2.1. External history of Occitan and Catalan

By the tenth century, Latin-based varieties spoken in Gaul had become strongly diversified, with the most significant division within Gallo-Romance between the langue d’oc (often referred to as ‘Occitan’) in the south, and the langue d’oïl, in the north (see Figure 1). The modern langue d’oc area is commonly divided into six main dialects (Bec, Reference Bec1963:37; see Figure 2), with Gascon in the southwest, including the Béarnais sub-dialect (the focus of the present study).

Figure 1.

Gallo-Romance languages (Mooney, Reference Mooney2016: 9)

Figure 2.

Gallo-Romance dialect areas (Mooney, 2016: 9)

The linguistic classification of Catalan remains controversial (cf. Badía Margarit, Reference Badía Margarit1955 and the use of the term lengua puente), and it is most frequently described as Ibero-Romance, rather than Gallo-Romance. Catalan-speaking areas were subject to unique external pressures, such as the presence of an Arabic-speaking elite in the south of the territory. Most Catalan-speaking regions came under Castilian influence from the early fifteenth century, and the earliest processes of transfer from this powerful superstate are documented around this time. Catalan is divided into two broad dialect areas, Western and Eastern (Veny and Massanell, Reference Veny and Massanell2015), and Northern Catalan (the focus of this article) is a sub-dialect of Eastern Catalan.

Although the classification of Occitan as Gallo-Romance and Catalan as Ibero-Romance is common, some scholars criticize the justification for this decision as superficial. Posner is not alone in maintaining that ‘the end-result looks suspiciously like areal grouping’ (Reference Posner1996: 24), and Juge (Reference Juge2007) defends the notion that Catalan and Occitan should in fact form their own separate sub-branch of Romance. Many studies, including Bec (Reference Bec1963) and Sumien (Reference Sumien2006), have emphasized the common diachronic developments in, and strikingly similar synchronic structures of, Occitan and Catalan: ‘whereas a “Proto-Occitan-Catalan” is a quite plausible concept, a “Proto-Occitan-French” (excluding Catalan) certainly is not’ (Harris, Reference Harris, Harris and Vincent1988: 16). On this basis, ‘Occitano-Romance’ has been proposed as a sub-language family in its own right, distinct from Gallo- and Ibero-Romance, or, in some cases, as a sub-grouping within Gallo-Romance. Bec (Reference Bec1963: 55–58) takes this argument further, proposing two supra-dialectal areas within Occitano-Romance, on the basis of common phonological developments in each area: alverno-méditerranéen (later arverno-méditerranéen; Bec, Reference Bec1973: 18–19), including nord-occitan and Provençal, and aquitano-pyrénéen, including (Pyrenean) Lengadocian, Gascon, and Catalan. Within the aquitano-pyrénéen grouping, which is said to be ‘centré autour du gascon’ (Bec, Reference Bec1973: 18), Pyrenean Lengadocian is identified as ‘le “pont” naturel entre gascon et catalan’ (Bec, Reference Bec1973: 19), but the rest of the Lengadocian dialect is singled out as particularly conservative, emphasizing the structural parallels between the Gascon dialect and Catalan.

2.2. Sociolinguistic profile of Occitan and Catalan

The introduction of the Lois Jules Ferry (1880–1886), which provided free and compulsory education in French, as well as national conscription during the two World Wars, constituted a serious blow to the regional languages of France and instigated a widespread process of language shift throughout metropolitan France. As such, both Occitan and Catalan are in an increasing state of language obsolescence in France.

It is difficult to propose speaker numbers for Occitan, and indeed for the langues de France in general, but recent work has cautiously estimated that there are around 500,000 native speakers in total (Martel, Reference Martel2007). In the present study, data will be drawn from speakers of the Béarnais sub-dialect of Gascon (see Figure 3), the historically Romance-speaking part of the Pyrénées-Atlantiques département. In contrast to Martel’s (Reference Martel2007) estimate, Moreux (Reference Moreux2004: 25) suggests that, in the entire historically Gascon-speaking area alone, there are approximately 500,000 speakers in total, ranging from 3 per cent of the population in Bordeaux to 30 per cent in Béarn. This estimate includes, however, both native and semi-speakers and Moreux states that, for the Béarn region at the end of the twentieth century, only about 14 per cent of the adult population were fluent speakers (approximately 40,000 speakers), with more than 50 per cent of these speakers over the age of sixty and rural-dwellers. Active competence among younger generations is rare, but since the 1980s, some public schools and private Calandretas (immersion-education schools) have offered bilingual Occitan-French primary education. The effect of the eight Calandretas on the linguistic competence of the younger generations in the Béarn region is negligible, however, with only approximately 2,000 children in attendance (Moreux, Reference Moreux2004: 33).

Figure 3.

Areas of study (adapted from Wikimedia Commons user Giro270)

Northern Catalonia is the name given to the area of France where Catalan is spoken, and largely corresponds to the Pyrénées-Orientales département. Regarding speaker numbers, the government of the Autonomous Community of Catalonia supports distribution of language competence surveys in the area, in conjunction with local partners. A 2015 study indicated that 35.4 per cent of the population of the département claimed to speak some Catalan (approx. 167,000 speakers), and that 20.6 per cent (approx. 97,000) could speak the language either ‘well’ or ‘very well’ (Baylac Ferrer, Reference Baylac Ferrer2016: 35). However, the self-reported nature of these findings may call into question their accuracy. As in Béarn, there is an extremely limited presence of Catalan in the education system, with only 2,452 children receiving some form of Catalan instruction, either in the private Bressola immersion system, or in occasional public language classes.

3.1. Phonological distribution

In the transition from Late Latin to Proto-Romance, intervocalic /ɡl/ (< original –GL– and voiced –CL–) palatalized, leading to the existence of a palatal lateral /ʎ/ in word-final position, post-syncope, e.g. OCULUM > *OCLUM > Gasc. uèlh /weʎ/, Cat. ull /uʎ/, Old Fr. ueil /ɥɛʎ/ ‘eye’. This sound change yielded the same results as later developments in Gallo-Romance involving intervocalic /jl/ and /lj/ (Müller, Reference Müller2011: 99; Ford, Reference Ford1966: 35), e.g. Vulgar Latin TRIPALIĀRE > Occ. tribalhar /tɾibaˈʎa/, Cat. treballar /tɾəbəˈʎa/, Old Fr. traveillier /traveʎɛr/ ‘to work’. In word-initial position, /lj/ sequences also gave rise to palatal laterals in some varieties of Occitan (e.g. southern Gascon), and Catalan, e.g. LEVĀRE > Gasc. lheba /ˈʎebɔ/, Cat. lleva /ˈʎebə/, and remained /lj/ in Old French, e.g. il lieve /il ˈljɛvə/ ‘he rises’; in other Occitan varieties, the word-initial lateral is commonly alveolar.

In Standard Modern French, the palatal lateral /ʎ/ does not survive, having evolved to [j] in intervocalic and final position, e.g. œil /øj/ ‘eye’, travailler /tʁavaˈje/ ‘to work’. In word-initial position, Old French distinguished a stressed present tense with a /lj/ sequence stem in verbs such as lever, e.g. tu liev-es /ty ˈljɛv-əs/ ‘you (s.) raise’, and an unstressed stem with no palatal glide, e.g. vous lev-ez /vuz ləˈv-ɛts/ ‘you (pl.) raise’. Modern French has a stressed stem in which the /lj/ sequence is not present, e.g. tu lèves /ty ˈlɛv/ ‘you (s.) raise’, due to a regular /jɛ/>/ɛ/ change which occurred in thematic presents (Foley, Reference Foley1979: 107–108), leaving /l/ in initial position, e.g. il lève /il lɛv/.Footnote ¹ The /ʎ/ phoneme was realized variably as [ʎ] or [j] throughout the seventeenth and eighteenth centuries, with the [j] variant ‘widespread in lower-class Parisian speech in the seventeenth century’ (Lodge, Reference Lodge2004: 189). While the use of the [j] variant was the subject of much metalinguistic commentary, the /ʎ/>[j] change only came to completion at the beginning of the nineteenth century (Ayres-Bennett, Reference Ayres-Bennett1996: 228, 233; Lodge, Reference Lodge2004: 215). Therefore, in Modern French, /l/ now occurs in word-initial position /#_V/, and /j/ occurs in intervocalic /V_V/ and final positions /V_#/ (see Table 1).

Table 1.

Distribution of the palatal lateral in Occitan (Gascon) and Catalan, with French for comparison

Occitan, on the other hand, has been markedly more conservative in its preservation of the palatal lateral consonant. It is, however, useful to distinguish between retention of the palatal lateral in initial /#_V/, intervocalic /V_V/, and final /V_#/ positions when discussing the high level of interdialectal variability in retention rates. In initial position, the palatal lateral is rare cross-dialectally, but in Gascon, it is maintained in a small set of words, namely lhevar ‘to rise’, (di-)lhèu ‘perhaps’, and lhèit ‘bed’ (forming a minimal pair with lèit ‘milk’ in most Béarnais dialects). In intervocalic positions, /ʎ/ is maintained throughout the langue d’oc continuum except Provençal, where it has merged with /j/ (Wheeler, Reference Wheeler, Harris and Vincent1988: 249; Ford, Reference Ford1966: 20), as in Modern French, e.g. trabalhar /tɾabaˈʎa/ > Prov. [tʁ̥avaˈja] ‘to work’. In word-final position, /ʎ/ is depalatalized to [l] in Lengadocian (Bec, Reference Bec1973: 46; Wheeler, Reference Wheeler, Harris and Vincent1988: 249; Oliviéri and Sauzet, Reference Oliviéri, Sauzet, Ledgeway and Maiden2016: 326), e.g. talh /taʎ/ > Leng. [tal] ‘cut’, while in Provençal has both [j] and [w] in word final position, e.g. genolh /d͡ʒeˈnuʎ/ > Prov. [d͡ʒeˈnuj] ‘knee’ and filh /fiʎ/ > Prov. [fj̥ew] ‘son’. In Gascon, /ʎ/ is retained in all positions (see Table 1), violating the general Occitan phonotactic constraint of /ʎ/ being impermissible in coda position; this is characteristic of the conservative Occitano-Romance supra-dialectal grouping of aquitano-pyrénéen (Oliviéri and Sauzet, Reference Oliviéri, Sauzet, Ledgeway and Maiden2016: 327; see Section 2.1); /ʎ/ may also occur in a tautosyllabic consonant cluster with /s/ in coda position, where /-s/ is the inflectional morpheme for the pluralization of common nouns (see Table 1). In Cardaillac Kelly’s (Reference Cardaillac Kelly1973) study of the Gascon variety spoken in Donzac (Tarn-et-Garonne), she notes that /ʎ/ has ‘no variants and no distributional limitations’ (Reference Cardaillac Kelly1973: 32), even noting that [ʎ] transfers variably into the regional variety of French spoken by her bilingual informants, where it replaces French /j/ in intervocalic position (Reference Cardaillac Kelly1973: 69), e.g. veille /vej/ [vɛʎə]. This suggests that [j] is not used as a variant of /ʎ/ in Gascon.

The phonological distribution of the palatal lateral in Catalan is the same as in Gascon, with /ʎ/ retained in all positions (see Table 1). There are, however, two different phenomena that cause [ʎ] to appear in variation with [j]. The first is known as ieisme (literally, ‘[j]-ism’), and results in words that descend from -LI- or -C’L- being variably realized as [ʎ] or [j]. This does not affect words with Latin L- (and as such, does not occur word-initially) or -LL-, is diatopically restricted, and has been attested for centuries (Wheeler, Reference Wheeler2005: 34–35). Within Northern Catalonia, ieisme is traditionally only found in the valley of Vallespir (Recasens, Reference Recasens1996: 322), an area not represented in the present data. The second phenomenon refers to a wholesale replacement of /ʎ/ with [j], and has been the subject of metacommentary for some time, but not systematically studied (Wheeler, Reference Wheeler2005: 35). Recasens (Reference Recasens1996: 324) speculates that this phenomenon might be underway in Northern Catalan due to transfer from French. More recently, he advances that ieisme in Northern Catalonia may be favoured by younger, urban speakers (Recasens, Reference Recasens2017: 325), but offers no empirical support for either explanation.

In summary, Gascon and Catalan are similar in that they retain the phonemic palatal in all positions. Given that all Gascon and Northern Catalan speakers are also bilingual with French, we must highlight the structural correspondences between the surface phonologies of their languages. Flege’s (Reference Flege and Winitz1988, Reference Flege and Kent1990, Reference Flege1991) ‘Speech Learning Model’ (SLM) posits that the phonemes of a bilingual’s languages are equated as cognates and so have the potential to influence each other, in the form of phonological or phonetic transfer (see Mooney, Reference Mooney2017 for discussion). In the case of Gascon-French and Catalan-French bilinguals, word-initial /ʎ/ in Gascon and Catalan is phonemically cognate with /l/ in French, while intervocalic and word-final /ʎ/ is phonemically cognate with /j/ in French (see Table 1).

3.2. Acoustic characteristics

The articulatory gestures associated with the production of lateral segments gives them a characteristic acoustic profile (adapted from Colantoni, Reference Colantoni, Auger, Clements and Vance2004: 86–87): their first formant (F1) is relatively low frequency (300–400 Hz), the second formant (F2) varies according to the location of the constriction, usually falling somewhere in the region of 2000 Hz (Zampaulo, Reference Zampaulo and Haug2015: 75), and the third formant is relatively high in frequency. Tabain et al. (Reference Tabain, Butcher, Breen and Beare2014: 1), for example, report the following average formant frequencies for [ʎ] in Central Australian languages: F1 (339 Hz); F2 (2093 Hz); F3 (2891 Hz); F4 (4001 Hz). Two resonating chambers are involved in the production of the lateral, one formed by the openings on both sides of the tongueFootnote ² and the other by the cavity on top. This second cavity causes antiformants or zeroes to appear in the spectrum in the F2–F4 range (Fant, Reference Fant and Malmberg1968). Palatal laterals are distinguished from adjacent vowels by sharp changes in the formant frequencies (Fry, Reference Fry1979: 137), though their formant structure is relatively stable and tends not to be affected significantly by vocalic context. Finally, relatively speaking, palatal laterals are long segments when compared with other consonants. The precise acoustic characteristics of the palatal lateral differ cross-linguistically and according to a host of language internal factors, such as stress, for example (see Colantoni, Reference Colantoni, Auger, Clements and Vance2004: 86–87 for an overview of their acoustic profile in Spanish and Portuguese).

4.1. Contact-induced linguistic change

Situations of language contact provide ideal conditions for language change. Weinreich (Reference Weinreich1968 [1953]: 18–19) presents a range of potential processes that may take place when languages with different phonemic systems co-occur in the bilingual mind. These include, for example, under-differentiation, where a phonemic distinction is lost in a speaker’s less dominant language (for example, /ʎ/~/j/ in Occitan and Catalan), when it does not occur in the speaker’s dominant language (here, French). Additionally, Occitan shows signs of adopting elements of the groupe rhythmique prosodic structure with phrase-final stress (Sichel-Bazin et al., Reference Sichel-Bazin, Buthke, Meisenburg, Braunmüller and Gabriel2012: 361), while Occitan (Mooney, Reference Mooney, Ayres-Bennett and Carruthers2018) and Northern Catalan (Hawkey, Reference Hawkey2018: 116–126) display extensive use of uvular rhotics, under influence from French.

There is evidence to suggest that Occitano-Romance /ʎ/ is variably replaced by [j] and this is commonly conceived to be due to linguistic transfer from dominant contact languages, as a result of pervasive bilingualism (Wheeler, Reference Wheeler2005: 35). In France, where Gascon and Catalan have found themselves in long-term language contact with French, the overall outcome of unilateral L2-to-L1 contact-induced change would be a reduction in the use of the [ʎ] variant, replacing it with [j]. We might also expect, however, to see an increased probability of transfer in specific linguistic contexts, with language-internal constraints conditioning transfer on the basis of phonemic cognates in the surface phonology of the contact language, French (see Section 3.1): in initial position, Gascon and Catalan /ʎ/ corresponds to /l/ in French; in intervocalic and final position, the corresponding French phoneme is /j/. As such, we may expect to find a statistically significant preference for /ʎ/ to be realised as [j] in non-initial contexts, for example, if the change is due to contact with French.

4.2. Regular sound change

The Neogrammarians of the late nineteenth century held the view that sound changes, such as /ʎ/>[j], occurred in a strictly regular fashion, progressing incrementally from one sound to another, and motivated by internal concerns, such as ease of articulation (Hale, Reference Hale, Brain and Janda2017). Such gradual sound changes do not operate above the level of conscious awareness, and thus do not initially acquire social meaning: these internally motivated ‘changes from below’ (Labov, Reference Labov1972: 178–180) are not triggered by language contact, and can occur in monolingual speech communities. They originate in a subset of the speech community, before being adopted by other groups, after which point they may start to show stylistic variation and be endowed with social meaning.

Palatal laterals have featured in the phonological inventories of most Romance languages; Ford notes that they are traditionally ‘found in all the Romance languages with the exception of Sardinian and Roumanian’ (Ford, Reference Ford1966: 35). The /ʎ/>[j] change is well attested in Romance, but the extent to which the change has progressed in the different languages is highly variable: ‘from almost complete loss in French (Nyrop, Reference Nyrop1923) to preservation in Portuguese (Silva, Reference Silva1999) and Catalan (Recasens et al., Reference Recasens, Farnetari, Fontdevila and Pallarès1993; Recasens and Pallarès, Reference Recasens and Pallarès2001)’ (Colantoni, Reference Colantoni, Auger, Clements and Vance2004: 83). Some studies have suggested that contact cannot (fully) account for the distributional patterning of the variants of /ʎ/, such as Moratal (Reference Moratal2011), where Catalan-Spanish bilinguals had different rates of maintenance in each of their languages, and, indeed, this change has occurred in many Romance varieties that do not find themselves in a situation of long-term contact, such as French (Wheeler, Reference Wheeler2005: 35) or Argentine Spanish (Colantoni, Reference Colantoni, Auger, Clements and Vance2004: 83). Bec (Reference Bec1973: 47) has argued that the explanation is to be found in the low functional load of the /ʎ/~/j/ contrast, therefore advancing a language-internal motivation for the /ʎ/>[j] change. As such, while it seems plausible that the /ʎ/>[j] change in Gascon and Catalan may be due to contact with French, we must acknowledge the fact that it could be occurring as a change from below, ‘probably due to the [phonetic] similarity of the palatal lateral and the glide’ (Colantoni, Reference Colantoni, Auger, Clements and Vance2004: 97; see Section 4.3). In light of this, we must consider the possibility that Gascon and Catalan, as Romance languages, may exhibit variable maintenance of the palatal lateral as a result of internally motivated sound change.

4.3. ‘Hidden variation’ theory (Ohala, Reference Ohala, Breivik and Jahr1989)

The /ʎ/>[j] change may (at least partially) be motivated by the phonetic similarity between the two sounds. Colantoni (Reference Colantoni, Auger, Clements and Vance2004) has shown that, in Argentine Spanish, a palatal approximant [j], or glide-like segment, may emerge in the CV transition from the palatal lateral [ʎ] to the following vowel, as a result of coarticulation.Footnote ³ She argues that this glide-like transition may be the motivation for the /ʎ/>[j] change, interpreting its presence within the framework of Ohala’s (Reference Ohala, Breivik and Jahr1989) ‘hidden variation theory’, sometimes referred to as the ‘listener oriented approach to sound change’. The basic premise of Ohala’s theory is that synchronic phonetic variation is a precondition for sound change:

Therefore, sound change is seen to take place when the listener reinterprets secondary acoustic cues that are present in the signal as important for parsing and producing the segment. These secondary cues can, over time, become phonologized and integrated into the speaker’s lexical representation of words (Ohala, Reference Ohala, Breivik and Jahr1989: 179). Within this framework, we may view the presence of a contextually-conditioned glide-like segment between the palatal lateral and the following vowel as an inevitable consequence of articulatory constraints; the dynamic tongue movement involved in the transition from the configuration for the lateral to that of the following vowel produces a glide segment, and listeners may then interpret this as intentional: ‘If listeners are not able to recover the acoustic signal that identifies the lateral component of [ʎ], they may identify it as a palatal [j] and reinterpret it as such in their sound system’ (Zampaulo, Reference Zampaulo and Haug2015: 75).

Considering the change in Spanish and Portuguese in an Optimality Theory framework, Zampaulo (Reference Zampaulo and Haug2015) argues that it is motivated by a difference in the constraint ranking of the grammars of the listener and speaker, building on Colantoni’s (Reference Colantoni, Auger, Clements and Vance2004) argument that a glide-like segment in the CV transition may motivate the substitution of [ʎ] by a glide. Colantoni notes, however, that ‘the presence of a glide-like CV transition would not account […] for the loss of the lateral’ (Reference Colantoni, Auger, Clements and Vance2004: 84), advancing two alternative (but not mutually exclusive) hypotheses to account for the reduction of the palatal lateral: (i) that the glide becomes longer, and the lateral shorter, until the lateral eventually disappears; (ii) that the constriction of the lateral becomes wider and then more similar to the glide (Colantoni, Reference Colantoni, Auger, Clements and Vance2004: 84). Testing these hypotheses on data from Argentine Spanish, but only considering tokens where the [ʎ] variant was maintained, Colantoni (Reference Colantoni, Auger, Clements and Vance2004) found evidence to support both hypotheses, arguing, however, that the second hypothesis is the most likely explanation for the change.