typology and mental architecture

Tue Jan 19 08:04:50 UTC 2010

Dear Bill,

Thanks for your further comments.

> I agree that some, perhaps many, aspects of languages are functional in a way
that does not reflect any underlying
> specifically linguistic mental structure. However, I don't see functional
> explanations and an underlying mental grammar as an unbridgable dichotomy in
> the way I infer (perhaps incorrectly) from your remarks. I imagine that some
> aspects of languages are entirely structural and instantiate underlying mental
> grammar, while others are functional and are not manifestations of a mental
> grammar at all. 

Yes, it's quite possible that some general properties of languages are due
to the "underlying mental structure/archirecture" (=cognitive constraints,
or what I've called the "cognitive code"), while others are due to
functional constraints.

The question is: How do we tell whether a general property is due to one or
the other factor? It seems to me that if we find a functional motivation for
a generalization (e.g. in terms of frequency of use; thus, since Croft
(1991) we have known that systematic noun-verb-adjective distinctions in
languages are due to different usage frequencies of thing, action and
property words in reference, predication and modification functions), then
we have no evidence that a mental/cognitive constraint is involved.

The problem with the generative approach is that it tends to jump to a
conclusion about the mental architecture without considering simpler
functional explanations. (It's as if one tried to infer the genetic code
from the phenotype of organisms, without taking adaptation into account.)

> But I suspect that a great many aspects of language are functionally motivated
*and* built into an underlying mental architecture (in
> fact built into the architecture precisely because they are functionally
> required). Or to put it another way, a considerable part of the underlying
> architecture of language is functionally motivated and requires attention to
> functional issues to properly understand.

Maybe, but how do we know? Most proposals about the mental architecture
(=UG) are based on typology (i.e. existing languages), but if existing
languages are a tiny subset of cognitively possible languages (just as
existing organisms are a tiny subset of genetically possible organisms),
then this evidence is quite useless for that goal. ("Typology is irrelevant
to grammatical theory", as Newmeyer put it: cf. his 2005 book "Possible and
probable languages".)

> However, if we assume that such a mental architecture exists, then typology
> has a crucial role to play in the investigation of that.  We could take an
> alternative view - that there is no specifically linguistic mental structure,
> and that language is the result of something else, such as the complex
> interaction of a set of generalised cognitive functions. (If we do take that
> view, I must say I can't see why a separate discipline of linguistics is
> appropriate, but that's another matter.) If we do want to take that view we
> need to be up front about it, and say our definition of typology applies in a
> perspective in which no underlying specifically linguistic modality exists. On
> the other hand if we want to assume that there is a specifically linguistic
> modality, then the goal of linguistics must at least in part be to model that.

I don't know whether there is a "specifically linguistic modality", and I
don't know about possible cognitive constraints on language structure. As I
said earlier, I think experiments with artificial languages are perhaps the
best way of finding out about them (just as biologists first learned about
genetics from "artificial organisms", i.e. breeding experiments). So far we
seem to know very little, and it's best to be agnostic, rather than take a
particular view from the outset.

Thus, typology should not try to be like genetics, but like comparative
zoology and botany: Go out to the field (documentation), collect specimens
for museums (description and creation of typological databases such as
WALS), and build a theory of adaptation that accounts for the general
patterns that are found.

Best,
Martin

>>>> Martin Haspelmath <haspelmath at eva.mpg.de> 18/01/10 10:32 PM >>>
> Thanks, Bill, for this clear and succinct statement.
> 
> But unfortunately, not all linguists agree with it: Many would say that
> typology is of little relevance to modelling the structure of language in
> the mind ("formal theory"), because most universals seem to have functional
> explanations and are not due to any restrictions on the mental grammars that
> we can acquire and represent mentally. As is shown by the natural
> acquisition of language games and artificial languages, as well as by
> artificial learning experiments, we can acquire a lot more than we normally
> find in natural languages.
> 
> In generative linguistics, the role of "formal theory" is "modelling the
> underlying mental structure of language in such a way that it allows for all
> attested languages, while not predicting the unattested languages", but on a
> functional approach, predicting unattested languages falls on "functional
> theory".
> 
> Greetings,
> Martin
> 
> P.S. I have written about these matters in my 2004 article (Haspelmath,
> Martin. 2004. Does linguistic explanation presuppose linguistic description?
> Studies in Language 28. 554-579.)
> 
> 
> On 18/01/2010 02:15, "Bill Palmer" <Bill.Palmer at NEWCASTLE.EDU.AU> wrote:
> 
>> Dear Yuri
>> 
>> I see the role of typology in the following way. The core research agenda of
>> linguistics is to model the structure of language in the mind. A language is
>> a
>> set of almost entirely identical structural systems in the minds of each of
>> its speakers, and every language is somehow an instantiation of the
>> underlying
>> mental language capacity.
>> 
>> Four layers of research activity contribute to this core agenda: formal
>> theory, typology, description, and documentation. Modelling the structure of
>> language in the mind is the role of formal theory. However, to do this,
>> formal
>> theory depends on typology and description: in different ways, each tells
>> formal theory what it must attempt to account for. Theories of language
>> cannot
>> be adequate if they are not based on an understanding of what occurs in
>> language, both in terms of the range of possibilites across languages, and
>> the
>> details of structures within individual languages. The central role of
>> typology is to provide the former, and description the latter. Typology is
>> about understanding the boundaries around the possible in language.
>> 
>> Typology does this in three ways: 1) by identifying the full range of what is
>> possible in each area of language structure by observing it in an actual
>> language or languages; 2) more problematically, by inferring that some
>> logically possible structures cannot occcur, on the basis of observing their
>> complete absence from any actual language (although this may result from an
>> accidental gap - such structures may exist in languages as yet undescribed,
>> or
>> in languages that became extinct last week or a thousand years ago, or will
>> evolve out of a currently existing language - and attempting to distinguish
>> between impossible structures and accidental gaps is also an issue facing
>> typology as well as formal theory); and 3) also problematically, by
>> determining which structures are universal, which common, and which rare.
>> Formal theory then has the role of modelling the underlying mental structure
>> of language in such a way that it allows for all of 1), while not predicting
>> any of 2), and hopefully also accounting for 3) (particularly how a structure
>> can be possible, but be what Corbett calls "vanishingly rare").
>> 
>> The role of language description is to provide as detailed as possible a
>> description of the possible and impossible structures and their functions in
>> individual languages. This provides formal theory with the very detailed data
>> it needs to build thorough models, but it also provides typology with
>> adequate
>> descriptions on which to base its generalisations. Good description is
>> theoretically and typologically informed, but should not in itself have a
>> theoretical or typological focus (although good descriptive work often also
>> results in research that address theoretical or typological issues).
>> 
>> So, formal theory depends on good typology and on good description, and
>> typology also depends on good description. Description in turn depends on
>> good
>> documentation: as in any science, the more detailed, extensive and accurate
>> the raw data, the more accurate the description of the phenomenon under
>> observation, and ultimately the more adequate the theories developed on the
>> basis of that data.
>> 
>> cheers
>> Bill
>> 
>> 
>> Dr Bill Palmer
>> Convenor, Pacific Languages Research Group
>> Linguistics Research Higher Degree and Honours coordinator
>> School of Humanities and Social Science
>> University of Newcastle
>> Callaghan NSW 2308
>> Australia
>> email bill.palmer at newcastle.edu.au
>>>>> Yuri Tambovtsev <yutamb at MAIL.RU> 16/01/10 10:57 PM >>>
>> Dear Typology colleagues, reading journals on typology I came to a conclusion
>> that different linguists and edotors understand typology in their own way. It
>> especially concerns editors who may return a manuscript on the pretext that
>> it
>> is not on typology. Therefore, it occurred to me to discuss the term
>> "typology" in order to realise how it is understood in modern linguistics.
>> Looking forward to receiving your definitions to yutamb at mail.ru  Yours
>> sincerely Yuri Tambovtsev
>> 
>> 
> 
> 
> 
> 