[Lingtyp] comparative concepts

Sat Jan 23 09:45:01 UTC 2016

I never took Philosophy 101, but I am trying to understand differences 
and similarities between languages, and I feel that using descriptive 
categories for comparative purposes often leads to confusion.

On 22.01.16 22:25, Östen Dahl wrote:
>
> I agree with Matthew that the issues are becoming too complex and at 
> the same time with Dan that this is Philosophy 101. But the questions 
> asked in that course are often the most difficult ones. The quotation 
> from William James may be relevant here. It seems to me that what 
> Martin is saying about “cleric” as a comparative concept means that 
> pace James this concept cannot be an objective one, since apparently 
> it is just a "product of our cognitive system", as opposed to chemical 
> elements, which are something more. (But Locke said that all 
> categories are products of our cognitive system??)
>

I think Locke was wrong there (at least for the practical purposes of 
science) – some categories are NOT just the products of our cognitive 
system, but exist in the world, independently of any observers. 
"Hydrogen" and "photon" are such categories, as are "red fox" and 
"Lezgian language" (otherwise foxes wouldn't be able to interbreed and 
speakers would be able to interact and conform to norms). I have talked 
about such categories as "natural kinds" (though I haven't read the 
philosophical literature about them), and I think they constitute 
discoveries of science, not just instruments for discoveries.

According to generative grammar, categories such as vP or [+wh] or 
[+coronal] are also natural kinds – existing in the world independently 
of the observers, and (at least potentially) discoveries of generative 
grammar. This is a possibility that we should take very seriously, but 
my interim conclusion is that this approach doesn't work well – to 
compare languages fruitfully, one needs concepts that are set up 
specifically for the purposes of comparison, as *instruments* for 
further discoveries. (Balthasar Bickel has sometimes said that 
typological variables are like measures – I like this analogy, because 
it's clear that concepts such as "meter" are not discoveries, but are 
nevertheless crucial to science.)

Edith Moravcsik asked: "It is impossible in principle for constructions 
in two languages to be members of the same descriptive category?" I 
would say yes, because descriptive categories are set up on a 
language-specific basis ("distributionally", to use Bill Croft's word) 
for the purposes of description (or analysis). Hypothetically one could 
imagine two languages that have exactly the same grammar (but different 
words), and in that case, one might say that they share descriptive 
categories. Perhaps at a lower level, this situation is actually found – 
so maybe with respect to the behaviour of property words, Italian and 
Spanish are indeed close to identical. In that case, it would not do any 
practical harm to say that they have the same descriptive category. But 
we normally describe each language separately (e.g. we do not skip the 
description of Spanish adjective syntax and point to an already existing 
description of the same facts in an Italian grammar), i.e. we treat each 
grammar as an indivisible unique system.

Östen continues:
>
> I am not quite sure I am following here. Does the difference between 
> “cleric” and “hydrogen” depend on the differences in well-definedness 
> or on the fact that “cleric” is based on social constructs whereas 
> “hydrogen” is a natural phenomenon? I get confused when Martin 
> mentions the theoretical possibility of an innate mental category of 
> “cleric” – that would seem irrelevant to me.
>

Hydrogen is a discovery (a natural kind), but "cleric" (in the 
comparative sense) is a construct of comparative religion scholars, i.e. 
an instrument for discoveries. "Well-definedness" is an issue only for 
comparative concepts. We don't need definitions for natural discoveries 
such as "red fox" or "hydrogen" or "Lezgian language", or "vP" (note 
that generativists never define their categories, which is completely 
consistent with their claimed status as natural kinds). The suggestion 
of an innate mental category "cleric" was meant to be patently absurd, 
but maybe for other categories of social organization (e.g. grandfather 
or marriage), this is less absurd.

Volker Gast has made an intriguing suggestion:

> This discussion seems to show a broad consensus that we do not 
> actually generalize over linguistic data, but over comparative 
> concepts/linguists' classifications. ... So I wonder if WALS should be 
> renamed to the 'The World Atlas of Comparative Concepts'? ;-)

I wouldn't say that we don't generalize over "linguistic data", but it 
is true that we don't generalize over grammars (in the strict sense), 
just as comparative biologists (e.g. scholars concerned with the 
relative wing ratio of bats and birds 
<http://jmammal.oxfordjournals.org/content/17/4/412.2.abstract>) do not 
generalize over genomes. (As a historical anecdote, it may be 
interesting that the very first working name of WALS was "The grammar 
atlas" – I still have an e-mail from David Gil from November 1998, when 
we were starting to work on this project, which has this in the subject 
line).

So if grammars are analogous to genomes, the totality of our linguistic 
behaviour (within a speech community) is analogous to phenotypes. I 
would say that typologists generalize over these phenotypes – compare 
Matthew's point that usage frequencies also play an important role for 
word order typology. Whether these phenotypes are well described by the 
term "structures" is indeed a good question – I must say that I would 
associate the term "structure" more with "grammar", perhaps using 
"grammatical patterns" for the phenotypes. "World Atlas of Grammatical 
Patterns"?

Best,
Martin

-- 
Martin Haspelmath (haspelmath at shh.mpg.de)
Max Planck Institute for the Science of Human History
Kahlaische Strasse 10	
D-07745 Jena
&
Leipzig University
Beethovenstrasse 15
D-04107 Leipzig

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/lingtyp/attachments/20160123/69dd37ce/attachment.htm>