2 CONTEXT: SUPRALOCALIZATION AND REGIONAL FRENCH

As noted above, the turn of the millennium saw an increase in variationist research on regional dialect levelling (or, more accurately, regional accent levelling (Foulkes and Docherty, Reference Foulkes, Docherty, Foulkes and Docherty1999b: 5)). A number of studies in the late 1990s and early 2000s examined varieties of British English, such as Foulkes and Docherty (Reference Foulkes and Docherty1999a), Watt and Milroy (Reference Watt, Milroy, Foulkes and Docherty1999), Williams and Kerswill (Reference Williams, Kerswill, Foulkes and Docherty1999) and Dyer (Reference Dyer2002) (and cf. Kerswill (Reference Kerswill, Britain and Cheshire2003), discussed above); and in the early 2000s, several UK-based researchers applied a similar perspective to their work on the regional French of France. (For example, Armstrong (Reference Armstrong2001, Reference Armstrong2002), Temple (Reference Temple, Hintze, Pooley and Judge2001), Hornsby (Reference Hornsby, Jones and Esch2002, Reference Hornsby2006, Reference Hornsby2007), Pooley (Reference Pooley, Jones and Esch2002) and Boughton (Reference Boughton2003).)

The project reported in Boughton (2003; see also Reference Boughton2005) was undertaken in this scientific context as well as in response to a growing number of impressionistic observations in the literature alluding to the attrition of diatopic variation in metropolitan French, especially since the mid-twentieth century. To give but one example, Léon (Reference Léon1993: 222) remarks that a study carried out in 1980 involving perceptual judgements of speakers from a number of regions already showed a high degree of convergence with the supralocal norm at that time. In order to test such remarks by empirical means, Boughton (Reference Boughton2003) carried out fieldwork in Nancy and Rennes in 1997–1998 to gather a large corpus of interviews from 64 speakers (32 from each city) stratified according to age, gender and social class in addition to region. The principal findings, both behavioural and perceptual, led to the conclusion that supralocalization was indeed quite advanced in the Nancy–Rennes corpus, especially in Rennes, and that this was probably also the case in other similar cities and therefore in northern, urban French more generally.

This hypothesis was taken up by Pooley (Reference Pooley2006) in a survey article which sought to examine the progress of regional accent levelling in French in the langue d’oïl area and Franco-Provençal zone by comparing well-known accounts based on speakers born in the first half of the twentieth century – such as Martinet (Reference Martinet1945), Walter (Reference Walter1982) and Carton et al. (Reference Carton, Rossi, Autesserre and Léon1983) – with more recent analyses of speakers born after 1965, who would have been aged 40 or under at the time of writing in 2006. These analyses were drawn from studies of locations in eastern France, plus Brittany, Normandy, the Nord–Pas-de-Calais, Picardy, Paris and the Auvergne. Some of these, e.g. Tifrit’s (Reference Tifrit2003) study of Dijon and Guézennec’s (Reference Guézennec2003) investigation of the Ⓘle-de-Sein, Brittany, were carried out under the aegis of the Phonologie du Français Contemporain (PFC) project which, while not primarily variationist in its original intent (Durand, Laks and Lyche, Reference Durand, Laks, Lyche, Pusch and Raible2002: 96; see also Pooley, Reference Pooley2006: 369), has done much in recent years to provide new data relating to regional phonological variation in French. In a series of related maps, based partly on Pottier (Reference Pottier and Martinet1968) and Armstrong (Reference Armstrong2001), Pooley illustrates his proposed metaphor of an ‘oïl slick’ representing the diffusion of supralocal French in areas where divergence was previously much more entrenched. He concludes that ‘[e]ven allowing for the peripheral resistant areas of northern France (both east and west), the loss of regiolectal features and the consequent convergence towards Oïl or supra-local French is remarkable. Nowhere else in western Europe are phonological regiolectal features levelled to such a degree over such a large area’ (p. 386). It is notable that Alsace is marked as one of the ‘peripheral resistant areas’ where ‘divergent forms occur variably’ among speakers born in or after 1965. Pooley (Reference Pooley, Jones and Hornsby2013: 196) later notes that this conclusion was drawn ‘partly on the basis of a lack of evidence and partly on informal communication with colleagues’. Indeed, it was in this context that the research reported here found a key element of its motivation: to break new ground by carrying out an extensive variationist survey in a border region of the Hexagon that had not been previously studied in the Labovian framework (and was little studied from any theoretical perspective). We return to this shortly below.

Research investigating supralocalization in French has continued in recent years. It is, for example, the chief focus of Armstrong and Pooley (Reference Armstrong and Pooley2010), which extends the debate to francophone Europe; and within France, the frame of reference has been pushed beyond the northern two thirds of the country to consider southern and central regions of the langue d’oc. While Pooley (Reference Pooley2006) offers some consideration of the linguistic situation in the Limousin and the Auvergne and concludes, on the basis of studies by Potte (Reference Potte, Taverdet and Straka1977) and Sobotta (Reference Sobotta2003), that the available evidence suggests ‘considerable levelling’ (p. 382) in the northern langue d’oc area, in a companion piece, Pooley (Reference Pooley2007) provides a review of levelling in more southerly regions. A number of accounts are synthesized to build a picture of a ‘Dominant Southern Pattern’ (p. 43), marked primarily in vocalic features (such as schwa, nasal vowels and mid-vowels), which Pooley concludes is ‘very much alive’ despite ‘many indications that it is markedly regressive’ (p. 61): ‘convergence towards supra-local norms appears to have gained ground among speakers born in the latter part of the 20^th century to the point where many young adults from the region are no longer immediately recognisable as southern’ (pp. 62–63).

Durand, Eychenne and Lyche (Reference Durand, Eychenne, Lyche, Jones and Hornsby2013), responding in part to Pooley (Reference Pooley2006) and Armstrong and Pooley (Reference Armstrong and Pooley2010), bring to bear on the question of convergence an authoritative overview of 10 years of PFC data, focusing on southern varieties. They argue that ‘while southern French is indeed changing in the direction of northern-based norms, substantial differences still exist’, and ‘even within northern France, we do not observe the homogeneity which is often assumed in the literature’ (see also Boughton, Reference Boughton, Jones and Hornsby2013; Hall and Hornsby, Reference Hall, Hornsby, Davies and Ziegler2015). The authors propose examples of features which do not currently appear to be aligning straightforwardly with supralocal norms, namely the loi de position constraint on mid-vowels, and variable schwa. This prompts us to reflect that a focus on supralocalization can lead us to imply that the overall phonological picture is purely one of loss of variation, of simplification and homogenization. Of course, this is certainly not the case: variation is persistent and is often more complex than it might first appear. Even if there is general agreement regarding the loss of vestigial localized features, this does not mean that all speakers in all regions of France simply adopt the supralocal norm, or in the same way (cf. Armstrong and Pooley, Reference Armstrong and Pooley2013).

Mooney (Reference Mooney2016) demonstrates this amply by way of a sociophonetic analysis of a corpus gathered in the Béarn region. Inspired by Hornsby’s (Reference Hornsby2006) re-evaluation of the concept of regional French, Mooney examines outcomes of both language contact and dialect contact in an area that could be expected to be among the most resistant to processes of convergence: the peripheral southwestern corner of the Hexagon. While Mooney observed evidence for the adoption of supralocal features, this was not entirely straightforward: younger speakers were found to combine supralocal and regional features in an innovative way that led to increased, rather than decreased, linguistic diversity (pp. 109–110, 124).

It seems therefore that the processes and outcomes of linguistic change may differ according to factors such as whether the regional variety in focus is located in the supralocal area or outside it, whether in the south or in peripheral areas of the north of France, and perhaps also according to the nature of the accent features studied (in particular, whether they are vocalic or consonantal). While the majority of studies of regional French over recent decades provide evidence in support of the related ongoing processes of (i) attrition of traditional accent features and (ii) diffusion of supralocal norms and behaviour, we have seen that it is not always the case. But what of Alsace?

Relatively little is known about accent variation and change in the regional French of Alsace, in spite of its unique sociolinguistic profile. A border zone from both the political and linguistic viewpoints, several varieties are in contact there, including: French; various sub-dialects of the Germanic regional language, Alsatian; standard German; a Romance variety known locally as Welsch; and a number of immigrant languages. It is therefore a prime site for the observation of language contact phenomena. For this reason, much previous research has tended to focus on issues such as language choice, use and attitudes, and code-switching between French and Alsatian (Gardner-Chloros, Reference Gardner-Chloros1991; Vassberg, Reference Vassberg1993; Vajta, Reference Vajta2004) as well as on Alsatian itself (Philipp, Reference Philipp1965; Beyer and Matzen, Reference Beyer and Matzen1969; Bothorel-Witz et al., Reference Bothorel-Witz, Philipp and Spindler1984) and rather less on the characteristics of the French spoken in the region.

Pooley (Reference Pooley2006: 371) refers to the ‘continued vitality’ of the Alsace accent. However, he cites only ‘informal confirmation’, along with the work of Bickel-Kauffmann (Reference Bickel-Kauffmann1983), as evidence of the continued use of regional features described in key sources which also date from the early 1980s (Walter, Reference Walter1982; Carton et al., Reference Carton, Rossi, Autesserre and Léon1983) and which are based on very small numbers of speakers (three and one respectively). This results in what might be described as a ‘maximalist’ account of regional French in Alsace in the 1980s. Features nevertheless mentioned by Pooley, citing Bickel-Kauffmann (whose primary focus was in fact phonological transfer from French to Alsatian rather than vice versa (Gardner-Chloros, Reference Gardner-Chloros1991: 8)), are: aspiration of initial voiceless plosives in stressed syllables; realization of initial orthographic <h> as [h]; and the voicing of voiceless plosives, especially when word-initial.

With regard to studies of the substrate influence of Alsatian on the pronunciation of French, in addition to Walter (Reference Walter1982: 115–118) and Carton et al. (Reference Carton, Rossi, Autesserre and Léon1983: 14–18), the principal contribution is that of Philipp (Reference Philipp1965, Reference Philipp1967, Reference Philipp and Salmon1985), whose work on the variety of Alsatian spoken in the village of Blaesheim, near Strasbourg, also encompassed features of the French spoken there by bilingual individuals. Philipp (Reference Philipp1967, drawing on Reference Philipp1965) focuses on the distinctive prosody of five of her Blaesheim speakers rather than on segmental features and emphasizes the perceptual salience of non-standard stress patterns: ‘[c]’est probablement l’accentuation caractéristique du bilingue alsacien qui le “trahit” plus que toutes les autres interférences’ (p. 67), an aspect of regional French in Alsace which has also been the object of more recent research (Boula de Mareüil et al., Reference Boula de Mareüil, Adda-Decker, Woehrling, Bardiaux, Simon and Simon2012). Philipp (Reference Philipp and Salmon1985), based on the same fieldwork, provides details of some other features. These include the devoicing of voiced plosives and fricatives in word-final position (whereby, for example, vide and vite become [viːt] and [vit], demonstrating also that a lengthened vowel is realized before a plosive that would be voiced in standard French); and inversion of schwa and liquid in final obstruent-liquid clusters, as in faible [fεːpəl] and coudre [kʰuːtəʀ]. The latter examples also show devoicing of medial plosives with concomitant lengthening of the preceding vowel. Finally, Philipp observes the neutralization of voicing opposition in initial plosives (differentiated by the aspiration of canonically voiceless segments, as in un beau pot [œ̃popʰo]) and fricatives, of which the most remarked-on example concerns /ʒ, ʃ/, whereby joue and chou are both realized as [ʃu].

Devoicing of obstruents, along with aspiration and non-standard stress, is one of the features analysed by Bonnot, Bothorel-Witz and Huck (Reference Bonnot, Bothorel-Witz, Huck and Lüdi1993) in the scripted speech of two bilingual informants, one female aged 49 and one male aged 51, both of whom had left school at the age of fourteen. Consonant devoicing showed a gender difference, with the male speaker devoicing 78 tokens and the female, 31; the authors therefore suggest the possibility that ‘le marqueur phonétique “régional” le plus sensible est la désonorisation’ (p. 36). Gardner-Chloros (Reference Gardner-Chloros1991: 8) also notes the ‘apparent confusion’ of voiced and voiceless plosives, attributing this to ‘a transfer from Alsatian lenis stops’, alongside initial stress, as the principal characteristics of the regional accent.

Other aspects of the variety, such as morpho-syntactic and lexical interference (see, for example, Matzen, Reference Matzen1973; Jadin, Reference Jadin and Salmon1985, and other papers in the same volume) have received attention, and more recently, Marchessou (Reference Marchessou2018) has reported on an ethnographic study of the speech of 24 young Strasbourgeois.e.s (aged 16–21) in a multi-ethnic working-class neighbourhood, with a focus on lexical innovations, indirect interrogatives, quotatives and discourse markers. However, to our knowledge, no other large-scale quantitative study, based on a stratified sample, has been carried out on the phonology of the regional French of Alsace; and most existing research on the regional accent is based on rural varieties, while urban usage remains relatively neglected (cf. Durand, Eychenne and Lyche, Reference Durand, Eychenne, Lyche, Jones and Hornsby2013: 68).

The present study addresses this neglect. In particular, we focus on the following research questions: (i) do we observe in Alsace the increased diversity in regional French found in other recent work outside the supralocal area (cf. Durand et al, Reference Durand, Eychenne, Lyche, Jones and Hornsby2013; Mooney, Reference Mooney2016), or supralocalization?; (ii) what is the effect of the fundamental extra-linguistic variables of age, gender and socioeconomic status?; and (iii) are there differences in patterns of sociolinguistic variation according to whether the speech community is urban or rural? We present and discuss selected results which seek to shed light on these questions, after first setting out some methodological information in the following section.

3 THE ALSACE CORPUS

The data on which the present analysis is based are drawn from a large corpus of spoken French resulting from surveys (carried out by Pipe, under the supervision of Boughton) at two research sites in AlsaceFootnote ¹ in 2011–2012. The primary survey location was the conurbation of Strasbourg, or Communauté Urbaine (since renamed the Eurométropole) de Strasbourg, which, in 2019, is the eighth largest city in France by population size.Footnote ² The principal urban centre in the region, Strasbourg was chosen as it has not previously been the object of variationist research and for the important practical reason that the fieldworker had lived there previously for seven months while working as a lectrice at the university and had subsequently returned to carry out a pilot study for a prior related project (Pipe, Reference Pipe2010). She was therefore well acquainted with the context and had built up a substantial network of personal contacts, invaluable for successful sociolinguistic fieldwork.

This consideration also played a role in the selection of the second fieldwork site, the small village of HelsheimFootnote ³ (population approximately 700), located to the north of Strasbourg in a rural part of the same department, the Bas-Rhin, and where the fieldworker had contacts willing to provide accommodation and personal introductions to other inhabitants of the village. In addition, the two locations are in the same Alsatian dialect area,Footnote ⁴ minimizing the likelihood of linguistic differences due to substrate factors, but they are extremely different with regard to population size. The rural context was chosen as offering the greatest contrast to the urban centre, to test the hypothesis that the extent of supralocalization may be conditioned by the size and type of speech community and to investigate potential differences in patterns of variation according to the urban–rural dimension. Furthermore, Helsheim is situated in the area north of the Forest of Haguenau, known as the Outre-Forêt, which has the reputation of being the part of Alsace in which the local dialect and cultural traditions have been preserved to the greatest extent. During the fieldwork, it was observed that Alsatian was indeed regularly spoken by the villagers to one another, including within the younger generation. We might therefore reasonably expect supralocalization to be less advanced there, since Alsatian is frequently used and may influence the local variety of French (cf. Mooney’s (Reference Mooney2016: 43) discussion of linguistic transfer in a situation of Béarnais–French bilingualism).

In both locations, quota (or judgement) samples of speakers were for the most part contacted through personal networks, using the ‘friend of a friend’, or ‘snowball’ method (Milroy and Gordon, Reference Milroy and Gordon2003: 32), a technique which can serve to reduce the risk of refusal to participate and also to encourage a more informal speech style during interviews. This approach was very successful in Helsheim, where the villagers all knew each other and where a member of the host family accompanied the fieldworker on a first visit to all potential participants. It was equally effective among the existing network of contacts in Strasbourg, though these were predominantly middle class. The samples were stratified by age and gender, and in Strasbourg, also by socioeconomic group: participants were categorized as broadly working class or middle class according to whether their occupation (or parents’ occupation in the case of young people still in education) was manual (‘blue collar’) or non-manual (‘white collar’) (cf. Trudgill, Reference Trudgill1974; Chambers and Trudgill, Reference Chambers and Trudgill1998: 49). While a simplification of the complex web of factors that determine socioeconomic status, it is generally agreed among social scientists that occupation remains ‘the best single indicator’ of social class (Chambers, Reference Chambers2009: 51; see also Harrison, Reference Harrison, Jones and Hornsby2013: 14), even though many sociolinguistic surveys in francophone contexts have preferred level of formal education as a proxy measure, or have not attempted to sample for social class at all (cf. Durand, Eychenne and Lyche, Reference Durand, Eychenne, Lyche, Jones and Hornsby2013: 59). The rural sample, however, was not differentiated according to class. In part this was in order to keep the overall sample size manageable, but chiefly it was because in the tightly-knit social network of the village, the majority of inhabitants were employed either in agriculture or in unskilled industrial jobs and there was little division of the speech community into groups based on occupation or other indicators of social status. In Strasbourg, it proved difficult to obtain a full quota of working-class participants, females especially, partly owing to network factors – the fieldworker’s contacts (or their parents) were predominantly employed in white-collar jobs, as were the further participants those contacts were able to suggest – and partly to the demographic characteristics of the city, which thrives on tourism, the university and the European institutions. Many blue-collar jobs are carried out by immigrant workers and there is little industry, which is concentrated rather in Mulhouse, so the population of manual workers is relatively small in general (Howiller, Reference Howiller2008: 23, 45, 121). The additional sampling criterion of ‘nativeness’ also restricted the potential pool of participants: to enhance comparability across the sample, all speakers were born in Alsace (preferably at the research site, or at least in the Bas-Rhin department) or had moved there before their fifth birthday (cf. Labov, Reference Labov, Hungerford, Robinson and Sledd1970: 288–289). Their native language was either French or Alsatian: French for the urban speakers aged 30 and under, Alsatian for almost all the others.

In all, data were collected from 40 informants in Strasbourg and 16 in Helsheim, resulting in an overall sample of 56 speakers, structured as shown in Table 1. The three age groups represent three broad life stages (Milroy and Gordon, Reference Milroy and Gordon2003: 38–39). In Helsheim, only the youngest and oldest groups were sampled. Again, this approach served to keep the overall sample size and data collection manageable, but it also helps to draw out any age contrasts (previous studies of supralocalization which have adopted a similar sampling procedure include Watt and Milroy (Reference Watt, Milroy, Foulkes and Docherty1999); Docherty and Foulkes (Reference Docherty, Foulkes, Foulkes and Docherty1999)). As mentioned above, the urban working-class (female) sample is incomplete and so results for those cells are hereafter treated with due circumspection, and shown in parentheses. A near-full sample was nevertheless obtained for participants aged over 61 in all sub-groups, and the rural sample is comparable with the urban middle-class upper and lower age groups, which is useful with a view to contrastive analysis of variation in the corpus.

Table 1.

Size and structure of the Alsace speaker sample

Data were collected in sociolinguistic interviews, conducted in French, with either one or two participants. (The fieldworker is a completely fluent non-native speaker of French). These were relatively informal conversations, usually taking place in the participant’s home or, occasionally, their workplace. The same topic was proposed in each case: Alsatian cultural and linguistic practices. This had the advantage of placing the interviewer in the position of a learner (as recommended originally by Labov (Reference Labov, Baugh and Sherzer1984: 40)) and also allowed for the elicitation of qualitative data relating to such issues as language use and attitudes, as well as life histories. Full information and the set of guiding questions used are given in Pipe (Reference Pipe2014: 230–232). Scripted speech styles (e.g. reading passages, word lists) were not elicited, in order to keep the conversations as informal as possible, to minimize the effect of the Observer’s Paradox by avoiding the closer monitoring of pronunciation that is entailed by reading aloud, and because the primary focus of the study is the analysis of linguistic variables in spontaneous speech according to the participants’ social characteristics.

5 SYNTHESIS AND CONCLUSION

In this final section, we synthesize the patterns observed for each of the two variables, noting similarities and differences between them, and in doing so we revisit the research questions set out at the end of section 2: (i) do we observe in Alsace the increased diversity in regional French found in other recent work outside the supralocal area (cf. Durand et al., Reference Durand, Eychenne, Lyche, Jones and Hornsby2013; Mooney, Reference Mooney2016), or do we observe supralocalization?; and (ii) what is the interaction between linguistic behaviour on the one hand and our speaker variables (age, gender, class and urban or rural location) on the other? We conclude by reiterating our principal findings and suggest some potential avenues for further research.

The findings of the quantitative analyses of (h) and (ʒ) are broadly in line with expectations for traditional, non-standard regional accent features: rural and working-class speakers generally have higher rates of the non-standard variants, while middle-class and urban participants tend to use more of the standard variant in each case. While the results for gender appear rather more unusual – an observation we revisit below – it is incontrovertibly speaker age that has the greatest interaction with behaviour in the Alsace corpus. For both variables, we find a very striking decrease in rates of the traditional regional variant, in apparent time. This result is illustrated by Figure 6, which allows us to compare these trajectories of decline.

Figure 6.

Relative frequencies of [h] and non-assimilatory devoicing of /ʒ/ according to speaker age groups.

While we observe in Figure 6 a similar pattern for both variables, we see also that the dramatic decline in non-assimilatory /ʒ/-devoicing, which now appears at best vestigial but more probably lost, is not fully reflected for (h): the non-standard variant is still maintained to some extent even by the youngest age group. This difference in the findings for the two variables may indicate a degree of complexity in this variety of regional French, as opposed to a straightforward picture of supralocalization across the board. In addition, it may be the case that [h]-realization has become more restricted in its lexical as well as social distribution, but analysis of the linguistic constraints on the variable is needed to test this hypothesis. For both variables, we note in particular the considerable difference between those aged over and under 60 years. Many of the participants in the 61+ group had spent less time in formal education than is now possible, and all of them spoke Alsatian as a first language. Those aged 61 at the time of the fieldwork in 2011 were born in 1950; participants older than 61 were born in the first half of the twentieth century. The many social differences between this and subsequent generations have undoubtedly had an important impact on various aspects of behaviour, including linguistic (cf. Armstrong and Pooley, Reference Armstrong and Pooley2010); and as we noted at the outset, it is especially since the second half of the twentieth century that regional dialect levelling processes in metropolitan French have gathered pace. The results for the variables analysed here therefore provide new evidence in support of the hypothesis that supralocal French continues to gain ground, both geographically and socially, though with some differences in usage and in the degree of attrition according to the feature studied.

The predominant importance of age is illustrated further by Table 5, which shows the speaker groups arranged in hierarchical order by rate of use of each traditional variant, from highest to lowest percentage. In both cases, it is the older rural males who have the highest rate. For (h), there is a greater spread in rates of usage; for (ʒ), all of the younger groups and most of the 31–60 groups have (near-) zero rates of the localized variant. The dotted line in bold represents the averages for the sample as a whole: (h), 44.53% and (ʒ), 8.85%. While the patterns for gender, class and location are not so readily discerned in the rank order, for both variables, all groups aged 61+ have above-average rates of both non-standard variants, and all groups aged 60 or younger have below-average rates.

Table 5.

Rates of [h] and non-assimilatory devoicing of /ʒ/ by speaker group arranged in order from highest to lowest relative frequency

(Key: ORM = older rural male; MMF = middle-aged middle-class (urban) female; YWM = younger working-class (urban) male; etc.)

While these results for age point clearly to a decline in traditional regional features and a concomitant increase in the adoption of supralocal variants, our findings in relation to gender appear atypical. Some previous studies of supralocalization have shown females, especially younger, middle-class females, to be at the forefront of the adoption of supralocal norms (see, for example, Armstrong and Unsworth, Reference Armstrong and Unsworth1999 on southern French schwa; and Armstrong and Low, Reference Armstrong and Low2008 on the spread of /ɔ/-fronting) and yet our data show certain females – urban females including those from white-collar backgrounds – with significantly higher preferential rates of non-standard variants when compared with their male counterparts. This reverse gender pattern (cf. Armstrong and Pooley, Reference Armstrong and Pooley2010: 256) is highly significant (chi-square (df, 1) = 6.96, p = 0.008) for (ʒ) among the older, middle-class Strasbourg participants (males 10.63%, females 15.34%; cf. Table 4) and is also significant (chi-square (df, 1) = 4.30, p = 0.038) for (h) for the older and younger middle-class speakers from Strasbourg grouped together, where all cells are of equal size (males 31.2% [h], females 45.6%; cf. Table 3). We note the partially similar ‘distribution sexolectale atypique’ observed by Pooley (Reference Pooley, Armstrong, Bauvois, Beeching and Bruyninckx2001) for word-final consonant devoicing among older women in Roubaix. He accounts for this with reference to specific social network factors affecting women born before 1938, who worked predominantly in the textile industry alongside Flemish-speaking Belgians. In this case, there is a clear link between gender, age and occupation-based networks; in the present study, however, such conditions do not seem to apply. Indeed, in the Helsheim data, where speaker networks are dense and multiplex, we find the usual gender pattern for (ʒ) among the older group and not the reverse (older rural males 50.14%, older rural females 17.66%; chi-square (df, 1) = 124, p = 0.000, highly significant). In addition, it is notable that in our Alsace data we observe reverse gender patterns for middle-class rather than working-class females, as was the case in Roubaix.

It is difficult to offer a conclusive explanation of these atypical findings without further research on degrees of social awareness of the regional accent, and of these features in particular (Labov, Reference Labov2001: 196–197). Both seem to be above the level of conscious awareness, but their relative salience and indexical meaning may nevertheless differ. As we have noted, non-standard voicing patterns, including (ʒ), are frequently mentioned both in the linguistic literature on the regional French of Alsace and in folk sources and metalinguistic discussions, to the extent that they have attained the status of stereotype or even shibboleth (Chambers and Trudgill, Reference Chambers and Trudgill1998: 75–76). In some senses it is strange that awareness of (h) seems lower, since application of the variable involves the introduction of an additional segment, albeit one that is represented orthographically, and can also occur in standard French, such as in hop(-là)!: both aspects which might render it less stigmatized. It is also the case that /ʒ/-devoicing can occur much more frequently than h aspiré, notably in the pronoun je (+verb), and, in Alsace, in all word positions, so there is an increased chance of its being noticed and remarked on, and therefore potentially censured and avoided. A relevant piece of qualitative data is provided by one of the older, working-class Strasbourg females. Her rate of non-assimilatory devoicing of /ʒ/ was near-zero (0.04%) and yet she had one of the highest rates of [h] in the whole sample, at 88.6%. This elderly informant had lived and worked in Paris for a few years as a young adult, so it is plausible that processes of accommodation had led to the adoption of supralocal norms for voicing while (h) was retained. Further research is needed to tease out a more precise understanding of both variables. For example, a targeted study could include scripted styles to explore levels of awareness, and attitudinal experiments could be devised to elicit direct and indirect perceptions and evaluations of these and other accent features to ascertain their status in relation to both overt and covert prestige.

It is known that gender preferential patterns are disrupted when linguistic change is in progress. If the change is above the level of awareness, women prefer the standard variant; if below, women tend to show higher rates of the innovative variant than men, even if that form is non-standard. The latter tendency has been referred to as the ‘gender paradox’, whereby ‘[w]omen conform more closely than men to sociolinguistic norms that are overtly prescribed, but conform less than men when they are not’ (Labov Reference Labov2001: 293). However, with regard to the features of the regional French of Alsace examined here, the long history of proscription and high level of overt awareness is somewhat at odds with this theory. While our principal findings relating to the attrition and loss of traditional regional features in apparent time emerge clearly, the results concerning female–male differences (and the lack of them, where these are not statistically significant) require further investigation, in order better to understand the nature of the interaction between speaker gender and processes of supralocalization in regional varieties of metropolitan French.