Second Dialect Acquisition

Second dialect acquisition (SDA) can potentially occur whenever one interacts with members of a different social group who speak a different way. Often, the new dialect is a regional variety of a speaker's first language (L1), encountered as a result of mobility; for example, native speakers of Canadian English acquiring Southern Standard British English in Oxfordshire, England (Chambers, Reference Chambers1992), or New York City English in Manhattan (Nycz, Reference Nycz2013). However, speakers may also be exposed to new dialects of an L2 (Drummond, Reference Drummond2013) or to dialects associated with different ethnic groups or social classes (Fix Reference Fix2014; Hazen, Reference Hazen2001; Siegel Reference Siegel2010). Acquisition can be evidenced by changes in speech perception and/or speech production, though changes in production may be more salient to both speakers and researchers (for more discussion of perception in dialect acquisition and what it means to “have” an accent, see Nycz, Reference Nycz2015; Walker, Reference Walker2018). While the phrase “second dialect acquisition” may imply that a whole new linguistic system is learned (or is the target of learning), the sociolinguistic research in this field indicates that this rarely if ever happens. Actors may strive to emulate a whole dialect or accent in preparation for a role, perhaps with the assistance of a dialect coach, but this explicit goal is not typically shared by ordinary people in everyday life, whose language ideologies may lead them to view those who change accents as inauthentic (e.g., Eckert, Reference Eckert2003). Instead, most speakers variably and gradiently adopt some new dialect features while maintaining many first dialect norms, with change or lack thereof mediated by various linguistic and social-attitudinal factors (for a review, see Nycz, Reference Nycz2015).

Mechanisms of Dialect Shift and Change in the Individual

SDA could in principle occur via explicit or implicit learning. Speakers might become aware of dialect differences that are especially salient and socially marked, and attempt to change features that are either stigmatized or that simply mark them as being “not from around here” (though, see Nycz, Reference Nycz and Babel2016, for cases in which such attempts appear to be unsuccessful). While specific stereotyped features may be intentionally avoided or adopted, explicit learning of new dialects typically does not take place in the same way that it does for L2 learning. Because mutually intelligible dialects by definition share most of their grammar, phonology, and lexicon, speakers can usually communicate successfully with little conscious effort beyond learning the occasional new word or new meaning for an existing word; “instances of non-understanding” (Varonis & Gass, Reference Varonis and Gass1985) that lead L2 learners to notice specific linguistic features are comparatively rare in dialect contact. Instead, most of the learning that takes place does so implicitly, with speakers gradually shifting their realization of particular words and sounds toward that of the surrounding dialect.

This phenomenon is often referred to as accommodation or convergence. Accommodation is the term favored by sociolinguists; it encompasses any linguistic changes that a speaker makes because of their audience. Sociolinguistic studies, unsurprisingly, tend to highlight how accommodation is used to negotiate social distances and establish solidarity in interaction (e.g., Bell, Reference Bell1984; Bell & Johnson, Reference Bell and Johnson1997; Bourhis & Giles, Reference Bourhis, Giles and Giles1977; Coupland, Reference Coupland1984; Shepard, Giles, & Le Poire, Reference Shepard, Giles, Le Poire, Robinson and Giles2001). As such, both convergence toward and divergence from an interlocutor's linguistic patterns (real or expected; see, e.g., Fasold, Reference Fasold1972) are considered forms of accommodation. While accommodation has been theorized in a variety of ways within the discipline (e.g., Bell, Reference Bell1984; Giles, Coupland, & Coupland, Reference Giles, Coupland, Coupland, Giles, Coupland and Coupland1991), the one crucial element in all accounts is interaction between individuals who have some relationship, however fleeting.

Convergence, meanwhile, is the term more often used in the psycholinguistic and laboratory phonology literature. It refers specifically to cases in which speakers adjust their speech to become more similar to that of a target voice. Convergence is construed as a mostly automatic, unconscious process resulting from an alignment of linguistic and/or gestural representations (Pickering & Garrod, Reference Pickering and Garrod2004), and in laboratory contexts, convergence has been demonstrated for various aspects of speech and other types of gesture (e.g., Chartrand & Bargh, Reference Chartrand and Bargh1999; Honorof, Weihing, & Fowler, Reference Honorof, Weihing and Fowler2011; Shockley, Sabadini, & Fowler, Reference Shockley, Sabadini and Fowler2004). However, it can also be mediated by linguistic factors such as linguistic category boundaries and word frequency (Goldinger,Reference Goldinger1998; Nielsen, Reference Nielsen2011), social factors like gender and interactional role (Pardo, Reference Pardo2006; Namy, Nygaard, & Sauerteig, Reference Namy, Nygaard and Sauerteig2002), attitudes about speaker groups (Babel, Reference Babel2010), and initial distance between the speaker's variety and the target voice (Kim, Horton, & Bradlow, Reference Kim, Horton and Bradlow2011; Nycz & Mooney, Reference Nycz and Mooney2017). A key finding of such experimental work is that, contrary to the sociolinguistic view outlined above, interaction does not seem to be a prerequisite for convergence: Talkers readily converge toward disembodied voices presented over headphones or speakers (Delvaux & Soquet, Reference Delvaux and Soquet2007; Goldinger, Reference Goldinger1998; Nielsen Reference Nielsen2011), though convergence is facilitated by a visible interlocutor (Dias & Rosenblum, Reference Dias and Rosenblum2011). Moreover, some studies have documented convergence effects in laboratory contexts lasting for as long as 7 days after exposure (German, Carlson, & Pierrehumbert, Reference German, Carlson and Pierrehumbert2013; Goldinger, Reference Goldinger, Cutler, McQueen and Zondervan2000).

Accommodation or convergence may happen to some extent whenever a hearer is exposed to linguistic input (Babel, Reference Babel2010), but the effects of any onetime exposure on the adult speaker are often short-lived, undone or obscured by the effects of many other inputs. However, when a consistent change in linguistic input occurs—for example, when a person settles in a new region or is otherwise embedded in a new speech community—many small shifts in the same direction over time can result in long-term changes to pronunciation and observable acquisition of new dialect features (Munro, Derwing, & Flege, Reference Munro, Derwing and Flege1999; Nycz Reference Nycz2013; Trudgill, Reference Trudgill1986; Walker, Reference Walker2014). The question is, can media be a source of input (in naturalistic, nonlaboratory conditions) that causes similar long-term shifts?

The Role of Broadcast Media (TV)

Television has been a main focus in sociolinguistic discussions of possible media effects on language use (for an overview, see Sayers, Reference Sayers2014), usually in the context of debunking the widely held idea that localized accents and dialects are being (or will be) leveled in favor of a media-promulgated standard. Large-scale dialectological studies have shown that, in fact, this leveling is not taking place: indeed, ongoing regional sound changes are leading to divergence across wider regions (Labov, Ash, & Boberg, Reference Labov, Ash and Boberg2006).

One possible explanation for television's apparently limited role in accent change is that while television watching may provide input, it does not usually involve interaction. Studies of L2 acquisition have found that interaction plays a crucial role in learning, both reinforcing existing knowledge and providing opportunities for correction (Gass, Mackey, & Pica, Reference Gass, Mackey and Pica1998). Linguists who have focused on variation within a language, meanwhile, have long observed that patterns of interaction strongly predict patterns of similarity and difference at the group or community level (Bloomfield, Reference Bloomfield1933; Labov, Reference Labov2001); put simply, people talk like the people they talk to. While certain salient and simple features—lexical items, mostly (Trudgill, Reference Trudgill2014)—can be explicitly borrowed by anyone with limited exposure to a source dialect (say, via television), vowel chain shifts and other more complex aspects of phonological and syntactic structure seem to require sustained local contact in the right social networks if they are to be learned (Labov, Reference Labov2007; Milroy, Reference Milroy, Cain and Russom2007). This is not to say that television input is completely irrelevant; Stuart-Smith, Pryce, Timmins, and Gunter (Reference Stuart-Smith, Pryce, Timmins and Gunter2013) argued that, once a nonlocal feature has already diffused into a community via traditional face-to-face contact, its use within that community may be “accelerated” by broadcast media, if those media contribute to or enforce social meanings attached to that feature. Similarly, Kristiansen (Reference Kristiansen2014) suggested that television may shape ideologies around specific language varieties and groups of speakers, which again may facilitate the borrowing and spread of new dialect features in face-to-face interaction. But there is little evidence that exposure to new dialect forms through passive television consumption alone is enough to cause change in the individual speaker.

As discussed in the previous section, however, speakers have been shown to converge toward mediated speech input in the absence of any interaction or even a visible interlocutor, at least in laboratory contexts. This suggests that the crucial factor favoring convergence and dialect learning may not be interaction so much as engagement with new linguistic input, whether that engagement comes from social pressures within an interaction or task-motivated attentiveness in the context of an experiment; indeed, Stuart-Smith et al. (Reference Stuart-Smith, Pryce, Timmins and Gunter2013) stressed that engagement with a particular television show is a key factor influencing the extent to which that show acts as an “accelerant” of change.

New Broadcast Media: YouTube and Podcasts

The internet offers several types of new “broadcast” media, which, like television, are technically one-way and non-(immediately)-interactive; if interaction is the crucial factor allowing dialect learning, we might expect these media to pattern with television as a (non)influencer of dialect. If engagement is more important, however, then some of these media may provide a more favorable context for dialect learning. In this section I discuss two types of new broadcast media—YouTube videos and podcasts—and suggest that the most successful examples of these may offer their consumers a higher engagement experience and thus more potential for dialect learning.

Videos on YouTube are superficially similar to those on television; indeed, many videos on YouTube are copies of those presented on television, provided by corporations in a corporate voice. Due to YouTube's low barriers to participation and strong user support, however, many more videos are contributions from diverse individual “YouTubers” with varying levels of fame and notoriety (Chau, Reference Chau2010), who may also represent a larger range of accents and dialects than found on corporate broadcast media. The most successful YouTubers invite audiences not just to passively consume content but also to “like” and “follow” their channels and to further engage in the “participatory spectacle” (Androutsopolous, Reference Androutsopolous, Tannen and Trester2013) via commented responses and even video replies. Any interaction that occurs will necessarily be asynchronous (with comments or response videos posted after the initial video goes live) and may not even involve “talking,” but could still evoke some of the same mechanisms underlying accommodation in face-to-face interaction. The influence of a speaker in any particular video will be transient, but if a viewer consistently listens to and engages with the other-dialect talk of a favorite YouTuber, the phonetics literature reviewed above suggests that the viewer will eventually show convergence toward that target voice. We might also expect that details of specific videos would favor or disfavor convergence—for example, varying audio quality, which may result in some signals that are too degraded to support learning, or varying access to visual information, which has been shown to improve speech perception (Sumby & Pollack, Reference Sumby and Pollack1954; Havenhill, Reference Havenhill2018).

Podcasts occupy a similar hybrid space between traditional broadcast media and interactive conversation. Podcasts are similar to broadcast media in being one-way and not synchronously interactive, transmitted to listeners who cannot immediately respond. Podcasts, like YouTube videos, may also be created and released by a wide range of people, though the most popular ones (according to iTunes rankings) are typically produced by media companies, either traditional (e.g., National Public Radio) or podcast-focused (e.g., Gimlet, Radiotopia), resulting in high-quality audio (and speech) input. Individual podcasts are usually led by a consistent host or hosts who present information in a conversational format.

Podcasts typically lack video input, which may hinder speech perception and thus the uptake of new dialect variants. However, popular podcast formats may favor engagement and thus accommodation among listeners. The casual, conversational tones of many podcasts may lead listeners to feel like they are hanging out with friends (Edison Research, 2016); communicatively competent listeners know, implicitly, that speakers style-shift to less formal styles with intimates (Bell, Reference Bell1984). Moreover, speakers adjust their speech to become more similar to those who they feel close to (Pardo, Gibbons, Suppes, & Krauss, Reference Pardo, Gibbons, Suppes and Krauss2012). Taken together, these two pieces suggest that podcasts (particularly those that are unscripted and casual in style) would elicit more convergence among listeners. Podcasts may additionally invite engagement in the form of listener emails or voicemails, which are sometimes read and responded to on the podcast, or even synchronous interactions during recordings of live shows.

As with videos experienced via both traditional TV and YouTube, listeners typically do not verbally respond to podcasts as they listen to them. However, if voices producing limited tokens of individual words and heard in a laboratory context can elicit phonetic convergence to specific accent features, it is reasonable to suspect that voices producing natural conversational speech, listened to with interest and engagement in one's home, could do the same, with repeated exposure over time leading to measurable production changes in the listener. At the very least, however, YouTube and podcasts may serve a similar role as television in accelerating the use of dialect features that have already been adopted by the speaker or listener via conventional means.

Mediated Conversation: VCMC

Video computer mediated communication (VCMC) is even closer than TV or YouTube to face-to-face conversation on the mediated spectrum. Skype, FaceTime, Google Hangouts, and other software offer people who are not physically copresent the opportunity to hold synchronous multimodal conversations that are in many ways similar to those that take place face to face. To the extent that these conversations are more intentional (participants must plan to sign on at a particular time) and more likely to take place between intimates, VCMC might be expected to favor convergence behavior compared to the wide range of face-to-face conversations in which people engage. At the same time, there are several differences between VCMC and face-to-face communication that may result in less accommodation via video chat. First, video interlocutors do not have access to the same visual information as those in face-to-face interaction. The unmarked view of one's partner in VCMC is a headshot (Licoppe & Morel, Reference Licoppe and Morel2012); to the extent that participants are not able to fully entrain to the gestural and postural patterns of their partner because of this limited view, this may have some inhibitory effect on linguistic convergence (Shockley, Santana, & Fowler, Reference Shockley, Santana and Fowler2003). Second, variation in the quality of both audio and visual signal in VCMC could result in imperfect perception or encoding of linguistic information (De Decker & Nycz, Reference De Decker and Nycz2011). Third, while the shallow time depth of these technologies means that clear behavioral norms have yet to emerge, at least some VCMC participants may hold more flexible views regarding what constitutes appropriate behavior in an interaction, in ways that may not facilitate accommodation. For example, while adult users find it rude when interlocutors multitask during a VCMC conversation (Ames, Go, Kaye, & Spasojevic, Reference Ames, Go, Kaye and Spasojevic2010; Molyneaux, O'Donnell, & Milliken, Reference Molyneaux, O'Donnell and Milliken2012; O'Hara, Black, & Lipson, Reference O'Hara, Black and Lipson2006), teenagers may not mind if their VCMC partners simultaneously text, cook, or engage in other activities (Buhler, Neustaedter, & Hillman, Reference Buhler, Neustaedter and Hillman2013; DiDomenico & Boase, Reference DiDomenico, Boase, Tannen and Trester2013). If VCMC speech is subject to less engagement and attention for these reasons, then this would lead us to expect less convergence among younger users who hold these views. Finally, most VCMC software displays each participant's own video image to themselves during the call (de Vasconcelos Filho, Inkpen, & Czerwinski, Reference de Vasconcelos Filho, Inkpen and Czerwinski2009); again, to the extent that this visual feedback draws attention away from one's audience or calls attention to one's own speech (Labov, Reference Labov1972), this may impede convergence.

Conclusion and Questions for Future Research

I have very briefly reviewed what is known and what is yet to be known about the role of media in SDA. Sociolinguists tend to agree that while salient and linguistically unembedded dialect features such as words may be explicitly picked up from TV or other media, structurally more complex features (e.g., vowel space configurations) that form most of the basis for dialect differences are unlikely to diffuse via broadcast media, requiring face-to-face interaction. However, media influence on language use is difficult to empirically establish (e.g., Tagliamonte, Reference Tagliamonte2014), and sociolinguistic work on convergence toward new accent features has yet to examine new broadcast media and social media technologies like VCMC. Lab studies indicate that even low-interactive mediated speech input can result in fine-grained phonetic shifts toward a target voice, but these effects have yet to be documented over the long term after continued exposure to a new dialect.

These gaps in our knowledge have an upside, however: wide-ranging opportunities for future research to uncover the mechanisms underlying convergence and dialect learning in mediated speech contexts, as well as to disentangle various factors that may favor convergence or lack thereof. For example, studies comparing face-to-face conversations and video chats could attempt to control for the ways in which these two modalities might elicit difference types of behavior and conversation (e.g., by having dyads complete the same type of task in each condition), later adding manipulations to tease apart the roles of factors such as attention, attitude, and topic and how these interact with modality. Similarly, longitudinal studies of new-media consumers could track whether exposure to the same podcast host or YouTuber over time correlates with shifts in specific vowels and consonants that initially differ between listener and media personality, and how this might be mediated by measures of attitude, engagement, or traditional “social” factors. Research along these lines will not only shed light on the linguistic, social-attitudinal, and cognitive conditions most favorable to dialect (and language) learning with or without technology, but will also help language learners and teachers make the best use of technology available.