Cleaning up a mess: modal constraints

[John Cowan <cowan@SNARK.THYRSUS.COM>]

While revising the textbook, I bumped up against the section headed
"Comparative sumti tcita" in Lesson 6, sublesson 3.  This introduces
the now-obsolete cmavo "mo'u" in sentences like:

1)      la djan. *mo'u semau la betis. nelci la meris.
        John more than Betty likes Mary
        John likes Mary more than Betty does

vs.

2)      la djan. nelci la meris. *mo'u semau la betis.
        John likes Mary more than Betty.
        John likes Mary more than he likes Betty.

(Throughout this message, I will star "*mo'u" to avoid confusion with the
current cmavo "mo'u" of selma'o ZAhO, which has nothing to do with what's
going on here.)

The textbook explains "*mo'u" as introducing a "modal constraining phrase",
which is said to be neither truly restrictive nor truly non-restrictive.
It is not restrictive because (in Example 1) "more than Betty" in no way
restricts or specifies the meaning of "la djan."; John is already fully
identified.  But it is not truly non-restrictive either, because the presence
of "semau la betis." in this sentence does not merely add ancillary information.
If John likes Mary more than Betty, it does not follow that he likes Mary
(without reference to Betty).  This may be easier to see with "less than";
if I like George less than Frank, I may not like George at all.

When the textbook was reviewed by pc, he protested the notion of "neither
restrictive nor non-restrictive" as a bogosity.  As a result of other
reviews, some changes were made:  the cmavo "pe" and "ne" of restrictive
and incidental identity were changed to the current "no'u" and "po'u"; and
the cmavo "*mo'u", explained above, and the cmavo "po'u" of restrictive
association were changed to "ne" and "pe" respectively.  In this way,
"ne" came to signify non-restrictiveness.  This change was announced in
JL10.

When I read JL10, the muddy typeface caused me to read "mo'u" as "no'u", and
so I interpreted the change as a simple swap "no'u" <-> "ne", whereas in
fact "no'u" had previously been unassigned: there had been no non-restrictive
equivalent of the former "po'u", now "pe".  I later made erroneous remarks here
on Lojban List as a result.  In addition, my thinking was influenced by
this error into ignoring the meaning of "*mo'u", now supposedly "ne", and
instead inferring the meaning of "ne" as simply "non-restrictive pe".

In particular, when I argued for the abolition of the BAI cmavo "ci'a",
meaning "written by...", I constructed the following pair of sentences:

3)      mi nelci la .apasionatas ne fi'i la betoven.
        I like the Appassionata, created (finti) by Beethoven.

4)      mi nelci la .apasionatas. pe cu'u la rubenstain.
        I like the Appassionata performed (cusku) by Rubenstein.

In Example 3, the "created by Beethoven" was truly incidental information;
the "ne fi'i la betoven." meant the same as "noi la betoven. finti".
Removing it in no way changed the truth of Example 3.  In Example 4,
however, the corresponding "pe cu'u la rubenstain." meant "poi la
rubenstain. cusku", and pinned down exactly which (version of) the
Appassionata was meant.

What I did not see was that neither of these uses matched the semantics
of "ne" in Examples 1 and 2, as I had not looked back at Lesson 6 for
a long time.

Greg Higley spotted the anomalous use of "ne semau" (and the corresponding
"ne seme'a") in one of his September 1992 papers, and proposed that
"mau" and "me'a" be moved to selma'o JOI.  In his analysis, Example 1
would be replaced by something like:

5)      la djan. *mau la betis. nelci la meris.
        John more-than Betty likes Mary.

where "*mau" is a non-logical connective that cannot be split up.
Greg also pointed out a number of other ways in which comparison
could be used (between selbri, tenses, etc.) if the change were made.

Mark Shoulson and Colin Fine were the only people to comment; Mark
thought the analysis made sense but opposed it as involving too much
change to existing text; Colin proposed two sentences showing the use
of "semau" as a regular sumti tcita, not attached with "ne" to another
sumti:

6)      mi gleki semau tu'a le prujeftu
        I am-happier than last week

7)      ta xunre semau le karce
        That is red, more than the car

I do not understand Example 7: I can only take the English to mean "That
is redder than the car is" which requires "ta *mo'u le karce cu xunre".
Example 6 is more substantial, however: it says that the whole bridi
(my being happy) is more than some bridi not expressed, but using the sumti
"le prujeftu"; presumably "mi gleki ca le prujeftu" is the intent.
So no change was made.

The original problem was still lurking, however.  Elsewhere in JL10, in
lojbab's response to a letter (ironically also by Higley), he uses the
sentence:

8)      mi ne pu la djan. klama
        I, before John, came

as a way of saying "I came before John did."  Higley had rightly criticized
"mi klama pu le djan.", used in Lesson 4A, as invalid sumti-raising
(though without using that name), and lojbab was proposing several
alternatives, including an explicit abstraction and Example 8 among others.
("tu'a" was not yet born, and the whole sumti-raising problem was yet to be
understood.)

When discussing Example 8 with lojbab today, I saw in a flash that it
really meant:

9)      le nu mi klama cu purci le nu la djan. klama
        The event-of (I come) is-before the event-of (John comes)

and that the real function of this "ne" was the same as the "*mo'u" in the
textbook: it turned the underlying form, namely Example 9, inside out!
Similarly, the underlying form of Example 1 was:

10)     le ni la djan. nelci la meris. cu zmadu le ni la betis. nelci la meris.
        The amount-of (John likes Mary) is-more-than the amount-of
                (Betty likes Mary)

So in effect Higley was correct:  "ne semau" was functioning as a non-logical
connective in Examples 1 and 2.  But furthermore, "ne pu" was functioning
in the same way in Example 8!  This was altogether different from the
behavior of "ne fi'i" in Example 3.  So there was a genuine difference
between "ne", working as a non-restrictive associator, and the
connective "*mo'u" that needed recognition with a separate cmavo.  It was
not enough to move "mau" and "me'a" to JOI; potentially every modal tag,
including tenses, could be a connective.

Example 8 would then be interpreted "I (who am before John) come" as it
stands; the "ne pu la djan." would mean no more than "noi purci la djan."
To get the connective interpretation, the form would be:

11)     mi *mo'u pu la djan. klama

This change amounts to introduce a new kind of joik:  "*mo'u" followed by
any simple-tag (the grammar term for a modal or tense).
Making a formal grammar change would also allow some of Higley's other uses
for comparison to appear:

12)     ti xunre *mo'u semau narju
        This is more red than orange.

13)     mi pu *mo'u semau ca nelci loi vanju
        I in-the-past more-than now like to drink the-mass-of wine.

The cmavo "mo'u" is already committed to selma'o ZAhO, where it has a good
memory hook in "mulno".  Therefore, a new cmavo is needed.
Among the remaining unassigned cmavo, "ne'e" probably makes the most sense.
It suggests something related to, but distinct from, "ne".  However, lojbab
felt that "ne'esemau" might be thought too verbose, and we went on a search
to see if any monosyllabic CVV cmavo might be moved or freed.

The candidates:  "dai" (selma'o KOhA), "nau" (selma'o BY), "zau" (selma'o BAI).
None of these has seen live use.

"dai" belongs to the set of which "di'u" and "di'e" are the best known:
they refer to specific utterances (the last utterance, the next-to-last,
the next, etc.)  "dai" is the vague member of the set; it refers to
"something somebody said", past or future, actual or potential.  As such,
it is almost as unconstrained as "ko'a" or "da poi se cusku"; it is too
unanchored.  I would be in favor of dropping it altogether.  Alternatively,
if someone feels that this meaning must be represented as a cmavo, then
"do'i" is available.

"nau" cancels all shifts (upper case, lower case, different alphabets,
different fonts) that may be in effect for the letteral words of selma'o BY.
It can be used to clear out the current state when you don't know what it is.
This meaning probably does not need a monosyllable; in speech you can figure
out what shifts are in effect by context, and in writing you can look back
and see.  "na'a" would be a usable substitute.

"zau" is the BAI for "zanru", and could be replaced by "zi'u".

Of these, I like "nau" best; it begins with the right letter.  If "dai" were
also freed, it could be used as a short label for the "empathy attitudinal"
I talked about in a previous posting.  (The upshot of responses to that
posting was that we will have an empathy attitudinal; it will attach to
any other attitudinal to indicate that the speaker is expressing someone
else's feelings, usually but not necessarily the listener's; in particular,
in a story it is usually the characters of the story.)

So the points for comment are these:

        Should "*mo'u" return, making a modal/tense into a joik?
        If so, should it have "dai" or "nau" or "ne'e" or "zau"?
        If "dai", should "do'i" replace the current "dai"?
        If not "dai", should "dai" be used as the empathy attitudinal?


--
John Cowan      cowan@snark.thyrsus.com         ...!uunet!cbmvax!snark!cowan
                        e'osai ko sarji la lojban.