Message-Id: <199602042325.BAA21958@xiron.pc.helsinki.fi>
Date:         Sun, 4 Feb 1996 17:26:31 -0600
Reply-To: "Steven M. Belknap" <sbelknap@UIC.EDU>
Sender: Lojban list <LOJBAN@CUVMB.BITNET>
From: "Steven M. Belknap" <sbelknap@UIC.EDU>
Subject:      <lojban polti> JCB on change
To: Veijo Vilva <veion@XIRON.PC.HELSINKI.FI>
Content-Length: 15423
Lines: 260

coi

Regarding my <polti> construction, I want to point out that JCB wrote a bit
on single-source primitives (gismu in lojban). For what its worth, he
apparently accepted this activity for Loglan.

There has been a lot of carping about JCBs management of Loglan. I think we
need to also acknowledge the admirable audacity and genius of the man for
the mammoth task he undertook, and also acknowledge that perhaps his
approach may have been well thought out, if perhaps not perfectly carried
out. If we are going to learn from past errors, lets get all the different
perspectives on these past errors. I urge all interested in lojban to
obtain a copy of loglan 1 (fourth edition) by James Cooke Brown. There's a
lot of interesting stuff in there. You can get info from The Loglan
Institute about obtaining Loglan publications from the Loglan www page,
which has a hyperlink from the lojban www page. (Noblesse oblige)

It is a damn shame that JCB, McIvor and many others with much to contribute
are not working on lojban today. Perhaps if we were all fluent
Loglan/lojban speakers we could have been more successful in our <polti>.
This excerpt from his book could admirably serve as a guide for how a
lojban academy ought to deal with change.

quick scan summary of JCB on change:

"In the end some change in the fundamental structure of the language will
almost certainly be necessary. This may seem unfortunate; but the
alternative policy of freezing its early forms could well mean its early
death. For only by accommodating productive change is the language likely
to struggle through its own early traditions and come alive and grow."

Full text of JCB on change:

Throughout this chapter we have taken for granted that new content words
could be added to Loglan without changing the langusage in any essential
way. This is true; and it is true because, apart from their major partition
into names and predicates, content words in Loglan do not differ
grammatically in any other way. Thus, in any position in which one name
appears, any other name may also appear...perhaps not sensibly but always
grammatically. Similarly, and far more startingly, any predicate may be
replaced by any other predicate, and the replacement operation will leave
the basic structure of the sentence quite unchanged. As a consequence,
adding new names and predicates to Loglan *cannot* change the grammer of
the language. For the differences between new content words and old ones
can have nothing to do with grammar.

What this means is that prediciates and names constitute in Loglan just two
huge, rather amorphous "parts of speech." For their members are
grammatically interchangeable. But though they comprise the bulk of the
vocabulary, these are only two of about fifty parts of speech, or
word-classes, in Loglan grammer (24) Most of the other forty-odd
word-classes typically have only a small number of members, and about
half-of the word-classes, which includes all the punctuators, have only
one. But the main point is that approximately 48 of the 50 word-classes of
Loglan are occuped by little words and their compounds. We have called
these grammatically finely divided words the "structure words" of the
language. We must now consider what it means to change a structure word, or
to add to the existing set of them.

Let us review briefly what we mean by "structure word." In the three
preceding chapters of this book we have been exclusively concerned with
questions of grammar. What this has meant is that we have considered in
turn how each distinct group of little words coulld be used to determine
the over-all structure of Loglan utterances. Thus, whether an utterance was
a question, an answer, a claim about individuals, a claimm about masses, a
connection between claims, an imperativve, and so on, depended entirely on
the pattern and kind of the little words that appeared in it. Consider the
following utterance:

(1) Le mrenu pa ditca          The man was a teacher.

In both languages we know what is going on grammatically because of the
pattern of little words. We can reveal that pattern most cearly by
withdrawing the content words altogether:

(2) Le...pa...        The...was a...

Let us call such content-free expressions "grammatical frames". Now the
point is that thousands of content words in both languages can be freely
used to fill the blanks in such grammatical frames without in any way
changing the structure of their remarks. But what is that structure? In
Loglan it is the information carried by the little words in frame (2) that
tells us that a certain individual or set of individuals, whom the speaker
means to designate by mentioning one da's properties, was in the past
characterized by a certain property or quality, or by a relationship with
other individuals or sets which is here expressed incompletely. As a whole,
therefore, the frame alone expresses a kind of claim; moreover, the kind of
claim it makes is a predicting claim, not an identifying one. So much can
be told without any content whatsoever, that is, without looking at any
names or predicates.

Now it is true that words like le in Loglan and 'the' in English *can* be
replaced by certain other little words without significantly changing the
structure of the remark. For example, the grammatical frame below:

(3) Levi...na...        This...is now a...

does not differ signifficantly from that in (2). In fact, for most
purposes, the grammer of these two frames may be regarded as identical. For
in Loglan, at least, the two words le and levi, as well as the pair of
tense operators na and pa, may replace each other wherever they occur
without ever turnig a grammatical sentence into an ungrammatical one, or
*vice versa*. Thus, le and levi, like any two predicates, are also
grammatically interchangeable. But the important thing is that the group of
interchangeable words to which le and levi belong, unlike the group of
predicates, is a very small one.

Now what happens if you change, withdraw, or add members to a small class
of structure words? Some idea of the effect can be obtained by considering
the following playful changes in English. Suppose you try replacing the
word 'the' by the new word 'foo' in a few sentences. You may find it nearly
impossible even to begin. But now try replacing the word 'book' by the
similarly nonsensical word 'thoo' in a few remarks. This is not only
possible, but it can even be done with a certain heedless glee. (25) The
point is that structure words are too intimately associated with
fundamental linguistic habits for us to suffer their changes gladly.
Content words, however, can be added to, redefined, even wholly replaced,
without causing much pain to the human mind. A student in a first-year
course in almost any university subject spends half da's time learning new
content words; and some of them will be perfectly ordinary old words which
da must now learn to uderstand in a new way. But it is apparently quite
easy to do this. Similarly, any apprentice to a new craft must learn dozens
of new content words a month; and again some of the words will have old
meanings de must now reject. But no craft and no scholarly discipline can
dispense with words like 'the' or 'and' or 'but' and very few dare to
change or redefine them. None but the most impudent of sciences,
mathematics and logic, dare to change them. Logic abandons the structure
words of ordinary language almost completely...and therefore, perhaps,
provokes the greatest linguistic bewilderment in its students.

If we look at the grammatical types of the words that occur as slang, we
see the samme fundamental difference. There are plenty of slang nouns,
verbs, adjectives, and adverbs in English; but there are very few slang
articles or prepositions and no slang connectives that I know of. The
reason is clear. Structure words change very slowly. They may be added to
from time to time; but even this is historically resisted in most
languages. For example, the forms 'ain't' and 'this-here' have been around
for a long time and are still not quite accepted as "good English." In
contrast, note how short a time it has taken the word 'astronaut' to
install itself in the very lap of the language.

Now what this means for Loglan is that the arrangement of structure words
with which its lfe commences will (i) almost undoubtedly have to be changed
or added to, but (ii) those changes, at least, will be mightily resisted.
The second point is now obvious; but what about the first:

Let us first make the categorical prediction that such changes will be
necessary. For while the laboratory testing of the grammar--in a certain
sense, is engineering--has long been completed, the actual construction of
a human grammar inside human minds is a continuous, never-finished process;
and we will almost certainly find that there are unsolved structural
problems still. Mistakes will be found; ambiguities will arise; important
natural functions will turn out to be inadequately served. (26) A human
grammer is an immense affair, a mansion of many rooms, and we cannot claim
to have examined every nook and cranny of the structure we have designed.
Not only that, but somme questions--and some of which we are now aware--can
only be settled by use. For example, I am interested to see how loglanists
will handle their potentially "preposition-free" language. Will they use
case-tags? Or ignore them? If they use them, under what circumstances will
they do so? On another level, I sometimes fear that the connective system I
have designed may prove too intricate for human minds to handle in the
"real time" world of speech; and so on. If so, parts of that vast structure
will remain unused by the majority of speakers. Adjustments may have to be
made.

As questions of this kind are gradually being answered by experience or
experiment, some speakers of the language will want to change it to bring
its structure into harmony with those answers as they emerge. If such
changes involve its most deeply embedded structure words, there is some
question in my mind whether they can actually be executed without exciting
open rebellion among those who have learned these "erroneous" structures.
(27)

Some changes of course will be relatively easy to accomplish. Additions to
the list of modal operators--tie, lia, sea, and the like--should be nearly
as easily accomplished as the addition of a predicate. The list of
discursive operators--bea, pou, suit, and so on--should also prove easy to
augment. But additional case tags will, I predict, be more difficult to
introduce. Compound indicators may also be added with great freedom. And
the less frequently used tense and location operators may prove relatively
easy to augment or change. Perhaps even the rarer "punctuation marks" may
yield to pressure. But the connective system, or the tense system, or the
system of descriptive operators, will probably prove more stubborn. Let us
hope that these more fundamental structures of the grammar will now prove
to be well-engineered, and that such changes as take place in them will be
augmentative rather tha corrective.

In the end some change in the fundamental structure of the language will
almost certainly be necessary. This may seem unfortunate; but the
alternative policy of freezing its early forms could well mean its early
death. For only by accommodating productive change is the language likely
to struggle through its own early traditions and come alive and grow.

Notes

(24) In the technical argot of Loglan workers, classes of words whose
members are grammatically interchangeable with one another are called
*lexemes*, and members of any one such class are called its allolexes. So
the question in hand may be rephrased as, What happens to the grammer if
you add or delete an allolex from some lexeme? The answer: if there is
still one allolex left, nothing.

(25) 'This is a good thoo. I like these two thoos. This thoo is bigger than
that thoo' come easily enough. But not 'Foo man gave foo book to foo girl'.
One can hardly get one's tongue around the latter sequence, let alone one's
mind! The difficulty experienced in making such "allollexic"
replacements--as measured by overall time, hesitation, and the errors
committed in making them--mmight well be used as an experimental measure of
the size of the lexemes in a natural language, i.e., the number of their
allolexes." Loglan, having lexemes of known size, offers an opportunity to
test this notion.

(26) These predictions--originally made in 1975--have already been once
fulfilled. But the "important natural function" that was not well-served by
the 1975 language turned out not to be grammatical but morphological: the
1975-style complex predicates were not regularly decipherable, and so had
to belearned as quasi-primitives. This defect had wide-ranging
consequences, the most critical of which was that it slowed the acquisition
of this portion of the vocabulary. There were others. The discovery of this
raft of morphological problems during the 1977-78 learning trials led to
what was called the "Great Morphological Revision", a program to build and
test the best possible set of "decipherable affixes", an engineering
program that was to last through 1982. The unsolved problem in the grammar,
in contrast, was well-known in 1975: Loglan grammer was not formally, but
only heuristically, unambiguous. Discovering a formal proof of its freedom
from ambiguity, and disambiguating it where it was not, was a task that had
been necessarily left undone for want of powerful enough tools. This lack
was filled in 1975--the very year of Loglan's second public debut--by Aho,
Ullman and Johnson (1975). With the "autommatic parser generators" that
issued from their work, Loglan grammer was made, and demonstrated to be,
unambiguous in 1982, and has remained so every since.

(27) In fact, something like this actually happend in the course of the
re-engineering of Loglan morphology during the years 1978-1985, a project
called the "Great Morphological Revision" or GMR. A rebellion did occur--or
did very nearly--in response to the publication of Notebook 2, a report on
the early findings of the GMMR research. Some loglanists did not welcome
such fundamental changes in the morphological structure of their language.
See Brown (1983c)) for some reflections on why resistance of this kind
might be more likely to occur when changes in morphology are proposed than
when the changes proposed affect only grammear. I argue in that paper that
this observed difference in resistance to change, if it turns out to be a
real one, could very well reflec the very different evolutionary histories
of these two compartments of the human language-handling apparatus, the
early Pleistocene one, which was arguably evolved to handle the mimetic
acquisition of morphology and a proto-lexicon, as opposed to the late
(geologically) Recent one, evolved to handle the inductive acquisition of
grammar. I argue that mimesis and induction involve very different kinds of
mental programming, the first (in several senses) being essentially
conservative and change-resisting, the second, essentially exploratory and
change-welcoming.


Steven M. Belknap, M.D.
Assistant Professor of Clinical Pharmacology and Medicine
University of Illinois College of Medicine at Peoria

email: sbelknap@uic.edu
Voice: 309/671-3403
Fax:   309/671-8413