Introduction
Awareness, Variation, and Second Dialect Acquisition
Moving to and settling in a region that is different from the one a person grew up in is a fairly common experience in North America. Observations that one’s native region and adopted region differ in accent or dialect seem to be just as common, followed closely by claims about how one’s accent has or hasn’t changed as a result of spending time in the new community. Mobile adults do change some aspects of their speech after exposure to new dialects (e.g. Munro et al. Reference Munro, Derwing and Flege1999; Bowie Reference Bowie2000; Conn and Horesh Reference Conn and Horesh2002; Foreman Reference Foreman2003; Evans Reference Evans2004; Sankoff Reference Sankoff and Fought2004; Evans and Iverson Reference 78Evans and Iverson2007; Bigham Reference Bigham2010; Ziliak Reference Ziliak2013; Walker Reference Walker2014). But to what extent do such changes rely on explicit speaker awareness of dialect differences? In this chapter, I examine the relationship between awareness and acquisition of new dialect features in a study of Canadians who have moved to the New York City region, describing the extent to which these speakers have changed their use of two dialect features subject to markedly different levels of awareness. While explicit awareness may enhance or attenuate adoption of new dialect features, I argue that such awareness is not necessary for dialect change, which is likely driven by unconscious, automatic processes; moreover, awareness of a dialect feature does not imply control over the use of that feature.
Sociolinguists have long posited a link between speaker awareness of a linguistic variable and behavior with respect to that variable. Labov (Reference Labov1963:8) remarks that in selecting linguistic variables for sociolinguistic study, “we would like the feature to be salient, for us as well as the speaker, in order to study the direct relations of social attitudes and language behavior. But on the other hand, we value immunity from conscious distortion, which greatly simplifies the problem of reliability of the data.”Footnote 1 The assumption is that if a feature is too salient, speakers are more likely to change the way they use it in erratic and generalization-muddling ways. Labov (Reference Labov1972a) more formally links awareness and variability, distinguishing indicators – variables used by a particular group which are below the level of conscious awareness and do not vary stylistically – from markers and stereotypes, which rise above the level of conscious awareness and show stylistic variation. Because they are subject to conscious awareness, markers and stereotypes may also be subject to conscious distortion if speakers become too focused on linguistic form; this point has led to the development of methods which reduce attention paid to speech and thus awareness of language form at a local conversational level, so that patterning of variables will be as unaffected by conscious processes as possible (Labov Reference Labov1972b). To the extent that these methods are successful, variationists can observe systematic patterns in speech which reveal the implicit knowledge (or unconscious awareness) that speakers have about the use of sociolinguistic variables (see Squires and Preston in this volume for additional discussion of implicit vs. explicit knowledge).
Scholars interested in second dialect (D2) acquisition have similarly claimed a connection between awareness of features and the likelihood that speakers will change their behavior given new dialect input. In such cases, awareness can intervene at two points: speakers may drop or modify features of their first dialect (D1) or accommodate towards features of the D2. Trudgill (Reference Trudgill1986) addresses both points, stating that “in contact with speakers of other language varieties, speakers modify those features of their own varieties of which they are most aware” (p. 11), and that “accommodation does indeed take place by the modification of those aspects of segmental phonology that are salient in the accent to be accommodated to” (p. 20), although he also notes that linguistic constraints and social factors may play a mediating role. For Trudgill, awareness means conscious awareness, while salient describes features which are “most prominent in the consciousness” of speakers (p. 12). For Auer et al. (Reference Auer, Barden and Grosskopf1998), salience is a complex construct comprising objective linguistic characteristics (such as lexicalization) as well as subjective characteristics which essentially reflect conscious awareness (style differences in read vs. interview speech, representation in writing, stereotyping). Similar to Trudgill, Auer et al. claim that “salience is a necessary but insufficient condition for dialect loss and acquisition,” as the “attitudinal polarity” of social meaning attached to a variable may protect it from change (p. 184). Siegel (Reference Siegel2010) agrees, concluding that “in order to be acquired, a variant must be salient enough to be noticed.” Noticing is a term used in second language learning research to refer to conscious awareness and subjective experience of a linguistic feature; it is a step beyond mere perception of a feature, where speakers may have awareness of its patterning that is “not necessarily conscious” (Schmidt Reference Schmidt1990).Footnote 2 Although terminology varies, the focus of the work reviewed here is the same: conscious awareness and its role in D2 acquisition. Explicit awareness of a feature is seen as prior to, and indeed a prerequisite for, change with respect to that feature. At the same time, awareness (presumably explicit?) of the feature’s social meaning and the speaker’s attitudes around that meaning may inhibit its maintenance or adoption.
Preston (Reference Preston1996) presents a more detailed discussion of the relationship between awareness and language behavior, which on one reading may seem to support this view. He outlines four “modes of awareness” which capture different aspects of how non-linguists think about and use linguistic variables. Three of these modes represent facets of the knowledge that speakers have about linguistic form. A feature may have more or less availability to a speaker as a topic of explicit linguistic discussion; this mode more or less maps onto the notion of awareness as discussed in the literature reviewed above. In addition, a speaker’s grasp of how that feature patterns linguistically or socially may vary in its degree of accuracy and detail. The fourth mode captures the degree to which speakers consciously control their use of a variety or feature, reflecting differences in behavior. The modes are conceptually independent, and Preston gives concrete examples of cases in which they do not align: in one case, a speaker claims to be a proficient speaker of African-American Vernacular English (AAVE), but declines to perform the variety when prompted by her parents and an interviewer, stating that she can’t really speak it unless among other AAVE speakers. This case illustrates the dissociation between availability, which in this case is high (the speaker freely talks about AAVE as a variety and even claims to use it) and control, which is low (she cannot perform this variety when asked to do so). Yet, while availability and control are not clearly correlated (one could easily imagine or even call to mind specific cases of speakers for whom AAVE is highly available and who are able to shift into the variety with ease), an implicational relationship still seems to hold: try to find a speaker who is completely oblivious to the existence of AAVE, and ask them to perform this variety. If dialect change is largely a matter of controlled linguistic behavior – that is, a conscious choice to jettison old features or adopt new ones – then the claim that conscious awareness (high availability) is necessary for such change seems to follow.
Revisiting his modes (this volume), Preston points out that there is much more going on under the surface with respect to both awareness and behavior. The behavior he specifically addresses is language perception – the ways in which listeners interpret speech signals in a given ostensible social context or react to particular signals (or indeed, to more abstract mentions of specific varieties). It seems clear that listeners do not consciously control their perceptions or reactions in relevant cases; moreover, the explicit post hoc rationales listeners may give for their responses likely reflect only a small portion of the complex network of ideologies and associations underlying them (see also Campbell-Kibler, this volume).Footnote 3 The same is surely true of language production: the vast majority of “decisions” that speakers make about which variant to use when in real-time speech cannot be subject to conscious reflection, and speakers’ grasp of the reasons behind these decisions must be incomplete at best.Footnote 4
There is growing evidence that accommodative processes in speech production are similarly subject to largely unconscious, automatic forces. Laboratory studies of spontaneous imitation (e.g. Goldinger Reference Goldinger2000; Delvaux and Soquet Reference Delvaux and Soquet2007; Nielsen Reference Nielsen2011) demonstrate that speakers alter their realizations of particular sounds to converge towards that of heard voices, even though they are not instructed to imitate those voices. That is, speakers change aspects of their accent without any indication of conscious control directing this change; moreover, they are not consciously aware of the relevant differences as they do so. Unconscious convergence of this kind is not inevitable: the tendency to converge may be reined in by social or attitudinal factors which favor accent maintenance or even divergence (Babel Reference Babel2010). Such evidence suggests a different view of the role of explicit awareness in D2 acquisition: it is not a prerequisite to individual dialect change, but instead may act as a filter on the unconscious accommodative processes which set dialect change in motion.
Canadians in the New York City Region
Nycz (Reference Nycz2011, Reference Nycz2013a) presents a study of dialect variation in native speakers of Canadian English (CE) who have moved as adults to the New York City area. This work focused on speakers’ fine-grained phonetic realization of two features which distinguish their native dialect from that of their new home, but also probed awareness of and attitude about these features through qualitative means.
One of the features I examined is the quality of the diphthong /aʊ/ before voiceless consonants, in words like about and mouse. CE is characterized by so-called Canadian Raising, in which the nucleus of this allophone of /aʊ/ is raised in the vowel space compared to /aʊ/ in non-pre-voiceless position (Joos Reference Joos1942). Canadian Raising occurs in both /aɪ/ and /aʊ/ and is not limited to Canada. Raising of /aɪ/ has been found in Martha’s Vineyard (Labov Reference Labov1963), Philadelphia (Labov Reference Labov1994), the Inland North (Eckert Reference Eckert2000), and Ocracoke Island (Schilling-Estes Reference Schilling-Estes1998). Raising of /aʊ/ has also been documented in Martha’s Vineyard, and in Virginia (Kurath and McDavid Reference Kurath and McDavid1961). However, /aʊ/-raising is still largely associated with CE by non-linguists (Niedzielski Reference Niedzielski1999); the phrase out and about, produced with hyper-raised nuclei (oot and aboot) is a popular, if phonetically inaccurate, stereotype of CE. The English spoken in and around New York City (henceforth NYaE), in contrast, does not exhibit raising of either /aɪ/ or /aʊ/. Labov et al. (Reference Labov, Ash and Boberg2006) note the “conservative character of New York City upgliding vowels,” observing that the nuclei of /aɪ/ and /aʊ/ are no higher than those of the low vowels /æ/ and /ɑ/.
The second feature I analyzed is the structure of the low back vowel system – whether there is evidence for two vowel categories, typically transcribed /ɑ/ and /ɔ/,Footnote 5 in the low back region of the vowel space, or just a single category. The Atlas of North American English (ANAE) includes Canada within the large region which does not distinguish words such as cot and caught in perception or production (Labov et al. Reference Labov, Ash and Boberg2006), and according to Boberg (Reference Boberg and Schneider2008), “virtually all native speakers of Canada today” have this merger, which has been present in Canadian English for several generations (p. 150). In contrast, New York City is located in an area where the Cot/Caught distinction remains robust; here, the raised quality of the vowel in Caught helps to maintain the contrast. In neighboring New Jersey, these vowels are also distinct. Coye (Reference Coye2009) reports, based on questionnaire data, that the merger of these vowels is “gaining a solid foothold in New Jersey;” this may be true of counties in the northwest, where around 30 percent of questionnaire respondents report that the vowels of Don and Dawn sound the same, but the responses for the majority of counties in New Jersey are overwhelmingly (>85 percent) distinct.
These two features are subject to very different levels of awareness among speakers of English in North America.Footnote 6 Canadian Raising of /aʊ/, as a stereotype of CE, has high (or common) availability in Preston’s terms, but generally low phonetic accuracy, as the quality of the stereotyped vowel does not reflect its actual quality in CE. Awareness with respect to detail may be incomplete: metalinguistic commentary tends to focus on the words out, about, and house, rather than the conditioning context or the vowel category.Footnote 7 Finally, there seems to be a high level of control, at least with respect to performances of the stereotype; whether this control extends to more authentic realizations is unclear.
The low back vowel system is not subject to similar levels of explicit awareness. Mergers and distinctions per se are typically below the level of conscious awareness and receive no social evaluation (Labov Reference Labov1994); studies of the low back merger in particular do not contradict this conclusion (e.g. Baranowski Reference Baranowski2013). Thus, the distinction (or lack thereof) is usually not available for commentary. The quality of the individual vowels is a different story: speakers may comment on the way that New Yorkers say coffee, for example, imitating a high, back, and diphthongal vowel in this word and revealing at least a limited awareness of NYaE /ɔ/.
This difference between the two features allows us to formulate some soft predictions. If explicit awareness is a prerequisite for dialect change, then we expect that the Canadian speakers in this study would show no acquisition of the D2 feature of low back vowel distinction per se. If they have some explicit awareness of the quality of the vowel in specific words like coffee, it is possible they may alter their pronunciation of these words towards the D2 realization – or pointedly not do so. Behavior with respect to Canadian Raising is harder to predict: high explicit awareness will either lead to maintenance of this D1 feature (because of positive associations with Canadian identity) or its eradication (if there is a desire to assimilate).
If explicit awareness is not a prerequisite for dialect change, but instead functions as a filter on change, then we should find evidence of convergence towards the D2 in both features. This convergence may be attenuated for Canadian Raising, depending on the attitude and desire of the individual speaker. Convergence towards the low back vowel system, meanwhile, should be unaffected, if speakers indeed lack explicit awareness of the distinction per se; however, words like coffee may show less convergence than words like cot, if speakers are aware of the stigma associated with the local production of /ɔ/.
The remainder of this chapter addresses the following questions: Are the specific individuals in this study explicitly aware of either (or both) of these features? If a feature is available for comment, how do they talk about it – is the social evaluation positive or negative, and do they express a desire to change that feature? How are each of these features actually used by the speakers in this sample? Finally, what do the observed patterns tell us about the relationship between awareness and dialect change?
Methods
The data are drawn from sociolinguistic interviews I conducted with seventeen native speakers of Canadian English in 2008. All of these speakers were born and grew up in Canada and later moved to New York City or nearby towns in New Jersey after the age of 21. The twelve women and five men hail from a variety of Canadian provinces and vary in age at time of interview and age of move (Table 3.1).Footnote 8 All but two interviews took place in New York City; the remaining two interviews were held in New Jersey, in venues near the speakers’ homes. My own dialect in 2008 more or less reflected what was at the time my regional background (born and lived seventeen years in New Jersey, less than an hour’s drive to New York City, followed by four years in New Hampshire and seven years in New York City) and social network structure (not very dense or multiplex): I natively produce a low back vowel distinction and a low /aʊ/ nucleus in pre-voiceless contexts, but my production varies according to audience and other contextual factors.Footnote 9 In my communication with speakers before, during, and after the interview, I presented myself as an American Canadaphile who has knowledge and appreciation of some regions (e.g. southeastern Ontario) and aspects of Canadian culture (musical and culinary) and is eager to learn more about other regions and aspects (the Western provinces; politics and national identity).
Table 3.1 Speakers in the Study Described by Gender, Age, Number of Years Spent in the New York City Area at Time of Interview, and Region of Origin
| Speaker | Gender | Age | Years in NYC area | From |
|---|---|---|---|---|
| LC | female | 30 | 1 | Ottawa/Toronto |
| LW | female | 31 | 10 | New Brunswick |
| PW | male | 32 | <1 | Vancouver/Toronto |
| BW | male | 37 | 2 | Toronto |
| NW | female | 39 | 14 | Alberta |
| TM | female | 41 | 3 | Toronto/Manitoba/Ottawa |
| ES | male | 42 | 5 | Manitoba/Alberta |
| JF | female | 45 | 14 | Manitoba |
| LG | female | 46 | 7 | Northern Ontario/Toronto |
| JC | male | 48 | 18 | Montreal |
| EW | male | 50 | 16 | Saskatchewan |
| BK | female | 54 | 21 | Ottawa/Montreal |
| GH | male | 54 | 15 | Montreal/Toronto |
| CW | female | 54 | 28 | Montreal |
| SS | female | 54 | 27 | Montreal |
| DB | female | 58 | 11 | Halifax/Toronto |
| VJ | female | 70 | 44 | Toronto |
Each speaker participated in four activities. First, we engaged in a one-to-one conversation about their life as a Canadian in the New York region. Topics in these conversations included their hometown in Canada, their experiences growing up there, their reasons for moving to the United States, and their feelings about their home and adopted countries. Next, the speaker read words from a word list presented with flashcards and completed a sociolinguistic minimal pair task. Then the speaker completed a minimal pair judgment task, in which they were asked to look again at the minimal pair list and say whether they thought there were any pairs which people from the New York region would have different judgments on or pronounce differently. Finally, the conversation resumed with additional discussion of linguistic features and impressions of their native dialect and the local variety. The conversational data were phonetically analyzedFootnote 10 to determine how each speaker produced the features of interest, while the metalinguistic commentary generated by all tasks was used to assess each speaker’s awareness of these features.
Results
Canadian Raising
Awareness of Canadian Raising
The results regarding awareness of Canadian Raising in /aʊ/ are easily summarized: every speaker interviewed is consciously aware of this D1 feature. Many speakers mentioned raising in /aʊ/ as a feature of CE before they were even asked about language in the conversation; for the rest, raising was the first or secondFootnote 11 feature they responded with when asked about features of their native dialect. In addition to high availability, participant knowledge of this feature is also characterized by a high level of accuracy and detail: while the speakers acknowledge the phonetically inaccurate American oot and about stereotype, they accurately perform the phonetics of this feature, can distinguish their own naturally raised productions from the stereotype, and can characterize the set of words that contain it.
While /aʊ/ is subject to high awareness among these speakers, its social evaluation is not straightforwardly positive or negative. Instead, speakers consistently note simply that it is a very Canadian feature, one that instantly outs them to Americans as “not from here” (1). For this reason, several speakers, assuming high amounts of control over this feature, claim to have consciously reduced their use of it since moving to the States, as JF does in (2):
(1) Honest to God I think “about” – I don’t say it with a Canadian pronunciation … except sometimes in conversation people who know Canadians will go “Aha! I caught you!” … Sometimes apparently I will say ab[ʌʊ]t (ES, 42)
(2) I feel like I wanna be the one who reveals I’m Canadian, I don’t want people to hear it in my speech. So that’s why I’m saying “ab[a:ʊ]t … so I have changed, consciously changed … (JF, 45)
It is important to note here that speakers are not in any way ashamed of being identified as Canadian – indeed, many expressed pride in their Canadian identity. However, they would like to have some control over how this information is revealed,Footnote 12 and claim to take this control by deliberately manipulating their use of salient dialect features. In this group of speakers, we thus have across-the-board explicit awareness of a D1 feature, and in several cases a desire to attenuate it, as well as a belief that this has successfully occurred.
Use of Canadian Raising
Formant measurements were taken at the F1 maximum of the diphthong nucleus in 1,210 /aʊ/Footnote 13 words across all speakers. An Analysis of Variance (ANOVA) using treatment contrasts was carried out for each speaker, with F1 as the outcome variable and a predictor variable dividing the tokens into four groups: instances of about, instances of out, tokens of other raising context words such as shout or gout, and non-raising context words like loud or how. The non-raising group was set as the control level against which the other three raising context groups were compared. What results from each of these analyses is two pieces of information per raising context group: (1) an assessment of whether the average F1 for the group is significantly different from that of the non-raising group (i.e. whether that group shows significant raising); and (2) a coefficient indicating the magnitude of this difference.
The results of these analyses are summarized in Figure 3.1, which plots the regression coefficient (in Hz) associated with each word group for each speaker. A horizontal line is drawn at –60 Hz, reflecting the threshold value used by ANAE to categorize tokens as raised or not. Seven of the speakers are consistent raisers, with each of the three groups having a significantly lower F1 value than the baseline non-raising group; these differences are also all above the 60 Hz threshold. Four speakers show significant raising only in about and out (again, with both of these word classes sitting above the ANAE threshold), while their other raising context words are not significantly different from the non-raising baseline. Finally, four speakers show significant raising only in about.
Figure 3.1 /aʊ/ Raising in About, Out, and Other Raising Words among Canadians in the New York City Area
Black type indicates a significant difference between the word group and the non-raising context baseline group.
A closer look at the other-raising group, however, indicates that lexical items within this group showed raising to varying extents: higher word frequency was associated with higher F1, suggesting that the greater frequency of exposure to D2 tokens of these items has resulted in them shifting towards D2 realizations at a faster pace.
To summarize, nearly all of the speakers continue to use Canadian Raising in /aʊ/ in at least some contexts. An implicational trend can be seen: about is the most raised group, typically followed by out, then other raising-class words; this trend holds even for the two speakers who showed no significant differences between the raising groups and non-raising baseline. That is, the words which are most strongly associated with Canadian raising – via the phrase out and about – are also the words which most exhibit this feature, despite speakers’ specific desire to attenuate it. Within the other raising-class words, higher frequency items show more convergence towards D2 realizations.
Low Back Vowels
Awareness of Low Back Vowel Contrast
The idea that mergers and distinctions per se are not typically subject to conscious awareness and overt commentary is largely borne out in the current data. Speakers failed to comment on the low back vowel distinction during a conversation about language, as well as a task which specifically calls attention to potential contrast. No speaker mentioned this feature in their interview, either on their own or when asked to describe the features of NYaE they have noticed.Footnote 14 Moreover, in the ordinary minimal pair task everyone demonstrated a merger in both production and perception: word pairs such as cot/caught and don/dawn were produced homophonously and speakers accordingly noted that they sound the same.
More interesting results emerged from the minimal pair judgment task. When asked to comment on how New Yorkers might pronounce or judge the pairs on the minimal pair list, seven speakers revealed an awareness of the Cot/Caught distinction in their new dialect. These speakers produced many or all of the relevant pairs with a phonetically exaggerated distinction, using an extremely high and often diphthongal vowel for Caught words. In contrast, four other speakers showed no explicit awareness of this feature in the judgment task: they picked out other pairs as sounding different in the ambient dialect (e.g. pointing out that the words higher and hire, a distractor pair, would be produced without /r/s), but passed over the low back vowel pairs without comment. The remaining six speakers showed some explicit awareness of low back vowel differences, but this awareness seemed limited to specific lexical items. VJ, for example, commented that doll may be produced with a more “drawn out” vowel than tall, but otherwise did not spot any low back differences, nor generalize to other words. The low back vowel distinction thus has low availability even for the speakers who have some explicit awareness of it, insofar as discussion of this feature can only be elicited using very targeted prompts. Moreover, the accuracy and detail of this awareness varies widely for those speakers who demonstrate it.
For those speakers who do comment on the low back vowel pairs, social evaluation focuses on the quality of the vowel in Caught. For the most part, these evaluations are rather neutral – while a few speakers think the New York /ɔ/ is “grating” or “annoying,” the individuals in this study are mostly amused by the difference.
Unlike Canadian Raising, then, the low back vowel system is subject to variable awareness among the expat Canadians in this study: some speakers are explicitly aware of the distinction as such, some speakers are explicitly aware of differences only in specific lexical items, and others reveal no explicit awareness of this difference.
Use of Low Back Vowels
Formant measurements were taken at the F1 maximum of the low vowel in 3,288 tokens of Cot words and 2,052 tokens of Caught words across all speakers. Mixed effects linear regression was used to determine whether each speaker produced a significant distinction between these two groups in spontaneous speech. For each speaker, two analyses were run, one with F1 as the dependent variable, and one with F2. For each formant, a model containing phonological predictors such as following place and following manner and a random effect of word was compared to a similar model containing those same terms plus a factor coding word class. When the model including word class is found to be significantly better than the model lacking this factor, this indicates that there is variation between the word classes which cannot be attributed to phonological factors alone, and that the speaker distinguishes these word classes in production on the relevant formant dimension. The effect size associated with word class in this more complex model can be interpreted as the magnitude of this distinction.
The results of the analyses of conversational data are shown in Figure 3.2, which plots the effect size (in Hz) associated with word class obtained in the F2 and F1 analyses of each speaker. Eleven of the seventeen speakers produce a significant (if in some cases small) difference between Cot and Caught in some dimension in spontaneous speech. Yet, this behavior is not clearly related to awareness of the feature. While five of the speakers who display a contrast also have a general explicit awareness of the contrast, the other six speakers in this group appear to have acquired a distinction with limited explicit awareness or no awareness of it. Meanwhile, two speakers who have a general explicit awareness of the contrast do not exhibit a significant difference between these word classes.
Figure 3.2 The Effect Size (in Hz) Associated with the Word Class Factor Obtained in the F2 and F1 Analyses of Each Speaker
Speakers with a large difference along both dimensions are plotted further away from the origin, while those with little or no difference between word classes are plotted closer to the origin.
There are also significant frequency effects associated with the realization of these tokens. Table 3.2 shows the results of analyses of each word class across speakers. Higher frequency Cot words tend to be lower and fronter (that is, more D2-like) than lower frequency words of this class, while higher frequency Caught words are higher (again, more D2-like) than lower frequency Caught words. These effects are not symmetrical: the frequency effects associated with the Cot words are somewhat greater than and more significant than those of the Caught words.
Table 3.2 Frequency Effects on F1 and F2 for Each Word Class
| Effect (Hz/Count) | p-value | ||
|---|---|---|---|
| Caught | F1 | –0.38 | <0.05 |
| (F2) | (–0.03) | (1) | |
| Cot | F1 | 0.52 | <0.01 |
| F2 | 1.72 | <0.001 |
To summarize, the majority of speakers show evidence of having acquired a distinction between Cot and Caught words in their conversational speech, and in several cases this acquisition occurs without explicit awareness of the D2 feature. In addition, higher frequency words of both classes show greater evidence of convergence towards D2 realizations, although these effects are less pronounced for the Caught words.
Discussion and Conclusion
Canadian Raising in /aʊ/ and the low back vowel distinction differ greatly in the type of awareness they are subject to. The speakers in this sample, like many North Americans, are explicitly aware of Canadian Raising and its social meaning(s), and in many cases believe that they have control over this feature. Awareness with respect to the low back vowel distinction is more varied: some speakers have an explicit awareness of the distinction per se, others reveal a limited explicit awareness connected to particular lexical items or vowel qualities, and others show no evidence of explicit awareness. However, even speakers who accurately characterize this feature only do so when prompted in a particular way, so the distinction does not have high availability. As described in “Use of Canadian Raising,” above, expectations about how these features should pattern will depend on whether explicit awareness is, as some scholars have claimed, a prerequisite for loss of a D1 feature or acquisition of a D2 feature, or rather a filter which may attenuate automatic, unconscious processes of accommodation.
Canadian Raising, the D1 feature subject to high across-the-board awareness, exhibits both stability and change in the speech of these study participants. On the one hand, all speakers show robust raising in the words out and about, and several speakers continue to raise in other raising-context words. On the other hand, there is evidence of change, in that higher frequency lexical items (setting aside out and about) are realized with lower nuclei, indicating that speakers are accommodating to the D2 in a patterned, lexically gradual manner. While this analysis cannot address the relationship between explicit awareness and dialect change (as speakers are uniformly aware of Canadian Raising), the data reveal a disconnect between awareness of a feature and control over that feature. If speakers are aware that this feature marks them as Canadian and want to suppress its use, then not only might we expect less raising overall, but the stereotype items out and about should be realized with the lowest nuclei rather than the highest. The patterns described here show that explicit awareness (and will to change) does not necessarily translate into control.
The stability of Canadian Raising is unexpected in both the prerequisite and the filter view of awareness, given the explicitly stated attitudes and desires of the speakers. To understand these results, the conversational context in which these data were collected must be considered. As a researcher who specifically recruited expatriate Canadians and expressed an interest in learning about each speaker’s experience as a Canadian in the United States, I no doubt established a setting that was favorable to eliciting CE features; in this context, the risk of [ʌʊ]ting oneself as Canadian was a non-issue. Of course, I was still an American conducting interviews in New York or New Jersey, a context which should prevent wholesale style shifting back into the D1. Indeed, total style-shifting back into D1 cannot be happening for most speakers, given their decidedly non-CE realization of the low back vowel system. So while use of certain CE features like Canadian Raising might be especially favored in the data collected here, it is unlikely that this speech is qualitatively different from the speech of these speakers in everyday contexts. We might expect that in other conversational contexts, characterized by non-Canadian topics and/or less explicitly welcoming American interlocutors, the speakers in this study might suppress Canadian Raising to a larger extent. Such intraspeaker variation would indicate that explicit awareness and attitude interacts with other factors such as topic and audience to influence how features are used in particular contexts. Further research which examines topic- and audience-based variation is needed to address this issue.
The low back vowel findings are more informative. The speakers in this sample vary in their awareness of the D2 low back vowel distinction, allowing a test of the idea that explicit awareness is necessary for adoption of the D2 feature: if this idea is correct, those who have explicit awareness should show more acquisition of this feature (although perhaps less consistently than one might expect for a more available feature), while those with no explicit awareness should display no distinction. In the current data, this prediction fails to hold: while eleven out of seventeen speakers show evidence of having acquired a difference between the Cot and Caught words, three indicate no explicit awareness of this feature and three others have fairly limited explicit awareness (including the speaker who exhibits the greatest distinction between word classes!). These results indicate that explicit awareness is not a prerequisite for acquisition of this feature. Instead, automatic accommodative processes likely drive this acquisition: speakers unconsciously adjust their realizations of relevant words in the direction of those in the ambient dialect, resulting in the separation of the two word classes.
At the same time, these accommodative processes may be attenuated by awareness of a dialect feature and the evaluations that accompany it. The speakers in this study are often explicitly aware of the phonetic quality of the NYaE Caught vowel and of the stigma that it carries, while the quality of Cot typically goes unnoticed. The findings presented here suggest that convergence towards the D2 is somewhat less advanced for Caught than for Cot, as frequency effects on the former are both smaller and less significant. Automatic convergence towards NYaE Caught may be stunted by explicit awareness of vowel quality and its meaning, while convergence towards Cot proceeds unimpeded.
