RSS feed

The Constraint of Relevance in Subtitling

Łukasz Bogucki, University of Lodz


This article discusses the concept of subtitling, a variety of screen translation, within the framework of Relevance Theory and Translation Studies. The constraints that operate in the process of subtitling are threefold; firstly, technical limitations as imposed by subtitling companies, secondly, abstract constraints as operative in any kind of translation, and finally, the meta-constraint of relevance.


subtitling, constraints, norms, relevance, quality assessment, strategies, procedures, semiotics

1. Introduction – a taxonomy of screen translation.

Screen translating, or audiovisual translation (AVT), is conventionally taxonomised into subtitling, dubbing and voice-over. Synchronized captions are dominant in Nordic countries, the Netherlands, Belgium, Portugal, Greece and Israel (Gottlieb, 1992:169), while dubbing is more popular in southern Europe. Fawcett (1996:84) remarks that, in principle, "large countries dub and small countries subtitle". There is a political and economic basis for this. Dubbing is much more expensive (about tenfold), as actors have to be hired to read each role, and they are not infrequently first-rate stars. Hollywood's box-office success, Shrek, starred Eddie Murphy and Cameron Diaz (the film is animated, so intralingual dubbing had to be used), whereas the Polish version had Jerzy Stuhr, Zbigniew Zamachowski and Adam Ferency, all very well known actors.

Susan Bassnett may be jumping to conclusions when she claims that dubbing is a form of government-regulated political censorship in that the original is never heard, unlike subtitling, which makes comparison possible and is therefore more democratic (Bassnett and Lefevere 1998), but political reasons behind the choice of audiovisual translation techniques cannot be entirely ruled out. Voice-over, where the target text is read out by a lektor and superimposed on the original, which is also audible, originated in the former Soviet Union and was brought to Poland, where it still enjoys popularity. According to recent research (a poll by Inst. SMG KRC Poland, 2002) 50.2% of Poles prefer voice-over and 43.4% opt for dubbing; subtitling has only 8.1% supporters. A staggering 72.1% of Poles, when asked which type of AVT was the worst, chose subtitling. The latter is a standard in Polish cinemas (intralingual subtitles seem to be gaining ground on Polish television in documentaries with authentic utterances played back from a low-quality recording, e.g. telephone conversations), and dubbed cartoons as well as certain commercials are gaining popularity, but documentaries and foreign films for television are voiced-over. This technique may be beneficial for foreign language learners, - although subtitling is undoubtedly a better choice in this respect (Brett, unpubl.) - and less costly than dubbing in that only one reader is hired, but its imperfections are many. Notwithstanding these, it remains the main mode of transferring foreign programmes onto the Polish television market, because of target audience expectations.

As we have just demonstrated, the nature of AVT is constrained. Matricial norms (Toury, 1995:58-9) stipulate when a given ST is dubbed, subtitled or voiced-over. However, this decision is one facet of the constrains on (screen) translation. The specificity of AVT may mean that the concept of translational constraints as proposed by Toury needs redefinition when applied to AVT. On the one hand, norms assist the translator in the daunting task of making choices. On the other hand, the temporal, spatial and other restrictions imposed on the screen translator narrow down the range of possible equivalents and translation strategies to an absolute minimum.

What needs to be underlined is that constraints on AVT are not to be equated with technical restrictions on the process, however significant the latter may be. Ascheid (1997) claims that dubbing is a non-constrained type of AVT in that it consists in an exchange of one voice for another, resulting in a new text that owes nothing to the original. However, this perception of constraints is clearly reduced to procedural limitations; in our understanding, dubbing is subject to translational constraints per se, as a translation activity.

2. The concept of constrained subtitling.

It seems that the type of AVT where translational constraints are at their most vivid is subtitling. In dubbing, (and, to some extent, also in voice-over) restrictions on the form of the target text stem mostly from the fact that equivalents have to match the visual component, that is the image, so what the audience hear is more or less consistent with what they see. Subtitling, as a cross-medium activity (spoken to written) is much more complex, therefore translational loss is practically an occupational hazard.

Gottlieb (1992) speaks of formal, or quantitative, and textual, or qualitative, constraints on subtitling; the former are imposed on the subtitles by the visual context of the film, and the latter are space and time factors. In practice, the visual context dictates that the verbal component is limited to what is not shown on screen, to minimise redundancy; in terms of quality, space and time restrictions may have a detrimental effect. Therefore subtitling has to do without a range of complex structures. It is quite hard to preserve stylistic effects or politeness patterns in this kind of translation. To exemplify, let us briefly look at a fragment from a popular Polish series (Czterdziestolatek) and its English subtitles from TV Polonia:

ST: „Panie inżynierze, no... no znamy się tak długo że chyba mogę być z panem szczera. Niechżeż się pan do tego nie miesza (...)." „Zosiu, mieliśmy mówić do siebie na ty." „Ale mnie to jakoś przez usta nie przechodzi."

(lit.: "Mr Karwowski, well... we have known each other for so long that I hope I can be frank with you. Stay out of this (...)." "Zosia, we were supposed to be on a first-name basis." "But I can't utter that").

TT: "Can I be frank with you? Stay of of this (...)" "Please not so formal" "I can't help it".

Clearly, the target text (leaving aside the language mistake in the first line) is pruned and rid of everything but the bare gist of the message. In terms of language, it is uncomplicated, with short, simple sentences.

The premise that underlies the subsequent discussion is that while all translation is necessarily constrained, subtitling is constrained in a more intricate and complex way. The technical restrictions on the length of the subtitle, the additive nature of subtitling and the visual complementation of the text on screen mean that the traditional approach to translational constraints, though an excellent starting point, cannot possibly account for all the facets of this unique type of interlingual translation. The gap, it is argued, can be filled by the concept of relevance. A feasible combination of relevance, traditional constraints and technicalities limiting the subtitler's freedom of choice can yield a new approach to subtitling procedures.

3. The pros and cons of Relevance Theory.

In his seminal work on the nature of translation George Steiner (1975:238) opines 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". Indeed, translation theory continually brought to the fore the same notions – equivalence, translatability, translation as science, art or craft. Since then, translation theory has broadened its scope.

Gutt (1990:135) stipulates that

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.

This is a somewhat bold statement to make. Indeed, to claim that a single principle can function in lieu of an entire translation theory seems preposterous. However, a few years earlier another approach wrought havoc with translation theory as it was known then. Manipulation School, also known as Translation Studies, dismissed the prescriptive issues of correct translation and equivalence. Another influential approach within the broad framework of Translation Studies, Toury's (op cit) model of translational constraints, has a bearing on the topic in hand. Although TS started out as a primarily literature-oriented model, it lends itself to universal application. Equally contentious and adaptable, Relevance Theory has already proved as popular as Manipulation School was in its heyday. However, in the case of Sperber and Wilson's model, versatility may be double-edged.

Their groundbreaking proposition has received criticism. The authors are not indifferent to it (Sperber and Wilson, 1997:145). Gutt's application of the theory to translation, by extension, has also had its fair share of scorn (Derrida, 2001). The main issue under contention seems to be the very concept of 'relevance', a trade-off between the benefit of acquiring new information and the cost of the processing effort necessary to derive this information. Indeed, to speak of 'relevant' information in translation may seem to be as imprecise and infelicitous as to label translations as 'good' or 'correct'.

Originally, that is in Sperber and Wilson's locus classicus, Relevance Theory addressed primary rather than secondary communication (translation). The practical distinction is that the speaker's intention is not directed at the translator, thus the meanings they infer are not destined for their own end-use. The translator merely recreates the speaker's intention, making allowances for the different resources for and constraints on the expression of conceptual content that the target language-cum-culture offers. Gutt (1991:101-2) offers an explanation by making a distinction between the descriptive and the interpretive use of language. Reporting on what another person has said or written, translation involves the latter type of language use and its goal is to achieve interpretive resemblance to the source language message:

If we ask in what respects the intended interpretation of the translation should resemble the original, the answer is: in respects that make it adequately relevant to the audience – that is, that offer adequate contextual effects; if we ask how the translation should be expressed, the answer is: it should be expressed in such a manner that it yields the intended interpretation without putting the audience to unnecessary effort.

This is perhaps where one can see a superficially weak point of Relevance Theory. By introducing the concept of relevance, it gives the outdated axiom of translation quality assessment a new lease of life. However, at the core, the syllogism does not differ from what Nida (1964) proposed in the form of his principle of equivalent effect. The tarnished notion of equivalence, criticised and nomen omen manipulated by Toury and the Manipulation School advocates, could be defined as the relationship between the source text and its translation, but not that of absolute identity of meaning, but of correspondence or matching, after all the differences between the source addressee and the target recipient have been taken into account (Hermans, 1991:157). To this end, Relevance Theory may at best be offering a new guise, yet to the same age-old bone of contention.

However, the fact that Relevance Theory is only ostensibly novel is not seen as a major drawback. That there should be a link between Nida's and Gutt's approach only reinforces the validity of the former.

Relevance Theory argues that the translator's output should not be geared towards achieving as high a degree of explicitness as possible, but should rather take account of the addressees' immediately accessible assumptions and the inferences they can readily draw (Carston, 1999: 105). The claim seems particularly valid in the case of AVT, where the recipients are forced to follow the target text (the subtitles) at a fixed pace, together with the film that they are watching. The subtitler has no choice but to liberate themselves from the stigma of literal translation and formal equivalence. Choices from among the members of individual paradigms rarely involve word-for-word translations. Verbal opulence is hardly possible in this type of AVT. Assuming, rather imprecisely, that 'not everything gets translated', Relevance Theory is seen as a tool to investigate what it is that can be done away with for the target text to meet the conventional criteria for acceptability.

Kovačič (1996) reminds us that, in the Hallidayan terms of the ideational, interpersonal and textual functions of language, subtitles favour the ideational - they are informative, whereas in dialogue it is often the interpersonal that is important. Whereas in authentic, spoken language the phatic dimension is a known fact, in the written rendition of film dialogue there is little, if any, room for it. It can then follow that our perspective on AVT, specifically on subtitling, cannot be reduced to mere technical hindrances on the process of producing the target text; the content is determined also by what is linguistically relevant in the visual context. The next section discusses translational constraints, bringing back the notion of relevance seen precisely as a constraint.

4. Translation as a constrained activity.

That translation is a multifaceted phenomenon, Venuti (2000:5-12) suggests, can be accounted for by, on the one hand, its relative autonomy, and on the other, the concepts of equivalence and function. While the translator can take the liberty to choose an equivalent that pleases him, the choice is far from arbitrary. Translation is not unmediated or transparent communication. This is what distinguishes it from the foreign text. Translation is interpretive; translations are, after all, foreign language versions of source texts. It follows, then, that the process of their production must be restrained and subject to certain regularities.

To interpret the text for the TL audience, the translator must ask about the content and purpose of the original work. Steiner (1975:8) refers to the thorough reading of the original text as "potentially unending". It is important to understand who the target audience is, as this influences choice of vocabulary and syntax. Higher education levels ensure better reading skills and broader general knowledge. In translation proper, this allows for higher register and more academic style; in AVT, it allows for denser subtitles both in their content and quantity. Audiovisual translators must also comprehend the implicit information that the author leaves in the dialogue or the visual part of the film and decide whether it needs to be made explicit. It is their task to decide whether any collateral material is needed to enable the TL audience to follow the narrative. Most recipients in AVT constitute non-specialist and non-homogeneous audiences, therefore a viable balance needs to be struck between what is presupposed and what is openly stated.

It has to be reiterated that translation is always subject to constraints. Mayoral et al (1988:361) speak of non-constrained or ordinary translation of a text in prose vs. constrained translation of films. However, their perception of constraints is limited to technical hindrances observable in the process of translation. In our understanding, the concept lends itself to a much wider application. The principle of relevance is in fact not merely an excellent solution to many translational problems, but another constraint on (interlingual) communication. Producing information in a way that does not require unnecessary processing effort on the part of the audience curtails the paradigm that the speaker, or in the case of translation the translator, has at their disposal (Gutt, 1990). Chesterman's (2000:69) taxonomy of constraints includes the social norm of communication, highlighting the translator's task to ensure optimum communication between the parties. This constraint can be linked to the principle of relevance, as the rationale behind both is basically the same.

5. The nature of AVT constraints.

Fawcett (1996:72) has looked into subtitlers' work sheets and found them "messy, with doubts and hesitations, flashes of insight and blind spots, as the translator searches for the equivalences which will fit the constraints of each given situation". Thus the process of subtitling may be thornier and more intricate than the process of translating due to the multitude of factors influencing it. On the other hand, it may be more predictable and thus subject to more detailed analysis.

What seems to be conspicuous is, clearly, curtailment due to space limits on screen and time limits dictated by the human brain's limited capability. Fawcett (ibid) says that

making comprehension of a written text almost instantaneous will produce grammatical and lexical structures of great simplicity (...) which can be verified by simply juxtaposing a page of dialogue for a dubbed and subtitled version of the same script.

The shape of individual characters is not unimportant either; i, l or t are narrow, while m and w are much wider. This can influence the choice of lexicon, as one searches for the shortest synonym with as few wide characters as possible. The fact that should a longer subtitle be necessary it can be divided into two lines provides some comfort, but it has to be borne in mind that the division is not entirely at the subtitler's free will. According to SBS Style Guide 2000, the following units must not be divided: subject and verb, verb and object, article and noun, adjective and noun, preposition and the rest of a phrase and conjunction and the remainder of the sentence.

Another technicality is that the subtitle must not pre-empt the plot line. For example, the dialogue may have been constructed so as to build up tension and the key part has been pushed to the very end of the utterance. Subtitles must respect this suspense and the written text must not appear before the spoken word.

Moreover, it is not infrequent for subtitles to be further adjusted by technicians. Thus the product of translation is given its finishing touch not by the translator, but, with due respect, by mere craftsmen entirely ignorant at translating.

Chesterman (2000:66) spoke of prescriptive expectancy norms, governed by translation tradition and genre conventions. These are established by the expectations of the audience, sometimes "validated by a norm-authority of some kind.". In the case of subtitling, companies issuing in-house stylebooks with normative guidelines may act as the norm-authority. Karamitroglou's (1998) attempt at a subtitling standard might be meta-authoritative in this sense. However, in Poland expectancy norms in subtitling are virtually nonexistent, due to the lack of a national subtitling standard and very little tradition to date. A formerly binding Polish norm permitted two lines and thirty-two characters per line of subtitle, with at least eight frames between two subsequent subtitles. At present, the minimum number of frames between subtitles is four. Some Polish companies (Laser Film Text, Film Service) allow 40 or even more characters per line. What used to be a hard and fast rule (one subtitle per one screenshot) has gone with the ever-increasing pace of action in modern film productions. Subtitles will now remain on screen for two, rarely 3-4 screenshots. A single subtitle must remain on screen for at least twenty frames, enough for monosyllables to be read and comprehended by most literate members of the audience. The maximum time for a subtitle to remain on screen is 128 frames[i].

Constraints may result from repeated practice. Not infrequently, regular occurrence of a translational solution is coupled with target audience expectations, where the addressees of a target text 'expect' the translator to come up with a particular solution. A notorious example is that of the action movie Die Hard. The Polish translator, at a loss as to how to render the title, focused on a particular scene where the protagonist has to tread barefoot on broken glass, and came up with Szklana pułapka (lit. Glass Trap). The film had two sequels; the Polish titles were, respectively, Szklana pułapka II and III, even though there was no reference to glass whatsoever in the later parts. Had the translator changed the title(s), though, the films would have suffered promotion-wise, as fewer Polish addressees would have recognised them to be sequels of the box office hit. The issue was quite interestingly echoed in a later film, a parody of American action movies, titled Szklanką po łapkach (Spy Hard, a pun on Szklana pułapka).

Restrictions of concision and omission are standard operating practice in subtitling. Pisarska and Tomaszkiewicz (1996:193-4) suggest areas that are subject to curtailment, viz. reformulations, greetings and goodbyes, information recoverable from the image, or politeness phrases. One might add the disappearance of characteristics of spontaneous speech, such as false starts, corrections, explanations, incomplete sentences, breaking messages into several parts, and so on.

It must also be borne in mind that the issue in question is language-specific. Whereas English abounds in short words (except in formal registers), and clipped forms are signum temporis (op, flu, bra, ad, high-tech, pop and many more), Polish, for example, often uses longer words, mainly due to inflection. While no research as to the length of the translation vis-à-vis the original as regards Polish and English is known to the author, the general tendency appears to be for the Polish to be somewhat longer than the comparable English.

6. A three-tier model of constrained subtitling.

The principle of relevance is seen here as a meta-constraint on subtitling that, in a manner of speaking, exists and operates independently of other constraints – norms / conventions and technical limitations. Thus, the product is a function of the operation of three types of restrictions.

What is ostensibly a technical hindrance (for instance the limited number of characters per line of subtitle) will have the subtitler searching for a coherent, yet economical equivalent. The function of textual-linguistic norms that govern the choice of lexis is strongly related. Whatever technicalities obstruct the subtitler's actions and whatever norms and conventions are at work, the choices that result from the subtitling process are filtered through the meta-constraint of relevance. The prefix requires explanation.

The versatility of the concept of relevance means that it conditions all translational actions. There are instances of subtitling solutions where the specificity of the process does not have a detrimental effect – or any analysable effect whatsoever – on the product. We must bear in mind that film dialogue constitutes a skilfully structured text, where economy is also a factor. If we allow the simplification that the main problem in subtitling is the necessity to trim down the target text, it has to be said that at least parts of the dialogue are not affected by this difficulty, as the original utterance may be concise enough to render the technical restrictions on the target text irrelevant. In certain films such a situation may even be the norm. What it means is that technical or procedural limitations may at times be inoperative, or – rarely – flouted.

In the tradition of Sperber and Wilson's work, verbal redundancy is scarce in subtitling. However, this principle cannot be misapplied, being reduced to the simple rule of 'the less you translate, the better your translation is'. The next section will discuss the issue of quality assessment in subtitling.

7. Subtitling quality assessment.

The idea of assessing translation has circulated in translation theory from time immemorial. However, a number of evaluative studies have fallen victim to being overtly prescriptive or annoyingly imprecise in their pronouncements. While most theoreticians acknowledge the need to translate 'accurately' and 'correctly', few can explain what they mean by that. Subtitling is no exception. To exemplify, let us briefly examine the model put forward by Brondeel (1994:29), who investigates three levels of equivalence that he sees at work in subtitling, viz. informative, semantic and communicative. Consequently, he asks three crucial questions, that is whether all the information has been transferred, whether the meaning has been transferred correctly and whether the subtitle also transfers the communicative dynamism. However, a perusal of his examples shows that the questions, albeit relevant, are too general and imprecise to yield satisfactory responses.

Broadly speaking, Nida's (1964:164) four basic requirements of a translation, viz. making sense, conveying the spirit and manner of the original, having a natural and easy form of expression and producing a similar response, are wholly applicable to AVT. While they can make a viable starting point, though, they need to be supplemented with other criteria. In order to generate them, a study into the nature of AVT is required.

To start with, one must take into account the multi-dimensionality of film, where the spoken word is but one element of the semiotic Gestalt. Any film is polysemiotic in texture. As Gottlieb (2001:6) aptly puts it

a screen adaptation of a 100,000 word novel may keep only 20,000 words for the dialogue, leaving the semantic load of the remaining 80,000 words to the non-verbal semiotic channels – or to deletion.

These figures are arbitrary and tentative, but it remains a fact that actual reduction in subtitling is neutralized due to the complementation of the other three elements of filmic message (Tomaszkiewicz, 1993). A linguistic analysis of the subtitled text is thus never sufficient. In the words of Jorge Díaz-Cintas (2001:189-190)

The degree of excellence of an audiovisual translation has to be measured against what the subtitlers decide to translate as well as what they decide not to translate because they prefer to exploit other semiotic dimensions of the film.

The image on screen acts as situational context, thus understatements and lack of cohesion is acceptable in film translation; the visual supplements the verbal in complete comprehension of screen production.

In most cases of cinema subtitling, the target text coupled with the other elements of the filmic message is aimed at pleasing the audience, regardless of whether the immediate feeling is amusement (comedy), fear (horror) or anything else. Therefore a significant factor in assessing subtitling quality is inconspicuousness. Bakewell (1987:16) remarks that "the best possible response from the audience would be for them never to be aware that we had done anything at all"; though the comment refers to dubbing, it seems to be equally applicable to subtitling. Jaskanen (1999: 23) comments on the invisibility of subtitling, explaining that subtitling should not be obstructive to the eye and should ideally blend in with the film. She goes on to give an example of the audience in Finnish cinemas who react to a joke in a subtitle not after having read the subtitle but after the corresponding utterance on screen has been completed: "(...) the TL audience feel they don't have a 'licence' to laugh before the SL audience do" (ibid:46). It may be due to the fact that the audience want to treat the subtitles as a 'necessary evil', in a sense subconsciously ignoring them. On the other hand in many cases the subtitles are there, in a manner of speaking, 'just for the record', as most of the audience actually follow the original; the Finnish are known for their excellent command of the English language. However, Jaskanen should be given credit for noticing that the benchmark for success in subtitling is how the audience see the product, quite like in translation, but arguably even more importantly.

Subtitling, where comparison of the original and the translation is immediate as long as the viewer has a working command of the source language, is continually subject to unprofessional 'quality assessment'. As Nornes (1999:13) remarks, "all of us have, at one time or another, left a movie theater wanting to kill the translator." He (ibid:18) goes on to contemptuously speak of 'corrupt subtitlers', who

(...)accept a vision of translation that violently appropriates the source text, and in the process of converting speech into writing within the time and space limits of the subtitle they conform the original to the rules, regulations, idioms, and frame of reference of the target language and its culture.

Therefore, in this respect subtitling is necessarily a target-oriented activity, but, Nornes goes on to explain, only until new technologies make 'abusive subtitling' possible. This type of AVT is source-oriented, not as constrained as 'corrupt' subtitling, and draws on the fact that currently movie-literate audiences can process multicoloured captions which are much longer than the conventional 35 characters, sometimes even to the extent of becoming like footnotes; therefore the subtitler has many more resources at his disposal. However interesting and innovative, Nornes' controversial idea has yet to find its place into full-scale moviemaking; as it is, we now have to contend ourselves with imperfect, 'corrupt' subtitling.

Other scholars, though, realise that imperfection is subtitlers' calculated risk; Shochat and Stam (1985:46) remark that "subtitles offer the pretext for a linguistic game of 'spot the error'." That is largely because the audience are given but one chance to understand the information that comes across to them. Neither can they re-read the unclear chunks of the target text (unlike in written translation) nor is there normal speech redundancy (unlike in spoken production, where salient information can be given a number of times). The optimum subtitle is thus one which is easy to process, yet fraught with content. Too little information, even when coupled with stimuli coming from the other semiotic channels, renders the audience confused as to what the intended meaning of the communication was. Too much information either constitutes a breach of (in most cases hard and fast) rules of subtitle production, or, where that does not apply, renders the audience even more confused as their information processing ability is stretched to the limit. The pertinent question, then, is what 'too little' and 'too much' actually mean.

Fluent speakers of English watching the satellite channel TV Polonia used to get perverse joy from finding fault with English subtitles of Polish films. At present their quality may leave little to be desired, but in the early days of Polish subtitling they would contain blunders like 'Do wójta nie pójdziemy' vs. 'We won't go to the commune leader' (the scene from Sami swoi shows two characters haggling over the price of a commodity. The seller eventually announces that an agreement can be reached, which is in Polish expressed by the above. The idiom is rooted in rural community customs, where commune leaders used to be called in to arbitrate in arguments between peasants. The effect that the original has on the source addressee is that no arbitration will be necessary, whereas the semantic translation is entirely lost with the English-speaking recipient.)

This extreme example of decontextualised translation, 'translating words rather than what they mean', is a typical translational error. Had the translator been presented with a book instead of a movie script, they could have made the same mistake. Whether it is due to the translator's inadequate language / translation competence or to the peculiarities of the given assignment and the restrictions of the length of the product is pure guesswork (although in all probability I would not be mistaken to lay the blame on the former). What has to be said, though, is that there will be overlapping between TQA in translation proper and that in subtitling. However, there will just as well be strategies unique to AVT or even specifically to subtitling.

8. The semiotics of subtitling.

The polysemiotic nature of film makes film translation a testing task. According to one of the most influential figures in translation, it was even thought to be impossible (Catford, 1965:53). Modern translation theories, however, can account for it .

Baker (1998:245) states that film is a semiotic composition of four channels: verbal auditory, non-verbal auditory, verbal visual and non-verbal visual. The model can be simplified to comprise, respectively, dialogue, sound, subtitles and image. Baker states that two factors motivate the choice what to translate and what to leave out: intersemiotic redundancy and intrasemiotic redundancy. For example, both the visual and auditive channels may convey almost identical information and the spoken word can be regarded as redundant. Alternatively, the same information may be repeated within the sound track. Each subtitle has to work both as a unit as well as part of a larger polysemiotic whole in order to maximise retrievability of the intended meaning.

A potential problem with the complementation of the visual and the verbal in subtitling is best seen in dialogues where a question is posed and confirmation or denial is expected. In cases where answers are visualised rather than verbalised, i.e. the character nods or shakes their head on screen, there is usually no addition in subtitling. However, the following hypothetical dialogue:

Do you mind if I sit down?

No, not at all.

contains a negative answer which, semantically speaking, indicates agreement rather than disagreement. Therefore its rendition into Polish is unlikely to contain negation:

Przepraszam, czy mogę usiąść?

Tak, bardzo proszę.

(lit.: 'Excuse me, may I sit down?' 'Please do').

If such a dialogue was part of a film script, the corresponding image on screen might clash with the subtitle, as the second speaker in the interaction might shake his or her head. The subtitler would have to take this into account, translating the text through the prism of the image. It may be a minor issue in the sense that no head movement whatsoever may well be made, but most literate audiences will recognise expressions for 'no' and 'yes' in many languages, and may wonder whether the translation is actually correct.

Subtitling can be described as diasemiotic (Gottlieb, 2001:8) in that its most outstanding characteristic is the shift in mode from speech to writing. This has the result that certain features of speech (non-standard dialect, intonation, style-shifting) will not automatically be represented in the written target text. Kovačič (1995) discusses excessive reliance of subtitling on norms of written language; the target text, albeit in writing, is far from typical for written language. Thus the shift is to be understood in ideational rather than factual terms. The actual original is written, as subtitlers work with a script. However, the language resembles the spoken rather than written variety.

The distinguishing features of spoken and written language are widely discussed in the literature of linguistics (e.g. Halliday 1990). However, in the film genre the original dialogue is not real, merely purporting to be so. As we have just indicated, it is first produced in the written mode, in the script: then, in the case of translation for dubbing, the original text is translated 'written word' for 'written word' and then 'spoken' by dubbing actors. In the case of subtitles, the original script is merely transposed to a different written form.

The fact that ostensibly spontaneous speech in film dialogues is acted rather than genuine means that its production is subject to (intralingual) constraints of synchrony. Mayoral et al (1988:359) taxonomise it into five types. Broadly speaking, the text has to conform to the other communicative elements of a filmic message in terms of time, space, content, lip synchronisation and character unity. Thus, while it has most of the features of natural spoken language, such as pauses, repetition, self-correction, slips of the tongue, and interruption as well as overlapping in the case of dialogue, all of the above are studiously prepared.

Another facet of this unnaturalness is that characters do not construct their utterances with their interlocutors in mind, but it is the task of the script-writer to construct discourse with the audience in mind. In Bell's (1984) taxonomy of categories of text receiver, film characters can be called addressees, but the audience can be termed auditors, ratified participants in the speech event who are not being directly addressed.

9. Subtitling strategies.

Translation strategies, and by extension subtitling strategies, should be treated descriptively rather than prescriptively, as "a posteriori analytic tools" (Fawcett 1996:69). The rationale behind their conceptualisation is to systematise commonly made choices in translation as a decision-making process.

A single instance of translation as a process can be seen as a transaction. Within that transaction, "the various modalities and procedures to go with it presuppose choices, alternatives, decisions, strategies, aims and goals" (Hermans, 1996:26). Thus, what happens as a result of the process, and as part of it, is largely predictable. However, it is also arbitrary to a certain extent. The absolute transparency of translation as science is a thing of the past. Already more than two decades ago Newmark (1981) realised the applicability of universal translation solely to a handful of text types. It has now become clear that the specificity of a particular translation assignment impinges on the character of choices made in the process. Whereas the daily grind of translating last wills and testaments or contracts may largely be rule-governed, there are kinds of translation which allow for greater freedom of choice, where paradigms are larger, choices from among their members less restrained, and the variety of applicable translation procedures greater.

However comprehensive an approach to translational procedures is applied in the case of a particular translation analysis, there will always be instances of overlapping, or even solutions that fall into none of the listed categories. Even more importantly, there will be decisions that are beyond any explanation; Hermans (1999:80) insists that translators' choices do not always have to be conscious or rational. Nevertheless, certain solutions are more frequent than others.

Studies of translation strategies are frequently descriptions of handling non-equivalence. Mona Baker (1992) lists eight strategies of coping with lack of equivalence at word or phrase level. All the solutions she suggests seem to bring about translational loss, in one form of another. They detract from the target text in terms of accuracy (translation by a superordinate, a more neutral or less expressive word, or by cultural substitution) or in the experience of the target readership, who have to interrupt the continuity of reading to look at footnotes (translation using a loan word with a subsequent explanation). They could be downright corner cutting (omission). Finally, they can unnecessarily lengthen the target text (paraphrase).

The linear set of translation strategies proposed by Vinay and Darbelnet (1977/1995) has turned out to be comprehensive and applicable to most translational actions. The linearity of the approach manifests itself in the seven procedures (borrowing, calque, literal translation, transposition, modulation, equivalence and adaptation) ranging from the most semantic to the most communicative one. An interesting feature of Vinay and Darbelnet's model is the dichotomy of obligatory and optional shifts. It follows that the translator is at liberty to make certain adjustments if he deems them appropriate (optional shifts), but to produce a target text that adheres to whatever notion of correctness is valid for the particular commission, he is forced to make certain other adjustments (obligatory shifts). This distinction runs parallel to Toury and Hermans' (op cit) arrangements of constraints in order of force.

Salkie (2001:437) explores modulation and transposition, the two most productive strategies, with reference to Relevance Theory. His distinction between the two is more practical than Vinay and Darbelnet's:

In the case of transposition, the guiding question is 'how would the target language naturally express it?'; with modulation the question is 'how would a speaker of the target language naturally conceive of it?'. Thus a translator who proceeds by way of modulation has considered an extra dimension which transposition does not involve – not just the words but what they refer to.

Chesterman (2000:85) aptly compares translation strategies to means that are used in order to meet ends, that is translation norms. It is also in this paper that this relationship is cultivated, albeit with the introduction of a third variable, namely relevance. Thus, by adopting appropriate strategies the subtitler adheres to valid norms, simultaneously obtaining relevance in the context of the whole of the filmic message. The process produces optimum translation.

A perusal of subtitling strategies (Gottlieb, 1992:166) shows that the constraints that operate in the process of subtitling result in a greater degree of loss than it is the case in translation proper. Out of ten procedures that Gottlieb sees at work in subtitling at least three entail considerable reduction. The meta-constraint of relevance, therefore, works as a filter, making sure that what is lost in the process is irrelevant or does not prevent the audience from appreciating the resulting product.

10. Conclusions.

The starting point for this discussion was that audiovisual translation, in particular subtitling, constituted a peculiar type of translation. While the obvious similarities that subtitling bears to interlingual translation proper do not prevent general translation theory from providing a surface theoretical framework, it does not suffice to account for the intricacies of this kind of screen translation. Relevance Theory and Translation Studies, specifically the latter's understanding of translational constraints, appear to be of some assistance in this task.

Having probed into the foundations of both approaches, we believe that the constraints on subtitling are threefold. Firstly and quite predictably, in the case of subtitling the length of the target text is limited. The text itself is made accessible to the target audience for a matter of seconds, after which time it is irrevocably gone. The aim of the target text is to aid the audience in comprehending and appreciating a filmic message. Taking the above into account, the subtitler maximally simplifies their product, so that the message it conveys gets across to the intended recipient, yet the process of taking it in is not too strenuous.

Secondly, a parallel is drawn between the assumption above and the principle of relevance. We find that since both can be reduced to essentially the same premise, the constraint of relevance is as applicable to subtitling as the technical restrictions that we have just brought up.

Thirdly, since subtitling, however unique, is after all an instance of translation, it should be subject to translational constraints in the more abstract, Tourian sense of the word. They should not be confused with technical hindrances, as by definition Tourian constraints are meant to help rather than encumber the translator.

Subtitles cannot exist independently, in the same fashion as a film script that is unintelligible without the corresponding image and soundtrack. Thus, what the target text contains is a result of the application of multifaceted constraints and relevant cultural filters. Ideally, when presented to the audience together with the visual, verbal and sonic stimuli of the original (subtitling is additive, the TT does not replace the ST but supplements it), the effect it produces is maximal comprehension and appreciation at minimal processing effort.


Ascheid, A. (1997). "Speaking tongues: voice dubbing in the cinema as cultural ventriloquism," The Velvet Light Trap, no. 40 (Fall 1997).

Baker, M. (1992). In Other Words. A Course in Translation London and New York: Routledge.

Baker, M. (1998). Routledge Encyclopedia of Translation Studies. London : Routledge.

Bassnett, S. and A. Lefevere (1998) Constructing Cultures: Essays on Literary Translation Clevedon: Multilingual Matters.

Bell, A. (1984). "Language style as audience design," Language in Society 13, pp. 145-204.

Brett, P. "The rationale for the usefulness of multimedia for language learning," on-line at

Carston, R. (1999). "The semantics/pragmatics distinction: a view from Relevance Theory," in K. Turner (ed.) The Semantics/Pragmatic Interface From Different Points of View Elsevier Science, pp. 85-124.

Catford, J. C. (1965). A Linguistic Theory of Translation Oxford: OUP.

Chesterman, A. (2000) 'Teaching strategies for emancipatory translation,' in C. Schäffner and B. Adab (eds) Developing Translation Competence Amsterdam and Philadelphia: John Benjamins, pp. 77-89.

Derrida, J. (2001). "What is a 'relevant' translation," translated by L. Venuti, Critical Inquiry Winter 2001, pp. 174-200.

Díaz-Cintas, J. (2001). "The value of the semiotic dimension in the subtitling of humour," in L. Desblache (ed.) Aspects of Specialised Translation Paris: La Maison du Dictionnaire, pp. 181-190.

Fawcett, P. (1996). "Translating film," in G.T.Harris (ed.), On Translating French Literature and Film, Amsterdam/Atlanta: Rodopi, pp. 65-86.

Gottlieb, H. (1992). "Subtitling – a new university discipline,' in C. Dollerup et al (eds) Teaching Translation and Interpreting Amsterdam: John Benjamins, pp. 161-170.

Gottlieb, H. (2001). Screen Translation. Six studies in subtitling, dubbing and voice-over Copenhagen: Kopi Service.

Gutt, E.-A. (1990). "A theoretical account of translation - without a translation theory," Target 2, pp. 135-164.

Gutt, E.-A. (1991). Translation and Relevance. Cognition and Context London: Basil Blackwell.

Hermans, T. (1991). "Translational norms and correct translations" in K.M. Van Leuven-Zwart and T. Naaijkens (eds) Translation Studies: The State of the Art, Amsterdam: Rodopi, pp. 155-169.

Leuven-Zwart (1996). "Norms and the determination of translation: a theoretical framework" in Álvarez and Vidal (eds) Translation, Power, Subversion Clevedon: Multilingual Matters, pp. 25-51.

Leuven-Zwart (1999). Translation in Systems Manchester: St. Jerome.

Jaskanen, S. (1999). On the Inside Track to Loserville, USA: Strategies Used in Translating Humour in Two Finnish Versions of Reality Bites, pro gradu thesis, Helsinki University.

Karamitroglou, F. (1998). "A proposed set of subtitling standards in Europe," Translation Journal Vol. 2 No. 2.

Kovačič, I. (1995). "Reinforcing or changing norms in subtitling," in C. Dollerup and V. Appel (eds) Teaching Translation and Interpreting III, Amsterdam: Benjamins, pp. 105-109.

Kovačič, I. (1996). "Subtitling strategies: a flexible hierarchy of priorities," in C. Heiss and R.M. Bollettieri Bosinelli (eds) Traduzione multimediale per il cinema, la televisione e la scena Bologna: Clueb, pp. 297-305.

Mayoral, R., D. Kelly and N. Gallardo (1988). "Concept of constrained translation. Non-linguistic perspectives on translation,' Meta XXXIII, vol. 3, pp. 356-367.

Newmark, P. (1981). Approaches to Translation Oxford: Pergamon Press.

Nida, E. A. (1964). Towards a Science of Translating Leiden: Brill.

Nornes, A. M. (1999). "For an abusive subtitling," Film Quarterly 52(3), pp. 17-34

Pisarska, A. and T. Tomaszkiewicz (1996). Współczesne tendencje przekładoznawcze Poznań: UAM.

Salkie, R. (2001). "A new look at modulation," in M. Thelen and B. Lewandowska-Tomaszczyk (eds), Translation and Meaning Part 5, Proceedings of the Maastricht Session of the Third International Maastricht – Łódź Duo Colloquium on 'Translation and Meaning' Maastricht: Hogeschool Zuyd/Maastricht School of Translation and Interpreting, pp. 433-441.

Shochat, E., and Robert Stam (1985). "The cinema after Babel: language, difference, power," Screen XXVI, pp. 35-58

Steiner, G. (1975). After Babel. Aspects of Language and Translation Oxford: OUP.

Sperber, D. and D. Wilson (1995, 2nd edition) Relevance: Communication and Cognition Oxford: Blackwell.

Sperber, D. and D. Wilson (1997). "Remarks on relevance theory and the social sciences," Multilingua 16, pp. 145-151.

Tomaszkiewicz, T. (1993). Les Opérations Linguistiques qui Sous-tendent le Processus de Sous-titrage des Films, Poznań: UAM.

Toury, G. (1995). Descriptive Translation Studies and Beyond Amsterdam/Philadelphia: John Benjamins.

Venuti, L. (ed.) (2000). The Translation Studies Reader London and New York: Routledge.

Vinay, J.P. and J. Darbelnet (1977). Stylistique comparée du français et de l'anglais Paris: Didier. In English as (1995). Comparative Stylistics of French and English, translated by J. C. Sager and M. J. Hamel, Amsterdam and Philadelphia: John Benjamins.

I am indebted to Elżbieta Gałązka for this information.

Łukasz Bogucki is adjunct professor and teacher of translation and interpreting at Institute of English Studies, University of Lodz, Poland.

He obtained a doctorate in translation theory in 1997. He is currently working on a post-doc project in audiovisual translation. He can be reached at