Journal ofhttp://eng.sagepub.com/ English Linguistics Regional Phonetic Differentiation in Standard Canadian English Charles Boberg Journal of English Linguistics 2008 36: 129 DOI: 10.1177/0075424208316648 The online version of this article can be found at: http://eng.sagepub.com/content/36/2/129 Published by: http://www.sagepublications.com Additional services and information for Journal of English Linguistics can be found at: Email Alerts: http://eng.sagepub.com/cgi/alerts Subscriptions: http://eng.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav Citations: http://eng.sagepub.com/content/36/2/129.refs.html >> Version of Record - May 12, 2008 What is This? Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 Regional Phonetic Differentiation in Standard Canadian English Journal of English Linguistics Volume 36 Number 2 June 2008 129-154 © 2008 Sage Publications 10.1177/0075424208316648 http://eng.sagepub.com hosted at http://online.sagepub.com Charles Boberg McGill University Taking as a point of departure the preliminary view of regional phonetic differentiation in Canadian English developed by the Atlas of North American English, this article presents data from a new acoustic-phonetic study of regional variation in Canadian English carried out by the author at McGill University. While the Atlas analyzes mostly spontaneous speech data from thirty-three speakers covering a broad social range, the present study analyzes word list data from a larger number of speakers (eighty-six) drawn from a narrower social range, comprising young, university-educated speakers of Standard Canadian English from all across the country. The new data set permits a more detailed view of regional variation within Canada than was possible in the Atlas, which focuses on differentiating Canadian from neighboring varieties of American English. This view adds detail to the established account in some respects, while suggesting a revised regional taxonomy of Canadian English in others. In particular, this article reports on several phonetic isoglosses that divide Canada’s Prairie region from Ontario, thereby splitting the “Inland Canada” region of the Atlas into western and eastern halves. In fact, the data presented here suggest a division of Standard Canadian English into six regions at the phonetic level, rather than the three proposed by the Atlas: British Columbia, the Prairies, Ontario, Quebec (Montreal), the Maritimes, and Newfoundland. This taxonomy corresponds to the six major regions identified in the study of lexical data reported in Boberg (2005b). Keywords: accents of English; acoustic phonetics; Canadian English; dialectology; dialects of English; vowels Canada in the Atlas of North American English In the Atlas of North American English, Labov, Ash, and Boberg (2006), henceforth LAB, offer the first national view of Canadian English phonetics. A previous national Author’s Note: A preliminary version of this article was presented to the American Dialect Society annual meeting in Anaheim, CA, on January 6, 2007. Thanks are due to several students who assisted with the research, particularly Anicka Fast, Deena Fogle, Ellen House Kogut, and Erika Lawrance, and to the editors and reviewers of the manuscript who suggested valuable improvements to it. Financial support was received from the Social Sciences and Humanities Research Council of Canada (Standard Research Grant No. 410-02-1391). 129 Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 130 Journal of English Linguistics survey of Canadian English (Scargill & Warkentyne 1972) used written questionnaires, thereby excluding the possibility of studying phonetic variation, while previous phonetic studies had been restricted to individual communities (e.g., Clarke 1991; De Wolf 1992; Gregg 1957; Woods 1991) or to single variables (e.g., Chambers 1973; Clarke, Elms, & Youssef 1995; Esling & Warkentyne 1993; Joos 1942). Unlike these previous studies, LAB examine a wide range of variables in a nationwide sample, using data from acoustic analysis of telephone interviews with thirty-three participants in urbanized areas from coast to coast. This allows them to develop a set of criteria for distinguishing Canadian from neighboring varieties of American English at the phonetic level as well as to put forward at least a superficial view of regional and social variation within Canadian English. Identifying at a broad level the major sound changes at work in Canadian English and the place of Canadian English within a taxonomy of North American English dialects was necessarily the main concern of a work dealing with the study of variation and change on a continental basis: finer details of the internal geographic or phonetic structure of Canadian English were left for future studies with a more Canadian focus, like that reported here. In addition to the low-back merger of /o/ and /oh/ (cot and caught; see Figure 2 for a full explanation of phonemic symbols), which is well established in Canada as in several regions of the United States, LAB report that Canadian English is characterized by several distinctive phonetic patterns not found to the same degree or with the same regularity in neighboring American varieties (2006, 217): (1) the Canadian Shift, a lowering and retraction of the short front vowels /i/, /e/, and /æ/ (bid, bed, and bad) in response to the low-back merger, which was first identified by Clarke, Elms, and Youssef (1995); (2) Canadian Raising, the pronunciation of the diphthongs /aw/ and /ay/ (house and tight) with nonlow nuclei before voiceless obstruents, first noted by Joos (1942) and further analyzed by Chambers (1973), among others; (3) the articulation of the long vowels /ey/ and /ow/ (day and go) with tense, peripheral nuclei that approach monophthongal forms; and (4) the articulation of /aw/ (cow) in relatively back position, a feature shared with some neighboring parts of northern United States. In keeping with the general practice of the Atlas, LAB give each of these variables a quantitative, phonetic definition derived from acoustic measurement of the nuclear values of the first and second formants in tokens of each vowel produced by each participant: the Canadian Shift is defined as the F1 of /e/ (bed) being greater than 650 Hz, the F2 of /æ/ (bad) being less than 1825 Hz, and the F2 of /o/ (cot) being less than 1275 Hz; Canadian Raising is defined as the difference between the raised and unraised nuclei of /aw/ and /ay/ (house vs. cow; tight vs. tie) being greater than 60 Hz in the F1 dimension; front /ey/ (day) is defined as F2 greater than 2200 Hz and back /ow/ (go) as F2 less than 1100 Hz; and back /aw/ (cow) is defined as F2 less than 1550 Hz. Within the political boundaries of Canada, a coincidence of these Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 Boberg / Regional Differentiation in Standard Canadian English 131 five phonetic isoglosses defines an “Inland Canada” region that constitutes the main type of Canadian English, containing the large majority of Canada’s English-speaking population. The outer periphery of this region, defined by the Canadian Shift, extends some 2,800 miles (4,500 km) from Vancouver in the west to the Englishspeaking community of Montreal in the east. Its inner core, defined by the remaining isoglosses, extends from Edmonton in the northwest to Toronto in the southeast (224). Atlantic Canada, including the Maritime provinces and Newfoundland, is excluded from Inland Canada, forming a separate subregion of uncertain status. Regional divisions within Canada arise from several features: the Canadian Shift, which excludes Atlantic Canada (219); the merger of /e/ and /æ/ before /r/ (marry and merry), which excludes Montreal and Newfoundland (219); fronted /ahr/ (car, hard), which distinguishes Atlantic Canada (222); and the allophonic distribution of /æ/, which is raised before nasals in the Loyalist-settled regions of Ontario and the Maritimes but not in Montreal or the West (223), and raised before /g/ west of Quebec (182). A vestige of a Mid-Atlantic-style split short-a (tense /æh/ before fricatives and voiced stops contrasting with lax /æ/ elsewhere) was found in some older Maritime speakers in Saint John and Halifax (223). These regional divisions amount to a tripartite view of Canada, with major isogloss bundles separating Inland Canada from Vancouver in the west and from Atlantic Canada in the east, as seen in LAB’s map 15.7 (224), reproduced here as Figure 1. Several aspects of LAB’s study of Canada suggest that further research might produce a modified or at least more detailed view of regional Canadian English phonetics. To begin with, thirty-three participants was a reasonable sample for distinguishing Canadian from neighboring varieties of American English, LAB’s main purpose, but not for achieving more than a superficial view of regional divisions (much less social divisions) within Canada. Uncertainties arising from the small sample are compounded by the method of recruitment, which was essentially random, involving interviews with the first two people in each city who answered the telephone, were willing to be interviewed, and met the criteria established for avoiding nonlocal speakers. This resulted in a socially diverse sample that, while beneficial in some respects, introduced an important confound in a regional analysis based on small numbers: there was no guarantee that what appeared to be a regional difference between two places might not instead be a social difference aligned with age, sex, social class, or some other factor. A final uncertainty in LAB’s view of Canadian English arises from the type of data analyzed: aside from a few minimal pairs for each speaker, the Atlas data come mostly from spontaneous speech. This has obvious advantages, bringing the analysis closer to the ultimate object of description, but it introduces a further confound in interspeaker and regional analyses, as there is only limited control over the kind or frequency of allophonic environments in the data from individual speakers. Some crucial environments, like those involved in the analysis of Canadian Raising, were sometimes underrepresented in the data from spontaneous speech. Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 132 Journal of English Linguistics Figure 1 Inland Canada Edmonton St. John's Vancouver Calgary Saskatoon Sydney Regina Winnipeg St. John Thunder Bay Montreal SSMarie Canadian Shift F2(ow) < 1100 Hz F2(ey) > 2200 Hz F2(aw) < 1550 Hz Canadian raising of /ay/ and /aw/ Halifax Arnprior Toronto Ottawa London Windsor Source: Atlas of North American English (Labov, Ash, & Boberg 2006, 224, map 15.7). Reprinted with permission. These sources of uncertainty give rise to several questions for further research. The most important of these is whether an in-depth follow-up study using different methods and a different sample would confirm the Atlas view of Canadian regional phonetics or suggest a different view. Might it reveal additional phonetic divisions, perhaps challenging the tripartite model of Canadian English dialect geography proposed by LAB? These questions were investigated by a new study conducted by the author at McGill University, titled Phonetics of Canadian English (PCE), which is the basis for this article. Method: The PCE Study The PCE project is a study of regional, sex, and attitudinal effects on vowel production in Standard Canadian English. It involved tape-recorded sociolinguistic interviews with fifty-one female and thirty-five male McGill University undergraduate students from across Canada, conducted by fellow students. A breakdown of the sample by region and sex is shown in Table 1. The regional breakdown was motivated by the dialect divisions identified by the Atlas, which were to be tested in the new study, as well as by traditional and well-established regional divisions of Canada along geographic, political, demographic, cultural, and historical lines. The division between British Columbia and the Prairies, coinciding with the sparsely populated Rocky Mountains, will examine the status of the western edge of LAB’s Inland Canada region. That between the Prairies and southern Ontario, coinciding with a 1,300-mile (2,100-km) sparsely populated expanse of the Canadian Shield, will examine the major nonlinguistic regional division within LAB’s Inland Canada region. Those between southern Ontario, greater Toronto, and eastern Ontario will Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 Boberg / Regional Differentiation in Standard Canadian English 133 Table 1 Sample of Phonetics of Canadian English Participants, by Region and Sex Region British Columbia Prairies (Alberta, Saskatchewan, Manitoba, northwest Ontario) Southern Ontario Greater Toronto Eastern Ontario Quebec (mostly greater Montreal) Maritimes (New Brunswick, Nova Scotia) Newfoundland Total Female Male Total 8 8 4 7 12 15 4 4 5 8 9 5 51 3 4 4 5 7 1 35 7 8 9 13 16 6 86 examine the linguistic status of the three major regions of Ontario, Canada’s most populous province; those between eastern Ontario, Quebec, and the Maritimes will examine the eastern edge of LAB’s Inland Canada region; and that between the Maritimes and Newfoundland will reflect the special status of Newfoundland as the best-known linguistic enclave in Canada. The participants of the present study all grew up entirely in the regions they represent, as did their parents; their first extensive and long-term exposure to nonlocal peer groups came with their moving to Montreal to attend McGill, usually at the age of eighteen. They can therefore be taken to be good representatives of the local speech of each region, at least for their generation and at the middle-class social level typical of McGill’s undergraduate population. The interviews elicited demographic information on each participant, a word list, conversation, and a written opinion survey. They were recorded on Type II (CrO2) analog cassette tapes using Marantz PMD 221 cassette recorders and Audiotechnica AT 803b omnidirectional lavalier microphones. In order to eliminate phonetic, prosodic, lexical, and other linguistic variables from consideration, the regional analysis presented here will focus exclusively on data from the word lists, which provide a uniform set of data from each participant: 145 productions, covering all of the vowels of English in fully stressed position in a range of allophonic environments. In the word list, each vowel appears at least once before /t/, /d/, /n/, /l/, and /r/, while long vowels also appear in final position. Allophonic environments of particular interest are represented by more than one token, such as those involved in Canadian Raising and the Canadian Shift. The complete word list, which also includes variables of phonemic incidence and consonantal variables not of interest here, bringing the total to 180 words, appears as the appendix. The 145 tokens from each participant’s word list representing the allophonic range of each vowel were subjected to acoustic analysis, using the same equipment Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 134 Journal of English Linguistics and method of analysis as in the Atlas (LAB 2006, 36-40): single-point nuclear measurement of F1 and F2 using linear predictive coding analysis in CSL 4400 (Kay Elemetrics). Within each vowel nucleus, measurements were made at a diagnostic point in the trajectory of one or both formants: at the maximal value of F1 in the case of a vowel whose central tendency is the lowering and raising of the tongue; at a point of inflection in F2 in the case of a vowel whose central tendency is movement of the tongue toward, then away from the front or rear periphery of the vowel space. In the absence of clear points of inflection, a measurement point was selected within the steady state extending through the middle of the nucleus. As in the Atlas, the data from each participant were then normalized, using the additive point system of Nearey (1978), in which the raw formant values of each speaker in a group are adjusted (up for men and down for women) by a scale factor derived from the difference between the natural log means of the speaker’s and the group’s formant values. In the PCE database, the group mean of F1 and F2 values taken together was 1119 Hz, of which the natural log is 7.02; scaling factors ranged from 0.84 for the woman with the highest voice to 1.2 for the man with the lowest voice. Following normalization, mean formant values for each vowel and allophone were calculated for each participant (a mean of the individual values for several tokens), whereupon regional means for each vowel and allophone were calculated from the mean values of the participants in each regional group. Finally, multivariate analysis of covariance (MANCOVA) tests were run in SPSS to determine the significance of the effect of region on the phonetic measures, independent of sex. As the balance of male and female participants in each region was similar, with the exception of Newfoundland, the effects of sex on vowel production will be set aside for future treatment; we will be concerned here only with regional differences. The lack of male participants from Newfoundland will also be of little concern, as the speech of Newfoundland—a subject deserving a much fuller treatment of its own—lies outside the focus of this report. A methodological comparison of the Atlas and the present study is presented in Table 2. Essentially, the present study involves substantially more data from a more tightly controlled sample, allowing the analysis to focus with greater confidence on regional differences to the exclusion of other factors. The cost of this greater control is that fewer social types of speech are represented and an apparent-time analysis of age differences is not possible, but social and diachronic variation will not concern us here. Results The basic results of the acoustic analysis of the 145 word list tokens for each Canadian participant are given in Figure 2, which shows the mean values of F1 (height) and F2 (advancement) for each vowel and its important allophones for the eighty-six participants. The symbols used for broad phonemic transcription closely follow the Atlas (LAB 2006, 11-15) and the work of Labov more generally; they are Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 Boberg / Regional Differentiation in Standard Canadian English 135 Table 2 Methodological Comparison of the Atlas and PCE Studies Criterion Sample size Social characteristics Technology Source of data Atlas PCE 33 Diverse Telephone to tape, CSL Spontaneous speech, minimal pairs 86 Young, middle class In person to tape, CSL Word list (for present analysis; other data available) Note: Atlas = Atlas of North American English (Labov, Ash, & Boberg 2006); PCE = Phonetics of Canadian English explained in the key below the figure. The numerical data on which the figure is based are given in Table 3. Both sets of data reflect a regionally balanced mean—a mean of the eight regional means—thereby avoiding any bias in favor of regions represented by a greater number of participants. This is the vowel system of Standard Canadian English as spoken by young, university-educated Canadians from across the country. It provides a framework in which the developments discussed below can be understood to be taking place. As such, it is of course an abstraction developed for purposes of comparative analysis. While many aspects of this system are common to most or all Canadian English speakers, it should be emphasized that this is not the vowel system of any one region of Canada, much less of any individual, but serves rather as a national average against which the systems of particular regions or individuals can be compared, and which can itself be compared to vowel systems that occur in other parts of the English-speaking world. Of particular importance are the three-way merger of /ah/, /o/, and /oh/ (father, cot, and caught) in low-back position; the separation of the main distribution of /æ/ in low-front position from its allophones before /g/ and nasals in lower-midfront position; the separation of the main distributions of /aw/ and /ay/ in low-central position from their allophones before voiceless obstruents in lowermidcentral position; the clearly northern orientation of the front vowels, with /i/ and /e/ much lower and more retracted than the corresponding long vowels, /iy/ and /ey/, which are strongly peripheral; and the large difference between /uw/, which shows advanced fronting compared to its allophone before /l/ in high-back position, and /ow/, which shows much more moderate advancement compared to its allophone before /l/ on the rear periphery. While the large majority of Canadian English speakers share this system, several of the means shown in Figure 2 nevertheless conceal significant regional variation, which will be the main concern of the following sections. The Canadian Shift A MANCOVA of the effect of region and sex on the F1 and F2 of the six short vowels /i, e, æ, o, ∧, u/ found no statistically significant influence of region on the Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 136 Journal of English Linguistics Figure 2 Mean F1 and F2 Measurements for Vowel Phonemes and Major Allophones of Standard Canadian English (Balanced Mean of Eight Regions) F2 (Hz) 2600 2400 2200 2000 1800 1600 1400 1200 1000 800 300 iy uwl uw 400 uwr ohr u er eyr ær 500 owl, oy, owr 600 ow awT æN or 700 e ahr ayT æg F1 (Hz) r i ey ah, o, oh 800 ay awn æ 900 aw 1000 Key to Symbols Short vowels i e er æ æg æN ær o or r u Front upgliding sit set berry sat bag band, ham carry cot, bother borrow cut curry cook iy ey eyr ay ayT oy see say pair tie, side tight, spice toy Back upgliding aw awn awT ow owl owr uw uwl uwr cow, loud down doubt, house go, code cold pour do, food pool poor Monophthongs ah ahr oh ohr father, spa bar, dark caught, saw short bird Canadian Shift as a whole (F = 1.162; p = .193). In general, then, this appears to be a pan-Canadian development, at least among middle-class youth, contrary to the report of LAB that Atlantic Canada does not participate in it. However, region did have (marginally) significant effects on the F2 of /e/ (p = .061) and the F1 of /u/ (p = .024). The former measure reflects the retraction of /e/, found by Boberg (2005a, 141) to be the most active phase of the Shift in current Montreal English. The regional means Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 Boberg / Regional Differentiation in Standard Canadian English 137 Table 3 Mean F1 and F2 Measurements (in Hz) for Vowel Phonemes and Major Allophones of Standard Canadian English, with Standard Deviations (Balanced Mean of Eight Regions) Vowel Class /iy/ /ey/ /eyr/ /ay/ /ayT/ /oy/ /aw/ /awn/ /awT/ /ow/ /owl/ /owr/ /uw/ /uwl/ /uwr/ /i/ /e/ /er/ /æ/ /æg/ /æN/ /ær/ /o/ /or/ /∧/ /∧r/ /u/ /ah/ /ahr/ /oh/ /ohr/ /3±/ M F1 SD M F2 SD 401 573 603 844 734 520 870 814 731 612 521 514 422 404 444 563 732 591 884 774 717 630 774 575 760 536 582 777 736 768 526 561 9 10 15 20 12 16 25 26 33 21 23 15 9 14 23 13 10 11 12 48 42 64 21 34 18 13 19 20 17 19 16 10 2494 2189 2113 1429 1652 928 1603 1841 1704 1294 936 940 1734 965 964 2043 1883 2019 1724 1951 2089 1956 1224 1049 1501 1433 1332 1211 1404 1211 1003 1543 43 44 35 30 23 17 15 46 54 44 30 19 76 49 86 33 35 36 32 84 66 55 26 52 40 40 36 18 81 18 27 33 for these measures are shown in Table 4. The F2 of /e/ was found to be significantly higher in the Prairie region than in all Ontario regions, and higher in Quebec than in southern Ontario, suggesting that Quebec and the Prairies are somewhat behind Ontario in at least this phase of the Shift. Nevertheless, the regional profile of the Canadian Shift is far from clear: while participants from Toronto and Vancouver did number among the group of a dozen speakers with the lowest F2 values for /e/ (lower than 1800 Hz), that group also included two young women from Newfoundland and Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 138 Journal of English Linguistics Table 4 Regional Differences in the Advancement of /e/ and the Height of /u/ (in Hz) Region British Columbia Prairies Southern Ontario Toronto Eastern Ontario Quebec Maritimes Newfoundland National mean M F2 (e) SD M F1 (u) SD 1890 1943 1844 1861 1854 1916 1893 1888 1891 23 20 29 27 26 22 19 42 89 619 564 586 607 576 567 571 541 581 14 12 17 16 15 13 11 25 47 at least one participant from every other region, suggesting that this change is proceeding on a nationwide basis. The regional identity of those with higher, more conservative F2 values for /e/ (higher than 2000 Hz) is somewhat clearer: this group contains six participants from the Prairie region, two from Quebec, and one from a small city in southern Ontario. While leadership of the Canadian Shift is regionally diffuse, resistance to it—or failure to adopt advanced forms of it—is found mostly in areas that are somewhat isolated from the main centers of English Canadian urban culture in Toronto and Vancouver. While the phonetic position of /u/ has not generally been regarded as related to the Canadian Shift, the F1 of /u/ was found to be higher in British Columbia than on the Prairies or anywhere east of Toronto; it was significantly higher in Toronto than on the Prairies or in Newfoundland. This suggests a lowering of /u/ in words like cook, foot, and stood that is centered particularly in Toronto. This development may indeed be completely independent of the Canadian Shift, since lowering of /u/ is not related to the lowering or retraction of the front short vowels in any obvious way. Of the nine participants with the highest F1 values for /u/ (greater than 640 Hz), three are from Toronto, three from Vancouver, two from Nova Scotia, and one from Ottawa, indicating a strong Ontario/British Columbia urban bias for this feature. By contrast, participants with low F1 values for /u/ tend to come from Atlantic Canada, Quebec, and the Prairies. Canadian Raising Although Canadian Raising is probably the best-known feature of Canadian English, LAB found that it was “not uniform enough to serve as a defining feature of the dialect of Canada” (2006, 221), both because it is variable within Canada and because it also occurs, especially with /ay/, in some neighboring varieties of American English. Nevertheless, LAB include an isogloss delimiting reasonably consistent Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 Boberg / Regional Differentiation in Standard Canadian English 139 Canadian Raising of /ay/ and /aw/ among the five that identify Inland Canada (224). As mentioned above, the analysis of Canadian Raising is particularly subject to fluctuations in the incidence of allophonic environments in natural speech data, making word list data particularly useful. The PCE word list allowed for a comparison of five tokens of /aw/ (cow, foul, loud, proud, sour) with four of /awT/ (doubt, house, shout, south) and of seven tokens of /ay/ (file, rider, side, sign, tide, tie, tire) with five of /ayt/ (fight, sight, spice, tight, writer); tokens of /aw/ before /n/ (down, gown, town) were treated as a separate allophone from the main distribution of /aw/ and not included in the analysis of Canadian Raising, since they tend to be fronted relative to the main distribution. As displayed in Figure 2 and Table 3, the PCE data reveal a difference of 142 Hz between the mean F1 values of raised and unraised /aw/, considerably more than the threshold of 60 Hz established for Canadian Raising by LAB. For /ay/, which is articulated further back than /aw/, as in midland and southern dialects of American English and in contrast to the relative positions of these diphthongs in the Inland North, Canadian Raising has a bigger effect on F2 (raised tokens are on average 229 Hz further forward) than on F1 (raised tokens are 110 Hz higher), though this degree of F1 difference is also well above LAB’s critical level. Contrary to LAB and other previous reports [Chambers and Hardwick (1986); De Wolf (1992: 91, 99); Hung, Davidson and Chambers (1993); Woods 1993: 159-167], which found both regional differences among major cities and a recession of raising among younger speakers, the PCE data indicate that Canadian Raising is a largely uniform feature of Canadian English, perhaps all the more remarkable given the young age of the PCE informants: 88 percent of them showed a difference between the mean F1 values of raised and unraised /aw/ of 50 Hz or greater, 84 percent showed the equivalent difference for /ay/, and 92 percent showed a difference of greater than 50 Hz between the mean F2 values of raised and unraised /ay/. This relative uniformity makes it surprising that raising of the two vowels is only weakly correlated: a correlation test of F1 difference measures produces a value of only r = .51, while there is virtually no correlation between F1 differences for /aw/ and F2 differences for /ay/ (r = .20), suggesting that those speakers who raise /aw/ the most do not necessarily also raise /ay/ the most and that these vowels should be analyzed separately. A MANCOVA of the effect of region and sex was run on four measures of Canadian Raising: the F1 distance between /aw/ and /awT/; the F1 and F2 distance between /ay/ and /ayT/; and the F2 position of /awT/, which appeared to vary regionally in its relative advancement. Region was found to have a significant overall effect on this group of dependent measures (F = 1.815; p = .009). Within the group, significant regional effects were found for two of the /aw/ measures: F1 distance (F = 2.143; p = .050) and F2 position (F = 3.217; p = .005). Regional means and standard deviations for these measures are shown in Table 5. The significant effect of region on F1 distance principally reflects the variable character of raising in Newfoundland: of the six PCE participants from Newfoundland, three produced F1 (/aw/-/awT/) distance measures greater Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 140 Journal of English Linguistics Table 5 Regional Differences in the Canadian Raising of /aw/ and in the Advancement of Raised /awT/ (in Hz) Region M F1 distance (aw–awT) SD M F2 (awT) SD 123 168 176 128 156 117 153 37 142 21 18 26 24 23 20 17 38 72 1623 1617 1747 1760 1754 1684 1702 1750 1692 32 27 39 36 35 29 26 56 109 British Columbia Prairies Southern Ontario Toronto Eastern Ontario Quebec Maritimes Newfoundland National mean than 50 Hz while three did not. The Quebec–Prairies comparison of F1 distance was also found to be marginally significant (p = .056), reflecting the difference between uniformly strong raising on the Prairies and variable raising in Montreal, where two participants had distance measures of less than 50 Hz. Though comparisons of British Columbia with other regions were not found to be significant, three of the eight participants from the Vancouver-Victoria region in particular were also found to have F1 distance measures of less than 50 Hz, possibly indicating a weakening of the Canadian pattern in the urban area of southwestern British Columbia, previously reported by Chambers and Hardwick (1986: 35–37), that would be substantiated by a larger sample. By contrast, only one of eight Toronto-area speakers showed raising of less than 50 Hz. Nevertheless, the isogloss drawn by LAB excluding Vancouver and Montreal from the domain of Canadian Raising in the core area of Inland Canada does not find support in the PCE data. Regional variation in Canadian Raising of /aw/ extends beyond the question of whether raising takes place to the phonetic quality of the raised allophone. While Chambers and Hardwick (1986) and Hung, Davison, and Chambers (1993) report impressionistic data indicating a fronting of unraised allophones of /aw/ in apparent time in cities across the country, the PCE data show a significant regional difference in the advancement of raised /awT/. In particular, British Columbia and the Prairies have /awT/ significantly further back than all other regions except Quebec. The data in Table 5 suggest that raised /awT/ is produced about 150 Hz further forward in southern and eastern Ontario than in western Canada, corresponding to phonetic values approximating [ε> ] and [∧< ], respectively. This regional difference was also noted among young Vancouverites by Chambers and Hardwick (1986: 37-41), who label it rounding, but was not directly addressed by LAB (a reflection of it can be found in Map 10.30, 107). Among the PCE participants, the six with the highest F2 values for /awT/ (greater than 1860 Hz) are almost all from Toronto or southern Ontario; only one comes from elsewhere, specifically from Gander, Newfoundland. Of the fifteen participants with F2 values under 1600 Hz, by contrast, ten come from ó ó Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 Boberg / Regional Differentiation in Standard Canadian English 141 western Canada and the other five from east of Ontario: there is not a single southern or eastern Ontarian participant in this group, emphasizing the strong association between fronted /awT/ and Ontario. At the other end of the range, the most extreme backing of /awT/ (F2 less than 1450 Hz) is found in two participants from Edmonton: here, the stock phrase out and about sounds not like oot and aboot, the American stereotype of Canadian speech, but like oat and a boat. Emeneau (1936) noted a similar pronunciation in Lunenburg, Nova Scotia, which apparently led to a merger of /awT/ and /owT/, such that couch and coach were homonyms. Of the fifteen PCE participants with F2 values for /awT/ under 1600 Hz, two were indeed from Halifax, the capital and largest city of Nova Scotia, confirming that back /awT/ is a general feature of mainland Nova Scotia English, but another was from Saint John, New Brunswick, suggesting that this feature may have a broader distribution in the Maritimes. However, none of the participants with back /awT/ shows evidence of the couch–coach merger. Position of Long, Up-gliding Vowels MANCOVAs of the effect of region and sex on the F2 of the long, up-gliding vowels /iy, ey, ay, oy/ and /uw, ow, aw/ found no significant overall effect on either set: in general, the positions of these vowels shown in Figure 2 are relatively uniform across the country. However, a significant individual effect was found for one measure in the front up-gliding set, the F2 of /ey/ (F = 2.641; p = .018), whereby /ey/ was most peripheral on the Prairies (mean F2 = 2263 Hz), with a phonetic quality approaching monophthongal [e:j]: Prairie values were significantly higher than those to the west and east in British Columbia, southern Ontario, Toronto, or the Maritimes, where /ey/ has a more diphthongal quality closer to [εj]. This supports the exclusion of Vancouver by LAB’s isogloss for F2 (ey) but introduces a division within Inland Canada not present in LAB, distinguishing the Prairie region, with extremely peripheral /ey/, from Ontario, with less peripheral values. As for the back up-gliding vowels, the identification of peripheral /ow/ and back /aw/ with Inland Canada in LAB was not supported by the PCE data: these variables appear to have no clear regional profile. However, a marginally significant effect was found for the F2 of /uw/ (F = 2.043; p = .061). This was assessed with six word list tokens covering the wide allophonic range across the top of the vowel space that is occupied by /uw/ when not followed by /l/. In ascending order of mean F2, these tokens are boots (F2 = 1343 Hz), food (1397 Hz), tooth (1648 Hz), too (1702 Hz), soon (1749 Hz), and do (1973 Hz). The most conservative tokens, boots and food, are only moderately advanced beyond the mean F2 of /uwl/ (967 Hz), still clearly in the high-back quadrant of the vowel space; tooth, too, and soon show advancement well into the center of the vowel space; do, the most advanced token, while retaining its identity as a back up-gliding vowel, has practically reached the F2 position of /i/ (2051 Hz), clearly in the high-front quadrant of the vowel space. Again, such Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 142 Journal of English Linguistics extreme effects of phonetic environment emphasize the value of word list data for purposes of strict comparison. The regional analysis of this set of tokens as a whole found that /uw/ is significantly further back in Atlantic Canada (mean F2s less than 1650 Hz) than in Toronto, southern Ontario, or British Columbia (mean F2s greater than 1800 Hz), suggesting that Atlantic Canada lags behind the rest of the country in the fronting of /uw/. Unlike in LAB (2006, 101, 156), New Brunswick was not found to be a general exception to the Atlantic Canadian resistance to fronting: of five New Brunswick speakers, only two had advanced values for the F2 of /uw/; the other three had values consistent with the lower means for Atlantic Canada. The lowest values for the F2 of /uw/ were found in both speakers from Cape Breton (northern Nova Scotia) and two speakers from Montreal (where Boberg [2004, 554–56] showed back /uw/ to be especially characteristic of Italian-origin Montrealers) as well as in single speakers from St. John’s, Newfoundland, and Edmonton, Alberta. Speakers most advanced in the fronting of /uw/ do not show a clear regional concentration, suggesting that the position of this vowel is affected as much by social as by regional factors, along with phonetic environment. Merger of /e/ and /æ/ before /r/ Most of Canada displays a conditioned merger of /æ/ and /e/ before /r/ in midfront position, so that marry sounds like merry, Harold like herald, and so on; in addition, both of these classes are merged with /ey/ before /r/, so that marry and merry both sound the same as Mary. A MANCOVA of the effect of region and sex on the F1 of /ær/ (F = 4.271; p = .000) confirmed LAB’s report that the regional exceptions to the merger of /ær/ and /er/ (though not of /er/ and /eyr/) are Quebec and, less consistently, Newfoundland, as shown in Table 6. At 764 Hz, the mean F1 of /ær/ is significantly lower in Quebec than every other region, while that in Newfoundland is significantly lower than that of the Prairies. Table 6 also shows that the distance between the prerhotic allophones of /æ/ and /e/ is over 100 Hz in Quebec and Newfoundland, while it is virtually nil in the rest of Canada, indicating a complete merger. The absence of this merger is a feature that Montreal English shares, rather mysteriously, with the midAtlantic coastal region of the United States; however, resistance to the merger in Montreal could just as well represent a retained feature of British English (also a likely explanation in Newfoundland) as indicate a linguistic affinity with New York City or Philadelphia. In fact, resistance to the merger is a consistent feature only of urban Montreal English, not of Quebec English in general: of the thirteen Quebec participants, the nine from greater Montreal have a mean F1 distance between /ær/ and /er/ of 218 Hz, with individual mean distances ranging from 112 Hz to 310 Hz, indicating solid distinctions; the four participants from smaller English-speaking communities outside greater Montreal have a mean F1 distance of only 35 Hz, similar to that in the rest of Canada, with only one speaker showing a distinction, and this of only 99 Hz difference between the F1 means of /ær/ and /er/, below the bottom of the Montreal range. Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 Boberg / Regional Differentiation in Standard Canadian English 143 Table 6 Regional Differences in the Height of /ær/ and in the Height Difference between /ær/ and /er/ (in Hz) Region Quebec Newfoundland Rest of Canada M F1 (ær) SD M F1 distance(ær–er) SD 764 692 598 110 111 14 161 108 7 105 84 40 Fronting of /ahr/ LAB (2006, 111, 222) report that the relative advancement of /ahr/ varies across Canada, with backer variants in the West and fronter in the East; in fact, fronted /ahr/ is the only positive feature that LAB found to be associated with their Maritime Canadian region (141, 222). On the PCE word list, /ahr/ was represented by five tokens: car, dark, harp, star, and start. A MANCOVA of the effect of region and sex on the phonetic position of /ahr/ in these data confirmed and refined the Atlas view, with a significant effect of region (F = 4.973; p = .000) that is entirely due to F2, or advancement (F = 8.983; p = .000), rather than to height. In fact, the relative advancement of /ahr/ is one of the strongest regional indicators of Canadian English, with mean F2 values increasing steadily from west to east across Canada, except for a sharp decline in Quebec, as shown in Figure 3. Along this spectrum, statistically significant differences were found between the Prairies and southern Ontario and between Quebec and its neighbors to the west and east. Ontario and Atlantic Canada, then, show significantly fronter values for /ahr/, approaching [å®], than the West or Quebec, which have values closer to [∧®]. Looking at individual participants, the lowest values for the F2 of /ahr/ (twenty participants with means of less than 1300 Hz) all come from western Canada and Quebec, with only two exceptions from southern or eastern Ontario; the most extreme values at this end of the scale (less than 1200 Hz) come from northwestern Ontario (Sault Ste. Marie and Sudbury), indicating that the centralized /ahr/ of Ontario and Atlantic Canadian speech stops abruptly somewhere north of greater Toronto. The highest values for the F2 of /ahr/ (1500 Hz and greater) all come from southern Ontario and eastward, with a particular concentration in Nova Scotia; the most extreme speaker, with a mean F2 of 1706 Hz, is from Truro, Nova Scotia. LAB’s map 10.34 (2006, 111) suggests that this regional difference may have originated in the historical settlement of Canada, with fronted (or, to be more historically accurate, unbacked) variants of /ahr/ spreading from New England northward into Loyalist-settled parts of Canada and westward into Yankee-settled parts of the American Inland North. These territories now form a continuous belt from Nova Scotia through Massachusetts to southeastern Minnesota, while less fronted variants Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 144 Journal of English Linguistics Figure 3 Mean F2 of /ahr/ (in Hz) by Region from West to East across Canada 1550 1500 Mean F2 (ahr) 1450 1400 1350 1300 1250 F2 (ahr) BC PR SO TO EO QC MT NL 1303 1315 1439 1416 1438 1344 1511 1524 Region Note: BC = British Columbia; PR = Prairies; SO = southern Ontario; TO = greater Toronto; EO = eastern Ontario; QC = Quebec; MT = Maritimes; NL = Newfoundland. predominate elsewhere. Today, the (r) isogloss is superimposed on this pattern, so that Atlantic Canada and the Inland North have an /r/-ful variant of the fronted vowel, while eastern New England has an /r/-less variant. Allophones of /æ/ LAB show the phonemic status and phonetic conditioning of æ in words like bat, bad, bag, and band to be one of the most intricate and regionally diagnostic variables in North American English (2006, 173-84). For instance, these vowels are split into lax /æ/ and tense /æh/ phonemes in the Mid-Atlantic region (bat with /æ/ and bad and band with /æh/; bag with /æh/ in New York City but /æ/ in Philadelphia) but tensed and raised to midfront position as a single phoneme, /æh/, in the Inland North, the initiating phase of the Northern Cities Shift. In most of New England and the Midland, a “nasal system” prevails, in which a single phoneme, /æ/, is raised only but always before the front nasals /n/ and /m/. In most of the West, the nasal system gives way to a variety of less clearly organized systems involving a continuous range of Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 Boberg / Regional Differentiation in Standard Canadian English 145 allophones, though with nasals always showing the most raising on the continuum. Only the northwestern United States and Canada show an exception to this pattern of conditioning: in a region extending from Wisconsin to Seattle to Anchorage, Alaska, and across western and central Canada from Vancouver to Ottawa, allophones of /æ/ before /g/ are also raised and fronted, joining the prenasal tokens in midfront position. Raising before /g/ is phonetically distinct from raising before front nasals, arising not from tensing and raising of the nucleus per se but from assimilation of nuclear height and advancement to the high-front transition demanded by a following voiced velar consonant. The direction of the glides in the two types of raising is opposite: those before front nasals have in-glides, toward [´], whereas those before voiced velars, /g/ and /N/, have up-glides, toward [j]. This glide reversal causes a change across subsystems, so that short /æg/ effectively becomes long, up-gliding /eyg/, merging with the /eyg/ of plague, vague, and so on, a neutralization similar to the one that prevents tense–lax distinctions before /N/ in all English dialects (this merger was first noticed by Zeller [1997]). In some places, /eg/ is also involved in this development: either because the merger of /æg/ and /eyg/ is avoided by a laxing of /eyg/ to /eg/ (so that bag is [bεjg] or [bε:g] but vague is [vεg] or [vε:g]), or because /eg/ shows a similar phonetic development to that of /æg/, so that lag and leg, and bag and beg, are indistinguishable, each having [-εjg] or [-e:g]. The merger of /æg/ and /eyg/ was shown in the Atlas to occur regularly in Wisconsin and Minnesota and neighboring communities in Canada (Thunder Bay and Sault Ste. Marie, Ontario) as well as more sporadically across the Canadian Prairies (2006, 182, map 13.5). The PCE word list includes four tokens of /æN/ (band, ham, stamp, and tan) and three of /æg/ (bag, gag, and tag) as well as six tokens of /æ/ in the “elsewhere” environment (bad, sack, sad, sat, tally, and tap). Analyses of these tokens produced results that largely support the Atlas view of (æ) in Canada. All regions of Canada show raising of prenasal /æ/, with the mean Cartesian or diagonal distance between /æ/ and /æN/ ranging from between 300 and 400 Hz in Western Canada, Quebec, and Newfoundland to over 400 Hz throughout Ontario and in the Maritimes; the most extreme separation of the allophones occurs in southern and eastern Ontario, where the Cartesian distance between them exceeds 500 Hz. The lowest and backest variants of /æN/ occur in Quebec, where Boberg (2004, 556) shows that raising of /æN/ is ethnically conditioned, with British-origin speakers raising like Ontarians but Italian and Jewish-origin speakers producing very little raising at all. Turning to /æg/, the PCE data reveal a similar pattern: all regions show raising of /æg/, with substantially less raising in Quebec and Newfoundland: the Cartesian distance between /æ/ and /æg/ is generally between 100 and 200 Hz in Quebec and Newfoundland but over 200 Hz in the rest of Canada. However, unlike prenasal raising, raising before /g/ is greatest not in Ontario and the Maritimes, but in western Canada, including the northwestern region of Ontario identified in the Atlas as displaying the merger of /æg/ and /eyg/: on the Prairies, the mean Cartesian distance between /æ/ and /æg/ is over 350 Hz. Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 146 Journal of English Linguistics Table 7 Regionally Significant Measures of Allophonic Conditioning of /æ/, in Hz (Standard Deviations in Parentheses) Region British Columbia Prairies Southern Ontario Toronto Eastern Ontario Quebec Maritimes Newfoundland National mean Mean Cartesian Distance (æ–æN) Mean Cartesian Distance (æ–æg) Mean Difference Between (æ–æN) and (æ–æg) 392 (147) 354 (143) 517 (143) 458 (73) 526 (148) 312 (175) 412 (128) 286 (81) 404 (151) 323 (129) 352 (161) 323 (115) 244 (92) 311 (95) 149 (78) 243 (75) 147 (88) 270 (130) 70 (189) 3 (171) 194 (195) 213 (150) 215 (132) 164 (157) 169 (143) 139 (72) 135 (172) A MANCOVA was carried out to test the effect of region and sex on three measures of the Cartesian distance between /æ/ and its allophones before front nasals and /g/ along the front periphery of the vowel space: /æ/ to /æN/, /æ/ to /æg/, and the difference between /æ–æN/ and /æ–æg/ (the first two distance measures). Region had a significant overall effect on this group (F = 2.581; p = .000) as well as on each measure individually. Regional means for these measures are displayed in Table 7, along with standard deviations. The latter are fairly large, particularly in western Canada, indicating considerable interspeaker variability in the raising of /æ/; nevertheless, the MANCOVA identified many significant regional differences despite this variability. The MANCOVA confirmed that raising before nasals is strongest in Ontario: the Prairies and Newfoundland have significantly less raising than southern and eastern Ontario, while Quebec has significantly less raising than all of Ontario. Raising before /g/ was found to be strongest on the Prairies, where Cartesian measures of the distance between /æ/ and /æg/ are significantly higher than those in Toronto, Quebec, or Atlantic Canada. On the Prairies, a combination of moderate raising before nasals and extreme raising of /æ/ before /g/ eliminates any difference between the positions of these allophones, reflected in the low figure of 3 Hz in the third column of Table 7; British Columbia’s figure of 70 Hz is also substantially lower than those of central and eastern Canada, which are all well over 100 Hz. The MANCOVA confirmed that the Prairie value for this measure is significantly lower than that of every other Canadian region except British Columbia and Newfoundland. These data indicate an important isogloss dividing Ontario from western Canada: east of this division, /æ/ is raised more before nasals than before /g/; west of it, raising of /æ/ before /g/ approximates prenasal raising. Regional differences in the relation between prenasal and prevelar raising of /æ/ are displayed graphically in Figure 4, which plots the Cartesian distance in Hertz Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 Boberg / Regional Differentiation in Standard Canadian English 147 Figure 4 Raising of /æ/ (in Hz) before Nasals and /g/, by Region Cart.dist. (Hz) /æ/-/æg/ 400 PR 350 BC SO 300 EO 250 200 QC MT TO 150 NL 100 300 350 400 450 500 550 Cart.dist.(Hz)/æ/-/æN/ Note: BC = British Columbia; PR = Prairies; SO = southern Ontario; TO = greater Toronto; EO = eastern Ontario; QC = Quebec; MT = Maritimes; NL = Newfoundland. between /æ/ and /æN/ against that between /æ/ and /æg/ for each region. The graph clearly shows four pairs of regions, well separated by this diagnostic. In the lower left are Quebec and Newfoundland, which show relatively little raising in either environment. In the upper right are southern and eastern Ontario, which show the strongest prenasal raising of /æ/ together with considerable prevelar raising. In the upper left are British Columbia and the Prairies, which show the strongest prevelar raising of /æ/, approximating their moderate raising of prenasal /æ/. In the middle are greater Toronto and the Maritimes, which show moderate values for both types of raising, Toronto with slightly more prenasal raising, the Maritimes with slightly less. For 17 of the PCE participants, the distance between /æ/ and /æg/ subtracted from the distance between /æ/ and /æN/ actually produced a negative figure, indicating greater separation of /æg/ from /æ/ than of /æN/ from /æ/, a striking reversal of the usual order of constraints on /æ/ raising. Such speakers occur in every region except Newfoundland, but the preponderance—ten of the seventeen—are from western Canada. By contrast, only one such speaker can be found in each of southern Ontario, Toronto, eastern Ontario, or Quebec, and three in the Maritimes. This indicates that Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 148 Journal of English Linguistics the raising of /æg/ beyond /æN/ is largely a western Canadian phenomenon. The five most extreme speakers in this regard, with /æg/ more than 150 Hz higher than /æN/, are all from the West: Vancouver, Edmonton, Vernon (British Columbia), Red Lake (northwestern Ontario), and Swift Current (Saskatchewan); the last speaker produced a measure of –383 Hz, a complete reversal of the usual positions of these allophones, with strong raising of /æg/ but comparatively little raising of /æN/. Crucially, however, the four speakers from communities across northwestern Ontario—Thunder Bay, Red Lake, Sault Ste. Marie, and Sudbury—do not show a consistent pattern in the allophonic conditioning of /æ/ that would allow them to be categorized clearly with either the Prairies or with southern Ontario, as was possible with (ahr). Measured in terms of the height of /æN/, the speaker from Red Lake shows relatively low, Prairie-like values (higher F1); the speaker from Sault Ste. Marie shows relatively high, Ontario-like values (lower F1); while the speakers from Thunder Bay and Sudbury show medial values. This suggests that northwestern Ontario forms a transition zone between the speech of southern Ontario and that of the Prairies, the nature of which could be effectively elucidated only by a local study of that region with a larger sample. Discussion The PCE data support a somewhat different view of regional phonetic variation in Canadian English than that of LAB. The Canadian Shift appears to be a national development, without a sharp isogloss excluding Atlantic Canada, but with Ontario ahead of Quebec and the Prairies in the retraction of /e/, one of its main components. With respect to Canadian Raising, the PCE data show above-average raising of /aw/ on the Prairies, in southern and eastern Ontario, and in the Maritimes. This is only partially consistent with the LAB isogloss (2006, 222, map 15.5), which shows no raising in New Brunswick, variable raising on the Prairies, and strong raising in Toronto. LAB do not report variation in the F2 of raised /aw/, which was found here to be a strong regional indicator, with fronter vowels in Ontario and backer vowels in the West and the Maritimes. The PCE data do support the LAB view of all of Canada except the Atlantic region having advanced fronting of /uw/, but find a minor regional difference in the peripherality of /ey/ that was not detected by LAB (95, map 10.18), with peripheral /ey/ most characteristic of the Prairies. The PCE data also support the LAB view of the regional distribution of the merger of /ær/ and /er/, but motivate a different view of /ahr/, without the variability in Alberta or the fronting in Montreal reported by LAB (111, map 10.34). Finally, the PCE data on the allophones of /æ/ match the corresponding LAB data closely, with raising before nasals strong in Ontario and the Maritimes but weak in Quebec and the Prairies (223), and raising before /g/ excluding Quebec and Newfoundland (182, map 13.5), though the PCE data find raising before /g/ to be strongest on the Prairies, a regional difference not identified by LAB. Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 Boberg / Regional Differentiation in Standard Canadian English 149 Table 8 F2 (æ) – F2 (uw) by Region Region British Columbia Prairies Southern Ontario Toronto Eastern Ontario Quebec Maritimes Newfoundland National mean F2 (æ) – F2 (uw) SD –100 122 –95 –101 –76 35 121 230 7 70 59 87 81 76 65 57 125 251 At this point, a number of individual regional differences have been identified, but a more general picture of regional differentiation has not been presented. For instance, where do the most important regional divisions lie within Canadian English, and which parts of Canada can be seen as more innovative or more conservative? It is possible to develop a broader view in two ways. First, the variables under study can be categorized in terms of their status as traditional features or active changes in progress, based on the evidence gathered by LAB of the dynamic processes affecting North American English as a whole and Canadian English in particular. This evidence suggests that three of the variables discussed above—Canadian Raising, the fronting of /ahr/, and the allophonic conditioning of /æ/—are traditional features of Canadian English, while two others— the Canadian Shift and the fronting of /uw/—are changes in progress, the first apparently confined mostly to Canada, the latter affecting North America on a continental level. The combined effect of these changes operating in tandem is to reduce the distance in the F2 dimension between /æ/, traditionally a front vowel, and /uw/, traditionally a back vowel, so that the nuclei of the two vowels come to occupy similar positions, somewhat behind the front periphery of the vowel space. The conjunction of these changes—an index of innovation—can therefore be measured as the F2 of /uw/ subtracted from the F2 of /æ/: positive values indicate a relatively back /uw/ and front /æ/, the historical pattern, while negative values indicate a relatively front /uw/ and retracted /æ/, the innovative pattern. It turns out that this measure shows a clear regional pattern, revealed in a MANCOVA of the effect of region and sex on F2 (æ) – F2 (uw) (F = 2.495; p = .024). The regional index scores are displayed in Table 8. This measure shows the close approximation of /æ/ and /uw/ in F2 space across the country as a whole, with an index score of close to zero, but with a great deal of variation on either side (a standard deviation of 251 Hz). In fact, the national mean is a compromise between strongly positive scores on the Prairies, in the Maritimes, and in Newfoundland, indicating that these regions are more conservative in this respect, and strongly negative scores in British Columbia and the three regions of Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 150 Journal of English Linguistics Ontario, suggesting that these are the most phonetically innovative regions of the country. Quebec falls on the conservative side with the Prairies and Atlantic Canada, but less emphatically so. Perhaps not surprisingly, the regions showing the most innovative patterns are those focused on the two largest urban centers of English Canada, Toronto and Vancouver. Another more general view of the PCE data can be obtained by examining the number of phonetic differences that are aligned with each regional division. In this respect, the tripartite view of Canadian dialect geography developed by LAB receives some support from the PCE data: important isoglosses do separate British Columbia from the Prairies, and Atlantic Canada from Ontario, with Quebec an often distinct interstitial pattern. However, the PCE data indicate that LAB’s Inland Canada region contains an important internal division—not evident in the Atlas maps—between the Prairie region and Ontario, while Quebec emerges as a clearly distinct region in the PCE data. Specifically, British Columbia is divided from the Prairies by the F2 of /ey/, which is lower in British Columbia and higher on the Prairies, but united with the Prairies by low values for the F2 of raised /aw/ and the F2 of /ahr/, and by the proximity of the prenasal and pre-/g/ allophones of /æ/; these three features together define a western Canadian subregion spanning all four western provinces and parts of northwestern Ontario, thereby diminishing the primacy of the western border of LAB’s Inland Canada region. A greater challenge to LAB’s Inland Canada region comes from the substantial bundle of isoglosses that split it in half, dividing the Prairies from the southern regions of Ontario: participants from the Prairies have less retracted values for /e/, less fronted variants of raised /aw/, more peripheral values for the F2 of /ey/, less fronted variants of /ahr/, and less raising of prenasal /æ/. These five isoglosses, together with the features that unite the Prairies with British Columbia, suggest that the basic differences between the settlement histories of Ontario, founded by American Loyalists followed by large groups of Ulster Irish and Scots, and western Canada, peopled by a mix of Canadian, British, American, and European settlers from a wider range of origins, continue to have linguistic consequences today. While these differences are not of a similar magnitude to those found among the major regional varieties of American English, much less of British English, they nevertheless create subtle differences between the types of English that can be heard in Ottawa or Toronto and in Calgary or Vancouver; these differences are not necessarily detectable only to the ears of a trained phonetician. This presents a challenge, if only a minor challenge, to the conventional view that Canadian English is geographically homogeneous over the vast territory extending from Vancouver to Ottawa. It naturally comes as less of a surprise that there should be differences between the types of English spoken in Ontario, Quebec, the Maritimes, and Newfoundland, given the longer history of settlement and greater diversity of linguistic influences in these regions. The PCE data find that the English-speaking population of Montreal, isolated from those of Ontario to the west and New Brunswick to the east by the Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 Boberg / Regional Differentiation in Standard Canadian English 151 minority status of English in Quebec and by considerable geographic distances (especially to the east), exhibits only variable presence of three Canadian features: Canadian Raising of /aw/, the merger of /ær/ and /er/, and the raising of prenasal /æ/; Quebec English also has lower values for the F2 of /ahr/ than its neighbors. This suggests that these features were inherited from American English by means of the Loyalist settlement of the Maritimes and Ontario during and after the American Revolution, which had comparatively little influence on the establishment of Montreal’s English-speaking population. (While Quebec was a temporary refuge for large numbers of Loyalists immediately after the war, most of these were soon resettled further west in Ontario; most of Montreal’s English speakers came directly from Britain.) Atlantic Canada is distinguished by its resistance to the fronting of /uw/, while Newfoundland patterns with Quebec in showing the absence or variable presence of Canadian Raising of /aw/ and raising of /æ/ before /g/; their unique histories make these regions less Canadian, in linguistic terms, than the rest of Canada. The apparently limited extent of Newfoundland’s divergence from the rest of Canada in these data requires some qualification. As is well known, the phonetic and other distinguishing features of the varieties of English spoken in Newfoundland are many and varied, going well beyond the scope of this study. The few phonetic variables studied here, which have a general importance for Canada as a whole and are embedded in a general theory of the dynamic processes affecting North American English on a continental level, are not necessarily the most important of the dozens that could be studied in a focused investigation of the unique properties of Newfoundland English (see Clarke [2004] for a recent review of many of these properties). Moreover, the problems associated with analyzing the speech of university students living outside the province are particularly acute in the case of Newfoundland, so that the data presented here, whatever their value in studying middle-class Newfoundland English, cannot be taken as representative of the wide range of local varieties spoken by the majority of the population of Newfoundland. More generally, it must be admitted that the restrictions of the PCE sample— young, university-educated, mostly middle-class Canadians—tend to constrain the applicability of the PCE data to Canadian English as a whole, even beyond the obvious case of conservative enclaves like Newfoundland, Cape Breton, Lunenburg, or the Ottawa Valley. A study of working-class speech in any of the regions or communities discussed above might reveal even greater regional differences than those reported here, or additional variables that do not appear in middle-class speech. The PCE data likely represent the minimal extent of regional variation in the phonetics of Canadian English; the real picture can be assumed to be even more vivid once a broad range of social levels is taken into account. Nevertheless, it is interesting to know how much regional variation exists at this minimal, middle-class level; if it can survive where regional identities are weakest, its survival at more locally oriented social levels would seem to be assured. Several considerations, by contrast, moderate the social limitations of the sample. One is the observation that social variation in the sound of North American English Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 152 Journal of English Linguistics tends to involve consonants—(r), (ng), (-t, -d), (th, dh), and so on—whereas regional variation tends to involve vocalic variables, such as those studied here or in the Atlas of North American English. The variables of vocalic nuclear quality analyzed here have by far the most pervasive and systematic effect on the perception of regional speech: more than consonantal variables or morpho-syntactic variables, which vary socially across regions and apply only among a minority of speakers at the lower end of the social class spectrum, and more than lexical variables, which occur only sporadically in spontaneous speech. Another consideration is that many of the vocalic variables examined here seem to operate below the level of conscious awareness, being indicators of regional identity rather than markers of social class that respond to stylistic variation. The likelihood that any of the PCE subjects might have learned to control the Canadian Shift of /e/, the Canadian Raising of /aw/, or the allophonic conditioning of /æ/ is extremely small: for these variables, word list data are closely representative of vernacular speech. Finally, the PCE data presented here, though based on word list productions only and on a highly constrained social sample, show a close correspondence to the data of LAB, which reflect the spontaneous speech of a broader range of social types. This convergence of data from different sources suggests not only that word list data are of considerable value in the study of Standard Canadian English, but that the data obtained from the socially restricted PCE sample are not broadly different from those obtained from a wider social sample like that of the Atlas. Unlike in Labov’s famous study of New York City, in the case of modern, urban, Standard Canadian English, we are not dealing with a “sink of negative prestige” (1972, 136), introducing an element of chaotic distortion of the vernacular grammar in formally elicited speech. On the contrary, with the exception of a minority of nonstandard speakers in older rural areas, specific enclaves, and urban lower-class neighborhoods, most native Englishspeaking Canadians are comfortable with the way they speak, believing it to be inferior, perhaps, to the “best” British English but certainly superior to most, if not all, varieties of American English and very close to an imagined global standard. In this context, the PCE data have considerable value as a controlled comparison of regional productions of vocalic variables in Standard Canadian English. As such, the PCE data suggest that the primary dialect divisions of Canada at the phonetic level are between six rather than three regions: the West, Ontario, Quebec, the Maritimes, and Newfoundland, with a secondary division within the West between British Columbia and the Prairies. Northwestern Ontario constitutes a wide and sparsely populated transition zone between the West and southern Ontario. This picture closely matches the regional taxonomy based on lexical data developed in Boberg (2005b), suggesting that it applies to Canadian English as a whole, rather than only to the phonetic level of analysis. Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 Boberg / Regional Differentiation in Standard Canadian English 153 Appendix Word List Used in Phonetics of Canadian English Study PLEASE READ THE FOLLOWING WORDS AS YOU WOULD NORMALLY SAY THEM bar sit file student collar stare pajamas lasagna strong tally sort too sell sat fork start still sterile sorry palm down drama south bother doubt void fool caller caught bad steer dawn sour bold go monitor cow pair proud tide star sod Pakistani lager charity rider stamp spa saw horrible sale see coin Colorado steel tag whine stir soprano which tire stood band tin poor bird toe sack talk lava cut cot tooth seed stayed carry step mafia boots writer pour dark llama worry Slavic dead side relative town foot father soon due barrel sign foil sawed top Don sun gown loud sight foul borrow stain bag duck girl did bang new state code dull cool food stone house hanger gag Iraq singer tan whale dirt tour plaza say full tie ham car care tube sure cold stole sick sanity fight tap macho stud Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013 sad sore cook do spirit tight turn ferry set deck coat cup veto core seen tool curry pasta shout spice panorama façade taco berry sock Picasso harp avocado short toy tip ten boat hurry seat calm 154 Journal of English Linguistics References Boberg, Charles. 2004. Ethnic patterns in the phonetics of Montreal English. Journal of Sociolinguistics 8. 538-568. Boberg, Charles. 2005a. The Canadian Shift in Montreal. Language Variation and Change 17. 133-154. Boberg, Charles. 2005b. The North American regional vocabulary survey: Renewing the study of lexical variation in North American English. American Speech 80(1). 22-60. Chambers, J. K. 1973. Canadian raising. Canadian Journal of Linguistics 18. 113-135. Chambers, J. K. & Margaret Hardwick. 1986. Comparative sociolinguistics of a sound change in Canadian English. English World-Wide 7. 21-46. Clarke, Sandra. 1991. Phonological variation and recent language change in St. John’s English. In Jenny Cheshire (ed.), English around the world: Sociolinguistic perspectives, 109-122. Cambridge, UK: Cambridge University Press. Clarke, Sandra. 2004. Newfoundland English: Phonology. In Edgar W. Schneider, Kate Burridge, Bernd Kortmann, Rajend Mesthrie, & Clive Upton (eds.), A handbook of varieties of english, vol. 1: Phonology, 366-382. Berlin: Mouton de Gruyter. Clarke, Sandra, Ford Elms & Amani Youssef. 1995. The third dialect of English: Some Canadian evidence. Language Variation and Change 7. 209-228. De Wolf, Gaelan T. Dodds. 1992. Social and regional factors in Canadian English. Toronto, Canada: Canadian Scholar’s Press. Emeneau, Murray Barnson. 1936. The dialect of Lunenburg, Nova Scotia. Language 11. 140-147. Esling, John H. & Henry J. Warkentyne. 1993. Retracting of /æ/ in Vancouver English. In Sandra Clarke (ed.), Focus on Canada, 229-246. Amsterdam: John Benjamins. Gregg, Robert J. 1957. Notes on the pronunciation of Canadian English as spoken in Vancouver, B.C. Journal of the Canadian Linguistics Association 3. 20-26. Hung, Henrietta, John Davison & J. K. Chambers. 1993. Comparative sociolinguistics of (aw)-fronting. In Sandra Clarke (ed.), Focus on Canada, 247-267. Amsterdam: John Benjamins. Joos, Martin. 1942. A phonological dilemma in Canadian English. Language 18. 141-144. Labov, William. 1972. Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press. Labov, William, Sharon Ash & Charles Boberg. 2006. Atlas of North American English: Phonetics, phonology and sound change. Berlin: Mouton de Gruyter. Nearey, Terrance Michael. 1978. Phonetic feature systems for vowels. Bloomington: Indiana University Linguistics Club. Scargill, M. H. & Henry J. Warkentyne. 1972. The survey of Canadian English: A report. English Quarterly 5. 47-104. Woods, Howard B. 1991. Social differentiation in Ottawa English. In Jenny Cheshire (ed.), English around the world: Sociolinguistic perspectives, 134-149. Cambridge, UK: Cambridge University Press. Woods, Howard B. 1993. A synchronic study of English spoken in Ottawa: Is Canadian English becoming more American? In Sandra Clarke (ed.), Focus on Canada, 151-178. Amsterdam: John Benjamins. Zeller, Christine. 1997. The investigation of a sound change in progress: /æ/ to /e/ in midwestern American English. Journal of English Linguistics 25. 142-155. Charles Boberg is Associate Professor of Linguistics at McGill University in Montreal, Canada. He is co-author, with William Labov and Sharon Ash, of the Atlas of North American English (Mouton de Gruyter, 2006). His research focuses on variation and change in North American English, particularly English in Canada. Downloaded from eng.sagepub.com at MCGILL UNIVERSITY LIBRARIES on February 12, 2013