THE THEORY OF TRANSLATION
5. Linguistic models of the process of translation
5.5. The transformational model
According to the simplest version of the transformational model, it is not reality translators identify the source language signs with but the target language signs, since the target language equivalents to the source language words are already giv
en in their minds. In this theory, translation is viewed as simple substitution; the substitution of source language signs with target language signs.
This model, however, would only work if the system of the source language and the target language were identical regarding the number, the distribution and operation of elements. That is, each source language sign would have only one equivalent in the target language, and this would link to the rest of the target lan
guage signs according to the same rules as the source language sign to the rest of the source language signs.
In the case of natural languages this is obviously impossible. But the fact that there exist no two languages whose elements and their distribution would be the same does by no means imply that there are no elements in any two languages
which would have the same distribution and would function following the same rules. According to the transformational model) this common field serves as the basis for interlingual translation) which the translator reaches via a series of intralin
gual) i.e. language internal transformations.
5.5.1. The antecedents of the transformational model
The enormous influence that Noam Chomsky’s works exerted on the developers of transformational translation models can also be detected in their terminology.
Rozentsveig) in his 1964 study on the transformational model of translation) used the term jadro (‘core’)) altough he did not use the concept of deep structure, which is not surprising, since Chomsky’s Aspects of the Theory of Syntax, introduc
ing the notion of “deep structure” is only published in 1965.
The influence of Chomsky’s generative grammar is clearly detectable in the works of another outstanding representative of the transformational translation model, Eugene Nida. In his first book. Science of Translating (1964) he describes his attitude to Chomsky’s ideas in the following way:
For the translator especially, the view of language as a generative device is important, since it provides him first with a technique for analysing the process of decoding the source text, and secondly with a procedure for describing the generation of the appropriate corresponding expressions in the receptor language. Certain comparativists and descriptivists who are working with a limited corpus of written texts may find more traditional techniques somewhat easier to apply, but for the translator, who perhaps more than anyone else must take language in its dynamic aspect, a view of grammar as a generative device has many distinct advantages (Nida
1964: 60).
In his 1964 book Nida uses the following terms: kernels (“the kernel construc
tions in any language are the minimal number of structures from which the rest can be most effectively and relevantly derived”) and basic structure (“remark
able similarities between the basic structures of different languages are increasing
ly becoming an object of study by linguists”), but does not use the term “surface structure” and “deep structure”. In 1964, he describes the process of translation as consisting of the following stages:
(1) Analyse the source language expression in terms of the basic kernel sen
tences (...),
(2) Transfer the kernel forms of the source language to the equivalent kernel forms of the receptor language (...) ,
(3) Transform the kernel utterances of the receptor language into the stylisti
cally appropriate expressions (Nida 1964: 69).
In 1969, by the time of Nida’s second seminal book written together with Charles Taber The Theory and Practice of Translation, the Chomskyan notions of “surface struc
ture” and “deep structure” were widely used in the literature and Nida’s descrip
tion of the process of translation changed accordingly (see below in 5.5.2.). He
5. Linguistic models of the process of translation
made use of the term “surface structure", but he did not use the term “deep struc
ture” (see the following footnote, where Nida referred to the term “near-kernel level”):
For this book on the theory and practice of translation we are not advocat
ing that the translator go below the level of the kernels to the underlying bases, the “deep structure”. There are certain theoretical interests in such an approach, but practically, the bases are neither useful nor advisable, since these bases cannot be readily manipulated. When the message is transferred, it is not however on precisely on the kernel level for if this were the case, the connections between the kernel elements would be lost or obscured. Therefore, the transfer is made at a near-kernel level, in which the relevant connections between the kernels are explicitly marked (Nida
1969: 39-40, emphasis by K.K.).
It should be noted, however, that the terms taken over from generative grammar do not have the same meaning in the work of Rozentsveig’s or Nida’s as in Chomsky’s.
As distinct from the latter, the former theorists use these terms metaphorically and their method of analysis also fails to meet the strict formal requirements dic
tated by generative grammar.
5.5.2. The process of translation in the light of the transformational model
According to the transformational model (Revzin and Rozentsveig 1964, Komis
sarov 1980), in the first phase of the process, the analysis phase, the translator goes back from the source language surface structure to the source language core sentences or deep structure via a series of transformations (intralingual trans
formation); in the second phase these are replaced by the equivalent core sen
tences or deep structure of the target language (interlingual transformation);
and in the third phase, the synthesis phase, the translator reaches the target lan
guage surface structure from the target language core sentences or deep structure via a series of transformations (intralingual transformation).
A similar description of the process of translation can be found in Nida and Taber (1969). They call the analysis phase “back-transformation”, the synthesis phase “restructuring” and they assume a “transfer” phase in between.
(1) analysis, in which the surface structure is analysed (...),
(2) transfer, in which the analysed material is transferred in the mind of trans
lator from language A to language B, and
(3) restructuring, in which the transferred material is restructured in order to make the final message fully acceptable in the receptor language. (Nida and Taber 1969: 33).
As a result of such analysis “kernel or core structures” are created in this system as well, which are a composite of four structural categories: objects (e.g., house, dogy man, sun, stick, water, spirit, etc.), events (e.g., run, jump, kill, speak, shine, appear,
grow, die, etc.), abstracts (e.g., red, blue, small, many, quickly, etc.), and finally rela
tions (e.g., because, and, not).
These structural categories cannot be identified with word classes. The word promise, for instance, which is a noun on the source language surface, is an event name in the kernel structure, and can also be realised as a verb on the receptor language surface. Kernel structures are such that the relationship between the ele
ments within them is expressed in the clearest and simplest possible way, ready for transfer.
5.5.3. The advantages and disadvantages of the transformational model
The advantage of transformational models is that by introducing the concept of intralingual transformation and describing their operation they provide a realistic picture of the various stages of the process of transformation. The everyday routine of translators, the constant decisions between the various target language options can be considered transformations. However, in comprehending the target lan
guage text and analysing its content, one is also conducting language internal transformations. It often happens, for instance, that complex nominal structures with multiple modifications are broken down into several simple sentences in the mind.
The idea that translation basically comprises a series of language internal trans
formations may be useful in translator training. Back-and-forth transformation of the surface sentences of the source text contributes to a better understanding of the text, and transformations carried out in producing the target language surface contribute to translators’ awareness of the target language devices.
Taking into consideration language internal transformations may also be help
ful in the assessment of translations, in determining the degree of equivalence.
Sometimes a source language and target language element sign cannot be directly equated with each other because in addition to occupying different places in the systems of the two languages, they also have a different “transformational histo
ry”. The Hungarian noun fejlődés (‘development’) is not always equivalent with the Russian noun razvitie (‘development’) because razvitie is a nominal transform of four Russian verb forms (1) razvivafsyja (imperfect indefinite), (2) razvitysya (per
fect indefinite), (3) razvivaty (imperfect definite), (4) razvity (perfect definite), while behind the Hungarian noun fejlődés (‘development’) there is only the intransitive verb, namely fejlődik (‘to develop’), while the nominal transform of the transitive verb fejleszt (‘to develop’) is another noun: fejlesztés (‘development’). A similar comparison can be also made between English and Hungarian: the English noun development has two Hungarian equivalents: fejlődés and fejlesztés, because of differ
ences in their derivational history.
The transformational model plays an important role in machine translation as well. If there exists a common area between the two languages, i.e. if there are lex
ical units and grammatical structures in the two languages whose equivalence can be predicted, then within this area there is no need for the skills or previous expe
rience of a translator. This kind of transfer can be done even by a machine, and so machine translation can be accomplished with some pre- and post-editing.
A defect of the transformational model is that, similarly to the denotative mod
5. Linguistic models of the process of translation
el, it creates a drastic separation between the source language text and the target language text. It does not take into account the important role of the source lan
guage form in creating the target language surface.
In addition, it often happens in translation that certain source language and target language structures can be treated as equivalent despite the fact that they cannot be traced back to the common area between the two languages. These may be classified as equivalences on the basis of the denotative-situational model; when the linguistic form plays almost no role in the translation: English: Beware of the dog! French: Chien méchant!, German: Bissiger Hund\, Hungarian: Vigyázz, a kutya harap! (‘Be careful, the dog bites’).