Re: [engelang] Xorban: Semantics of "l-" (and "s-" and "r-") - engelang messages

On Tue, Sep 11, 2012 at 5:25 PM, And Rosta <and.rosta@hidden.email> wrote:

Mike S., On 11/09/2012 20:46:

> On Mon, Sep 10, 2012 at 8:53 PM, And Rosta <and.rosta@hidden.email
> <mailto:and.rosta@hidden.email>> wrote: Mike S., On 09/09/2012 20:27:

>
>
>> I really don't see a problem with Xorban's implicitly bound
>> variables, which provide such a simple, succinct and
>> easy-to-understand means at leveraging the verbiage of previous
>> restrictions. They are much better defined than "he/she/it/they"
>> in English, and tend to obviate the problems found in things like
>> "John helped his[=Bob's] father fix his=[John's] car." Maybe there
>> is some greater value that comes from repeatedly rebinding the same
>> variables with the same restrictions sentence after sentence, but I
>> am not seeing it.
>
> The problem is that it requires the hearer to remember the most
> recent restriction for every vowel morpheme. That's a huge burden on
> memory. Most of the memory effort is wasted. And it violates the way
> human language works, because the mind throws away nonsemantic info
> as soon as it's been processed.
>
> Remembering or noticing that "a" explicitly loses its restriction
> outside the formal scope of its binding - which would happen
> routinely in even relatively simple Xorban sentences - seems a
> greater burden on human abilities, and is a far greater violation of
> the way human language works, than remembering simply that "a" is
> bound by "la mlta" after someone says "la mlta".

On the contrary! Syntax is highly sensitive to hierarchical syntactic structure. The syntactic aspect of the Xorban binding relation is perfectly normal natlang syntax -- in broadbrush terms --, whereas the name-based method is not.

> Imagine if we had to remember or notice that "he", "she" and "it"
> changed meaning based on something like the scope of its referent
> (and only scope) instead of based on introducing a newer, more
> salient referent (i.e. rebinding).

You mean, imagine a system like English anaphora? Like pronouns bound by quantifiers in English? Like wh-phrases binding their traces? Even personal pronouns are sensitive to syntactic scope -- sometimes, tho not always, to scope alone).

Like English anaphora with nonspecific referents as in "John likes cats. Alice likes them too", but without barring specific readings either. Xorban variables are resumptive and phonological. Traces of wh-phrases are the opposite, which makes it impossible to use them more than once. Even pronouns bound by universal quantifiers can seemingly appear outside their scope:

ra prna le lsre djnake. di djni tvlika.
Every person knows the story. John told them.

ra prna se pzzslce ctkake. si mnika so nmdoke ctkiko. le skne klmake.
Every person ate a slice of pizza. Some of them ate two [slices]. Then they went to the movies.

> ra je frmra se xsle pnsake drxake.
>
> It's hard to imagine that the average human has the mental parsing
> acuity to automatically take that to mean "Every farmer who owns a
> donkey beats things",

What can I say? You grossly underestimate human parsing ability -- beyond what seems to me the limits of reasonable difference of opinion. Humans have pretty much equal parsing ability (so not correlated with 'acuity'), do it automatically, and don't make mistakes (of assigning an ungrammatical structure to a grammatical sentence).

I accept that it may turn out that any natlang that satisfies the basic criterion of loglanghood must behave in unnatural ways, not parsed like ordinary human language. But hopefully that's not the case, and we'll only find out whether it is by trying to create designs based on the premise that a loglang can be processed like human language.

The implicit binding rule doesn't make Xorban any less a loglang, and it does make it more useful and briefer, given that "o'e" and rebinding variables are always available when something different is wanted. Therefore, even if it were possible for humans to avoid it, the implicit binding rule would still be worth supporting as it stands.

> especially since this is available:
>
> ra je frmra se xsle pnsake drxa[ko'e].

I think you're maybe thinking not of parsing so much as diachronic change. If the evolution of Xorban grammar were left to the vagaries of usage, would it come to pass that "ra je frmra se xsle pnsake drxake" would mean "beat the donkey"? Maybe, but very possibly with a concomitant change that allows drxake to attach within the scope of ra.

It takes no diachronic change at all to make "lu fu sa xrma bjra vska'aku. [la na xrma] xkra" un-Gricean when "xkro'e" is available and practically begging to be used. If implicit restrictions are officially discouraged (which is a separate question from whether they are practicably avoidable), then no one speaking correctly and cooperatively is going to say "sa xrma bjra. [la na xrma] xkra", ever, and a huge number of otherwise relatively briefer, useful, sensible sentences in normal contexts effectively become un-Gricean/joke sentences.

Just out of curiosity, of the two choices:

1. Free variables have implicit restrictions.
2. Free variables are illicit with a few sanctioned exceptions.

... which do you think is better?

> The implicit binding scheme gives a simple and useful and probably
> inevitable interpretation for afterthought free variables.

John's notion -- or, at least, a notion I arrive at through discussion with him -- of variables continuing with their binding beyond the syntactic scope of their binder, until rebound (or forgotten) strikes me as more simple, more useful, more usable and more inevitable than the implicit binding with perseverative restriction scheme. Not that I am yet ready to advocate such a scheme, but I am more persuaded of its cognitive naturalness.