Language change in social structure
Luca Onnis and Matthew Lou-Magnuson
Nanyang Technological University, Singapore
Human behaviour, including language, is highly social and occurs within communities organised in stratified social networks at difference scales. Social networks are composed of a “web of ties” between individuals, and the structure of a network will vary depending on the types of connections it is composed of. Linguists have long posited that the structure of social networks, and the interactions between members within the networks, are a driving force behind language change, maintenance, and loss. Yet the specific mechanisms whereby the social context of learning and use affects the evolutionary trajectory of a language over time and geographical spread are hard to elucidate. Most of the research so far is descriptive at best, and different anecdotal claims concerning a given relationship between the history of some particular people, and the grammar of the language they speak, for example, cannot be easily subjected to systematic comparison and falsifiability.
Several practical problems in the study of language change in social networks seriously limit a deeper understanding at present: first, structural changes such as grammaticalisation patterns occur over long timeframes across generations of speakers, and as such are impossible to fully record; fundamental changes in extant languages began centuries or millennia before the beginning of their historical records. At best one can hypothesise current trends that are indicative of possible “phase transitions”, e.g. in patterns of Singlish change across generations of speakers.
Second, the fine-grained social structure of the community that produced a linguistic phenomenon is hardly recoverable. Even in fortunate cases when rich primary data are available, as in historic “longitudinal” corpora, these repositories collapse across the individual sources and are unable to reconstruct the network connectivity of the community that generated them. Thus, they can at best describe the overall frequency trajectory of use of a certain linguistic construction from, say Time 1 to Time 2. In this “grand-averaging”, the reconstructive processes and mechanisms that lead to that particular change are left to the linguist’s final guesswork, and cannot be linked to “primary data” about the social structure that generated it.
A third related practical problem is that structural changes occur over many interactant speakers, and so far it has not been possible to obtain indexed linguistic data from hundreds, if not thousand of speakers of a given community. Even if it was, it would be impractical or unethical to manipulate hundreds or thousands of peoples’ linguistic behaviours in an attempt to study mechanisms of global language changes (although see Bond et al., 2012).
Recent technological advances and new formalisms, however, have seen a new Network Science (NS) develop, and provide insights into the emergence of global, system-scale properties. Crucially, NS offers metrics, formalisations, and computer implementations of social networks that potentially allow a quantitative and qualitative analysis of language behaviours of real and simulated communities of speakers. Here we discuss how concepts and tools from Network Science can become increasingly relevant to the study of language change in relation to social structure. In particular, we first show that agent-based computer simulations can provide explorations of possible language trajectories, testing of possible causes and effects, and proof-of-concepts for hypothesised processes of language change. For example, Lou-Magnuson & Onnis (submitted), provided the first causal explanation of the long standing correlation between social structure and morphemic complexity.
A second method leveraging on network science is a new breed of human simulations of iterated learning (IL). Language is transmitted and changes via a process of cultural evolution iterated over generations. At each generation, a learner must observe the linguistic behaviour of others, and use their endogenous abilities to discover a grammar that accounts for this input. It is possible to adapt IL and incorporate both vertical cultural transmission over time, and horizontal transmission in terms of negotiation of conventions between peers in a community of learners. We do so by creating a multiplayer online language learning game (LLG). Human simulations can help narrow down the set of computer generated hypotheses, and provide experimentally controlled evidence that in principle certain aspects of the structure of a society can impact language trajectories.
A third approach strives to obtain higher ecological validity and scalability for the previous methods. It involves extracting publicly available patterns of human relations from volunteers’ online social networks (Facebook, Twitter, etc.). Social networks analyses of individuals can then be used as a predictor of the specific language uses of participants.
We hope to encourage a discussion of how these methods can be brought to fruition in combination with existing methods and findings from historical linguistics and the grammaticalisation literature. We dub Network Linguistics the new cross-disciplinary area that can arise (Onnis, Lou-Magnuson, & Oh, in prep.), and argue that it can provide an unprecedented opportunity to bridge the social and diachronic aspects of language with the psycholinguistic behaviour of individual speakers.
References:
Bond, R. M., Fariss, C. J., Jones, J. J., Kramer, A. D., Marlow, C., Settle, J. E., & Fowler, J. H. (2012). A 61-million-person experiment in social influence and political mobilization. Nature, 489(7415), 295-298.
Lou-Magnuson, M. & Onnis, L. (submitted). Social networks limit language complexity.
Onnis, L., Lou-Magnuson, & Oh, P. (in preparation). Language change in social structure: The birth of Network Linguistics.