Andrew Chesterman


2000b.     Translation typology. In A. Veisbergs and I. Zauberga (eds), The Second Riga Symposium on Pragmatic Aspects of Translation. Riga: University of Latvia, 49-62.






One goal of an empirical translation theory is to establish a comprehensive  d interpretingh translation typology. There are many kinds of translation, some more typical and widespread than others, and we need a system of classification both for theoretical and practical purposes.

       In the first place, theoretical research needs a typology because we need to be able to structure our concept of translation in some way or other; and because different kinds of generalizations apply to different kinds of translation. Much traditional work in translation research seems inadequate nowadays because it made unjustifiable generalizations from a particular type of translation (such as literary or biblical translation) to translation universally. Indeed, for some scholars, theories of literary translation are evidently assumed without further question to be theories of all translation (compare e.g. the title and contents of Schulte and Biguenet (eds, 1992), Theories of Translation).

       In the second place, clients and translators need a typology in order to specify the translation brief. You do not go into a furniture shop and ask the salesman to sell you “a chair” — you specify the kind of chair you want. Both clients and translators are naive if they assume that they can speak of “a translation” without specifying what kind of translation they have in mind. Of course, professional translators often assume that they “know” what kind of translation is wanted, and clients assume that translators know this; but this assumption is not always correct, and seldom made explicit. Clients and professional translators might both benefit from a mutually agreed typology.

       In what follows, I will first outline some of the earlier work on translation typology and then make a more detailed proposal of my own.


Previous approaches


       Most traditional thinking about translation typology has been binary: two main types are set up, mostly as opposite ends of a continuum. The most common parameter has been “free vs. literal”, or “word-for-word vs. sense-for-sense”. A modern version of this distinction is the one proposed by Newmark (1981) between semantic and communicative translation. Semantic translation is closer, more literal; it gives highest priority to the meaning and form of the original, and is appropriate to translations of source texts that have high status, such as religious texts, legal texts, literature, perhaps ministerial speeches. Communicative translation is freer, and gives priority to the effectiveness of the message to be communicated. It focuses on factors such as readability and naturalness, and is appropriate to translations of “pragmatic” texts where the actual form of the original is not closely bound to its intended meaning. These are texts like advertisements, tourist brochures, product descriptions and instructions, manuals.

       A major problem with this kind of distinction is how to measure the degree of literalness, closeness, or distance, freedom. One solution has been to analyse and count the various kinds of changes (shifts, strategies) that have taken place from source to target text. For one of the most detailed and successful proposals of this kind, see Leuven-Zwart (1989/90).

       A slightly different kind of binary typology was proposed by Juliane House (1977): covert vs. overt translations. Covert translations are those that are intended not to be recognized by target readers as translations. In other words, they are so natural target language (and probably therefore fairly free translations) that they do not seem distinguishable from non-translated texts of the same kind in the target language. Examples include advertisements, technical texts, newspaper texts. Overt translations, on the other hand, are obviously translations, and intended to be recognized as such, because they are more closely linked with the source culture. Examples are translations of political speeches, poems, sermons.

       Corpus studies have shown that covert translations may contain linguistic features that have statistically different distributions as compared to non-translated, parallel texts (see e.g. Laviosa 1997). Even covert translations therefore seem to be textually different from non-translations, which suggests that they may be some universal features of translated texts.

       A similar distinction has been made by Nord (e.g. 1997), who sets up an opposition between documentary and instrumental translation. A documentary translation is manifestly a document of another text, it is overtly a translation of something else. Insofar as it presents itself as a report of another communication, it is a bit like reported speech. Instrumental translation, on the other hand, functions as an instrument of communication in its own right, it works independently of a source text, and is judged on how well it expresses its message. So instrumental translation is a bit like direct speech. A translation of a computer manual, for instance, is normally instrumental: the point of the translation is to make sure that the reader understands how to install and use the computer; the point is not to produce a maximally accurate representation of the original text.

       The typological problem becomes more complex when text types are introduced. Reiss and Vermeer (1984) argued that the translation method depended on the text type concerned as well as on the purpose of the translation. Reiss proposed four basic types, the first three being very traditional: informative texts, expressive texts, operative (i.e. persuasive, instructive) texts, and audio-visual (multi-medial) texts. Dubbing and subtitling, for instance, are clearly special types of audio-visual translation. However, we need to be careful not to confuse classifications of text types as such with classifications of translation types, for there is quite a lot of terminological overlap. Labels such as “biblical translation”, “literary translation” or “poetry translation”, for instance, really seem to be referring to text types — the text type that is being translated.

       A different approach is taken by Folkart (1989), whose central criterion is that of reversibility: that is, the extent to which back-translation leads to a text that is the same as the original. She proposes four main types of translation, but she is really talking about text types. The first, most reversible type she calls mathematical texts. These are so highly dependent on particular fixed expressions, for example describing elements of an equation or a formula, that translation is highly predictable and back-translation works well. Type two is technical texts, which are also fairly formulaic. Type three is “constrained texts”, i.e. domain-specific texts such as legal documents, or notices like “Wet paint!” which have well-established, fixed translations. And type four covers all other texts, general and literary, where predictability and reversibility are lowest. What we have here is of course a continuum — as with the other distinctions discussed above.

       A wider set of criteria is proposed by Sager (e.g.1993, 1997). In his latest contribution (1998) he has six: the existence (or not) of situational antecedents in the target culture; the familiarity of the target language document type in the target culture; the purpose of the translation (same as or different from the purpose of the original); the relative status of the source and target texts; the awareness (or not) by the reader that the target text is a translation; and the existence (or not) of standardized translation solutions from previously translated texts. On the basis of these criteria, he ends up with three major translation types: Bible translation, literary translation, and non-literary (technical etc.) translation. Here again, despite Sager’s criteria, the resulting classification seems actually to be one of text types.

       Classifications of translation types obviously depend on what criteria you use. The proposal I outline below borrows from several of the contributions discussed above, and also from Gouadec (1990) and Ibrahim (1994). It is more complex than any of these, but, as we shall see, can be simplified. The proposal is based on Chesterman (1998).


A new proposal


I distinguish first between four sets of variables, A-D:


       A) Equivalence variables (having to do with the relation between source text and target text)

       B) Target-language variables (having to do with the style of the target text)

       C) Translator variables

       D) Special situational variables


These variables are ways in which translations can vary, parameters along which clients and translators can make choices. I will now look at each set in turn, outlining the main variables in each case. There may well be other variables that might be included in a more refined analysis, but I suggest that these are the main ones. (The rest of this section, together with the following one on default values, is taken almost verbatim from Chesterman 1998: 205-209.)


A) Equivalence variables


       A1) Function: same or different? — Is the main function of the target text intended to be “the same” as that of the source text, or not? If not, what? (Different function leads to an adaptation of some kind.)

       A2) Content: all, selected, reduced or added, or some combination of these? — Does the translation represent all the source content, or select particular parts of it (keyword translation) or reduce the content overall (summary translation, gist translation; subtitling), or add some elements such as explanations (exegetic translation)?

       A3) Form: what are the formal equivalence priorities, what formal elements of the source text are preserved? — The main ones are text-type (“same” or different? Different genre, e.g. verse to prose, sonnet to lyric?); text structure; sentence divisions (full-stops preserved; a common interpretation of what is meant by literal translation); word/morpheme structure (gloss translation, linguistic translation); other (e.g. sounds — phonemic translation, transliteration, transcription; or lip-movements — dubbing).

       A4) Style: evidently intended to be “same” or different? — If different, in what way (another sense of adaptation)?

       A5) Source-text revision for error correction: evident or not (implicit or explicit)? Minimal or major? — Has the translator “edited” the source text during translation, corrected factual errors, improved awkward style and communication quality, or is the source reproduced without corrections or improvements? This is the “cleaning-up transediting” mentioned by Stetting (1989). (For cultural transediting, see under B2.)

       A6) Status: is the status of the target text, with respect to the status of the source text, autonomous, equal, parallel or derived? (Sager 1993: 180.) — This status is autonomous if the source text had only provisional status, such as a draft letter or notes; equal if both texts are functionally and legally equal, such as legislation in bilingual countries, official EU texts; parallel if the translation appears alongside the source text and is functionally parallel to it, e.g. in multilingual product descriptions (incidental translation); derived in other cases. To these status categories we might add one that we could call subordinate, referring to cases where the source text is co-present, as in gloss or interlinear translation, but the target text is not functionally parallel. Yet another aspect of status, occurring together with any of the above-mentioned ones, is whether the source text actually used in the translation is the original text (direct translation) or some intermediary version in a third language (indirect translation); in the latter case, the status of the target text might be said to be once-removed (or even twice-removed, etc.).


B) Target-language variables


       B1) Acceptability. — A small number of subtypes can be distinguished here.

              (i) Good native style: fluent and readable, may involve editing (communicative translation).

              (ii) 100% native style: no signs of translationese, conforms to target text-type norms (covert translation).

              (iii) Deliberately marked, resistant to target stylistic norms (foreignized translation).

              (iv) Grammatical: grammatically faultless but clearly a translation, features of translationese (overt translation, whether by intention or not).

              (v) Intelligible: comprehensible, but with grammatical and stylistic weaknesses. Usually not publishable without native revision.

              (vi) Machine translation (with or without postediting).

              (vii) Unintelligible.

(Some of these subtypes thus require a competent native speaker of the target language.)

       B2) Localized or not? — Is the translation adapted to local cultural norms (localized translation, yet another sense of adaptation)? Stylistic norms such as British or American English also come in here.

       B3) Matched or not? — Is the translation matched with a defined set of previous texts, e.g. those produced by the client’s company, to conform to client-specific norms (e.g. via the use of a translation memory system)? (EU “hybrid translations”, for instance, or translations that have to be standardized to a particular format.) An extreme form of literary translation might even seek to match the style of a particular individual writer (parody translation).


C) Translator variables


       C1) Visibility. — Is the translator visible, e.g. in footnotes, a commentary or preface, via inserted terms from the source text in brackets, via evidence of the translator’s own particular ideology (learned translation, philological translation, commentary translation, thick translation; feminist translation, polemical translation)?

       C2) Individual or team? — Are there indications suggesting that the text was translated by more than one translator?

       C3) Native speaker of target or source language, or neither ( — inverse translation if the translator is a native speaker of the source language)?

       C4)  Professional or amateur? This is obviously a complex continuum, not a simple binary difference. At the professional end we expect to find, for instance, evidence of adequate world and domain knowledge, adequate background documentation, adequate technical equipment, adequate knowledge of intended readership, etc. Are there indications of non-professional translatorial behaviour, such as carelessness?


D) Special situational variables


The number of situational variables is virtually infinite, and many (such as client helpfulness, actual availability of documentation...) may leave no visible traces in the translation. Here are three main ones:—


       D1) Space: constraints of layout, screen space, speech bubbles, total pages...

       D2) Medium: same (written or spoken) as source text, or not? (E.g. sight translation, from written to oral.) Also: use or presence of other semiotic systems, other media, diagrams... (screen translation, dubbing, Gouadec’s (1990) diagrammatic translation...).

       D3) Time: are there indications suggesting that the translation had to be done in an unusual hurry? A careless translation might (rightly or wrongly) give such an impression, for instance.


Default values


Mathematically, the total number of possible combinations of these variables is enormous. Fortunately, by no means all of them are equally likely to occur, and not all the configurations are even logically possible. We might start reducing the number of possible and reasonably likely types to a more manageable size, first, by specifying default values for each of the variables, for example as follows (where “same” is of course understood to have quotes around it):


       A1:  Same function

       A2:  All content

       A3:  Same text-type and structure

       A4:  Same style

       A5:  Minimal implicit source-text revision

       A6:  Derived status, direct translation

       B1:   Good native

       B2:   Not localized

       B3:   Not matched

       C1:   Invisible translator

       C2:   Individual translator

       C3:   Target language native

       C4:   Professional

       D1:  No special space constraints

       D2:  Same, written, medium

       D3:  Adequate time...


This set of default values would thus correspond to what we might call the default prototype concept of a translation, in the minds of most clients or readers of translations. It purports to represent the “folk view” of what a typical translation is. (Note: the view of professional translators, on the other hand, might well be rather different with respect to some of these variables... Such possible differences would in themselves be worth researching.) Strictly speaking, of course, these default values actually represent no more than my hypothesis about what such a folk view might be, and as such they are open to testing and possible refutation or refinement.




As was suggested at the beginning of this article, a translation typology has both theoretical and practical use. If the general goal of translation theory is to understand and explain translation, we need to be able to propose and test hypotheses about the causes and effects of particular features of translations and types of translations. At first, we may have to limit such hypotheses to individual cases, as follows (1-2):


       (1) I hypothesize that the occurrence of this particular feature in this translation was caused by such-and-such.

       (2) I hypothesize that this particular feature in this translation will have such-and-such an effect on the readers (for instance: typical readers will not like it).


The next stage is to generalize from individual cases to types of translation, and for this we need a typology. So we will find hypotheses like this (3-5):


       (3) I hypothesize that this feature, which I find in this kind of translation, has that cause / will have that effect.

       (4) I hypothesize that conditions of this kind (including source text type) will lead to translations of such-and-such a type.

       (5) I hypothesize that the effect of a given translation type is caused by such-and-such formal features in the target texts of this kind.


Comparisons between type-bound hypotheses — if they are corroborated by tests — may then allow us to develop hypotheses which account for translations more generally, perhaps even universally, so that we can propose general translation laws (as in Toury 1995). Conversely, we might find that hypotheses which have been thought to apply to all translations actually concern only particular kinds of translations.

       As also suggested earlier, some version of a typology — probably a simplified one — would be useful for negotiating with clients about the kind of translation they want in a particular case. Clients might be asked whether they wanted the default values or not, and if not, what? Frequently occurring combinations of values could be defined as separate translation types, perhaps varying with particular clients, to simplify the initial negotiation: do you want translation type a, b, c, or d?

       For instance, one translation type might be called “Source-edited”; this kind of translation would involve major source-text revision (A5 in the set of variables above), which might affect the default values of other variables. Another type might be called “Absolute translation” (Gouadec’s term); this would explicitly exclude any kind of source-text revision (A5), so that all errors etc. were preserved, perhaps marked with a “[sic]”. Yet another common type would be “Summary translation”. This would have default values for A1, A6, B, C and  D; reduced content for A2, and A3-5 as affected by A2.

       One type might need more time to produce than another, such as an extra revision process or the use of a client-specific terminology bank. It might therefore be more expensive — you get what you pay for. Clients might be sent typical types and definitions in advance, for future reference. This way of gradually increasing the client’s awareness of the complexity of the translation process, an awareness that not all translations are of the same kind, would contribute both to client education and to the social visibility of the translator’s profession.





Chesterman, Andrew 1998. “Causes, Translations, Effects”. Target 10, 2, 201-230.

Folkart, Barbara 1989. “Translation and the Arrow of Time”. TTR 2, 1, 19-50.

Gouadec, Daniel 1990. “Traduction Signalétique”. Meta 35:2, 332-341.

Ibrahim, Hasnah 1994. “Translation Assessment: a Case for a Spectral Model”. Cay Dollerup and Annette Lindegaard, eds. Teaching Translation and Interpreting 2. Amsterdam: Benjamins, 151-156.

House, Juliane 1977. A Model for Translation Quality Assessment. Tübingen: Narr.

Laviosa, Sara 1997. “How Comparable Can ‘Comparable Corpora’ Be?” Target 9, 2, 289-319.

Leuven-Zwart, Kitty van 1989/90. “Translation and Original: Similarities and Dissimilarities” I and II. Target 1, 2, 151-181 and 2, 1, 69-75.

Newmark, Peter 1981. Approaches to Translation. Oxford: Pergamon Press.

Nord, Christiane 1997. Translating as a Purposeful Activity. Manchester: St. Jerome Publishing.

Reiss, Katherine and Vermeer, Hans J. 1984. Grundlegung einer Allgemeinen Translationstheorie. Tübingen: Niemeyer.

Sager, Juan C. 1993. Language Engineering and Translation. Consequences of automation. Amsterdam: Benjamins.

Sager, Juan C. 1997. “Text Types and Translation”. Anna Trosberg, ed. Text Typology and Translation. Amsterdam: Benjamins, 25-41.

Sager, Juan 1998. “What Distinguishes Major Types of Translation?” The Translator 4, 1, 69-89.

Stetting, Karen 1989. “Transediting — a New Term for coping with a Grey Area between Editing and Translating”. Graham  Caie et al., eds. Proceedings from the Fourth Nordic Conference for English Studies. Copenhagen: Department of English, University of Copenhagen, 371-382.

Toury, Gideon 1995. Descriptive Translation Studies and Beyond. Amsterdam: Benjamins.

Schulte, Rainer and Biguenet, John 1992. Theories of Translation. An anthology of essays from Dryden to Derrida. Chicago: University of Chicago Press.





Back to [Publications list]                     Back to [Homepage]