A language is not lost when its last speaker dies but a generation earlier, when children stop it.
By the time the loss is visible, the decisive event is decades in the past.
That is why projects so often arrive in time to record a language and too late to keep it.
stops for reasons that are rarely about the language itself.
Parents make a calculation about which language will serve a child in a school and a labour market.
The calculation is usually correct, which is the part that programmes must address honestly.
Telling a parent that a language is precious does not alter the arithmetic she has already done.
Programmes that have worked changed the arithmetic instead, usually by attaching the language to employment or schooling.
Revival is possible and it is not restoration, and the two produces disappointment.
A revived language is acquired by learners rather than infants, and it changes accordingly.
Its simplifies towards that of the dominant language and its borrows heavily.
Purists describe the result as artificial, and speakers of it describe it as their language.
Both descriptions are accurate and only one of them belongs to people who will use it tomorrow.
Documentation remains valuable even where fails, and its value is frequently overstated by linguists.
A recorded preserves structure and loses the thing a language mainly is, which is a set of habits shared by a community.
Communities that hold their own recordings use them differently from those whose recordings sit in a university.
The first group treats them as a resource and the second is frequently unaware they exist.
Where the archive is held turns out to matter more than what is in it.