Language extinction: it's real, it's serious, and it's hard (but getting easier) to assess

In response to last month's Ethnoblog post, a reader commented that the topic of language extinction and the way Ethnologue reports on it is confusing.  The answer to that comment merits more than a quick reply, so I'm making it the topic of this month's post. It also means that this post is going to be a bit longer than usual.

First off, I should clarify that many of the assertions about language endangerment and language extinction that are being more and more reported in the press didn't originate with the Ethnologue.  Not that we don't believe that the situation is serious, but credit should be given where credit is due. It wasn't us.  Rather, in the early 1990s a major session at the Linguistic Society of America conference was dedicated to the topic of language endangerment and several well-known and very knowledgeable linguists called for a concerted response and focusing of attention on the plight of many lesser-known languages that were on the verge of disappearing. The estimate was that only 10% of languages seem safe in the long term, up to 50% may already be moribund, and the remainder are in danger of becoming moribund by the end of this century.  That evaluation was based on the best sources available at the time (which included the Ethnologue) but it also focused primarily on the locations where language death was most obviously and seriously occurring--Australia and the Americas. Other parts of the world weren’t adequately considered and, as it turns out, not in such dire straits as the general warning seemed to imply. Gary Simons, our executive editor, and I have written a paper "The world's languages in crisis: a 20-year update," which cites the original discussion in the 1990s and then uses data from Ethnologue to document the current situation (online preprint here).

The Ethnologue had for many years categorized languages as "living,"  "nearly extinct," and "extinct" (along with a couple of other categories) but as interest in language endangerment grew that set of categories was too broad and not very helpful. Simply looking at the total number of languages in each of those categories didn’t provide us with enough data about whether specific languages were expanding or shrinking. Each year the ISO 639-3 process, for example, identifies some new languages either as the result of discovering new languages through field work or because further study of a particular language and its variants seems to warrant that one or more of those variants be split off and identified as a separate language. That increases the total count of living languages, but without knowing anything about the vitality of those newly identified languages, we can’t really say much about the endangerment trends.

Because many scholars were looking for that kind of data, we developed and began to use the EGIDS, our measure of language endangerment and development, as a more fine-grained way to describe the status of each language. It is a fairly quick and easy framework that can be applied to all languages in the world and so has facilitated the kind of global analysis that can support or disprove those earlier claims. In each Ethnologue language entry, the EGIDS assessment is included under the label "Status," and an overview of the EGIDS profile of all of the languages in a country is provided, as a bar graph, on each country page.

The EGIDS gives us a way to begin to track the trends of language endangerment, but since we have only been using it for a few years, we don’t have enough history at this point to be able to describe a rate of language loss based solely on the EGIDS status numbers.  There has been some work done by others which takes historical Ethnologue data and analyzes it with a view towards seeing the patterns of language diversity loss over the decades since Ethnologue began. That is a noble pursuit, but one difficulty with that approach is that there are changes to the inventory of languages from one edition to another that may be for reasons other than language death.  It may not be entirely straightforward for readers to determine why a language has disappeared from the Ethnologue’s list of living languages though this process is documented in the ISO 639-3 updates. Within a few more years and with adequate feedback and prompt updating of the EGiDS numbers, we may be able to say more about that ourselves.

And that points out another reason why our reporting may seem to be a bit muddled.  We publish what we know each year at the time of publication, incorporating changes from our ongoing effort to elicit updates and corrections.  While we make tens of thousands of changes with each edition, there are many language entries that remain unchanged.  It seems to be especially the case that our data on extremely endangered languages doesn’t get updated in a timely manner, so it may well be (and we wish it weren’t so) that some of the languages which we are identifying as very weak or which have very few remaining users have already gone out of use.

With the implementation of the EGIDS, we also implemented a major change in our “rules of accounting” and re-classified a number of languages which previously had been considered “Extinct” (EGIDS 10) as “Dormant” (EGIDS 9). Dormant languages are those which no longer are used for day-to-day activities but which still have a strong association with a group’s ethnic identity. They are used in very limited ways for ceremonial or symbolic purposes but usually those who use the language are reciting memorized language forms and don’t have sufficient proficiency to use the language for daily life.  With that one-time change, we began to include those “sleeping” languages in our counts of living languages.  That, of course, resulted in a sudden increase in the number of languages, which though extremely endangered, are still living for statistical purposes.

And one more factor to consider is that for very practical reasons, we only track and report on languages that have gone out of use since 1950.  The Ethnologue began in 1952, so that threshold represents the earliest scope of our research. Many languages became extinct in the years prior to 1950 and we don’t list them nor include them in our tallies of the various EGIDS levels. The result is that our profiles of language status by country, especially in the parts of the world that are most prone to language shift and death, are somewhat rosier than they might be if, for example, we included all of the languages that were alive and well in North America or Australia at the time of first contact with outsiders.

We are optimistic that, as we continue to collect vitality data over a longer period of time, some clear trends will emerge. As always, we encourage “those who know” to provide feedback on our data either via our Contact Us page, or better yet, by registering as an Ethnologue user and giving direct feedback on a particular language or country. That helps us to present as clear and accurate a picture as possible.


Submitted by Mark Szymcik on Fri, 2015-09-04 19:06
Thanks for the explanation and clarifications. I'm glad Ethnologue is not buying into e.g., 50% of living languages will definitely be gone in 50 or 75 or whatever years.
Submitted by linguistic_serv... on Thu, 2015-10-15 13:09
Actually, Ethnologue isn't trying to make any predictions at all, but with the EGIDS system simply trying to document current vitality. That's a common source of confusion in discussion of endangered languages--the difference between a descriptive and predictive perspective. For some people, "endangered" means that the language is already dying, because children are no longer learning it (EGIDS 6b-8). Others use the term to mean that the language might reasonably be expected to go extinct in the next 100 years. In other words, some languages that are ranked by EGIDS as "vigorous" (or stronger) may still be considered endangered when considering the future. But, again, we're not trying to make predictions as to how many there may be.