A Socially Neutral
Definition of Translating
Brian Mossop ©2003
[Note: Since this was
written over 10 years ago, I have modified my views on defining translation, as
will appear in future articles on the subject]
A
well known objection to any attempt to define translation is that,
historically, translation has been different things in different times and
places. That is, social norms have dictated that the label ‘translation’ be
applied to one set of texts at one time/place, but a different set of texts at
another time/place. As a result, definitions of translation tend to be covert
statements of what a society, or subgroup within a society, sees as the proper
way to translate. A classic example is Nida
&Taber’s definition: “translating consists in reproducing in the receptor
language the closest natural equivalent of the source-language message, first
in terms of meaning and secondly in terms of style” (1969:12). The expression
‘closest natural equivalent’ makes this statement prescriptive: only dynamic,
audience-oriented texts are translations.
Clearly,
if we wish to have a neutral definition of translation, one that does not
reflect any particular recommended practice, we cannot start from some set of
texts identified as translations. The set of texts to which people have applied
labels such as ‘translation’ or ‘version’ (or a member of a similar word set in
another language) is a multifarious hodge-podge, in which we cannot hope to
find a small set of distinguishing features.
In
this article, I propose a definition of translation sufficiently broad to be
compatible with varying social practices. The definition is stipulative;
that is, rather than starting from some observable activities which this or
that society labels ‘translating’, and attempting a definition which picks out
just those activities, I simply proclaim that any language production activity
which has certain characteristics will be deemed to be translational. Thus the
definition will be theoretical rather than empirical. The definition conjures
into being, so to speak, a theoretical entity, the value of which is discussed
at the end of the article.
This
approach makes it possible to achieve a universal definition (valid at all
times and places) and it also places translating within a larger scheme of
language production activities. Further, it avoids prescriptivism, for it says
nothing at all about how translators ought to work.
It
will be seen that I have already departed in one important way from traditional
approaches to defining translation: the definition I propose will pick out not
a set of completed texts but rather a set of language production activities. As
the argument proceeds, it will become apparent why a definition in terms of
process rather than product is valuable.
Translating
as understood here has three defining characteristics, all of which must be
present:
The
first thing to say about translating is that it is an instance of quoting. This
statement is not to be taken figuratively: translating is not like quoting; it is quoting, that is, translating is reported discourse. Thus
translating is in the first instance to be distinguished from language
production in which the speaker does not
intend to represent other language.
Most of the language we produce concerns states of affairs in the world
rather than what others have said. But the ‘metarepresentational’
capacity—the capacity to use language in order to represent language—is a vital
and universal feature of human languages, and translating is a manifestation of
it.
Quoting
is commonly thought of as reproducing the exact words of the source; in this
sense of the term, translating obviously cannot be quoting, since all or almost
all the words are changed. In fact, however, the reproduction of exact words is
a highly specialized form of quoting, required in certain written contexts (in
law, scholarship and journalism). Most quoting does not involve the
reproduction of exact words. Quoting is perhaps best understood as a kind of
dramatization 1: the speaker or writer puts the source on a
metaphorical stage and makes her speak in the first person:
I ran into Gwendolyn yesterday and she said I hear you’re not going to the translation
conference; I said No I don’t think
so; she said I really think you
should.
More
specifically, Clark & Gerrig (1990) argue that an
act of quoting is one in which the speaker purports
to demonstrate — rather than describe — selected aspects of something that has
been said or written. Consider:
(1) “I’ll get it translated
by tomorrow”, said Gwendolyn slowly.
(2) “I’ll...get... it... translated...by...tomorrow”, said
Gwendolyn.
(3) Gwendolyn promised to translate it by the next day.
The speaker of (1)
describes the fact that Gwendolyn spoke slowly whereas the speaker of (2)
demonstrates it (not very successfully in the above written version; in the
oral equivalent, the quoter would speak slowly). The
speakers of both (1) and (2) demonstrate the illocutionary force (promising) of
what Gwendolyn said, whereas the speaker of (3) describes it.
Clark & Gerrig provide a partial
list of the things a quoter can decide to demonstrate2,
including voice pitch, speech defects, level of formality, propositional
content, illocutionary force, exact words uttered (just one option among
others) and, of course, the language used by the original speaker:
(4) “Je le ferai traduire
pour demain”, said Gwendolyn.
(5) Speaking in French, Gwendolyn promised she’d have it
translated by Friday.
(6) “I’ll get it translated by tomorrow”, said Gwendolyn in
French.
The speaker of (4) demonstrates the fact that Gwendolyn spoke in
French. The speaker of (5) describes this fact, as well as the illocutionary
force and propositional content of what she said. Interlingual
indirect reports like (5) do not demonstrate anything. Even if we consider the
utterance She’ll have it translated by
Friday, this too is descriptive rather than demonstrative in relation to Je le ferai traduire pour demain. It is
thus non-quotational and—if quoting is a defining
feature of Translating—non-translational. (It is true that such sentences may
be used in interlingual communicating situations, by
people called translators, but, as will be clarified further on, not everything
a person called a translator says counts as translating.)
Turning to sentence (6), here the speaker describes the fact that
Gwendolyn spoke in French, but demonstrates the force and content of what she
said. The portion inside quotation marks is therefore translational. Clark
& Gerrig’s interpretation of sentences like (6)
is vastly superior to that offered by Ann Banfield,
according to whom people take them as paradoxes (the quotation is in English,
not French) and then resolve the paradox through an “implicit belief in...a
kind of universal language which can be represented by particular languages”
(1982: 248-9). I suggest, following Clark & Gerrig
(1990: 798), that people do not take such sentences as paradoxes at all, but
instead as perfectly ordinary instances of quoting.
Now most translations, unlike the one in (6), are not presented as
quotations. Nevertheless (6) is a model of what every translator is doing.
Given a French text containing:
(10) A-B: ...Je le ferai traduire pour demain...
I can demonstrate various aspects of it, for example its
illocutionary force and propositional content, some of its lexico-syntactic
structure, and its level of language, by producing:
(11) X-C: ...I’ll get it translated by tomorrow...
Conceiving Translating as quoting has the virtue of strengthening
one of the classic answers to the old question about the possibility of
translation. Aside from the common bio-ecology all human groups share (which
provides extralinguistic reference points for
two-language communication), translation is made possible in part by the
structural universals of human language. As a supplement to this latter point,
we can now say that translation is possible because quoting is possible: all
languages provide lexico-syntactic devices for
demonstrating, or dramatically representing, the discourse of others.
The
second defining feature of the act of translating is that the language producer
first reads or hears a chunk of text in one language; then, within a very short
time interval, produces some text in another language and then moves on to the
next chunk of text in the first language, and so on. The chunks can be of
various sizes (within the limitations set by memory), and they can be processed
with more or less attention to cotext and context.
The
time gap involved ranges from 3-4 seconds in simultaneous interpretation to perhaps
x seconds in consecutive interpretation without note-taking. In between are the
various kinds of written translation, where the time is that required to read a
chunk ranging from a phrase to a longish sentence, or a couple of relatively
short ones.
Now
what of cases where, after reading a chunk of text, the producer of a written
translation stops for a relatively lengthy period of reflection or research?
These periods are in fact irrelevant because after they end, the producer
re-reads the original chunk of ST and then immediately starts composing the
translation. Thus the translational act is the second reading followed by
composition in the TL. If the translator does not re-read ST but rather relies
entirely on memory after the period of reflection or research, then the
language production act does not count as translating under the definition
proposed here.
In
speaking of a small time gap, I refer to the gap between reception of a chunk
of ST and production of the ‘first draft’ of the translation. Of course,
corrections may be made in this first draft, even in oral translation. In
written translation, the corrections may take place hours, days, weeks or even
years after the first draft was composed. Such corrections, however, constitute
separate acts of language production. Here we see one reason why a definition
is best phrased in terms of production rather than in terms of the final
product.
Thus
translating is not just any “text-derived text production”. Rather it is a very
specific kind in which production occurs immediately after reception of a
relatively small chunk of ST.
Sequentiality means that the ST is processed as text (as a sequence of wordings), not
as a mere source of ideas. It is quite possible, when writing a text in one
language, to use a text in the same or a different language as a source of
inspiration. Those ideas may then be combined with ideas drawn from other
texts, and from the writer’s own stock of ideas. The result may, in some
passages, bear a certain resemblance to one of the texts that provided a source
of ideas; there may a resemblance in meaning and even a resemblance in the
sequential order of ideas. However this does not satisfy the sequentiality criterion for SIQ unless the passage in
question was actually created by reading a chunk of ST wording, then quoting it
imitatively in TL (see next section on the third defining criterion of SIQ),
then reading the next chunk of ST, and so on.
That
said, sequentiality does not imply completeness; that
is, one is still engaging in SIQ even if some chunks of the source text are
omitted from the translation.
The
third defining feature of translating is an intention to preserve the meaning
of the source text rather than change it. The key word here is ‘intend’. I am
translating if I intend the text in the second language to mean more or less
the same as the source text. Whether I
succeed or not is utterly irrelevant. To put it another way: what others may think about my translation
is utterly irrelevant. For various reasons (lapses in attention,
limitations of TL, defects in my knowledge of SL or TL, deadlines), my output
may be full of deviations both major and minor from various aspects of the
meaning of the source text. The process that gave rise to this output still
counts as translating, as long as the deviations are not deliberate. There is
an unfortunate tendency in much translation theory, arising from a misguided
desire to be ‘useful’, to confuse translation with good translation, for example,
to talk about the cognitive processes involved in translation as those
processes which lead to a good translation. In the definition set out here, no
distinction is made between good and bad translations.
Much quotation is non-imitative: the quoter deliberately changes the meaning of what the source
said or wrote, often putting words into the source’s mouth. The same is true of
two-language quoting, a process usually dubbed ‘adapting’. Now it may happen
that, even though I intend to make significant changes in the meaning, current
social norms still label the result a translation. Indeed, as in the era of the
belles infideles
in France, norms may actually require changes in meaning, for example making
sexual or political content in a novel conform to the sensibilities of TL
readers or censors. But that is irrelevant to our definition. If I am aiming to
change meaning in order to satisfy such a norm, then I am not translating under the definition to be used here. Note that
this exclusion is in no way prescriptive: the social label ‘translation’ can
still be attached to a non-imitative TL text; our definition, it will be
recalled, is not an attempt to capture all those texts to which this or that
society attaches the label ‘translation’.
The
concept of imitation is the way in which meaning-sameness is introduced into
our definition. It is a subjective, translator-centred
approach to meaning-sameness, in terms of intention. Recall, once again, that
the definition proposed here concerns the moment of translation production, not
that later moment in which some ‘critic’ (a translation teacher, a quality
controller or reviser) compares the translation to the source and must employ a
definition of meaning-sameness that is at least partially objective. Since imitation
(meaning preservation) is defined in terms of the translator’s intent at the
moment of production, not in terms of the final product, it can change from
moment to moment. For example, at one moment I may believe that a chunk of ST
means x, so that if I write a TL string which I believe expresses x, then I am
imitating ST. However a few seconds later, I may change my mind and decide that
the ST chunk actually means y. At that moment, x ceases to be
meaning-preserving, and in order for my action to count as translating I must
write a new TL string which I believe expresses y.
To
sum up our three-feature definition, translating is an instance of quoting
which is sequential in method and imitative in intent. An act of language
production counts as translating if and only if it meets all three of these
defining criteria. Let us call the language production act so defined ‘SIQ’:
sequential imitative quoting.
Readers
may find it odd that no mention is made in the definition of the fact that the
quoted and imitated material is in another language. That is because SIQ is
practically always interlingual. I have been able to
think of only two intralingual cases that may
qualify, and they are closely related to each other. The first is exact-words
quoting; the second is
plagiarizing: a text is cited without attribution, and some of the wording is
changed in hopes of avoiding detection. Neither of these constitute clearcut cases of intralingual
SIQ: they are not necessarily sequential (an exact-words quoter
or a plagiarist need not operate chunk-by-chunk) and it is somewhat doubtful
whether exact-words quoting can be called imitative.
Perhaps
a more obvious candidate for intralingual SIQ would
be a film made in rural Quebec dialect which is subtitled in Standard French
for viewing in France? However this is interlingual,
not intralingual. By ‘interlingual’
I mean that ST and translation are in different ‘linguas’.
‘Different lingua’ does not mean ‘different standard written language’; it
means a speech form that is sufficiently different in its lexicogrammar
and phonology from the speech of the receivers (in the above case, the film’s
viewers in France) that they need help, not because the subject matter of the
text is beyond them, but because they lack the lexicogrammatical
and phonological knowledge necessary to understand what is being said. The fact
the film’s characters are speaking in what society labels a ‘dialect of French’
rather than a different langauge is irrelevant: the
dialect/language distinction is a matter of cultural labeling; it does not
correspond to any objective lexico-syntactic
distinction.
Aside
from the possible exceptions mentioned (exact-words quoting and plagiarism),
the various forms of intralingual production fail to
have all three defining features of
SIQ. Thus style editing to make a text read more smoothly is sequential but it
is not quotational; its purpose is not to demonstrate
features of the original wording. Oral intralingual
quoting is sometimes imitative, but it is not sequential: people simply report
the gist of what someone said minutes, hours or days earlier; they can’t
remember the sequence of words or even the sequence of ideas in detail.
As
for paraphrasing a text (for a non-expert audience for example), this is
non-imitative. Paraphrasing is both formally and functionally very different from translating.
Formally, it is different in that there is no ‘literal paraphrasing’, that is,
paraphrases (and indeed intralingual quotations
generally) have no tendency to be formally similar to their sources; nor do
they show signs of ‘interference’ from the lexico-grammatical
structures of the source text in the way literal translations do. Functionally,
paraphrases have a clarifying function which translations do not have. One
could, in theory, produce an imitative paraphrase, but there would be no point
to it. People paraphrase because they believe the source text was not clear.
Translations, on the contrary, do not by their nature clarify. A translation of
an article in advanced particle physics will not be clear to non-physicist
readers. Translators may of course chose to clarify (a non-translational act)
at the same time as they translate.
It
is true that people may repeat what
they or others have said, using either the same or different words, and it is
true that repetitions do nto have the funciton of clarifying. However repetition is non-quotational (i.e. it is non-demonstrative); it is discussed
briefly in the next section.
Non-SIQ language
production
Definitions
by their nature exclude as well as include, and consideration of the exclusions
can be helpful in understanding the definition Thus SIQ can be grasped by its
difference from other kinds of language production. First, quotation may be
non-sequential or non-imitative. Second, there are of course many kinds of non-quotational language production, both unilingual and
two-language. In the following
sections we’ll look at (1) non-quotational
two-language production, (2) non-quotational
unilingual production, (3) non-imitative unilingual quoting, and (4)
non-imitative or non-sequential two-language quoting.
|
SIQ |
non-sequential
or non-imitative quoting |
|
||
|
quotational lg
production |
non-quotational lg production |
|||
|
two-language |
unilingual |
two-language |
||
|
|
4 |
3 |
2 |
1 |
People
wishing to communicate a message in more than one language can prepare the
various language versions in parallel. For example, the English and French
versions of some of Canada’s laws have been prepared in this manner. Someone
conveys the general content of the proposed law to English and French drafters,
who then go away and separately prepare the two versions. There is no source
text, and thus no quoting. Later, in a separate writing act, the two versions
are brought into close conformance with each other with respect to the detailed
legal content. In other kinds of co-writing (e.g. multilingual advertising
campaigns), there may be no such later stage, that is, there may be no concern
with the question of whether the various versions ‘mean the same’. Some
co-written texts are very similar in meaning, some widely divergent.
There
are of course a vast number of forms of non-quotational
unilingual language production. Here we’ll look at a few which are of interest
because they are in some sense close to SIQ. First, I’ll look at five forms of
production which start from a source text. Rather than being quoted, however,
the source is repeated, copied, recited, transcribed or transliterated.
Sometimes the result is formally identical to quoting (it may contain the same
words as a quotation), but its function is not demonstrative.
Repeating
People
repeat either the wording or the meaning of what they themselves or others have
said because they think receivers may not have heard the first time, or because
they want to emphasize a point. Thus repeating differs functionally from
quoting even if there is no formal difference (a repetition of what someone
said and a quotation of what that
person said may contain exactly the same words). Repetitions are non-quotational in that they do not aim to demonstrate features
of the wording of the source text.
Copying
This
is the act of creating an extra physical token of the forms that constitute an
existing text. This can be done on a photocopying machine, but more interesting
for our purposes is the use of a computer word processor’s Copy function.
Translators often Paste into their translations copies of wordings taken from
TL document banks (including banks of previous translations into TL). The
Pasting may result in a translation
(i.e. the same output wording could have obtained by SIQ), but the act of
Pasting is of course not itself translational because it is not quotational or imitative.
The
use of copying by translators is changing the translation process. During translation as it has
been defined here, the quotational process involves
composing an imitative demonstration out of the mental store of TL lexicogrammatical material. With copying however, things
are quite different. This is perhaps best seen in the use of translation memory
programs, where the copies of previous documents are made and pasted into the
draft translation automatically. The translator's mental process thus begins
not with a single text in SL but with two on-screen texts, one in SL and one in
TL (a sentence pasted in from the store of previous translations). The task is
then essentially one of revision: deciding whether the pasted item is a valid
translation of the ST passage, accepting it if it is, or making corrections if
it is not. This is not a quotational procedure.
When
a newsreader or actor speaks the words of a script, their performance—whether
mechanical or creative—is not a demonstration to the audience of any features
of the script. The written script functions rather as a prop for the performance,
specifically a mnemonic device. Much reciting involves a modal switch from
writing to speech, though this is not a necessary feature: Speech-to-speech
reciting occurs when someone administers an oath, saying “repeat after me”.
The
modal switch in the opposite direction, from speech to writing, occurs for
example in the work of a court reporter, or when a student taking notes in a
lecture writes down some or (through shorthand) all of the professor’s spoken
words. Transcribing differs functionally from exact-words quoting in that its
sole aim is to record linguistic form; the student may not have understood the
meaning at all. In exact-words quoting, the preservation of the exact words is
not an end in itself but rather a specialized way of demonstrating various
aspects of meaning (in scholarly work for example). Note in passing that if our
student does not record the professor’s words but rather paraphrases the
professor’s meaning, then he is quoting the professor, demonstrating features
of what the professor said either to his later self (when studying for an exam)
or possibly to other students to whom he lends the notes.
In
this language production act, one form of graphic inscription is replaced by
another, following some set of one-to-one rules. Examples would be putting into
the Latin alphabet a text written in the Cyrillic alphabet, turning a written
Chinese text into the International Phonetic Alphabet, putting a text into
Braille, or encrypting a text. There may be several systems available (for
Romanizing Cyrillic say) but once you have selected a system, there is only one
possible transliteration of the text. A quoter might
draw on transliteration (e.g. use a phonetic spelling to convey the sound of a
character’s speech in a novel), but such a demonstrative purpose is not
inherent in transliterating.
The
functions of transliteration are quite varied. Thus transcribing a text into
IPA makes a text in a foreign language pronounceable by language learners, even
if they do not know the meaning. Romanization makes it possible to record a
written text in cases where suitable fonts (in Cyrillic say) are not available.
Putting a text into Braille makes the meaning of a text accessible (to the
visually impaired). Encrypting a text has the opposite effect—making its
meaning inaccessible to the unauthorized.
Machine
translation (MT) is best understood as an extremely sophisticated kind of
transliteration. Given a particular set of rules (and in the case of
‘example-based’ MT, a particular database of examples), the output wording is
completely predetermined. (In so-called ‘statistical’ MT, the form-changing
rules contain a probabilistic element, in that alternative forms are ranked,
but the outcome is still predetermined for a given sentence, and the ranking
rules do not involve the machine in any consideration of meaning.) The very
complicated set of rules for turning the input string of marks to the output
string is cleverly designed so that the output will (the programmers hope)
coincide reasonably well with
forms of human–produced writing in a natural language. As with Cutting
and Pasting, the result may sometimes
be useable as a translation, but the process
of creating the output is definitely not translational. The machine, being
nothing but an elaborate calculator, is not demonstrating anything at all about
the input text, or operating on meaning in any way. The role of meaning in the
activity of computers is zero, even if some portion of a set of MT rules are
dubbed a ‘semantic component’ by the programmers. It’s important not to confuse
what happened in the minds of the programmers with what is happening in the
machine as it creates MT output. For example, the programmers’ world knowledge
is sometimes encoded to some degree in dictionary entries, but once one sees
how this is done, it becomes clear that the computer can in no sense be said to
be picking out world knowledge that is relevant to the passage at hand as it
creates the ‘translation’.
An
interesting type of language production is the creation of subtitles for
television interviews with people who are either speaking their second language
with poor pronunciation or else speaking a dialect that differs from the
listeners’ dialect in phonology. In other words, the interviewees’ vocabulary
and grammar are perfectly adequate, but their pronunciation would prevent the
message from getting across without the use of subtitles. Now, is the
production of these subtitles an instance of transliterating, of transcribing, or
of translating? As in some kinds of transliterating, the subtitles give access
to pronunciation, but there is no pre-existing one-to-one system available. As
in transcribing, the subtitles take us from speech to writing and they do
record form, but is that their sole purpose? Once the pronunciation is
accessible, lexico-syntax becomes accessible to
anyone who knows the language in which the subtitles are written, and indeed
that is the purpose of creating them. Indeed, the restriction to phonology is
accidental: if the interviewee happened to make a syntactic or lexical error
while speaking his second language, or if he used some dialect-specific
lexicon, the subtitler would replace these with
proper forms from the listeners’ dialect. In other words, this is genuine
translation. As already explained, there is no reason to limit the term
translation to cases where the ST is in one official standard language and the
translation in another. The SL in translation can be a dialect or a second
language speaker’s interlanguage.
Let’s
now turn to cases of non-quotational unilingual
language production in which, unlike the above five cases, there is no source
text. Rather the speaker is producing language on behalf of someone else: there
is a source person whose ideas are being conveyed, even though there is no
source text.
A
would-be autobiographer describes her life to a professional writer, who then
writes the autobiography. The ghostwriter may occasionally quote from
tape-recordings or from notes taken during interviews with the subject, or from
documents she provided, but this is not inherent in ghostwriting.
At
a conference, participants may discuss specific topics in small groups. Later
each group appoints a rapporteur, who prepares an
oral summary of the group’s discussion for a plenary session. These summaries
may later be turned into written form.
The
term paraphrase is used in many different ways. Here we will use it to mean intralingual quotation. As already noted, if we leave out
exact-words quoting and plagiaristic quoting, when people quote others intralingually, their goal is to clarify what has been
previously said. Thus paraphrasing is not imitative with respect to previous
speech. For example, a nurse in a hospital may paraphrase what a doctor has
just said in language that enables the patient to understand, through
explanatory expansions and use of non-technical language.
Paraphrasing
sometimes occurs as a post-translation production process. Given a text in one
language, a text in another can be created in two steps, the first of which is
translational and the second non-translational. For example, in theatre
translation, a translator may produce a very close translation of the SL
script. The output of this first stage will generally not be considered useable
by the intended receivers. It is therefore reworked, intralingually,
by a dramaturge to make it suitable for performance on a TL stage. The
reworking can in principle be done by someone who does not know SL.5
(4) Non-imitative
two-language quotation
Two-language quotation may fail to be SIQ because it is non-imitative or
non-sequential. The most obvious case of this is one where a text in one
language is adapted for readers in another language who require the addition of
explanations (translation for children or for non-experts). The language
producer quotes sequentially but
clarifies as he works. He may experience
this as a single act, but it is perhaps best
conceived as two language production acts going on at once: translating +
paraphrasing.
The
activity sometimes known as ‘transediting’, where
translators ‘rewrite’ poorly written source texts as they translate, may
include a certain amount of translating + paraphrasing if ST is not suited to
its TL audience. However most transediting work
consists not of clarifying but of ‘smoothing’ the text so that it is
mechanically easier to read (by re-organizing the order of presentation,
adjusting poor fous, eliminating confusing
redundancies, creating coherent inter-sentence connetions,
and so on). These transediting tasks are not
demonstrative, i.e. transediting is for the most part
non-quotational. The language production work
involved might be described, as the term ‘transediting’
suggests, as simultaneous translating + editing (revising).
Finally,
as already discussed in connection with the criterion of sequentiality,
it’s possible to quote (i.e. demonstrate features of) a text in another
language non-sequentially when that text is being used merely as a source of
inspiration. If one particular text in another language were to be used as the
main source of inspiration, a comparison of that text with the newly written
text might describe the latter as a translation with massive additions and
subtractions. However, whatever the outcome might look like, the method of
production was not sequential: the text in another language was simply mined
for inspiration at various points during the composition process; it was not processed
as a sequence of text units (read one unit – translate – read next unit –
translate, and so on).
A
production-oriented approach yields a far more satisfying account than one
which tries to classify every text
–every product of language production—as either a translation or a
non-translation. A classification of texts will either have arbitrary boundaries (where is the cut-off
between ‘free translations’ and ‘adaptation’?) or it will have to resort to the
unsatisfying notion of a ‘cline’ from translations to non-translations. With
the approach proposed here, we can simply describe each chunk of the text as having been produced either translationally or non-translationally,
depending on whether it was produced by SIQ or not. We are thus
freed from deciding whether a given text is translational or not. A text can be
partly translational under this approach, for a language producer can
switch back and forth from SIQ to non-SIQ as he processes a single text. That is, at any point, he can stop
quoting or stop imitating or stop processing sequentially, or any combination
of these. A text can thus be
described as having been produced by a mixture of SIQ and non-SIQ activities.
6
A
second benefit is that we avoid any hint of prescriptivism. The fact that an
instance of language production fails to meet one or more of the criteria for
SIQ in no way prevents a society from labeling the result a translation. For
example, a society might require the texts it calls translations to be non-imitative,
or it might recognize indirect-speech renderings as translations. This would
have no bearing whatsoever on our definition, which is completely independent
of social labeling and does not seek to encompass the various social norms we
find in history.
To
be very clear: the process ‘translating’, as defined here, is not that
process which leads to a text which some society deems to be a ‘translation’.
There is no reason to think that some single process results in just that set
of texts which are deemed to be translations. All the following are possible:
translating
(SIQ) occurred output
is deemed a translation
yes yes
yes no
no yes
no no
The
first possibility in this list (translating occurred + output is deemed a
translation) is the one which creates the connection between SIQ and
translation in the social sense. The concept of SIQ would be arbitrary and
pointless if it were not the case that much of what is socially described as
translation is in fact sequential imitative quoting.
Just
as a text can be recognized by a society as a ‘translation’ even if it is not
produced by SIQ, so can its producer be recognized as a ‘translator’. A person
can switch from SIQ to non-SIQ language production within a single text; this
may or may not disentitle him to the social label ‘translator’. Thus a
(socially defined) ‘translator’ is not necessarily someone who is engaging in
SIQ, and someone who is engaging in SIQ may or may not be a (socially defined)
translator. An example, briefly mentioned earlier, is indirect-speech
translation. Since indirect speech is not demonstrative, it does not count as
quoting under our definition. Clearly, however, translators often use indirect
speech. Under the approach taken here, such individuals are serving (socially)
as translators but they are not engaging in SIQ.
A
third benefit is that definition is applicable applicable
in all times and places and to all the modes and media of translation: oral,
written and signed; dubbing, subtitling, dialogue interpreting, simultaneous
and consecutive interpreting, and document translating.
A
fourth benefit is that the definition lends itself to predictions. For example,
one might predict that those parts of a text which are produced by SIQ will contain
‘interference’, mainly because of the short time interval between reception of
ST and production of the translation.
Endnotes
1
The dramatization metaphor suggests the creative aspect of the
translator’s work. However it does have a drawback in that it may suggest a
fictive type of language production, which is not intended. In a discussion of
the uses of quoting in conversation, Tannen (1989:
98ff) emphasizes the creative aspect of quoting but declines to pick out those
cases in which the quotation refers back to a specific real prior discourse.
Instead, her discussion is very target-context-oriented: quoting is seen as
having a variety of functions in the rapporteur’s
context. For example, the quotation may simply be a way of livening up an otherwise
dull narrative: the dramatized persona did not actually say anything at all. Or
the rapporteur is speculating about what someone was
thinking, and presents the thought as a quote, as in the colloquial “Peter was
like: why are you doing that?”, meaning that Peter reacted in such a way that
one could speculate that he was thinking “why are you doing that?”. These kinds
of fictive quoting work — reporting of imagined thought or speech — are not
relevant here. In Translating there is always a real textual source.
2
The
demonstrative approach to quoting yields a satisfactory account of what Catford
(1965: 22-4 and 56-63) calls restricted translation. One of his examples is
actors mimicking foreign accents: an actor in an English film who is playing a
French character uses English words and syntax but French phonology. Thus ‘the’
is pronounced ‘ze’ and so on. This Catford calls phonological translation.
Catford’s account often strikes readers as somewhat
quirky. At first, one might wonder why he did not simply use an interlingual example: an actor can say ‘ze’
instead of ‘the’ not just in an original English film, but in a dubbed French
film, with the dubbing actor retaining French phonology while using the lexicosyntax of the TL. However, there is nothing wrong in
principle with what Catford was doing here: including
both interlingual and non-interlingual
language production within a single theory. The approach taken in this present
article also relates interlingual and non-interlingual production, but through the concept of
quotation.
Catford’s mistake was I think twofold. In part, he made a
poor terminological decision: it was not a good idea to use the term
‘translation’ for the non-interlingual case where an
actor in an original English film imitates a French accent. A more serious
problem with his account is his need to invoke a special concept of ‘restricted
translation’ for cases such as the one under consideration, where only
phonology is ‘translated’. If we understand translation as quotation, and
quotation as demonstrative, then we do not need any special notion of
restricted translation (in this case, demonstrating phonology but not lexis or
syntax), because restriction is inherent in the demonstration theory of
quotation. Quotation is understood as demonstrating certain features of the
source but not others. So in our dubbed French film with French accent
retained, the phonology of ST is demonstrated, as is its illocutionary force
and propositional content, but not its lexicosyntax.
(The case of the English actor using a French accent in an original English
film is of course non-quotational, since there is no
source text. But one can imagine a unilingual case which is quotational:
I quote a Francophone who was speaking in English, and use a French accent as I
report what she said.)
3
If
paraphrasing is not translating (i.e. not SIQ), that means rejecting Jakobson’s well-known concept of ‘intralingual
translation’, or rather, rejecting the widespread misinterpretation of the
relevant passage in Jakobson’s essay “Linguistic
Aspects of Translating” (Jakobson 1959). Jakobson does say that a
translation is a reported speech, but a close reading reveals that the article
is actually about stating in L2 the meaning of an L1 expression. Consider a
French text on fruit harvesting that contains the phrase blessures mécaniques de pommes ‘mechanical injuries of apples’. I can
state the in-context meaning of blessures mécaniques either in French (‘lésions
à la suite d’un accident plutôt qu’une
infection’) or in some other language (‘defects of mechanical rather than
infectious origin’). Now suppose I produce this English, talking either to
myself or to a colleague: “He means defects of mechanical rather than
infectious origin”. Such a statement is not functioning as a report; it is a
heuristic device whereby I spell out the meaning of the passage in hopes of
eliciting a useable translation (such as ‘bruising’). As Jakobson
puts it:
...the
meaning of any linguistic sign is its translation into some further, alternative
sign, especially a sign “in which it is more fully developed”, as Peirce, the
deepest inquirer into the essence of signs, insistently stated. The term
‘bachelor’ may be converted into a more explicit designation, ‘unmarried man’,
whenever higher explicitness is required. We distinguish three ways of
interpreting a verbal sign: it may be translated into other signs of the same
language, into another language, or into another, nonverbal system of symbols.
(1959: 232)
Jakobson immediately contrasts this
with reporting:
Most frequently,
however, translation from one language into another substitutes messages in one
language not for separate code-units but for entire messages in some other
language. Such a translation is a reported speech. (1959: 233)
Nothing more is said in the article about reported speech, and it
is hard to see any connection between Jakobson’s
mention of it and the main line of his discussion, where a ‘translation’ is an
elaborated statement of meaning.
4
It
is a fact rarely noted (but see
Berman ?1986) that the degree of imitation which is possible interlingually is typically greater than what can be achieved within a language. Thus if
a word is replaced with a synonym, the synonym will almost always be more (or
less) formal, or more (or less) technical, or re (or less) literary, or more
(or less) current, and so on. But if the word is translated, it will very often
be possible to find a word at the same level of technicality, formality etc.
5
Note
that consecutive interpretation based on note-taking, although certainly a
two-step process, is not an instance of non-translational production because a
unilingual speaker of TL does not (and in principle cannot) perform the second
step. The notes, typically in a mixture of SL and TL and various non-linguistic
symbols and graphics, are generally understandable only to the individual who
produced them.
6
To
see the value of a production-oriented approach, let’s contrast it with Gutt’s Relevance Theoretic approach (Gutt
2000), which attempts to encompass both production and reception.
According
to Sperber & Wilson, interpretive uses of
language “achieve relevance by informing the hearer of the fact that so-and-so
has said something” (1986:238). In
other words, it is not enough for the speaker/writer to intend to convey
information about the prior utterance; the speaker/writer must also intend
listeners/readers to recognize this intention to represent the prior utterance.
Now
in the various forms of oral translation, this recognition of intention by
listeners is automatic: the listener has headphones at a conference, or can see
the translator standing next to the ST speaker in a courtroom. However written
translations may not be recognized as such. Indeed, it is very frequently the case
in written translation that translators try to conceal the fact that the text
they are producing is a translation. In other words, they do not want the fact that they are
representing a prior text to be recognized. Frequently, they succeed. Either
readers do not realize at all that the text is a translation, or their
realization of it recedes from consciousness as they read. One reason this can
happen is that, unlike what happens in conversational reported speech,
translations are usually not embedded in a larger text, or introduced by a quotational phrase, as in: Speaking in French, Mr Chretien said “We are
maintaining Canadian sovereignty”. Thus English translations of Anna Karenina do not begin Leo Tolstoy wrote “Happy families are all
happy in the same way...”. Of course, the source-text author and the
translator may be identified on the title page, but as people read along in the
text and get wrapped up in the story, the fact of translation disappears from
their minds.
Now,
one could argue that the presentation of a novel not as an original story but
as a translation of a classic by Leo Tolstoy is important to readers. However,
the great majority of translated texts nowadays are not translations of
literary works but scientific, technical, commercial, legal and administrative
documents. Most often, neither the original author (who may be a committee) nor
the translator are identified. If there are very clear instances of unidiomatic
language in the document, readers may suspect that it is a translation;
otherwise they will not. But even if they do so recognize it, that is of no
importance: one may care that one is reading not an English novel but a
translation of a great classic of Russian fiction, but who cares whether the
instructions for operating a CD-ROM player were originally written in English
or in Japanese?
From
the point of view of Relevance Theory, if a reader takes an English translation
of a French astronomy book to be just a book about astronomy, and does not
understand it as a representation of the original French, then its status is no
different from an original English book about astronomy. In other words, the book achieves
relevance by virtue of representing the writer’s thoughts about black holes and quasars, not by virtue of
resembling another text about
astronomy. According to Ernst Gutt, such a book does
not count as a translation. Since
Relevance Theory insists on the reader’s recognition of the writer’s
intention to represent another utterance, Gutt is
reduced to saying that such a document ‘involves’ translation at the production
phase, but then somehow ceases to be translation at the time of reception
(2000:215-221). This extremely
unsatisfying conclusion does seem to follow from Gutt’s
adherence to Relevance Theory.
Consider
a sentence that might appear in such a translation--“The nearest black hole is 90 light years away”. This sentence will be taken descriptively (as a statement about black holes) rather than
interpretively (as a representation of the original French book) even
if the reader happens to be aware, at the moment of reading, of the fact of
translation. The reason is simple: since the translator is not an
astronomer, no one is interested in what the translator thinks about black
holes. The only people who are really interested in translations as
representations of other utterances are translators themselves, translation
teachers, translator certifying agencies,
and scholars who study translation or reported speech. The average
reader will only think about the fact of translation if there is a suspicious
passage—if a sentence does not seem to make sense, a reader who is aware of the
fact of translation might stop to wonder if the text in the original language
actually said that. Otherwise, the
average reader will not be concerned with the degree of resemblance between
translation and source, or with the fact of translation itself.
This
will be true even when the fact of translation is evident. Most noteworthy here
are cases where the fact of translation is not signaled by the unidiomatic
wording of the translation, or the presence of a quotational
phrase, but by other means. For example the physical presence of the court
interpreter in the courtroom makes it clear that she is representing in one
language what the witness or attorney or judge has just said in another
language. The fact of translation is similarly made clear by the voice coming
through headphones at a conference where simultaneous interpretation is
provided, and by the subtitles appearing on a movie screen as characters speak
in a foreign language. In all these cases, the translator does not have to do
anything at all to convey the intention to represent a prior utterance; he or
she simply speaks or writes, and the fact of translation is evident.
Yet
despite the evident fact of translation in all these cases , most people will
take no interest in it. At a movie, people who are bilingual may compare the
characters’ speech with the subtitles, but unless something goes wrong that
calls the translator’s competence into question (a subtitle makes no sense),
monolingual moviegoers will pay no attention to the fact of translation or to
the question of resemblance between translation and original. This is because,
as Pym points out (1992), equivalence defines translation in the public mind.
In other words, during reception, people think: it’s a translation - that means
it’s equivalent to ST. This being so, there is no point in giving any thought
to the fact of translation.
The
upshot of this discussion is that a useful theory of translation must separate
what happens at the moment the translation is produced from what happens at the
moment it is read or heard. Translations are produced as reported speech, but
they are very often not received as such. Relevance Theory is not suited to
this situation, since it is inherently reception-oriented. That is, it starts
from what happens when people are trying to understand language, and only
brings in production by implication: if people understand things in such and
such a way, then speakers must produce language tailored to that way of
understanding.
References
Berman
?1986 “L’essence platonicienne
de la traduction”?
Catford 1965, A Linguistic Theory of Translation
Clark
& Gerrig 1990, “Quotation as Demonstration” in
Language 60:4
Gutt 2000, Translation and Relevance
Jakobson 1959, “On Linguistic Aspects of Translation” in
Reuben Brower (ed) On Translation.
Pym
1992, Translation and Text Transfer
Sperber & Wilson 1986, Relevance: communication and
cognition.
Tannen 1989, Talking Voices