Language ecology, language endangerment, and relict languages: Case studies from Adamawa (Cameroon-Nigeria)


 As a contribution to the more general discussion on causes of language endangerment and death, we describe the language ecologies of four related languages (Bà Mambila [mzk]/[mcu], Sombә (Somyev or Kila) [kgt], Oumyari Wawa [www], Njanga (Kwanja) [knp]) of the Cameroon-Nigeria borderland to reach an understanding of the factors and circumstances that have brought two of these languages, Sombә and Njanga, to the brink of extinction; a third, Oumyari, is unstable/eroded, while Bà Mambila is stable. Other related languages of the area, also endangered and in one case extinct, fit into our discussion, though with less focus. We argue that an understanding of the language ecology of a region (or of a given language) leads to an understanding of the vitality of a language. Language ecology seen as a multilayered phenomenon can help explain why the four languages of our case studies have different degrees of vitality. This has implications for how language change is conceptualised: we see multilingualism and change (sometimes including extinction) as normative.


Introduction
A commonly cited cause of language endangerment across the globe is the dominance of a colonial language. The situation in Africa is often claimed to be different, with the threat being more from national or regional languages that are themselves African languages, rather than from colonial languages (Batibo 2001(Batibo : 311-2, 2005Brenzinger et al. 1991;Connell 1998Connell , 2015Mufwene and Vigouroux 2008;Nettle 1996, discussed below, offer opposing views). In Africa, the replacive lect¹ is more likely to be an African language. As Maarten Mous puts it: Another reason why language endangerment is less dramatic in Africa is because, contrary to the situation in Latin America, in Africa former colonial languages are not a major factor in language loss. Colonisation has not led to the marginalization of the original peoples as it has in the Americas, and peoples speaking endangered languages in Africa are  often not economically worse off than their neighbours speaking healthier languages. In the majority of African countries the colonial language is the official language but this has little impact on the every-day communicative situations compared to, say, Spanish in Latin America. (Mous 2003a: 158). This situation is particularly acute in multi-ethnic towns and cities. Among rural communities, different processes are underway. Individual languages are becoming extinct and are being replaced by other local languages. For example, both Sombә and Njanga, two of the languages we examine here, like a number of other languages in the region, have given way not to a colonial language, nor have their speakers adopted a regional lingua franca as their primary language, though one is available. Rather, other local languages have taken on the functions they formerly filled. The study of the dynamics of language shift such as has occurred in the cases of Sombә and Njanga adds a layer of complexity to our knowledge of the factors involved in language endangerment and loss.
In this paper, we use language ecology as a framework to analyse and discuss the specifics of the somewhat different ecologies of three endangered languages: Njanga, Sombə, and Oumyari Wawa, in contrast to that of Bà Mambila, which is not under imminent threat. The endangered languages face differing degrees of threat (two are on the brink of extinction) due to a combination of several factors, without a single factor being predominant. And while their situations share similarities, each is unique. Their endangerment can best be summarised as resulting from changes in the ecologies of these three languages. In Section 2, the importance of studying the language ecology of a region to understand linguistic evolution and the fortunes of individual languages is discussed and the details of the approach we adopt are presented, including discussion of the model introduced in the recent paper by Ludwig et al. (2019) (henceforth LMP). In Section 3, we present general characteristics of the area in the Cameroon-Nigeria borderland where we carry out our fieldwork, the Mambiloid region, which roughly corresponds to the outer ring of the LMP model, which represents the language macro-ecology of the region (see Section 2.4). Further details of this are found in Section 4. In Sections 4-7, a detailed description of the settings of each of the four case studies is presented. The Mambiloid language group is introduced in Section 3.2 below. Our examination of the four languages (Bà Mambila, Njanga, Oumyari Wawa, and Sombə) follows the template given in Section 2.5 to determine salient features of their specific ecologies; i.e. characteristics situated on the middle ring of the LMP model and to facilitate comparison among them. Following this, Section 8 draws comparisons among the four case studies, showing how their micro-ecologies differ, and substantiating the view that detailed consideration of the language ecology of a (sub-)region, or of a specific language, can give insight to the differing fortunes of the languages of the region. The concluding section, Section 9, addresses the relations between language vitality and language ecology from a more general standpoint and offers a more general perspective on understanding the causes of language endangerment. We argue that this approach has implications for how language change is conceptualised: we see multilingualism and change (sometimes including extinction) as normative. Phenomena such as monolingualism and long-term persistence are unusual in the wider scheme of things and need explanation rather than being taken for granted as the default background, though this is not the place to explore these issues.

Language change, ecology, and linguistic evolution
Linguists and anthropologists, among others, have come to recognise the importance of linguistic diversity and the threat posed by the current accelerated rate of language loss. Work on language endangerment has tended to be directed primarily at the documentation of endangered languages, as a means of preserving their uniqueness "beyond the grave." A second, though not totally separate, strand of research is focused on issues involved in language maintenance and development and, with respect to language documentation, its importance in providing the materials necessary for language revitalisation programs. Further strands of research are concerned with the processes involved in language attrition and death, principally from a structuralist standpoint, and with identifying (and remediating) the causes of language endangerment. Although much has been pronounced on the issue, still comparatively little research has been African, rather than the languages of former colonial powers. Connell (2015) offers a more nuanced view, providing evidence that the influence of colonial languages is growing, but that this is among the more general consequences of colonialism. Both colonialism and regional conquest² imposed new political environments on local social structures, which, among other things, typically included the favouring of one "unifying" language as essential to nation building, thereby eroding traditional reliance on multilingualism. It led not only to greater prestige for colonial languages, but also to the rise of national languages. So, for example, Swahili's place in Tanzania is a direct attempt to replicate colonial structures (ironically, while trying to eliminate colonialism). Moreover, given the rapid rate of urbanisation throughout much of Africa, this situation is particularly acute in multi-ethnic towns and cities. Among rural communities, however, languages are also under threat, though different processes are underway, leading to a situation in which individual languages are becoming extinct and are being replaced by other local languages. Our case studies illustrate that there are many different ways in which replacement may happen. The characterisation that one language has replaced another does not imply that the replacing language "drove out" the previous one, but only that one is used in situations and domains where another was formerly used.³ If the social contexts in which a language is used are reduced then speakers make choices from their linguistic repertoires which end up with their repertoires being reduced, since there is never occasion to use a particular language.

Background orientation
Following Silverstein (1998: 407) and Grinevald and Bert (2011), we distinguish between language community and speech communitythe latter being a group who speak to one another using whatever "linguistic resources" (languages) are locally available, and the members of which are intimately familiar with the "rules" for using language appropriate to a given situation; the former is a group defined by the use of a particular language (the members of one language community may belong to many speech communities). Our starting points are particular villages in the Nigeria-Cameroon borderland, which comprise distinct speech communities and have distinct language ecologies.⁴ Many of the languages within these communities are endangered to varying degrees, and in some cases, are near extinction. We recognise, along with Grinevald and Bert (2011), that the notions of language community and speech community are problematic in situations of endangerment. The range of competence/fluency that exists among speakers of an endangered language makes problematic the identification of someone as a speaker (of language X), and hence who is member of a language community; this makes defining the language community of an endangered language difficult, even if the boundary issues in establishing the limits of a speech community are more easily resolvable. We see these as legitimate concerns but ones that are resolvable in practice in the field and when dealing with research data.
In terms of linguistic varieties, a kind of conceptual continuum can be recognised running from Idiolects -Registers -Sociolects -Dialects -Languages. Zeitlyn and Connell (2003) argue that there is a similar continuum of scales or levels for the basis of ethnic identity, each level related to what they describe as a fractal history (see also Gausset 2010).⁵ There is a range of different identities available to language users, and different ones of these become active in different contexts. This is intimately and complexly connected to language since, all the qualifications notwithstanding, it remains the case that language is routinely taken as indicative of identity. When strangers meet at markets, for example, it is a reasonable inference that someone who speaks language X with apparent native-like fluency is of ethnicity X.
There is a parallel set of problems for both identity and languageboth are analytic categories, which apply to social groups but not directly, or straightforwardly, to the individuals who make up those groups. Any label, whether for a language or an ethnic group, is a roughshod generalisation. An individual speaks differently with different interlocutors, and at times might accept one social label, at others another (see Gausset 2010). Silverstein (1998) considers some of these issues⁶ concluding that, "the language community, and hence its language, can be seen as a precipitate of sociocultural process" (1998: 402 our emphasis). He is helpfully clear about the distinction between analytic categories (he talks of "synchronic abstractions" p. 415) and the events, with their actors, from which our generalisations are abstracted.
So to be clear, we are not essentialist about either languages or ethnic identities. There are local stereotypes that bundle together languages and ethnicity (often in francophone Cameroon using talk of tribu), but these emic categories, while sometimes mapping easily onto groupings on the ground, are as often statements about ideology and aspiration rather than accurate summaries of how people group politically; of their shared and disparate histories, let alone the multiple languages that people speak (see Di Carlo and Good eds, 2020 for case studies of rural multilingualism, mainly from a little to the south of where our research was undertaken). There is considerable variation in the language ideologies in the different case studies we present: use of Njanga has been shaped by local political considerations; of Wawa by religion and Fulɓeisation; and for the Sombə language is connected to occupation. We think that the comparative framework provided by Language Ecology can help unpack some of this variation.
We recognise that analytic terms are approximations and that linguists too often talk of entities in ways that suggest they have sharp or clearly defined edges when in fact there are no such entities and therefore no edges at all.⁷ But these approximations do have purchase on events in the real world that would be happening and will continue to happen, whether or not our research took place. So we continue, but cautiously, with these issues in mind. Our definitions are ostensive not essential. We can identify some clear (albeit rare) examples and use these as exemplars to help us make sense of the more commonly occurring unclear cases. We are concerned with language in the world not in the mind. This leads us to argue for the use of a language ecological model and consider it natural for a speech community to pick and choose among different resources in the linguistic repertoires available to them (Busch 2012(Busch , 2017).

Language endangerment and language ecology
This is not the place for a survey of the history of language (or linguistic) ecology theory.⁸ However, a few words are in order to help our readers appreciate the background. An ecology can be defined as a network of interactive relationships among organisms and their environment (including complex and competing feedback loops). In this sense, the term is widely used in biology and is well-known among the general public. That languages form an ecology has been much less considered, even among linguists, than their biological counterparts, and so a brief explanation as to its meaning within linguistics may be useful. The notion that languages, like biological systems, exist in an ecology is credited to Haugen (1972) and has been explored  6 More recently their application to language endangerment has been discussed in the journal Language Sciences (vols 40 and 41, 2013; e.g. Orman and other contributors). 7 See Hudson (1996) for interesting discussion of this problem with respect to the notions of "language" and "speech community." and developed by linguists such as Mackey (1980), Mühlhäusler (1992Mühlhäusler ( , 1996, and Mufwene (2004), to name but a few.⁹ In summary, the ecology of a language is the network of relationships that a language (i.e. its speakers) has with the other languages it comes in contact with, together with other aspects of its environment that might exert an influence on it. A succinct definition of language ecology is provided by Raith (1984: 6): "the totality of relations that obtain between language and environment, i.e. between the factors and conditions which make language possible or which affect or influence language." In examining the ecology of a language, then, one takes into account such factors as the following (cf Ludwig et al. 2019;Mühlhäusler 1996;following Haugen 1972): the classification of the language (i.e. its relatedness to other languages); the demography of its speakers; its internal variation (regional and social); its domains of use; other languages used by its speakers; their degree of bi-or multilingualism; its degree of development; whether it has institutional support; the attitudes of its users; as well as the many nonlinguistic aspects of its social environment and natural environment. This list is not meant to be exhaustive; cf Edwards (1992) who proposes a similar, much expanded, framework. One might also consider the language ecology of a region, and the set of languages spoken in a region, rather than that of a specific language (cf Voegelin et al. 1967). In Silverstein's terms (above), this is to think about the language ecologies of speech communities. Ludwig et al. (2019), in the introduction to a collection entitled Linguistic Ecology and Language Contact, summarises the literature and advances the notion of language ecology in a way that accommodates the range of complexity found in the cases we are considering. They argue for an analytic framework which does not make a priori assumption about which factors are salient or indeed present in all cases. The point about an ecological network of interconnected and reciprocally influencing factors is that one (or several) may be dominant in one situation but others in another situation (even nearby ones, as we shall see).
Another way of characterising their approach might be to say they describe a "meta-model" of language ecology rather than a model. By definition, a model can never cover all parameters at work in the processes or entities it is supposed to illustrate. We assembled in our model what we consider to be significant parameters at each level for each dimension. In accordance with our statements above (especially statement 2 [language-systems are dynamic]) we cannot offer a fixed set of parameters and no universal method for an ecological investigation of language and language contact. We assume that from a theoretically indefinitely large number of parameters only some become active, and of these only some become dominant in a specific contact situation at a given time (2019: 17). This helps us consider the social contexts within which language use occurs. Changes in the cultural or social setting of a languageor other aspects of a language's ecologyparallel the effect of environmental changes on biological species. The eventual effects of such changes cannot be predicted at their outset. The possibility of drastic consequences on a language is summed up by Mackey (1980: 36-37): [T]he interrelated sequences of causes and effects producing changes in the traditional language behaviour of one group under the influence of another, results in a switch in the language of one of the groups.
[…] The causes and the process of such language shifts are worthy of study […] because they are intimately related to the life and death of languages.
In short, the fate of a language can be influenced by a multitude of factors, and it is misleading to assume it can be attributed simply to any one of these.¹⁰ In the next sections of the paper, we take up Mackey's  9 For more recent work see e.g. Hornberger (2002), chapters collected in Creese et al. (2008) and in Ludwig et al. (2019), especially the introduction. For discussion specifically addressing situations of language endangerment, see e.g. Connell 2010a, b, c, Nettle andRomaine (2000, chap 4), Grenoble (2011), Romaine (2013, and Ludwig et al. (2019). 10 Attempts to understand why a language comes under threat, or what needs to be done to revitalize a threatened language implicitly or explicitly assume just one or two factors need be addressed. Austin and Sallabank (2011), following Wurm (1991), identify four types of cause: natural disaster, war, political repression and cultural hegemony. Our work can be seen to some extent as fleshing out the dynamics of, especially their fourth type, cultural hegemony, though this is not to suggest that factors of cultural hegemony are exclusively important in our case studies. challenge that these are worthy of study and examine the situation of our examples, in an effort to shed some light on the causes of their failing fortunes. The approach of language ecology generally provides both a way of characterising a speech community at a point in time and how it is changing. The Ludwig, Mühlhäusler, and Pagel (LMP) model allows us to accommodate the range of factors that have determined the destiny of the languages we consider here. LMP summarise their model graphically as a set of concentric circles existing in multidimensional space, reproduced above in Figure 1; as can be seen, each layer represents a broader or more general set of possible influences. Arrows away from and toward the centre reflect the existence of factors existing primarily on one level influencing those on other levels.

An ecological matrix
Based on our collective field research and following the considerations of language ecology as presented, for example, in Haugen (1972), elaborated in Edwards (1992) and endorsed in, e.g. Mühlhäusler (1992) and LMP (2019), we have developed the following list of parameters which, when applied to the situations of the four different languages in our study, form an "ecological matrix" that characterises each speech community in question. The list was developed based on our own detailed familiarity with the region, particularly the village of Somié, having worked there collectively for several decades. The set of factors included here is not necessarily the only possible set, nor is it expected that all are necessarily relevant or of equal importance in each case. This may be true especially of the functional domains listed at (7) which were selected to include salient aspects of daily life and to allow for comparability across the four case studies. Thus, the matrix allows for dynamism by considering change within each domain separately and then through interconnections between them. Besides Haugen (1972) and Edwards (1992), inspiration also comes from the matrix model for codeswitching as elaborated by Carol Myers-Scotton (1993), although it differs in important ways from her conception. One important difference is that we make no assumption of priority that there is what she calls a matrix or main language (Myers-Scotton 2011). As she does say, it is a question of figure and ground, and our approach is consistent with hers in this respect. It would be possible to undertake a separate Myers-Scottonian codeswitching analysis of each functional domain; the differences in pattern between domains would be revealing of the dynamics of the overall system (Table 1).¹¹ At the top of the diagram in Figure 1 is the heading "Macrocultural competences/attitudes." LMP do not discuss this in any detail. Granted our limited, focused comparative frame, it is worth discussing what might be called the underlying cultural foundations, identifying the cultural base of the Mambiloid region. This is to ask whether linguistic classifications have identifiable cultural correlates, which is both a large question and potentially a hugely contentious one.¹² Having acknowledged the dangers, and keeping these very clearly in mind, it does seem possible to identify a set of "macrocultural competences and attitudes." The most general (and most prosaic) macrocultural competences concern the economic bases of the people we are discussing: they undertake subsistence agriculture, broadly along the escarpment of the Adamawa Plateau. In historical terms, they were sedentary, and primarily sorghum cultivators. Such basic material cultural commonalities link all speakers of Mambiloid languages and many others in the region. What seems to be more particular to the Mambiloid group are some equations in the traditional religion. Such is the paucity of comparative knowledge about the traditional religion of the various Mambiloid languages that we cannot be confident as to the extent of this, but we have some evidence. The key connection is between masquerades and oath taking. This seems to hold for at least Mambila, Kwanja, and another neighbouring group to the south of the Mambiloid area, e.g. the Yamba who speak a Bantoid Grassfields language. So the cultural equation holds for neighbouring groups who do not speak Mambiloid languages. Table 1: Template for an ecological description, comprising a list of functional domains and the languages used in each of them along with other relevant factors 1. General description of physical geography and setting 2. Socio-economic summary 3. Historical summary 4. Linguistic overview 5. Speech communities 6. Language-identity relationships 7. Functional domains: home; in the village; market; education; religion (formal/informal); outside the village; town; formal; situations; informal/casual situations; intimate situations; entertainment/media 8. Internal variation, regional 9. Internal variation, social 10. Status/solidarity 11. Attitudes of users 12. Written tradition, standardisation, institutional support 13. Association between language and economic/political advancement It may form a cultural base, but we need to be very clear that the distribution of languages and cultural features may overlap but do not form a tight mapping. There is no possibility of inferring directly from one to the other or vice versa.
It seems that the terms for "masquerade" and "oath" are cognate across much of the group; examples include Bà Mambila (suàgà for both), Súndànɨ́Kwanja sɔp (masquerade), and səə́r (oath). In neighbouring Yamba, we find sɔ; nwe/ŋweʔ (masquerade) and sɔta/sɔtap (oath). We lack data for "oath" for some of the languages under study, but it should be noted that in the northern part of the Mambiloid region related practices have been abandoned (see our case study of Oumyari Section 6) following the arrival and adoption of Islam over the course of the latter half of the nineteenth century up until the middle of the twentieth.

A note on methods
A wide range of methods have been used by the authors in the course of ongoing research in the area. Zeitlyn has been undertaking participant observation and ethnographic research in Somié since 1985 which includes the analysis of naturally occurring conversation. He has been collaborating with Connell since 1994, and together and separately, they have undertaken language use surveys, questionnaire, and observational-based research on linguistic issues on Mambila lects in Cameroon (mainly, but not only, in Somié village) and also in Nigeria. They have worked on documentation of extinct or near extinct lects such as Njerep (Connell and Zeitlyn 2000) Kasabe and Camba (Connell 2002a, b)¹³, both varieties of Mambila, as well as Somyev (Sombə; Connell 2010b). Historical material comes from collaboration with local researchers such as the late Mial Nicodeme , oral historical work with elders in many Mambila villages, and archival research in the national archives of both Cameroon and Nigeria, as well as in various archives in Europe and the USA (particularly concerning the early missionaries to Mambila, Paul Gebauer, and Gilbert Schneider). Funding from the AHRC in the UK enabled the other authors to undertake doctoral research in Njanga and Oumyari, working as sociolinguists principally concerned with endangered language documentation.
Most local discussion of language matters occurs indirectly through talk about (the loss of) "tradition" and in some cases, ethnic identities. Two clear examples of this can be cited; the first, (with video recording) occurred a little before 8 am on 14 January 2001. Degah François, the then Chief of Somié, addressed the village population as part of a dāmə̀blessing undertaken in the run up to the ŋgwùn ritual. He pointed out that ritual is a core of identity and that these blessings are done so that good things come and bad things pass. As he said, "Without doing such rites we are not Fulɓe, not Mambila or not Europeans" and then turned to do the dāmə̀. It is important to note that he was an observant Muslim. His rhetoric links ethnic identity to the performance of "tradition" which has clear implications for language. The second, documented in Connell and Zeitlyn (2000, Appendix 2), came while working with Bondjie, at the time one of the six remaining semi-speakers of Njerep. Bondjie had been asked to translate a song she had sung in her language, but rather than a literal translation gave a reflection on the history of the Njerep; her account of a second song, again rather than a literal translation, was a lament for the loss of Njerep identity as they had become absorbed by the more numerous Mvop.
Since our research involvement in the different case studies that follow has varied dramatically (from over 30 years to approximately 1 year), our participation in local discussions about language change and language loss has also varied. So we acknowledge that, when writing up our results to bring out the comparisons within our relatively restricted geographical area, we have not included what could be called the voices of the inhabitants of these ecologies, as discussed immediately below. We hope future work with our colleagues will allow them to develop and publish their opinions on these matters.


13 Although the name is pronounced the same, they bear no connection to and should not be confused with the Chamba Leeko who invaded the region in the late eighteenth century (Fardon 1988). LMP give several postulates for how ecological linguistics should be undertaken. Their postulate 5 is "The speech situation, or the speakers" situated talk, must play a key role in linguistic analysis. Empirical data collecting and corpus analysis are the preferred methods of operation in ecological linguistics' (p. 31). Although we agree with LMP about the importance of natural language data, we do not follow LMP's postulate 5 here, because the comparative approach adopted does not lend itself to this type of investigation. (We are not conducting a "linguistic analysis" in the sense they seem use.) One of the effects of our sort of comparative approach (what Candea (2019) calls "lateral comparison"comparing different non-Western cases without using a Western reference point) is that local voices drop out since they are so disparate. Moreover, although our cases are close geographically, they are not in close touch (as our interaction matrix in Table 3 below makes clear). So people in Somié do not really know about the case of Sombə and vice versa, and though people in Somié may know about Wawa, this is superficial at best. At the price of some repetition, we have tried to lay out the ways in which (even within a limited geographical area) the linguistic ecologies may differ.

Four case studies
Our stance and aims are both descriptive and analytic.¹⁴ We present four case studies that illustrate the situation in one rural area of Africa and show more generally that even an adequate understanding of language endangerment and the factors that drive it can best be achieved through consideration of the overall ecology of a language (or, the linguistic ecology of a region) and its dynamics, i.e. the dynamics of the speech communities in which it figures. Our case studies involve several related languages found in a relatively small and remote area in the Nigeria-Cameroon borderland. The study sites are located relatively close together. These languages reflect various degrees of vitality which, using the terminology proposed in Krauss (2007),¹⁵ range from stable to critically endangered. They exemplify the range of vitality found among Mambiloid languages and provide an interesting set of varying circumstances. They are: (1) the village of Somié on the Tikar Plain in Adamawa Region (formerly Province) in Cameroon; (2) the neighbouring but much smaller village of Mbondjanga; (3) the village of Oumyari, also in Adamawa Region but further north on the Adamawa Plateau; (4) the village of Kila Yang on the Mambila Plateau in Nigeria. We present our subject initially in terms of locations (of speech communities) and describe these locations in more detail below. We then summarise and discuss the similarities and differences among the four. The final section of the paper moves beyond these specific examples, to return to more general issues.  Krauss's (2007) framework offers criteria to determine the following degrees of language vitality: safe; stable; instable, eroded; definitely endangered; severely endangered; critically endangered; extinct. Letter grades (A-D) with possible +/allow for further refinement of a classification.
The four case study villages, their languages and their interactions, are summarised in the following tables. Table 2 lists the four villages together with language name and ISO code, an estimate of the number of speakers, its endangered status following the system introduced in Krauss (2007), and a listing of the languages most commonly used at each location, in order of frequency. Further details are given in each of the case studies Table 3 gives a matrix showing, in rough terms, the extent of interaction among the four languages themselves examined in this study, with Y indicating regular interaction, O occasional or limited interaction, and X no substantial direct interaction. These assessments are based on our own field research going back to 1985, but do not preclude the possibility of greater interaction in earlier times. Some oral traditions, for example, suggest closer contact between the Sombə and the Wawa (though not specifically Oumyari) than has existed in recent decades. The relative lack of contact among our four case studies suggests or reflects a degree independence that exists between them, found in the two inner circles of the LMP model, despite the commonalities of the outer circle.
The locations of these villages and the areas in the Nigerian-Cameroonian borderlands where the languages concerned are spoken are given in the following maps (Figures 2, 3, 4, and Table 4).
The languages in question are all a part of the Mambiloid group (discussed in Section 3.2 below), uncontroversially accepted as part of Bantoid, and so relatively closely related to the well-known Bantu languages. Figure 5 gives in broad outline the traditional view of the classification of the Bantoid languages shown on the map in Figure 3. There is considerable disagreement among scholars as to the relationships among the various groups represented in Figure 5. Most contentious is the division between North and South Bantoid: Boyd (1994) rejects it and indeed disagrees that Dakoid is to be included; Piron (1997) has Mambiloid relatively close to Tiv, thus also abnegating the North-South division. As this is not the place for detailed discussion of these issues, we simply note that recent work by Grollemund et al. (to appear) also rejects the North-South division; there, Ekoid and Bendi are grouped together as the first branching off Bantoid; Mambiloid is embedded more deeply in the tree forming a node with Tivoid and together these two form a node with Beboid.
A general point to be made from the conjunction of the Bantoid tree with the maps above is that the area where Mambiloid languages are spoken is in the wider area of what is regarded to be the homeland of Bantu languages, so our cases fall within the broad category of the "Bantu who stayed at home." Thus, the ancestors of the speakers of Mambiloid languages ("pre-Mambiloid") today are a part of the wider Bantoid grouping and have inhabited the region, conceivably, for millennia. This is not to suggest that they have been static (there have been local movements, e.g. to and from the Mambila and Adamawa plateaus), and additionally, there have been incoming groups or influences; the relatively recent arrival of the Fulɓe being but the most obvious and the growing presence of the Yamba, incomers from the Northwest Region of Cameroon, the most recent. It is also interesting to note, given their geographic proximity, that in this revised version, Tikar is more closely related to Grassfields than it is to Mambiloid (although this is not evident in the tree of Figure 5, see Figure 6 and discussion below).
While here we consider in some detail four local languages, many others are and have been spoken in the area and play a part in our discussion. The four principal languages are Bà Mambila, Njanga, Oumyari, and Sombə. Of these languages, Bà Mambila is a West Mambila lect (see Section 3.2) spoken in and around Somié and several other neighbouring villages, with minor variation across villages; nearby are Njerep, an East Mambila lect also spoken in and around Somié that, with just three or four¹⁶ rememberers, is now critically endangered, if not actually extinct, and Kasabe, another East Mambila lect now extinct, but most recently spoken in the immediate vicinity of Somié and other Bà Mambila speaking villages. Njanga, also critically endangered, is nominally a Kwanja lect known by about ten people¹⁷ and which was formerly the language of the village of Mbondjanga but has now given way to Sundani, another Kwanja lect. A near neighbour is Cambap, an East Mambila lect whose remaining 20-30 speakers are scattered in several Kwanja speaking villages, including Mbondjanga, and is likewise critically endangered. Oumyari is the variety of Wawa spoken in the small village of Oumyari which, though still transmitted intergenerationally, is under serious threat especially from the regional lingua franca Fulfulde and to a lesser extent Vute; in Krauss's scheme it would be considered "eroded." And finally, Sombə was a blacksmith language spoken in the village of Kila Yang on the Mambila Plateau; this language had just two remaining speakers in 2010. By Krauss's criteria, Sombə is critically endangered, though it differs from Njanga (and Njerep) in that both speakers are fluent in their language. The primary language of Kila Yang is Maberem, an East Mambila lect; at least two other semi-speakers (perhaps more appropriately, "rememberers"¹⁸) of Sombə were also living  16 It may seem trivial to count or identify the number of speakers when there are so few; the issue here is not so much with identifying speakers (though in endangerment situations this can indeed be problematic), as with the degree of competence a given person may need in order to qualify as a "speaker." 17 This was the count at the time of Robson's research (Robson 2011) in the village. Of these, only two were fluent speakers, another two were semi-speakers and the remaining six could only be considered "rememberers." The two fluent speakerswho were not only related by kin (an uncle and nephew), but also through friendship and through a political relationship since one was chief and the other advisorused Njanga for private conversations. Unfortunately in 2011 one of them passed away. In his absence it is highly unlikely that the lect is used in any situation anymore. 18 See Grinevald and Bert (2011: 51), broadly those who have had some competence and may be able to recall some information about a language. in the village of Hore Taram Torbi in Adamawa Region, Cameroon, at the time of our field research. Fulfulde, as noted, is the regional lingua franca. It is the language of the largely nomadic and pastoralist Fulɓe (also called Fulani, Peul) and has spread across the West African Sahel from its homeland in Senegal as far east as Sudan. The Fulɓe arrived in the Adamawa region beginning in the middle years of the nineteenth century bringing with them Islam. Fulfulde is classified as Atlantic, an early offshoot of Niger-Congo, and so is only very distantly related to the languages of the Mambiloid region.

General description of physical geography, demography
The area under consideration straddles the Nigeria-Cameroon border and can be split into three altitude levels: the Tikar Plain (alt. c. 700 m) shown in dark green and turquoise at the bottom half of Figure  terms, "dissected plateaus;"¹⁹ the escarpment that marks the boundary between the Mambila Plateau on one hand and both the Adamawa Plateau and Tikar Plain on the other also marks the international frontier. Both the Adamawa Plateau and Tikar Plain are found in Cameroon's Adamawa Region (the Tikar Plain also extends into North West Region), while the Mambila Plateau is a part of Taraba State in Nigeria. As the highest part of Nigeria, it has a relatively cool climate and has seen the arrival of increasing numbers of cattle with their attendant Fulɓe herders since the 1920s. There are now sizeable Fulɓe populations on the Mambila Plateau, some of whom are settled in the towns, as well as some Mambila ownership of cattle, blurring the ideologically neat distinctions between nomadic Fulɓe cattle owners and sedentary Mambila farmers. Indeed, the entire area we are concerned with may be characterised by agro-pastoral disputes, and associated stereotypes of enmity between agriculturalists (of various ethnicities) and Fulɓe pastoralists. The Tikar Plain is highly fertile, and until recently has only had cattle temporarily during the dry season transhumance, although that has begun to change this century. The farmers grow maize and groundnuts both for subsistence and as cash crops, and in partially cleared forests, there are extensive coffee plantations established since the 1950s. The Mapé dam, near Magba, south and west of Somié, was completed in 1987 and has had an as yet unexplored but nonetheless undoubted influence on the ecology, both biological and linguistic, of the region. The flooding of part of the plain precipitated a number of important changes in the region, from the moving of villages, the creation of a new trunk road, the consequent establishment of a new market along the roadand consequently bringing more speakers of different languages into contactto introducing changes in the local diet (e.g. greater availability of fish as a source of protein).
A tin mine was active on the Adamawa Plateau from 1933-1986 near Mayo Darlé, a former Kwanja village. While mining was underway, it drew in migrant workers from throughout the region and beyond, with linguistic consequences for everyone. We infer that Fulfulde, the local lingua franca, became more widely used in Mayo Darlé than in other villages nearby since the population became more linguistically heterogeneous with the influx of migrants to the mine. For the farmers in the neighbouring villages, an indirect consequence of the mine was to make farming for cash crops more profitable since there was a local market for agricultural produce (see Keen and Zeitlyn 2007).

General linguistic background
The Mambila Plateau, and the Mambila region generally, is home to substantial linguistic diversity: several small languages and a multitude of varieties of Mambila, the language which has lent its name to the genealogical grouping, Mambiloid, which includes these languages (see Connell, 2021). Several of these languages are very small, having only a few hundred speakers at most. Some of them are much smaller than even that, as already mentioned.
Mambiloid is a subgroup of Bantoid, as shown in Figure 5, and so part of the Benue-Congo branch of Niger-Congo. The status of Mambiloid as a valid subgrouping within Bantoid has been argued in Blench (1993) and Connell (2010c) among others, but challenged by Güldemann (2017). Connell (2021) presents evidence based on the comparative method and on lexical evidence which supports the integrity of the grouping. The dendrogram in Figure 6 shows a subgrouping of the Mambiloid languages, based on a phylogenetic analysis of the lexical evidence. In the analysis of this more recent study, Tikar was included as one of the outgroups and thus is included here. The languages of the four case studies of the present work are in bold.
The Mambila constitute the largest of the various ethnolinguistic groups that make up Mambiloid; they number around 100,000, perhaps more,²⁰ and the bulk of their population inhabit the Mambila Plateau on   21 Since many or most of the languages mentioned here are unknown in the literature, we include ISO 693-3 codes at first mention. Not all have as yet been given an ISO code. The division of Mambila into two clusters, "Nigerian" and "Cameroon" as in Eberhard et al. (2020) is purely on geographical grounds; the assignment of two ISO 693-3 codes is ill-founded.
international border. Mutual intelligibility is restricted within each cluster and the various lects tend to form a dialect chain or continuum in this respect; between the two clusters mutual intelligibility does not exist, although it is not uncommon to find speakers who, due to exposure, can understand a wide range of (also known in the literature as Somyev), spoken at Kila Yang, is the subject of the fourth case study presented below and is on the verge of extinction, having just two remaining speakers in 2013. As mentioned above, there are also Mambila lects on the Cameroon side of the international frontier. In addition, Kwanja [knp], spoken by about 10,000 people, has two principal varieties, Sundani and Ndung; Njanga, subject of our second case study, is considered to be a dialect of Kwanja, and the now extinct Bung is sometimes locally considered to have been fourth lect, though see Connell (2017) for discussion of the status of these two latter languages. Wawa [www] has a total population of approximately 3,000 speakers divided among four principal lects, one of which, Oumyari, is our third case study. Vute [vut] has a total population of approximately 21,000 speakers and at least eight different varieties and geographically is the most widespread of the Mambiloid group. Finally, in the north-eastern extreme of the Mambloid area is Nizaa [sgi], spoken in and around Galim by approximately 10,000 people. The geographical distribution of the Mambiloid languages is shown in Figure 7. There are also a number of other, non-Mambiloid, languages in the region. Most notable of these is Fulfulde ([fub]; Atlantic), the lingua franca of Northern Cameroon and adjoining parts of Nigeria (at least the Mambila Plateau), and so throughout the Mambiloid region except for the lower half of the Vute speaking area. Tikar ([tik]; Bantoid) overlaps the southwest of the Mambiloid area; it was formerly more widespread, but over the last few centuries has been displaced by expanding Mambila and, more recently, Kwanja speakers. English and French are the languages of administration and education in Nigeria and Cameroon, respectively; Hausa ([hau], Chadic) is growing in importance and influence on the Mambila Plateau, though it is unlikely to replace Fulfulde as lingua franca in the foreseeable future. Hausa has also  gained some influence in the Banyo area, as discussed below; Yamba ([yam], Grassfields Bantoid), an adjacent Grassfields Bantu language to the west, is now increasingly expanding into the Mambila and Kwanja areas of Cameroon and the Mambila Plateau. Yamba speakers also bring with them increased exposure to English and Cameroon Pidgin [wes]. Gbaya [gya], Mbum [mdd], and Peere [pfe] (all Adamawa) border the northeast and north of Mambiloid region and are frequently encountered in the Nizaa and northern Vute areas. As the foregoing discussion implies, multilingualism is the norm, but the heterogeneous linguistic situation also means substantial language contact exists beyond this in the daily lives of the region's inhabitants, and this is a pervasive feature in each of the four case studies we present.

Somié and the Bà Mambila language 4.1 General description
The village of Somié is located on the Tikar Plain, in the southern part of Adamawa Region at the base of the Mambila escarpment. The village is situated at the end of a local road and as such is relatively remote, though improvements in communications, particularly in road maintenance, have progressively reduced its isolation since the 1980s. The road is passable especially during the dry season, though during the rainy season travel can become difficult. Other roads leading out of the village, towards the north of Cameroon and to Nigeria, are under construction, and though only partially completed, have led to improved communications with these regions.²² Similarly, the road from Sabongari (and beyond that connecting to Ndu on the ring road) in Northwest Region to Atta, west of Somié, (opened c. 2000) while not reaching all the way to Somié, has led to greater ease of contact between the Cameroon Grassfields to the south and the Mambila region.
The village itself is comprised of "Somié Centre," which has identified neighbourhoods or "quartiers," as well as a number of outlying hamlets. Somié Centre has a population of approximately 2,500; the overall population of the village is closer to 3,000. Some of the quartiers are associated with groups speaking (or having spoken) particular languages, the clearest examples being Njerep quartier where the last Njerep speakers lived, and Tor Luo hamlet where the Kasabe language²³ was spoken. However, all quartiers have inhabitants who reflect connection to several different Mambila groups and exemplify the fractal nature of local histories (Connell and Zeitlyn 2000;Zeitlyn and Connell 2003;Zeitlyn et al. 2000).
Village amenities are few: a gravity-based water system has supplied a network of village standpipes since 2002; until very recently, there was no distributed electricity service, although a small but growing number of residents now have petrol-powered generators. Several villages including Somié are included in a scheme for rural solar electricity that was operational by late 2017. By mid-2007, the cellular telephone network coverage had been expanded so that it included the village, though successful communication required an external antenna. It is slowly being improved, and limited internet access was possible by 2015.

Aspects of the social environment
The indigenous people of the region are peasant farmers who to some extent supplement their livelihood through cultivation of cash crops such as coffee and maize. The society is not stratified, and language  22 Many online and paper maps show a non-existent motorable road connecting Somié directly to Mayo Darlé. The populations of both centres would be very happy if reality corresponded to the maps. In 2013/2014 a new path, passable by motor bikes was cut from Somié to Mayo Darlé by the communal work of men from Somié. This needs a large bridge to be built before it is useable by cars or lorries. 23 Njerep is critically endangered; Kasabe is now extinct. knowledge or use does not serve as a social marker of rank (or seniority). As discussed below, different languages serve different purposes, and the choice of language in a given situation is primarily pragmatic, governed by practical considerations and not straightforwardly by issues of prestige. It is not, however, a diglossic (or polyglossic) situation in a narrow understanding of the term, as different languages may be used in the same situation depending, for example, on who is present and there is no attendant status difference accorded to the different languages. This is examined below, in the discussion of functional domains. So, of all the languages used, none can be said to be of greater prestige than the others, though the status of French appears to be changing: in recent years, through increased opening of the region to outside influences, and as recognition of the importance of education grows, impressionistically at least, the status of French has been enhanced, and its use increased. For similar reasons, the use of English appears to be on the increase, at least in certain situations. Finally, Islam and Christianity exist side-by-side in Somié. While most Bà Mambila speakers are Christian, many of the immigrant Mambila, largely from Nigeria, are Muslim; the Fulɓe are Muslim and the Yamba are almost exclusively Baptist Christians. As a result, there is a mosque²⁴ and three churches, Protestant, Catholic, and Baptist.

Synoptic history of Somié
The complex history of Somié village has been explored in Zeitlyn (1992), Zeitlyn and Connell (2003), and Zeitlyn, Mial and Mbe (2000). We stress we have no reason to think that its history is exceptional in its complexity; rather we see it as typical and work with the assumption that all local histories will have similar degrees of complexity unless there is compelling evidence to the contrary. A synopsis of the Somié case follows.
There is overwhelming consistency in the oral histories we have collected over more than 30 years, as well as some archival evidence, concurring that Mambila on the Tikar Plain came from the Mambila Plateau and the adjoining areas of the Adamawa Plateau.²⁵ It is within this context that the histories of languages such as Bà Mambila and Njerep are to be understood. The present population of Somié is the result of several (at least three, and possibly four) waves of immigration by different groups of Mambila down from the Mambila Plateau onto the Tikar Plain. No clear and detailed picture emerges from the oral history concerning these events, as accounts from different informants are contradictory (see sources cited at beginning of this section). The named groups of immigrants are Liap, Ndeba, Njerep, and Mvop. It is a matter of controversy whether the Liap or the Ndeba were the first arrivals, though both may have had a hand in pushing the Tikar away from the base of the escarpment and further into the plain. Little is now known of the Liap other than that they came from the area around Chichale mountain, near Guessimi on the Adamawa Plateau. In Somié, some people are still occasionally described as Liap through patrilateral descent. Little is also known of the Ndeba, other than that they too came from the Guessimi area, and that they gave their name to the village. They are said to be the people who dug the trenched forts visible in aerial photographs and on the ground at Gumbe and in the forest of Duabang (see Saulieu et al. 2015), hamlets near to Somié.
The early proto-Mambila immigrants were conquered or absorbed in their turn by subsequent waves of Mambila invaders.²⁶ The second wave of immigrants to the present Somié area (considering both Liap and Ndeba as part of the first) is itself divided into different groups. All accounts agree this wave originated on  24 There is a main mosque on the central square and several smaller ones scattered through the village. The Baptist church is not in the village centre unlike the Protestant and Catholic churches. 25 Of course, once on the plain there has been assimilation of people from neighbouring groups through marriage and other forms of inter-village movement. 26 Some carbon dates have been reported from the Somié area from as early as 131 CE but these cannot be associated with a specific group (Nizésété and Zeitlyn 2008; see also Saulieu et al. 2015 for more recent dates). Granted the relatively secure oral histories of the final wave of immigrants (the Mvop) which started mid to late nineteenth century the earliest proto-Mambila immigrants must have been at least a century if not several centuries before this.
the Adamawa Plateau, probably in the area of Djeni Mountain (Aigue Mboundo on current maps of Cameroon; 6°37′51.46″N, 11°33′13.07″E) and introduced the Njerep and Kasabe (aka Luo). The linguistic evidence (see Connell, 2021;Zeitlyn and Connell 2003) establishes that Njerep and Kasabe, and a third language, Cambap (spoken by the Camba) were closely related.²⁷ However, elicited accounts differ as to the degree of differentiation between the Njerep and the Kasabe. The most common version suggests these two groups were most closely related to the Camba (aka Twendi; see Connell 1998, Connell 2002a, perhaps only differentiating when they left the villages of Sango and Tchamba (intermediate stops, already on the Tikar plain) fleeing to the hills at the foot of the Mambila escarpment to escape raiders on horseback. The horsemen in most accounts are Fulɓe invaders, though at least one local historian suggests this movement was earlier and precipitated by the Chamba invasion of a slightly earlier period.²⁸ Although descendants of the Njerep and Kasabe now live at the foot of the Mambila escarpment, on the Tikar plain, they still identify the villages in the mountains which they established, and one local mountain bears the name Tor Luo (tɔ̀r = "hill, mountain"; Luo = Kasabe). What is unclear is whether these two differentiated themselves from Camba in the way just described, or at an earlier time, possibly when leaving the Adamawa Plateau homeland. Perhaps one pointer with some bearing on the question is that the Kasabe claim to have songs not known to the Njerep. This alone cannot be taken as evidence of a long history of separation, but does point to a certain degree of cultural differentiation. Whatever the case, the Njerep and Luo are now separate and appear to have been so since they reached their present locations. Their separate accounts of relations with their neighbours, as discussed below, confirm this especially since the Njerep (but not, apparently, the Kasabe) claim to have conquered the Ndeba.
The last wave of Mambila immigrants onto the Tikar Plain were the Mvop who came down the escarpment from the village of Mvor on the Mambila Plateau, southwest of Dorofi. The site of Mvor has been documented by Jean Hurault (1979Hurault ( , 1986Hurault ( , 1988. Oral tradition tells us that a group of children of Tulum, their common ancestor in Mvor, founded the villages of Sonkolong, Somié and Atta. Some of the grand-children²⁹ of Tulum were ruling as chiefs when the Germans first arrived in the area at the beginning of the twentieth century.

Linguistic overview
The home language and language of daily life of most inhabitants of Somié is Bà Mambila,³⁰ as discussed above, a variety of Mambila. Somié is the smallest of three major villages in which Bà Mambila is spoken, the other two being Sonkolong and Atta. There are several smaller hamlets situated around and between the three villages where Bà Mambila is also spoken. Although some variation exists across the Bà Mambila speaking area, lexically, phonologically, and grammatically, this is minor and does not impede intelligibility or recognition that all speak the same variety of Mambila.
Most, if not all, Bà Mambila speakers also have good command of Fulfulde, the lingua franca of Northern Cameroon. There is a small number of Fulɓe resident in the village, as well as a seasonal influx  27 As well as Yeni, another language whose existence is known through mention in a song but little more survives than the name; see Connell (1998). Gausset (2010) identifies its (former) speakers as the Yeka, now absorbed by the Kwanja. 28 Mial Nicodeme, citing the published work of Eldridge Mohammadou (1990Mohammadou ( , 1991. There is potential for confusion between the two names "Camba" and "Chamba;" both are pronounced the same, [tʃamba], though at present there is no evidence of a connection between the two. 29 It is unclear to us how literally we should take the designation of "grandchild:" if literally, it sets limits on the arrival of the Mvop and places it squarely in the middle of the nineteenth century but the term could be being used more figuratively to mean "descendant," in which case the Mvop may have arrived much earlier. We note the date of the Ndeba fort at Somié appears to be late nineteenth century (see Saulieu et al. 2015). 30 "Bà" is the name this Mambila sub-group use to designate themselves; "Ju Bà" literally is "language (of) Bà," which is often also simply expressed "Bà." of nomadic Fulɓe³¹ in the region. Given the status of Fulfulde as a regional lingua franca, very few of the resident Fulɓe have learned to speak Bà Mambila. Fulfulde is used by Bà Mambila speakers primarily in communicating with Fulɓe or strangers who appear to be from the wider region. French is also widely known, particularly by younger generations as it is the language of instruction in the schools. Apart from its use in education, it is used in official situations and in addressing strangers who appear to be from further away or who are non-African.
Several other languages are spoken in Somié, in addition to Bà Mambila, Fulfulde, and French. First among these are several other varieties of Mambila, the most numerous of which is Mbar, also of the West Mambila cluster, with approximately 150-200 native speakers. The main population of Mbar is on the Mambila Plateau in Nigeria (in the village of Mbar), but recent immigration has resulted in a substantial Mbar population settling in Somié and adjacent hamlets. One hamlet in which the Mbar immigrants have settled is Njerep. As was said above, Njerep isor wasalso one of the settlements in which the East Mambila lect of the same name was spoken. At the time of our initial research on Njerep, in the late 1990s, there were seven people who might have been called speakers; all were best classified as rememberers, or in the case of one, perhaps two, as semi-speakers, hence its categorisation above as critically endangered (D); just four were still alive in 2010. The Njerep ethnic identity has been absorbed into the local Mambila (Bà) population (although there remain some local traditions which describe one or two quartiers as being predominantly Njerep), and Bà Mambila is the principal language of all ethnic Njerep; with the exception of retained knowledge of Njerep, their linguistic profile is now virtually identical to that of Bà Mambila.
Other Mambila immigrants from Nigeria have settled in and around the village;³² though they are not numerous in terms of individual language varieties, they represent both East and West Mambila clusters. Recent immigration has also brought a substantial number of Yamba speakers into the vicinity of Somié, coming from the neighbouring Northwest Region of Cameroon. They have settled outside the village proper, forming their own hamlets or neighbourhoods (this seems to be a general pattern for the Yamba, though in nearby Atta village, the main Yamba quartier is now contiguous with the Mambila centre). The Yamba brings with them not only their own language which, like Mambila, is Bantoid,³³ though of the Grassfields subgroup, but also English or, more accurately, English as represented in Cameroon Pidgin,³⁴ and almost all are members of the Cameroon Baptist Church (CBC), which is also relatively new to the area.
In addition to these, the languages of neighbouring people are to some extent spoken or understood, including the two principal varieties of Kwanja (Sundani and Ndung) and Tikar, all of which are Bantoid; Kwanja is part of the Mambiloid subgroup and is discussed below in the section on Njanga, while Tikar forms a branch on its own within the larger Bantoid grouping. In addition to these languages, many people have learned or at least been exposed to additional languages through travel to other parts of Cameroon.

Speech communities
The characterisation given here is of the main population of Somié, the Bà Mambila. In addition to their principal language, all Bà Mambila speak Fulfulde fluently (except the very young) and most have some fluency in French; those who have been to school are, of course, quite fluent in French. In addition to these  31 Also known as Mbororo. The relationship between Fulɓe and Mbororo is tendentious and complicated. There is a substantial literature on this, but largely irrelevant to this paper (see Stenning 1959 for a classic study to the north and Pelican 2015 for a more recent study somewhat to the south of the area we have been working in). 32 While we refer here to recent and ongoing immigration, with its own specific causes, it may be seen in context of movements spanning centuries. See Zeitlyn and Connell (2003). 33 The relationship between Yamba and Mambila, though both are Bantoid, is not particularly close. It would be roughly on the order of that between English and Swedish, both of which are Germanic languages. Tikar, mentioned immediately below, is even more distant, though still Bantoid. 34 Also known as Kamtok or "Wes Cos" (see Ayafor and Green 2017). three languages, many inhabitants of Somié will have some degree of fluency in one or more other languages, depending on their personal background. Having more or less basic competence in five other languages is regarded as unexceptional in the village. Many have some familiarity with Kwanja, whose territory is adjacent. Those who have travelled and lived away from Somié typically return with knowledge of the language spoken where they were. This characterisation does not necessarily apply to other groups. For example, with few exceptions, the Fulɓe do not learn Mambila as their language is the regional lingua franca, and others speak it;³⁵ although many Fulɓe also speak French, they do not master it or use it to the same extent as do the Mambila. Fulɓe and Muslim Mambila also know Arabic, at least to the extent it is necessary for religious purposes.

Language-identity relationships
As has been suggested in the introduction, the relationship between language and (ethnic) identity is complex and hard to characterise partly because the entities themselves (language and (ethnic) identity) are not suitable analytic terms for the region under study (they are emic not etic terms). However, there is a clear central local archetype or exemplar (e.g. of a Mambila person born to Mambila-speaking Mambila parents in the village where they too were born) and this is stereotypically extended. Beyond the archetype, it becomes complicated, unclear, and subject to negotiation very, very quickly.³⁶ The existence of the stereotypes provides flexibility and opportunity for language switching: a person may argue that "since I am unambiguously Mambila I can demonstrate my sophistication by speaking French, or my international Islamic credentials by speaking Fulfulde." Di Carlo (2018) discusses parallel cases from Lower Fungom some 150 km away.

Functional domains
A number of functional domains has been identified, exploration of which help describe and characterise the social relations among the different languages used in Somié and the wider region (which languages are used in which situations and by whom). The list presented here is not intended to be definitive or exhaustive; it is worth noting more generally that the set of functional domains significant for a given society is determined by the nature of that society.
Home: Bà Mambila is spoken as a home language among all family members, of all generations.
In the village: Bà Mambila is similarly used with friends, and with elders; with anyone known to be from the village or elsewhere in the Bà Mambila speaking area. With strangers, either Fulfulde or French is used; the choice is dependent upon whether the stranger appears to be from the wider region, or whether he/she initiates the exchange using French or Fulfulde. Visiting Mambila from Nigeria, or who may be speakers of an East Mambila lect, will generally be addressed in Fulfulde.
 35 A reviewer suggests that the real reason behind so few Fulɓe learning Mambila (or, we presume, other local languages) is "due to political and ideological concerns which have, in part, led to Fulfulde becoming a lingua franca;" we agree that certainly at some stage political dominance played an important role in establishing Fulfulde as the lingua franca, but we don't see these factors as necessarily of the same importance in the present time. See also Akumbu and Chie (2020). 36 We continue to use terms such as Mambila (for both language and identity) because we recognise that we risk some classic problems of academics telling people on the ground that they should be paying more attention to abstruse academic arguments and behaving differently. We can easily imagine the reaction we would get in Somié if we were to say there is no such thing as being Mambila. To be clear, if we have one message it is a different (albeit also classic) academic message: yes of course there are Mambila people speaking the Mambila language but it is not as simple as it might seem at first. Our message is that it is not clear-cut, it is messy! Market: Language choice in the market at Somié (a weekly market) is governed by a number of factors, but is generally subject to negotiation. Details are reported in Connell (2009). In that study, 14 languages were reported as being used in Somié market on one given market day (16 July) in 2005. Of the observed transactions, Fulfulde was used in 42%, Bà Mambila in 20%, Mbar Mambila in 13%, and French in 7%.
Education: As mentioned earlier, there are three official primary schools in Somié, originally a government school and a Protestant school in both of which French is the language of instruction. Since 2010, a secondary school has been opened in Somié which is officially bilingual (French, English). It predominantly uses French as the language of instruction, with an increasing use of Pidgin reflecting the increasing number of Yamba children. The children of Moslem parents are sent to an unofficial Quranic school in the village, which uses Fulfulde and Arabic.
Religion: The presence of both Christianity and Islam was discussed above. Within the Christian communities, Catholic and Protestant each have their own church, as do the Baptists (who are exclusively anglophone Yamba³⁷). There is a strong association between language and religion: Fulfulde and to some extent Arabic are used in the mosque among Muslims; services in the Catholic and Protestant churches are primarily in French, though portions of the service, including some hymns, may be performed in Bà Mambila, while Pidgin and Yamba are used in the Baptist church. The New Testament has been translated into Bà Mambila (Sonkolong variety) as well as the Gembu dialect of Mambila.
Three components of Mambila traditional religion should also be noted for their linguistic aspects: the sùàgà ritual oath, the Ŋgwùn ritual surrounding the chief, which includes a war dance, and ŋgàm divination. Bà Mambila continues to be used almost exclusively in the practice of these rituals. Court hearings often end with a sùàgà oath which is sealed either by licking the Chief's ritual stool (symbolising the office of the Chefferie) or by a sùàgà sacrifice. These oaths are accompanied by oratory which Zeitlyn (1994) reports need not be in Mambila (although all those witnessed have been). Two of the ritual chiefs who perform the secret rites for the chief at the biennial Ŋgwùn festival are chosen from Nigerian Mambila immigrants to Somié. We interpret this as a deliberate strategy to integrate the immigrant community into the village. However, a consequence of this has been that much of the mundane conversation between the five ritual chiefs about the ritual has been conducted in Fulfulde because of the limits on Bà Mambila -Mbar dialect inter-comprehension. Ŋgàm divination is always undertaken in Bà Mambila, although Zeitlyn has been told it understands all languages. Since some clients are non-Bà Mambila speakers, discussion between client and diviner about the questions to be posed and the responses received may be in French or Fulfulde.
Outside the village: Since, as described earlier, Bà Mambila is spoken in several villages in the surrounding area, the language choices involved for Bà Mambila speakers when travelling outside the village, but remaining in the area, are essentially identical to those given above for in the village, though with some modification. The village of Sonkolong is more mixed linguistically than other Bà Mambila-speaking villages; for example, it includes more Tikar speakers than other Mambila villages. As a result, Fulfulde, as the local lingua franca, may play a bigger role. Journeys further afield, e.g. further north in Cameroon or to the Mambila speaking part of Nigeria (the Mambila Plateau), require use of Fulfulde, while trips south will involve use of French or Cameroon Pidgin.
Town: There are two major towns in the wider Mambiloid region, both of which are administrative centres: Banyo to the north, which is a day's journey distant by vehicle, and Bankim to the south, half a day away. Two smaller towns, northeast and south respectively, are Mayo Darlé and Nyamboya. In none of these towns will a Somié resident use Bà Mambila, unless meeting other people already known to be Bà Mambila speakers, whether visiting or resident. This is possible since these towns are now all linguistically heterogeneous, though Bankim is predominantly Tikar-speaking. Banyo was originally Vute-speaking but now Fulfulde predominates. Mayo Darlé³⁸ and Nyamboya are both originally Kwanja-speaking; again, as  37 In the early 2010s the first marriages between Mambila women and Yamba men were recorded so this statement may not hold into the future. 38 Mayo Darlé became an important multi-ethnic settlement following the development of a tin mine nearby in the 1950s.
Fulfulde is the regional lingua franca, it is the language of choice in these settings, except when official or governmental business is being pursued (discussed below). Bankim is the exception, where both French and Fulfulde serve, the former more than the latter. Otherwise, Bà Mambila speakers who happen also to speak Tikar, Vute, or Kwanja will tend to use these languages when encountering speakers of them, or codeswitch between them and Fulfulde.
Formal situations: The relationship between younger and older, e.g. children and parents, is marked linguistically by common patterns such as parents using names for their children but being addressed with kin terms (e.g. tele "father" and mií "mother;" see Zeitlyn 1993Zeitlyn , 2005. Other than that there is relatively little formality to be considered. There are constraints on the use of language between senior and junior inlaws especially in the years immediately after a marriage is established. So, a new wife will not speak in the presence of her mother-in-law for several years and will avoid uttering her name for a long period continuing even after they have begun to speak. A new husband similarly avoids the name of his wife's mother. Apart from this, there are basically three types of formal situation a Bà Mambila speaker might encounter, each of which involves different choices with regard to language: (1) public gatherings, e.g. at a burial or funeral celebration some time after a death; (2) meetings with village elders; and (3) meetings with official or governmental interactions. The first, public gatherings are conducted in Bà Mambila, although Christian services at burials mix French and Bà Mambila. There are also dāmə̀rites, at which blessings are invoked on a family and a compound. These are conducted entirely in Bà Mambila. The second, meetings with village elders may be illustrated by the village court, which are conducted mainly in Bà Mambila, but with some codeswitching to Fulfulde, French, or Pidgin. When a dispute is heard in front of the Chief and the elders at the Chief's Palace, it will be in a mixture of languages spoken by the parties. If all are Bà Mambila speakers, then this will be used primarily with occasional code switching into French and Fulfulde (typically by the Chief, for example, when threatening to get the civil authorities (police) involved if an agreement cannot be reached). In this example, contact with the other languages acts as a metonym for contact with other people. If the disputing parties include non-Bà Mambila speakers, then Fulfulde is increasingly being used; French or Pidgin may also be used, but more rarely. The third situation can be summed up succinctly by saying simply that all official or governmental business is carried out in French. There are two exceptions: if the official with whom one is interacting happens also to be a Bà Mambila speaker, it is likely that discussions might happen in Bà Mambila, but this is not necessarily the case; if a non-Bà Mambila speaking colleague were present, they might well speak in French. Alternatively, if the official(s) concerned come from Banyo or other places in the north, the spoken interaction may occur in Fulfulde though any documents will be written in French.
Informal/casual situations: As suggested by descriptions given above for different domains, to the extent they include informal and/or casual situations, Bà Mambila is overwhelmingly used in these settings. Exceptions might occur if/when a non-Bà Mambila speaking participant is present (including Nigerian Mambila (non-Bà Mambila) speakers), in which case a common language would be chosen, Fulfulde or French, in that order.
Intimate situations: Intimate situations would include conversations or discussions between husband and wife, siblings or close friends, as well as other sorts of private conversation. Again, Bà Mambila is used overwhelmingly, even exclusively in these situations, if it is the first language of the participants. We have no detailed data on the use of Mbar in intimate situations, but given its patterns of usage in other settings, and until evidence to the contrary comes to light, we assume that it is used in intimate situations between Mbar speakers as well.
Entertainment/Media: Mambila traditional singing and dancing has decreased in frequency since Zeitlyn started his research in Somié in 1985. Then, women would regularly perform and sing a clapping dance (mgbè die: "the chief falls") every Friday evening outside the Chief's palace. There were also performers who would sing using mouth bows (kɔgɔŋ) and mvet style chest harps (njáŋ). Although the singing at the big festivals (e.g. the sùàgà masquerades and the Ŋgwùn festival) and to celebrate a marriage (njuŋ) is in Bà Mambila, the regular dancing on market day is now to non-Mambila music. Since approximately 2005, there have been generator-powered TVs in the village using satellite dishes for the signal. There are several (as many as five) places where for a small fee people can sit and watch TV. Football matches (with commentary in French or occasionally English), some TV soaps, mainly South American or Nigerian dubbed into French (via Chanel2), and action movies played from DVDs (VCDs) also with French sound tracks are the most popular. There is no FM radio signal in the village, though those who have a shortwave radio can tune in a range of broadcasting. The most popular station is Africa Numero Un broadcasting in French from Libreville, though Spanish language broadcasts from Equatorial Guinea are also sometimes heard, as is English via the BBC. There is a limited circulation of religious material. Some Islamic sermons circulate on CD and USB memory sticks and the Protestant Church is using sound recordings of extracts from the New Testament in Bà Mambila and in French.

Internal variation, regional
As mentioned, Bà Mambila is spoken in three main villages: Somié, Sonkolong, and Atta, and several smaller villages or hamlets in the surrounding area. Across these villages, there is mutual intelligibility, with everyone recognising they speak (more or less) the same dialect of Mambila: Bà. There is, however, some degree of variation, at least lexically and phonologically, which can be used to signal village identities. No detailed studies of cross-village variation have been conducted as yet, though comparison of the language of the different villages was done by the SIL bible translators who ultimately settled on the Sonkolong dialect as being maximally comprehensible. Sonkolong is situated between Somié and Atta; however, although closer in distance to Atta, the Sonkolong variety of Bà Mambila is closer to that of Somié, again at least with respect to lexicon and phonology. The inter-village differences are likely due, in large part, to different settlement patterns as described above (see Zeitlyn and Connell 2003 for more detail).

Internal variation, social
Informal age groups make limited use of in-group slang and nicknames to signal friendship and to exclude others who might overhear. Although most commonly used by teenagers and youngsters, such restricted codes reappear during major festivals where their use by respected "seniors" both scandalises and amuses the youth. We have not observed language variation by gender. Respect is due by the young to their elders and especially by recently married people to their in-laws. One mark of this is that, as discussed above, recently married people will not address their senior in-laws (particularly marked in the son-in-law, mother-in-law pairing). As a marriage stabilises and children are born, this is relaxed but even when they do address each other directly there is restraint and circumlocution about "delicate" topics. The practice, however, does not approach the degree found in hlonipha among the Zulu and neighbouring people (see Finlayson 1995;Herbert 1990).

Status/solidarity
Status and/or solidarity among speakers is often reflected in use of pronouns as is demonstrated in research going back to the seminal work of Brown and Gilman (1960). In Bà Mambila, there is now widespread use of the second person plural pronoun as singular (so-called T/V markers after "tu vous" usage in French); so the use of 2 PL bí rather than 2 SG wò is relatively common when addressing a senior individual, e.g. the Chief, used as a mark of respect. Older speakers report this as an innovation, so we suspect this is an influence from French as taught in school (where the teachers all receive vous), and note that use of vous is the practice when speaking French. Comparison with Nigerian varieties of Mambila, who are exposed to English rather than French, would provide additional insight into this.

Written tradition, standardisation, institutional support
None of the languages discussed in this section receives official or institutional support; none can be considered standardised or to have a written tradition. Bà Mambila has been given an orthography and a lexicon has been compiled and, as already mentioned, a New Testament translation based on the Sonkolong variety of Bà Mambila has been published. This work was undertaken by Protestant missionaries together with trained local scholars. Little has been done, however, to propagate these documents or use of the orthography, and in Somié, fewer than a dozen people can read or write the language with any degree of competence.

Attitudes of users
The attitudes of Bà Mambila (or Mbar) speakers towards their own and other languages can be summed up as generally one of equality. In view of the earlier discussion that the various languages in a speaker's repertoire have relatively fixed roles, this is to say that none is accorded greater prestige than the others. There are some indications that French may be gaining in prestige, particularly among the young; see below and Connell (2009).

Association between language and economic/political advancement
With Cameroon achieving independence in 1960-1961,³⁹ Fulfulde began to assume a certain degree of prestige politically; for those who aspired to political office, particularly in the north of the country, speaking Fulfulde was important. However, very few people among the local populations would then have had political ambitions, hence the degree of prestige of Fulfulde was limited. Moreover, given the animosity between incoming Fulfulde-speaking pastoralists and the subsistence agricultural practices⁴⁰ of the local people described in Section 3.1, any prestige Fulfulde may have gained has been countervailed. In recent years, largely through increased opening up of the region to outside influences, and as recognition of the importance of education has grown, the status of French has become enhanced, and its use increased relative to Fulfulde. For similar reasons, the use of English or Cameroon Pidgin appears to be on the increase, at least in certain situations.

Summary: Language contact in Somié
In Somié, the clearest cases of language contact are in the village market (see Connell 2009), at the dispensary (where more qualified staff are not locals and use Fulfulde, Cameroon Pidgin, and French with native speaker staff acting as translators where needed), at the local schools (officially using only French and English), and the Chief's court. Change is clearly apparent in the latter and brief discussion of this serves to illustrate changes in the language ecology of Somié that reflect increasing language contact. In the mid 1980s, most cases were heard in Bà Mambila with occasional codeswitching into Fulfulde and French. By the mid 2010s, in response to increased immigration from non-Bà Mambila speaking Mambila from Nigeria as well as Yamba from North West region of Cameroon, the use of Fulfulde as the local lingua  39 The parts of Cameroon under the French and British Mandates became independent at different dates. 40 The agricultural practices were described in Section 3.1. franca had increased considerably. By contrast with the 1980s (when research data starts), the presence of (Francophone) police officers is now not unusual when cases are heard. In 2012, a log of 214 cases heard in the Chief's court recorded the main language used. This had the following breakdown: Pidgin English 2 cases; French 24 (inc. one mixed with Fulfulde); Fulfulde 64 (not including the mixed cases); Bà Mambila 124 (inc. one mixed with Fulfulde). We do not have comparable quantitative data from the 1980s, but our fieldnotes from that period suggest that the only other language apart from Bà Mambila used for entire hearings was Fulfulde and that even this was at the time not common. It would appear then that a shift in language use in this domain is underway.
It is not difficult to find apparent loanwords in Bà Mambila. Instances from Fulfulde include saa, used alongside Bà Mambila njua, both meaning luck (there are hints that contrastive uses are developing, njua to mean "uncontrollable random luck," while saa can be crafted) as well as words in a diverse range of semantic fields: pātú "cat" (Ful paturu); ŋɡàm "because" (Ful ŋɡam); and jɔ̀ŋ "life" (Ful yonki). Possible loanwords from Tikar are also present; we note here mbáŋ "granary" and ntàn "hippopotamus," but with our present state of knowledge in several cases it is difficult to tell whether a given form has been borrowed since cognates are also present elsewhere in Mambiloid. Similarly, comparison between Bà Mambila and Tikar draft dictionaries reveals a number of common borrowings from Fulfulde which are unsurprising given the influence of Fulfulde in the region. And, again, there are words which are common throughout Mambiloid and others common to the wider Bantoid group.
Finally, we note two characteristics of Bà Mambila which it shares with other West Mambila lects: its predominant CV(C) word structure and its analytic nature. These may be compared to East Mambila, whose lects are primarily CVCV and feature a greater degree of inflection, especially in the verb phrase. This division is also shared with the two main Kwanja varieties (see below) and is considered to be a result of language contact at a much earlier period, perhaps a substratum influence. Detailed investigation of what languages this substratum may have comprised has yet to be undertaken, but speculation suggests other Bantoid languages. The West Mambila lect Len shows influence from nearby Grassfields (Connell 2007), which is not obviously due to direct, adstratal, influence. Beyond this, the reduced structure and analytic nature noted for West Mambila is somewhat reminiscent of Tikar.
So, despite the high degree of multilingualism in Somié, including the fact that virtually all Bà Mambila speakers are fluent in the regional lingua franca Fulfulde, and that fluency in French is also widespread, Bà Mambila remains relatively stable. The various languages that comprise the repertoires of the inhabitants of Somié, in particular the three most prominent (Bà Mambila, Fulfulde, and French), cannot be said to be in "competition" in a strict sense in that their rolesthe situations/domains in which each is usedare reasonably well-defined. Exceptions to Bà Mambila being spoken in those domains where it is the preferred language come about only when non-Bà Mambila speakers are present.

General description of physical geography
Mbondjanga village, where Njanga is spoken, is located on the Tikar Plain. Like Somié, it is relatively isolated and found at the end of a six km access road. It is surrounded by bush land: to the south are forest and savannah as well as land that has been cleared and used for farming; to the west more savannah and hills leading towards the Mambila Plateau; to the north the forested escarpment of the Adamawa Plateau; and to the east more forest and farmed land. There are over forty rivers or streams flowing through Mbondjanga's territory. The area is rich in plant and animal life including monkeys, various species of antelope, other mammals, birds, and reptiles. It is known in the area for its bananas, fish, and Atari kola. The village is also known for the ndúngína tree from which the bark that the Kwanja traditionally use in healing rituals is taken. This tree is significant for the village's identity and customs.
The regional concept of "village" denotes an area or a territory governed centrally by a traditional chief, but comprising the surrounding uninhabited land which nonetheless belongs to the people of that territory. Consequently, farmland rights and those for hunting, fishing, and harvesting trees/wood fall to the people of the territory (specifically to the chief). Mbondjanga's surrounding territory covers at least 18 square kilometres, stretching from the Adamawa Plateau in the north to the savannah of the Tikar Plain in the south. It borders that of the Camba (at the village of Tchamba), the Kwanja of Yimbere, the Kwanja of Mayo Djinga, the Taceme-Mambila of Ribao, the Bà-Mambila of Somié, and the Tikar of Kongui, as well as the land of the former Mayo Darlé mine, called Koudja or simply Mayo Darlé Mine.
To the south, a narrow dirt road six km long (a one hour walk) leads out to the crossroad village of Carrefour Mbondjangaa much newer and busier village which grew out of the Tchamba settlement when a new road was built linking Sonkolong to Nyamboya after the road between Sonkolong and Bankim was cut by the waters of the Mapé river dam in 1987. Like many "crossroads towns," Carrefour Mbondjanga is heterogeneous both ethnically and linguistically. To the north, the village territory rises up the Adamawa escarpment bordering onto the Kwanja village of Mayo Djinga, linked by a path. To the west, villagers can also walk 12 km to Somié via bush track, a two hour walk in the dry season.

Socio-economic summary
Mbondjanga, like many villages of the region, is largely monostratal from a socio-economic standpoint, though multi-ethnic and multilingual. Most inhabitants are peasant farmers; both men and women cultivating cash crops especially maize and coffee. Other forms of income include trade (there are three shops in the village), the small-scale sale of farmed and crafted goods at local markets, and small-scale husbandry. Like the other locations discussed here, the only electricity services are personally owned; water is gathered from the territory's rivers and streams in the absence of a village well. In 2010, the village had no mobile phone reception and most inhabitants had never had access to the Internet. The village has a primary school as well as a church, but no mosque or medical centre.
There is relatively low emigration from the region. Until the opening of the secondary school in Somié, young male Kwanja sometimes left for the closest towns and cities (Bankim, Banyo and, further afield, Ngaoundéré) for secondary education; a few adults (mainly men) also go to these towns and to the cities in the southe.g. Yaoundé and Doualafor work. Workers in these situations return regularly to the Kwanja region, if not permanently. Similarly, few people leave the village of Mbondjanga on a day-to-day basis except to attend local markets. This can be done easily on foot (to Somié, Carrefour Mbondjanga, or to Mayo Darlé, a five hour walk up over the hills). Other reasons why people might (less regularly) leave the village are as follows: young women marrying outside of the village who move to live with their husbands;⁴¹ visiting relatives in other villages (for example if the relative is ill); or to participate in celebrations or festivities in other villages. Villagers who have family ties outside the village can be obliged to participate in such events even when they are not easily able to afford the travel costs. One may also have to leave the village to visit a doctor, traditional healer, hospital, or the authorities at Bankim, for example to get a licence for a shop, or if one has been accused of a crime and been called to a hearing either at the police or the subdivision's offices (the "sous-prefecture").

Historical summary
Gausset (2010) provides a general picture of Kwanja history within which Mbondjanga is situated. Over the last 100 years or more, they have moved south in response to the arrival of the Fulɓe, specifically the  41 Patrilocal marriage is the regional norm: with few exceptions it holds for all four of our cases studies. establishment of the Lamidate⁴² of Banyo. While they used to live on the Adamawa Plateau, Kwanja are now found both on the Plateau and on the Tikar Plain. Mbondjanga exemplifies this: there have been three established village sites over the last 100 years or more. The first village was on the mountain of Dontap which overlooks the current village and was conceivably established as the Njanga sought refuge from the invading Fulɓe; the second was in the forested area, 1-2 km behind the current site; the current location is the third.
The oral history of Mbondjanga village and the Njanga language reaches back to 1890 and the reign of Wanjang. There are various conflicting stories about the origin of Wanjang's power over the territory of Mbondjanga. However, a plausible sketch⁴³ describes Wanjang as a leader of a small group of people living on Dontap mountain. Responding to the threat of war, Wanjang's father took him to be trained in a neighbouring region. On Wanjang's return, he fought a number of battles, including a battle for control of territory against two other Kwanja groups (known by the names of their leaders: Nyamakper and Nyambih). Having beaten his rivals, Wanjang took control of the territory but the rivals remained, leaving a legacy of tensions that underlie conflicts in the village to the present time. The main families of the village consider themselves descendants either of Wanjang or of Nyambih.

Linguistic overview
The Kwanja language [knp] has two main dialects, Ndung and Sundani, which are mutually intelligible, with Ndung the larger of the two. A third, much smaller language, also considered to be a variety of Kwanja (see Connell 2017 for discussion), is Njanga, spoken in 2010 by between just four and ten elderly people, and is considered by the local population to be the original language of Mbondjanga. Cambap ([twn] aka Twendi) is sometimes referred to as a dialect of Kwanja (e.g. Eberhard et al. 2020), though this is mistaken; Cambap is more appropriately classified as part of the East Mambila cluster (Connell 2001(Connell , 2002a(Connell , b, 2021. Kwanja villages are locally grouped by whether they are Sundani or Ndung speaking, or mixed-dialect villages (as many as fourteen⁴⁴ can be considered mixed-dialect). Six villages are considered Sundanispeaking villages by the Kwanja, and seven Ndung-speaking. In some cases, the reason for this perception is due to historical politicse.g. an Ndung-speaking chief founded the villagewhereas in others the development has been more organic, e.g. a larger number of Ndung speakers than Sundani speakers inhabit the area. Language choice per village is not uniform; e.g. a person may live in Yimbere (Sundani-speaking), but prefer to speak Ndung; however, more often than not, a person will speak both dialects. This is a result of their shared history and years of language contact between Ndung and Sundani speakers including substantial intermarriage between the two groups. It is common to find Kwanja couples in which one speaks Ndung and the other Sundani and each uses their own lect to address their children. Consequently, many people are bilingual in both. In these environments, language contact is mundane and taken for granted.

Speech communities
Mbondjanga residents speak the two main Kwanja lects: Sundani and Ndung. Mbondjanga can, however, be considered a "Sundani-speaking village" to the extent that it is the preferred lect of a majority (61%) of  42 The seat of the regional ruler; the Lamidate of Banyo is subservient to the Sultan of Ngaoundéré. The Lamido of Banyo recognizes the suzerainity of the Sultan of Ngaoundéré. 43 Discussed in more detail in Robson (2011). people living in the village and is considered a marker of the village identity by both inhabitants and outsidersi.e. both residents of Mbondjanga and other Kwanja call Mbondjanga "a Sundani village." In addition to the central village, the village territory comprises six additional neighbourhoods or "quartiers" of various ethnicity (Kwanja, Mambila, Yamba, and Fulɓe). The wider Mbondjanga village, incorporating all of the chief's territory, is a highly multilingual area. A language census carried out by Connell in 2001 (aspects of which are summarised in Connell 2009Connell , 2015 involving a door-to-door household survey and a classroom survey counted a total of 11 languages spoken, not distinguishing different varieties of Mambila or Yamba. All people, including children, in Mbondjanga speak Fulfulde to some level, many have some knowledge of French, and others have knowledge of Mambila or Cambap or other neighbouring and regional languages. Use of Yamba and Mambila by native Yamba and Mambila is in general restricted to the quartiers and households where these speakers live, and Fulfulde, French, and occasionally Pidgin are used for communication between these groups at the wider village level. Njanga is critically endangered. In 2010, there were four to ten speakers of Njanga, ranging in age from mid 40 s to late 60 s. The two best speakers and the only fluent speakers of Njanga were the chief and his elder uncle (now deceased). In addition to these speakers were two semi-speakers, both reasonably competent in Njanga and of the same generation as the chief, although around 15 years younger than him. One of the semi-speakers was the eldest son of the chief's advisor and the other, the premier notable of the chief. He is also related to the other three speakers via his mother, who is a descendent of Wanjang. The remaining five speakers, essentially rememberers, are a paternal and maternal aunt to the chief, the older sister of the chief, another cousin of the chief, and another distantly related relative of the chief who considers himself not a descendent of Wanjang, but rather of Wanjang's "rival" Nyambih.

Speakers and demography
Mbondjanga is a much smaller village than Somié (in both population and territory). It is the traditional village of the Njanga peoplea small ethnic group, mostly descendent from the chief Wanjang who ruled from c. 1880 to 1930. The main families of Mbondjanga (including those identifying as "Njanga" and other well-established families identifying with other descent lines) inhabit the central neighbourhood of the village. In 2008, the central village population was estimated at approximately 300 (Robson 2011) and while the population has likely grown since then, there have been no major developments in the area (e.g. new roads, new services, changes in power) that would precipitate a major population growth.
Socio-historically, the residents of Mbondjanga village have developed familial and marital and social and linguistic connections to the people of each of the territories it borders. For example, many Njanga residents have married Bà-Mambila of Somié and some Mbondjanga residents understand and/or speak Bà Mambila and use it when they travel to the weekly market at Somié. However, despite this proximity and the interconnections, Mbondjanga comprises its own language ecology as its inhabitants principally speak a variety of Kwanja and largely identify themselves as being of Kwanja ethnicity.

Language-identity relationships
Njanga speakers typically have in common one or more of the following characteristics: patrilineal descent from Wanjang, involvement or interest in village politics and (political) tradition, age, and male gender. Robson (2011) concludes that over the course of the twentieth century, as power in Mbondjanga village became polarised, Njanga became associated with the specific lineage of Wanjangthe descent group of the chief. "Ordinary" members of the villagei.e. those with no possibility to become chiefshied away from a lect so exclusively associated with a particular identity. Only men in the village eligible to become chief saw value in maintaining the lect as an authentication of their status as descendent of Wanjang and validation of their suitability for chief.
On this account, Njanga became moribund because of an increasingly limited functionality, which, within the multilingual ecology of the region, was not viable. Njanga's small population, its restriction to only one village together with its specific association with Njanga patrilineal descent, meant that the lect, once it came into closer contact with other varieties (viz, Sundani and Ndung), became functionally reduced and thus its use and subsequently its transmission went into decline. Njanga is today primarily a symbol of a particular kin-group identity, that of the Wanjang line, only maintained by those who have a (micro-political) interest in maintaining it (see the general discussion of ethnic and other identities in Section 2.2).

Functional domains
In present-day Mbondjanga, four languages, Sundani, Ndung, Fulfulde, and French, are used, sometimes in the same domain, and sometimes including codeswitching. Only institutional domains such as meetings with staff from the national administration (including the school), or meetings with regional traditional authorities, are limited to French and/or Fulfulde. Kwanja is the principal daily language of the village; it is used in church (in addition to French and Fulfulde), as well as at traditional and village meetings; it is used with other Kwanja elsewhere, for example at the nearby weekly markets at Carrefour Mbdonjanga and Somié. Kwanja is in some cases also used at Somié between Kwanja and Kwanja-speaking Mambila. Further, while Kwanja is the main language used in the household, if non-Kwanja friends are present, speakers will switch to Fulfulde or another common language; that is, speakers will typically strive to accommodate those present.
Use of Njanga is now highly limited in terms of domains, and the number of speakers very small. Indeed, loss of speakers and shrinking domains are self-reinforcing. Today, Njanga has no function (daily or ceremonial) which is not met by other lects spoken in the village (notably Sundani and Ndung). At the time of Robson's research (2009), natural use of Njanga was limited to conversations between the chief of the village and his advisor, and sometimes they also used it with the premier notable. It is perceived as a male preserve; although Robson (2011) was able to document women's knowledge of the language, many were reluctant to admit to any knowledge of it. Other speakers only used it in artificial settings created by her investigation into Njanga.
Robson (2011)'s reconstruction of Njanga's sociolinguistic history estimates that Njanga ceased to function as a fully transmitted quotidian lect around the early 1940s. This was around the time that village moved from its second site to its present location, bringing them into greater contact with other groups and closer to areas outside of the village, an event that changed the dynamics of the village and has plausibly contributed to the demise of the language (see Robson 2011). Prior to this, Njanga was spoken at the village level and used in several domains (household, the market, chiefly meetings). Since then, the function of Njanga shifted from the practical to the symbolic: its primaryperhaps even solefunction now is to mark Njanga patrilineal identity in opposition to other identities in the village. Whereas in the past Njanga may have also been spoken by different groups within Mbondjanga, it is now only used by those who see themselves as representatives of the Njanga descent groupi.e. the descendants of Wanjang.

Internal variation, regional
Little research has been conducted into variation in Kwanja. As mentioned earlier, the three varieties of Kwanja, Sundani, Ndung, and Njanga,⁴⁵ are distributed unevenly across the Kwanja speaking region:  45 Connell (2012/2017) argues Njanga is a variety of Kwanja as a result of convergence, in parallel to Gausset's (2010) argument that Kwanja as an ethnic identity came about through convergence. This is not to suggest that Njanga is itself a mixed language arising through convergence similar to e.g. Ma'á/Mbugu (Mous 2001(Mous , 2003b(Mous , 2013. Sundani and Ndung are spoken on both the Tikar Plain and the Adamawa Plateau, with Ndung speakers being predominant in both areas. For both varieties, there is some evidence to suggest that there may be variation according to age and the distance of the village from the main road with the more "isolated villages" (typically those on the Adamawa Plateau) being more conservative. There is also variation within Kwanja between the variety spoken on the Adamawa Plateau and the variety spoken on the Tikar Plain. Such variation comprises differences between grammar and lexicon; Robson (2011) provides full discussion.
Variation internal to Njanga may be due to the loss of functionality, in that it is rarely used, and although the majority of Njanga speakers today learnt the lect as a first language, this occurred just as the language went into decline so they have lost full competency in the language. These possibilities aside, impressionistically we can say that Njanga is subject to a substantial degree of influence from Sundani, the latter now being their primary language; Njanga speakers are uncertain of the differences that exist between the two. So, Njanga speakers borrow from Sundani or produce forms which are partially assimilated to or influenced by Sundani to the extent that "hypercorrection" takes place on an individual level; see Robson (2011) for details. It is also possible, indeed likely, that Njanga always demonstrated a certain amount of cross-speaker variation, as it was always a small language and never standardised (see Connell 2002a, b;Good et al. 2011).⁴⁶

Internal variation, social
Given the monostratal nature of Kwanja/Njanga society, as discussed above in Section 5.2, it is not surprising that little has been found with respect to socially determined variation in Mbondjanga, just as was the case for Somié, reported above. We note, however, that preliminary work by Robson suggests the possibility of age-related variation: that more conservative language may be used among older people living in more remote villages. This is an area for further investigation.

Status/solidarity
No particular prestige is accorded to different Kwanja lects, or to Fulfulde from the perspective of Kwanja speakers; the Fulɓe may well see things differently. Language choice use is closely tied to domain, though this is overridden by concerns for accommodation. On the other hand at least among the young and educated, French appears to be growing in prestige. Njanga, as discussed above, has to a large extent become emblematic of the Wanjang lineage; to speak Njanga is indicative of being of this clan; the reverse, however, is not necessarily true, as not all of Wanjang's descendants in Mbondjanga speak Njanga. Still, its use serves as a marker of solidarity among those who are able to speak it.

Written tradition, standardisation, institutional support
As with most other languages of the region, Njanga has never received institutional or governmental support; there was never a standard variety, nor was it ever given an orthography. Sundani has been given  46 Connell (2002a, b) discusses variation in nearby Cambap, challenging the view that the attrition process leads to increased variation. Variation exists at different levels in many languages that are small, spoken by relatively close knit communities and have never been subject to standardizationbut are not endangered. an orthography and been used in a translation of the New Testament, the work of Lutheran missionaries together with a local language committee.

Attitudes of users
Given its critically endangered state, and that many women speakers, at least, were reluctant to admit to any knowledge of it, it would be difficult to argue that a positive attitude exists toward Njanga. Robson (2011) argues that negative attitudes toward the lect, and what it represented, played some role in its decline. Conversely, a positive attitude towards the lect among the handful of remaining speakers has helped to maintain Njanga to the present time. At the time of Robson's research (2007Robson's research ( -2009, knowledge and awareness of Njanga varied among inhabitants of Mbondjanga and among Kwanja speakers in the wider region. Those who did not recognise the language tend to consider it to sound old-fashioned or complicated. Those who did recognise the lect as Njanga associated it with the chief and the lineage of Wanjang. More generally, given the multilingual nature of Mbondjanga and the wider Kwanja (or indeed Mambiloid) region, it is interesting to consider relations among the various languages from the perspective of who learns which languages. Kwanja people are highly likely to speak both Sundani and Ndung, as well as Fulfulde, French (less likely for women) and, possibly, Mambila, Tikar, Pidgin, or other Cameroonian languages. Perrin (1988: 1) suggested, "there is a clear pecking order for the language groups; Tikar first, then Mambila, then Kwanja. People are unlikely to learn the language of those considered inferior." While it is true that among the local languages some are more likely to be second languages than others, the "pecking order" claimed by Perrin is not borne out in our experience. Rather it is again considerations of exposure and accommodation that appear to determine which languages are learned as second languages; a Kwanja speaker who is regularly in contact with or living among Mambila or Tikar speakers will be likely to learn these languages; the same would be true of a Mambila or Tikar speaker who is regularly in contact with Kwanja speakers; he or she would learn Kwanja. It is only the Fulɓe who rarely learn the language(s) of others, this given the position of Fulfulde as the regional lingua franca, as discussed earlier (Section 4.5, fn. 34).

Association between language and economic/political advancement
As is clear from the foregoing discussion, none of the local languages are crucial to possible political or economic advancement except at the very local level. Both French and Fulfulde to some extent of course do contribute to this, as discussed in Section 4.12.

Mbondjanga summary
The language ecology of Mbondjanga is characterised by the presence of several languages and multilingualism on the part of its inhabitants. In Section 5.5 above, we outlined the complexity of the speech communities found in the village of Mbondjanga and its surrounding territory; these include, principally, the two Kwanja lects Sundani and Ndung, varieties of Mambila (mainly Bà Mambila given the proximity of Mbondjanga to Somié, but also some Nigerian lects), and several different varieties of Yamba in addition to Njanga and the severely threatened Cambap. In total, 11 different languages were reported being used at Mbondjanga. As is common to the region, contact with Fulfulde is strong. From the perspective of the few surviving Njanga speakers, Sundani especially, but also to some extent Ndung, can be seen as more prominent, given its status as the language of the home and primary daily use. If on the other hand one were to take the perspective of a Sundani speaker in Mbondjanga, Njanga would constitute only a minor source of contact, Fulfulde being the major source. Robson and Griffiths (2010: 232) report the influence of Fulfulde on Kwanja and Wawa numeral systems, colour terms, and the days of the week. In the case of the latter, a young female Kwanja speaker is cited as giving a mix of Fulfulde, Pidgin, and Kwanja, in which only two days had Kwanja names. An older informant offered Kwanja names for all 7 days, but these were disputed by other elders. No distinctly Njanga day names could be elicited. The older, traditional, 10-day week, in which days reflected the rotation of markets, had already been replaced by a westernised 7 day week. Robson (2011) discusses the of influence of Sundani and Ndung (as well as Fulfulde) on Njanga citing apparent borrowings at different levels: lexical, phonological, and morphological, though at the same time being careful to point out the difficulty in distinguishing what is borrowed from what is inherited. As an example, in comparing a corpus of 1,628 words elicited for Njanga and Sundani, 76.3% were found to be cognate across the two languages, though it was indeterminate how many of these would be actual borrowings. Similarly, at the morphological level, Robson gives several instances of possible cognates between Njanga and Sundani, some of which may result from contact, other may be retentions from a common parent. Space does not permit a review of all of these, but one example is the definitive marker,/ -ra/in Njanga,/-re/in Sundani, with the latter now also being used in Njanga.
Being the primary village language for daily life, Kwanja is in some respects dominant but this is true only in a restricted sense, both geographically and in functional domains. Unlike Somié, where the traditional village language remains the primary language, Njanga, the traditional language of Mbondjanga, is now critically endangered. It is no longer regularly used in any functional domain and now appears to serve primarily as emblematic of the traditional ruling clan of the village. The decline of Njanga is summarised by the following historical hypothesis.
Mbondjanga village moved its location in the 1940s. This produced greater contact with other Kwanja speakers, the eventual result being that the village lost political autonomy. The village chiefship and its associated language went into a decline which has now more-or-less terminated. There is still a chief, but for at least the past 20 years he has no longer been the influential player he presumably was at an earlier time, i.e. before the late 1930s. By the end of the 1950s, Njanga had declined to the extent that it was no longer the language daily used at Mbondjanga, but had given way to Sundani Kwanja. It remained the emblematic language of the village at the time of Robson's research.
By moving, they came into the ambit of the village Yimbere (a Sundani-speaking village) and thus became loosely allied to them, as well as villages on the Adamawa Plateau, through both political and kinship alliances. Since the creation of the new road in 1987 (see Section 5.1), they have also been in closer contact with the village of Nyamboya and Ndung speakers. Hence, most people in Mbondjanga village once spoke Njanga, but now they mainly speak Sundani (see Gausett 2010, Robson 2011. However, on our view of the Kwanja language ecology, this doesn't fit exactly with the usual argument of "X language replaces Y language because the speakers of X language are more politically dominant." Njanga died because its speakers started trying to be singulari.e. tied to one identity, one exclusive group of people (Sundani speakers). Njanga speakers (people who would have been Njanga speakers) came to select other languages from their repertoires. Mbondjanga, being historically affiliated and politically allied to Sundani-speaking villages, e.g. Yimbere, began to identify (as it does still today) as a Sundani-speaking village so Njanga was dropped.
6 Oumyari village and the Wawa language 6.1 General description of physical geography Oumyari is situated some 15 km west of Banyo, the administrative headquarters of Mayo-Banyo Division in Adamawa Region. It is one of nine villages and several smaller settlements where Wawa is spoken, all of which are in the region between Banyo and the Nigerian frontier, ranging from the southwest of Banyo to its northwest. The village and the larger Wawa region are situated on the Adamawa Plateau, where vegetation is sparser than is found on the Tikar Plain, described in the preceding sections. Except for within the town of Banyo, and certain stretches of the N-8 trunk road, all the roads in the area are untarred, including the roads linking Oumyari with Banyo and with other Wawa villages.

Socio-economic summary
The nearest administrative centre is the town of Banyo, where there is a daily market, and a weekly cattle market. Until the mid-1980s, Oumyari Wawa used a traditional 10-day week (see Griffiths and Robson 2010). When this was still in use, a local market was held once every 10 days in the Oumyari quartier of Dame. The main weekly market is now held every Friday in Banyo. The village of Mbenguedje maintains a market on the traditional 10-day cycle and a small market is held every Friday in the village of Ndi (see Figure 4 for locations).
Like the Mambila and Kwanja described above, the Wawa are primarily peasant farmers with the community being mostly self-sufficient in food. Maize is the staple crop, supplemented by other crops such as groundnuts, chilli pepper (Capsicum frutescens), and beans, and some limited livestock husbandry (cattle, goats, poultry). Other occupations exist (trading, bee-keeping, etc.), though everyone farms to some extent. Many men also gain an income from cutting firewood in the savannah, which is then sold in Banyo. A few men also work in town on market days as motorbike taxi drivers.
Although its location is relatively remote, Oumyari is not as isolated as this might imply. Banyo is approximately 30 minutes by road, so many people visit on a regular basis (e.g. a truck used to come on a weekly basis to bring people to and from the Banyo market, though this service has stopped and has been replaced by motorcycle taxis). Migration in and out of Oumyari is another factor that brings Oumyari and its inhabitants in contact with the wider region. There are a number of non-Wawa women who have moved to the village having married Wawa men. These women are mainly Vute or Mambila, though some have a mixed ethnic background. Although their first language is not Wawa, they speak it well and use it on a daily basis. Wawa women from Oumyari also marry men from other villages in Cameroon and Nigeria. This entails the women changing residence and leaving Oumyari (and adopting a new primary language). Many young men seek their fortune in town (Banyo or other cities) to save sufficient money to marry. Indeed, many young men are no longer satisfied with being self-sufficient peasants and want to find work, but there are only limited employment opportunities in Oumyari or even Banyo and so they migrate further away. Impressionistically, such dissatisfaction seems to be on the increase, and a decline in fertility of the soil in the area is one reason given for this increase in out-migration. It has become increasingly difficult to grow enough maize, to support and feed one's family, harvest-to-harvest.
The usual form of motor transport in the Wawa villages is motorcycle taxi, but most people walk. As money is in short supply, the people using motorcycles are most commonly men. Most Wawa villages are within a day's walking distance from each other. In the past, Oumyari village was visited once a week by a truck, picking up people, their merchandise/produce, and firewood to take to town for the Friday market; however, this was suspended in 2012. Since then the number of motorcycles in the village has risen to approximately ten by 2014.
Three Wawa villages, Oumyari, Gandoua, and Ndi, are within a day's walking distance from each other and they are relatively close to a major trading route connecting Banyo to Nigeria going through Boudjounkoura at the border. The Wawa villages are situated along smaller tracks between 3 and 10 km from this road. The tracks to Oumyari and Gandoua are in bad shape and only suitable for motorbikes, not cars. The village of Ndi has a regional dispensary and the road to it is in a much better state. The fourth Wawa village, Mbenguedje, is somewhat further away, to the south and can be reached by motorcycle from a junction near Allat, on the main road between Banyo and Mayo Darlé.
Oumyari village does not have tap-borne water, though there are mechanical water pumps in two quartiers; similarly, there is no grid-based electricity, though some individuals own petrol-powered generators. There is a primary school just outside the main quartier of the village, with the nearest secondary school being located in Banyo. The language of instruction is French (as per government policy), and only people who have attended school speak French to any extent. Teachers are typically from elsewhere in Cameroon and in our experience tend to look down on the local language and culture. Most families cannot afford to send all their children to school. Generally speaking, parents prefer to send boys to school rather than girls, as girls are expected to help more in the household and in the fields. This leads to many more males than females being able to speak French. However, this seems to be changing, and many parents are proud of their daughters going to school. During the dry season, the classrooms are full, but during rainy season only a few children attend classes as they have their own fields to cultivate and are expected to help their parents farming.

Historical summary
Little is known in detail of Wawa history before the arrival of the Fulɓe; for example, how and when the various villages were established. It seems that they are descendants of one group of the original (Mambiloid or pre-Mambiloid) population of the area (see Figures 3 and 4 and associated discussion). The Fulɓe arrived bringing jihad approximately 150 years ago. Whereas other groups fled and/or were scattered, the Wawa tended to be more willing to accommodate the invaders and accept their dominance (Gausset 1999(Gausset , 2010. Exceptions may be the existence of a Wawa village on the Mambila Plateau in Nigeria and those families at Hore Taram Torbi, mentioned below in Section 7.3; oral tradition suggests they fled the Fulɓe. According to some oral traditions, when the Fulɓe arrived in the Banyo area in the nineteenth century, the Wawa avoided being enslaved by helping them conquer Banyo (Gausset 1999(Gausset , 2010. The Fulɓe are still regarded as foreigners and strangers by older Wawa; however, they are the most relevant non-Wawa group for the Wawa population. Despite their early cooperation with the Fulɓe, it was only in the 1950s and 1960s that the Wawa generally converted to Islam. Consequently, while the Wawa have had language contact with the Fulɓe for well over a century, this contact has intensified in the 50 years since Cameroon's independence in 1960 both as a result of their conversion and because the country's first President (a Muslim) brought Islam and therefore Fulfulde to greater prominence. Linguistically, the result has been considerable influence on Wawa from Fulfulde, in both the domains of lexicon and grammar (Martin 2012), though this influence is variable across lects. The findings presented in Thomae et al. (2013) suggest that the relations between the Wawa and the Fulɓe have improved through prolonged contact. For more detailed information on the historical aspects of relations between the Wawa and the Fulɓe, see Gausset (1998Gausset ( , 1999.

Linguistic overview
Like the other languages discussed in this article, Wawa is a Mambiloid language. It is among the most morphologically complex of Mambiloid languages, with complex nominal and verbal inflection and derivation. There is no standardised form of Wawa, though some Wawa people say the Gandoua dialect is the "true" Wawa. The dialect of Oumyari appears to have undergone considerable loss morphologically and syntactically, and many grammatical constructions are considered "heavy" (i.e. old or archaic) even by older speakers; for instance, the passive mood is not used anymore in everyday speech, and the structure of the verb phrase has been affected by contact with Fulfulde (Fulfulde prepositions and adverbs have replaced Wawa verbal extensions). However, the Fulfulde spoken by Wawa speakers displays some Wawa influences, for example, the replacing of implosive stops with geminate oral stops and the introduction of tonal contrasts (Martin 2012).
Wawa is spoken primarily in the four villages mentioned above and smaller villages or hamlets within their orbit, all to the west of Banyo, in an area bounded by the national road (to Tibati, carrying many trucks) to the east and the escarpment of the Mambila Plateau to the west. Linguistically, it is bordered to the east by Vute (different varieties), to the south by Kwanja and Mambila, and to the west by Mambila, again in different varieties. As stated above, in the nineteenth century, the Fulɓe conquered a large area which is today part of northern Cameroon and north-eastern Nigeria. Hence, in the Banyo area Fulfulde used to be the only administrative language until the Fulɓe lamidates became part of European colonies at the end of the nineteenth century (Mbaku 2005: 36); it was partially replaced by German and officially, for much of the past century, by French. However, many of the Fulɓe structures are still in place and the use of Fulfulde with them. The Fulɓe still have great influence in parallel to the power exercised by the Cameroonian state. Most important decisions which affect the community in Oumyari will be discussed with or in the presence of a Fulɓe Mallam (a religious leader in the mosque). Therefore, these discussions will be held in Fulfulde. However, this only affects the men. Children and women do not participate in such discussions.
All Wawa in Oumyari are multilingual. Even small children speak Fulfulde to some degree, along with Wawa. Many people also know some Vute, due to intermarriage between the villages of Oumyari and Ndi, which has a large Vute population, as well as with other villages in the area (many of which speak Vute as their main language). Many men also know some Hausa and some Pidgin.⁴⁷ Immigrant spouses use their native Vute and Mambila when conversing with members of their own group. Non-knowledge of the local language among the town-Wawa is not perceived as negative by other Wawa; it is merely accepted as fact and does not determine their ethnic identity.
Within Wawa, four varieties are usually identified (Griffiths 2010;Martin 2012;Starr 1989), each named for the main village in which they are spoken; from south to north, these are: Mbenguedje (see Todd 2011), Oumyari, Ndi, and Gandoua. Other Wawa villages include Yabam, Mbassewa, Kassala, Gaoula, and Dembesse. People from these latter five villages usually identify themselves as using the Oumyari variety of Wawa. Wawa is also spoken by a few families at Hore Toram Torbi (see Section 7.3, below) and a Wawa village reportedly exists on the Mambila Plateau in Nigeria. Mutual intelligibility across Wawa dialects is limited, especially between Gandoua on one hand and Oumyari and Mbenguedje on the other, and it is more usual for speakers of different dialects to use Fulfulde when speaking together; it is described as being easier, that it takes more effort to understand the other person's dialect, than to switch to use Fulfulde. A similar phenomenon occurs among the Wawa community living in Banyo who in 2010 created an organisation to promote their language and culture: the Association des Wawa du Cameroun (AWACAM). Given the lack of mutual intelligibility among Wawa dialects, and that most ethnic Wawa living in Banyo have only limited command of the language, the language of communication at AWACAM meetings, at least initially, was Fulfulde.
Early studies of the languages of the area classified Wawa as a dialect of Vute. Wawa and Vute bear similarities to one another, and studies by Guarisma (1987) and Starr (1989) compared Wawa to the Vute dialect of Banyo, a dialect geographically close to Wawa and so the possibility of their similarities being due to contact cannot be ruled out. However, studies using the comparative method (Connell 2010b(Connell , 2021) have thus far not revealed any shared innovations that would be diagnostic of a Wawa-Vute grouping. So, the Wawa are here treated as a separate linguistic and ethnic group, a view that is generally shared by the Wawa themselves. Nonetheless, Vute and Wawa people recognise the cultural and linguistic similarities shared by the two groups. Generally, the much more numerous Vute consider the Wawa to be an ethnic subgroup, and their languages to be a dialect of Vute. Griffiths and Robson (2010) looked at the influences of the highly multilingual environment in which the Wawa people live. Martin (2012) and Griffiths (2010) separately have considered different aspects of the Oumyari Wawa language system.  47 Given mixed influences from Nigeria and Cameroon, without a detailed study it is not possible to say more about this variety of Pidgin.
In Krauss' (2007) framework, the endangerment status of Oumyari Wawa is "instable/eroded;" the language is still transmitted to and spoken by children on a daily basis; However, the younger generations mix it heavily with Fulfulde and codeswitching is common.

Speakers and demography
Oumyari village comprises several quartiers or hamlets, which are associated with different Wawa kin groups or other ethnolinguistic groups, including Fulɓe, Vute, Mambila, and Yamba. Some of these are more or less contiguous with central Oumyari and others are as far as 4 km from the village centre. A detailed discussion of this village structure is beyond the scope of the present paper; however, it does have striking similarities to the Nizaa situation described in Endresen (1999). The village centre has a population of approximately 700 inhabitants, while population estimates for the Wawa region range from approximately 3,000 to 5,000 speakers.

Language-identity relationships
Some Wawa people, especially those living in Banyo, are aware of the low status of Wawa in the eyes of the Fulɓe. They also are aware that their children are losing touch with their cultural history by not speaking the language. They recognise that this is their responsibility, but are also conscious that Fulfulde is all that is needed in town. As we noted above, especially for those in Banyo, the language is not seen as necessary for ascription or maintenance of Wawa ethnic identity. How long this will persist is a matter of considerable uncertainty.

Functional domains
In considering the functional domains comprising the language ecology of Oumyari, in addition to Wawa, a number of other languages need to be taken into account. Fulfulde is of course the lingua franca of the region, though we note it does not have official status in Cameroon; French has official status and so both of these fill roles that are not open to Wawa. Other languages which might be considered to have greater relevance in certain domains relative to Wawa are Arabic (as a vehicular language of religion) and Hausa (a vehicular language used as a trade language and for radio broadcasts both spoken and sung⁴⁸ in the area), as well as Vute and Mambila which both have larger populations of speakers. However, we concentrate on Fulfulde in the following, as this is the most prominent second language within the language ecology of Wawa. Of course, language choice is not only dependent on the domain, but also on the languages available to people present in a given situation. That is, Wawa could be spoken in most, if not all, domains, as long as the participants are all Wawa speakers. That Wawa is not spoken in the mosque is due to the presence of Fulɓe men coming for the prayers; for the same reason, it is not spoken in school meetings.
Home. Wawa is spoken at home and passed on to all children. Some Wawa who live in towns such as Banyo send their children to their home village for prolonged periods of time so that they can learn the language from their peers and grandparents. Therefore, the Wawa language predominates in the domestic domain. Even when a woman is not a Wawa speaker before marriage, the language of the household will be  48 We note that some broadcast songs were in Fulfulde varieties from Niger which some people in Oumyari thought to be Hausa because they couldn't understand it.
Wawa and she will normally learn it. Wawa is not, however, the only languages used in the home; Wawa women these days spend a considerable amount of time teaching their children Arabic and how to read the Quran. Therefore, in mother-child interaction from a certain age onwards, there will be a degree of interaction which involves Arabic. Use of Arabic has entirely to do with Islam and the Quran and is not used in non-religious contexts; although further study is needed, it does not appear to be an extension of Shuwa Arabic spoken further north.
In the village: Within the village, the public sphere is dominated by use of Wawa and Fulfulde. As long as there are no Fulɓe or other non-Wawa present, Wawa is spoken in public among ethnic Wawa. Three different sub-domains in Oumyari illustrate the nature of language choice: (1) There are two open meeting places on the main square: a big mango tree at the entrance to the village under which older men like to sit and chat, and a bench in the middle of the village, under a smaller tree, where younger men gather. In these places, Wawa is the main medium of communication, unless a non-Wawa speaker is present. When Fulɓe men stop by to chat or trade, the language of communication changes to Fulfulde (with speakers of other languages, a common vehicular language is chosen, most commonly, but not always, Fulfulde).
(2) The language used in religious gatherings is always Fulfulde, as there are always some Fulɓe men present, e.g. the local Imam is Fulɓe. All prayers are of course recited in Arabic, but the language of communication is Fulfulde. We have already mentioned domestic religious training which mixes Wawa and Arabic.
(3) In school meetings, the languages of communication are French and Fulfulde. This is discussed in greater detail below.
More generally, even though Wawa is the primary language of Oumyari, children and young adults mix Fulfulde words when they are speaking Wawa. For example, when counting, only the numbers one to five are in Wawa; for any higher number, the Fulfulde words are used (except "ten," which is borrowed from Vute; see Griffiths and Robson (2010) for details. Children are often unaware that they are mixing Wawa and Fulfulde. It happens occasionally among children that a story is begun in Wawa and ended in Fulfulde, without their noticing. Older Wawa also make regular use of Fulfulde and codeswitch more consciously. Arabic and Fulfulde greetings are used much more frequently than traditional Wawa greetings due to the pressures of Fulɓeisation (see below) and Islamic traditions.
Market. There is no village market in Oumyari and people travel to Banyo to the large general market on Friday. There, Wawa speakers mostly speak Fulfulde while doing their trading, but as in all other scenarios described above, they also use other languages depending on the person they are talking to. There are a few Wawa traders at the Banyo market, with whom Wawa speakers from Oumyari use Wawa. Some Wawa speakers also use Vute with traders who have come to Banyo from one of the Vute-speaking villages. Traders who originate from the West or Northwest Regions generally speak Fulfulde and use that with Wawa people, unless the Wawa person has enough French or Pidgin to speak those instead. At the market in Ndi village, Wawa speakers use Wawa when speaking with other Wawa from their village. However, some Wawa speakers may choose to speak Fulfulde to Wawa coming from villages with whose dialect they are less familiar. They speak Fulfulde with Fulɓe, and possibly Hausa, with Hausa speakers. They may even use some Pidgin to merchants from Nigeria or anglophone Cameroon.
Education. The official language of education in the francophone part of Cameroon is French, as mentioned earlier, and this is used in the school at Oumyari. At meetings with parents, the head teacher, who anecdotally is fluent in Fulfulde, speaks only French to the assembled fathers (only men attend). This is then translated into Fulfulde by another French speaker. Wawa is not used at all in these meetings. Only people who have attended school speak some French. As school attendance is not a great priority for girls, this leads to many more males being able to speak French than females. In Oumyari village, all men and women study the Quran and are able to read Arabic script. Children are sent to the unofficial Quranic school in the village and we have also witnessed women teaching their children how to read and write Arabic.
Religion. The dominant religion on the Adamawa Plateau is Islam, introduced by the Fulɓe, as discussed earlier, and all Wawa are Muslim. This means Arabic is an integral part of the religious life of the Wawa, as well as the Fulɓe. (Outside this domain Arabic is only used in greetings. These are more formal than other greeting forms.) As a result, virtually everyone, even girls, can read and write Arabic script to the extent it is needed for religious purposes. Both Arabic and Fulfulde are used in the mosques. Wawa is not used in the mosque, as there are always non-Wawa present which results in a language switch to Fulfulde. Gausset (1999: 11) claims that Islam in Adamawa poses a problem of identity, since it is part of the pressure on people to "Fulɓeise," i.e. to reject traditional rituals and identity to gain access to a respected identity (see Keen and Zeitlyn 2007, for discussion of Fulɓeisation in nearby Mayo Darlé). By and large, the Wawa have Fulɓeised; they have adopted a Muslim way of dress and when in town speak Fulfulde. They also have adopted Fulɓe rituals and rites of passage. For example, traditional naming practices have been lost in favour of Fulɓe/Muslim names. Nevertheless, Wawa have not given up their language and most still claim a Wawa identity.
Outside the village. The Wawa language is not much spoken outside the Wawa villages. Beyond the village borders, communication takes place predominately in Fulfulde unless all involved are Wawa; although as mentioned above, when together, speakers of Gandoua and Oumyari dialects would at times converse in Fulfulde. This affects the entire speech community as all Wawa people on occasion leave the village to travel to town for the market. This does not necessarily lead to a switch to speaking Fulfulde. With Vute speakers, some Wawa (those who can) switch to Vute rather than Fulfulde.
Town. The description of language use in town (Banyo) is in large measure the same as that given earlier for the market, since attending the market in Banyo is one of the principal reasons for visiting the town. The languages spoken by Wawa people when in town are Wawa, Fulfulde, Hausa, Pidgin, French, and Vute. Subject to issues of mutual intelligibility between Wawa varieties, Wawa speakers from the village use Wawa to speak to other Wawa people in town, if they know that the person is able to speak the language. In most cases, such a person will be a Wawa born in one of the villages who moved to Banyo. Otherwise, the most spoken language is Fulfulde; it is used between Wawa speakers and ethnic Wawa who do not speak the language, and as the default choice with non-Wawa. If a Wawa person speaks some Hausa, they may use it when communicating with Hausa merchants in Banyo. Pidgin, if known, is used when talking to Yamba people and merchants from the Northwest Region of Cameroon or from Nigeria. The broad pattern is the same as for use in markets, especially as "town" generally is held to equal "market," and the patterns of usage in market holds even for those Wawa who have moved to town.
Formal situations. Most formal communication (e.g. with the police, administrators, or doctors) takes place outside the village and, as already mentioned, in this case Fulfulde will be the language of choice. However, within the village, formal education is done in French. In the primary school, the children are taught in French from day one without their having any prior knowledge of French. Finally, a form of name avoidance practice exists between husband and wife, which might be considered "formal;" it is discussed below in Section 6.9, "social variation." Informal/casual situations. Discussion of the distinction between formal and informal situation follows from previous points regarding communication in the village. Given that there are very few formal situations to be observed, formal and informal situations are hard to contrast and compare. Most of the communication is informal. One could invoke aspects of the notion of a society of intimates (Givón 2002;Trudgill 2011).
Intimate situations. Wawa is the language of intimacy. Conversations of a private nature and those between friends and family are conducted in Wawa, when the speakers are Wawa. As has been said, women marrying into the village learn Wawa and use it to talk to their husbands.
Entertainment/Media. The Wawa people have Fulɓeised to a considerable extent; this has meant the abandonment of many traditions, and Fulfulde (and to some extent Arabic) has become the main code for certain traditional activities (see Gausset (1992 and for some of the traditions no longer practised). Often both Wawa and Fulfulde have to compete with French as a language of modernity. The Wawa prefer Fulɓe and Hausa pop music over Western style music, but Ivorian music is also very popular (which is often sung in French along with Ivorian languages). However, school, administration, mobile phones, and similar aspects of modern life are associated with French.
The four main Wawa dialects appear to be very similar in terms of phonology. Morphosyntactically, as mentioned earlier, they are among the most complex of the Mambiloid languages, though Oumyari may have lost some of this complexity relative to the other varieties. Within Oumyari, there is some further variation, as it is spoken across several villages. The differences are, however, confined to morphology and possibly syntax; the phonology appears to be largely consistent across the villages concerned.

Internal variation, social
It will already be apparent that in the absence of substantial social differentiation, there is little socially determined linguistic variation. In Oumyari, however, it is customary for married women to avoid the use of their husband's name in conversation, i.e. a form of avoidance practice is used. An avoidance name is used (e.g. Dídí or Dídí Àló) or the woman simply paraphrases the name as "the father of my son/daughter" (i.e. a form of teknonymy). This is not reciprocal: men do not avoid their wives' names.

Status/solidarity
The preference for Wawa to use their own language in the situations described above can to some extent be attributed to a desire to maintain their own identity, to show solidarity as Wawa. It has also been observed that even chiefs of other villages, e.g. speakers of Vute, will switch to Wawa when they are in a Wawa village. Sometimes speakers will adapt their language choice (e.g. switch from Fulfulde to Wawa) when discussing things they do not want outsiders to understand. A division of labour between French and Fulfulde exists in the area indexing power and solidarity. Officials, such as police, may sometimes switch to French to exert authority, despite usually speaking Fulfulde. Fulfulde will be used as the language of solidarity in this case.

Written tradition, standardisation, institutional support
Until recently, there had been no explicit effort to maintain or develop Wawa. Like the other languages discussed in this paper, it has no written tradition, is not standardized, nor has it received governmental support. The older generation is conscious, however, that the children mix more Fulfulde into their Wawa than they themselves do. As was mentioned above, the AWACAM organisation was formed to promote Wawa culture and language and prevent their loss. This may lead to some maintenance efforts, as the association recognises the endangerment of Wawa and wants to strengthen the language use among the younger speakers. There were plans to organise regular cultural festivals.
There is no written tradition for Wawa. If something needs to be kept in writing, Arabic script (Ajami) in some circumstances is preferred over the Latin alphabet to record the information and it is used to take notes in Fulfulde. Examples of such notes are items such as phone book inventories (including only names) and shopping lists usually written in Fulfulde.

Attitudes of users
The Wawa generally claim they have a positive attitude towards their ethnicity and language. To a great extent they have Fulɓeised, in a broad sense of the term. However, despite having converted to Islam, they do not claim Fulɓe ethnicity/ancestry and have not completely given up their language in favour of Fulfulde, although it is threatened. They use it as a marker to separate and to distinguish themselves from the Fulɓe or other ethnic groups when appropriate, and as we have said above, they use it as their home and community language. However, they accommodate non-Wawa speakers by shifting language during official discussions, using Fulfulde if there is even one non-Wawa person present, because Wawa is not generally acquired by non-ethnic Wawa people, except by women who marry Wawa men and the occasional anthropologist or linguist.

Association between language and economic/political advancement
There is a clear power relationship between Wawa and Fulfulde with Fulfulde being the dominant language related to social and political power. However, Fulfulde in turn is dominated by French in administrative contexts and Arabic in the religious domain. Both Fulfulde and Hausa are connected to economic affairs.

Summary: Language contact in Oumyari Wawa
The Wawa people live in a multilingual ecology. The greatest degree of language contact, and of linguistic influence on Oumyari Wawa, comes from Fulfulde, as intensive interaction with local Fulɓe people living in the village occurs on a daily basis. Language contact with Vute is also common among people who have family members living in Vute hamlets or the neighbouring village of Ndi, where there are many Vute residents. The villagers' contact with French is limited to interactions with the resident school teachers and to school meetings, rarely with non-Fulfulde-speaking traders in Banyo. Contact with Hausa, Mambila, and Pidgin speakers is rarer and the language of communication with speakers of these languages defaults to Fulfulde. The effects of such constant codeswitching can easily be recognised. On an individual level, the amount of Fulfulde lexicon used by Wawa speakers varies by age. Children and younger people (under 25) use more Fulfulde words than older people (over 50 years of age). On a morphosyntactic level, many Wawa verbal extensions are considered archaic by younger speakers, who prefer to use Fulfulde adverbials, interjections, and conjunctions, instead. One example is the habituative extension/-to/, which has been replaced in everyday speech with the Fulfulde adverbial tum "always." Thus (1) would commonly be expressed as in (2), or without any marking as in (3) In discourse, many structures have entered via Fulfulde, for example the topic markers kâ (general focus) and bó, the imperative weakener lē, the discourse markers to᷆ ("well, okay") and ndá ("so"), and the emphasis markers ôn and há (both "very much"); tò "if" is used to express conditional clauses.
Many narratives are similar in structure to Fulfulde narratives. All direct speech in narratives is sung and is either in Vute or Fulfuldenever in Wawa. While this is not a rare phenomenon across Africa, it shows how common codeswitching is. In everyday discourse, the authors noticed that children would sometimes start a story in Wawa and finish in Fulfulde.
The Wawa language is still a language of daily communication. There are few occasions or domains in which Fulfulde (or French) would be the language of choice. In case of Wawa and Fulfulde, the language choice is usually determined by the village border (Wawa within, Fulfulde without), though within specific functional domains, Fulfulde may win out. Regardless of setting or domain, Fulfulde will be the language of choice if there are non-Wawa speakers present. There is a clear power relationship between Wawa and Fulfulde with Fulfulde being the dominant language. However, in the wider ecology, Fulfulde is dominated by French in administrative contexts and Arabic in the religious domain.

General description
The village of Kila Yang is situated on the Mambila Plateau, near the town of Mayo Ndaga in Nigeria. This constitutes the only one of our case studies located outside of Cameroon, though as discussed below the speakers of Sombə find their origins in Cameroon. It is adjacent to the village of Kuma and is in some sources considered to be a part of, or enclaved by, Kuma. Hurault (1998) gives a detailed description of the geographical and climatic characteristics of the region. The Mambila Plateau is largely devoid of trees, with the exception of small forests found in valleys and along streams or riverbanks. The climate is relatively drier than that found below on the Tikar Plain described earlier, though the rainy season has approximately the same duration (April to November). Many smaller roads and tracks may become impassable at the height of the rains.
Sombə, or Somyev, was by 2011 spoken for certain by just two elderly men, though when we first encountered this language in the mid-1990s there were approximately 20 speakers. The language is variously known in the literature and locally as "Kila," "Kilayen," "Kila Yang," Somyev, or "Fur." The first two of these are the Fulfulde words for "blacksmith," singular and plural; "Kila Yang" is presumably an anglicisation of the Fulfulde plural form; "Somyev" is the Mambila word for this group and their language. "Fur" is their own word for "person" or "man." Blench (1993) also mentions the term "Somyewe;" however, this was not familiar to the speakers we worked with. Sombə is the name they use to refer to themselves and their language, and as we have elsewhere used this term rather than the somewhat better known "Somyev," we continue to use it in this report. Sombə is of interest not just for its endangered status, but also because of its unique linguistic and cultural status in the area: Sombə is the language of a blacksmith group. It is not a trade jargon or sociolect, but is (or rather, was) the primary language of daily use for the smiths and their families, both in the home and outside. This seems different from the other reports of blacksmiths in the wider region such as Van Beek's (1991 study of the Kapsiki smiths to the north of the Mambiloid region and Fowler's study of the Babungo to the south (1990,1995), or Kastenholz's (1998) study of smiths (and leatherworkers) among the Mande, much farther west, having their own caste-specific jargon. The traditional blacksmithing trade has now vanished, bringing substantial change to the lifestyle of this group. Sombə was first discussed in terms of its endangered status in Connell (1998), where it was suggested that a probable reason for its demise was the disappearance of the blacksmith trade and the arrival of the Fulɓe and their effect on the language ecology of the region. The effect on the specific language ecology of Sombə was that it was brought into contact with at least two other languages which had a major influence on its fortunes: Fulfulde, which of course became the lingua franca of the region, and Maberem, the variety of Mambila that has replaced Sombə. More detailed discussion of the demise of Sombə in relation to its language ecology is found in Connell (2010a).
The two speakers mentioned in the previous paragraph live in the village of Kila Yang, so our discussion focuses mainly on aspects of the language ecology of Kila Yang. Since there were also two semispeakers/rememberers in the Cameroon village of Hore Taram Torbi, where the people and their language were known as Zhuzhun. We also discuss the situation in this second village and the connection between the two villages, or sets of speakers.

Socio-economic summary
The population of Kila Yang, in 2007, was said to be about 3,000 (this is not an official figure, and may overstate the actual population). The inhabitants of Kila Yang comprise several ethnicities: principally Mambila, but Fulɓe, Hausa, and Yamba, as well as Sombə, are also there in numbers worth mentioning. This mix has a corresponding linguistic heterogeneity, with the exception of Sombə; one estimate suggests the ethnic Sombə population to be as much as 50% of the village population. The village has a primary school, a medical centre, and both a church and a mosque. There is no electricity service except for privately owned (i.e. personal), petrol-powered generators, nor is there pipe-borne water; water is collected either at a village pump or a nearby stream. Peasant farming is the principle occupation of the inhabitants of Kila Yang and surrounding area; virtually every family has its own fields. In the past, as just mentioned, blacksmithing was the livelihood of the men (i.e. the Sombə portion of the population), but even then the smiths would have engaged in subsistence farming. As far as is known, all males grew up in the trade; it is of course possible that some may have chosen other paths and left the group. The village also supports many traders and other occupations, though people engaged in such occupations are also farmers. To a lesser extent, cattle-herding is practised; those with cattle are typically relatively well-off, given the high value of cattle.

Synopsis of Sombə/Kila Yang history
Approximately 125 years ago (or perhaps as early as late eighteenth century), the Sombə migrated from the Adamawa Plateau to their present location (we note that this east to west direction is in the opposite direction from Mambila traditions). The Sombə no longer remember their place of origin, only its direction. Their oral traditions assert this was in search of better farmland, since in addition to smithing they are also farmers, and it may be noted that the soil at their present locale is indeed more fertile than on adjacent parts of the Adamawa Plateau. In contrast, it is possible that pressure from invading Chamba Leeko (eighteenth century; Fardon 1988) or Fulɓe (nineteenth-twentieth centuries) played a large role in precipitating their move. Indeed, the oral traditions collected at Hore Taram Torbi are explicit in claiming that invading Fulɓe uprooted local peoples, including the Sombə/Zhuzhun, causing them to seek refuge, and that is how the Zhuzun came to be at Hore Toram Torbi. In contrast to the Sombə, the Zhuzhun have a name for their place of origin, identified as Sagura. ⁴⁹ Hurault (1998) presents a scenario also involving the incoming Fulɓe on the Mambila Plateau, arguing that local groups were scattered and found refuge in the small forests of the Plateau, and it was by this means that villages such as Kila Yang (and, we add, Hore Taram Torbi on the Adamawa Plateau) were established, that are comprised of different clans or ethnolinguistic groups. A plausible scenario, then, is to suggest that invaderswhether Chamba Leeko or later Fulɓeprecipitated a movement that led to some Sombə seeking refuge at what is now Hore Taram Torbi, joining together with the Torbi (and later, Wawa speakers), while others fled further, to the Mambila Plateau where they ultimately found refuge at what is now Kila Yang. The Sombə claim to be the first inhabitants of Kila Yang, and that they were later joined by Mambila (the Maberem speakers); this has some credibility given that, as mentioned, "kila" is the Fulfulde word for "blacksmith." The current oldest generation of blacksmiths were the first among the Sombə to learn Maberem (the local variety of Mambila); indeed some of them never did come to speak it. This generation would have been born some 70-90 years ago, i.e. approximately in the 1920-1930s. As such, they would likely have been the second generation to grow up living together among the Maberem Mambila, assuming the discussion and chronology in Hurault (1998) are accurate.

A note on Hore Taram Torbi and Sombə
The village of Hore Taram Torbi is located east of Mayo Ndaga, across the border on the Adamawa Plateau in Cameroon. It is isolated and of difficult access in the mountains that form the Mambila escarpment, a region that is itself already remote. The village is small, with an official population of just 357 in 1997. The "Torbi" of its name refers to the largest of the three ethnic groups that comprise this small village, the two others being Wawa and Zhuzun. The Wawa component of the village was, according to our information, originally just one family though it has since attracted others; now there appear to be at least three, who have reportedly migrated from the separate but closely related villages of Oumyari, Ndi, and Tawa.
The Torbi speak a variety of Mambila,⁵⁰ though on our most recent visit we learned that only elderly people use the language, so it too is severely endangered. Some, perhaps many, Torbi also speak Wawa; we do not have more specific information on this, though several of the Torbi we spoke to said they spoke Wawa. Many of the Wawa inhabitants no longer speak their language on a daily basis. The few remaining Zhuzun apparently speak both Wawa and Torbi and can only be considered semi-speakers, or perhaps rememberers of their own language. (A speaker interviewed by Connell in 1997, now deceased, was said to be the last fluent speaker of Zhuzun in the village.) Already this suggests a fairly complex language ecology for the village, but in fact this is just a part of the story: the principle language of the village is said now to be the Banyo variety of Vute, although it is interesting to note that the village reportedly has no Vute residents. Finally, most inhabitants, perhaps all, also speak Fulfulde.

 49
No such name appears on modern maps of the region, but the Map of Cameroon edited by M. Moisel, E2, Banyo (1913) now in the Basel Mission Archives includes "Saguri" to the north and east of "Terám." 50 The Torbi have been referred to in the literature as a group separate from the Mambila who inhabit the eastern part of the Mambila Plateau. Linguistically however, people in the villages named as Torbi villages by Meek (1931) are speakers of various East Mambila lects. Meek suggests they may be that portion of the Mambila population who accepted the suzerainty of the Fulɓe during the conquest, and proposes a mixed Fulfulde-Sombə provenance for the term. Percival, following Meek, claims the name comes from the Mambila word for blacksmith, "turube" which the Fulɓe used for all Mambila, having first encountered a community of smiths (viz, Sombə). This strikes us as unlikely, not least because this word is not attested in any variety of Mambila known to us, but also because the name of the village (Kila Yang) and one of the local names for the Sombə community (Kila), used even by Meek) are Fulfulde, and mean "blacksmiths."

Linguistic overview
Mambiloid and its affiliations can only be tentatively stated. Connell (2010Connell ( , 2017Connell ( , 2021 proposes on the basis of comparative evidence (phonological, morphological, and lexical) that, within Mambiloid, Sombə forms a subgroup with Tep, Mvanip, Ndunda, and Mbongno. As discussed earlier, the Mambiloid grouping is part of Bantoid, and these languages are in the general area of what consensus holds to be the homeland of Bantu. It is thus a reasonable assumption that the progenitors of the Sombə have been in this region perhaps for millenia. An alternative scenario suggests the Sombə, relatively recently, recognised a need for their services, migrated into the area and eventually adopted the local language, and became more or less assimilated. While there are oral traditions relating to migration, as discussed, these accounts refer only to movements within the Mambiloid area; there is none that suggests an outside provenance. Moreover, there is little linguistic evidence to suggest their language is a result of wholesale language shift and subsequent modification; to the contrary, commonalities Sombə shares with Mbongno, Mvanip, Ndunda, and Tep indicate these five form a subgroup within Mambiloid (Connell, 2021). It also seems reasonable to suggest, given the language was used only by this blacksmith community, and that the demand for their services would always have been restricted, that the speaker population would always have been small.

Speech communities
Today, everyone in Kila Yang speaks Maberem and Fulfulde. Maberem Mambila is the primary language of the village; to a lesser extent people also speak Hausa (perhaps increasingly a lingua franca, though it is too soon to tell whether Hausa will replace Fulfulde). Many people speak or at least have a degree of passive knowledge of Tungba (the variety of Mambila spoken at Gembu, the largest urban centre on the Mambila Plateau). English is spoken by people who have been to secondary school (i.e. mostly those younger than about 50).

Language-identity relationships
With the demise of Sombə and the reduced state of the blacksmith culture, it is difficult to argue that the language has a strong associative value with Sombə identity. Still, people in the village are aware of its existence even if they rarely hear it spoken and they are certainly aware of their heritage as blacksmiths. Even today, young children can be seen banging on rocks pretending they are working at an anvil, replaying the work of their ancestors. It is not difficult, however, to make a case for the language as emblematic of Sombə identity in the past, when the language and the blacksmith culture were still vibrant. Being the language of the blacksmiths, and their being the only ones to have used it, it would have been at the heart of what it meant to be a Sombə blacksmith.

Functional domains
The language is no longer used on a daily basis, in any domain. However, it bears mention that the two people who do speak it are fluent and used it when speaking together. It is not clear from our research how frequently they used it, though they showed no reticence in performing monologues and dialogues for us, and they were overheard speaking it casually between themselves. We can perhaps assume that it is a language of intimacy for them; used, for example, when they do not want others to understand, though we note again that there are people of the next younger generation who are said to have a passive knowledge of the language, as did, for example, our interpreter when working in Kila Yang.
Sombə, spoken at Kila Yang, and Zhuzhun at Hore Taram Torbi are the same language, as discussed above. Only lexical data have been collected for Zhuzhun, and so our observations here are necessarily constrained. However, based on the data available, we find little difference between the two varieties; what differences do exist are in line with expectations for language varieties that have been separated for 150 years or perhaps longer: some differences in pronunciation, and occasional lexical differences.

Internal variation (social)
Given the state of this language, it is not possible to comment on the extent of internal variation that might have existed when it was still viable. The four remaining representatives of Sombə that we interviewed in 2010the two fluent speakers at Kila Yang and the two semi-speakers at Hore Taramare all of the same generation; three of four are of the same gender; the language is not used on a daily basis in different functional domains.

Status/solidarity
Again, little can be said with respect to how status or solidarity may have been reflected in Sombə usage; we might speculate that the occasions on which the two speakers at Kila Yang do use the language reflect a degree of solidarity between them not shared with other members of the community.

Written tradition, standardisation, institutional support
Throughout its history, Sombə never received institutional or governmental support; there was never a standard variety, nor was it ever given an orthography.

Attitudes of users
It is difficult to say much in detail about the attitude of Sombə speakers towards their language; the language now is on the edge of extinction, which suggests it has lost whatever associations it once had. That the language must have survived for centuries, likely always as a very small language spoken only by blacksmiths, suggests it may never had great prestige outside its own speaking group. By the same token, however, we might assume it held great significance for its speakers and was probably more important as emblematic of identity for them than for other ethnolinguistic groups who were their neighbours.

Association between language and economic/political advancement
Clearly, there is no economic or political benefit today in knowing or using Sombə. Connell (1998Connell ( , 2010a has argued that at least in part the demise of Sombə was brought about by the demise of the local blacksmithing trade. This itself came about partly through a change in traditional marriage customs in the region and through an increase in imported goods. In former times, it was customary for a prospective groom to provide bridewealth consisting of, among other items, a number of hoes, the manufacture of which was the stock-intrade of the smiths. The waning of this custom had a substantial effect on the blacksmith trade. To this extent, the fortunes of the language have been tied to economic considerations, though it is safe to say the language never was a vehicle for either economic or political advancement. At the same time (late nineteenthearly twentieth centuries), imported iron goods became available that could be reworked, e.g. to make knives and hoes, so undercutting the need for smelting and its contribution to economic aspects of Sombə livelihood. 7.14 Summary: Language contact in Kila Yang and Hore Taram Torbi According to oral traditions, the Maberem-Mambila joined the Sombə not long after they (the Sombə) founded the village. This indicates that close language contact has been a feature of the village since its founding, or shortly thereafter, but it is difficult to say at this point to what extent these two languages influenced each other; it is known that it was only the current and last generation of Sombə speakers in Kila Yang who shifted to Maberem as their principal language of daily use. Prior to this, there were presumably varying degrees of accommodation between speakers of the two languages, but much communication would have been (and to some extent still is) in Fulfulde, once it became established as the regional lingua franca. And, as mentioned in Section 6.5, there is considerable exposure to Tungba-Mambila as well as English and, to a lesser extent, Hausa and other regional languages.
At Hore Taram Torbi, the situation is similar, though somewhat more complex, in that three different groups came together to form the village; Sombə speakers joined together with a Mambila speaking group; these were then joined by a Wawa speaking group. Both Fulfulde and Vute are also known by most inhabitants and French by some. Unusually for the region, Vute, and not Fulfulde, is reported as being the lingua franca of the village (and this despite there being no Vute residents).
The upheaval introduced to the area with the Fulɓe conquest led to changes in the language ecology of Sombə speakers, most obviously in that it divided their community, some taking refuge at what is now Hore Taram Torbi, others moving onto the Mambila Plateau to establish Kila Yang. As a result, they were brought into close contact with languages with which they had little, if any, contact previously. Joining together with Maberem Mambila in village life would have brought on subtle and not so subtle changes in the blacksmith culture. Among these, we note that changes in marriage customs, brought on possibly by the influence of Islam arriving from the north and both Christianity and colonial rule from the south, meant (10 or more) hoes⁵¹ were no longer needed as gifts to future parents-in-law ("bridewealth" in the anthropology literature). This may be connected more to the arrival of money than influence from world religions (see Bohannan's 1955 classic paper on spheres of exchange among the Tiv and Rehfisch (1960) on multiple Mambila marriage systems before WW2), but either way a substantial part of the Sombə's trade was lost. Similarly, it has been noted that the arrival of imported metal goods, which could be reworked when no longer useful, obviated the need for smelting and this aspect of the blacksmiths' livelihood ceased. These changes and no doubt others, undocumented, led to the loss of functionality of Sombə and its ultimate demise.

Discussion
In the preceding sections, we have presented case studies of four languages having different degrees of vitality. One of these languages, Bà Mambila, is relatively stable and following Krauss' (2007) classification  51 It should be noted that often what are called "hoes" are sheets of iron in the approximate shape of hoe blades rather than parts of functional tools. They are reservoirs of value.
can be labelled as "Stable" (A). The other three exhibit lesser degrees of vitality; Oumyari Wawa is "Eroded" (A-), while both Njanga and Sombə are "Critically Endangered" (D/D-), though in detail their statuses and situations differ in interesting ways. All of these languages, on one level, constitute part of a regional language ecology (cf Section 3), yet the differences among them reflect differences in their own specific language ecologies (Sections 4-7). These can be insightfully diagrammed using the LMP model introduced in Section 2, represented as Figure 1, reproduced here as Figure 8 to facilitate our discussion. The outer circle, having "macrocultural competences/attitudes," "macrospaces" (i.e. large scale spatial distribution as shown in the maps above, relating to where different lects are spoken), discourse traditions/history, and language types/families as its components, encompasses aspects of the languages under study which are shared. In our case, the outer circle represents the historically constituted macro-ecology; the Adamawa region and the "Mambilophony:" including its history, cultural traditions, and linguistic characteristics: all are Mambiloid languages and are typologically similar; all are in contact with Fulfulde as the lingua franca and secondarily with either French or English. Their speakers have lived in the region possibly for millennia and have, grosso modo, a shared history and culture base (some features of which have been discussed in Section 2.4), which has led to commonalities in discourse traditions.
Since our principal aim is to discover why languages which share so much have taken different trajectories with respect to their vitality and sustainability, our focus has been to look at their specific situations. These, schematically, are located in the middle ring of the three in the LMP model, which has as its components group-specific competencies and attitudes, considerations of not only physical space, but different social groups and interactional types, discourse patterns, and codes. The details, to the extent these are available to us, were presented in Section 3; detailed study of discourse strategies, for example, has not been undertaken for any of the languages we've examined and effectively is no longer possible for Njanga and Sombə. These details revealed considerable differences among the four languages, not least in the specific patterns of language contact. That is, they represent different reactions or responses to changes in the relations between features found primarily in the two outer rings of the LMP model, as summarised in the following paragraphs. However, as the model implies, they also carry through to (and/or spread out from) the inner circle. The comparative approach adopted in this work has led away from discussion at this level.
The key features of Bà Mambila as spoken at Somié that set it apart from the other cases are to some extent its size, and arguably of greater importance, its "membership" in the larger Mambila community. With approximately 3,000 speakers, its population is by no means large, but in itself is nevertheless roughly equal to all Wawa dialects combined; it is probably substantially larger than Njanga or Sombə were at any time in their histories. As for being part of a larger Mambila community, this is to be recognised on two levels: first, Somié is part of a wider Bà Mambila-speaking community, in that there are two other villages and several hamlets which are Bà Mambila-speaking; and second, the existence of Mambila as a "macrolanguage;" though there are several dialects, with substantial differences among them, their speakers all see themselves as ethnolinguistically Mambila.
Oumyari Wawa speakers, too, form part of a larger community in that there are other varieties spoken in their immediate region, though all are small and the area they inhabit is also small. Apart from its size, one aspect of its ecology that distinguishes it from the other languages studied here is found in its history, or how its speakers reacted to historical events; the Wawa, unlike most of the neighbouring people, chose to accommodate rather than resist the invading Fulɓe. As a result, despite the fact that intergenerational transmission is still strong, the language has undergone and continues to undergo influence from Fulfulde. As mentioned, there are situations in which Wawa children appear not to distinguish between the two languages. The fact that the Wawa, unlike the other groups discussed here, have converted completely to Islam has also almost certainly facilitated the extensive borrowing of Fulfulde into the language and brought the language to its present endangered state. And while one of the strengths of a Wawa identity for Oumyari may be that it is part of a wider network of Wawa villages, a "support system," the situation described here is part of an ongoing process and how long the language, or indeed a Wawa identity, will survive is uncertain.
Njanga and Sombə have commonalities not shared by the other two languages, particularly in that both are critically endangered. Unlike the other two, both are exceptionally small, not only now, but perhaps always were.⁵² This certainly must be the case for Sombə, as argued earlier; it is not possible to reach back into the history of Njanga to any depth, but as far as we have been able to, there has only ever been one Njanga speaking village. So, unlike Bà Mambila and Oumyari Wawa, neither Njanga nor Sombə had a wider "support system." The fate of both languages can also be linked to how their speakers reacted to historical events. If the oral traditions collected in Hore Taram Torbi are accepted, the Sombə fled from the invading Fulɓe, thus scattering an already small group. For the Njanga, we found no traditions specifically asserting they fled from the Fulɓe, but as we noted earlier, the first village site referred to was conceivably established as they sought refuge from the invading Fulɓe. Hurault (1964Hurault ( , 1979 argues that mountain refuges were commonly sought as the indigenous people fled the Fulɓe. Despite the similarities between Njanga and Sombə, there are also substantial differences. Njanga survived due to the symbolic role it played, while at the same time it may be argued it has reached its demise now it has come to have only a symbolic role. Part and parcel of this is that many, indeed most, of its remaining speakers were at best semi-speakers. This differs from Sombə where the last two speakers are fluent, and even within our time in the region there were as many as 20, and possibly more, fluent speakers. That is, the passage of Sombə into obsolescence took a different route than Njanga. While Njanga withered as the political importance of its speakers waned, Sombə (its speakers) never had political importance; and though their trade was central to their existence as an autonomous group, and hence that of their language,  the blacksmiths with whom we worked were of the first generation to learn Maberem, the Mambila variety which has replaced Sombə.

Conclusions
As one moves from immediate documentation to tracing patterns of language contact and change over the longer term, one inevitably becomes concerned with linguistic evolution. Socioculturally, the Mambiloid languages are changing to fit new demographic and technological niches. This makes evolutionary phenomena visible. Our different case studies, which are of related languages situated close together geographically, show a range of responses made by these highly multilingual speakers. These can seem like creative blends (Fauconnier and Turner 2008) of the linguistic material available as well as straightforward borrowing and supplanting of lexical items or entire linguistic systems. Such creativity has been discussed for colours and numerals of Wawa and Njanga by Griffiths and Robson (2010). Our ecological approach considers change as resulting from the complex interaction of many factors (consonant with Trudgill (2011) who notes that often contact-induced change has been misleadingly seen as quite one-dimensional by linguists).
The linguistic ecology that we have described for the Mambiloid languages provides speakers with many different materials to choose from, linguistic resources that help them adapt to the new situations in which those speakers find themselves. However, the pressure is on for these languages to adapt to changing political, social, demographic, cultural, and cognitive aspects of daily life. Growing population, the introduction of formal education as well as money becoming ubiquitous have all changed the language ecologies that we have been considering.
The LMP framework provides an interesting and helpful analytic tool to give purchase on the complexity of language ecologies and enables discussion and comparison across differences. This is an important advance and one that has helped us bring into focus aspects of our own data.
Indeed, the LMP approach underpins our conclusion that linguists need to "change the frame." Linguistic theory needs to resist the temptation to take languages such as contemporary English, Mandarin, Spanish, or French as normative. Their situations and ecologies are almost certainly not "normal" in global historical terms and at a theoretical level they are misleading. Our belief is that for most of its history the world has been more like Somié, Mbondjanga, Oumyari, or Kila Yang than Western Europe today, where national language equates to the nation-state at the expense of small and minority languages; the theoretical framework used by linguists should reflect this. We find a language ecological approach, particularly the LMP model, provides a positive starting point. Four poles, explained below, define our general response to their model.
(1) Multilingualism is normal/normative (2) Language Change is normal/normative (3) Complex maps have complex histories (often fractal ones, see Zeitlyn and Connell 2003), i.e. complexity is normative (4) Different responses are possible. Human history is not a deterministic system.
Multilingualism is normal/normative. That is, monolingualism should be taken as unusual and needing of explanation rather than as the default. Evidence for this may be found in combinations of linguistic and genetic research. other words, the evidence is that the people of the Lower Fungom valley have been intermarrying for generations while maintaining the different languages spoken in their villages. So the parties to marriages must have often been speakers of different languages. The situation in the Cross River area of Nigeria is similar at a somewhat larger scale, both of population size and linguistically (Veeramah et al. 2010). So we conclude that the default (from which exceptions may occur) is that people normally speak several languages.
Language Change is normal/normative. Work on language endangerment forces one to confront the reality of language change. However, historical linguistics provides substantive documentation of the many different ways languages change over time, both internal and external change. The dominant view in language endangerment scholarship is shaped by situations in which a language is threatened by an incoming language, often colonial, sometimes a recently promoted national language and that this is somehow not "normal." A language ecological view undermines some extreme positions on language purity and allows one to understand the commonalities between situations involving colonial languages (in that these too can be understood through the LMP model) and those described here; and we might add historical evolution or change of a language in its external aspects. The complication, and why we remain committed to working on endangered languages, is the issue of what could be termed "linguistic legacies to the future." Although no one now speaks Old English, it is easy to understand how that language has contributed to the forms of English being spoken in the twenty-first century. Old English has a clear linguistic legacy. The concern for historical linguistics about endangered languages such as the ones we have discussed here is that lects such as Njanga or Sombə may leave no discernible trace on the languages that will be found spoken in the area one or two centuries hence. In the absence of such legacies, future linguists may struggle to understand the complex language ecologies that we have documented.⁵⁴ The two final points are best discussed together, since they are closely intermeshed. They result from teasing out some of the implications of the truism that the world is a complicated place. Not only is language ecology itself complicated, but it fits into a dizzying array of factors that all have implications for how languages change over time. Furthermore, it seems there is complexity at all scales as we summarise in our heading "Complex maps have complex histories." We also note that these are often fractal ones (see Zeitlyn and Connell 2003). If complexity is normative, the results do not produce a deterministic system. Human history, linguistic history, is not predictable. Even when there are external upheavals, their consequences (including their linguistic implications) cannot be predicted with confidence or read off the preceding conditions. Different influences, whether invasions and colonisations or varying degrees of language contact, may have multiple and different results in neighbouring, similarly situated locations. All of these point to the value of specific historical research of the type we have provided above.
In sum, what we have shown is the need for understanding language ecologies as an essential part of comprehending language change and the differing fortunes of closely related languages. Even closely related groups that share key factors may weight them in different ways. Contact is but one factor in language change. We cannot assume that the same factors are important everywhere. This has theoretical significance that can be accommodated in models such as that proposed by LMP (2019).
colleagues and collaborators, friends, and supporters in Cameroon and Nigeria, in particular Professor Ngessimo M Mutaka of the University of Yaoundé 1 for assistance in the early stages of the research, as well as the Institute of Mambila Studies and York University, Toronto. Research in Cameroon was done under research permits 042, 043, and 044 of 19/6/2007, issued by the Ministry of Scientific Research and Innovation. We are particularly grateful for the detailed comments received from an anonymous reviewer and to Prof Peter Mühlhäusler and the editors of Open Linguistics.
Funding information: Research supported by AHRC grant AR112306 "Documentation of Endangered languages" to David Zeitlyn.
Author contributions: All authors have accepted responsibility for the entire content of this manuscript and approved its submission.