The hidden dimensions of a change from below: Consequence markers in Montreal French

Abstract This article examines the rise of vernacular consequence marker ça-fait-que (CFQ), often realized as [fɛk] or [fak], at the expense of its standard counterparts donc and alors in Montreal French. The apparent-time analysis is based on a 2012 corpus of semi-directed interviews collected in Montreal. Previous studies treated the CFQ/donc/alors alternation as a purely lexical sociolinguistic variable. Our analysis shows how a vernacular variant (CFQ), initially associated with the working class and stigmatized, comes to compete, develop as a default form, and eventually crowd out forms at the other end of the social prestige scale (alors and donc). We rely on new socio-phonetic considerations to unveil a reconfiguration of the variable. The integration of the sociophonetic dimension sheds light on a complex process of diffusion, where a change from below is propelled by an additional change, but from above. Our article shows the key role played by women in both changes.


INTRODUCTION
Thanks to the interrelated sociolinguistic corpora collected since the 1970s, apparenttime and real-time research on the variety of French spoken in Montreal has been fruitful in helping us to understand linguistic change, in particular the relationship between apparent-time change and real-time change (Sankoff 2018). As we will see below, the Montreal corpora are also well suited to deepening our understanding of the complex relationship between changes from below and changes from above, which are often seen as independent phenomena. This article builds on previous research, and reports new results on language variation in Montreal French, providing further evidence of change in the use of the consequence markers ça fait que (CFQ), often reduced and realized as [fak], and its standard counterparts alors and donc. This linguistic variable, which was conditioned by social factors such as age and socio-economic status in the 1970s (Dessureault-Dober 1974), is carefully re-examined through an apparent-time, quantitative analysis of the variation in a corpus collected in the 2010s in the sociallydiverse Hochelaga-Maisonneuve neighbourhood of Montreal (Blondeau et al. 2012).
In this article, we expand the analysis of the lexical variation of CFQ/donc/alors with the aim of better understanding the relationship between the two standard variants donc and alors, associated with a more formal level of language (Dessureault-Dober 1974), and their vernacular counterpart CFQ, typical of spoken Quebec French. We deepen our understanding of the dynamics of the change by discussing the tension between change from above and change from below. One of our objectives is to explain how the vernacular variant CFQ comes to compete, develop as a default form, and eventually crowd out forms at the other end of the social prestige scale. To do so, we use new insights to unveil the reconfiguration of the variable.
Our analysis integrates a new dimension related to the socio-phonetic realization of the markers with the aim of providing a better explanation of the diffusion of CFQ at the community level and the social ramifications of this change. This phonetic aspect has never been systematically and quantitatively analyzed from a variationist perspective. We show that the success of the lexical variant CFQa change from belowis intertwined with a hidden change from above, involving a vowel alternation, namely the replacement of the stigmatized variant [fak] by the more neutral variant [fɛk].
A discussion of two longstanding issues in sociolinguistics is followed by a brief overview of previous variationist studies on consequence markers. The methodological section focusses on data collection, the description of the stratified sample, and circumscription of the variable context. The results are presented in two separate sections detailing apparent-time analyses. The first section focuses on the social configuration of the three lexical variants in competition; the second examines the sociophonetic aspects of variation. The discussion section sheds new light on the explanation of linguistic change affecting the speech community by considering socio-phonetics in the general interpretation of the variation observed. The social conditioning of the diverse phonetic realizations of CFQ allows us to revisit the canonical tripartite model of the variable.

TWO LONGSTANDING PUZZLES IN SOCIOLINGUISTICS
This section addresses two longstanding issues in sociolinguistics, namely the relation between change from above and change from below, and the gender paradox.

Change from above and change from below
Starting with Labov's pioneering work (1972), a longstanding line of research in sociolinguistics distinguishes between two types of linguistic change: change from above, and change from below. On the one hand, changes from above are characterized by a conscious process and involve the diffusion of incoming prestigious forms. The incoming forms, generally associated with the higher levels of the society, are often promoted in the community through the speech of middle-class speakers, who often use these linguistic forms to differentiate themselves from lower-class speakers. When other social classes adopt these forms, the change diffuses within the community. On the other hand, changes from below are generally characterized by a process that takes place at the subconscious level. These changes enter the language primarily through the vernacular and spread throughout the community without speakers' awareness. However, when changes from below rise to the level of awareness, they are frequently stigmatized and rejected by the higher classes (Labov 2001). At least two questions are thus raised: what makes a change from below rise to the level of awareness and how do some variants become stigmatized?
While changes from above refer to changes resulting from conscious social pressure, changes from below are initially subconscious and therefore more difficult to perceive. Sociolinguists seek to connect the cognitive dimension of linguistic awareness (or saliency) to inter-individual social factors (social origin, age, gender, ethnicity) or intra-individual factors (style, situation, speaker agency).
The definition and the criteria used to distinguish the two conceptschange from below and change from aboveremain subject to interpretation. As defined, the two types of change seem independent from each other (conscious/salient/standard/higher class vs. unconscious/non-salient/vernacular/lower class), and generally are not expected to jointly contribute towards a single outcome. In this article, we revisit this binary division between change from above and change from below. We consider a well-documented change from below (the rise of CFQ), and use apparent-time data to unveil a hidden change-from-above dimension to this variable.

The gender paradox
Another longstanding issue in sociolinguistics is the role played by gender in language variation and change. In sociolinguistic theory, the concept of gender is viewed as a social construct. The fact that gender differentiation of speech often plays a powerful role in linguistic change and that this social differentiation relates to sociointeractional patterns in everyday life is one of the main generalizations discussed by Labov in his early work on the social setting of language variation and change, and this pattern has remained relevant for many speech communities (Labov 1972(Labov , 2001Meyerhoff 2006). While the original observations concerning the gender paradox (Labov 1990) continue to be an important object of inquiry in the study of linguistic variation and change (Eckert 2000), the interpretation of the role of gender, and its relation to other social factors and to social meaning have evolved. Milroy and Milroy (1993) discussed how closely gender is related to class and social network, and its role in patterns of change. In addition, further development in sociolinguistics and gender studies in linguistics have led to a critique of the traditional methodologies on gender and language (Cheshire 2002), and of the binary approach. Furthermore, according to Buchholz (2002: 33), all sociolinguists must understand gender 'not as a variable that transcends particular situations but as a complex and context-specific system for producing identities and ideologies'.
In many speech communities, it has been observed that for stable sociolinguistic variables, women use the prestige variants at a higher rate than men do overall. Eckert's interpretation (2000) suggests that symbolic resources are particularly useful for women to position themselves within or in opposition to a group. When linguistic variables are involved in a change in progress (whether a change from below or from above), women often lead the change and act as innovators.
In cases of change in progress involving an incoming variant that is positively evaluated, women tend to use the innovative variant. Such results have been shown by Lennig (1978) for the modern Paris vowel system in which speakers from the upper middle class initiated the change, followed by the lower middle class and then the working class. In this change in progress, women were about a generation ahead of the men. Since changes from above involve a degree of consciousness, speakers are aware of the prestige associated with a variant and are 25 consequently prone to hypercorrection. The role of women in change from above is in line with the situation for stable variation (Labov 2001): in both cases, women favour the more prestigious/formal variant.
However, women also play a crucial role in change from below, as they seem to lead men in the use of incoming non-standard variants that are below the level of counsciousness. Changes from below challenge societal norms; women, especially those who are socially entrenched and involved in their community, often lead this linguistic change.
The differential role of gender clearly appears when the stages of a particular change are taken into consideration. It has been observed that 'the mechanism of the change crucially involved the initiating role of women at the outset, and the later adoption of the change by men.' (Labov 2001: 283). However, is the leading role of women confirmed in the case of a non-binary variable involved in a series of changes? Is it also confirmed when both types of changefrom above and from beloware intertwined?

CONSEQUENCE MARKERS IN LAURENTIAN FRENCH
Previous studies of Quebec French and other varieties of Laurentian French 1 have considered consequence markers from the complementary perspectives of semantics, pragmatics and sociolinguistics. While some studies have provided a comparison of their semantic values and a description of their pragmatic functions (Léard 1983, Dostie 2006see Blondeau et al. 2019 for an extensive review), other studies have examined consequence markers from the perspective of variationist sociolinguistics by treating CFQ, alors and donc as competing variants of a single linguistic variable.
While the forms alors and donc are used in Standard French (both written and spoken), CFQ is a vernacular form (Dessureault-Dober 1974). Some studies have mentioned that CFQ [sa fɛ kə] can be reduced to [fɛk] (Dessureault-Dober 1974) and may be phonetically realized as [fak] (Léard 1983), following a lowering process in word-final position. According to Dumas (1978), this phenomenon is restricted to a number of lexical items, such as fait. Because the form with lowering has been lexicalized (Dessureault-Dober 1974), it has been maintained as such in the community, despite the fact that this type of vowel lowering is in decline and characteristic of older speakers from working-class backgrounds expressing themselves in informal communication situations (Deshaies-Lafontaine 1974, Dumas 1978. Dessureault-Dober (1974), the first sociolinguistic study to examine consequence markers, was based on a sub-sample of 24 speakers from the 1971 Sankoff-Cedergren corpus (Sankoff et al. 1976). This variationist analysis showed that the variable was socially conditioned: the CFQ variant was associated with the lower end of the socio-economic ladder, male speakers and youth. Adopting an apparent-time perspective, Dessureault-Dober interpreted this regular age distribution as a change in progress in favour of CFQ at the community level. In this analysis, alors was considered a declining form associated with older speakers and social prestige. Note that in Dessureault-Dober's sample, the variant donc was too rare to be included in the quantitative analysis.
Thirteen years later, the Montreal 1984 corpus (Thibault and Vincent 1990) provided the opportunity to follow changing sociolinguistic configurations over time. In their study of discourse markers, Thibault and Daveluy (1989) re-examined the trajectory of alors. Their study, based on the entirety of the 1971 (120 speakers) and 1984 corpora (72 speakers), showed an increase in the use of alors, which they interpreted as an age-grading phenomenon. Although they focused on only one form and thus did not take into account all three variants, their results appeared to refute Dessureault-Dober's initial hypothesis of a declining alors at the community level.
Recent analyses (Blondeau et al. 2019, Martineau 2019) reexamined the situation in two interrelated varieties of Laurentian French based on two sub-samples of the 2012 FRAN corpus, providing new insights on this variable. These analyses of the Montreal data reintegrated donc into the model, confirming the diffusion of CFQ and the sharp decline of alors, and thus challenged the age-grading interpretation (Thibault and Daveluy 1989). Furthermore, the results from Blondeau et al.'s 2019 study showed a surprising gender asymmetry. While the strong increase in the use of CFQ appeared to be a change initiated by women, the replacement of alors with donc was a change initiated by men from a high or intermediate socioeconomic background. Additionally, this analysis demonstrated that CFQ had continued to develop as the default variant. These results point toward a re-shaping of the role and usage of CFQ by youth: social variation was observed for young male speakers, but not among young women. Villeneuve et al. (2019) shed light the effect of style in the variation by looking at two factors: the formality of the interview, and the pronominal forms of address (vous vs tu). Comparing two sets of television interviews, they showed that CFQ was more frequent in the less formal interviews and among tu users. As for donc and alors, they were more frequent in the more formal interviews. They also established a relationship between a more frequent use of alors and the vous users, which would indicate that alors would be more formal than donc, a more 'neutral' variant. Such interpretation regarding the stylistic status of alors contrasts with previous claims that donc is considerably more formal than alors (Rehner and Mougeon 2003, Mougeon et al. 27 BLONDEAU AND TREMBLAY 2009) and questions the interpretation that the stylistic formality of alors in spoken French is not strong enough for this form to withstand the vigorous rise of CFQ (Martineau 2019). Under Martineau's analysis, alors would be less formal than donc and rather an equivalent of CFQ. Finally, in a real-time study over more than 40 years, (Blondeau, et al., 2021a) confirmed the obsolescence of alors and its replacement by donc as a form of social prestige. Based on the entire corpus of 120 speakers from 1971, the analysis demonstrated that despite its rarity, donc was already part of the variable context in 1971 and that this incoming form had also already begun to develop at the expense of alors. It therefore seems that, despite their more or less formal status, alors and donc have been competing for a long time on the scale of social prestige, which could partially explain the variation observed by Villeneuve et al. (2019).
This linguistic variable has also been the subject of attention in other varieties of Laurentian French, namely in Ontario, where the contact with English is more intense, and so, a borrowing from English, is part of the variable context. In Ontario, so is a strong contender and competes with CFQ, while alors and donc compete as standard variants (Mougeon and Beniak 1991, Rehner and Mougeon 2003, Mougeon et al. 2009). Moreover, Bigot and Papen (2021) sheds new light on the variable by analyzing the effect of identity. For communities of Laurentian French differentially affected by language contact, linguistic practices and norms regarding consequence markers need to be nuanced according to the social and ethnocultural identity of the speakers.
To summarize, research over the last forty years has shown the dynamics of consequence markers in Canada. In all communities, CFQ is associated with informal speech, and donc and alors with more formal speech. In Montreal, we observe the replacement of alors, which is becoming obsolete, by both donc and CFQ. These Montreal results are surprising for a number of reasons. First, how can we explain the rise of vernacular CFQ at the expense of standard alors? Second, while the rise of donc is a change from above led by men, the rise of CFQ is a change from below led by women. Given the traditional association of women with standard speech, the leading role of men in the diffusion of donc is unexpected. Finally, two changes appear to compete within the variable: a change from below (the rise of CFQ) and a change from above (the rise of donc). Why do men and women play distinct roles in the diffusion of these two changes? Can this be attributed to the tripartite nature of the variable itself, or does it reflect new gender roles in language change?

METHODOLOGY
This methodological section provides a description of the corpus and the variable context, as well as information regarding exclusions, coding, and analyses.

The Hochelaga-Maisonneuve sub-corpus of the corpus FRAN
The current analysis is based on speech data from sociolinguistic interviews collected  (Blondeau et al. 2019). Consequently, the corpus was purposely designed using the same social categoriesage, gender and socioeconomic status (SES)as previous Montreal variationist corpora, thereby ensuring comparability. In this article, we provide an apparent-time analysis of the situation, which is based on a synchronous comparison of age groups in 2012. According to the labovian apparent-time construct, differences among generations mirror actual diachronic developments in a language (Bailey et al. 1991).
While the greater Montreal area still bears the evidence of the traditional split between Francophone and Anglophone communities, most neighbourhoods are increasingly culturally and socially diverse. Hochelaga-Maisonneuve is one of the neighbourhoods that best lends itself to comparison with previous corpora; it is a predominantly French-speaking, socially diverse urban neighbourhood. Situated in the eastern part of the city where Francophones constitute a majority, Hochelaga-Maisonneuve is a working-class neighbourhood that has become gentrified, leading to social mixing (Germain and Rose 2010). This cohabitation of populations from different social backgrounds has allowed the collection of a socially stratified corpus, which facilitates comparisons with the former socially-stratified 1971 Montreal French corpus. (Sankoff et al. 1976).
Data was collected from 50 participants of whom 23 were female and 27 were male. Participants ranged from 18 to 89 years of age: 13 participants (18-25), 12 participants (26-39), 12 participants (40-60), and 13 participants (61 +). They were required to have grown up in the Great Plain area of Montreal, to have been educated in French and to have been living in the neighbourhood for at least five years. For the purposes of comparability, the SES categories were based on the Montreal-1984 categorization scheme (Thibault and Vincent 1990), according to the occupational history of the speakers or their families : SES High (Liberal profession/ Business Person/Bachelor Degree/Intellectual), SES Intermediate (Technician /Supervisor/ White-collar/Office worker), and SES Low (Blue-collar/Manual worker/History of no stable employment). As shown in Table 1  speaker sample was somewhat hampered by the fact that it was difficult to recruit older participants situated at the higher end of the social hierarchy in the oldest age group, particularly among women. The data was collected via face-to-face sociolinguistic interviews. All 50 interviews were transcribed and audio-aligned. Thirty-eight of the interviews were transcribed using ELAN at the Université de Montreal under the supervision of M. Tremblay, and the remaining 12 interviews were transcribed using PRAAT at the University of Ottawa, under the supervision of F. Martineau.

The variable context
This study focuses on three variants, alors and donc, and their vernacular counterpart CFQ. These connectors can fulfill grammatical and discursive functions. While their grammatical function corresponds to the expression of a consecutive relationship between two propositions, as in (2-4), their discursive function corresponds to cases where the connector is used to engage a turn of speech or introduce a new topic (5) or to mark the end of a speech turn (6-8).
(2) on ramassait pas la neige alors les bancs de neige, on sautait du balcon du deuxième là on sautait dans le banc de neige en bas. (FRAN-Montréal HOMA13_911M67) 'they didn't collect the snow so, (there were) the snow banks, we jumped off the balcony of the second floor, we jumped into the snow bank below.' (3) La ruelle euh la la neige ils la ramassaient pas donc il y avait plein de neige l'hiver (FRAN-Montréal HOMA13_027F60) 'The alley, uh, the snow, they didn't collect it so there was a lot of snow in the winter.' In section 6, we will consider how phonetic variation is intertwined with lexical variation and further refine the variable context.

Exclusions, coding and analyses
Only the forms that can be used interchangeably to fulfill both grammatical and discursive functions were included in the analysis, resulting in a total of 4120 tokens (with an average of 82.4 tokens per participant). We excluded other uses such as alors que 'while' and the intensifier donc [dɔ] in (9) (Bertrand 2014).
'Wow. It's so beautiful.'  We also excluded the one token of the form du coup (10), frequent in Hexagonal French, due to its rarity.
(10) Ben on dirait que du coup je suis comme gêné fait-que j'ai comme de la misère. 'Well it seems like I'm like embarrassed so I have like a hard time.'(FRAN-Montréal HOMA13_015M18) All tokens were coded for three social factors: age, gender and SES (high (SES 1 +2), intermediate (SES 3+4) and low (SES 5+6)) as well as for consequential or discursive function.We present the results of multivariate analyses conducted with the variable rule program GoldVarb X (Sankoff, Tagliamonte and Smith 2005) and further explore the source of the variation by contrasting gender, age and socioeconomic status based on cross-tabulations of the results. Similar analyses were conducted on the phonetic realization of the three variants for a subset 2 of the corpus, comprising the 37 speakers aged 18-60 as displayed in Table 1 (3679 tokens, with an average 99.4 tokens per participant).

RESULTS: LEXICAL VARIATION
This section focuses on the results of the lexical variation between alors, CFQ and donc.

General tendencies
The following analyses are based on a total of 4120 tokens of the three connectors from the 50 speakers of the 2012 corpus. 3 Table 2 provides the distribution of each of the three variants.
While alors is rare, barely reaching a rate of 2.6%, donc is used 18,4% of the time. Neither of the two variants associated with Standard French (alors and donc) reaches the 20% threshold. CFQ, generally associated with vernacular speech, is by far the most common, with a rate of 78.9%. These results confirm a previous analysis based on a sub-sample of the corpus suggesting a change from below (Blondeau 2 One of the limitations of this study is that we were unable to conduct an analysis of the phonetic realization of each variant, since we do not have access to the sound files and transcriptions of the 12 older speakers interviewed by France Martineau and Yves Frenette. These older speaker data are only searchable online for lexical items via a search engine.
3 Unlike previous work on consequence markers in Montreal 2012, which were based on sub-samples (Blondeau et al. 2019, Martineau 2019, this study considers the corpus in its entirety. The distribution of donc in the 2012 corpus is higher than in the 1971 data: it appears to have benefited from the decline of alors and progressed within the speech community. However, this rise requires closer scrutiny, beginning with the role of pragmatic function.

Pragmatic function
As previously discussed, alors, donc and CFQ are considered functionally equivalent and can be used as consequence markers or discursive markers. Table 3 shows the distribution of the variants according to the pragmatic function. While CFQ favours a discursive function, the two others do not, possibly indicating that the more frequent a form, the closer it is associated with a discursive function. This suggests that the spread of CFQ can be explained by its inclination to fulfill a discursive function.
The next section examines the social configurations of the variation through the lens of apparent time. We first examine the decline of alors (section 5.3), and then the two progressing forms: donc and CFQ (section 5.4).

Alors: A dying star
As shown in Table 2, in 2012, alors only represents 2.6% of the variable context (N = 109), and is therefore too rare to sustain a multivariate analysis. However, an examination contrasting gender with age based on cross-tabulations of the results is   informative. In Figures 1 and 2, we present the use of alors in comparison with the other variants combined. Note that while each cell represents a small number of participants (usually two), the number of tokens considered is relatively high. Taken together, these results are revealing of community trends. As shown in Figure 1, the use of alors is retained only by men belonging to the oldest age group, and, among this group, its usage is socially stratified. Moreover, alors is practically absent from the speech of men in the 18-60 age group, and no social stratification is observed. It is also worth noting that in the 61+ age group, alors is only used by speakers of the high (H) and intermediate (M) SES, and almost absent from the speech of the speakers with a low (L) SES status.
In contrast, Figure 2 shows that older female speakers seldom use the variant alors and no evidence of social stratification is detectable, as is the case for males . For women, the use of alors is also rare in the oldest age group. 4 It is slightly more present in the speech of the 40-60 age group, but there is no sign of SES stratification. In the younger group (18-39), alors is almost completely absent.
Previous studies have debated whether the use of alors by older speakers was a sign of language change or an age-grading phenomenon. The 2012 apparent-time analysis sheds light on this issue. With a community rate below 3%, alors appears to have become obsolete. 5 In colloquial French, alors is only retained by the oldest male speakers and only in this age group is regular social stratification apparent. We propose that this association of alors with older male speakers with a high or mid SES exemplifies the vestiges of its use as a historically prestigious variant. Two solid contendersstandard donc and vernacular CFQare competing to fill the void left by the decline of alors.   Table 4 shows a sharp contrast between speakers over the age of 60, who use donc only 10.4% of the time, and speakers in the 40-60 age group, who use donc 23.6% of the time. This difference, together with the link between donc and a high SES, would support the claim that donc is replacing alors as the prestigious variant. This is additionally supported by Figure 3, which shows similar social stratification around the use of donc by both male and female speakers in the 40-60 and 61+ age groups, though this is less apparent in the latter.

Two variants on the rise
While Table 4 showed an association between male speakers and the use of donc, it is worth noting that in the 40-60 age group, this variant is associated with female speakers, who use donc at a rate of 30%, while men from the same age group use donc at a rate of 14%. Within this age group, donc clearly appears to be replacing alors as the prestigious form; its association with female speakers and its social stratification are indications of a change from above led by women.
This contrasts sharply with the use of donc by speakers between the ages of 18 and 40. Figure 4 shows that, in comparison with female speakers (40+), female speakers (18-39) abruptly abandon donc in favour of CFQ, most notably in the 26-39 age group.
Surprisingly, unlike women of the same age group (18-39), men have increased their use of donc. Comparing Figures 3 and 5, we see that male speakers from the high SES between the ages of 26 and 39 use donc at a much higher rate (63%) Figure 3: Rate of donc according to age group and SES (men and women 40+) than male speakers from the 40-60 age group (36%). Comparing the 18-25 age group with the 26-39 age group, we see an increase from 14% to 46% (intermediate SES), and from 3% to 18% (low SES). In the 18-25 age group, the social stratification of donc becomes irregular as the higher SES group discards this form 6 and selects the same form (CFQ) as female speakers in the same age range.
The sudden (and surprising) abandonment of donc in favour of CFQ is seen in female speakers aged 18 to 39 -who represent an age group that has never exhibited high rates of use of alorsand in male speakers aged 18 to 25 with a high SES.
To summarize, donc appeared to be a good candidate for the replacement of alors as the prestigious variant. However, our apparent-time analysis shows that its usage was of short duration, favoured by speakers aged 40-60 with high SES and mainly women. A closer look at the younger age groups suggests a rapid abandonment of donc by female speakers. The success of donc was very limited in scope and this can only be understood in light of the story of CFQ.  6 An anonymous reviewer remarks that discourse marker use can be highly speaker-specific and that it is possible that there is something particular about the two speakers. We agree that one limitation of the present analysis is that each cell represents a small number of participants (2). However, the number of tokens considered in the high SES 18-25-year-old males category is relatively high (N = 160), and the near-categorical use of CFQ by these two speakers is consistent with the apparent-time data which show that gradual abandonment of donc in the community. 36 CJL/RCL 67(1/2), 2022

CFQ: A rising star
We saw in Table 2 that in 2012, CFQ was used 79% of the time. This is an increase of 26% from the 1971 data reported in Blondeau et al. (2021b). The multivariate analysis provided in Table 4 shows that in 2012, as in 1971, this form is favoured by female speakers, young speakers, and individuals with a low SES. However, a closer look at the use of CFQ shows that its association with low SES individuals is limited to older speakers. Figure 6 shows that for speakers aged 40 to 60 and 61+, this form is strongly favoured by speakers with low SES. Figure 6 shows that male speakers of the 40-60 age group use CFQ systematically more frequently than women (Men: H (62%) M (78%), L (98%); Women: H (43%), M (56%), L (92%). However, Figure 7 shows the reverse situation for female speakers aged 18-25 and 26-39. CFQ is favoured by women, who display a near-categorical selection of CFQ; consequently, neither of the two groups (women 18-25 and 26-39) shows social stratification related to their use of the vernacular variant. In contrast, men between the ages of 18-25 and 26-39 still demonstrate variation in their use of CFQ. In the 26-39 age group, men with high SES continue to eschew CFQ. However, the change observed in female speakers aged 18 to 39 appears to be spreading to young male speakers, in particular those with a high SES, who use of CFQ 99% of the time.
To summarize, the results of our apparent-time data analyses indicate that women were at the forefront of two successive linguistic changes. Female speakers first initiated a change from above with the replacement of standard form alors by Figure 6: Rate of CFQ according to age-group and SES (women and men 40+) 37 standard form donc. Donc was then quickly abandoned and replaced through a change from below in the form of the vernacular CFQ. Male speakers also contributed to the replacement of alors with donc. Despite being slower in their adoption of donc, they continued to use it after the form was abandoned by women in favour of CFQ. This delayed change in male speaker behavior accounts for the link between men and the use of donc visible in the multivariate analysis provided in Table 4.
The rise of CFQ and the decline of alors and donc raise questions regarding the socio-stylistic value of the CFQ variant. If the vernacular form extends across all social groups, then this also raises the question of the sociolinguistic status of this variant. Is the spread of the vernacular variant CFQ to all social groups truly the result of a change from below? The next section addresses these questions by considering the socio-phonetic variation in the use of CFQ.

SOCIO-PHONETIC VARIATION
In this section, we show how phonetic variation is intertwined with lexical variation to convey both linguistic and social information. We begin by describing the subcorpus used for the analyses, examine the different phonetic realizations of all three lexical variants, and finish with a focused analysis of the impact of linguistic and social factors on the choice between two of the primary phonetic variants of CFQ, [fak] and [fɛk].

The sub-corpus
The transcription of the interviews with software that provided time-aligned annotations allowed further coding of socio-phonetic variation in the use of the three lexical variants. In this section, we focus on the 37 speakers, ages 18-60, and study the 3,679 tokens from the available sound-aligned transcriptions. 7 As seen in Table 5, the distribution of the three lexical variants in the sub-corpus (37 speakers) is similar to that of the main corpus (50 speakers).

Phonetic realizations of alors, donc and CFQ
In the sub-corpus, alors is found to have two different phonetic realizations: [ Table 6 shows the results of a GV analysis on the effect of the presence of a word-final schwa on the selection of the variant for consequential or discursive function. 8 In Table 6, it is evident that the final schwa is more frequently present when the form in question is functionally discursive, a phenomenon also observed by Dessureault-Dober (1974  The socio-phonetic study is based on auditory phonetic transcriptions. The different phonetic realizations of alors, donc and CFQ were coded by two native speakers of Montreal French, with undergraduate training in phonetics. After searching the orthographic transcriptions for the relevant forms, each token was then associated with a broad auditory phonetic transcription. 8 A reviewer suggests also considering the phonetic context as a conditioning factor for presence of a word-final schwa.This is not an unreasonable suggestion, given that schwa is often inserted to break up or avoid complex consonant clusters in French. However, wordfinal schwa can also be inserted to signal turn of speech, change of topic or the end of a speech turn, which accounts for its association with the discursive function of consequence markers. 39 epiphenomenal and reflects the fact that discursive function favours CFQ (as seen in Table 3).
As mentioned previously, there is considerable variation in the phonetic realization of CFQ.   Phonological context was not considered as a factor, despite the fact that, as an anonymous reviewer pointed out, in written French, final schwa in ça fait que is deleted before a vowel as in ça fait qu'elle… vs ça fait que la femme… Spoken French differs from written French in that final schwa deletion can also occur before a consonnant: ça fait que la femme is generally  [a] with older speakers from working-class backgrounds in Quebec French (see section 3), we wondered whether the choice of vowel in CFQ is a potential sociolinguistic variable, a possibility supported by the fact that both fak and fek are found in informal writing, such as text messages (13) (from Blondeau and Tremblay, to appear). (13) a. Tasaccess mais faut que tu aille un compte fakjaidemande au gars de ten faire un.
T'as accès mais faut que tu aies un compte fak j'ai demandé au gars de t'en faire un. 'You have access but you need an account so I asked the guy to create one for you.' (Texto4Science, user_110250) b. jc pas akelle heure jva revenir fekfie toi pas sur moi pour faire de koi Je sais pas à quelle heure je vais revenir fek fie toi pas sur moi pour faire de quoi. 'I don't know when I'll come back so don't count on me if you want to do something. ' (Texto4Science, user_111030) The next section examines the linguistic and social conditioning of this vowel alternation.

Defining a new variable context: The [a]/[ɛ] alternation in CFQ
This section focuses on the vowel alternation between [a] and [ɛ] in CFQ. Both vowel options appear frequently in the sub-corpus, making them good candidates for sociolinguistic analysis. All CFQ forms were coded by two native speakers. Each CFQ was searched in the audio-aligned orthographic transcription and matched to a possible phonetic representation from a list, which was adapted as needed. Given the saliency of the [a]/[ɛ] alternation in Quebec French (Riverain-Coutlée 2014), and consequently the fak/fek alternation, this coding process was straightforward, and no particular difficulty was reported. Reliability tests were not conducted. However, unclear cases were discussed between the two coders and in doubt, ambiguous tokens were excluded from the analysis.
pronounced ça fait qu'la femme [safɛklafam]. Moreoever, final schwa deletion in ça fait que is not obligatory before a vowel and thus in Quebec French ça fait qu'elle can be realized as [safɛka] (Table 9). This result is much lower than the 13.4% reported by Dessureault-Dober (1974: 6). This is an indication that the reduced form has continued to progress on the path to lexicalization. Table 10 shows that the presence of [sa] or its reduced form [s] is not associated with the consequential or discursive function of the variant.   presence of ça triggers a near-categorical selection of [ɛ] forms, which justifies the exclusion of ça forms from the analysis. The final sub-corpus comprises 2 690 tokens representing 92% of the occurrences of the CFQ variant.

fɛk variants vs fak variants: Social factors
The following analyses were performed on 2,690 tokens across 37 speakers aged 18-60. The  Table 12 displays the results of a GV analysis on the effects of social factors on the fɛk realization and the fak realization of the reduced forms (without ça). The application value is the fɛk variants.
The three factor groups considered were selected as significant. With a range of 42, the age factor plays the most important role in the variation. The variable is characterized by a regular age distribution where the fɛk variants are favoured by younger speakers. The SES factor group is also significant with a range of 23: the fɛk variants are favoured by speakers from the intermediate SES and disfavoured by the other two groups. Gender is also selected as significant, with a range of 9: female speakers slightly favour the fɛk variants.
Given the role of gender in lexical variation discussed in section 5, separate GV analyses were performed for each gender group in order to better understand the socio-phonetic variation at play. Table 13 provides the results of the analysis based on 1,712 tokens for female speakers, and 978 tokens for male speakers.
Both analyses identify age as the most important factor for both female and male speakers. The fɛk variants are clearly preferred by younger speakers of both genders. However, for the two other age groups, female and male speakers behave differently.      (Labov 1972). The situation for male speakers provides the reverse portrait. Men from the intermediate SES group do not favour the fɛk variants, while speakers with a high SES or low SES favour it. Note that the order of the percentage values and corresponding factor weights are inverted, as a result of interactions discussed below in the cross-tabulation analysis.

Women as leaders of the change in favour of the fɛk variants
To clarify the relationship between gender and SES, cross-tabulations were performed. Figure 8 shows cross-tabulations of the results for the female speakers by age group. Comparing results across three age groups provides a better apparent-time perspective on the introduction of the fɛk variants and its diffusion among the female speakers.
In the 40-60 group, the fɛk variants are almost entirely absent for low SES speakers and marginally present for speakers with high SES. However, they are clearly associated with intermediate SES speakers. For the 26-39 group, the fɛk variants are used by all three socioeconomic groups, and again we observe the propensity for intermediate SES speakers to use these forms. This suggests that the middle-class women 40-60 are the leaders of the change. Finally, among the younger age group (18-25), all social groups are users of the fɛk variants, and there is no longer a clear social stratification. 12

Men
Similar analyses were performed on data gathered from male speakers. Figure 9 is based on cross-tabulations of the results for the male speakers by age group.
Young male speakers aged 18 to 25 from the high and intermediate SES are strong users of the fɛk variants. However, the other two age groups exhibit a very different behavior. The fɛk variants do not reflect a regular socioeconomic distribution within the 40-60 age group. In particular, the use of fɛk by low SES is restricted to the older male speakers and remains puzzling, and is probably the source of the discrepancy identified in Table 13. For speakers from the 26-39 age group, the fɛk variants are used infrequently and again without a visible regular socioeconomic distribution.
To summarize, the results of lexical variation presented in section 5 confirmed the rise of CFQa change from belowbut left unexplained the sudden adoption of a vernacular form by speakers with high SES. To resolve this issue, we further examined in section 6 the distribution of two phonological forms of CFQthe fak variants and the fɛk variantsand demonstrated the social conditioning of the variable: fɛk variants are favoured by women and youth, while fak variants are associated with speakers with a low SES. In the next section, we propose a new model which reintegrates the variants alors and donc and combines them with the fɛk forms of the CFQ variant, which we label the 'All' variant.

THE SOCIOLINGUISTIC STATUS OF FεK VARIANTS
The ultimate success of CFQ is attributed to the rise of the fɛk forms. As shown in Table 14, alors and donc are standard language forms. Fak forms are both vernacular and stigmatized, especially among women. The ultimate success of CFQ is really the ultimate success of fɛk variants, which represent non-stigmatized vernacular forms. Fɛk first replaced donc as a prestigious variant, but then spread to speakers with low SES. Fɛk is not socially restricted because it is more neutral, which explains its spread across all levels of SES.   In the analysis of lexical variation in section 5, we observed that the speakers who were previously users of donc or alors abandoned these two standard variants in favour of the CFQ variant. A clearer picture can be obtained by opposing the new All variant (i.e., alors + donc + fɛk) with the fak variant. In particular, this allows for a better understanding of the situation observed for male speakers (section 6.4.2). The analyses provided here are based on 3,459 tokens across 37 speakers aged 18-60, of which 1,403 tokens were from male speakers and 2,056 tokens were from female speakers. Figures 10 and 11 illustrate the inclusion of the other variants and offer interesting insights into these situations. Figure 9 showed that for speakers from the 26-39 age group, the fɛk variants were used infrequently and without a visible regular socioeconomic distribution. As we can see in Figure 10, the All vs fak variant model provides a better representation of the usage of consequence markers observed in male speakers of this age group than the simple comparison between the fak and fɛk variants. For the 26-39 age group, the All vs fak variant model provides an understanding of the sociolinguistic distribution of the forms. This new level of comparison demonstrates that the nonstigmatized variant of CFQ, fɛk, patterns with alors and donc.

Men
Looking at the situation from this perspective offers a better way to understand the path of the change. It also provides a framework for the interpretion of the variation from the perspective of a binary opposition between a stigmatized vernacular

Women
A similar analysis is performed on the data from female speakers (Figure 11). A comparison with the data provided in Figure 8 above (section 6.4.1 Women) shows that the model provides a slightly different perspective on the change affecting consequence markers. The difference is essentially only applicable to the 40-60 age group, which is the generation of donc users. Looking at this change on an age continuum from the perspective of apparent time, a clearer picture of the change emerges. For the 40-60 age group, speakers with middle or higher SES use fɛk and donc, while fak was the main choice for lower SES speakers. The disappearance of donc from the speech of the 26-39 age groupas observed in section 4.4.1made room for a binary opposition between the fɛk and the fak variants. Finally, the fɛk variant represents the new normal and the default variant for the younger age group.

DISCUSSION
This article contributes to a better understanding of the two theoretical constructs -Change from above and Change from belowby showing how the two may coexist within a given variable. Our results illustrate how a lexical form first associated with the lower SES (CFQ) was able to spread to the higher SES, and replace the two competing variants (alors and donc). As the vernacular variant CFQ can phonologically encode social variation, it blurred the distinction between change from below and change from above. Because the phonological variant /fɛk/ was less stigmatized than the traditional vernacular variant /fak/, it was massively adopted by the higher SES speakers, and then spread across all SES at the same time, possibly eliminating its old rival fak. Thus, the ultimate success of CFQ in this lexical competition can be attributed to the fact that what originated as a change from below, specifically the replacement of alors and donc by CFQ, ultimately became a change from above with the replacement of fak by fɛk. Thus, this 'change-from-above' component to the variable is the key to understanding the overall success of the change from below, and explains the crucial role played by the higher SES in the diffusion of CFQ. In the case of consequence markers in Montreal French, a change from below (the lexical change) and a change from above (the sound change) are intertwined within the same variable. 13 Our results suggest that different components of the same change may be of a very different nature, including some parts as change from below, some as change from above.
Another contribution of our apparent-time analyses is that they allow us to disentangle the respective roles played by men and women in the diffusion of both types of change. Counter to Blondeau et al's 2019 proposal that the replacement of alors with donc was initiated by men, we show that the cascading series of changes resulting in the diffusion of donc and fɛk was driven by female speakers. At first, alors competed with fak. Then alors was gradually being supplanted by donc. Finally, donc was gradually being supplanted by fɛk. The apparent-time analyses suggest that fɛk is now spreading across all SES, possibly eliminating fak. This could indicate that consequence markers are on the verge of losing their sociolinguistic marker status.
Alors/fak → donc/fak → fɛk /fak → fɛk So, not only did women lead the change involving the sudden rise and fall of donca phenomenon of short durationbut they also contributed to the diffusion of CFQ by propulsing the fɛk variant throughout the speech community. This highlights once again the active role of women in the diffusion of innovations (Labov 1990).
In contrast, men demonstrated a delay in the adoption of the new forms. This delay explains their present use of both donc and fɛk and by extension, their deviation from the traditional push-chain pattern. As late adopters, male speakers appear to have followed a two-step process in the replacement of donc by fɛk. The first step was the introduction of fɛk, at which point donc and fɛk co-existed as non-stigmatized alternatives to CFQ. This was followed by the abandonment of donc.

CONCLUSION
In this article, we further documented the change in the use of consequence markers in Montreal French. We made a methodological contribution by showing how circumscribing the variable context can be an analytical and dynamic process. We first considered a tripartite lexical variable consisting of the two standardized variants alors and donc, and a vernacular variant CFQ. Digging deeply into the patterns of this variation led us to unveil a hidden phonological dimension, and to divide the vernacular variant (CFQ) in two subvariants, fak and fɛk.
In basing our apparent-time analysis of the lexical variation on a 2012 corpus and by integrating a socio-phonetic dimension to the analysis of a subset of the data, we have deepened our understanding of a change in progress at the community level. The comparison between the lexical variation in 2012 and past analyses confirms that CFQ was not only the rising star but that this variant has diffused across the speech community over time. The 2012 apparent-time perspective suggests that women with intermediate SES were at the forefront of the change. Additionally, our analysis allows a reassessment of the role of alors. The rarity of this form in 49 the 2012 corpus does not support an age-grading explanation for alors. If alors played such a role before (Thibault and Daveluy 1989), its scarcity in 2012 suggests that this form is being phased out. The third variant donc behaved like a shooting star. Due to its association with the standard language, donc appeared to be a good candidate for the replacement of alors as the prestigious variant. Marginal in the 1970s data, its presence has increased over time. However, our apparent-time analysis of the 2012 data shows that its usage was of short duration, favoured by speakers aged 40-60 with high SES, most of whom were women. A closer look at younger age groups suggests an abandonment of this form by women aged 18-39 and men aged 18-25. The short duration of the success of donc can only be understood in light of the ascent of CFQ.
Our results confirm the rise of CFQ and provide an explanation for its use by speakers with low and high SES at the expense of the standard variants alors and donc. We attribute the success of CFQ in this lexical competition to its ability to phonologically encode social variation. A careful examination of the social distribution of the two groups of phonological variants of CFQ (fak variants and fɛk variants), revealed that the choice between the two was socially conditioned: fɛk variants are favoured by women and youth, and fak variants are linked with a low SES. Nowadays, the lowering of the lower-mid vowel /ɛ/ in word-final position is perceived as archaic, and stigmatized (Riverain-Coutlée 2014). This makes the lexicalized fak variants particularly salient, and prone to avoidance by younger and more educated speakers. After a change in the phonogical realization of the lexical variant CFQ (a → ɛ), CFQ was no longer stigmatized, potentially returning the variant to subconscious status, allowing it to continue to thrive in the community. Our article shows the key role played by women in all three phases of the change: 1) the rise of CFQ (a change from below); 2) the phonological tweak to make it socially acceptable by the middle class (a change from above), and 3) the propulsion of the new variant, now devoid of stigmatization (a change from below).
To conclude, our results further show that traditional first wave sociolinguistic categories are still useful to track the social trajectory of changes in progress. Such community trend studies provide a necessary background for third wave sociolinguistic studies, which explore the role of lifespan change, including agency and age grading, aspects that we are currently exploring in panel studies. As directions for future research, a closer look at the intra-individual variation through the lense of style might shed light on the social meaning associated with each variant. In addition, perceptual analyses of the values associated with the sociophonetic variants of CFQ, as well as with alors and donc, might be informative about the sociolinguistic marker status of this variable.