Jörg Rhiemeier, On 14/04/2011 22:04:
> On Thu, 14 Apr 2011 17:31:05 +0100, And Rosta wrote:
>
>> It's questionable if anybody has learnt to speak Lojban, for at least two
>> reasons. Firstly, the formal specification of Lojban syntax is highly
>> unnatural and is untaught. Most Lojban users use some kind of ad hoc
>> natlang-like grammar they've induced largely unconsciously. Some of the most
>> expert ones use (-- judging by their self-reports) some kind of idiolectal
>> simplified version of the official one. Secondly, Lojban has some elements
>> that imitate natural language and some elements that very much don't. Lojban
>> usage is heavily biased towards the bits that imitate natural language. One
>> unnatural-seeming feature that does see usage is phrase-final terminators,
>> but I suspect that this is evidence that the feature is in fact susceptible
>> to a naturalistic analysis.
>
> Oh, I didn't know that. If that is the case, it is likely that
> Lojban is in fact a Type I impossible language rather than just a
> Type II impossible one.
This conclusion would be too hasty. The formal specification of Lojban is incomplete -- the correspondence between sentence forms and sentence meanings is specified only informally. So Lojban qua formal specification simply isn't a grammar or a language or a blueprint for a language. And there is a speech community of people speaking what they call 'Lojban', as you noted, so if Lojban is what they speak then it is speakable.
I'd say that there are elements of the Lojban specification that are Type I impossible because of unnaturalness, namely the so-called formal grammar, and others that are barely usable because of over-reliance on short-term memory, namely its reference-tracking methods.
Brett Williams, On 15/04/2011 22:16:
> On Thu, Apr 14, 2011 at 12:31 PM, And Rosta<[log in to unmask]> wrote:
>>
>> It's questionable if anybody has learnt to speak Lojban,
>
> Erm. hello! I speak fluent Lojban.
That claim is questionable for the reasons I gave before, though it also depends what counts as 'Lojban'. See below.
>The only difficulty I ever have
> in understanding text in Lojban is with not remembering vocabulary,
> unless it uses the grammar really maliciously. I did have a lot of
> trouble with Lojban's grammar when I first started, but it's been
> about fifteen years since then, so I don't even really still remember
> what was so confusing about it. I'm not even sure which parts of the
> grammar seem unnatural from an outside perspective.
It's several years since I was accustomed to looking at Lojban usage, so some of my observations might be out of date (this is a hasty list drawn up just for this message, so I had better reserve the right to refine or modify some of these points in future discussion...):
* The formal grammar. It may be that expert speakers have induced something with with equivalent weak generative capacity, i.e. generating the same set of sentences, but I don't believe they assign the same structures to sentences as the formal grammar does. The formal grammar generates highly unnatural structures ('parse trees').
* Putatively syntactic structures assigned by the formal grammar often don't have semantic import (so the syntactic compositionality doesn't reflect semantic compositionality) and some important meanings are encoded by word-order alone and not by structures assigned by the formal grammar. In natural language, syntax is the mechanism that builds sentence meanings. So Lojban is either utterly unnatural or has an undocumented 'real syntax' or has no real syntax at all (with speakers just getting by on the kinds of heuristics used by nonfluent L2 learners).
* SE is highly unnatural, in that it not only promotes an argument to x1 but also demotes the former x1 to the former argument place of the new x1. I.e. it swaps argument places. Multiple SEs are mindboggling.
* FA is not at all unnatural, but sees relatively little usage (relative to FA-less sumti), so even though FA is fully part of the language specification, I suspect that because it is little used, speakers aren't fluent with it.
* Xorlo has rescued usage from being the garbage it used to be when following CLL rules, but achieves this by making it possible to do without quantifiers in logical forms: in pre-xorlo days, the meaning of the sentences people said was completely different from the meanings speakers intended to encode. That problem will have ameliorated, but I'd still suspect that the problem still persists, and that there exists a situation that can be described either as one in which hardly anybody is fluent or as one in which the language implicit in actual parole is widely deviant from the documented specification (aka 'the prescription').
* Although Lojban has several mechanisms for indicating when multiple predicates share the same argument, all are so clunky that in usage they're generally not indicated (-- unlike in predicate logic notation).
* As formally specified, all clause internal orders other than predicate-initial are equally unmarked, but in usage there's a strong tendency to use predicate-second order, and an even stronger tendency to avoid predicate-later (rather than predicate-sooner) orderings in structures with many nested clauses. Again, I suspect that 'fluent' speakers are fluent with frequent patterns and not with infrequent patterns. (This may seem obvious, but I think it's far more characteristic of 'intermediate' speakers of L2s than it is of fluent speakers of natural languages.)
* For some areas of meaning, such as conditionals and interrogatives, the correspondence between surface sentence form and logical form is not even informally documented; so these correspondences exist only in the langue implicit in parole.
* Lojban's obligatory pauses are completely unnatural. Reinterpreting these as a phoneme whose primary allophone is a glottal stop is more natural, but this phoneme is required in places where glottal stops are articulatorily or acoustically problematic. Again, I suspect that it's unlikely that many are fluent in production or comprehension of this phenomenon.
--And.
