<html>
<head>
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
On 23.03.19 14:33, Bernhard Wälchli wrote:<br>
<blockquote cite="mid:f1c5698ac4ba404abf04ee0932245358@ling.su.se"
type="cite">
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
<style type="text/css" style="display:none;"><!-- P {margin-top:0;margin-bottom:0;} --></style>
<div id="divtagdefaultwrapper" style="font-size: 12pt; color:
rgb(0, 0, 0); font-family: Calibri, Helvetica, sans-serif,
EmojiFont, "Apple Color Emoji", "Segoe UI
Emoji", NotoColorEmoji, "Segoe UI Symbol",
"Android Emoji", EmojiSymbols;" dir="ltr">
<div>Dear Edith, dear Martin,</div>
<div><br>
</div>
<div>><span style="font-family: Calibri, sans-serif;
font-size: 18.6667px;">Martin is right</span></div>
<div><span style="font-family: Calibri, sans-serif; font-size:
18.6667px;"><br>
</span></div>
<div>No, Martin is wrong. </div>
</div>
</blockquote>
<br>
Edith basically said what I meant (and I agree that the issue of
variation is orthogonal), but let me briefly explain why I think
it's so important to note that a statement like "language L has
phenomenon P" is a comparative statement.<br>
<br>
If there were no comparative definition of concepts like "palm tree"
or "panda" in biology, or "ideophone" or "incorporation" in
linguistics, then statements like the following would make no sense:<br>
<br>
– Europe has palm trees<br>
– India has pandas<br>
– English has ideophones<br>
– Vietnamese has words (not just morphs and clauses)<br>
<br>
The second statement should actually be questioned, because while
India has animals of the <i>Ailurus</i> genus ("red panda"), it has
no animals of the <i>Ailuropoda</i> genus ("giant panda"). These
two genera are unrelated, so saying that "both India and China have
pandas" makes no sense (like saying that Mérida is larger in
Venezuela than in Spain).<br>
<br>
So if we don't have a general definition of "ideophone" that makes
use of the same criteria in all languages, then it makes no real
sense to claim that English has (or lacks) ideophones. And if we
don't have a definition of "word" that makes use of the same
criteria in all languages, then what does it mean to claim that
Vietnamese has (or lacks) words?<br>
<br>
All this is orthogonal to Bernhard's (very laudable) preoccupation
with language-internal variation (India and China also have
country-internal variation with respect to how many pandas there are
in which regions, etc.)<br>
<br>
Best,<br>
Martin<br>
<br>
<blockquote cite="mid:f1c5698ac4ba404abf04ee0932245358@ling.su.se"
type="cite">
<div id="divtagdefaultwrapper" style="font-size: 12pt; color:
rgb(0, 0, 0); font-family: Calibri, Helvetica, sans-serif,
EmojiFont, "Apple Color Emoji", "Segoe UI
Emoji", NotoColorEmoji, "Segoe UI Symbol",
"Android Emoji", EmojiSymbols;" dir="ltr">
<div>Martin would be right under the premises</div>
<div>(i) that language-internal variation is always negligible
and </div>
<div>(ii) that variable properties across languages are always
best captured in terms of discrete and simple (binary)
features. </div>
<div>However, these premises are not acceptable (even though
typology has a strong bias toward investigating properties
where these premises somehow arguably do not do much harm;
see, e.g., Wälchli 2009), and they are certainly mistaken for
ideophones in Lithuanian. I agree with Martin that it is
useful to start with clear definitions. Let us assume we have
a suitable definition for ideophones. We will then (depending
on how exactly we define ideophones probably) find in
Lithuanian that certain texts abound with ideophones while
there are many others where there is just nothing nada niente
(and that that distribution is not at all random, but has
interesting extra-linguistic correlates) and probably that
different speakers of Lithuanian have different inventories of
ideophones. Some maybe none at all or just very few.</div>
<div><br>
</div>
<div>Martin rejects the idea of UG that features are a priori
given and argues that pre-established categories do not exist.
Fine! But why then retaining the idea that typological
features should be discrete (even though this may be
convenient when using reference grammars as data source)? It
is strange that many typologists who have given up the
premises of UG still exclusively or almost exclusively
conceive of structural properties as discrete features
inherent in languages. There are exceptions such as
Koptjevskaja-Tamm (2013) recognizing alternatives: “Discrete
classifications, or typologies, operate with a restricted
number of types (typically 2 – 6, cf. the chapters in the
WALS) and are opposed to continuous typologies, which involve
quantitative characterizations of phenomena.” In many
instances, discrete classifications are nothing else but
tremendous data reduction that make claims about properties in
languages entirely non-falsifiable ( “Lithuanian has
ideophones” is as true as “Lithuanian has no ideophones”
depending on what Lithuanian data you happen to look at and
where your threshold is for recognizing the presence of
certain properties as a feature, even if everybody agrees
about the comparative concept).</div>
<div><br>
</div>
<div>It still puzzles me and will probably never stop puzzling
me with which self-evidence many typologists –
occasionally the same people who favor terms such as
“diversity linguistics” – neglect language-internal
variability despite works such as Miller & Weinert (1998)
and Kortmann (2004). Cross-linguistic diversity is just one
kind of variability in language (the one that typologists
happen to be most interested in). Languages are not
homogeneous (the idea of homogeneity is probably a heritage
from the Romantic roots of typology when languages were
considered to be organisms). When investigating a property,
the null-hypothesis for typology should be: language-internal
variability is as relevant as cross-linguistic variability.
Ideally, the typologist should then demonstrate that
cross-linguistic variation actually matters more than
language-internal variation and that that null-hypothesis can
be rejected. It is not self-evident for all structural
properties that “language” is the most relevant or only
variable, certainly not for ideophones. (And that
cross-family variability is as relevant as family-internal
variability. The omnipresent idea of stratified sampling
considered to be good methodology testifies of this. If the
property investigated happens to be diachronically stable,
fine! But what is the point of stratified sampling if you
happen to come across properties that are maximally unstable
diachronically?)</div>
<div><br>
</div>
<div>An observation about a single language does not provide
certainty about that language. </div>
<div><br>
</div>
<div>Bernhard Wälchli</div>
<div><br>
</div>
<div>Miller, Jim & Weinert, Regina. 1998. Spontaneous Spoken
Language. Oxford: Clarendon.</div>
<div>Koptjevskaja-Tamm, Maria. 2013. Typology, theories and
methods. In Schierholz, Stefan J. & Wiegand, Herbert Ernst
(eds.) Wörterbücher zur Sprach- und Kommunikationswissenschaft
(WSK) Online: Theories and Methods, ed. by B. Kortmann.</div>
<div>Kortmann, Bernd (ed.). 2004. Dialectology Meets Typology
(Dialect Grammar from a Cross-Linguistic Perspective). Berlin,
New York: Mouton de Gruyter.</div>
<div>Wälchli, Bernhard. 2009. Data reduction typology and the
bimodal distribution bias. Linguistic Typology 13.1: 77-94.</div>
<div><br>
</div>
<br>
<div style="color: rgb(0, 0, 0);">
<div>
<div style="">
<pre style="margin: 0in 0in 0.0001pt; font-size: 10pt; font-family: "Courier New"; color: black;">
</pre>
</div>
</div>
</div>
</div>
</blockquote>
<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
Institut fuer Anglistik
IPF 141199
D-04081 Leipzig
</pre>
</body>
</html>