Modeling language shift.

It is estimated that there are as many as 6,000 distinct languages currently spoken, but this cultural diversity is rapidly disappearing. In terms of the fraction of the total being lost, the estimated current rate of language extinction exceeds the rate of loss of biodiversity (1–3). Small geographical ranges and small speaker populations, especially in parts of the world experiencing high economic growth, are among the strongest indicators of language extinction risk (4). Most of the recent language extinction events are caused by language shift rather than the extinction of the population speaking this language (5). Language shift is the process whereby members of a community in which more than one language is spoken abandon their original vernacular language in favor of another. Especially in language contact situations, people are confronted with choices about which language to speak. Processes of globalization, urbanization, and long-distance economic migration, have led to increased interactions between groups speaking different languages, and therefore to a need of a common language of communication. Historically, in situations of warfare, disaster, or repression, the dominant language may be forcibly imposed as a deliberate policy of its native speakers. In other cases, inequality may be expressed less directly. Often the language seen as more modern, useful, or giving access to greater social mobility and economic opportunities is chosen as the lingua franca, thereby driving the process of shift (6). Research into language shift has been conducted in a variety of scientific disciplines, from linguistics and sociolinguistics to economics and physics. In one approach, language shift is modeled as a competition between linguistic communities for speakers, analogous to the competition dynamics between interacting species in ecology (but with speakers capable of “switching species”). These models make the simplifying assumption that languages are fixed entities, and that only speaker frequencies in the population can vary. Prochazka and Vogl (7) contribute to this literature by developing a cellular automata framework to model language shift on a fine spatial scale. Linguists recognize that frequency of use, proficiency, and perceived value are among the most important immediate factors affecting persistence or endangerment of minority languages (8), and Prochazka and Vogl’s (7) model assumes that such factors are captured by a measure of the local abundance of speakers. The authors show that neighborhood concentrations of speakers of the minority language are good predictors of its persistence (or of a slowing of its rate of decline). Modeling enables us to gain a better understanding of the process of language shift, by identifying demographic, socioeconomic, cultural, and linguistic processes that can generate spatial and temporal patterns similar to those seen empirically. The comparison of theory with data is one of the crucial components of this literature, although this does not remove the problem of equifinality (especially where the data are sparse). Subsequent formal analysis of the models can also shed light on possible future outcomes of language shift. Additionally, models can be used as an artificial experiment. In the past, serious concerns over the loss of linguistic diversity have driven governments and international organizations to actively engage in the maintenance of endangered languages (1, 4, 9). Models can provide useful information about the potential success of intervention strategies; for example, they can inform about the total intervention strength needed to achieve the desired goal of language planners (10). Modeling frameworks describing language shift can be roughly divided into two categories: agentbased and equation-based approaches. By simulating specific rules about individual behaviors, agent-based models often focus on elaborating the necessary interaction and transmission mechanisms (e.g., refs. 11–13), whereas by transforming these behaviors into mathematical principles, equation-based models usually concentrate on how to describe the shift dynamics at the population level (14). Following an influential paper by Abrams and Strogatz (15), most of the recent approaches have been equation-based. These studies have emphasized the importance of bilingual or diglossic