<html>
<head>
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
On 15.11.17 13:42, Eitan Grossman wrote:<br>
<blockquote
cite="mid:CAA00bNkw=-vOeyepk-G6LcFw2F-Zx7558bo_Je_U70A72_-G+g@mail.gmail.com"
type="cite">
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
<div dir="ltr">
<div class="gmail_extra">
<div class="gmail_quote">
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">
<div bgcolor="#FFFFFF" text="#000000"><span class="">David
wrote:<br>
</span></div>
</blockquote>
<div> </div>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">
<div bgcolor="#FFFFFF" text="#000000"> I would just add
that would-be substantive notions such as, say,
"sonorant consonant" or "kinship term" are themselves
every bit as abstract as purely formal notions such as
domains, or syntactic categories. (After all these
years working on Indonesian, I still can't make up my
mind whether it even HAS kinship terms ...)<span
class=""><br>
</span></div>
</blockquote>
<div>I agree, and it strikes me that the term 'abstract' is
too loose to be useful without being careful about its
scope. "Sonorant consonant" involves several, perhaps many
layers, of abstraction. First of all, over individual
tokens of events in speech (and even the notion 'segment'
has been argued about in phonetics and phonology),
resulting in something like a phone [n] or a phoneme /n/
(the latter often involving another stage of abstraction)
within a particular language; so even descriptive
categories are abstractions.</div>
<div><br>
</div>
<div>Comparing such such types across language involves even
more abstraction - and maybe we need a third type of
brackets for that kind of comparative concept. Bundling
together things like [m], [n], [l] and so on into
'sonorant' is yet another abstraction. And this goes all
the way up.</div>
</div>
</div>
</div>
</blockquote>
Yes, it's true: "abstract" is too broad – what I mean is concepts
(such as zero, or transformation, or rule ordering, or phonological
domain, or syntactic category) that don't have a straightforward
connection to phonetic or semantic substance. (All of Matthew's
examples inhis recent message are substantive in this sense.)<br>
<blockquote
cite="mid:CAA00bNkw=-vOeyepk-G6LcFw2F-Zx7558bo_Je_U70A72_-G+g@mail.gmail.com"
type="cite">
</blockquote>
One might of course try to typologize on the basis of such
non-substantive notions (e.g. languages with zero or languages
without zero, languages with "late merge" vs. languages with "early
merge", languages with syntactic categories and categoryless
languages), but usually such typologies don't work well, if at all.<br>
<br>
(Incidentally, it's very odd to say that "even descriptive
categories are abstractions" – because descriptive categories CAN
EASILY be very abstract/nonsubstantive, while comparative concepts
must normally be more substantive. It's the substantive aspects that
carry over to other languages, not the abstractions. Concepts like
transformations and zeroes are important for description, but not
for typology. In my view, this is the main reason for the failure of
generative typology.)<br>
<br>
Martin<br>
<pre class="moz-signature" cols="72">--
Martin Haspelmath (<a class="moz-txt-link-abbreviated" href="mailto:haspelmath@shh.mpg.de">haspelmath@shh.mpg.de</a>)
Max Planck Institute for the Science of Human History
Kahlaische Strasse 10
D-07745 Jena
&
Leipzig University
IPF 141199
Nikolaistrasse 6-10
D-04109 Leipzig
</pre>
</body>
</html>