, to the textual levels at which target text generation and revision occur, to the
pauses the occur while translating, is captured in video format and made readily
available for retrospective analysis in a highly-visual, naturalistic manner. Some of
the paramount problem indicators in screen recordings include extended pauses
in screen activity, information retrieval (using software or online), and revision
behavior.
My students, almost unanimously, stated that screen recording was their
preferred method (over IPDR logs and recorded verbalizations) when asked to
self-reflect on translation processes. They appreciated the naturalness of workflow
rendition and the fact that they did not have to deviate from the way they usually
translate for purposes of protocol compilation. Furthermore, my students felt it
was easiest to track their performance when watching screen recordings (as op-
posed to when analyzing IPDR logs or listening to recorded verbalizations) thanks
to the verisimilitude that the video playback offered, mirroring each step of the
process from beginning to end in a granular fashion.
112 Erik Angelone
The primary objective of this study is to document the relative efficacy of IPDR
logs, recorded verbalizations, and screen recordings as revision tools used by
4.2 Participants
This study tracked the performance of six first-year M.A. students (German–
English Translation) at Kent State University over the course of a semester (Fall
2011) in a translation practice course. Five of the students were English L1 and one
was German L1. All of the students had had formal coursework in translation at
the undergraduate level. Prior to taking the course in which data were collected,
none of the six participants had ever taken part in IPDR, think-aloud, or screen re-
cording activities. Although this study was not set up to explore the potential of an
expertise effect, it is worth noting that two of the students had some professional
experience (< five years with 100% of income earned through translation activity
on a regular basis), one working as a project manager and the other as a freelance
translator. The other four students had no professional experience.
Over the course of the semester, the students were asked to create and then self-
reflect on one of the following three protocol types in conjunction with a cor-
responding translation task: (1) an IPDR translation log, (2) a recording of all
articulated thought processes while translating, and (3) a screen recording docu-
menting all screen activity while translating. All protocols were created on the stu-
dents’ own computers in their own natural working environments. IPDR logs were
created using a word processing application. Think-aloud protocols were created
using a pre-installed voice recording application in conjunction with a built-in or
114 Erik Angelone
Each of the nine source texts was a self-contained, full-length text, approximately
250 words in length and of comparable difficulty. The translation tasks were not
timed, but students were asked to complete the translation task, corresponding
self-reflection activity, and any desired revisions in one sitting. Each text was care-
fully selected based on its exhibiting a balance of lexical, syntactic, and stylistic
Rich Points (PACTE 2011: 38) that were intuitively expected to pose challenges
in translation. Ultimately, for each self-reflection activity type, students translated
(1) a primarily operative text, (2) a balanced operative/informative text, and (3) a
primarily informative text. The self-reflection activity types were deliberately stag-
gered in such a manner so that students were using all three types at the beginning,
middle, and end of the semester. After making any desired revisions, students sub-
mitted their translations along with their respective process protocols.
2. http://www.bbsoftware.co.uk/BBFlashBack_FreePlayer.aspx
3. Classification based on the text typology model proposed by Reiss (1976).
The impact of process protocol self-analysis 115
After the students submitted their materials, errors in the translation products
were codified in terms of type (punctuation, spelling, lexis, syntax, style, mis-
translation) and frequency of each type. In this study, lexical errors included such
phenomena as using incorrect terminology, false cognates, and incorrect or weak
collocations. Syntactic errors involved errors in word order, verb tense, incom-
plete or run-on sentences, and subject-verb agreement. Stylistic errors stemmed
from inappropriate register (too formal or too colloquial), lexical or grammatical
inconsistency (when the text would call for consistency), or lexical and grammati-
cal redundancy (when the text would call for variation). Finally, mistranslation
errors resulted from inappropriate additions or omissions and errors resulting in
incorrect meaning transfer. Furthermore, errors were tracked in the following two
manners: (1) by individual student, according to error type/frequency and activity
type used, and (2) collectively, according to error type/frequency and activity type
used.
14
12
10
Frequency
8
IPDR Log
6 Think-aloud
Screen recording
4
0
Student 1 Student 2 Student 3 Student 4 Student 5 Student 6
Participants
Figure 1. Punctuation and spelling error frequency by self-reflection activity type
In Figure 1, the data show that four of the six students made the most punctu-
ation and spelling errors when using IPDR logs for purposes of self-reflection.
116 Erik Angelone
14
12
10
Frequency
8 IPDR Log
Think-aloud
6 Screen recording
0
Student 1 Student 2 Student 3 Student 4 Student 5 Student 6
Participants
In Figure 2, as well, the data reveal that four of the six students made the most
lexical errors when using IPDR logs, perhaps further suggesting that this particu-
lar method is less helpful in catching errors at the level of smaller textual units.
As it turns out, the vast majority of problems documented in IPDR logs by most
of the students had to do with syntactic and stylistic issues. Comments pertinent
to terminology issues appeared early in their logs, but then tapered off. On many
occasions, my students would ask exactly how much content should be included
in an IPDR log (“Do I really need to document each and every word that caused
problems”?). It almost seems as if students preferred documenting “bigger issues”
when entering content rather than entering each and every problematic lexical
item. Since there is no other point of departure besides the self-generated con-
tent found in a log, it is quite likely that undocumented lexical problems went
The impact of process protocol self-analysis 117
unnoticed in the product during retrospection and revision and, hence, remained
as errors.
The majority of students made the fewest number of lexical errors when us-
ing screen recording. In light of this method’s efficacy in mitigating punctuation,
spelling, and lexical errors, it seems as if screen recording engages a visual process-
ing system for relatively granular problem recognition, successfully guiding atten-
tion to detail at the level of such smaller textual units.
14
12
10
Frequency
8 IPDR Log
Think-aloud
6 Screen recording
0
Student 1 Student 2 Student 3 Student 4 Student 5 Student 6
Participants
14
12
10
Frequency
8 IPDR Log
Think-aloud
6 Screen recording
0
Student 1 Student 2 Student 3 Student 4 Student 5 Student 6
Participants
At the level of style (see Figure 4), the students’ utilization of both IPDR logs
and recorded articulations tended to yield relatively high frequencies of errors.
Interestingly, stylistic problems tended to be well-documented in the IPDR logs, yet
were seldom articulated in the recorded verbalizations. In the case of the latter, it
is plausible that errors occurred in the product because stylistic problems were not
directly articulated in the context of the process protocol. When it came time for
self-reflection, perhaps the students were looking for such direct (or indirect) artic-
ulations pertaining to stylistic problems, and were less inclined to look for problem
indicators as found in the form of speech disfluencies or extended periods of silence.
The fact that stylistic errors occurred when IPDR logs were used in spite of
the fact that problems of this kind were well-documented is intriguing. A higher
frequency of documentation did not result in fewer errors. Perhaps the external,
self-documentation format inherent to problem reporting when creating IPDR
logs is less conducive to the mitigation of stylistic errors, regardless of degree of
documentation.
As far as the mitigation of stylistic errors is concerned, once again, the utiliza-
tion of screen recording during retrospective analysis proved to be most effica-
cious for most of the students. At this stage of data analysis, it starts to become
evident that screen recording was the most conducive of the three self-reflection
activity types at mitigating errors not only at the sub-sentential level, but also at
the supra-sentential level.
The Figure 5 data indicate that the translations of four of the six students con-
tained the most mistranslation errors when recorded verbalizations were used for
purposes of self-reflection. While listening to the recordings in conjunction with
errors documented in the corresponding products, it quickly became apparent that
The impact of process protocol self-analysis 119
14
12
10
Frequency
8 IPDR Log
Think-aloud
6 Screen recording
0
Student 1 Student 2 Student 3 Student 4 Student 5 Student 6
Participants
students articulated at a steadily declining rate as the task progressed, while mis-
translation errors appeared at an increasing rate. In other words, during retrospec-
tive analysis, students generally had nothing concrete to go on besides extended
periods of silence in trying to locate potential mistranslation problems/errors,
which may explain why errors of this type went largely unnoticed. The utilization of
screen recording yielded considerably fewer mistranslation errors for the majority
of the students, further warranting a closer examination of what potentially makes
this method particularly conducive vis-à-vis IPDR logs and recorded verbalizations.
An examination of error frequencies in aggregate for each of the five error
types documented in this study (see Figure 6) indicates that the utilization of
screen recording during retrospective self-reflection resulted in the fewest number
70
60
50
Spelling/Punc.
Frequency
40 Lexis
Syntax
30 Style
Mistrans.
20
10
0
IPDR Log Think-aloud Screen recording
Self-reflection activity type
Figure 6. Overall error frequency by self-reflection activity type
120 Erik Angelone
of errors across the board. At this stage, it is important to posit some further pos-
sible explanations for the efficacy of screen recording insofar as problem recog-
nition and error mitigation are concerned. This starts with a close examination
of what transpires during retrospection involving screen recording that does not
transpire in a similar fashion when using IPDR logs and recorded verbalizations.
First and foremost, it is possible that the guided (cued) eye movement that is
activated when analyzing a screen recording makes problem indicators (and er-
rors) embedded in the process protocol more salient. When working with screen
recordings, the students’ visual attention is inherently drawn to problems as they
unfold in real-time. As a result, even in the absence of concrete problem indicators
(in the case of screen recording, look-ups, extended pauses in screen activity, and
revision behavior, among others), this guided visual attention catalyzes a height-
ened state of cognitive awareness, to the extent that a highly effective problem
recognition monitor remains activated throughout the retrospection session. The
results of this study suggest that guided visual attention is a more optimal channel
for error mitigation than the the audio processing system activated when analyz-
ing recorded verbalizations and the more offline processing system activated when
analyzing IPDR log content.
There are additional possible explanations for the greater overall success of
screen recording as a revision tool above and beyond this inherent guided visual
cuing system. For example, screen recording is the only method of the three ex-
amined in this study that did not require some degree of parallel processing on
the student’s part during translation and concurrent process protocol generation.
The protocol was basically generated automatically by the software application
and students were not required to perform a simultaneous task while translating
(documenting problems and problem solving in the case of IPDR logs and articu-
lating thought processes in the case of recorded verbalizations). This may have
resulted in a lower cognitive load, and, in turn, less translation errors in the first
place. Finally, another possible explanation for screen recording’s success might be
that the problem indicators encoded in this particular process protocol type (look-
ups, extended pauses in screen activity, and revision behavior) are somehow more
overt in the sense that they draw greater attention. A more granular follow-up
analysis of problem indicator frequencies in the various protocol types in relation
to revisions made (or not made) to corresponding translation passages in their
presence will hopefully shed more light in this regard.
The impact of process protocol self-analysis 121
The data obtained in this study point towards screen recording as being the most
efficacious self-reflection activity type for purposes of error mitigation. Larger-
scale follow-up studies involving a higher number of participants and additional
language pairs are needed at this stage to further assess the efficacy of all three
self-reflection activity types (as well as others not included in this pilot study).
Although this study is primarily geared towards translation pedagogy, it might
also be interesting to explore the impact of various translator and text variables on
error patterns in conjunction with the three process protocol types. For example,
is screen recording equally effective as a revision tool for problem recognition and
error mitigation when used by professionals as opposed to students? What im-
pact does text length have on process protocol efficacy? How does directionality
(translating into L1 or L2) correlate with process protocol efficacy? In summary, it
is hoped that the findings obtained here will encourage translator trainers who are
dedicated to a process-oriented approach to consider integrating screen recording
as a fundamental tool to foster the learner’s problem recognition capacities, pro-
cess awareness, and metacognition in general.
References
Alves, Fabio, and Tania Liparini Campos. 2009. “Translation Technology in Time: Investigating
the Impact of Translation Memory Systems and Time Pressure on Types of Internal and
External Support.” In Behind the Mind: Methods, Models and Results in Translation Process
Research: Copenhagen Studies in Language 37, ed. by Susanne Göpferich, Arnt L. Jakobsen,
and Inger M. Mees, 191–218. Copenhagen: Samfundslitteratur.
Angelone, Erik. 2010. “Uncertainty, Uncertainty Management and Metacognitive Problem
Solving in the Translation Task.” In Translation and Cognition, ed. by Gregory M. Shreve,
and Erik Angelone, 17–40. Amsterdam: John Benjamins. DOI: 10.1075/ata.xv.03ang
Angelone, Erik. 2013. “Watching and Learning from ‘Virtual Professionals’: Utilising Screen
Recording in Process-oriented Translator Training.” In New Prospects and Perspectives
for Educating Language Mediators, ed. by Don Kiraly, Silvia Hansen-Schirra, and Karin
Maksymski, 139–155. Tübingen: Gunter Narr.
Angelone, Erik, and Gregory Shreve. 2011. “Uncertainty Management, Metacognitive Bundling
in Problem-Solving, and Translation Quality.” In Cognitive Explorations of Translation, ed.
by Sharon O’Brien, 108–130. London: Continuum.
Dam-Jensen, Helle, and Carmen Heine. 2009. “Process Research Methods and their Application
in the Didactics of Text Production and Translation: Shedding Light on the Use of Research
Methods in the University Classroom.” trans-kom 2 (1): 1–25. http://www.trans-kom.eu/
ihv_02_01_2009.html
Dragsted, Barbara, and Inge G. Hansen. 2008. “Comprehension and Production in Translation:
A Pilot Study on Segmentation and the Coordination of Reading and Writing Processes.”
122 Erik Angelone
In Looking at Eyes: Eye Tracking Studies of Reading and Translation Processes, Copenhagen
Studies of Language 36, ed. by Susanne Göpferich, Arnt L. Jakobsen, and Inger M. Mees,
9–29. Copenhagen: Samfundslitteratur.
Gile, Daniel. 1994. “The Process-Oriented Approach in Translation Training.” In Teaching
Translation and Interpreting 2, ed. by Cay Dollerup, and Annette Lindegaard, 107–112.
Amsterdam: John Benjamins. DOI: 10.1075/btl.5.17gil
Gile, Daniel. 2004. “Integrated Problem and Decision Reporting as a Translator Training Tool.”
JoSTrans 2: 2–20. http://www.jostrans.org/issue02/art_gile.php.
Göpferich, Susanne, and Riitta Jääskeläinen. 2009. “Process Research into the Development of
Translation Competence: Where are We, and Where Do We Need to Go?” Across Languages
and Cultures 10 (2): 169–191. DOI: 10.1556/Acr.10.2009.2.1
Hansen, Gyde. 2006. “Retrospection Methods in Translator Training and Translation Research.”
JoSTrans 5: 2–41. http://www.jostrans.org/issue05/art_hansen.php.
Jakobsen, Arnt Lykke. 1999. “Logging Target Text Production with Translog.” In Probing the
Process in Translation, ed. by Gyde Hansen, 9–20. Copenhagen: Sammfundslitteratur.
Jääskeläinen, Riitta. 2000. “Focus on Methodology in Think-aloud Studies on Translating.” In
Tapping and Mapping the Processes of Translation and Interpreting, ed. by Sonja Tirkkonen-
Condit, and Riitta Jääskeläinen, 71–82. Amsterdam: John Benjamins.
DOI: 10.1075/btl.37.08jaa
Kujamäki, Pekka. 2010. “Auf der Suche nach treffenden Worten. Bildschirmvideos als Mittel zur
Analyse von studentischen Übersetzungsleistungen.” In Infinite Kontrastive Hypothesen.
Beiträge des Festsymposiums zum 60. Geburtstag von Irma Hyvärinen, ed. by Leena
Kolehmainen, Hartmut Lenk, and Annikki Liimatainen, 141–164. Frankfurt: Peter Lang.
Massey, Gary, and Maureen Ehrensberger-Dow. 2011. “Commenting on Translation:
Implications for Translator Training.” JoSTrans 16: 26–41. http://www.jostrans.org/issue16/
art_massey_ehrensberger_dow.php.
O’Brien, Sharon. 2008. “Processing Fuzzy Matches in Translation Memory Tools: An Eye-
Tracking Analysis.” In Looking at Eyes: Eye Tracking Studies of Reading and Translation
Processes, Copenhagen Studies of Language 36, ed. by Susanne Göpferich, Arnt L. Jakobsen,
and Inger M. Mees, 79–102. Copenhagen: Samfundslitteratur.
PACTE. 2005. “Investigating Translation Competence: Conceptual and Methodological Issues.”
Meta 50(2): 609–619. DOI: 10.7202/011004ar
PACTE. 2011. “PACTE Translation Competence Model: Translation Project and Dynamic
Translation Index.” In Cognitive Explorations of Translation, ed. by Sharon O’Brien, 30–56.
London: Continuum.
Pym, Anthony. 2009. “Using Process Studies in Translator Training: Self-discovery through
Lousy Experiments.” In Methodology, Technology and Innovation in Translation Process
Research: Copenhagen Studies in Language 38, ed. by Susanne Göpferich, Fabio Alves, and
Inger M. Mees, 135–156. Copenhagen: Samfundslitteratur.
Reiss, Katharina. 1976. Texttyp und Übersetzungsmethode: Der operative Text. Kronberg:
Skriptorverlag.
Sharmin, Selina, Oleg Spakov, Kari-Jouko Räihä, and Arnt L. Jakobsen. 2008. “Where on the
Screen do Translation Students Look While Translating, and for How Long.” In Looking
at Eyes: Eye Tracking Studies of Reading and Translation Processes, Copenhagen Studies
of Language 36, ed. by Susanne Göpferich, Arnt L. Jakobsen, and Inger M. Mees, 31–51.
Copenhagen: Samfundslitteratur.
The impact of process protocol self-analysis 123
Sarah Eardley-Weaver
1. Introduction
Opera translation is evolving to embrace all audiences, including the blind and
partially-sighted, and the process of translating this multisemiotic art form is ad-
vancing towards widened accessibility. As numbers of people with sight loss are
rising in aging populations, innovative translation techniques to provide for this
growing minority group are becoming more widespread in various audiovisual
doi 10.1075/bct.77.08ear
2015 © John Benjamins Publishing Company
126 Sarah Eardley-Weaver
(AV) media, such as television and opera (Díaz-Cintas and Anderman 2009).1
These developments reflect an increasing general acknowledgement of accessibil-
ity issues in today’s society, also demonstrated by legislation such as the European
‘Audiovisual Media Services Directive’ (European Parliament 2010) and the UK
‘Disability Discrimination Act 2005’ (Great Britain Parliament 2005). Similarly,
in the field of translation studies, the concept of translation and the scope of AV
translation research are expanding in line with social and politico-economic ad-
vancements to encompass media accessibility (Remael 2010). However, the pro-
cess of raising awareness about access facilities, such as opera translation for blind
and partially-sighted patrons (BPSPs), is still in progress (Greening and Rolph
2007: 127; Díaz-Cintas and Anderman 2009: 6). It is within this sociocultural con-
text that this chapter investigates the process of translation of live opera for BPSPs
in the UK, employing an original research design. This research design is based
on the notion of the translation process as a network (see Section 2) and takes
into account that analyzing audience reception is an integral part of translation
process research (see Künzli and Ehrensberger-Dow 2011: 188; Tiselius and Jenset
2011: 273; Mateo 1997: 99–102). The central focus in this chapter is the examina-
tion of audience feedback on opera translation facilities for BPSPs and of the im-
pact of audience reception on the translation process.
Firstly, the notion of translation in this study is problematized in relation to
the two main opera translation modalities targeted at BPSPs in the UK: audio de-
scription (AD) and touch tour (ToTo). The theoretical framework is established,
and the translation processes of ToTos and of the different types of AD currently
available in UK opera houses are examined. Secondly, the audience’s response to
the translations is investigated with reference to the reception project carried out
in collaboration with the UK opera company Opera North (ON) at performances
of Bizet’s Carmen in May 2011. Finally, the study findings and methodological ap-
proach are discussed in relation to the wider research framework, and conclusions
are drawn regarding their impact on the translation process.
1. Results show “more than a doubling (115% increase over 2010) in the numbers of people with
partial sight and blindness in the UK, to nearly 4 million people by 2050” (Access Economics
2009: 44).
Opening eyes to opera 127
2. Ethnographic research is characterized by the collection of data in natural settings using a
variety of techniques to explore and understand human social worlds from the perspective of
the research participants (Hammersley and Atkinson 2007: 2; Hubscher-Davidson 2011: 4).
128 Sarah Eardley-Weaver
accessible mode of reading aloud, or voicing, subtitles.” Since these methods are
rarely used in the UK, the focus in this chapter is restricted to AD and ToTos.
AUDIO
DESCRIPTION
(AD)
Audio
Audio ‘through’
introduction
description
(AI)
Live Pre-
recorded
Condensed Extended
version (to be version (hard
played in the copy &/or
theatre) online)
Figure 1. Subdivisions of opera AD used in the UK.
130 Sarah Eardley-Weaver
Thus, the roles of translator and target receiver merge. This merging of roles is rare
in other translation contexts, although a parallel can be drawn with fansubbing,
“the most important form of crowdsourcing” translation in which members of a
community of fans provide subtitles of a film or program and make them avail-
able online for this community (European Commission 2012: 45). By recogniz-
ing that fans are not passive consumers and are confident about the contributions
they can make to the translation process, the term ‘prosumer’ has been used to
describe fansubbers, as the boundary between producer and consumer is blurred
(European Commission 2012: 31). There is a similarity with ToTo participants,
as they are given opportunities to ask questions and are active in the translation
process. Another parallel can be drawn with crowdsourcing, as a ToTo is a joint
translation process performed by the multifarious network of the aforementioned
agents. The decisions of the various translating parties will be mutually influ-
ential, and this is similar to crowdsourcing translation in which work is carried
out collaboratively to help other members of the community to access products
(European Commission 2012: 23).
132 Sarah Eardley-Weaver
Where used, ToTos have certainly become part of the multisemiotic opera
translation process for BPSPs. Like any translation process, social and contextual
factors have a significant impact on the format and route of the ToTo. For instance,
the choice of elements to touch may be determined by safety concerns (e.g., set
fragility or dangerous objects) or by time constraints. Similarly to any other trans-
lation strategy — such as the impossibility of using computer-aided translation
(CAT) tools for certain language combinations — practical issues determine the
feasibility of consistently providing ToTos because some opera houses do not have
an accessible stage. In these situations, guiding BPSPs may prove to be dangerous.
(UK). The reception project aimed to evaluate the effectiveness of these opera ac-
cess facilities from the audience’s perspective. In this chapter, the objective is to
analyze the translation processes of AD and ToTos by studying BPSPs’ responses to
these modalities. The hypothesis is that while existing translation modalities meet
audience needs to a certain degree, there are individual elements, especially relat-
ing to the notion of a shared and inclusive experience, which could be improved.
The overall research design for this reception project drew on Oppenheim’s model
for data collection and survey (1992: 6–8). A combined methodology using a two-
fold design (Creswell 2003: 114) which involved questionnaires and feedback via
discussion was employed. These data collection methods were chosen as they were
considered the most ecologically valid and reliable way to investigate reception of
the actual translation techniques in their natural environment, given the live nature
of AD and ToTos and the exploratory approach of this study. Feedback was gath-
ered from users of the AD and/or ToTo, including BPSPs and their sighted com-
panions. Respondents were asked to complete two questionnaires: a preliminary
questionnaire prior to the performance to gather demographic details and other
personal information, and a post-performance “group-administered” question-
naire (Oppenheim 1992: 103) to gain feedback on the access facilities provided.3
In all questionnaires there were both closed-ended and open-ended items.
In the post-performance survey, questions required graded responses using
a variation on the Likert scale technique which replaced the standard set of re-
sponses (strongly agree to strongly disagree) with the numerical scale 1=not at all,
2=not very much, 3=so-so, 4=quite a lot, and 5=very much. Each question on the
hard copy post-performance questionnaires also included room for comments,
and further qualitative data was gathered from informal group discussion.
In this chapter, the initial analysis of the questionnaire results is presented,
combining quantitative and qualitative data.4 Firstly, the profile of the whole group,
including BPSPs and their sighted companions, is briefly discussed. Secondly, data
collected regarding the BPSPs’ reception of the AD and ToTo is analyzed. The fo-
cus of this data analysis is three-fold, concerning the audience reception of: (1)
different types of AD; (2) the translation of features provoking an emotional reac-
tion, such as humor and shock; and (3) the ToTo. Humor and shock are focused on
because this production of Carmen included dark comedy and explicit references
4. The AD and ToTo at the survey performances were provided by VocalEyes (VocalEyes 2011).
134 Sarah Eardley-Weaver
The profile of the whole group is discussed here because the opera experience is
a social activity shared among BPSPs and their companions, who may, moreover,
be agents in the translation process (see Section 2.2). In total, 32 people partici-
pated in the questionnaire on AD and ToTo: 11 sighted patrons and 21 BPSPs,
of whom 4 were totally blind, 16 were legally blind and 1 was partially-sighted.5
Among the BPSPs, 8 were male and 13 were female. Among the sighted patrons,
there were 5 males and 6 females. Figure 3 shows the distribution of age across the
whole group: half of the participants were aged 66 and above. These results could
arguably suggest a lack of awareness among younger people of the availability of
Age groups
10 – 17 18 – 25
26 – 35
36 – 45
10 − 17
18 − 25
26 − 35
36 – 45
66 + 46 – 55
46 – 55 56 – 65
66 +
56 – 65
Figure 3. Distribution of age in the participants using AD and touch tour.
5. Categories are based on definitions of visual acuity by the RNIB (Access Economics
2009: 11–12).
Opening eyes to opera 135
these access facilities, although many other factors could explain the low numbers
of younger AD and ToTo users, such as the opera genre itself, or this particular
production of Carmen, which was given a 12+ rating by ON. Interestingly, com-
ments made by respondents highlighted a general lack of awareness and publicity
regarding access facilities. One participant commented that “access is not always
advertised, you have to ask” and another remarked “I’d forgotten about that” with
reference to the pre-recorded preparatory AI.
The results analyzed in subsequent sections refer to responses from the 21
BPSPs participating in the survey. The analysis involves examination of quanti-
tative and qualitative data collected from the post-performance questionnaire in
the form of graded responses and comments. This accumulated quantitative and
qualitative data of 21 BPSPs is significant, because on average 27 BPSPs register
to use AD and/or ToTo annually at ON productions.6 Due to the nature of the
ethnographic research approach adopted, results are necessarily situated and cau-
tion is needed in making inferences (Hubscher-Davidson 2011: 13). However, as
Blommaert points out, referring to Pierre Bourdieu’s views, “there is a clear sug-
gestion that single cases, even if they don’t speak to the totality of the population
or the system, can speak to theory” (2005: 228).
Figure 4 displays the average scores given by the BPSPs to questions regarding the
amount and helpfulness of the different types of AD listened to. Standalone AI was
not provided at the reception project performances, so this translation modality
cannot be investigated. However, live AI and pre-recorded preparatory AI were
5 5
4 4
Through description
3 3 Live AI
Preparatory AI
2 2
1 1
Figure 4. Average perceived amount and helpfulness of different types of AD listened to.
This observation highlights the benefit of increased familiarity with the opera and
production gained by listening to the preparatory AI. Similar comments empha-
sizing this element were made regarding the live AI, such as “it’s a good reminder
even if you’ve heard it before.” Several participants noted the value of both of these
types of AD in setting the scene. Regarding through-description, participants
commented on the helpfulness of the immediacy of the description and commu-
nication of details as they happen.
The average scores given by the BPSPs to questions concerning emotional response
to the opera are displayed in the first three columns in Figure 5.
The first column in Figure 5 shows the average response to a question asking
participants if they felt emotionally engaged with the opera on this occasion. The
score of 3.7 reveals a fairly positive response, although it also suggests that some
participants did not feel that they were able to fully experience the emotion of the
opera. The following remark by a survey participant referring to the second-hand
experience in AD for BPSPs, although emphasizing the social role of translation
in contributing to inclusion, might provide a partial explanation for this result.
With AD you are experiencing something second-hand. As a listener you don’t
really have an opinion, you don’t have your own reaction to it, you have the audio
describer’s reaction to it. But it really helps to be able to share visual aspects with
Opening eyes to opera 137
4.5
3.5
2.5
1.5
1
Emotion Humor Shock Enjoyment of Helpfulness of
touch tour touch tour
Figure 5. Average response to the translation of emotional aspects of the opera and to the
touch tour
This comment highlights the paradoxical nature of the reception of AD, the in-
evitable subjectivity of the audio describer’s role and the interpretative element
in the AD process as a whole. The AD allows the BPSP to interpret the opera
independently from other audience members, but the BPSP’s interpretation is not
independent because it is influenced by the audio describer’s rendering; it is sec-
ond-hand. Another BPSP alluded to this distancing effect, but this time in relation
to the translation of humor in the AD. The participant stated:
It’s quite hard to convey humor in the third person because it’s reported speech.
It’s hard to portray something that’s funny because of the timing of the punchline
— there’s a delay in the delivery of the humor. […] The reaction of the audience is
not simultaneous and so it loses impact.
The average scores of 3.8 and 4.0 given to questions about the AD’s helpfulness
in conveying humorous and shocking aspects, respectively (see columns 2 and 3,
Figure 5), indicate a positive response overall, although participants’ comments
reveal mixed reviews. Acclamations included, for example, “AD has brought the-
ater to life for me. I received the humor of Carmen well because of the AD,” and “I
liked the comic moment in the AD where it said something like ‘Zuniga sits there
in all his massiveness.’ I don’t mind subjectivity — it’s inevitable so why not have
fun!” On the other hand, there were also mildly critical comments such as “I don’t
think that was as good. I imagine humor is difficult to convey in AD.” Similarly, re-
garding shocking elements, several BPSPs’ comments revealed a lack of awareness
138 Sarah Eardley-Weaver
of any shocking details, such as nudity and allusions to sexual violence, suggesting
that the translation in the AD was not entirely successful for some. In particular,
the lack of a shared reception of this feature was noted. For instance, one partici-
pant highlighted the difference in verbal and visual communication:
If I compare my feelings with my [sighted] wife’s feelings, my wife found it a vio-
lently disturbing production. I didn’t receive this idea of physical violence. If one
is told ‘Zuniga kicks Carmen after she has fallen over’ it has less impact than if
you see it. I didn’t wince but it did make me think. I did notice the use of shock-
ing language in the AD to convey the shock of the production, for example ‘she
punched him in the balls’ — that was a clue that the gloves were off. However,
although it gave a hint, still the language used didn’t help me receive the same
impact of visual violence.
This remark also raises questions concerning the notion of a shared experience
which is particularly significant in the context of live performance received in a
group context, as demonstrated by other participants’ comments. There were also
very positive remarks regarding the helpfulness of AD in conveying shocking as-
pects, such as:
The AD was very helpful for this, not just for the stabbing of Carmen but also the
sinister characters Escamillo and the leader of the guards Zuniga. I got a sense of
the sinister characters, and also the fact that Carmen is a bit of an outcast — the
idea of racism. That shocking aspect came across loud and clear. The use of lan-
guage and not using euphemistic language was also effective — they [the audio
describers] said it as it was ‘she kicked him in the balls.’
Columns 4 and 5 in Figure 5 show the average scores given by the BPSPs who at-
tended the ToTo (11 out of 21) to questions about enjoyment and helpfulness of
the ToTo. The average score given for enjoyment is marginally higher than that
for helpfulness. This result might be explained by the respondents’ emphasis on
the “fun” aspect of the ToTo, as mentioned by several participants, referring, for
example, to the enjoyment of speaking to the actress who attended the ToTo. Also,
Opening eyes to opera 139
while some explicitly remarked on the helpfulness of the ToTo in providing op-
portunities to ask questions and in visualizing the stage, the following participant
comment highlights a limitation of the ToTo: “it depends on the set as to whether
the ToTo is helpful.” Further audience reception research is needed regarding fac-
tors affecting the helpfulness and translation processes of ToTos, such as the set,
agents involved, and numbers and ages of patrons. Studies focusing on BPSPs’
preferences regarding interpretative description and explanation of elements of
the director’s and design team’s vision are also required. From this perspective, the
following comment is also noteworthy:
I have some reservations about ToTos but I can’t resist them. It’s so nice to be part
of the business. It adds to the experience of the performance. But it’s a bit like
seeing a film of a book you like. In some ways I prefer to create my own mental
image of the stage from the AD without interference of the tactile experience and
details on the ToTo. […]
The same BPSP continued as follows, emphasizing the social aspect of the transla-
tion process in providing an inclusive experience:
As a totally blind person who has never seen, all the details given about color, cos-
tumes, props, and so on are not particularly helpful for me. I realize though that
audio describers have to cater for all audiences with differing degrees and types
of visual ability. For me, the important part of the opera experience is what you
talk about with other members of the audience. That is what is so valuable about
ToTos and AD — that you can share the experience with the rest of the audience.
This remark also emphasizes the need for further research into the varying re-
quirements of a diverse audience.
The results support the initial research hypothesis that, while current translation
processes in AD and ToTos meet BPSPs’ needs to a certain degree, improvements
could be made. The quantitative and qualitative data suggest a generally positive
response from BPSPs to AD and ToTos, but comments highlight some of their
perceived limitations. While it must be acknowledged that these findings are con-
text-bound due to the ethnographic research approach adopted (Hammersley and
Atkinson 2007: 159; Glaser and Strauss 1967), they are clearly valuable, especially
given the pioneering nature of the reception project.
In relation to the actor-network theoretical framework discussed in Section 2,
the findings corroborate the idea that AD and ToTos form part of a translation
network. For example, this notion is tacitly understood by BPSPs, as demonstrated
140 Sarah Eardley-Weaver
engagement? Are there core features that patrons consider fundamental to any
ToTo or other details that should not be revealed? Furthermore, the theoretical
framework drawing on actor-network theory and grounded theory provides a
foundation from which to develop larger-scale audience reception studies, also
examining opera AD and ToTos in conjunction with other facilities such as au-
dio subtitling and braille program notes. This theoretical framework may also be
transferable to other translation contexts in view of the aforementioned analogies
between AD or ToTos and other forms of translation such as crowdsourcing and
sign interpreting. In addition, research bringing together investigation into the
translator’s role and audience reception is needed in order to unite and promote
dialogue between these agents in the translation network and to further explore
the sociocultural, collaborative aspects of the translation process. As Gambier
states: “in this changing mediascape, the translators have and will have a major
role, if they fully realize their socio-cultural function” (2006: 7).
4. Concluding remarks
mutual understanding among the agents in the network, which may contribute to
changes in translation practice. In addition, reception research heightens aware-
ness of the access facilities in general and of any innovations in the field, thus
contributing to a universally inclusive, multisemiotic opera experience. In sum-
mary, audience reception studies may open eyes to the collaborative, complex, and
dynamic nature of the opera translation network.
References
Access Economics. 2009. “Future Sight Loss UK (1): The Economic Impact of Partial Sight and
Blindness in the UK Adult Population.” RNIB. http://www.rnib.org.uk/knowledge-and-
research-hub/research-reports/general-research/future-sight-loss-uk-1.
Blommaert, Jan. 2005. “Bourdieu the Ethnographer: the Ethnographic Grounding of Habitus
and Voice.” The Translator 11 (2): 219–236. DOI: 10.1080/13556509.2005.10799199
Braun, Sabine. 2008. “Audiodescription Research: State of the Art and Beyond.” Translation
Studies in the New Millennium 6: 14–30.
Buzelin, Hélène. 2005. “Unexpected Allies: How Latour’s Network Theory Could Complement
Bourdieusian Analyses in Translation Studies.” The Translator 11 (2): 193–218.
DOI: 10.1080/13556509.2005.10799198
Cabeza, Cristóbal. 2010. “Opera Audio Description at Barcelona’s Liceu Theatre.” In New
Insights into Audiovisual Translation and Media Accessibility, ed. by Jorge Díaz-Cintas, and
Josélia Neves, 227–238. Amsterdam: Rodopi.
Chesterman, Andrew. 2006. “Questions in the Sociology of Translation.” In Translation Studies
at the Interface of Disciplines, ed. by João Ferreira Duarte, Alexandra Assis Rosa, and Teresa
Seruya, 9–27. Amsterdam: John Benjamins. DOI: 10.1075/btl.68.03che
Chiaro, Delia. 2006. “Verbally Expressed Humour on Screen: Reflections on Translation and
Reception.” JoSTrans 6: 198–208.
Chiaro, Delia. 2007. “The Case of Dubbed Comedy in Italy.” In Doubts and Directions in
Translation Studies, ed. by Yves Gambier, Miriam Schlesinger, and Radegundis Stolze, 137–
152. Amsterdam: John Benjamins. DOI: 10.1075/btl.72.16chi
Creswell, John W. 2003. Research Design: Qualitative, Quantitative, and Mixed Methods
Approaches. Thousand Oaks, CA: Sage.
De Koster, Karin, and Volkmar Mühleis. 2007. “Intersensorial Translation: Visual Art Made up
by Words.” In Media for All: Subtitling for the Deaf, Audio Description and Sign Language,
ed. by Jorge Díaz-Cintas, Pilar Orero, and Aline Remael, 189–200. Amsterdam: Rodopi.
Denton, John. 2007. “Audience Reception of Translated Audiovisual Texts and the Problem of
Visual and Verbal Shared/Unshared Knowledge.” In La Lingua Invisibile. Aspetti Teorici e
Tecnici del Doppiaggio in Italia [Invisible Language. Theoretical and Technical Aspects of
Dubbing in Italy], ed. by Giuseppe Massara, 23–36. Roma: Nuova Editrice Universitaria.
Díaz-Cintas, Jorge, and Gunilla Anderman (eds). 2009. Audiovisual Translation. Language
Transfer on Screen. Basingstoke: Palgrave Macmillan.
Díaz-Cintas, Jorge, Pilar Orero, and Aline Remael (eds). 2007. Media for All: Subtitling for the
Deaf, Audio Description and Sign Language. Amsterdam: Rodopi.
Opening eyes to opera 143
Dickinson, Jules, and Graham H. Turner. 2008. “Sign Language Interpreters and Role Conflict in
the Workplace.” In Crossing Borders in Community Interpreting: Definitions and Dilemmas,
ed. by Carmen Valero Garcés, and Anne Martin, 231–244. Amsterdam: John Benjamins.
DOI: 10.1075/btl.76.12dic
Di Giovanni, Elena. 2011. “Connecting the Dots in Audiovisual Translation Research:
Translation, Reception, Accessibility and Children.” In Between Text and Receiver:
Translation and Accessibility, ed. by Elena Di Giovanni, 7–19. Frankfurt am Main: Peter
Lang.
Eardley-Weaver, Sarah. 2010. “Opening Doors to Opera: The Strategies, Challenges and General
Role of the Translator.” InTralinea 12: 1–8.
Eardley-Weaver, Sarah. 2014. Lifting the Curtain on Opera Translation and Accessibility:
Translating Opera for Audiences with Varying Sensory Ability. Durham: Durham theses,
Durham University. Available at Durham E-Theses Online: http://etheses.dur.ac.uk/10590/.
European Commission. 2012. “Crowdsourcing Translation.” Directorate General for Translation.
http://bookshop.europa.eu/en/crowdsourcing-translation-pbHC3112733/.
European Parliament, Council. 2010. “Audiovisual Media Services Directive 2010/13/EU.”
European Parliament. http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=CELEX:32
010L0013:EN:NOT.
Gambier, Yves. 2006. “Multimodality and Audiovisual Translation.” MuTra 2006 — AV
Translation Scenarios: Conference Proceedings. 1–8. http://www.euroconferences.info/
proceedings/2006_Proceedings/2006_Gambier_Yves.pdf.
Gambier, Yves, and Henrik Gottlieb (eds). 2001. (Multi) Media Translation: Concepts, Practices,
and Research. Amsterdam: John Benjamins. DOI: 10.1075/btl.34
Gerzymisch-Arbogast, Heidrun. 2005. “Introducing Multidimensional Translation.” MuTra
2005 — Challenges of Multidimensional Translation: Conference Proceedings. 1–15. http://
www.euroconferences.info/proceedings/2005_Proceedings/2005_GerzymischArbogast_
Heidrun.pdf.
Glaser, Barney G., and Anselm L. Strauss. 1967. The Discovery of Grounded Theory: Strategies for
Qualitative Research. New York, NY: Aldine de Gruyter.
Gottlieb, Henrik. 2005. “Multidimensional Translation: Semantics turned Semiotics.” MuTra
2005 — Challenges of Multidimensional Translation: Conference Proceedings. 1–29. http://
www.euroconferences.info/proceedings/2005_Proceedings/2005_Gottlieb_Henrik.pdf.
Great Britain Parliament. 2005. “Disability Discrimination Act 2005: Chapter 13.” The Stationery
Office. http://www.legislation.gov.uk/ukpga/2005/13.
Greening, Joan, and Deborah Rolph. 2007. “Accessibility — Raising Awareness of Audio
Description in the UK.” In Media for All: Subtitling for the Deaf, Audio Description and Sign
Language, ed. by Jorge Díaz-Cintas, Pilar Orero, and Aline Remael, 127–138. Amsterdam:
Rodopi.
Hammersley, Martyn, and Paul Atkinson. 2007. Ethnography: Principles in Practice. New York,
NY: Routledge.
Hansen, Gyde. 2008. “The Dialogue in Translation Process Research.” In Translation and
Cultural Diversity: Proceedings of the Eighteenth FIT World Congress, Shanghai, 2–7
August 2008, 386–397. Shanghai: Foreign Languages Press.
Holland, Andrew. 2009. “Audio Description in the Theatre and Visual Arts: Images into Words.”
In Audiovisual Translation. Language Transfer on Screen, ed. by Jorge Díaz-Cintas, and
Gunilla Anderman, 170–185. Basingstoke: Palgrave Macmillan.
144 Sarah Eardley-Weaver
doi 10.1075/bct.77.10not
2015 © John Benjamins Publishing Company
Index
recorded verbalizations 105, self-reflection 107, 108, 109, 113, transcoding 37, 55
106, 108, 110, 111, 112, 118, 114, 115, 116, 117, 118, 119, 121 transfer 11, 12, 26, 46, 54, 55, 87,
119, 120 self-report personality test 70 102, 106, 107, 108, 112, 115, 126
recursive model of translation self-revision 4, 16, 85, 87, 88, 89, transfer links 43, 54, 59
3, 21, 24, 35, 37 90, 91, 92, 93, 94, 95, 96, 97, transfer processing 56
reflex-like equivalence relations 98, 99, 100, 101, 102 transference 21, 24, 29
55 semantic representations 24, 28 translation act 2, 3, 7, 8, 10, 11,
regeneration 27, 29, 37 semi-experts 78 13, 14, 15, 16, 64, 67, 71
regeneration hypothesis 26, 27 sensitivity 75, 76 translation brief 31, 88
regressions 26 sensory (communication) translation competence 14,
retrieval 13, 27, 43, 55, 56, 57, 58, channel 127, 130 66, 72
108, 110, 111 shared representations 21, 23, translation event 2, 3, 4, 7, 8, 9,
retrospection 106, 108, 116, 24, 25, 26, 27, 29, 30, 31, 34, 14, 15, 16, 17, 18
117, 120 36, 37, 38, 39 translation model 7, 22, 24
retrospective analysis 111, 118, sight translation 12 translation practice 2, 7, 17,
119 sign interpreting 132, 141 113, 142
retrospective interview 15, 49, situated cognition 8 translation problem 4, 7, 10, 11,
50, 51, 52, 53 Skopos Theory 16 12, 13, 23, 66, 76, 85, 101
retrospective methods 18 sociological (translation) event translation purpose 31
retrospective self-reflection 2, 3, 4, 7, 8, 10, 15 translation routines 55
107, 108, 119 solution evaluation 108 translation strategies 23, 134
reverse engineering (models) somatic response 65, 73, 74, 78 translational conflict 54
12, 15, 18, 108 sources of disturbance 107 translator training 4, 14, 18, 105,
reverse-engineered processes speech disfluencies 12, 110, 118 106, 107, 108, 111, 112
10, 12, 15 standalone version 129, 130 Translog 4, 13, 88
revision 4, 9, 16, 17, 86, 88, 89, strategy execution 107
90, 91, 92, 93, 94, 95, 96, 97, strategy observation 110 U
98, 99, 100, 102, 105, 106, 110, stylistic errors 115, 118 unjustified cognitive activity 70
111, 112, 114, 117, 121 stylistic problems 118
revision behavior 108, 111, 120 subliminal skills 78 V
revision process 16, 91, 92, 93, surtitles 132 verbal reports 79, 110
95, 100 syntactic error 115, 117 verbalization 13, 63, 64, 73, 74,
revision tool 105, 106, 111, 112, syntactic representations 24 75, 76, 105, 106, 108, 110, 111,
116, 117, 120, 121 112, 118, 119, 120
Rich Points 114 T verbatim representation 24,
routine transfer links 43 T-move 89, 92, 93, 94, 95, 96, 27, 29
97, 98, 100, 101 vertical method 26
S TAP see think-aloud protocol vertical model 22, 36
S-move 89, 92, 93, 94, 95, 96, Theory of Translational Action vertical processes 37, 38
97, 98, 99, 100, 101, 102 16 vertical translation 22, 23, 24
screen recording 4, 105, 106, think-aloud protocol 8, 13, 23, virtual processes 10, 11, 15
111, 112, 113, 114, 116, 117, 118, 36, 63, 71, 72, 74, 79, 82, 83, visual attention 112, 120
119, 120, 121 110, 113 visual interference 12
segmentation 57 think-aloud recording 114 visual processing system 117
segmenting tendencies 112 through-description 129, 136
self-analysis 105, 106, 112, 114 ToTo see touch tour W
self-awareness 111 touch tour 125, 126, 127, 128, working memory 37, 67
self-correction 89 129, 130, 131, 132, 133, 134, 135,
self-projection 76 137, 138, 139, 140, 141