Patrik Austin, On 29/07/2014 07:34:
>> The main objective of my PhD is to make a formalization of a
>> (non-Chomskyan) minimalist grammar, but obviously it's theoretically
>> a bit broader than that.
>
> [And:] > How is this not an already-solved problem? To put that another way, in what way does predicate logic not count as a minimalist grammar?
>
> It's not a grammar, is it? But if you can make a minimalist grammar
> based on predicate logic, then that would definitely count.

It's virtually a grammar. To turn it into a grammar you just have to devise a syntax and -- of course, but not really relevantly to this discussion -- a way for expressing it in phonological form. Since predicate logic notations already exist, it's blindingly obvious how at least some syntaxes for it can work. That's why I say it's a solved problem.

> You'd have to explain how to make questions, for instance, by giving
> grammatical rules, and whatever you're going to use. How would you
> formalize "If a tree falls in a forest, does it make a sound?"
> (etc.)?

Simple: just translate them into predicate logic, and then encoding that in our loglang is trivially simple. Asking "how can one formulate such and such in predicate logic" is worthwhile and interesting, but there's a whole discipline of formal semantics that addresses it, and the answers don't seem to be directly relevant to the question of what makes a minimalist grammar.

>> [And:] > (Are you shortly to embark on this PhD, or is its commencement still some years off? Have you actually had a proposal approved by a supervisory committee?)
>
> It's been approved – everything's fine - but they suggested I find
> one more supervisor. I could start this September but I decided to
> delay it to get a chance to apply for a scholarship. I think I'll
> write my first article any time, though. As soon as I'm ready.

Maybe reading the article would make me change my mind, but currently it looks to me as though it's not a viable project, because it has no evident prospect of finding a better answer than the one we think we already know. I don't know anybody more interested in this stuff than me, so this is not dismissiveness born of lack of interest. I suppose there might be a PhD on the question of why natlangs don't look like minimalist loglangs.

My own PhD proposal was, as I remember it, a mixture of unwitting flimflam, vagueness and wheel-reinvention, certainly not anything I would now judge a viable basis for a PhD, and indeed I then spent six years writing and abandoning theses on a range of different topics before the pending responsibilities of imminent paternity forced me to sit down and write one on whatever it was going to be easiest to sit down and write a thesis about. The experience of just about everybody tells me that a PhD goes much more smoothly if from the outset its plan is readily focusable, and looks to be achievable and sufficiently original.

> [And:] > The focus of much engelanging in this area has been on
> finding a concise and readily human-usable notation notation for it.
> The concisest notation I know of is 2D Livagian. The concisest
> possibly-human-usable notation whose conciseness is guesstimatable is
> probably Xorban.
>
> I'm having a look at Livagian on
> http://eaworld.conlang.org/relays/relay9/ring3/livagian/ . The
> problem for using it in a PhD is that Livagian is unpublished and
> it's hard to find information. Other than that the basic grammar
> looks simple so it could become a candidate.

Livagian is just the name of my conlang throughout its many incarnations and its many periods of disincarnation.

2D Livagian, tho, is a different thing, a notation and was described in this thread a few days ago.

> Xorban is published at least in Wordpress even if it's not finished.
> However, there seems to be a logical problem with both languages
> because their dictionaries have word-specific rules. This could mean
> that if you have a dictionary with an infinite number of words,
> you'll end up having an infinite number of grammar rules.

I don't think any inventor of a logical language would be so foolish as to allow this logical problem. Either there are no word-specific rules or word-specific rules pertain only to words in closed-classes (which necessarily constitute a finite grammar).

It's a shame that Xorban ground to a halt without more closely approaching a finished and acceptably documented form, but its main relevance to the discussion here is that it aimed to be concise rather than to have a maximally minimal grammar, tho I guess it did have an optimally minimal grammar. If your interest is in maximally minimal grammars of languages that unambiguously encode logical forms, you don't need to study prior art in conlangs, you just need to study logical forms (which is the work of an afternoon rather than years).

> [And:] > In what way is being a cultural phenomenon incompatible with being based on logical necessity? Who currently maintains that human language is essentially a cultural phenomenon not shaped by logical necessity?
>
> That's just what the professors keep telling me, but we two know it's ridiculous.

It's your professors that strike me as having the eccentrically rare view. I wonder if there is some contrarian Finnish school of linguistics that I have not previously encountered. Are there particular works or scholars that your professors like to cite as best expounding their view?
  
> [Daniel:]
Logan, rather.
> Figuring out the simplest possible human-usable
> encoding of predicate logic would probably count as an answer (as long
> as you could prove that it was simplest).
>
> Ok. But falsification has to work the other way: I will present my
> solution which can in turn be falsified by others. I will argue in
> detail why the model I've chosen cannot be made any simpler, but I
> can be mistaken, and there may be other possibly better models that
> me and the professors are not aware of.

If you intend to propose what you think is the simplest possible lossless human-usable encoding of predicate logic, that would certainly be very interesting. But the PhD's worth of interest would, it seems to me, be in psycholinguistic investigations of human-usability -- processability.
  
> [Logan:]
Me, rather.
> No sustained human-usability tests have been done, but my
> sense is that so far we have found no lossless human-usable encoding,
> and my money is on our eventually being able to prove that no
> lossless human-usable encoding is possible, tho I would love to lose
> that bet by discovering a lossless human-usable encoding. Obviously
> we do have lossless encodings, so what I am saying is that I think
> none of them are human usable. (Lojban has user-controllable
> lossiness, and users always go for high lossiness because less
> lossiness would be too much to cope with.) Partly this is because
> losslessness will overload working memory, and partly it is because
> the greater effort of using it would almost always outweigh the
> communicative benefit of losslessness.
>
> That's all true. If it turns out the grammar has to be a really
> tricky or 'alien' one, then that will be the conclusion of my study.

I take it that this tricky 'alien' grammar is not human-usable? How would you conclude that it *has* to be not human-usable?

> I think the question is genuinely interesting and I'm 100 % convinced
> we will find the answer if we try.

In what way will you find "the answer"? That is, what is the question such that you will find for it a better answer than any we currently have?

>My study could serve as a starting  point.  

And in what way would your study serve as a starting point? Among us we have many decades of experience in devising minimalist grammars (with unlimited expressive capability, plus unambiguity and losslessness). I think that experience shows us the following:

1. The simplest notation for the information a language must encode is 2D Livagian. (It uses the minimal number of symbol tokens and symbol types. But it does represent event arguments explicitly; I guess it might be possible to find an even simpler notation that doesn't.)

2. It is trivially easy to find a lossless unambiguous human-utterable encoding for this -- e.g. just speak standard predicate logic notation aloud -- but the probably unsolved and probably unsolvable problem is to find one that is reasonably ergonomic.

--And.