What sort of criteria are used to deem a language description complete?

Question

I'm not a field linguist but my hobby overlaps with it, particularly the descriptive part.

At the far ends, with remnant langauges (Virginian Algonquin) and well studied mega-lingua francas like English or French the answer is clear. Given limited time and resources, what constitutes enough dictionary, corpus and other descriptive material to say a language is documented?

One book by a field linguist I read said about six pages of well meticulously glossed text covered most of a languages details. Are there any other rules of thumb for deciding when one has hit the point of diminishing marginal returns for documenting a language?

(I'm not really looking for the answer "the job is never done!", since it isn't really useful)

Answer 1

You might enjoy a 2005 web publication by Johanna Nichols, which discusses what one can expect to learn about a language, given the size of the available corpus.

In my particular experience, the most efficient situation is where one has a fair-sized text corpus along with follow-up elicitation based on patterns encountered in the text. Having a purely text-based corpus makes the number of words required considerably larger. For one project I've worked on, with a ~280,000 word corpus with free translations and no supporting materials, myself and my colleagues could cover all of the basic topics in a grammatical description. A current project has a corpus of ~35,000 words with extensive notes from follow-up elicitation, and I'd consider the potential coverage better, but not by much, than for the language with the larger corpus.