1 INTRODUCTION
The Endangered Languages Project reports that over 40 per cent of the world's 7,000 languages are at risk of disappearing (Catalogue of Endangered Languages, 2017). One way a language may become endangered is when it is no longer a valued vehicle of communication for its speakers, who have opted to shift towards another socially or economically dominant language (e.g. Spolsky, Reference Spolsky, Austin and Sallabank2011). In the case of Louisiana French (LF), the focus of the current investigation, a shift towards English began in the 1800s and has gradually continued to the present day, with numbers of French speakers dwindling statewide.
Obsolescing languages often demonstrate distinctive linguistic patterns, including increased variability and rapid change (e.g. Wolfram, Reference Wolfram, Chambers, Trudgill and Schilling-Estes2004; O'Shannessy, Reference O'Shannessy, Austin and Sallabank2011). In the dialect of LF spoken by the Pointe-Au-Chien (PAC) IndiansFootnote 1 of Pointe-Aux-Chênes,Footnote 2 Louisiana, both effects can be observed. For example, regarding subject expression, there exist (at least) 12 different ways of expressing the first-person-singular pronoun, based on morphological and phonological sources of variation. Moreover, the pronoun system appears to be in the process of changing. Previous research centered separately on the morphological (Rottet, Reference Rottet1996; Dubois, Reference Dubois and Donabédian2001; Dajko, Reference Dajko2009) and the phonological (Salmon, Reference Salmon2007; Dajko, Reference Dajko2009; Carmichael, Reference Carmichael2017) variability in the pronoun system suggests that the variation is not random. However, a unified explanation of pronoun use across the remaining population of speakers has yet to be provided.
In a language death context, it is common to find a final generation of non-fluent speakers (semi-speakers, or SSs), passive bilinguals and rememberers, all of whom have some knowledge of the language, though their production and understanding may be incomplete (e.g. Dorian, Reference Dorian1977). This range of language ability, which differs from ‘healthy’ language contexts, is due in part to children receiving more limited exposure to the target language, thus resulting in incomplete acquisition. We target the speech of these SSs for analysis, drawing connections between second language acquisition (SLA) and language death research in order to reconceive the SSs’ patterns as an independent linguistic system (cf. Selinker, Reference Selinker1972). We explore the inventory of first-person-singular forms across speakers of differing fluency ranges, characterizing the extent to which the patterns we observe carry across groups and across individuals. This approach enables us to see whether SSs feature fewer forms and/or innovations not found in the older fluent norm. Based on our analyses, we propose that concomitant morphological and phonological changes in progress in the pronoun system of fluent LF speakers may have led to reanalysis of first-person-singular subject expression in the speech of some SSs in our corpus.
2 BACKGROUND
In this section, we first introduce the sociohistorical context of French in Louisiana. Next, we describe the LF pronoun system, focusing on the inventory of first-person-singular subject forms within the LF spoken in Pointe-Aux-Chênes. Then we describe the ways that LF features manifest common processes documented in other situations of language death. Finally, we draw connections between language death and SLA research to explain how insights from SLA can inform our analyses of first-person-singular subject expression in the context of declining LF, proposing our research questions in light of these points.
2.1 The sociohistorical context of French in Louisiana
Francophone colonizers began arriving in the Louisiana territory in the late seventeenth century. Though the colony was slow to expand, settlers trickled in over the next hundred years, bringing with them a range of French dialects, some coming directly from France, and others indirectly from colonies in Canada and the Caribbean (Klingler, Reference Klingler, Mathis-Moser and Bischof2009; Dajko, Reference Dajko2012). Two waves of AcadianFootnote 3 settlers arrived in Louisiana in the latter part of the eighteenth century (Brasseaux, Reference Brasseaux1992). The Acadians intermarried and intermingled with other settlers throughout the region, including Europeans, American Indians and Africans of various linguistic backgrounds. Louisiana briefly became a Spanish colony, and then in 1803 was purchased by the United States. However, up until the Civil War, the French linguistic and cultural heritage remained strong in South Louisiana. The early twentieth century brought pro-Anglophone policies, such as the 1921 revised state constitution banning schooling in any language but English (Picone, Reference Picone1997). This time period was marked by severe stigmatization of LF. In the 1960s, the Council for the Development of French in Louisiana (CODOFIL) was established, and the state seemed to embrace its Francophone roots. Despite these efforts, the population of French speakers has declined in Louisiana, portending the gradual death of LF (Picone, Reference Picone1997; Bankston and Henry, Reference Bankston and Henry1998; Carmichael, Reference Carmichael2008).
In Pointe-Aux-Chênes, both Cajuns and Indians use LF on a daily basis, however most of these speakers are elderly. In contexts of gradual language death, individuals in a speech community shift away from speaking the traditional language until intergenerational transmission is interrupted, at which point a language is considered endangered (Campbell and Muntzel, Reference Campbell, Muntzel and Dorian1989). Without social changes or revitalization efforts which encourage the youngest generation in the speech community to adopt the traditional language again, an endangered language will eventually have no native speakers; it will become extinct. And indeed, this is the case in Pointe-Aux-Chênes, which is in danger of extinction in the next few generations (Carmichael, Reference Carmichael2008).
2.2 LF Pronoun System
Linguistically, LF features a number of lexical items, grammatical constructions, and pronunciations that distinguish it from Metropolitan FrenchFootnote 4 (MF), such as use of alveolar tap /ɾ/ rather than velar fricative /ʁ/, use of après + infinitive to express present progressive (il (est) après venir ‘he is coming’), and use of ‘autres’ in plural pronouns (vous-autres; nous-autres) (Papen and Rottet, Reference Papen, Rottet and Valdmen1997; Rottet, Reference Rottet, Valdman, Auger and Piston-Hatlen2005; Carmichael, Reference Carmichael2008). While these features may also be heard in varieties of French spoken in Acadia and elsewhere in Canada, due to shared settlement history, LF demonstrates some features not found in Canadian French varieties. For example, lexical items deriving from local American Indian terms, including chaoui ‘raccoon’ and bayou ‘slow-moving stream’ (Dajko, Reference Dajko2009). Other distinctive lexical items in LF derive from liaison reanalysis, for example noncle to mean ‘uncle’ (cf. MF mon oncle ‘my uncle’) and zoiseau to mean ‘bird’ (cf. MF des oiseaux ‘some birds’) (Blainey, Reference Blainey2017). Similarly, words such as éyoù ‘where’ (cf. MF où) and équand ‘when’ (cf. MF quand) feature prothesis, or the addition of an initial sound (Papen and Rottet, Reference Papen, Rottet and Valdmen1997).
Another key difference between MF and LF lies in the pronoun system. In both LF and in MF, there exist clitic pronouns (e.g. MF/LF je ‘I’ and il ‘he’), as well as tonic/disjunctive pronouns (e.g. MF moi ‘me’ and lui ‘him’; LF mon ‘me’ and lu ‘him’). These are sometimes referred to as weak and strong pronouns, respectively, since in MF ‘weak’ subject pronouns je and il cannot take prosodic stress or appear without a conjugated verb, in contrast with ‘strong’ moi and lui. As in many language death contexts (cf. Wolfram, Reference Wolfram, Chambers, Trudgill and Schilling-Estes2004), there is a preference in LF for more analytic sentence structure (Rottet, Reference Rottet1996), thus the unbound ‘strong’ object pronouns are often used rather than clitic object pronouns (compare MF je leur ai parlé ‘I spoke with them’ with LF j'ai parlé à eusse). That means that tonic pronouns already appear somewhat more frequently in LF than in MF. They also occur in both LF and MF in what is referred to as subject doubling, in which optional tonic pronouns may be used in addition to clitic pronouns, as either left or right dislocations (e.g., MF moi, je préfère le rouge ‘me, I prefer the red one’; je ne l'ai jamais vu, moi ‘I never saw him, me’) (Carroll, Reference Carroll and Lefebvre1982; Barnes, Reference Barnes1985; de Cat, Reference de Cat2007). Subject doubling occurs in higher rates in LF compared to MF (Rottet, Reference Rottet1996). Perhaps due to the high rate of use, subject doubling is treated as an emblematic feature of LF and of the French-influenced variety of English spoken in South Louisiana in which this phrasing pattern is calqued (e.g. ‘me, I went to the store’ or ‘he's not so friendly, him’; Dubois and Horvath, Reference Dubois and Horvath2002; Carmichael, Reference Carmichael2013). In addition to subject doubling, in LF ‘strong pronouns’ may also appear alone as the subject of a verb, with the clitic pronoun omitted entirely (e.g. LF mon mange ‘me eat’ or lu mange ‘him eats’) (Rottet, Reference Rottet1996; Dubois, Reference Dubois and Donabédian2001; Girard, Reference Girard, Salvesen and Helland2013). Such constructions would be ungrammatical in MF. Another form of pronoun expression that exists in LF and other Romance languages (cf. Carvalho, Orozco and Lapidus Shin, Reference Carvalho, Orozco and Lapidus Shin2015, for Spanish) but would generally be ungrammatical in MF are null-subject forms (e.g. MF je veux aller au magasin ‘I want to go to the store’ may be expressed in LF as simply veux aller au magasin) (Rottet, Reference Rottet1996; Girard, Reference Girard, Salvesen and Helland2013).
Beyond the morphosyntactic differences in pronominal systems described above, there is also phonological variation in some varieties of LF that impacts the pronunciation of the first-person-singular clitic pronoun je ‘I’ in particular. As in MF, je may be pronounced either as /ʒ/ or, before voiceless segments, /ʃ/. In the variety of LF spoken in Pointe-Aux-Chênes, these forms may also be depalatalized to /s/ or /z/, again impacted by the voicing of the following segment.Footnote 5 While depalatalization is most common in first-person-singular subjects, it is not exclusive to these environments; it is also sometimes heard in other words, like changé as chanzé (Dajko, Reference Dajko2009). Another form of allophonic variation that is distinctive within the LF spoken in Pointe-Aux-Chênes is the lenition of /ʒ/ to /h/.Footnote 6 Similarly to depalatalization, it is most common in first-person-singular forms, but it is also heard in words like jambe ‘leg’ (pronounced as /hɑ̃m/) and déjà ‘already/before’ (pronounced as /dehɑ/) (Rottet, Reference Rottet1996; Carmichael, Reference Carmichael2008, Reference Carmichael2017; Dajko, Reference Dajko2009). As a result of the morphological and phonological variation within this variety of French, first-person-singular subjects may be variously expressed in (at least) 12 different ways. Table 1 presents this variation in terms of both the morphological and phonological differences.
LF first-person singular subject expression in Pointe-Aux-Chênes
Previous research has found social patterning to first-person-singular subject expression in LF according to gender, ethnicity, age, fluency, and register. Mon je appears in higher rates in casual speech than careful speech and is significantly preferred by younger speakers and Indians (Rottet, Reference Rottet1996). Mon is used more by less fluent speakers (Dubois, Reference Dubois and Donabédian2001). The /h/ variant is used more frequently by older fluent speakers in casual speech and is retained longer by women (Salmon, Reference Salmon2007; Carmichael, Reference Carmichael2007, Reference Carmichael2008; Dajko, Reference Dajko2009). And null subjects and the /z/ variant are used more by Indian speakers than by Cajuns (Dajko, Reference Dajko2009). While this research demonstrates that the variation in first-person-singular subject expression is systematic, each study has either focused on the phonological variation in the clitic or in the presence of the disjunctive pronoun mon, without examining the potential relationship between the two. In the current study, we suggest that both ought to be considered to better understand the changes in first-person-singular pronoun subject expression in LF.
2.3 Variation and change in language death contexts
Previous research on linguistic obsolescence has put forth a few key characteristics observed across many dying languages (e.g. Dressler, Reference Dressler, Peranteau, Levi and Phares1972; Dorian, Reference Dorian1989; Schmidt, Reference Schmidt1985a,Reference Schmidtb; Wolfram, Reference Wolfram, Chambers, Trudgill and Schilling-Estes2004; O'Shannessy, Reference O'Shannessy, Austin and Sallabank2011; Palosaari and Campbell, Reference Palosaari, Campbell, Austin and Sallabank2011):
1) Stylistic shrinkage
2) Reduction/simplification
3) Increased variability
4) Rapid change
(1) Stylistic shrinkage is a reduction in meaningful linguistic shifts across social circumstances. Naturally, when there is reduction in domains across which a language is used, there is less need for linguistic features that indicate a shift in register. (2) Reduction and simplification may manifest as loss of phonological distinctions, especially those not found in the dominant language or those with a low functional load (Campbell and Muntzel, Reference Campbell, Muntzel and Dorian1989; O'Shannessy, Reference O'Shannessy, Austin and Sallabank2011). One may also encounter the replacement of irregular or otherwise marked morphosyntactic features with more regular, linguistically unmarked forms, and a preference for analytic sentence structures with more rigid word order (Dorian, Reference Dorian1978; Schmidt, Reference Schmidt1985a; Campbell and Muntzel, Reference Campbell, Muntzel and Dorian1989; O'Shannessy, Reference O'Shannessy, Austin and Sallabank2011; Palosaari and Campbell, Reference Palosaari, Campbell, Austin and Sallabank2011). (3) Increased variability may occur in phonological or morphological realizations (Wolfram, Reference Wolfram, Chambers, Trudgill and Schilling-Estes2004) and can be motivated by contact with a dominant language, as words and structures are borrowed into the dying language, sometimes with specific functional distributions (Campbell and Muntzel, Reference Campbell, Muntzel and Dorian1989; Dajko and Carmichael, Reference Dajko and Carmichael2014). (4) Rapid change may be driven by the reduction in domains of use, convergence with the dominant language, and use of ‘attritional’ forms by younger and non-fluent speakers (SSs) (Schmidt, Reference Schmidt1985a,Reference Schmidtb; Cook, Reference Rottet1995; Dorian, Reference Dorian1973, Reference Dorian1978; Dubois and Noetzel, Reference Dubois and Noetzel2005). All four features are attested in LF.
Stylistic shrinkage may be seen in the LF pronoun system. MF observes a T/V distinction in second-person singular, in which the form tu is used with intimates while vous is the formal second-person-singular pronoun. Though historically Louisiana French features this same distinction, it is virtually absent in present-day speech (Rottet, Reference Rottet1996; Dajko, Reference Dajko2009). Notably, there is evidence of LF speakers style-shifting with phonetic variables (Carmichael, Reference Carmichael2007, Reference Carmichael2017; Dajko, Reference Dajko2009), thus it is important to note that stylistic shrinkage is not synonymous with a complete lack of style shifting (see King, Reference King and Dorian1989; Wolfram, Reference Wolfram, Chambers, Trudgill and Schilling-Estes2004).
Simplification can be observed within the LF verb paradigm, which has become leveled across subject pronouns and regularized across verb types (Girard, Reference Girard, Salvesen and Helland2013). This means, for example, that the conjugation of the verb aller is pronounced /va/ across all person and number conjugations.
The feature of interest in this article, first-person-singular subject expression, is arguably the result of increased variability and rapid change. Increased variability occurs on the phonological level, in that /ʒ/ may be variably pronounced as /h/, /s/, /z/, or /ʃ/ (Rottet, Reference Rottet1996; Dajko, Reference Dajko2009; Carmichael, Reference Carmichael2017). Rottet (Reference Rottet1996) and Dubois (Reference Dubois and Donabédian2001) have documented a striking intergenerational increase in presence of the tonic pronoun mon in the LF pronoun system, both in terms of subject doubling and use of the tonic pronoun alone to express first-person-singular referents. Both suggest that these changes to the LF pronoun system may be related to the proficiency of younger speakers, attributing these attritional forms to incomplete acquisition. Dubois and Noetzel (Reference Dubois and Noetzel2005) similarly note that age and proficiency impact LF locative prepositions, finding that younger and less fluent speakers in fact produce increased variability in locative prepositions and some innovative forms that differ significantly from the older fluent norm. Innovations driven by younger and non-fluent speakers have also been found in other language death situations (Dorian, Reference Dorian1978; Schmidt, Reference Schmidt1985a,Reference Schmidtb), thus making age and fluency of particular interest in this study.
2.4 Language Death and SLA
To better understand declining varieties, linguists have made connections between language obsolescence and the field of SLA. For example, Andersen (Reference Andersen and Dorian1989) shows that insights gained from SLA can enhance the observations that language death researchers make. Specifically, he suggests that ‘the linguistic consequences of language contraction and death are not simply ‘oddities’ of little relevance to ‘normal’ language bahavior [sic], but linguistic phenomena which can be seen in other areas of [second] language acquisition, use, and disuse as well’ (393). Similarly, Wolfram (Reference Wolfram, Chambers, Trudgill and Schilling-Estes2004) has argued that variability in a declining language is similar to that of other varieties, including interlanguage. Interlanguage (Selinker, Reference Selinker1972) is a foundational term in SLA because it conceives of a non-native linguistic system as an independent linguistic system that is systematic and dynamic, just as a first language is. One thing that second-language (L2) learners and SSs arguably have in common is a limited linguistic system. Thus, although SSs’ variety is ‘imperfect’, it is possible to investigate it as a viable linguistic system in its own right, just as SLA researchers do with interlanguage. While our intention is not to suggest that language death and adult L2 acquisition are identical (cf. Dorian, Reference Dorian and Andersen1983), a goal in the current study is to continue to apply insights from SLA to the study of first-person-singular subject expression in LF.
In the same vein, we highlight two specific connections between language death—the case of SSs in particular—and L2 acquisition that seem especially relevant for the present investigation. We focus on one approach to SLA that has its origins in Labovian sociolinguistics: variationism (Geeslin and Long, Reference Geeslin and Long2014). First, researchers have identified two types of variability in interlanguage (Rehner, Reference Rehner2002). Type I variation occurs between forms that native speakers use and those that they do not (i.e. forms that only learners use). Type II exists between forms that native speakers use. Research on both types of variation has demonstrated that variability in a non-native language, regardless of whether it is present in the input, can be rule-governed. Thus, since Harrison and Anderson (Reference Harrison, Anderson, Harrison, Rood and Dwyer2008) have argued that SSs can incorporate innovations into their speech, it is worth examining SSs of LF to determine whether they exhibit patterns that diverge from those of more fluent speakers with regard to first-person-singular subject expression and whether this language use is systematic. The second connection concerns individual variability. Growing evidence in L2 variationist research suggests that analyses of aggregate data mask different patterns of variability among individual learners (Geeslin, Linford, Fafulas, Long and Díaz-Campos, Reference Geeslin, Linford, Fafulas, Long, Díaz-Campos, Amaro, Lord, de Prada Pérez and Aaron2013). In other words, documenting individual variability provides an important piece to understanding the observations made by aggregate data about non-native linguistic systems. Just as Dorian (Reference Dorian1994: 634) has attested to the value of examining ‘personal-pattern variation’ in language death situations, we are motivated to pay particular attention to possible differences among speakers.
As we move forward to examine first-person-singular subject expression among LF speakers, applying the aforementioned insights from SLA may lead to new understanding regarding SSs’ linguistic systems. In this vein, we pose two research questions:
1. What forms do LF speakers in Pointe-Aux-Chênes use to express first-person-singular subjects, and how does the inventory of forms differ across speakers in varied age and fluency groups?
2. What are individual strategies for expressing first-person-singular subjects, and how might they be related to the overall inventory of morphological and phonological forms found in this speech community?
To address these questions, we asked participants in this community to translate a list of sentences featuring first-person-singular constructions in English and analysed the resulting translations according to the inventory of forms presented by individual speakers of varying age and fluency backgrounds. Below we describe our data collection and analysis methods.
3 METHODS
3.1 Research site
The data for this study was collected in and around Pointe-Aux-Chênes, a rural fishing town located along Bayou Lafourche in South Louisiana. Pointe-Aux-Chênes is located southeast of Houma, marked in Figure 1 along with New Orleans and the state capital of Baton Rouge.
Pointe-Aux-Chênes, Louisiana
Pointe-Aux-Chênes is predominantly populated by Cajuns and Indians. In general, the portion of the town located closer to the commercial centre of Houma tends to be more Cajun, while further down the bayou towards the Gulf of Mexico tends to be more heavily concentrated with Indians, many of whom are members of the PAC Indian Tribe. The tribe is an independent group that does not belong to the neighbouring United Houma Nation nor to the Chitimacha Confederation of Muscogees, although tribal members consider themselves related to the Chitimacha, Choctaw, Acolapissa, Biloxi and Atakapa tribes. The French spoken by PAC Indians has been examined in several studies over the past 20 years (Rottet, Reference Rottet1996, Reference Rottet, Valdman, Auger and Piston-Hatlen2005; Carmichael, Reference Carmichael2007, Reference Carmichael2017; Dajko, Reference Dajko2009), in part because of the relative robustness of LF in this community compared to other towns in Louisiana.
3.2 Participants
Since Pointe-Aux-Chênes is a small, insular community and not all residents speak French, it was not possible to generate a random sample of participants. Instead, participants were recruited with a focus on balancing across age and fluency groups based on Dorian's (Reference Dorian1981) classifications within language death contexts: older fluent speaker (OFS), younger fluent speaker (YFS) and semi-speaker (SS). Dorian coined the term ‘semi-speaker’ in her research on East Sutherland Gaelic, to categorize individuals in a language death situation who are perceived as non-fluent or otherwise ‘aberrant’. She elaborates as follows:
Unlike the older Gaelic-dominant bilinguals, the semi-speakers are not fully proficient in Gaelic. They speak it with varying degrees of less than full fluency, and their grammar (and usually also their phonology) is markedly aberrant in terms of the fluent-speaker norm. . .[m]ost semi-speakers are also relatively halting in delivery, or speak Gaelic in rather short bursts, or both; but it is not manner of delivery which distinguishes them, since semi-speakers of comparable grammatical ability may speak with very different degrees of confidence and ‘fluency.’ At the lower end of speaker skill, semi-speakers are distinguished from near-passive bilinguals by their ability to manipulate words in sentences. (107)
Thus, the classification of SS encompasses a range of linguistic skill, making recruitment and classification of SS participants a challenge. Part of the definition of a SS relies on community perceptions (Dorian, Reference Dorian1977), in that their deviations from OFS speech patterns are viewed as mistakes, unlike linguistic deviations of YFSs (Rottet, Reference Rottet1995). So, while searching for SSs, community members were asked if they knew anyone they considered to speak baroque (‘broken, strange’) or manière drôle (‘sort of funny’) or simply pas bien (‘not well’). In some cases, such inquiries led to individuals only able to exchange basic greetings in French, who were not included in this study. Those with enough French to be included often stated explicitly that they considered their French to be bad, rusty, or incomplete. On the low end of the SS spectrum, individuals able to produce novel utterances—however halting their speech was—were considered to be SSs and included in this study. On the high end, some SSs could carry on fluid conversations about a variety of topics but featured neologisms or linguistic structures seeming to result from incorrectly parsing or misremembering a given expression. Thus, classification as a SS was based on a combination of community standards of linguistic ability, personal assessment of linguistic ability, and a more general linguistic comparison of SS language patterns with community norms.
The final sample includes 28 informants across three speaker groups: 12 OFSs, eight YFSs, and eight SSs, with each group evenly split between men and women.Footnote 8 Demographic information for individual participants is provided in Appendix A. OFSs ranged from 54–73 years old, while the YFSs (37–44 years) and SSs (28–44 years) were similar in age. All OFSs acquired French as their first language, and many were French dominant. They reported learning English at school, though some explained that they acquired some English in advance of starting school from older siblings who had been to school already. Six out of eight YFSs also reported learning French before English, with the other two explaining that they learned French and English simultaneously in the home. Four YFSs stated that they did not learn English until attending school. SSs, in contrast, all spoke English as their primary language at home. Many reported learning French from an older relative like a grandparent, and often they were the only one of their siblings to speak any French. Some SSs claimed that the person they learned French from was the only person with whom they spoke French, and for a handful of SSs that individual had passed away, meaning they no longer actively used French outside of the interview completed with them. Thus, the experiences of younger speakers in the community were quite different as the decline of French progressed, and for SSs in particular, outlets for speaking in French became limited.
3.3 Data collection and analysis
Data for this study were collected in and around Pointe-Aux-Chênes in 2007–2008. Participants completed a translation task consisting of 50 brief sentences in English read aloud by the first author and verbally translated to LF by the participant. Half of the sentences contained a first-person-singular construction; the full text of these target sentences is provided in Appendix B. The tasks were completed in the homes of the participants, in the presence of a second interviewer with extensive experience in the community.
Recordings were transcribed and coded for expression of LF first-person-singular subjects in all potential contexts. Coding resulted in a total of 707 tokens (20–30 per speaker; 200–250 per speaker group) across the 12 possible realizations of first-person-singular subjects introduced in Table 1. To examine patterns across the dataset, we generated cross-tabulations for subject expression, both according to speaker group and individual speaker. Because of the large amount of variants and the small size of the dataset, it was not possible to submit the data to statistical analysis, thus we report on general trends observed in the cross-tabulations.
4 RESULTS AND DISCUSSION
In this section, we first report on aggregate results across the three age and fluency groups in this study, identifying the inventory of first-person-singular forms used by speakers in each group. We note general trends, then examine individual patterns. We situate these patterns with respect to previous studies of first-person-singular subject expression in this community and adjoining areas (Rottet, Reference Rottet1995; Salmon, Reference Salmon2007; Carmichael, Reference Carmichael2017; Dajko, Reference Dajko2009). We ultimately propose that consideration of the interrelationship between phonological and morphological variation in subject expression may shed light on the patterns observed in the linguistic system of the SSs. Our framing of the analysis draws on conceptual observations in SLA that recognize ‘imperfect’ linguistic systems as autonomous (analogous to the interlanguage of adult L2 learners) and that acknowledge the presence of different types of variability in multilingual contexts.
While all speakers were posed with the same set of sentences to translate, individual speakers varied both in their interpretation of the English sentences and in how they chose to translate specific words or phrases into French. There was also a range of first-person-singular forms in the translations. Some examples of typical subject forms across speaker groups are provided below, where participants all translated the sentence ‘I only planted sweet potatoes in the garden’.
OFS3: J'ai [ʒe] juste planté des patates douces dans le jardin.
OFS6: J'ai [he] planté des patates douces dans le jardin.
YFS1: J'a [sa] juste planté des patates douces dans le jardin.
YFS6: Ai (∅) planté juste les patates douces dans le jardin
SS7: Mon j'a [sa] planté. . .des sweet patates. . .dans le jardin.
SS8: Mon a juste a planté la patates douces dans le jardin.
Table 2 presents the patterning of first-person-singular subject expression across OFS, YFS, and SS groups for all translation sentences. Forms that comprise over 10 per cent of subject expression for a given group are bolded.
Pronoun use in translation task across age and fluency groupsFootnote 9
Table 2 illustrates that in this task, the OFS and YFS groups each demonstrate an inventory of six forms for first-person-singular subjects: five phonological variants of the clitic je as well as ∅. In contrast, the SS group makes use of ten forms. Thus, in the case of first-person-singular subject expression in LF, we do not see reduction or simplification of forms overall; rather, we potentially see a case of innovation by the SSs, resulting in a greater inventory of forms for these non-fluent speakers compared to the OFS and YFS norms.
There also appear to be preferences for specific forms across these groups: OFS and YFS both tend to use (ʒ), (h), and ∅ more than other forms. Situating the results in Table 2 in terms of previous research on LF subject expression, we see surprisingly high rates of ∅ for OFSs and YFSs at 21–26 per cent rates, compared to Rottet's 1–7 per cent rates in 1996 and even Dajko's 10–15 per cent rates in 2009 in similar translation tasks. Rates of (h), (s) and (z) across OFS and YFS are however comparable to other studies in the area (Dajko, Reference Dajko2009).
In contrast to the OFS and YFS data, the SSs show a strong preference for forms featuring mon. For SSs, (mon) and (mon s) comprise over two-thirds of the tokens, while OFSs and YFSs never use mon forms in the translation sentences, with or without the clitic je. Moreover, while ∅ is used 110 times across OFS and YFS data, it only appears six times in SS data. Similarly, while the form (h) is used over one-third of the time by OFS and YFS (175 tokens), it appears only 17 times in SS translations. Overall, it is clear from the aggregate data that the OFSs and YFSs in this community share norms for first-person-singular subject expression, while SSs largely depart from those norms.
To better understand how to interpret the trends in the aggregate data and to avoid assuming uniformity across individuals within this speech community, it is useful to examine individuals separately. In Tables 3, 4 and 5, percentages above 10 per cent are bolded, and percentages above 50 per cent are shaded, to provide a visual indicator of overall trends. Because none of the OFSs or YFSs use mon, we report only on the clitic and null forms in the respective tables.
Rate of use of first-person-singular subject forms by OFS speaker
Rate of use of first-person-singular subject forms by YFS speaker
Rate of use of first-person-singular subject forms by SS speaker
In the OFS data, individual speakers generally show preference for one or two forms, though each speaker makes use of three to five forms across the task. Moreover, all OFSs except for OFS10 make use of ∅, and four speakers (OFS1, OFS2, OFS9 and OFS11) use ∅ more than any other form. A final note about individual OFS data is that no OFS is categorical in their subject-expression use. In contrast, in Table 4 we see that YFS6 categorically uses ∅ in the translation task. Beyond this example, however, YFS individual patterns generally resemble OFS data: individual preference for one or two forms–generally ∅, (ʒ), or (h) –with each speaker demonstrating an overall inventory of three to five forms.
The SS data differs from OFS and YFS patterns in several key ways. First, every SS uses (mon) at least once, while no OFS or YFS use this form in the translation task. Furthermore, individual SS inventories range from 1–7 forms. Thus, not only do SSs as a group make use of more forms than fluent speakers, individual SSs also demonstrate a greater range of variability in their inventory of forms for first-person-singular subject expression.
Table 5 shows that, as with the YFS data, there is a SS speaker who is categorical in his pronoun usage: SS2 uses only (mon). SS3 is also nearly categorical, using (mon s) except for a single token of (mon). When examining the individual data, we see that all the tokens of (h) in SS data derive from a single speaker, SS1. SS1 is also the most fluent SS in the corpus, followed by SS5. Both speakers were able to speak at length in French about certain topics, despite struggling at times with vocabulary, misunderstanding some interviewer comments, and producing non-targetlike morphological forms compared to fluent speakers in the community. SS1 and SS5 each have the highest rates of the most common variants used in OFS and YFS speech: ∅, (ʒ), and (h). Moreover, they are the SSs who use the fewest tokens featuring mon, either alone or with a clitic. In contrast with SS1 and SS6, the categorical (mon) user SS2 is the speaker who was least able to carry on a fluid conversation without switching to English. Thus, even within the SS group, speakers’ fluency levels may have an impact on how closely they converge on the fluent speaker norm.
Comparing the individual results across groups, we can make a few observations. First, there are individual speakers in OFS and YFS groups whose inventories on this task do not include all three of the most frequent forms (ʒ), (h) and ∅; therefore, it is not the case that every OFS resembles every YFS. Second, SSs on the whole look different from the older fluent norm. Two SSs, though, resemble overall patterns in OFS/YFS groups: SS1 and SS5, who show a preference for (ʒ), (h) and ∅ and who use mon tokens the least out of all SSs. Both of these speakers feature a larger inventory of forms (six to seven different forms) than OFS/YFS speakers, who generally use four or fewer forms, so they still differ from the fluent norm. It is not the case that all SSs demonstrate a greater inventory of forms than fluent speakers, however, since some SSs use only one to two different forms. Thus, a third observation we can make deals with the inventories of forms across the speaker groups: Individual SSs feature a broader inventory of forms than OFSs and YFSs do. We can interpret this pattern in terms of previous research on language death situations, which note the paradoxical presence of both increased variability as well as reduction/simplification. And indeed, the greater inventory of forms amongst SSs could be taken as an example of increased variability, while the pattern of certain individuals to use only one or two forms may reflect reduction/simplification. Another common linguistic result of language death is rapid change, which we can observe in comparing the OFS/YFS data with the SS data, where we see a drastic shift towards the presence of mon forms, with and without the clitic je. For all but SS1 and SS5, the mon forms are the most commonly used by SSs in translations. The question arises: if mon is not used by OFS and YFS in the translation task, why is it the dominant form for so many SSs in this same task? And how can we reconcile the fact that tonic pronouns, clitic pronouns, and ∅ are all used as forms of first-person-singular subject expression within this speech community?
Research on subject pronouns in French has not come to a consensus about the status of tonic pronouns like moi versus clitic pronouns like je. Auger (Reference Auger, Parodi, Quicoli, Saltarelli and Zubizarreta1994, Reference Auger2003) and others have suggested that clitic pronouns may be analysed as agreement markers on the verb, thereby explaining why subject doubling (moi, je vais au magasin) is observed and null-subject forms (*∅ vais au magasin) are not. In response, de Cat (Reference de Cat2007) has used syntactic and prosodic diagnostics to argue that attested instances of left dislocation in spoken contemporary French violate several assumptions that follow from classifying clitic pronouns as agreement markers. The issue of null-subject forms in French is also a topic of debate. Kato (Reference Kato1999) describes French as a non-null-subject language, pointing out that Old French once did feature null subjects, but this construction fell out of use when the verbal system stopped distinguishing phonologically between the verb conjugations for each person/number. However, LF features even more verb paradigm-leveling than MF, so following this logic LF should continue to block unexpressed subjects, which of course it does not. Also posing problems for this conclusion is LeBlanc's (Reference Leblanc1996) examination of null-subject forms in the French spoken on Îles-de-la-Madeleine, where she found the highest rates of ∅ in third-person singular and plural contexts—those which would be most ambiguous in terms of the subject. Both Kato (Reference Kato1999) and de Cat (Reference de Cat2007) point to the prevalence of unexpressed subjects and ‘strong-pronoun’ constructions (e.g. moi mettre ça) in children's speech, suggesting that children are in the process of learning the appropriate rule to populate the PRO position within a sentence. While we do not purport to make theoretical claims about the syntactic status of first-person pronouns or unexpressed subjects in LF, comparing the inventories of fluent and non-fluent LF speakers can shed light on the distinctive distribution of these forms within the linguistic system of individual SSs. Below, we propose a potential reanalysis of the tonic pronoun mon as a standalone subject pronoun, which may have been further facilitated within this particular community by the presence of null-subject expression and the phonological variant /h/ in the fluent norm. To provide some context for our analysis, we now turn to previous research on LF subject expression.
Focusing on the same speech community as we analyse in the current study, Rottet (Reference Rottet1996) examined the distribution of (je) (all overt phonological realizations collapsed), (mon), (mon je) and ∅ across three age groups of speakers. Notably, since his data were collected 14 years earlier, our OFSs are equivalent to his middle-aged group, while our YFSs and SSs would be comparable to his youngest group of speakers. Rottet found that speakers ‘at the lower end of the age and proficiency continua’ (131) tended to use higher rates of (mon), (mon je) and ∅, in comparison with the older fluent norm. He cites the use of (mon) as an innovation that is not found in the older fluent norm. As a means of explaining this phenomenon, he turns to hypotheses relating to language acquisition and language contact. Like Kato (Reference Kato1999) and de Cat (Reference de Cat2007), Rottet points out that children acquiring French as their first language acquire stressed pronouns such as moi/mon before clitics, and it could thus be the case that speakers with less exposure to LF remain in an earlier acquisitional phase. He also indicates that English influence may be at play, since English features only non-clitic pronouns and it would presumably be the dominant language for younger and less fluent speakers.Footnote 10 We acknowledge the potential contribution of these effects, though our examination offers several points of departure from Rottet's that may allow for greater nuance in interpreting the patterns observed. To begin with, we do not collapse younger and non-fluent speakers, thus we can tease apart the effects of age (natural language change, albeit in a language death situation) and proficiency (incomplete acquisition). Dubois (Reference Dubois and Donabédian2001) also drew attention to this issue, by examining SSs as a distinct group. In Dubois’ sample of speakers from St Landry Parish, (moi/mon) was specific to SSs and not found in the speech of either OFSs or YFSs. And while Rottet explains that ∅ is on the rise with younger and less fluent speakers, our dataset reveals that SSs in fact use very few tokens of ∅ compared to fluent speakers from older and younger generations of speakers. Thus, ∅ appears to be an innovation centered within the fluent-speaker population, not one motivated specifically by lower proficiency of the speakers, while (mon) appears to be specifically tied to speaker fluency.
Drawing on SLA research (e.g. Rehner, Reference Rehner2002), we can understand this difference in language usage between the YFSs and the SSs in terms of Type I variation (see also Nance, McLeod, O'Rourke and Dunmore Reference Nance, McLeod, O'Rourke and Dunmore2016). As mentioned previously, Type I variation distinguishes between targetlike forms (i.e. those that proficient, adult native speakers use) and non-targetlike forms (i.e. those that they do not use but learners do). That is, Type I variation is representative of variation introduced as a part of the language-learning process and is not found in the fluent-speaking population's speech. The rise in ∅ documented by Rottet does not appear to be an example of Type I variation, since non-fluent speakers make very little use of this feature. Use of (mon), however, does seem to be an example of Type I variation, since it is not present in fluent speakers’ translations of the sentences provided in the task.
It is significant to note that another form exists in the fluent speakers’ pronoun inventories that is nearly absent in the SSs’ speech: that of (h), arising from allophony between /ʒ/ and /h/. Only one SS makes use of (h); none make use of (mon h). In other words, (h) is a case of Type II variation (e.g., variability among targetlike forms) that the majority of SSs have not acquired. Unlike the rise in null-subject forms documented by Rottet (Reference Rottet1996), Carmichael (Reference Carmichael2008, Reference Carmichael2017) and Salmon (Reference Salmon2007) demonstrate a decrease in use of the /h/ allophone amongst younger speakers in the community. Recall furthermore that allophony between /ʒ/ and /h/ in LF is not specific to first-person-singular contexts; first-person clitic je is simply a high-frequency environment for /ʒ/ in LF, compared to many of the other words featuring this phoneme (e.g. jambe ‘leg’, giraumon ‘pumpkin’, nager ‘swim’, jardin ‘garden’). Dajko (Reference Dajko2009) demonstrated that Indians from Pointe-Aux-Chênes featured significantly higher rates of /h/ for first-person-singular subjects (36% in careful speech, 48% in casual speech) in comparison to /h/ in other contexts (12% in careful speech, 6.5% in casual speech). This means that as part of the input, SSs in this community are hearing /h/ as a subject form more than they are hearing it as a general phonological variant for /ʒ/. Therefore, it seems that SSs are likely being exposed to few instances of the allophonic variation between /ʒ/ and /h/ beyond first-person-singular position. This characteristic of the input may mean that the tokens of /h/ in first-person-singular position may be difficult to distinguish from unexpressed subjects, if there is not a pre-existing awareness of (h) as a potential substitute for the clitic je, since /h/ is voiceless with weak formant structure. DuboisFootnote 11 (Reference Dubois and Donabédian2001) further notes that in St Landry Parish, (mon) in SS LF occurs in similar environments to (mon je) in fluent speaker LF. She argues that the common phonological reduction of ‘je’ within (mon je) may have contributed to the reanalysis of (mon) as standalone pronoun. Such reduction is also common in Pointe-Aux-Chênes and may be an additional contributing factor. Considering these potential phonological effects alongside the points that Rottet makes related to the markedness of (mon) allows us to build a case for a potential interaction between the phonological and morphological forms attested in Pointe-Aux-Chênes. We propose that the patterns of (h) and ∅ in the target speech within Pointe-Aux-Chênes may contribute to the increased presence of (mon) in the linguistic system of SSs in our sample.
Examination of individual SS results provides further evidence for this interpretation. Above it was noted that only two speakers use (h) or ∅ in the translation task: SS1 and SS5. That is, none of the other SSs feature these forms in their individual inventories of first-person-singular forms. Given the community-wide patterns described above, we propose that the Type II variation present in the community amongst (ʒ), (h), and ∅ is simply not being acquired by some individual SSs. Instead, these SSs appear to be noting the presence of the more marked ‘strong pronoun’ mon and reinterpreting that particle as a single, standalone pronoun: (mon). SS2, SS4, SS6 and SS8 all feature high rates of (mon), while SS3 and SS7 feature high rates of (mon s). Notable within SS3 and SS7’s data is the comparably low rate of (mon). Based on this distribution, we suggest that these two speakers may have interpreted (mon s) as a single morpheme. Additional evidence for this interpretation comes from prosodic/segmentation factors: these speakers also featured instances in their speech where they began sentences with (mon s) pronounced as a single element /mɔ̃s/, pausing before the next word. Therefore, within our SS group we have two SSs who seem to align more closely with the fluent-speaker norm in their first-person-singular subject-expression inventories (SS1 and SS5), while the other six may have introduced Type I variation into their linguistic system in the form of the reanalysis of (mon) and (mon s) as potential first-person-singular forms. We have suggested that a contributing factor in this reanalysis may be the presence of null-subject forms and the general allophony between /ʒ/ and /h/, both of which prevented acquisition of the clitic form (je).
5 CONCLUSION
In this study, we sought to compare the use of first-person-singular subject-expression forms by OFSs, YFS and SSs, focusing on this feature where increased variation (in terms of forms of subject expression) interacts with reduction/simplification (in terms of the verb paradigm) in a language death context. In our examination of a dataset of translated sentences, we found that OFSs and YFSs draw from similar strategies for expressing first-person-singular subjects–namely the clitic pronouns (ʒ) and (h) and null-subject forms–even though individual speakers may not use each variant in identical rates. In contrast, many SSs used (mon) or (mon s) in high rates, with very few tokens of (h) or ∅ present across SSs. We also noted that when examining individual SSs, the only two to use (h) or ∅ did not use (mon) or (mon s) at high rates. Taken in conjunction with previous research demonstrating a rise in null-subject forms and a concomitant reduction in overall use of the /h/ variant (within first-person-singular constructions, as well as in other contexts), we proposed that the variation in the fluent-speaker forms may have facilitated the reanalysis of the pronoun form, as the input for SSs is ambiguous in terms of the presence of clitic pronouns. Three methodological and conceptual developments compared to previous studies, which were motivated by SLA research, allowed us to make the case for this reanalysis: (1) the conception of SSs’ linguistic system as an autonomous system analogous to adult language learners’ interlanguage, (2) the difference between Type I and Type II variation and (3) the importance of examining individual trends. The application of these SLA insights to the analysis of SS speech lends additional support to Wolfram's (Reference Wolfram, Chambers, Trudgill and Schilling-Estes2004) claim regarding language variation and change in a language death situation. Wolfram argues that there is little empirical data to support the dominant view that variability in language death is different from that in healthy language communities. In contrast, he suggests that variation and change in these contexts is systematic and similar to processes observed in ‘healthy’ speech communities. He notes, however, that the rate of change in a language death context is more evident and widespread. Some of what influences the rapidity of linguistic change are the social changes that go on in a declining language's speech community, as younger speakers and non-fluent speakers—or SSs—have less access to the dying language than the older fluent generation. We suggest that in examining SS speech as a linguistic variety worthy of study in its own right, we can draw conclusions about linguistic processes in the broader community. In order to continue with this line of research, it is therefore crucial to analyse spontaneous speech to determine whether the current study's observations generalize to these speech contexts. A dataset of spontaneous speech that is large enough to undergo statistical analysis will also enable us to investigate the linguistic and extralinguistic factors that condition the variation of first-person-singular subject forms in LF.
PARTICIPANTS’ DEMOGRAPHIC INFORMATION
APPENDIX B: TRANSLATION SENTENCES FEATURING FIRST PERSON SINGULAR SUBJECTS
1. I already paid the young man.
2. I broke my leg.
3. I have to change my shirt before we go to Houma.
4. I was ashamed to speak French at school.
5. I ate jambalaya today.
6. I would rather eat a peach than an orange.
7. I fixed the car today.
8. I ate until I could not eat anymore.
9. I threw away the shoes she bought us.
10. I like to sing at church.
11. I eat eggs and ham for breakfast.
12. I listen to the radio every day.
13. I will never change my mind.
14. I will never forget my first day of school.
15. The shrimp I ate made me sick.
16. I bought something for the children.
17. I walked until my legs hurt.
18. I never dive into the bayou.
19. I try not to judge my friends.
20. I chose the red pants.
21. I only planted sweet potatoes in the garden.
22. The milk was sour so I threw it away.
23. When I was young, I used to go to school on the school boat.
24. I said hello to the old men.
25. I have to wait until Thursday.
