1. Variation and Change in the French of France

This section examines in detail some recent developments in the pronunciation of the French of France, based to some extent on studies adopting a number-crunching or variationist viewpoint. Subsequent sections consider the varieties of French spoken in Belgium and Switzerland; we may note in passing the seeming gap in the literature on Luxembourg French. To provide a recent historical perspective, it may be useful to begin by mentioning the work of L. Milroy (1987) in Belfast, which had a narrower focus than the broader categories of gender, social class, age, ethnicity, and region. The research was prompted by the wish to know why speakers in cohesive social networks resisted the pervasive pressure of the standard, in view of the advantages accruing to those who controlled the dominant variety of the language. Given the leveling trends observed over the last few decades, the question posed by Milroy now seems less compelling, or at least it seems to require a new focus. This article reviews some research highlighting what seem to be local pockets of resistance to national leveling processes. In France, these networks can provide conditions in which vernacular features avoided by the majority of speakers are maintained.

When one attempts to assess the French situation, the nature of the documented evidence means that a strict paralleling of trends with other languages is problematic. Although early work on U.S. and U.K. English focused on behavior buttressed to some extent by speaker perceptions, as exemplified by work by Trudgill (1974) on over- and underreporting, studies on French have by contrast often dealt with variation from a perceptual optic—notably, in Martinet (1945), but also in more recent work such as that of Von Nolcken (2002), Eloy et al. (2003), Bulot (2005), Kuiper (2005), Woehrling and Boula de Mareüil (2006), Scherrer et al. (2015), and Boula de Mareüil et al. (2017). This is to some degree complemented by two major phonological overviews, that of Walter (1982) and the body of evidence still being accumulated by the large-scale Phonologie du Français Contemporain (PFC) project (cf. Durand et al., 2009). The relatively limited sociolinguistic perspective of the PFC protocol has proved capable of extension by reference to fuller and more detailed social information, as in Hambye (2005) for Belgium.

Leveling, a notion often referred to, is defined by Williams and Kerswill (1999, p. 149) as “a process whereby differences between regional varieties are reduced, features which make varieties distinctive disappear, and new features emerge and are adopted by speakers over a wide geographical area.” For some scholars (e.g., Trudgill, quoted by Watt, 2002, p. 59), leveling can refer to two conceptually separate processes, capable of occurring independently of each other: (a) koinéisation, equivalent to the German term Nivellierung ‘the outcome of dialect mixing resulting from in-migration, characterized by a reduction in the number of forms available in the mixture’, and (b) Dialektausgleich ‘whereby dialect areas enlarge, and dialect diversity is lost to a certain extent’. Leveling generally leads to convergence in the neutral sense that growing numbers of speakers are differentiated by fewer and fewer features. For France, Armstrong and Pooley (2010) argue that the term advergence, the convergence toward a ‘prestige’ norm, seems apt, whereas this is much less true in the United Kingdom, for example. The results reviewed here point to a more complex situation, as the pronunciation of ‘hexagonal’ or mainland French has been affected by both leveling and standardization (Kerswill, 2001). In France, it may be claimed that there is a significant degree of advergence toward features of a variety generally referred to as ‘Reference French’ (RF) (cf. Lyche, 2009, for a discussion), the reference group in question being middle-class Parisians. The next section, in turn, examines variation in French consonants and vowels.

2. Macrolevel Changes in the French of France

2.2 Vowels

Figure 1 shows the traditional inventory of vowels of RF, corresponding to what the majority of Martinet’s Parisian, non-southern, and mobile informants claimed to use and to what Müller (1985) called the ‘prescriptive norm’. It contains the set of distinctions held up until fairly recently as the model pronunciation for L2 learners of French, at least in the transcriptions of bilingual dictionaries. It is quite a large inventory, comprising 12 oral and 4 nasal vowels, and mergers in colloquial styles between the midvowels and the two reflexes of the /a/ phoneme, realization of schwa as a front rounded midvowel, and compression of the prescriptive four-vowel nasal system to three result in a much reduced system compared to the more formal maximal one. This is again in contrast to languages having variation manifested through greater or lesser degrees of regional marking.

Open in new tab

Figure 1. The vowels of Reference French.

For a significant minority of the subjects of Martinet (1945), particularly in eastern France, including Burgundy, the major difference from contemporary RF was that they self-reported as differentiating certain otherwise homophonous pairs by vowel-length distinctions, many of which had forms with or without word-final <e> (e.g., si [si] and scie [siː]) although not all (e.g., [bɛl] belle and [bɛːl] bêle). Some speakers, particularly in the North, claimed not to distinguish /a/ from /ɑ/, and many more (Paris, Centre, Burgundy, North, Normandy, and West) claimed to merge /ɛ᷉/ and /œ̃/. More recent work (e.g., Lyche, 2009) admits as RF what earlier scholars such as Müller (1985) had called the norme d’usage ‘statistical norm’. This notion evokes the most frequently used forms, an inventory reduced by the mergers of /a/ and /ɑ/ and /ɛ᷉/ and /œ̃/, where /a/ and /ɛ᷉/ have taken over in almost all contexts. Hansen (1998) has a very comprehensive account of variation in the nasals in Parisian French. The phonetically different realizations of the midvowels are increasingly in complementary distribution, with the higher variants occurring in open syllables and the lower in closed. The front pair /ø/ and /œ/ not only differentiate very few minimal pairs, but their complementary distribution in pairs like peu [pø] and peur [pœʁ] is no longer quite so systematic given the growing acceptability of forms like [pøʁ] which were at one time emphatically a feature of vernacular speech.

In many cases, /e/ and /ɛ/ and /o/ and /ɔ/ are clearly in complementary distribution, with the higher vowel of each pair being used in open syllables and the lower in closed: the so-called loi de position. There are, however, a number of potential minimal (or near-minimal) pairs differentiated by /e/-/ɛ/ in open syllables such as ballet [bale] and balai [balɛ] and similarly épée-épais, poignée-poignet and /o/-/ɔ/ in closed syllables as in Pol [pɔl] and pôle [pol] and similarly cotte-côte, sol-saule.¹ The /e/-/ɛ/ distinction is certainly maintained by some speakers (cf. its crucial role in Gueunier et al., 1978) to differentiate grammatical endings such as aimer, aimé, and aimerai, which have the higher vowel, and aimais, aimai(en), and aimerai(en)t, which, at least in traditional prescriptive accounts, take the lower variant. For the majority of speakers too, /ɛ᷉/ and /œ̃/ have merged, or perhaps more accurately /ɛ̃/ has taken over much of the lexical space occupied by /œ̃/, although the frequency of the one item un may mean that some speakers, like the subject of Østby (2007), maintain the rounded variant in that context. As two of the distinctions, /ɛ᷉/-/œ̃/ and /œ/-/ø/, permit the differentiation of very few minimal pairs, it is unsurprising that merged forms are now widely acceptable. One is inclined to claim that this is also true of /e/-/ɛ/ in lexical items but to remain slightly more guarded with regard to verb inflexions, although many speakers follow the so-called loi de position and use the higher vowel for all first-group endings. Concerning /o/-/ɔ/, the distinction is more generally maintained despite its relatively limited productivity and the obvious effects of supralocalization. This pair seems to be undergoing a phonetic change, fronting toward [ø], as in Martinet’s classic phrase c’est jeuli le Mareuc, and substantiated with instrumentally analyzed behavioral data gathered in Roanne and reported by Armstrong and Low (2008). Most of these changes may be classified as coming from below (i.e., middle-class speakers adopting what used to be considered vernacular forms), relegating to more highly monitored styles the maintenance of both variants, either as phonologically distinctive or, as is more often the case, as phonetically different and socially significant. For instance, back /a/ has been shown to be a regional and/or vernacular variant in Paris (Jamin, 2005), Lille (Pooley, 2001), and Brittany (Guézennec, 2003), whereas the distinction is maintained in Normandy (Hall, 2008).

In addition to the fronting of /o/, the most striking innovation is the emergence of prepausal schwa, which appears to be emanating from Paris (Hansen & Mosegaard Hansen, 2003). Whereas the most quoted example, bonjour-euh, illustrates a case of intrusive schwa, the vowel often corresponds to orthographic <e>, as shown in italics in example (1):

(1)

Prepausal schwa appears at present to convey certain pragmatic functions, and it is too early to tell whether its presently embryonic use will increase to take on social indexing. Schwa in other phonetic contexts (more accurately mute-e, given its usual realization as a front rounded midvowel) appears to have little indexical value in differentiating social groups but has been reported as distinguishing speech situations very sharply, in particular spontaneous and scripted styles (Hansen, 2000, discussed in Armstrong, 2007).

Most of the phenomena described in this section may be classified as examples of ‘downward’ leveling, the adoption of vernacular forms into standardized speech, in particular /o/ fronting. In the cases of /a/ replacing /ɑ/ and the generalized use of /e/ in lexical items and the merger of /ø/ and /œ/, one could argue that the prevailing forms are socially neutral and may thus be positively perceived and at the same time widely adopted. Perceptual studies further substantiate the emergence of this leveled supralocal variety. As early as the 1940s, the responses of Martinet’s informants suggested that the majority of subjects from all non-southern areas, as well as geographically mobile ones, could not be differentiated. He noted too the lack of strong differentiation between subjects from Paris and those from the immediately contiguous Centre-Nord. More recent perceptual studies (Armstrong & Boughton, 1999; Woehrling & Boula de Mareüil, 2006) show that ordinary speakers are unable to locate the origins of people from northern France, construed as the vast area that constitutes the greater part of francophone Europe in terms of both territory and population. In Armstrong and Boughton’s study, informants from Rennes proved incapable of identifying speakers from Nancy, a sizable town some 600 kilometers distant.

3. A New Emerging Multiethnic Vernacular?

The French banlieues or high-rise estates, inhabited by many French citizens of second- and third-generation Maghrebi origin, are undoubtedly a source of linguistic innovation in youth slang, many new items being manifestly ‘ethnic’ with sources in dialectal Arabic, Berber, Ro, and various sub-Saharan African languages. There is also considerable inventiveness in the use of French elements, both supralocal and markedly vernacular, as well as of regional languages and English. In pronunciation, it is, however, harder to claim segmental innovations.

Table 3 compares a number of the marked features of the speech of a 22-year-old delivery driver from the northern suburbs of Paris recorded in the 1970s (Carton et al., 1983, p. 84) with the speech of informants in the 15–25 and 30+ age range in the multiethnic community of La Courneuve, northeast of Paris, in the late 1990s (Jamin, 2005). Arguably, the three most important features are 7, 8, and 9, which show strongly in both corpora. The lexical distribution of back /a/ may be considerably different, but that is also the case in Normandy and the Nord-Pas-de-Calais. Two of the three features (10 and 12) listed in part (c) (late 1990s only) appear to be becoming increasingly acceptable in mainstream varieties, which suggests another possible change from below. The closing of /ɛ/ toward /e/ ((11)) is a widespread vernacular feature observed in other regions, such as the Nord.

One feature strongly alleged to be an ethnic feature is the pharyngealized realization of /r/, which is more used by speakers of Maghrebi and non-Meridional French origin in the La Courneuve banlieue (Jamin, 2005) and Perpignan (Pickles, 2001). As Pickles’s informants were not particularly disadvantaged, it would be interesting to gather further comparisons in other cities.

Jamin (2005) also found that the palatalization/affrication of voiceless stops showed a similar social distribution across the ethnicities, particularly among his younger informants still immersed in street culture. Comparison with Trimaille’s findings in Grenoble and those of Gasquet-Cyrus in Marseille with regard to palatalization/affrication led to the plausible hypothesis of intercity convergence among multiethnic youth in their divergence from supralocal forms (Jamin et al., 2006). Trimaille’s studies of young people of Maghrebian extraction in Grenoble showed extremely heavy use of palatalization, and Gasquet-Cyrus found that palatalization in Marseille was strongly associated with the Quartiers Nord where many people of migrant origin live. More recent work by Trimaille (2010) has, however, shown that mainstream speakers (e.g., politicians in formal situations) make considerable use of palatalization, which points to its increasing acceptability in standard usage. A recent volume by Fagyal (2010) provides convincing evidence of the influence of Maghreb Arabic on the French of young banlieue speakers but none on whether outward-looking speakers are exporting some of these features into mainstream French.

4. Overview of Non-Southern French

Open in new tab

Figure 2. Documented spread of supralocal French.

Source: Pooley and Armstrong (2010). Reproduced with permission from Palgrave Macmillan.

The overview was intended to bring some light and shade to the picture of the geographical spread and phonological uniformity of the variety, or set of varieties, describable as leveled or supralocal French. The implication is that in France, over an unusually large area in European terms at least, the usual colloquial speech form used by the vast majority of French citizens born since 1965 contains no regional element in its phonology that is readily recognizable, at least by ordinary speaker-hearers. Figure 2shows the (mainly PFC) research locations that bear out this claim. This supralocal variety has spread beyond the historic boundaries of the Oïl and Francoprovençal areas into the northern Oc regions (e.g., Puy-de-Dôme) (Potte, 1977), leaving a number of peripheral regions with speakers, particularly from the lower social classes, who diverge from this norm only in a small number of features.²

The salience of divergent features currently used by speakers born since 1965 varies from region to region, and variable awareness appears to correlate with social-class differences. Awareness seems not to go beyond a fairly immediate radius, and representations of vernacular speech within larger regions, such as Normandy and the Nord-Pas-de-Calais, show a considerable time lag in relation to recently documented behavior, such that patois or marked regional accents are believed to show greater vitality than they actually do. Apparent-time indicators point unmistakably to a reduced number of distinctive features, few of which are exclusive to any given region. A structurally unusual combination or lexicophonological distribution of nonunique features may produce a perceptually striking effect, as in the case of backed variants of /a/ in the Nord-Pas-de-Calais compared to Paris vernacular or Normandy. Whether this effect is capable of proving indexical in the usual sociolinguistic sense is a different matter. The variants showing vitality are or were heard in Parisian vernacular and appear to conform to a ‘city-hopping’ model, reaching Lille before the Pas-de-Calais, Rouen before the Cotentin, and no doubt Quimper before the Île-de-Sein. In some instances, up-leveling and down-leveling pressures coincide, as in the case of (e) in Normandy, or may be sociolinguistically ambiguous as in the case of back /a/, which is a socially split variable characteristic of old-fashioned Standard French, Northern, and, with increasing exclusivity, vernacular French but overwhelmingly absent in middle-class usage.³

5. Southern French

Open in new tab

Figure 3. Southern France showing localities investigated in a sociolinguistic perspective.

Source: Pooley and Armstrong (2010). Reproduced with permission from Palgrave Macmillan.

Figure 3 sketches the area where accents readily identifiable as southern may commonly be heard and have been documented in fairly recent phonological and sociolinguistic studies. It comprises the départements abutting the Mediterranean coast and Pyrenean borderlands and the immediately contiguous inland regions. The map indicates all localities studied using comparable approaches, including Brun (1931) for Marseille and Séguy (1951) for Toulouse, as well as the home areas of the subjects in Walter (1982) and Carton et al. (1983). Behavioral sociolinguistic work covers the Béarn (Diller, 1978), Fos-sur-Mer (Chauvin, 1985), Pézenas (Durand et al., 1987), Aix-en-Provence (Taylor, 1996), Carcassonne and Lézignan-Corbières (Armstrong & Unsworth, 1999), and Perpignan (Pickles, 2001). Researchers working in the PFC project have studied Saint-Jean-Pied-de-Port in the Basque Country (Aurnague & Durand, 2003), Douzens (Durand & Tarrier, 2003), Rodez and Salles-Curan in the Tarn (Pustka, 2007; Sobotta, 2003), and Lacaune (Meisenberg, 2003). Also of interest are the perceptual studies on Salses and Sigean (Wanner, 1993), Provence (Kuiper, 2005), and Marseille (Binisti & Gasquet-Cyrus, 2003; Woehrling & Boula de Mareüil, 2006).

This wide range of localities covering several of the more southerly subregions contrasts markedly with Martinet’s two-part division of the region into the Midi, defined as the traditional Oc region excluding a comparatively small area which he called the southwest (the Bordelais and the Périgord), which in terms of claimed pronunciation was southern rather than northern but considerably less so than the Midi. Thus 58% of his southwestern informants claimed to pronounce word-final schwa in laque compared to 90% in the Midi and 21% in the nonmeridional areas. There are grounds for supposing considerable overreporting by these informants of non-southern features, as these claims are largely undermined by the available behavioral evidence. Most of the traditional contrasts of the prescriptive norm were claimed by a largish minority: (41%) for /a/-/ɑ/ (patte-pâte) and a smallish majority (53%) for /o/-/ɔ/ (saute-sotte) and /e/-/ɛ/ piqué-piquait (60%). Other claims inspire greater confidence, in particular the maintenance of the /œ̃/-/ɛ᷉ contrast (89%) and the use of a nasal consonant appendage, as in chanter [ʃante] (58%).

While descriptions of marked regional varieties spoken by those born between the late 1800s and the mid-20th century suggest subregional differences, work on the speech of informants born since 1965 persuade readers to follow Martinet and consider the Midi a single region. This approach can be justified for three principal reasons: first, the difficulty experienced by outsiders identifying the precise geographical origins of southern speakers. Thus a recent study carried out using PFC data (Woehrling & Boula de Mareüil, 2006) suggests very strongly that while southern accents of various degrees of broadness are consistently recognized as southern in contrast to northern (Normandy and Vendée) and Swiss (Vaud), subregional varieties could not be readily distinguished by a panel of young well-educated Parisian judges. This was despite the fact that they were asked to choose between three specifically named southern accents (Basque Country, Languedoc, and Provence) exemplified by PFC informants of all ages, including the very oldest speakers, who retained some of the most locally marked features already discussed (e.g., the use of initial /h/ by a speaker from the Basque Country). Second, one particular variable has probably been studied more than all the others combined, namely, the variable retention of mute-e. Despite the wide range of research sites, attempts to localize the various phonetic realizations of the variant have been largely limited to anecdotal observations, of less precision than the observations of Brun in interwar Marseille. For instance, Coquillon (1997, cited in Durand & Tarrier, 2003) suggests that schwa tends to be longer in Toulouse than in Marseille (Brun, 1931, § 2.7) and Durand and Tarrier (2003) believe that Toulousain French is more ‘conservative’ in this respect. Such comments have never been tested using variationist techniques, which have been applied only to the retention or deletion of mute-e in various contexts, which, as Armstrong and Unsworth (1999) point out, is socially significant.⁴ Third, few if any potentially salient subregional differences have been investigated, presumably because they are highly recessive. For instance, the regional differences in the consonantal element for nasal vowels in word-final prepausal position (bilabial in Languedoc, velar in the southwest and Provence) noted by Carton et al. (1983) show strong signs of leveling toward the velar in more recent work.

The review of behaviorally based studies of Southern French will therefore be based on the salient pan-Midi variables: (a) mute-e, (b) nasal vowels, (c) oral vowels, and (d) consonants. This survey is complemented by some observations on perceptions.

A number of studies confirm the continuing variable retention of mute-e, but in every case, frequencies of occurrence are reported as decreasing in correlation with a variety of social factors. First, speakers with higher levels of education have lower realization rates than their less well-educated counterparts, as shown in a study by Diller (1978) of older speakers (born ca. 1905–1920) from a small village in the Béarn. Second, speakers from the higher social classes delete mute-e more than those from the lower strata (Chauvin, 1985; Taylor, 1996). Among 11- and 12-year-olds in Fos-sur-Mer and Martigues, Chauvin noted the effect in tension of social class and geographical mobility. While locally born middle-class children tended to have the supralocal pattern of mute-e deletion, working-class subjects born outside Provence had the local, heavily mute-e pronouncing pattern. Detailed research by Taylor (1996) in Aix-en-Provence found use of marked local mute-e variants (e.g., [ɒ] and [ɐ]) to correlate broadly with retention rates. Speakers showing the strongest retention rates and using the broadest regional variants were overwhelmingly older manual workers with the fewest educational qualifications and were often able to speak Provençal, from which some of the backed and rounded realizations derive. Third, there are clear indications that mute-e realizations are diminishing down the generations, in parallel with the decreasing use of ancestral languages: A study by Durand et al. (1987) of the small Languedoc town of Pézenas found decreasing mute-e retention among four generations of female speakers from the same family. Pickles’s (2001) work on Perpignan reports far lower rates of realization among his teenage informants born in the mid-1980s compared to a speaker born in the early 1960s and used as a model of a local accent. Studies by Sobotta (2003) and Pustka (2007) of three groups of Aveyronnais found retention to be greatest among the over 60s subjects who were fluent in Occitan.⁵ Pustka’s study draws attention to two other factors, the first of which might be called the town-village divide. Realizations of mute-e were more frequent in the small village of Salles-Curan (pop. 1,100) than in the larger town of Rodez (town 23,000, urban area 61,000). Other PFC research suggests that even the inhabitants of tiny villages are significantly affected by the leveling factors, as is the case of Douzens (Aude) (Durand & Tarrier, 2003), where the most markedly regional consonant variants also appear recessive. Some investigations suggest greater stability. Aurnague and Durand (2003) suggest that retention rates in Saint-Jean-Pied-de-Port in the Basque Country are comparable with those reported by Walter (1982) for her informant from Hasparren, but the apparent stability perhaps takes insufficient account of demographic differentials: Hasparren (pop. 5,303) is a small town and Saint-Jean-Pied-de-Port (pop. 1,417), rather, a village.

Pustka’s work also underscores the importance of migration, by showing that Paris-based Aveyronnais delete mute-e more than their counterparts from Rodez or Salles-Curan. Pickles notes a considerable difference in realization between adolescents of local origin and others from a different ethnic or geographical background (Catalonia, Maghreb, and non-southern France). It is unclear whether the nonassimilation of non-southern speakers in Perpignan, compared to Fos-sur-Mer/Martigues a generation earlier, is an indicator of ongoing leveling or whether Pickles’s subjects were from the median classes.

A study by Armstrong and Unsworth (1999) of students in Aude identifies two other factors: gender and regional loyalty. This study identified four contexts, which do not fully correspond to those defined by Martinet. Students from two towns in Aude were investigated: Lézignan-Corbières (pop. around 8,000) and Carcassonne (pop. around 44,000). A clear pattern for schwa retention in the three selected contexts emerges, as can be seen in Table 4.

Girls delete schwa considerably more than boys do in every context studied, and some differences displayed statistical significance. By comparison, social-class differences are not significant. This pattern of females leading change in the direction of supralocal (but not necessarily prestige or standard) norms echoes findings in the United Kingdom, for instance, in /t/-glottalling in Tyneside (Milroy et al., 1994) and Cardiff (Mees & Collins, 1999). The gender differences overlap significantly with an index of regional attachment, based on a series of five questions with possible scores ranging from 0 to 15. The female index scores were significantly higher than those of the males (7 compared to 3.75) whereas differences based on social class proved to be minimal. Pickles’s (2001) findings in Perpignan take on considerable interest in light of what is known to be occurring elsewhere in the south. His ‘local’ informants of the 37 from seven schools reveal similar deletion rates in phrase-final position to those shown for Lézignan-Corbières/Carcassonne. In word-final phrase-medial position, the overall rate in Perpignan is nearly identical to that of the Aude girls (around 63%), whereas the boys from the same area delete much less in that context.

Both Taylor (1996) and Armstrong and Unsworth (1999) refer to prepausal ‘schwa-tagging’ but do not see it as affecting the speech behavior of southern speakers at the time their corpora were gathered (1980s for Taylor; 1990s for Armstrong and Unsworth). Word-final prepausal position is not only the most favorable environment for schwa in the southern varieties but also the context where schwa-tagging occurs in Paris and other northern areas. Most commentators on southern varieties point out, however, that schwas generally correspond to orthographic <e>, whereas this is not necessarily the case for schwa-tagging, which may be intrusive as in bonjour [bɔ̃ʒuʁə]. Finding no examples of an intrusively realized schwa, Armstrong and Unsworth hypothesized that their informants may be shifting to a more conservative form of the supralocal variety.

PFC projects confirm the consistent realization of four distinct qualities of nasal vowels in Salles-Curan (Aveyron), Saint-Jean-Pied-de-Port (Basque Country), and Douzens (Aude). In all three locations, /a᷉/ is generally fronted toward /a᷉/, and in Douzens /ɛ,/ is raised toward [ẽ] (Durand & Tarrier, 2003). The bilabial and dental/alveolar consonantal elements, as in pain [pɛɛ᷉^m] or mettons [metɔɔ᷉ⁿ], noted by Carton et al. for the exemplar selected for the Languedoc from Gaillac (Tarn), appear to have been regionally leveled in favor of a velar, as noted by Meisenburg (2003) in Lacaune in the same département and in Douzens (Durand & Tarrier, 2003) and Rodez and Salles-Curan (Pustka, 2007).

Taylor (1996) presents a detailed analysis of Aix-en-Provence nasal vowels, focusing on degree of nasalization and the presence and character of the nasal appendage. Table 5 shows that canonical nasals in the very broadest regional accents may occur without nasalization or any consonantal element at all.

Marked regional accents are or were characterized by oral vowel + full nasal consonant sequences, while nasals in what Taylor calls the regional standard generally have a weaker consonantal element, often a consonantal appendage, with possibly a degree of vowel nasalization. Indeed, a fully nasalized vowel followed by a slight offglide, she claims, is or was more highly regarded than the fully nasalized (‘pointu’) vowel of standard varieties. The broadest regional forms are again used mainly by the least-well-educated speakers and those in manual occupations. Subjects who had a baccalaureate might use regional standard forms, and those with higher qualifications tend toward supralocal norms. The most highly supralocal forms were associated with business executives having the most contacts outside the region, particularly in Paris. Older speakers (born 1905–1920) tended to use the broader regional features the most, while use of the velar nasal proved to be a highly masculine feature. Other regional identities (e.g., Provençal, pays d’Aix, and the traditional maritime variety of Marseille), although sensitive to gender differences, also correlated with the broader regional varieties.

The nasal vowels are given a different treatment by Pickles (2001), whose analysis focuses on two parameters: presence or absence of a nonnasalized vowel followed by a nasal consonant and ‘nasal-vowel modification’, particularly the realization of /a᷉/ as [ɛ᷉], as in [perpinjɛ᷉] Perpignan. His data suggest that the vowels occurring most frequently in the corpus are most likely to be realized by localized V+N sequences. The modified nasal, however, proved to be a highly marginal variant with a mean of 3% for both locals and Maghrebians with a range of 0%–8%. Regrettably, given that the two regional variants discussed do not cover anything like the majority of cases, the study does not make it entirely clear which forms the school pupils used most of the time.

With regard to oral vowels, four of the shibboleth minimal pairs of the prescriptive norm are generally not realized (e.g., Durand & Tarrier, 2003, p. 119): Expressed in archiphonemic notation, /A/ is generally front, /E/ generally high, /œ/ generally high, and /O/ generally low. Several researchers (Taylor, Meisenburg, Durand and Tarrier, and Aurnague and Durand,) note the pertinence of the loi de position, although the phonetic difference between the vowels is less marked than in supralocal French, and a number of exceptions have been noted. The most problematic context is what Taylor calls pseudo-closed syllables, those which would be closed in supralocal French but in southern varieties would be followed by a syllable whose nucleus is mute-e, as in mère, belle. For this context Taylor tentatively claims variability. Durand and Tarrier (2003, p. 119) suggest that the realization of schwa may cause syllables of this type to behave as if closed, as in fête [fɛtə], paume [pɔmə], jeûne [ʒœnə], and bêtement [bɛtəmã]. Tables 6 and 7 show findings by Taylor (1996, pp. 103ff) regarding the variable phonetic characteristics of the front and back midvowels /e/-/ɛ/ and /o/-/ɔ/ according to socioregional variety.

For the E archiphoneme, the broadest regional accents are characterized by close variants (which may be as close as [e͎]), which many speakers use categorically. A slightly more open variant, which Taylor calls ‘emphatic’, as in c’est vrai! [sevʁe͎], occurs in varieties classed as regional, if not of the broadest type. Variable use of [ɛ], on the other hand, occurs only in the most standardized forms. In nonfinal closed and open syllables, Taylor (1996, p. 68) speaks of ‘free variation’, although certain subsets, for example, items containing the clusters /ks/ and /ɡz/ like excuse and exemple, were consistently pronounced with close variants. In word-final closed syllables, the raised variant [e͎] does occur and open variants are variable in regionally marked accents. For back-midvowels in nonfinal open syllables, [o] is clearly the dominant form (Taylor, 1996, pp. 72–73), but some influence of the supralocal variety can be seen, since items that have /o/, such as beauté, are realized with a greater proportion of close variants overall (89.6%) than items like botter, which has /ɔ/ in mainstream varieties (75.5%). In word-final closed (and pseudo-closed) syllables, open realizations of [ɔ] or closer ([ɒ]) were used in 90% of cases overall in contexts where /ɔ/ would occur in standard varieties (e.g., robe). In items where standardized varieties have [o], such as aube, 81.3% of realizations were more open than standard /o/.

The most marked regional variant, [ɒ], is used across a considerable part of the community in the aube set but only in the broadest regional accents in the robe set. In the more standardized varieties, the distribution of the variants is close to that of more regional ones, although the close vowel is variably lowered somewhat ([o̞]) in relation to the former. Curiously, Taylor’s data point to hypercorrections in the robe [ʁob(ə)] but not in the aube set. In items containing syllables closed by /z/, like chose, in which only the close variant occurs in standardized varieties, the main variants were [ʃɒz(ə)] (56%) and [ʃɔz(ə)] (34%), and again the broadest regional variants were associated with males, manual workers, and Provençal speakers. The hypercorrect forms mostly correlated with claims of a town identity and usage by women.

The consonantal features identified as southern by Martinet (1945) do not figure very prominently in recent sociolinguistic work, often meriting only cursory mention. Word-final consonant devoicing was observed in a minority of cases among a group of middle-class informants from Bordeaux (Temple, 2001, pp. 157–158). Apical /r/ was consistently used only by the oldest speakers in Douzens (Durand & Tarrier, 2003), Salles-Curan (Pustka, 2007), and Saint-Jean-de-Port (results reported by Aurnague & Durand, 2003), with middle-aged and younger speakers using only uvular variants. A strongly fricative variant of /r/ is noted in Perpignan (Pickles, 2001) This variant is shown to be a strong minority feature, with ‘locals’ showing 0% to 18% usage and the heaviest user, a Maghrebian, scoring 42%.

Use of the palatal nasal /ɲ/ is also losing ground to the sequence [nj], for example, in Douzens (Durand & Tarrier, 2003), although its (possibly imminent) demise may simply parallel a change that has occurred in supralocal French. Hiatus in items like nier and lion are noted in PFC studies, particularly in the speech of the oldest informants, in Lacaune (Lonnemann & Meisenburg, 2006), Saint-Jean-Pied-de-Port, and Douzens (Durand & Tarrier, 2003). The glottal fricative is much rarer, with only a single occurrence from the oldest informant from Saint-Jean-Pied-de-Port in a reading passage, in the item le hasard [løhazar].

The lateral fricative, noted by Carton et al. (1983) and Walter (1982), but not by Martinet, occurs in the usage of three (older) subjects in Saint-Jean-Pied-de-Port in both reading and conversational styles, distinguishing pairs like étriller [etriʎe] and étrier [etrije]. Unsurprisingly, the younger generations of this family do not use it.

High levels of migration to the south are no doubt an important factor in promoting leveling, but it may also be true that locally born young people are also converging toward supralocal norms. A study by Wanner (1993) of two small villages in the traditionally Catalan- and Occitan-speaking areas (Salses, Pyrénées-Orientales, and Sigean, Aude) is highly suggestive in this respect. In response to the question, ‘Le français tel qu’on le parle dans le Languedoc-Roussillon est-il différent du français parlé à Paris?’, the highest proportion of positive responses came from the oldest respondents. This corresponds to their own linguistic practices, for, in Wanner’s words, ‘Déjà les jeunes parlent souvent un français parisien presque parfait’ (1993, p. 81). Kuiper (2005, p. 46) offers similar observations for his younger Provençal informants. These testimonies suggest strongly that among the generations born in the 1980s, convergence toward supralocal norms has moved on apace, with middle-class and more highly educated young adults leading the way.

6. Summary of the Situation in Southern France

What one might call the dominant southern pattern (DSP) shows considerable vitality, although there are many indications that it is starting to recede, as younger and better-educated speakers increasingly converge toward using supralocal norms. This tendency is further promoted by many northerners moving south and showing little inclination to adopt southern norms, particularly in the middle or higher strata of society. However, it may be that working-class youngsters are more likely to be assimilated to the DSP (Chauvin, 1985), although new large-scale industries attracting large numbers of people from other parts of France are rare. The DSP includes only vocalic features, with the possible exception of the velar nasal element in nasal vowels, and highly recessive consonantal features are heard only in the speech of the elderly, mostly living in small villages.

Linguists have made much of the southern oral-vowel distribution, and the truism that most speakers conform variably to the loi de position, with close variants in greater use than in mainstream varieties, is very largely borne out by the evidence. Some traditional varieties, like the Marseille speech described by Brun (1931) and exemplified by the Provence speaker in Carton et al. (1983), continue to show a degree of divergence, but the vitality of the stereotypical ‘real Marseillais’ features (Binisti & Gasquet-Cyrus, 2003) is hardly confirmed by a study now three-quarters of a century old. Indeed, Brun noted a considerable degree of continuing convergence toward the supralocal variety, an observation that can only be reinforced by more recent evidence.

The dominant nasal-vowel variants seem to be slightly more weakly nasalized than is usual in supralocal French, possibly with a slight nasal offglide, which is generally velar even in rural areas, where bilabial and dental variants were noted a few decades earlier.

Southerners continue to pronounce more schwas than those in the north, although they are doing so to a lesser degree than in the past, with middle-class, younger, and female speakers lead the charge. Although some scholars have hinted at subregional differences in the pronunciation of schwa, such detailed indications as are available (e.g., Taylor, 1996) suggest that local variants are indicative only of the broadest accents. It should be borne in mind that the realization of schwa as a rounded central midvowel [œ] or [ø] may well be a feature shared by most varieties of French.

A study by Martinet (1945) of middle-class men born between 1880 and 1920 showed that at that time, many southerners perceived their speech to be closer to the prescriptive norm than the DSP on many points, like the /a/-/A/ distinction (31% or 44%) and realization of nasal vowels without nasal consonant (42%). Perceived deletion rates for schwa were low, but the study showed variability. In contrast to this considerably converged vowel system, Martinet’s informants professed high use of regional consonant variants: apical and geminate /r/ (46%) and hiatus in words with standard yod (72%), which recent studies have shown to be relic features. It is tempting to draw a parallel with what has occurred in northern England over the past few decades. Supralocal consonantal features like th-fronting have spread to the whole of England (Britain, 2005) while generalized northern vocalic features are displacing more localized variants (Watt & Milroy, 1999). If convergence toward supralocal norms affects consonants first, as the description of Toulouse speech by Séguy (1951) already suggested more than half a century ago, the dominance of the DSP in the middle decades of the 20th century (Séguy, 1951; Walter, 1982) would point to the equally plausible inference that a broad regional norm had emerged in the vowel system—for example, speakers may be recognizably southern but not recognizably Provençal, Gascon, or Languedocien for most outsiders (Woehrling & Boula de Mareüil, 2006). This is in line with one aspect of the leveling phenomenon examined here.

7. Belgian French

The territory of contemporary Belgium had been continuously in the hands of foreign powers until its declaration of independence in 1830: in modern times, Austria (1713–1792), France (1792–1815), and finally the Netherlands (1815–1830). French control ended after Waterloo (1815) and France was obliged to accept the borders of 1792, under Dutch rule. In 1830, a revolt was led by a French-speaking alliance (consisting of Catholic and Liberal L1 francophone elites and French-educated Flemings), who disapproved of William of Orange’s imposition of Dutch as the language of the state but more still of his attempts to impose Protestantism. Following France’s example, Belgium adopted a centralist francophone regime, initially de facto and later (1839) officially, when a small Brussels-based francophone elite held power. Within the first few years of the country’s existence, Flemish voices began to be raised against the inequity of this situation and demanded that Flemish should be recognized as a co-official language and used for official functions in Flanders. The equally uneasy contemporary situation shows a Flemish-speaking majority (some 60% of the population) in the now more prosperous north of the country, the Brussels enclave where more than 90% of inhabitants speak or claim to speak French, and francophone Wallonia in the south. This brief and incomplete historical sketch is designed to some extent to explain the unease felt by francophone Belgians over the Flemish threat; on the other hand, the hexagonal French standard remains a constant looming presence. These pressures might be thought to have the effect of galvanizing Belgian francophones to support a common cause, perhaps with linguistic consequences, although one must allow that a shared sense of belonging does not exclude regional variation in speech. Any such consequences would, however, have to undo centuries of differences between the two centers of influence, Brussels and Wallonia.

Bearing these factors in mind, this section summarizes the descriptions of the lone Walloon informant by Walter (1982) and four of the Walloon exemplars from Francard (1989), comparing their findings with the inventories of Belgian vernacular features: Grégoire (1956), Remacle (1969), Baetens-Beardsmore (1971, 1979), Piron (1979, 1985), Pohl (1983; 1985, 1986), Reuse (1987), Warnant (1997), Ball (1997), Klinkenberg (1999, 2000a, 2000b), Pöll (2001), Francard (2001), and Hambye et al. (2003). Most of the listings (exceptions to this are Grégoire, whose aim is to correct ‘les vices de la parole’, Baetens-Beardsmore, who focuses on Brussels, and Reuse, who focuses on Charleroi) are descriptive in intent and attempt to cover the whole of francophone Belgium, perhaps encouraging a more unitary perspective than prevailing discourses relating to regional identity might suggest. Figure 4shows the Belgian dialects of the Walloon region (Table 8).

Open in new tab

Figure 4. The traditional Belgian dialect regions of Wallonia.

Source: Lechanteur (1997, p. 85); Rossillon (1995), in Pooley and Armstrong (2010). Reproduced with permission from Palgrave Macmillan.

Walter’s informant, MLL, from la Roche-en-Ardenne, born in 1923, was a French-Walloon bilingual whose parents spoke only Walloon, although they understood French. The daughter of a farmer, married to a carpenter, she had received limited secondary education and had followed her mother by working from home as a seamstress and later as a hotel chambermaid.

The non-supralocal French features are listed in the table in descending order of likelihood of occurrence, and also, it might be added, of stereotypical perception. Word-final consonant devoicing may be exemplified by Belges pronounced /bɛlʃ/. Although some caution is required in interpreting the tabulation as an implicational hierarchy, given the possibility of chance gaps, comparison with the overviews listing Belgian regional and vernacular features suggests several things. For example, in general, items occurring in three or more of the lists shown in Table 3 can be considered characteristic of most Belgian vernacular varieties, whereas features occurring in only two columns may be considered regionally restricted, and the two features occurring in only one column appear to be more supralocal, but not generalized, in the speech of the selected subjects largely because of a number of lexical restrictions.

Some of these features, notably WFCD, the maintenance of the /a/-/ɑ/ and /œ̃/-/ɛ᷉/ contrasts, hiatus in items like scier, and the absence or restricted distribution of /ɥ/, are compatible with both prestige varieties and vernacular Brussels accents. In some cases, the features need to be more narrowly defined to show acknowledged regional or social differences. With regard to geographical restrictions, a velarized realization of back /a/ is likely to be heard in western Wallonia (Picard substrate) or Brussels. Weakened realizations of vowels in polysyllables, such as téléphone [tɛləfɔn], are characteristic of both Brussels and eastern Wallonia. The fronting of /a᷉/ to [a᷉] and the raising of /ɛ᷉/ to [ẽ] seem to be characteristic of eastern Wallonia (La Roche-en-Ardenne, La Calamine), as is the realization of [h] as in hêtre (La Roche-en-Ardenne, Verviers).

Hiatus carries much more social marking if reinforced by a glide, as in scier [sije], tuer [tywe], or louer [luwe]. Vowel lengthening can justly be described as old-fashioned standard if it marks masculine-feminine distinctions, as in fumée [fymeː] and amie [amiː], but not if reinforced by a glide as in [fymeːj] or [ami:j]. Cases where vowel lengthening occurs without phonemic contrast, as in [pʁɔblɛːm] problème, [aʁɛːn] arène, [ɛːl] aile, and [leːs] laisse, are generally construed as vernacular. Similarly, although lax realization of high vowels has been shown to be compatible with Brussels middle-class usage (Walter, 1982), lax realization of midvowels tends to be more marked (e.g., téléphone [tɛlœfɔn] and crapule ‘crook’ [kχapœl‎]), except in the case of common monosyllables such as les, mes, and ces pronounced, for example, [lɛ], where they can be interpreted as old-fashioned standard.

The neutralization of /o/-/ɔ/ may be taken as a sign of leveling rather than divergence from supralocal French, since historically the distinction was capable of occurring in open syllables, permitting differentiation of pot and peau and sot and seau. Not only do such pronunciations run counter to the loi de position, but other forms are also divergent, such as fosse [fɔs] or restaurant [ʁɛstoːʁa᷉ː], both of which can occur in generalized vernacular French.

Lengthening is widely noted for the nasal vowels but not the rather weak nasalization which seems to emerge from analysis of the exemplars in Francard (1989).

The traditional stigmatization of marked Walloon accents can be encapsulated in the term wallonner as characterized prescriptively by Remacle (1969, p. 70): “To describe the general way that Walloons speak French, we can say […] that they talk thickly and heavily” Pour caractériser en général la manière dont les Wallons prononcent le français, on peut dire qu’ils ‘wallonnent’, ou, en d’autres termes, qu’ils parlent avec une lourdeur pâteuse.” Remacle adds that the impression of a slow speech rate is largely conveyed by vowel lengthening, but this has been shown to be sociolinguistically complex, ranging from realizations compatible with the prestigious to the highly stigmatized. This is not to say that stigmatization is an automatic precursor to loss, as demonstrated by the study of Thiam (1995) of apical /r/ in the traditional mining district of the Borinage. One may add that a perception by outsiders of slow speech rate is generally the linguistic reflex of a negative social stereotype—for the French, Belgians are notoriously figures of fun. In reality, speech rates differ little across languages and varieties, as opposed to speech styles.

The distinction between prestigious and generalized vernacular Belgian features is far from clear-cut, as three mnemonic sequences show. Pohl (1985) proposed L’ourse brun pâle est enrouée [luʁsbʁœ̃paːlɛta᷉ʁueː] where WFCD and realizations such as [lwi] lui are not exemplified, in contrast to the model sentence in Francard, 2001, p. 266), Ces huit Belges n’aiment pas les sots [sewibɛlʃnɛmpalesɔ], which also features low-mid /ɔ/ in an open syllable alongside /ʒ/-devoicing. An earlier sequence used by Pohl (1983, p. 36), huit chaînes trop larges [wiʃɛ᷉ntʁɔlaʁʃ], widens the sociolinguistic spectrum further by adding nasalized vowels.

It is unfortunate that no very recent work sheds much light on variation in Brussels, the largely French-speaking capital with more than 1 million inhabitants, representing some 10% of the country’s population. A description by Walter (1982) of the speech of a Brussels informant contains few features that differ strikingly from the list given above. Table 9 summarizes more recent sites selected for study in a variationist perspective. These only partly reflect the regional distribution of accents, and they cover only a fairly narrow range of variables (six in total, if only those studied in conversational style by Bauvois, 2002a, 2002b, are taken into account).

In addition to this rather restricted linguistic range, only apical /r/ in the Borinage and vowel lengthening in penultimate syllables in Liège can be interpreted as regional features within Belgium, the others having social-class value (schwa, WFPOLD, and strongly articulated variants of /r/). WFCD is pan-Belgian but also occurs in Switzerland and certain regions of France, where its social meaning is radically different from that in Belgium. The concluding section provides an overview of the Belgian situation.

8. Swiss French

9. Concluding Remarks on Belgian and Swiss French

Thus it appears that both Belgian and Swiss francophones recognize the prestige of French norms (which they often perceive as Parisian) but take some pride, nevertheless, in at least some of their divergences from them. Both are, however, likely to modify their speech to supralocal norms in direct interaction with a French person, being no doubt aware of the attitudes expressed in the study by Kuiper (2005, pp. 38–42), in which Swiss and Belgian accents were placed 23rd and 24th below those of each of the 22 regions of France in regard to correctness and pleasantness as well as to degree of difference from the Parisian norm.

Within their own countries, the majority would, however, admit to a certain pride and pleasure in the use of forms associated with either national or more narrowly local identities. In both territories, it is the higher social classes, most exposed to interactional contact with French speakers, who are the least linguistically secure. The perceived greater degree of security among the lower ranks of society and older persons is undoubtedly due in part to a lower level of awareness of external norms connected to less mobility and in part to social pressures not to speak in a way that is not too ‘Parisian’. At the same time, dialectally marked speech is heavily stigmatized and avoided, even if, particularly in Belgium, the traditional endogenous languages are more highly valued than in the past.

Hambye (2008) provides a useful broad summary of the Belgian situation. His main title, “Convergences et Divergences,” refers to ongoing changes in practice as well as in attitudes toward standard French. Hambye suggests, on the basis of data drawn from actual behavior as well as the elicitation of the ‘linguistic imaginary’, that standard Belgian varieties are converging to each other as well as diverging from the French of the hexagon. The qualification ‘standard’ is important. From the linguistic detail presented here, a fairly complex pattern of continuing social differentiation emerges, expressed in part through allegro forms such as /l/ deletion in words like milieu. These seem not to be localized. At the same time, social-regional differentiation finds expression in localized forms like vowel lengthening, though this dimension of differentiation appears to be receding. Convergence to an endogenous norm, functioning in parallel with divergence from supralocal French, is signaled perhaps most notably by WFCD.

All this may be characterized as a form of leveling, in the sense used here. The sociohistorical and cultural factors that have shaped Belgium, and the linguistic expression found there, cannot be thought about without taking into account the French presence across the border. Nor can the importance of Flemish-Walloon tensions be overestimated. The result is ‘leveling’ construed in a rather negative or reactive sense—the obverse of the positive sense of identity just mentioned.

Perhaps the most striking distinctive feature of the Swiss situation is the wide reach in the broadcast media of ‘legitimized’ (to use Bourdieu’s term) or ‘supralocal discourses’, as Thibault (1998) in particular suggests. In Belgium, comparable public discourse on RTBF news (Belgium’s French-language public broadcaster) is more divergent from supralocal French norms, suggesting that a range of features is compatible with prestige pronunciation. This is despite the lack of agreement as to the existence, or even the recognition of, the specific features of the Belgian variety. Although in both countries, perceptual studies provide strong indications of regional differences, this is not borne out by recent behavioral evidence, at least in the case of Belgium, where a fair amount of data is available. It is possible that in Switzerland, concentration on the Vaud distorts the overall sociolinguistic picture, whereas in Belgium the fuller behavioral evidence points to the vitality and general lack of stigma attached to some non-supralocal forms shared by Walloons and Bruxellois, in particular the most studied feature, devoicing of final consonants.

10. Concluding Remarks

A major element that has been apparent throughout this article is the looming presence of the former Parisian variety of French. In France, the variety has swamped most others, with the unusual result that much variation appears to have been displaced to grammar and lexis. In Belgium and Switzerland, the supralocal norm plays a role often attested in other countries, that of the variety embodying the less personal and intimate aspects of social identity. It may be, therefore, that at the heart of all sociolinguistic variation lies a tension between the local and supralocal. It has been argued elsewhere (Armstrong & Mackenzie, 2018) that if variation due to local or regional affiliation is understood as being a theoretical primitive—a basic fact of human nature—the observable statistical relationships between language and global demographic variables such as gender or class may to a certain extent be secondary, surface by-products of relatively complex cultural interactions that have localized ideologies as their primary point of reference. From that perspective, a major task of the sociolinguist in any given study is to unpick the local–global nexus which underlies the relevant statistical correlations. And from the same perspective, the French situation is highly unusual, at least in phonology.