External knowledge as an acceptable argument?

[historicallinguist]

Hello linguists. I know this is not much a problem for natural language semantics (formal semantics), experimental phonetics, and computational linguistics, etc (because these areas generally deal with English, or some other programming languages), but it is quite an epistemological issue particularly relevant to problems dealing with language typology and language documentation. In particular, the core of the linguistic epistemological problem lies in whether it is a valid argument to base one's result (at least partially) on something beyond the data. For example, if I am dealing with Swedish data and I happen to know Swedish (as a L2 speaker, or native speaker), is it a valid argument that say that I know X is right because I know Swedish even if there is no evidence in data suggesting X? I raised this epistemological problem to different profs. Generally, I got two camps of profs. Theoretical folks gave me a staunch no, and said that any argument about the data must be based on the data in the dataset. They reasoned that, because we are dealing with data, conclusion must be based on data, and nothing else. Sociolinguistic and anthropological linguistic folks gave me a yes, and said that it is okay to use external knowledge beyond the data as an argument for some conclusion about the data in a dataset. They reasoned that, because our knowledge about linguistic universal (they are referring to Greenberg Universals) and specific languages is part of our general knowledge about language, there is no reason why we should block our general knowledge and confine ourselves to the data in the dataset. 

I personally incline to the theoretical folks, but admit that the sociolinguistic and anthropological linguistic's argument worths entertaining further. What is your take on this epistemological problem? 

[vonham]

I'm not sure if I follow the question. What do you mean by "right". If you mean using native knowledge to judge a sentence as grammatical, then yeah that's totally fine and theoretical linguists use native judgements all the time. If you mean using native knowledge of a language to argue that a particular structure/rule/POS/etc exists in a language, well then I'd say you need to go beyond that knowledge as "proof". Like you'd need to make an argument based on real data, and ideally other native speakers' judgements as well. If you can't find data to corroborate your argument then maybe your initial judgement is incorrect, or mediated by prescriptivism or some sort of language ideology. If linguistics were based solely on native judgements then it wouldn't be much of a discipline. 

[fuzzylogician]

Not sure I followed all of that. That said, 

1. I don't know why you think that formal semantics, experimental phonetics, and computational linguistics (what an odd collection of subfields!) only study English. That aside, 

2. Any argument you make has to be based on data. Your "knowledge" is insufficient; facts need to be established through argumentation. That may be based on data from your native(!) judgments (I would be less interested in your judgments on non-native languages, in most cases) and best through additional data collection for any non-obvious claim, be it judgment work involving multiple speakers, experimental evidence, corpus-based evidence, arguments from language acquisition, etc. 

3. Where do you think Greenberg's universal came from, if not an argument from data? You can use such a universal as an assumption in your analysis, that's for sure. You'd need to give appropriate citations to those who've argued for it and, if necessary, cite and discuss arguments against it. Your own argument will be only as strong as the premises it relies on.