In this paper I argue that the phenomenon commonly referred to as "translation" can be accounted for naturally within the relevance theory of communication developed by Sperber and Wilson (1986a): there is no need for a distinct general theory of translation. Most kinds of translation can be analysed as varieties of interpretive use. I distinguish direct from indirect translation. Direct translation corresponds to the idea that translation should convey the same meaning as the original. It requires the receptors to familiarise themselves with the context envisaged for the original text. The idea that the meaning of the original can be communicated to any receptor audience, no matter how different their background, is shown to be a misconception based on mistaken assumptions about communication. Indirect translation involves looser degrees of resemblance. I show that direct translation is merely a special case of interpretive use, whereas indirect translation is the general case. In all cases the success of the translation depends on how well it meets the basic criterion for all human communication, which is consistency with the principle of relevance. Thus the different varieties of translation can be accounted for without recourse to typologies of texts, translations, functions or the like.
The amount of literature on translation is vast - people have written on this subject for about two millennia. However, the bulk of the literature that came to be written over the centuries does not necessarily indicate the depth of understanding that has been reached on this topic. Thus Steiner states that "despite this rich history, and despite the calibre of those who have written about the art and theory of translation, the number of original, significant ideas in the subject remains very meagre" (1975, p. 238). Levy observed that the penetration of subject matter was lacking especially on the theoretical side:
"Only a part of the literature on the problem of translation moves on the theoretical plane. Until today most studies and book publications, especially on literary translation, have not gone beyond the limits of empirical deliberations or essayistic aphorisms." (Levy 1969, p. 13; translation my own)
About half-way through this century things began to change. Scholars increasingly began to call for a well-founded scientific study of translation. At first linguistics seemed to offer the framework needed, but it soon became clear that it would not be adequate on its own. So today there is a strong call for a multidisciplinary investigation: linguists, psycholinguists, sociolinguists, semioticians, anthropologists, teachers and, of course, translators are all called upon to tackle the problem together.
The approach generally advocated for this multidisciplinary research is essentially an inductive-descriptive one: by examining the phenomena found in translation, one aims to discover regularities that can be stated and will then form the science of translation.
However, even at this early stage questions have arisen about the value of the likely outcome of this effort. Firstly, translations seem to be so varied and the number of factors on which they depend so large that it is not clear that more than statistical generalisations can be made. Secondly, given the variety of domains that need to be considered, what sort of a science is likely to evolve from this enterprise - will it be anything coherent at all? Thirdly, since the outcome of such inductive investigations will be crucially determined by its input, how can one avoid the risk of circularity? In other words, how can one avoid the danger that the concept of translation to be developed will be merely a reflection of what one took it to be in the first place - that is, something dependent on the investigator's opinion?
Related to this last point is the problem of evaluation and decision-making in translation: it is difficult to see how an inductive-descriptive approach can deal adequately with the problem of evaluating translation since by nature it describes what is rather than what should be. Yet the concern for quality control in translation seems to be one of the major driving forces behind the search for systematic accounts or theories of translation: it is hoped that the explicit and systematic treatment of the subject matter will make possible the setting of objective standards.
Most of these attempts at the scientific treatment of translation have followed the structuralist approach to language, relying heavily on categorization, especially of text and translation types. While this in itself proved to be a major challenge, matters became more complex still when extra-linguistic factors like the function and purpose of a text, and even particular interests of the target audience had to be considered.
The following example from Neubert (1968), discussed in Wilss (1982), gives an idea of the difficulties involved in capturing all these factors in a single theory of translation. The example concerns a passage from John Braine's novel "Room at the Top" where the colour of the sky is described as "the grey of Guiseley sandstone". Summarizing Neubert's discussion of this example, Wilss points out that the expression "Guiseley sandstone" could be translated into German either as "Guiseley-Sandstein" or simply as "Sandstein", and he claims that the decision as to which rendering is the right equivalent will depend on the interest of the receptors:
"If this interest is exclusively focussed on literary aspects of the original, the translator can confine himself to the reproduction of 'Guiseley sandstone' by 'Sandstein', ... . If, on the other hand, the translator must reckon with additional interests of the reader in area studies, he must react accordingly, because in a case like this only a translation containing an explicit reference to 'Guiseley sandstone' would meet TE [=translation equivalence] expectations." (Wilss 1982, p. 145)
Wilss concludes from such examples that "TE [= translation equivalence] cannot possibly be integrated in a general translation theory (...), but must be looked upon as part of specific translation theories which are at best text-type-related or, even more restrictedly, single-text- oriented" (1982, p. 135)
It is interesting that Wilss does not discuss the fact that this view entails a reductio ad absurdum as far as theoretical concerns go; after all, one of the main points of theory-construction is that it should allow us to explain complex phenomena in terms of simpler ones, that is, one of its main motivations is to make generalisations about phenomena. But if it turns out that each individual phenomenon, that is, each text, or even each instance of its use with a particular audience, may require its own theory, then this means that the phenomena in question are not accounted for in terms of generalisations at all, but that they actually fall outside the scope of theory.
One recent reaction to this development is Snell-Hornby's "integrated approach" to translation (1988). She feels that the category-based approach is more of a hindrance than a help because "In its concrete realization language cannot be reduced to a system of static and clear-cut categories" (p. 31) Therefore in her integrated approach "... the rigid typology of the objectivist and reductionist tradition will therefore be replaced by the prototypology, a dynamic, gestalt-like system of relationships, whereby the various headings represent an idealized, prototypical focus and the grid-system gives way to blurred edges and overlappings" (p. 31).
While one sympathizes with Snell-Hornby's criticism and rejection of category- and typology-based approaches, it is not clear what is gained, in the final analysis, by the changeover to prototypology. Recognizing the existence of "blurred edges and overlappings" is commendable, but without further explication the translator is left to his own devices as how to move along the cline between the prototypes.
Furthermore, one wonders what the theoretical or practical value of the prototypes themselves is. Consider the following statement:
"In this concept the historical dichotomy has been replaced by a fluid spectrum, whereby, for example, prototypically literary devices such as word-play and alliteration can be accommodated both in 'general' newspaper texts (...) and in the language of advertising, and conversely prototypically technical terms from the language of science or culture-bound items from the 'general' area of politics or everyday living can be explained and interpreted as literary devices (...)." (Snell-Hornby 1988, p. 33)
There seems to be little point, apart from descriptive convenience, in labelling word-plays or alliterations as a 'prototypically literary device' when these labels do not predict their distribution in texts, but actually cross-cut the prototypology of texts. If anything, such cross-cuts suggest that factors other than prototypology are at work.
Snell-Hornby is right when she calls for "a basic reorientation in thinking", but she does not go to the core of the problem when she sees this as "a revision of the traditional forms of categorization" (op. cit., p. 26). The problem is not the form of categorization used, but reliance on categorization as such. Categories - whether they are rigid as in traditional typologies or 'blurred' as in 'prototypologies' - are helpful for the organisation of data and the description of phenomena, but explanation and theoretic penetration require an understanding of the properties in virtue of which the phenomena interact.
As Snell-Hornby herself says, translation is a cross- cultural event - it is part of cross-cultural communication, and communication is an event in which people share their world of thought with others. Therefore, the account of translation I propose is embedded in an explanatory theory of communication that focuses on how people share thoughts with one another.1 This theory is the relevance theory of communication developed by Sperber and Wilson (1986a).2
2 The framework
In the space given, I have to limit myself to a brief sketch of those ideas of relevance theory that have a bearing on the topic in hand.
The theory offers an empirical, cognition-based account of human communication.3 It views communication as primarily an inferential process: the central task of the communicator is to produce a stimulus - verbal or otherwise - from which the audience can infer what set of thoughts or assumptions the communicator intended to convey.
Since the range of inferences one can make from any phenomenon is huge and open-ended, there needs to be some constraint that helps the audience to identify those assumptions which the communicator intended to communicate.
This constraint is provided by the principle of relevance, which amounts to the following, twofold presumption: the set of assumptions which the communicator intends to convey will be adequately relevant to the audience, and the stimulus produced is such that it avoids gratuitous processing effort on the audience's part.
This presumption of optimal relevance is necessarily communicated by every instance of ostensive communication - it is part of our human psychology. Thus whenever a communicator claims someone's attention indicating that he intends to communicate something, it is assumed by both parties that the communicator is not putting the audience to work gratuitously, but that he believes a) that what he intends to communicate is adequately relevant to the audience, and b) that the audience can recover it without unnecessary processing effort. In effect this means that the audience is entitled to assume that the first interpretation of the stimulus found to be consistent with the principle of relevance is the one intended by the communicator.
The notion of relevance itself is defined as a cost-benefit relation: the cost is the amount of mental processing effort required to interpret the stimulus, and the pay-off consists in the contextual effects derived from it. Hence the less effort the processing of a stimulus requires and the more contextual effects it has, the more relevant it will be.
Contextual effects result when information conveyed by the stimulus is inferentially combined with contextual assumptions, that is, with information already available to the audience, perhaps from memory or perception. This accounts for the intuition that for successful communication it is not enough for the information conveyed to be new; rather it must lead to some alteration of the knowledge possessed previously.
Thus suppose I told you out of the blue:
(1) There is butter available at the foreign currency store.
While I am quite sure that this information would be completely new to most of my readers, I think that they would have problems in making sense of my utterance. Firstly, you would not know which store I am talking about, and secondly you would not know what to do with this information. Relevance theory accounts for this reaction: for most readers utterance (1) does not link up with any other information readily available to them, so no contextual effects are achieved, and the utterance is felt to be irrelevant.4
By contrast, if I told (1) to a colleague in the capital of a certain African country, he would find this information very relevant: butter is not always easy to get there, and therefore (1) would readily link up with information already known to him, yielding a number of contextual effects:
(2) Information already known:
(a) Butter is in short supply.
(b) When something is in short supply, one needs to buy it quickly when it is available.
Combined with the information contained in (2), the information supplied by (1) yields assumption (3) as a contextual effect, more specifically, as a contextual implication:5
(3) One needs to buy butter quickly from the foreign currency store.
It is important to note that not all contextual assumptions are equally accessible to the audience at all times; for example, as you read this paper, the information contained in the last sentence or so will be highly accessible to you, whereas information you read at the beginning of the paper might be much less so. You may be able to recall that information too, but it would require greater effort. This relationship between accessibility of contextual information and processing cost is important for the process of context selection: under the principle of relevance it induces the audience to work with the most highly accessible contextual assumptions that will yield adequate contextual effects.
As mentioned above, relevance theory applies to both verbal and non-verbal communication alike. Verbal stimuli differ from non-verbal ones in that they typically encode semantic representations in virtue of their linguistic properties. However, these semantic representations are usually incomplete - they provide schemas or "blueprints" (Blakemore 1987) for propositions which need to be inferentially enriched and developed through the use of contextual information in order to yield mental representations with a fully propositional form. This process includes such aspects as disambiguation, reference assignment, interpretation of semantically vague expressions like "soon" or "some" and so forth. Again, this process of developing the semantic representation of an utterance into a propositional form is controlled by the criterion of consistency with the principle of relevance.
Thus utterance (1) does not by its semantic properties specify which foreign currency store is being talked about; it is the hearer's task to find the intended referent. He will do this by searching the contextually available information for a referent that is highly accessible in his mind and that will yield an interpretation with adequate contextual effects.6 As soon as he finds a referent that meets these two criteria he will assume it to be the speaker-intended one.
Returning to our concern with translation, let us start from the hypothesis that all instances of human translation can be accounted for as instances of ostensive-inferential communication. As we consider different kinds of translation, we will be testing the validity of this assumption.
3 "Incidental" translation
The first kind of translation we want to look at is one that we encounter very frequently in our daily lives, for example, when we read the English text on the label of a foreign product, the English instructions for a machine manufactured abroad, or an English tourist brochure in a foreign country. Very often these texts are produced on the basis of an original text in the foreign language, hence they are normally considered translations, and dealt with in writings on translation, as, for example, in House (1981), Hönig and Kußmaul (1984), or Picken (1983).
However, it seems doubtful that we need to refer to a theory of translation to account for such cases. Suppose, for example, that your company has produced photocopiers for export to an Eastern African country, and produces an operating manual in, say, Swahili. Now looking at it first from the customers' point of view, what counts for them is that the Swahili manual tells them clearly all they need to know for operating the photocopier. It is completely inconsequential to them whether there was an English original of this manual and whether the Swahili manual faithfully represents the information of that original. In fact, they may need to be given more or different information than the customers in England, perhaps because the conditions under which they use the copier differ from those in England, or because the background knowledge they bring to the machine differs from those of the average customer in England and so forth. Thus the customer's criterion for the quality of the manual will be how well it enables them to operate the copier.
Similarly from the producer's point of view: he wants his customers to be satisfied with the product, and this again requires that the customers know how to handle that product. So the aim of the producer, too, is that the manual provide all the information which that particular group of customers needs to operate the copier appropriately. It is not his primary interest to inform them of what the English manual says. Of course, the producer may find it very convenient to use the English original as a starting point for the Swahili manual, but this fact is incidental rather than essential for the success of the Swahili manual: he could just as well appoint a Swahili-speaking technician to produce a Swahili manual for that copier from scratch, and again the quality of the manual would be judged by how helpful it proved to the customers.
Put in general terms, such instances are characterized by the fact that the receptor language text is produced and presented to the target audience not because it faithfully represents the contents of some source language original, but in its own right; the existence of a source language text in such situations is incidental rather than necessary for the interlingual communication act to succeed.
These cases are clearly instances of ostensive-inferential communication: a communicator wants to communicate certain thoughts to a target audience - the only complication is that the source language communicator does not master the receptor language. Therefore he needs the help of a bilingual person to produce a receptor language stimulus that will communicate his informative intention. In other words, the process of stimulus production is shared between (at least) two individuals, but there is only one stimulus that is significant, and that is the receptor language one.
Seeing that such cases of interlingual communication the source language stimulus plays an auxiliary rather than a central role, one wonders how appropriate it is to refer to them as 'translation' at all.
4 Conveying the original message
If one were to ask around what people think a translation should achieve, the most frequent answer would probably be that it should communicate the meaning of the original. This has not always been so, but since the middle of this century this view has been adopted increasingly by translation theorists. Accordingly, the quality of a translation is now often judged in terms of its comprehensibility and impact on the receptors.
This re-orientation has probably found its fullest development in circles concerned with the translation of the Bible, though it is not limited to this enterprise. The first and probably most influential approach along these lines is that of "dynamic equivalence" translation developed by Nida and Taber (Nida 1964; Nida and Taber 1969). These scholars state clearly that for them the meaning, or "message" of the original takes first priority:
"Translating must aim primarily at 'reproducing the message'. To do anything else is essentially false to one's task as a translator." (Nida and Taber 1969, p. 12)7
This commitment to reproducing the "message" of the original was taken up by others, for example in the "idiomatic approach" of Beekman and Callow (1974) which has been extended by Larson (1984) to cover the translation of non- biblical literature as well.
What do these approaches mean by the "meaning" or "message" of the original? There are no explicit definitions given, but it is clear from what is said that the notions held are very comprehensive; they include both the "explicit" and "implicit" information content of the original, and extend to connotations and other emotional aspects of meaning as well.8 For reasons of space, I shall concentrate here on the idea that a translation should convey the same information as the original.
According to relevance theory, the assumptions the communicator intends to communicate can be conveyed in two different ways: as explicatures or as implicatures. Explicatures are a subset of assumptions that are analytically implied by a text or utterance; more specifically, explicatures are those analytic implications which the communicator intended to communicate.9 Implicatures are a subset of the contextual assumptions and contextual implications of an utterance or text - again that subset which the communicator intended to convey. Both explicatures and implicatures are identified by the audience on the basis of consistency with the principle of relevance.
With this framework in mind, the demand to preserve the information content of the original amounts to the demand that the explicatures and implicatures of the translation should be the same as the explicatures and implicatures of the original.
Straightforward as this demand may sound, there is a rather serious problem here, and this lies in the logical interdependence between explicatures, implicatures, and the potential context - or, more technically, the cognitive environment - in which a text or utterance is processed. This is, of course, one of the most basic characteristics of inferential communication. For example, the statement "There is a police car over there" could be used to communicate rather different ideas on different occasions: in a context where people are looking for help with a broken-down car it may be used to imply, "Let's go there and ask for help"; by contrast, in a context of someone driving a car with only one headlight working it might mean "Let's turn off this road quickly before they see us".
In both cases, the propositional form of the utterance and its explicatures may be the same, indicating that there is a police car at a certain distance in the environment; the implicatures, however, can be very different indeed, depending on what contextual assumptions are accessible in the mutual cognitive environment of speaker and hearer.
One consequence of this is that whenever a given stimulus is interpreted in a potential context that differs in information content from the one envisaged by the original communicator, misunderstandings are likely to arise. Let us use the term secondary communication situation for such instances. Since most translation is done in secondary communication situations, it is not surprising that it has run into difficulties along these very lines.
For example, the Gospel of Mark reports an incident where four men lowered a paralyzed man through an opening in the roof in order to get him to Jesus; in one language it was found that a translation of this passage implied a miracle: "Since no indication was given of how four men, carrying a paralyzed friend, could get onto a roof (and the language helper tended, naturally enough, to think in terms of his own familiar steep thatched roof), the language helper assumed a miracle, ..." (Beekman and Callow op. cit., p. 47) Many such problems have been reported.
Unfortunately, the approaches advocating "same-meaning- translation" have failed to understand the inferential nature of this problem; mistaking it for a language problem rather than one of mismatch in contextual knowledge, they have proposed that the principle of keeping the meaning constant obliges the translator to express himself in such a way that misunderstandings will not arise. In practice this has meant one of two things: either the translator can "explicate" the implicit information needed to arrive at the correct interpretation of the text; thus in the example given it is suggested that he may have to add the information that the men climbed up stairs that led to the roof. That is, he would express in the translation a contextual assumption of the original. Or he can, in certain cases at least, change the meaning expressed in the text. This latter practice is subject to some other constraints and is mostly suggested for the rendering of non-literal uses of speech, such as metaphors or irony.
As it has turned out, these solutions have succeeded only in part: it has not always been possible to prevent misinterpretation by either explication or semantic changes in the text. In the light of relevance theory, this is not surprising because the demand that a translation should convey the same interpretation as the original in secondary communication situations is at variance with one of the most basic requirements of successful communication; this is the requirement that to be communicable an interpretation has to be consistent with the principle of relevance. Since consistency with the principle of relevance is always context-dependent, what this means is that it is not necessarily possible to communicate a given set of assumptions to any audience, regardless of what their context might be.10 Communication is not just a matter of finding the right stimulus for what one wants to say - it crucially involves determining what one can communicate to a particular audience, given their particular background knowledge.
We see, then, that the idea that the basic characteristic of translations is that they convey the meaning of the original to the target audience is too simplistic. It has already been variously questioned in the literature (cf. e.g. Reiss and Vermeer 1984, Frawley 1984), but relevance theory helps us to see why it cannot be upheld as a general condition: it does not meet the requirements for successful communication in secondary communication situations.
The point we have reached in our discussion so far, then, is the following: it seems essential that a translation represents an original text in another language, but the demand that it communicate the meaning of that original to the target audience runs into problems in secondary communication situations. Does relevance theory provide some more appropriate way of characterizing what translation is about?
5 Translation as interpretive use
The obvious place to look for an answer to this question within relevance theory is the notion of interpretive use. As Sperber and Wilson (1986a) have shown, there are two fundamentally distinct ways in which utterances, and representations more generally, can be used:11 they can be used descriptively, that is, as true descriptions of some state of affairs, and they can be used interpretively, and this means they are used in virtue of their resemblance with some other representation.12
As an example, take a book review. Reviewers often start out with summarizing the essential ideas of the book, as the original author presented them, and in many cases these ideas are given in the form of simple affirmative statements rather than being embedded every time in a matrix clause like: "X thinks that ..., X writes that ..., X claims that ..." etc.. This is done even where the reviewer actually disagrees with these claims, as his comments in the evaluative section may show.
In terms of relevance theory, the presentation of the ideas of the book author is an instance of interpretive use: the statements that summarize those ideas are presented because they interpretively resemble the statements of the original author, that is, because they share explicatures and/or implicatures of the original work.
This contrasts with descriptive use, where the statements are presented because the communicator believes them to be true of some state of affairs. Take, for example, a geography teacher giving his class a lesson on China. Though the information he passes on to his class will have come out of books he has read, he is not presenting it to the class as 'something the books say', but as facts that he himself believes to be true. Again, he will not introduce each statement by an explicit claim like, "It is true that ...".
Since the distinction between descriptive and interpretive use need not be marked linguistically, one of the tasks of an audience is to determine whether the utterances made are presented interpretively or descriptively.
The potential importance of this distinction can be seen from an incident that took place recently in West Germany. There Jenninger, then president of the West German parliament, delivered a speech on the 40th anniversary of the founding of the Federal Republic of Germany. In this speech he outlined some of the Nazi thinking, without, however, always marking these statements as quotations. This invited the misunderstanding that he was, in fact, voicing his own opinion - in other words, that these statements were instances of descriptive use, expressing what he himself believed to be true. Not surprisingly, his speech caused considerable uproar, and in the end Jenninger had to resign from his presidency, though not only for this reason.
Thus the distinction between interpretive and descriptive use is an important one. Direct and indirect speech quotations, irony, and many other uses of language all rely on interpretive resemblance: the utterances in question are always presented in virtue of the fact that they interpretively resemble another representation, be it a text or thought. Since translations are also texts presented in virtue of their resemblance with an original, it seems they fall naturally under the category of interpretive use.
From a purely theoretical point of view, the ideal solution for translation theory would be the null hypothesis - that is, that translation simply is interpretive use, the only difference from other instances of interpretive use following from the fact that the original and its report happen to be in two different languages.
Now one important point about interpretive resemblance is that it is not an absolute, but a comparative notion: that is, utterances can interpretively resemble one another to varying degrees, and this will depend on the number of implicatures and/or explicatures they share.
Suppose I had been stopped by a stranger in the street and had talked with him briefly while my friend walked on slowly. After a brief conversation, I would catch up with my friend, and he might ask me what it was all about. Now if it was a reasonably short conversation, I might give my friend a verbatim report. My report would be an instance of interpretive use that involved a very high degree of resemblance, where the report would share virtually all implicatures and explicatures with the original.
On the other hand, I might answer my friend with one very brief statement like, "Oh, he wanted some money." Now as long as this statement shared at least one explicature or implicature with the original, it would interpretively resemble it. It could, for example, be that the stranger had not actually said anything like, "I want money from you." Perhaps he had described in very vivid terms his current financial problems, only implicating that I should help him with some money. In this case my summary statement would still resemble the original in that it shared one of its implicatures.
In other words, interpretive use is a very flexible notion, covering, for example, the verbatim report of a conference session just as much as a ten-line summary of it in a newspaper.
From one point of view, this flexibility seems desirable for a theory of translation - after all, in the course of time the term "translation" has been applied to virtually any kind of speech reporting across languages, including summaries.
However, the fact that it does cover such a wide range of texts may be seen as a disadvantage, in that it would not allow us to account for the common intuition that somehow a "translation" is something different than a "paraphrase" or an "abridgment". It is, of course, possible that this intuition will turn out to be elusive, but it seems worth examining.
The question, then, is: considering that the idea of translation as "interlingual interpretive use" may seem too wide for this purpose - can we narrow down the notion of interpretive resemblance in a way that will make translation clearly distinct from freer forms of interlingual communication?
From a theoretical point of view, the problem is that interpretive use as such seems too lax in the kind or degree of resemblance it demands with the original: the sharing of but one implicature or explicature would be sufficient for a receptor language text to interpretively resemble the original.
What one would like to demand is the sharing of all explicatures and implicatures, but, as we saw above, this is not possible in secondary communication situations, the problem being that utterance interpretation is context- dependent. Is there any other way in which translation could be defined in a more definite way?
6. Translating communicative clues
It will be recalled from section 2 that not the whole meaning of an utterance is context-dependent; we said that in the interpretation process contextual information is used to enrich and develop a semantic representation that is determined by the linguistic properties of the utterance. Would it not, therefore, be possible to set up a theory of translation aimed at reproducing the linguistically determined, semantic properties of the original utterance or text?
Something very close to this has, in fact, been proposed, for example, by Kade (1968) of the "Leipzig School" of translation, and I myself proposed a solution along these lines in Gutt (1987, 1988). However, the problem is that such a definition does not capture all that one would need to take care of in translation; one reason for this is that not all expressions of natural language have a semantic representation in the linguistically specified sense; for example, proper names, greetings like hello, discourse connectives like so or therefore (cf. Blakemore 1987), and onomatopoeia would not be covered.
The same would be true of a number of stylistic features, such as foregrounding and backgrounding, the connotative "meaning" of words like "daddy" as compared to "father", or even the distinction between assertions and yes-no questions: none of these aspects would be covered by such a definition, and yet they would normally be considered important aspects of translation.
The problem is that these other aspects make themselves felt in the contextual implications of the translation - but as we have seen, the involvement of contextual information creates a problem in all secondary communication situations.
However, is it not possible to widen the theoretical base of such a translation theory to include these additional features? Could not, for example, direct quotation serve as a model for a constrained translation theory? As Sperber and Wilson (1986a) have pointed out, in direct quotations one produces another token of the original sentence (p. 227f). Is it not, in some sense, possible to produce in the target language another token of the source language sentence, that is, to produce a target language sentence that has all the intrinsic properties of the original sentence?
It does not take too much to see that such a demand cannot be upheld generally. For example, English has contrastive stress, but other languages, like Amharic, do not. Therefore, the Amharic translation of an English sentence cannot share with it the property of contrastive stress. Thus it is immediately clear that this notion of sharing all the intrinsic properties of the original will not do for a narrow definition of translation - the two stimuli will not be the same precisely because they belong to two different languages.
Yet if what we said just now is right, then the importance of preserving the properties of the original does not lie in their intrinsic value, but in the influence they have on the interpretation of the stimulus. Thus although English and Amharic do not share the property of having contrastive stress, the clue which contrastive stress provides for interpreting the English utterance, that is, that the stressed constituent is foregrounded, can also be provided in Amharic by the syntactic means of clefting. In fact, it seems that one of the remarkable things about languages is that while they do differ in their concrete properties, they resemble each other with regard to the clues they are able to provide for the interpretation of an utterance. Let us refer to such clues as communicative clues.
These considerations open up the possibility of defining translation in terms of the communicative clues shared between the original and the receptor language text; the most stringent condition possible would be that a translation must provide the same communicative clues as the original. In view of its relatedness to direct speech quotation, let us refer to this kind of translation as direct translation.
The virtue of direct translation, like that of direct quotation, would be that it provides for the target audience all the communicative clues needed to arrive at the intended interpretation of the original.
An important question is, of course, what sort of things the notion 'communicative clue' can cover. Gutt (1989) surveys a number of aspects. To take just one example, let us have a look at the opening paragraph of Dickens' Tale of Two Cities.
(4) "It was the best of times, it was the worst of times, it was the age of wisdom, it was the age of foolishness, it was the epoch of belief, it was the epoch of incredulity, it was the season of light, it was the season of darkness, it was the spring of hope, it was the winter of despair, we had everything before us, we had nothing before us, we were all going to heaven, we were all going direct the other way ..."
On this passage Chukovskii comments: "There is an almost poetic cadence in this excerpt. The sound symmetry conveys its ironic tone extremely well" (1984, p. 144). By contrast, he feels that a translation into Russian along the following lines misses these effects:
(5) "It was the best and worst of times, it was the age of wisdom and foolishness, the epoch of unbelief and incredulity, the time of enlightenment and ignorance, the spring of hope and the winter of despair."13
Chukovskii feels that the problem is that "... [the translators] did not catch the author's intonations and thus robbed his words of the dynamism stemming from the rhythm" (op. cit., p. 144). Chukovskii apparently attributes the special effect achieved by the original to such phonological properties as "sound symmetry" and "rhythm".
While it seems unlikely that the "ironic tone" and the "dynamism" here are phonologically conditioned, we can give an explicit account of these effects if we pay attention to the syntactic structures involved, the main difference here being that the translation combines into single, coordinated sentences what were independent pairs of sentences in the original. One of the effects of using such a string of independent sentences is that each can be interpreted as a separate statement, "echoing" perhaps the opinion of a particular group of people.
In fact, such an echoic interpretation seems to have been intended here for two reasons; firstly, it resolves the apparent contradictions between Dickens' statements; secondly, as Sperber and Wilson (1986a) have shown, the thrust of echoic utterances is not only to report what someone thought or said but typically to express an attitude towards it. Here both the exaggerated form of the statements and the fact that each is followed by its exact opposite suggest that Dickens' considered these evaluations ridiculous - hence the note of irony perceived by Chukovskii.
If this is correct, then we can understand why the translation cited does not get the irony across: the coordinated form gives the impression that each pair of evaluations constitutes a single, paradoxical statement, and hence fails to provide an important clue to the intended ironical interpretation.
The survey in Gutt (1989, forthcoming) deals with clues arising from a wide range of properties: semantic representations, syntactic properties, phonetic properties, discourse connectives, formulaic expressions, stylistic properties of words, onomatopoeia and phonetic properties that give rise to poetic effects. The overall result of the survey seems to be that within the framework of relevance theory, the notion of direct translation, defined in terms of shared communicative clues, is helpful and allows explicit treatment of many problems of translation, including the more subtle ones, like poetic effects, that have often been claimed to be beyond the scope of objective analysis.
7 Translation, faithfulness and successful communication
Thus it seems that we have arrived at two possible ways of defining translation: on the one hand there is the comparatively narrow, stimulus-oriented notion of direct translation; on the other there is the much wider interpretive-use notion, which we might want to refer to as indirect translation, in contrast to direct translation.
This state of affairs might be considered acceptable from a practical point of view, but from the theoretical point of view at least two important problems remain: firstly, the notion of "communicative clue", though useful, lacks an explicit definition, and secondly it remains unclear why there should be two such ways of defining translation, rather than three, four or twenty five.
To answer these questions, let us first have another brief look at direct and indirect speech quotations. This time we shall ask what the conditions are under which they can lead to successful communication.
Beginning with indirect quotation, as an instance of interpretive use, any indirect speech quotation creates a presumption of faithfulness; as Sperber and Wilson (1986a) have shown, the speech reporter creates a presumption that the interpretation he intends to convey resembles the interpretation of the original closely enough in relevant respects. This presumption of faithfulness is a derived notion. It follows from the nature of interpretive use on the one hand and the principle of relevance on the other; as an instance of interpretive use, an indirect quotation is used in virtue of its interpretive resemblance with the original; by the principle of relevance it creates a presumption that the interpretation offered will be adequately relevant under optimal processing. Thus we see that relevance theory comes with a ready-made notion of faithfulness, that exists independently of translation.
Hence when a communicator engages in indirect quotation, he will tend to communicate those thoughts of the original interpretation that he believes to be adequately relevant, and he will express himself in such a way that the audience will be able to recover those thoughts in consistency with the principle of relevance. All the audience needs to do is to go ahead with processing: it can expect that by using the contextual assumptions to hand, the first interpretation consistent with the principle of relevance will be the one intended by the speech reporter.14
With direct speech quotation matters seem rather different: the audience cannot simply use the most accessible contextual assumptions to arrive at an authentic interpretation; rather, in order to recover the intended interpretation of the original, it will have to use the contextual assumptions envisaged by the original communicator. This point is not only common sense, but also well-recognized in literary circles; one of the preconditions of authentic literary interpretation is a reconstruction of the historical, cultural and sociological background against which that piece of literature was created.
Correspondingly, one would expect this same principle to be applied to translated works, especially those that aim to follow the original very closely. Strangely, in translation circles the importance of this requirement has not really been understood. Translated works are regularly criticized for failing to convey implicatures that really depend on the availability of the original context.
If the same requirements were made of direct quotations, then someone wanting to quote from Shakespeare should word the quotation in such a way that the audience could interpret it correctly, no matter how different their background might be from that of the original audience.
I want to suggest that this somewhat absurd situation has arisen from an inadequate understanding of the nature of language and communication; more specifically, these assumptions seem to be rooted in the code-model view of language and communication; in that view successful communication of the original message would depend on the proper use of the code (except for "noise" in the channel), and so, if the translation led to misunderstandings, the most likely cause would be a coding mistake on the translator's part.
However, even if the stimulus used is a coded one, in human communication it does not convey an interpretation except by inferential combination with a context. In ostensive communication, there is a causal interrelation between stimulus, context and interpretation, established by the principle of relevance, and I believe that the failure to see this interdependence has been one of the main reasons for the stagnation in the translation debate, if not its main cause.
In fact, a clear recognition of this causal relation opens the way to a coherent, explicit account of translation, and this account will integrate the notion of direct translation into the framework of interpretive use.
Let us approach this solution via the following new definition of direct translation:
(6) Direct translation
A receptor language utterance is a direct translation of a source language utterance if and only if it purports to interpretively resemble the original completely.
In order to see how this definition relates to our earlier notion of direct translation, let us take a closer look at what it entails.
First of all, it defines translation independently of the potential context of the receptors - in fact, it defines it with regard to the context envisaged by the original author. This follows from the fact that the intended interpretation of a given text cannot be arbitrarily communicated to any audience regardless of their cognitive environment, but requires that the target audience process it with regard to the originally envisaged context. This means that the presumption of complete interpretive resemblance can be taken to hold only with regard to the original context, and we just saw that this is no extraordinary requirement, though at variance with a widely accepted view in the field of translation.
This first entailment has two very important effects. From the receptor audience's point of view it means that they can expect to derive an authentic interpretation of the translation only with regard to the original context; in other words, if they want to find out the original interpretation, the onus is on them to familiarize themselves with the cognitive environment of the original. Thus it is possible in principle to communicate the originally intended interpretation by translation, as the approaches discussed in section 4 had thought, but it requires that the translation is processed with regard to the originally envisaged context. In practical terms this means that generally direct translations may need to be interpreted in a very different way from indirect translations, just as direct quotations may need to be interpreted differently from indirect quotations.15
Correspondingly, from the translator's point of view it means that he need not adapt the translated text to avoid misunderstandings likely to arise from contextual differences, because he can work on the assumption that the translation will be interpreted with regard to the original context. In fact, he should not make such adaptations because if processed in the original context, such adaptations would lead to an interpretation different from that of the original. Thus, the presumption of complete interpretive resemblance rules out the explication of implicit information, summarizing and other changes in explicit content.
However, direct translation constrains not only the explicit content - it also determines the other properties of the translated text; again this follows from the causal interdependence of stimulus, context and interpretation: in order to achieve complete interpretive resemblance the translated text will have to convey not only the same explicatures as the original but also its implicatures, and so it will have all the properties needed to convey these implicatures as well.
And here we have the link-up with our earlier definition of direct translation: what we tried to capture intuitively with the notion of "communicative clue" is just this causal aspect of the stimulus, that is, its potential to convey the intended interpretation of the original in the original cognitive environment. Thus we wanted the "communicative clues" to take care of all those properties of the original that affected its interpretation without, however, demanding identity in those properties, and we also wanted them to be independent of the receptor language context. Our new definition of direct translation captures all of these characteristics, and it does so without reliance on the notion of "communicative clue". This means that the notion of "communicative clue" has no theoretical status, though it may well prove helpful to the translator as an auxiliary concept for his practical work. Thus he may want to evaluate his translation by a comparison of communicative clues.
Our new definition allows us also to bring out a very important difference between direct quotation and direct translation that we have not commented on yet. Direct quotations can, in principle at least, be produced without a full understanding of the originally intended interpretation - simply by producing another token of the same sentence type. Thus a child can report an utterance verbatim without a full grasp of what it was intended to convey.
In direct translation this is not possible - the translator cannot produce a target language text that interpretively resembles the original completely without having first recovered the full interpretation of that original himself. This is true even of the communicative-clue-based account of direct translation. The translator cannot determine whether a given property of the original is a "communicative clue" without knowing what, if any, effect it was intended to have on the original interpretation; this, in turn, means that he has to first find out what the originally intended interpretation was. To return to Chukovskii's translation example: it seems that the translator had missed the echoic nature of the intended interpretation, therefore he failed to recognize the communicative clue provided by the syntactic structure of the original, and so he bungled the translation.
Thus we see that our relevance theory account explains the common demand in translation that translation presupposes a good grasp of the intended interpretation of the original. It is not difficult to see that this demand applies also to indirect translation - as we said above, indirect translation is based on the notion of interpretive resemblance, hence also presupposes a grasp of the originally intended interpretation.
And now we can see that indirect and direct translation are not as different as they looked at first: they both turn out to be instances of interpretive use; in other words, the notion of interpretive use provides a unified account for both direct and indirect translation. The essential difference between them is that direct translation is committed to complete interpretive resemblance, whereas indirect translation presumes only adequate resemblance in relevant respects.
It would seem to me that the recognition that translation is dependent on interpretive resemblance has far-reaching consequences for people involved in machine translation who work largely on the basis of transcoding. If relevance theory is right, then the progress toward fully adequate translation will require programs that can derive and compare interpretations of texts, which presupposes, among other things, that they can handle considerations of relevance.
As to the question why there should be just these two explicit notions of translation, the answer would seem to follow straightforwardly from the framework of relevance theory: as Sperber and Wilson (1986a) have pointed out, interpretive resemblance covers the full range from no resemblance at all to complete resemblance; however, while there is no principled cut off point at the lower end, the upper limit is clearly definable in principle: that is, as complete resemblance. Hence it is not surprising that there should be two distinct notions of translation, corresponding to the general notion of interpretive resemblance and its limiting case respectively.
One problem that we raised above, but have not really answered concerns the apparent vagueness of the notion of indirect translation: how can one work with a notion of translation that comprises anything from a summary statement to an expanded presentation of the original interpretation? Should one not break up this continuum by setting up appropriate subtypes?
To this question two answers can be given; firstly, while one can no doubt try to distinguish different types of indirect translation, such a typology would always be arbitrary, because there is no non-arbitrary way of breaking up a continuum.
Secondly, while such a typology may be interesting from a descriptive point of view, it is unnecessary for ensuring communicative success by indirect translation. Indirect translation, like any other instance of interpretive use, comes with a presumption of faithfulness: the translator presents his translation on the presumption that its interpretation adequately resembles the original in respects relevant to the target audience. The communication act succeeds where the translation lives up to this presumption, and it turns out to be inadequate where it falls short of the presumption.
What the translator has to do in order to communicate successfully, is to arrive at the intended interpretation of the original, and then determine in what respects his translation should interpretively resemble the original in order to be consistent with the principle of relevance for his target audience with its particular cognitive environment. Nothing else is needed.
Thus our relevance-theoretic account provides a principled answer to problems like the "Guiseley sandstone" example we discussed in the introduction above: the answer whether or not the translator will include the geographical reference or not is determined by considerations of relevance with regard to the particular context which the audience brings to the translated text.
In fact, it seems that the various rules, principles, and guidelines that have been proposed for translations of different sorts are all applications of the principle of relevance. Consider e.g. the following overview given by Newmark:
"A technical translator has no right to create neologisms..., whilst an advertiser or propaganda writer can use any linguistic resources he requires. Conventional metaphors and sayings ... should always be conventionally translated (...) but unusual metaphors and comparisons should be reduced to their sense if the text has a mainly informative function.... The appropriate equivalents for keywords... should be scrupulously repeated throughout a text in a philosophical text. ... In a non-literary text, there is a case for transcribing as well as translating any key-word of linguistic significance, ..." (1988, p. 15)
It is not difficult to see that each of these rules is an application of the principle of relevance, spelling out what aspects of the original the translator should preserve for a particular audience in order that his translation adequately resemble the original in respects relevant to them. Perhaps this is one of the most important features of the solution proposed here: the notion of translation simply as "interlingual interpretive use" makes clear predictions about what translation would be appropriate in any given situation, without reliance on typological props.16
8 On the limits of direct translation
Now from a theoretical point of view it may seem fine to define direct translation in absolute terms - but what about the "messy" reality of natural languages? Can we assume that direct translation is generally achievable, that is, that it can be achieved for just any text or utterance between any pair of languages? If not, of what use is this notion?
This question brings us to the issue of translatability - which in turn would lead on to the question of effability; we cannot go into this here because it would seem to require a dissertation of its own. Personally I believe that Sperber and Wilson (1986a, pp. 191f) are right in arguing that effability in the strong sense does not exist, and I think there are good reasons to assume that translatability does not generally exist either, at least not in the strong sense entailed by direct translation.
However, it would seem to me that little depends on the answer to these questions as far as our account is concerned, because our definition relies on a presumption - not a guarantee of success. As Sperber and Wilson (1986a) have pointed out, "the principle of relevance does not say that communicators necessarily produce optimally relevant stimuli" (p. 158) - in other words, it does not guarantee the success of an act of communication; however, it does lay down the conditions for successful communication.
By the same token, the presumption of complete resemblance in direct translation does not guarantee its success - but lays down the conditions for its success. Put in concrete terms, it specifies that a direct translation will be successful if and only if it conveys the interpretation of the original when interpreted with regard to the original context. To the degree that it does not, it will have fallen short of its presumption, and risk misinterpretation.
In this way, our definition of direct translation provides the frame of reference for its own evaluation, and at the same time it spells out clearly the risks involved in direct translation: the presumption of complete interpretive resemblance entitles the receptors to maximal assumptions about resemblance, hence they will be likely to draw inferences from all sorts of stylistic and other details of the translated text. At the same time, language differences may make it impossible to achieve complete interpretive resemblance - and hence our account predicts that in such instances some of the inferences of the receptors will be mistaken, and that without knowledge of the source language the receptors will not be able to spot such misinterpretation, unless the translator alerts the receptors to such problems.
This prediction seems to capture exactly what happens in practice: to the extent that linguistic differences between receptor language and source language make complete interpretive resemblance impossible, the interpretations of translations will always differ from the original interpretation, even if the receptors have taken the greatest care to familiarize themselves with the historical, cultural, etc. context of the original, and hence the receptors generally need to remember that a translation is not an original, even in direct translation.17
It should be noted that within the framework of ostensive- inferential communication, the frame of reference extends also to the manner of expression, presuming that the stimulus used is the most economical one to get the intended interpretation, that is, that of the original, across. This would account for certain intuitions about "unnaturalness" or "literalism"; for example, the use of unusual or even ungrammatical syntactic structures tends to make the receptor language stimulus more costly to process; if these complications were not outweighed by an increase in relevance with regard to the intended interpretation, they would make the stimulus less than optimally relevant.
Similarly, where the translator cannot preserve all the explicatures and implicatures but has to select, consistency with the principle of relevance would require that he give priority to a rendering that will achieve an optimum of relevance. Thus even in situations where full success is not possible due to language differences, our account makes predictions about the optimal translation.
In fact, there is no a priori reason why a translator should follow, for example, the direct translation approach consistently throughout a text. Thus he may aim for complete resemblance in some parts of the text, but be less ambitious in others. What he needs to consider all the time, though, is that, whatever he does, it will have affect the success or failure of his translation - this follows from the causal interdependence of cognitive environment, stimulus and interpretation.
This means that the account of translation provided here is not normative - it does not tell the translator what to do. Neither is it descriptive - it is not interested in how different kinds of translation can be characterized. Rather it is meant to be explanatory: it aims at explaining how people can communicate via translation, and what the conditions for communicative success are. A better understanding of these conditions will help the translator to be more successful in his task; it will help him to anticipate potential misunderstandings, and to take measures to counteract them effectively.
In Gutt (1989, forthcoming) a number of such measures are discussed. Suffice it here to mention just one obvious means the translator can use to work for communicative success; this is to inform his audience clearly of what he is trying to achieve. It seems that much of the criticism leveled against translations stumbles on this very point: the intentions of the translator do not meet the expectations of the target audience, and so miscommunication results. Thus rather than relying on the label "translation" somewhere in the front of the book - a label that has no generally agreed content - the translator can increase the prospects for communicating successfully if he takes care to explain to his audience what he is trying to achieve.
In conclusion, we see that relevance theory enables us to provide what translation theorists have been looking for - an explicit framework for accounting for the phenomena commonly subsumed under the term 'translation'. We saw that it covers 'incidental' translation, that is instances of translation where the existence of a source language original is not essential to the communication process. The other instances are covered by relevance theory as two clearly distinct instantiations of interpretive use: indirect translation is simply interpretive use between stimuli from two languages; direct translation, on the other hand, is the special case of interpretive use that creates a presumption of complete interpretive resemblance between stimuli from two languages. Placed in a historical perspective, these two notions could perhaps be seen as the spelling out of the century-old intuition that there is a dichotomy between "literal" and "free" translation.
1 This article is based on my doctoral dissertation (Gutt 1989), which is to be published as Gutt (forthcoming).
2 Sperber and Wilson (1986a) contains the fullest presentation of relevance theory as a whole; specific issues have been dealt with in Sperber and Wilson (1986b), Wilson and Sperber (1988a) and (1988b). For a brief introduction, peer comments and a reply by the authors see Sperber and Wilson (1987).
3 More precisely, relevance theory is concerned with ostensive communication, where ostensive behaviour is defined as "behaviour which makes manifest an intention to make something manifest" (Sperber and Wilson 1986a, p. 49)
4 Of course, (1) is not actually irrelevant - the subsequent explanation provides a context in which it does achieve contextual effects as an example. This illustrates another important characteristic of human communication: the relevance of an utterance, hence its intended interpretation, need not be immediately obvious upon first encounter, but may be recovered with the help of subsequent information.
5 Contextual effects can be achieved in three different ways: the inferential combination of the information expressed in the utterance and of some previously known information can yield an implication not obtainable from the utterance alone nor from the previously known information alone, but only by the combination of the two; such implications are called contextual implications, like the one in the example just given. Secondly, the inferential combination of utterance and previous knowledge can also lead to the cancellation of information previously believed - in other words, a previously held belief gets corrected. Thirdly, the inferential combination can strengthen a previously held belief, so that one is more certain of its being true. For more details see Sperber and Wilson (1986a).
6 More precisely, he will look for a referent that the communicator could reasonably have believed to be highly accessible in his mind, and to yield an interpretation with adequate contextual effects.
7 In fact, Nida and Taber demanded that equivalence should not be restricted to the information content alone, but also to the "dynamics" of the texts, and these dynamics were to be measured in terms of audience response. It is this notion that gave the approach its name - "dynamic equivalence translation".
8 Nida and Taber gloss the term "message" as follows: "Message: the total meaning or content of a discourse; the concepts and feelings which the author intends the reader to understand and perceive." (op. cit., p. 205)
9 Analytic implications are characterized by the fact that they have been derived by analytic inference rules only; an analytic inference rule is formally defined as a rule that "takes only a single assumption as input" (Sperber and Wilson 1986a, p. 104). The analytical implications of an utterance are determined by its propositional form. In practical terms, "the analytic implications of a set of assumptions are those that are necessary and sufficient for understanding it, for grasping its content." (Sperber and Wilson 1986a, p. 105)
10 For a more extensive discussion and illustrations of this point see Gutt (1989) or forthcoming.
11 These two different usages are assumed to reflect two different ways in which our minds entertain representation.
12 More correctly, this applies to representations with logical properties. See Sperber and Wilson 1986a for more information.
13 Bobrov, S.P. and M.P. Bogoslovskaja, Povest' o dvukh gorodakh, Sobranie sochinenii, 1957-63; vol XXII, p. 6, as cited in Chukovskii (1984, p. 144)
14 Note that in interpretive use the contextual assumptions available to the audience may well include assumptions about the original representation.
15 How different the processing of direct translations will be will depend on how different the cognitive environment of the target audience is from that of the originally envisaged audience.
16 This is not to say that translation rules that make e.g. text-typological generalisations cannot be helpful, especially for the training of translators. However, what must be clearly borne in mind is that such rules do not have a value of their own but are valuable only in so far as they are valid applications of the principle of relevance. In other words, any such rule may need to be set aside if consistency with the principle of relevance for a particular audience requires this.
17 It might seem more realistic to re-define direct translation as presuming maximal rather than complete interpretive resemblance; however, the notion of "maximal" interpretive resemblance is undefined, and such a redefinition would seem to obscure the very point just made: that misinterpretations are likely to arise in direct translation wherever linguistic differences make complete interpretive resemblance impossible.
Beekman, J. and J. Callow (1974) Translating the Word of God, vol. 1, Zondervan, Grand Rapids, Mich.
Bassnett-McGuire, Susan (1980) Translation Studies, Methuen, London.
Blakemore, Diane (1987) Semantic constraints on relevance, Blackwell, Oxford.
Chukovskii, K. (1984) The Art of Translation, (trl. & ed. Lauren G. Leighton), University of Texas Press, Knoxville.
Frawley, William (1984), 'Prolegomenon to a theory of translation'. In W. Frawley (ed.), Translation: Literary, linguistic and philosophical perspectives. Associated University Press, London, pp. 159-175.
Gutt, Ernst-August (1987a), 'What is the meaning we translate?' Occasional Papers in Translation and Text- linguistics no. 1, January 1987, pp. 31-58.
Gutt, Ernst-August (1988), 'From translation to effective communication'. Notes on Translation vol. 2, no. 1. pp. 24-40.
Gutt, Ernst-August (1989), Translation and relevance. University College London doctoral dissertation.
Gutt, Ernst-August (forthcoming), Translation and relevance. Blackwell, Oxford.
Hönig, Hans G. and Paul Kußmaul (1984), Strategie der Übersetzung. Narr, Tübingen.
House, Juliane (1981), A model for translation quality assessment. Narr, Tübingen.
Kade, O. (1968) Zufall und Gesetzmäßigkeit in der Übersetzung, VEB Verlag Enzyklopädie, Leipzig.
Krings, H.P. (1986) Was in den Köpfen von Übersetzern vorgeht, Gunter Narr Verlag, Tübingen.
Larson, Mildred L. (1984) Meaning-based translation: A guide to cross-language equivalence. University Press of America, New York.
Levy, Jiri (1969), Die literarische Übersetzung: Theorie einer Kunstgattung. Athenäum, Frankfurt.
Lyons, John (1969), Introduction to theoretical linguistics. Cambridge University Press, Cambridge.
Newmark, P. (1988) Approaches to Translation, Prentice Hall, Hemel Hempstead.
Nida, E.A. (1964) Toward a Science of Translating. With Special Reference to Principles and Procedures Involved in Bible Translating, E.J. Brill, Leiden.
Nida, E.A. and C. Taber (1969) The Theory and Practice of Translation, E.J. Brill, Leiden.
Picken, Catriona (1983), The translator's handbook. Aslib, London.
Reiss, Katharina and Hans J. Vermeer (1984), Grundlegung einer allgemeinen Übersetzungstheorie. Niemeyer, Tübingen.
Schulte, R. (1987) "Translation Theory: A Challenge for the Future", Translation Review, no. 23, (Special Theory Issue), 1987.
Snell-Hornby, Mary (1988) Translation Studies: An integrated approach, John Benjamins, Amsterdam.
Sperber, Dan and Deirdre Wilson (1986a) Relevance: Communication and Cognition, Blackwell, Oxford.
Sperber, Dan and Deirdre Wilson (1986b), 'Loose talk'. Proceedings of the Aristotelian Society 1985/6, vol. 86, pp. 153-171.
Sperber, Dan and Deirdre Wilson (1987), 'Précis of Relevance: Communication and Cognition'. Behavioural And Brain Sciences vol. 10, pp. 697-754.
Steiner, George (1975) After Babel. Aspects of Language and Translation, Oxford University Press, Oxford.
Wilson, Deirdre and Dan Sperber (1988a) "Representation and Relevance" in R.M. Kempson (ed), Mental Representations: The Interface between Language and Reality, Cambridge University Press, Cambridge, pp. 133-153.
Wilson, Deirdre and Dan Sperber (1988b), 'Mood and the analysis of non-declarative sentences'. In J. Dancy, J. Moravcsik and C. C. W. Taylor (eds), Human agency: language, duty and value. Stanford University Press, Stanford, pp. 77-101.
Wilss, W. (1982) The Science of Translation. Problems and Methods, Gunter Narr Verlag, Tübingen.