2.1 Motivation

In an obsolescing language variety, the driving force behind sociolinguistic research may be different in at least three ways from research in a healthy language community. First, it will likely not be possible for the endangered-language researcher to add to literature on the critical period or the transition between adolescence and adulthood (e.g., Cukor-Avila Reference Cukor-Avila2002, Wagner Reference Wagner2012, Rickford and Price Reference Rickford and Price2013), since an endangered language is characterized by the cessation of transmission of the variety to children. Fluent Louisiana French speakers, for example, are generally over 60 years old; the youngest speaker in this case study was 58 when he was interviewed in 2010. Likewise, studies of intergenerational transfer may not be possible (but see Dubois and Noetzel Reference Dubois and Noetzel2005 for work with Louisiana French). However, endangered varieties do present interesting opportunities for research into language variation among working adults, retirees, and elders, potentially in the absence of normative language-internal pressures; this is the motivation for studying liaison in Louisiana French. Such research responds to a need for more variationist analysis in post-adolescent language variation and change (Eckert Reference Eckert and Coulmas1997).

Second, as with endangered indigenous languages, the researcher's position within and responsibility to the language community will likely be different from what it would be in a situation of linguistic stability (Dorian Reference Dorian2010). Naturally, all sociolinguists are called to give back to the communities they study (Wolfram Reference Wolfram, Chambers and Schilling2013), but the variationist may be called to participate in and contribute to larger community efforts revolving around language revitalization and language rights, depending on the wishes of community members (Nettle and Romaine Reference Nettle and Romaine2000). While this relationship is more involved and complicated, it can also be more rewarding and mutually beneficial than a traditional researcher-participant relationship: the linguist is in a position to contribute positively to the lives of those who are providing valuable sociolinguistic information, and can use social-scientific research to advocate for legislative and institutional support for the endangered language community (Dorian Reference Dorian1987). Such action can build trust between the language community and the researcher, facilitating longitudinal research and making a substantial impact outside of academia (see Good Reference Good, Austin and Sallabank2011 for details on language resource creation, and Rice Reference Rice2006 for an overview of the ethics of endangered language documentation).

Third, the factors conditioning language variation patterns may not include the usual sociolinguistic factors of age, sex, socioeconomic status, or education (Dorian Reference Dorian1994); indeed, work with indigenous minority-language communities suggests that clan, exogamy, gender-based access to prestige, and majority-language contact are also influential extralinguistic factors (Stanford and Preston Reference Stanford and Preston2009). In French-speaking Louisiana, however, common sociolinguistic factors appear to be active: some studies have found that sex and speech style are most important for explaining certain language variation patterns (Carmichael Reference Carmichael2007, Reference Carmichael2008; Blainey Reference Blainey and Mailhot2009, Reference Blainey, Neveu, Toke, Klingler, Durand, Mondada and Prévost2010) while others have found that age is the strongest explanatory sociolinguistic factor (Salmon Reference Salmon2007, Reference Salmon2009). Still others have reported geographically-based variation patterns (Baronian Reference Baronian2005, Dubois Reference Dubois, Valdman, Auger and Piston-Hatlen2005, Dajko et al. Reference Dajko, Crossland, Fort, French, Goss, Johnson, Hinrichs, Powers, Pulwer, Schneider, Shipps, Snider and Williams2008, Blainey Reference Blainey2015), and those examining questions of ethnicity have underscored its importance for language variation patterns in Francophone Louisiana (Dajko Reference Dajko2009, Reference Dajko2012)Footnote ¹ . As a result, it is important to consider all potential extralinguistic factors at the analysis stage. The case study of Louisiana French liaison included the sociolinguistic factors of sex, generation, speech style, and home town.

2.2 Goals and questions

Because of these potential differences in motivation, the goals and questions of variationist studies in threatened varieties of Western languages are also likely to be distinct from those of studies in vibrant language varieties. One line of questioning centres around the amount of variation that exists in obsolescing language varieties. Will research on threatened varieties of Western languages confirm that language endangerment consistently results in increased rates of language change, as asserted in Cook (Reference Cook1995) and O'Shannessy (Reference O'Shannessy, Austin and Sallabank2011), or will it show that different endangered language communities have different degrees of language variation and change, as put forward in Connell (Reference Connell2002)? If differing levels of variation are uncovered, it will be important to verify that the type of linguistic variable that is investigated (e.g., syntactic, morphological, phonological) and the study design are not correlated with the level of reported variation. In the case of Louisiana French liaison, the level of variation is lower than for other sound patterns, such as the substitution of [h] for /ʒ/ (Carmichael Reference Carmichael2007) or /ε/ lowering before /r/ (Blainey Reference Blainey2015). This may lend support to Connell's (Reference Connell2002) hypothesis.

Another question driving sociolinguistic research in threatened language communities has to do with the relation between language contact and language variation.Footnote ² How can we tell if the (absence of) variation is due to language contact, and that it is not simply the nascent, incipient or completed change observable in any language? Syntactic evidence from contact varieties of Western languages provides a nuanced view of this question. At one end of the spectrum, in their examination of minority French varieties spoken in the Channel Islands and in North America, Gadet and Jones (Reference Gadet and Jones2008) demonstrate that virtually none of the syntactic changes that they examine are due to language contact alone, in keeping with the “multiple causation” principle defined in Thomason and Kaufman (Reference Thomason and Kaufman1988: 57, cited in Gadet and Jones Reference Gadet and Jones2008: 239) and with Silva-Corvalán's (Reference Silva-Corvalán, Seliger and Vago1991, Reference Silva-Corvalán2008) research on Spanish in the United States. In the middle, Neumann-Holzschuh's (Reference Neumann-Holzschuh2009) work on Acadian and Louisiana French syntax shows that some processes corroborate the results of Gadet and Jones (Reference Gadet and Jones2008), but that others, such as preposition stranding and accusative object pronoun postposition, appear to be due to language contact and convergence alone (Neumann-Holzschuh Reference Neumann-Holzschuh2014)Footnote ³ . At the other end of the spectrum, Mougeon et al. (Reference Mougeon, Nadasdi and Rehner2005) focus on post-verbal prepositions and certain verbs in Ontario French, demonstrating that all eight of the innovations they observe are solely due to language contact. Moreover, they conclude “that there are thresholds of contact with English below which or above which innovations that deviate more or less from the traditional norm will or will not emerge” (Mougeon et al. Reference Mougeon, Nadasdi and Rehner2005: 113).

Based on these studies, it appears that certain variable parts of a language's syntax, such as French prepositions, may be prone to copying in language contact situations. Indeed, in Louisiana it is the study of intergenerational transfer of preposition use that reveals effects of language contact and bilingualism (Dubois and Noetzel Reference Dubois and Noetzel2005). However, much remains to be understood about how contact affects variation, and which components of the grammar are most easily affected by language shift. As a phenomenon that implicates both French phonology and syntax, liaison can involve prepositions (e.g., dans_un parc ‘in a park’). This makes French liaison an intriguing subject of investigation for questions of language contact and obsolescence.

In sum, it is important for the researcher of an obsolescing language variety to be cognizant of potential differences in motivation and goals at the outset of the research project, so that analyses resulting from the research can be properly contextualized within the sociolinguistic literature. The Methods sections explores how researchers can adapt protocols designed for vibrant language varieties to suit the context of an endangered language variety, by taking into account questions of available speakers, literacy, age, past language-related trauma, and power dynamics with community outsiders.

3.1 Variables to study

In sociolinguistic fieldwork, “[t]he most frequently studied variables are phonological and morphological” (Feagin Reference Feagin, Chambers and Schilling2013: 23), but in language death research, “most work on the role of convergence in attrition has focused on aspects of the morpho-syntactic restructuring of the attriting language” (Bullock and Gerfen Reference Bullock and Gerfen2004: 95). Fortunately, more recent work features phonetic and phonological analyses of endangered languages (e.g., Stanford and Preston Reference Stanford and Preston2009, Smakman and Heinrich Reference Smakman and Heinrich2015). The current study of Louisiana French liaison contributes the perspective of a complex phonological process in an endangered variety of a Western language.Footnote ⁴

Liaison is a sandhi phenomenon: a liaison consonant appears between two words which, when said in isolation, do not contain the liaison consonant (e.g., in Louisiana French, mes ‘my’ [me] + amis ‘friends’ [ɑ̃.mi] → mes amis ‘my friends’ [me.zɑ̃.mi]). By convention, the first word in the pair (e.g., mes ‘my’) is referred to as W1, while the second word (e.g., amis ‘friends’) is referred to as W2. Variable liaison environments have been linked to language learning and literacy (Laks Reference Laks2005), making an orally transmitted variety such as Louisiana French an interesting testing ground for variability in French liaison.

3.2 Addressing questions of literacy in the endangered language variety

In order to create a workable research methodology for studying an endangered variety of a Western language, it is often necessary to modify a larger project's general protocol so that literacy is not a prerequisite for participation. Regarding French liaison, reading exercises can be helpful for examining the effect of the written word on adults’ variable liaison use. However, only oral exercises are appropriate with Louisiana French speakers. This presents the uncommon and exciting opportunity to explore the evolution of French liaison in the absence of a written norm. It is important to remember as well that some elicitation tests may be inappropriate in a situation of language endangerment. For example, tests of linguistic insecurity (e.g., as outlined in Labov Reference Labov, Baugh and Sherzer1984) are not effective in a community that does not have an overarching prestige variety, such as the Louisiana French community.

The most recent recordings of 20 speakers for the current case study (Blainey Reference Blainey2013) rely on Klingler and LaFleur's (Reference Klingler, Lafleur, Detey and Nouveau2007) adaptation of the global Phonology of Contemporary French project (Phonologie du français contemporain [PFC], Durand et al. Reference Durand, Laks, Lyche, Pusch and Raible2002, Reference Durand, Laks, Lyche, Durand, Laks and Lyche2009) for French-speaking Louisiana. The PFC has as its general goal to capture French as it is spoken around the world, focusing on the dynamics of French schwa and liaison patterns. Part of the linguistic interview requires the speaker to read a word list and a short passage in “Standard” French. As this would not be appropriate in Francophone Louisiana, where speakers do not read and write in French and where use of “Standard” French serves to remind them of the self-perceived inferior status of their own French, Klingler and LaFleur (Reference Klingler, Lafleur, Detey and Nouveau2007) employ translation exercises as substitutions that maintain the formality and structural consistency of the reading exercises. Due to widespread lexical variation in Louisiana French, Blainey (2013) further tailors the translation activities in order to make them suitable for Lafourche Parish Louisiana French speakers, a method that will likely be necessary for any endangered-language-variety fieldworker using a protocol designed for a larger language.

3.3 Archival data collection

Recent sociolinguistic research has greatly benefited from longitudinal language data (see Sankoff Reference Sankoff, Ammon, Dittmar, Mattheier and Trudgill2005, Wagner Reference Wagner2012 for overviews), allowing researchers to test (either by using panel or trend studies) the apparent-time hypothesis (Labov Reference Labov1963, Reference Labov1966), as well as to examine the different types of change that can happen over a person's lifespan (see Labov Reference Labov1994, Sankoff and Blondeau Reference Sankoff and Blondeau2007). However, the reason a language variety is endangered is often that it has been historically devalued, as a method of communication, as a mark of identity, and as an object of social-scientific study. This lowers the chance that previous linguistic recordings exist for the language variety. Of course, it is always possible that previous linguistic work has been carried out, especially for endangered varieties of Western languages. For example, the earliest phonetic transcriptions of Louisiana French used in the current case study come from the appendix of John Guilbeau's doctoral dissertation, from seven speakers recorded in the 1930s and 1940s (Guilbeau Reference Guilbeau1950). Nevertheless, it is crucial for the sociolinguist to explore all possible sources of diachronic information. One potential trove of past recordings, particularly for enclave varieties of Western languages, may come from non-linguistic academic research. In Louisiana, for example, folklorists such as Alan Lomax and Irene Whitfield made recordings of Louisiana French musicians singing in their native tongue as early as the 1930s (Whitfield Reference Whitfield1939, Caffery Reference Caffery2014). As part of the Louisiana Project (Projet Louisiane), Canadian anthropologists and sociologists also conducted interviews with Louisiana French speakers in the 1970s in order to better understand the social networks and ways of life that held Francophone communities together in the state (Breton and Louder Reference Breton and Louder1979, Louder and Waddell Reference Louder and Waddell1979). Such musical compendia and social-scientific interview corpora can sometimes be found with the help of university archivists and reference librarians. Again referencing the case of Louisiana French recordings, the extensive Archives of Creole and Cajun Folklore at the University of Louisiana at Lafayette's Center for Louisiana Studies (2011) provide a searchable database of Louisiana French recordings from various research initiatives, such as those mentioned above. The current case study draws from these archives, selecting 12 interviews made in 1977 by Alain Larouche (Larouche Reference Larouche1979, Reference Larouche1980) as part of the Louisiana Project. Conathan (Reference Conathan, Austin and Sallabank2011) includes further ideas on where to find archival language materials, as well as recommendations to researchers on how to create durable, archive-ready materials.

Other sources of longitudinal linguistic data may be community cultural centres, churches, and libraries. Libraries provide a drop-off point to community members who discover old cassettes or communication records; cultural centres may have historical displays of local lifeways and language use; and churches often contain valuable birth, death, and marriage records. Employees at these centres often have a passion for digging up old records and recordings, and can also prove to be good allies for making contact with native speakers. Furthermore, unpublished or locally published historical accounts of the speech community may only be available at these locations, revealing to the researcher useful lines of questioning for sociolinguistic interviews.

Online, researchers can consult video-sharing sites such as Youtube, or social media sites like Facebook, to find publicly available recordings of speakers. There are many video recordings of older fluent Louisiana French speakers available on Youtube, for example. In terms of more general demographic information, ancestry-building websites can prove useful in the event that public records are scarce. This is true for Louisiana French, since hurricanes, heat, and humidity have led to the loss or degradation of public records: genealogical information furnished by Louisiana natives and distant relatives in the French-speaking world can help to piece together past marriage, living, and migration patterns of the Louisiana French speech community.

Finally, by consulting with community members directly, researchers can sometimes benefit from family records, old recordings, or personal letters (if the language variety has a written code). This brings us to the question of finding, selecting, and contacting native speakers for interviews.

3.4 Discovering the demographics of the speech community

The sociolinguistic methodology for interviewing speakers of a healthy language variety normally specifies X speakers per pre-defined subgroup (e.g., generation, sex, socioeconomic status, level of education), but this model is not straightforward for threatened language varieties. The small number of speakers available for interview can be a daunting reality for the fieldworker's enterprise with an endangered variety of a Western language. Of those community members who are fluent speakers of the language variety in question, only a subset are typically willing to speak their native tongue in the presence of outsiders. Which regions have the highest concentrations of fluent native speakers? How many fluent speakers who are older or younger, male or female, urban or rural (and so forth) can one hope to find and interview? This calls to mind Labov's famous characterization of historical linguistics as “the art of making the best use of bad data” (Labov Reference Labov1994: 11). The challenge for the researcher is to use effective planning and strategies to maximize the pool of potential native speakers.

Census records and previous studies working to establish the parameters of the speech community can be enormously helpful. For instance, the United States census provides the basis for general estimates of the vitality of French varieties in Louisiana. Census statistics must be approached with caution in the case of Louisiana, both because of varying use of language labels across the state, and because of changing wording over time on the census itself, but they do provide a general picture of French language use in the State. Data for the current case study come from Golden Meadow, Lafourche Parish, due to a much higher than average concentration of native French speakers in this town of 2,200 people (28% French-speaking in Golden Meadow versus 3% French-speaking in Louisiana (Census Bureau 2011)). To give two examples of academic studies scrutinizing the sociolinguistic context in Louisiana, Neumann-Holzschuh (Reference Neumann-Holzschuh2014) provides a better understanding of the history of language and dialect contact in Francophone Louisiana, and Blyth (Reference Blyth and Valdman1997) lays out the contemporary sociolinguistic situation of Louisiana French. As for the documentation of field methods and speaker attitudes, Dubois details her extensive survey of four Louisiana parishes, looking at linguistic fluency, ethnic identity, and language attitudes, notably showing that “specific sociolinguistic forces (i.e., age, gender, and geographic factor) influence the attitude of the respondents” (Dubois Reference Dubois and Valdman1997: 68). The work includes excellent recommendations for fieldwork methods in Louisiana. Taken together, these resources reveal to the sociolinguist what kinds of demographic challenges and particularities the enclave speech community may present.

One such particularity is the question of how to discover and define generational divides. As with all linguistic research, speakers will not likely fall into neat decade-based generational categories. Rather, generational divides correspond to significant changes in the community's relationship with the contact language. In Louisiana, one such change was the arrival of the petroleum industry in rural communities (Larouche Reference Larouche1979). There are government-funded projects on how large industries have impacted communities along the Gulf Coast; as a result, the effects of the oil and gas industry on the area have been painstakingly studied and documented, complete with firsthand accounts (McGuire Reference McGuire2004). Furthermore, Dubois (Reference Dubois and Valdman1997: 48) establishes a Linguistic Ability and Background index, defining different generations of speakers based on their language learning environment as children (Dubois Reference Dubois, Valdman, Auger and Piston-Hatlen2005: 288–290). This research sheds light on how generational divides can be classified, based on socioeconomic and sociolinguistic changes in the community through time.

Such resources also aid the sociolinguist in the essential task of assessing each speaker's fluency in the enclave variety. Correctly assessing speaker fluency is important because semi-speakers (Dorian Reference Dorian1977) have markedly different grammars than older fluent speakers, as has been shown in Louisiana French (Rottet Reference Rottet2001). Depending on the project and its aims, as well as the demographics of the speech community, the researcher may need to draw exclusively on the expertise of older fluent speakers (as is true for the PFC project), but for larger documentation projects such as Rottet (Reference Rottet1995) and Dubois (Reference Dubois and Valdman1997), it is best to interview the widest array of speakers possible. The simplest approach, when seeking out potential interviewees, is to rely on the expertise of a local native speaker. The next section focuses on how to find such a person.

3.5 Community co-investigator(s)

Working with a community insider can help the field researcher avoid many pitfalls, and this is particularly true in communities of endangered language varieties. In order to make contact with a local native speaker who is willing to collaborate on a research project as a co-investigator, there are several options available. First, depending on the amount of existing literature on the language variety in question, the researcher can contact other academics who have interviewed native speakers, asking for suggestions regarding potential local contacts; this was the method followed for the 2010 Louisiana French interview recordings. Second, the researcher can contact community or government agencies linked to language empowerment, maintenance or revitalization efforts, where interested local activists and fluent speakers are likely to convene. These types of introductions to potential co-investigators through an established person or agency are optimal, because they help to mitigate natural feelings of mistrust on the part of locals meeting an outsider with a microphone and a project protocol.

If such intermediaries are not available, a third option is to travel to the community directly, and to speak to community leaders such as shop owners, religious leaders, or political leaders, explaining the project's aims and asking for introductions to local native speakers who might be willing to participate in the project as co-investigator(s). Note that while the third option may work very well in a healthy language variety (Labov Reference Labov, Baugh and Sherzer1984), the political or religious power structures in place may be the very forces that are driving language endangerment, so that there may not be feelings of trust between authorities and local people when it comes to questions of language. However, having a local native speaker as an ally and co-investigator in the community is very important, making the selection of at least one of these options imperative.

A brief note on the researcher's sex: it is advantageous (depending on social norms, local customs, and availability in the language community) to find a co-investigator of the opposite sex, because endangered language researchers have reported discrepancies in speaker participation that appear to depend on the interviewer's sex. For example, Nagy (Reference Nagy2000: 149) notes that “the men in [her] age range were inhibited, to some extent, by social norms, which do not permit [one-on-one and small-group conversations] between ‘marriageable’ members of opposite sexes, unless an engagement is meant to be the outcome.” Likewise, in the case of Louisiana French, Rottet (Reference Rottet1995: 309) reports that “most (if not all) of the people who refused to be interviewed were women. This could be related to the fact that the researcher is male, but it could also be indicative of language attitudes.” Since sex can still be an active sociolinguistic factor in situations of language shift (Cavanaugh Reference Cavanaugh2006), the sex of the co-investigator (assuming that there is a choice of co-investigator) is an important consideration as the researcher prepares to go into the field, so that a balanced speaker sample may be obtained.

3.6 Making contact and minimizing discomfort

By working with a community insider, the researcher can also minimize other instances of native speaker embarrassment and refusal. In the heart of Francophone Louisiana, it can suffice to approach individuals during cultural festivals, religious events, or in the community itself, simply asking vous-autres parle cadien? (‘do you speak Cajun?’) (e.g., Dajko et al. Reference Dajko, Crossland, Fort, French, Goss, Johnson, Hinrichs, Powers, Pulwer, Schneider, Shipps, Snider and Williams2008), but not all Louisiana French speakers feel comfortable using their native tongue with strangers. Moreover, in other language contact situations, such behaviour may intimidate speakers and/or violate cultural norms. Consultation with a community insider co-investigator before approaching native speakers can circumvent such problems, and the co-investigator's presence at initial meetings can encourage participation and trust on the part of native speakers.

The question of power dynamics and language use is of utmost importance when the researcher approaches and converses with endangered language speakers. It is crucial to discover which languages or varieties will provoke a negative reaction, and the best course of action is again to seek advice from local language activists, authorities in the linguistic community, and the community insider(s) with whom the researcher is collaborating. In the case of Lafourche Parish Louisiana French, for example, English language use was forced upon children in the school system beginning in the 1920s (State of Louisiana Constitution, Art. 12, §12, 1921), and upon adults working in the oil and gas industry beginning in the 1930s (Larouche Reference Larouche1979). While some speakers avoided the school-based language indoctrination (along with the physical and emotional abuse it entailed) by not attending class, the influx of foreign Anglophone workers to the oil fields resulted in considerable intercultural and interlinguistic conflict:

As a result, one might assume that English language use should be avoided with Louisiana French speakers. However, beginning in the late 1960s, language revitalization efforts promoted foreign varieties of French in the school system (Henry Reference Henry and Valdman1997: 194). Because of this, older fluent Louisiana French speakers were confronted with grandchildren who attempted to speak to them in French, but who used non-native vocabulary, constructions, and pronunciations; this served to drive home feelings that Louisiana French was a broken and inferior dialect that did not deserve to be taught in school (Ancelet Reference Ancelet1988: 351). Thus, it is “Standard” French that is the shame-inducing speech variety, while English is now seen as a neutral form of communication.

Learning this lesson can be costly. In one early instance, the author began interviewing a fluent Louisiana French speaker with two other L2 French interviewers, and accidentally used a construction that does not exist in Louisiana French (qui fais-tu pour la vie? instead of qui tu fais pour la vie?, ‘what do you do for a living?’). For the native speaker, this was sufficient evidence that the author spoke le bon français and therefore would not understand this person's French, effectively ending the interview. Fortunately, the Louisiana French speaker agreed to talk to the other interviewers, answering the author in English for the remainder of the interview.

So then, what if the researcher speaks the very speech variety that inflicts feelings of shame on older fluent speakers, making them unwilling to converse in their native tongue? The onus is on the researcher to avoid using the offensive language variety, working to use the local variety as much as possible. One strategy is to present the researcher as an interested outsider who would like to learn the language, and to ask speakers’ forgiveness in advance for any errors. This places native speakers in their rightful respected place as language experts and teachers. It also firmly establishes the researcher as a community outsider in contrast to the community insider co-investigator, encouraging a stylistic differentiation between insider- and outsider-directed speech (Bell Reference Bell1984).

By the same token, translation or reading activities may have a negative effect on speakers because they recall childhood traumas suffered in the school system. In French-speaking Louisiana, many speakers express feelings of silliness when they are asked to complete counting exercises (e.g., une école, deux écoles, trois écoles… ‘one school, two schools, three schools…’), or to translate strange words or sentences into French. Worse, speakers can become frustrated and upset that they cannot remember how to translate a word or phrase. The field researcher must work to reassure native speakers that there are no right or wrong answers, and that the responses that they have provided are valuable for research and for future language learners. In the case of Louisiana French liaison, for instance, counting exercises helped to confirm that liaison [z] functions as a morphological plural marker on nouns.

Finally, it can be awkward for the community insider to carry on an informal conversation with other native speakers, precisely because in smaller, tight-knit communities, the co-investigator may already feel that theyFootnote ⁵ do not have many questions to which they do not already know the answers. In such instances, going through potential questions (such as those modeled in Labov Reference Labov, Baugh and Sherzer1984) with the co-investigator in advance may provide them with a set of go-to questions. Talking with the co-investigator about any curiosities they may have about social and family dynamics, customs, or lifeways in the community's past may also lead to questions they consistently ask informants. In Louisiana, for example, a virtually fail-proof line of questioning revolves around food preparation and consumption; this is a recurrent theme in the community insider interviews of Blainey (Reference Blainey2013).

4.1 Tests of statistical significance

Tagliamonte (Reference Tagliamonte2006) provides an excellent guide for variationist statistical tests, focusing on logistic regressions in Goldvarb (Sankoff et al. Reference Sankoff, Tagliamonte and Smith2005). Contexts of language endangerment, however, may not provide the quantity or type of sociolinguistic data necessary for logistic regression analysis. Furthermore, with the increased likelihood that individual speakers have “personal-pattern variation” (Dorian Reference Dorian1994: 635) that cannot be explained by larger sociolinguistic groupings such as age, sex, or level of education, it is important for sociolinguistic research to incorporate the individual in statistical analyses whenever possible. Mixed-effects models using R (R Core Team 2015) or R-based variationist platforms like Rbrul (Johnson Reference Johnson2008) can control for factor group interactions and include the speaker as an external factor group. Again, such models require larger numbers of tokens than may be available from sociolinguistic fieldwork in an endangered community. In the case of Louisiana French liaison, despite a reasonable number of tokens for 38 speakers (3,317 tokens), a mixed-effects model was not possible due to variation in the number of tokens contributed by each speaker (ranging between three and 103) and forms of invariability that further eliminated most tokens from the analysis, which will be discussed in section 4.2. As a result, the analysis relied on bivariate parametric tests of significance. While mixed-effects models may be the ideal analytical tool for variationist sociolinguistic analysis, research involving endangered varieties of Western languages may not be in a position to use them.

4.2 Data characteristics

After collecting, coding and analysing the data, the sociolinguist may discover that there is not as much variation as previously thought, as has sometimes been the case in healthy language varieties (D'Arcy and Tagliamonte Reference D'Arcy and Tagliamonte2015). This is certainly true of Louisiana French liaison patterns in Lafourche Parish. As outlined in section 3, the analysis considered the speech of 38 Louisiana French native speakers living in South Lafourche Parish, recorded at three points in time: 1939, 1977, and 2010 (Guilbeau Reference Guilbeau1950, Larouche Reference Larouche1979, Blainey Reference Blainey2013).Footnote ⁶ When there does not appear to be variability in the data set, despite predictions to the contrary, it is important to verify that nothing has been overlooked. In the case of Louisiana French liaison, multiple aspects of the token set proved to be invariable and imbalanced, which greatly reduced the number of tokens for analysis. First, there was the question of which words could form liaison pairs, in terms of the syntactic identity of the first word (W1) and the second word (W2) (e.g., Noun-Determiner, Pronoun-Verb, Preposition-Determiner). Of 97 syntactic W1-W2 combinations, 84 were invariable, and only four of the variable W1-W2 pairs (representing 34 tokens) had liaison rates between 33% and 85%. Indeed, the apparent variability of those four W1-W2 types may have simply been a result of low token numbers. Second, when only the variable W1-W2 combinations were considered, there was no change in the rate of liaison over time, and chi-square tests did not reveal significant effects of other sociolinguistic factors such as sex or generation, even though phonological research on Louisiana French liaison has noted age-based changes in liaison use (e.g., Boutin and Lyche Reference Boutin and Lyche2014 ).

Third, when the liaison consonants (/n, t, z/) were compared for each variable W1-W2 combination type, in the majority of cases one or more of the liaison consonants was categorical. Fourth, once the liaison consonants were separated within the variable W1-W2 pairs, lexically-based categorical behaviour became apparent. Indeed, the three most frequent W1s (on, en, les) accounted for over half the tokens in this subset of the data. Thus, while the invariability encountered within the data set ruled out the possibility of using mixed-effects models by eliminating too many tokens, it also revealed considerable stability in a situation of language contact and shift. The case study of Louisiana French liaison demonstrates that variationist research involving obsolescing varieties of Western languages may produce this type of unexpected result, further enriching knowledge about language contact and variation.

The sociolinguist may also find similarities to healthy language varieties in the patterns present in endangered language varieties. In Louisiana French, cases of invariable (and nearly invariable) W1-W2 liaison pairs reflected liaison trends in other varieties of French. For example, combinations such as a determiner followed by an adjective (the first two words in un autre homme ‘another man’) always liaised, while pairs such as a verb followed by a preposition (the last two words in il parlait avec ‘he spoke with’) never liaised. Moreover, the majority of categorical W1-W2 liaison behaviour (78 of 84 W1-W2 pairs) represented non liaison, and largely corresponded to environments where liaison is infrequent in other French varieties. The (nearly) invariable patterns in Louisiana French thus represent a highly regularized version of trends in other varieties of French, where liaison has been described as both “profoundly variable and unstable” (Durand et al. Reference Durand, Laks, Calderone and Tchobanov2011: 130) as well as “extraordinarily constrained and limited” (Durand et al. Reference Durand, Laks, Calderone and Tchobanov2011: 131).