[Lingtyp] comparative concepts & What do glossing labels stand for?

[David Gil]

First, let me bring to your attention a handout for a talk given by 
Eitan Grossman which raises many of the issues discussed in this 
thread.  (Eitan seems to be too modest to engage in such 
self-advertisement, so with his agreement, I am doing this on his behalf.)

https://www.academia.edu/20715205/What_else_are_comparative_concepts_and_descriptive_categories_good_for


Next, my reactions to three recent postings by Östen, Matthew and Bill ...


On 26/01/2016 05:27, Östen Dahl wrote:
> Here is a question that I would like to pose to the members of the ALT list. If we accept the distinction between "descriptive categories" and "comparative concepts", what do the labels we use in glossing example sentences stand for - in particular, the labels defined in the Leipzig glossing rules? I have some thoughts about this myself but would like to hear what others think first.
> östen
Great question.  My own answer: obviously both!  Glosses have to be 
faithful to the language-specific categories of the language.  But at 
the same time they have to convey information about the language that is 
intelligible to someone who doesn't speak it.

The question of glosses is thus reminiscent of the issue of bilingualism 
that has already come up earlier in this discussion. And like 
bilingualism, it seems to me that glosses pose a serious problem for the 
distinction between language-specific categories and comparative concepts.


On 26/01/2016 02:13, Matthew Dryer wrote:
> [...] hypothesized absolute universals fall into two categories. They 
> are either trivial (like every language must have at least three 
> phonemes) or untestable. What this means is that even if there are 
> absolute universals, trying to identify what they are is fruitless.
I don't share Matthew's skepticism here.  Among others, I think there's 
a class of absolute universals of the rather abstract formal kind that 
are neither trivial nor untestable, and, equally importantly, probably 
true.  A couple of my own favourites:

(1) Grammars don't count.  For example, no language has a rule that 
specifies that a certain constituent must consist of exactly 10 
syllables.  Note that this cannot be due to a general cognitive 
constraint, since, outside of grammar, we can count syllables, e.g. when 
writing poetry in iambic pentameter.

(2) Grammars don't reverse the order of elements.  For example, no 
language has a rule that inflects a word for a certain category by 
reversing the order of segments in the word.  Again, note that this 
cannot be due to a general cognitive constraint, since some ludlings do 
precisely this.

Lots more examples like this can be constructed.


On 26/01/2016 00:18, William Croft wrote:
> David's response makes an assumption that I, and possibly Martin, would not agree with: that comparative concepts are distributionally-defined categories just like language-specific categories are.
I'm not assuming that comparative concepts ARE distributionally-defined 
categories, just that they CAN BE.  Others can clearly refer to 
phonological or semantic features.
> David proposes a hypothetical world in which two languages have identical distributions of equivalent units across equivalent constructions, and suggests that these would be comparative concepts in Martin's and my, or at least my, sense. David already raises a major problem: the world isn't like that, so why should we waste time theorizing about it?
Because, as I argued further on, understanding how we might deal with 
this hypothetical world is necessary for understanding how we should 
deal with aspects of our real world — such as bilingualism, dialectal 
variation — that bear a resemblance to that hypothetical world.
> [...]
>
> I did look at your 2000 paper, David. The paper constructs a categorial grammar model to define crosslinguistically valid language-specific categories. Just one symbol, S, is assumed as a starting point. But to define that crosslinguistically, you appeal to Boas' semantic definition of 'a complete idea'.
No I don't.  S is defined purely distributionally, with no reference to 
semantics.  I have no idea what the notion of "complete idea" refers to, 
and certainly would not wish to base a crucial definition on it.  (My 
reference to Boas was an aside, making the point that my theory 
resembles Boas' approach and for that matter also that of the 
generativists in that it takes the sentence, rather than, say, the word, 
to be the starting point of grammatical description.)
> [...]
>
> Finally, David discusses language-internal variation. But his solution is to slice up a language into homogeneous grammars in order to preserve some kind of categorial consistency. This is the same analysis used by generative grammarians for diglossia. Taken to its logicla conclusion, this leads to a model of each speaker having zillions of grammars in their head in order to capture all the variation across all situations of use (Croft 2000:51-53). Instead we should embrace a single variable grammar for a speaker to model her knowledge about language-internal variation (ibid.).
I did not propose this as a solution.  I raised this as a logical 
possibility, which, in the extreme case, leads to precisely the kind of 
absurd "zillions of grammars" situation that Bill rightly rejects.  My 
point, however, is that because there is no clear cut distinction 
between dialects and languages, there is no obvious cut-off point, no 
easy way to draw the distinction between micro-variation and "zillions 
of grammars" on the one hand, for which case we should probably "embrace 
a single variable grammar", and bilingualism on the other, for which it 
would be more reasonable to the existence of two distinct grammars in 
the speaker's head.

The difficulty in distinguishing between these two cases is dramatically 
(if not exclusively) evident in the case of "large" languages with lots 
of sociolinguistic and geographical variation, such as 
Malay/Indonesian.  When people ask me how many languages I speak, I tell 
them I don't know how to answer the question, not only because I don't 
know how to count my varying degrees of non-native proficiency in 
different languages, such as my imperfect French, but also because I am 
familiar with several varieties of Malay/Indonesian, such as Kuala 
Lumpur Malay, Riau Indonesian, Siak Malay, Jakarta Indonesian and Papuan 
Malay.  Should I count these as one or as five?  Somehow neither answer 
seems satisfactory.  In addition I am familiar with Minangkabau, which 
in many respects feels less different from Riau Indonesian or Siak Malay 
than Papuan Malay does.  But Minangkabau has the cultural and political 
clout to earn it the label of a "different language", though this should 
not be relevant to the issues under consideration here.  Although the 
above auto-biographical example involves an obviously 
non-native-speaker, it's not exceptional:  many if not most Indonesians 
are adept, to varying degrees, in a wide range of varieties of 
Malay/Indonesian, straddling the boundaries between different dialects 
and different languages.  And of course such issues are not specific to 
Indonesia.

So how do we deal with this state of affairs?.  Bill seems to be 
offering us a stark dichotomy between single variable grammars 
associated with multiglossia, and distinct incommensurate grammars 
associated with multilingualism.  My view is that the real-life 
continuum between dialects and grammars should be reflected in a 
continuum between these two modes of representation.  How exactly to do 
this I do not know; my remarks at this point are largely programmatic.  
But here's how I see things.  Using the recently introduced term 
"languoid" to generalize over dialects and languages, given two 
languoids, I would envisage their respective descriptions as involving a 
mix of (a) specific stuff that is unique to each description and 
incommensurate with anything in the other description; and (b) common 
stuff that is shared by both descriptions.  Bill's two choices provide 
us with the two logical endpoints:  while single variable grammars have 
none of the specific stuff in (a), distinct incommensurate grammars have 
none of the common stuff in (b).  However, the dialect-language 
continuum suggests that for languoids straddling the dialect-language 
fence it is reasonable to countenance a mix of the two.  And once we 
allow for that, we might as well acknowledge the possibility of such a 
mix, with appropriately different relative balances, also for very 
closely related dialects on the on hand, and for very different 
languages on the other.

Let us now consider more closely the latter case, that of very different 
languages, whose respective descriptions will have lots of specific 
stuff unique to each description, but also some common stuff, shared by 
both descriptions.  Such common stuff would constitute, simultaneously, 
elements of the respective descriptions of each language, but at the 
same time also a basis for cross-linguistic comparison, thereby, for 
cases such as these at least, casting doubt on the necessity for 
distinguishing between language-specific categories and comparative 
concepts.

-- 
David Gil

Department of Linguistic and Cultural Evolution
Max Planck Institute for the Science of Human History
Kahlaische Strasse 10, 07745 Jena, Germany

Email: gil at shh.mpg.de
Mobile Phone (Indonesia): +62-812-73567992