Show Summary Details

Page of

PRINTED FROM the OXFORD RESEARCH ENCYCLOPEDIA, LINGUISTICS (linguistics.oxfordre.com). (c) Oxford University Press USA, 2016. All Rights Reserved. Personal use only; commercial use is strictly prohibited. Please see applicable Privacy Policy and Legal Notice (for details see Privacy Policy).

date: 24 November 2017

Social Network Approach in African Sociolinguistics

Summary and Keywords

The Social Network Analysis approach (SNA), also known as sociometrics or actor-network analysis, investigates social structure on the basis of empirically recorded social ties between actors. It thereby aims to explain e.g. the processes of flow of information, spreading of innovations, or even pathogens throughout the network by actor roles and their relative positions in the network based on quantitative and qualitative analyses. While the approach has a strong mathematical and statistical component, the identification of pertinent social ties also requires a strong ethnographic background. With regard to social categorization, SNA is well suited as a bootstrapping technique for highly dynamic communities and under-documented contexts. Currently, SNA is widely applied in various academic fields. For sociolinguists, it offers a framework for explaining the patterning of linguistic variation and mechanisms of language change in a given speech community.

The social tie perspective developed around 1940, in the field of sociology and social anthropology based on the ideas of Simmel, and was applied later in fields such as innovation theory. In sociolinguistics, it is strongly connected to the seminal work of Lesley and James Milroy and their Belfast studies (1978, 1985). These authors demonstrate that synchronic speaker variation is not only governed by broad societal categories but is also a function of communicative interaction between speakers. They argue that the high level of resistance against linguistic change in the studied community is a result of strong and multiplex ties between the actors. Their approach has been followed by various authors, including Gal, Lippi-Green, and Labov, and discussed for a variety of settings; most of them, however, are located in the Western world.

The methodological advantages could make SNA the preferred framework for variation studies in Africa due to the prevailing dynamic multilingual conditions, often on the backdrop of less standardized languages. However, rather few studies using SNA as a framework have yet been conducted. This is possibly due to the quite demanding methodological requirements, the overall effort, and the often highly complex linguistic backgrounds. A further potential obstacle is the pace of theoretical development in SNA. Since its introduction to sociolinguistics, various new measures and statistical techniques have been developed by the fast growing SNA community. Receiving this vast amount of recent literature and testing new concepts is likewise a challenge for the application of SNA in sociolinguistics.

Nevertheless, the overall methodological effort of SNA has been much reduced by the advancements in recording technology, data processing, and the introduction of SNA software (UCINET) and packages for network statistics in R (‘sna’). In the field of African sociolinguistics, a more recent version of SNA has been implemented in a study on contact-induced variation and change in Pana and Samo, two speech communities in the Northwest of Burkina Faso. Moreover, further enhanced applications are on the way for Senegal and Cameroon, and even more applications in the field of African languages are to be expected.

Keywords: sociolinguistics in Africa, social network analysis, variation studies, under documented contexts, linguistic innovation

1. Background

The field of general sociolinguistics has to acknowledge that African societies are in many ways different from those of the Global North. In Africa, the term nation state covers a wide range of political constructions that most often stand at odds with the Western concept of state and society. Ethnicity is not a result of racial segregation and social identity construction, but in most African socio-political settings, ethnic groups are grounded in primordialism, distinct cultures, and languages thus adhering strongly to ethnic nationalisms.

Nevertheless, present African postcolonial societies have little to do with the early descriptions of European travellers. While tradition, family, and descent are still important, African societies also appropriate modern life-modes and identity perceptions. Although slave trade and colonialism, as well as modernity and postcolonial globalization, have impacted massive social and political changes in Africa, lively syncretism and active appropriation document a deep-rooted resistance to the life modes of industrialized societies. As social categories like class, gender roles, and others have emerged under different historical conditions one cannot, for instance, equate a traditional social category as clan with the classic Marx/Weber social class concept and proceed with the standard variationist approach.

Compared to Western contexts, variationist studies in Africa are rare. This fact is also due to the specific colonial and post-colonial research history in Africa. Given the linguistic diversity on the continent of approximately 1,900 distinct languages, of which the majority are still under-documented, the academic focus lays on language descriptions and on historical and applied linguistics. Furthermore, the strong tradition of African anthropological linguistics is less concerned with language and society and more engaged with traditional knowledge and lexical semantics. However, the descriptive gap is slowly diminishing, which is a prerequisite for variationist studies.

Currently, most Africanists acknowledge the fact that linguistic diversity is a reason for variation as such. The high degree of societal and individual multilingualism often results in code switching according to social domains. Studies on language use in multilingual settings show that variation affects all languages of a speaker’s repertoire most often according to situational factors. A growing body of evidence also points to cases of ‘organic multilingualism’ where mixing, switching, and intertwining of elements from the highly diversified repertoire is the default in communication. As linguistic interferences add to sociolinguistic parameters, sociolinguistics in Africa is rarely concerned with variable rules of one language only, but most often is confronted with contact-induced variation. Thus, linguistic variation in Africa is a highly complex matter where societal and individual multilingualism lead to linguistic interferences governed by specific social parameters.

With the introduction of Social Network Analysis (SNA) to variation studies by the Milroys (Milroy & Milroy, 1978, 1985), the authors not only widened our understanding of the mechanisms of language variation and change but also introduced a powerful ethnographic method to the field. As a matter of fact, SNA is an ideal bootstrapping technique for social categorization in highly dynamic communities and under-documented contexts, and it is extremely helpful for the identification of pertinent social relations in a given speech community. For sociolinguistics in Africa it therefore offers a framework to understand and explain the patterning of linguistic variation and the mechanisms of language change.

2. History of SNA and Current Developments

Social network analysis as a formal representation of social structure is based on the quantification of actor ties, which are conceived as the most basic social relations. In a predefined group, all existing ties between the members (the actors) according to a social activity or relation (e.g., being friends, going to a regional market together, seeking advice) are aggregated into a coherent representation: the network. This approach has been a central issue to the school of British sociology in the second half of the last century (Barnes, 1954) and also has precursors in anthropology (Radcliffe-Brown, 1940, Schnegg, 2010a). It has been successfully applied in Africa by Kapferer (1969, 1972) and in various other studies by the anthropological school of Schweizer (1989, 1996; and see Schnegg, 2010a, p. 26). Lately, its influence as a methodological tool in anthropology diminished because the dynamic and open characteristics of networks in modern societies are very difficult to operationalize (Hannerz, 1992, p. 51; see Schnegg 2010b, p. 861). While the traditional fields of applications in anthropology have mostly been kinship and economic exchange relations, pragmatics and communicative ties were not perceived as social activity, and consequently, were excluded from ethnographic description. However, many procedures and measures from the ethnographic SNA can be taken over for the sociolinguistic approach.

Social networks are also central in theories that model innovation and diffusion of material and non-material goods in social groups. In these approaches, network principles are part of an explanatory framework that allows determining the mechanisms of diffusion of innovation and norm retention alike (Schweizer, 1996, p. 11ff.). Weak ties, low density, and structural holes in the social network are held responsible for the lack of control over individual actors and counter the development of a common identity in a given community (Granovetter, 1973). Since Rogers’ “Diffusion of innovations” (1962), the diffusion problem is by itself a vast field that has tremendously developed over the last decades and by now “includes more than 5,000 studies. No other field in the behavioural and social science represents more effort by more scholars in more nations” (Danowski, Gluesing, & Riopelle, 2011, p. 128). Whether referring to material culture or linguistic features, the acceptance of innovations by individuals appears as a rather complex phenomenon, and related theories have multiplied.

The adaption of SNA to sociolinguistic variation studies was facilitated by the view that speech communities are social systems and that an actor’s position is based on her/his communicative and social acts. Thus, social activity equals communication, and social ties are established through verbal exchange. By now, this view has become a central component in social and in linguistic theory (Austin, 1962; Bourdieu, 1991; Coupland, Sarangi, & Candin, 2001; Habermas, 1981; Luhmann, 1984; Searle, 1994). According to this concept, language variation in all its facets is understood as grounded in everyday social interaction, and language change can be regarded as a function of the need for social differentiation and identity construction of individuals and speech communities.

Based on the assumption that communication is social activity, Milroy and Milroy (1978) and Milroy (1980) first added SNA to the sociolinguistic toolkit in their study of vernacular speech forms in three Belfast communities. The application of the concept to this urban context also benefited from Kapferer’s (1969, 1972) work in Zambia. In the beginning, however, the network idea was just “a set of procedures rather than a fully fledged theory” (Milroy, 1980, p. 46). Again relying on Granovetter’s (1973) principles of diffusion of innovation via weak ties and its counterpart, the resistance to change in close-knit networks, they developed the aspect of structural opponents in the network to be able to tell innovative from conservative speakers. Based on SNA measures from a range of actors from the targeted Belfast speech communities L. Milroy formulated a general hypothesis stating that “closeness to vernacular speech norms correlates positively with the level of integration of the individual into local community networks” (Milroy, 1980, pp. 133–134). Subsequently, they refined their method to compare the individual characteristics of linguistic actors and also tried to define general features of social networks that govern the processes of language change in a speech community. Their general hypothesis regarding the correlation between the structure of a network and its proneness to change reads like this: “Linguistic change is slow to the extent that the relevant populations are well established and bound by strong ties, whereas it is rapid to the extent that weak ties exist in populations” (Milroy & Milroy, 1985, p. 375).

The quest for and the description of the leaders of linguistic innovation has also been a component of other sociolinguistic research. Labov’s sociometric (Labov, 1972) analysis and discussions of personal attributes of innovators refers to the same principles of diffusion via social networks. In later studies, SNA was fruitfully applied to a variety of contexts by various authors (e.g., Gal, 1979; Labov, 2001; Lippi-Green, 1989), most of them, however, located in the Western world.

Other attempts to use SNA in linguistic research focus on the dynamic character of social networks. For instance, when combining two representations of the same network at two different time points, one can display the spread of any given feature. This idea lies behind Ross’ (1997) representations of social networks, which he uses in a metaphorical way to display his ideas about the division of Proto-Malayo-Polynesian into its various sub-families. He also tries to back-up the graphic representations of the supposed developments with reconstructions of speech-community events that help to establish a relative chronology of the developments. As interesting as this is, it is void of the ethnographic basis needed to establish a synchronic ‘real’ network, which in turn is indispensable data needed to develop realistic models of whatever speech community in question.

Such dynamic modelling on the basis of real life networks represents the most recent development in SNA. While until recently sociolinguistic network studies primarily focused on synchronic linguistic variation in small speaker groups, they commonly stay mute regarding the question on how social networks affect language change at a larger historical scale. To model the effects of social networks over extended time periods, recent approaches use concepts of large-scale networks and innovation diffusion to account for language dynamics and linguistic changes from a computer-aided simulation perspective. According to this perspective (Fagyal, Swarup, Escobar, Gasser, & Lakkaraju, 2010; Ke, Gong, & Wang, 2008; Nettle, 1999) linguistic change is based on innovations that need to overcome a threshold of acceptance in the network to successfully spread and become new norms. Once the parameters for the overcoming of the threshold are defined, computer simulations manipulate further features of the networks to make guesses as to how language changes at a large time-scale. However, this line of SNA studies is still at the beginning and especially needed are empirical studies on “linguistically-relevant large influence networks” (Fagyal et al., 2010, p. 2067).

3. SNA: Praxis and Theory

The social network approach consists of two related components: network theory and network analysis. Somehow, the term social network analysis has become the cover term for the whole approach. While the theory lays the general foundation for research, argumentation, and the motivation of basic model components, the analytical part is concerned with the implementation of data collection, the calculation of network measures and the graphical representation of the results. Network theory is also a cover term that includes various views on how to explain social mechanisms by concepts such as weak and strong ties or actors’ embeddedness in the social tissue.

A sociolinguistic study that draws on the SN approach usually comprises an ethnographic description and documentation of a given social structure and strives to identify the social parameters that govern linguistic variation. Alongside this analytical part, theoretical concepts and theories on diffusion and change can also be tested. The social network approach as advocated here goes way beyond a metaphorical use of the term network and provides a formal representation of the social structure of a group or speech community. One should also keep in mind that communicative ties are not specially treated in SNA, as the concept of a tie between actors implies the flow of information and verbal interaction. Tie strength indicates frequency or personal emotional values in a given relation but not quality in terms of successful verbal interaction. The social network model does not reproduce interaction in discourse.

SNA relies on the following basic assumptions:

  1. 1. Actors and their actions are viewed as interdependent rather than independent units.

  2. 2. Relational ties between actors are channels for transfer or flow of resources (material or non-material).

  3. 3. The network structure provides opportunities for and constraints on individual actions.

  4. 4. Network models conceptualize structure, such as social, economic, political as patterns of relations among actors (Huber, 2007, p. 43).

Inherent to social network data is the abstraction from what constitutes intuitively or knowingly a personal relation in a given social setting. A network model uses a formalization technique to store relations as edges and individual actors as nodes in matrices. Ties can be of multiple kinds and of different levels of strength. Network research also records primary individual attributes of actors such as age and level of education.

The network data can be gathered with different strategies like survey, ethnographic, or documentary research. The social group under scrutiny may be predefined through common constant membership such as colleagues at a workplace, school class, or family members. A common technique for other cases where the boundaries are unclear or too wide is the snowball method. One starts with an individual actor who is considered central in a certain respect and then proceeds along her/his social relations in a snowball manner (Jansen, 1999, p. 65ff.). In actual research, one often uses a blend of both attempts if, for example, a village population cannot be investigated as a whole but sampled. In such a case, several central actors from different subgroups are taken as points of departure.

The basic approach for the analysis of social network data is mathematical graph theory and matrix algebra. The resulting representations of overall structure, sub-groupings and patterns are achieved with matrix summation of the edges in a network. The related network measures like, for instance, power, influence, or embeddedness are calculated through formalized algorithms defined by social network theory (Rogers, 1962; Valente, 1995, 2005; cited after Friemel, 2010; see also Hall, 2006).

A matrix contains either binary information about a relation, like “X is a friend of Y” or “X is not a friend of Y” (Table 1), or valued data, like “the grade of friendship between X and Y” (Table 2). Binary relational data can be directed or non-directed, indicating if a relation is claimed by one actor or both actors. For example, “X likes Y” and “Y likes X,” or “Z likes W,” but “W does not like Z.”

Table 1. Adjacency Matrix (Binary) Indicating Friendship Relations Between Individual Actors

Actor/Actor

1

2

3

4

5

6

7

1

0

0

0

0

0

0

2

0

1

0

0

0

0

3

0

0

1

0

0

0

4

0

0

1

0

0

0

5

0

0

0

0

1

1

6

0

0

0

0

1

0

7

0

0

0

0

1

0

Table 2. Valued Adjacency Matrix Indicating Friendship Relations Between Individual Actors

Actor/Actor

1

2

3

4

5

6

7

1

0

0

0

0

0

0

2

0

3

0

0

0

0

3

0

0

3

0

0

0

4

0

0

3

0

0

0

5

0

0

0

0

4

3

6

0

0

0

0

4

0

7

0

0

0

0

4

0

Relations can be characterized as a link between the same kind and different kinds of actors. A matrix containing information about the same kind of actors is called an adjacency matrix—for example, friendship relations between individuals (see Tables 1 and 2). In an incidence matrix, the data about relations between different kinds of actors is stored; for example, individual actors attending local market places (Table 3).

Table 3. Incidence Matrix; Relations Between Actors And Marketplaces.

Actor/Marketplace

Mara

Din

Poro

Nassari

1

1

1

0

0

2

1

1

1

1

3

1

1

1

1

4

1

1

0

1

5

1

1

1

1

6

1

1

1

1

7

1

1

1

1

8

1

1

0

1

9

1

0

0

1

10

1

1

0

0

11

1

0

1

1

12

0

1

1

1

13

0

0

0

0

14

1

1

1

1

15

1

1

0

1

16

1

1

1

1

In terms of sociolinguistic research, SNA provides the means for the categorization of speech group and individual actors on an empirical basis, which than serves for correlations with linguistic variables. On the group level, it allows one to measure density and the grade of heterogeneity of the speech community as a whole. Both measures are of major value for comparison and discussions of social forces leading to linguistic variation. Moreover, as these measures are the sum of individual relations in the network, SNA is independent from any external categorization and only relies on emic categories. However, for a concrete implementation, the issue is somehow shifted to the question of what pertinent social ties in a given community are.

Thus, besides mastering the formal methods and concomitant SNA statistics, the success of applying SNA heavily depends on achieving ethnographic adequacy for the conceptualization of the social relations. For the implementation of SNA, gaining knowledge through ethnographic methods such as participant observation is an important step prior to the operationalization of the social ties. There is, however, no generalized procedure to do so, because social relations that are regarded as primary by a certain speech group may be regarded as irrelevant by another community. For example, the concept of neighborhood plays an important role as a social category in Philadelphia (Labov, 2001) and Belfast (Milroy & Milroy, 1978), and it is also the case in the West African Samo and Pana speech communities studied by Beyer and Schreiber (see Section “The Pana-Samo Project”). But it is not assumed by principle that this factor is relevant in all cases for the social network.

Network data can be analyzed either qualitatively, in terms of graph interpretation, or quantitatively, in terms of network scores and their statistical correlation with linguistic variables. Both aspects complement each other and thus represent two ways of looking at the same thing.

3.1 The Quantitative Approach

Since the introduction of the SNA to sociolinguistics, the original method has developed in many ways. The rise of digital communication networks has made network data available in a previously unthinkable dimension. This led to intensive research on new statistical approaches and the mathematical foundations of network modeling (see Further Reading). In this context, new standalone software and statistical packages for scripting languages (e.g., ‘sna’ for R) allow far more approaches to data analysis than in the early times of statistical database software. Current SNA software like UCINET (Borgatti, Everett, & Freeman, 2002) or VennMaker (Schönhuth, Kronenwett, Gamper, & Stark, 2014) offer a range of ready-to-use standard statistical procedures and adapted visualization tools of network structure.

A usual procedure in the quantitative approach aggregates the number, strength, and type of relations into a network score and then correlates it as a numeric vector with the linguistic variables. This procedure was employed, for example, in the studies of Milroy and Milroy (1978) and others (see Further Reading). In these attempts, an individual actor’s score is calculated with relational network data and then treated as any other individual attribute, like age or gender. The quantitative network approach, therefore, differs from other variationist approaches as some attributes are based on aggregated network measures and then correlated with linguistic variables.

Moreover, network scores are continuous data. This allows the use of regression methods if variables are operationalized likewise as continuous, for example as frequencies of variable rule outcomes. As already shown in early sociolinguistic studies (Labov, 1969), most speakers vary with regard to variable rules, and the range of their variation depends on external factors. Speakers may thus be categorized with regard to their use of variable rules, which results in a frequency distribution applicable for more elaborate inferential statistical analysis. This second approach requires a linguistic interview, in which the same variable occurs in a controlled setting and in sufficient token numbers.

The semi-categorical character and the social nature of language in combination with the empirically grounded method of data collection thus involve a quantitative method that is different from usual standards of statistics. Sociolinguistics is generally more occupied with exploratory than with inferential statistics. Similarly, statistics in SNA is likewise different form traditional inferential approaches as social network data are inherently dependent. Correlations between network indices also require specific statistical methods (Borgatti, Everett, & Johnson, 2013, p. 126ff). However, as social network measures are independent from linguistic variables, both can be correlated.

The choice of SNA measures in sociolinguistic research is also motivated by theoretical considerations. According to network theory and diffusion of innovation research, centrality measures are considered salient for detecting opinion leadership and for displaying structural equivalences of actors (see Section “History of SNA”). Still, the transfer of traditional sociolinguistic concepts onto SNA measures is not straightforward. For instance, a concept like prestige is differently conceived in sociolinguistics and network analysis. In SNA, prestige was mainly understood as a combination of an actor's centrality in the network and the power s/he can execute by virtue of this SN position. Bonacich (1987) proposed to differentiate between two kinds of prestige in the SNA paradigm, namely centrality-based and power-based prestige. He claimed that an actor only has power-based prestige if s/he is connected to others who are not so well connected, not powerful by themselves, while traditionally it was assumed that central actors with many ties are powerful as such (Hanneman & Riddle, 2005). However, only a few sociolinguistic studies take up recent trends in social network theory and apply newly developed measures to their data (but see Marshall, 2004).

Another strand of the quantitative network approach goes beyond individual speaker’s networks scores and attributes. Properties of the network itself, like density and cohesion, have an impact on the support of local linguistic norms on the community level. For instance, the Milroys found out that dense and multiplex networks have a norm conserving quality, whereas loose networks seem to favor language change (see Section “History of SNA” and: Milroy & Milroy, 1985, p. 375). Therefore, SNA offers an empirical testing ground for diffusion simulations in different kinds of networks, or for integrated models of language contact that relate community attributes such as group cohesion, with the outcome of contact-induced change (see section 4.2).

3.2 The Qualitative Approach

In the qualitative approach, the network data is transferred into a sociogram, a graph displaying the network structure as individual relations between actors. The purpose is to indicate not only the amount of ties of an actor quantitatively, but also his/her position or embeddedness in certain groupings. Social distance between actors is displayed then as spatial distance in a graphical representation of the network. In a network graph, individuals are represented as nodes (points, dots, vertices) and social relations as ties (connections, edges, lines) (Figure 1).

Social Network Approach in African SociolinguisticsClick to view larger

Figure 1. Graphic representation of basic relations and actor attributes.

When indicating a valued relation of strength, the ties may be displayed in different sizes to show the attribute. Different colors or geometric forms may be used to display different qualities like kinship or neighborhood. Arrows show the direction of a given relation (Figure2).

Social Network Approach in African SociolinguisticsClick to view larger

Figure 2. Graphic representation of tie attributes.

Density is visualized with the spacing between individual nodes in the network. Clusters and cliques appear in the graph as actors who are strongly connected with some but not all actors of the network (Figure 3).

Social Network Approach in African SociolinguisticsClick to view larger

Figure 3. Visualization of network density: (a) close-knit cluster vs. (b) loose network structure.

If actors are connected via different kinds of relations like kinship and neighborhood, the network is characterized as multiplex (Figure 4).

Social Network Approach in African SociolinguisticsClick to view larger

Figure 4. Visualization of network multiplexity.

Reciprocity or symmetry indicates whether the directed ties are mainly reciprocal or unidirectional (Figure 5).

Social Network Approach in African SociolinguisticsClick to view larger

Figure 5. Visualization of network symmetry.

The sociogram is the result of graph theoretical algorithms on the basis of a data matrix (Figure 6).

Social Network Approach in African SociolinguisticsClick to view larger

Figure 6. Visualization of a social network by Netdraw, based on the sum of all recorded ties.

The resulting sociogram is a means to identify the roles of individual actors that are not necessarily indicated by quantitative network measures of centrality. For example, some marginal actors are only loosely integrated into the network and characterized, for example, by a low rate of centrality. They may function, however, as connecting actors or brokers between subgroups and fulfil an important social role (Figure 7).

Social Network Approach in African SociolinguisticsClick to view larger

Figure 7. Types of actor roles in a social network.

Mapping attributes like the use of linguistic variants onto the network structure complements statistical correlations, indicating such diffusion mechanisms that are triggered not by common actor attributes but by mere social proximity. A graphic visualization of sociolectal distribution is particularly interesting if diffusion occurs, say, in only one subgroup of the whole network (Figure 8).

Social Network Approach in African SociolinguisticsClick to view larger

Figure 8. Example distribution of a linguistic variable in a sample network.

It is also a tool for data exploration in order to develop hypotheses that are then to be tested with quantitative means.

4. Applications in Africa

A range of publications explicitly mentions social networks but do not use the approach in the analytical and/or theoretical sense referred to here. For instance, Owens, in his study of variation in the spoken Arabic of Maiduguri, Nigeria, uses SNA as a synonym for social micro-groups: “data from micro-groupings which would probably qualify as closely-knit social networks if a formal definition of them in these terms were to be attempted” (Owens, 1998, p. 259). SNA is also prominent in another title by Graham, “A Look at the Acts of Identity Theory Through a Social Network Analysis of Portuguese-Based Creoles in West Africa” (Graham, 2001). Here, the author uses network methods as a tool to measure dialectal variation as he models social relations as lexical similarity. As there are no real social network data involved, the original idea of SNA in variationist’s studies is somehow inverted. In a sociolinguistic survey, Language Use in Ethiopia from a Network Perspective (Meyer & Richter, 2003), the authors claim that higher education students show “manifold network relations.” (Meyer & Richter, 2003, p. 45). This is deduced from the fact that students leave their original homesteads to attend university cities and therefore are per definition in a ‘weak-tie’ situation, leading to multiple contacts and relations. Moreover, the cities are understood as innovative hotspots that crystallize out of the agglomeration of loosely knit actors, which adds in turn, to their ever growing attraction. However, no empirical data to support such claims are produced.

4.1 SNA Studies in Africa

The first study employing the social network approach in a more straightforward way is Joan Russell’s (1981) “Communicative Competence in a Minority Group. A Sociolinguistic Study of the Swahili-Speaking Community in the Old Town Mombasa” The author compares two Swahili varieties that are spoken side by side in the urban center of Mombasa, in Kenya. While Kimvita is the dialect traditionally spoken in the Old Town of Mombasa, the nationwide standard Swahili is taught in schools and is prevalent among younger people and newcomers. Russell reduces the dialectal difference to two phonological markers: /t̪/ and /n̪d̪/ of Kimvita, as opposed to /č/ and /nǰ/ of standard Swahili. Being an exclusive feature of the Old Town variety, the Kimvita phonemes are considered a signal for “Afro-Arab Mombasa Swahili-speaking group membership, and of identity with that group” (Russell, 1981, p. 74). Relying on her acquaintance with the social fabric in the city, she is furthermore convinced “that socio-economic class, even if this could have been established with any accuracy, would be of far less importance as a determinant of linguistic variation than age, sex, style of discourse, or status (insider vs. outsider) of interlocutor” (Russell, 1981, p. 64). On this backdrop, the local social networks are conceived as potential channels of communication and therefore as loci of linguistic variance (Russell, 1981, p. 66).

However, while her linguistic data from 24 adults and 12 pupils rely on a variety of text genres and interviews, the concomitant social network is not established in a systematic way. The SNA is only used for the selection of the subjects for the study as being part of the closed and multiplex network of old-established families in Old Town Mombasa. There is neither an objective characterization of the entire network nor of individuals’ networks or positions therein (Russell, 1981, p. 66ff.). The study, therefore, is only a first approximation to the potential of SNA. Its results show how the standard Swahili pronunciation is gaining ground among Kimvita speakers, showing that mostly elderly women are its last strongholds (Russell, 1981, p. 107).

Salami (1991) employs SNA in a study of variation in the Yoruba of Ile-Ife, Nigeria. He establishes the ego networks of 31 women and 39 men and searches for correlations with linguistic markers that are measured against the backdrop of standard Yoruba (Salami, 1991, p. 218ff.). This standard variety is opposed to three main dialects, which are, like the standard, very well described in the literature. He finally analyzes four phonetic variables that all show at least one variant known to be stigmatized as backcountry in the city (Salami, 1991, p. 221ff.). His hypothesis follows the Milroy thesis regarding vernacular speech norms (Section “History of SNA”): “the predominant use of a local and/or nonstandard dialect variant correlates with the greater degree of the individual speaker's integration into local community networks” (Salami, 1991, p. 224). The individual’s network relations are researched with a questionnaire that covers the domains of relatives, neighborhood, friendship, work, and church communities. From the answers, the author derives figures characterizing the individual networks in terms of density, closeness, and multiplexity, which are aggregated in turn into an individual network score for each respondent (Salami, 1991, p. 225).

Salami’s analysis shows, first of all, that social parameters like age, gender, and education do not correlate with individuals’ network scores, thus proving that integration into the local community is independent from these attributes. In terms of correlation with linguistic markers, he finds one gender-specific variant and some correlations between the linguistic variables, the town quarter, and the interlocutors’ geographic family background (Salami, 1991, p. 230f.). Two of the variables show clear-cut correlations with the network score. The stigmatized [s] of the /S/ variable and the stigmatized [mɔɔ] of the {maa} variable both correlate positively with high network scores of women (Salami, 1991, p. 233f.). There are, however, some more linguistic markers for good integration into local community networks that only correlate with specific age and gender classes. Salami concludes that specific linguistic variants have a double function in Ile-Ife: they signal membership in a specific social group and at the same time they set the norm of pronunciation in the related community networks. He also states that the linguistic variables vary in terms of their normative power. While the speakers are aware that some elements do signal specific group memberships, other variants are produced below the level of consciousness. Moreover, Salami sees no confirmation for the generalization from Western studies (Labov, 2001; Milroy, 1985; Trudgill, 1978) that women tend to use standard or prestige forms while men stick to the conservative forms. In Ile-Ife, as in Mombasa, it seems that women stick to the local vernacular norm more than men (Salami, 1991, p. 242). This different view on gendered speech behavior is a direct consequence of the ethnographically underpinned SNA application. Unlike a more standard approach to sociolinguistics, SNA would expect that the social meaning of gender and linguistic behavior of males versus females varies between communities.

The study suffers from some methodological shortcomings such as the unclear context of the linguistic data gathering and the enigmatic computation of the network score. Moreover, it is at least questionable whether correlations with only four cases (n = 4) are statistically reliable at all (Salami, 1991, pp. 236–237). Notwithstanding this, the study shows that sociolinguistic research in Africa is fruitfully enlarged with the concept of social networks, and even more so when no other measures of social stratification are available.

There are currently ongoing some very promising projects that use SNA as part of their toolkit. They are all connected to multilingual contexts and the questions of contact-induced language development (see Section “Outlook”).

4.2 The Pana-Samo Project

The most recent research project using SNA methodology also targeted contact-induced variation in a multilingual contact zone in West Africa. In this research (Beyer, 2010, 2014, 2015; Beyer & Schreiber, 2013; Schreiber, 2009, 2014), the authors used SNA to throw light on ongoing language contact processes and also set out to evaluate the adaptability of SNA-methodology to predominantly oral societies with low norm enforcing powers and highly dynamic populations. On the backdrop of a rural and longstanding contact area on the border between Burkina Faso and Mali, the research focused on two neighboring village communities that entertain manifold social and economic relations, albeit the dominant languages in both villages are from completely different language families. While one village, Pini, is part of the Northern Samo (Mande) language area, the other, Donon, belongs to the endangered minority of Pana (Gur) speakers.

The linguistic data included various tests of phonological variables, morphological variation, and a set of lexical and discourse elements, all of which are specific to the targeted language types and the linguistic area. In terms of an exploratory strategy, the whole set of potential variables was investigated with different lexical and morpho-syntactic environments and in different text genres (such as word lists, translation tasks, telling of picture stories, free speech). All of the variables were correlated with all kinds of speaker attributes, including SNA measures. As another extension of the variationist approach, linguistic variables were classified as innovation or retention, based on comparative data from the literature. By quantifying the use of an innovation or retention over the whole dataset, an innovation/retention score was calculated for each actor. Furthermore, a local vernacular descriptive norm (DVN) was constructed for each variable or item by taking the distribution pattern in the social network and other sociolinguistic parameters into consideration. The rate of adherence to the so-defined norm, or the deviation from it, were likewise calculated for the whole dataset and transferred into a so-called ‘norm-adherence’ score.

First of all, the Pana-Samo research demonstrated that the method is very well adaptable to rural village environments, under the prerequisite that insights into the local cultures inform the questionnaire. A combination of network questions along basic local categories like neighborhood and kin relations, added up with questions on mutual help and advice networks worked very well. The main body of SNA data produced a network of 41 actors in Pini (a second, larger network was constructed with 86 actors for correlations with a reduced set of variables) and 94 actors in Donon. The derived network representations brought out, on one hand, what was intuitively reported by community members about some individual actor’s positions in the social tissue of the communities and, on the other hand, displayed central hubs, cliques, gate-keepers, but also marginal actors very clearly. It also became clear that the functions of the two village communities are partly identical according to the most influential networks (kin and clan relations), but they also differ in some important aspects, not predictable in the first place.

In Donon, a network score was constructed based on a multitude of criteria (see Beyer, 2010) that clearly correlated with linguistic markers. The correlations underline the connection between good integration into local community networks and closeness to vernacular speech norm. The higher a given speakers network score, the lower the percentage of unusual pronunciations, use of lexical loans, and foreign discourse markers (Beyer, 2015). The research also made clear that a great number of linguistic variations cannot be explained with just one social correlation. It is very often the case that correlations only come out when specific network features and general attributes are connected in a meaningful way.

For example, the proneness of some Samo speakers from Pini to consonant depalatalization (CDEPAL) correlates significantly with specific network measures such as in-eigenvector (InARD = In-eigenvector with Average Reciprocal Distance, r = 0.52) and related eigenvector measures. In SNA terminology, this means that actors with higher numbers of incoming ties from other relatively central actors (i.e., eigenvector) significantly prefer the fricative [ʒ] rather than [dʲ], the palatalized plosive. In sociolinguistic speech this reads as: Socially more influential actors of the speech community prefer the (areal linguistic) innovation [ʒ], while peripheral speakers stick with the conservative pronunciation.

Moreover, the data also show a significant relation (r = 0.4) of CDEPAL with the Pana affinity score. This score describes actors in the Samo speaking village of Pini who have strong relations with Pana speakers from neighboring villages. The correlation signals that speakers with high Pana affinity produce the fricative less frequently. This statistical result suggests an instance of contact-induced variation. Though, it is hard to decide whether Pana L1 speakers impose their interference on Samo speakers related to them, or whether these Samo speakers stick to the same variant in terms of retention as suggested by the above cited statistics. In any case, the social network structure seems to prevent a (re-)diffusion of [dʲ] to the whole speech community as actors with strong Pana affiliations and the ones showing a high network score of ‘betweenness’ favor the fricative release. In this sense, it can be concluded that the network factor is more dominant than the language contact one (Figure 9).

Social Network Approach in African SociolinguisticsClick to view larger

Figure 9. Distribution of the depalatalization variable in the Pini sample network (frequency of CDEPAL rising from light to dark blue and black)

In this sense, SNA provides empirical arguments that contact-induced change is not necessarily an occurrence of interference alone but relates to a socially driven actuation of diffusion in the speech community.

Another insight from the Pana-Samo project is that variables from virtually all parts of the linguistic system may become markers of identity. For instance, the above mentioned foreign discourse markers (highly correlated with low network score speakers), which are at first unconsciously and accidentally used by some marginal actors, seem to spread through the network and thus are on the way to become general means of discourse structure in the Donon community (Beyer, 2014, 2015). The same is true for some morphological features.

Schreiber (2014) describes the choice of borrowed discourse markers in the Pini community as depending on personal style and attributes, the communicative context, and again, an individual actor’s network roles and positions. Based on social data and personal biographies, the analysis shows that no relation exists between linguistic competence in the donor languages and usage of borrowed discourse markers. In the Pini case, the use of French and Jula discourse markers is not related to the actual personal repertoire (Figure 10).

Social Network Approach in African SociolinguisticsClick to view larger

Figure 10. Frequency of use of the discourse marker “bon” in the Pini sample network (black circle = frequent use).

On the basis of a descriptive vernacular norm (DVN) (see Section “The Pana-Samo Project”), the position of different speaker types in the network becomes visible. Based on the qualitative analysis of SNA, cliques of innovative speakers are identified in the Pini network (Figure 11).

Social Network Approach in African SociolinguisticsClick to view larger

Figure 11. Distribution of speech behavior in the Pini sample network (black circle = innovative speakers).

In such a way, classified variables help to test hypotheses on the relation of speech community network attributes and contact-induced variation by correlating network-based coherence and heterogeneity scores with a linguistically motivated variation score and empirical innovation rate.

5. Outlook

Given the prevailing dynamic multilingual conditions, most frequently on the backdrop of less standardized languages, SNA is a promising framework for variation studies in Africa due to its methodological advantages. It is highly flexible and, when correctly implemented, produces a wealth of real, live data that allow in-depth analysis of current processes of language variation and change that would be hard to obtain with the classical variationist research design.

Currently, at least two ongoing research projects in Africa implement SNA as part of their methodological toolkit. Both projects are situated in rural contexts where multilingualism is ubiquitous and deeply entrenched in everyday communication. In such contexts of organic multilingualism, SNA applications are very promising in order to disentangle speakers’ actual variations in the use of their highly diversified repertoires and the repercussions this may have on language developments and linguistic norm building. Both the London based ‘Crossroads’ project and the Buffalo based ‘Lower Fungom’ project currently gather SNA data in their respective areas, the Senegal province of Casamance and the northwestern parts of the Cameroonian Grassfields. Together with further planned research and the ongoing SNA data analysis from the Pana-Samo project, a first step to a comparative appraisal of SNA data from non-Western context is on the way.

Albeit SNA is demanding in terms of implementation and data handling, the rich data mining possibilities and the deep insights in a given speech community are worth the effort. Moreover, recent developments in software and data science allow the treatment of far more data than in the early times of SNA. However, apart from methodological and technical requirements, a further potential obstacle is the pace of theoretical development in SNA. Since its introduction to sociolinguistics, various new measures and statistical techniques have been developed by the fast growing SNA community. Receiving this vast amount of recent literature and testing new concepts is also a challenge for SNA application in sociolinguistics, but—again—worth the effort.

Further Reading

Borgatti, S. P., Everett, M. G., & Johnson, J. C. (2013). Analyzing social networks. Los Angeles, CA: SAGE.Find this resource:

Fagyal, Z., Swarup, S., Escobar, A. M., Gasser, L., & Lakkaraju, K. (2010). Centers and peripheries: Network roles in language change. Lingua, 120, 2061–2079.Find this resource:

Gal, S. (1979). Language shift: Social determinants of linguistic change in bilingual Austria. New York: Academic PressFind this resource:

Granovetter, M. S. (1973). The strength of weak ties. The American Journal of Sociology, 78(6), 1360–1380.Find this resource:

Hanneman, R. A., & Riddle, M. (2005). Introduction to social network methods. Riverside: University of California Press.Find this resource:

Jansen, D. (1999). Einführung in die Netzwerkanalyse. Grundlagen, Methoden, Anwendungen. Opladen, Germany: Leske + Budrich.Find this resource:

Lippi-Green, R. (1989). Social network integration and language change in progress in an alpine rural village. Language in Society, 18, 213–234.Find this resource:

Milroy, J., & Milroy, L. (1985). Linguistic change, social network and speaker innovation. Journal of Linguistics, 21, 339–384.Find this resource:

Milroy, L. (1980). Language and social networks. Language in society, 2. Oxford: Blackwell.Find this resource:

Milroy, L., & Gordon, M. (2003). Sociolinguistics: Method and interpretation. Malden, MA: Blackwell.Find this resource:

Milroy, L., & Llamas, C. (2013). Social networks. In J. K. Chambers & N. Schilling (Eds.), The handbook of language variation and change (2d ed., pp. 409–427). Malden, MA, Oxford: Wiley-Blackwell.Find this resource:

Wassermann, S., & Faust, K. (1994). Social network analysis: Methods and applications. Cambridge, U.K.: Cambridge University Press.Find this resource:

References

Austin, J. L. (1962). How to do things with words. Oxford: The Clearendon Press.Find this resource:

Barnes, J. (1954). Class and committees in a Norwegian island parish. Human Relations 7, 39–58.Find this resource:

Beyer, K. (2010). Language contact and change: Social factors in an African rural environment. Journal of Language Contact, 3, 131–152.Find this resource:

Beyer, K. (2014). Multilingual actors: Examples from a West-African contact zone. In de Féral, C., Kossmann, M. G., & Tosco, M. (Eds.)̩, In and out of Africa. Languages in Question. In Honour of Robert Nicolaï. Language Contact and Language Change in Africa (Vol. 2, pp. 15–35). Leuven: Peeters.Find this resource:

Beyer, K. (2015). Multilingual speakers in a West-African contact zone. An integrated approach to contact-induced language change. In G. Stell, & K. Yakpo (Eds.)̩, Code-switching at the crossroads between structural and sociolinguistic perspectives (pp. 237–258). New York: de Gruyter.Find this resource:

Beyer, K., & Schreiber, H. (2013). Intermingling speech groups: Morpho-syntactic outcomes of language contact in a linguistic area in Burkina Faso, West Africa. In I. Léglise & C. Chamoreau (Eds.)̩, The interplay of variation and change in contact settings (pp. 107–134). Amsterdam, The Netherlands: Benjamins.Find this resource:

Bonacich, P. (1987). Power and centrality: A family of measures. American Journal of Sociology, 92(5), 1170–1182.Find this resource:

Borgatti, S. P., Everett, M. G., & Freeman, L. C. (2002). UCINET for Windows: Software for social network analysis. Version 6. Lexington, KY: Analytic Technologies.Find this resource:

Borgatti, S. P., Everett, M. G., & Johnson, J. C. (2013). Analyzing social networks. Los Angeles, CA: SAGE.Find this resource:

Bourdieu, P. (1991). Language and symbolic power. Cambridge, MA: Polity Press.Find this resource:

Coupland, N., Sarangi, S., & Candlin, C. (2001). Sociolinguistics and social theory. Language in social life series. New York: Longman.Find this resource:

Danowski, J. A., Gluesing, J., & Riopelle, K. (2011). The revolution in diffusion theory caused by new media. In A. Vishwanath & G. A. Barnett (Eds.), The diffusion of innovations (pp. 123–144). New York: Peter Lang.Find this resource:

Fagyal, Z., Swarup, S., Escobar, A. M., Gasser, L., & Lakkaraju, K. (2010). Centers and peripheries: Network roles in language change. Lingua, 120, 2061–2079.Find this resource:

Friemel, T. N. (2010). Diffusionsforschung. In: C. Stegbauer & R. Häußling (Eds.), Handbuch netzwerkforschung (pp. 825–834). Wiesbaden, Germany: VS Verlag für Sozialwissenschaften.Find this resource:

Gal, S. (1979). Language shift: Social determinants of linguistic change in bilingual Austria. Academic Press.Find this resource:

Graham, S. (2001). A look at the acts of identity theory through a social network analysis of Portuguese-based creoles in West Africa. Journal of Pidgin and Creole Languages, 16(1), 1–51.Find this resource:

Granovetter, M. S. (1973). The strength of weak ties. The American Journal of Sociology, 78(6), 1360–1380.Find this resource:

Habermas, J. (1981). Theorie des kommunikativen Handelns. Frankfurt, Germany: Suhrkamp.Find this resource:

Hall, B. H. (2006). Innovation and diffusion. In J. Fagerberg & D. C. Mowery (Eds.), The Oxford handbook of innovation (pp. 459–485). New York: Oxford University Press.Find this resource:

Hanneman, R. A., & Riddle, M. (2005). Introduction to social network methods. Riverside: University of California Press.Find this resource:

Hannerz, U. (1992). Cultural complexity. Studies in the social organization of meaning. New York: Columbia University Press.Find this resource:

Huber, F. (2007). Social networks and knowledge spillovers. Networked knowledge workers and localised knowledge spillovers. Frankfurt, Germany: Lang.Find this resource:

Jansen, D. (1999). Einführung in die Netzwerkanalyse. Grundlagen, Methoden, Anwendungen. Opladen, Germany: Leske + Budrich.Find this resource:

Kapferer, B. (1969). Norms and the manipulation of relationships in a work context. In C. Mitchell (Ed.), Social networks in urban situations (pp. 181–244). Manchester: University of Manchester Press.Find this resource:

Kapferer, B. (1972). Strategy and transaction in an African factory. Manchester: Manchester University Press.Find this resource:

Ke, J., Gong, T., & Wang, W. S.-Y. (2008). Language change and social networks. Communications in Computational Physics, 3(4), 935–949.Find this resource:

Labov, W. (1969). Contraction, deletion, and inherent variability of the English copula. Language, 45, 715–762.Find this resource:

Labov, W. (1972). Language in the inner city. Studies in the Black English vernacular. Philadelphia: University of Pennsylvania Press.Find this resource:

Labov, W. (2001). Social factors. Language in society, 29. Malden, MA: Blackwell.Find this resource:

Lippi-Green, R. L. (1989). Social network integration and language change in progress in a rural social network integration and language change in progress in a rural alpine village. Language in Society, 18, 213–234.Find this resource:

Luhmann, N. (1984). Soziale systeme: Grundriss einer allgemeinen Theorie. Frankfurt, Germany: Suhrkamp.Find this resource:

Marshall, J. (2004). Language change and sociolinguistics. Rethinking social networks. Palgrave studies in language variation. New York: Palgrave Macmillan.Find this resource:

Meyer, R., & Richter, R. (2003). Language use in Ethiopia from a network perspective. Frankfurt, Germany: Peter Lang.Find this resource:

Milroy, J., & Milroy, L. (1978). Belfast: Change and variation in an urban vernacular. In Peter Trudgill (Ed.), Sociolinguistic patterns in British English (pp. 19–36). London: Arnold.Find this resource:

Milroy, J., & Milroy, L. (1985). Linguistic change, social network, and speaker innovation. Journal of Linguistics, 21, 339–384.Find this resource:

Milroy, L. (1980). Language and social networks. Language in society, 2. Oxford: Blackwell.Find this resource:

Nettle, D. (1999). Linguistic diversity. Oxford: Oxford University Press.Find this resource:

Ngom, F. (2006). Lexical borrowings as sociolinguistic variables in Saint-Louis, Senegal. Munich, Germany: Lincom Europa Academic.Find this resource:

Owens, J. (1998). Neighborhood and ancestry: Variation in the spoken Arabic of Maiduguri (Nigeria). Impact: Studies in Language and Society Nr. 4. Amsterdam, The Netherlands: John Benjamins.Find this resource:

Radcliffe-Brown, A. R. (1940). On social structure. The Journal of the Royal Anthropological Institute of Great Britain and Ireland, 70(1), 1–12.Find this resource:

Rogers, E. M. (1962). Diffusion of innovations. New York: Free Press of Glencoe.Find this resource:

Ross, M. (1997). Social networks and kinds of speech-community event. In R. Blench & M. Spriggs (Eds.), Archaeology and language (pp. 207–261). One world archaeology, 27. London: Routledge.Find this resource:

Russell, J. (1981). Communicative competence in a minority group: A sociolinguistic study of the Swahili-speaking community in the Old Town, Mombasa. Leiden, The Netherlands: Brill.Find this resource:

Salami, O. L. (1991). Diffusion and focusing: Phonological variation and social networks in Ile-Ife, Nigeria. Language in Society, 20(1), 217–245.Find this resource:

Schnegg, M. (2010a). Die Wurzeln der Netzwerkforschung. In C. Stegbauer & R. Häußling (Eds.), Handbuch Netzwerkforschung (pp. 1–8). Wiesbaden, Germany: VS Verlag für Sozialwissenschaften.Find this resource:

Schnegg, M. (2010b). Ethnologie. In C. Stegbauer & R. Häußling (Eds.), Handbuch Netzwerkforschung (pp. 869–968). Wiesbaden, Germany: VS Verlag für Sozialwissenschaften.Find this resource:

Schönhuth, M., Kronenwett, M., Gamper, M., & Stark, M. (2014). VennMaker. Version 1.4.0.

Schreiber, H. (2009). Social networks, linguistic variation, and micro change in an African context. A case study in the borderland of Mali and Burkina Faso. Sprache und Geschichte in Afrika, 20, 209–230.Find this resource:

Schreiber, H. (2014). Imports and exports on linguistic markets in the West African Sahel. In C. de Feral, M. Kossmann, & M. Tosco (Eds.), Language contact and language change in Africa: Hommage à Robert Nicolai (Vol. 2, pp. 251–268). Paris: Presses Universitaires de l’École Normale Supérieure.Find this resource:

Schweizer, T. (1989). Netzwerkanalyse. Ethnologische Perspektiven. Berlin: Reimer.Find this resource:

Schweizer, T. (1996). Muster sozialer Ordnung. Netzwerkanalyse als Fundament der Sozialethnologie. Berlin: Reimer.Find this resource:

Searle, J. R. (1994). Intentionality. An essay in the philosophy of mind. Cambridge, U.K.: Cambridge University Press.Find this resource:

Trudgill, P. (1978). Sociolinguistic patterns in British English. London: Arnold.Find this resource:

Valente, T. W. (1995). Network models of the diffusion of innovations. Cresskill, NJ: Hampton.Find this resource:

Valente, T. W. (2005). Network models and methods for studying the diffusion of innovations. In P. J. Carrington, J. Scott & S. Wasserman (Eds.), Models and methods in social network analysis (pp. 98–116). Cambridge, U.K.: University Press.Find this resource: