Film is a combination of many signifying codes, broadly classified as visual or acoustic.¹ Although the specificity of film lies in the interaction of these multiple codes, some of them are, in practice, more equal than others. Film scholars and cinephiles have traditionally privileged the visual over the acoustic; it is well known, for instance, that Henri Langlois exhibited prints of films at the Cinémathèque française without translation. By contrast, scholars of translation are most interested in film dialogue, and so modes of audiovisual translation tend to be defined in terms of how they treat dialogue: subtitling translates dialogue by superimposing written translations of the dialogue on screen; dubbing replaces the dialogue track with a re-recorded dialogue track in the target language; voiceover overlays a new dialogue track on the old one, usually spoken by one, or at most two, speakers.

The assumption here is that each mode of translation transposes specific, selected codes in a film – by contrast with, say, book translation, where we would normally expect all the different elements (text, typographical codes, layout, illustrations, cover design, etc.) to be transposed in some form. Subtitling leaves acoustic codes unaffected, altering only the visual element of the film by the superimposition of the titles. Dubbing and voiceover leave visual codes unaffected, only substituting or supplementing acoustic elements. Of course these are generalisations, but they have a powerful structuring effect on the ways in which we carry out research in film translation, and, more importantly, on the ways in which film translation is presented to the public, e.g. as a set of options on a DVD. Films on DVD are often presented with multiple subtitle and/or audio tracks. Through interactive menus, the viewer is given the choice of which version of the film to watch. This presentation depends on the assumption that the film text, in other words the succession of images or shots, is invariant, and needs merely to be overlaid by subtitles or have its audio track substituted with a dubbed track to function coherently as a film in the target language.

But we don’t have to look very deeply to see how fragile these assumptions about the “unaltered” elements of film in translation are. For example, it has been argued that subtitling allows us access to an “authentic” original film. As Jorge Díaz Cintas observes, “[o]ne of the advantages generally attributed to subtitling is that this mode respects the original, as it ensures that the product remains intact, with the written T[arget] L[anguage] text simply being added”.² Richard Kilborn likewise suggests that “the original text remains intact beneath the subtitled overlay”.³ But quite apart from the fact that watching a subtitled film is a very different experience to watching the same film unsubtitled,⁴ the film text itself may be altered. A film may be cut, as well as subtitled, for release abroad, for reasons of marketing or censorship (or both).⁵ The aspect ratio may be altered when a film is published in a new format, which affects the presentation of the image as well as the visibility or layout of the subtitles.

Equally, when a film is dubbed, more than the soundtrack may be altered. Digital technologies allow various forms of manipulation of the image; Japanese anime dubbed for release in the United States, for instance, could feature “digikinis” or digitally added garments to reduce the amount of nudity on screen.⁶

The unspoken words of cinema

Long before digital technologies, certain kinds of image were already the subject of manipulation in translation: what we might call the unspoken words of cinema. By unspoken words I mean written text in the image: what Dirk Delabastita calls the “verbal visual” element and what Frederic Chaume refers to as “graphic codes”.⁷ These are a common feature of film. They may be in the form of filmed written texts or captions superimposed on the image. For instance, credit sequences provide metadata. Posters, signs or advertisements of various kinds may function as elements of set dressing. Captions or intertitles offer necessary contextual information (“meanwhile, back at the ranch…”). Characters may write and read written or printed texts of different kinds.

This last category is really what I want to focus on here: texts that are located within the story world (diegetic texts) and to which our attention is explicitly drawn, for instance in insert shots.⁸ Text inserts like this were already common in the silent period as an alternative to dialogue titles, or to extra-diegetic intertitles with their omniscient narrative voice. F.W. Murnau’s late silent film Tabu: A Story of the South Seas (1931) is conspicuous by its preference for diegetic rather than extradiegetic text on screen. Carl Dreyer’s early sound film Vampyr (1932) is a good example of a film which depends heavily on text inserts: shots of pages from a book about vampires have an essential expository function in the narrative. Insert shots of newspaper headlines, letters, telegrams, notices and other written texts (and, more recently, telephone and computer screens) have continued to be widely used by filmmakers for dramatic effect and to convey key plot points. In the heyday of classical cinema they were prominent in the work of filmmakers such as Fritz Lang and Alfred Hitchcock.

Text inserts do not generally constitute a problem for subtitling; their content can be represented in subtitles in the same way as the dialogue can. The subtitler must decide whether the text on screen is semantically “active” and meaning-bearing for the target audience, in which case it should be subtitled if possible, or whether it is just visual noise, in which case it can be left unsubtitled.

Dubbing is a different matter. Dubbed speakers move about the story space speaking the target language, but the story space is potentially still peopled by diegetic signs, notices and inscriptions in the original language in which they were filmed. The composition of the shot and the mise-en-scène indicates to viewers to what extent the information is relevant; sometimes information presented in this way will be trivial, but sometimes it will be essential for understanding the plot. Dubbing aims by definition to offer spectators a spoken translation, so some dubbed films translate the relevant inserts through a voiceover.⁹ This leaves the source language display on screen, a potential distraction for the viewer or a source of linguistic incoherence.

Another, more radical, option might be to excise the insert shot, perhaps compensating by providing the information elsewhere in the dialogue. Nowadays, with more flexibility in modes of audiovisual translation, a dubbed film might choose to subtitle the important text inserts. But in the early decades of sound film, a favoured solution was to recreate the inserts in the target language. Reshooting text inserts and editing them into the original image track created a story world which was potentially entirely adapted, both visually and acoustically, for the target market. Edmond Cary argues that dubbing is a more complete form of translation than other forms: “Will it not be the glory of our century to have invented a type of translation […] which […] accepts all the constraints of other types of translation, which can lay claim to the title of total translation.”¹⁰ (His preoccupation is with moving away from a text-based definition of translation, and he does not explicitly address the dubbing of the story world in the article, but we might bear in mind that as someone writing in the 1950s and 1960s, Cary would have grown up with translated insert shots as standard practice in Hollywood films.) If multilingual production was a form of “dubbing the body of the actor” as Ginette Vincendeau has argued, the reshooting of text inserts had the effect of “dubbing” the physical environment.¹¹ We can treat this practice as a pre-digital form of localization, following Reinhard Schäler’s definition of localization as “the linguistic and cultural adaptation of [...] content to the requirements and the locale of a foreign market”.¹²

Figures 1-4 which follow provide some examples of what these localized text inserts looked like. All the examples in this article are from live action film. The case of animated film is rather different and will not be covered here.

It was not only diegetic text which received this treatment. Credit sequences were routinely recreated in the target language and indeed continue to be so. Prefatory material such as prologues and epigraphs also needed to be translated:

This practice of reshooting inserts raises questions of some importance. Under what conditions was the work done, where, and at what stage of production? My assumption has been that these localized versions were part of the first dubbed versions for theatrical exhibition, but this remains an assumption, in the absence of documentary evidence. What technological and other constraints governed this practice? As Karin Littau has argued, tools and technologies have a direct effect on the practice and theory of translation.¹³ They also, of course, govern film production. What relationship, if any, existed between the status of text inserts as part of film style, and the practices of audiovisual translation?¹⁴ What aspects of translation theory may best explain the translation decisions made? How do these translated elements interact with other elements of a film text? What impact, if any, did this practice have on film reception?

Production and translation

First of all, we should emphasise that this issue is both one of production and one of translation. In a sense, these shots could be treated like any other pickup shot. In some films they might only fill a few moments of screen time; in other films they have a major role to play. The Testament of Dr. Mabuse contains some 35 written texts, not including the credits, which needed recreation or transfer in English. Only a few of these are repeat shots or close-ups of wider shots. They reflect the thematic and documentary importance to Lang of signs in public space (see also his previous film M), as well as the importance to the plot of Dr. Mabuse’s mad scribblings. Hitchcock’s Suspicion contains about 25 such shots, which create suspense and surprise, and convey key plot points.¹⁵ From a technical point of view, the volume of pickup shots needed for localisation would have required a solid infrastructure with provision for casting (e.g. for the hands which are often visible in images and letters), lighting, set dressing, graphic design and so on. The shots would need careful editing and splicing into the print of the original version sent to the dubbing studios (see below). All of these are aspects of the overall localization process.

Text inserts also, in a very concrete sense, need translation. As we can see from the prologue sequence from Spellbound (figure 6) in English and Italian, text inserts could be quite substantial. The poem from Blackbeard the Pirate (figure 5) which functions as an epigraph to the film poses some challenging issues to the translator, who must decide whether and to what extent to reproduce the textual effects such as rhyme and period language. The poem has a humorous, ironic tone which would need to be conveyed in translation.

While some shots (e.g. those from Le Procès Paradine in figure 1) seem to reproduce the English material as mimetically as possible, including in layout and typography, others adapt the material. In the epigraph from Barbe-Noire le pirate, we can identify textual shifts of various kinds; the rhyme scheme and metre are modified, the mock attribution and the quotation marks disappear. The production values seem lower in the translation, e.g. with the less elaborate initial capital and the more limited colour palette.

Names are often adapted, as they are, of course, in the dubbing scripts themselves. In Suspicion Johnnie Aysgarth’s employer, George Melbeck, is Georges Melbeck in French and Giorgio Melbeck in Italian. Even fairly trivial elements may conscientiously be transposed, as in the case of this addressed envelope from Suspicion (figure 7) which is adapted to a degree which we would find unusual today, e.g. rendering “High Street” literally as “Rue Haute”. The culturally specific designation “Esquire” has been adapted to “Capitaine”.

The name of Lina McLaidlaw, the heroine of Hitchcock’s Suspicion, played by Joan Fontaine, is rendered in a variety of ways: Lina Mac Laidlan in Italian; Lina Mc Kinlaw in French (figure 8).

Her husband’s friend and business partner is referred to variously as Gordon Cochrane Thwaite (English), Gordon Thwaite (Italian) and Gordon Wells (French) (see figure 3). In the French version of While the City Sleeps (Fritz Lang, 1956), the character of Edward Mobley is rechristened Jack Mobley, and the reshot text inserts are consistent with this.

We could consider the adaptation of names through the prism of Lawrence Venuti’s theory of domestication and foreignisation.¹⁶ Rather than facilitating a foreignising effect, these translation decisions seem to position the translated film as much as possible as a naturally occurring target-language text. Inconsistencies can, however, be found. In Spellbound, the male protagonist, played by Gregory Peck, is introduced as “Dr Anthony Edwardes”, but it is soon clear that he is not Edwardes and is suffering from traumatic amnesia. It is the comparison of his signature with Edwardes’ own authentic signature which proves to Dr Constance Petersen, played by Ingrid Bergman, that Edwardes is not who he thinks he is. In the Italian version Io ti salverò, Edwardes’ first name is culturally adapted to “Antonio”¹⁷ This causes a problem for the presentation of the signatures:

As we can see from figure 9, the Italian inserts add an “io” to the name, while apparently (oddly) retaining the long tail of the y. When it comes to the comparison of the signatures, however, the Italian version merely retains the English shot of the two signatures side by side (figure 10):

There seems to be no reason for this inconsistency since the Italian profilmic objects had already been created and could have been juxtaposed in the Italian as well as in the English. It may simply have been an oversight.

Text inserts and editing

Diegetic text on screen does not always occur conveniently in insert shots.¹⁸ More complex problems of editing and continuity are posed by text which is visible as part of the filmed action of a scene. In Blackbeard the Pirate, ship’s surgeon Robert Maynard is operating on Blackbeard under duress when a note is slipped to him urging him to kill the pirate. As we can see from the hands, the insert in French is reshot using a different actor. The shot transitions in the French are jerky and awkward, reflecting the difficulty of integrating the reshot insert seamlessly into the rest of the action.

In Hitchcock’s Spellbound, the heroine goes to the hospital’s library to find a copy of the book Labyrinth of the Guilt Complex by the new director Dr. Anthony Edwardes. We see a shot of a row of books and a hand which reaches out and takes one of the books down. The close-up shot of the row of books is recreated in the Italian (figure 12):

In the English version this is part of a single shot beginning with a close-up of the book spines and continuing as a hand reaches out and takes down Edwardes’ book. In Italian, the shot of the row of Italian books is interpolated, and then there is a cut back to the original footage; it can plainly, if momentarily, be seen that the book taken off the shelf is the English copy.

A similar problem occurs several times in While the City Sleeps, titled in French La Cinquième Victime. Towards the beginning of the film a detective is investigating a crime scene where the murderer has scrawled words on the wall. The detective is seen in a medium shot with the graffiti visible in the background. The camera cuts to a close-up of the graffiti. In the French version, the establishing shot is from the original English version, but the insert of the graffiti is reshot in French.

Later in the same film, we see a newspaper headline, partly obscured by a shot glass. The close-up of the newspaper headline is reshot in French and interpolated, but the rest of the shot, where the hand lifts away the shot glass, is retained from the English version (figure 13).

The editing together of elements in both languages potentially presents a dissonance for the viewer, but we suppose that the foregrounding of the translation through insert shots creates a sort of “relevance effect” which the audience would have understood much in the manner of a subtitle; rather than substituting for the “source-language” shot it supplements it. This relevance effect may also help to explain images which are only partly localized, e.g. this newspaper page from Spellbound, where only the main headline is translated into Italian; the rest of the page remains in English:

In a fascinating article on M le Maudit, the partly-dubbed, partly-reshot French version of Fritz Lang’s M, we are told in relation to the film’s many instances of in-vision text that “it is impossible systematically to substitute a French filmed equivalent”.¹⁹ The French version was released in April 1932; this makes it part of the heyday of multilingual production, so it seems a little surprising that more rigorous localization did not take place, as, for instance, with Lang’s subsequent film Le Testament du Docteur Mabuse. Certainly by the late 1930s it seems to have been common for extensive pickup shooting of text inserts to be carried out as part of the localization process.

The context of insert localization

From a historical point of view, it is of interest to understand in what contexts this reshooting of texts was carried out, and for how long. It seems to have been a common practice in the major European territories, though few studies that I have consulted acknowledge this localization practice.²⁰ Michel Chion suggests that it was common practice in France up until the 1950s.²¹ This date range may be a little conservative; the examples I have come across run from the early 1930s to the early 1960s. By the 1950s there is some evidence that pickup shooting was more selective, and we find superimposed captions on screen as well as reshot inserts. In the case of animated films the practice was certainly common for much longer. With regard to location, it is presumed that in most cases the localization is carried out as part of the overall dubbing process. Several countries (France, Italy) had legislation in place to ensure that dubbing was carried out in the respective territories as a measure to protect jobs in the local film industry, and it seems safe to assume that the text localization would have taken place alongside the dubbing. This may also help to explain differences in approach between different language versions (see e.g. the examples from Suspicion in figures 3, 7 and 8 above). Admittedly, the case of Suspicion is slightly unusual. According to the account by Gerardo di Cola, at the time of the Armistice of Cassibile in September 1943, when Italy formally surrendered to the Allies, a group of Italian actors were in Spain shooting the feature film Dora, o le spie (Raffaello Matarazzo, 1943). After the Armistice they were unable to return, and were forced to remain in Spain until the end of the war. During this time they dubbed a number of films for Twentieth Century-Fox.²²

The technical issues raised by the practice of reshooting text inserts are intriguing. In credit sequences or prologues (Barbe-Noire le pirate, Le Procès Paradine) we find that text which in the original film is overlaying moving images is sometimes provided over a static image in the localized version (figure 15):

The fact that some reshot sequences have no moving images underlying the text suggests that “clean” footage was not made available for recaptioning, or at least not always. Localization would have been done uniquely through editing new footage out and in as appropriate. In the Italian version of Suspicion (Il sospetto) it seems in some cases as though the text inserts are drawn in the Italian, rather than filmed (see e.g. the train ticket in figure 3 above). If this is the case, it may have been a function of the unusual circumstances in which this dubbed version was made.

Implications of insert localization

There are many implications of this practice for the way that we think about film “in translation”. As with any source-text/target-text pair, the reshot inserts raise questions of equivalence and fidelity to the source shots. We may think about the extent to which the target elements of the shot (shots of hands, pages, backgrounds, typography and layout) are an “accurate” reproduction of the source shot, or are consonant with their target text environment. What kinds of standard translation shifts and adaptations do we observe; what are the norms of insert localization? We could speak about “source hands” and “target hands” with hardly a trace of flippancy: the non-verbal visual elements of the film are content to be translated, just as the words are, and they are an aspect of the final dubbed film which must be taken into account when we look at these films as translations.

Reshot inserts also raise important issues of textuality. They compellingly illustrate François Thomas’ observation that “it is the rule, and not the exception, that a film will be shown in many alternate versions in the course of its lifetime”.²³ Today, these localised versions seem largely forgotten. They are apparently so disregarded by distributors that they can cause interesting anomalies for publishers of VHS and DVD editions. There may be multiple versions of films on the market in different formats; VHS editions and broadcast versions with reshot inserts may have been replaced, over the years, by DVD editions produced from source language masters. A film may be “fully” dubbed (including both acoustic verbal and visual verbal elements), or partly dubbed, where a dubbed soundtrack is added to a source-language version of the film. In the latter case the acoustic verbal elements will have been transposed into the target language but credit sequences, captions and inserts will still be in the source language. The audience may be mystified at close-up shots of letters or notices whose duration, and perhaps musical accompaniment, underlines their importance for the plot, but for which no translation is supplied on the dubbed dialogue track. It may thus happen that a film is presented on DVD as available in both dubbed and subtitled versions, but only one of these labels is really accurate, depending on the master that has been used. My French DVD copy of Le Procès Paradine, which was the original spur to this research project, says that the film can be watched in “VO” [Version Originale, i.e. the English version] or “VF” [Version Française, i.e. the dubbed French version], but the credit sequences and inserts are in French only, meaning that the claim to offer a “VO” is at best only partly true. In the case of this particular film we are only speaking of a couple of text inserts, and the effect on the viewer would be small; in the case of other films the plot could become difficult or impossible to follow. The reshot insert functions as a site of textual instability, to be addressed through careful scholarship.²⁴

“Originality” is also put into question. In any case we must treat with some scepticism the very existence of “an” original for film,²⁵ and particularly so when it comes to translation.²⁶ Which is the original? The “ideal” source-language text? The film as seen by the first viewers in the source culture? The version seen by the first viewers in the target culture? As pointed out by Michel Chion and by François Albera, Claire Angelini and Martin Barnier, it is often forgotten by today’s critics that the version of a given film they are watching may not be the version which was first exhibited and which governed the early reception of that film in the target market.²⁷ Interestingly, in an era when many classic films have been redubbed, there is now nostalgic interest in the “doppiaggio storico” [historical dubbing] or the “doublage d’époque” [period dubbing] as an original in itself, to be preferred to later and inferior redubbings.²⁸ The criticisms of later dubbings rehearsed by online communities mirror the preoccupation with fidelity which characterises a lot of traditional translation criticism.

Reshot inserts represent points of friction which invite us to look again at the implications of audiovisual textuality for translation research. What structuring categories do we use? Most film historians, as well as translation historians, would make a fairly clear distinction between multiple-language production and dubbing, for instance. But reshot inserts add an element of multilingual production to otherwise dubbed films. Some films, e.g. M le Maudit, used a combination of dubbing and reshooting, not only of text inserts but also of whole scenes.²⁹ Films such as M le Maudit, or The Crimes of Dr Mabuse, which does not reshoot scenes but is heavily recut in relation to Lang’s German original and which extensively localizes its inserts, shows how porous the boundary between multiple-language versions and dubbed versions can be; not distinct categories, but a continuum.

Lastly, there are implications for archiving and the study of film. The films which I have drawn on for images here are sometimes available in old VHS or cheap DVD editions, replaced as soon as commercial considerations warrant it by more “authentic” source-language master copies, or are chance-found off-air recordings from broadcast versions. There seems to be no system for collection, storage or study of localized film copies, which means that much of the material in this field may already have been lost. Archival collections of “film in translation”, which would combine theatrical, broadcast, VHS and DVD copies for study, would allow us to open a new chapter in research into the ways in which films travel between languages and cultures.

Note on sources:

Data is given as fully as possible, where available.

The French inserts for Le Procès Paradine are from an undated Region 2 French DVD offering English and French audio and French subtitles published by Aventi, copyright ABC, Inc. The image track is that of the first French dub carried out in 1949, according to the visa in the opening credits. This dub was carried out by Lingua Synchrone.

The French inserts for Barbe-Noire le pirate are from a 2004 Region 2 French DVD offering English and French audio and French subtitles published by Editions Montparnasse in the “Collection RKO”.

The inserts for Soupçons are from a French VHS tape released in 1986 by Ciné-Collection. Cf. Pascal Laffitte’s blog post of 22 August 2012 on the Objectif Cinéma site: http://www.objectif-cinema.com/blog-doublage/index.php/2012/08/22/365-soupcons-1941.

The inserts for Il sospetto seem to be from a version broadcast on the Italian television channel RAI 3. An edited selection of inserts is available at at time of writing.

Das Testament des Dr. Mabuse and Le Testament du Docteur Mabuse stand to each other in the relation of full-scale multilingual productions. They can be found on the 2004 Criterion Collection edition (spine no. 231) of The Testament of Dr. Mabuse. The dubbed and recut 1951 American release The Crimes of Dr. Mabuse is available on the NTSC DVD The Diabolical Cinema of Dr. Mabuse Volume 2, All Day Entertainment Deluxe Collectors Edition, released in 2000.

The Italian version of Spellbound, entitled Io ti salverò, is available at at time of writing.

The copy I consulted of La Cinquième Victime (While the City Sleeps) is from a French broadcast.

I am indebted to Francesco Caminiti for invaluable information about dubbing and insert localization in Italy, and for making available to me inserts from a number of Italian versions. Thanks to Ghislaine Weidmann for making available to me the copies of Soupçons and La Cinquième Victime.

About the author

Carol O’Sullivan is a Senior Lecturer in Translation Studies at the University of Bristol. Her research interests include audiovisual translation, translation history and literary translation. She is the author of articles on the translation of classical literature, censorship in translation, Victorian translation publishing and subtitling. She is a co-editor with Luc van Doorslaer, Denise Merkle and Michaela Wolf of The Power of the Pen: Translation & Censorship in Nineteenth-Century Europe (Vienna: LIT Verlag, 2010). Her monograph Translating Popular Film was published by Palgrave Macmillan in 2011. She is Associate Editor of the journal Translation Studies.

E-mail: carol.osullivan@bristol.ac.uk