On April 17, 2009, the Administrator signed proposed endangerment and cause or contribute findings for greenhouse gases under Section 202(a) of the Clean Air Act. EPA held a 60-day public comment period, which ended June 23, 2009, and received over 380,000 public comments. These included both written comments as well as testimony at two public hearings in Arlington, Virginia, and Seattle, Washington. EPA carefully reviewed, considered, and incorporated public comments when issuing the final Findings on December 7, 2009.
Language diversity is under threat. While each language is subject to specific social, demographic and political pressures, there may also be common threatening processes. We use an analysis of 6,511 spoken languages with 51 predictor variables spanning aspects of population, documentation, legal recognition, education policy, socioeconomic indicators and environmental features to show that, counter to common perception, contact with other languages per se is not a driver of language loss. However, greater road density, which may encourage population movement, is associated with increased endangerment. Higher average years of schooling is also associated with greater endangerment, evidence that formal education can contribute to loss of language diversity. Without intervention, language loss could triple within 40 years, with at least one language lost per month. To avoid the loss of over 1,500 languages by the end of the century, urgent investment is needed in language documentation, bilingual education programmes and other community-based programmes.
endangerment
Our analysis seeks the best set of variables, from 51 candidate variables, to explain variation in endangerment level (the dependent variable), over and above covariation due to relationships between languages, spatial autocorrelation and contact between language distributions, and allowing for interactions between predictor variables and region. We reduced the number of variables by grouping variables according to their pairwise correlations, identified independent variables with significant predictive power on a proportion of the data (training dataset), then evaluated the fit of the model on the remaining data (test dataset). We then estimated model parameters on the full dataset (see Methods for details).
Five predictors of language endangerment are consistently identified at global and regional scales: L1 speakers, bordering language richness, road density, years of schooling and the number of endangered languages in the immediate neighbourhood. Each of these predictors highlights a different process in language endangerment; taken together, they paint a picture of the way interactions between languages shape language vitality.
Number of first-language (L1) speakers is the greatest predictor of endangerment. It is important to emphasize that not all small languages are endangered, and that language loss does not necessarily result from a reduction in number of people in a particular culture or population, but often occurs when people shift from using their heritage language to a different language1,30. Therefore the multilingual setting in which each language is embedded (referred to as the language ecology) plays a key role in endangerment, by influencing whether speakers shift to another language or adopt additional languages in their multilingual repertoires31. Our results suggest that direct contact with neighbouring languages, as reflected in the number languages with overlapping or touching distributions, is not in itself a threatening process. In fact, languages whose distributions are directly in contact with a greater number of other autochthonous languages have lower average endangerment levels (Fig. 1). This may reflect a common observation that communities in regular contact with speakers of other Indigenous languages may be multilingual without necessarily giving up their L1 language31. If ongoing language contact was a threat to language vitality, then we might expect that more isolated languages, such as those on islands, would be less endangered, but this is not the case (Supplementary Fig. 7). Similarly, we find no evidence that barriers to human movement that might be expected to reduce contact between nearby speaker populations, such as steep or rough terrain, are associated with reduced endangerment. We conclude that being in regular contact with speakers of another language does not in itself usually endanger Indigenous language vitality. Instead there are other more complex social, economic and political dynamics influencing language endangerment that may co-occur with language contact but are not synonymous with it.
There is consistent global support for higher average levels of schooling being associated with greater language endangerment (Fig. 1). The association between schooling and language endangerment cannot be interpreted as a side effect of growing socioeconomic development, because years of schooling is a much stronger predictor of endangerment patterns than other socioeconomic indicators. Instead, it is consistent with a growing number of studies showing a negative impact of formal schooling on minority language vitality, particularly where bilingual education is not supported or, in some cases, is actively discouraged38,39,40. Yet having a minority education policy is not globally associated with reduced threats to language diversity, possibly due to variation in the extent and manner of provision of bilingual education for speakers of minority languages. For example, the Bilingual Education Act of the United States (1968) was primarily concerned with improving access to mainstream education for students from non-English speaking backgrounds by using heritage language as a bridge to English acquisition, rather than being designed to allow students to maintain their first language41.
If a language is no longer being learned by children, we can use demographic information to predict when, in the absence of interventions to increase language transmission, there will be no more living L1 speakers. We can combine the current L1 speaker population size with endangerment score (which tells us the relative generational distribution of L1 speakers and whether the number of L1 speakers is declining; Supplementary Table 1), and use demographic information on age structure of the population (Supplementary Table 8) to predict how many L1 speakers will be alive in the future (see Supplementary Methods 5 for details). Our analysis is conservative in that it only considers change in L1 speakers in languages identified as having reduced transmission to younger generations (see Supplementary Table 1): we did not model change in speaker number for languages currently considered to be stably transmitted, even though they may become endangered in the future.
The crisis of language endangerment has prompted worldwide efforts to recognize, document and support language diversity45, reflected in the UNESCO International Decade of Indigenous Languages, beginning in 2022. Every language represents a unique expression of human culture, and each is subject to idiosyncratic influences of their specific history and local sociopolitical environment. By identifying general factors that impact language vitality, or areas at greatest risk of language loss, we may be better placed to direct resources for maintenance of language diversity.
We emphasize that these predictions are not death knells, but possible outcomes in the absence of investment in language vitality. For example, while our model predicts Alutiiq (Pacific Gulf Yupik ems) in Alaska to increase in endangerment level, the community has instituted a language revitalization programme that may counter the predicted trend. Identifying external factors associated with language endangerment can focus attention on areas where language vitality might become threatened. For example, some areas with the greatest predicted increase in road density, such as Nigeria, Papua New Guinea and Brazil48, are predicted by our model to have the highest potential loss of languages (Extended Data Fig. 4). Since increasing road density also has negative impacts on biodiversity, focusing mitigation efforts on areas of increasing road density may be beneficial for both language vitality and biodiversity49,50.
In addition to identifying correlates of language endangerment that are likely to change in the future, such as land use, we also identify factors that are open to intervention to reduce loss of language diversity. Currently, more years of formal schooling are associated with greater rates of language endangerment (Fig. 1). Research suggests that bilingual education, where students learn part or all of the curriculum in their first language, typically results in greater overall academic achievement without sacrificing proficiency in the dominant national language51, but emphasis on high-stakes testing for competency in the national language can contribute to erosion of heritage language proficiency42. Having provision for bilingual education enshrined in legislation, or official recognition of minority languages in government or in education, is not sufficient to reduce language endangerment (Supplementary Fig. 7). Implementation requires genuine commitment to bilingual education, and support from community members who can bring heritage language to the classroom. The benefits of providing support to enhance Indigenous language vitality, in terms of wellbeing52,53 and socioeconomic outcomes54, are likely to far outweigh the costs. Implementation of support for Indigenous language vitality at all levels of governance and within speaker communities is urgent, given the predicted loss of L1 speakers who can aid language vitality and transmission (Fig. 3).
Many previous analyses of global patterns of language endangerment relied on speaker population size and geographic distribution as proxies of endangerment status4,20,66. While low speaker number is the ultimate outcome of endangerment, current population size may not always provide a reliable indicator of language vitality or risk of loss67,68. Small localized languages may be stable and vigorous, for example some Papuan languages are confined to one or a few villages with only hundreds of speakers, yet are not considered endangered (for example, Neko ISO 639-3: nej, Mato met), and large widespread languages are not secure if they are not being reliably transmitted to younger generations (for example, Domari rmt, an endangered Indo-European language with over a quarter of a million speakers). Using population and range size to represent endangerment also conflates endangerment and diversity: range and population size correlate with number of languages per unit area17, so an area with more languages may, all things being equal, also contain a larger number of endangered languages4,20. Our analysis emphasizes global trends and general patterns over specific language trajectories or local histories. Use of global databases provides an overview of language diversity and vitality, but it is not possible to verify current speaker numbers, endangerment and distributions without expert knowledge of each individual language. Some regions or language families may be less well represented in global databases (for example, Australian languages have patchy representation and would benefit from expert revision on speaker numbers and endangerment levels). Furthermore, there is often no clear line between a dialect and a language, and this can result in variation in assigning L1 speakers to languages (Supplementary Methods 2.1.2). Our results should therefore be interpreted as providing general patterns and broad-brush predictions rather than specific detail on particular languages. 2ff7e9595c
Comments