+1 862 207 3288 summary for the uploaded article 4
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH
GILLIAN SANKOFF HE´ LE`NE BLONDEAU
University of Pennsylvania University of Florida
We address the articulation between language change in the historical sense and language
change as experienced by individual speakers through a trend and panel study of the change from
apical to dorsal /r/ in Montreal French. The community as a whole rapidly advanced its use of
dorsal [R]. Most individual speakers followed across time were stable after the critical period,
with phonological patterns set by the end of adolescence. A sizeable minority, however, made
substantial changes. The window of opportunity for linguistic modification in later life may be
expanded with rapid change in progress when linguistic variables take on social significance.*
1. INTRODUCTION. Reintegrating synchronic and diachronic linguistics by studying
language change in progress, Weinreich, Labov, and Herzog (1968) demonstrated that
change necessarily involves diffusion. They showed that language change implies variation
(though not the reverse), and that as an innovation diffuses through the speech
community, the social structure of that community conditions its path. Subsequent
research has not, however, directly addressed the question of the articulation between
language change in the historical sense and language change as experienced by individual
speakers. We take this as our central concern in analyzing a phonological change
in Montreal French—the community shift in the pronunciation of /r/—on the basis of
data collected in 1971 and 1984. We address three major goals in our exploration of
the problem. First, we look at the two major types of longitudinal research: trend and
panel studies, and evaluate the relative contribution of each. Second, we assess the
extent to which predictions based on an apparent-time interpretation of our 1971 age
distribution represent a change in progress that has continued, as measured against realtime
longitudinal data. Third, we reconsider the role of the critical period in language
acquisition as it relates to language change.
The analysis of the relation between vocalized and consonantal /r/ in New York City
was a crucial early study (Labov 1966). Labov cites several earlier observers whose
descriptions of New York City held it to be chaotic. According to Hubbell, for example,
‘The pronunciation of a very large number of New Yorkers exhibits a pattern in these
words [those containing post-vocalic /r/] that might most accurately be described as
the complete absence of any pattern’ (Hubbell 1950:48, cited in Labov 1966:36). Labov
replaced this idea of New York City speech as chaos by explaining variation in /r/ (and
* We thank the National Science Foundation for funding this research (Language Change Across the
Lifespan, Grant BCS-0132463 to Gillian Sankoff). This joint research was begun when He´le`ne Blondeau
held a postdoctoral fellowship from the Fonds pour la Formation de Chercheurs et l’Aide a` la Recherche,
Gouvernement du Que´bec, at the University of Pennsylvania in 1999–2001. We gratefully acknowledge the
invaluable contributions of David Sankoff and Henrietta Cedergren in codesigning and implementing (along
with the first author) the original Montreal study in 1971; of Pierrette Thibault and Diane Vincent in carrying
out the 1984 followup study; and of Diane Vincent, Marty Laforest, and Guylaine Martel in undertaking
the 1995 followup. We are especially grateful to Pierrette Thibault for her help in making materials of many
kinds available for our present research, often at short notice, and thank her and Bill Labov for discussion
of theoretical and methodological issues, as well as questions of substance. For their assistance in coding
and verification of the data, we thank Anne Charity, Alice Goffman, Daniel Alejandro Gonzales, Sarah
Moretti, and Sergio Romero. Michael Friesner and Damien Hall not only helped with final coding and
verification, but also made many useful suggestions on an earlier version of the manuscript. We thank Malcah
Yaeger-Dror for checking data on some of the speakers whose vowels she analyzed in the 1971 and 1984
corpora and for helpful comments on the manuscript. We also thank Miriam Meyerhoff, Kirk Hazen, Brian
Joseph, and two anonymous referees for many constructive comments.
560
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 561
the four other phonological variables he examined) in terms of ongoing language
change, structured according to the social stratification of the speech community.
The current article is also a study of /r/. It examines a change in progress in the
pronunciation of /r/ in Montreal French, during a period when the older, apical [r]
pronunciation was ‘in rapid regression’ (Santerre 1978:117) in favor of ‘posterior’ or
dorsal [R].1 The rapidity of this change during the 1960s and 1970s contrasts markedly
with the slow pace of diminishing (r) vocalization in New York (Labov 1994). Before
proceeding to the Montreal case, it is important to explain how the three specific
goals of our article fit into the larger project of understanding the relationship between
historical change and individual speakers.
Though most sociolinguistic studies over the past forty years have necessarily been
synchronic, making use of the present to explain the past (Labov 1975), there are now
a number of longitudinal research projects that have produced valuable results. These
projects, designed mainly as TREND STUDIES, in which communities that had been studied
a decade or more previously were resampled, include Brink & Lund 1979, Cedergren
1988, Trudgill 1988, Thibault & Daveluy 1989, Ashby 2001, Blake & Josey 2003, and
Josey 2004. A second type of longitudinal study is the PANEL STUDY, in which the same
individuals are followed across time. In general, the considerable logistical problems
and consequent expense involved in such research have precluded large-scale panel
followups; several researchers, however, have carried out such studies, usually with
rather smaller samples (e.g. Baugh 1995, Cukor-Avila 2002, Hernandez-Campoy 2003,
Nahkola & Saanilahti 2004). Some studies have included both trend and panel components
(Paunonen 1996, Blondeau 2001, Naro & Scherre 2002, de Paiva & Duarte 2003,
Zilles 2005). Trend studies clearly constitute the best use of resources if the object is
to track language change in progress. However, the combination of trend and panel
research is essential for the interpretation of change, especially as regards the relationship
of change to individual grammars (see also Blondeau 2006 on this point). Both
types of longitudinal sociolinguistic research are crucial in building more informed
models of the role of individual speakers, over their lifetimes, in language change.
Stability of the grammars of individuals after initial language acquisition is the premise
that underlies the concept of APPARENT TIME, introduced by Labov in his early studies
of Martha’s Vineyard (1963) and New York City (1966) as part of a demonstration of
the integration of synchronic and diachronic analysis in linguistics. In both cases, he
used observations made at earlier periods to verify that, by and large, the speech of
older people tends to reflect the state of the language when they acquired it. Table 1
represents Labov’s review of the logical possibilities for interpretation of two different
patterns that might be observed in a synchronic study. When use of a particular variant
is plotted against speaker age, we may see either a flat pattern or a regular increasing
or decreasing slope with age.2
If a synchronic study shows no age differentiation (the ‘flat’ pattern), we can infer
either that no change is occurring—both individual speakers and the community as a
whole are stable (interpretation #1), or that all the speakers in the community are
1 We use [r] as a convenient symbol to represent the apical flap. Tousignant (1987a,c) discusses various
pronunciations of dorsal /r/, including velar fricative ([W] or [x]), uvular fricative [²], and uvular trill [.],
mentioning also the environmentally conditioned variability in degree of voicing. We use the symbol [R]
to include all dorsal variants.
2 Other age-related patterns might of course be possible, but would not be of interest in making temporal
inferences.
562 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
SYNCHRONIC PATTERN INTERPRETATION INDIVIDUAL COMMUNITY
flat 1. Stability stability stability
regular slope with age 2. Age grading change stability
regular slope with age 3. Generational change stability change
( apparent-time interpretation)
flat 4. Communal change change change
TABLE 1. Patterns of change in the individual and the community (adapted from Labov 1994:83).
changing together at the same rate—both older and younger speakers are at the same
stage in a change affecting them equally (interpretation #4). By contrast, a regular slope
with age may mean that generation after generation, individuals change as they get
older, yet the community remains stable over time. According to this interpretation
(#2, AGE GRADING), as each cohort of speakers ages, it steadily increases its use of one
variant of the variable. Labov (1964) provides an example of this in analyzing stages
that children go through in the acquisition of standard English. Alternately, individuals
may retain their childhood patterns, with each age cohort of speakers registering an
increasing use of the variant upon entering the community. This generational change
corresponds to the classic apparent-time interpretation (#3) of change in progress.3
Judicious use of the apparent-time interpretation of age-graded variability has been
a hallmark of sociolinguistic research for the past thirty years. However, given the
largely synchronic nature of most studies, it has not been possible to verify models of
the locus of change with respect to individual speakers. Sankoff 2005, 2006 suggested
a distinction between two separate patterns involving speaker change in a situation
where we observe a regular slope correlated with speaker age. In the first or classic
case, we can use the term age grading in the traditional sense, assuming a generational
pattern that is cyclic or repeats as a function of cultural dictates of what is appropriate to
speakers of a given age. But another possibility is ‘lifespan change’, in which ‘individual
speakers change over their lifespans in the direction of a change in progress in the rest
of the community’ (Sankoff 2005:1011). Such a pattern would be historical (rather
than cyclic) in character. Labov’s (1994) discussion of Cedergren’s (1988) restudy of
(ch) variation in Panamanian Spanish demonstrates that the age distribution includes
both an apparent-time and an age-grading component. It leaves open, however, the
question of whether the age grading observable from the restudy is a matter of ageappropriate
frequencies that would be cyclical in character, or a change in which individuals
over their lifespans are swept along with historical language change in the wider
community. The model put forward in Table 2 represents the addition of a lifespanchange
component distinct from the traditional concept of age grading in that it does
not necessarily imply cylicity. We explore the extent to which such a model is justified
empirically in the present case.
Last, we are interested in how the change is experienced by individual speakers past
puberty, and in the relationship of our study to the critical period. The current literature
modeling language change tends to locate it within the scope of first language acquisition
(e.g. Lightfoot 1999), although this is more a tacit assumption in the field than a
much-researched topic. In a recent treatment of the relationship between brain develop-
3 Figure 1, which is discussed in further detail below, serves as an example of these two possible interpretations.
According to the apparent-time interpretation, the curve represents a change in progress in which [R]
is replacing [r] in the community; according to an age-grading interpretation, young speakers currently using
[R] would progressively replace it with [r] as they age.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 563
SYNCHRONIC PATTERN INTERPRETATION INDIVIDUAL COMMUNITY
flat 1. Stability stability stability
regular slope with age 2a. Age grading change stability
regular slope with age 2b. Lifespan change change change
regular slope with age 3. Generational change stability change
( apparent-time interpretation)
flat 4. Communal change change change
TABLE 2. Addition of a pattern reflecting lifespan change that accompanies change at the level
of the community.
ment and individual phonological systems, Anderson and Lightfoot argue that puberty
poses an important limit:
In the early years of life, the brain develops vast numbers of new synapses, peaking at about 2 years
of age. Over the next several years, neurons die and synapses wither, resulting in a specifically limited
organization. Neuron death levels out at around age 7, while overall synaptic organization is determined
by around the age of puberty. If we assume that the role of experience is to shape and mold this specific
organization, then whatever is in place by puberty is what we are ‘stuck with’. (2002:209; scare quotes
in the original)
Analyzing speakers age fifteen and above across the thirteen-year interval between
1971 and 1984, we verify how labile individuals are in later life, and how our results
relate to what is known about the critical period.
2. /r/ IN MONTREAL FRENCH. As a prelude to his study of /r/ variation in contemporary
German, Wiese (2001) surveys rhotics crosslinguistically. He notes that rhotics may
have at least three places of articulation (‘dental/alveolar; postalveolar/retroflex; and
uvular’), and that ‘manner of articulation . . . is at least as variable, [including] trills,
fricatives, approximants, taps/flaps, and vowels’ (Wiese 2001:11). As in numerous
other languages, Que´be´cois French includes many of these variants. Morin points out,
with justice, that although the articulation of /r/ was variable in Paris in the seventeenth
century, the first French colonists to arrive in New France would probably have been
users of apical [r] (Morin 2002:51). At the end of the nineteenth century, apical [r]
was still the norm in the Province of Quebec. In 1901, Laval University Professor
Adjutor Rivard opined that the dorsal articulation was a ‘vice de prononciation [qui]
vient quelquefois d’une certaine paresse de la langue, plus souvent de l’affectation, ou`
encore d’une habitude contracte´e de`s l’enfance’4 (Morin 2002:51, citing Rivard 1928
[1901]). As of the mid-twentieth century, however, there was a clear geographical split.
Vinay’s survey of 1947–1950 confirmed that Western Quebec, including Montreal,
was characterized by the traditional apical [r], tapped or trilled (Vinay 1950). The
observations of Father Laurent Tremblay in 1941 confirmed the normative acceptability
of this variant at the time (Verreault & Lavoie 1999); Charbonneau (1955) also described
[r] as the Montreal norm. A dorsal [R] (velar fricative or uvular trill) was typical
of the eastern part of the province, a zone comprising all points east of Trois Rivie`res
including Quebec City (Vinay 1950).5
4 Dorsal articulation was a ‘defect of pronunciation which may stem from a certain linguistic laziness,
more often from affectation, or else from a habit contracted from childhood’ (Gillian Sankoff translation).
5 There is no generally agreed upon date as to when Eastern Quebec may have adopted dorsal [R]. Noting
/r/ variability over several centuries on both sides of the Atlantic, Morin proposes a possible scenario for
the eventual establishment of dorsal [R] as the norm in Quebec City and elsewhere in Eastern Quebec. On
this view, apical [r], brought by colonists in the 1600s, would have been established earliest, but dorsal [R]
might have been dominant among colonists settling in the capital (Quebec City) during the second century
of colonization (between 1700 and 1760, when French administration ended). Rivard’s 1928 [1901] remarks
about dorsal [R] being the result of laziness or childhood habits may attest to its already having been quite
well established in Eastern Quebec by that time. Morin (2002:51) notes that in the Atlas linguistique de l’est
564 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
Canonical /r/ as [r] or [R], then, depends largely on the region. Though in the course
of change in Montreal, many speakers vary between the two, we observe only minimal
phonetic conditioning between these variants for the variable speakers (Sankoff &
Blondeau 2007). But a number of other variants occur in rapid speech. These are
generally conditioned according to principles detailed in the careful phonetic studies
of Tousignant (1987a,b,c). Briefly, these variants include:
(i) DELETION: As in other varieties of French, final clusters with liquids are
typically simplified, especially when followed by a consonant-initial word.
Though it is possible to say the compound boucle d’oreille ‘earring’ slowly
and carefully as [buklUd:rejU], it is virtually always pronounced as [bukd:rej].
Similarly the phrase l’autre gars ‘the other guy’ could be pronounced as
[lotrUg:], but is in fact characteristically rendered as [lotg:]. Note that the
careful pronunciation, using schwa, involves resyllabifying such that the /r/s
in these cases are actually in onsets. Deletion is also usual before a pause.
Thus Lise R. (038),6 age seventeen in 1971, in answer to a question about
whether her neighborhood has changed much since she was a small child,
says Bien peut-e^tre (‘well maybe’), deleting the final /r/ in this case as in all
others in pause and preconsonantal environments. Que´be´cois French also
frequently deletes liquids before a vowel in a following word. For example,
Yannick C. (126), seventeen years old in 1984, said Nous aut(res) on va aller
viv(re) avec l’ami de ma me`re (‘We’re going to go live with my mother’s
boyfriend’), deleting the /r/s in the parenthesized material. This was at the
beginning of the interview when Yannick was speaking rather carefully.
(ii) VOCALIZATION: In codas, especially after a final diphthong or lengthened
vowel, /r/ may be lenited to the point of vocalization. Examples of words
where vocalization frequently occurs include boire ‘to drink’, pe`re ‘father’,
pour ‘for’, and sur ‘on’.
(iii) RETROFLEX /r/ is a fairly rare variant that occurs sporadically, often in Englishassociated
words, especially although not only proper nouns like Steinberg
(a now-defunct supermarket chain) and Montre´al (Tousignant 1987b). In coda
position, retroflex /r/ occurs most frequently after the rounded vowels /oe/,
as in peur ‘fear’, conge´lateur ‘freezer’, and Steinberg, and /y/, as in su^r
‘certain’ and pur ‘pure’.
Since these forms did not involve tongue contact with either the alveolar ridge or
the velar/uvular region, they were not considered to represent consonantal targets and
were therefore not included in the quantitative analysis. We follow the practice of the
earlier studies cited above in this regard. Between 1950 and 1970, canonical /r/ in
Montreal began to change, with the dorsal [R] making strong inroads. The conclusion
that by 1970 [r] was in ‘rapid regression’ (Santerre 1978) was confirmed in an apparenttime
calculation by Clermont and Cedergren (1979), henceforward C&C, who found
that the innovative [R] was increasingly common among speakers below age thirtydu
Canada, all the data points from the Eastern Quebec region of Nicolet, Rivard’s original home, were
registered as dorsal [R] in the 1951 survey.
6 As with all the other speaker names cited in this article, ‘Lise’ is a pseudonym. Speaker identification
numbers from the original corpus are also provided since many speakers are referred to by number only in
previous publications.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 565
five. Figure 1, adapted from that study, represents their study of all 120 speakers in
1971.
FIGURE 1. [R]/([R] [r]) by age for 120 speakers in 1971, plotted by age group (data from Clermont &
Cedergren 1979).
Summarizing the results from that study, Cedergren reports that a multiple regression
with rate of [R] use as the dependent variable shows that ‘the single most important
factor . . . was the speakers’ age, followed by social class membership, sex, and educational
level of attainment’ (1988:49). Cedergren verified these apparent-time inferences,
showing that change had already begun at least a decade earlier, using a sample of
twenty-nine interviews from the Bibeau-Dugas corpus established in 1963.
Our first attempt at tracing change since 1971 was strictly limited to a panel comparison
(Sankoff, Blondeau, & Charity 2001, henceforward SBC), and its central issue was
the lability of speakers’ linguistic systems in later life. To this end, we considered data
from the 1984 and 1995 restudies of Montreal (Thibault & Vincent 1990, Vincent et
al. 1995). We examined speaker trajectories for fourteen speakers recorded in both
1971 and 1984, for all twelve of the 1971 speakers who were carried through 1984
to 1995, and for the two additional younger speakers added in 1984 who were also
reinterviewed in 1995.
In the current article, we use an expanded panel, and also make trend comparisons
of speakers sampled in 1971 and 1984, allowing us to document both real-time change
in the speech community as well as the trajectories of /r/ pronunciation of a number
of individual speakers who were followed across that time span. We focus on the social
factors that differentiate speakers in terms of their use of the two canonical variants,
[r] and [R]. The phonological factors that partially condition the alternation between
these variants are dealt with in Sankoff & Blondeau 2008, where we also consider
stylistic factors, which prove to be almost without influence on the alternation.
3. DESIGN AND METHODOLOGY OF THE CURRENT STUDY. It is widely acknowledged
that the most reliable method for studying language change in a community is that of
the trend study (Trudgill 1988, Labov 1994). In the ideal case, the followup study
would resample the community, using the same sampling criteria as the first (de Paiva &
566 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
Duarte 2003, Zilles 2005). In our case, the 1984 survey was designed mainly as a panel
study (Thibault & Vincent 1990); however, the twelve new younger speakers who were
added (see below) were used as a nucleus in designing matched subsamples that allow
us to make trend comparisons. The logic of these comparisons, involving selected
subsamples of thirty-two speakers each (see Figure 2), is explained in §§3.1–3.4.
PANEL SAMPLE
32 speakers
1971
TREND SAMPLE
1971
32 speakers
16 M 16 F
TREND SAMPLE
1984
32 speakers
16 M 16 F
1984
16 M
16 M
16 F
16 F
FIGURE 2. Design of the trend and panel samples for the current study.
The 1971 sociolinguistic study of Montreal consisted of a random sample of 120
speakers, stratified by age, sex, and social class (Sankoff & Sankoff 1973). In 1984,
Thibault and Vincent succeeded in relocating and reinterviewing sixty of the original
interviewees, making this the largest and most representative panel study in sociolinguistics.
Since the youngest speakers of 1971 were by then twenty-eight years old,
they added twelve new speakers between age fifteen and twenty-five, again stratified
according to sex and socioeconomic level. To examine change across individual lifespans,
we could have studied all sixty of the reinterviewed subjects, comparing their
1971 data with their 1984 data. Similarly, we first thought that we might study changes
at the community level by comparing the sixty speakers of 1971 who were NOT reinterviewed
with the different sixty of 1984. However, the subsample of speakers who
were reinterviewed was not comparable to those who were, according to their relevant
social characteristics. First, the mean age of the sixty reinterviewed speakers was thirteen
years older than the speakers of 1971. Second, as a group, the reinterviewed
speakers overrepresented people of a higher social class.7 Last, we decided to sample
from the ninety-five speakers age fifty-seven and under in 1971, mainly on the grounds
that relatively few speakers above that age appeared in the reinterviews of 1984. We
weighted our subsamples toward speakers under forty, who would be most reliably
resampled.
In order to be able to construct matched subsamples of different speakers, we set up
a sampling stratification scheme that would allow us to resample from the pool of
available speakers according to age, sex, and social class. The largest subsamples we
could create, given the constraints described above, consisted of thirty-two individuals
each, as shown in Tables 3–6 below. To represent a range of social classes, we divided
7 Twenty of the original subjects of the 1971 sample were selected from eight census tracts with the
lowest mean family income in the city, a neighborhood known as ‘Centre Sud’ near the Port of Montreal
containing some 27,000 people. Between that time and 1984, a massive highway project had destroyed most
of this neighborhood, and it was very difficult to locate former residents (though Thibault and Vincent were
successful in some cases). It was obviously easier to relocate professional people who could be found through
their business addresses even if they had changed their place of residence.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 567
the population into three groups according to their scores on the LINGUISTIC MARKET
(LM) index.8
3.1. THE PANEL SUBSAMPLE. Constructing a subsample for the panel of speakers to
be followed between 1971 and 1984 necessarily involved selecting from among the
sixty original speakers who were reinterviewed in 1984. Setting aside those over age
fifty-eight as described above, we selected thirty-two speakers to fill the subsample
grid. For most cells, we had more than the required number of speakers available, so
our choice was determined by two factors. First, we always selected the most voluble
speakers (those with the longest interviews), so as to be sure to have plenty of tokens
for analysis. Second, if the sound quality for any speaker’s recording was poor (unfortunately
the case for three or four of the speakers we initially selected), that person was
replaced by another speaker with matching social characteristics. The structure of the
final 1971 panel sample is provided in Table 3.
SOCIAL CLASS
WC (LM 22) LMC (LM 23–59) UMC (LM 60)
AGE M F M F M F TOTAL
15–24 2 2 2 2 3 2 13
25–39 3 2 1 2 2 2 12
40–56 1 2 1 1 1 1 7
TOTAL 6 6 4 5 6 5 32
TABLE 3. Composition of 1971 panel subsample by age, sex, and social class (scores on linguistic market
index divided into 3 groups, corresponding to Working Class, Lower Middle Class,
and Upper Middle Class).
By definition, the panel included a constant group of speakers, so the 1984 panel
comprised the same individuals, thirteen years older, and analyses of the panel compare
the 1984 data to the same individuals’ data from 1971. Since the youngest panel speakers
were twenty-eight years old in 1984, there was no one in the youngest age group of
the subsample grid for that year, as indicated in Table 4. The oldest 1971 panel speaker,
Lucille R. (084), was fifty-six in 1971 and sixty-nine in 1984.
SOCIAL CLASS
WC (LM 22) LMC (LM 23–59) UMC (LM 60)
AGE M F M F M F TOTAL
15–24 0 0 0 0 0 0 0
25–39 4 3 3 2 3 3 18
40–69 2 3 1 3 3 2 14
TOTAL 6 6 4 5 6 5 32
TABLE 4. Composition of 1984 panel subsample by age, sex, and social class (scores on linguistic market
index divided into 3 groups, corresponding to Working Class, Lower Middle Class,
and Upper Middle Class).
8 Having discovered that the six-point socioeconomic index of Thibault and Vincent (1990) and the
linguistic market index devised by D. Sankoff and Laberge (1978), on which scores between 0 and 100 were
assigned to every speaker in 1971, were highly correlated (r .71), we decided to use the more highly
gradated LM index in choosing our subsample. In selecting the speakers, we cut the scale in such a way as
to represent approximately one-third of the speakers in each group, which are labeled (as a heuristic) in
Tables 3–6 as UMC (Upper Middle Class), LMC (Lower Middle Class), and WC (Working Class). The
UMC group certainly also includes a number of Upper Class speakers; and the WC group includes many
members of the underclass—people who in 1971 were unemployed, some of whom were illiterate, had a
year or less of formal schooling, and had lived their entire lives on welfare.
568 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
3.2. THE TREND SUBSAMPLES. The subsample from 1971 that would be used as a
basis for trend comparisons had to be selected from among the sixty speakers who
were NOT reinterviewed in 1984. From this pool of speakers, using the same criteria
outlined above, we selected thirty-two individuals, stratified by age, sex, and social
class as shown in Table 5. For 1971, then, our trend and panel subsamples were selected
from the two halves of the sample of 120 according to the same criteria. Although the
two halves were somewhat disparate for reasons explained above, the subsampling
criteria allowed us to select subsamples that matched each other fairly closely, as
becomes evident in Tables 8 and 9 below.
SOCIAL CLASS
WC (LM 22) LMC (LM 23–59) UMC (LM 60)
AGE M F M F M F TOTAL
15–23 2 2 4 3 3 2 16
24–39 2 2 1 1 1 2 9
40–56 1 1 1 2 1 1 7
TOTAL 5 5 6 6 5 5 32
TABLE 5. Composition of 1971 trend subsample by age, sex, and social class (scores on linguistic
market index divided into 3 groups, corresponding to Working Class, Lower Middle Class,
and Upper Middle Class).
The trend subsample for 1984 had to be selected as if we had returned to the speech
community and taken a new sample according to the criteria of 1971. We of course
included the twelve younger speakers who were added to the study in 1984, matching
them to their 1971 age mates. There were, however, only eleven speakers age twentythree
and under in 1984, compared to sixteen in that age range for the 1971 trend
subsample. For the middle age group, we included the one new speaker age twentyfive,
and had to fill out this group from the ranks of the younger reinterviewed speakers.
Luckily, there were a good number of these people under age forty in 1984. Since the
1984 trend subsample was to be compared to its 1971 counterpart involving different
speakers, but would not be directly compared to the 1984 panel subsample, it was
economic to complete the 1984 trend subsample with speakers who were also part of
the 1984 panel. The 1984 trend subsample, selected to match its 1971 counterpart
insofar as was possible, is shown in Table 6.
SOCIAL CLASS
WC (LM 22) LMC (LM 23–59) UMC (LM 60)
AGE M F M F M F TOTAL
15–23 2 2 2 2 1 2 11
24–39 3 3 2 2 1 2 13
40–56 2 2 1 1 1 1 8
TOTAL 7 7 5 5 3 5 32
TABLE 6. Composition of 1984 trend subsample by age, sex, and social class (scores on linguistic market
index divided into 3 groups, corresponding to Working Class, Lower Middle Class,
and Upper Middle Class).
3.3. DATA CODING AND ANALYSIS. For each speaker in each year, our goal was to
code approximately one hundred tokens, beginning about ten minutes into the interview
(usually starting at the fifth page of the transcript) and continuing to code each /r/ that
occurred. For many speakers, however, deleted and vocalized /r/ constituted a fairly
large proportion of the tokens, and for this reason we often increased the total token
size to about 120, sometimes more. We could usually locate the desired number of
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 569
tokens within about twenty to thirty minutes of speech. Two coders listened to each
tape and noted codes on separate spreadsheets, after which the number of instances of
[R] and [r] were noted and the percentage calculated as [R]/([R] [r]), setting aside
the other variants in this calculation. This procedure is illustrated in Table 7, which
provides the codings of He´le`ne Blondeau and Sarah Morettifor Bernard L. (087’71).
All variants of /r/ were coded for each speaker, but following C&C, only canonical
[R] and [r] entered into our calculations.
CODER [R] [r] VOCALIZED DELETED RETROFLEX INDISTINCT TOTAL [R]/([R][r])
Blondeau 39 44 29 16 2 7 137 47.0%
Moretti30 28 23 35 4 13 133 51.7%
TABLE 7. Codings for Bernard L. in 1971 by He´le`ne Blondeau and Sarah Moretti.
We then ran a chi-square test on the 2 2 table highlighted on the left, and if the
difference between the two coders was not significant at the 0.05 level, as in this case,
we usually considered the value for that speaker in that year to be the mean of the two
codings.9 If the codings were significantly different, a third coder recoded the data,
after which two of the coders (always involving at least one of the two coauthors of
this article) listened to the tape together and reconciled the differences. There were two
major sets of problem cases. First, tokens with a very weak dorsal articulation might
be heard as ‘vocalized’ or ‘indistinct’ by one coder and as the dorsal [R] by another
coder; joint relistening usually resolved these cases. The second type involved the
manner of articulation. Both apical and uvular /r/ can be trilled, and trills made it very
difficult to hear the place of articulation. Spectrographic examination with Praat did
not help increase reliability on the whole, and we therefore followed the original procedures
of C&C and of Tousignant, using auditory coding supplemented by spectrographic
analysis in only a small number of problem cases. Since (as in the case illustrated in
Table 7) the two coders often arrived at slightly different total numbers of tokens in
listening to the same stretch of speech, we can provide only an approximation rather
than a definitive count of the total number of tokens coded for this project. Our estimate
is that the results given in this article represent the analysis of approximately 12,400
tokens, or approximately 115 per speaker sample.
3.4. THE LOGIC OF COMPARISONS. We selected our panel and trend subsamples from
the 120 speakers of 1971 using the same criteria of age, sex, and social class as described
above. Drawing two separate stratified subsamples from the same community, we would
hope to derive the same results. In analyzing the data, we thus first compare the two
1971 subsamples to see whether indeed their behavior is matched in the way we would
expect. We make separate comparisons across older and youngers speakers in the subsamples.
Next, we look at the panel as a group and examine the extent to which these
individuals did or did not change. Finally, the panel’s behavior is set in the context of
the real-time change as revealed by comparing the two trend samples.
4. THE SITUATION IN 1971. The linguistic situation in 1971 with respect to the major
variants of /r/ in Montreal was well documented (C&C, Santerre 1978, Tousignant
1987a, SBC). C&C demonstrated that speakers of higher socioeconomic status and
9 In this particular case, Alice Goffman had also coded the same recording and her result was 49% [R].
Gillian Sankoff had occasion to listen to that tape again later in the coding procedure, and reconciled the
three codings. The data point finally entered for Bernard L. in 1971 was in fact not the mean of all three,
but the figure of 47% [R] originally arrived at by He´le`ne Blondeau.
570 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
younger speakers were most likely to use a higher proportion of the innovative [R]
variant. In selecting our two subsamples from among the 120 speakers they studied,
the first question we ask in the analysis is whether we are justified in treating the two
1971 samples of thirty-two speakers as two halves of a split sample of sixty-four. Can
we reasonably think of them as two sets of people drawn at random to represent the
same community? Tables 8 and 9 make the comparison in terms of means within each
subsample.
1971 PANEL SPEAKERS AGE RANGE MEAN AGE MEAN LM % [R]
16 younger 15–24 19.1 47.1 63.8%
16 older 25–56 37.1 43 41.3%
All 32 speakers 15–56 28.1 45 52.5%
TABLE 8. Mean social characteristics and mean [R] %, 1971 panel sample.
1971 TREND SPEAKERS AGE RANGE MEAN AGE MEAN LM % [R]
16 younger 15–24 18.8 44.5 68.5%
16 older 25–56 39.2 32.6 34.3%
All 32 speakers 15–56 29.0 38.5 51.4%
TABLE 9. Mean social characteristics and mean [R] %, 1971 trend sample.
We have more to say later about individual speakers. However, mean results for the
two groups show that they are indeed quite comparable. Both samples contained thirtytwo
speakers, sixteen men and sixteen women, and both contained from ten to twelve
speakers in each of the three socioeconomic groups. The mean age of the 1971 panel
speakers was 28.1; that of the 1971 trend speakers was 29. The socioeconomic linguistic
market index values were 45 in the case of panel speakers and 38.5 for trend speakers.
Although we tried hard to select the most closely matched sample, we ended up with
speakers of a slightly higher socioeconomic level in the panel, reflecting the fact that
such speakers were somewhat easier to relocate in the 1984 study.10
In selecting our subsamples, we made sure to represent three age groups of speakers,
but for analytical purposes, we divide each subsample into sixteen younger and sixteen
older speakers, whose age ranges are provided in the relevant tables. In both subsamples,
speakers age twenty-four and under show means of [R] usage considerably higher than
speakers twenty-five and older. The mean [R] percentages for older speakers in the
two samples are seven percentage points apart (41.3% for the panel sample and 34.3%
for the trend sample), whereas the younger speakers’ means are within five percentage
points of each other (63.8% for the panel sample and 68.5% for the trend sample).
Grouping younger and older speakers together in each subsample, the means are only
one percentage point apart (52.5% for the panel and 51.4% for the trend sample). This
result allows us to proceed with confidence in studying change over time.
5. CHANGE ACROSS THE LIFESPAN: THE PANEL SPEAKERS OVER TIME. An initial look
at the panel as a group can be obtained by considering the group means for the 1984
panel speakers: the thirty-two individuals of the 1971 panel sample, all thirteen years
older, as shown in Table 10.
1984 PANEL SPEAKERS AGE RANGE MEAN AGE MEAN LM % [R]
16 younger 29–37 32.1 47.1 77.8%
16 older 38–69 50.1 43 50.8%
All 32 speakers 29–69 41.1 45.1 64.2%
TABLE 10. Mean social characteristics and mean [R] %, 1984 panel sample.
10 This point is elaborated in n. 7.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 571
Comparing Table 10 with Table 8 reveals that as a group, these speakers have increased
their use of [R] over the period. The younger speakers’ group mean was 63.8%
in 1971; these same people show a mean of 77.8% in 1984, an increase of more than
ten percentage points. Perhaps more surprising, given the widespread acceptance of
the idea that adults’ linguistic systems are relatively stable, is the increase of almost
ten percentage points, from 41.3% to 50.8%, for the older group. The means, however,
give a misleading picture of the sample members as a homogeneous group. As we
discovered in examining the smaller panel sample across the twenty-four-year span
from 1971 to 1995 (SBC), the increase is far from uniform when we look at individuals.
To separate the speakers into those who were stable and those who changed, we did
a chi-square test comparing each speaker’s total apical and dorsal /r/ tokens in their
1971 and 1984 samples. Ten of the thirty-two speakers were designated as having
changed significantly over the period, nine in the expected direction. The tenth speaker,
Christine Q. (004), appears to be a counterexample, since her change was in the opposite
direction, from 92% to 87% [R]. However, the chi-square probability of 0.05 cannot
be considered since one would expect at least one such movement by chance in a group
of thirty-two speakers. Further, since she remains in the near-categorical range of Table
11, we treated her as one of those speakers who remained stably in that category across
time.
We consider first in Table 11 a comparison between the speakers who were stable
over time: the ten (including Christine Q.) who maintained near-categorical [R], the
innovative variant, and the ten who maintained near-categorical [r], the traditional
Montreal variant. The two groups have very different social profiles: the group we
label ‘early adopters’ of the innovative form are much younger (just a little more than
half the age of the speakers who retain the traditional form) and have a higher mean
social-class index on the linguistic market scale. More striking is the very disproportionate
sex ratio: drawn from a pool of sixteen male and sixteen female speakers, eight of
ten early adopters are female; and seven of ten traditionalists are male, a statistically
significant difference.11 Gender difference is then a third variable, in addition to age
and social class, that is important in determining which speakers are early adopters of
the innovative form.
MEAN SEX
TRAJECTORY NO. OF MEAN AGE RATIO
RANGE TYPE RANGE OF [R] TYPE SPEAKERS LM 1971 FM
categorical or near- 85–100% stable 10 39.1 16.9 82
categorical [R]
categorical or near- 0–17% stable 10 29.0 30.3 37
categorical [r]
TABLE 11. The 20 stable speakers who remained categorical or near-categorical users of one of the two
variants across the study, 1971–1984.
The twelve remaining speakers were those whose behavior was more variable. Midrange
speakers are defined as those whose use of [R] fell between 17% and 84% at
either period. The mid-range speakers stand out in Figure 3, which plots the percentage
of [R] for each of the panel speakers in both 1971 (filled diamonds) and 1984 (hollow
squares). The overall increase of 12.3 percentage points is clearly not attributable to
most speakers having slightly increased their use of the innovative variant. Rather, the
increase is due to nine of the thirty-two individual speakers having made a fairly dra-
11 The significance level of p 0.05 was calculated by Fisher’s exact test.
572 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
matic shift between 1971 and 1984. The trajectories of these nine individuals are designated
by arrows connecting their 1971 with their 1984 data points.
FIGURE 3. Individual percentages of [R]/([R][r]) for the 32 panel speakers for 1971 and 1984. Trajectories
plotted for all speakers who showed a significant difference between the two years.
At the bottom of the graph, we see two groups of the 1971 speakers in the ellipses:
a tight cluster of seven individuals under age thirty, none of whom uses more than
10% [R], and another group of five speakers between the ages of thirty-five and fifty,
and who range between 0% and 17%. Ten of these twelve people form the stable group
of Table 11 in their use of the conservative form. Their 1984 values are displayed in
the two dotted ellipses to the right of each of the 1971 groups. Two of them, however,
Lysiane B. (007) and Alain L. (104), make substantial changes, abandoning the virtually
categorical conservative pattern they displayed in 1971. Along with the other midrange
speakers, their data appears individually in Table 12. Lysiane and Alain are
among the nine speakers whose trajectories appear in Fig. 3.
A majority of mid-range speakers in 1971 (seven out of ten) had moved to the
categorical or near-categorical use of innovative [R] by 1984. They are people we call
‘later adopters’ of the innovative variant. Five of the seven are young; five of the seven
are male. As a group, we would characterize their behavior as catching up with their
peers, the early adopters. The next three people listed in Table 12 are those mid-range
speakers who were stable, somewhat older, behaving more as we had expected older
individuals to behave. Of note is that one of them, Andre´ L. (065), a professional actor
age twenty-seven when we met him in 1971, was the only speaker who exhibited
stylistic variation (Sankoff & Blondeau 2008). In 1995 he maintained virtually the same
overall level of [R] as he had in 1984: 69%. At the bottom of the table are the two
individuals who moved from virtually categorical use of [r] to dominant use of [R],
occupying the upper part of the variable range. Lysiane B. (007) is a case of exceptional
upward social mobility, a twenty-four-year-old factory worker when we met her in
1971, a businesswoman in 1984, and when last interviewed in 1995, a successful realtor.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 573
AGE % [R] % [R]
SPEAKER TRAJECTORY TYPE 1971 1971 1984
Louis-Pierre R. (092) variable › categorical 16 64% 100%
Paul D. (064) variable › categorical 16 21% 95%
Lise L. (071) variable › categorical 17 73% 90%
Guy T. (088) variable › categorical 19 48% 100%
Christian B. (025) variable › categorical 27 68% 98%
Diane R. (066) variable › categorical 40 72% 97%
Joseph R. (075) variable › categorical 53 48% 79%
Andre´ L. (065) stably variable 27 61% n.s. 66%
Henriette N. (015) stably variable 38 77% n.s. 83%
Lucille R. (084) stably variable 56 83% n.s. 75%
Lysiane B. (007) categorical › variable 24 7% 65%
Alain L. (104) categorical › variable 45 17% 66%
TABLE 12. Percentages of [R] use for 12 speakers whose [R] use fell into the mid-range of
variation in either or both 1971 and 1984. Note that ‘categorical’ here refers to
categorical or near-categorical.
Although her own characterization of the major change in her speech over this period
is that she has learned English, it is clear that she has adopted many features characteristic
of middle class speakers during her adult life (Blondeau et al. 2002). As far as Alain
L. (104) is concerned, we were quite surprised that a man of forty-five would have
shown such a remarkable change by the time he was fifty-eight. A car salesman, Alain
was interviewed in 1984 at the car dealership where he worked, and much of the talk
focused on questions of language and of taste. To attribute his relatively high level of [R]
to formality, however, is probably too hasty. We have been unable to show stylistically
conditioned variation for any speaker other than the professional actor Andre´ L. (065),
who was part of the long-term longitudinal sample followed from 1971 through 1995
and who retained his upper mid-range value throughout (SBC, Sankoff & Blondeau
2008).
Table 12 and Fig. 3 show clearly that the mid-range use of both variants is not stable.
Not only do most speakers cluster at the top or the bottom of the range, but mid-range
speakers also tend very strongly to become categorical over time.
We return to a discussion of the matter of individual stability/instability and related
issues in the final section. To conclude our study of the panel results, we find that they
confirm our previous research on a smaller sample (SBC). Looked at in terms of the
critical age, the evidence from this panel is that twenty-three of the thirty-two speakers
(72%) showed stability over the period of their post-fifteen-year-old lives that we examined;
the other nine (28%) increased their use of the innovative form significantly.
Perhaps a more telling comparison is to look at the pool of speakers who had not yet
adopted the innovative form in 1971. Of the twenty-two speakers who might have
changed toward the innovative form, nine, or 41%, did so.
We might now ask: is 41% a lot, or a little, in terms of our expectations about
speakers’ malleability in adult life? Are the speakers who make this adult transition
the leaders of the change in the community? In the next section, in which we present
the results of the trend study, we put these questions into the community perspective.
6. CHANGE IN THE COMMUNITY: THE EVIDENCE OF TREND COMPARISONS. We noted in
§3.2 (comparing Tables 5 and 6) that the 1984 trend sample was quite similar to the
1971 trend sample in terms of the social characteristics of the speakers. In selecting
that subsample, however, we were not able to match the numbers of 1971 for the
574 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
youngest group, since only twelve new young speakers were interviewed in 1984, one
of whom was twenty-five years old. When distributing our trend sample of 1984 according
to the two age groups selected for analysis, there is an unequal number of ‘younger’
and ‘older’ speakers. For this reason, we pay most attention to the comparison of
younger to younger, and older to older, in tracing the trends between 1971 and 1984.
The figures for older and younger speakers reported in Table 9 are repeated in Table
13, along with the comparable results from the 1984 subsample.
TREND SUBSAMPLES AGE RANGE MEAN AGE MEAN LM % [R]
16 younger 1971 15–24 18.8 44.5 68.5%
11 younger 1984 15–24 17.9 60.9 92.5%
16 older 1971 25–61 39.2 32.6 34.3%
21 older 1984 25–61 38.6 40.3 74.9%
TABLE 13. Mean social characteristics and mean [R] %, 1971 and 1984 trend subsamples.
The eleven younger speakers who were added to the sample in 1984 were a good
match for their 1971 counterparts in age, but we note an increase in the mean LM
scores, at least some of which is attributable to overall socioeconomic change in the
community. By 1984, the economic recession of the early 1970s had turned around,
and more families could afford to keep their teenagers in school longer. Although one
would expect higher [R] scores for speakers of higher socioeconomic status, the jump
from 68% [R] in 1971 to 92.5% in 1984 among speakers under twenty appears to
represent a real-time change in the community. Perhaps even more significantly, the
twenty-five-to-sixty-one-year-old trend subsample in 1984, more closely matched for
socioeconomic status with their 1971 counterparts, shows an even greater increase: from
34.3% in 1971 to 74.9% in 1984. Adults of working age in the Montreal community in
1971 were quite likely to be predominant users of [r], but this was no longer the case
in 1984. These trend comparisons clearly indicate that change was vigorous in the
community in the 1970s and 1980s.
The trend study, representing a real-time reassessment of the community after a
thirteen-year interval, is the basis for assessing real-time change. As such, the community
mean indicates a massive shift toward the replacement of [r] by [R] during this
period. Although the mean increase of panel speakers seemed to indicate progress
toward the incoming form, the trend study confirms that at the community level, the
rate of change is much greater, as is evident from the comparisons in Table 14.
MEAN [R], YOUNGER SPEAKERS MEAN [R], OLDER SPEAKERS
Panel, same Two separate trend Panel, same Two separate trend
speakers samples speakers samples
1984 77.8% 92.5% 50.7% 74.9%
1971 63.8% 68.5% 41.3% 34.3%
mean increase 14.0 24.0 9.4 40.6
TABLE 14. Comparison of trend and panel subsamples over time.
Our examination of panel speakers as a group across time confirmed that such group
means may give a misleading picture of what is going on. The examination of individual
speakers led us to infer that the change is being implemented by people who alter their
pronunciation quite rapidly, rather than a steady, incremental raising of levels across
individual lifespans. An examination of the individual speakers in the two trend samples
further illuminates these initial inferences, as can be seen in Figure 4.
Figure 4 indicates that the real-time change reflected in the community means of
Table 13 is a product of a very different distribution of speakers in 1971 and 1984.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 575
FIGURE 4. Individual percentages of [R]/([R][r]) for the 32 trend speakers for 1971 and for the 32 different
trend speakers for 1984.
Seven speakers under age thirty-five in 1971 (filled diamonds inside the dotted rectangle
at the lower left) were predominant users of the conservative [r]; and only six were in
the near-categorical [R] range on the upper left. By 1984, only one speaker under age
thirty-five was a predominant user of the conservative form, and with one mid-range
exception, all of the other younger speakers were clustered within the near-categorical
[R] range. Among those over thirty-five, we see that in 1971, no speaker showed more
than 50% [R]; most were clustered at the bottom of the figure in the categorical [r]
range. In contrast, by 1984 there were six majority users of [R] among those older than
thirty-five.
The panel study showed that early adopters remained stable across the 1971–1984
period. Once having adopted innovative [R], they continued as categorical or nearcategorical
users of this form. By the time they got to be age thirty-five and older,
these early adopters were joined by a number of later adopters, who caught up with
the change in later life. But whereas many younger speakers in 1971 were still majority
users of the conservative form, this was no longer the case in 1984.
7. THE PROCESS OF CHANGE.
7.1. THE RESULTS OF OUR STUDY OF [r] N [R]. What generalizations can be drawn
from the detailed results reviewed in §§5 and 6, and how do they relate to our understanding
of language change? Let us begin by summarizing the major observations
drawn from the trend and panel studies, along with the results of C&C (1979) based
on the entire sample of 120 speakers from 1971.
THE SITUATION IN 1971. Three major generalizations emerge regarding the situation
in 1971. (i) In 1971, speaker age was a highly determinative factor. Of the 120 speakers,
C&C found that among the twenty-five oldest (those age fifty-eight and older in 1971),
twenty were categorical users of [r] and only one had more than 4% [R]. At the other
576 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
end of the spectrum, many speakers younger than thirty-five in 1971 were categorical
or near-categorical users of [R]. (ii) In both trend and panel samples, most speakers
tended to be near-categorical users of either INNOVATIVE [R] (with [R] values ranging
from 85% to 100%), or of CONSERVATIVE [r] (with [R] values ranging from 0% to 17%).
(iii) A minority of speakers exhibited variable behavior, and those variable speakers
tended to occupy the 50% to 80% [R] range.
THE SITUATION IN 1984. By 1984, a number of important changes had ensued. (i)
The real-time evidence from the trend samples shows a massive increase of [R] by
1984. Speakers under twenty-five in 1984 were twenty-five percentage points ahead
of their 1971 counterparts in their use of [R], and speakers between age twenty-five
and fifty-six in 1984 were forty-one percentage points ahead of their 1971 age mates.
(ii) The panel showed a more modest gain. As a group, younger panel speakers increased
their use of [R] by fourteen percentage points as compared with their earlier selves;
older speakers increased by nine percentage points. In terms of change across the
lifespan, panel members divided into three groups whose members behaved differently:
(a) Of the twelve majority-[r] speakers of the 1971 panel, ten were stable through 1984,
and two moved up dramatically to the mid-range, both registering over 60% [R] use.
(b) All ten speakers who registered 85% or more [R] in 1971 stayed within that range
in 1984. (c) Of the ten mid-range speakers of 1971, seven significantly increased their
use of [R], finding themselves in 1984 in the 85% to 100% near-categorical range,
whereas the three remaining mid-range speakers were stable across the thirteen-year
time span. (iii) Among 1984 trend speakers, only one under thirty-five years old used
less than 70% [R]; almost all of these younger speakers were virtually categorical users
of [R].
7.2. RESOLVED AND UNRESOLVED QUESTIONS. By the evidence reviewed in the preceding
section, two possible trajectories can be ruled out as underlying the implementation
of this change in the Montreal speech community: (i) the change does NOT proceed
mainly by speakers slowly increasing their use of [R] across their lifetimes, and (ii)
the change does NOT proceed by successive cohorts of new members entering the speech
community bearing a slightly higher value of [R] usage than their older peers. Rather,
the inevitable conclusion is that change in the community is a result of individual
speakers, especially younger speakers, being added to the pool of majority-users or
categorical users of [R].
What is the process by which new speakers get added to the pool? We saw that both
in 1971 and even more in 1984, many speakers in the fifteen-to-nineteen age range
were already categorical [R] users when they first entered our study. One important
question that we can try to answer on the basis of indirect evidence is: did these young
speakers acquire categorical [R] usage as part of their initial language acquisition as
young children, or did they join the pool of [R] users sometime between first language
acquisition in early childhood and the age of fifteen? Given the fact that almost everyone
over the age of thirty-five in 1971 was a categorical user of [r], we would have to
assume that input from parents of virtually all children in 1971 was categorical [r]. By
1984, the situation would have changed somewhat. Some adolescents born between
1965 and 1970 would have had parents born after 1945 who themselves were predominant
users of [R]. Nevertheless, it seems that the majority of our [R] users whose parents
were native Montrealers would have heard the [r] variant as their primary input in first
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 577
language acquisition. They would thus have differentiated themselves from their parents
in later childhood or early adolescence.
Indeed, we have some direct evidence on this point. Almost all of the interviews
with adolescents were carried out in the family home. Often a parent was present, and
often those parents participated to some extent in the interview, being called on by
their sons or daughters, for example, to supply information about where their grandparents
came from. In some of these cases, we were fortunate to be able to establish the
/r/ pattern of the parent. In sixteen of the thirty-one interviews carried out with young
people between fifteen and twenty (including all 1971 and 1984 interviews), parents
were present, usually the mother. Although we had much smaller samples of these
parents’ speech, we found no cases where the parent appeared to be anything but an
apical [r] user.
We must conclude that a child with two parents who are categorical users of [r]
would have [r] as an initial target form. It follows that in 1971 and to a large extent
also in 1984, most people who arrived at age fifteen or sixteen as categorical users of
[R] were speakers who had actually switched in childhood or early adolescence. It is
possible that in 1971, sixteen-year-old Louis-Pierre R. (092), with 64% [R], was
recorded in the late stages of such a switch, which was complete by 1984, when he
had 100% [R]. This might also be true of some of the other young speakers in Table
12, given in 1.
(1) Paul D. (064) age 16–29 21% N 95%
Lise L. (071) age 17–30 73% N 90%
Guy T. (088) age 19–32 48% N 100%
Christian B. (025) age 27–40 68% N 98%
The trajectories of these later adopters, whom we were able to observe, might simply
be paralleling a similar trajectory made earlier in life by their earlier-adopting peers
whose change we did not observe. Indeed it might also be correct to view the two older
speakers who registered similar increases as ‘very late’ adopters: Diane R. (066), age
40–53, with a 72% N 97% increase, and Joseph R. (075), age 53–66, with a 48% N
79% increase. That a change from virtually categorical [r] to strongly majority [R] over
a lifetime is possible is shown by the case of Lysiane B. (007), whom we were lucky
to be able to follow through 1995. Lysiane went from 7% to 65% to 78% [R] over the
twenty-four-year span from 1971 to 1995 (SBC).
The norm is clearly one of relative stability after adolescence, however, as we saw
from the fact that a majority of panel members did not change so radically. Our conclusion
is that most of those speakers observed to be categorical users of [R] in late
adolescence in fact made the switch earlier in life. Although it is possible that some
children were influenced by older siblings who had already adopted the innovative
form, the fact remains that there is a fairly sharp break between speakers older and
younger than thirty-five in the 1971 data. So while it is possible that some younger
siblings of early adopters may have acquired the innovation as part of their L1 acquisition,
the older siblings themselves must have gone through the switch process.
In considering the source of the innovation and how it spread initially, it is important
to look at both geographic and socio/stylistic considerations. C&C pointed out that
among the Montrealers with parents from a traditionally [R]-pronouncing area of Quebec,
those over forty showed no particular tendency to be early adopters of [R]; however,
they did observe such a tendency for younger speakers. It seems that once the change
578 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
was in progress in Montreal, it was acceptable and even appropriate for younger speakers
to retain the pronunciation of their parents in cases where the parents were from
[R]-pronouncing regions.12
It is clear that we are dealing here with a change from above (the change has a
source outside the community; early adopters are more likely to be female and of a
higher socioeconomic status). But just how aware are community members of these
competing forms of /r/, and what is their stylistic status? We did not carry out subjective
reaction tests at the time when the spontaneous speech was recorded, but there were
some speakers who mentioned regional variation in /r/ pronunciation during the part
of the interview when they were questioned about the differences between Montrealers
and people from elsewhere. In the excerpt in 2 from a 1971 interview, Denise, the wife
of interviewee Alain L. (104) and a Montrealer in her early forties at the time, was the
one who brought up /r/ pronunciation when the topic of ‘whether you can tell where
someone is from by the way they speak’ was introduced.
(2) Denise L.: Mes parents moi venaient du Lac St-Jean. Ils grasseyaient.
‘My parents came from Lac Saint-Jean. They used to pronounce
a ‘‘guttural’’ /r/.’
Moi-me^me quand j’aicommence´ a` aller a` l’e´cole puis—je faisais
rire de moi,
‘Me too when I started at school and—I got laughed at.’
fait-que je me suis dompte´. ‘Je vas arre^ter de rouler les r!’13
‘So I was crushed. ‘‘I’m going to stop rolling my r’s!’’ ’
Interviewer: Vous avez pris vite l’accent de Montre´al?
‘Did you pick up the Montreal accent quickly?’
Alain L. Et bien ma femme est ne´e a` Montre´al,
‘Yes well my wife was born in Montreal,’
mais ses parents et les plus vieux de sa famille sont ne´s au Lac
St-Jean.
‘But her parents and the older people in her family were born
in Lac Saint-Jean.’
Despite a few remarks of this sort that showed awareness of regional differences,
particularly on the part of people who had crossed the east-west (R/r) line in moving
from one area to another, stylistic variation among our 1971 speakers was marginal at
best (Sankoff & Blondeau 2008).
By 1984, the mid-range speakers we examined still did not behave consciously as
if the [R] variant held any particular cachet—for most of them, in the upper range of
variable usage, it had become the unmarked variant. By contrast, for some younger
speakers, [r] seemed to have taken on meanings of being old-fashioned, having the
12 In documenting consonantal changes in the mid-Rhine region of Germany over a thirty-year period,
Bellman and colleagues (1999) found a number of changes moving in the direction of Hochdeutsch, but the
towns where change was occurring were those that bordered on areas that already had these features. The
actual prestige of the Hochdeutsch forms would presumably have been uniform throughout the region, but
the effect of prestige was found only on the boundary, that is, with speakers who had contact with those
who already used the prestige form. In our sample, Yannick C. (126), a younger speaker (seventeen years
old in 1984), was typical of his generation with 100% [R]. One early influence on his speech, however, may
have been the fifteen months he spent with his family in Quebec City at the age of seven when he was in
first grade.
13 Most people contrast grasseye´ (dorsal) with roule´ (apical), but despite the apparent confusion of these
terms on the part of Denise, she clearly means she was under social pressure to stop using the dorsal variant.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 579
ring of one’s grandparents, of one’s teachers, of the clergy. Some younger speakers
used it for comic effect. In 3, Lucie A. (125), age twenty-two in 1984, is talking about
the kind of luxurious home she might like to have. She is very serious throughout
the passage, using her normal [R] in all of the /r/-words: pouvoir, superbes, pouvoir,
re´nover, regarde. Then she pulls herself up short from this reverie and the last three
lines are very ironic, when she contrasts her true situation with this dream of luxury.
In both peut-e^tre and trop, the word is drawn out and the apical [r] clearly signals a
comic effect in which she pokes fun at herself for having these visions of another way
of life.
(3) Lucie A., age 22 (1984)
. . . ils e´taient bien chanceux de pouvoir [R]
‘. . . they were very lucky to be able’
se payer des loyers comme c¸a la`,
‘to pay rents like that,’
puis—des maisons superbes [R] la`. J’aime—
‘and the superb homes. I like—’
j’aime c¸a, la—la la vie qu’il-y-a la`-dedans,
‘I like that, the—the lifestyle there,’
dans ces maisons la`. Ou bi en
‘in that kind of houses. Or else’
de pouvoir [R] acheter une maison
‘to be able to buy a house’
pui s de la re´nover [R] puis c¸a, c¸a—
‘and to renovate it, and that, that—’
En tout cas souvent je pense a` c¸a tu-sais
‘In any case I often think about that y’know’
puis je regarde [R] moicomment ce-que je vis
‘and I look at myself, how I live,’
puis je me dis ‘Oui.’
‘and I say to myself, ‘‘Yeah, right!’’ ’
Peut-e^tre [r] un jour! [R]
‘Maybe one day!’
Je sais pas trop [r] la`.
‘I don’t really know.’
A further mystery concerns the nature of the allophonic variation both between the
two allophones in competition, and between them and the other variants of /r/ (C&C,
Tousignant 1987a, SBC, Sankoff & Blondeau 2008). For variable speakers, there is a
strong tendency to retain [r] in onsets, the most salient position, whereas the innovative
[R] seems to appear first in codas in which /r/ is in any case subject to lenition and
deletion. It would seem that a high-prestige pronunciation ought to occur first in the
most salient position, but in listening to the speech of our variable speakers, it is almost
as if [R] creeps in through the back door, where the least attention is being paid.
Speakers who tend toward categoriality of either the conservative or the innovative
variant tend to have a higher proportion of [r] and [R] tokens overall. That is, these
speakers have many fewer cases of vocalized or deleted forms than do the intermediate
speakers. It is almost as if the intermediate speakers no longer have a clear consonantal
target for /r/. Further details on allophonic conditioning of the other /r/ variants can be
found in Sankoff & Blondeau 2008.
580 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
The issues of allophonic variation and of stylistic significance may help to shed
further light on the implementation of the change, but it is too late to be able to study
the situation of L1 acquisition in families where children ended up with a different
allophone from that of their parents. Nevertheless, we think that there are two important
facts that need to be underlined here: (i) apical [r] must have been the target and the
first-acquired form in L1 acquisition for the vast majority of teenagers and young adults
in Montreal in 1971 who were early adopters of dorsal [R], given the input they would
have had from their parents; and (ii) all of the mid-range speakers under age thirty in
1971 had become categorical or near-categorical users of [R] by 1984. In tracing their
trajectories as late adopters, it is reasonable to infer that their early-adopting peers had
traced the same trajectory earlier in their lives. Figure 5 sketches the trajectories we
envisage for the early adopters, after the pattern of those we observed.
FIGURE 5. Actual trajectories for all mid-range speakers under age 30 in 1971, and putative trajectories for
earlier stages.
7.3. MONTREAL [r] N [R] IN A WIDER PERSPECTIVE. How do our findings relate to
other cases of language change, and to clarifying the relative usefulness of trend and
panel studies in longitudinal analysis? Before drawing what we believe to be the general
implications of our study for the understanding of change in progress, we must point
out two respects in which the case we are dealing with here may differ from that of
other changes. We then conclude by stepping back from the current analysis and making
some generalizations we believe may be applicable to research in other communities.
THE DISCRETE CHARACTER OF MONTREAL /r/. Our findings indicate that this change
differs from others described in the literature in one important way: the relative lack
of stable variation. More speakers tended to be categorical than variable, and those
who changed did so very rapidly.
We noted in §§5 and 6 that (i) most speakers in both 1971 and 1984 were majority
users of one of the two variants, rather than occupying the mid portions of the range,
and that (ii) panel results revealed variable speakers as tending toward categoriality
over time. Such a pattern contrasts markedly with that of most studies of change and
variation in vowels, in which gradualism is the order of the day. Gradual changes in
frequency have been characteristic of postvocalic /r/ in New York City and New England,
where a former r-vocalization pattern is now variable, but where change has been
slow; age differences among subjects interviewed for the Atlas of North American
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 581
English (Labov et al. 2006) were not significant. There is, however, a striking parallel
to the Montreal case in the rapid diffusion of postvocalic /r/ in the American South,
where ‘r-pronunciation has swept through the region in all styles of speech so that
younger white speakers are consistently r-ful’ (Labov et al. 2006:47).14
Another aspect of our findings that differentiates this change from others that have
been studied is that the change in the phonetics of /r/ does not appear to interact with
other aspects of Montreal French phonology. As far as this second point is concerned,
we note the conclusion drawn by van Hout and Van de Velde about crosslinguistic
variation with respect to /r/: ‘Languages may often display almost all of the /r/ sounds
commonly mentioned. Transition from one type of /r/ into another is frequent, not unidirectional,
and irrelevant to the linguistic system’ (2001:3).
Continuing in this vein, they highlight the observation of Wiese (2001) that for
German and other Germanic languages, ‘the pattern of /r/ . . . shows that the phonotactics
is undisturbed by the segmental variations’ (van Hout & Van de Velde 2001:3).
Although we do not agree that variation in /r/ is always without systemic effects (see
for example Labov 1966 on vocalic mergers in New York City that are inhibited by
/r/ pronunciation), it seems to be true that in the Montreal French case at least, the
change in the phonetics of [r] does not have systemic phonological consequences.
Beginning with Labov 1963, many studies that have used the apparent-time construct
as an indicator of change in progress have been studies of vowels and vowel systems;
change is often systemic, with chain shifts and parallel shifts more common than isolated,
individual changes. It would seem that the more embedded any given change is
as part of a structural system, the less susceptible it might be to individual initiative
of the sort we saw on the part of speakers like Lysiane B. Preliminary work on the
diphthongization of the mid vowels /o/ (as in chose) and /eu/ (as in beurre) across
individual lifespans in Montreal (Sankoff et al. 2006) has shown a smaller degree of
change than is the case with /r/.15
THE IMPLICATIONS OF MONTREAL [r] N [R]. Our study has three major implications
that are relevant for other research on change in progress, and for the relationship of
change in the community to change across individual lifespans.
(i) Trend vs. panel comparisons in the assessment of change in progress. Our
study has provided clear evidence for an important methodological principle in longitudinal
research: in order to verify inferences about apparent time drawn from synchronic
studies, the most important tool is a carefully designed trend study. We saw that the
increases made by panel-study members were only a pale reflection of what was going
on in the wider community. In using [R] more frequently, older speakers are by no
means leading the change, although some are behaving somewhat like younger members
of the community.
14 This rapid change, inferred from a coefficient of 14 (probability 0.0001) for an age difference of
twenty-five years (Labov et al. 2006:48), means that a speaker age twenty-five with 100% postvocalic /r/
might have a grandparent age seventy-five who had only 72%.
15 In longitudinal research on Montreal French mid vowels and on the back /&/ i n -ation words, Kemp
and Yaeger-Dror (1991), Yaeger-Dror and Kemp (1992), and Yaeger-Dror (1989, 1994, 1996) have traced
developments in real time through the use of early-twentieth-century radio broadcasts as compared with
similar recordings from the same speakers twenty to forty years later. They also used data from the 1971
and 1984 Montreal corpora. Summarizing the results of panel comparisons involving eight speakers recorded
in 1971 and 1984, Yaeger-Dror states that speakers may continue ‘to advance their [vowel phonology] toward
a newer phonology well into middle age’ (1994:284–86).
582 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
(ii) Implications for the use of the apparent-time construct. In several studies of
longitudinal change reviewed in Labov 1994 (New York City, Panama, Norwich), the
results were similar to those presented here in that the age distribution at time B was
analyzable in terms both of community change since time A and of the component
referred to by Labov as age grading. Although the results reviewed by Labov were
based on trend, not panel, studies, this second component clearly has to do with change
over the lifespan. Labov’s conclusion comparing the restudies of New York City and
Panama City presages the results we found:
In each case, age-grading is the dominant configuration. In succeeding generations, speakers follow the
same pattern across apparent time. But there is also a steady increment of the process at a lower level,
showing that a real-time change is taking place. This suggests that in trying to decide which model is
correct for a given process, we may have been setting up a misleading opposition between age-grading
and generational change. It is possible that age-grading is involved in the mechanism of real-time change
for certain types of linguistic change. (Labov 1994:97)
Our findings, along with this reasoned opinion of Labov’s, motivate our proposal to
add a new category labeled LIFESPAN CHANGE (as proposed in Table 2) to the model of
the relationship between generational change and age grading. This concept differs
from age grading (which generally involves cyclicity), as discussed in §1.
In our research reported here, the trend study was invaluable in establishing both
the direction and the rate of change. However, in most cases where we are interested
in assessing change in progress, we do not have the luxury of a trend study. In such
cases, we want to know how much we can rely on apparent-time inferences. Here the
findings of the panel study must be considered. Insofar as the movement among panel
speakers was decidedly directional, with some older speakers going along with the
change, there is an important conclusion to be drawn about apparent-time inferencing.
If a significant minority of older speakers has actually gone along with the change in
adult life, then means derived from a cohort of older speakers actually overestimate
their rates of use of the variable as children. In an apparent-time calculation, we would
normally assume that if, for example, speakers age forty-five to fifty are using the
innovative variant at a rate of 50%, whereas fifteen-to-twenty year olds are using it at
a rate of 80%, then there has been an increase of thirty percentage points over the
thirty-year period, an average of ten percentage points per decade. But if indeed that
cohort of forty-five-to-fifty year olds have themselves inched up ten percentage points
as a group since their teens, then the true value of that cohort thirty years earlier
would not have been 50% (the rate they show today), but only 40%. The conclusion
is inescapable: to the extent that older speakers change in the direction of change in
progress in their adult lives, apparent time UNDERESTIMATES the rate of change. A major
implication of our study for other research that relies on inferences from systematic
synchronic-age distributions is that apparent-time calculations correctly identify the
direction of change, but may underestimate the rate of change, a conclusion also reached
by Boberg (2004) in a study of Montreal English.16
16 According to Boberg, ‘the behavior of older speakers in late adoption contributes to the speed with
which an innovative feature replaces an obsolescent feature in the community as a whole. In this sense, late
adoption accelerates rather than retards changes in progress’ (2004:258). Contrasting the two phonological
variables he studied with numerous lexical variables, Boberg’s judicious assessment was that late acquisition
by adults was ‘most characteristic of lexical variation’ (ibid, p. 250), rather than phonological variation as
we have seen here.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 583
We concur with Bailey et al. 1991, Bailey 2002, and Tillery & Bailey 2003 that
earlier worries about the validity of apparent time in identifying change in progress
have proven to be exaggerated.
(iii) Implications for the relevance of the critical period to language change.
Expecting relative stability in later life, we were surprised to see such a substantial
minority of speakers making major changes in their pronunciation. Of the twenty-two
speakers who might have changed, nine (about 40%) did so. The most stable group
was the categorical and near-categorical users of [r]: only two of the twelve departed
from this initial pattern. Seven of the ten mid-range speakers increased significantly,
however, moving to the categorical/near-categorical range of use of dorsal [R]. They
included all of those speakers under age thirty in the original study. This and other
evidence led us to conclude that most of the majority users of [R] in the 1970s and
1980s had changed their pronunciation at some time later than the period of initial L1
acquisition.
This evidence should not be interpreted as a blow against the reality of the critical
period, but as an indication of the necessity for greater attention to be focused on the
degree and kind of lability that occurs in later life. Relative stability of individual
speakers in panel studies has been documented by Brink and Lund (1975), Labov and
Auger (1998), and Nahkola and Saanilahti (2004). By contrast, Kemp and Yaeger-Dror
(1991), Baugh (1995), Naro and Scherre (2002), and Zilles (2005) have shown that
some individuals have made significant changes in later life. Guy and Boyd (1990)
presented a cogent scenario of how English speakers deepen their knowledge of the
morphology of tense marking across their lifespans: in the ‘semiweak’ verb class (kept,
told), children’s final t/d deletion patterns with that for monomorphemes; adults, however,
apparently deepen their analysis and treat the t/d of semiweak verbs like past
tense morphemes. And whereas comprehension of second-dialect features has been
shown to increase in adult life (Sankoff 1980), production of second-dialect features
is a piecemeal affair for both adolescents and adults, generally lacking the systematic
character of first language acquisition and subject to considerable interspeaker variability
(Chambers 1992, Bowie 2000, Kerswill&Williams 2000, Sankoff 2004). Meyerhoff
and Walker (2007) have demonstrated that long-term ‘urban sojourners’ from the Caribbean
island of Bequia have retained native grammatical constraints on copula variation
despite considerable exposure to other dialects as adults. It is likely that the adoption
of changes characteristic of younger speakers in one’s own community has many commonalities
with this process.
Our results require us to rethink the relationship between L1 and later language
acquisition, and between acquisition and change. Are late-adopting young adults simply
an epiphenomenon with respect to community change, following along behind the
younger leaders? Or do they have any influence on the children growing up after them?
In the case of a change from above like this one, the model of child innovators with
no adult input seems unlikely.17 Certainly the fact that adults use dorsal [R] at higher
frequencies contributes to the overall input from the community that children hear as
they form their phonological systems. Hay and colleagues (2006) show that in a change
in progress (the fear/fare merger in New Zealand), community members have quite
accurate perceptions of generational and social-class differences between speakers who
17 Such a scenario, with child innovators in the lead and adults trailing behind, seems clearly the correct
one in a case like the change from below involving the spread of the new verb of quotation, be like (Cukor-
Avila 2002, Tagliamonte & D’Arcy 2005).
584 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
do and do not produce the merger, and use this knowledge in differentially decoding
the same acoustic signal. And in a study of caregivers’ speech to children in a Scottish
community where a local phonetic variant of the diphthong (au) is in variation with a
standard variant, Smith and colleagues (2007) have shown that caregivers who use the
local variant almost exclusively in their general spontaneous speech nevertheless use
the standard at very high rates to young children, especially in contexts of teaching
and discipline.
(4) It’s r[u:]nd the wrong way, no turn it r[RA]nd!
(Caregiver to child in Buckie, Scotland; from Smith et al. 2007:70)
The youngest children replicate these very high rates, whereas slightly older children
use more of the local variant. Clearly the (au) variable is highly salient in Buckie, in
contrast to the situation of /r/ in Montreal; however, this study points to social differentiation
in usage patterns during L1 acquisition of a type that may well have an impact
on language change.
A further question involves how to model what is going on for speakers who show
a dramatic range in their trajectories over only thirteen years. Should we view these
people as having restructured their phonologies? Adjustment in the frequencies of two
variants speakers already possessed from childhood seems perhaps like a relatively
minor change. But several people replaced the unmarked, vernacular [r] that they appeared
to use spontaneously and unreflectingly with the innovative [R] characteristic of
speakers younger than themselves. As we demonstrate elsewhere (Sankoff & Blondeau
2008), these speakers have also altered their constraint hierarchies: majority users of
[R] seem to retain [r] only in onsets, which was not characteristic of their earlier
phonological systems as inferrable from their speech production. Pace Anderson and
Lightfoot (2002), our results lead us to believe that adults’ abstract phonological systems
at age thirty, or forty, or fifty are different from what they were in childhood or at
puberty.18 Change is dramatic in childhood, but it does not stop there. To understand
the dynamics of change in the speech community, we must follow language across the
lifespan.
REFERENCES
ANDERSON, STEPHEN, and DAVID LIGHTFOOT. 2002. The language organ: Linguistics as
cognitive physiology. Cambridge: Cambridge University Press.
ASHBY, WILLIAM J. 2001. Un nouveau regard sur la chute du ne en franc¸ais parle´ tourangeau:
S’agit-il d’un changement en cours? French Language Studies 11.1–22.
BAILEY, GUY. 2002. Real and apparent time. The handbook of language variation and
change, ed. by Jack K. Chambers, Peter Trudgill, and Natalie Schilling-Estes, 312–32.
Oxford: Blackwell.
BAILEY, GUY; TOMWIKLE; JAN TILLERY; and LORI SAND. 1991. The apparent time construct.
Language Variation and Change 3.241–64.
BAUGH, JOHN. 1995. Dimensions of a theory of econolinguistics. Towards a social science
of language: Papers in honor of William Labov. Vol. 1: Variation and change in
language and society, ed. by Gregory R. Guy, Crawford Feagin, Deborah Schiffrin,
and John Baugh, 397–419. Amsterdam: John Benjamins.
BELLMANN, GU¨ NTER; JOACHIM HERRGEN; and JU¨ RGEN ERICH SCHMIDT. 1999. Mittelrheinischer
Sprachatlas, Band 4: Konsonantismus (Dialektalita¨t. Konsonanten des westgermanischen
Bezugssystems. Sproßkonsonanten). Tu¨bingen: Niemeyer.
BLAKE RENEE, and MEREDITH JOSEY. 2003. The /ay/ diphthong in a Martha’s Vineyard
community: What can we say 40 years after Labov? Language in Society 32.451–85.
18 Adult phonology is probably also influenced by literacy. See for example Labov 1964 and Treiman et
al. 2002.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 585
BLONDEAU, HE´ LE`NE. 2001. Real-time changes in the paradigm of personal pronouns in
Montreal French. Journal of Sociolinguistics 5.453–74.
BLONDEAU, HE´ LE`NE. 2006. Panel studies and language variation. Encyclopedia of languages
and linguistics, 2nd edn., vol. 9, ed. by Keith Brown, 150–54. Oxford: Elsevier.
BLONDEAU, HE´ LE`NE; GILLIAN SANKOFF; and ANNE CHARITY. 2002. Parcours individuels et
changements linguistiques en cours dans la communaute´ francophone montre´alaise.
Revue que´be´coise de linguistique 31.13–38.
BOBERG, CHARLES. 2004. Real and apparent time in language change: Late adoption of
changes in Montreal English. American Speech 79.250–69.
BOWIE, DAVID. 2000. The effect of geographic mobility on the retention of a local dialect.
Philadelphia: University of Pennsylvania dissertation.
BRINK, LARS, and JO/RN LUND. 1975. Dansk Rigsma°l I–II: Lydudviklingen siden 1840 med
s+rligt henblink pa° sociolekterne i Kobenhavn. [Standard Danish I–II: The phonetic
development since 1840 with special regard to the sociolects in Copenhagen.] Copenhagen:
Gyldendal.
BRINK, LARS, and JO/RN LUND. 1979. Social factors in the sound changes of modern Danish.
Proceedings of the ninth International Congress of Phonetic Sciences, vol. 2, 196–203.
Copenhagen: University of Copenhagen.
CEDERGREN, HENRIETTA. 1988. The spread of language change: Verifying inferences of
linguistic diffusion. Language spread and language policy: Issues, implications, and
case studies (Georgetown University Round Table on Languages and Linguistics 1987),
ed. by Peter H. Lowenberg, 45–60. Washington, DC: Georgetown University Press.
CHAMBERS, JACK K. 1992. Dialect acquisition. Language 68.673–706.
CHARBONNEAU, RENE´ . 1955. Recherches d’une norme phone´tique dans la re´gion de Montre´al.
E´
tudes sur le parler franc¸ais au Canada, 83–98. Sainte-Foy, QC: Presses de l’Universite
´ Laval.
CLERMONT, JEAN, and HENRIETTA CEDERGREN. 1979. Les ‘R’ de ma mere sont perdus dans
l’air. Le franc¸ais parle´: Etudes sociolinguistiques, ed. by Pierrette Thibault, 13–28.
Edmonton, AB: Linguistic Research.
CUKOR-AVILA, PATRICIA. 2002. She say, she go, she be like: Verbs of quotation over time
in African American Vernacular English. American Speech 77.3–31.
DE PAIVA, MARIA, and MARIA EUGENIA DUARTE. 2003. Mudanc¸a lingu¨i´stica em tempo real.
Rio de Janeiro: Capa.
GUY, GREGORY, and SALLY BOYD. 1990. The development of a morphological class. Language
Variation and Change 2.1–18.
HAY, JENNIFER; PAULWARREN; and KATIEDRAGER. 2006. Factors influencing speech perception
in the context of a merger-in-progress. Journal of Phonetics 34.458–84.
HERNANDEZ-CAMPOY, JUAN MANUEL. 2003. Complementary approaches to the diffusion of
standard features in a local community. Social dialectology: In honour of Peter Trudgill,
ed. by David Britain and Jenny Cheshire, 23–37. Amsterdam: John Benjamins.
HUBBELL, ALLAN FORBES. 1950. The pronunciation of English in New York City: Consonants
and vowels. New York: King’s Crown Press.
JOSEY, MEREDITH. 2004. A sociolinguistic study of phonetic variation and change on the
island of Martha’s Vineyard. New York: New York University dissertation.
KEMP, WILLIAM, and MALCAH YAEGER-DROR. 1991. Changing realizations of A in (a)tion
in relation to the front A-back A opposition in Quebec French. New ways of analyzing
sound change, ed. by Penelope Eckert, 127–84. New York: Academic Press.
KERSWILL, PAUL, and ANN WILLIAMS. 2000. Creating a new town koine: Children and
language change in Milton Keynes. Language in Society 29.65–115.
LABOV, WILLIAM. 1963. The social motivation of a sound change. Word 19.273–309. [Revised
as Ch. 1 in Sociolinguistic patterns. Philadelphia: University of Pennsylvania
Press, 1973.]
LABOV, WILLIAM. 1964. Stages in the acquisition of standard English. Social dialects
and language learning, ed. by Roger Shuy, 77–103. Champaign, IL: National Council
of Teachers of English. [Reprinted in Readings in Amerian dialectology, ed. by
586 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
H. B. Allen and Gary Underwood, 473–98. New York: Appleton-Century-Crofts,
1971].
LABOV, WILLIAM. 1966. The social stratification of English in New York City. Washington,
DC: Center for Applied Linguistics. [2nd edn., Cambridge: Cambridge University Press,
2006.]
LABOV, WILLIAM. 1975. On the use of the present to explain the past. Proceedings of the
11th International Congress of Linguists, ed. by Luigi Heilmann, 825–51. Bologna:
Il Mulino. [Reprinted in Linguistics at the crossroads, ed. by Adam Makkai, Valerie
Becker Makkai, and Luigi Heilmann, 226–61. Padova: Liviana, 1977.]
LABOV, WILLIAM. 1994. Principles of linguistic change, vol. 1: Internal factors. Oxford:
Blackwell.
LABOV, WILLIAM; SHARON ASH; and CHARLES BOBERG. 2006. Atlas of North American
English. Berlin: Mouton de Gruyter.
LABOV, WILLIAM, and JULIE AUGER. 1998. The effect of normal aging on discourse: A
sociolinguistic approach. Narrative discourse in neurologically impaired and normal
aging adults, ed. by Hiram Brownell and Yves Joanette, 115–34. San Diego, CA:
Singular.
LIGHTFOOT, DAVID W. 1999. The development of language: Acquisition, change, and evolution.
Oxford: Blackwell.
MEYERHOFF,MIRIAM, and JAMES A.WALKER. 2007. The persistence of variation in individual
grammars: Copula absence in ‘urban sojourners’ and their stay-at-home peers, Bequia
(St Vincent and the Grenadines). Journal of Sociolinguistics 11.346–66.
MORIN, YVES-CHARLES. 2002. Les premiers immigrants et la prononciation du franc¸ais au
Que´bec. Revue que´be´coise de linguistique 31.39–78.
NAHKOLA, KARI, and MARIA SAANILAHTI. 2004. Mapping language changes in real time: A
panel study on Finnish. Language Variation and Change 16.75–92.
NARO, ANTHONY, and MARIA MARTA PEREIRA SCHERRE. 2002. The individual and the community
in real-time linguistic change: Social dimensions. Paper presented at New Ways
of Analyzing Variation (NWAV) 31, Stanford, CA, October 13.
PAUNONEN, HEIKKI. 1996. Language change in apparent time and in real time. Samspel &
Variation: Sprakliga studier tilla¨gnade Bengt Nordberg pa 60-arsdagen, 375–86. Uppsala:
Uppsala Universitet, Institut fo¨r Nordiska Sprak.
RIVARD, ADJUTOR. 1928 [1901]. Manuel de la parole. 2nd edn. Quebec: Garneau.
SANKOFF, DAVID, and SUZANNE LABERGE. 1978. The linguistic market and the statistical
explanation of variability. Linguistic variation: Models and methods, ed. by David
Sankoff, 239–50. New York: Academic Press.
SANKOFF, DAVID, and GILLIAN SANKOFF. 1973. Sample survey methods and computerassisted
analysis in the study of grammatical variation. Canadian languages in their
social context, ed. by Regna Darnell, 7–64. Edmonton, AB: Linguistic Research.
SANKOFF, GILLIAN. 1980. Mutual intelligibility, bilingualism and linguistic boundaries. The
social life of language, by Gillian Sankoff, 133–41. Philadelphia: University of Pennsylvania
Press.
SANKOFF, GILLIAN. 2004. Adolescents, young adults and the critical period: Two case studies
from ‘Seven Up’. Sociolinguistic variation: Critical reflections, ed. by Carmen Fought,
121–39. Oxford: Oxford University Press.
SANKOFF, GILLIAN. 2005. Cross-sectional and longitudinal studies in sociolinguistics. An
international handbook of the science of language and society, vol. 2, ed. by Ulrich
Ammon, Norbert Dittmar, Klaus J. Mattheier, and Peter Trudgill, 1003–13. Berlin:
Mouton de Gruyter.
SANKOFF, GILLIAN. 2006. Age: Apparent time and real time. The encyclopedia of language
and linguistics, 2nd edn., vol. 1, ed. by Keith Brown, 110–16. Oxford: Elsevier.
SANKOFF, GILLIAN, and HE´ LE`NE BLONDEAU. 2008. Instability of the [r] [R] alternation in
Montreal French: The conditioning of a sound change in progress. ‘r-atics 2 (Special
issue of E´ tudes & Travaux), ed. by Hans Van de Velde and Roeland van Hout, to
appear.
SANKOFF, GILLIAN; HE´ LE`NE BLONDEAU; and ANNE CHARITY. 2001. Individual roles in a realtime
change: Montreal (r N R) 1947–1995. In Van de Velde & van Hout, 141–57.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 587
SANKOFF, GILLIAN; MICHAEL FRIESNER; DAMIEN HALL; and HE´ LE`NE BLONDEAU. 2006. Le
facteur temps et le changement phone´tique dans le franc¸ais montre´alais. Paper presented
at Les franc¸ais d’ici, Queen’s University, Kingston, ON, June 5.
SANTERRE, LAURENT. 1978. Les /R/ montre´alais en re´gression rapide. Prote´e 2.117–30.
SMITH, JENNIFER; MERCEDES DURHAM; and LIANE FORTUNE. 2007. ‘Mam, my trousers is
fa’in doon!’: Community, caregiver, and child in the acquisition of variation in a Scottish
dialect. Language Variation and Change 19.63–99.
TAGLIAMONTE, SALI, and ALEX D’ARCY. 2005. He’s like, she’s like: The quotative system
in Canadian youth. Journal of Sociolinguistics 8.493–514.
THIBAULT, PIERRETTE, and MICHELLE DAVELUY. 1989. Quelques traces du passage du temps
dans le parler des Montre´alais, 1971–1984. Language Variation and Change 1.19–45.
THIBAULT, PIERRETTE, and DIANE VINCENT. 1990. Un corpus de franc¸ais parle´. Montre´al
84: Historique, me´thodes et perspectives de recherche. Sainte-Foy, QC: De´partement
de langues et linguistique, Universite´ Laval.
TILLERY, JAN, and GUY BAILEY. 2003. Approaches to real time in dialectology and sociolinguistics.
World Englishes 22.351–65.
TOUSIGNANT, CLAUDE. 1987a. La variation sociolinguistique: Mode`le que´be´cois et me´thode
d’analyse. Montreal: Presses de l’Universite´ du Que´bec.
TOUSIGNANT, CLAUDE. 1987b. Relations sociales entre le /r/ ve´laire et le /r/ re´troflexe anglais
de Montre´al. Revue que´be´coise de linguistique the´orique et applique´e 6.39–55.
TOUSIGNANT, CLAUDE. 1987c. Les variantes du /R/ montre´alais: Contextes phonologiques
favorisant leur apparition. Revue que´be´coise de linguistique the´orique et applique´e
6.73–113.
TREIMAN, REBECCA; JUDY BOWEY; and DERRICK BOURASSA. 2002. Segmentation of spoken
words into syllables by English-speaking children as compared to adults. Journal of
Experimental Child Psychology 83.213–38.
TRUDGILL, PETER. 1988. Norwich revisited: Recent linguistic changes in an English urban
dialect. English World Wide 9.33–49.
VAN DE VELDE, HANS, and ROELAND VAN HOUT (eds.) 2001. ‘r-atics: Sociolinguistic, phonetic
and phonological characteristics of /r/ (Special issue of E´ tudes & Travaux 4).
VAN HOUT, ROELAND, and HANS VAN DE VELDE. 2001. Patterns of /r/ variation. In Van de
Velde & van Hout, 1–9.
VERREAULT, CLAUDE, and THOMAS LAVOIE. 1999. ‘La langue de nos gens’ du pe`re Laurent
Tremblay: Une premie`re synthe`se sur la variation ge´olinguistique du franc¸ais parle´ au
Que´bec au de´but des anne´es 1940. Langues et Linguistique 25.145–212.
VINAY, JEAN-PAUL. 1950. Bout de la langue ou fond de la gorge? French Review 23.489–98.
VINCENT, DIANE; MARTY LAFOREST; and GUYLAINE MARTEL. 1995. Le corpus de Montre´al
1995: Adaptation de la me´thode d’enque^te sociolinguistique pour l’analyse conversationnelle.
Dialangue 6.29–46.
WEINREICH, URIEL; WILLIAM LABOV; and MARVIN HERZOG. 1968. Empirical foundations for
a theory of language change. Directions for historical linguistics, ed. by Winfred Lehmann
and Yakov Malkiel, 97–195. Austin: University of Texas Press.
WIESE, RICHARD. 2001. The unity and variation of (German) /r/. In Van de Velde & Van
Hout, 11–26.
YAEGER-DROR, MALCAH. 1989. Real time vs. apparent time change in Montreal French.
York Papers in Linguistics 13.141–53.
YAEGER-DROR, MALCAH. 1994. Sound change in Montreal French. Phonological structure
and phonetic form: Papers in laboratory phonology 3, ed. by Patricia Keating, 267–92.
Cambridge: Cambridge University Press.
YAEGER-DROR, MALCAH. 1996. Phonetic evidence for the evolution of lexical classes: The
case of a Montreal French vowel shift. Towards a social science of language: Papers
in honor of William Labov. Vol. 2: The linguistic structure of variation and change,
ed. by Gregory R. Guy, Crawford Feagin, John Baugh, and Deborah Schiffrin, 263–87.
Amsterdam: John Benjamins.
YAEGER-DROR, MALCAH, and WILLIAM KEMP. 1992. Lexical classes in Montreal French:
The case of (.:). Language and Speech 35.251–93.
588 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
ZILLES, ANA MARIA S. 2005. The development of a new pronoun: The linguistic and social
embedding of a gente in Brazilian Portuguese. Language Variation and Change
17.19–53.
Sankoff [Received 10 August 2006;
Department of Linguistics accepted 8 February 2007]
University of Pennsylvania
619 Williams Hall
Philadelphia, PA 19104-6305
[gsankoff@earthlink.net]
Blondeau
Department of Romance Languages and Literatures
University of Florida
170 Dauer Hall
P.O. Box 117405
Gainesville, FL 32611-7405
[blondeau@ufl.edu]
PLACE THIS ORDER OR A SIMILAR ORDER WITH US TODAY AND GET AN AMAZING DISCOUNT 🙂
summary for the uploaded article 4
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH
GILLIAN SANKOFF HE´ LE`NE BLONDEAU
University of Pennsylvania University of Florida
We address the articulation between language change in the historical sense and language
change as experienced by individual speakers through a trend and panel study of the change from
apical to dorsal /r/ in Montreal French. The community as a whole rapidly advanced its use of
dorsal [R]. Most individual speakers followed across time were stable after the critical period,
with phonological patterns set by the end of adolescence. A sizeable minority, however, made
substantial changes. The window of opportunity for linguistic modification in later life may be
expanded with rapid change in progress when linguistic variables take on social significance.*
1. INTRODUCTION. Reintegrating synchronic and diachronic linguistics by studying
language change in progress, Weinreich, Labov, and Herzog (1968) demonstrated that
change necessarily involves diffusion. They showed that language change implies variation
(though not the reverse), and that as an innovation diffuses through the speech
community, the social structure of that community conditions its path. Subsequent
research has not, however, directly addressed the question of the articulation between
language change in the historical sense and language change as experienced by individual
speakers. We take this as our central concern in analyzing a phonological change
in Montreal French—the community shift in the pronunciation of /r/—on the basis of
data collected in 1971 and 1984. We address three major goals in our exploration of
the problem. First, we look at the two major types of longitudinal research: trend and
panel studies, and evaluate the relative contribution of each. Second, we assess the
extent to which predictions based on an apparent-time interpretation of our 1971 age
distribution represent a change in progress that has continued, as measured against realtime
longitudinal data. Third, we reconsider the role of the critical period in language
acquisition as it relates to language change.
The analysis of the relation between vocalized and consonantal /r/ in New York City
was a crucial early study (Labov 1966). Labov cites several earlier observers whose
descriptions of New York City held it to be chaotic. According to Hubbell, for example,
‘The pronunciation of a very large number of New Yorkers exhibits a pattern in these
words [those containing post-vocalic /r/] that might most accurately be described as
the complete absence of any pattern’ (Hubbell 1950:48, cited in Labov 1966:36). Labov
replaced this idea of New York City speech as chaos by explaining variation in /r/ (and
* We thank the National Science Foundation for funding this research (Language Change Across the
Lifespan, Grant BCS-0132463 to Gillian Sankoff). This joint research was begun when He´le`ne Blondeau
held a postdoctoral fellowship from the Fonds pour la Formation de Chercheurs et l’Aide a` la Recherche,
Gouvernement du Que´bec, at the University of Pennsylvania in 1999–2001. We gratefully acknowledge the
invaluable contributions of David Sankoff and Henrietta Cedergren in codesigning and implementing (along
with the first author) the original Montreal study in 1971; of Pierrette Thibault and Diane Vincent in carrying
out the 1984 followup study; and of Diane Vincent, Marty Laforest, and Guylaine Martel in undertaking
the 1995 followup. We are especially grateful to Pierrette Thibault for her help in making materials of many
kinds available for our present research, often at short notice, and thank her and Bill Labov for discussion
of theoretical and methodological issues, as well as questions of substance. For their assistance in coding
and verification of the data, we thank Anne Charity, Alice Goffman, Daniel Alejandro Gonzales, Sarah
Moretti, and Sergio Romero. Michael Friesner and Damien Hall not only helped with final coding and
verification, but also made many useful suggestions on an earlier version of the manuscript. We thank Malcah
Yaeger-Dror for checking data on some of the speakers whose vowels she analyzed in the 1971 and 1984
corpora and for helpful comments on the manuscript. We also thank Miriam Meyerhoff, Kirk Hazen, Brian
Joseph, and two anonymous referees for many constructive comments.
560
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 561
the four other phonological variables he examined) in terms of ongoing language
change, structured according to the social stratification of the speech community.
The current article is also a study of /r/. It examines a change in progress in the
pronunciation of /r/ in Montreal French, during a period when the older, apical [r]
pronunciation was ‘in rapid regression’ (Santerre 1978:117) in favor of ‘posterior’ or
dorsal [R].1 The rapidity of this change during the 1960s and 1970s contrasts markedly
with the slow pace of diminishing (r) vocalization in New York (Labov 1994). Before
proceeding to the Montreal case, it is important to explain how the three specific
goals of our article fit into the larger project of understanding the relationship between
historical change and individual speakers.
Though most sociolinguistic studies over the past forty years have necessarily been
synchronic, making use of the present to explain the past (Labov 1975), there are now
a number of longitudinal research projects that have produced valuable results. These
projects, designed mainly as TREND STUDIES, in which communities that had been studied
a decade or more previously were resampled, include Brink & Lund 1979, Cedergren
1988, Trudgill 1988, Thibault & Daveluy 1989, Ashby 2001, Blake & Josey 2003, and
Josey 2004. A second type of longitudinal study is the PANEL STUDY, in which the same
individuals are followed across time. In general, the considerable logistical problems
and consequent expense involved in such research have precluded large-scale panel
followups; several researchers, however, have carried out such studies, usually with
rather smaller samples (e.g. Baugh 1995, Cukor-Avila 2002, Hernandez-Campoy 2003,
Nahkola & Saanilahti 2004). Some studies have included both trend and panel components
(Paunonen 1996, Blondeau 2001, Naro & Scherre 2002, de Paiva & Duarte 2003,
Zilles 2005). Trend studies clearly constitute the best use of resources if the object is
to track language change in progress. However, the combination of trend and panel
research is essential for the interpretation of change, especially as regards the relationship
of change to individual grammars (see also Blondeau 2006 on this point). Both
types of longitudinal sociolinguistic research are crucial in building more informed
models of the role of individual speakers, over their lifetimes, in language change.
Stability of the grammars of individuals after initial language acquisition is the premise
that underlies the concept of APPARENT TIME, introduced by Labov in his early studies
of Martha’s Vineyard (1963) and New York City (1966) as part of a demonstration of
the integration of synchronic and diachronic analysis in linguistics. In both cases, he
used observations made at earlier periods to verify that, by and large, the speech of
older people tends to reflect the state of the language when they acquired it. Table 1
represents Labov’s review of the logical possibilities for interpretation of two different
patterns that might be observed in a synchronic study. When use of a particular variant
is plotted against speaker age, we may see either a flat pattern or a regular increasing
or decreasing slope with age.2
If a synchronic study shows no age differentiation (the ‘flat’ pattern), we can infer
either that no change is occurring—both individual speakers and the community as a
whole are stable (interpretation #1), or that all the speakers in the community are
1 We use [r] as a convenient symbol to represent the apical flap. Tousignant (1987a,c) discusses various
pronunciations of dorsal /r/, including velar fricative ([W] or [x]), uvular fricative [²], and uvular trill [.],
mentioning also the environmentally conditioned variability in degree of voicing. We use the symbol [R]
to include all dorsal variants.
2 Other age-related patterns might of course be possible, but would not be of interest in making temporal
inferences.
562 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
SYNCHRONIC PATTERN INTERPRETATION INDIVIDUAL COMMUNITY
flat 1. Stability stability stability
regular slope with age 2. Age grading change stability
regular slope with age 3. Generational change stability change
( apparent-time interpretation)
flat 4. Communal change change change
TABLE 1. Patterns of change in the individual and the community (adapted from Labov 1994:83).
changing together at the same rate—both older and younger speakers are at the same
stage in a change affecting them equally (interpretation #4). By contrast, a regular slope
with age may mean that generation after generation, individuals change as they get
older, yet the community remains stable over time. According to this interpretation
(#2, AGE GRADING), as each cohort of speakers ages, it steadily increases its use of one
variant of the variable. Labov (1964) provides an example of this in analyzing stages
that children go through in the acquisition of standard English. Alternately, individuals
may retain their childhood patterns, with each age cohort of speakers registering an
increasing use of the variant upon entering the community. This generational change
corresponds to the classic apparent-time interpretation (#3) of change in progress.3
Judicious use of the apparent-time interpretation of age-graded variability has been
a hallmark of sociolinguistic research for the past thirty years. However, given the
largely synchronic nature of most studies, it has not been possible to verify models of
the locus of change with respect to individual speakers. Sankoff 2005, 2006 suggested
a distinction between two separate patterns involving speaker change in a situation
where we observe a regular slope correlated with speaker age. In the first or classic
case, we can use the term age grading in the traditional sense, assuming a generational
pattern that is cyclic or repeats as a function of cultural dictates of what is appropriate to
speakers of a given age. But another possibility is ‘lifespan change’, in which ‘individual
speakers change over their lifespans in the direction of a change in progress in the rest
of the community’ (Sankoff 2005:1011). Such a pattern would be historical (rather
than cyclic) in character. Labov’s (1994) discussion of Cedergren’s (1988) restudy of
(ch) variation in Panamanian Spanish demonstrates that the age distribution includes
both an apparent-time and an age-grading component. It leaves open, however, the
question of whether the age grading observable from the restudy is a matter of ageappropriate
frequencies that would be cyclical in character, or a change in which individuals
over their lifespans are swept along with historical language change in the wider
community. The model put forward in Table 2 represents the addition of a lifespanchange
component distinct from the traditional concept of age grading in that it does
not necessarily imply cylicity. We explore the extent to which such a model is justified
empirically in the present case.
Last, we are interested in how the change is experienced by individual speakers past
puberty, and in the relationship of our study to the critical period. The current literature
modeling language change tends to locate it within the scope of first language acquisition
(e.g. Lightfoot 1999), although this is more a tacit assumption in the field than a
much-researched topic. In a recent treatment of the relationship between brain develop-
3 Figure 1, which is discussed in further detail below, serves as an example of these two possible interpretations.
According to the apparent-time interpretation, the curve represents a change in progress in which [R]
is replacing [r] in the community; according to an age-grading interpretation, young speakers currently using
[R] would progressively replace it with [r] as they age.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 563
SYNCHRONIC PATTERN INTERPRETATION INDIVIDUAL COMMUNITY
flat 1. Stability stability stability
regular slope with age 2a. Age grading change stability
regular slope with age 2b. Lifespan change change change
regular slope with age 3. Generational change stability change
( apparent-time interpretation)
flat 4. Communal change change change
TABLE 2. Addition of a pattern reflecting lifespan change that accompanies change at the level
of the community.
ment and individual phonological systems, Anderson and Lightfoot argue that puberty
poses an important limit:
In the early years of life, the brain develops vast numbers of new synapses, peaking at about 2 years
of age. Over the next several years, neurons die and synapses wither, resulting in a specifically limited
organization. Neuron death levels out at around age 7, while overall synaptic organization is determined
by around the age of puberty. If we assume that the role of experience is to shape and mold this specific
organization, then whatever is in place by puberty is what we are ‘stuck with’. (2002:209; scare quotes
in the original)
Analyzing speakers age fifteen and above across the thirteen-year interval between
1971 and 1984, we verify how labile individuals are in later life, and how our results
relate to what is known about the critical period.
2. /r/ IN MONTREAL FRENCH. As a prelude to his study of /r/ variation in contemporary
German, Wiese (2001) surveys rhotics crosslinguistically. He notes that rhotics may
have at least three places of articulation (‘dental/alveolar; postalveolar/retroflex; and
uvular’), and that ‘manner of articulation . . . is at least as variable, [including] trills,
fricatives, approximants, taps/flaps, and vowels’ (Wiese 2001:11). As in numerous
other languages, Que´be´cois French includes many of these variants. Morin points out,
with justice, that although the articulation of /r/ was variable in Paris in the seventeenth
century, the first French colonists to arrive in New France would probably have been
users of apical [r] (Morin 2002:51). At the end of the nineteenth century, apical [r]
was still the norm in the Province of Quebec. In 1901, Laval University Professor
Adjutor Rivard opined that the dorsal articulation was a ‘vice de prononciation [qui]
vient quelquefois d’une certaine paresse de la langue, plus souvent de l’affectation, ou`
encore d’une habitude contracte´e de`s l’enfance’4 (Morin 2002:51, citing Rivard 1928
[1901]). As of the mid-twentieth century, however, there was a clear geographical split.
Vinay’s survey of 1947–1950 confirmed that Western Quebec, including Montreal,
was characterized by the traditional apical [r], tapped or trilled (Vinay 1950). The
observations of Father Laurent Tremblay in 1941 confirmed the normative acceptability
of this variant at the time (Verreault & Lavoie 1999); Charbonneau (1955) also described
[r] as the Montreal norm. A dorsal [R] (velar fricative or uvular trill) was typical
of the eastern part of the province, a zone comprising all points east of Trois Rivie`res
including Quebec City (Vinay 1950).5
4 Dorsal articulation was a ‘defect of pronunciation which may stem from a certain linguistic laziness,
more often from affectation, or else from a habit contracted from childhood’ (Gillian Sankoff translation).
5 There is no generally agreed upon date as to when Eastern Quebec may have adopted dorsal [R]. Noting
/r/ variability over several centuries on both sides of the Atlantic, Morin proposes a possible scenario for
the eventual establishment of dorsal [R] as the norm in Quebec City and elsewhere in Eastern Quebec. On
this view, apical [r], brought by colonists in the 1600s, would have been established earliest, but dorsal [R]
might have been dominant among colonists settling in the capital (Quebec City) during the second century
of colonization (between 1700 and 1760, when French administration ended). Rivard’s 1928 [1901] remarks
about dorsal [R] being the result of laziness or childhood habits may attest to its already having been quite
well established in Eastern Quebec by that time. Morin (2002:51) notes that in the Atlas linguistique de l’est
564 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
Canonical /r/ as [r] or [R], then, depends largely on the region. Though in the course
of change in Montreal, many speakers vary between the two, we observe only minimal
phonetic conditioning between these variants for the variable speakers (Sankoff &
Blondeau 2007). But a number of other variants occur in rapid speech. These are
generally conditioned according to principles detailed in the careful phonetic studies
of Tousignant (1987a,b,c). Briefly, these variants include:
(i) DELETION: As in other varieties of French, final clusters with liquids are
typically simplified, especially when followed by a consonant-initial word.
Though it is possible to say the compound boucle d’oreille ‘earring’ slowly
and carefully as [buklUd:rejU], it is virtually always pronounced as [bukd:rej].
Similarly the phrase l’autre gars ‘the other guy’ could be pronounced as
[lotrUg:], but is in fact characteristically rendered as [lotg:]. Note that the
careful pronunciation, using schwa, involves resyllabifying such that the /r/s
in these cases are actually in onsets. Deletion is also usual before a pause.
Thus Lise R. (038),6 age seventeen in 1971, in answer to a question about
whether her neighborhood has changed much since she was a small child,
says Bien peut-e^tre (‘well maybe’), deleting the final /r/ in this case as in all
others in pause and preconsonantal environments. Que´be´cois French also
frequently deletes liquids before a vowel in a following word. For example,
Yannick C. (126), seventeen years old in 1984, said Nous aut(res) on va aller
viv(re) avec l’ami de ma me`re (‘We’re going to go live with my mother’s
boyfriend’), deleting the /r/s in the parenthesized material. This was at the
beginning of the interview when Yannick was speaking rather carefully.
(ii) VOCALIZATION: In codas, especially after a final diphthong or lengthened
vowel, /r/ may be lenited to the point of vocalization. Examples of words
where vocalization frequently occurs include boire ‘to drink’, pe`re ‘father’,
pour ‘for’, and sur ‘on’.
(iii) RETROFLEX /r/ is a fairly rare variant that occurs sporadically, often in Englishassociated
words, especially although not only proper nouns like Steinberg
(a now-defunct supermarket chain) and Montre´al (Tousignant 1987b). In coda
position, retroflex /r/ occurs most frequently after the rounded vowels /oe/,
as in peur ‘fear’, conge´lateur ‘freezer’, and Steinberg, and /y/, as in su^r
‘certain’ and pur ‘pure’.
Since these forms did not involve tongue contact with either the alveolar ridge or
the velar/uvular region, they were not considered to represent consonantal targets and
were therefore not included in the quantitative analysis. We follow the practice of the
earlier studies cited above in this regard. Between 1950 and 1970, canonical /r/ in
Montreal began to change, with the dorsal [R] making strong inroads. The conclusion
that by 1970 [r] was in ‘rapid regression’ (Santerre 1978) was confirmed in an apparenttime
calculation by Clermont and Cedergren (1979), henceforward C&C, who found
that the innovative [R] was increasingly common among speakers below age thirtydu
Canada, all the data points from the Eastern Quebec region of Nicolet, Rivard’s original home, were
registered as dorsal [R] in the 1951 survey.
6 As with all the other speaker names cited in this article, ‘Lise’ is a pseudonym. Speaker identification
numbers from the original corpus are also provided since many speakers are referred to by number only in
previous publications.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 565
five. Figure 1, adapted from that study, represents their study of all 120 speakers in
1971.
FIGURE 1. [R]/([R] [r]) by age for 120 speakers in 1971, plotted by age group (data from Clermont &
Cedergren 1979).
Summarizing the results from that study, Cedergren reports that a multiple regression
with rate of [R] use as the dependent variable shows that ‘the single most important
factor . . . was the speakers’ age, followed by social class membership, sex, and educational
level of attainment’ (1988:49). Cedergren verified these apparent-time inferences,
showing that change had already begun at least a decade earlier, using a sample of
twenty-nine interviews from the Bibeau-Dugas corpus established in 1963.
Our first attempt at tracing change since 1971 was strictly limited to a panel comparison
(Sankoff, Blondeau, & Charity 2001, henceforward SBC), and its central issue was
the lability of speakers’ linguistic systems in later life. To this end, we considered data
from the 1984 and 1995 restudies of Montreal (Thibault & Vincent 1990, Vincent et
al. 1995). We examined speaker trajectories for fourteen speakers recorded in both
1971 and 1984, for all twelve of the 1971 speakers who were carried through 1984
to 1995, and for the two additional younger speakers added in 1984 who were also
reinterviewed in 1995.
In the current article, we use an expanded panel, and also make trend comparisons
of speakers sampled in 1971 and 1984, allowing us to document both real-time change
in the speech community as well as the trajectories of /r/ pronunciation of a number
of individual speakers who were followed across that time span. We focus on the social
factors that differentiate speakers in terms of their use of the two canonical variants,
[r] and [R]. The phonological factors that partially condition the alternation between
these variants are dealt with in Sankoff & Blondeau 2008, where we also consider
stylistic factors, which prove to be almost without influence on the alternation.
3. DESIGN AND METHODOLOGY OF THE CURRENT STUDY. It is widely acknowledged
that the most reliable method for studying language change in a community is that of
the trend study (Trudgill 1988, Labov 1994). In the ideal case, the followup study
would resample the community, using the same sampling criteria as the first (de Paiva &
566 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
Duarte 2003, Zilles 2005). In our case, the 1984 survey was designed mainly as a panel
study (Thibault & Vincent 1990); however, the twelve new younger speakers who were
added (see below) were used as a nucleus in designing matched subsamples that allow
us to make trend comparisons. The logic of these comparisons, involving selected
subsamples of thirty-two speakers each (see Figure 2), is explained in §§3.1–3.4.
PANEL SAMPLE
32 speakers
1971
TREND SAMPLE
1971
32 speakers
16 M 16 F
TREND SAMPLE
1984
32 speakers
16 M 16 F
1984
16 M
16 M
16 F
16 F
FIGURE 2. Design of the trend and panel samples for the current study.
The 1971 sociolinguistic study of Montreal consisted of a random sample of 120
speakers, stratified by age, sex, and social class (Sankoff & Sankoff 1973). In 1984,
Thibault and Vincent succeeded in relocating and reinterviewing sixty of the original
interviewees, making this the largest and most representative panel study in sociolinguistics.
Since the youngest speakers of 1971 were by then twenty-eight years old,
they added twelve new speakers between age fifteen and twenty-five, again stratified
according to sex and socioeconomic level. To examine change across individual lifespans,
we could have studied all sixty of the reinterviewed subjects, comparing their
1971 data with their 1984 data. Similarly, we first thought that we might study changes
at the community level by comparing the sixty speakers of 1971 who were NOT reinterviewed
with the different sixty of 1984. However, the subsample of speakers who
were reinterviewed was not comparable to those who were, according to their relevant
social characteristics. First, the mean age of the sixty reinterviewed speakers was thirteen
years older than the speakers of 1971. Second, as a group, the reinterviewed
speakers overrepresented people of a higher social class.7 Last, we decided to sample
from the ninety-five speakers age fifty-seven and under in 1971, mainly on the grounds
that relatively few speakers above that age appeared in the reinterviews of 1984. We
weighted our subsamples toward speakers under forty, who would be most reliably
resampled.
In order to be able to construct matched subsamples of different speakers, we set up
a sampling stratification scheme that would allow us to resample from the pool of
available speakers according to age, sex, and social class. The largest subsamples we
could create, given the constraints described above, consisted of thirty-two individuals
each, as shown in Tables 3–6 below. To represent a range of social classes, we divided
7 Twenty of the original subjects of the 1971 sample were selected from eight census tracts with the
lowest mean family income in the city, a neighborhood known as ‘Centre Sud’ near the Port of Montreal
containing some 27,000 people. Between that time and 1984, a massive highway project had destroyed most
of this neighborhood, and it was very difficult to locate former residents (though Thibault and Vincent were
successful in some cases). It was obviously easier to relocate professional people who could be found through
their business addresses even if they had changed their place of residence.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 567
the population into three groups according to their scores on the LINGUISTIC MARKET
(LM) index.8
3.1. THE PANEL SUBSAMPLE. Constructing a subsample for the panel of speakers to
be followed between 1971 and 1984 necessarily involved selecting from among the
sixty original speakers who were reinterviewed in 1984. Setting aside those over age
fifty-eight as described above, we selected thirty-two speakers to fill the subsample
grid. For most cells, we had more than the required number of speakers available, so
our choice was determined by two factors. First, we always selected the most voluble
speakers (those with the longest interviews), so as to be sure to have plenty of tokens
for analysis. Second, if the sound quality for any speaker’s recording was poor (unfortunately
the case for three or four of the speakers we initially selected), that person was
replaced by another speaker with matching social characteristics. The structure of the
final 1971 panel sample is provided in Table 3.
SOCIAL CLASS
WC (LM 22) LMC (LM 23–59) UMC (LM 60)
AGE M F M F M F TOTAL
15–24 2 2 2 2 3 2 13
25–39 3 2 1 2 2 2 12
40–56 1 2 1 1 1 1 7
TOTAL 6 6 4 5 6 5 32
TABLE 3. Composition of 1971 panel subsample by age, sex, and social class (scores on linguistic market
index divided into 3 groups, corresponding to Working Class, Lower Middle Class,
and Upper Middle Class).
By definition, the panel included a constant group of speakers, so the 1984 panel
comprised the same individuals, thirteen years older, and analyses of the panel compare
the 1984 data to the same individuals’ data from 1971. Since the youngest panel speakers
were twenty-eight years old in 1984, there was no one in the youngest age group of
the subsample grid for that year, as indicated in Table 4. The oldest 1971 panel speaker,
Lucille R. (084), was fifty-six in 1971 and sixty-nine in 1984.
SOCIAL CLASS
WC (LM 22) LMC (LM 23–59) UMC (LM 60)
AGE M F M F M F TOTAL
15–24 0 0 0 0 0 0 0
25–39 4 3 3 2 3 3 18
40–69 2 3 1 3 3 2 14
TOTAL 6 6 4 5 6 5 32
TABLE 4. Composition of 1984 panel subsample by age, sex, and social class (scores on linguistic market
index divided into 3 groups, corresponding to Working Class, Lower Middle Class,
and Upper Middle Class).
8 Having discovered that the six-point socioeconomic index of Thibault and Vincent (1990) and the
linguistic market index devised by D. Sankoff and Laberge (1978), on which scores between 0 and 100 were
assigned to every speaker in 1971, were highly correlated (r .71), we decided to use the more highly
gradated LM index in choosing our subsample. In selecting the speakers, we cut the scale in such a way as
to represent approximately one-third of the speakers in each group, which are labeled (as a heuristic) in
Tables 3–6 as UMC (Upper Middle Class), LMC (Lower Middle Class), and WC (Working Class). The
UMC group certainly also includes a number of Upper Class speakers; and the WC group includes many
members of the underclass—people who in 1971 were unemployed, some of whom were illiterate, had a
year or less of formal schooling, and had lived their entire lives on welfare.
568 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
3.2. THE TREND SUBSAMPLES. The subsample from 1971 that would be used as a
basis for trend comparisons had to be selected from among the sixty speakers who
were NOT reinterviewed in 1984. From this pool of speakers, using the same criteria
outlined above, we selected thirty-two individuals, stratified by age, sex, and social
class as shown in Table 5. For 1971, then, our trend and panel subsamples were selected
from the two halves of the sample of 120 according to the same criteria. Although the
two halves were somewhat disparate for reasons explained above, the subsampling
criteria allowed us to select subsamples that matched each other fairly closely, as
becomes evident in Tables 8 and 9 below.
SOCIAL CLASS
WC (LM 22) LMC (LM 23–59) UMC (LM 60)
AGE M F M F M F TOTAL
15–23 2 2 4 3 3 2 16
24–39 2 2 1 1 1 2 9
40–56 1 1 1 2 1 1 7
TOTAL 5 5 6 6 5 5 32
TABLE 5. Composition of 1971 trend subsample by age, sex, and social class (scores on linguistic
market index divided into 3 groups, corresponding to Working Class, Lower Middle Class,
and Upper Middle Class).
The trend subsample for 1984 had to be selected as if we had returned to the speech
community and taken a new sample according to the criteria of 1971. We of course
included the twelve younger speakers who were added to the study in 1984, matching
them to their 1971 age mates. There were, however, only eleven speakers age twentythree
and under in 1984, compared to sixteen in that age range for the 1971 trend
subsample. For the middle age group, we included the one new speaker age twentyfive,
and had to fill out this group from the ranks of the younger reinterviewed speakers.
Luckily, there were a good number of these people under age forty in 1984. Since the
1984 trend subsample was to be compared to its 1971 counterpart involving different
speakers, but would not be directly compared to the 1984 panel subsample, it was
economic to complete the 1984 trend subsample with speakers who were also part of
the 1984 panel. The 1984 trend subsample, selected to match its 1971 counterpart
insofar as was possible, is shown in Table 6.
SOCIAL CLASS
WC (LM 22) LMC (LM 23–59) UMC (LM 60)
AGE M F M F M F TOTAL
15–23 2 2 2 2 1 2 11
24–39 3 3 2 2 1 2 13
40–56 2 2 1 1 1 1 8
TOTAL 7 7 5 5 3 5 32
TABLE 6. Composition of 1984 trend subsample by age, sex, and social class (scores on linguistic market
index divided into 3 groups, corresponding to Working Class, Lower Middle Class,
and Upper Middle Class).
3.3. DATA CODING AND ANALYSIS. For each speaker in each year, our goal was to
code approximately one hundred tokens, beginning about ten minutes into the interview
(usually starting at the fifth page of the transcript) and continuing to code each /r/ that
occurred. For many speakers, however, deleted and vocalized /r/ constituted a fairly
large proportion of the tokens, and for this reason we often increased the total token
size to about 120, sometimes more. We could usually locate the desired number of
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 569
tokens within about twenty to thirty minutes of speech. Two coders listened to each
tape and noted codes on separate spreadsheets, after which the number of instances of
[R] and [r] were noted and the percentage calculated as [R]/([R] [r]), setting aside
the other variants in this calculation. This procedure is illustrated in Table 7, which
provides the codings of He´le`ne Blondeau and Sarah Morettifor Bernard L. (087’71).
All variants of /r/ were coded for each speaker, but following C&C, only canonical
[R] and [r] entered into our calculations.
CODER [R] [r] VOCALIZED DELETED RETROFLEX INDISTINCT TOTAL [R]/([R][r])
Blondeau 39 44 29 16 2 7 137 47.0%
Moretti30 28 23 35 4 13 133 51.7%
TABLE 7. Codings for Bernard L. in 1971 by He´le`ne Blondeau and Sarah Moretti.
We then ran a chi-square test on the 2 2 table highlighted on the left, and if the
difference between the two coders was not significant at the 0.05 level, as in this case,
we usually considered the value for that speaker in that year to be the mean of the two
codings.9 If the codings were significantly different, a third coder recoded the data,
after which two of the coders (always involving at least one of the two coauthors of
this article) listened to the tape together and reconciled the differences. There were two
major sets of problem cases. First, tokens with a very weak dorsal articulation might
be heard as ‘vocalized’ or ‘indistinct’ by one coder and as the dorsal [R] by another
coder; joint relistening usually resolved these cases. The second type involved the
manner of articulation. Both apical and uvular /r/ can be trilled, and trills made it very
difficult to hear the place of articulation. Spectrographic examination with Praat did
not help increase reliability on the whole, and we therefore followed the original procedures
of C&C and of Tousignant, using auditory coding supplemented by spectrographic
analysis in only a small number of problem cases. Since (as in the case illustrated in
Table 7) the two coders often arrived at slightly different total numbers of tokens in
listening to the same stretch of speech, we can provide only an approximation rather
than a definitive count of the total number of tokens coded for this project. Our estimate
is that the results given in this article represent the analysis of approximately 12,400
tokens, or approximately 115 per speaker sample.
3.4. THE LOGIC OF COMPARISONS. We selected our panel and trend subsamples from
the 120 speakers of 1971 using the same criteria of age, sex, and social class as described
above. Drawing two separate stratified subsamples from the same community, we would
hope to derive the same results. In analyzing the data, we thus first compare the two
1971 subsamples to see whether indeed their behavior is matched in the way we would
expect. We make separate comparisons across older and youngers speakers in the subsamples.
Next, we look at the panel as a group and examine the extent to which these
individuals did or did not change. Finally, the panel’s behavior is set in the context of
the real-time change as revealed by comparing the two trend samples.
4. THE SITUATION IN 1971. The linguistic situation in 1971 with respect to the major
variants of /r/ in Montreal was well documented (C&C, Santerre 1978, Tousignant
1987a, SBC). C&C demonstrated that speakers of higher socioeconomic status and
9 In this particular case, Alice Goffman had also coded the same recording and her result was 49% [R].
Gillian Sankoff had occasion to listen to that tape again later in the coding procedure, and reconciled the
three codings. The data point finally entered for Bernard L. in 1971 was in fact not the mean of all three,
but the figure of 47% [R] originally arrived at by He´le`ne Blondeau.
570 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
younger speakers were most likely to use a higher proportion of the innovative [R]
variant. In selecting our two subsamples from among the 120 speakers they studied,
the first question we ask in the analysis is whether we are justified in treating the two
1971 samples of thirty-two speakers as two halves of a split sample of sixty-four. Can
we reasonably think of them as two sets of people drawn at random to represent the
same community? Tables 8 and 9 make the comparison in terms of means within each
subsample.
1971 PANEL SPEAKERS AGE RANGE MEAN AGE MEAN LM % [R]
16 younger 15–24 19.1 47.1 63.8%
16 older 25–56 37.1 43 41.3%
All 32 speakers 15–56 28.1 45 52.5%
TABLE 8. Mean social characteristics and mean [R] %, 1971 panel sample.
1971 TREND SPEAKERS AGE RANGE MEAN AGE MEAN LM % [R]
16 younger 15–24 18.8 44.5 68.5%
16 older 25–56 39.2 32.6 34.3%
All 32 speakers 15–56 29.0 38.5 51.4%
TABLE 9. Mean social characteristics and mean [R] %, 1971 trend sample.
We have more to say later about individual speakers. However, mean results for the
two groups show that they are indeed quite comparable. Both samples contained thirtytwo
speakers, sixteen men and sixteen women, and both contained from ten to twelve
speakers in each of the three socioeconomic groups. The mean age of the 1971 panel
speakers was 28.1; that of the 1971 trend speakers was 29. The socioeconomic linguistic
market index values were 45 in the case of panel speakers and 38.5 for trend speakers.
Although we tried hard to select the most closely matched sample, we ended up with
speakers of a slightly higher socioeconomic level in the panel, reflecting the fact that
such speakers were somewhat easier to relocate in the 1984 study.10
In selecting our subsamples, we made sure to represent three age groups of speakers,
but for analytical purposes, we divide each subsample into sixteen younger and sixteen
older speakers, whose age ranges are provided in the relevant tables. In both subsamples,
speakers age twenty-four and under show means of [R] usage considerably higher than
speakers twenty-five and older. The mean [R] percentages for older speakers in the
two samples are seven percentage points apart (41.3% for the panel sample and 34.3%
for the trend sample), whereas the younger speakers’ means are within five percentage
points of each other (63.8% for the panel sample and 68.5% for the trend sample).
Grouping younger and older speakers together in each subsample, the means are only
one percentage point apart (52.5% for the panel and 51.4% for the trend sample). This
result allows us to proceed with confidence in studying change over time.
5. CHANGE ACROSS THE LIFESPAN: THE PANEL SPEAKERS OVER TIME. An initial look
at the panel as a group can be obtained by considering the group means for the 1984
panel speakers: the thirty-two individuals of the 1971 panel sample, all thirteen years
older, as shown in Table 10.
1984 PANEL SPEAKERS AGE RANGE MEAN AGE MEAN LM % [R]
16 younger 29–37 32.1 47.1 77.8%
16 older 38–69 50.1 43 50.8%
All 32 speakers 29–69 41.1 45.1 64.2%
TABLE 10. Mean social characteristics and mean [R] %, 1984 panel sample.
10 This point is elaborated in n. 7.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 571
Comparing Table 10 with Table 8 reveals that as a group, these speakers have increased
their use of [R] over the period. The younger speakers’ group mean was 63.8%
in 1971; these same people show a mean of 77.8% in 1984, an increase of more than
ten percentage points. Perhaps more surprising, given the widespread acceptance of
the idea that adults’ linguistic systems are relatively stable, is the increase of almost
ten percentage points, from 41.3% to 50.8%, for the older group. The means, however,
give a misleading picture of the sample members as a homogeneous group. As we
discovered in examining the smaller panel sample across the twenty-four-year span
from 1971 to 1995 (SBC), the increase is far from uniform when we look at individuals.
To separate the speakers into those who were stable and those who changed, we did
a chi-square test comparing each speaker’s total apical and dorsal /r/ tokens in their
1971 and 1984 samples. Ten of the thirty-two speakers were designated as having
changed significantly over the period, nine in the expected direction. The tenth speaker,
Christine Q. (004), appears to be a counterexample, since her change was in the opposite
direction, from 92% to 87% [R]. However, the chi-square probability of 0.05 cannot
be considered since one would expect at least one such movement by chance in a group
of thirty-two speakers. Further, since she remains in the near-categorical range of Table
11, we treated her as one of those speakers who remained stably in that category across
time.
We consider first in Table 11 a comparison between the speakers who were stable
over time: the ten (including Christine Q.) who maintained near-categorical [R], the
innovative variant, and the ten who maintained near-categorical [r], the traditional
Montreal variant. The two groups have very different social profiles: the group we
label ‘early adopters’ of the innovative form are much younger (just a little more than
half the age of the speakers who retain the traditional form) and have a higher mean
social-class index on the linguistic market scale. More striking is the very disproportionate
sex ratio: drawn from a pool of sixteen male and sixteen female speakers, eight of
ten early adopters are female; and seven of ten traditionalists are male, a statistically
significant difference.11 Gender difference is then a third variable, in addition to age
and social class, that is important in determining which speakers are early adopters of
the innovative form.
MEAN SEX
TRAJECTORY NO. OF MEAN AGE RATIO
RANGE TYPE RANGE OF [R] TYPE SPEAKERS LM 1971 FM
categorical or near- 85–100% stable 10 39.1 16.9 82
categorical [R]
categorical or near- 0–17% stable 10 29.0 30.3 37
categorical [r]
TABLE 11. The 20 stable speakers who remained categorical or near-categorical users of one of the two
variants across the study, 1971–1984.
The twelve remaining speakers were those whose behavior was more variable. Midrange
speakers are defined as those whose use of [R] fell between 17% and 84% at
either period. The mid-range speakers stand out in Figure 3, which plots the percentage
of [R] for each of the panel speakers in both 1971 (filled diamonds) and 1984 (hollow
squares). The overall increase of 12.3 percentage points is clearly not attributable to
most speakers having slightly increased their use of the innovative variant. Rather, the
increase is due to nine of the thirty-two individual speakers having made a fairly dra-
11 The significance level of p 0.05 was calculated by Fisher’s exact test.
572 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
matic shift between 1971 and 1984. The trajectories of these nine individuals are designated
by arrows connecting their 1971 with their 1984 data points.
FIGURE 3. Individual percentages of [R]/([R][r]) for the 32 panel speakers for 1971 and 1984. Trajectories
plotted for all speakers who showed a significant difference between the two years.
At the bottom of the graph, we see two groups of the 1971 speakers in the ellipses:
a tight cluster of seven individuals under age thirty, none of whom uses more than
10% [R], and another group of five speakers between the ages of thirty-five and fifty,
and who range between 0% and 17%. Ten of these twelve people form the stable group
of Table 11 in their use of the conservative form. Their 1984 values are displayed in
the two dotted ellipses to the right of each of the 1971 groups. Two of them, however,
Lysiane B. (007) and Alain L. (104), make substantial changes, abandoning the virtually
categorical conservative pattern they displayed in 1971. Along with the other midrange
speakers, their data appears individually in Table 12. Lysiane and Alain are
among the nine speakers whose trajectories appear in Fig. 3.
A majority of mid-range speakers in 1971 (seven out of ten) had moved to the
categorical or near-categorical use of innovative [R] by 1984. They are people we call
‘later adopters’ of the innovative variant. Five of the seven are young; five of the seven
are male. As a group, we would characterize their behavior as catching up with their
peers, the early adopters. The next three people listed in Table 12 are those mid-range
speakers who were stable, somewhat older, behaving more as we had expected older
individuals to behave. Of note is that one of them, Andre´ L. (065), a professional actor
age twenty-seven when we met him in 1971, was the only speaker who exhibited
stylistic variation (Sankoff & Blondeau 2008). In 1995 he maintained virtually the same
overall level of [R] as he had in 1984: 69%. At the bottom of the table are the two
individuals who moved from virtually categorical use of [r] to dominant use of [R],
occupying the upper part of the variable range. Lysiane B. (007) is a case of exceptional
upward social mobility, a twenty-four-year-old factory worker when we met her in
1971, a businesswoman in 1984, and when last interviewed in 1995, a successful realtor.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 573
AGE % [R] % [R]
SPEAKER TRAJECTORY TYPE 1971 1971 1984
Louis-Pierre R. (092) variable › categorical 16 64% 100%
Paul D. (064) variable › categorical 16 21% 95%
Lise L. (071) variable › categorical 17 73% 90%
Guy T. (088) variable › categorical 19 48% 100%
Christian B. (025) variable › categorical 27 68% 98%
Diane R. (066) variable › categorical 40 72% 97%
Joseph R. (075) variable › categorical 53 48% 79%
Andre´ L. (065) stably variable 27 61% n.s. 66%
Henriette N. (015) stably variable 38 77% n.s. 83%
Lucille R. (084) stably variable 56 83% n.s. 75%
Lysiane B. (007) categorical › variable 24 7% 65%
Alain L. (104) categorical › variable 45 17% 66%
TABLE 12. Percentages of [R] use for 12 speakers whose [R] use fell into the mid-range of
variation in either or both 1971 and 1984. Note that ‘categorical’ here refers to
categorical or near-categorical.
Although her own characterization of the major change in her speech over this period
is that she has learned English, it is clear that she has adopted many features characteristic
of middle class speakers during her adult life (Blondeau et al. 2002). As far as Alain
L. (104) is concerned, we were quite surprised that a man of forty-five would have
shown such a remarkable change by the time he was fifty-eight. A car salesman, Alain
was interviewed in 1984 at the car dealership where he worked, and much of the talk
focused on questions of language and of taste. To attribute his relatively high level of [R]
to formality, however, is probably too hasty. We have been unable to show stylistically
conditioned variation for any speaker other than the professional actor Andre´ L. (065),
who was part of the long-term longitudinal sample followed from 1971 through 1995
and who retained his upper mid-range value throughout (SBC, Sankoff & Blondeau
2008).
Table 12 and Fig. 3 show clearly that the mid-range use of both variants is not stable.
Not only do most speakers cluster at the top or the bottom of the range, but mid-range
speakers also tend very strongly to become categorical over time.
We return to a discussion of the matter of individual stability/instability and related
issues in the final section. To conclude our study of the panel results, we find that they
confirm our previous research on a smaller sample (SBC). Looked at in terms of the
critical age, the evidence from this panel is that twenty-three of the thirty-two speakers
(72%) showed stability over the period of their post-fifteen-year-old lives that we examined;
the other nine (28%) increased their use of the innovative form significantly.
Perhaps a more telling comparison is to look at the pool of speakers who had not yet
adopted the innovative form in 1971. Of the twenty-two speakers who might have
changed toward the innovative form, nine, or 41%, did so.
We might now ask: is 41% a lot, or a little, in terms of our expectations about
speakers’ malleability in adult life? Are the speakers who make this adult transition
the leaders of the change in the community? In the next section, in which we present
the results of the trend study, we put these questions into the community perspective.
6. CHANGE IN THE COMMUNITY: THE EVIDENCE OF TREND COMPARISONS. We noted in
§3.2 (comparing Tables 5 and 6) that the 1984 trend sample was quite similar to the
1971 trend sample in terms of the social characteristics of the speakers. In selecting
that subsample, however, we were not able to match the numbers of 1971 for the
574 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
youngest group, since only twelve new young speakers were interviewed in 1984, one
of whom was twenty-five years old. When distributing our trend sample of 1984 according
to the two age groups selected for analysis, there is an unequal number of ‘younger’
and ‘older’ speakers. For this reason, we pay most attention to the comparison of
younger to younger, and older to older, in tracing the trends between 1971 and 1984.
The figures for older and younger speakers reported in Table 9 are repeated in Table
13, along with the comparable results from the 1984 subsample.
TREND SUBSAMPLES AGE RANGE MEAN AGE MEAN LM % [R]
16 younger 1971 15–24 18.8 44.5 68.5%
11 younger 1984 15–24 17.9 60.9 92.5%
16 older 1971 25–61 39.2 32.6 34.3%
21 older 1984 25–61 38.6 40.3 74.9%
TABLE 13. Mean social characteristics and mean [R] %, 1971 and 1984 trend subsamples.
The eleven younger speakers who were added to the sample in 1984 were a good
match for their 1971 counterparts in age, but we note an increase in the mean LM
scores, at least some of which is attributable to overall socioeconomic change in the
community. By 1984, the economic recession of the early 1970s had turned around,
and more families could afford to keep their teenagers in school longer. Although one
would expect higher [R] scores for speakers of higher socioeconomic status, the jump
from 68% [R] in 1971 to 92.5% in 1984 among speakers under twenty appears to
represent a real-time change in the community. Perhaps even more significantly, the
twenty-five-to-sixty-one-year-old trend subsample in 1984, more closely matched for
socioeconomic status with their 1971 counterparts, shows an even greater increase: from
34.3% in 1971 to 74.9% in 1984. Adults of working age in the Montreal community in
1971 were quite likely to be predominant users of [r], but this was no longer the case
in 1984. These trend comparisons clearly indicate that change was vigorous in the
community in the 1970s and 1980s.
The trend study, representing a real-time reassessment of the community after a
thirteen-year interval, is the basis for assessing real-time change. As such, the community
mean indicates a massive shift toward the replacement of [r] by [R] during this
period. Although the mean increase of panel speakers seemed to indicate progress
toward the incoming form, the trend study confirms that at the community level, the
rate of change is much greater, as is evident from the comparisons in Table 14.
MEAN [R], YOUNGER SPEAKERS MEAN [R], OLDER SPEAKERS
Panel, same Two separate trend Panel, same Two separate trend
speakers samples speakers samples
1984 77.8% 92.5% 50.7% 74.9%
1971 63.8% 68.5% 41.3% 34.3%
mean increase 14.0 24.0 9.4 40.6
TABLE 14. Comparison of trend and panel subsamples over time.
Our examination of panel speakers as a group across time confirmed that such group
means may give a misleading picture of what is going on. The examination of individual
speakers led us to infer that the change is being implemented by people who alter their
pronunciation quite rapidly, rather than a steady, incremental raising of levels across
individual lifespans. An examination of the individual speakers in the two trend samples
further illuminates these initial inferences, as can be seen in Figure 4.
Figure 4 indicates that the real-time change reflected in the community means of
Table 13 is a product of a very different distribution of speakers in 1971 and 1984.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 575
FIGURE 4. Individual percentages of [R]/([R][r]) for the 32 trend speakers for 1971 and for the 32 different
trend speakers for 1984.
Seven speakers under age thirty-five in 1971 (filled diamonds inside the dotted rectangle
at the lower left) were predominant users of the conservative [r]; and only six were in
the near-categorical [R] range on the upper left. By 1984, only one speaker under age
thirty-five was a predominant user of the conservative form, and with one mid-range
exception, all of the other younger speakers were clustered within the near-categorical
[R] range. Among those over thirty-five, we see that in 1971, no speaker showed more
than 50% [R]; most were clustered at the bottom of the figure in the categorical [r]
range. In contrast, by 1984 there were six majority users of [R] among those older than
thirty-five.
The panel study showed that early adopters remained stable across the 1971–1984
period. Once having adopted innovative [R], they continued as categorical or nearcategorical
users of this form. By the time they got to be age thirty-five and older,
these early adopters were joined by a number of later adopters, who caught up with
the change in later life. But whereas many younger speakers in 1971 were still majority
users of the conservative form, this was no longer the case in 1984.
7. THE PROCESS OF CHANGE.
7.1. THE RESULTS OF OUR STUDY OF [r] N [R]. What generalizations can be drawn
from the detailed results reviewed in §§5 and 6, and how do they relate to our understanding
of language change? Let us begin by summarizing the major observations
drawn from the trend and panel studies, along with the results of C&C (1979) based
on the entire sample of 120 speakers from 1971.
THE SITUATION IN 1971. Three major generalizations emerge regarding the situation
in 1971. (i) In 1971, speaker age was a highly determinative factor. Of the 120 speakers,
C&C found that among the twenty-five oldest (those age fifty-eight and older in 1971),
twenty were categorical users of [r] and only one had more than 4% [R]. At the other
576 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
end of the spectrum, many speakers younger than thirty-five in 1971 were categorical
or near-categorical users of [R]. (ii) In both trend and panel samples, most speakers
tended to be near-categorical users of either INNOVATIVE [R] (with [R] values ranging
from 85% to 100%), or of CONSERVATIVE [r] (with [R] values ranging from 0% to 17%).
(iii) A minority of speakers exhibited variable behavior, and those variable speakers
tended to occupy the 50% to 80% [R] range.
THE SITUATION IN 1984. By 1984, a number of important changes had ensued. (i)
The real-time evidence from the trend samples shows a massive increase of [R] by
1984. Speakers under twenty-five in 1984 were twenty-five percentage points ahead
of their 1971 counterparts in their use of [R], and speakers between age twenty-five
and fifty-six in 1984 were forty-one percentage points ahead of their 1971 age mates.
(ii) The panel showed a more modest gain. As a group, younger panel speakers increased
their use of [R] by fourteen percentage points as compared with their earlier selves;
older speakers increased by nine percentage points. In terms of change across the
lifespan, panel members divided into three groups whose members behaved differently:
(a) Of the twelve majority-[r] speakers of the 1971 panel, ten were stable through 1984,
and two moved up dramatically to the mid-range, both registering over 60% [R] use.
(b) All ten speakers who registered 85% or more [R] in 1971 stayed within that range
in 1984. (c) Of the ten mid-range speakers of 1971, seven significantly increased their
use of [R], finding themselves in 1984 in the 85% to 100% near-categorical range,
whereas the three remaining mid-range speakers were stable across the thirteen-year
time span. (iii) Among 1984 trend speakers, only one under thirty-five years old used
less than 70% [R]; almost all of these younger speakers were virtually categorical users
of [R].
7.2. RESOLVED AND UNRESOLVED QUESTIONS. By the evidence reviewed in the preceding
section, two possible trajectories can be ruled out as underlying the implementation
of this change in the Montreal speech community: (i) the change does NOT proceed
mainly by speakers slowly increasing their use of [R] across their lifetimes, and (ii)
the change does NOT proceed by successive cohorts of new members entering the speech
community bearing a slightly higher value of [R] usage than their older peers. Rather,
the inevitable conclusion is that change in the community is a result of individual
speakers, especially younger speakers, being added to the pool of majority-users or
categorical users of [R].
What is the process by which new speakers get added to the pool? We saw that both
in 1971 and even more in 1984, many speakers in the fifteen-to-nineteen age range
were already categorical [R] users when they first entered our study. One important
question that we can try to answer on the basis of indirect evidence is: did these young
speakers acquire categorical [R] usage as part of their initial language acquisition as
young children, or did they join the pool of [R] users sometime between first language
acquisition in early childhood and the age of fifteen? Given the fact that almost everyone
over the age of thirty-five in 1971 was a categorical user of [r], we would have to
assume that input from parents of virtually all children in 1971 was categorical [r]. By
1984, the situation would have changed somewhat. Some adolescents born between
1965 and 1970 would have had parents born after 1945 who themselves were predominant
users of [R]. Nevertheless, it seems that the majority of our [R] users whose parents
were native Montrealers would have heard the [r] variant as their primary input in first
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 577
language acquisition. They would thus have differentiated themselves from their parents
in later childhood or early adolescence.
Indeed, we have some direct evidence on this point. Almost all of the interviews
with adolescents were carried out in the family home. Often a parent was present, and
often those parents participated to some extent in the interview, being called on by
their sons or daughters, for example, to supply information about where their grandparents
came from. In some of these cases, we were fortunate to be able to establish the
/r/ pattern of the parent. In sixteen of the thirty-one interviews carried out with young
people between fifteen and twenty (including all 1971 and 1984 interviews), parents
were present, usually the mother. Although we had much smaller samples of these
parents’ speech, we found no cases where the parent appeared to be anything but an
apical [r] user.
We must conclude that a child with two parents who are categorical users of [r]
would have [r] as an initial target form. It follows that in 1971 and to a large extent
also in 1984, most people who arrived at age fifteen or sixteen as categorical users of
[R] were speakers who had actually switched in childhood or early adolescence. It is
possible that in 1971, sixteen-year-old Louis-Pierre R. (092), with 64% [R], was
recorded in the late stages of such a switch, which was complete by 1984, when he
had 100% [R]. This might also be true of some of the other young speakers in Table
12, given in 1.
(1) Paul D. (064) age 16–29 21% N 95%
Lise L. (071) age 17–30 73% N 90%
Guy T. (088) age 19–32 48% N 100%
Christian B. (025) age 27–40 68% N 98%
The trajectories of these later adopters, whom we were able to observe, might simply
be paralleling a similar trajectory made earlier in life by their earlier-adopting peers
whose change we did not observe. Indeed it might also be correct to view the two older
speakers who registered similar increases as ‘very late’ adopters: Diane R. (066), age
40–53, with a 72% N 97% increase, and Joseph R. (075), age 53–66, with a 48% N
79% increase. That a change from virtually categorical [r] to strongly majority [R] over
a lifetime is possible is shown by the case of Lysiane B. (007), whom we were lucky
to be able to follow through 1995. Lysiane went from 7% to 65% to 78% [R] over the
twenty-four-year span from 1971 to 1995 (SBC).
The norm is clearly one of relative stability after adolescence, however, as we saw
from the fact that a majority of panel members did not change so radically. Our conclusion
is that most of those speakers observed to be categorical users of [R] in late
adolescence in fact made the switch earlier in life. Although it is possible that some
children were influenced by older siblings who had already adopted the innovative
form, the fact remains that there is a fairly sharp break between speakers older and
younger than thirty-five in the 1971 data. So while it is possible that some younger
siblings of early adopters may have acquired the innovation as part of their L1 acquisition,
the older siblings themselves must have gone through the switch process.
In considering the source of the innovation and how it spread initially, it is important
to look at both geographic and socio/stylistic considerations. C&C pointed out that
among the Montrealers with parents from a traditionally [R]-pronouncing area of Quebec,
those over forty showed no particular tendency to be early adopters of [R]; however,
they did observe such a tendency for younger speakers. It seems that once the change
578 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
was in progress in Montreal, it was acceptable and even appropriate for younger speakers
to retain the pronunciation of their parents in cases where the parents were from
[R]-pronouncing regions.12
It is clear that we are dealing here with a change from above (the change has a
source outside the community; early adopters are more likely to be female and of a
higher socioeconomic status). But just how aware are community members of these
competing forms of /r/, and what is their stylistic status? We did not carry out subjective
reaction tests at the time when the spontaneous speech was recorded, but there were
some speakers who mentioned regional variation in /r/ pronunciation during the part
of the interview when they were questioned about the differences between Montrealers
and people from elsewhere. In the excerpt in 2 from a 1971 interview, Denise, the wife
of interviewee Alain L. (104) and a Montrealer in her early forties at the time, was the
one who brought up /r/ pronunciation when the topic of ‘whether you can tell where
someone is from by the way they speak’ was introduced.
(2) Denise L.: Mes parents moi venaient du Lac St-Jean. Ils grasseyaient.
‘My parents came from Lac Saint-Jean. They used to pronounce
a ‘‘guttural’’ /r/.’
Moi-me^me quand j’aicommence´ a` aller a` l’e´cole puis—je faisais
rire de moi,
‘Me too when I started at school and—I got laughed at.’
fait-que je me suis dompte´. ‘Je vas arre^ter de rouler les r!’13
‘So I was crushed. ‘‘I’m going to stop rolling my r’s!’’ ’
Interviewer: Vous avez pris vite l’accent de Montre´al?
‘Did you pick up the Montreal accent quickly?’
Alain L. Et bien ma femme est ne´e a` Montre´al,
‘Yes well my wife was born in Montreal,’
mais ses parents et les plus vieux de sa famille sont ne´s au Lac
St-Jean.
‘But her parents and the older people in her family were born
in Lac Saint-Jean.’
Despite a few remarks of this sort that showed awareness of regional differences,
particularly on the part of people who had crossed the east-west (R/r) line in moving
from one area to another, stylistic variation among our 1971 speakers was marginal at
best (Sankoff & Blondeau 2008).
By 1984, the mid-range speakers we examined still did not behave consciously as
if the [R] variant held any particular cachet—for most of them, in the upper range of
variable usage, it had become the unmarked variant. By contrast, for some younger
speakers, [r] seemed to have taken on meanings of being old-fashioned, having the
12 In documenting consonantal changes in the mid-Rhine region of Germany over a thirty-year period,
Bellman and colleagues (1999) found a number of changes moving in the direction of Hochdeutsch, but the
towns where change was occurring were those that bordered on areas that already had these features. The
actual prestige of the Hochdeutsch forms would presumably have been uniform throughout the region, but
the effect of prestige was found only on the boundary, that is, with speakers who had contact with those
who already used the prestige form. In our sample, Yannick C. (126), a younger speaker (seventeen years
old in 1984), was typical of his generation with 100% [R]. One early influence on his speech, however, may
have been the fifteen months he spent with his family in Quebec City at the age of seven when he was in
first grade.
13 Most people contrast grasseye´ (dorsal) with roule´ (apical), but despite the apparent confusion of these
terms on the part of Denise, she clearly means she was under social pressure to stop using the dorsal variant.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 579
ring of one’s grandparents, of one’s teachers, of the clergy. Some younger speakers
used it for comic effect. In 3, Lucie A. (125), age twenty-two in 1984, is talking about
the kind of luxurious home she might like to have. She is very serious throughout
the passage, using her normal [R] in all of the /r/-words: pouvoir, superbes, pouvoir,
re´nover, regarde. Then she pulls herself up short from this reverie and the last three
lines are very ironic, when she contrasts her true situation with this dream of luxury.
In both peut-e^tre and trop, the word is drawn out and the apical [r] clearly signals a
comic effect in which she pokes fun at herself for having these visions of another way
of life.
(3) Lucie A., age 22 (1984)
. . . ils e´taient bien chanceux de pouvoir [R]
‘. . . they were very lucky to be able’
se payer des loyers comme c¸a la`,
‘to pay rents like that,’
puis—des maisons superbes [R] la`. J’aime—
‘and the superb homes. I like—’
j’aime c¸a, la—la la vie qu’il-y-a la`-dedans,
‘I like that, the—the lifestyle there,’
dans ces maisons la`. Ou bi en
‘in that kind of houses. Or else’
de pouvoir [R] acheter une maison
‘to be able to buy a house’
pui s de la re´nover [R] puis c¸a, c¸a—
‘and to renovate it, and that, that—’
En tout cas souvent je pense a` c¸a tu-sais
‘In any case I often think about that y’know’
puis je regarde [R] moicomment ce-que je vis
‘and I look at myself, how I live,’
puis je me dis ‘Oui.’
‘and I say to myself, ‘‘Yeah, right!’’ ’
Peut-e^tre [r] un jour! [R]
‘Maybe one day!’
Je sais pas trop [r] la`.
‘I don’t really know.’
A further mystery concerns the nature of the allophonic variation both between the
two allophones in competition, and between them and the other variants of /r/ (C&C,
Tousignant 1987a, SBC, Sankoff & Blondeau 2008). For variable speakers, there is a
strong tendency to retain [r] in onsets, the most salient position, whereas the innovative
[R] seems to appear first in codas in which /r/ is in any case subject to lenition and
deletion. It would seem that a high-prestige pronunciation ought to occur first in the
most salient position, but in listening to the speech of our variable speakers, it is almost
as if [R] creeps in through the back door, where the least attention is being paid.
Speakers who tend toward categoriality of either the conservative or the innovative
variant tend to have a higher proportion of [r] and [R] tokens overall. That is, these
speakers have many fewer cases of vocalized or deleted forms than do the intermediate
speakers. It is almost as if the intermediate speakers no longer have a clear consonantal
target for /r/. Further details on allophonic conditioning of the other /r/ variants can be
found in Sankoff & Blondeau 2008.
580 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
The issues of allophonic variation and of stylistic significance may help to shed
further light on the implementation of the change, but it is too late to be able to study
the situation of L1 acquisition in families where children ended up with a different
allophone from that of their parents. Nevertheless, we think that there are two important
facts that need to be underlined here: (i) apical [r] must have been the target and the
first-acquired form in L1 acquisition for the vast majority of teenagers and young adults
in Montreal in 1971 who were early adopters of dorsal [R], given the input they would
have had from their parents; and (ii) all of the mid-range speakers under age thirty in
1971 had become categorical or near-categorical users of [R] by 1984. In tracing their
trajectories as late adopters, it is reasonable to infer that their early-adopting peers had
traced the same trajectory earlier in their lives. Figure 5 sketches the trajectories we
envisage for the early adopters, after the pattern of those we observed.
FIGURE 5. Actual trajectories for all mid-range speakers under age 30 in 1971, and putative trajectories for
earlier stages.
7.3. MONTREAL [r] N [R] IN A WIDER PERSPECTIVE. How do our findings relate to
other cases of language change, and to clarifying the relative usefulness of trend and
panel studies in longitudinal analysis? Before drawing what we believe to be the general
implications of our study for the understanding of change in progress, we must point
out two respects in which the case we are dealing with here may differ from that of
other changes. We then conclude by stepping back from the current analysis and making
some generalizations we believe may be applicable to research in other communities.
THE DISCRETE CHARACTER OF MONTREAL /r/. Our findings indicate that this change
differs from others described in the literature in one important way: the relative lack
of stable variation. More speakers tended to be categorical than variable, and those
who changed did so very rapidly.
We noted in §§5 and 6 that (i) most speakers in both 1971 and 1984 were majority
users of one of the two variants, rather than occupying the mid portions of the range,
and that (ii) panel results revealed variable speakers as tending toward categoriality
over time. Such a pattern contrasts markedly with that of most studies of change and
variation in vowels, in which gradualism is the order of the day. Gradual changes in
frequency have been characteristic of postvocalic /r/ in New York City and New England,
where a former r-vocalization pattern is now variable, but where change has been
slow; age differences among subjects interviewed for the Atlas of North American
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 581
English (Labov et al. 2006) were not significant. There is, however, a striking parallel
to the Montreal case in the rapid diffusion of postvocalic /r/ in the American South,
where ‘r-pronunciation has swept through the region in all styles of speech so that
younger white speakers are consistently r-ful’ (Labov et al. 2006:47).14
Another aspect of our findings that differentiates this change from others that have
been studied is that the change in the phonetics of /r/ does not appear to interact with
other aspects of Montreal French phonology. As far as this second point is concerned,
we note the conclusion drawn by van Hout and Van de Velde about crosslinguistic
variation with respect to /r/: ‘Languages may often display almost all of the /r/ sounds
commonly mentioned. Transition from one type of /r/ into another is frequent, not unidirectional,
and irrelevant to the linguistic system’ (2001:3).
Continuing in this vein, they highlight the observation of Wiese (2001) that for
German and other Germanic languages, ‘the pattern of /r/ . . . shows that the phonotactics
is undisturbed by the segmental variations’ (van Hout & Van de Velde 2001:3).
Although we do not agree that variation in /r/ is always without systemic effects (see
for example Labov 1966 on vocalic mergers in New York City that are inhibited by
/r/ pronunciation), it seems to be true that in the Montreal French case at least, the
change in the phonetics of [r] does not have systemic phonological consequences.
Beginning with Labov 1963, many studies that have used the apparent-time construct
as an indicator of change in progress have been studies of vowels and vowel systems;
change is often systemic, with chain shifts and parallel shifts more common than isolated,
individual changes. It would seem that the more embedded any given change is
as part of a structural system, the less susceptible it might be to individual initiative
of the sort we saw on the part of speakers like Lysiane B. Preliminary work on the
diphthongization of the mid vowels /o/ (as in chose) and /eu/ (as in beurre) across
individual lifespans in Montreal (Sankoff et al. 2006) has shown a smaller degree of
change than is the case with /r/.15
THE IMPLICATIONS OF MONTREAL [r] N [R]. Our study has three major implications
that are relevant for other research on change in progress, and for the relationship of
change in the community to change across individual lifespans.
(i) Trend vs. panel comparisons in the assessment of change in progress. Our
study has provided clear evidence for an important methodological principle in longitudinal
research: in order to verify inferences about apparent time drawn from synchronic
studies, the most important tool is a carefully designed trend study. We saw that the
increases made by panel-study members were only a pale reflection of what was going
on in the wider community. In using [R] more frequently, older speakers are by no
means leading the change, although some are behaving somewhat like younger members
of the community.
14 This rapid change, inferred from a coefficient of 14 (probability 0.0001) for an age difference of
twenty-five years (Labov et al. 2006:48), means that a speaker age twenty-five with 100% postvocalic /r/
might have a grandparent age seventy-five who had only 72%.
15 In longitudinal research on Montreal French mid vowels and on the back /&/ i n -ation words, Kemp
and Yaeger-Dror (1991), Yaeger-Dror and Kemp (1992), and Yaeger-Dror (1989, 1994, 1996) have traced
developments in real time through the use of early-twentieth-century radio broadcasts as compared with
similar recordings from the same speakers twenty to forty years later. They also used data from the 1971
and 1984 Montreal corpora. Summarizing the results of panel comparisons involving eight speakers recorded
in 1971 and 1984, Yaeger-Dror states that speakers may continue ‘to advance their [vowel phonology] toward
a newer phonology well into middle age’ (1994:284–86).
582 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
(ii) Implications for the use of the apparent-time construct. In several studies of
longitudinal change reviewed in Labov 1994 (New York City, Panama, Norwich), the
results were similar to those presented here in that the age distribution at time B was
analyzable in terms both of community change since time A and of the component
referred to by Labov as age grading. Although the results reviewed by Labov were
based on trend, not panel, studies, this second component clearly has to do with change
over the lifespan. Labov’s conclusion comparing the restudies of New York City and
Panama City presages the results we found:
In each case, age-grading is the dominant configuration. In succeeding generations, speakers follow the
same pattern across apparent time. But there is also a steady increment of the process at a lower level,
showing that a real-time change is taking place. This suggests that in trying to decide which model is
correct for a given process, we may have been setting up a misleading opposition between age-grading
and generational change. It is possible that age-grading is involved in the mechanism of real-time change
for certain types of linguistic change. (Labov 1994:97)
Our findings, along with this reasoned opinion of Labov’s, motivate our proposal to
add a new category labeled LIFESPAN CHANGE (as proposed in Table 2) to the model of
the relationship between generational change and age grading. This concept differs
from age grading (which generally involves cyclicity), as discussed in §1.
In our research reported here, the trend study was invaluable in establishing both
the direction and the rate of change. However, in most cases where we are interested
in assessing change in progress, we do not have the luxury of a trend study. In such
cases, we want to know how much we can rely on apparent-time inferences. Here the
findings of the panel study must be considered. Insofar as the movement among panel
speakers was decidedly directional, with some older speakers going along with the
change, there is an important conclusion to be drawn about apparent-time inferencing.
If a significant minority of older speakers has actually gone along with the change in
adult life, then means derived from a cohort of older speakers actually overestimate
their rates of use of the variable as children. In an apparent-time calculation, we would
normally assume that if, for example, speakers age forty-five to fifty are using the
innovative variant at a rate of 50%, whereas fifteen-to-twenty year olds are using it at
a rate of 80%, then there has been an increase of thirty percentage points over the
thirty-year period, an average of ten percentage points per decade. But if indeed that
cohort of forty-five-to-fifty year olds have themselves inched up ten percentage points
as a group since their teens, then the true value of that cohort thirty years earlier
would not have been 50% (the rate they show today), but only 40%. The conclusion
is inescapable: to the extent that older speakers change in the direction of change in
progress in their adult lives, apparent time UNDERESTIMATES the rate of change. A major
implication of our study for other research that relies on inferences from systematic
synchronic-age distributions is that apparent-time calculations correctly identify the
direction of change, but may underestimate the rate of change, a conclusion also reached
by Boberg (2004) in a study of Montreal English.16
16 According to Boberg, ‘the behavior of older speakers in late adoption contributes to the speed with
which an innovative feature replaces an obsolescent feature in the community as a whole. In this sense, late
adoption accelerates rather than retards changes in progress’ (2004:258). Contrasting the two phonological
variables he studied with numerous lexical variables, Boberg’s judicious assessment was that late acquisition
by adults was ‘most characteristic of lexical variation’ (ibid, p. 250), rather than phonological variation as
we have seen here.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 583
We concur with Bailey et al. 1991, Bailey 2002, and Tillery & Bailey 2003 that
earlier worries about the validity of apparent time in identifying change in progress
have proven to be exaggerated.
(iii) Implications for the relevance of the critical period to language change.
Expecting relative stability in later life, we were surprised to see such a substantial
minority of speakers making major changes in their pronunciation. Of the twenty-two
speakers who might have changed, nine (about 40%) did so. The most stable group
was the categorical and near-categorical users of [r]: only two of the twelve departed
from this initial pattern. Seven of the ten mid-range speakers increased significantly,
however, moving to the categorical/near-categorical range of use of dorsal [R]. They
included all of those speakers under age thirty in the original study. This and other
evidence led us to conclude that most of the majority users of [R] in the 1970s and
1980s had changed their pronunciation at some time later than the period of initial L1
acquisition.
This evidence should not be interpreted as a blow against the reality of the critical
period, but as an indication of the necessity for greater attention to be focused on the
degree and kind of lability that occurs in later life. Relative stability of individual
speakers in panel studies has been documented by Brink and Lund (1975), Labov and
Auger (1998), and Nahkola and Saanilahti (2004). By contrast, Kemp and Yaeger-Dror
(1991), Baugh (1995), Naro and Scherre (2002), and Zilles (2005) have shown that
some individuals have made significant changes in later life. Guy and Boyd (1990)
presented a cogent scenario of how English speakers deepen their knowledge of the
morphology of tense marking across their lifespans: in the ‘semiweak’ verb class (kept,
told), children’s final t/d deletion patterns with that for monomorphemes; adults, however,
apparently deepen their analysis and treat the t/d of semiweak verbs like past
tense morphemes. And whereas comprehension of second-dialect features has been
shown to increase in adult life (Sankoff 1980), production of second-dialect features
is a piecemeal affair for both adolescents and adults, generally lacking the systematic
character of first language acquisition and subject to considerable interspeaker variability
(Chambers 1992, Bowie 2000, Kerswill&Williams 2000, Sankoff 2004). Meyerhoff
and Walker (2007) have demonstrated that long-term ‘urban sojourners’ from the Caribbean
island of Bequia have retained native grammatical constraints on copula variation
despite considerable exposure to other dialects as adults. It is likely that the adoption
of changes characteristic of younger speakers in one’s own community has many commonalities
with this process.
Our results require us to rethink the relationship between L1 and later language
acquisition, and between acquisition and change. Are late-adopting young adults simply
an epiphenomenon with respect to community change, following along behind the
younger leaders? Or do they have any influence on the children growing up after them?
In the case of a change from above like this one, the model of child innovators with
no adult input seems unlikely.17 Certainly the fact that adults use dorsal [R] at higher
frequencies contributes to the overall input from the community that children hear as
they form their phonological systems. Hay and colleagues (2006) show that in a change
in progress (the fear/fare merger in New Zealand), community members have quite
accurate perceptions of generational and social-class differences between speakers who
17 Such a scenario, with child innovators in the lead and adults trailing behind, seems clearly the correct
one in a case like the change from below involving the spread of the new verb of quotation, be like (Cukor-
Avila 2002, Tagliamonte & D’Arcy 2005).
584 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
do and do not produce the merger, and use this knowledge in differentially decoding
the same acoustic signal. And in a study of caregivers’ speech to children in a Scottish
community where a local phonetic variant of the diphthong (au) is in variation with a
standard variant, Smith and colleagues (2007) have shown that caregivers who use the
local variant almost exclusively in their general spontaneous speech nevertheless use
the standard at very high rates to young children, especially in contexts of teaching
and discipline.
(4) It’s r[u:]nd the wrong way, no turn it r[RA]nd!
(Caregiver to child in Buckie, Scotland; from Smith et al. 2007:70)
The youngest children replicate these very high rates, whereas slightly older children
use more of the local variant. Clearly the (au) variable is highly salient in Buckie, in
contrast to the situation of /r/ in Montreal; however, this study points to social differentiation
in usage patterns during L1 acquisition of a type that may well have an impact
on language change.
A further question involves how to model what is going on for speakers who show
a dramatic range in their trajectories over only thirteen years. Should we view these
people as having restructured their phonologies? Adjustment in the frequencies of two
variants speakers already possessed from childhood seems perhaps like a relatively
minor change. But several people replaced the unmarked, vernacular [r] that they appeared
to use spontaneously and unreflectingly with the innovative [R] characteristic of
speakers younger than themselves. As we demonstrate elsewhere (Sankoff & Blondeau
2008), these speakers have also altered their constraint hierarchies: majority users of
[R] seem to retain [r] only in onsets, which was not characteristic of their earlier
phonological systems as inferrable from their speech production. Pace Anderson and
Lightfoot (2002), our results lead us to believe that adults’ abstract phonological systems
at age thirty, or forty, or fifty are different from what they were in childhood or at
puberty.18 Change is dramatic in childhood, but it does not stop there. To understand
the dynamics of change in the speech community, we must follow language across the
lifespan.
REFERENCES
ANDERSON, STEPHEN, and DAVID LIGHTFOOT. 2002. The language organ: Linguistics as
cognitive physiology. Cambridge: Cambridge University Press.
ASHBY, WILLIAM J. 2001. Un nouveau regard sur la chute du ne en franc¸ais parle´ tourangeau:
S’agit-il d’un changement en cours? French Language Studies 11.1–22.
BAILEY, GUY. 2002. Real and apparent time. The handbook of language variation and
change, ed. by Jack K. Chambers, Peter Trudgill, and Natalie Schilling-Estes, 312–32.
Oxford: Blackwell.
BAILEY, GUY; TOMWIKLE; JAN TILLERY; and LORI SAND. 1991. The apparent time construct.
Language Variation and Change 3.241–64.
BAUGH, JOHN. 1995. Dimensions of a theory of econolinguistics. Towards a social science
of language: Papers in honor of William Labov. Vol. 1: Variation and change in
language and society, ed. by Gregory R. Guy, Crawford Feagin, Deborah Schiffrin,
and John Baugh, 397–419. Amsterdam: John Benjamins.
BELLMANN, GU¨ NTER; JOACHIM HERRGEN; and JU¨ RGEN ERICH SCHMIDT. 1999. Mittelrheinischer
Sprachatlas, Band 4: Konsonantismus (Dialektalita¨t. Konsonanten des westgermanischen
Bezugssystems. Sproßkonsonanten). Tu¨bingen: Niemeyer.
BLAKE RENEE, and MEREDITH JOSEY. 2003. The /ay/ diphthong in a Martha’s Vineyard
community: What can we say 40 years after Labov? Language in Society 32.451–85.
18 Adult phonology is probably also influenced by literacy. See for example Labov 1964 and Treiman et
al. 2002.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 585
BLONDEAU, HE´ LE`NE. 2001. Real-time changes in the paradigm of personal pronouns in
Montreal French. Journal of Sociolinguistics 5.453–74.
BLONDEAU, HE´ LE`NE. 2006. Panel studies and language variation. Encyclopedia of languages
and linguistics, 2nd edn., vol. 9, ed. by Keith Brown, 150–54. Oxford: Elsevier.
BLONDEAU, HE´ LE`NE; GILLIAN SANKOFF; and ANNE CHARITY. 2002. Parcours individuels et
changements linguistiques en cours dans la communaute´ francophone montre´alaise.
Revue que´be´coise de linguistique 31.13–38.
BOBERG, CHARLES. 2004. Real and apparent time in language change: Late adoption of
changes in Montreal English. American Speech 79.250–69.
BOWIE, DAVID. 2000. The effect of geographic mobility on the retention of a local dialect.
Philadelphia: University of Pennsylvania dissertation.
BRINK, LARS, and JO/RN LUND. 1975. Dansk Rigsma°l I–II: Lydudviklingen siden 1840 med
s+rligt henblink pa° sociolekterne i Kobenhavn. [Standard Danish I–II: The phonetic
development since 1840 with special regard to the sociolects in Copenhagen.] Copenhagen:
Gyldendal.
BRINK, LARS, and JO/RN LUND. 1979. Social factors in the sound changes of modern Danish.
Proceedings of the ninth International Congress of Phonetic Sciences, vol. 2, 196–203.
Copenhagen: University of Copenhagen.
CEDERGREN, HENRIETTA. 1988. The spread of language change: Verifying inferences of
linguistic diffusion. Language spread and language policy: Issues, implications, and
case studies (Georgetown University Round Table on Languages and Linguistics 1987),
ed. by Peter H. Lowenberg, 45–60. Washington, DC: Georgetown University Press.
CHAMBERS, JACK K. 1992. Dialect acquisition. Language 68.673–706.
CHARBONNEAU, RENE´ . 1955. Recherches d’une norme phone´tique dans la re´gion de Montre´al.
E´
tudes sur le parler franc¸ais au Canada, 83–98. Sainte-Foy, QC: Presses de l’Universite
´ Laval.
CLERMONT, JEAN, and HENRIETTA CEDERGREN. 1979. Les ‘R’ de ma mere sont perdus dans
l’air. Le franc¸ais parle´: Etudes sociolinguistiques, ed. by Pierrette Thibault, 13–28.
Edmonton, AB: Linguistic Research.
CUKOR-AVILA, PATRICIA. 2002. She say, she go, she be like: Verbs of quotation over time
in African American Vernacular English. American Speech 77.3–31.
DE PAIVA, MARIA, and MARIA EUGENIA DUARTE. 2003. Mudanc¸a lingu¨i´stica em tempo real.
Rio de Janeiro: Capa.
GUY, GREGORY, and SALLY BOYD. 1990. The development of a morphological class. Language
Variation and Change 2.1–18.
HAY, JENNIFER; PAULWARREN; and KATIEDRAGER. 2006. Factors influencing speech perception
in the context of a merger-in-progress. Journal of Phonetics 34.458–84.
HERNANDEZ-CAMPOY, JUAN MANUEL. 2003. Complementary approaches to the diffusion of
standard features in a local community. Social dialectology: In honour of Peter Trudgill,
ed. by David Britain and Jenny Cheshire, 23–37. Amsterdam: John Benjamins.
HUBBELL, ALLAN FORBES. 1950. The pronunciation of English in New York City: Consonants
and vowels. New York: King’s Crown Press.
JOSEY, MEREDITH. 2004. A sociolinguistic study of phonetic variation and change on the
island of Martha’s Vineyard. New York: New York University dissertation.
KEMP, WILLIAM, and MALCAH YAEGER-DROR. 1991. Changing realizations of A in (a)tion
in relation to the front A-back A opposition in Quebec French. New ways of analyzing
sound change, ed. by Penelope Eckert, 127–84. New York: Academic Press.
KERSWILL, PAUL, and ANN WILLIAMS. 2000. Creating a new town koine: Children and
language change in Milton Keynes. Language in Society 29.65–115.
LABOV, WILLIAM. 1963. The social motivation of a sound change. Word 19.273–309. [Revised
as Ch. 1 in Sociolinguistic patterns. Philadelphia: University of Pennsylvania
Press, 1973.]
LABOV, WILLIAM. 1964. Stages in the acquisition of standard English. Social dialects
and language learning, ed. by Roger Shuy, 77–103. Champaign, IL: National Council
of Teachers of English. [Reprinted in Readings in Amerian dialectology, ed. by
586 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
H. B. Allen and Gary Underwood, 473–98. New York: Appleton-Century-Crofts,
1971].
LABOV, WILLIAM. 1966. The social stratification of English in New York City. Washington,
DC: Center for Applied Linguistics. [2nd edn., Cambridge: Cambridge University Press,
2006.]
LABOV, WILLIAM. 1975. On the use of the present to explain the past. Proceedings of the
11th International Congress of Linguists, ed. by Luigi Heilmann, 825–51. Bologna:
Il Mulino. [Reprinted in Linguistics at the crossroads, ed. by Adam Makkai, Valerie
Becker Makkai, and Luigi Heilmann, 226–61. Padova: Liviana, 1977.]
LABOV, WILLIAM. 1994. Principles of linguistic change, vol. 1: Internal factors. Oxford:
Blackwell.
LABOV, WILLIAM; SHARON ASH; and CHARLES BOBERG. 2006. Atlas of North American
English. Berlin: Mouton de Gruyter.
LABOV, WILLIAM, and JULIE AUGER. 1998. The effect of normal aging on discourse: A
sociolinguistic approach. Narrative discourse in neurologically impaired and normal
aging adults, ed. by Hiram Brownell and Yves Joanette, 115–34. San Diego, CA:
Singular.
LIGHTFOOT, DAVID W. 1999. The development of language: Acquisition, change, and evolution.
Oxford: Blackwell.
MEYERHOFF,MIRIAM, and JAMES A.WALKER. 2007. The persistence of variation in individual
grammars: Copula absence in ‘urban sojourners’ and their stay-at-home peers, Bequia
(St Vincent and the Grenadines). Journal of Sociolinguistics 11.346–66.
MORIN, YVES-CHARLES. 2002. Les premiers immigrants et la prononciation du franc¸ais au
Que´bec. Revue que´be´coise de linguistique 31.39–78.
NAHKOLA, KARI, and MARIA SAANILAHTI. 2004. Mapping language changes in real time: A
panel study on Finnish. Language Variation and Change 16.75–92.
NARO, ANTHONY, and MARIA MARTA PEREIRA SCHERRE. 2002. The individual and the community
in real-time linguistic change: Social dimensions. Paper presented at New Ways
of Analyzing Variation (NWAV) 31, Stanford, CA, October 13.
PAUNONEN, HEIKKI. 1996. Language change in apparent time and in real time. Samspel &
Variation: Sprakliga studier tilla¨gnade Bengt Nordberg pa 60-arsdagen, 375–86. Uppsala:
Uppsala Universitet, Institut fo¨r Nordiska Sprak.
RIVARD, ADJUTOR. 1928 [1901]. Manuel de la parole. 2nd edn. Quebec: Garneau.
SANKOFF, DAVID, and SUZANNE LABERGE. 1978. The linguistic market and the statistical
explanation of variability. Linguistic variation: Models and methods, ed. by David
Sankoff, 239–50. New York: Academic Press.
SANKOFF, DAVID, and GILLIAN SANKOFF. 1973. Sample survey methods and computerassisted
analysis in the study of grammatical variation. Canadian languages in their
social context, ed. by Regna Darnell, 7–64. Edmonton, AB: Linguistic Research.
SANKOFF, GILLIAN. 1980. Mutual intelligibility, bilingualism and linguistic boundaries. The
social life of language, by Gillian Sankoff, 133–41. Philadelphia: University of Pennsylvania
Press.
SANKOFF, GILLIAN. 2004. Adolescents, young adults and the critical period: Two case studies
from ‘Seven Up’. Sociolinguistic variation: Critical reflections, ed. by Carmen Fought,
121–39. Oxford: Oxford University Press.
SANKOFF, GILLIAN. 2005. Cross-sectional and longitudinal studies in sociolinguistics. An
international handbook of the science of language and society, vol. 2, ed. by Ulrich
Ammon, Norbert Dittmar, Klaus J. Mattheier, and Peter Trudgill, 1003–13. Berlin:
Mouton de Gruyter.
SANKOFF, GILLIAN. 2006. Age: Apparent time and real time. The encyclopedia of language
and linguistics, 2nd edn., vol. 1, ed. by Keith Brown, 110–16. Oxford: Elsevier.
SANKOFF, GILLIAN, and HE´ LE`NE BLONDEAU. 2008. Instability of the [r] [R] alternation in
Montreal French: The conditioning of a sound change in progress. ‘r-atics 2 (Special
issue of E´ tudes & Travaux), ed. by Hans Van de Velde and Roeland van Hout, to
appear.
SANKOFF, GILLIAN; HE´ LE`NE BLONDEAU; and ANNE CHARITY. 2001. Individual roles in a realtime
change: Montreal (r N R) 1947–1995. In Van de Velde & van Hout, 141–57.
LANGUAGE CHANGE ACROSS THE LIFESPAN: /r/ IN MONTREAL FRENCH 587
SANKOFF, GILLIAN; MICHAEL FRIESNER; DAMIEN HALL; and HE´ LE`NE BLONDEAU. 2006. Le
facteur temps et le changement phone´tique dans le franc¸ais montre´alais. Paper presented
at Les franc¸ais d’ici, Queen’s University, Kingston, ON, June 5.
SANTERRE, LAURENT. 1978. Les /R/ montre´alais en re´gression rapide. Prote´e 2.117–30.
SMITH, JENNIFER; MERCEDES DURHAM; and LIANE FORTUNE. 2007. ‘Mam, my trousers is
fa’in doon!’: Community, caregiver, and child in the acquisition of variation in a Scottish
dialect. Language Variation and Change 19.63–99.
TAGLIAMONTE, SALI, and ALEX D’ARCY. 2005. He’s like, she’s like: The quotative system
in Canadian youth. Journal of Sociolinguistics 8.493–514.
THIBAULT, PIERRETTE, and MICHELLE DAVELUY. 1989. Quelques traces du passage du temps
dans le parler des Montre´alais, 1971–1984. Language Variation and Change 1.19–45.
THIBAULT, PIERRETTE, and DIANE VINCENT. 1990. Un corpus de franc¸ais parle´. Montre´al
84: Historique, me´thodes et perspectives de recherche. Sainte-Foy, QC: De´partement
de langues et linguistique, Universite´ Laval.
TILLERY, JAN, and GUY BAILEY. 2003. Approaches to real time in dialectology and sociolinguistics.
World Englishes 22.351–65.
TOUSIGNANT, CLAUDE. 1987a. La variation sociolinguistique: Mode`le que´be´cois et me´thode
d’analyse. Montreal: Presses de l’Universite´ du Que´bec.
TOUSIGNANT, CLAUDE. 1987b. Relations sociales entre le /r/ ve´laire et le /r/ re´troflexe anglais
de Montre´al. Revue que´be´coise de linguistique the´orique et applique´e 6.39–55.
TOUSIGNANT, CLAUDE. 1987c. Les variantes du /R/ montre´alais: Contextes phonologiques
favorisant leur apparition. Revue que´be´coise de linguistique the´orique et applique´e
6.73–113.
TREIMAN, REBECCA; JUDY BOWEY; and DERRICK BOURASSA. 2002. Segmentation of spoken
words into syllables by English-speaking children as compared to adults. Journal of
Experimental Child Psychology 83.213–38.
TRUDGILL, PETER. 1988. Norwich revisited: Recent linguistic changes in an English urban
dialect. English World Wide 9.33–49.
VAN DE VELDE, HANS, and ROELAND VAN HOUT (eds.) 2001. ‘r-atics: Sociolinguistic, phonetic
and phonological characteristics of /r/ (Special issue of E´ tudes & Travaux 4).
VAN HOUT, ROELAND, and HANS VAN DE VELDE. 2001. Patterns of /r/ variation. In Van de
Velde & van Hout, 1–9.
VERREAULT, CLAUDE, and THOMAS LAVOIE. 1999. ‘La langue de nos gens’ du pe`re Laurent
Tremblay: Une premie`re synthe`se sur la variation ge´olinguistique du franc¸ais parle´ au
Que´bec au de´but des anne´es 1940. Langues et Linguistique 25.145–212.
VINAY, JEAN-PAUL. 1950. Bout de la langue ou fond de la gorge? French Review 23.489–98.
VINCENT, DIANE; MARTY LAFOREST; and GUYLAINE MARTEL. 1995. Le corpus de Montre´al
1995: Adaptation de la me´thode d’enque^te sociolinguistique pour l’analyse conversationnelle.
Dialangue 6.29–46.
WEINREICH, URIEL; WILLIAM LABOV; and MARVIN HERZOG. 1968. Empirical foundations for
a theory of language change. Directions for historical linguistics, ed. by Winfred Lehmann
and Yakov Malkiel, 97–195. Austin: University of Texas Press.
WIESE, RICHARD. 2001. The unity and variation of (German) /r/. In Van de Velde & Van
Hout, 11–26.
YAEGER-DROR, MALCAH. 1989. Real time vs. apparent time change in Montreal French.
York Papers in Linguistics 13.141–53.
YAEGER-DROR, MALCAH. 1994. Sound change in Montreal French. Phonological structure
and phonetic form: Papers in laboratory phonology 3, ed. by Patricia Keating, 267–92.
Cambridge: Cambridge University Press.
YAEGER-DROR, MALCAH. 1996. Phonetic evidence for the evolution of lexical classes: The
case of a Montreal French vowel shift. Towards a social science of language: Papers
in honor of William Labov. Vol. 2: The linguistic structure of variation and change,
ed. by Gregory R. Guy, Crawford Feagin, John Baugh, and Deborah Schiffrin, 263–87.
Amsterdam: John Benjamins.
YAEGER-DROR, MALCAH, and WILLIAM KEMP. 1992. Lexical classes in Montreal French:
The case of (.:). Language and Speech 35.251–93.
588 LANGUAGE, VOLUME 83, NUMBER 3 (2007)
ZILLES, ANA MARIA S. 2005. The development of a new pronoun: The linguistic and social
embedding of a gente in Brazilian Portuguese. Language Variation and Change
17.19–53.
Sankoff [Received 10 August 2006;
Department of Linguistics accepted 8 February 2007]
University of Pennsylvania
619 Williams Hall
Philadelphia, PA 19104-6305
[gsankoff@earthlink.net]
Blondeau
Department of Romance Languages and Literatures
University of Florida
170 Dauer Hall
P.O. Box 117405
Gainesville, FL 32611-7405
[blondeau@ufl.edu]
PLACE THIS ORDER OR A SIMILAR ORDER WITH US TODAY AND GET AN AMAZING DISCOUNT 🙂
