On Mon, May 2, 2011 at 8:03 PM, And Rosta <[log in to unmask]> wrote:
[...]
>> I don't understand why. It seems to me that all of the arguments
>> against Lojban apply equally to Fith.
>
> The rules of Fith grammar serve to take phonological words as input and
> yield predicate-argument structure as output. By contrast, Lojban's grammar
> assigns meaningless baroque structures to sequences of phonological words:
> the baroque structures are unlike anything found in natlangs, and they have
> no relationship to meaning; the so-called grammar is not really linguistic
> at all.

Ah. I understand the objection now.

>> But the reasons are entirely different- simple lack of skill with a
>> particular language vs. lack of the neurological machinery required to
>> develop that skill.
>
> No, I meant that Fithians would be much better than humans at speaking not
> only Fith but also natlangs, because of their superior neurological
> machinery. My general point being that the essential differences are in the
> memory capacities of the two species, not in the grammars of their
> languages.

Perhaps. I suspect that Fithians would have trouble with a lot of
aspects of human natlangs that we don't worry about much, though. It's
not just a matter of different memory capacity, it's a matter of
different underlying models of processing and data representation.

>> Yes. Not without additional rules to allow for re-ordering to take
>> advantage of optimization techniques, and additional rules introducing
>> additional levels of abstraction (I'm pretty sure this is why all
>> natlangs that I know of, although they vary somewhat in syntactic
>> classes, all have more than 2 of them). The same concepts could be
>> broken up into multiple simple sentences with shallow trees, but
>> sentences with analogous structure require far more tree complexity in
>> Palno than in English.
>
> What are "levels of abstraction" and "analogous structure"?

No abstraction: All you have are generic "concepts". Some of them are
complete on their own, others have "holes" in their definitions that
can be filled in by other concepts. The structure in which you link
concepts together by using some concepts to fill in holes in others
defines larger compound concepts. When there are no unfilled holes,
you have a sentence.

One level up: Some concepts are atomic, and some are predicates.
Predicates are concepts for relations between other concepts, and
usually have holes to fill (argument structures).

Another level: Some full-shell concepts are suitable as sentences
(clauses), and others are only suitable as arguments to other concepts
(phrases).

Another level: There're different kinds of predicates that make
different kinds of phrases or clauses, which can only be joined into
larger concepts in specific ways depending on what their type is
(moving semantic distinctions in function into grammatical
distinctions).

More different kinds of predicates (verbs, adpositions, adjectives,
adverbs) and more different kinds of clauses and phrases represent
higher and higher levels of abstraction, because I can rely on an
abstract class definition rather than the semantics of a particular
morpheme to determine function and predict structure.

Etc., etc. Of course, this is only a partial ordering, and I'm sure
there are lots of other ways of defining the basic levels of
abstraction besides the one I just chose to outline.

I'm using "analogous structure" to mean roughly structures in one
language that will replicate as closely as possible the same parse
trees as structures in another language.

> Could you give an example of a Palno sentence that requires a much deeper
> parsing stack than its English equivalent?

"More different kinds of clauses represent higher levels of abstraction."

Transformed into Palno-with-conjunctions grammar (two, variants
depending on how I handle the attributives):
"Kinds-nom.-top. different-nom., more-nom., and clauses-acc. of-nom.
levels-nom-top. higher-nom. and abstraction-acc. of-acc. represent."
"Kinds-nom., which-nom. different, more and clauses-acc. of,
levels-acc., which-nom. higher and abstraction-acc. of, represent."

This isn't bad (after all, it's including the modifications I made to
make it not too bad- they flatten out the parse tree). But in the
original two-syntactic-class variant:

"(((Kinds-top. different) more) clauses of) ((levels-top. higher)
abstraction of) represent."

Parenthesized to make the nesting levels obvious. This is a short
sentence, so it's not getting unwieldy yet, but you can start to see,
every additional adjective, preposition, or adverb in English adds an
extra nesting level to the strict-Palno equivalent. (Relative clauses
somewhat surprisingly don't add an extra level, because the Palno
versions just re-arrange the nesting order; conjunctions don't add any
extra X-bar levels, but they do require more memory capacity.) The
sullied-Palno versions have a lot more grammatical marking, but the
parse trees end up a lot flatter, pretty darn close to English
equivalents.

Note that I'm using topic marking as an equivalent for relational
operators or variable declaration without having to actually extract
multiple independent sentences. Without the topic marking, this:

"(((Kinds different) more) clauses of) ((levels higher) abstraction
of) represent."

means something like

"The fact that there are more instances of kinds being different
represents the fact that instances of levels being higher are of
abstraction."

Which is really confusing in English, and really easy to say in Palno,
but then, that's not the sort of thing one would want to say very
often, and I don't consider a strength that a language is really good
at tersely encoding things that you would never want to say, and less
good at things that you would like to say.

-l.