[Lingtyp] spectrograms in linguistic description and for language comparison

[Adam Singerman]

I'd like to thank everyone for their responses. I'm going to respond
to some of Randy's reply, as there are important points of
disagreement and my own comments seem to have been misinterpreted in
places.

Randy wrote: "The beauty of working inductively is that you are only
responsible for what is in your data. You don't have to make broad
generalisations about the language that in many cases turn out to be
problematic, you just say this is what is and is not in my data. Of
course the more data you have the stronger the generalisations you can
make."

This is a major philosophical disagreement. I don't want my work on
Tupari to be responsible only for what happens to be in my corpus; I
want it to make accurate predictions about the linguistic behavior of
Tupari speakers *going forward*. Asking speakers for acceptability
judgments is a tool for teasing apart what can be said from what can't
be said, and I need this tool to enhance the predictive power of my
descriptions and analyses. Especially where endangered languages
(unlikely to be researched by future generations of linguists) are
concerned, to tell fieldworkers "you are only responsible for what is
in your data" feels like bad counsel.

Regarding methodology, Randy wrote: "Asking people to translate word
lists or sentences will not give you useful data. What you will get
back are the categories of the working language." I agree that using
translation tasks as one's main data collection strategy is not best
practice, but this strategy is not what I called for in my post.

Quoting Karl Popper: "Whenever a theory appears to you as the only
possible one, take this as a sign that you have neither understood the
theory nor the problem which it was intended to solve." My
interpretation is that Popper is talking about theories that make
testable predictions; he is encouraging the reader to always seek ways
to disprove predictions in order to falsify the theory at hand. But in
my post I did not advocate any such theory. (The closest I came to
doing so was discussing the methodological carefulness of various
fieldworker-semanticists, but I did not promote their models.) What I
*did* advocate was an inclusive methodology that uses acceptability
judgments alongside corpus data. This methodological choice is
compatible with a very wide range of theories in Popper's sense.

I mentioned "control vs raising structures, pied-piping, islands, gaps
in inflectional paradigms" as some areas of grammar where corpora,
even very large corpora, aren't enough; to see the structural patterns
we need negative evidence. It is of course any linguist's right to
say, as Randy did, that "these are not things that concern me." But
the fact that these topics don't concern a particular linguist does
not constitute an argument against investigating them nor against
using acceptability judgments to do so. To not research certain topics
simply because we can't get at them through corpora alone would be to
allow a dispreference for acceptability judgments to limit our
discipline. This strikes me as backwards.

Finally, I genuinely did not mean to say that anyone does linguistics
wrong. I agree that there are many different ways to do linguistics
and so I really do apologize for this misunderstanding. The only thing
I consider wrong (in the sense of "not true") is the claim that
corpora are all we need for good description to take place.

— Adam