The layers of subtitling

The study of subtitling, although widely practiced over the past 20 years, has generally been confined to comparative studies focusing on the product of subtitle translation, with little or no consideration of the conditions of creation and reception. Focusing on the process of subtitle production, occasional studies have touched upon the cognitive processes accompanying it, but no study so far has related these processes, and the resulting products, to various degrees of translators’ competence. This is precisely what this essay does, focusing on the different layers of subtitle translation provided for two different films and in two different contexts. By analysing the first and second versions of subtitle translations, we shall reflect on the acquisition, and application, of different subtitling competences. Subjects: Language & Linguistics; Language & Literature; Translation & Interpretation


Introduction
Subtitling is a form of translation known the world over and yet a general audience's knowledge is limited to the subtitles' bare functionality, preferably actualized in as invisible a way as possible (Bannon, 2009).
As it is the first activity falling within current definitions of AVT to have been the object of scholarly interest 1 , its study still appears to lack methodological sophistication and breadth owing to the fact that it is often confined to case study-based, contrastive analyses. Such studies are indeed frequent, as many students and scholars in a variety of domains enjoy observing what is often perceived as a "familiar" activity. Thus, subtitling, if observed in full circle, moves paradoxically from being universally known by the general audience as an aid to the reception of foreign audiovisual material, to ABOUT THE AUTHOR Elena Di Giovanni is Associate Professor in English and Translation at the University of Macerata. She holds a degree in Specialized Translation and a PhD in English and Audiovisual Translation. She has taught and researched audiovisual translation and media accessibility for over 15 years, giving talks and workshops all over Italy but also in Spain, the United Kingdom, Ireland, France, the USA and Egypt. For the past 9 years she has been Visiting Lecturer at Roehampton University, London, MA Course in Audiovisual Translation. Her main research interests include subtliting, dubbing, access services for television, cinema and live events. She also specializes in research on translation in post-colonial settings and writing as translation.

PUBLIC INTEREST STATEMENT
The translation of audiovisual texts is so pervasive as to be part of our daily lives in many ways. Subtitling, used in more and more contexts and across media, is practiced the world over but its study is relatively recent and partly underdeveloped. This essay aims at promoting a more comprehensive study of subtitling by focusing on the development of translation competences in subtitlers and the impact experience has on the actual production of subtitle translations. Analysing several versions of the same subtitle translation for two films screened at film festivals in Italy, the author's aim is to shed light on the cognitive processes accompanying the creation of subtiles and their correlation with subtitlers' competence and experience.
being perceived as a universally approachable object of research undertaken by the young and notso-young in a variety of fields.
As Perez Gonzalez points out (Perez Gonzalez, 2014, pp. 92-93), the study of audiovisual translation (AVT), which emerged in the early Nineties 2 , derived from the very practice of subtitling, the first observers and analysts being practitioners themselves. Still, as Perez Gonzalez notes, audiovisual translation studies-and the study of subtitling within it-have generally moved reluctantly from practice to full-fledged theorization for a number of reasons which include the difficulty of building homogeneous and comparable corpora, issues of copyright, the often extremely lengthy processes behind the collection, transcription, and analysis of audiovisual translated dialogue (Ibid,p. 94). The move beyond descriptivism (i.e. the contrastive analysis of source and target texts) has so far proved to be equally difficult. And if large-scale reception studies in AVT have already been described as complex and time-consuming (Di Giovanni, 2012), although they are extremely useful for feeding back into methodological and empirical research, studies focusing on the very process of subtitle creation have also been scant. This may be ascribed to the multidisciplinary competences needed for such analyses, which range from psychology to statistics, but it is also likely to be due to the highly diverse contexts involving applications and modes of subtitle creation. Last but not least, the tight deadlines in the subtitling business, coupled with the fact that subtitlers work as freelancers, mostly from home, make these studies all the more challenging.
In the following pages, we shall explore the different skills and constraints involved in subtitling, focusing on the subtitler's competence and how it emerges through the consideration of several layers of the same translation. By analysing two subsequent Italian versions of the same films, we shall reflect upon subtitling competence and what it entails, while also shedding light on contemporary trends in subtitling which point to more linearity with the source text.
After focusing first of all on subtitling, its study and practice, we shall then move on to an examination of cognitive studies on translation and the development of translation competence as posited by several scholars. All this will then be applied to subtitling and to the analysis of two layers of subtitle translations for Go tell the Spartans (1978) and Other men 's women's (1931), whose different versions have been kindly provided by SubTi Ltd.

Defining, learning, and studying subtitling
As Egoyan and Balfour suggestively put it, "subtitles embed us" (Egoyan & Balfour, 2004, p. 30). In their words, subtitles allow us access to the narrative of foreign films, while projecting us into the films themselves.
Besides the cinema, subtitling is with us on a number of other contemporary devices, for an increasingly diversified set of purposes, from educational material to online gaming or DYI videos.
Unique to subtitling is the turning of language delivered orally into a written text, appearing on screen simultaneously with the dialogues, which remain fully audible. Thus, subtitling performs the twofold task of adding a channel of communication and ensuring the co-existence of two languages.
To add yet another suggestive definition, Béhar states that subtitling is "a form of cultural ventriloquism", where "the focus must remain on the puppet, not the puppeteer" (Béhar, 2004, p. 85). While adding a cultural connotation to his definition of subtitling, Béhar also reinforces the commonly shared belief that subtitling has to be as unobtrusive as possible to be effective. On a similar, but less evocative note, Bannon (2009, p. 3) remarks that the success of subtitles "is measured by how little viewers notice them." These statements, although pointing to the functionality of subtitling, evoke the long-standing, reductive notion of translation as inferior to the original (Bassnett & Trivedi, 1999), with the source text remaining untouchable and primary, while the translation is ancillary, strictly functional, and perishable.
The statements quoted above also lead us to the equally long-standing and widely discussed notion of equivalence in translation, which was very interestingly revived by Theo Hermans in The Conference of the Tongues. As Hermans says, "any reminder that the text in question is in fact a translation threatens the assumption of equivalence and tells the reader: oh yes, this is only a translation" (Hermans, 2007, p. 24). As far as Hermans is concerned, translations can claim authority only when they become canonized or, as he says, authenticated: "Establishing equivalence amounts to an act of authentication. A translation thus proclaimed to be equivalent to its parent text ceases to be a translation.
[…] A translation that has not ceased to be a translation cannot be equivalent to its source" (Ibid, p. 27).
By virtue of its unique nature, which implies perpetual coexistence with its original, subtitling can neither claim equivalence nor authority. It remains confined to the role of a relay service, whose invisibility is always primary.
Nonetheless, subtitling is a highly complex, multifarious activity. It implies several stages, requires a number of skills amongst which are linguistic and cultural competences. Moreover, subtitling demands that all the skills and competences noted above are activated simultaneously and with an equal degree of awareness.
Along with criteria and concepts drawn from cognitive studies on translation, the analysis presented in the following sections also focuses on technical, linguistic, and cultural criteria unique to, and ineluctable for, subtitling. These criteria, however, are hardly ever monolithic: technical, linguistic, and cultural features often concur. Let us briefly outline them, drawing inspiration from several studies on subtitling and audiovisual translation in general.
In The Semiotics of Subtitling, De Linde and Kay remind us that subtitling activities are strongly determined by spatial restrictions and temporal restrictions (De Linde & Kay, 1998, pp. 6-7): both these, which the subtitler has to consider and comply with simultaneously, limit the amount of text which can be provided in a subtitle, therefore implying a third, essential parameter: reduction (Kovacic, in De Linde, 5;Georgakopoulou, 2003;etc.). Reduction in itself necessitates linguistic, cultural, and intersemiotic skills, as deletion in the target text of elements present in the source text requires paraphrasing skills, thorough analysis of the multisemiotic textual unit for appropriate deletion, and so on. Another parameter vital for subtitling is segmentation, which again involves linguistic and technical skills and an awareness of spatial restrictions. Creating one or two liners that function as semantic units is one of the most important features of quality subtitling, ensuring smoothness in reading and perception. De Linde and Kay define text cohesion as one more essential parameter to ensure appropriate reception: cohesion within each subtitle and over several subtitles, possibly the entire film. Not too distant from text cohesion is what De Linde and Kay and also Díaz Cintas and Remael (2007) define as language style, i.e. recourse to appropriate linguistic structures and lexical choices in consideration of the film narrative. Moving away from essentially linguistic criteria, Göpferich and Jääskeläinen (2009) refer to the appropriate use of resources, which is indeed an essential technico-cultural feature for subtitling. On the technico-linguistic front, synchronization is certainly a feature that cannot be overlooked: being in synchrony with the images and dialogues is an all-important, universally recognized feature of subtitling (as well as other forms of AVT). To conclude, although many other features could be listed here, we will take into consideration only one more culture-linguistic parameter discussed by Bannon, i.e. the need to understand the context in order to bridge two cultures appropriately. This applies to temporal and geographic features, as well as diastratic and diaphasic variety.
With these parameters in mind, let us move on to an exploration of studies dealing with the process of translation, which will, then, be applied to the following analysis.

Mapping translation competence
Studies focusing on the making of translation, the cognitive efforts thereof, and the acquisition of translation competence have been carried out for over 25 years, although, as previously mentioned, infrequently applied to AVT. This may be due to the relative youth of audiovisual translation studies within TS, as well as the complexity of mapping the processes of AVT making, considering the variety of settings where it is performed and the tight deadlines to which audiovisual translators work.
In translation studies, one of the most comprehensive volumes devoted to the discussion of the process of translation to date is Sharon O'Brien's edited collection Cognitive Explorations of Translation (O'Brien, 2009), whose manifold contributions provide profound insights and empirical evidence of what goes on in the minds of translators while doing their job. Most interestingly, the volume offers new perspectives, through innovative empirical research (corpus and key-logging analysis, eye tracking studies) and sound reference to previous studies. The latter are also duly surveyed in Hurtado Albir and Alves' chapter on "Translation as a cognitive activity" for the Routledge Companion to Translation Studies (Hurtado Albir & Alves, 2009). These resources will form the basis for this section and the ensuing analysis.
As a point of departure, a distinction has to be made between the two main lines of investigation established so far: on the one hand, scholars have focused on the cognitive efforts and the phases involved in such processes, whereas, on the other, attention has been geared towards the acquisition of translation competence. These two approaches have given way to a host of studies that are indeed interrelated since no study of competence acquisition by translators can take place without considering the cognitive processes involved in a translator's job. For the purpose of our analysis, we will only refer to one theoretical approach to translation as a cognitive activity, while focusing at length on the development of translation competence.
One of the most influential and fruitful models for the exploration of translation processes is the so-called interpretive theory of translation (ITT), developed by Seleskovitch and Lederer even before translation studies had been acknowledged as a discipline (see Seleskovitch, 1968, 1975, and Lederer, 1981expanded in Lederer, 1994). The ITT model revolves around three phases, which can be identified in every translation (and interpreting) process, i.e. understanding, deverbalization, and reverbalization. Understanding concerns the generation of sense, as explained in 2009 by Hurtado Albir and Alves in their reprise of the ITT model (Ibid, p. 55). It involves different types of knowledge defined by the ITT theorists, namely linguistic, encyclopaedic, and contextual knowledge. With specific reference to audiovisual translation for subtitles, the latter could be further divided into the knowledge (1) of the context evoked in the audiovisual text to be translated, (2) of the context in which the translation is being created (a film festival, a TV studio), and of course (3) the context of reception. Deslisle added a further step to the three phases noted above, specific to written translation and described as a second interpretation, aimed to "verifying the exactness of the provisional solutions found" (Deslisle, 1988, p. 55). It seems that this phase could be paired with proofreading, which is indeed a second interpretation and a verification of the solutions found by a second translator. Proofreading is central to many translational activities, especially subtiting, and it will be the object of our analysis in the next sections.
Regarding translation competence acquisition, theoretical frameworks and models have been set forth in the years since 1976 (Wills, 1976). Over the years and across theories, a recurrent standpoint has been that translation competence is made up of sub-competences, variously defined and ranging from linguistic to extralinguistic and contextual features. This is all the more so of subtitling, which, as stated above, requires not only the sub-competences referred to as general translation activities in models such as the TransComp model Göpferich, 2008), but additional skills which move from complex technical expertise to the ability to comply with strict space and time limitations, as seen in the previous section.
One model, which seems to be particularly valuable for our study, was proposed by the PACTE group in PACTE, 2003 and subsequently revised and validated. In the report on the validation process (PACTE, 2009), Beeby et al., (alias the PACTE group), defined several dependent variables to be observed against one independent variable, i.e. the degree of translation competence of the participants in their experiments, which they determined through criteria such as years of experience and the centrality of translation as a work activity. The dependent variables, on the other hand, were: (1) knowledge about translation, i.e. the individual's knowledge of the principles of translation and the translation profession; (2) the efficacy of the translation process, which refers to the relationship between the time taken to complete a translation task and the acceptability of the solutions; (3) decision-making, which is the most complex variable and refers to the decisions made during the translation process as a result of a recourse to external support (documentation); (4) the translation project, which refers to the subject's approach to the translation of a specific text and the unit it comprises; (5) the identification and solution of translation problems, i.e. the difficulties encountered by the subjects when carrying out their translations; and (6) the use of instrumental resources, namely recourse to dictionaries, glossaries, encyclopaedias, etc. (2009, pp. 36-37). These variables are all interesting but also complex in nature, since they do not isolate linguistic from extralinguistic competences and they refer to the overall translation process to the overall translation process. However, there seems to be a certain degree of overlap, especially between 3 and 5, whereas, in the PACTE model, 1 is directly linked with the independent variable, i.e. the translator's competence.
For the purpose of our analysis, the independent variable becomes the goal, through a bottom-up approach.
Only variables 2, 5, and 6 will be considered, in a different order and integrated by two extra variables as specified below: (1) Identification and solution of translation problems (2) Awareness of, and compliance with, technical criteria (3) Awareness of, and compliance with, linguistic criteria (4) Use of external resources (5) Overall efficacy of the translation.
Within each of these variables, we shall consider the subtitling parameters listed in Section 2 above.
With reference to the identification and solution of translation problems (variable 1), we will focus mainly on the linguistic or cultural issues reflected in translational choices. These will include the treatment (understanding and translating) of cultural references, linguistic style, and the use of appropriate register, deixis and, more generally, translation errors arising from lack of comprehension. In relation to variable 2 (awareness of, and compliance with, technical criteria), we shall refer to the first two parameters listed in Section 2 above, i.e. spatial and temporal restrictions, amalgamated in the evaluation of the length of the subtitle text and the time available to read it. Synchronization is a further parameter falling within variable 2, although nowadays subtitlers are hardly ever entrusted with this task. Variable 3, which focuses on the awareness of, and compliance with, linguistic criteria, will offer reflections on such parameters as text reduction and segmentation. Variable 4 corresponds to one of the extralinguistic parameters identified for subtitling above and will be analysed precisely in these terms, whereas variable 5, connected with the overall efficacy of the translation, will lead us to reflect on the overall understanding and relaying of the context evoked in the source text and the text cohesion (or lack thereof) emerging from the target texts.

Subtitling competence in novice and experienced translators
As stated in Section 1, although it is generally perceived as inconspicuous, subtitling is a clearly visible part of an audiovisual product, a service whose smooth reception is ensured by its very unobtrusiveness. Yet, every set of subtitles, accompanying every film in cinemas or DVDs, is seen by millions of people, its impact being generally underestimated. And if subtitles are de facto steeped in invisibility, subtitlers are all the more so. It is well known that most audiovisual translation techniques-not only subtitling-involve translators as mere links in a chain (Díaz Cintas & Remael, 2007, p. 81), and that most other stages beyond linguistic transfer generally receive more attention. Moreover, subtitlers hardly ever work alone on the linguistic operations involved in the overall subtitling process. Proofreaders are called into play as interpreters of the main translator's work and as providers of a second interpretation (see Deslisle, above) that moves from the first translator's choices to a definitive version of the target text. This double interpretation-the two, sometimes three layers 3 in the translation of a subtitle-has hardly received any attention in research on subtitling.
Proofreaders are themselves subtitlers, generally recruited by virtue of their experience and expertise in subtitling. The layers of subtitling are, therefore, provided by at least two translators with different translation competences.
In the following sections, we shall try to highlight subtitling competence by looking at the two layers behind the translation of subtitles for two different films. In These two films and their two Italian versions were chosen for analysis by virtue of the radically different degrees of intervention during the proofreading phase. Each table below contains examples from the original script (ST), the first translation (T1), and the second version (T2). As the company, which provided the script and the translations, operates on a pre-segmented script, with the English dialogues already inserted in subtitle boxes, synchronization, a parameter connected with variable 2 above, will hardly ever be mentioned in the analysis.
Excerpts are taken from the first 15 minutes of each film. No information about the subtitlers and proofreaders was requested and obtained until the analysis was completed.

Go Tell the Spartans
A ground-breaking American film, Go Tell the Spartans was dubbed into Italian in 1978, the year of its American release. In more than 37 years, at least a dozen different subtitled versions are likely to have been produced in Italy, but as is often the case, subtitles are reproduced, rather than purchased and reused. The reasons behind the remaking, rather than reusing, of subtitle translations are manifold, mostly connected with issues of copyright and lack of communication/collaboration among the operators working in the audiovisual translation industry. Commissioning a new translation is often far less expensive and time-consuming than locating a previous translation, purchasing and adapting it for new purposes. Interestingly, however, new translations are occasionally inspired by previous ones, especially if these are available on the market on DVDs. Subtitles created for film festival screenings, on the other hand, are very hard to locate, as they mostly remain in the private archives of festivals and subtitling companies. This situation thus confirms the ephemeral character of subtitle translations as opposed to original versions, but also, paradoxically, as opposed to dubbing translations. The latter are seldom the object of complete recreation, so that dubbing is more stable, whereas subtitling is more volatile.
Subtitle retranslations in themselves constitute sets of layers, both interesting and valuable, pointing to a host of translational strategies and cross-relations among the different versions. Although extremely hard to obtain, these retranslations, which bear witness to the genetics of film translation, ought to be made the object of systematic investigation. For the purpose of this study, however, we shall limit ourselves to the analysis of the two, synchronic layers of subtitle translations performed for one and the same purpose, i.e. a film festival screening. In the following pages, we shall comment upon the two Italian versions created by SubTi Ltd. for the Torino Film Festival in 2014. Since the first translation reveals a number of errors and inappropriate solutions, we shall discuss the first and the second version together, referring back to the source text when appropriate. Table 1 presents the source text (ST), translation (T1), and proofread version (T2) of the written information provided in the opening scene. One interesting piece of data which immediately strikes the eye is the length of T1 as opposed to T2: although proofreaders are very often called to apply further reduction to a first version, in this case the second version is longer overall than the first one. Timing for the subtitles allows for longer sentences, so that added explicitness does not hamper readability.
With reference to the identification and solution of translation problems, the excerpt below highlights the correction of translation errors from the first to the second Italian version. Through a process of erroneous over-specification, "a rebellion" is turned into "the civil war," and restored to "a rebellion" in T2. Then, in a reversed process, the sending of "military advisors" is replaced by "the support of American militaries" in T1 and turned again into the sending of "military advisors" in T2. After the reference to the Communists, who are to be fought in the ST, chased in T1 (subtitle 4) and fought again in T2 (subtitle 5), the closing line in the ST gives way to an overall error in translation, whereby a little, confused and faraway war becomes "still ongoing, and without winners" (subtitle 5). This total shift of focus and meaning is redressed in T2, which restores the sense and linearity of the original.
As for awareness of, and compliance with, technical criteria within this passage, the duration-tolength ratio is on the whole acceptable in both versions. Subtitles 1 and 3 in the second version are slightly longer than they are in the first, this choice being probably based on the duration of the two subtitles, which is approximately five seconds, and allows for further explicitness without hampering readability. Adding one subtitle in T2 to rewrite what appears only as subtitle 4 in T1 is, on the other hand, the result of an effort to restore meaning, which is inappropriately rendered in the first translation (the sending of military advisors from the US to fight the Communists is translated as army soldiers sent to chase out the Communists).
Moving onto variable 3-the awareness of and compliance with linguistic criteria-let us focus on segmentation only and the creation of semantic units in T1 and T2. Subtitle 1 contains the first signs of positive amendments to segmentation in the proofread version: the verb "lost" and its direct object "the war" are no longer split in T2, resulting in more coherence not only in line 1, but also in line 2, where "their colonies in Indo-China" becomes a unit and not a piece of information divided into two subtitles. Subtitle 3 in T1 offers a clear example of poor segmentation across subtitles, ending in "with" and thus not providing any unity of meaning.
Regarding variable 4, or the use of external resources in translation, a comparison between T1 and T2 highlights the fact that the first translator has not sought out facts beyond the ST and has turned a rebellion into a civil war, and a war which, in 1964, is still limited in scope into an ongoing war with no winners points to the superficial research-if any-carried out by the translator. A more accurate use of resources is palpable in T2, where more attention to the facts and their linguistic transfer can be clearly detected.
To conclude with variable 5, the overall efficacy of the translation is higher in T2, with this version also pointing to a greater awareness of the coexistence on screen of the written ST and the written translation on the part of the proofreader. Such a case clearly deserves special attention and indicates that skill is likely to be acquired through subtitling experience as it is, in fact, in other types of translation. Table 2 contains the original version and the first and the second layers of the translation for the initial exchange between Major Barker (Burt Lancaster) and one of the local Vietnamese soldiers supporting the American army, called Cowboy by the American soldiers.
In terms of the identification and solution of translation problems (variable 1), the use of appropriate register (language style) and deixis prove particularly interesting in this excerpt. Regarding linguistic style, T1 reflects an unnecessary use of simplified and incorrect English for Cowboy's lines. Since his original voice can be heard while the subtitles are being read, and his grammatical mistakes in English do not convey a meaning to be transferred linguistically, the solution adopted in T2 seems to be the most appropriate. The proofreader has opted for standard English, redressing the grammatical mistakes in T1 and avoiding the oversimplified, childlike register that can be seen, for instance, in subtitle 6 of T1. The translator's choice of broken language to translate Cowboy's line may be due to lack of expertise and specific training, whereas the proofreader's intervention points to familiarity with this particular translation issue. Another proof of expertise can be found in the use of deixis for the revision of subtitle 1 in T2: for further reduction (in a subtitle whose duration is 2.45 s), "out of the water" is replaced by "out of that." "That" stands for the jug and the water, which can be clearly seen on screen. On the other hand, an incorrect use of deixis is made in T1 and redressed in T2 with reference to the ST sentence "put him back in the cage." Subtitle 7 in T1 uses "in his place," where "place" is supposed to stand for "cage." However, since no visual and verbal reference is provided in the ST to refer "place" to "cage," the proofreader has restored full referentiality by using "cell." In terms of awareness of, and compliance with, technical criteria (variable 2), we have already hinted at instances of reduction that enhance readability and the appreciation of the film as a whole. The use of deixis and the deletion of unessential elements ("un bel" in subtitle 3 and "signore" in subtitle 9) allow for the optimization of duration-to-length. Adherence to linguistic criteria in subtitling (variable 3) lead us to note an overall tendency to revise segmentation during the passage from T1 to T2, by avoiding two-line subtitles when they are deemed unnecessary. This is the case, for instance, for subtitles 1, 7, and 8, all of which become one-liners in T2. The choice made by the proofreader in this respect reveals a greater awareness of readability issues and of the cognitive efforts involved in reading two lines rather than one. With regard to variable 4-i.e. the use of resources-this brief excerpt poses no particular problems, as it does not require any specific research. Perhaps a check made on the use of the French word "instamment" would have shown that  it is not in any way connected to the military sphere and could have been avoided in the translation, also by virtue of its not having any specific value and impact on comprehension.
By way of conclusion, variable 5 (overall efficacy of the translation) reveals greater cohesion in T2, especially between the dialogues and the images on screen. Cohesion within the text is ensured by eliminating non-essential and inappropriate grammatical mistakes and child-like expressions (subtitle 6, T1). However, of additional note here is that the proofreader has restored several elements originally deleted in T1 to his/her T2, and not only for reasons of comprehension and more appropriate translation. Indeed, unlike two decades ago, one of the most common trends in professional subtitling today points to less text reduction and more linearity with the source text. Awareness of this trend is yet another proof of more thorough expertise in subtitling on the part of the proofreader.

Other Men's Women
The choice of Other Men's Women and its two layers of subtitles was based upon two main criteria: the inherent complexity of the ST in terms of translation, and the limited amount of intervention in the first version by the proofreader. The complexity of the ST is clear from the outset of the film, with the first lines being chosen as the initial sample for our analysis.
The film was released as early as 1931 and the use of English is remote from contemporary speech and full of puns, metaphors, and allusions. Moreover, as is often the case with films from the first decades of international cinema production, dialogues are frequent and speech delivery rates are extremely high, thus requiring considerable condensation. Since there are fewer proofreader's changes than in Go Tell the Spartans, we shall examine our five variables with more explicit reference to the passage from ST to T1, and we shall then apply them to the transfer from T1 to T2.
Starting with the opening exchange, featured in Table 3, in terms of the first transfer process (ST to T1), variable 1 points straight to a high level of linguistic and translational skills on the part of the translator. S/he has succeeded in appropriately decoding and succinctly relaying expressions such as "hog-wild baby, no fooling" in subtitle 3 and "scramble three and a cup of jamocha" in subtitle 4, to mention but two. The linguistic style-use of register-is appropriate in T1, notwithstanding the extra effort required to transfer a linguistic style that goes back almost a century. As for translation problems and solutions, the reference to the "bun" in ST proves particularly challenging, involving several lines delivered both by Bill and the waitress, as reflected in subtitles 5, 6, and 7 in T1. The use of "bun" in the ST is linked initially to "bread" and "toast," nonetheless evoking a date, a partner. An equivalent solution in Italian is not to be found, as no bread-related word could be made to refer to a partner. Thus, the translator opted for "caffè corretto" [laced coffee], making a connection with the request for coffee Bill has just made, and with the translation chosen for the second line of subtitle   Reference to a partner is made by the waitress in subtitle 7, then repeated and discussed further in the following lines outside the excerpt above. Variable 2, i.e. an awareness of technical criteria in subtitling and compliance with them, is evident in the movement from ST to T1: considering the fast pace of the original dialogue, the translator has striven to condense most lines, while keeping essential pieces of information in each. This is the case for the translation of subtitle 2, whose duration is slightly over three seconds and whose condensation in the translation allows for appropriate decoding through the subtitles. Similarly, subtitle 4 condenses the two lines delivered by Bill and the waitress, deleting a repetition ("scrambled") and also "in a hurry," which is reinstated (and duly translated) slightly later (see subtitle 9). With reference to variable 3 and the compliance with linguistic criteria, T1 shows appropriate segmentation overall. Text reduction, as intimated above, is constantly applied when necessary, as is the case for the text appearing all in one line in subtitle 7. A clear reference to Bill's date and the waitress' jealousy is made in only five words. The use of punctuation is appropriate and only highlights the addition of a few exclamation marks to follow the emphatic delivery made by the two characters. The use of resources, as in variable 4, has indeed been appropriately made by the translator, at both lexical and cultural levels. On the whole, the translation conveyed in T1 appears cohesive and respectful of the context evoked in the film, thus enhancing smooth perception. The second layer of the translation, as displayed in T2 above, reveals minimum adjustments. In terms of translation solutions (variable 1), no changes have been made from T1 to T2 for the excerpt above, the only exception being the reference to "bun," which appears as "caffè corretto" [laced coffee] in T1 and is translated as "focaccia" [pizza bread] in T2. The proofreader has restored the reference to breadlike food, to ensure coherence within the line delivered by the waitress and perhaps to ensure smoother (cultural) reception by the Italian viewers by choosing a familiar item such as "focaccia." However, this solution is no longer coherent with the following line, which leaves T1 unchanged except for the addition of "No" at the beginning of the sentence. Thus, the waitress asks Bill whether he wants bread, toast, or pizza bread, and he replies with an unrelated "No, I drank too much last night." With reference to variable 2-i.e. compliance with technical criteria and especially duration-to-length ratio-T2 in subtitle 2 is more condensed than T1, which proves to be an appropriate solution. On the other hand, T2 in subtitle 4 sees the reintroduction of a repetition (the word "scrambled"), adding considerable extra length in spatial terms ("strapazzate" is an 11 type-word) and could have easily been avoided. All the other subtitles remain virtually unchanged in terms of technical criteria. As for linguistic criteria (variable 3), the proofreader has spelt out basic numbers (one to five), in compliance with the subtitling company's rules. The reduction has been revised, or rather expanded, with the reintroduction of certain elements appearing in the ST: this is the case of "you devil!" in subtitle 6, and "or maybe" in subtitle 5. Segmentation and punctuation remain virtually unchanged in the proofread version.
In terms of the use of resources, nothing significant can be detected in the passage from T1 to T2. This is probably not due to a lack of attention on the part of the proofreader, but rather to the appropriate solutions found by the translator for T1. On the whole, the second layer of this subtitle translation does not yield significant differences from T1, thus preserving the overall efficacy of the first translation (variable 5), with occasional reprises from the ST. Table 4 contains a brief exchange between three characters; an angry landlady who is playing host to Bill White, Bill himself (heavily drunk), and Jack, Bill's friend and colleague. Just as the landlady is kicking Bill out, Jack passes by and rescues his friend.
As with the previous excerpt, we shall first examine our five variables-and the related subtitling parameters-with reference to the first translation in order to then observe the changes made by the proofreader.
Once again, the passage from the ST to T1 reveals a fairly high degree of accuracy in complying with timing, linguistic criteria, and translation solutions. Starting from the latter, as with variable 1, linguistic style comes into play in the choice of the physical reproduction, in the subtitles, of the landlady's stuttering by means of suspension marks. This choice is arguable, since the stuttering remains audible in the ST and the use of suspension marks in T1 becomes excessively visible as well as hampering the reception of the subtitles. On the other hand, the characters' overall register, as well as their humorous lines, have generally been preserved and occasionally replaced with well-known Italian idiomatic expressions. This is the case, for instance, in subtitles 9 and 12, whose idiomaticity in T1 is worth noticing. Where the strong idiomaticity of the ST has not been as strongly conveyed (see subtitle 6), the translation solution is nonetheless acceptable, although weaker. In terms of readability (duration-to-length ratio, the main feature of variable 2), T1 has duly condensed the ST to comply with the timing for each subtitle, although on the whole this excerpt was not particularly problematic in this respect. However, it may be worth pointing out that, in subtitle 4, the translator has opted for text distribution on two lines, perhaps considering the long time available to read the subtitle. Variable 3 points to appropriate reduction and segmentation, although the appreciation of both features in T1 is partly dismissed by the use of suspension marks to indicate the lady's stuttering. Suspension marks are also used for the passage from subtitle 10 to 11, a choice which is only occasionally made by the translator and which could also have been avoided in this case. The use of resources (variable 4) in T1 highlights an appropriate recourse to dictionaries, glossaries, and other tools for the correct decoding of idiomatic expressions no longer in use. To conclude with variable 5, the overall efficacy of the translation, it seems that T1 appropriately relays the context evoked in the film, mainly through an appropriate use of register. The use of suspension marks hampers cohesion, but on the whole the subtitles allow for the appropriate reception of the ST.
Moving onto the second layer of subtitling for this excerpt, the linguistic style and register (variable 1) are indeed the object of revision on the part of the proofreader in T2. The latter deletes all the suspension marks used in T1, those used for the landlady's stuttering as well as those for the passage from subtitle 10 to 11. Only subtitle 3 in T2 retains the marks used for the genuine suspension of the line uttered by Bill, interrupted by the landlady. In terms of translation solutions, the proofreader occasionally reinserts words appearing in the ST, but not in T1 (such as "piccola", "little", in subtitle 9), or substitutes lexical items in T1 by slightly more fluent solutions ("l'acqua dentro" instead of "l'acqua che scorre" in subtitle 1, or "senta" rather than the more formal "ascolti" in subtitle 3). Subtile 6 is paraphrased in T2, although this revision does not imply any meaningful change. On the whole, the solutions adopted in T1 are all maintained in T2, with only minor changes (see "furfante" instead of "canaglia"). As for variable 2, no particular intervention is detected in the second layer of the translation, whereas variable 3 points to interesting revisions worthy of discussion. A slightly more accurate use of punctuation is introduced, mostly visible in the appropriate insertion of 5 exclamation marks in T2 (subtitles 1, 2, 4, 6, 7). Where the unnecessary suspension marks are deleted (subtitle 10), a comma is introduced to accompany the shift from 10 to 11. In terms of segmentation, T2 reveals the unnecessary alteration of the line break in subtitle 11, where "before" is awkwardly moved to the first line, thus hampering coherence. In subtitle 4, the proofreader prioritizes the principle of having oneline subtitles whenever possible, thus dispensing with the translator's wish to provide a two-line subtitle for a long six-second slot. Variable 4 does not lead to any noticeable change, whereas variable 5 points to added cohesion in T2, with the deletion of the unnecessary suspension marks.

Discussion
The analysis of the two layers of subtitling for the two films above has perhaps been excessively detailed, but it has revealed the validity of a bottom-up approach to framing subtitling competence. By looking at the passage from ST to T1, and subsequently from T1 to T2, different stages of subtitling competence, at technical, linguistic, extralinguistic, and cultural levels, have been outlined. In a more detailed, or subsequent study, it may be worth counting the occurrences of positive or negative interventions by the translator and proofreader, and perhaps also further classifying them along a continuum that moves from a very appropriate to a totally inappropriate solution.
In his analysis of translation competence (Chesterman, 1997) defines it as a process of gradual automatization, relating it to Dreyfus' five stages of skill acquisition (Dreyfus & Dreyfus, 1986). These comprise novice, advanced beginner, competence, proficiency, and expertise, with novices normally recognizing only predefined features and rules, advanced beginners being aware of non-defined but relevant features, competent translators being able to cope with hierarchical and goal-oriented decision-making, proficient translators possessing intuitive understanding and taking deliberate action, and expert translators providing a fluid performance and applying deliberative rationality. This classification is indeed useful, as it allows for the appropriate nuancing of competences while catering to the most relevant stages in the acquisition of skill and expertise. If we apply this classification to the two layers of the two translated films above, it would seem plausible to assume that Go Tell the Spartans had a novice translator for T1, whereas the T2 provider seems to fall within the fifth category, that of expert translators. Other Men's Women reveals a slightly more complex scenario, with the T1 translator certainly being proficient and his/her competence also revealing a certain degree of expertise. The proofreader working on T2, on the other hand, is clearly an expert translator, although the fact that s/ he is working on a high-quality first layer of translation leads him/her to make some inappropriate choices, as with subtitles 3 and 4. Although translating and proofreading are two very distinct tasks, it is extremely common in the subtitling industry-and not only in the subtitling industry-for the same professionals to undertake the translations, at least when it comes to proofreading. This job is normally entrusted to senior translators who, however, continue to practise translation.
When the analysis illustrated above was completed, a request was made for information concerning the four translators who worked on the two films, namely their years of overall professional subtitling activity as well as the length of their collaboration with the subtitling company who hired them for the two films. It was thus discovered that T1 for Go Tell the Spartans had, indeed, been provided by a novice subtitler with no previous experience with the company and a very limited record in terms of subtitling activities on his/her CV. T2, on the other hand, had been entrusted to one of the two most experienced translators within the company, with an overall 25 years of subtitling activity and considerable experience as a proofreader. As for Other Men's Women, T1 had been performed by a subtitler with 10 years of experience and a host of other collaborations in addition to this company, whereas T2 was the work of the same proofreader who had worked on Go Tell the Spartans. This last piece of information is particularly interesting and leads us to two hypotheses concerning the work of proofreaders. On the one hand, they seem to feel compelled to leave traces of their work even if a translation has been performed by a proficient subtitler, which explains the occasionally irrelevant, or slightly inappropriate revisions. On the other hand, proofreaders always show a more thorough knowledge of the guidelines and requirements of the company they work for, so that their revisions very often reveal a thorough compliance with them. However, compliance with company requirements seems to go hand in hand, with an overall tendency to remain loyal to the source text, avoiding deletions as much as possible and opting for greater linearity (see Subsection 4.1 above) in subtitle translation, an aspect which certainly deserves further investigation.

Conclusion
As Díaz Cintas and Remael have noted (see Section 4), subtitlers are mere links in a chain, but if isolated and analysed in depth, the translation link itself also reveals a chain, a sequence of two or more layers. An examination of the layers involved in the translation for subitles-supposing that we are able to obtain them at all-is useful, in that it sheds light on subtitling competence, how it is acquired and applied. Although a quantitative analysis within the framework outlined in this essay is hardly conceivable, further comparable studies are certainly to be encouraged, as is a more systematic classification of the parameters presented here within each variable.
Moreover, in order to enhance genetic studies with reference to subtitling further, research ought to be undertaken concerning the retranslations of the same films from a diachronic perspective. Although extremely difficult to find, these subsequent subtitle translations would contribute significantly to shedding light on changes in linguistic use, extralinguistic criteria, and, above all, in terms of loyalty to, and linearity with, the source text.