[Lingtyp] Traditional view of language and grammar in indigenous societies

[Adam Singerman]

Dear Randy,

Many thanks for these clarifications. I agree 100% that being trained
in a majority/dominant language's grammatical tradition doesn't always
translate into success when it comes to making sense of the structure
of a minority language, especially one that is typologically different
from (or unrelated to) the majority language. I have seen this myself
in Brazil, where the grammatical categories of Portuguese can be
easily misapplied to Indigenous languages.

Just to clarify my own terminology, I use the terms "acceptable" and
"unacceptable" when referring to speakers' judgments but reserve the
terms "grammatical" and "ungrammatical" for talking about linguists'
analyses. Speakers don't tell the linguist whether something is
grammatical or not; they only tell us whether something is acceptable
or not. And if a speaker says something is unacceptable, it's up to
the linguist to figure out why (morphosyntactic ill-formedness;
semantic anomaly; pragmatic infelicity, including impoliteness; etc).
This is why when I teach field methods I insist that my students pay
attention to the distinction between (un)acceptability and
(un)grammaticality.

With this terminology clarified, I would like to ask about this
statement of yours: "There is no such thing as ungrammatical sentences
in natural linguistic data." I don't see how this claim can be
maintained, at least not in the very strong version you put forward
here? Speakers make mistakes — repetitions, disfluencies, etc. Suppose
a speaker tells me a narrative and afterwards, when we review the
recording together, I ask the speaker about the meaning of a sentence
where there's an agreement paradigm I didn't expect, and the speaker
says "oh, I said that wrong" and then offers a revised version.
Wouldn't this be a case of an ungrammatical sentence (what I would
call an unacceptable sentence) in naturally produced data? Also, if
there is variation between speakers of a language, as there always is,
it's possible that something in our corpus will sound right to a
particular group of speakers but not to another group, so depending on
one's perspective there is in fact an unacceptable/ungrammatical
sentence in the data.

Can you clarify for me what you mean by your claim that natural
linguistic data do not include ungrammatical sentences? Are you and I
disagreeing about terminology or about something deeper?

Thank you,
Adam