During the 2000s, a plethora of publications in the field of AVT addressed the study of different features of orality related to the interactional character of dubbed dialogues, comparing dubbed dialogues with either film dialogues in their OV counterparts or corpora of natural speech. The vast majority of studies focused on the Italian and Spanish dubbed renderings of English phraseological units, discourse markers and terms of address.

Scholars in the field who have approached the study of audiovisual orality in fictional products using a corpus linguistics methodology agree in the formulaic character of dubbed dialogues. Such fixedness seems to be due to the use of calqued syntactical structures such as specific cleft sentences (Pavesi 2016), weak connectors, the use of progressive tense and split structures (Freddi 2008).

A review of previous literature shows how most studies have tackled the analysis of orality features on the screen from the prism of discourse analysis and phraseology, with intensifiers and discourse markers being the most widely studied phenomena. Results obtained in previous research reveals that orality markers such as you know or I mean tend to be less present in Spanish dubbing than in the English original version or domestic Spanish productions (Chaume 2004; Baños 2014). Such omissions, together with the use of vague hedges, have been attributed to achieving brevity and speeding up communication (Chaume 2004; Quaglio 2009). According to the study carried out by Baños (2014), these orality markers also seem to be more stereotypical and unidiomatic due to interference from the source language; however, the author also identifies some orality markers that were not present in the original dialogues as well as other phraseological units, prefixes and shortening phenomena, in what is regarded as an attempt to achieve more credibility.

Romero Fresco (2007; 2009; 2012) analyses the use of intensifiers and discourse markers in colloquial conversation, comparing their frequency of use in three different sources: the dubbed version of a sitcom in Spanish, a sitcom originally produced in Spanish and a corpus of real-life conversation. According to the author’s findings, some translations provided for intensifiers, such as de veras (really in the ST) and transition markers like está bien (fine or okay in the ST), which are recurrently used in dubbed scripts and play-acted contexts despite being virtually absent in real talk, something that can lead to asymmetries between the original and dubbed versions in terms of social distance between the characters. Similarly, Marzà (2016) analyses intensifying traits such as prefixes and suffixes, like súper- and –mente, and other figures of speech, such as hyperbole, and finds a higher frequency of use of intensified segments in audiovisual corpora and a majority of unnatural intensifiers in dubbing. Moreover, according to Romero Fresco (2007), Spanish fillers such as pues or hombre, which are rather frequent in spontaneous conversation, are barely present in dubbed dialogues. In the Italian context of dubbing, Zanotti (2014) also examines general extenders in TV language, which are defined as expressions with vague semantic content such as and all that, and everything. The author observes a reduction of these markers in translation due to omission being the preferred strategy; however, direct translation was also used to a certain extent in the corpus (19% of cases).

Although these discrepancies can be interpreted as a lack of naturalness, authors justify the presence of these unnatural features by referring to the humoristic purposes of the original product (Marzà 2016) as well as to the notion of suspension of linguistic disbelief, by which consumers of dubbed products cease to question the actual authenticity of the language used on the screen (Romero Fresco 2009: 68).

Another feature of orality that has drawn the attention of researchers in the field of AVT is the translation of terms of address in dubbing. Romance languages such as Spanish and Italian use morphology to convey familiarity or social distance between the speakers through the pronouns lei or usted, which do not exist in English. This lack of grammatical markers is compensated in English with terms of address, which is a common feature in casual conversation. On the one hand, authors such as Pavesi (2012) have found shifts between the personal pronouns tu and lei in Italian dubbing which signal differences in the character’s linguistic positioning and the emotional intensity of important narrative events. On the other hand, some researchers in AVT (Antonini and Chiaro 2009; Naranjo 2015) have noted that dubbed corpora in Spanish and Italian often present literal translations of these terms of address, which are not actually frequent in real-life talk, such as amico [friend], figluiolo [son], hermano [bro, brother], chica [girl]. From a broader approach above the word level, Quaglio (2009) examines the inner structure of conversations in the form of turn-taking and identifies some features in Italian dubbing, such as the absence of overlaps and interruptions, the presence of unexpected features (e.g. pragmatic failure for humoristic purposes (Quaglio 2009: 147), as well the co-occurrence of vague linguistic devices (e.g. sort of) and overly elaborated explanations (Quaglio 2009: 148). In natural conversation corpora, however, he detects longer turns due to the fact that narrative segments are followed by single-word utterances such as sure, right or okay.

Finally, moving away from the corpus linguistics methodology, Pérez-González (2007) proposes a systemic functional approach to studying the naturalness of original and dubbed dialogues that allows researchers to analyze extended stretches of interaction from the perspective of interpersonal relationships. In his analysis, the author observes how spontaneous parts of the original dialogues are sometimes “neutralized by the overall artificiality of the interactional dynamics in the target language” (Pérez-González 2007: 34). Examples extracted from his corpus show how the role or portrayal of some characters in the original version (e.g. the dominant character of one of the speakers) is diluted in the dubbed version since the exchange in the target language fails to reproduce the “sequential build-up” (Pérez-González 2007: 31) found in the original.

Together with the discourse markers and other features of interactional orality, taboo words have also received a lot of attention from scholars. Instances of taboo expressions analyzed in previous works mainly include abusive and derogatory swearing, as well as other terms related to sexuality, crime, death or killing (see Zanotti 2012; Soler-Pardo 2013; Ávila Cabrera 2016).

A review of the existent literature on this trait in the Italian and Spanish dubbing context reveals a general downgrading trend of the original taboo expressions, either by omission or euphemization and authors have associated this trend with translators’ self-censorship with a view to mitigate the impact on the viewers, broaden the target audience and obtain the approval of the board of censors (Ranzato 2009; Zanotti 2012; Soler-Pardo 2013). It is also worth mentioning that some scholars such as Ávila-Cabrera (2016) have detected a shift in translators’ behavior between productions from the 90s and those from the 2000s, with a lower percentage of not technically justified omissions of offensive language in more recently released films.

Some more recent works have attempted a more fine-grained analysis, finding similar results but with some differences among different categories of taboo formulas. Giampieri (2017; 2018) provides an account of the translation of screen insults (such as ‘motherfucker’), as well as expressions with scatological, religious and sexual connotations. The author finds that a large number of the original taboo words containing sexual references were omitted or lessened in her corpus, whereas some religious offences seem to have been replaced with vulgar language.

Another trend detected in more recently released films is a literal translation together with the routinization of taboo expressions consisting of repeatedly using the same expression in the target language as a fixed equivalent for specific ST structures. In this regard, authors (Ávila-Cabrera 2016;Giampieri 2017; 2018) have reported on the fixedness of the expressions used, which reflect the fact that translators do not resort to the wide variety of creative solutions available in the target language.

Finally, some works have also revealed a dissonance between the low presence of taboo expressions in dubbed products compared to real conversations. In their study, Antonini and Chiaro (2009) carried out an experiment in which a group of randomly selected viewers were asked to watch clips extracted from fictional programs dubbed in Italian and then fill in a questionnaire to check their understanding of specific sections and assess the likelihood that certain “turbulent” language-specific features occurred naturally in Italian. Taboo words were one of the features present in their corpus of film clips based on the premise that in Italy, at the time, these taboo words were frequently omitted in dubbed language even when they are a common feature in everyday Italian. Indeed, their results show that taboo words were rated as one of the closest features to spontaneous talk in Italian.

This tendency has even been spotted in the translation between culturally close language pairs such as Italian and Spanish. Zamora (2016) examines the Spanish dubbed renderings of the word cazzo present in original Italian productions, concluding that syntactic calques and euphemization were the most used strategies even though the degree of social tolerance, conflictivity and frequency of use do not correspond to their literal equivalents in Spanish everyday language.

To sum up, studies so far have revealed two essential patterns that seem to generally appear in the translation of conversational language and orality traits in both Italian and Spanish dubbed dialogues, namely: 1. interference from the source language; 2. fixedness or lack of linguistic variety in the renderings proposed. This tendency coincides with the conclusions drawn by Pavesi (2008; 2018) who argues that audiovisual dialogues have undergone a process of routinization over the years, i.e., a tendency across languages by which commonly recurrent ST conversational expressions are consistently translated in the same way in the TT, often through the use of calques and borrowings from the ST.

59 students (22 male and 37 female) with a mean age of 19.14 participated in this study. All participants were in their first year at university and were not pursuing any degree related to translation or foreign languages. Two main criteria were considered for the sampling. Firstly, the participants had to belong to the same generation of viewers, which guaranteed, to a certain extent, that they had had similar viewing experiences and expectations of dubbed products. Secondly, they must not be pursuing a degree in translation or foreign languages so as to ensure that they did not have specific knowledge or training in audiovisual translation, which would probably avoid excessive alertness or strong judgments about the quality of dubbing triggered by their knowledge and training in translation. Participants were randomly distributed into two groups: Group A with 9 males and 21 females and Group B with 13 males and 16 females. Most participants indicated having at least a daily or weekly habit of watching dubbed audiovisual products

Two scenes representing different emotional situations (anger and sadness) were chosen from the film Like Crazy, originally shot in English (see more details in Appendix 1). For each scene, two versions (natural and play-acted) were translated into Spanish by a professional translator and dubbed by junior voice actors-in-training. Both scripts in Spanish differed slightly in the pragmatic and discursive choices, while their semantic content remained intact (see Appendix 2). We drew on previous literature to design materials that adhered as much as possible to the traits observed in dubbese vs. spontaneous talk. In this regard, in the natural version, we tried to detach from the routinization tendency detected in previous works by introducing a variety of discourse markers, which would reflect the pragmatic attitudes of the characters in each scene and we instructed the voice actors to avoid a dramatized ‘reading-for-acting’ style when speaking the lines. Conversational Spanish markers, such as en plan, pues, venga and es que were introduced whenever the lip-synchronization restrictions allowed for it. In the scene representative of anger, we also replaced the empathic articulation present in some imperatives of the OV to convey higher levels of emotional arousal with the Spanish intensifier que (e.g. ¡que me mires y me lo digas! vs. ¡mírame y dímelo!), as this is a more common strategy in this type of communicative exchanges in Spanish. To help actors detach from their play-acted routines, we asked them to imagine how they would talk and react to what the other character was saying if they were going through the same situation in real life.

On the other hand, the translation for the play-acted version was closer to the original and tried to preserve a dubbese style with calqued formulaic routines and the actors were provided with instructions to ensure that they intentionally included some vocal and paralinguistic traits identified in play-acted and dubbed speech by previous studies, such as vocal fry, breathiness, fluctuation of pitch, exaggerated articulation and elongantion of sounds, especially in inter-word diphthongs where the visuals allowed for it. The style of dubbed scenes in productions traditionally broadcast during the weekends by the Spanish media corporation Atresmedia in its Multicine section were also used as a benchmark to guide voice-actors as clear examples of some of these features.

Participants in Group A watched the two dubbed versions of the scene representing anger (titled ‘FIGHT’) and participants in Group B watched the two versions of the scene representing sadness (titled ‘PHONE CALL’). Participants in Group A watched the play-acted version first, followed by the natural version, while the reverse order was applied for Group B to counter-balance any potential order effects. Afterwards, participants were asked to complete an ad hoc questionnaire of seven questions to determine whether they noticed any difference between the two versions in terms of naturalness as well as to find out about their preference for one style over the other. The first questions were aimed at determining whether participants were really able to detect the differences without conditioning them beforehand with any sort of clues as to where the difference was to be found. It was not until later in the questionnaire that they were specifically instructed to assess the naturalness of the versions. It is important to point out that we provided a specific definition of “naturalness” in the questionnaire as “more similar to Spanish everyday talk,” so that the concept was not confusing for our participants.

The experiment was conducted in situ in a university classroom. Both the video clips and the questionnaires were provided through a GoogleForm and participants’ answers were collected online from participants’ own laptops or tablets. Participants also used their own earphones.

Both quantitative and qualitative analyses were carried out with the collected data. Descriptive and inferential statistical analyses were conducted to find out about perceived naturalness and preferences for the two dubbed versions. For this purpose, within-group comparisons for each version of each scene were performed by running the non-parametric test Mann-Whitney, since the samples did not comply with the required criteria of normality and homogeneity of variance. For inferential statistics, a score of 0 or 1 point was assigned depending on whether participants chose the analyzed scene as the most natural/preferred or not, respectively, as they had to opt for one or the other based on its naturalness and their preference. Percentages were also calculated for the occurrences of tags or descriptors provided by participants when they were asked to indicate specific differences they perceived between both versions (e.g. stiff, realistic, emotional).

Both versions were perceived by participants as being different from each other. A total of 100% and 86.2% of participants in Groups A and B respectively reported having noticed a difference. Participants who indicated having perceived a difference were then asked to use their own tags or descriptors to explain how the two versions were different. They were to formulate their answers following the scheme “Scene 1 is more/less + adj. + than 2.” In Tables 1 and 2 we collected and sorted by categories all the descriptors provided by participants in both groups for both the natural (marked with a ‘N’) and the play-acted (marked with a ‘P’) versions. The frequency with which these descriptions appear in participants’ answers is also provided in the tables.

Two broad categories were considered to classify participants’ descriptors: dubbing quality and emotional intensity. In the first case, as displayed in the section “Poor Dubbing Quality” of Table 1, no negative tags (0%) were found when participants assessed the natural version of the scene (‘FIGHT_N’) in terms of the quality of dubbing; however, 28.6% of participants attributed negative tags to the play-acted version (‘FIGHT_P’), mainly highlighting that it was indeed overplayed and stiff or forced.

On the other hand, positive tags for the natural scene outweighed those for the play-acted scene (52.9% vs. 23.8%) with most participants labeling it as more natural and realistic. Nonetheless, 23.8% also perceived the play-acted version as having some positive characteristics (e.g. understandable, better dubbed, professional).

Concerning emotional intensity, descriptors were classified into two subcategories: high and low intensity, depending on their semantic connotations. As Table 1 shows, a higher frequency (41.18% vs. 19%) of high-intensity descriptors, such as colloquial and vulgar, were used by participants to refer to characteristics of the natural scene in comparison to the play-acted scene. As for the low-intensity descriptors, they were more frequently found in participants’ answers associated with the play-acted version than the natural version (28.6% vs. 5.88%). Therefore, from these results we can infer that the natural scene was perceived as having a higher emotional intensity, allegedly due to its colloquial and even vulgar character, whereas the play-acted scene was perceived as being more literal, external, cold, robotic, formal or slow.

Table 2 below displays the results obtained for the scene PHONE CALL. The same categories of dubbing quality and emotional intensity were used for this analysis. Firstly, contrarily to the scene FIGHT, participants in this group did not use any descriptors (0%) associated with poor quality for the play-acted scene, whereas in the natural scene we find that 26.32% of the answers contain negative tags such as overacted. Secondly, a higher frequency of positive tags related to the naturalness of the scene (natural) was found associated with the play-acted version (85.71%) compared to the natural version (47.37%). Finally, while the natural scene was more frequently assessed with high-intensity descriptors, such as colloquial or dramatic (19%) than with low intensity descriptors, such as formal (5.26%), its play-acted counterpart received little attention in terms of emotional intensity with 0% cases of high-intensity descriptors and only one low-intensity tag (dry).

In this section, we carried out inferential statistics that allow a comparison of the level of naturalness perceived by participants between the play-acted and the natural version for each scene. Table 3 shows descriptive data: mean (Mn), median (Md) and standard deviation (SD) for each data sample, as well as the results of the Sapiro-Wilk normality test (S-W test), Levene’s homogeneity of variance test and the Mann-Whitney test. The p-value obtained after carrying out this last test indicates whether the differences between the two versions (natural vs. play-acted) are statistically significant (<0.05) or not significant (>0.05).

As displayed in Table 3 below, very significant differences between the samples of both versions in Group A (p=0.00) reveal that the version of the scene ‘FIGHT’ which was originally recorded with a more natural style was, in fact, perceived as more natural by participants than the play-acted version. However, in Group B no significant differences were found between both versions for the scene ‘PHONE CALL’ (p=0.09). For this scene, a higher percentage of participants actually perceived the play-acted version as being more natural.

When participants were asked more precisely about the specific traits that made them able to tell the differences between both versions, three different aspects emerged: 1) differences in the language used in the script, 2) paralinguistic features and 3) some specific moments of the scene. As Table 4 shows, ‘FIGHT’ differences are perceived mainly in terms of the type of language used, with most participants noticing specific changes which made the natural version more colloquial. Differences in paralinguistics were also perceived mostly in terms of intonation and the speed with which participants talked. Also, almost 38% of participants referred to the moment in which participants started shouting as the point in which differences were more noticeable. On the other hand, in ‘PHONE CALL,’ paralinguistic changes (especially prosodic) were the most clearly perceived as differentiating elements between the two versions, with 70% of participants reporting changes at this level. 26% of participants also detected changes in the type of language used and only a small percentage (3.7%) mentioned a specific point in the scene.

When asked about their preference for one dubbing style over the other, statistical analyses only revealed significant differences for the scene ‘FIGHT.’ As illustrated in Table 5, whereas in ‘FIGHT’ there is an preference for the natural version over the play-acted version with a p-value of 0.000, in ‘PHONE CALL’ no statistical differences were found, with a p-value >0.05. The percentages show that approximately half of participants preferred the natural version (44.83%) and the other half (55.17%) preferred the other.

Finally, participants were asked to justify such preferences in an open-ended question. In both scenes, the quality of dubbing in terms of naturalness was the most frequently reported reason for preference of one version over the other by arguing that it was “more realistic,” “more similar to everyday speech” or “less acted/theatrical.”