Unnatural bedfellows? The sociolinguistic analysis of variation and language documentation

ABSTRACT

Sociolinguistics and language documentation are both sub-fields of linguistics dealing with variation in natural language, but how they approach variation differs. This paper suggests that sociolinguistic studies of probabilistic variation can contribute not only to more thorough documentation of the structure of a lesser-known language, but also contribute to the advancement of linguistic theory and language change over time. Conversely, central concepts in sociolinguistics may need to be reconsidered when researchers begin to look beyond urban, monolingual communities of speakers. Two case studies are presented: the first case study of Bislama pronouns and subject agreement shows that documenting the structure of a lesser-studied language can help elucidate what might have happened in the past. It also shows how an analysis of variation can address outstanding questions about the fundamental linguistic structure of an under-documented language and the ways that variation feeds into grammaticalisation. The second case study presents an analysis of another variable in Bislama that problematises the alleged simplicity of creole languages, and proposes a general principle of transformation of constraints in situations of language contact. The paper suggests that, although the combination of sociolinguistics and language documentation may always be uneasy, they can make surprisingly congenial bedfellows.

KEYWORDS:

• Sociolinguistics
• Vanuatu languages
• language variation
• language change
• language documentation

Introduction

A major component of the field of sociolinguistics is the study of language variation and change.¹ All languages change, and all speakers of languages exhibit variation in their use of language (Meyerhoff 2019). Speakers make choices (conscious or unconscious) about the kind of sentence structure that will be used to express an idea, and not every word is pronounced the same way, even by the same speaker, every time they utter it. Synchronic variation in language is not only of descriptive interest to linguists but it is also of interest because of what synchronic variation tells us about diachronic variation and the trajectories of language change. Much as evolutionary biology posits speciation to arise from individual mutation, synchronic variation is a necessary precursor to, and hence the root of, diachronic (longitudinal) change. Variationist sociolinguists try to discover what factors might influence inter- and intra-speaker variation, and the field rests on a solid foundation of empirical evidence.

From this empirical foundation we know that synchronic variation is influenced by aspects of a language’s structure but also by social factors. Hence, in order to account for variation we need to attend to both linguistic and social factors. By teasing apart the relative importance of social and linguistic constraints on synchronic variation we can, by application of the uniformitarian principle (Labov 1994), extrapolate about what factors have influenced the diffusion of language change in the past. This extrapolation has been validated in several studies of the history of English (Nevalainen and Raumolin Brunberg 2017). We also gain a richer picture about how subtle aspects of linguistic structure are shaped by interlocutors’ perception of the variation around them. We learn about how competing forms might be reanalysed innovatively and become grammaticalised as new constructions.

Linguists working in the field of language documentation, too, are necessarily engaged with language variation, but from a somewhat different perspective. Instead of starting from a well-understood and well-documented language grammar, documentary linguists are using their skills to (first) extrapolate away from the inter-individual variation that characterises all speakers and every speech community in order to adequately describe the structure of a lesser-known or poorly-documented language.

We can characterise the difference in what variation means to the two groups like this: On the one hand, for sociolinguists, the analysis of variation allows the researcher to recuperate non-normative ways of speaking, linguistic forms that have been set aside as irrelevant because they are not categorically distributed, but rather, occur only more or less frequently in different contexts or with different speakers. These variants are perceived to be non-standard or to be as marginal as the social groups who typically use them. A variant generally becomes salient to the sociolinguist because they have extensive (near) native speaker knowledge of the variety at hand, and because that knowledge allows them to recognise the variant as somehow non-normative. The purpose of recuperating these variants is not only to validate individual speakers’ choices but also to build up a repository of case studies that enrich our understanding of the ways in which language change typically diffuses across individuals and over time.

On the other hand, documentary linguists are unlikely to have the detailed linguistic or ethnographic knowledge that is necessary for variationist sociolinguistic analysis (Sankoff 1988, Stanford 2008, Stanford and Pan 2013). They will inevitably encounter variation between and within individuals as they collect their data, but the primary purpose of engaging with this variation is to allow the documentary linguist to provide a clear account of the non-probabilistic loci of variation in the grammar of a language (e.g. which plural endings are used with nouns of different noun classes, how a word stem is pronounced depending on whether it is preceded by a consonant or a vowel, etc.). They may (and increasingly try to) document generalisations about whether certain forms are more or less likely to be used by, say, younger speakers rather than older ones (Evans 2003, 2008; Himmelmann 2006, Nagy 2009, Jauncey 2011, Childs et al. 2014, Von Prince 2015). However, it is still rare to see the wholescale incorporation of probabilistic generalisations into grammars.

Like sociolinguists’ analysis of synchronic variation, the documentation of variation has links to historical linguistics. But instead of being concerned with how variation illuminates the manner in which variants percolate through a community of speakers, it is concerned with how variation sheds light on how the language being described fits into larger families or clades, or how it informs our knowledge about linguistic typology. Haig et al. (2011) show how more precise typological generalisations about, for example, word order can be made if documentary linguists are able to draw on balanced corpora, and the primary linguistic database associated with Michaelis et al.’s (2013) survey of pidgin and creole languages attempted to provide relative frequencies for different lexical and grammatical variants.

Most sociolinguistic research on language variation has focused on speakers of well-described languages, that is, predominantly on people who are located in the urban centres of modern nation-states, and generally on speakers who are monolingual (Nagy and Meyerhoff 2008, Stanford and Preston 2009, Childs et al. 2014, Smakman and Heinrich 2015, 2018, Mansfield and Stanford 2017). This raises the question of what can be gained from deploying the methods and principles of variationist sociolinguistics in the documentation of lesser-known languages. Are variationist sociolinguistics and language documentation concerned with such fundamentally different kinds of variation that they will always be uncomfortable bedfellows, or can they lie harmoniously together?

In the next section, I present the case for the analysis of variation in the service of both local communities’ and linguists’ interests in situations of language documentation. I then introduce two case studies from my own documentation of Bislama, the creole spoken in Vanuatu, which focus primarily on linguistic interests. The first case study illustrates some of the points raised in the introduction, specifically, it exemplifies the way that a study of synchronic variation in the course of documenting the structure of a lesser-studied language can help elucidate what might have happened in the past. It also shows how an analysis of variation can address outstanding questions about the fundamental linguistic structure of an under-documented language and the ways that variation feeds into grammaticalisation.

The second case study delves deeper into the structure of the same language and presents an analysis of variation that problematises the alleged simplicity of creoles. I will suggest that if we adopt methods drawn from the the probabilistic traditions of variationist sociolinguistics, linguists may be able to propose innovative generalisations about the constraints on language change.

The paper concludes with some observations on the way documentation advances sociolinguistic goals, emphasising the mutual benefit of engagement between both research traditions.

Esoteric and exoteric reasons for analysing variation

Most linguists working on the documentation of a lesser-known language are confronted by community language ideologies that highlight certain aspects of variation and change. These may be as innocuous as descriptive observations that younger speakers, or speakers from a different village, use one variant more than the speaker him/herself does. But attention to variation may also be framed in ways that reflect powerful constraints on local social structure (Coto Solano 2017). If a speaker characterises children as ‘confusing’ or ‘mixing up’ the words for ‘snake’ and ‘eye’ in Tamambo (Malo Island, Vanuatu, where /m^wata/ ‘snake’ is often pronounced the same as /mata/ ‘eye’ among speakers born after 1990), this is not just a descriptive observation. It includes normative and evaluative information about the variation, too.

Like many people engaged in language documentation, I have heard younger speakers express a reluctance to use, or even shame about using, the local language because they are (rightly or wrongly) worried about their elders hearing their speech as full of ‘mistakes’ and being ‘bastardised’ or ‘imperfect’.² In other words, communities of speakers are often aware of language variation and see it as something that needs to be addressed. These are some of the esoteric reasons, arising from within the community itself, for incorporating the sociolinguistic analysis of variation into language documentation.

There are also exoteric reasons (from outside the community) for taking the time to engage with probabilistic variation in a language that is being documented. Grenoble (2010, p. 83) observes that at a practical level ‘Linguists can help educate speakers in dialect awareness, to understand that variation is the natural result of language change and is found in vital languages which are robustly spoken’. And the benefits are mutual: linguistic theory also profits since ‘more research is pressingly needed in the area of contact-induced change and language attrition’ (Grenoble 2010, p. 67). For instance, Dorian (2010) has suggested that certain patterns of variation may be diagnostic of obsolescence (rather than the ‘natural’ language change alluded to by Grenoble, above).

Filtering out the noise

If, as we have noted, variation is a ubiquitous aspect of everyday language, how might a documentary linguist decide which types of variation afford careful scrutiny, and which variation can be safely extrapolated away from and treated as noise? In better known languages, this filtering is done by reference to previous grammar descriptions (which establish what forms are default, unmarked or normal – the benchmarks against which variation is generally defined), by ethnographic observation of the comments and practices that speakers themselves engage in, and by the informed insights of (near) native speaker linguists.

All of these pose difficulties for the linguist undertaking language documentation. First, there is likely to be no existing grammatical desciption (this is generally what their documentation will generate), second, they may not be able to spend extensive periods in the field as a participant observer, and third, there may be a paucity of trained native speakers to assist with the task. Increasingly, there are productive partnership between linguists and native speakers of under-documented languages (Stanford and Pan 2013, Taho and Krifka 2013, Meakins and Algy 2016, Van den Bos et al. 2017) and funding agencies such as the Rausing Foundation’s Endangered Languages Documentation Project encourage and financially support extensive periods in the field as well as the linguistic training of local researchers. It is worth noting the non-independence of extensive fieldwork and the linguistic training of local research partners.

These difficulties are not insuperable, but they must be acknowledged as points of difference between the two fields of linguistics.

Two cases studies of variation in the service of documentation

In this section, I offer two case studies in which subtle and probabilistics patterns of sociolinguistic variation proved to be relevant to broader linguistic issues (descriptive and theoretical) because of gaps in the prior linguistic description of the language. The purpose of the case studies is to show that serious engagement with synchronic vairation can make a positive contribution to linguistic description and theory.

Bislama pronouns and subject verb agreement

I begin with a specific descriptive problem in the analysis of Bislama, the analysis of which sheds crucial light on the nature of subject-verb agreement in the language. Bislama is the English-lexified creole spoken widely in Vanuatu, a country with around 110 distinct indigenous vernacular languages and two official languages of education (English and French).³ Within this magnificent linguistic playground, Bislama is recognised as the national language of Vanuatu, and the Constitution gives it primus inter pares status in relation to the other vernacular languages spoken in the country. That is, the Constitution says that it can be replaced at any time by one of the other indigenous languages of Vanuatu but the likelihood of this actually happening seems rather remote, given that Bislama, by virtue of not ‘belonging’ to any particular group (as vernacular languages do), has become a very successful index of belonging for all Ni-Vanuatu.

Since the modern form of Bislama stabilised in the second half of the 20thC, there has been debate among linguists about what the most appropriate analysis is for the particles that occur immediately to the left of the finite verb (i and oli in Table 1). The analysis of these particles in turn impacts on the analysis of Bislama pronouns.

Table 1. Bislama verb paradigm with the verb laf ‘laugh’.

	Singular	Dual ‘two’	Trial ‘three’	Plural ‘three or more’
1 (incl.) ‘me plus you’	–	yumitu laf	yumitri laf	yumi laf
1 (excl.) ‘not you’	mi laf	mitufala i laf ‘me + s.o. else’	mitrifala i laf ‘me + 2 others’	mifala i laf ‘me + others’
2	yu laf	yutufala i laf ‘you + s.o else’	yutrifala i laf ‘you + 2 others’	yufala i laf ‘you + others’
3	hem i laf	tufala i laf	trifala i laf	olgeta oli laf

The crucial forms are actually the ones that the literature (Crowley 1990, 2000, Camden 1993) spends the least time talking about, that is, the forms where the subject is 1st person singular or 2nd person singular. These can take the shape of mi mi and yu yu or simply of mi and yu. The disagreement in the literature was whether the basic form is the doubled string mi mi/yu yu, in which the first mi/yu is a pronoun and the second one is a homophonous verb prefix, or whether the basic form is the singleton mi/yu and there is a zero or null verb prefix. Under the latter analysis, doubled sequences like mi mi and yu yu are focused constructions, analogous to 3rd person singular constructions where the subject has been topicalised on the left. Something similar occurs in English, a 1st person pronoun can be moved to the left just as a proper noun can be. Compare: Gill, she sent the papers out and Me, I had nothing to do with it. The main purpose of Meyerhoff (2000) was to provide an analysis of variation in the omission of subject pronominals in Bislama, so it was necessary first to determine the best analysis of these particles, before an analysis of the linguistic and social constraints on pronoun deletion could be undertaken.

At the time I started my work on Bislama there were two clear camps, the first, which held that mi mi and yu yu sequences are a pronoun followed by a verb prefix, was favoured by (among others) Camden (1977). The second, which held that there was a single pronoun and that doubled forms were actually focused nouns (as above) was favoured by (among others) Tryon (1987).

The analysis for 3rd person singular and plural is, by comparison, much less controversial. Historically, it is clear that the i particle that occurs before the verb in 3rd singular (and by extension the oli form that occurs in 3rd plural) did emerge from a strategy of subject doubling (Crowley 1990 and cf. G. Sankoff 1984 for Tok Pisin). However, despite originating as pronouns, it is apparent that since the mid to late 20thC, i and oli have been affixes on the verb.

This creates an eminently testable situation. Meyerhoff (2000) examined the cases in 3rd person where it is clear that we have a fronted NP, a pronoun and the verb phrase and compared their distribution in a corpus of everyday speech with the distribution of mi mi and yu yu in the corpus. This allows us to see if the doubled construction behaves more like clauses with a focussed third person singular subject (e.g. Vosale, hem i kam ‘It was Vosale who came’) or like a simple pronoun plus i (e.g. hem i kam ‘she came’). If they are distributed in discourse like the focussed 3rd person singular forms, then we could conclude that the first mi or yu is a focused constituent and not the pronoun subject. On the other hand, if they are distributed in discourse like the simple 3rd person singular clauses (hem i V), then we would conclude that they are more like pronouns followed by homophonous preverbal markers. A careful consideration of the distribution of forms across different contexts led to the conclusion that Tryon’s analysis was linguistically more accurate, that is, that mi mi and yu yu subjects were in fact focus constructions (Meyerhoff 2000).

This finding was not the end of the significance of this analysis of variation. It raised further questions for linguistic theory, as it became clear, therefore, that 3rd person subjects and 1st and 2nd person subjects behave rather differently from one another.

Clearly, the historical origins of i as a 3rd singular pronoun continue to influence its synchronic distribution. The pronoun ‘he’ may have grammaticalised into a 3rd person singular agreement form i, but if we consider which pronouns i occurs with, it is clear that it still carries a persistent sense of ‘3rd person’. That is, it only occurs where the subject necessarily includes someone other than the speaker and the hearer.

In this case study, the close analysis of variation shed new light on a long-standing linguistic debate about the analysis of Bislama. Here, the analysis of variation in a minority language was able to resolve a descriptive problem that had not been settled in earlier attempts to document aspects of the syntax of Bislama. In doing so, it was also able to enrich linguistic theories of grammaticalisation and to explain how earlier variability had given rise to the pronominal system and the structure of the verb phrase in current day Bislama.

Bislama and Tamambo argument deletion: the effects of language contact

The second case study speaks to more general questions about how processes of language contact have shaped the synchronic structure of Bislama; it is, therefore, both descriptive and theoretical.

In this work I have used the exploration of variation as a means of evaluating possible sources of contact-induced change. I argue that transferring variation is an exceedingly complex cognitive task, because it requires analysis of not only the linguistic system of both the donor and the recipient language, but also may require analysis of a wide range of social and linguistic constraints. Speakers have to co-ordinate what they know about the social and linguistic systems in both languages, all the time keeping track of the constraints on variation. Given this burden, it is entirely to be expected that we may sometimes see evidence that speakers have reallocated, or even reversed, constraints on the variation that is being transferred from one language to another (Meyerhoff 2009).

In a striking pattern (that has now been replicated in several other quite different contexts of language contact), I found that on Malo island in Vanuatu, the community languages Bislama and Tamambo both allow speakers to optionally omit the object in a transitive clause. In Tamambo, this omission is more likely when it refers to an object that is an inalienable possession. In the local variety of Bislama, there is also a significant effect for the alienable/inalienable status of the object, but in Bislama alienable possessions are more likely to be omitted.

I suggested that where we see this kind of partial transfer of variation from one language to another, and there is evidence suggesting that the ranking of some of the detailed constraints have been altered or even reversed that this reflects a very general principle of ‘transformation under transfer’. We find examples like the Tamambo/Bislama data for (in)alienability in a growing number of case studies of variationist studies of languages that are in contact with each other. For example, in joint work with Erik Schleef (Schleef et al. 2011) we have found it in the way teenage Polish migrants seem to be gradually picking up local norms of variation in English. Blondeau and Nagy’s (2008) work on French-English contact in Montreal finds similar patterns, and Buchstaller and D’Arcy (2009) observed the same kind of phenomenon in their cross-dialect study the new quotatives in English. It would appear that transformation under transfer is a very basic principle in situations of language contact.

Why this matters for the study of creole languages

By taking the systematics of variation in these two case studies seriously, we can also engage with theoretical debates over the inherent nature of pidgin and creole languages. If we look beyond the categorical, and examine stochastic components of the grammar, we can rebut a claim often promulgated in the literature on pidgins and creoles, namely that creoles like Bislama do not mark subject-verb agreement or relations like alienable/inalienable possession (and hence, are somehow simpler than the languages that fed into them). A variationist perspective shows that they may encode these features probabilistically, rather than as categorical rules, but it would be wrong to say these features are absent. Hence, if we ignore probabilistic variation (as the majority of accounts of pidgins, creoles and contact languages do), linguists risk short-changing the relative complexity of human language in speech communities with histories of profound language contact.

If variation is a necessary prerequisite for change, then it’s obvious that looking at variation in contact languages has the potential to be enormously rewarding. The variation may be able to show us how complexity develops out of what (in the spirit of Dahl 2004) we might call ‘immature’ language phenomena.

It also means we can provide a fresh perspective on the question of whether pidgins and creoles are necessarily ‘simpler’ than non-pidgin and creole languages. Some recent discussions about linguistic complexity and simplicity have suggested that they are forwarding a daring, revisionist view by claiming that some languages are simpler than others. But these discussions miss an important point. Claims of equicomplexity across languages originated in a claim that all humans (ergo all human languages) can express equally complex semantic concepts. However, since we can’t directly observe semantics, we have to choose some surface phenomenon to look at, and whether semantics is viewed through the lens of syntax, morphology or lexis is a theory-internal decision. Comparisons of different measures of linguistic simplicity or complexity are therefore comparisons of different theories of language, not comparisons of linguistic simplicity or complexity itself, and the mathematics for comparing the explanatory power of different theories is very different from the mathematics we usually use in language modelling and quantitative studies of the distribution of specific forms in corpora. Moreover, given that many linguistic theories specifically exclude from their investigation observations of how language is used in everyday interaction, we cannot derive measures of complexity from pragmatics or probabilistic patterns of variation; they have been excluded a priori from our calculations.

This is not just an empirical point of principle. It might well be central to deriving a more informed notion of complexity because probabilistic patterns might well be considered more complex than deterministic patterns. Take for instance, the kinds of case marking relations documented by Meakins (2011) for Gurindji Kriol. In Gurindji Kriol, ergative marking is optional (not obligatory, as it is in Gurindji) and when it is used, it does not just mark thematic relations, it marks a suite of pragmatic and interactional factors. By most standards, this would appear to be a lot more complex than a system where ergative marking invariably and predictably marks the subject of intransitive and the object of transitive clauses. If we were to ignore variable processes in the grammar, we might conclude that Gurindji Kriol does not have the same richness of case marking that Gurindji does, in the same way that we might have concluded that Bislama does not encode a distinction between alienable and inalienable possessions. And this would be wrong.

Further opportunities

The way the case studies were presented in the previous section naturally reflects my personal training and disposition as a variationist, and they may give the (unintended) impression that the interplay between documentation and sociolinguistics involves the sociolinguistic study of language variation elaborating other branches of linguistics. To the extent that this is true, it is largely an artefact of my own discovery process, and it records the means by which I have used the skill set of sociolinguistics to contribute to linguistics more broadly. But documentary linguistics is equally able to extend sociolinguistics by problematising central assumptions or principles of sociolinguistics.

A key one is the centrality of the notion of the speech community in sociolinguistics. Traditional sociolinguistics identifies a speech community on the basis of shared patterns of variation – that is, co-membership in a speech community is defined not only by speakers alternating between the same forms, but by evidence that speakers share similar evaluations of when it is or is not appropriate to use those forms (Labov 1972). Generally, shared evaluations have been inferred from uniform patterns of style shifting. That is, even if one speaker is a heavy user of a non-standard or innovative variant and another speaker uses it very infrequently, they can still be identified as members of the same speech community if they both shift away from their baseline frequency of the innovative form in formal or careful contexts. In other words, a definition of the speech community should include objective measures of production and subjective measures of perception in some way or another.

The study of an undocumented language presents several challenges to this principle. While the notion of a ‘standard’ or ‘norm’ is central to methods in sociolinguistics, an undocumented language is most unlikely to have experienced formal standardisation, e.g. for use in education (cf. Vandenbussche 2007). Consequently, there may be a multiplicity of norms and standards within the community (Nagy 2009, Sallabank 2012, Coto Solano 2017). If we define the speech community as speakers with shared norms and shared evaluations of those norms, then it may be difficult or impossible to determine whether speakers of an undocumented language are indeed orienting to shared evaluations of norms. Where there are multiple norms, speakers’ patterns of style-shifting may not neatly line up with the patterns typically observed in previous literature that has concentrated on urban, and largely monolingual, speakers.

Even if speakers do seem to produce interpretable patterns of style shifting, it’s not clear that we should expect shared evaluations of norms to produce the same secondary patterns of variation we are familiar with from urban social dialectology. Considerable attention is currently being paid to this matter. We may not find that the same form consistently marks style, gender and socio-economic measures in the same way that variables do in urban social dialect surveys. Women may not, in all communities, lead changes that are progressing below the level of conscious awareness (Labov 2015, and papers in Kasstan and Horesh 2018). While we have some reports of women leading change from below in minority speech communities (e.g. Ravindranath 2009), we have other studies dating back to the 1980s and 1990s that indicate that once we look beyond the WEIRD (Henrich et al. 2010) speech communities, there is a complex interaction between gender and social network types and language use, especially where there is extensive contact with a dominant language (Dubois and Horvath 2000).

Recent advances in sociolinguistics have emphasised the emergence of new ways of thinking about what language variation means (Eckert 2018). Eckert’s research has demonstrated that variationists can engage happily with intellectual problems (such as what ‘meaning’ is) by looking closely at how and when individuals use variation in everyday interaction. This so-called ‘third wave’ approach to sociolinguistics concentrates on a much more ‘local’ perspective on variation, and it enables sociolinguists to learn more about how variation acquires different associations or meanings and how these are negotiated and shift over time and space.

A study of language variation that emerges from the process of language documentation is possibly the most ‘local’ analysis one can imagine. In this sense, the documentation enterprise takes sociolinguistics beyond even the third wave, to a position where the researcher’s questions may be guided by the community (cf. Cameron et al. 1992 on the need for sociolinguistic research with speakers, as opposed to research on and research for).

In recent years, working with the community of Hog Harbour/Vüthiev in East Santo, Vanuatu, I have found that when language documentation is the primary goal of my fieldwork, this frees me up (as a sociolinguist) to be more creative in what kinds of data I collect and how data collection and data analysis can be structured so as to strengthen not only the community but my analyses of variation. To give but two examples: one of our first projects was to create a children’s picture book that documented the official opening of a tarseal road along the eastern coast of Santo. The process of adding captions to all the photos in the picture book highlighted some interesting (for me) differences in lexical choice and agreement marking on verbs used by older and younger, fluent speakers of the local language, Nkep.

A second project involved working with the entire community to film re-enactments and oral history interviews of a machine gun attack on the village that took place during the 1980 Santo Rebellion. The families of Hog Harbour/Vüthiev saw this as a crucial step in the documentation of the local language and culture, and by allowing them to direct the documentation process, we produced a DVD that involved the whole community, is owned by the whole community and that asserts the language as an identifier of community strength, vitality and unity (in a time of considerable social change). Sociolinguists often talk about the significance of language variation and language use as an act of identity (Le Page and Tabouret-Keller 1985). Marrying documentation with the study of language variation in this case resulted in what I would call acts of legitimation – activities and events that do not merely involve speakers identifying with their language and place, but also involve them asserting temporal and moral claims to those identifications. In short, the unification of the different traditions generated a more agentive and creative space for the speakers of Nkep than the traditional role of speakers as sociolinguistic subjects offers.

The purpose of this paper has been to explore some of the practical and theoretical implications of an approach to linguistic research that draws on both the traditions and methods of variationist sociolinguistics and the traditions and methods of language documentation. While both branches of linguistics cannot help but deal with variation, the kind of variation they focus on differs substantially. The detailed analysis of specific variables offers important insights into the nature of language variation and change. Incorporating the details of language variation into the description of a lesser-known language can resolve debates over what the underlying structure of that language is and it can make a substantial contribution to debates about the typology of contact languages.

In other words, bringing together the methods and principles of variationist sociolinguistics and the methods and goals of language documentation may create unusual bedfellows, but the accommodation may be far from an unpleasant or unnatural experience.

Disclosure statement

No potential conflict of interest was reported by the author.

ORCID

Miriam Meyerhoff http://orcid.org/0000-0002-2599-1870

Additional information

Funding

This work was supported by the Endangered Languages Documentation Programme (SOAS).

Notes

1 This paper reflects thinking that developed over a number of years and has benefitted from many people’s feedback and encouragement. I’d like to thank two anonymous Ngā Kete reviewers, the organisers and audience at the Australian Linguistic Society (2012), New Zealand Language & Society (2014) conferences, colleagues in Nick Evans’ Wellsprings of Linguistic Diversity Project at the Australian National University, Kate Burridge, Naomi Nagy, Jim Stanford, Gillian Sankoff and the increasing number of colleagues who are strengthening the connections between variationist sociolinguistics and language documentation. I am responsible for the paper’s remaining shortcomings.

2 My thanks to Naomi Nagy for sharing her fieldwork observations with Faetar in southern Italy. The terms ‘bastardised’ and ‘imperfect’ come from her recordings. I am grateful to Nagy for conversations that inform much of this section.

3 The recent policy change allowing the Year 1–3 curriculum to be taught in a vernacular language did not change the overall French/English focus of the education system.