[Lingtyp] To include xenophones or not

[Adam Singerman]

Dear Ian,

I've just gone back to read your original question, the one that got
this whole thread started. I do understand that there are drawbacks to
incorporating xenophones / loan phonemes into your database, just as
there are logistical challenges when constructing any large
crosslinguistic sample. My question is: what, exactly, do you want to
be able to use the database *for* in the long run? As the
contributions to this discussion have shown, excluding xenophones will
flatten out the data and make the entire database less accurate. So
what's to be gained by making a database of 700 Eurasian languages if
in exchange for including all 700 you have to exclude relevant
phonological information? This seems to me to be one of those cases
where quantity (the number of languages in your database) should not
outrank quality (the accuracy and nuance of the data for each
language). Is there a reason why you couldn't work with a smaller set
of languages (100 or 200 — which, in my view, would still be a serious
undertaking) and incorporate more information, including xenophones?

Best,
Adam