AAA - Arbeiten aus Anglistik und Amerikanistik Band 32 (2007) Heft 2 Gunter Narr Verlag Tübingen Canadian Raising in Victoria, B.C. An Acoustic Analysis Ingrid Rosenfelder This article presents the results of a study of the phonetic realization of Canadian Raising, as produced by 39 speakers from Victoria, B.C. A total of 2489 tokens of the diphthongs / a / and / a / in nine different phonetic environments were analyzed acoustically by measuring the first and second formants (F1/ F2). Contrary to the findings of Chambers & Hardwick (1986) and Hung, Davison & Chambers (1993), which predicted the “Americanization” of Canadian speech and a decline in the use of raising, Canadian Raising is still in existence for both diphthongs. Raising chiefly occurs in the phonetic environment of a following voiceless consonant, including the ‘flapped’ or ‘voiced’ / t/ , and involves a lower F1 (i.e. raising) of the offglides as well. Moreover, a trend toward fronting of the onsets can be discerned, with / a / being predominantly realized as fronted already, and / a / increasingly becoming so, especially in the phonetic environment of a following obstruent. 1 Introduction The term “Canadian Raising”, coined by Chambers in 1973, refers to the phenomenon by which the onsets of the diphthongs / a / and / a / are raised to [ ] and [ ], respectively, when followed by a tautosyllabic voiceless consonant (Chambers 1973: 94; Paradis 1980: 40/ 41). Canadian Raising has long been a stereotypical feature of Canadian speech and was already commented on as early as 1935 (Emeneau 1935). It received wider linguistic attention in 1942, when it was commented upon as a phonemic problem (Joos 1942) involving a possible phonemic split of the diphthongs in front of a following ‘voiced’ / t/ . Canadian Raising has been shown to go back as far as the 1880s (Thomas 1991: 162), but much controversy surrounds the origin of the phenomenon, for which various mechanisms have been proposed. While Ingrid Rosenfelder 258 some authors follow Joos (1942) in accounting for Canadian Raising through the phonetic mechanism of pre-fortis clipping and minimalization of articulatory distance between onset and offglide, other have argued for it to be a remnant of the diphthongization of Middle English and in the Great Vowel Shift (Gregg 1973), pointing out that one should speak more correctly of “lowering” instead of “raising”. A compromise explanation has been advanced by Trudgill (1984), who suggested that Canadian Raising can be seen as the result of dialect mixture and subsequent phonological reallocation. Furthermore, acoustic analyses of diphthongs perceived as “raised” by Thomas (1991) revealed that these diphthongs also featured higher offglides. Thus, according to another model (Moreton & Thomas 2004), the origin of Canadian Raising can be accounted for by the peripheralization of offglides before voiceless consonants, with the onsets of the diphthongs shortened and raised through coarticulation. The sociolinguistic aspects of Canadian Raising came into focus in the 1980s, when a sound change in progress was observed for the diphthong / a / in Toronto, Vancouver and Victoria, B.C. (Chambers & Hardwick 1986; Hung, Davison & Chambers 1993), which involved the fronting, non-raising and rounding of the onset. Fronting correlated with age, sex and phonetic environment, fronted onsets being favored by young, female respondents, and in the so-called “elsewhere” environment (i.e. other than before voiceless consonants). Non-raising, though a less significant phenomenon, clearly correlated with age in Vancouver, but only weakly so in Victoria. Rounding of the onset to [o] was restricted exclusively to Vancouver. The motivation for the sound changes was explained as an increasing “Americanization” of Canadian speech for fronting and non-raising, and phonetic assimilation for rounding. Despite its name, Canadian Raising also occurs in the Northern United States, as has been demonstrated by Vance (1987) for the diphthong / a / in Rochester, NY, and for both diphthongs in Ann Arbor, MI, by Dailey-O’Cain (1997). The latter found evidence for a sound change in progress, already completed for / a / but still going on with respect to / a / , with correlations between the raised variants and gender and age for the latter diphthong. She also found some degree of raising in phonetic environments other than those ‘traditionally’ predicted by Chambers (1973), namely before nasal-voiceless consonant clusters, and before / r/ . The present study examines the situation of the Canadian Raising diphthongs today, with special regard to the two sound changes mentioned above. Although the main focus is on the phonetics of Canadian Raising, sociolinguistic aspects are also taken into account. Moreover, while Canadian Raising has traditionally been mostly studied ‘by ear’, the present study attempts to shed new light on the phenomenon through acoustic analysis, by measuring the frequencies of the first and second formants (F1 and F2) of Canadian Raising in Victoria, B.C. 259 1 A reading passage was chosen over free conversational speech for purely practical reasons, as the easiest way of obtaining the same number of diphthong tokens in the same phonetic environments for each speaker. 2 Both terms for the same phenomenon appear in the literature. The term ‘flapped’ will be used throughout the present study. 3 These environments correspond to those investigated by Dailey-O’Cain (1997), allowing a comparison with the results of her study. the diphthongs. Following Dailey-O’Cain (1997), the phonetic realization of the onsets of both / a / and / a / is investigated in nine different phonetic environments. Furthermore, the acoustic realization of the offglides is examined, and statistical tests are performed to determine whether the use of the raised or fronted diphthongs correlates with age and/ or gender. 2 Methods 2.1 Data Collection Data for the present study were collected during a short visit to Victoria in September 2004. Subjects were students or faculty members of the University of Victoria, who were asked to read aloud a fairy tale 1 containing a number of / a / and / a / -tokens in different phonetic environments: • vlessC: before a voiceless consonant • flappedT: before a ‘flapped’/ ’voiced’ / t/ 2 • voicedC: before a voiced consonant other than nasals, / l/ and / r/ • Nasal: before a nasal • N+vlessC: before a nasal-voiceless consonant cluster • R: before / r/ • L: before / l/ • #: word-finally • V: before a vowel The reading passage also contained two ‘reference vowels’ in the words rather (/ a / ) and ungodly ([ ]) 3 . Recordings were made digitally at a sampling rate of 22050 Hz in a sound-treated recording room in the Phonetics Laboratory of the University of Victoria. Participation in the study was voluntary, and all participants were unaware of the purpose of the experiment. Speakers were also asked to fill out a background questionnaire soliciting information about their age, sex, native language(s) and geographical background (where the speakers and their parents had grown up; length of residence in Victoria; prior place(s) of residence). Ingrid Rosenfelder 260 4 Since speakers who had grown up outside of what Chambers (1973: 84) has defined as “heartland Canada” were excluded from the present study, the label ‘Can’ in Table 1 must be understood in the sense of ‘heartland Canada without British Columbia’. Figure 1: Age distribution of speakers. Although the aim was to record speakers from a broad range of ages, the majority of the subjects were college students in their early twenties (see Fig. 1). Speaker ages ranged from 17 to 50, with a median age of 22. Proportions of male/ female speakers were approximately equal, with a larger number of women among the younger subjects. Parents grown up in … Vic BC Can other Vic 1 BC 1 7 Can 3 2 14 other 1 2 1 7 Total 6 11 15 7 39 Table 1: Geographical background of speaker family. (Answers to questions “Where did your father/ mother grow up? ”) [Vic = Victoria, BC = British Columbia, Can = Canada] Only about a quarter of the subjects had actually grown up in Victoria. The remainder divide up approximately equally between British Columbia and the rest of Canada 4 . Within these categories, the proportions of male/ female Canadian Raising in Victoria, B.C. 261 5 i.e. the one most frequently used in everyday life 6 Second languages mentioned were French, Hungarian and Russian. speakers were approximately equal, with the exception of British Columbia, where women constituted about two thirds of the speakers who had grown up there. All speakers listed English as their first language 5 . Although for a substantial proportion of speakers (11), one or both parents had grown up outside of Canada, only few (3) speakers indicated bilingualism 6 . The distribution of parental geographical background can be found in Table 1. 2.2 Acoustic analysis Recordings for the present study were analyzed on a personal computer using Praat (Boersma & Weenink 2004). In order to eliminate the effect of adjacent speech sounds on the diphthongs, it was attempted to identify what Hertz (1991: 99) has called the “interior transition”, i.e. that part of the diphthong from the point where the transition from the preceding speech sound ceases (because the target has been reached) and the transition towards the second target of the diphthong begins, to the point where the target of the offglide has been reached and the transition towards the following speech sound begins. Formant measurements were taken at these two points, with formant settings (maximum number of formants, maximum formant frequency) adjusted so as to yield continuous formant trajectories overlaying the formant tracks observed in the spectrogram. Formant values were extracted from the labeled sound files with the help of a Praat editor script, which automatically adjusted formant settings as specified in the labels, and then summarized and plotted for each speaker using various Perl scripts. In total, 2489 tokens were analyzed, with four measurement values each (F1/ F2, onset/ offglide). Formant measurement error was determined to range approximately 100 Hz in both directions, mainly due to the decision of where to place the points of measurement. 2.3 Statistical analysis In order to determine whether differences in formant measurements were indeed significant, and whether any significant correlations existed between the variables involved, a number of statistical tests were conducted using the SPSS statistics software package. Differences in mean F1 and F2 values between the phonetic environments were investigated using two-way ANOVAs, with both speaker and phonetic environments as independent variables. A Levene test examined homogeneity of variances for the different groups. Furthermore, post-hoc Ingrid Rosenfelder 262 7 Exceptions are the following: For the men, vlessC and Nasal differ only at the .01 level, and vlessC and voicedC do not differ significantly at all. For the women, vlessC and voicedC differ at a level of .05 only. In addition, F1 values for flappedT are low enough for this group of speakers to differ from those for vlessC at a level of .01. tests were conducted to determine which of the groups differed significantly from each other. As neither group sizes nor variances were equal for the groups investigated, a Games-Howell test was employed. Pearson correlation coefficients were calculated to examine the relation between F1 values measured at the onsets and offglides of the diphthong, as well as to determine whether raising and fronting of the diphthong onsets, i.e. lower F1s and higher F2s, correlated with speaker age. 3 Results A general overview of the acoustic realizations of the Canadian Raising diphthongs, differentiated by phoneme and speaker sex, can be found in Figs. 2 and 3, which display median F1 and F2 values for the phonetic environments investigated in the present study. Error bars indicate second and third quartiles, i.e. the inner 50% of the data. Measurements of the two reference vowels in the words rather and ungodly are also displayed. Median and quartiles were chosen for plotting rather than mean and standard deviations because of the former’s greater robustness with respect to extreme outliers. Moreover, as quartiles are not necessarily symmetric, they give a better indication of the direction of variation. It should be noted, however, that all statistical tests employed in the analysis required the use of formant means. As differences between medians and means were generally small, this does not pose much of a problem. 3.1 F1 - Raising 3.1.1 Phoneme / a / ? Figs. 4 and 5 show box-and-whisker plots for F1 at the onset of the diphthong / a / for male and female speakers. Two lines of reference are given by the median formant values of the two vowels [a ] and [ ] in rather and ungodly. Most noticeably, both plots show that the F1 values for those environments in which raising would be expected according to the literature, vlessC and flappedT, differ from the other environments in that they lie below the line of reference provided by the median first formant values of the reference vowels. For the majority of cases, differences in mean F1 values between these and the other phonetic environments are indeed statistically significant at the .001 level 7 . Thus, there seems to be a general pattern among both Canadian Raising in Victoria, B.C. 263 men and women to favor vlessC and flappedT for the raising of the onset of / a / , with flappedT as the ‘leading’ raising environment. Lowest raising scores are found for R and #. Some raising also seems to occur for voicedC, whose F1 values show considerable overlap across the reference lines. All other environments lie somewhat in between and do not differ significantly from one another. 3.1.2 Phoneme / a / Median F1 values for the onset of / a / in the nine phonetic environments are shown in Figs. 6 and 7 for male and female speakers, respectively. Once again, median F1 values are lowest for flappedT, and the most striking differences are found between the two ‘traditional’ raising environments, vlessC and flappedT, and the remaining groups. All of these differences are highly significant (p = .001) for the men. For the women, the pattern is interrupted by the two ‘nasal’ environments (Nasal and N+vlessC), where no significant differences in mean F1 to the two Raising environments could be found. Thus, some degree of raising seems to take place in these environments among the female speakers but not among the men. In contrast to the findings for phoneme / a / , a clear differentiation takes place between voicedC and the Raising environments for both male and female speakers, which is significant at the .001 level for both. Ingrid Rosenfelder 264 Figure 2: Formant plot of diphthong onsets and offglides: Phoneme / a / . Symbols indicate median formant values in the different phonetic environments. Error bars represent second and third quartiles, i.e. the inner 50% of the data. (Note that the offglide before / l/ (L) lies in the onset region of the other phonetic environments.) Canadian Raising in Victoria, B.C. 265 Figure 3: Formant plot of diphthong onsets and offglides: Phoneme / a / . Symbols indicate median formant values in the different phonetic environments. Error bars represent second and third quartiles, i.e. the inner 50% of the data. Ingrid Rosenfelder 266 Figure 4: Median and quartiles for F1 (onset) of phoneme / a / , male speakers. Circles and asterisks indicate extreme values (located more than 1.5 box lengths from the median) and outliers (distance from median more than three box lengths), respectively. Reference lines representing median F1 values of the vowels in rather (solid line) and ungodly (dashed line) are also shown. Canadian Raising in Victoria, B.C. 267 Figure 5: Median and quartiles for F1 (onset) of phoneme / a / , female speakers. Circles and asterisks indicate extreme values (located more than 1.5 box lengths from the median) and outliers (distance from median more than three box lengths), respectively. Reference lines representing median F1 values of the vowels in rather (solid line) and ungodly (dashed line) are also shown. Ingrid Rosenfelder 268 Figure 6: Median and quartiles for F1 (onset) of phoneme / a / , male speakers. Circles and asterisks indicate extreme values (located more than 1.5 box lengths from the median) and outliers (distance from median more than three box lengths), respectively. Reference lines representing median F1 values of the vowels in rather (solid line) and ungodly (dashed line) are also shown. Canadian Raising in Victoria, B.C. 269 8 Note that, strictly speaking, this is not exactly the same relation as claimed by Thomas (1991: 156), who stated that diphthongs with raised offglides had onsets that sounded Figure 7: Median and quartiles for F1 (onset) of phoneme / a / , female speakers. Circles and asterisks indicate extreme values (located more than 1.5 box lengths from the median) and outliers (distance from median more than three box lengths), respectively. Reference lines representing median F1 values of the vowels in rather (solid line) and ungodly (dashed line) are also shown. 3.1.3 Offglides Another objective of the present study was to examine whether raising of the diphthongs is indeed a process involving both diphthong onsets and offglides, as claimed by Thomas (1991) and Moreton & Thomas (2004). According to their hypotheses, it is likely that raised diphthongs should feature higher degrees of raising in their offglides as well, as the original phonetic motivation for the raising process was the peripheralization of diphthong offglides before voicless consonants, with corresponding raising of the nuclei through coarticulation 8 . Therefore, possible correlations between Ingrid Rosenfelder 270 higher. Nor does this theory necessarily demand higher offglides for raised diphthongs, as Canadian Raising is seen as a subsequent phonologization of the mechanism outlined above, with differences in the nuclei originally conditioned by mere phonetic principles reinterpreted as phonological target differences (Moreton & Thomas 2004). However, finding a correlation between the two variables would lend support to this theory about the origin of Canadian Raising. the F1 values measured at diphthong onset and offglide were also investigated in the present study. Box-and-whisker plots of median F1 offglides for both phonemes and sexes are displayed in Fig. 8. It can be seen that F1 values in the raising environments are lower on average than those of the other environments. Highest median formant frequencies, i.e. lowest diphthong offglides, are found for / a / in the environments of following / r/ and / l/ , which differ very markedly from all other values, especially among the female speakers. High frequencies also occur before nasals, though to a much lesser extent. These effects, however, are not so much due to Raising as they are to a tendency to monophthongize / a / in these environments, especially before / l/ . Although correlations significant at the .01 level between the F1 values for onset and offglide were obtained across all speakers for the ‘traditional’ raising environments vlessC and flappedT, the same was true for a number of other environments, some of which did not show raising at all. It must therefore be concluded that correlations between the raising of diphthong onsets and offglides are not a simple function of the mean overall degree of raising. Moreover, these correlations might be partially induced by the fact that ‘raw’, i.e. unnormalized formant data were used, which are strongly dependent on the size of the vocal tracts of the individual speakers. Canadian Raising in Victoria, B.C. 271 9 Slightly lower significances are found only between voicedC and N+vlessC, as well as between voicedC and R, which are significant at a level of .01 for the men, and not at all for the women. Figure 8: Median and quartiles for F1 offglides, by phoneme and male/ female speakers. Circles and asterisks indicate extreme values (located more than 1.5 box lengths from the median) and outliers (distance from median more than three box lengths), respectively. 3.2 F2 - Fronting 3.2.1 Phoneme / a / Fronting of / a / , about which little has been previously reported in the literature, is shown in Figs. 9 and 10. As for F1, medians and quartiles are displayed for each of the phonetic environments investigated, together with the two reference lines denoting median F2 values for the vowels in rather and ungodly, which delimit the average extent of the vowel space in the frontback dimension for the low vowels. Highest median F2 values occur for the environments of vlessC, flappedT, and voicedC, whose differences to the remaining environments are almost universally significant at the .001 level for both male and female speakers 9 . Ingrid Rosenfelder 272 10 Differences in mean F2 values between the phonetic environments of the two abovementioned groups are much more marked among the women, where they are almost universally significant at the .001 level, with the exception of a following voiceless consonant (vlessC), which differs at a .01 level only. For the men, differences pattern similarly but at lower p values. As, in this study, the latter category includes only the phonemes / z/ and / d/ , fronting of / a / seems to be favored most before obstruents. Fronting also occurs to a lesser degree before a following nasal-voiceless consonant cluster, and before / r/ , as well as word-finally and before nasals. Following / l/ and vowels show the lowest fronting scores. 3.2.2 Phoneme / a / Fronting of / a / , as shown in Figs. 11 and 12, does not seem to follow the pattern established for / a / , neither for the male nor for the female speakers. By and large, fronting of / a / takes place to a much greater extent, with the majority of environments showing median F2 values on or above the reference line provided by reference vowel [a: ]. This is especially true for the female speakers. Fronting environments for / a / can be divided into two large groups, encompassing vlessC, flappedT, Nasal, N+vlessC and # on the one hand, and voicedC, R and L on the other, with diphthongs followed by a vowel lying somewhere in between. Differences between the phonetic environments of these two groups are statistically significant for both male and female speakers 10 . Overall, fronting of / a / appears to be much more widespread than fronting of / a / , in terms of phonetic environment. 3.2.3 Offglides Box-and-whisker plots of median F2 measurements at the offglides of both / a / and / a / are displayed in Fig. 13 for both male and female speakers. These plots are of interest insofar as they can serve to shed further light on the hypothesis that the origin of the Canadian Raising mechanism lies in the hyperarticulation of syllable nuclei before voiceless consonants (Moreton 2004). Canadian Raising in Victoria, B.C. 273 Figure 9: Median and quartiles for F2 (onset) of phoneme / a / , male speakers. Circles and asterisks indicate extreme values (located more than 1.5 box lengths from the median) and outliers (distance from median more than three box lengths), respectively. Reference lines representing median F2 values of the vowels in rather (solid line) and ungodly (dashed line) are also shown. Ingrid Rosenfelder 274 Figure 10: Median and quartiles for F2 (onset) of phoneme / a / , female speakers. Circles and asterisks indicate extreme values (located more than 1.5 box lengths from the median) and outliers (distance from median more than three box lengths), respectively. Reference lines representing median F2 values of the vowels in rather (solid line) and ungodly (dashed line) are also shown. Canadian Raising in Victoria, B.C. 275 Figure 11: Median and quartiles for F2 (onset) of phoneme / a / , male speakers. Circles and asterisks indicate extreme values (located more than 1.5 box lengths from the median) and outliers (distance from median more than three box lengths), respectively. Reference lines representing median F2 values of the vowels in rather (solid line) and ungodly (dashed line) are also shown. Ingrid Rosenfelder 276 Figure 12: Median and quartiles for F2 (onset) of phoneme / a / , female speakers. Circles and asterisks indicate extreme values (located more than 1.5 box lengths from the median) and outliers (distance from median more than three box lengths), respectively. Reference lines representing median F2 values of the vowels in rather (solid line) and ungodly (dashed line) are also shown. Canadian Raising in Victoria, B.C. 277 Figure 13: Median and quartiles for F2 offglides, by phoneme and male/ female speakers. Circles and asterisks indicate extreme values (located more than 1.5 box lengths from the median) and outliers (distance from median more than three box lengths), respectively. As can be seen from the plots, / a / offglides are realized with a higher F2 for the environment of a following voiceless consonant by speakers of both sexes. It should be noted that at least for flappedT and voicedC, a clear difference to the F2 values of vlessC can be detected for / a / . Thus, in contrast to the results obtained in the previous sections, flappedT does not cluster with vlessC but rather with its voiced counterpart. These findings - that front-gliding diphthongs are produced with more fronted offglides before voiceless consonants - are in agreement with the theory that hyperarticulation plays a role in the production of the Canadian Raising diphthongs. Peripheralization of pre-voiceless coda syllable nuclei would entail not only raising of the diphthong offglides, but also an exaggeration of the front/ back dimension (Moreton 2004: 4). However, the same cannot be said about the diphthong / a / , whose offglides show higher F2s as well (i.e. less rather than more peripheralization) in the raising environments. Ingrid Rosenfelder 278 3.3 Sociolinguistic Aspects In order to investigate the sociolinguistic aspects of Canadian Raising, correlation coefficients were calculated to determine whether the use of the raised and fronted diphthongs varies with age or gender. Since it is not possible to directly compare men and women using ‘raw’ formant frequencies, correlation coefficients were calculated separately for these two groups. The results are shown in Table 2. Unfortunately, it is difficult to draw representative conclusions from this sample of subjects, as they do not constitute a truly random sample. Nevertheless, some indications of trends can be obtained from the data. 3.3.1 F1 - Raising Table 2 indicates a clear difference between the use of raising among male and female speakers. While onset values show no significant correlations with age for the female speakers, they do so for the men in a number of environments, including the ‘traditional’ raising environments vlessC and flappedT, but also others (see Table 2). All of these correlations have the same (negative) sign, indicating lower degrees of raising among the younger speakers. These results are somewhat confusing given the fact that in many of these environments raising does not take place to any significant degree (see sections 3.1.1, 3.1.2). While it might have been possible to regard these correlations as artefacts of the statistical input involved (due to the fact that there was only one older speaker in the male subject group), an alternative analysis excluding this speaker still retained the correlations at the same level of significance for the two ‘traditional’ raising environments, (vlessC and flappedT) although weakening the level of significance to p < .05 for some of the others. Thus, younger men appear to use less raising than their older counterparts in these two environments. F1 (onset) / a / / a / male female male female vlessC -.397** -.104 -.446** -.168 flappedT -.436** -.125 -.386** -.153 voicedC -.160 .017 -.425** -.106 Nasal -.429** -.110 -.199 -.143 N+vlessC .167 -.097 -.216 -.070 R -.405** .030 -.086 -.002 L -.279* -.125 -.196 .020 # -.425** .007 -.287* .061 V -.294* .099 .010 .082 Canadian Raising in Victoria, B.C. 279 F2 (onset) / a / / a / male female male female vlessC -.083 -.083 .257* -.124 flappedT .162 -.272* .292** -.264** voicedC -.007 -.404** .115 -.297* Nasal -.042 -.422** .170 -.022 N+vlessC .132 -.316* .197 -.077 R .028 -.086 .297* -.195 L .011 -.260* .041 -.153 # -.084 -.382** .133 -.003 V -.002 -.215 .153 -.045 Table 2: Pearson correlation coefficients for F1/ F2 (onset) [Hz] vs. age, by phoneme and male/ female speakers. Asterisks indicate p < 0.01 **; p < 0.05 *. 3.3.2 F2 - Fronting A completely different situation emerges from the data for fronting. Here, it is the female speakers who show significant correlations in their use of fronted onsets, while hardly any are observed for the male speakers (see Table 2). Moreover, fronting appears to be more strongly age-graded for the diphthong / a / . As the correlation coefficients displayed in Table 2 have opposite signs for men and women, younger female speakers have higher F2 values, i.e. more fronted diphthong onsets, in those environments that show significant agegrading, while the opposite holds true for the men. 4 Discussion 4.1 Raising Raising mainly occurs in the environments predicted by the ‘traditional’ rule for raising (Chambers 1973). This applies to both phonemes and speakers of both sexes, the data showing a clear differentiation in mean F1 values between the phonetic environments of a following voiceless consonant or a flapped / t/ , and all other environments. Interestingly, it is before the latter consonant that speakers show the most extreme values of F1. Thus, while the status of a following flapped / t/ was anything but clear with respect to raising in the original situation described by Joos in 1942, today it seems to be considered even ‘more voiceless’ than the other, truly voiceless consonants, at least with respect to the F1 or high/ low dimension only. A look at Ingrid Rosenfelder 280 11 It should be noted, however, that all of the above studies considering differences in pronunciation more systematically than in just a few selected lexical items have their main (Boberg 2000) or exclusive (Zeller 1993) focus on the eastern regions of Canada and the United States, where the variety of American English involved in the comparisons is currently undergoing a sound change known as the Northern Cities Chain Shift (Labov 1991; Labov, Ash & Boberg 2006). Moreover, clear differences exist with respect to the cot/ caught merger investigated by Zeller (1993: 187/ 188) between this Northern variety of American English and the General American area in the western United States, which should be considered as the adjacent American speech area for Victoria. Figs. 2 and 3, however, shows that this is only part of the whole truth. The two categories are clearly differentiated by the location of their respective offglides, which are peripheralized to a much greater degree before a following voiceless consonant, which holds true for both / a / and / a / . As discovered by Thomas (1991), these offglides play an important role in the perception of raising, raising being perceived for diphthongs with higher offglides. Therefore, the ‘leading’ phonetic environment for raising appears to be a following flapped / t/ only when solely considering the F1 or high/ low dimension, but a following voiceless consonant when taking into account the overall acoustic realization of the diphthongs. The differentiation with respect to mean F1 values (i.e. raising) between the ‘raising’ and ‘other’ environments is furthermore more pronounced among the female than among the male speakers, as well as to a greater degree for / a / than for / a / . The hypothesis that Canadian English is becoming more American therefore has to be rejected, at least with respect to this variable, as it is in fact / a / -raising which is the perceptually more salient feature of Canadian speech, rather than the raising of / a / . The continued use of / a / raising therefore indicates that Canadian speakers clearly contrast their pronunciation with that of the adjacent American dialect, possibly to emphasize their Canadian identity. At first sight, the above findings seem to contrast with a number of studies providing evidence for an Americanization of Canadian speech (see e.g. Zeller 1993; Nylvek 1992). However, these studies mainly dealt with the choice of lexical items or the pronunciation of selected individual words like either. Where systematic differences in the vowel system were investigated, clear differences between Canadian and American informants emerged (Zeller 1993: 187/ 188). With respect to the vowel system as a whole, marked border effects have also been shown to exist between the United States and Canada by Boberg (2000) 11 . The above findings are also compatible with the original study by Hung, Davison & Chambers (1993), which showed Non-Raising to be only a marginal phenomenon in Victoria. Age-grading of Non-Raising could only be seen in the data by ignoring the Non-Raising Index for the adult male speakers: “Victoria would then show a certain degree of age stratification, but the Canadian Raising in Victoria, B.C. 281 Index scores (all below 10) are too low to suggest that any change is occurring” (Hung, Davison & Chambers 1993: 255). Evidence for the age-stratification of Non-Raising in this study was found for Vancouver only. Even there, the possible decline in the use of raising has not taken place, as shown by the data from Pohler (1994), which suggests that raising is still widespread among young Vancouverites in 1993. Thus, American influence on Canadian English notwithstanding, the Canadian Raising diphthongs appear to be quite resistant to this influence. Regarding the acoustic realization of the Canadian Raising diphthongs with respect to their conditioning phonetic environments further confirms the findings that Canadian English, as far as the Canadian Raising diphthongs are concerned, is distinct from the adjacent American variety. Comparing the data from the present study with the results obtained by Dailey-O’Cain (1997), it becomes clear that Canadian Raising patterns quite differently in Canada than it does in the northern United States. While the studies agree to the extent that they show raising to occur predominantly before voiceless consonants, including flapped / t/ s, they differ in their results for the other phonetic environments investigated. High percentages of raising (approximately 50%) were found in Ann Arbor for / a / in the environments of a following nasal-voiceless consonant cluster and / r/ (Dailey-O’Cain 1997: 111), whereas the latter phonetic environment is in fact the one exhibiting the highest F1 values (i.e. the least raising) in Victoria. On the other hand, raising of / a / in Ann Arbor was found to a small degree only before voiced consonants, while low F1 values found for speakers from Victoria indicate that the rule for raising may have become extended to this environment, with some degree of raising taking place when / a / is followed by a voiced consonant. Raising of / a / , which has been interpreted as progressing in the northern United States, with young females leading the change (Dailey-O’Cain 1997: 117), seems to have extended to include the phonetic environments of a following nasal and nasal-consonant cluster in Victoria. Moreover, there is tentative evidence from the present study to indicate that a change in the opposite direction (less raising) might be taking place for the male speakers in Victoria. 4.2 Fronting Fronting of the diphthongs is widespread among both male and female speakers in Victoria. The phenomenon is more pronounced for / a / than for / a / , and more so for women than for men. The two diphthongs pattern differently with respect to the phonetic environments favoring fronting: for / a / , the highest mean F2 values are observed for following obstruents (including voiced consonants), and, to a lesser degree, for diphthongs preceding nasal-voiceless consonant clusters, / r/ , as well as word-finally. For / a / , Ingrid Rosenfelder 282 fronting is favored before voiceless consonants and flapped / t/ s, nasals and nasal-voiceless consonant clusters, as well as word-finally. In contrast to / a / , onsets show lower F2 values, i.e. less fronting, before voiced consonants. These findings agree with the results obtained in the 1980s by Hung, Davison & Chambers (1993: 254/ 255), who found the youngest generation of speakers from Victoria to front nearly equally in all environments, in contrast to the findings obtained for the other cities investigated, where the “elsewhere” (non-raising) environment was favored for the fronting of / a / . Significant correlations with age found for the diphthong / a / in a number of phonetic environments suggest that a sound change in progress may be going on among female speakers in Victoria, with younger women using more fronted onsets than older women. This concerns primarily the ‘other’, i.e. non-raising environments, although a weak correlation was detected for a following flapped / t/ . At first, the fact that no correlation with age was found for the diphthong when preceding voiceless consonants seems to contradict this hypothesis. However, a look at the data (Fig. 10) shows that this is the environment with the highest degree of fronting already, with an overall mean F2 value lying well above the reference line provided by the vowel in rather. Moreover, it can be gathered from the same figure that the amount of individual variation is also highest for onsets in this environment, encompassing a range of nearly 1000 Hz, which is in fact the complete range of variation for all environments. This suggests that fronting of / a / has already become the norm for this environment, but that this norm may not have established itself firmly yet. Thus, / a / appears to be currently undergoing a sound change in the direction of further fronting in the other environments which has already been completed before voiceless consonants. No such change in the same direction is visible for the diphthong / a / , which, however, consistently features higher F2 values for all environments for both men and women (compare Figs. 9 and 11, 10 and 12). Therefore, the sound change in progress observed in the 1980s by Hung, Davison & Chambers (1993: 251-254) seems to have been completed for / a / , with fronting of the onset of this diphthong having become the norm. The same sound change now appears to be extending to / a / as well. 5 Conclusion Canadian Raising of both diphthongs is still widespread, and still chiefly occurs in the expected phonetic environment, i.e. before a following voiceless consonant. This latter category includes the flapped / t/ . Moreover, the phonetic realization of the raised diphthongs not only includes raising of the onsets, but that of the offglides as well, which supports the theory that Canadian Raising in Victoria, B.C. 283 peripheralization of the offglides may have played a role in the historical origin of the phenomenon. Raising of both / a / and / a / remains firmly established, indicating that the predicted Americanization of Canadian speech has not taken place, as it is in fact the / a / -diphthong whose raising is perceptually most prominent for non-Canadians. If raising is moreover spreading in the northern United States, as reported by Dailey-O’Cain (1997), the two varieties of English are converging, but with respect to the phenomenon of raising, it would then be Canada that is influencing the United States linguistically, and not vice versa. Fronting of the diphthongs is also widespread in Victoria. For / a / , this pronunciation appears to have become the norm, with fronted onsets occurring in virtually all phonetic environments and quite uniformly across individual speakers. Thus, the sound change in progress observed in the 1980s involving this diphthong seems to have been completed, with a corresponding change now taking place for / a / , which appears to be on the increase among younger women in Victoria. In conclusion, Canadian Raising today is still in existence for both diphthongs, but a trend toward fronting of the onsets can be discerned, with / a / being predominantly realized as fronted already, and / a / increasingly becoming so. Acknowledgements This article presents the results of my Wissenschaftliche Arbeit (Lehramt an Gymnasien), written between October 2004 and February 2005 under the supervision of Prof. Beat Glauser at the University of Heidelberg, Germany. I would like to thank the following people for their support: Prof. Glauser for his supervision, especially for establishing connections to Canada for the collection of data; Prof. John H. Esling and Prof. Leslie Saxon at the University of Victoria, for supporting my research and allowing me to use the facilities of the Phonetics Laboratory there; Ricardo Serrano, for technical assistance with the recordings; Prof. Paul Boersma at the University of Amsterdam and Dr. Günther Sawitzki from the Department of Applied Mathematics at the University of Heidelberg, for answering my questions about Praat and statistics, respectively, and finally, my speakers, whose time and effort in participating in this study were greatly appreciated. References Boberg, Charles (2000). “Geolinguistic diffusion and the U.S.-Canadian border”. Language Variation and Change 12. 1-24. Boersma, Paul & Weenink, David (2004). Praat. A system for doing phonetics by computer (Version 4.2.06) [Computer software]. http: / / www.praat.org. [Accessed January 2005]. Ingrid Rosenfelder 284 Chambers, J.K. (1973). “Canadian Raising”. Canadian Journal of Linguistics 18. 113-135. (repr. in Chambers 1975. 83-100.) - and Hardwick, Margaret F. (1986). “Comparative Sociolinguistics of a Sound Change in Canadian English”. English World-Wide 7. 23-46. Dailey-O’Cain, Jennifer (1997). “Canadian raising in a Midwestern U.S. city”. Language Variation and Change 9. 107-120. Emeneau, M.B. (1935). “The dialect of Lunenburg, Nova Scotia”. Language 11. 140-147. Gregg, Robert J. (1973). “The diphthongs i and ai in Scottish, Scotch-Irish and Canadian English”. Canadian Journal of Linguistics 18. 136-145. Hertz, Susan R. (1991). “Streams, phones and transitions: toward a new phonological and phonetic model of formant timing”. Journal of Phonetics 19. 91-109. Hung, Henrietta; Davison, John and Chambers, J.K. (1993). “Comparative Sociolinguistics of (aw)-Fronting”. In: Sandra Clarke (ed.). Focus on Canada, Varieties of English Around the World 11. Amsterdam: John Benjamins Publishing Company. 247-267. Joos, Martin (1942). “A Phonological Dilemma in Canadian English”. Language 18. 141-144. (repr. in Chambers 1975. 79-82.) Labov, William (1991). “The three dialects of English”. In: Penelope Eckert (ed.). New Ways of Analyzing Sound Change. New York: Academic. 1-44. - Ash, Sharon & Boberg, Charles (2006). The Atlas of North American English. Phonetics, Phonology and Sound Change. New York: Mouton de Gruyter. Moreton, Elliott (2004). “Realization of the English postvocalic [voice] contrast in F1 and F2”. Journal of Phonetics 32. 1-33. - and Thomas, Erik (2004). “Origins of Canadian Raising in Voiceless-Coda Effects: A Case Study in Phonologization”. Abstract. 9th Conference on Laboratory Phonology. http: / / www.linguistics.uiuc.edu/ labphon9/ Abstract_PDF/ moreton.pdf. [Accessed October 21st, 2004]. Nylvek, Judith A. (1992). “Is Canadian English in Saskatchewan becoming more American? ” American Speech 67. 268-278. Paradis, Carole (1980). “La Règle de Canadian Raising et l’Analyse en Structure Syllabique”. Canadian Journal of Linguistics 25. 35-45. Pohler, Anja N. (1994). Canadian Raising in Vancouver. Unpublished M.A. thesis. University of Heidelberg, Germany. Thomas, Erik R. (1991). “The Origin of Canadian Raising in Ontario”. Canadian Journal of Linguistics 36. 147170. Trudgill, Peter (1984). “New-Dialect formation and the Analysis of Colonial Dialects: The Case of Canadian Raising”. In: H.J. Warkentyne (ed.). Papers from the Fifth International Conference on Methods in Dialectology. Victoria: University of Victoria. 35-46. Vance, Timothy J. (1987). “‘Canadian Raising’ in some Dialects of the Northern United States”. American Speech 62. 195-210. Zeller, Christine (1993). “Linguistic Symmetries, Asymmetries, and Border Effects within a Canadian/ American Sample”. In: Sandra Clarke (ed.). Focus on Canada, Varieties of English Around the World 11. Amsterdam: John Benjamins Publishing Company. 179-199. Ingrid Rosenfelder University of Freiburg Freiburg, Germany