And Rosta wrote:
>
> Jasp:
> #> The facts, as I see them, are that formal linguistics does a good
> #> job of modelling grammar and a bad job of modelling how language
> #> in the mind works. The cognitivist model that you and Joe and
> #> Dick espouse does a good job of modelling how language in the mind
> #> works, but a bad job of explaining where grammar comes from. The
> #> same goes for Minimalism: the more it takes UG seriously and tries
> #> to pare it down to something plausible, the less successfully it
> #> models actual grammars.
> #>
> #> The first thing I think one has to do, which few linguists do do,
> #> is recognize these facts. That's the stage I'm at. Where does one
> #> go from there? Well, obviously the task is to find some sort of
> #> bridge between formal grammar and cognitivist models of mental
> #> language that explains how the former exists transcendentally
> #> within the latter. My conjecture (because it is the only halfway
> #> decent one I can think of) is that grammar is an emergent property
> #> and that there is only a relatively tight solution space to the
> #> problem of how to communicate effectively in speech using the
> #> human brain as a processor. There's probably work being done on
> #> this that I don't know of; the name that comes to my mind is John
> #> Hawkins, with his attempts to explain syntactic phenomena as
> #> the product of very general parsing strategies.
> #>
> #> But what I do want to insist on is that grammar as this abstract(able)
> #> system does exist, and that its abstractability is inherent, not a
> #> delusion of the formal linguist.
> #
> #And this is just exactly where we differ. Actually, I think your
> #'conjecture' seems intuitively correct. I can't see why I can't say
> #something like that without straying from my cognitivist-wholist
> #cave-system. Wouldn't that count as an explanation of 'where grammar
> #comes from'
>
> yes, but...
>
> #(while bearing in mind that I don't recognise the existence
> #of grammar in the sense that you mean it!)?
>
> ... no. I don't think you can claim to explain X if you deny that
> X exists!
>

ok, but you recognise that this is just a problem associated with the linear nature of email messages, do you? I think your conjecture can explain the existence of grammar as I understand it. I think it would also explain grammar as you understand it if it had to, as it happens, just that it doesn't since grammar as you understand it doesn't exist.

> #I think that what you call grammar is exactly a delusion of the formal
> #linguist: it is a product of (eminently admirable) attempts to describe
> #the observed patterns of language. For these codifications to work (by
> #which I mean fit in a book roughly the size of Quirk et al) they must
> #be rational, finite and abstract.
>
> If true, the delusion itself would need an explanation.

If I knew enough about ancient Greece, I might be able to give you the beginnings of an historical explanation. We could also specualte that there is something reductivist inherent to human psychology, but I'd prefer not to.

> How is it that
> different linguists and models all tend to converge on the same solution?

When this does indeed happen, could it be because they are all clever folk, and are working on the same problem?

> How is it that new and better 'analyses' continue to be discovered' that
> appear to reveal hitherto unperceived patterns and regularities?

This can only be because the old analyses are so poor.

> How is
> it that abstract models are so successful? -- Structuralist semiotics
> set out with the research programme of applying the methods of linguistics
> to all domains of culture, and that programme was a dire failure precisely
> because there was no equivalent underlying system there to be discovered.
>
> #> #To risk another analogy (and to rob shamelessly from Jackendoff),
> #> #this strikes me as like the drunk man searching for his lost key
> #> #under the street lamp because it is lighter there and easier to look.
> #>
> #> I think I don't understand the analogy. What is the lost key?
> #> A theory of language? And what is the light area?
> #
> #The key is, as you say, a theory of language; the light is reductivism.
> #
> #Of course the analogy doesn't quite work because your reductivism allows
> #you to claim that whatever is found under the streetlight *is* the key.
> #Joe and Dick and I insist on it providing us with a way of getting into
> #the house without waking the wife.
>
> I'm still struggling a bit. We seem to disagree not about where to look
> for the key but rather about the nature of the house or the lock that
> we're seeking a key for, or whether there is a house there at all.
>

I think I'd rather drop the whole key thing, actually, if I might.

> It seems to me that the only thing we disagree about is whether reductive
> methods produce an essentially fictive object that is merely the
> inevitable product of those methods, or whether they turn up something
> real and demanding of an explanation.

Yes. And remember that reductive methods do inevitably produce a product of the kind you are advocating, so there is nothing unlikely about my position. In fact what is unlikely and in need of an explanation (if you are right) is the fact that a natural phenomenon should be exactly as predicted by a theory of science that owes its development to a series of social accidents. (I concede that the same explanation I gave above also applies here:  "they are all clever folk")

> It reminds me of the mathematics of biological forms -- e.g. leaf growth
> following fractal patterns, and shoots branching off from stems at
> intervals that follow the fibonacci sequence. If you're Chomsky, you'd
> say that the only possible explanation is that plants have genes
> that specify the fractal and fibonacci patterns. If you're you, you'd
> say that the patterns need no explanation -- that if you apply mathematical
> methods to nature you're bound to come up with something that, superficially,
> seems systematic, but if you varied nature in random and arbitrary ways the
> same mathematical methods would still find mathematical patterns. If
> you're me, you'd say that the patterns are real, and that they represent
> the optimal or most probable solution to the problem of how to grow.
>

Most probable yes (probability of 1).

> #> #> This said, I actually raised the chess analogy to try to
> #> #> illustrate my distinction between grammatical pattern and
> #> #> textual pattern, the former being like an abstract chessgame
> #> #> defined solely as a sequence of moves, and the latter being like
> #> #> (a category of) actual situated chessgames.
> #> #
> #> #This is the more interesting source of disagreement, since it
> #> #constitutes a claim on your part about the nature of language-in-the-mind.
> #> #
> #> #As I understand it you are distinguishing between those mental
> #> #representations that do and those that do not refer to 'situational'
> #> #properties. Does either kind of representation have any other
> #> #properties not shared by the other?
> #>
> #> Clearly more abstract, unsituated sort has many fewer properties
> #> and these properties are of many fewer sorts. Furthermore, the
> #> more abstract, unsituated categories belong to a complex system and
> #> the less abstract, situated ones don't.
> #
> #But these properties follow from the differences in schematicity, don't
> #they?
>
> Yes. The properties that the less schematic categories have that the
> more schematic categories lack do follow from the differences in
> schematicity. But the fact that the complex system (phonology, grammar)
> involves only the more schematic categories does not follow from the
> differences in schematicity.

If I understand you correctly (which I may not), then I dispute this 'fact'.

>
> #> #The difference between 'situational' and 'non-situational' properties
> #> #boils down to the type-token distinction as far as I can see (you may
> #> #object, saying that your distinction is between 'context-free/purely
> #> #linguistic' and 'paralinguistic/pragmatic/..' properties, but that
> #> #amounts to trying to do [[language-as-grammar]-in-the-mind], and I won't
> #> #let you get away with it), and the type-token distinction is just a
> #> #question of generality.
> #>
> #> I can't see why you think it boils down to the type-token distinction,
> #> and in fact WG rejects the type-token distinction. We mentally represent
> #> situation types, don't we -- e.g. the class of situations where one's
> #> spouse comes home from work tired. That's not a memory of a single
> #> event -- it's a generalization over similar events. But the more
> #> similar the events are, the fewer properties of individual events are
> #> going to be bleached away in the generalized representation. If
> #> one's spouse always behaves virtually exactly the same way when
> #> arriving home tired, then the general representation will closely
> #> approximate the representation of any single event.
> #
> #Fine. We hope that the generalisation process preserves all general
> #properties, not just some of them. Your distinction (below) requires us
> #to learn that _I_ isa noun (linguistic information) differently from the
> #way in which we learn that it refers to its speaker (contextual information).
>
> The argument I was trying to make is that we have concepts -- e.g. "to be or
> not to be" and "I'll be back" (The Terminator) -- that are not just
> abstract grammatical patterns and not even abstract grammatical patterns
> with certain contextual features added. The concepts are each defined by a
> prototypical utterance whose features can in principle comprise every
> possible feature an utterance can have; they do not form a small and finite
> set that can be expressed using attributes like [Grammatical Structure= ],
> [Speaker= ], etc.

How's about we tried treating these concepts in the same way as words? Then I could have a word _I_ll_be_back_ (which has -accent->cod_austrian) and assign it all the comical properties I associate with it

That would be just like my word ONE, which refers to its speaker who has -self_importance->very, and I'd only need the one learning system.

Jasper

>
> When we come to study grammar -- *system* --, we find that its categories
> are so 'schematic' that they are of a qualitatively different kind. The
> upshot is that unacceptability of utterances is due sometimes to
> ungrammaticality and sometimes to deviation from a textual prototype.
>
> --And.