Formulating Strategies for the Translator

nyone who combines commercial translation activities with working as an academic, involved in both translation studies (TS) research and training students to become professional translators, will be acutely aware of how problematic it is to formulate translation strategies for the benefit of translators. If you are in such a position, you are used to feeling equally uncomfortable about what you hear at times on the subject of strategies at academic conferences and professional translators' conferences. You are familiar with dismissive remarks from practitioners who reject wholesale the notion of any possible contribution of TS to the concrete resolution of problems at the 'wordface,' to borrow Wagner's term (Chesterman and Wagner 2002). Cross, for instance, deems Baker's Encyclopedia of Translation Studies to be "interesting, but irrelevant" (1998:27, quoted by Wagner in Chesterman and Wagner 2002:1). The contention is usually that every problem/text/translation situation is different, which means that generalizations, and therefore reusable strategies, are out of the question. It is also argued that the concepts TS relies upon are so complex, esoteric and remote from translation practice that translators could not even begin to apply them to their tasks (for further arguments, see Robinson 2001:162-163). As Wagner points out: "There can be few professions with such a yawning gap between theory and practice" (Chesterman and Wagner 2002:1), a worrying situation for a group of professionals who have a status and credibility problem, since they can still practise without any formal qualification.

The ability to deliver strategies to practitioners lies at the heart of the tension within TS itself between the so-called 'pure' and 'applied' branches of the discipline, with many theorists going out of their way to distance themselves from any form of prescriptivism. Chesterman points out how, for several decades, mainstream translation theorists have taken the view that they "should seek to be descriptive, to describe, explain and understand what translators do actually do, not stipulate what they ought to do" (Chesterman and Wagner 2002:2); they "see themselves as studying the translators, not instructing them" (id.). His outline of the current goals of translation theory leaves no room for prescriptivism:

Theory seems to end where applied considerations start, with the possible implication that applied TS has no theoretical component: "Applied research, or translator training, naturally focuses on what translators should do, on what translations should be like, prescriptively; but this is not the task of translation theory itself" (id., p. 52).

The formulation of translation strategies also bears on the relationship between TS and related disciplines: what is the place of the conceptual tools and metalanguage borrowed from linguistics, for instance, when it comes to providing translators with strategies?

Against this background, the purpose of the present article is:

(a) to explore the nature of possible strategies aimed at the translator for the translation from English into French of a problematic linguistic feature, namely emphasis;

(b) at a more general level, and using emphasis as the main source of illustration for the discussion, to reflect on how translation strategies aimed at translators might usefully be formulated in order to meet their practical requirements, and how they might differ from other translation strategies put forward by translatologists.

Developing Mailhac 1996a/b and Woolner 1998, the analysis will focus successively on procedures, parameters and strategies, but first of all we must clarify the nature of emphasis as a translation problem.

1. Emphasis as a translation problem

What constitutes emphasis is a notoriously thorny issue (Vautherin 1991:7, 47; Cadiot 1991:19). However, one common form is intonational stress (or its written representations as italics, bold, etc.) and this is the type we shall focus on. When translating from English into French, such emphasis turns out to be problematic for a number of reasons.

The first difficulty lies in detecting and interpreting correctly instances of emphasis. Typographical conventions are not always applied, so it is not unusual to find orally stressed words which are not indicated as such typographically. For instance, the following example taken from a training material script (end of an interview) did not indicate that in the English recording the second Thank you in fact carried a stress on the pronoun: "Thank you." "Thank you." This led to an erroneous rendering ("Merci." "Merci."), instead of the more appropriate: "Merci." "C'est moi qui vous remercie." (Mailhac 2000:415). Correct identification is all the more important since the same structure can take on opposite meanings through emphasis (e.g. I thought you were studying vs. I thought you were studying; Wood 1991:125). Identifying stress in speech is also fraught with difficulties, given the subtlety of the intonation patterns (Wood 1991) and the fact that the translator is not normally a native speaker of the SL. Grammatical words (see below) offer a particularly subtle continuum of possibilities in terms of emphasis (weak forms, unstressed normal forms, stressed normal forms, etc.; Wood 1991:129). Finally, written representations of speech will fail to reflect all the intonational meanings relevant to the translator's decisions (Wood 1991:124).

Although French does offer the possibility of intonational stress, it is neither as flexible nor as frequent as it is in English. Other types of resources tend to be used to convey similar meanings. Volsik (1991) observes a very high frequency of cleft constructions in the translation of emphasis into French. He also points out how translation can substantially modify the distribution of meaning. In the case of English-French translation, it can widen the range of interpretations by introducing ambiguities (see Roubichou-Stretz 1991:115 for a similar position), whilst in the other direction, it can shift what he refers to as the 'centre of gravity of the utterance' (e.g. moving the emphasis away from verbs to nouns or pronouns). Solutions often involve idiomatic equivalents which are semantically very subtle. Not unlike other languages (Anderman 1999:36), French resorts to what can be described as particles: short words such as donc, et, mais, bien, là, tiens, va, dis, enfin, aussi, alors, au fait. These connectors operate in a different way when conveying meanings equivalent to English intonational emphasis. They can lose their full logical value as part of a process which Abraham (1991, referred to by Anderman 1999:36) calls semantic 'bleaching.' The following illustrate possible contrasts of this nature in French:

"Donc comment va John?" (full connector; = "Therefore how is John?")

"Comment va donc John?" ('bleached' meaning; = "How is John?")

"Alors c'est de ta faute!" (full connector; = "Then it's your fault!")

"Alors là, c'est de ta faute!" ('bleached' meaning; = "That's your fault!")

"Aussi est-ce de ta faute." (full connector; = "Therefore it is your fault.")

"C'est ta faute aussi!" ('bleached' meaning; = "That's your fault!")¹

One of the translator's tools, the dictionary, turns out to be largely unusable for two reasons. First of all, emphasis frequently applies to grammatical words (Wood 1991:129-137) which one would not look up. Our analysis of the translation by Philippe Rouard (1984) of the first 100 italicised words found in Lewis Carroll's Through the Looking Glass reveals that 65 occurrences concern grammatical terms (all, am, are, at all, can, can't, could not, couldn't, did, do, five, four, he, here, I, is, it, I'll, I've, like, me, must, my, never, not, ought, our, shall, should, some, somebody, something, that, them, this, three, very, was, would, you, your, you'd) with the following distribution in terms of most frequent types: personal pronouns: 21; modals: 13; that/this: 11; possessive adjectives: 7; to be: 5; negation: 5; do/did: 2; to have: 2. Secondly, renderings involve very subtle pragmatic nuances which are heavily dependent on their context and will not therefore appear in dictionaries, or grammars for that matter. For instance, in the sentence John called Mary a republican then SHE insulted HIM (Cadiot 1991:21), which is impossible to translate literally into French, the meaning conveyed by the emphasised pronouns (i.e. that calling Mary a republican amounted to an insult), can only be retrieved and rendered in terms of its specific context.

Because of the nature of the problem and the mismatch in the way meanings are mapped in the SL and TL, emphasis is often not rendered adequately (Wood 1991) and there is a clear risk of producing forms of 'translationese.' Jakobsen (1986:104, quoted by Anderman 1999:40) refers to a "distinct awkwardness of style" stemming from lower frequencies of modal particle use in translated material. Considering the translation of certain types of emphasis, Volsik (1991:79) even regards as inevitable the existence of an 'interlangue linguistique' (linguistic interlanguage) exhibiting a degree of 'étrangeté résiduelle' (residual strangeness) resulting from interference.

Having established the nature of the translation problem which will be used to illustrate our discussion, we can now turn to procedures, parameters and subsequently strategies to explore how these should be formulated for the benefit of the translator and how they might differ from other translation strategies found in TS. As indicated earlier, Woolner (1988) will provide the starting point of our analysis as far as emphasis is concerned.




2. Procedures

A procedure is defined here as a means of translating a particular element as part of a strategy (e.g. cultural borrowing, calque, cultural substitution and definition are amongst the procedures available for the translation of cultural references). Some procedures are limited in scope (like the ones just mentioned), others can apply to wider units, including a whole text (e.g. exoticism with minimum/maximum presence of the translator in the case of cultural references; see Mailhac 1996a). A procedure is thus a tool to be exploited in the broader context of a strategy in order to solve a translation problem. In that sense, it is more akin to what Chesterman (1997, ch. 4) labels a 'strategy.' To the extent that the properties of a tool are determined by its intended use, procedures are goal oriented and, being part of the translational output, they are visible (e.g. one can see whether a culture-specific term has been borrowed, defined in a footnote, etc. in the translation).

In her dissertation on emphasis in translation, based on the general approach adopted by Mailhac (1996a) for cultural references, Woolner (1998) assesses the positions of Vinay and Darbelnet (1960), Astington (1983), Hervey and Higgins (1992) and Grellet (1993) in order to define a framework which is then applied to a corpus including two French translations (Papy, 1961; Parisot 1979) and two German translations (Teutsch, 1989; Enzensberger, 1998) of Alice's Adventures in Wonderland. Woolner's findings cannot be assessed in detail here, however a brief evaluation is necessary.

First of all, her sources do not include the special issue of Palimpsestes (1991) devoted to emphasis. In spite of the proposed theme (emphasis in terms of French/English translation or contrastive linguistics), some of the articles in this volume do not really seem to address the question (Berman, Chassigneux, Cadiot); some just touch upon it (Nice, Roubichou-Stretz); some deal with aspects which are not directly related to the kind of emphasis we are examining here: pauses, hesitations, etc. (Leclercq), the representation of action and activity (Guillemin-Flescher). However other contributions (Volsik, and particularly Wood) are directly relevant to our focus.

Looking more specifically at Woolner's study, the number of procedures (and parameters) she arrives at, irrespective of the inevitable limitations associated with her corpus (e.g. as a nineteenth-century written corpus, it does not contain any contemporary colloquial spoken language which might yield different results), is much more significant than what was gleaned initially from the TS literature reviewed. However, the demarcation line she proposes between lexical and syntactic procedures should be modified to avoid having 'lexical sequences' which include additional clauses (e.g. en ce qui me concerne... / il est vrai que...). Some minor reorganisation of the presentation order she adopts will make it possible to regroup items which are conceptually similar (e.g. additions). It seems preferable to use the term 'deletion' rather than 'omission' for cases where not rendering the emphasis is deliberate rather than accidental. The addition of particles could be incorporated in the list of lexical additions, since they represent a common solution. One should also add two types of procedure. The first one is phonetic (vowel/consonant lengthening; Wood 1991). The second could be described as 'descriptive label.' Although it is not present in the TS sources or the corpus, it is available in principle and is equivalent to 'descriptive characterization' (Mailhac 1999:135) which can be used to render marked speech. It involves adding metalinguistic comments in order to enlighten the reader. A statement in which emphasis conveyed surprise could thus be accompanied, if pragmatically feasible, by a description explicating the nuance (e.g. "...she replied, sounding surprised" or as a stage direction in the case of a play). The pragmatic constraints associated with it make it different from other forms of addition.

Allowing for these points, the following amended list of procedures can be put forward, regrouped into 10 types (as opposed to Woolner's 8):

(1) - Lexical procedures

- lexical repetition (It's very good > C'est très, très bien)

- use of lexical superlative/diminutive (I love > J'adore)

- addition of noun (hers > celle de Jane)

- addition of adverb or adverbial phrase (bel et bien)

- addition of interjection (pardon, voyons, par exemple, diable, etc.)

- addition of particle (va, dis, etc.)

- addition of adjective or adjectival phrase (one > seul et unique)

- addition of verb (but all he said was > mais il se contenta de demander)

- addition of conjunction (Who are you? > Et qui es-tu, toi?)

- addition of lexical sequence

- focus (Pour ma part / A mon avis)

- surprise (Quelle idée)

- lexical harmonisation (use of an idiomatic expression appropriate to the context: This time there could be no mistake > Cette fois il ne pouvait plus y avoir l'ombre d'un doute [= ... there could not be the shadow of a doubt])

(2) Syntactic procedures

- syntactic reprise (ante- and post-position: I know what you want : Ah, toi, je sais bien ce que tu veux! [literally: Ah, you, I know what you want])

- adding a clause (en ce qui me concerne)

- cleft sentences (I did it > c'est moi qui l'ai fait)

- change of sentence type (to exclamatory, negative, interrogative, etc.:You would tell Olivia ... > Qu'est-ce que tu avais besoin de dire à Olivia...? [= Why on earth did you have to tell Olivia...?])

(3) Morphological procedures

- stressed personal pronoun forms (moi, toi, lui, etc.); use of reinforcement (vous-même/vous autres)

- demonstrative pronouns (e.g. celui-ci/là)

- demonstrative adjectives (e.g. ce/cette ... -ci/là)

(4) Phonetic procedures (vowel/consonant lengthening: Dreadfully old-fashioned > Terrrriblement démodés (Wood 1991:128))

(5) Punctuation (commas, suspension points, dashes, exclamation marks, inverted commas)

(6) Typographical marker (e.g. bold, italics, underlining)

(7) Descriptive label (...she replied, sounding surprised)

(8) Compensation (e.g. transfer of emphasis from that to a reinforced negation—du tout—in the following example: Oh, I shouldn't like that! > Oh! je n'aimerais pas ça du tout!; see Harvey 1995 for the concept of compensation and Wood 1991:137-138 for examples related to emphasis)

(9) Combination of procedures (e.g. morphological + typographical: As if I would talk on such a subject > Comme si moi, j'allais parler d'une chose pareille!)

(10) Deletion (I have tasted eggs, certainly > J'ai certainement goûté à des oeufs: the emphasis on the auxiliary is not actually rendered because of the presence of certainly)

Faced with the task of selecting an appropriate procedure, the translator must consider the range of relevant parameters which will determine choices.




3. Parameters

A parameter corresponds to any factor which needs to be taken into account when choosing a procedure (e.g. communicative/pragmatic function and readership are among the parameters to be taken into account when translating culture-bound references). Parameters can apply to a small unit (e.g. word) or a larger unit (e.g. whole text). Given their role in the selection of procedures, parameters also act as evaluation criteria, since any factor relevant to the choice of procedures must be relevant to translation-quality assessment. Parameters will normally combine and interact with each other requiring the translator to assess their relationship in order to reach a decision about the most appropriate procedures.

The parameters having a bearing on emphasis are frequently ignored by authors who often stop short of making them explicit. Woolner's list (1998:32-34) includes five of them (type of word emphasized, sentence type, pragmatic function, presence of other emphasizing element, text type). Leaving aside points of detail and external factors (e.g. the translator's brief), and bearing in mind that what follows does not claim to be exhaustive, a number of additions could be made: linguistic medium, pragmatic context, readership, level of speech, linguistic frequency norms, style. Overall, the amended list includes 11 parameters, ranging from parameters which may obtain for the whole text to some which are more specific:

(1) Linguistic medium (spoken vs. written: this will affect the possibility of rendering intonation by intonation as opposed to some written equivalent)

(2) Pragmatic context (e.g. the existence of a narrator or the option of stage directions would make the use of descriptive labels possible)

(3) Nature of the text (e.g. one would expect options to be more restricted in a sonnet than a novel)

(4) Readership (e.g. there could be the possibility of slightly different use of typographical conventions for emphasis in children's literature)

(5) Style (e.g. nineteenth-century English prose; Carroll's highly frequent use of emphasis)

(6) Level of speech (e.g. colloquial language would alter the range of lexical and syntactic options in French; see Wood 1991:128)

(7) Linguistic frequency norms concerning various means of conveying emphasis in the SL and TL (e.g. the much higher frequency of cleft constructions in French compared to English; Volsik 1991:86)²

(8) Pragmatic function (e.g. expressing contrast, surprise, confirmation, challenge, contradiction, impatience, suggestion, order)

(9) Sentence type (e.g. the use of the interjection diable in exclamatory or interrogative sentences).

(10) Nature of word emphasised (e.g. emphasis on personal pronouns would often result in a morphological procedure; see Wood 1991:125)

(11) Presence of other emphasizing element (e.g. the presence of certainly in the following example makes it possible not to render the emphasis on the auxiliary without any real loss: I have tasted eggs, certainly > J'ai certainement goûté à des oeufs)

Having clarified the nature of procedures and parameters, it is now possible to address the question of strategies.



4. Strategies

As a term, 'strategy' is conceptually broader than 'procedure,' hence its use here to refer to a method employed to translate a given element/unit (including a whole text) making use of one or more procedures selected on the basis of relevant parameters. A strategy thus links procedures with the conditions which obtain when they are used, these being specified in terms of parameters. It can be either ad hoc, and be restricted to a specific context, or more general, and be reusable in a range of contexts, the latter type being naturally of greater interest to TS. When generalizable, a strategy can be construed as a rule, with the intrinsic ambiguity which characterizes this concept, as well as others such as 'norm' or 'law' (Mailhac 2006).

In its descriptive sense, 'rule' refers to some observed regularity ("X is what normally happens/As a rule, X happens."; cf. French epistemic use of il est de règle que + indicative). In its prescriptive sense, it refers to a norm to be followed ("You must do X./The rule is to do X"; cf. French deontic use of il est de règle que + subjunctive). The two senses are obviously connected ("X is what normally happens, therefore you must do X"). However, not every descriptive rule/norm/law can be associated with a prescriptive counterpart (the Archimedes principle is purely descriptive; physicists do not admonish particles to act according to the laws which characterise their behavior, etc.). This raises the question, which will be addressed later, of the relationship between descriptive and prescriptive strategies. Given that they are oriented towards the resolution of translation problems, strategies, be they descriptive or prescriptive, are teleological in nature.

Unlike procedures, strategies are not directly visible as part of the observable translation output. In principle, they fall into three categories: they can be conscious, potentially conscious (e.g. instinctive automatized translational behavior may be accessed through introspection, if required), or totally subconscious (e.g. as would be the case with undesirable strategies such as the ones resulting in various forms of translationese).³ Whenever strategies are not directly accessible through the translator, they need to be hypothesized from the available data.

As a discipline, TS operates across a range going from the non-applied to the applied. The non-applied level is concerned with the description, explanation and prediction of phenomena, and therefore translation strategies pertaining to this level have an essentially descriptive, explanatory and predictive role; they contribute to our understanding and knowledge of translation as an activity. They need to satisfy the usual requirements of descriptive and explanatory adequacy, verifiability, falsifiability, economy (accounting for the largest possible number of phenomena with the smallest possible number of explanatory facts), etc., and will normally be probabilistic. Their formulation is conditioned by their functions and a highly abstract and complex conceptual apparatus would be perfectly in order if it proved necessary to achieve the right level of adequacy.

At the other end, applied TS seeks to provide translation strategies to guide the translator in his/her task and offer a framework for quality assessment and developing translation skills. Such strategies will be prescriptive in nature rather than descriptive and explanatory as such (even if they contain an explanatory dimension, their function goes beyond explanation); they constitute decision-making tools based on choices and contribute to translation know-how. They will normally be probabilistic and acquired as explicit knowledge before some of them, at least, are internalized and applied instinctively by the translator. In this respect, they are very similar to the grammatical rules which the learner of a second language needs to memorize, internalize and apply. This similarity can assist in clarifying the nature of the criteria which need to be met for a prescriptive translation strategy to be usable.

When addressing the question "What criteria influence the level of difficulty learners are likely to experience in acquiring grammatical features as explicit knowledge?", Ellis (2002:28) puts forward the following six criteria (provided here without the examples which illustrate them):

Mutatis mutandis, this framework can be applied to explore what criteria influence the level of difficulty trainee translators are likely to experience in acquiring translation strategies as explicit knowledge. In practice, this amounts to discovering the features which prescriptive translation strategies should ideally possess to be translator-friendly, and this is what we will attempt in order to contrast them with the properties exhibited by non-prescriptive strategies of the type formulated within non-applied TS.

Formal complexity corresponds to the extent to which the strategy involves a large or limited number of procedures and parameters. The more numerous they are, the more difficult it becomes to apply or memorize strategies, particularly in view of the fact that procedures may combine with each other, as may parameters, thus multiplying the number of theoretically possible permutations. Unlike non-prescriptive strategies, prescriptive ones must therefore remain below a certain level of formal complexity to fulfil their function.

In our example, the association of 10 procedures with 11 parameters will clearly result in a fair level of complexity (higher than Woolner's who only had 8 and 5, respectively). This would reduce the chances of providing usable strategies. If the detail of the procedures and parameters which are categories (lexical/syntactic/phonetic procedures, pragmatic meaning, etc.) is provided in the formulation, then the level of complexity is significantly increased with the total number of items coming into play rising by an additional 24. This will result in a considerable multiplication of possible combinations; it is nevertheless possible to simplify formulations by adjusting the scope (see below).

On the positive side, some of the parameters hold for the whole text (e.g. linguistic medium, overall pragmatic context, nature of text), which means that, once factored in, and unless there are strong reasons to depart from them, they automatically apply to individual occurrences which makes their application easier.

Semantic complexity (the term 'semantic' is preferable to 'functional' in the context of translation) can be defined as the extent to which the meanings involved are transparent or opaque. In the case of emphasis, some of the meanings are particularly subtle, varied and difficult to identify in English, both in terms of the intonation pattern which expresses them and their actual semantic nature (Wood 1991). Similarly, French equivalents can be difficult to label and extremely idiomatic (e.g. particles). In order to retain their practical usefulness, strategies must refer to meaning types which are characterized by a reasonable degree of transparency, a constraint which does not apply in the same way to non-prescriptive strategies. Meanings referred to in non-applied theories must be clear too, but a modality which is so abstract that it could only be expressed through complex logical symbols, for instance, would not be of any practical use to a translator.

Reliability corresponds to the extent to which the strategy has exceptions. In view of the nature of the translation process, rules will normally be probabilistic and carry a number of exceptions. For example, not all instances of English prosodic stress will result in lexicalisation and statistical information about possible correlations between factors and procedures would be helpful to prioritize recommendations. On this particular criterion, prescriptive strategies remain close to their non-prescriptive counterparts, since, even at the non-applied level, claims about descriptive rules can only be made if the number of exceptions remains below a certain level.

Scope represents the extent to which the strategy has broad or narrow coverage. Here, there seems to be a trade-off in terms of usability. Compare the following possible formulations: (a) "When..., lexicalize"; (b) "When..., use an adverb"; (c) "When..., use précisément." It could be argued that (a), which has the broadest scope, is more usable, to the extent that it can be applied more frequently. However, it can also be argued that it is of less assistance to the translator compared to (c), for it does not offer a specific solution in the way (c) does (option (b), is clearly in the middle in terms of what it provides). In other words, the more general the strategy, the more usable it may prove in terms of potential frequency of use, but the less usable it may turn out to be if its broader coverage correlates with a greater lack of precision. Similarly, narrow scope may turn out to be helpful in yielding specific solutions, but, by nature, these will be very limited in their application.

As hinted earlier, varying the scope from specific equivalents to broader categories may constitute a means of simplifying strategies, reducing them to broad principles which may prove particularly helpful if combined with reliable frequency information (e.g. "When ..., the most frequent types of procedures to render English emphasis into French are, in order:....). Again, on this criterion, prescriptive strategies remain close to their non-prescriptive counterparts, as the descriptive and explanatory power may well weaken as coverage expands.

The extent to which a strategy can be formulated simply with minimum metalanguage will obviously be crucial to its acquisition and application. The terminology Woolner resorts to remains simple and, with the exception of the term 'deontic,' does not extend beyond basic grammatical labels. Although it examines a different type of emphasis from the one we are focusing on here, Cadiot's article (1991) provides a perfect example of the kind of linguistic metalanguage which may be required for descriptive and explanatory purposes, but would be highly problematic if used systematically in the formulation of applied strategies: détachement sans rappel, dislocation, détachement thématique, aboutness, apodose, protase, horizon thématique ouvert par le constituant détaché, topicalisation, acquis/given, statut sémantique et référentiel, clivage, référence déictique/générique, modalité constative/injonctive, diathèse passive/neutre, ancrage référentiel, rhématicité globale, biprédications, pronoms topiques, propositions incolores, bloc référentiel figé, mécanismes inférentiels, extériorité syntaxique, liberté référentielle, kairos, energeia, indexation situationnelle, déjà-là, cadrage de l'énonciation, etc. It is interesting to note that the author's analysis is very much a linguistic one, with very little said about translation as such. This example also illustrates how the possible contribution of linguistics to applied TS is constrained by the degree of technicality of the conceptual and terminological apparatus borrowed. The same problem would not arise with non-applied TS since, whenever necessary, its formulations can make use of a highly abstract and complex conceptual apparatus.

In theory, Ellis's L1/L2-contrast criterion could be reformulated in terms of SL/TL contrast: a SL feature that corresponds to a TL feature is easier to translate than a feature that does not. However, closer scrutiny reveals that the existence of a corresponding feature does not necessarily simplify the translation process. Both English and French can express emphasis through intonational stress, yet we have seen how problematic the translation of this linguistic feature can be. Also, the distance between languages and the resulting mismatch in the way meanings are mapped are not necessarily synonymous with difficulty. If we consider translation into English of what French tends to use to convey what would amount to prosodic stress in English (lexical/syntactic/morphological/phonetic/typographical resources, punctuation, descriptive label), being able to render all these by prosodic stress (or its written equivalents—italics, etc.) is quite economical: a whole range of very disparate French elements can be handled by a single English procedure. In other words, the SL/TL contrast has to be interpreted in terms of the number of possible procedures involved in rendering the corresponding SL problem (a stressed word in our case) in a given translation direction. Consequently, in the case of translation strategies, the SL/TL contrast will automatically correlate with the number of relevant procedures, which means that it will be subsumed within the formal-complexity criterion mentioned earlier and can be dispensed with as a separate criterion.

A further criterion to be added here concerns the order of terms in the formulation. Whereas descriptive characterisations could be formulated as "Procedure X is used when...," prescriptive statements should mirror the order of the actual translating process, starting with the conditions and ending with the choice of procedure: "When ..., select procedure X."

The criteria identified so far add up to what could be described as a 'minimax principle' of minimum effort for maximum usability to the translator,⁴ and the need for simplicity which they specify should be qualified to allow for the distinction between memorizable and consultable know-how, given that the latter can be more substantial and complex than the former. However, these criteria are not enough in themselves, since initially one has to ensure that the recommended strategy is actually a desirable one. It ought to meet certain quality standards in order to guard against the spread of undesirable translational behavior (e.g. interference, stylistic flattening, over-explicitation, etc.; Chesterman 1997:152).

In principle, desirable strategies could be construed as being simply the ones used by competent translators. In his discussion of what he calls 'normative laws,' i.e. laws "descriptive of the behavior of competent professionals" "who set the professional norms," Chesterman (1997:73-74) mentions possible criteria which might be used to identify this subset of translators. Amongst them are peer recognition and years of experience, "In other words, translator competence (on this view) is defined socially, not linguistically." If desirable translational behavior is identified as being simply what competent translators defined in this manner actually do, there is no guarantee that it will deliver quality because the criteria are not directly linked to the merits of the translation output and it assumes that such translators are generally unlikely to perform in a manner which is open to criticism. There is also a clear danger of circularity in Chesterman's position if translational competence is identified on the basis of the presence of certain behaviors, e.g. explaining culture-bound terms, to use his example: How do we know that a translator is competent? Because (s)he explains culture-bound terms. How do we know that such behavior is a sign of quality? Because that is what competent translators do.

It would therefore be more appropriate to determine what represents desirable strategies by applying empirically verifiable criteria. One could demonstrate for instance that, for a relevant set of readers, a given strategy to deal with cultural references has removed a degree of opacity which would have interfered with the communication of the message or, to use our example of emphasis, that the strategy applied has successfully conveyed the intended nuance (e.g. surprise) whilst satisfying other essential parameters. Woolner's study yields another example: 28% of cases of identified emphasis were not translated (1998:25-26). In order to establish whether we are dealing with a recommendable amount of deletion, a high level of undesirable omissions, or something in between, one could apply a combination of criteria amongst the ones which were identified (linguistic frequency norms, nature of text, pragmatic context, etc.).

It should be pointed out that prescriptive strategies need not correspond to attested translation strategies as a quality prerequisite. For instance, if no occurrence of the procedure we called 'descriptive label' has been identified in existing studies, it does not necessarily follow that using such a procedure, and therefore formulating a strategy based on it, would be inappropriate, since it could merely reflect the fact that it happened to be absent from the corpus or corpora used or, alternatively, that no one had thought of using it in spite of its obvious merits in certain contexts. Flexibility is necessary here to accommodate the possibility of new translation procedures and strategies.

At this point of the discussion a few remarks are called for concerning the way in which some of the notions used in our analysis relate to the concept of norm. We shall restrict ourselves to issues which are directly relevant to the kind of strategy we have been focusing on.

Given that strategies are rules, they share the ambivalence linked to this notion with the concept of norm in so far as both can exist in descriptive and prescriptive forms. Anything prescriptive, be it a strategy or a norm, will need to satisfy the desirability and minimax-principle criteria and anything described as a 'norm' must be based on a statistically significant volume of data. It follows from this that a descriptive norm of the type which would stipulate the procedure(s) selected when certain conditions are fulfilled in terms of relevant parameters would only differ from the corresponding descriptive strategy with regard to the statistical significance of the data on which it is based. A strategy (whether descriptive or prescriptive) can be based on a limited corpus, whilst a norm, by definition, cannot. It also means that, in practice, when a descriptive/prescriptive strategy is grounded on data which is statistically significant, it amounts to a descriptive/prescriptive norm.

As far as the relationship with parameters is concerned, certain norms can operate as parameters in view of the fact that they constitute factors which are relevant to the decision-making process. One example would be Chesterman's expectancy norms which are "established by the expectations of readers of a translation (of a given type) concerning what a translation (of this type) should be like" (1997:64). These expectations may cover: "text-type and discourse conventions, (...) style and register, (...) the appropriate degree of grammaticality, (...) the statistical distribution of text features of all kinds, (...) collocations, lexical choice, and so on." (id.).

On a first level, our analysis has enabled us to explore what makes the translation of intonational emphasis from English into French problematic, to identify 10 possible procedures and 11 parameters, and to comment briefly on some of the issues relating to relevant strategies.

On a second level, we have tried to demonstrate that prescriptive strategies, whose function is to guide the translator in his/her task and offer a framework for quality assessment and developing translation skills, should ideally comply with a minimax principle for the translator (minimum effort for maximum usability) and meet the six criteria which underlie this principle: the formal and semantic complexity, as well as the metalanguage, should be such that they do not interfere with the formulation; the number of exceptions to the rules should be limited and the scope as broad as possible; the order of formulation should mirror the translation process and start with the conditions to end with the choice of procedure. In addition, a desirability requirement, to be assessed by objective empirical criteria, should be satisfied in order to ensure quality.

Strategies put forward by the non-applied branch of TS differ from prescriptive strategies in a number of ways. Their function is to describe, explain and predict translational phenomena and therefore contribute to our understanding and knowledge of translation, as opposed to practical translation know-how. They do not need to meet the desirability and minimax-principle criteria (apart from the reliability and scope criteria) and, as a result, have a different relationship with neighboring disciplines such as linguistics, since borrowing highly complex concepts from them may well be appropriate at times.

Given the differing properties of descriptive and prescriptive strategies, translatologists should be clear about the nature and function of the strategies they endeavor to reconstruct and should not attempt to blur the differences between them, since this would undermine their specificity and, with it, the extent to which they can fulfil their respective functions.

Critics, such as Cross quoted above, who seemingly reject wholesale the contribution of TS to the work of the translator fall into the trap of a monolithic and reductionist view of TS which does not correspond to the reality. Applied TS can and does make a contribution to the practice of translation and it does so by exploiting, whenever appropriate, the findings of non-applied TS research.

¹ For a discussion of the translation of particles, see Anderman 1999. See also Newmark 1982:148.

² See the conflicting claims of Wood (1991:125) and Nice (1991:146) concerning the frequency of italics in English compared with French.

³ See Jääskeläinen 1993 for a discussion of issues related to degrees of consciousness.

⁴ The minimax principle proposed here and the framework associated with it are quite distinct from Levý's minimax principle (1967).