Some preliminaries on typology in general as it relates to Grambank
Linguistic typology, systematic cross-linguistic comparison, is a method used by linguists in a variety of different areas of languages’ characteristics: lexicon, semantics, phonology, morphosyntax, pragmatics etc. Just like other methods and schools of linguistics, such as parts of the generative enterprise, we are investigating the cognitive, areal, cultural, communicative and/or historical/evolutionary constraints on language diversity, disparity, complexity and change. Paraphrasing Boas (1938:132): all languages can express all concepts - what is interesting is to compare what they must do (i.e. what is grammatical/obligatory).
[the aim of linguistic theory] must be to show which structures are possible, in general, and why it is just those structures, and not others, that are possible
(Hjelmslev 1970 [1963]: 96)
Comparing languages is very different from describing or documenting languages, in comparison it is absolutely crucial that the thing compared is indeed applied consistently to every language in the sample. At times the comparative concepts applied in typology might be a very poor fit to capture a specific language appropriately, this is why, for example, Haspelmath, Bybee and Dahl have stressed the need to keep the comparative and language-specific enterprises separate (Bybee & Dahl 1989, Haspelmath 2010). Otherwise we might end up with descriptions that are too influenced by the typological framework so that we only find what we had thought to look for. It is also nearly impossible for typology to adequately accommodate all of the complexity of all language-specific situations. This is why terms and concepts used in comparison are necessarily not the same as those used in language-specific description.
Comparison is informed by description and often influences it as well, but the concepts being compared are defined in contrast to each other across the entire sample, not (or rarely) in relation to other language-specific concepts.
Languages can be considered both across time (diachronically) and in their current states (synchronically). The synchronic variation can be described in terms of different languages, dialects, sociolects, ideolects etc. As linguists we make decisions on where to draw up categories in this space and discuss these to reach agreement. In Grambank we are using the ISO 639-3 language codes for classifying what is and what is not a language. Within these language units we also find variation, and often we find change in motion. Concepts in morphosyntax, in fact in all language characteristics description, are rarely discrete and homogeneous. For example, demonstratives might grammaticalize into definite articles, auxiliaries become affixes, independent markers in general become bound, perfective aspect moves towards past tense and so on. Because the entities that we are targeting for comparison in this survey exist on these continua it is necessary for us to make explicit what criteria we use so that we are indeed comparing languages consistently. For this purpose we have created so-called 'clarifying comments' to aid coders.
It is inherent in the nature of comparative work that the variable that we are comparing changes as new data comes in. In early linguistics we started by looking for and comparing what we knew then, as we learned about more and more languages our perception of what is interesting to compare naturally changed. In the case of the Grambank, many of the features belong to the legacies of earlier cross-linguistic research projects and have been coded for up to 270 languages already. These features might not always be perfect; the search for perfect features is potentially neverending; in this enterprise we need to be pragmatic and consistent. Please take care in learning about the features and apply them consistently.
To summarize: comparative concepts are not the same as language-specific descriptive concepts. We need you as coders to understand the features very well and apply them consistently across the languages you’re coding.