Across Languages and Cultures 8 (1), pp.

1–16 (2007)
DOI: 10.1556/Acr.8.2007.1.1

ON THE IDEA OF A THEORY

ANDREW CHESTERMAN
MonAKO, Department of General Linguistics,
University of Helsinki
Unioninkatu 40, Helsinki 00014, Finland
E-mail: [email protected]

Abstract: This article is based on a lecture that has been given to several groups of
doctoral students at various times and in various places. It outlines five notions of what has
been taken to constitute a “theory”: myth, metaphor, model, hypothesis and structured
research programme. The most fundamental of these is the hypothesis. These different ideas
of what a theory can be are illustrated with examples from Translation Studies. Any theory
aims at description and explanation, and these two concepts are also discussed. A final
comment takes up the idea that translations themselves are theories, and that a translator is
thus a theorist or theôros.

Key words: theory, model, hypothesis, myth, metaphor, research

1. A WAY OF SEEING

The etymology of the word “theory” goes back to the Greek θεωρíα ‘theoria’
meaning ‘a way of looking at something’, in order to contemplate it and under-
stand it better. In this broad sense, we can say that a theory is a helpful point of
view. I take (better) understanding to be the general goal of any theory. A the-
ory of translation is thus a view of translation – or some part or aspect of it –
which helps us to understand it better.
This is an instrumental notion of “theory”: a theory is an instrument of un-
derstanding. Good theories are useful instruments; bad ones are eventually dis-
carded in favour of better ones. But theories are also goals or ends in their own
right, in the sense that they are conceptual structures that need to be designed,
formulated and tested. In other words, theories themselves are also forms of un-
derstanding. If understanding is the final goal, constructing a theory is an inter-

1585-1923/$ 20.00 © 2007 Akadémiai Kiadó, Budapest

2 ANDREW CHESTERMAN

mediate goal which can then serve as an instrument to that end, at the same time
as it gives this understanding a form. In this context, it is useful to compare the
notion of a method, from Greek meta hodos ‘after the way’. Methodology is,
etymologically, the business of proceeding along the way (hodos) to do some-
thing or to reach a goal. If a theory represents a form of understanding, methods
are the ways in which one actually uses, develops, applies and tests a theory in
order to reach the understanding it offers.

2. DESCRIPTION, EXPLANATION

The general goal of understanding a phenomenon means that we need access to

appropriate concepts, an ability to describe and explain, and often also the abil-
ity to predict. Appropriate concepts need to be defined, and justified in prefer-
ence to other “shadow” concepts that might have been used instead but were
not: i.e. the choice of concepts needs to be justified. (Compare the notion of
“shadow translations”, used e.g. by Johansson 2004.) The concepts also need to
be related to each other, as illustrated e.g. in the concept diagrams used in ter-
minology work. There are a great many concepts in Translation Studies (TS)
which have been given competing definitions: we do not have a consistent ter-
minology, in any language. Consider the variety of available definitions of the
key concepts of “translation”, “equivalence” and “strategy”, for instance; or the
different typologies of these concepts that have been proposed (on strategies,
see e.g. Chesterman 2005).
A description can be done from many perspectives. The linguist Kenneth
Pike (1959) suggested trying at least three such perspectives: seeing the phe-
nomenon as a particle (in isolation; and in comparison with similar particles); as
a wave (in which the phenomenon is seen as merging with other phenomena in
space and time; this perspective also includes the historical, diachronic view);
and as a field (where the phenomenon is related to its surrounding context). We
could also look at it from a functional perspective: what use is it, what is its
value? (Cf. Booth et al. 1995:40.)
Explanations also come in many forms. We can explain for instance how
something is possible (what are the necessary conditions?), or why it had to
happen (what are the sufficient conditions?). Explanations work by showing re-
lations of generalization, causality, or unification with other patterns of
phenemona. Hypotheses about translation universals, for instance, aim to ex-
plain the occurrence of, say, explicitation in a given translation by positing a
generalization: all (or most) translations manifest some explicitation, and this
makes it easier to understand why this particular translation does as well, if
there is indeed a general tendency. If the generalization is true, we are not sur-

Across Languages and Cultures 8 (1) (2007)

 ON THE IDEA OF A THEORY 3

prised when we find explicitation, because the generalization offers an explana-

tion of a kind. Causal explanations also increase our understanding of why
something happened, or why the form of something is as it is. And if we can
show the relations between the explanandum (that which needs to be explained)
and other phenemona, if we see it in the context of a more general pattern, we
also feel we have found some kind of explanation. For instance, we can better
understand why there have been several different German translations of the
classic Finnish novel Seitsemän veljestä (‘Seven brothers’, by Alexis Kivi) if
we look at the context of the different social and political conditions at different
periods in Germany, different translators with different intentions and ideolo-
gies, different publishers, different translation norms, different German images
of Finland at different times, and so on (Kujamäki 1998, 2001). These condi-
tioning factors are all weakly causal, but, as is often the case in the human sci-
ences, our sense of understanding here comes more from seeing the broad pic-
ture than from evidence of causality in a strict sense. To the extent that we can
relate these factors within a single model, we can arrive at an explanation via
unification. (On explanation via unification, see Salmon 1998.)
Predictions can be used to test explanations. But not all explanations imply
the corresponding ability to predict. However, even if precise predictions are
not possible, an explanation reduces our surprise that the phenomenon in ques-
tion has occurred, and perhaps allows us at least to anticipate other occurrences
in future. An explanation thus allows us to “make sense” of something. (For fur-
ther discussion of this general theme, see e.g. Chalmers 1999.)

3. KINDS OF THEORY

It is helpful, I suggest, to distinguish five types or notions of theory. I outline

them here. Examples from Translation Studies follow below.
Some of the first theories were myths: they were symbolic narratives which
offered some way of “making sense” of something mysterious. “We are mean-
ing-seeking creatures,” writes Karen Armstrong (2005:2); “[...] from the very
beginning we invented stories that enabled us to place our lives in a larger set-
ting, that revealed an underlying pattern [...]”. Think of the many myths of crea-
tion, in different cultures, for instance. Or the 20th-century Gaia myth which in-
vites us to see our planet as a living organism. A myth can have enormous ef-
fects on a person or a society.
Theories can also take the form of metaphors or similes. To see something
strange as something more familiar (or as being like something more familiar,
or as a kind of something more familiar) is one way of making sense of it.
Physicists have learned to see light, for instance, as particles; or as waves; or,

Across Languages and Cultures 8 (1) (2007)

4 ANDREW CHESTERMAN

nowadays, as both at the same time. “Metaphor is one of our most important
tools for trying to comprehend partially what cannot be comprehended totally,”
say Lakoff and Johnson (1980:193). But a metaphor only sheds light on one
side of a phenomenon: every metaphor also hides what it does not highlight. We
do not see the dark side of the moon.
Myths and metaphors are often said to represent the kind of knowledge
which Plato called mythos, as opposed to logos. Mythos is the form of knowl-
edge that is symbolic, intuitive, figurative, imaginative; logos is rational, ex-
plicit.
Scientific theories often involve metaphors and myths, but they typically
aim to be more explicit, based on evidence, and empirically testable. One kind
of scientific theory is a model, which seeks a relation of similarity with what-
ever it is a model of. Models represent what are taken to be the main elements
of a phenomenon, their main functions, and the main relations between the ele-
ments. Think of models of our planetary system, before and after Copernicus.
Or models of the brain (like the model of the brain as a black box, with input
and output); or models of cognitive processes (decision-making, for instance, is
sometimes represented as an algorithm of yes/no choices, a view which is im-
plicitly based on the metaphor of the mind as a computer).
For some philosophers, theories are seen as hypotheses. A hypothesis is,
roughly speaking, an educated guess at the best answer to a question, based on
the most reliable facts available. Popper’s general model (sic!) of scientific pro-
gress centres around the idea of generating and testing hypotheses (e.g. Popper
1959). To solve an initial problem, a tentative theory (a hypothesis) is proposed,
then this is tested (the stage of error elimination), and the result is usually a new
problem; and so the process continues. The original invention of the idea of a
hypothesis has been a major step in the development of the scientific method,
because it means that empirical claims need to be tested against evidence rather
than believed on the basis of the authority of the claimer, or on tradition, or on
intuition alone.
There are various basic types of hypothesis (illustrations coming below).

Descriptive hypotheses: all Xs have feature F / belong to class Y.

Explanatory hypotheses: X is caused by / made possible by, Y; Y explains X.
Predictive hypotheses: in conditions ABC, X will (tend to) occur.
Interpretive hypotheses: X can be (usefully) interpreted as Y.

The first three types are self-explanatory; they are all empirical hypotheses.
Predictive hypotheses are used to test (some) explanatory ones.
Interpretive hypotheses have a different status from the other types, be-
cause they are not tested directly against empirical evidence (are they true?) but
against pragmatic criteria (are they conceptually useful, insightful?). My claim

Across Languages and Cultures 8 (1) (2007)

 ON THE IDEA OF A THEORY 5

that there are five types of theory, for instance, is this kind of hypothesis. I find
it useful, as a way of clarifying the picture. But if someone proposes a compet-
ing hypothesis – say, that it makes more sense to classify types of theory into
seven, or seventeen groups – we simply have to see which classification seems
more useful, which “catches on”, which gives rise to better research. Interpre-
tive hypotheses are nevertheless an integral part of doing science. All defini-
tions and classifications, for instance, are based on them, not to speak of data in-
terpretion. If birds and dinosaurs are classified as belonging to the same zoo-
logical class, we shall probably “see” them rather differently than we would
otherwise. Do you define a tomato as being a fruit or a vegetable? Is Pluto a
planet nowadays or not? Alternative classifications illustrate the enormous vari-
ety of ways in which we human beings can conceptualize our world (for a clas-
sic study, see Lakoff 1987).
My fifth type of theory is the structured research programme. This view is
argued for especially by Lakatos (e.g. 1970). Such a theory has two main com-
ponents. At the centre, there is a “hard core” of fundamental principles and as-
sumptions which are not questioned within the programme but taken for granted
(although they may be questioned by people outside the programme). Outside
this core, there is a “protective belt” of supplementary assumptions and hy-
potheses to be tested, protecting the hard core.
Of these five types of theory, the hypothesis is the fundamental one. Myths
and metaphors can be seen as interpretive hypotheses, to be tested in use. Mod-
els are hypotheses too, to be tested against evidence; and the same can be said
of the cluster of hypotheses and assumptions which make up a research pro-
gramme: you try them out, and if they do not work adequately, you eventually
try something else.
Let us now look at how these kinds of theories are represented in Transla-
tion Studies.

4. MYTHS OF TRANSLATION

In the West, the dominant myth of translation is the Babel myth (Genesis
11). This myth has had major consequences for our perception of transla-
tion. We see it as tainted by the Fall from the Paradise of perfect commu-
nication, by failure; it is always second-best, never as good as the origi-
nal; and so on.
But we do not have to see translation through the eyes of this pessi-
mistic myth. In the East, apparently, translation can be seen in terms of
the myth of metempsychosis, or the transmigration of souls (Devy 1999).
It can thus be associated with the idea of rebirth, spiritual progress: as a
soul returns to live again in a new body, so a text may be born again in a

Across Languages and Cultures 8 (1) (2007)

6 ANDREW CHESTERMAN

new language, and perhaps in a form that is in some way better than its
previous existence. This is rather a more optimistic view, looking not at
what is lost in translation but what might be gained. Traces of this alter-
native myth can also be found in Renaissance writings on translation in
the west, as when translators are described as having some spiritual affin-
ity with their authors, as if the two had a shared soul (Hermans
1985b:127).

5. METAPHORS OF TRANSLATION

The history of Translation Studies is full of metaphors and images of transla-

tion. Here is a selection from the Renaissance period in Europe, all having to do
specifically with literary translation (paraphrased from Hermans 1985b, where
exact sources are given).

Translation is (like) imitation.

The translator is like a painter who shows a person’s body but not his soul.
The translator digests the original, turning it into blood and food.
To translate is to follow in the footsteps of the original.
A translation is the rival of the original, striving to beat it at its own game.
The translator is the servant or slave of the original.
A translator is a magpie among peacocks (i.e. trying to look like a peacock
but failing).
The translator is like Phaeton, who wanted to drive the chariot of the sun
and was punished for his presumption (i.e. if he dares to challenge the mas-
tery of the original).
A translation is the echo of a song.
A translation offers false pearls in place of diamonds.
A translation is like the reverse side of a tapestry.
A translation is a plain garment (whereas the original is a rich cloth).
A translation is like candlelight compared with the sun.
Translations are shadows of beautiful bodies.
Translation opens windows to let in the light.
The translator brings treasures which were hidden in the earth.
A translation is a jewel in the rough casket of language.
Translating is like pouring a precious liquid from one vessel into another.

Such conceptualizations reveal implicit theories about translatability, the

relation between meaning and form, translation equivalence and translation
quality.

Across Languages and Cultures 8 (1) (2007)

 ON THE IDEA OF A THEORY 7

My own preferred metaphor is from memetics (Chesterman 1997). I find it

interesting to see translations as ways of carrying the evolution of ideas
(memes) from one culture to another, like biological organisms carry the evolu-
tion of genes. Translations help memes to survive and spread. This metaphor
thus allows me to see translation within the framework of cultural evolution.
Translation has also been seen as a kind of something else which is more
familiar: as a kind of rewriting, for instance (Lefevere 1992). This placing of
translation under the more general term highlights its similarities with other
kinds of rewriting, such as paraphrasing, summarizing, or anthologizing (and
correspondingly downplays other kinds of similarities). Sometimes there are se-
rious disagreements about these hierarchies. There is currently some debate
about whether translation is a kind of localization, or whether localization is a
kind of translation (see e.g. Pym 2004). To some extent such interpretive hy-
potheses are expressions of power and status: if translation is the higher term, it
seems to be more important, and localization is demoted to being a mere type of
translation. And vice versa. So the debate is partly an institutional, territorial
one, with consequences concerning research funds, jobs, salary levels and so on.

6. MODELS OF TRANSLATION

I have elsewhere (Chesterman 2000) distinguished three types of models used in

Translation Studies.
Comparative models show translations in a relation to other texts, which
may be

• source texts (> the equivalence relation),

• non-translated comparable texts in the target language (> the naturalness
relation),
• other translations (> research on translation universals), or
• non-native texts, learners’ texts.

Examples of comparative models are those developed by for instance Cat-

ford, Koller, Vinay and Darbelnet; those based on contrastive research; and cor-
pus-based work on translation universals. Catford’s (1965) model is based on an
analysis of the various kinds of relations holding between source texts and
translations, on his theory of equivalence and his classification of types of shift.
An equivalence typology also lies at the centre of Koller’s (1979) model; he dis-
tinguishes denotative, connotative, text-type, pragmatic and formal types of
equivalence. Vinay and Darbelnet (1958) were primarily interested in the con-
trastive syntactic and rhetorical differences between English and French, and
proposed a number of “procedures” as ways of conceptualizing the kinds of

Across Languages and Cultures 8 (1) (2007)

8 ANDREW CHESTERMAN

changes that translators make when moving from one language to another. For
instance, the procedure of “transposition”, which involves changing the word
class. Although Vinay and Darbelnet framed their procedures in dynamic terms,
these are usually seen as ways of conceptualizing the results of these various
procedures. What we see, empirically, is the result of a change of word class,
for instance, when a verb in English is translated as a noun in French.
With the advent of corpus methodology, scholars began looking empiri-
cally at other kinds of relations. Translations into, say, English can be compared
to matched (i.e. with same subject matter and text type, etc.) non-translated
texts originally written in English, in order to see whether there are systematic
differences between the two corpora. In other words, we can measure the natu-
ralness of the translations. We can also compare translations into English with
translations into other languages, from a variety of source languages, and see
whether we find evidence of recurrent features (universals?) of the translations,
regardless of language or text-type, which might therefore be due to the transla-
tion process itself. (Examples will be given below of some claims that have
been made as a result of such research.) We might also want to compare transla-
tions to texts produced by non-native speakers: perhaps the effect of the extra
constraint of the translation situation is similar to that of having to produce a
text in a non-native language.
In all versions of these comparative models, then, we are dealing with two
sets of texts and looking at the relations between them. The relations are basi-
cally static and contrastive.
Process models, on the other hand, represent the temporal sequence of
stages in the translation process, either at the cognitive level of translation deci-
sions, or the sociological level of working procedures. An early and influential
process model was the one proposed by Nida (e.g. 1969). He suggested that
translation was like crossing a river (note the image!), but added that a transla-
tor does not actually cross at the point where the source text is. The translator
first analyses the source text (reduces it to a simplified form, i.e. looks for an
easier place to cross the river), then makes the transfer to the target language,
and then “restructures” the translation, which means stylistically polishing it so
that it corresponds to the stylistic profile of the original and to the expectations
of the audience. Nida thus modelled the translation process in terms of the three
stages of analysis, transfer and restructuring.
Other scholars have modelled the process differently. Sager (1993) bases
his analysis on four working procedures: specification (understanding the cli-
ent’s instructions, checking that the brief is appropriate and feasible); prepara-
tion (finding the necessary resources, terminology, and so on); translation; and
evaluation (revision). Nord (1991:32f) argues for what she calls a looping
model. Here, step one is analysis of the translation skopos; step two is analysis
of the source text; and step three is the production of the target text. Nord calls

Across Languages and Cultures 8 (1) (2007)

 ON THE IDEA OF A THEORY 9

this model a looping one, because it illustrates the way in which the translator
continually loops back and forth between the three elements of the model:
skopos, source text analysis and target text synthesis. This model is thus more
complex than a linear one going directly from source to target.
Process models are also evident in cognitive research using think-aloud
protocols (TAPs), where the aim is to model the translator’s decision-making
process (for surveys, see e.g. the special issues of Across Languages and Cul-
tures 3/1, 2002; and Meta 50/2, 2005). Work using the TRANSLOG program
also tracks the translation process: here, the computer records the translator’s
every key-stroke against a clock, producing a time record of the translator’s typ-
ing actions. When this is combined with a think-aloud protocol and perhaps a
video recording, and a retrospective interview, we can arrive at a fairly detailed
picture of the observable translation process and make some inferences about
what happens at the cognitive level. (For some recent TRANSLOG work, see
Hansen 2002 and other papers in the same volume; and Hansen 2005.)
On a much larger time-scale, a process model also underlies work on trans-
lation history. Here, though (as also with some of the cognitive research) there
is an overlap with causal models, in which the focus is not only on the temporal
relation of what came before or after what, but on the cause-and-effect relation.
Causal models, in their simplest form, look like this:

Causal conditions > Translations > Effects

Translations themselves are both the result of preceding causal conditions and
themselves the cause of consequences. These consequences may range from ac-
ceptance or rejection of the text by the client, or feedback from critics, to major
cultural developments. Think of the role played by translation in the history of
science, for instance. In the human sciences, however, causality can seldom if
ever be represented as a simple linear chain. What we usually find is a whole
complex of factors and contributory conditions, some of which are more power-
ful than others, and many of which also affect each other. In a very broad sense,
the causal conditions of translations range from the cognitive to the socio-
cultural. The immediate cause of a given word in a translation, for instance,
must be that the translator decided to write that word: the proximate cause is
cognitive. But we can continue asking “why?”. Perhaps he made that decision
because of his working conditions (a bad dictionary? not enough time?). Why
were the conditions like that? (The client didn’t care? The client chose a poorly
qualified, cheap translator?) What are the norms for that kind of translation task
in that culture? What are the underlying values and ideologies? And so on – the
questions never end, but answers to them can contribute something to our wish
to understand “why” something occurred in this translation. Similar lists of
questions can be made for other starting-points. Why, for instance, did a given

Across Languages and Cultures 8 (1) (2007)

10 ANDREW CHESTERMAN

translation have such-and-such an effect on the client / on the readers / on soci-

ety / on cultural development as a whole?
Causal models are used explicitly in several current theories. Skopos the-
ory, for instance, assumes as axiomatic that the single most important cause of
the form of a given translation is (or should be) its skopos, its purpose. (See e.g.
Vermeer 1996, Nord 1997.)
Scholars who have applied relevance theory to translation, on the other
hand, propose that the fundamental explanatory cause is the principle of rele-
vance itself. Roughly speaking, this is defined as the cost-benefit relation be-
tween the reader’s effort to understand (= cost) and the cognitive effects the
message has (= benefit): the more benefit and less cost, the more relevance; and
speakers are assumed to aim for optimal relevance. Relevance theory claims to
offer a cognitive explanation for why people communicate as they do (and also
why they translate as they do, since translation is seen here as being simply one
form of the interpretive use of language, i.e. like reported speech). (See Gutt
2000.)
Norm theory takes a different view of causation, looking more broadly at
the socio-cultural conditions and norms, including for instance the translation
tradition in a given culture, which affect translator’s decisions. (A classic source
is Toury 1995.) In translation training, translators are typically socialized into
the current norms in their society; but not all translators receive formal training,
and need to pick up the norms in other ways. Norms are based on values, which
in turn lead us to translation ethics: a translator’s professional ethics also affect
his decisions about how to act as a translator. Reception studies (also repre-
sented in Kujamäki 1998, cited above) also use a causal model to analyse why a
given translation had a given effect.
From this point of view, it is evident that all work on translation quality has
an underlying causal model, too. This is apparent in translation quality control,
where checking and documentation of procedures aim at assuring the quality of
the various stages of the translation process, on the assumption that if the pro-
cess is OK the resulting product will also be OK. (See e.g. the new CEN stan-
dard for translation services, EN 15038:2006, e.g. via <www.cen.eu>.) It is also
apparent in quality assessment, whether during training or in working life. A
given translation fails, or is rejected by the client, because it has given defects;
in other words, it causes a negative effect on someone. One could even say that
research on machine translation uses an implicit causal model: if the aim is to
produce a better automatic translation program, the focus is on adjusting the
program to produce an optimal-quality result – you vary the causal conditions
and check the effects.

Across Languages and Cultures 8 (1) (2007)

 ON THE IDEA OF A THEORY 11

7. HYPOTHESES OF TRANSLATION

Interpretive hypotheses abound, of course. All operationalization decisions are

based on them. An example might be a decision to interpret the concept of
translation quality, in a given research project, as “grades from 1 to 5 given by
translation trainers using a matrix of defined error types and weightings”; as
opposed for instance to “impressionistic grading by peers” or “acceptance or re-
jection by the client”.
Descriptive hypotheses have come much more to the fore with the interest
in corpus studies on translation universals, mentioned earlier. All claims about
universals are descriptive hypotheses. A well-known example is the explicita-
tion hypothesis, that translators tend to make explicit what was only implicit in
the source (originating from Blum-Kulka 1986). Another is the law of increas-
ing standardization, that translations tend to use a more unmarked style than
their source texts (Toury 1995). A third example is the claim that translators
tend to reduce repetition. (For a recent selection of work on these hypotheses,
see Mauranen and Kujamäki 2004.) Yet another well-tested example is the re-
translation hypothesis, that later translations of a given text into a given target
language tend to be closer to the source than its first or earlier translations. This
hypothesis does not seem to be finding much confirmation (see most recently
Brownlie 2006).
Explanatory hypotheses are also common, but have so far tended to refer to
individual case studies. A given translation, it is argued, has particular features
because of a number of given factors. Brownlie (2003), for instance, proposes
the following factors which seem to explain much of her material: the individ-
ual situation and context of production, including the translator’s attitudes; the
conditions determining the textuality of a translation, the way it is subject to dif-
ferent textual and intertextual constraints; translators’ norms; and the effects of
intersecting fields such as publishing and academia (in the case of academic
texts). An example of a more general explanatory hypothesis is the claim that
translators tend to start their mental processing with a literal translation: the lit-
eral translation hypothesis. This would explain the progression from more literal
towards freer versions evidenced in studies of the revision process, where the
analysis focuses on sequences of interim draft solutions (e.g. Englund Dimi-
trova 2005). Another explanatory hypothesis has emerged from TAP research:
that the quality of a translation can (sometimes) be (partly) explained by the
translator’s mood and emotional state (see e.g. Jääskeläinen 1999).
Predictive hypotheses are less common in Translation Studies, where ex-
perimental research is not widespread. (It is more usual in Interpreting Studies:
see e.g. Pöchhacker 2004.) One example is the claim (or assumption?) that if a
translator has more time the translation will be of a better quality. This is an
idea that has been challenged at least by Hansen (2002). Implicitly, of course,

Across Languages and Cultures 8 (1) (2007)

12 ANDREW CHESTERMAN

every translator proposes a predictive hypothesis when he or she delivers a

translation: the prediction is that the client will accept it as a good (or at least
adequate) translation; and the test is the reaction of the client.
Formulating a hypothesis explicitly is crucial. There is a big difference be-
tween for example “translation shapes the formation of cultural identity” and
“translation reflects the formation of cultural identity”. The first is an explana-
tory, cause-and-effect hypothesis, which suggests that translation is a (presuma-
bly contributory) condition of cultural identity formation. It is presumably not a
necessary or sufficient cause, however. And precisely when translation would
be a contributory condition is not made explicit in this formulation. The second
formulation is an interpretive hypothesis: translation can be interpreted as re-
flecting cultural identity formation. Anyone wishing to disagree would need to
counter-argue that, in their view, translations are more usefully regarded as
signs of something else, or that some other signs of cultural identity formation
are more important.
Good empirical hypotheses are testable against evidence; they are inter-
nally coherent, simpler or more powerful than competing hypotheses, bring
added value, and raise new questions and hypotheses. It is a good idea either to
start a research project with a hypothesis (better: a pair of contrasting hypothe-
ses), or else end with a claim that can function as a hypothesis to be tested in a
future research project.

8. STRUCTURED RESEARCH PROGRAMMES (?)

We can find some examples in TS which seem to be structured research pro-

grammes in Lakatos’ sense, but perhaps only to some extent. Polysystem the-
ory, for instance, has a set of “hard-core” assumptions, including the following
(from Hermans 1985a). We can analyse cultures in terms of polysystems, i.e.
systems of systems. A system is a structured complex of elements that can be
distinguished from its environment. Systems can be identified, but they are not
static; each system has a centre (e.g. of canonized forms) and a periphery, and it
has conservative and innovatory elements, which are in a constant state of flux.
Within a polysystem, different systems and subsystems struggle for central
status. One such system is that of literature, which itself contains other systems,
such as genres. Translations show what can happen at the interface of systems;
they can be part of the centre or of the periphery. Perhaps the most innovative
and provocative assumption in polysystem theory is that all translation involves
some manipulation, for a given purpose, such as the intended function of a
translation in the target system. Research based on these assumptions has given
rise to a large number of descriptive case studies, but general empirical hy-

Across Languages and Cultures 8 (1) (2007)

 ON THE IDEA OF A THEORY 13

potheses have not been much in evidence, with the exception of general claims
about the ubiquity of interference.
Corpus research on translation universals could also be said to share some
basic assumptions, such as the idea that features specific to translations are
worth seeking in their own right, not as evidence of poor translations; that trans-
lations constitute a text-type of their own, a “third code”, neither source nor
wholly target; that we have a (potentially) universal concept of translation
available which can apply to all cultures and all ages; and that there is some re-
liable way of deciding what translations to include in a corpus and what to ex-
clude (for instance on the parameters of professional vs. amateur, native vs.
non-native, or even good vs. bad: after all, a bad translation is still a translation,
of a kind...). (For some criticism of these assumptions, see Tymoczko 1998.)
The empirical hypotheses are then the various claims about possible universal
features, referred to above.
One lesson we can draw from this view of a theory is the importance of
distinguishing between what is assumed, what is being asked or tested, and
what is finally being claimed. Assumptions can be challenged, and hidden as-
sumptions can be exposed. Hypothesis-testing procedures and operationaliza-
tion decisions can also be challenged. And claims may be met with counter-
evidence and counter-claims, or suspected of being illogical. The difference be-
tween assumptions, hypotheses and claims is thus a question of their status in a
given research project, not of their propositional content as such.
But there is certainly not one shared paradigm in TS, which would allow
the development of a single research programme covering the whole field. (See
the Target debate, starting with Chesterman and Arrojo 2000.) We share very
general aims – to understand translation and the significance of the work of
translators – but we do not all work with the same sets of concepts, nor even the
same types of data. The lack of a single shared paradigm is also reflected in our
heterogeneous methodologies. (See Gile’s provocative division between the
Empirical Science Paradigm and the Liberal Arts Paradigm, on the EST website
at http://www.est-translationstudies.org/ > Research issues.)

9. CONTRIBUTING TO THEORY

The five types of theory I have outlined are not mutually exclusive. By this I
mean that a given research project may make use of different types. It might use
a causal model and propose or test an explanatory hypothesis, within a broader
structural framework of other assumptions and hypotheses, for instance. Or it
might focus on a descriptive hypothesis framed in a comparative model and
formulated in terms of a striking metaphor. One way of contributing to theory is

Across Languages and Cultures 8 (1) (2007)

14 ANDREW CHESTERMAN

indeed to show the relations between different existing theories and theory-
types, and thus try to develop a still more general theory.
One can also contribute to theory in other ways. By providing new data (in
order to test existing theories), for instance. Or by offering a better interpreta-
tion of existing data; by clarifying and relating concepts and assumptions. Or by
proposing a better model or a new hypothesis. Or by testing or refining a given
hypothesis, or proposing a new way of testing one, a new methodology. In other
words: one can contribute by suggesting a new or better answer to an interesting
question, by pointing out a new path which might lead to a better answer, or by
raising a new interesting question or problem.
Ultimately, the aim (as I see it) is to link various substantiated claims in
terms of more general laws, in the same field or in other fields, in steps towards
the possible unification of knowledge: consilience.

10. A TRANSLATION IS ITSELF A THEORY

This final comment is really only a footnote. I suggest that a translation itself
can be seen as a theory of how the source text can be translated, in all the five
senses outlined above.
It usually represents the source text as a linear narrative, like a myth.
It is usually taken “to be” metaphorically the source text (etymologically
trans-late = meta-phor). When I say ‘I have read Tolstoy’ (albeit in English
translation), I am not lying, for my War and Peace “is” the Tolstoy novel.
A translation usually models the source text isomorphically, both holisti-
cally and in terms of its parts.
It constitutes a hypothesis about the source text, about how it can be inter-
preted, usually about its meaning. A translator implicitly makes a claim that his
translation is equivalent in some relevant way to the source text, a claim that is
tested by the reception of the translation by clients, readers and perhaps critics
(cf. Pym 1995).
And the whole process of translating is like a structured research pro-
gramme, with a clearly defined problem to be solved, a set of assumptions (such
as translation norms), and a multitude of individual hypotheses to be tested, for
each segment of the text.
So translators theorize all the time (see also Pym 1992). But they theorize
also in a deeper sense. As Carol Maier (2006) suggests, the translator is a
theôros, i.e. a person who makes a “journey” in order to seek knowledge, a bet-
ter way of seeing; perhaps even wisdom. Maier writes (168–9): “the traveller
witnessed things and events with which he was unfamiliar. To see, then, in this
context was to theorize. [...] Moreover, to theorize those unfamiliar things was

Across Languages and Cultures 8 (1) (2007)

 ON THE IDEA OF A THEORY 15

to be affected by them, to be moved to wonder.” This wondering contemplation

takes us to the very roots of philosophical enquiry.

References
Armstrong, K. 2005. A Short History of Myth. Edinburgh: Canongate.
Blum-Kulka, S. 1986. Shifts of Cohesion and Coherence in Translation. In: House, J. and Blum-
Kulka, S. (eds.) Interlingual and Intercultural Communication: Discourse and Cognition in
Translation and Second Language Acquisition Studies. Tübingen: Narr. 17–35.
Booth, W. C., Colomb, G. C. & Williams, J. M 2005. The Craft of Research. Chicago: University
of Chicago Press.
Brownlie, S. 2003. Investigating Explanations of Translational Phenomena. Target Vol. 15. No.
1. 111–152.
Brownlie, S. 2006. Narrative Theory and Retranslation Theory. Across Languages and Cultures
Vol. 7. No. 2. 145–170.
Catford, J. C. 1965. A Linguistic Theory of Translation. Oxford: Oxford University Press.
Chalmers, A. [1976] 1999. What is This Thing Called Science? (3rd edition) Indianapolis: Hack-
ett.
Chesterman, A. 1997. Memes of Translation. Amsterdam: Benjamins.
Chesterman, A. 2000. A Causal Model for Translation Studies. In: Olohan, M. (ed.) Intercultural
Faultlines. Manchester: St. Jerome Publishing. 15–27.
Chesterman, A. & Arrojo, R. 2000. Shared Ground in Translation Studies. Target Vol. 12. No. 1.
151–160.
Chesterman, A. 2005. Problems with Strategies. In: Károly, K. & Ágota Fóris, Á. (eds.) New
Trends in Translation Studies. In Honour of Kinga Klaudy. Budapest: Akadémiai Kiadó.
17–28.
Devy, G. 1999. Translation and Literary History. In: Bassnett, S. & Trivedi, H. (eds.) Post-
colonial Translation. London: Routledge. 182–188.
Englund Dimitrove, B. 2005. Expertise and Explicitation in the Translation Process. Amsterdam:
Benjamins.
Gutt, E.–A. 2000. Translation and Relevance. Cognition and Context. Manchester: St. Jerome
Publishing.
Hansen, G. 2002. Zeit und Qualität im Übersetzungsprozess. In: Hansen, G. (ed.) Empirical
Translation Studies: Process and Product. Copenhagen: Samfundslitteratur. 29–54.
Hansen, G. 2005. Störquellen in Übersetzungsprozessen. Copenhagen: CBS.
Hermans, T. 1985a. Introduction. Translation Studies and a New Paradigm. In: Hermans, T. (ed.)
The Manipulation of Literature. Studies in Literary Translation. London. Croom Helm.
7–15.
Hermans, T. 1985b. Images of Translation. Metaphor and Imagery in the Renaissance Discourse
on Translation. In: Hermans, T. (ed.) The Manipulation of Literature. Studies in Literary
Translation. London. Croom Helm. 103–136.
Jääskeläinen, R. 1999. Tapping the Process. An Explorative Study of the Cognitive and Affective
Factors Involved in Translating. Joensuu: Joensuun yliopisto.
Johansson, S. 2004. Why Change the Subject? On Changes in Subject Selection in Translation
from English into Norwegian. Target Vol. 16. No. 1. 29–52.
Koller, W. 1979. Einführung in die Übersetzungswissenschaft. Heidelberg: Quelle und Meyer.
Kujamäki, P. 1998. Deutsche Stimmen der Sieben Brüder. Frankfurt am Main: Peter Lang.
Kujamäki. P. 2001. Finnish Comet in German Skies: Translation, Retranslation and Norms. Tar-
get Vol. 13. No. 1. 45–70.

Across Languages and Cultures 8 (1) (2007)

16 ANDREW CHESTERMAN

Lakoff, G. 1987. Women, Fire, and Dangerous Things. What Categories Reveal about the Mind.
Chicago: University of Chicago University Press.
Lakatos, I. 1970. Falsification and the Methodology of Scientific Research Programmes. In:
Lakatos, I. & Musgrave, A. (eds.) Criticism and the Growth of Knowledge. Cambridge:
Cambridge University Press. 91–196.
Lakoff, G. & Johnson, M. 1980. Metaphors We Live By. Chicago: University of Chicago Press.
Lefevere, A. 1992. Translation, Rewriting and the Manipulation of Literary Fame. London:
Routledge.
Maier, C. 2006. The Translator as Theôros. In: Hermans, T. (ed.) Translating Others. Vol. 1.
Manchester: St. Jerome Publishing. 163–180.
Mauranen, A. & Kujamäki, P. (eds.) 2004. Translation Universals. Do They Exist? Amsterdam:
Benjamins.
Nida, E. A. 1969. Science of Translation. Language Vol. 45. No. 3. 483–498.
Nord, C. 1991. Text Analysis in Translation. Amsterdam: Rodopi.
Nord, C. 1997. Translating as a Purposeful Activity. Manchester: St. Jerome Publishing.
Pike, K. L. 1959. Language as Particle, Wave and Field. Texas Quarterly Vol. 2. No. 2. 37–54.
Pöchhacker, F. 2004. Introducing Interpreting Studies. London: Routledge.
Popper, K. R. 1959. The Logic of Scientific Discovery. London: Hutchinson.
Pym, A. 1992. Translation and Text Transfer. Frankfurt am Main: Peter Lang.
Pym, A. 1995. European Translation Studies, une science qui dérange, and Why Equivalence
Needn’t Be a Dirty Word. TTR Vol. 8. No. 1. 153–176.
Pym, A. 2004. The Moving Text. Amsterdam: Benjamins.
Sager, J. 1993. Language Engineering and Translation. Consequences of Automation. Amster-
dam: Benjamins.
Salmon, W. C. 1998. Causality and Explanation. New York: Oxford University Press.
Toury, G. 1995. Descriptive Translation Studies and Beyond. Amsterdam: Benjamins.
Tymoczko, M. 1998. Computerized Corpora and the Future of Translation Studies. Meta Vol. 43.
No. 4. 652–659.
Vermeer, H. J. 1996. A Skopos Theory of Translation. Heidelberg: TEXTconTEXT.
Vinay, J.–P. & Darbelnet, J. 1958. Stylistique comparée du français et de l’anglais. Paris: Didier.

Across Languages and Cultures 8 (1) (2007)