2.1. Transmission

Transmission is defined by Labov (Reference Labov2007) in terms of the ‘family tree model’. The continuity of dialects and languages is said to be the result of children's ability to learn the language of older generations (including structural and social constraints) and to reproduce faithfully this language (2007: 346):

It is this ‘unbroken sequence’ of native-language acquisition by children that Labov terms transmission. The faithful replication of older generations’ language, however, need not be total. In Labov's view, children are capable of preserving ‘linguistic descent’ even when replication is imperfect – when language changes (Reference Labov2007: 346). Changes which take place internally over time are said to be generated by the process of incrementation whereby ‘successive cohorts and generations of children advance the change beyond the level of their caretakers and role models, and in the same direction over many generations’ (Labov, Reference Labov2007: 346). Incrementation takes place when children associate variability in the language of previous generations with the vector of age, advancing changes further along the same trajectory.

Labov illustrates the mechanism of transmission with evidence from the distribution of tense and lax short /æ/ in New York City which is governed by a complex array of phonological, grammatical, stylistic and lexical constraints. Younger generations are shown to faithfully replicate this system in New York and to preserve the integrity of the system as a whole, although more recent analyses have shown that the system of the youngest informants is also suggestive of change from below, via incrementation with respect to the traditional following phonological environment constraint (Becker, 2010). Labov also analyses evidence for the NCS in Chicago, a clockwise chain shift of six vowels originally initiated by the tensing and raising of short /æ/. This chain shift is free of lexical and grammatical constraints, applying to all instances of the vowels across the linguistic system, but the pattern is structurally complex in that it involves the interrelated and dependent movement of all vowels within the vowel space. Within the cities, such as Chicago, where the NCS was initiated, younger generations are shown not only to faithfully replicate the new vowel system in its entirety via transmission but also to advance the NCS changes via incrementation. Successive generations have also been shown to replicate and preserve structural and social constraints on variation during the transmission of grammatical features in Canadian English (Tagliamonte and Denis, Reference Tagliamonte and Denis2014), providing additional evidence for the ability of children to reproduce adult patterns with a high degree of structural accuracy.

2.2. Diffusion

Within the family tree model, as presented above, subsequent branches of the family tree are normally seen to become more distant from each other (cf. the development of the Romance languages from Latin). Similarities between parallel branches of the family tree are usually the result of contact between the speech communities involved and of the transfer of features from one to the other (Labov, Reference Labov2007: 347).

Diffusion is defined by Labov as the transfer of linguistic features across the branches of the family tree (2007: 347). The process which permits such transfer, and which underpins the mechanism of linguistic diffusion, is speech accommodation (Giles et al., Reference Giles, Taylor and Bourhis1973), whereby, in face-to-face interactions between adults, speakers converge to, or diverge from, the variety of language spoken by their interlocutor in order to gain social favour or to distinguish themselves socially (Trudgill, Reference Trudgill1986: 2).

The process of speech accommodation and the resultant diffusion of linguistic features from one variety to another is presented by Labov as secondary to transmission in that diffusing features are seen to replace traditional dialectal features which have been passed down from parent to child: ‘They are the result of a secondary process in which speakers of one particular dialect gain an ascendancy – commercial, political, or cultural – and the ensuing expansion of this dialect wipes out the intermediate forms of the original continuum’ (Reference Labov2007: 347). This phenomenon, whereby linguistic features spread out from culturally and economically dominant centres such as New York City or Paris, is part of the process of ‘regional dialect levelling’ (Kerswill, Reference Kerswill, Britain and Cheshire2003) which is claimed to be ‘leading to the loss of localised features in urban and rural varieties of English in Britain, to be replaced with features found over a wider region’ (Reference Kerswill, Britain and Cheshire2003: 223).

The spread of linguistic features, via diffusion, across branches of the family tree is traditionally thought to follow a wave-like and/or urban hierarchical pattern within the field of dialect geography. Wave-like diffusion is characterised by the adoption of linguistic features radiating outward from a central focus by nearby locations before those at greater distances. This wave-like model is often modified by the likelihood that nearby towns and cities will adopt diffusing features before more rural areas in between – urban hierarchical diffusion (Kerswill, Reference Kerswill, Britain and Cheshire2003: 223). The hierarchical diffusion of linguistic features is modelled in Trudgill's (Reference Trudgill1986) ‘gravity’ model: linguistic innovations are said to be leaping or ‘parachuting’ according to a defined hierarchical pattern, beginning in the largest urban centre and spreading to rural areas via smaller and smaller ‘satellite’ towns (Reference Trudgill1986: 39). Evidence from dialectology provides records of both diffusion and transmission acting simultaneously. In the urban centre where the linguistic feature originates, language change may take place via the mechanism of transmission (and incrementation) as younger speakers advance the change over time. As the linguistic feature diffuses outwards across geographical (and social) space, the linguistic forms adopted are not necessarily identical to the original form. This means that linguistic features spreading to outlying areas may be adopted ‘off the shelf’ (Milroy, Reference Milroy, Cain and Russom2007) without faithful replication of associated structural and social constraints on variability:

Labov demonstrates the inferior language-learning abilities of adults with results from two studies of geographical diffusion. Firstly, the New York City short /æ/ pattern is shown to have diffused geographically from New York to four other areas (Northern New Jersey, Albany, Cincinatti, and New Orleans) but, in each case, the adoption of this change from above has involved a loss of structural detail, ‘a transportation of the general phonetic basis for the NYC split, but not a faithful copy’ (Labov, Reference Labov2007: 360). Secondly, Labov examines evidence for the diffusion of the NCS from Chicago to St. Louis and demonstrates that St. Louis speakers approximate the NCS pattern rather than consistently replicating it, ‘indicating that the sound changes are diffusing individually, rather than as a system’ (Reference Labov2007: 383). Labov infers from the adoption of these individual features, in an ‘off the shelf’ fashion, that the presence of new vowels in St. Louis is not the structural consequence of the NCS (initiated by the tensing and raising of /æ/), but instead constitutes evidence for the borrowing of individual elements of the shift from Chicago. Likewise, Tagliamonte and Denis (Reference Tagliamonte and Denis2014) find that the diffusion of quotative be like from Toronto to surrounding Ontarian towns and villages has involved the loss of structural and social constraints on variation leading to ‘an attempt at, but imperfect replication of, the prevailing quotative system in Toronto, the diffusing model’ (Reference Tagliamonte and Denis2014: 129).

Following the adoption of diffusing features, subsequent transmission and incrementation may take place in the adopting varieties, leading to an array of incrementing regions ‘where each surrounding area exhibits incrementation at its own level, and the only difference between the big city and the small town is the time at which the process was initiated’ (Labov, Reference Labov2007: 350). Indeed, the structural and social constraints on variation and change will normally be absent or different in the new incrementing regions. Trudgill's (Reference Tuaillon1974) study of the diffusion of (æ) in the Norwegian dialects of the Brunlanes peninsula, for example, showed features adopted via diffusion to undergo subsequent linguistic change in the new towns because of incrementation on the part of children, though no analysis of structural constraints was included in this study.

Finally, we must make a distinction between linguistic features diffusing across geographical space and linguistic features diffusing across social groups within a defined geographical space. Both types of diffusion are underpinned by the process of speech accommodation and result in the weakening of the original diffusing pattern as well as the loss of structural complexity and sociolinguistic constraints on variation and change. For example, Labov (Reference Labov2014) shows that African American speakers in Philadelphia do not fully integrate the local dialect's structural constraints on the distribution of tense and lax /æ/, present in the speech of White informants, into their phonologies. Labov presents these findings as additional evidence of the distinction between ‘the nearly error-free transmission from parent to child, and the less accurate diffusion across the population’ (Reference Labov2014: 1), leading to the conclusion that, in cases of diffusion across both social and geographical space, ‘words and sounds may diffuse from one community to another, but systems do not’ (Reference Labov2014: 18).

2.3. The Transmission-diffusion dichotomy

In Labov's terms, change occurring during transmission can be viewed as a change occuring within the linguistic system, or ‘change from below’, whereas diffusion involves the importation of elements from other systems, or ‘change from above’ (Reference Labov2007: 347). Labov frames this dichotomy in terms of the differential language learning abilities of children and adults, respectively. Children are capable of replicating the adult system with a high degree of accuracy, acquiring linguistic features with all of their structural complexity and motivating change from below via the mechanism of incrementation when the features acquired are aligned with the vector of age. Adults, on the other hand, are less accurate in their language learning which places limitations on the mechanism of diffusion because most instances of dialect contact in face-to-face interactions are between adults, leading to a loss of structural detail during the acquisition process.

A clear dichotomy between these two mechanisms of language change is said to be dependent on the concept of a well-defined speech community with a common structural base and a unified set of sociolinguistic norms (Labov, Reference Labov2007: 347). Identifying and defining such a speech community is problematic in many respects (see Patrick, Reference Patrick, Chambers, Trudgill and Schilling-Estes2002 for a full discussion of these issues). Labov states, however, that any general view of language descent must be prepared to integrate the mechanisms of both transmission and diffusion, particularly when a well-defined speech community cannot be identified. For example, in western European dialectological studies (Auer and Hinskens Reference Auer and Hinskens1996; Trudgill Reference Trudgill, Guy, Feagin and Baugh1996; Kerswill Reference Kerswill, Britain and Cheshire2003), the contrast between transmission and diffusion is said to be less prominent because these studies have generally identified the transfer of well-known features of older established dialects as the main linguistic phenomenon taking place: ‘We rarely find reports of changes from below that depend on transmission through incrementation, as in the many new sound changes of North America’ (Labov, Reference Labov2007: 348). Additionally, Labov states that, hitherto, discussions of language change in European dialectology have largely examined linguistic variation and change in relatively simplified terms, focusing on isolated individual dependent variables without a full analysis of the transfer (via both transmission and diffusion) of structural and sociolinguistic constraints during dialect contact. Analysing lexical isoglosses or unconnected phonetic variables is said to inhibit a comparative examination of the outcomes of transmission and diffusion because, without a higher degree of abstraction, the preservation or loss of constraints on variation and change cannot be accurately identified (Labov, Reference Labov2007: 348).

3.1. Dialect levelling in metropolitan French

Regional varieties of French are often considered to have resulted from contact between the local languages of France and the more recently imposed French language: ‘In the first half of the twentieth century, as French began to make significant inroads into areas of provincial France where it had not previously been spoken, new varieties emerged from contact between local and national norms’ (Hornsby, Reference Hornsby2006: 3). Perhaps the most famous example of this is Tuaillon's assertion that regional French is ‘ce qui reste du dialecte quand le dialecte a disparu’Footnote ² (Reference Tuaillon1974: 576). Traditionally, dialectological studies of regional French have focused on the identification in French of substrate features from France's moribund languages and, subsequently, on the loss of these features in favour of non-local forms.

The loss of local features from regional varieties of French has been shown to result from the process of ‘regional dialect levelling’, of which geographical diffusion is a principal component (Pooley, Reference Pooley1996; Hornsby, Reference Hornsby2006; Boughton; Reference Boughton2006, Reference Boughton2013; Hall, Reference Hall2008; Armstrong and Pooley, Reference Armstrong and Pooley2010; Mooney, Reference Mooney2014a, Reference Mooney2016; and others). Hornsby (Reference Hornsby2006) notes that increasing urbanisation in France over the past century has gone hand in hand with geolinguistic homogenisation, particularly in the north of France, while Armstrong and Pooley (Reference Armstrong and Pooley2010: 12) view the ‘hypercephalic’ demography of France as promoting the adoption of Parisian speech forms. Social changes, particularly those related to the centralising forces of Paris, are taken to be the reason for a ‘tendance à l'uniformisation de la prononciationFootnote ³ en France métropolitaine’Footnote ⁴ (Borrell and Billières, Reference Borrell and Billières1989: 55).

The prevalent diffusing norm in metropolitan France is commonly referred to as ‘supralocal’ French, as opposed to ‘standard’ French. Where standard French is the reference variety of French codified in grammars and dictionaries and propagated through the education system as the national language of France, supralocal French, on the other hand, is a levelled northern urban variety of French which is, in phonological terms, closely equivalent to what Carton et al. (Reference Carton, Rossi, Autesserre and Léon1983) termed français standardisé (in opposition to français standard, in the normative sense) (see Figure 1). Supralocal French constitutes a statistical norm, in that it is the everyday speech form of the majority of the population in the northern two thirds of France which differs from standard French primarily in relation to its phonology (Pooley, Reference Pooley2006: 360).

Figure 1. Supralocal French (Pooley, Reference Pooley2006: 385).

While ‘supralocalisation’ involves the spread of the northern statistical norm via geographical diffusion (supralocal French or norme, cf. Garmadi, Reference Garmadi1981), ‘standardisation’, on the other hand, involves convergence towards the ‘ideologised’ prescriptive norm (standard French or sur-norme). It is, however, primarily the supralocal French norm that is involved in widespread dialect levelling in metropolitan French. This norm diffuses outward from Paris, both hierarchically and in a wave-like fashion, tending to affect urban areas, particularly in the northern two thirds of France, before spreading to the contiguous rural areas in between cities and towns.

There is evidence to suggest that the regionally neutral supralocal variety of French has also spread southward into the northern langue d'oc regions. Indeed, Pooley notes that it is ‘in particular the départements which contain the Mediterranean coastline and Pyrenean borderlands and the immediately contiguous regions, where accents readily identifiable as southern may be commonly heard’ (Reference Pooley2007: 40). Armstrong and Pooley (Reference Armstrong and Pooley2010) adduce high levels of migration to the south of France as a motivational factor in the adoption by young Méridionaux (literally, ‘Southerners’) of a northern accent. Based on the findings of small number of studies, such as Wanner (Reference Wanner1993) and Kuiper (Reference Kuiper2005), it is commonly assumed that younger speakers in southern regions of France are systematically adopting a non-local accent, what Hornsby and Pooley refer to as ‘le manque de méridionalité dans la prononciation des jeunes méridionaux’Footnote ⁵ (Reference Hornsby, Pooley, Hintze, Pooley and Judge2001: 510).

3.2. The region of Béarn

Béarn is the historically Romance-speaking part of the modern-day Pyrénées-Atlantiques département in southwestern France (see Figure 2). The region is primarily rural and the local langue d'oc variety historically spoken in Béarn, a sub-dialect of Gascon, is commonly referred to as Béarnais (see Mooney, Reference Mooney2014b). Over the course of the twentieth century, Béarnais was gradually ousted from all domains by the dominant French language. The subsequent rise of industrialisation, social mobility and in-migration to the region has led to a situation of dialect contact: the regional variety of French that had emerged from language contact has been in contact with incoming varieties of French for some time, with the most notable of these being the supralocal northern norm.

Figure 2. The region of Béarn (from Wikimedia Commons user Thomas Gun).

Béarn also contains the city of Pau, the second largest urban centre, after Bordeaux, in the region of Aquitaine. The central commune of the city had a population of 84,763 in 2009 (INSEE, 2012), but its greater urban area has a population of approximately 198,000 inhabitants. The demographic evolution of Pau was rapid in the latter half of the twentieth century: it had a population of only 48,320 in 1954 (INSEE, 2012). This rapid growth is due to large-scale in-migration following the discovery of natural gas at nearby Lacq in the 1950s. Pau is served by an international airport and the TGV (Train à Grande Vitesse) Atlantique high-speed rail network with links to Bordeaux in two hours. The national TGV network links Pau to Paris in five hours and intercity trains and motorways link it to other large urban centres such as Toulouse and the Bayonne-Anglet-Biarritz conurbation.

Pau is some 800km from Paris and, as such, it is unlikely that supralocal linguistic features diffusing from Paris in a wave-like fashion will be adopted by speakers in Béarn. A more likely scenario, given the geographical distance from the cultural and economic centre, is that Béarn residents may adopt features diffusing via a defined hierarchical pattern: it is possible that supralocal features may be diffusing to Pau via intermediate urban centres such as Bordeaux or Toulouse. Given the indirect nature of this diffusion, Labov's model predicts that supralocal features would be adopted into the regional French of Béarn with a loss of structural detail and of sociolinguistic constraints, since these have been shown to weaken with each successive instance of diffusion via (perhaps numerous) satellite towns.

5.1. Supralocal and Parisian nasal vowels

Previous studies of regional French have demonstrated that some convergence towards Parisian or supralocal linguistic norms is inevitable. The nasal vowel systems presented here will be essential to the discussion of geographical diffusion in the regional French of Béarn.

Where the standard French system distinguishes four nasal vowel phonemes (/ /, / /, / / and / /), the supralocal French system (also characteristic of Parisian French) contrasts only three nasal vowels (/ /, / / and / /) (see Figure 5) due to the merger of / / and / / to / / (Pooley, Reference Pooley2006: 368). This means that the words ‘brun’ (‘brown’) /bʁ / and ‘brin’ (‘sprig’) /bʁ / are both pronounced [bʁ ] by the majority of speakers in northern France.

Figure 5. Supralocal French nasal vowels.

Additionally, in contemporary Parisian French, the nasal vowels appear to be undergoing a counterclockwise chain shift (see Figure 6; Mettas, Reference Mettas1973; Walter, Reference Walter and Deyck1994; Hansen, Reference Hansen1998, Reference Hansen2001) in which / / approaches / /, / / approaches / /, and / / becomes very rounded and close, e.g., ‘bain’ (‘bath’) /b / →[b ], ‘banc’ (‘bench’) /b / →[b ], ‘bon’ (‘good’) /b / →[bõ]). Hansen notes that the counterclockwise movement of this shift contrasts with the ‘rotation [. . .] observée pour les voyelles du français canadien qui vont vers l'avant’Footnote ⁷ (Reference Hansen2001: 45), e.g., ‘bain’ /b / →[b ], ‘banc’ /b / →[b ], ‘bon’ /b / →[b ]) (cf. Carignan, Reference Carignan, Lee and Zee2011; Nicholas et al., forthcoming).

Figure 6. Chain shift in Parisian French.

The chain shift taking place in Parisian French constitutes a change from below in progress. Hansen (Reference Hansen2001) notes that the shift is being led by intermediate variants in certain contexts and that it was nowhere near complete in the early noughties. Hansen identified two structural factors accelerating the chain shift (2001: 45): variants occurring in stressed position (final syllable of the rhythmic group) were more advanced than variants occurring in unstressed environments (see Mettas, Reference Mettas1973; Fónagy, Reference Fónagy1989); within rhythmic groups, variants occurring in final syllables of polysyllabic words also exhibited more evidence for change in progress (see Léon, Reference Léon1983; Malderez, Reference Malderez1991).

In addition to the primarily sociolinguistic studies cited above, there have been a large number of studies focusing on the phonological status and phonetic quality of the (European) French nasal vowels. These studies are largely based on laboratory, rather than spontaneous, speech and provide a wealth of descriptive detail on the nasal vowels, examining them from a variety of perspectives: speech perception (Delvaux et al., Reference Delvaux, Demolin, Soquet and Kingston2004; Woehrling and Boula de Mareüil, Reference Woehrling and Boula de Mareüil2006; Delvaux, Reference Delvaux2009); articulatory phonetics (Maeda, Reference Maeda, Hardcastle and Marchal1990; Teston and Demolin, Reference Teston and Demolin1997; Montagu, Reference Montagu2004; Delvaux et al., Reference Delvaux, Metens and Soquet2002; Delvaux et al., Reference Delvaux, Demolin, Harmegnies and Soquet2008; Carignan, Reference Carignan2013); acoustic phonetics (Longchamp, Reference Lonchamp1979; Maeda, Reference Maeda1982, Reference Maeda, Huffman and Krakow1993; Montagu, Reference Montagu2007); phonology (Durand, Reference Durand1988, Reference Durand, Baronian and Martineau2009; Delais-Roussarie and Durand, Reference Delais-Roussarie and Durand2003; Durand and Eychenne, Reference Durand and Eychenne2011). In the fields of articulatory and acoustic phonetics, there is much emphasis placed on the mapping of articulatory gestures onto acoustic cues for nasalisation and thus many of the studies cited above fall into both categories.

5.2. Evidence for Transmission in Béarn

This section presents evidence for faithful linguistic transmission from parent to child as well as for gradual incrementation, characteristic of innovative ‘change from below’. The summative results for regional French nasal vowel quality presented here are based on mixed-effects regression models with two accepted levels of statistical significance: significant (p < .05); highly significant (p < .01).

Firstly, all generations in the study were shown to use a four-term nasal vowel system, characteristic of the traditional southern French pattern: regression analyses revealed all generations to distinguish four nasal vowels phonetically. The distribution of / / in the speech of the older generation is governed by syllable type on the F1 dimension (p < .01): final syllables favour lower variants of / / than medial syllables. This may be a result of the tendency for final syllables to be longer than medial syllables: speakers may therefore have more time to reach a lower target for this vowel. Nonetheless, this vowel height conditioning is replicated by the middle (p < .01) and young generations (p < .01) and the constraint ranking is the same in each case. This is evidence for the faithful replication of the / / distribution (along with structural constraints) in successive generations within the speech community. On the F2 dimension, the older speakers’ / / distribution is not governed by internal constraints. For the middle generation, on the other hand, vowel frontness/backness is constrained by the presence or absence of a nasal consonant (p < .05): vowels occurring without a nasal consonant coda are significantly centralised. The youngest generation shows a parallel development (p < .05), transmitted faithfully from the middle generation, with centralisation of / / when no nasal consonant is present. It is possible that no such constraint is evident in the older generation's system because there is less variability in nasal consonant presence, with the large majority of nasal vowels accompanied by a nasal consonant coda.

All three generations make a significant phonetic distinction between / / and / /, showing these phonemes not to be merged in the regional French of Béarn. This distinction is faithfully replicated by successive generations and no significant constraints on the distribution of / / were present in any of the generations.

The distribution of / / in the older generation is governed by syllable type on the F1 dimension (p < .01): final syllables favour lower variants than medial syllables. This syllabic constraint is successfully transmitted to subsequent generations (middle (p < .01) and young (p < .05)) who display the same constraint ranking on variation in / /.

Finally, the older speakers’ distribution of / / is constrained on the F1 dimension by syllable type (p < .01), with final syllables favouring lower variants than medial syllables. This constraint is faithfully replicated by the middle generation (p < .01) but not by the young generation. Additionally, the oldest generation's / / distribution is constrained by the presence or absence of a nasal consonant coda, with nasal consonant codas favouring variants further back in the vowel space (p < .05). This constraint is replicated in the system of the young generation (p < .05) but not that of the middle generation. This loss of structural detail in the chain of transmission to subsequent generations may, as we will see, be due to the fact that this vowel is involved in a ‘change from below’.

Despite the phonological stability in the nasal vowel system of the regional French of Béarn, the apparent-time study revealed various incremental phonetic changes to be taking place in the nasal vowels: / /-fronting; / /-backing; / /-centralisation. In each case, successive generations of speakers were shown to advance changes in vowel quality along the F2 dimension, aligning inter-generational variation with the vector of age.

The linear regression analysis presented in Table 2 shows the F2 values for / / to be undergoing change in apparent time. The baseline regression coefficient of -0.203 for the older speakers shows that their / / vowels have the lowest F2 values when compared with the other generations: / / vowels are significantly more frontal in acoustic space in each successive generation with the youngest generation leading the change. This pattern of change is characteristic of incrementation as defined by Labov (Reference Labov2007).

Table 2. Regression model for F2 (/ /) (with ‘speaker’ and ‘word’ as random effects)

The / / vowel is also undergoing change in apparent time in the regional French of Béarn, becoming significantly more back in the acoustic vowel space in each successive generation. Evidence for incrementation is presented in Table 3, where the linear regression coefficients indicate gradual F2-lowering: older speakers produce the frontest variants of / / with a positive baseline coefficient of 0.242, indicating that they produce the highest F2 values for this vowel. The negative regression coefficient returned by the analysis for the middle generation indicates F2-lowering, and the youngest speakers lead the change, as they have the lowest F2 values for / /.

Table 3. Regression model for F2 (/ /) (with ‘speaker’ and ‘word’ as random effects)

The final apparent-time change taking place in Béarn is / /-centralisation. The regression analysis in Table 4 returned ‘age’ as a highly significant predictor of the value of F2: younger generations realise / / as progressively more centralised than old speakers.

Table 4. Regression model for F2 (/ /) (with ‘speaker’ and ‘word’ as random effects)

Within the nasal vowel system of the regional French of Béarn, there is evidence for both transmission and for the incrementation of changes in apparent time. Structural constraints present in the system of the oldest generation, such as the syllabic conditioning of vowel quality, were shown to be successfully transmitted from parent to child in an unbroken chain of intergenerational transmission. Three clear cases of incrementation in apparent time were also identified, illustrating the gradual advancement of vowel quality changes by successive generations.

5.3. Evidence for Diffusion in Béarn

The changes taking place via incrementation in the nasal vowels of the regional French of Béarn cannot exclusively be said to constitute ‘change from below’ akin to the incremental changes presented by Labov et al. (Reference Labov, Ash and Boberg2006) for the NCS. There is varying evidence to suggest that at least some of these changes were initiated by the process of diffusion.

It seems logical to interpret the significant fronting of / / in apparent time as an instance of diffusion because the acoustic fronting of / / reduces the phonetic difference between this vowel and (stable) / /. This change may therefore be indicative of the prolific surpalocal merger of / / and / / diffusing to Béarn. If this is the case, it is clear that this change has not yet come to completion and that any apparent diffusion of the supralocal norm constitutes a ‘change from above’ in progress, since even the youngest generation was shown to make a significant phonetic distinction between the front nasal vowels.

The quality difference between fronted and backed / / traditionally distinguishes southern varieties of French from northern or supralocal ones. As such, the / /-backing change may be interpreted as an instance of supralocalisation whereby the traditional centralised variant of the older generation has been replaced, in apparent time, by a low back variant (approximately [ ]) by the mechanism of geographical diffusion. There is no evidence to suggest that the Parisian change shift, where / / raises to [ ], is diffusing to Béarn: the height of / / is stable across the generations, occupying a low position in the acoustic vowel space. Given the reduced levels of contact between children born into native Béarnais families and adult speakers from elsewhere during their formative years of early childhood development (0–5 years), it seems more likely that the supralocal backed [ ] variant was adopted into the regional French of Béarn by the middle generation of speakers. The difference between the older speakers’ conservative variant, approximately [ã], and the new supralocal variant adopted by the middle generation may then have been aligned by the youngest generation with the vector of age, causing them to interpret the situation as follows: the younger the speaker, the more advanced the F2 change. Labov has shown that an array of incrementing regions may exist post-diffusion, where one generation adopts the diffusing variant and successive generations advance the change at their own level (2007: 350). If this is the sequence of events, the backing of / / as a ‘change from above’ via diffusion by the middle generation may constitute a ‘change from below’ in Labov's terms for the youngest generation, as they advance the change via incrementation rather than adopting this feature via diffusion. Arguably, the 16–18 year olds in the youngest generation will now be aware of the overt prestige of this backed variant but, in their formative years, there is no reason to believe that speakers would interpret linear / /-backing as a supralocal change, as a result of reduced exposure to the supralocal norm. Armstrong and Pooley's (Reference Armstrong and Pooley2010) assertion that young Méridionaux have adopted supralocal features as a result of large-scale in-migration to the south must, in this scenario, be interpreted differently: while diffusion may be the reason younger speakers are adopting non-local features, it is possible that these features are adopted ‘indirectly’ from the previous local generation, who adopted these features ‘directly’ via diffusion from northern populations at an earlier stage.

/ /-centralisation was revealed in the apparent time study to be a significant change in progress. There is no evidence to suggest that this change is a direct result of geographical diffusion in that it has not hitherto been attested in supralocal French. Martinet (Reference Martinet1945; Reference Martinet1958) proposed a functional explanation for oral /ɔ/-fronting: the presence of /ɑ/ in the speech of northern informants was said to have caused crowding in the back of the vowel space leading to fronter realisations of /ɔ/. It is also possible, therefore, that / /-backing in the regional French of Béarn has caused / /-centralisation to maximise the phonetic distinction between the phonemes and to maintain a four-term nasal vowel system.

When we consider these three changes together, it appears that the nasal vowel system of the regional French of Béarn is experiencing a counterclockwise chain shift in apparent time, initiated by the backing of / /, as illustrated in Figure 7. Where the Parisian chain shift identified by Hansen (Reference Hansen1998, Reference Hansen2001) appeared to be initiated by the lowering and backing of / / (following a merger with / /), the chain shift in Figure 7 appears to have been initiated by the adoption of backed / / via diffusion by the middle generation of speakers, followed by subsequent incrementation by the youngest generation. From a functional perspective, this / /-backing in apparent time has caused parallel centralisation of / / and, therefore, the significant / /-fronting change discussed above may alternatively be interpreted as part of a wider systemic change, or chain shift (rather than a case of gradual supralocalisation via diffusion).

Figure 7. Chain shift in the regional French of Béarn.

The adoption of an individual feature from supralocal French, rather than the system as a whole, into the regional French nasal vowel system has therefore triggered a series of changes from below, internal to the system in Béarn. Some of these changes resemble supralocal norms but it is equally possible that these changes are internally-motivated or that, as younger speakers grow older and come into more intimate contact with the supralocal norm, multiple causation is more appropriate an explanation. This chain shift in the regional French of Béarn illustrates the transmission-diffusion interface, showing the two processes to interact and to overlap, leading to innovations in the sense that they are ‘spontaneous’ local developments that are not directly attributable to the process of diffusion.

These developments are also consistent with Labov's findings for diffusion. Much like the diffusion of the NCS from Chicago to St. Louis, individual features from the diffusing system, rather then the complexity of the entire system, are adopted into receiving varieties: the supralocal three-term nasal vowel system is not adopted and there is no evidence for the Parisian chain shift in Béarn. While the Parisian and Béarn chain shifts may be considered different responses to somewhat similar problems, from a functional perspective, the triggers are certainly different. We have seen this in Labov's study where the NCS was triggered by the raising of short /æ/ in Chicago but not in St. Louis; any incrementation of the chain shift involved the advancement of change for individual sounds (2007: 378).