How we learn about the languages’ structure so that we might code - grambank/grambank GitHub Wiki

There are many different ways to gather knowledge about the structure of languages. Typologist sometimes use translation questionnaires, stimuli, elicitation or spontaneous speech to collect primary data. Most commonly, however, they consult already existing language descriptions, particularly grammars and grammar sketches, but also papers and presentations describing specific grammatical features. Out of the more than 7,000 identifiable languages of the world, at least 4,702 languages have at least a grammar sketch (Hammarström et al. 2020). To code Grambank, we mostly rely on data and descriptions found in this literature.

When reading a grammar, we do not just take a description for granted. We cast a critical eye on the analysis of an author, and consider how the terminology and concepts an author uses corresponds to the comparative concepts we apply for the purpose of Grambank.

We sometimes consult experts working on particular languages. For previous versions of the questionnaire, when coding languages of Melanesia and Australia, the coding team corresponded with authors of grammars. Later, when African languages were coded for the Nijmegen Typological Survey, we also had extensive contact with active researchers. Discussions with language experts are particularly useful in cases of absence of evidence or when some arguments for a particular analysis are left explicit in the description. Before contacting experts, coders are advised to go through the existing literature. This way, many features have already been addressed, and more time can be devoted to the most challenging questions. Language experts must always have the chance to see the final coding and return with comments. We do not publish any data without the explicit consent of the expert.

Interacting with language experts when coding not only helps to bring up the accuracy of the coding, but also spreads awareness of the database and increases credibility in language-specific research communities. For more advice on how to interact with experts, go here.

Learning about a language from other languages or daughter languages

If the only source of information about the structure of a language is based on "it ought to be similar to a sister-langauge", then we discount that. We only want to record information that is known about that specific language.

However, if the language is an ancient language that related to extant languages, it is possible to use reconstruction by historical linguists. However, in such cases write "RECONSTRUCTION" in the comment field for each value of this kind.