JAM ES S HOLM ES

TRANSLATED!
Papers on Literary Translation and T ranslation Studies

with an introduction bv Raymond van den Broeck

S e c o n d E d itio n
 Describing Literary Translations: Models and M ethods1

1. No one with an interest in translation studies, on looking through any

standard bibliography ofliterary research, can fail to be struck by the fact
that, of all the diligence and m idnieht oil represented in its listings, such a
slight am ount was spent on the exam ination o f translations. F o r all their
prime im portance in the history o f E uropean literature, translations have
by and large been ignored as b â t a r d b rats beneath the recognition (let
alone concern) o f truly serious literary scholars.
Even so, over the past century there have been what m ust sum up to
hundreds of scholars who (to change the image) have moved out into what
they considered the marches and outlying regions o f literary studies,
producing m onographs that attem pted, in one way o r another, to describe
the relations between a literary text or set of texts and its or their translation
or translations. The appalling thing, really is not th at there are, com para­
tively, so few such studies, but th a t so m any of the studies th at have been
made are so haphazard, so piecemeal, so norm ative. And so naïve in their
methodology.
T h e naïveté in particular is striking. All too frequently one closes a study
of this kind with the feeling th at its au th o r, engrossed in his texts, failed to
take time out to consider just w hat the process is that manifested itself
between the existing original text and the new text produced by the transla­
tor. True, it is very useful to make a distinction between the product-,
oriented study of translations and the process-oriented study of translat­
ing. But this distinction cannot give the scholar leave to ignore the,
self-evident fact that the one is the result of the other, and that the nature of
the product cannot be understood without a com prehension o f the nature
of the process.
D uring the past quarter century, of course, scholars have also devoted a
great deal of thinking, if not of research, to the translation process as such.
F o r the most part, these have been scholars o f anotheri!k>,not philologians
but, prim arily, linguists, now and then with the aid o f a'm athem atician or
psychologist. Yet the results o f their thinking, to o , would seem to be in
large p art sim plistic and naïve, at least when applied to highly complex
entities of tEeTond that “ literary texts” 2 tend to be.
My point, then, is that if the emerging generation of scholars working
 82
with translations are to avoid the errors of their intellectual forebears, they
m ust develop an adequate model of the translation process before they can
hope to develop relevant m ethods for the description of translation
products.

2.1. The earliest explicit, more or less formalized models of the translation”
process were designed in the late forties and early fifties as bases for
program s of research into the feasibility o f so-called autom atic translation.
These models started from the notion that texts w.ere_strings of words (or
“ lexical items” ) which could, in the m ain, be translated item by item, if
only a few allowances were made for the unfortunate tendency of lan­
guages to exhibit language-pair differences in syntax and to create diver­
gent egocentric (that is, “ idiom atic” ) phrases.3 Later this basically lexical-
rank model was replaced by a sentence-rank model, in which (to cite the
terminology used by one of its forem ost advocates, Nida) a source-
language passage was converted into a receptor-language passage via a
tripartite process of analysis, kernel-level transfer, and restructuring.4

SOURCE L ANGUAGE RE CEPTOR L ANGUAGE

TEXT T R A N S L A T IO N

ANAL YS IS R E S TR U CT U RIN G

TRANSFER

Figure 1. Nida’s model o f the translation process (Nida 1969:484).

The shift from lexical-rank tosentence ra n k _was a significant step tow ards
sophistication, but the basic premise rem ained that a text is a string of
units, essentially serial in nature.

2.2. A fundam ental fact about texts, however, is that they are both serial
««¿.structural — that after one has read a text in time, one retains an array
of data about it in an instantaneous form. On these grounds, it has m ore
recently been suggested (though nowhere, as far as I know, clearly set out
in model form) th atjh etran slat-io n o ftex ts (or at least of extensive texts, or
at least of complex texts) takes place on two planes: a serial plane, where
one translates sentence by sentence, and a structural plane, on which one
abstracts a “ m ental conception” o f the original text, then uses th at mental
 83
conception as a kind of general criterion against which to test each sen­
tence during the form ulation of the new, translated text. This model might
be sketched as follows:

TEXTUAL
PROCESSIN
PLANES

TEXTUAL
PLANE

Figure 2. Two-plane text-rank translation model (TSI_ — source-language

text; T t l = target-language text).

Such a two-plane model would seem to come much closer than the eariier
serial models to describing the translation process as it takes place in the
tran slato r’s study. The introduction o f an abstract tex t-rank5 “ m ental
conception” — or, as I propose to call it henceforward, “ m ap” — would seem
to be a further step forward. v------- ’

2 .3 .1 would question, however, w hether one such m ap o r m ental concep­

tion is sufficient to model the actual translation process adequately. C on­
sider for a m om ent. Mr. X, who sometimes translates poetry into English,
has just reread a poem in French, say B audelaire’s “ La géante” . A m ong
the details in the m ap which he abstracts from the original poem will be (to
restrict myself to a few of the m ore elem entary features) th a t it is in sonnet
form , rhym ing abba abba cde cde, in syllabic verse, twelve (or thirteen)
syllables to the line. X, if he is like m ost English-language translators, will
not autom atically decide to “ retain’^tH frhym e scheme, the syllabic verse,
or the twelve- (or thirteen-) syllable lines. R ather, he has a num ber-of
options to select from . On the basis o f these selections (and a great m any
others) he in fact develops a second m ap, in various ways like the first, bu t
in others quite different. It is this second m ap, not the first, which he uses as
his criterion to guide him in carrying out his translation on the serial plane. -
If this really approxim ates the way in which the tran slato r works, then we
arrive at the following two-plane, tw o-m ap model:6 -
 84

RULES PLANE

TEXTUAL
PR O C E SSIN G
PLA N ES

TEXT PLANE

Figure 3. Two-map two-plane text-rank translation model (TSL = source-

language text; Tn = target-language text; DR = derivation rules; CR =
correspondence rules; PR = projection rules).

In my sketch of this model I have taken the further step of introducing

three sets o f rules by which specific phases of the translation process would
seem to be carried out. (It goes w ithout saying that in actual practice the
different phases are not always separated from each other in time; like
other hum an beings, the translator can be doing various things at once.7)
O f the three rule sets, the first, that of derivation rules (D R), determines the
way in which the tran slato r abstracts his m ap o f the source text from the
text itself, and the third, that of projection rules (PR), determines the way
in which he makes use of his map of the prospective target text in order to
form ulate the text, while the second, that of correspondence rules (CR) or
m atching rules (MR) — or, if one prefers, equivalence rules (ER) — deter­
mines the way in which he develops his target-text map from his source-text
m ap. It should be noted that the first o f the three phases described here the
tran slato r shares with every reader o f literary texts, the third with every
writer; the second, however, that of developing a target-text map from his
source-text m ap by means of correspondence rules, is uniquely a transla­
tional (or least a specific kind of m etatextual) operation, and as such
deserves our special attention.

2.3.1. It should be realized in this connection that the m ap of the source

text, if the translator-to-be who abstracts it is a skilled and experienced
reader, will be a conglom erate o f highly disparate bits o f inform ation. In
the first place, as a m ap of a linguistic artefact, it will contain inform ation,
at a variety of ranks, regarding features o f the text in its relation to the
linguistic continuum within which (or violating the rules o f which) it is
 85
form ulated, that is, contextual inform ation. Secondly, as a m ap of a
literary artefact, it will contain inform ation, at a variety of ranks,
regarding features of the text in its relation to the literary continuum within
which (or rebelling against which) it is form ulated, th at is, intertextual
inform ation. And third, as a m ap of a socio-cultural artefact, it will contain
inform ation, at a variety o f ranks, regarding features of the text in its
relation to the socio-cultural continuum within which (or transcending
which) it is form ulated, th at is, &irnatioflal-information.8

2.3.2. In the case of all three types o f inform ation, and at the various ranks
within each type, the tran slato r, as soon as he sets about seeking
correspondences with which to design his target-text map (or, in more
everyday terms, as soon as he starts thinking “ H ow am I going to translate
this?” ), is confronted by two dilemmas. } ,-
The first is that, for each feature in his source-text m ap, at least two
kinds of corresponding target-text m ap features will usually be available.
There will frequently be a feature which corresponds in form , but not in
function — a feature which m ight therefore be called (if one may borrow
the term from the biologists) a hom ologue. There will usually be a feature
which corresponds in function, but n o t in form — an analogue. A nd there
may also be a feature which corresponds in meaning, but in neither
function n o r form — should we, w ithout the help of the biologists, call this
a sem antologue or semasiologue? It is a rare thing when the three happen
to Coincide-across language barriers.
To return to my hypothetical tran slato r Mr. X. Should he, in his English
translation of “ La géante” , “ retain ” such features as syllabic verse, the
twelve- and thirteen-syllable line, the “ C ontinental” rhyme scheme, all of
them hom ologues, that is to say, in the English setting parallel in form to
the French, but clearly not in function? Or should he choose analogues:
syllabotonic verse, ten-syllable lines, the rhym e scheme of the English
sonnet? These are obviously m om entous choices, and which ones he is to ,
make and which to reject will be determ ined by the correspondence rulesL
which the translator has consciously or unconsciously established on the
basis of his confrontative knowledge of the French and English languages,
literatures, and cultures. It is clear th a t his choices will not always be of the
same kind for every type of inform ation, or at every rank. F or instance,
translators usually choose hom ologues when dealing with socio-cultural
features, but tend to choose analogues for m any linguistic features, while'
they would seem to exercise a great deal o f freedom of choice as regards the
various features of poetic form : one finds accepted (and, I should think,
acceptable) English translations th a t are form ally hom ological, analogi­
cal, or a m ixture of the two — and, indeed, m any translations that have
 86
abandoned all correspondence in this regard.
That last rem ark points to the second dilemma. The experienced
tran slato r will have discovered that there is a certain interdependence
am ong correspondences: the choice of a specific kind of correspondence in
connection with one feature of the source-text m ap determines the kind of
correspondence available for another or others, indeed in some cases
renders correspondence for certain further features infeasible or even
unattainable. The choice of blank verse instead o f the rhym ed couplet, for
instance, confounds correspondence at the phonic level and makes
correspondence on the end-stop/enjam bm ent axis for all practical
purposes impossible.
It follows that the translator, whether or not he is conscious of it,
establishes a kierarch))o f correspondences. Mr. X, for instance, may give
priority to homological correspondence at the rank of “ C ontinental"
sonnet form , and as a result of this strict form al choice be compelled to
reduce his correspondence requirements in regard to the semantic content
of the poem. Or he may assign close matching of the sem antic content such
priority th at he is forced to abandon any attem pt at correspondence at the
rank o f sonnet form (perhaps justifying him self in doing so by arguing that
free verse is a contem porary English analogue of the nineteenth-century
French sonnet).
In the case of many less complex text types, o f course, solutions to this
problem of correspondence hierarchy are fairly clear. The translator of
w hat (using for the m om ent a typology derived from the one developed by
the Czech structuralists)9 we may call an inform ative (or referential) text
will tend to give full priority to semantic correspondence, a n d will retain
other correspondences only when they do not interfere with that priority.
In the case of a vocative (appellative, conative) text, for instance a TV
com m ercial or a serm on, on the other hand, the translator (unless skewing
of function is required) will give priority to establishing^coiTespandeiicfiJif
a p p e a l, even at the cost of having to overhaul the sem antic message
completely. The literary text, however — and whatever we may call its
basic function: “ poetic” , “ esthetic” , “ reflexive” , “ fictive” — is a much
m ore complex entity, which may at various points (or indeed sim ultane­
ously) be informative, vocative, expressive, or for that m atter meta-lingual
or meta-literary. This makes the establishm ent of a hierarchy of
correspondence priorities a much less clear-cut problem , and various
translators will choose various solutions, none of which is dem onstrably
“ right” or “ wrong” (though the translator may think they are), but usually
“ somewhere in between” .

3. If this is a fair description of the literary translation process, in other

 87
w ords of the way in which the literary tran slato r goes about his business,
then the task for the scholar who wishes to describe the relationship
between the translated text and its original would seem to be obvious. He
m ust attem pt to determ ine the features o f the tran slato r’s two maps and to
discover his three systems of rules, those of derivation, projection, and,
above all, correspondence — in other words, the translator’s poetics.

3.1. T hat is m uch m ore easily said th an done, and I should like to devote
some tim e to considering this problem further. In a few instances the
analyst’s task may be m ade som ew hat less difficult for him. In the case of
contem porary translations, for example, he may be able to consult the
translato r himself. But many translators, even brilliant ones, are less than
eloquent in speaking or w n tin g a b o u t their craft, and in any regard there is
often a m arked discrepancy between the explicit poetics a translator avows
to subscribe to and the impjicit, often subconsciouspoetics he adheres to in
actual practice. Even in the case of older translations, there may be at least
some pointers tow ards a poetics to be found in the tran slato r’s prefaces,
com m entaries, notes, letters, and the like. A nd the limits within which an
individual poetics moves in a given literature in a given period can be
dem arcated grosso modo hy a study o f the translation theory of that
literature in th at period. Though in these instances, too, one must
constantly be on the lookout for discrepancies between theory and
practice. '

3.2. In m ost cases, however, the analyst is left with little o r no material
beyond the tw o (o r m ore) texts, the original and its translation(s), and it is
from these alone th at he must attem pt to derive his description. How can
he set about his task? A t the risk of simplifying the problem , I should like to
restrict myself here to considering what I have suggested is the m ajor
aspect of this task, that of attem pting to retrace the translator's two maps
and the correspondence rules determ ining their relationship.
Clearly, the analyst will have to approach this problem in a different way
from th a t of the tran slato r. The tran slato r, I have argued, derives a m ap of
the source text from the text itself, next applies a set of correspondence
rules, some of them m ore or less predeterm ined and some m ore or less ad
hoc, to develop a target-text map from the source-text map, and finally uses
this second m ap as a guide while form ulating his target text. The analyst,
on the other hand, starting from the two texts, will as a first step apply a set
o f derivation rules to each text in turn, in order to obtain maps of the two
texts. His next step will be, with the aid of a set o f com parison rules, to
com pare the two maps in order to determ ine the network ofcorrespond-
ences between their various features. This will then be followed by a third
 Figure 4. M odel o f the translation-descriptive process (T = text; SL =
source language; TL — target language; M = map; TR = translator; TS —
translation scholar; CN = network o f correspondences; CR — corres­
pondence rules; DR = derivation rules; CpR = comparison rules; AR =
abstraction rules).

step in which, with the aid of a set of abstraction rules, he derives a set of
correspondence rules and a correspondence hierarchy from the network of
correspondences.

Only in one phase of one of these steps does the work of the analyst parallel
th at o f the translator: in the operation of deriving the source-text map from
the text. The operation of deriving the target-text m ap, on the other hand,
is for the analyst the reverse of the operation perform ed by the translator
(though at the same time parallel to the analyst’s operation of deriving the
source-text m ap, and requiring com parable discovery7 procedures).
Similarly, the abstraction of a network of correspondences from '-he maps,
and of correspondence rules and a correspondence hierarchy underlying
th at netw ork, is an operation of quite a different kind from those
perform ed by the translator.
' A further com plication is one that applies to all studies of mental
: processes. Since in most cases there is little or no tangible evidence of what
, has taken place in the tran slato r’s “ m ind” except the text he has produced
as com pared to the original text, the scholar attem pting to trace the
relationship of the two texts likewise in m ost cases has no m aterial except
those two texts from which to derive his_conclusions. A nd since the
descriptive process he pursues is, though in a different way from the
tra n sla to r’s process, extremely complex, there is great danger th at the
 89
results of his analysis will be highly subjective and so o f little value to other__
scholars. Assuming th a t objectivity in any true sense is in such a m atter a
goal even m ore unattainable than in research dealing with tangible objects
a n d /o r events observable outside the “ m in d ” , one can nevertheless posit
that a high degree of intersubjectivity is an aim w orth striving after in a
research situation of this kind.

3.3. There would seem to be a choice for the analyst between two basic
working methods. In the first, the descriptive scholar, upon studying the
two texts, will derive from them a list o f distinctive features which strike
him as significant and deserving o f com parative analysis; frequently he will
also determine a hierarchical ordering of the features. The well-trained
analyst will, it m ust be assum ed, bring with him a detailed knowledge of
linguistic, literary, and socio-cultural theory1 such th at he can identify
contextual, intertextual, and situational elements in the texts in a m anner
acceptable to other scholars, and this, it m ust likewise be assum ed, will
provide at least a modicum of intersubjectiidty to his application of
linguistic, literary, and socio-cultural research m ethods. But the fact
remains th at none of the disciplines concerned with the nature of texts has
given us a generally accepted intersubjective m ethod for determ ining
distinctive features in a concrete text, so th at their selection remains’ to a
large extent an ad hoc operation. The result will consequently be th a t the
maps of the two texts derived by the analyst, like the tran slato r’s two maps,
will be incomplete: the analyst, for instance, m ay very likely discover blank
spaces (indications of terrae incognitae) in the tra n slato r’s m aps, but
overlook the blank spaces in his own — and precisely in such terrae
incognitae, as in parts of Africa in the old m aps, may be lions.
A second working m ethod, at least in theory, would be to circumvent the
problem of ad hoc selection of distinctive features by determining
beforehand a required repertory of features always to be analysed,
regardless of what specific text is involved. This m ethod, too, has at least
one major drawback: if its results are to lead to a m ap that is generally
acceptable as within reach of com pleteness, the repertory would have to be
quite extensive, and the task of providing full details on the texts would be
one that is arduous and tedious to the researcher and largely uninteresting
to the reader. fr '. j
The repertory m ethod w ould, however, assure a higher degree of
intersubjectivity to the results o f the analysis based on it — provided, of
course, th at scholars in the field could reach agreem ent as to w hat elements
should be included in such a repertory. L am bert has m ade an explorative
attem pt at a listing;10 it would seem to me th a t a further filling out and
structuring o f this listing should be one o f the m ajor foci o f research and
 90
discussion in the near future for scholars interested in translation
description. It is clear that the repertory m ust not only be quite complete,
but also cacnplex enough in structure to accom m odate a num ber of
paramsnic ases. Am ong these a m ajor one, o f course, is the axis-
microsim rnire-m esostructure-m acrostructure (from graphem e/m or- _
pheme via kiem e, sentence, and suprasentential units to text; in verse
moreover ri- line, stanza, and suprastanzaic units). But other axes .
i n t e r s t a t e one, notably th at o f form -m eaning-function (morphologue-
sem asiok^e-analogue) and that of (linguistic) contextuality - (literary)
in ten e x u srrr - (socio-cultural) situationality, and these axes too would
have 10 k incorporated.
The rase ci working out such a repertory would be enorm ous. But if
scholars to arrive at a consensus regarding it, in the way, for instance,
that txcsriso since Linnaeus have arrived at a consensus regarding
system ^- —ethods for the description of plants, it would then become
possible, f ” the first time, to provide descriptions o f original and
translated —cs, of their respective maps, and of correspondence networks,
rules, arc rrs'archies that would be mutually com parable. And only on the
basis of —rrnally com parable descriptions can we go on to produce
well-fo— studies of a larger scope: com parative studies of the
transiarircs of one author or one translator, or — a greater leap — period,
genre, coe-iinguage (or one-culture), or general translation histories.
Such sci-i. of course, the scholars o f our generation have tended to
rejecn ¿ s y seem to us unattainable, and so outside the range of our
less-thar-^rlting am bition. It is in any case certain th at they exceed the
grasp of ¿ e subjective, largely intuitive and impressionist methods still so
often beizx applied today. And only a more explicit, a m ore precise, a
stricter i s ; more intersubjective approach holds any prom ise o f greater
things ~ r ^ e .

Notes

1. T k pzn of this paper devoted to the m odel o f the tran slatio n process is an
outgrovc; cc a sem inar on the subject held with students at the University of
A m sie n sa in the spring o f l 975. This m odel also served as the them e fo r talks given at
the U n i v s j o f Iowa in Decem ber 1975, at the International C om parative Literature
Assodadcn's C olloquium on T ranslation T heory organized in B udapest on 18 and 19
Augcst and at the State U niversity o f New Y ork, B ingham ton, in Ja n u a ry 1977.
Remaig gg e ed u rin g discussion o f the talks a t Iowa C ity, B udapest, and B ingham ton,
as weB is tacse brought forw ard at the Leuven colloquium , have led to clarification of
points there in the first p art o f the paper.
2. I cse a>e term here assum ing th at we are m ore or less agreed on the core o f its
meanint. m e without attem pting anything approaching a definition.
 91
3. See e.g. the discussions in various early com puter-oriented studies.
4. In the various m odels developed at this level, the m ain difference o f opinion is in
just w hat is being transferred: syntactic elem ents (N ida’s kernel o r near-kem el
sentences) o r sem antic kernels. (Eugene A. N ida, “ Science o f T ran sla tio n ", Language,
45 [1969], pp. 483-498). T he fullest discussion o f these and o th er serial m odels o f the
translation process is to be found in V.N. K om issarov, Slovo operevode (Moscow: IM O ,
1973), a book w hich I am unfortunately-unable-te read, though m anuscript translations
o f several portions o f it m ade by various students a t the University o f A m sterdam have
given me confidence th at it is a work o f high significance which needs to be translated
into a W estern language in tolo.
5. O bviously, in the case o f longer texts there will also be m esostructural ranks,
ranging from those o f p aragraphs a n d /o r stanzas to those o f chapters o f novels, cantos
o f long poem s, scenes or acts o f plays.
6. O f course the charge can be m ade th at this m odel, too, is an oversim plification of
the translation process, ignoring as it does the m esostructural ranks. Eventually it m ay
therefore prove necessary to abandon it in favour'ofa'm ore com plex m odel introducing
a hierarchical series o f m aps, ranging from sentence-rank m aps via a num ber o f
m esostructural m aps to the text maps.
7. It also goes w ithoug saying th at there is a great deal o f feedback not indicated in
the m odel; details of the target-text m ap, and in som e cases even o f the source-text m ap,
may change drastically in the course of sentence-by-sentence (o r transem e-by-transem e)
translation.
8. On this term inology see A ndré Lefevere; “ The T ranslation o f L iterature: An
A pproach” , Babel, 16 (1970), pp. 75-79, and pp. 45-52 above.
9. A nd deriving originally from Biihler. Cf. e.g. R om an Ja k o b so n , “ Closing
Statem ent: Linguistics and Poetics” , T hom as A. Sebeok (ed.), S tyle in Language
(Cam bridge, Mass.: M .I.T. Press, 1960), pp. 350-377, esp. pp. 353-357.
10. José L am bert, “ Echanges littéraires et traduction: D iscussion d ’un p ro jet” ,
Jam es S H olm es, José L am bert, & R aym ond van den Broeck (eds.), Literature and
Translation: New Perspectives in Literary Studies (Leuven, A cco, 1978), pp. 142-160, esp.
pp. 154-155.