Re: syntactic rules ( was: Re: [engelang] Xorban: Termsets) - engelang messages

I think I have been misunderstanding & even more than I thought I had. I used to think we agreed at least that Lojban was a bad loglang because it was so difficult (if not impossible) to recover FOPL for the sentences, because of how far it deviated from logical standards. I now think that &'s point is almost exactly the opposite: that Lojban fails to be a loglang because it deviates from the standards for languages too far in the direction of FOPL to be a real language. Leaving aside the issue of what a loglang is to be on either of our views, I wonder what this leaves for &'s loglang project (which Xorban on its present course is apparently not going to satisfy). In particular, I wonder if he can still get clear presentation of logical structure (whatever that means for him) with language like rules (under his definitions) and without significant anaphora. Since he holds that place binding is an essential part of the structure of the language but does not allow that variable binding is the way to show that, we have to consider what other methods he has available.
Taking the simplest case, the binding of argument places of single predicate, there seem to be three broad approaches. The first is to take the notion of place fairly literally, correlating with a spatial order and then bind the places by putting the terms in the same order. This does not lend itself well to dealing with two predicates, since the order might be different for the second. But even in the simple case, it requires anaphora of some sort if one term has to occupy more than one place in the predicate or if interaction among the terms forces a different order ("Everyone's mother loves him", for example). The second approach, declensions, avoids the order problem by having each term tagged for the place it holds (according to some language specific scheme, I suppose), wherever it happens to fall in the word order. It might even be possible to allow for multiple positions to be marked, avoiding anaphora. The third possibility, conjugation, would have the predicate marked in some way at each place picking which of the available terms was to fill that place. The interesting question then is how the terms are to be identified -- by their positions in the sentence, by characteristics of their referents or of their physical form, or ... . This again avoids both earlier problems, but when we come to picking means of identifying the terms in each position, we find ourselves repeating most of the categories that are used to create anaphoric particles (the connection overt in Indo-European languages, of course), so we have, in effect, an anaphora system only without the pronouns (perhaps) and an order system (or some other way of identifying the places involved) as well. And, of course, the various approaches can be mixed in any number of ways. So, the idea is feasible, if somewhat more complicated that straightforward use of variables. How it might be developed, especially with reference to sentential fusion, is now a vaible topic for discussion.

From: John E Clifford <kali9putra@hidden.email>
To: "engelang@yahoogroups.com" <engelang@yahoogroups.com>
Sent: Tuesday, October 9, 2012 5:00 PM
Subject: syntactic rules ( was: Re: [engelang] Xorban: Termsets)

One of the reasons I dislike theological arguments is that they seem always to force people to say the most ridiculous things. So, in this discussion we have one discussant claiming to know how languages really work, another offering syntax rules which do not distinguish between a quantifier and a conjunction, or, again, one saying that, though binding is an important syntactical relation, variables are not syntactical objects, while the other holds that variables are not what is bound and aren't really there after all. Still, variables are unusual (if not non-existent) in the overt categories of natural languages and seem to be restricted to (even definitive of?) constructed loglangs. So the devices created for natural languages may not deal with them well and we are forced to use the devices created for logic to proceed. Variables bring together a number of different phenomena in natural languages (or, from another point of view, underlie them). We might try to use all of these phenomena (or an adequate selection of them) in our logical language, but that seems to go against the general idea of revealing the logical structure of the the sentences of the language (I take this to be at least one part of what a loglang is meant to do). So, we have variables as a syntactical category with many manifestations to deal with and the standard of FOPL as a way to deal with them. The issue seems to be how to fit that standard into something more linguistic and the first answer seems to be "Don't; fit linguistics into the FOPL mold." Carrying from that point is now the task (hopefully avoiding Montague along the way).

From: And Rosta <and.rosta@hidden.email>
To: engelang@yahoogroups.com
Sent: Tuesday, October 9, 2012 4:33 AM
Subject: Re: [engelang] Xorban: Termsets

Jorge Llambías, On 09/10/2012 02:18:
> On Mon, Oct 8, 2012 at 8:32 PM, And Rosta<and.rosta@hidden.email> wrote:
>>
>> My version:
>> Binder, with two complements, binding a bindee, which is generally an
>> argument-place complement (tho not in case of jek, ji'uk). The binding
>> relation is expressed in the phonological form of the binder and the
>> argument-place.
>>
>> Your version (if I understood right, which I probably didn't):
>> Binder has three complements, one of which is a variable. It generally
>> binds complements of predicates; complements of predicates are variables.
>> The variable has multiple forms, all of the same shape; each form cliticizes
>> to the form of the syntactic word the variable is a complement of.
>> Syntactically it's one entity, but phonologically it gets expressed as many
>> different forms, of the same shape.
>
> My version would be: Binder has two complements (each one a formula),
> and it binds zero or more variables of its complements.
>
> Variables are not syntactic objects by themselves (they are not
> words). Which variables a formula has, and which variables a binder
> binds, is determined by the phonological form of the syntactic
> objects.

I understand your version, but it is a misanalysis: if language works one way and you say Xorban works another, even though it can be analysed as working a different way that is consistent with the way language works, then I conclude yours is a misanalysis, i.e. I conclude that Xorban does not work the way you say it does. The binding relation is syntactic, because it's semantically interpreted. Indeed it's one of the two fundamental relations of logical form (which is the syntax of a loglang). The binding relation must therefore hold between syntactic objects. I don't think Xorban has or needs variables, but if variables do enter into the binding relation then they are syntactic objects in their own right -- they are words. In my analysis there are no variables, but argument-places are words in their own right, because they can be bound. The phonological form is determined by the syntax.

> An atomic formula is a word and has one or more free variables (and no
> bound variables).
>
> An operator may add to, remove (i.e. bind) or leave unchanged the free
> variables of its complement formula(s). Which of those things it does
> depends on which operator it is. For example "fV" will add (mormally
> one) free variable to those of its complement formula, "na" will
> neither add nor remove any free variables from its complement formula,
> "rV" will remove (bind) any free variable(s) from its complement
> formulas that coincide with the variables expressed in its
> phonological form. So the formula resulting after the application of
> an operator can have more, fewer or the same number of free variables
> as the operator's complement formula(s). Variables bound within the
> complement(s) of an operator are invisible to the operator and are
> unaffected by it.

If you're going to start stating and formalizing the rules of syntax, this would be one of the first ones to tackle.

--And.