Professional Documents
Culture Documents
Foreign (A) in North American English: Variation and Change in Loan Phonology
Foreign (A) in North American English: Variation and Change in Loan Phonology
research-article2020
ENGXXX10.1177/0075424219896397Journal of English LinguisticsBoberg
Article
Journal of English Linguistics
2020, Vol. 48(1) 31–71
Foreign (a) in North © The Author(s) 2020
Article reuse guidelines:
American English: sagepub.com/journals-permissions
DOI: 10.1177/0075424219896397
https://doi.org/10.1177/0075424219896397
Variation and Change journals.sagepub.com/home/eng
in Loan Phonology
Charles Boberg1
Abstract
Previous research has shown that Canadian English displays a unique pattern of
nativizing the stressed vowel of foreign words spelled with the letter <a>, like lava,
pasta, and spa, known as foreign (a), with more use of /æ/ (the trap vowel) and
less use of /ah/ (the palm vowel) than American English. This paper analyzes one
hundred examples of foreign (a), produced by sixty-one Canadian and thirty-one
American English-speakers, in order to shed more light on this pattern and its current
development. Acoustic analysis is used to determine whether each participant
assigns each vowel to English /æ/, /ah/, or an intermediate category between /æ/ and
/ah/. It reports that the Canadian pattern, though still distinct, is converging with
the American pattern, in that Canadians now use slightly more /ah/ than /æ/; that
men appear to lead this change but this is because they participate less than women
do in the Short Front (Canadian) Vowel Shift; that intermediate vowel assignments
are comparatively rare, suggesting that a new low-central vowel phoneme is not
emerging; that the Canadian tendency toward American pronunciation is not well
aligned with overt attitudes toward the United States and American English; and
that the national differences in foreign (a) assignment result not from structural,
phonological differences between the dialects so much as from a complex set of
sociocultural factors.
Keywords
acoustic phonetics, American English, Canadian English, Canadian Vowel Shift, foreign
(a), lexical borrowing, lexical diffusion, loan phonology, loanwords
Corresponding Author:
Charles Boberg, Department of Linguistics, McGill University, 1085 Dr. Penfield, Montreal, QC H3A
1A7, Canada.
Email: charles.boberg@mcgill.ca
32 Journal of English Linguistics 48(1)
1. Introduction
This article reports on a study of foreign (a) nativization in North American English.
Foreign (a) nativization (Boberg 1997, 1999, 2009) is the process by which the ortho-
graphic <a> of foreign languages (in either original or romanized orthography) is
adapted to the sound system of English by assigning it to one of the three English
phonemes that are also spelled with <a>: the /æ/ of trap; the /ah/ of palm (which for
most North Americans is the same as the /o/ of lot, resulting in a merged phoneme that
will be labeled /ah-o/); or the /ey/ of face.1 For example, for most North Americans,
the vowels of khaki and samurai are normally assigned to /æ/; those of falafel and
koala to /ah-o/; and those of potato and tornado to /ey/. Some words, like Pakistani or
panorama, present two instances of foreign (a), one under primary and one under sec-
ondary stress (those under tertiary stress, like the last vowel of panorama, are reduced
to a schwa and are therefore of no concern); this article will analyze foreign (a) in
primary-stress position. Though some foreign (a) words display widespread unifor-
mity in assignment patterns, others exhibit regional and social variation, including
fairly regular differences among national dialects of English, as well as change over
time. This article will focus on foreign (a) nativization in Canadian English—how it
differs from the American pattern and how it is changing over time.
These questions were examined by recording a sample of sixty-one Canadian and
thirty-one American undergraduate students reading one hundred sentences containing
foreign (a) words and by using acoustic phonetic analysis to measure the vowel quali-
ties they produced. The students were also asked to judge the similarity of foreign (a)
words to native words and to answer a questionnaire about their attitudes toward
American English and the United States. Because the foreign (a) vocabulary and its
relation to the sound system of English is fairly complex, and because this variable is
not widely known from previous research, the paper will begin with some background
information on foreign (a) nativization, including a review of the necessary phonologi-
cal concepts and of previous research dealing with the topic, before stating its research
questions in more detail. The following sections will present the study’s methods,
including how foreign (a) assignment was measured; the results of the acoustic analy-
sis, showing how foreign (a) assignment is correlated with national group and sex, and
how it is changing over time; the results of the perceptual study, showing how partici-
pants judge their own pronunciation; and finally a discussion of the results that seeks
to understand them in a larger sociolinguistic context, examining the role of both
structural-phonological factors and cultural or sociolinguistic factors in producing the
observed patterns of phoneme assignment.
other parts of speech also occur, like the adjectives ersatz, halal, khaki, macho, and
suave. In addition to common nouns like avocado, hijab, lava, latte, pajamas, pasta,
plaza, saga, and spa, the foreign (a) class includes thousands of proper nouns, com-
prising both geographic names (Gaza, Karachi, Zimbabwe) and personal names
(Gandhi, Kafka, Picasso). In the semantic domain of food and drink alone, Boberg
(2009:361) lists ninety examples, including items of originally German, Indian, Italian,
Japanese, Mexican, and Middle Eastern cuisine, from salami to masala to nachos to
kebabs, that are now commonly eaten all over the English-speaking world. Moreover,
the class is growing, with new additions spurred by ever-increasing contact among
cultures and their languages in today’s globalized world. Foreign (a) words are there-
fore among the most important cases of systematic phono-lexical variation—or varia-
tion in phonemic incidence across large sets of phonologically related words—in
present-day English.
merge with short /o/ as /ah-o/ (palm-lot), so that father and bother, for example,
rhyme (Labov, Ash & Boberg 2006:169-171, 230-231; Boberg 2010:127-130, 144-
146). The choice between /æ/ and /ah-o/ in American English is therefore more front
versus back than short versus long. American nativization relies instead on the per-
ceived foreign status of the words as grounds for assigning them predominantly to the
/ah-o/ class. While /æ/ occurs in many originally foreign but now fully Americanized
words like hamburger, Miami, and ranch, it seems to many Americans inappropri-
ate—even risible—when applied to more recent and markedly foreign words, as
found by Boberg (1999) in a study of American attitudes toward foreign (a) nativiza-
tion. There is even debate about some well-established American words: while
Montana is always /æ/ (though some British speakers use /ah/, rhyming it with
banana), Colorado and Nevada famously vary between /ah-o/ and /æ/. Beyond these
state names and another well-known case of variation, the older Hindi-Urdu loan
pajamas, the data presented below show that more recent and less Americanized
loans like angst, Cézanne, chianti, Iraq, soprano, and Yamaha can also vary.
Nevertheless, a larger set of relatively recent foreign (a) words is remarkably uniform
and displays an overwhelming preference for /ah/, a pattern noted in Shapiro (1997)
as well as in several general accounts of British-American differences (e.g., Wells
1982:122, 142-144; Gramley & Pätzold 1992:342; Trudgill & Hannah 2008:43).
Returning to the seven examples discussed above, lava, saga, spa, latte, macho, man-
tra, and pasta all usually have /ah-o/ in American English.
Lindsey (1990) was the first to note that the difference between short-long alterna-
tions in British English and consistent American use of what would be long vowels in
British English extends to other vowels as well. For instance, pita, Pedro, mocha, and
Buddhist are assigned respectively to the short vowels /i/, /e/, /o/, and /u/ (kit, dress,
lot, and foot) in British English but to long /iy/, /ey/, /ow/, and /uw/ (fleece, face,
goat, and goose) in American English, suggesting that dialect variation in vowel
assignment involves general phonological principles rather than the idiosyncratic
properties of a particular lexical class. Nevertheless, while there is some generality to
the /o/-/ow/ comparison (like mocha, for instance, are baroque, cognac, Costa Rica,
Kosovo, Molière, Prokofiev, risotto, Soviet, and yoghurt), the nativization patterns of
the other vowels seem to display less regularity than that of foreign (a), with fewer
data and plentiful exceptions. Foreign (a) variation will therefore be treated here, at
least provisionally, as a special property of the lexical class of foreign (a) words.
Indeed, it is difficult to identify purely phonological constraints on it, beyond the role
of vowel length in the British pattern.
Canadian English displays yet another distinct pattern of foreign (a) nativization,
which is the main focus of this paper. Observations of Canadian-American nativiza-
tion differences in small sets of individual foreign (a) words go back to Avis (1973:65),4
but the first systematic examination of the Canadian pattern in a larger set of words
was carried out by Boberg (2000), in a study of speech differences and geolinguistic
diffusion across the international boundary between Canada and the United States.
The traditional Canadian pattern takes the British use of /æ/ for short vowels as its
Boberg 35
default, but extends this treatment even to long vowels that have /ah/ in British English.
This is apparently by analogy with words of the bath class (staff, path, pass, cast, etc.),
which Canadians, exposed to a good deal of British English across their history,
learned to regard as a symbol of their cultural and linguistic independence from
Britain: for a native-born Canadian, pronouncing bath with /ah-o/ would seem pomp-
ous and affected. If, therefore, Canadians have /æ/ where the British have /ah/ in
bath—goes the analogy—they should also have /æ/ in drama and lava. Accordingly,
lava, saga, latte, macho, mantra, and pasta all traditionally have /æ/ in Canadian
English, a striking contrast with both the American and British patterns. Use of /ah-o/
in the traditional Canadian pattern is reserved for words like spa, since even in
Canadian English the constraint on short vowels in primary-stress final position
applies, discouraging the occurrence of /æ/ in bra, coup d’état, faux pas, Shah, etc.
The national dialect differences described above are summarized in Table 1.
Table 1. National Dialect Differences in Assignment of Foreign (a) Words to /æ/ (trap) and
/ah/ (palm), Exemplified with Three Words Featuring Different Phonological Contexts
Since Canadian English displays not only the /ah-o/ merger of most American dia-
lects but also the low-back merger of /o/ and /oh/ or lot-thought (Labov, Ash & Boberg
2006:58-65, 217-218; Boberg 2010:126-130, 144-146), Canadian /ah/ is involved in a
double-merger in low-back position (/ah-o-oh/ or palm-lot-thought). It is therefore too
far back to be phonetically appropriate as a nativization of foreign (a), except where the
context demands or encourages a long or back vowel, as in spa (final position) or Vivaldi
(before coda /l/). Nevertheless, American cultural influence suggests to Canadians that
/æ/ is also inappropriate in at least some foreign words, forcing them to abandon phone-
mic assignment in preference for extra-phonemic “imitation” of the foreign sound (Van
Coetsem 1988), or of the usual American rendering of that foreign sound. (Most foreign
(a) assignments by individual speakers are at several removes from the source language
and presumably reflect pronunciations learned from others in their communities, or from
schools or the media, rather than direct experience with the language in question.) This
strategy adapts the American preference for non-front nativizations to the phonetic prop-
erties of the Canadian vowel system, with its low-back /ah-o-oh/.
Boberg (2009) found that these intermediate productions comprised an average of
22 percent of the Canadian nativizations of twenty foreign (a) words, compared to
only 11 percent for the American comparison group. The frequency of intermediate
productions varied by word, but they rose above 40 percent in lava, façade, and plaza
(compared to 9 percent for Americans). That is, over 40 percent of Canadians pro-
duced a vowel in façade, for example, that could not be confidently identified as either
the /æ/ of native sad or the /ah-o/ of sod. This pattern, it was argued, suggested the
possible emergence of a new phoneme in Canadian English, a third low vowel between
/æ/ and /ah-o/ reserved for use in foreign (a) words.
The Canadian trend toward non-front nativizations is complicated by a second,
apparently unrelated change: the Canadian Vowel Shift, a lowering and retraction of
the short front vowels that appears to result from the low-back merger (Clarke, Elms
& Youssef 1995; Labov, Ash & Boberg 2006:130, 220-221). Since a similar set of
vowel shifts has been identified in western varieties of American English, the Canadian
Shift now bears several other names, including the Short Front Vowel Shift (Boberg
2019) and the Low-Back Merger Shift (Becker 2019). In keeping with the term
employed most recently in this journal by Boberg (2019), it will be referred to here as
the “Short Front Vowel Shift” (SFVS). One of the main stages or components of the
SFVS, first observed in Vancouver English by Esling and Warkentyne (1993), is a
retraction of native /æ/ from [æ] to [a], a quality very similar, if not identical, to the
inter-phonemic nativization of foreign (a) just discussed. Boberg (2009:376) sug-
gested two possible future resolutions of the competition between native /æ/ retraction
and inter-phonemic foreign (a) nativization for the low-central region of the Canadian
English vowel space. One was a merger of /æ/ and foreign [a], through the absorption
of the originally extra-phonemic tokens of foreign (a) into an increasingly retracted
native /æ/ phoneme. The other was a split, with a newly phonemicized foreign /a/
creating a barrier to the further retraction of /æ/, possibly even initiating a reversal of
this part of the SFVS as a way of increasing the “margin of security” between /æ/ and
(foreign) /a/ and thereby avoiding homophony. Which resolution would prevail in the
future was an open question, which will be addressed below.
Boberg 37
2. Method
2.1. The Sample
The data reported here were collected in 2012 from undergraduate students at McGill
University in Montreal, Quebec, Canada. McGill’s student population is drawn from
across North America, including a substantial number of American students, mostly
from the northeastern United States, as well as students from every region of Canada.
Student participants were recruited through the university’s online classified ad ser-
vice and offered twenty dollars as compensation for providing demographic data,
reading a list of two hundred words, judging the similarity of thirty-six word-pairs,
conversing with their interviewers about topics of general interest for 15-20 minutes,
and completing two two-page questionnaires, one dealing with lexical, phono-lexical,
and spelling variation, and the other with attitudes toward varieties of English and
various social and lifestyle matters. The ad specified that participants should be native
speakers of English with no formal education in linguistics. Interviews were con-
ducted by fellow students (see acknowledgments) in a variety of quiet locations on and
off campus, and recorded in Praat (Boersma & Weenink 2012), using plug-in micro-
phones and laptops.
38 Journal of English Linguistics 48(1)
Canadian
The project’s original design called for a sample of seventy-five participants: fifty
Canadians, divided by region, with twenty-five from western and twenty-five from
central and eastern Canada (reflecting a west-east division within mainland Canadian
English found to be important by Boberg 2008, 2010:199-209); and twenty-five
Americans, as a comparison group. By the time data collection was suspended, one
hundred fifteen interviews were completed, of which ninety-nine were with native
speakers of North American English, the subset reported on here. Owing to problems
of sound quality in a few interviews, it was possible to carry out acoustic analysis of
only ninety-two of these: sixty-one from Canada and thirty-one from the United States.
Of the Canadians, forty-one are eastern (mostly Ontario and Greater Montreal, with
only three from Atlantic Canada including one Newfoundlander) and eighteen western
(mostly British Columbia and Alberta, with only one from Manitoba and none from
Saskatchewan); two could not be categorized in this way and were omitted from the
regional analysis but included in the analysis of national differences. The Americans
represent a diverse set of regional origins from across the country, which will be dis-
cussed below. Of the Canadians, forty-five are women and sixteen men; a similar
female preponderance obtains among the Americans, with twenty-four women to
seven men. The sample is summarized in Table 2.
French (éclat, façade, panache), Germanic languages (angst, bratwurst, saga), Indic
and Persian (khaki, mantra, nirvana), Italian (finale, lava, pasta), Japanese (karate,
samurai, tsunami), Spanish and Portuguese (avocado, enchilada, llama), and other
languages (drama, koala, souvlaki). In terms of date of first attestation, which was
taken from the eleventh edition of Merriam-Webster’s Collegiate Dictionary (2003),
the words range from Renaissance loans (bravado, coup d’état, drama) to twentieth
century loans (macho, salsa, shiatsu); the mean date is 1828 and the median 1856.
In terms of word type, they include common nouns, like all of those just cited, as
well as place names (Iraq, Karachi, Uganda), names of famous people (Gandhi,
Kafka, Picasso), ordinary first names (Hans, Natasha, Vladimir), and brand names
(Armani, Mazda, Saab). By their very nature, foreign (a) words vary widely in real-
world frequency and familiarity, from daily vocabulary, like drama, pasta, spa, or
taco, to what for some speakers would be affected, esoteric, obscure, or rarely used
words, like éclat, élan, panache, or some foreign names and places. It is important
to acknowledge that this may influence how the words are pronounced, but that mat-
ter is not examined here; assuming the frequency and familiarity of individual for-
eign (a) words, which varies from one speaker to another, could be accurately
measured, its effect on nativization patterns would be an interesting subject for future
research.
The word list was also designed to include an equal number of non–foreign (a)
words, carefully selected to represent the native vowel system of each speaker; these
are shown in Appendix B, organized by phonetic environment. Given that participants
were already reading one hundred foreign (a) words, selection of other words was
constrained by the need to avoid unwanted effects of fatigue or boredom. Word choice
therefore focused on obtaining at least three tokens of each vowel phoneme in pri-
mary-stress position, along with extra tokens representing the most important vocalic
variables of Canadian English: lowering and retraction in the SFVS; Canadian Raising
of /aw/ and /ay/ before voiceless obstruents; fronting of back vowels; pre-nasal and
pre-voiced-velar allophones of /æ/; and, most importantly for present purposes, the
phonemic status and phonetic quality of the native low central and back vowels, /ah/
(palm), /o/ (lot), and /oh/ (thought).
In order to create a more natural context for the word list items, each was presented
as the final word of a short sentence, thereby turning the participant’s attention away
from pronunciation toward meaning and avoiding contamination effects between adja-
cent words. Moreover, in order to prevent participants from dwelling on comparisons
among foreign (a) items in particular, these were alternated with non-foreign (a) items,
so that the focus on foreign (a) was less obvious. Subject to this constraint, the whole
list was then alphabetized. For example, the first four lines of the word list, or rather
sentence list, are represented in (1)-(4).
experimental data set and would have skewed the representation of the vowel system
toward the bottom of the vowel space. Using only the non-foreign (a) data, the grand
mean of F1 and F2 for the ninety-two speakers was 1131 Hz, of which the natural log
is 7.031. (It should be kept in mind that this grand mean reflects a sample that is 75
percent female, which makes it somewhat higher than it would be with an equal repre-
sentation of men. For instance, the grand mean for the similar sample of Boberg
[2008:134], which was only 59 percent female, was 1119 Hz.) The scaling factor for
each speaker was obtained by calculating the exponent of the difference between the
natural log of the F1/F2 mean for the group and the natural log of the F1/F2 mean for
the speaker. The resulting scaling factors ranged from 0.91 for the woman with the
highest voice to 1.22 for the man with the lowest. These scaling factors were then
applied to the entire data set for each speaker, including the foreign (a) words, to
obtain the normalized data set for quantitative and statistical analysis.
Means of F1 and F2 for each native vowel were calculated for each speaker, over
the set of words containing that vowel, and the means of these means were then
calculated for the groups to be compared: Canadians and Americans; western and
eastern Canadians; and men and women. T-tests were carried out on each speaker’s
data, to determine whether significant differences existed between phonemes poten-
tially involved in merger, that is, the low-central and low-back vowels /ah, o, oh/
(palm, lot, and thought).6 The mean formant values of /æ/ were calculated using
only the main set of five /æ/ tokens (without the pre-nasal and pre-/g/ subsets, which
are allophones with distinct distributions). Since t-tests found that over 90 percent of
speakers from both countries did not display a statistically significant difference
between /ah/ and /o/, these sets were combined as /ah-o/, with the mean based on the
combined set of twelve tokens (five of /ah/ and seven of /o/; see section 3.1 for lists
of words in each set).
(a) token fell into the top or highest third for the speaker in question, it was classified
as /æ/; if it fell into the bottom or lowest third it was classified as /ah-o/; and if it fell
into the middle third it was classified as intermediate and potentially extra-phonemic.
While not without merit, this method could be criticized for the admittedly arbitrary
choice of the middle third of the F2 range as the definition of phonetically intermedi-
ate: taking only the middle fifth of the F2 range as intermediate territory, for instance,
would dramatically reduce the number of foreign (a) tokens classified as intermedi-
ate. A less arbitrary and more objective measure, which is adopted here, is standard
deviation: tokens falling within two standard deviations (about 95 percent of the
range) of the mean F2 value of /æ/ are categorized as tokens of /æ/; those falling
within two standard deviations of the mean F2 value of /ah-o/ are categorized as
tokens of /ah-o/; and those falling between these ranges, that is, higher than the top of
the /ah-o/ range but lower than the bottom of the /æ/ range, are categorized as
intermediate.
An important difference between these methods is that the first or “middle-third”
method of Boberg (2009) guarantees that all speakers have an intermediate range
between /æ/ and /ah-o/, regardless of how far apart the means of those vowels are or
how much variability in native vowel production exists around those means. The sec-
ond or “two-standard-deviations” method used here, on the other hand, potentially
eliminates the intermediate range for speakers whose native /æ/ and /ah-o/ are close
together and/or who display a wide range of variation in native vowel production. For
such speakers, classification of foreign (a) tokens as intermediate is impossible, hypo-
thetically reducing the overall number of tokens classified as intermediate by this
method. A comparison of the results of the two methods presented in section 3.6 shows
this hypothesis to be correct, though only for the Canadian sample. Despite this impor-
tant difference, the two methods produce similar results at a general level: both find
that /ah-o/ is the most common nativization outcome, followed by /æ/, followed by
intermediate. This lends credence to the analysis of Boberg (2009), though the smaller
number of intermediate nativizations identified by the two-standard-deviations method
casts some doubt on that study’s suggestion that intermediate nativizations may be
causing the emergence of a new low-central phoneme. Whereas the middle-third
method will be retained here for purposes of real-time comparison with the data of
Boberg (2009), the main analysis of national and sex differences will utilize the more
objective two-standard-deviations method.
3. Results
3.1. Phonemic Structure of the Low Vowel Space
Before examining variation in foreign (a) nativization, it is important to establish the
phonological framework in which it occurs: the phonemic structure of the native low
vowel space for each national group of participants. Figure 1 shows the mean positions
of the relevant vowels—/æ/, /ah-o/, and the foreign (a) class—together with a selection
Boberg 43
of other vowels that indicate their location relative to the larger vowel space (all based
on the word list data). Single means are shown for /uw/ and /uwl/ (goose and goose
before /l/), since there was no significant national difference for these vowels in either
height or advancement; all other vowels displayed a national difference significant in at
least one dimension (by a t-test, p < .05). In these cases, Canadian means are shown
as black circles and American means as white circles. It is remarkable that the front
vowels involved in the SFVS, /e/ and /æ/ (dress and trap), appear more shifted for
the Americans than for the Canadians, with higher F1 but lower F2, a matter explored
in Boberg (2019). The low vowels, in fact, are all significantly lower for the Americans
than for the Canadians, a surprising result that does not at present have an evident
explanation. That the American vowels are not also fronter indicates that this is not
simply a normalization problem wherein all of the American formant values are con-
sistently higher. In any case, the situation of both groups’ foreign (a) class with a
mean position directly between /æ/ and /ah-o/ is immediately apparent; moreover, the
Canadian foreign (a) mean is slightly closer to /æ/ and the American slightly closer to
/ah-o/, as would be expected given the national patterns of foreign (a) assignment
discussed in 1.2 (a difference significant at p < .001, as shown in Table 3).
Figure 1. Mean Canadian and American Formant Values (F1 and F2 in Hz) for Selected
Vowel Phonemes
Table 3. Mean Formant Data (in Hz) on Selected Vowel Sets for Canadian and American
Participants, with T-tests of National Differences
Table 3 provides the mean formant data on which Figure 1 is based, along with
the results of t-tests of the difference between the Canadian and American groups.
The mean of /æ/ is based on the words bad, pass, sad, sat, and staff; that of /ah/ on
blah, father, massage, calm, and palm; and that of /o/ on bother, cot, sod, collar,
bomb, don, and Tom. Separate means are given for /ah/ and /o/, whose F1 and F2
values overlap very closely for both groups. T-tests of these word sets found they are
in fact not significantly different for 97 percent of the Canadian participants (59/61)
and 94 percent of the Americans (29/31), so the number of participants in either
national group who have a phonemic contrast between /ah/ and /o/ is negligible. This
conforms closely to the results of previous research cited above and justifies the
representation of these vowels as a single phoneme (an average of /ah/ and /o/) in
Figure 1.
T-tests were also run on the low-back /o/-/oh/ (lot-thought) distinction, the lat-
ter vowel represented by the words boss, caught, flaw, saw, sawed, caller, and
dawn, with a similar result for the Canadian group: 95 percent (58/61) showed no
distinction between /o/ and /oh/ in height (F1) and 98 percent (60/61) showed no
distinction in advancement (F2), again matching the results of the previous research
on the low-back merger cited above. As might be expected, the Americans were
less uniform in this respect, with 26 percent (8/31) maintaining a low-back contrast
in height and a subset of half of these (13 percent or 4/31) also distinct in advance-
ment. The regional origins of these non-merged participants generally align well
with data on the geographic distribution of the low-back distinction in the Atlas of
North American English (Labov, Ash & Boberg 2006:60-61): three are from the
greater New York City region, two from Philadelphia, and one from Detroit, though
Boberg 45
two are from areas of New England where a merger would normally be expected.
Whereas most of the Americans (74 percent) are therefore phonologically similar to
the Canadians, a minority differ in this important respect. The possible effects of this
difference will be considered below in 4.5, but the diversity in the phonemic status of
/oh/ (thought) is the reason for its exclusion from Table 3 and from the vowel chart
in Figure 1.
The most important aspect of Table 3 for present purposes is the data on the foreign
(a) set, which involves mean F1 and F2 values across the hundred words. Although
these means are not particularly informative in the sense that they group together
words assigned to different phonemic categories in different proportions by each par-
ticipant, they nevertheless demonstrate the general national difference in phonemic
assignments discussed in 1.2. As was already evident in Figure 1, the Canadian F2
mean is 83 Hz higher than the American, a difference shown here to be significant at
p < .001, which reflects the Canadian tendency to make more assignments to /æ/ and
the American tendency to make more assignments to /ah-o/.
for regatta for Canadians and 45 percent for Yamaha for Americans. Beyond these
broad similarities, however, stark national differences emerge. Whereas Canadian
assignment to /æ/ remains strong at 72 percent for panorama, the last /æ/ word in
the Canadian top ten, it plunges more quickly to 50 percent for shiatsu in the
American list. Conversely, the top ten /ah-o/ words for Canadians fall off more
quickly—to 82 percent for llama—than for Americans, who have 90 percent for
kebab. Surprisingly, national differences in the frequency of intermediate assign-
ment show the opposite pattern from what would be expected based on the analy-
sis of Boberg (2009): Americans generally produce more intermediate assignments
than Canadians, with the lowest frequency in the American top ten (31 percent in
gala) higher than the highest frequency in the Canadian top ten (28 percent in
regatta).
Table 4. Top Ten Words in Each Assignment Category for Canadian and American
Participants
Canadians (N = 61) Americans (N = 31)
khaki 89% falafel 95% regatta 28% samurai 84% koala 94% Yamaha 45%
Pakistani 84% salsa 93% Picasso 28% fiasco 81% salsa 94% coup d’état 42%
fiasco 84% Guatemala 92% Slavic 27% khaki 77% mantra 90% teriyaki 42%
samurai 84% koala 92% Kafka 26% finale 71% llama 90% éclat 39%
Iraq 80% Zimbabwe 90% coup d’état 24% panorama 65% Zimbabwe 90% tsunami 35%
teriyaki 77% kebab 89% pasta 23% Pakistani 58% Yokohama 90% chianti 35%
panache 75% mafia 88% karate 22% Vladimir 58% mafia 90% Jacques 32%
Vladimir 75% Vivaldi 85% avocado 20% Bantu 52% Vivaldi 90% regatta 32%
angst 74% Bach 84% coup de grâce 19% Wagner 52% Milan 90% Uganda 32%
panorama 72% llama 82% Karachi 19% shiatsu 50% kebab 90% gala 31%
Note: Percentages indicate the proportion of respondents using the vowel at the top of each column in each word.
Most strikingly, the degree of overlap in the sets of words appearing under each
assignment for the two groups varies. For /æ/, six of the ten words are common: for
both groups, fiasco, khaki, Pakistani, panorama, samurai, and Vladimir usually have
/æ/. For /ah-o/, seven are in common: for both groups, kebab, koala, llama, mafia,
salsa, Vivaldi, and Zimbabwe usually have /ah-o/. Among intermediate assignments,
by contrast, only two words appear on both lists: coup d’état and regatta. That this
observation is not influenced by the arbitrary cut-off of ten words in each category can
be assessed by running Pearson correlation tests of the assignment frequencies for
each word for the two national groups, over the whole set of hundred words. These
tests reveal an international correlation of r = 0.784 for /æ/ assignments; 0.808 for
/ah-o/; but only 0.533 for intermediate assignments. There is less national agreement,
then, on which words should be treated as intermediate than on which words should
get /æ/ or /ah-o/.
Boberg 47
At a more general level, looking across the hundred foreign (a) words as a set, the
general pattern of national differences in the two main nativization outcomes reported in
previous studies (see 1.2) is strongly confirmed by the present set of data, as shown in
Figure 2.7 Americans make about 30 percent more assignments to /ah-o/ than Canadians
do, whereas Canadians make about twice as many assignments to /æ/ as Americans do,
differences that are significant at p < .001. With regard to intermediate assignments,
however, the method used here suggests that it is Americans, not Canadians, who make
the most intermediate assignments, the opposite of the national difference reported in
Boberg (2009:368). This is a small difference of only four percent, however, and its sta-
tistical significance depends on how it is measured. Comparing national averages for each
word across all speakers, the national difference in intermediate assignments is significant
at p = .004; comparing national averages of the totals for each speaker across all words,
it is not significant (p = .118). To err on the safe side, given its small size, the difference
in intermediate assignments will be taken to be non-significant; that is, the frequency of
intermediate assignments is roughly equivalent for Americans and Canadians.
Figure 2. Mean Frequency of Phoneme Assignment by National Origin of Participant (/æ/,
intermediate, or /ah-o/ over one hundred foreign-a words)
Note: National difference for /æ/ and /ah-o/ significant by two-tailed t-test at p < .001; difference for
intermediate not significant.
48 Journal of English Linguistics 48(1)
At the level of individual words, the pattern of Figure 2 is supported by the fact that
all of the major national differences involve higher Canadian frequencies of /æ/ and
higher American frequencies of /ah-o/; in fact, although Table 4 shows that some
words display approximately equal variant frequencies for the two national groups,
there is not a single instance of the opposite pattern, in which a significant national
difference involves Canadians using more /ah-o/ and less /æ/ than Americans. The
fourteen words that display the largest national differences (combined differences for
all three outcomes of 75 percent or greater) are shown in Table 5. All of these words
show higher Canadian use of /æ/ and higher American use of /ah-o/.
Despite the national differences just identified, the Canadian and American assign-
ment patterns are closely related. Figure 3 displays this correlation, which can be
quantified as r2 = 0.722, meaning that almost three quarters of each national pattern
can be accounted for by reference to the other, a result significant at p < .01. In other
words, while Canadians make fewer /ah-o/ assignments overall than Americans, when
they do assign to /ah-o/, it tends to be in words that Americans also assign to /ah-o/,
and vice versa.
subsample was too small to support a parallel analysis). A clear sex difference is evident,
significant at p < .001: assignment to /æ/, the more Canadian pattern, is led by women,
whereas assignment to /ah-o/, the more American pattern, is led by men. There is no
significant sex difference in the frequency of intermediate assignment.
Figure 3. Correlation Between Canadian and US Frequency of /ah-o/, by Word, with
Exponential Trend Line
well above the male level. By contrast, on the left side of the chart, where native /æ/ is less
retracted, the sex difference disappears (the two regression lines converge). The women
who make the most foreign (a) assignments to /æ/ are therefore those with the most
retracted native /æ/; since men do not retract native /æ/ as much as women, they make
fewer foreign (a) assignments to /æ/. A substantial number of low-central tokens that fall
into the intermediate or /ah-o/ range for men with a relatively front native /æ/ fall instead
into the /æ/ range of women whose native /æ/ has moved back toward /ah-o/ in the SFVS.
Figure 4. Mean Frequency of Phoneme Assignment for Canadians, by Sex of Participant
Note: Sex difference for /æ/ significant at p = .001; for /ah-o/ at p = .010; for intermediate ns.
assignments. Someone with a comparatively retracted /æ/ and /ah-o/, by contrast, might
be expected to show the opposite pattern, with more /æ/ assignments and fewer /ah-o/
assignments, like the Canadian women in Figure 5. It is therefore worth assessing the
relationship between the phonetic quality of native vowels and the frequency of foreign
(a) assignments to each of them. Table 6 shows the results of a series of Pearson correla-
tion tests of the correlations between the mean F2 frequency or advancement of the
native low vowels, /æ/ and /ah-o/, and the distance between them, and the frequency of
assignments to /æ/, intermediate, and /ah-o/ categories as determined by the two-
standard-deviations method, for the sixty-one Canadian participants.
Figure 5. Correlation Between Mean F2 of Native /æ/ and Frequency of Foreign (a)
Assignment to /æ/ for Canadian Female and Male Participants
Note: Values on the horizontal axis are reversed to conform to the F2 scale of a vowel chart, with
fronter /æ/ on the left and retracted /æ/ on the right.
Table 6. Correlations Between Mean F2 of Low Vowels /æ/ and /ah-o/ and Distance
Between These and Frequency of Assignments to /æ/, Intermediate, and /ah-o/, for sixty-one
Canadian Participants
Note: At 59 df, a two-tailed Pearson test that produces an r value ≥ 0.250 is statistically significant at
p < .05. The r values meeting that criterion are in bold.
52 Journal of English Linguistics 48(1)
Table 6 shows that the phonetic quality of the native low vowels does affect the
foreign (a) assignment patterns of the Canadian participants. As in the female data in
Figure 5, assignments to /æ/ are negatively correlated with the F2 of /æ/, but the first
row of Table 6 further shows that they are also negatively correlated with the distance
between /æ/ and /ah-o/: those with fronter /æ/ and a bigger /æ/-/ah-o/ space are less
likely to assign foreign (a) to /æ/. The bottom row of Table 6 shows the corresponding
effects on the frequency of /ah-o/ assignments, which is a positive correlation, as
would be expected: those with fronter /æ/ and a bigger /æ/-/ah-o/ distance are more
likely to assign foreign (a) to /ah-o/. Finally, the middle row of Table 6 shows that
intermediate assignments are positively correlated with both the F2 of /æ/ and the
/æ/-/ah-o/ distance, but negatively correlated with the F2 of /ah-o/. This means they
are more likely to occur when /æ/ is further front and therefore further from /ah-o/ and
less likely to occur when /ah-o/ is more central. In sum, Table 6 indicates that retrac-
tion of /æ/ toward /ah-o/ in the SFVS, creating a smaller /æ/-/ah-o/ distance, is associ-
ated with more /æ/ and fewer /ah-o/ and intermediate assignments: as native /æ/
moves back, it attracts some foreign (a) tokens that would be intermediate or in the
range of native /ah-o/ in less shifted speech.
Word /æ/ int. /ah-o/ /æ/ int. /ah-o/ /æ/ int. /ah-o/
avocado 33% 35% 33% 18% 52% 30% −45% 52% −10%
Colorado 47% 31% 22% 43% 31% 26% −10% 1% 20%
drama 47% 16% 36% 10% 26% 64% −79% 60% 76%
facade 2% 41% 56% 21% 41% 38% 779% −1% −33%
Iraq 81% 7% 11% 79% 13% 8% −3% 77% −26%
lager 16% 8% 76% 19% 16% 66% 19% 94% −14%
lasagna 4% 35% 62% 13% 33% 54% 261% −5% −12%
lava 41% 41% 19% 5% 25% 70% −88% −40% 281%
llama 34% 17% 49% 8% 11% 80% −76% −32% 64%
macho 4% 13% 84% 2% 13% 85% 203% 136% −29%
mafia 11% 7% 82% 31% 13% 56% −82% 79% 4%
pajamas 95% 4% 2% 54% 25% 21% −43% 576% 1072%
Pakistani 95% 4% 2% 79% 10% 11% −17% 170% 531%
panorama 82% 15% 4% 70% 16% 13% −14% 13% 261%
pasta 67% 20% 13% 48% 33% 20% −29% 64% 55%
Picasso 40% 27% 33% 30% 46% 25% −26% 68% −25%
plaza 38% 44% 18% 28% 38% 34% −27% −14% 89%
Slavic 41% 35% 24% 15% 45% 40% −63% 28% 66%
soprano 78% 11% 11% 59% 25% 16% −25% 125% 50%
taco 25% 25% 49% 8% 25% 67% −68% −3% 37%
MEAN 44% 22% 34% 32% 27% 41% −27% 23% 21%
Note: “int.” = phonetically intermediate between /æ/ and /ah-o/. These data reflect the middle-third
method of analyzing foreign (a) tokens used in Boberg (2009), for purposes of comparison, so the 2012
frequencies differ from those in Appendix A.
Figures 7 and 8 indicate that the shift from /æ/ to /ah-o/ is not uniform across the
twenty words; the pattern only emerges across the whole set. In Table 7, the largest
decreases in /æ/, as a percentage of the 1999 level, are found with lava, mafia, drama,
and llama, whereas façade, lasagna, and macho show increases in /æ/. The largest
proportional increases in /ah-o/ are found with pajamas, Pakistani, lava, and pan-
orama, whereas façade, macho, Iraq, and Picasso show comparatively small decreases
in /ah-o/. Combining these figures, the word lava stands out as the best representative
of the overall trend toward more /ah-o/ and less /æ/.
Another aspect of Figure 6 to note is how it compares with Figure 2, which shows
the overall frequency of the three nativization outcomes determined by the two-stan-
dard-deviations method. This is an inexact comparison, since Figure 6 represents only
twenty of the hundred words measured in Figure 2, but data on the whole set of one
hundred words produce a similar result (see note 7). If the 2012 data of Figure 6 are
54 Journal of English Linguistics 48(1)
compared to the Canadian data of Figure 2, it can be seen that the proportion of inter-
mediate assignments identified by the middle-third method is much higher than that of
the two-standard-deviations method. As discussed in 2.5, this is to be expected, since
the two-standard-deviations method does not guarantee that each speaker has an inter-
mediate range, whereas the middle-third method does, making it more likely that all
speakers will produce at least some intermediate assignments. Nevertheless, apart
from the difference in the frequency of intermediate assignments, the basic propor-
tions of the main /æ/ and /ah-o/ categories are similar and the ranking of the three
outcomes is the same: /ah-o/ is now the most frequent assignment, followed by /æ/,
followed by intermediate.
Figure 6. Mean Frequency of Canadian Phoneme Assignment by Date of Study for Twenty
Foreign (a) Words
Note: By paired t-test, difference for /æ/ significant at p = .001; for intermediate significant at p = .013;
for /ah-o/ marginally significant at p = .089.
word (ignoring intermediate assignments for simplicity’s sake), with large differences
indicating a clear majority pattern of one vowel over the other (e.g., 90 percent /æ/ and
10 percent /ah-o/, a difference of 80 percentage points) and small differences indicat-
ing substantial variation between the two vowels, with the proportions approximately
equal (e.g., 45 percent /æ/ and 55 percent /ah-o/, a difference of 10 percentage points).
Averaged across the one hundred words, the mean difference is 61 percent for
Americans but only 42 percent for Canadians (p < .001). In response to shifting stan-
dards, Canadians exhibit greater variability in vowel assignment than Americans, an
association between change and variation that accords well with the theory of lan-
guage change put forth by Weinreich, Labov, and Herzog (1968).
Figure 7. Comparison of Frequency of Assignment to /æ/ in 1999 and 2012, for Individual
Words (Ranked in Order of /æ/ 1999 Frequency)
in 2.3, in addition to reading the list of foreign (a) and native words, participants were
asked to give their judgments of the similarity of a smaller selection of foreign (a)
words to native words that were either minimal pairs or potential rhyming pairs. In
keeping with the main focus of this study, the analysis of these perceptual data will be
confined to the Canadian subsample. This can be increased from the sixty-one people
analyzed above to sixty-seven, which includes six participants whose interviews were
not of sufficient quality to support reliable acoustic analysis but did produce satisfac-
tory judgment data.
Figure 8. Comparison of Frequency of Assignment to /ah-o/ in 1999 and 2012, for Individual
Words (Ranked in Order of 1999 Frequency)
by uncertainty about the pronunciation of the <l>; in final, pre-pausal position where
<l> is not at issue, the low-central contrast virtually disappears, with 94 percent
(63/67) agreeing, for instance, that paw and spa rhyme. (Paw is historically an /oh/
word, not an /o/ word, but since /o/ cannot occur in this position and is merged with
/oh/, the low-central contrast is assessed in some positions with /oh/ instead.)
Turning to judgments of foreign (a) words, although many individual contrasts with
either /æ/ or /ah-o/ words were studied (the full set of pairs appears in Appendix C), the
analysis will focus on the most crucial case of foreign (a) words that had potential
rhymes or homophony with both /æ/ and /ah-o/ words. If a foreign (a) word were judged
by a substantial number of participants to contrast with both types of native word, its
vowel could be classified as extraphonemic, that is, distinct from the phonemes on
either side of it and therefore “outside” of the native phonemic system. Two-way con-
trasts in minimal pair or rhyming frames between /æ/, /ah-o/, and foreign (a) are not
easy to find, so the available data are limited, but five such cases from among the judg-
ment questions in Appendix C are listed in Table 8. The table focuses purely on partici-
pants’ judgments, regardless of the degree of contrast they actually produced: it shows
the proportion of Canadians claiming that each foreign (a) word sounds the same as
either of its native /æ/ and /ah-o/ pairs (which would reflect phonemic assignment to /æ/
or /ah-o/, respectively), the same as both native words (which could be called ambipho-
nemic), or different from both native words (extraphonemic). Ambiphonemic and
extraphonemic judgments indicate different types of uncertainty in phonemic assign-
ment: the former implies variable pronunciation, with the foreign (a) word matching
first one native word and then a second native word with a different vowel; the latter
implies a phonetically intermediate pronunciation, between the two native phonemes,
which is the sense of “extraphonemic” discussed above.
Table 8. Selected Word Pair Judgments for Sixty-Seven Canadian Participants
Comparison pairs = /æ/ ≠ /ah-o/ = /ah-o/ ≠ /æ/ = /æ/ & /ah-o/ ≠ /æ/ or /ah-o/
Dalí v. dally (A8) and 36% 18% 9% 37%
dolly (A10)
drama v. gamma (B6) 16% 64% 18% 1%
and comma (B2)
façade v. sad (B14) 13% 76% 9% 1%
and sod (B15)
Iraq v. rack (B12) and 76% 10% 10% 3%
rock (B13)
Jacques v. Jack (B9) 3% 82% 3% 3%
and jock (B10)
that many judge it to be the same as the /ah-o/ word dolly, but over a third judge it to
be different from both dally and dolly, suggesting an extraphonemic status for its
vowel. In the other four cases, by contrast, extraphonemic judgments drop to a negli-
gible level. In three of these, drama, façade, and Iraq, there is considerably more
ambiphonemic uncertainty, with perception of phonemic identity shifting from one
comparison to the next for 9-18 percent of participants, even if most of them judge
drama and façade to have /ah-o/ and Iraq to have /æ/. Though it provides a limited
view of only five cases, Table 8 does not, then, offer strong perceptual evidence of
extraphonemic assignment as a common outcome of Canadian foreign (a) nativiza-
tion. On the contrary, it appears to confirm the view motivated by the production data
in Figure 2: phonetically intermediate nativization outcomes are comparatively infre-
quent and most outcomes can be classified as either /æ/ or /ah-o/.
The fact that all young Canadians, not just women, have been shown in previ-
ous research to be participating in the SFVS makes it all the more surprising that
Figures 6-8 indicate a shift from /æ/ to /ah-o/ assignments. Table 6 revealed that the
phonetic developments of the SFVS, in particular the retraction of native /æ/ toward
/ah-o/, are associated with an increase in assignments to /æ/ and a decrease in inter-
mediate and /ah-o/ assignments, the opposite of the change toward /ah-o/ portrayed in
Figure 6. That /ah-o/ assignments are increasing at the expense of /æ/ assignments
despite the compensating development of the retraction of native /æ/ underlines the
dramatic nature of this change, which evidently responds to non-phonological pres-
sure on young Canadians to shift foreign (a) nativization toward a backer vowel
quality.
As mentioned in 2.1, each of the participants in the present study also answered an
opinion survey, which, among other things, asked about national varieties of English
and their associated cultures. While 76 percent of the Canadian respondents believed
Canadian English to be more similar to American than to British English, 60 percent
nevertheless thought British English is the more correct variety; 34 percent thought the
most correct variety is actually Canadian English, while only 4 percent gave that honor
to American English. Responses to the question of which variety sounds more pleasant
are similar: 58 percent said British, 40 percent Canadian, and only 1 percent American.
Moreover, when asked, “What are your general feelings about the United States and
American culture?,” a majority of the Canadians responded either “neutral” (43 per-
cent) or “negative” (27 percent); only 24 percent said “slightly positive” and only 6
percent said “very positive.” There is no evidence, then, of an overtly favorable view
among the young Canadians studied here of either American English or American
culture more generally. If American cultural influence exists, it must be of an indirect
kind at a covert, sub-conscious level (like the covertly positive evaluation of Low
Copenhagen speech by the Danish adolescents studied by Kristiansen [1998:125]).
The positive attributes of American English are evidently associated with a global
youth culture that happens to be centered in the United States, rather than with the
United States per se (for an earlier investigation of these issues with similar results see
Boberg 2010:34-35).
That there is no direct association of Canadian pronunciation with the overt pres-
tige of an American model is even more clearly shown by a lack of correlation between
the frequency of American-style /ah-o/ assignments and positive attitudes toward the
United States. When the participants were organized into groups based on their atti-
tude toward the United States, a t-test found that there were no significant differences
between the groups’ mean frequencies of /ah-o/ assignments. The choice to adopt the
American pattern of assigning foreign (a) words to /ah-o/ is evidently not associated
with overtly positive attitudes toward American culture.
On the other hand, perhaps a straightforward U.S.-Canada cultural difference is
not the criterion that Canadians are referencing in their decisions on phoneme
assignment. A study of Americans’ use of /æ/ versus /ah-o/ in the word Iraq found
that the choice is related partly to politics: Republicans prefer /æ/, a more nativized
vowel that reflects, perhaps, a kind of national pride more than respect for foreign
cultures, while Democrats prefer /ah/, a less nativized vowel that might be seen to
have the opposite connotation (Hall-Lew, Coppock & Starr 2010; see also Silva
et al. 2011).8 Though it is unwise to conclude too much from a study of only one
word, if Canadians absorb these ideological associations through American media,
perhaps the more politically liberal and therefore less pro-American among them
end up using more American vowels not because of their explicitly American cul-
tural associations but because of their association with liberal political views,
which are generally more widespread in Canada than in the United States and par-
ticularly strong among university students, like the participants in this study.
Perhaps /æ/, being the more native vowel, is associated in Canada as much as in the
United States with nativistic, anti-foreign attitudes and therefore rejected by those
Boberg 61
Appendix A. List of One Hundred Foreign (a) Words in Alphabetical Order, with Frequency
of Nativization Outcomes for U.S. and Canadian Participants (/æ/ = trap; /ah-o/ = palm-lot;
int. = intermediate between /æ/ and /ah-o/)
U.S. Canadian
Appendix B. Non-foreign (a) Words on Word List, Arranged by Vowel and Following
Environment (N = 100)
Vowel/Lexical set -vce stop -vce fricative +vce obstruent Final Nasal /l/ /r/
/i/ kit sit kiss did
/e/ dress set best dead ten
guess
/æ-æh/ trap, bath sat pass bad band barrel
staff bag ham carry
gag stamp tariff
sad tan
tag
/o/ lot cot bother bomb collar
sod Don
Tom
/ʌ/ strut cut bus stud
/u/ foot cook stood pull
foot
/iy/ fleece seat seed see
/ey/ face state stayed say
/ay/ price fight spice side pie
sight tide tie
tight
/oy/ choice choice void toy
/uw/ goose boots food do cool
too fool
tool
/ow/ goat boat tooth code go bold
coat toe cold
stole
/aw/ mouth doubt house loud cow down
shout south proud vow gown
town
/ah, ahr/ palm, father blah calm bar
start massage palm car
dark
start
/oh/ thought, caught boss sawed flaw dawn caller
cloth saw
/ɜ˞ / nurse dirt bird stir
Boberg 67
Section B. Rhymes (Words That May Sound the Same Except for the First Sounds)
Acknowledgments
This research would not have been possible without the able assistance of four student research
assistants at McGill University, whose contributions are deeply appreciated. Participants were
interviewed by Tara Glickman, Anna Prokofieva, and Madeleine Revill; acoustic analysis was
conducted by Thomas Kettig and Madeleine Revill. Thanks are also due to the editors and
reviewers of this journal and of Language Variation and Change (where the manuscript was
originally accepted, but length restrictions prevented its full publication), for many helpful sug-
gestions that have inspired important improvements in both analysis and discussion.
Funding
The author disclosed receipt of the following financial support for the research, authorship, and/
or publication of this article: The research presented in this paper was supported by the Social
68 Journal of English Linguistics 48(1)
Sciences and Humanities Research Council of Canada (Standard Research Grant # 410-2011-
1409). This project received ethical approval from McGill University’s Research Ethics Board
II (File # 447-0511).
ORCID iD
Charles Boberg https://orcid.org/0000-0003-0961-6138
Notes
1. Words in small capitals are the keywords established by Wells (1982:xviii-xix) for iden-
tifying and discussing classes of words that share a common vowel phoneme. The trap
set, /æ/ or “short-a,” in addition to the word trap, includes words like tap, back, badge,
scalp, hand, cancel, etc., while the palm set, /ah/ or “long-ah,” includes psalm, father, bra,
spa, lager, etc. The phonemic transcriptions between forward slashes used throughout this
paper are those adopted for broad transcription of English vowels by Labov (1991:7, 13)
and Labov, Ash, and Boberg (2006:12), based on the binary notation first developed by
Trager and Smith (1951): they represent phonemes, or contrastive elements at the phono-
logical level, not the precise phonetic qualities indicated by IPA symbols. For example,
/o/ is the “short-o” of lot, not the /o/ of Spanish todo (which would be English /ow/); /oh/
is the “long-open-o” of thought, with the “h” symbol designating not a voiceless glottal
fricative but the subclass of long, monophthongal or in-gliding vowels; in this system,
short vowels have one symbol and long vowels have two. Since these symbols differ from
some other notational practices, Wells’s keywords are also provided as equivalents at many
points in the text.
2. The distinction between /æ/ in trap and /ah/ in palm and bath attributed here to British
English is not of course universal in British English, but is found in Standard British
English, as well as in the southeastern English varieties on which it is based and in the
Southern Hemisphere varieties it influenced. This paper is not in a position to report on or
discuss foreign (a) nativization patterns in other varieties of British or World English.
3. A few words vary between /æ/ and /ey/, sometimes also including an /ah/ variant. For
example, Latinate vocabulary often shows variable application of the Great Vowel Shift:
data, patriotic, status, and strata can all have either /ey/ or /æ/. Amish, data, gala, tomato,
and vase can have all three vowels in at least some parts of the English-speaking world.
The first chapter of the first episode of Season 1 of the Canadian television comedy Corner
Gas (2004-2009), entitled “Gala or Gayla?,” includes an exchange between two of the
main characters, Brent and Hank, as to whether gala is pronounced with /æ/ or /ey/; in
British English the choice is between /ah/ and /ey/, while in American English, according
to Merriam-Webster’s Collegiate (2003), all three vowels are heard. While most North
Americans now say tomato with /ey/, rhyming with potato and contrasting with British use
of /ah/, there was once enough variation to inspire the Gershwin Brothers’ famous song,
Let’s call the whole thing off (1937). In Canada, the usual substitution of /æ/ for British
/ah/ produced an old form of tomato with /æ/, seldom heard today (Scargill & Warkentyne
1972:57). Use of /ah/ in vase still persists in the United States, where it is considered more
prestigious than /ey/; Labov (1972:117) includes it in his set of “linguistic insecurity” vari-
ables, and Mencken (1937:337) calls it “a fashionable affectation.” Canadian pronuncia-
tion of vase varies in both the vowel and the voicing of the final consonant, comprising
variants rhyming with bras, phase, and face (Scargill & Warkentyne 1972:54).
4. For even earlier comments on aunt, drama, garage, khaki, and vase, though not treated
cohesively as a set of foreign (a) words, and on Canadian use of /æ/ in palm and bath
Boberg 69
words in general, see Avis (1956). Garage rhyming with badge is a traditional Canadian
pronunciation (“frequently heard in Ontario” [Avis 1956:59]) that nicely illustrates the
older Canadian preference for /æ/ (Merriam-Webster’s Collegiate [2003] gives only /ah-o/
for American English and lists /æ/ as Canadian). Some Canadians even have /æ/ in collage
and mirage, but not normally in massage. British English has initial stress in many –age
words, which complicates inter-dialect comparisons of this word set.
5. Though several methods of inter-speaker vowel-space normalization have been developed,
the choice of the Nearey (1978) method was guided by the need for comparability with the
most important previous research on the topic, which also uses the Nearey (1978) method,
including both the previous acoustic study of foreign (a) of Boberg (2009) and the Atlas of
North American English of Labov, Ash, and Boberg (2006:39-40).
6. It is acknowledged that repeated use of t-tests can result in a Type I error (false positive), but
t-tests are the standard method of assessing the statistical significance of a phonemic contrast,
as in the Atlas of North American English, where the Plotnik vowel analysis program devel-
oped for the quantitative analysis of acoustic data “calculates a t-test of the statistical signifi-
cance of the difference between any two vowel means” (Labov, Ash & Boberg 2006:40). It
will be seen that the results of the t-tests used here, given the quantity of data they are based
on, conform very closely to expected patterns of phonemic contrast that are well known from
previous research, suggesting that any effect of Type I errors was negligible.
7. When applied to the present set of data, the middle-third method of categorizing foreign (a)
tokens used in Boberg (2009) produces similar results, as follows: Canadians have 34 percent
/æ/, 25 percent intermediate, and 41 percent /ah-o/; Americans have 15 percent /æ/, 19 percent
intermediate, and 66 percent /ah-o/. The main difference in the Canadian data is that the middle-
third method produces a higher proportion of intermediate tokens, with correspondingly lower
proportions of /æ/ and /ah-o/ tokens, so much so that Canadians have a higher proportion of
intermediate tokens than Americans, in contrast to Figure 2, produced by the two-standard-
deviations method. This difference is also affected by a change in the American proportion of
intermediate tokens, which falls slightly in the middle-third method instead of rising.
8. In Iraq and Iran it is not only the foreign (a) of the stressed second syllable that varies but
also the secondary-stress initial vowel, which can be either /ay/ (price), nativized output of
the Great Vowel Shift, as in irate, or an unshifted vowel, either short /i/ (kit) or long /iy/
(fleece), a quality closer to the foreign sound, as in irrational. The vowels in the two syl-
lables are not entirely independent: while /i/ or /iy/ can occur with either /æ/ or /ah-o/ in the
second syllable, /ay/ normally occurs with /æ/, as a fully nativized form, eye-rack, that is
generally shunned by well-educated speakers (as is the older example eye-talian for Italian).
References
Avis, Walter S. 1956. Speech differences along the Ontario-United States border. III.
Pronunciation. Journal of the Canadian Linguistic Association 2(2). 41-59.
Avis, Walter S. 1973. The English language in Canada. In T. A. Sebeok (ed.), Current trends in
linguistics, vol. 10, Linguistics in North America, 40–74. The Hague: Mouton.
Barber, Katherine (ed.). 1998. The Canadian Oxford dictionary. Toronto: Oxford University
Press.
Becker, Kara (ed.). 2019. The Low-Back-Merger Shift: Uniting the Canadian Vowel Shift, the
California Vowel Shift, and Short Front Vowel Shifts across North America. Durham, NC:
Duke University Press.
Boberg, Charles. 1997. Variation and change in the nativization of foreign (a) in English.
Philadelphia: University of Pennsylvania PhD dissertation.
70 Journal of English Linguistics 48(1)
Boberg, Charles. 1999. The attitudinal component of variation in American English foreign (a)
nativization. Journal of Language and Social Psychology 18(1). 49-61.
Boberg, Charles. 2000. Geolinguistic diffusion and the US-Canada border. Language Variation
and Change 12(1). 1-24.
Boberg, Charles. 2008. Regional phonetic differentiation in Standard Canadian English. Journal
of English Linguistics 36(2). 129-154.
Boberg, Charles. 2009. The emergence of a new phoneme: Foreign (a) in Canadian English.
Language Variation and Change 21(3). 355-380.
Boberg, Charles. 2010. The English language in Canada: Status, history and comparative anal-
ysis. Cambridge, UK: Cambridge University Press.
Boberg, Charles. 2019. A closer look at the Short Front Vowel Shift in Canada. Journal of
English Linguistics 47(2). 91-119.
Boersma, Paul & David Weenink. 2012. Praat: Doing phonetics by computer [computer pro-
gram]. Version 5.3.37, retrieved 1 March, 2013 from www.praat.org.
Chambers, J. K. 1995. The Canada-US border as a vanishing isogloss: The evidence of chester-
field. Journal of English Linguistics 23(1-2). 155-166.
Chambers, J. K. 1998a. Social embedding of changes in progress. Journal of English Linguistics
26(1). 5-36.
Chambers, J. K. 1998b. TV makes people sound the same. In Laurie Bauer & Peter Trudgill
(eds.), Language myths, 123-131. New York: Penguin.
Clarke, Sandra. 1993. The Americanization of Canadian pronunciation: A survey of palatal glide
usage. In Sandra Clarke (ed.), Focus on Canada, 85-108. Amsterdam: John Benjamins.
Clarke, Sandra, Ford Elms & Amani Youssef. 1995. The third dialect of English: Some Canadian
evidence. Language Variation and Change 7(2). 209-228.
Esling, John H. & Henry J. Warkentyne. 1993. Retracting of /æ/ in Vancouver English. In
Sandra Clarke (ed.), Focus on Canada, 229-246. Philadelphia: John Benjamins.
Gorman, Kyle, Jonathan Howell & Michael Wagner. 2011. Prosodylab-Aligner: A tool for
forced alignment of laboratory speech. Proceedings of Acoustics Week in Canada, Quebec
City, retrieved 1 March, 2013 from http://prosodylab.org/tools/aligner.
Gramley, Stephan & Kurt-Michael Pätzold. 1992. A survey of Modern English. London:
Routledge.
Hall-Lew, Lauren, Elizabeth Coppock & Rebecca L. Starr. 2010. Indexing political persuasion:
Variation in the Iraq vowels. American Speech 85(1). 91-102.
Kristiansen, Tore. 1998. The role of standard ideology in the disappearance of the traditional
Danish dialects. Folia Linguistica 32(1-2). 115-129.
Kroch, Anthony S. 1978. Toward a theory of social dialect variation. Language in Society 7(1).
17-36.
Labov, William. 1972. Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press.
Labov, William. 1981. Resolving the Neogrammarian controversy. Language 57(2). 267-308.
Labov, William. 1991. The three dialects of English. In Penelope Eckert (ed.), New ways of
analyzing sound change, 1-44. New York: Academic Press.
Labov, William. 2007. Transmission and diffusion. Language 83(2). 344-387.
Labov, William, Sharon Ash & Charles Boberg. 2006. Atlas of North American English:
Phonetics, phonology and sound change. Berlin: Mouton de Gruyter.
Lindsey, Geoff. 1990. Quantity and quality in British and American vowel systems. In Susan
Ramsaran (ed.), Studies in the pronunciation of English, 106-118. London: Routledge.
Mencken, Henry Louis. 1937. The American language. 4th edn. New York: Knopf.
Merriam-Webster’s Collegiate Dictionary. 2003. 11th edn. Springfield, MA: Merriam-Webster.
Boberg 71
Nearey, Terrance Michael. 1978. Phonetic feature systems for vowels. Bloomington: Indiana
University Linguistics Club.
Nylvek, Judith. 1992. Is Canadian English in Saskatchewan becoming more American?
American Speech 67(3). 268-278.
Rosenfelder, Ingrid, Joe Fruehwald, Keelan Evanini & Jiahong Yuan. 2011. FAVE (Forced
Alignment and Vowel Extraction) Program Suite, retrieved 1 March, 2013 from http://fave.
ling.upenn.edu.
Scargill, Matthew Henry & Henry J. Warkentyne. 1972. The Survey of Canadian English: A
report. English Quarterly 5(3). 47-104.
Shapiro, Michael. 1997. Broad and flat A in marked words. American Speech 72(4). 437-439.
Silva, David J., Sharon A. Peters, Fahad Ben Duhaish, Sok-Hun Kim, Yilmin Koo, Lana Marji
& Junsuk Park. 2011. Variation in the Iraq vowels outside the public forum: The indexing
of political persuasion reconsidered. American Speech 86(2). 179-191.
Stuart-Smith, Jane. 2007. The influence of the media. In Carmen Llamas, Louise Mullany &
Peter Stockwell (eds.), The Routledge companion to sociolinguistics, 140-148. London:
Routledge.
Stuart-Smith, Jane. 2011. The view from the couch: Changing perspectives on the role of the
television in changing language ideologies and use. In Tore Kristiansen & Nikolas Coupland
(eds.), Standard languages and language standards in a changing Europe, 223-239. Oslo,
Norway: Novus Press.
Trager, George L. & Henry Lee Smith Jr. 1951. An outline of English structure. Norman, OK:
Battenburg Press.
Trudgill, Peter. 1986. Dialects in contact. Oxford: Blackwell.
Trudgill, Peter & Jean Hannah. 2008. International English: A guide to the varieties of standard
English. 5th edn. London: Routledge.
Van Coetsem, Frans. 1988. Loan phonology and the two transfer types in language contact.
Dordrecht: Foris.
Warkentyne, Henry J. 1971. Contemporary Canadian English: A report of the Survey of
Canadian English. American Speech 46(3-4). 193-199.
Weinreich, Uriel, William Labov & Marvin I. Herzog. 1968. Empirical foundations for a theory
of language change. In Winfred P. Lehmann & Yakov Malkiel (eds.), Directions for his-
torical linguistics, 95-195. Austin: University of Texas Press.
Wells, J. C. 1982. Accents of English 1: An introduction. 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.
Author Biography
Charles Boberg is Associate Professor of Linguistics at McGill University in Montreal,
Canada. He is a coauthor, with William Labov and Sharon Ash, of The atlas of North American
English: Phonetics, phonology and sound change (Mouton de Gruyter, 2006) and the author of
The English language in Canada: Status, history and comparative analysis (Cambridge, 2010).
Most recently he coedited, with John Nerbonne and Dominic Watt, The handbook of dialectol-
ogy (Wiley Blackwell, 2018). His current research focuses on variation and change in the
vocabulary and phonetics of Canadian English, as well as on accent variation and change in
North American film and television.