Re: [lojban] Re: [lojban-announcements] Essay on the future of Lojban, with a simple poll for the community.

[Bob LeChevalier <lojbab@lojban.org>]

Christopher Doty wrote:
> I see two, maybe three, areas where there is a problem from a linguistic 
> perspective.  The first is that languages do not have verbs with more 
> than four unmarked slots for a predicate, and there are VERY few that 
> have four; the vast majority of verbs in the vast majority of languages 
> have three or less.  If you get more than four, you ALWAYS have some 
> sort of marking (most often as an oblique phrase; i.e., a preposition or 
> a postposition)) that indicates how the additional argument relates to 
> the predicate.  Yet, Lojban has gismu which take more than four 
> arguments.  If it were testable, I would put a LOT of money the fact 
> that, after Lojban was released into the wild, you could do a text count 
> and find that predicates rarely, if ever, have more than three arguments 
> in them, and that the three arguments pretty much always had the three 
> closest to the gismu.
>
> It is worthwhile to note, especially for those who like Lojban to be 
> mind-bending, that this fact likely has nothing to do with language, and 
> everything to do with cognition.  On average, working memory holds 
> something like 4-7 items (try using a phone menu with 9 items; it is 
> extremely annoying and frustrating, and makes it hard to do anything 
> except listen to the list of options).  It is thus no surprise that, in 
> languages, four is the maximum (three arguments and a verb, with a 
> couple verbs that take four), especially if one considers that 
> most utterances have more than just the verbs and the arguments. I think 
> this is what you meant by "processing depth"--the problem is that most 
> humans actually CAN'T PROCESS at the depth needed for a gismu with seven 
> places.  You could argue that this processing depth is learnable--maybe 
> it is, but I'd bet that learning to hold more in working memory is very 
> closely tied to how much you could process before any training.  This 
> also might be fine for a written language, since you can sit and look at 
> a sentence, but in speech, people just aren't going to be able to 
> process Lojban.

There aren't many Lojban words with more than 4 places, where there are, 
as with klama, the interrelationship of the places may help to keep them 
in mind destination-origin-route all go together, so when saying a 
Lojban bridi based on klama with all 5 places expressed, in my 
experience actually speaking the language, places 2-3-4 tend to unify 
mentally (and I have used all 5 places of klama in speech, though x5 
tends to be obvious most of the time and doesn't need to be spoken).  I 
think other verbs of motion that parallel klama work the same way.

It is easy to break this unity - terklama might be unusable as a 
predicate in fluent speech because it breaks up that unity.  We'll find 
out, won't we?

fanva is the other 5 placer that I have actually used in speech with all 
5 places.

I haven't used jutsi to express a full Linnean classification of a 
creature.  I doubt if I would do so other than in writing, but I imagine 
the ordering would allow the spoken form to be understood, even if there 
are 7 or 8 places (not sure of the maximum if one goes from subspecies 
to kingdom with all possible intermediates), and indeed omitting an 
intermediate would be what makes it incomprehensible no matter how well 
marked.

But the too-many-chunks-in-mind problem would already exist in Lojban, 
no matter  how many places are assigned to gismu.  Just start plunking 
abstraction bridi in a couple of the sumti of the main bridi, each with 
their own sumti, or add some relative clauses.  It is trivial to make a 
Lojban sentence "too complicated".

IIRC, the TLI equivalent of tikpa had 6 places.

> The second problem (or second half of this first problem) is that some 
> of the gismu seem to have tons of extra stuff in them that is not 
> something that would be included in the meaning of a word in any 
> language.  "Bucket," for example, contains a predicate slot for the 
> material the bucket is made from. This, as far as I could tell, was 
> thrown in to make the gismu have more slots.

No.  It was an attempt to be systematic, and to make the place 
structures easier to memorize.  A job I did imperfectly, I admit - the 
systematization came rather late.

The real reason is probably that JCB had a material place for some 
containers, and preserving compatibility with the TLI language for 
possible remerging was still a significant priority even after the 
original baselining of the gismu list, at which point people started 
actively fighting me if I wanted to change place structures even for 
good reasons.

Baselining and avoiding relearning were REALLY BIG issues in those days. 
 More people left the project, or refused to learn the language because 
we weren't yet willing to "stop tinkering, get the dictionary done, and 
let go of the language so people can use it", than for any other reason 
by far.  (CLL was originally a section of the dictionary that grew to 
book length).

> The material a bucket is
> made of has far less to do with bucketness than, say, all of the things 
> in klama have to do with going.  And why does "bucket" have it and not, 
> say, "bird"?

IIRC, all of the "container" gismu have a material place.  At least 
etymologically, I think that "glass" and "stein" are distinct from "cup" 
mostly in their material.  We had other examples, but it was a way to 
make one gismu cover a family of words in other languages.

The ability to use gismu in lujvo (including with SE and NU) to make 
multiple common words, in order to both keep the lexicon small and to be 
consistent with Zipf's law was a major factor in choosing the gismu.

> I can call something that isn't a living bird (say, a
> drawing of a bird), but why doesn't it a gismu slot to indicate it's 
> material?  If buckets get a slot for material, so should everything.

that is what BAI, and fi'o are for - to arbitrarily add any desired 
place to a predicate.

> Both of these things are easily fixed, though, without totally barfing 
> up Lojban.

The thought of even opening up the issue again brings precisely that 
phrase to mind.

Debates over change to the language get very heated very easily. 
"Fixing" is NOT easy for anyone who thinks about the people who aren't 
actively involved in the current discussion (Lojban List has over 400 
members; I doubt if any single thread ever gets more than 10% of them to 
comment.  CLL has sold over 500 copies, maybe even 600 by now, and I 
suspect that accounts for less than half of the people who try to learn 
the language.

> So, there 'tis--what the linguist doesn't like about Lojban (which, it 
> is worth noting, is far less than what he DOES like, but still).

The early debates, when linguistics considerations dominated more than 
logical perfection, would probably be interesting to you.  JL has a lot 
of it, early Lojban List has more, and references to Lojban on sci.lang 
and more rarely on Linguist List when I actively participated in both 
were especially relevant.

http://www.lojban.org/tiki/Why+Learn+Lojban%253F
has links to some of these at the bottom of the page.

--
Bob LeChevalier    lojbab@lojban.org    www.lojban.org
President and Founder, The Logical Language Group, Inc.

--
You received this message because you are subscribed to the Google Groups "lojban" group.
To post to this group, send email to lojban@googlegroups.com.
To unsubscribe from this group, send email to lojban+unsubscribe@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/lojban?hl=en.