Reappraising verbal language in audiovisual translation : From description to application

Focussing on the verbal code alone in audiovisual translation research is often criticised as it supposedly disregards the semiotic and cultural complexity of the audiovisual text. This article by contrast argues for the relevance of an in-depth analysis of the linguistic component of the multimodal complex. First, the article presents a model of key dimensions necessary to account for the space occupied by the language of dubbing, while placing it within a wider sociolinguistic context. Three pairs of dimensions are proposed: naturalness and register specificity, target language orientation and source language interference, and routinisation and creativity. Second, the article argues that translation for dubbing can be applied as a heuristic device to explore the translation of casual conversation. By analysing the rendering of a typical structure of conversational English in dubbed Italian, an illustration is provided of systematic crosslinguistic correspondences that potentially extend to spontaneous speech. Further sociolinguistic, diachronic and cross-cultural investigations are suggested as a way to push research into the language of dubbing forward.


Introduction
The Canadian-French film Juste la fin du monde (Dolan, 2016) tells a story of life and tragedy by zooming in on dialogue, images and sound.The verbal and the visual are co-protagonists, together with the pressing music, perfectively rendering the tension towards the main character's announcement of his looming death to his family.His sister-in-law is alone in attempting to break the wall of the unsaid.Her elliptical, unfinished and hesitant questions: Combien temps?, Combien temps ave-?, Combien temps?-equally elliptical in the English subtitles 'How much time?','How much time?','Will you?' -are the only verbal hints to the man's predicament.Verbal language is there, essential in the film's architecture to talk about daily life, family memories, conflicts and old grudges, at one level, but at another level the dialogue indirectly asks why the protagonist has come to visit his family after many years of absence; and by doing so, it dramatically expresses the anguish of the unspoken truth.
The aim of the present contribution is to reappraise the role of verbal language in audiovisual translation (AVT) by recognising the language of dubbing as a fully legitimate object of research and proposing a model to describe and account for it in a principled way.The application of dubbing research to the broader field of spontaneous spoken language offers further justification for the study of this translated language.In the remainder of the article, section 2 provides the general background against which dialogue takes central stage in audiovisual (AV) products.Section 3 presents a model of key dimensions of orality in AVT that moves beyond a reductionist and loss-oriented conception of dubbing.Section 4 explores how dubbing translation can be profitably applied as a heuristic device to study the translation of casual conversation, with subsection 4.4.2.illustrating the specific case of a typical spoken language construction.The conclusions briefly round up the main arguments of the article.

The background
After years of surprising neglect, AV dialogue and its translation have recently given way to a remarkable outpouring of scholarship, a wealth of topics and methodological approaches.As film dialogue is made up of highly contextualised language, it is necessary to consider the integration of all semiotic codes in the multimodal complex so as to achieve an ecologic account of how AV dialogue develops both diegetically and extradiegetically (e.g., Chaume, 2012;Pérez-González, 2014;Taylor, 2013).With multimodality coming to the foreground in AVT Studies, researchers have addressed themes such as space, characterisation and AV cohesion (Baumgarten, 2008;Guillot, 2017;Zabalbeascoa, 2012) to disclose not only how meaning may be lost during the transfer process but also effectively recreated thanks to the semiotic richness of film (Pérez-González, 2014;Taylor, 2013).For example, the shift from formal to informal pronouns (Meister, 2016;Pavesi, 2014Pavesi, -2015) ) strategically conveys greater emotionality through intersemiotic explicitation (Perego, 2009), whereby contextual cues and nonverbal behaviour in the original version are expressed linguistically in translation.Intersemiotic explicitation occurs, for instance, when a change in voice quality and volume in the original version is reproduced in translation accompanied by a change in grammatical form.Extra layers of meanings are also added in translation by exploiting interactional troping (Agha, 2007), which arises when two performed signs belonging to the verbal and nonverbal semiotic complexes "index models of social relation that are mutually inconsistent" (p.26).This occurs, for example, if respectful address is used when dubbing a scene of physical or psychological abuse (Pavesi, 2014(Pavesi, -2015)).
The attention to multimodality and film semiotics, however, represents one end of the continuum of research centred on the language of the screen.At the other end, it may be argued, the verbal code can be investigated by focussing on specific subsystems or language items, hence bringing to light salient features of the dialogue itself.This approach to research is often criticised within mainstream AVT research owing to its emphasis on the verbal component of film, vis-à-vis both the semiotic complexity and cultural embedding of the AV text.The contrast is evidenced in the following quotations by two leading figures in AVT research.
It is […] paradoxical that the dominant research perspective is linguistic, though AVT is actually a multisemiotic blend of many different elements such as images, sounds, language (oral and written), colours, proxemics and gestures.(Gambier, 2008, p. 11, emphasis added) [The] apparent lack of more academic contributions with a focus on the cultural angle of AVT is perhaps one of the many paradoxes in this field, […] given the wealth of cultural information conveyed by them and the fact that the linguistic fabric is only a part of the whole semiotic composite.(Díaz Cintas, 2012, p. 281, emphasis added) The linguistic approach to research on AV language can nonetheless be justified for reasons of feasibility, autonomy of the object of investigation, and social impact.Firstly, as for feasibility, there should be no objection to research that proceeds on the basis of the methodology available at the time and asks research questions that can be readily operationalised, leaving to more powerful theoretical and methodological approaches the task of bringing together and systematising previous results.That is, research on a complex and many-sided phenomenon such as AVT can benefit from the tackling of individual components to be later integrated into the larger picture.Secondly, the language of dubbing deserves to be studied on its own as a variety within the target repertoire, and with reference to other registers of the same language such as the language of original dialogue, the language of other AVT modalities, or the language of dubbing from different source cultures.The languages of dubbing should also be compared across different target languages and cultures, a crucial research avenue that is still under-investigated (but see for instance the comparative study of dubbing into European and Canadian French in Mboudjeke, 2016).From the complementary perspective of subtitling, Guillot (2016) argues that subtitles make up a text within the text and brings attention to their "meaning potential as text, that is [to] subtitles' capacity to mean on their own terms, as a necessary step for developing a better understanding of their interplay with other meaning-making resources from the broader semiotic context" (p.607).The idea that subtitles follow their own logic and their own agenda vis-à-vis original dialogues was already put forward by Remael (2003), in this way supporting the view that the verbal component of AVT merits close scrutiny.When treating telecinematic language per se, several approaches to analysis are available and contribute to a better understanding of how dialogue works.They include engaging with the sequential dimension of verbal language, seen as motivated series of connected utterances as with Conversation Analysis and Appraisal Theory (Pérez-González, 2007, 2014).They also comprise the examination of how the dialogue sprawls vertically from beginning to end within the AV text, developing its own norms during the process (Guillot, 2016).The language of AVT can also be addressed by looking at individual words and expressions repeated across different texts to identify registers, genres and ideology within corpus-based approaches to AVT (Baños, Bruti & Zanotti, 2013;Mouka, Saridakis & Fotopoulou, 2015;Pavesi, 2016a;Levshina, 2017, among others).Thirdly, the language of AVT is worth investigating in itself due to the strong social and sociolinguistic impact it exerts on viewers.Together with subtitling, the language of dubbing offers significant linguistic and sociolinguistic models to target audiences (Motta, 2015) and may trigger language change as a pervasive and influential form of language contact: "TV texts spread more widely and deeply than […] other kinds of texts because they enter into people's private space and often coexist with activities of everyday life, so the audience is more vulnerable, less concentrated, and more susceptible to interference" (Motta 2015, p. 954).It is in its social function that the role of the language of AVT can also be appraised as a means to foster first language literacy and metalinguistic awareness along with second language learning, not only in formal educational settings but also via naturalistic exposure through subtitled as well as dubbed media (Caruana, Ghia & Pavesi, 2014;Ghia & Pavesi, 2016;Motta, 2017, among others).

Focussing on the language of dubbing
Different perspectives have been taken on the language of dubbing, a kind of fictive orality (Brumme & Espunya, 2012).Mostly relying on a reductionist and loss-oriented conception of AVT, this variety is typically assessed with reference to its source language dependence, relative formality, lack of idiomaticity, repetitiveness and sociolinguistic under-characterisation.As a result, many definitions of dubbed language, or 'dubbese', have resorted to negatively connoted expressions such as fake, artificial, anti-realistic, stereotyped orality or language.However, if we move away from the initial view of telecinematic language − both translated and nontranslated − as inauthentic orality, and overcome the dichotomous conceptualisation of screen language being set between the written and the spoken poles, greater attention can be devoted to a sociolinguistic view of telecinematic discourse as a fully-fledged language variety (Locher & Jucker, 2017a).In this perspective the language of dubbing becomes an autonomous space that can be investigated analytically.
Since with dubbing the film's verbal message is meant to reach viewers as spoken language, orality is a central concern to this translation modality.A few variables or dimensions appear to be most relevant to our understanding of the dynamics underlining the representation of spokenness in AVT.Pavesi (2016b) put forward three pairs of complementary dimensions − naturalness and register specificity, target language orientation and source language interference, and routinisation and creativity.These are conceptualised as structured in pairs of poles of attraction that define intersecting planes of variation and delimit areas of overlap.The acknowledgement of multiple intersections and overlaps is needed to account for each language of dubbing and relate it to other neighbouring forms of fictive orality, such as non-translated AV speech, the language of subtitling, the languages of dubbing from different source languages into the same target language, as well as from the same source language into different target languages or national varieties.The same constructs can also be exploited to compare the language of dubbing with the reference norms of spontaneous conversation both in the original and the target language.It must be finally observed that the intrinsic variability of dubbed languages suggests a diachronic evolution in the norms that underlie their makeup.These changes across time can be studied by analysing redubs in the same target language, as shown by Zanotti (2015) working on a corpus of retranslations, and by constructing diachronic corpora of comparable AV products released at given time intervals.By drawing on such corpora, Minutella (2015Minutella ( , 2017) ) and Motta (2015) investigated source language interference in films and TV series across time.

Naturalness and register specificity
The opposition between naturalness versus register specificity defines the major axis of description and variation in telecinematic language.Naturalness as the degree of conformity to the spontaneous use of the spoken language can be contrasted to the register specificity of dubbed language as an independent variety.It can be viewed as a synonym of realism, although the two terms do not perfectly overlap, with realism being often interpreted as a culturally-embedded and subjective notion, dependent on the agreement among viewers belonging to a given community at one historical moment.Realism in this sense is bound to audiences' habits and expectations and naturalness refers to what is idiomatic or conventionally approved by viewers, who as native speakers recognise a linguistic choice as fitting in a given sociopragmatic situation (Romero-Fresco, 2009, 2012).Interestingly, Antonini andChiaro's (2005, 2009) and Bucaria's (2008) research on viewers' attitudes and perceptions towards Italian dubbed TV programmes showed that professionals and academics were generally stricter in their assessments of how probable linguistic features of dubbese were to occur in daily use.By contrast, ordinary viewers, and in particular younger viewers, had a rather lenient attitude towards typical English calques.This is presumably because of the familiarity they had acquired over the years with those expressions which, although unusual by everyday standards, were typical of dubbed language.
Realism can be thus investigated empirically through reception studies of viewers' responses and reactions to translated products, their mental decoding and understanding, along with repercussions, or viewers' attitudes and sociocultural consequences of AVT as extensively debated in Di Giovanni and Gambier (2018).Naturalness in the strict sense, on the other hand, is liable to analysts' quantitative inspections of phonological, lexical, syntactic and pragmatic features of AV language.For this reason, the major studies on the naturalness of dubbed languages, including Italian, Spanish and Catalan, all share a comparative, corpus-based methodology (e.g., Baños, 2014;Freddi & Pavesi, 2009;Marzá, 2016;Zamora & Alessandro, 2016).
But what exactly does naturalness entail?If we start from the assumption that conversation is the baseline to which screen dialogue should be compared, a first aim of research is to find out to what extent and through which linguistic devices translated AV dialogue aligns with spontaneous conversation.Baños (2014) investigated a wide selection of spoken features belonging to the various language levels in a few episodes of the Spanish dubbed version of Friends (Krane & Kauffman, 1994-2004) and the domestically-produced Spanish sitcom Siete Vidas (Garcia Velilla, 1999Velilla, -2006)).The results of her corpus analysis suggest that dubbed language overall is further removed from colloquial conversation and is hence 'less natural' than the language of domestic, nontranslated target language products.To ascertain the degree of alignment between dubbed Italian and spontaneous conversation, Pavesi (2008Pavesi ( , 2009) ) compared the frequency and functions of subsets of key conversational features including personal pronouns, marked word orders, connectors and spoken language subordinators in a parallel AV corpus and reference corpora of the target language.The findings substantiated a good degree of overlap between spontaneous speech and dubbed Italian for most of the syntactic features investigated.By examining a set of pragmatic markers including de veras (really) and veamos (let's see) in dubbed Spanish, Romero-Fresco (2009, 2012) conversely showed that the language of dubbing for this group of pragmatic features shifts towards the written and formal pole of register variation, hence exhibiting traits of unnaturalness in comparison to non-translated AV dialogue.Overall, results for the dubbed languages investigated so far reveal a variable degree of alignment between the language of dubbing and spontaneous speech.
But does AV dialogue need to mirror conversation faithfully and fully in order to sound realistic and plausible, or is a looser similarity enough for viewers to experience orality?Guillot (2012) in her application of Roger Fowler's (2000) Cognitive Theory of Mode to subtitling and Pavesi (2008Pavesi ( , 2009) ) looking at dubbing have both argued in favour of a selective view of speech representation or mimesis in AVT, in that just a few features in AVT, named "privileged carriers of orality" by Pavesi (2009), can be enough to allow access to the mental model of the appropriate mode or register and trigger the experience of orality in viewers.Similarly, Tomaszkiewicz (2001) has explained that viewers as well as subtitle readers reconstruct missing parts of the ideal, intended text by activating their pre-existing competence as conversationalists.
Selection is therefore constitutive of AV registers and, as ultimately resulting from the economy constraints of the medium (Valdeón, 2017), should not be regarded in terms of reduction or impoverishment.This part-for-the-whole or metonymic understanding of naturalness, however, leads us to the question of exactly which features are entrusted with orality in dubbing.Hence the significance of thorough mappings of spoken language features across language levels.In a recent study on naturalness in Valencian dubbing, Marzá (2016) discovered that there is an uneven distribution of target features expressing intensification in dubbing, which quantitatively displays more differences than similarities with colloquial conversation.However: Some specific intensifying mechanisms evoke the colloquial register and naturalise the dubbed text, such as the use of superlatives and diminutives, the semantic fields in the intensification of modality, the variety and colloquialness of phraseological units, the preferred use of molt and tot as simple modifiers (p.273, emphasis added).This field of enquiry can be widened by asking whether features evoking orality vary within the same language and culture according to source language, and in different target languages, or are shared across different dubbed languages.
Register specificity is the opposite side of the picture in the relationship between the language of dubbing and spontaneous spoken language.With dubbed language being generally described as further removed from spontaneous spoken language than non-translated dialogues (e.g., Baños, 2014;Marzá, 2016;Rossi, 1999), all the phonological, morphological, syntactic, lexical and discoursal tendencies that distinguish it from spontaneous conversation can be used to characterise it as a selfstanding register.Crucial to the distinctive characterisation of dubbed speech is intonation.Drawing on a corpus of audio files from the series How I Met Your Mother (Bays & Thomas, 2014-2015) dubbed into Spanish, Sánchez Mompeán (2017) successfully isolated the intonation patterns that set dubbing apart from spontaneous speech − elongation of sounds, syncopated rhythm caused by pitch fluctuations, tense and precise articulation, emphasis and variation in tempo −, hence empirically substantiating the claim of unnaturalness attributed to suprasegmentals in AVT.Register specificity can in fact be viewed as including a degree of (un)naturalness or conventional orality.From this perspective naturalness is a different construct from 'similarity to spontaneous spoken language' and applies to the perception of the language of dubbing as a kind of language use in its own right, whose plausibility should be assessed against what viewers expect dubbing to be like based on their previous experience.Along similar lines, Travalia (2012) argues that the reproduction of source models through supposed sociolinguistic equivalents would be disfavoured by target audiences, while the artificiality deriving from the neutralisation of social and geographic varieties leads to greater acceptance by the audiences in that sociolinguistic neutrality better matches the visual foreignness of the imported AV product.

Target orientation and source language interference
The link between naturalness and target orientedness is close, although the two notions do not overlap.Translated dialogue may be different from spontaneous speech and still come near the norms of the target community, notably with reference to the language of domestic products.The extent of target orientation or domestication and, complementarily, the extent of source language influence may also typify different media − with TV series being described as more subjected to interference phenomena than films (Motta, 2015;Pavesi, 2008).At a syntacticpragmatic level, an interesting case of overlap between target orientation and naturalness in Italian is given by marked word orders, including right and left dislocations and cleft clauses: the language of dubbing polarises towards the norms of both spoken Italian and Italian films, while it mainly shuns the patterns found in the Anglophone original texts (Pavesi, 2005(Pavesi, , 2008(Pavesi, , 2009(Pavesi, , 2016a).Motta's (2015) recent diachronic study of AVT in paleoTV versus neoTV programmes in Italy (before and after 1976) confirms the special status of marked word orders in Italian dubbing, which was already evident in films translated during the Italian neorealist period (Rossi, 1999) and in films and television series from the late eighties and early 1990s (Pavesi, 2005).Right dislocations of the type Ce l'ho già una scopa ('it I already have a broom' < I already have a broom) are systematically added to the target texts where they introduce a pattern of the informal spoken language, also exploited in domestic Italian productions (Motta, 2015;Rossi, 1999).
In translation criticism and AVT research, an ambivalent attitude prevails towards the role the source models play in the language of dubbing.If the levelling out of the original cultural matrix is emphasised as an undesired result, Anglicisation -as the import of Anglophone modules of speech -is equally criticised.Yet, AV texts are inherently source-bound and interference-prone.There are different reasons that account for the intrinsic permeability to features of the source language and source texts in dubbing.These reasons are both internal, having to do with the architecture of represented conversation on screen, and external, as growing working time constraints are associated to more than optimal reliance on the source text.Due to the structure of telecinematic discourse, cross-comparisons are performed at all the production stages of dubbing (e.g., Paolinelli & Di Fortunato, 2005).Turns are pre-assigned and visible on screen, length and timing of speaking must be reproduced, nonverbal emphasis must be considered, and some lip-synchrony must be allowed in the target text (Chaume 2012).As Toury (1995/2012) already pointed out, a piecemeal organisation of the translation process brings about more visible traces of interference in the target text.The turn-by-turn organisation of AV dialogue is hence biased towards carrying over source language features as well as avoiding natural target mechanisms that do not have direct equivalents in the source language.The latter phenomenon was discussed in relation to the Unique Item Hypothesis (Tirkkonen-Condit, 2002) but has been appropriately labelled calques of absence by Marzá in her study of naturalness in Valencian dubbing (2016, p. 274).
Lexis and phraseology have been universally found to be open to source language transfer in the form of semantic, structural, pragmatic and frequency calques in dubbing (cf.Minutella, 2015 for a recent overview).Calques appear to occur more readily than loanwords (e.g., Brincat, 2000;Alfieri, Contarino & Motta, 2003) as they disguise more easily in agreement with the domesticating tendencies of dubbing, and are still widespread in spite of the reported greater vigilance against obvious structural transfer in contemporary AV products (Bruti & Zanotti, 2018;Motta, 2015).At the same time, direct borrowing appears to be on the increase, in particular in AVT into Italian of Anglo-American feature films, animated films and TV series aimed at young audiences (Minutella, 2017).Both calques and loanwords offer readymade solutions to isochrony problems.

Routinisation and creativity
The influence of the source language may give way to repetitiveness in the dubbed text through translational routines, that is "recurrent solutions to translation problems which tend to become overextended" in time (Pavesi, 2008(Pavesi, , p. 94, 2018)).Routinisation in dubbing also derives from the prefabricated nature of AV dialogue (Chaume, 2001, p. 81;Tomaszkiewicz, 2001) along with the formulaicity and predictability of Anglophone filmic speech (Freddi, 2011;Taylor, 2008), when fixed, reiterated patterns in the source language are translated with fixed, reiterated patterns in the target language.That is, formulaicity is reproduced in the dubbed versions when some English conversational formulaic expressions are translated with recurrent phraseological calques in dubbing.In this way, several calques have become translational routines in Italian dubbing, including: sì signore < yes sir, no signore < no, sir; posso aiutarla < can I help you?; è tutto < that's all ; fine della storia < end of story (Motta, 2015;Pavesi, 2005Pavesi, , 2016b)).
But routinisation and formulaicity are only one side of the coin; their counterpart, creativity, a more elusive concept, has also attracted attention in dubbing research (Ranzato, 2010(Ranzato, , 2011)).As a problem-solving process, creativity is most visible when an obstacle is stumbled upon, both locally and generally in the areas where mismatches occur cross-linguistically and crossculturally.The rendering of the typically intermingled regional and social dialects, for example, calls for creative solutions by dubbing professions if the dominant levelling out is to be avoided, as exemplified in Minutella's (2016) analysis of the creative domestication of the animated film Gnomeo & Juliet (Asbury, 2011) dubbed into Italian.Importantly, creativity can shift from an individual act to a social process, in this way moving from innovation to established norms.As for German dubbing, for example, Queen (2004) has shown that a distinctive street culture language spoken by urban youth is creatively and yet systematically used to translate the speech of young African-American male characters.

Audiovisual translation to explore the translation of spontaneous spoken language
As the language of dubbing undergoes multiagency mediation processes, its translational facet makes it inappropriate for the investigation of either spontaneous spoken language or fictive orality tout court.As made clear in Corpus Linguistics and Corpus-based Translation Studies (e.g., McEnery & Xiao, 2007), this is due to the limitations of any description of language X or contrastive study of languages X and Y, where language X is translated, language Y is non-translated.Such contrastive studies take great advantage from comparable corpora of original languages, spontaneous or fictive, as in Zago's (2018) comparative study of 3-and 4-grams in English and Italian original film language carried out on the comparable component of the Pavia Corpus of Film Dialogue (Pavesi, 2014).
The systematic investigation of the language of AVT, and of dubbing in particular, can however be of significant use as a heuristic procedure for a different goal, namely exploring the translation of spontaneous conversation, i.e. informal, unguided interactions in various settings (at home, at work, in cars, etc.).This largely underexplored application of research on AVT has great potentials for both translation and interpreting studies and nearby disciplines such as linguistics, applied linguistics, language learning and teaching.As spontaneous conversation is the basic type of humans' verbal communication, the need is there to gain information about the transferability of its key features from one language to another.Yet, in its informal, impromptu and phatic realisations, conversation typically remains untranslated.Admittedly, there are approximations to the translation of casual conversation.In television interviews with foreign guests, for example, a seemingly spontaneous verbal exchange is staged for the benefit of the non-participating audience.Nevertheless, several factors limit the approximation of this genre to conversation.Among them, we can mention the lack of synchrony in the interpreter's intervention, which is necessarily detached from the original language production, and the interpreter's focus on the referential rather than the phatic aspects of communication.Moreover, only one of the two interlocutors is usually translated, while the unstructured, two-way interaction found in natural conversation is clearly not staged in these cross-linguistic encounters.More recently live TV programmes (e.g., dog-training programmes, cooking programmes) are being dubbed.They offer some overlap with conversation and could be used for the study of its translation, limitedly however to language-in-action, one genre of spontaneous spoken language where speech is in support of what people are doing with reference to the physical entities present in the situational context.These instances of media communication bring us close to the dubbing of AV dialogue.This translation modality entails the substitution of the original soundtrack with an equivalent target language soundtrack in films, TV series, soap operas, etc., that is all those products that are structured around various forms of fictive dialogue.Following Richet (2001), it may be contended that cinema and television offer a privileged means for the study of features of spontaneous spoken language in translation.Films and television series represent a natural and matchless source of parallel texts, in which the target language dialogue is expected to reproduce the modality of delivery as well as the content of the source dialogue (Tomaszkiewicz, 2001).In AVT, therefore, we are given the possibility of examining, turn by turn or utterance by utterance, what can be considered functionally equivalent expressions in the source and target texts.Also, texts are anchored to fixed scenes representing locations, situations, movements, gestures, gazes, etc., a whole set of semiotic signs that do not change when the language changes (Chaume, 2004(Chaume, , 2012)).Dubbed films hence offer unique opportunities for the synchronous reproduction of verbal-nonverbal combinations in invariant, although culturally not equivalent situations.

Representing conversation
The use of AVT data to explore the translation of conversation depends on the suitability of AV dialogue to investigate conversation.Several authors have in fact issued a call to use the language produced by filmmakers and heard on TV as a source of reliable data for linguistic analysis.Murphy back in 1978 highlighted the sociolinguistic richness of AV language, whereas Lakoff & Tannen (1984) shifted the attention to the artistic realism of film dialogue.While analysing conversational strategies in Ingmar Bergman's film Scenes from a marriage (Bergman, 1973), they argued that "there is much to be gained by looking at artificial conversation first, to see what these general, unconsciously-adhered-to assumptions are; and later returning to natural conversation to see how they may actually be exemplified in literal use" (p.323).In a similar vein, having investigated the TV series Star Trek (Roddenberry, 1966(Roddenberry, -1969)), Rey (2001) suggested that AV dialogue represents a valuable indicator of how conversation is perceived, whereas Alvarez-Pereyre (2011) discussed film dialogue as a linguistic specimen, that is, language data that can contribute to our knowledge about language.While ultimately espousing a position in favour of the essential distinction between non-fictional and fictional texts, Locher and Jucker (2017a) present the borders between the two as fuzzy and slippery: "there is nothing in the syntax or morphology of a sentence or in the choice of vocabulary which systematically differentiates between fictional and other texts" (p.4).Importantly, if AV dialogue exhibits register specific features (see the recent contributions in Locher & Jucker 2017b), and is characterised by few disfluency phenomena, increased discourse immediacy and reduced vagueness (Bublitz, 2017;Levshina, 2017;Quaglio, 2009), as well as increased emotionality and conflictuality (Freddi, 2011;Bednarek, 2010Bednarek, , 2012;;Levshina, 2017), quantitative corpus investigations have confirmed that it overlaps to a considerable degree with spontaneous conversation (Forchini, 2012;Quaglio, 2009).More recently close alignment has also been established between spontaneous British and American conversation and the language of English subtitles (Levshina, 2017), once again showing that English AV conversation shares high involvement, interactivity, and informality with face-to-face conversation.

Dubbing as a discovery procedure
Since research has brought to the fore the sociolinguistic and pragmatic relevance of the data coming from the screen, provided due attention is paid to AVT constraints, we can suggest that features performing a key role both in speech and AV dialogue can be usefully examined in dubbing to bring to light pragmatic correspondences across spontaneous spoken languages.Just like the "representation of speech in fiction (whether in the form of a play or dialogue in prose) has served as one of the best proxies for actual, historical data" (Denis & Tagliamonte 2017, p. 554), dubbing translation to some extent can serve as a surrogate for the translation of casual conversation in everyday contexts.A few phenomena that correlate with the social and situational parameters of impromptu communication have in fact been profitably addressed in dubbing research.They include: questions, as related to turn-taking and the interactivity of face-to-face conversation (Ghia, 2014); politeness in speech acts owing to the social and interpersonal nature of spoken language (Bruti, 2013); deictic expressions as linked to the contextualisation of speech in space and time as shared by interlocutors (Pavesi, 2008(Pavesi, , 2009)).

A short illustration
To round up the arguments in favour of verbal language and fictive orality as central concerns in AVT research, a short illustration is provided of how dubbing can shed light on the translation of discoursal-syntactic structures of face-to-face conversation.Among different types of speechrelated focussing strategies, demonstrative clefts of the type That's what I see; That's what it means; That's exactly what Tom said; That's why I'm here, deserve special attention due to their premium role in spontaneous spoken English, where they occur quite frequently and have a formulaic character (Calude, 2008).In speech, demonstrative clefts highlight relevant segments of discourse and perform a summative, conclusive function while often introducing a gloss or an explanation.They are thus very useful in assuring textual cohesion and linking up interlocutors' turns together.In translation, they are interesting features from a cross-linguistic perspective, in that formal equivalents rarely occur in other European languages including German and Italian (Johansson, 2001;Miller, 2006).Working with the Pavia Corpus of Film Dialogue, a parallel corpus of naturalistic dialogues (Pavesi 2014), Pavesi (2016a) found that demonstrative clefts in English filmic speech align in both formulaicity and frequency with the same constructions in real life conversational Englishabout 90 occurrences per 100,000 words.They also share the same functions as they draw attention to portions of text or context, summarise and conclude sections of discourse, and make explicit what was implicit or unsaid.These functions are illustrated in the extract below (Example 1), where the first cleft has a highlighting, cataphoric function, the second one is conclusive and the third one is both conclusive and explanatory.The naturalness of AV dialogue prepares the ground for exploring how these conversational structures are transferred into another language.
Interestingly, only four English demonstrative clefts were calqued into Italian (Pavesi, 2016a).By contrast, the highest percentage (40%) of the 215 demonstrative clefts in the English dialogue were translated with more natural it-clefts equally headed by a demonstrative (Example 2).
These results uncover viable pragmatic equivalents in a spoken language that does not have the same structural choices as English.
Other strategies came to the fore for the translation of English demonstrative clefts in Italian (Pavesi, 2016a).Apart from omissions and reformulations, which circumvent the translation problem and account for 19% of all translation solutions, another frequent translation pattern was given by unclefted, one clause utterances containing a demonstrative.The occurrence of these structures with both unmarked (Example 3) and marked word orders (Example 4) accounted for a high percentage of cleft translations (25% of all instances).Importantly, these single clauses confirm the key role of deixis in spoken language, while uncovering other syntactic strategies by which it can be effectively transferred across languages.More specifically, utterances exhibiting marked word orders and a demonstrative pronoun represent a notable subpattern which couples two focussing strategies together: syntactic reordering of constituents and a strong form of reference.Syntactic reordering may include both the fronting of the demonstrative pronoun, as in the example (4) above, and dislocations, as in the example below (5).What should be noticed here is that Italian can just use reordering of words to give prominence to discourse entities when English resorts to demonstrative clefts.Finally, translations also resort to focussing particles to render the highlighting function of English demonstrative clefts (13% of all instances), in agreement with the universal tendency whereby these devices are productive in languages where clefts are not available or rarely occur (Miller, 2006).However, many of these particles are deictic in the dubbed renderings, starting with ecco, 'here'/'there', used as a presentative in spoken Italian (Example 6): translation for dubbing confirms the primacy of deixis in face-to-face communication and shows the way it can be transferred from the source language by taking advantage of specific devices of the target language.
In this short illustration of how translation for dubbing can unveil cross-linguistic correspondences pertaining to spontaneous conversation, a spoken English construction was chosen that is employed in AV dialogue similarly to real life conversation.It was shown how demonstrative clefts in English are systematically rendered in dubbing via a few Italian strategies that transfer the focusing functions of the original constructions.Not all aspects of spontaneous spoken language are amenable to such analysis and care must be taken in the interpretation of dubbing patterns whose frequency or infrequency may be motivated by synchronisation constraints.With this provision in mind, data from dubbing can furnish valuable indications of resources available in target language.

Conclusions
This contribution has presented a reappraisal of the spoken component of AVT, an object worthwhile of investigation in itself for reasons of feasibility, autonomy and social impact.Once we overcome the reductionist conception of AVT language as artificial orality, or a midway blend between speaking and writing, dubbing can be examined systematically drawing on a model that accounts for its complexity.The model proposed in this contribution builds on three pairs of complementary dimensions − naturalness and register specificity, target language orientation and source language interference, and routinisation and creativity.Although most investigations on the language of AVT focus on the construct of naturalness, all the dimensions identified in the proposed model are necessary to account fully for individual dubbed languages.The need is there, moreover, for studies on diachronic variation, different AV registers as well as a wider array of dubbed languages in Europe and beyond.As for applications, AVT can also be capitalised on to gain insight into the translation of casual conversational language.Provided attention is paid to the intrinsic limitations of translated and non-translated fictive orality as a mirror of spontaneous language, research on dubbing can unveil recurrent cross-linguistic correspondences that would be difficult to gain access to considering the scarcity of relevant translational data.As illustrated with demonstrative clefts, a typical feature of casual conversation, if we exploit the similarities between spontaneous and fictive orality, the analysis of dubbed texts can bring to the fore discoursal and pragmatic devices available in different spoken languages to perform the same or similar functions.As a result, the knowledge we draw from dubbing research can be applied to the broader fields of Cross-cultural Pragmatics, Conversation Analysis, Spoken Language and Translation Studies at large.

Biographical note
Maria Pavesi, PhD, is a Professor of English language and Linguistics at the University of Pavia, where she also teaches audiovisual translation.Her research has focussed on orality in dubbing with special attention to personal, spatial and social deixis, and second language acquisition via audiovisual input.For the past 12 years she has developed the Pavia Corpus of Film Dialogue, a parallel and comparable corpus now comprising about 700,000 words of Anglophone and dubbed and original Italian film transcriptions.Her most recent publications include "Translational routines in dubbing: taking stock and moving forwards", Routledge (2018) and "Corpus-based audiovisual translation studies: ample room for development" in The Routledge Handbook of Audiovisual Translation Studies (2018).