"mi nu'o citka ro dacti", "For any object, I would eat it." vs "mi nu'o citka piro loi dacti", "I would eat the whole of the mass of all objects."

At least, I think that's what you wanted. If you wanted "I would eat any one thing, but no more", that would be "mi nu'o citka ro lo pa dacti", I think.

>>995

nu'o ku ro da zo'u mi citka da
It could be the case that, for every x, I eat x
"I could eat everything."

ro da nu'o ku zo'u mi citka da
For every x, it could be the case that, I eat x
"I could eat anything."

I would say "mi nu'o citka ro da" is the first one. For the reverse scope: "mi ro da nu'o citka".

"Would" is a bit more tricky than "could", as it involves some kind of implicit "if". Perhaps with "va'o":

va'o ku ro da zo'u mi citka da
(Under the right circumstances) it would be the case that, for every x, I eat x
"I would eat everything."

ro da va'o ku zo'u mi citka da
For every x, (under the right circumstances) it would be the case that, I eat x
"I would eat anything."

>>996

> I would say "mi nu'o citka ro da" is the first one. For the reverse scope: "mi ro da nu'o citka".

I think both of those mean "nu'o ku ro da zo'u". To get "ro da nu'o ku zo'u" I think you'd have to say "mi citka ro da nu'o ku" or "mi ro da nu'o ku citka", in the same way that you have to use "na ku" instead of "na" to get the parallel effect.

Note that this contradicts what I said in >>995. I was wrong in >>995. I'm beginning to think Lojban would be much more intuitive if "CAhA BRODA", "NA BRODA", and similar constructs implied a sumti at the right-hand end of the prenex instead of the left. Quantifier boundaries would then only be created if explicit sumti (e.g., "naku" or "nu'oku") were used.

>>1004

>> I would say "mi nu'o citka ro da" is the first one. For the reverse scope: "mi ro da nu'o citka".
>I think both of those mean "nu'o ku ro da zo'u".

That would be sort of consistent with the CLL na-rule, though I don't think there's much support for it in CLL. In fact, there are places where CLL wants "na" to have scope over anything else, so that "na nu'o citka" and "nu'o na citka" would end up being the same
http://jbotcan.org/cllc/c5/s13.html
but there are other places where a little room for sanity is left
http://jbotcan.org/cllc/c15/s2.html

> To get "ro da nu'o ku zo'u" I think you'd have to say "mi citka ro da nu'o ku" or "mi ro da nu'o ku citka", in the same way that you have to use "na ku" instead of "na" to get the parallel effect.

Fortunately, we don't disagree about those.

> I'm beginning to think Lojban would be much more intuitive if "CAhA BRODA", "NA BRODA", and similar constructs implied a sumti at the right-hand end of the prenex instead of the left.

"Intuitive" is hard to pin down, because depending on whether you use "any" (which has wide scope) or "every" (which has narrow scope) for "ro" you may get different intuitions. And that's just coming from English, other languages may have other ways of dealing with such scope issues. I'm satisfied with a simple rule that works for all cases and doesn't need new exceptions for special cases.

> Quantifier boundaries would then only be created if explicit sumti (e.g., "naku" or "nu'oku") were used.

"Always move selbri operators to the end" would have the mirror problem of "always move selbri operators to the beginning", because it is also possible for two selbri to share tail-terms, although in actual use that case is much less common.

>>1006

My guess is http://jbotcan.org/cllc/c5/s13.html is less authoratitive than http://jbotcan.org/cllc/c15/s2.html. The contradiction should be added to the CLL errata.

> "Always move selbri operators to the end" would have the mirror problem of "always move selbri operators to the beginning", because it is also possible for two selbri to share tail-terms, although in actual use that case is much less common.

I'm a little unclear on what the issue is. Could you give an example of the problem?

>>1008

> I'm a little unclear on what the issue is. Could you give an example of the problem?

su'o da ge na dunda ro de poi cukta gi na lebna ro de poi bolci vau su'o di

This is how I would interpret it, following always the same rule that all operators have scope over following operators:

su'o da zo'u ge na ku ro de poi cukta ku'o su'o di zo'u da dunda de di gi na ku ro de poi bolci ku'o su'o di zo'u da lebna de di

>>1010

> su'o da ge na dunda ro de poi cukta gi na lebna ro de poi bolci {ku'o} vau su'o di

Now hold on a minute. If you start with "da ge na dunda ro cukta gi na lebna ro bolci vau di" (omitting su'o for brevity, which I hope you think is allowed), I think the CLL interpretation is "da zo'u ro dexipa poi cukta zo'u ro dexire poi bolci zo'u di zo'u ge naku zo'u da dunda dexipa di gi naku zo'u da lebna dexire di". Binding the same variable to two different things in the same expression is probably a bad idea unless you explicitly put the quantifiers in prenexes with non-overlapping scopes.

>>1011

> Now hold on a minute. If you start with "da ge na dunda ro cukta gi na lebna ro bolci vau di" (omitting su'o for brevity, which I hope you think is allowed),

Let's not, in these discussions, for the sake of clarity.

>I think the CLL interpretation is "da zo'u ro dexipa poi cukta zo'u ro dexire poi bolci zo'u di zo'u ge naku zo'u da dunda dexipa di gi naku zo'u da lebna dexire di".

I don't remember CLL having any interpretation for a bridi that includes both logical quantifiers and connectives. Does it say anywhere which has scope over which? I take these two:

la djan .e la meris cu prami su'o da

ro le re prenu cu prami su'o da

as being more or less equivalent. Do you take the first as saying that there is someone that both John and Mary love? (If so, how would you say that each of them loves someone?)

> Binding the same variable to two different things in the same expression is probably a bad idea unless you explicitly put the quantifiers in prenexes with non-overlapping scopes.

With my interpretation they have non-overlapping scopes.

>>1017

> I don't remember CLL having any interpretation for a bridi that includes both logical quantifiers and connectives. Does it say anywhere which has scope over which?

http://jbotcan.org/cllc/c16/s10.html#e10d3
Examples 10.3 through 10.5, and the paragraph immediately following.

From a different perspective:
http://jbotcan.org/cllc/c16/s14.html

>I take these two:
> la djan .e la meris cu prami su'o da
> ro le re prenu cu prami su'o da
> as being more or less equivalent. Do you take the first as saying that there is someone that both John and Mary love? (If so, how would you say that each of them loves someone?)

Yes, I take them as different. "There exists X such that John and Mary love X", and "For each of two persons Y, there exists an X such that Y loves X".

la djan prami su'o da ije la meris prami su'o de
la djan ce'e su'o da pe'e je la meris ce'e su'o de prami
nu'i ge la djan su'o da nu'u gi la meris su'o de nu'u prami

Ick! I think the system for termset logical connection in Lojban is horrible. (Why is the forethought connective inside the forethought termset?!?)

ro me la djan a la meris cu prami su'o da
ro lu'a la djan ce la meris prami su'o da

"The two people love the same thing."
ro le re prenu ce'e su'o da prami
su'o da zo'u ro le re prenu prami da
su'o da selpa'i ro le re prenu

>>1022

> http://jbotcan.org/cllc/c16/s10.html#e10d3

Ugh. So according to CLL "mi na prami roda" = "I don't love everyone", but "mi na prami roda .ije ke'u mi na prami roda" = "I don't love anyone". Horrible.

OK, so CLL does seem to have a rule that quantifiers scope over connectives. That's one more thing to add to my list.

>>1027

> Ugh. So according to CLL "mi na prami roda" = "I don't love everyone", but "mi na prami roda .ije ke'u mi na prami roda" = "I don't love anyone". Horrible.

That's why I was saying it would be more intuitive if "na" implied a "naku" at the end of the prenex.

>>1029

> That's why I was saying it would be more intuitive if "na" implied a "naku" at the end of the prenex.

OK, I agree that would be a slight improvement. But there is still weird stuff that doesn't depend on "na":

"do prami noda .ije mi prami da" = (according to CLL rules) "there isn't anyone you and I both love".

I am definitely not following the CLL rules of scope for logical operators.

> "do prami noda .ije mi prami da" = (according to CLL rules) "there isn't anyone you and I both love".

Correct. It's identical in meaning to "do e mi prami no da", which seems like a good idea to me. I guess you think it should mean "You love nothing and I love something.", which would be "do prami noda ije mi prami de" in CLL Lojban. How would you expand "do e mi prami no da" to a logical connection between bridi without changing the meaning?

>>1055

> How would you expand "do e mi prami no da" to a logical connection between bridi without changing the meaning?

ge no da zo'u do prami da gi no da zo'u mi prami da

Same meaning as: ro lo re mi'o prami no da

In fact ".e" is always equivalent "ro lo re", and ".a" to "su'o lo re". The full list of equivalences is:

.e -- ro lo re
.a -- su'o lo re
na.enai -- no lo re
na.anai -- me'i lo re
.onai -- pa lo re
.o -- vei ro .a no lo re

(The last one is a bit ugly.) All other connectives are not symmetrical, and so don't have a quantifier equivalent.

>>1057

Argh. I forgot that in your version of Lojban "do e mi prami no da" means "You love nothing and I love nothing" instead of "There is nothing you and I both love."

ge no da zo'u do prami da gi no da zo'u mi prami da
CLL: do a mi prami no da
You: do e mi prami no da

no da zo'u ge do prami da gi mi prami da
CLL: do e mi prami no da
You: ?

The simple fact of it is that it does not bother me that in English one says "You and I love nothing." to mean "You love nothing and I love nothing." instead of "There is nothing you and I both love.", whereas the CLL Lojban (oh how I wish I could stop qualifying "Lojban") structural equivalent of "You and I love nothing" means "There is nothing you and I both love" instead of "You love nothing and I love nothing". Lojban is different from English, and I'm OK with that. I don't see any practical objection to using the CLL interpretation other than "It doesn't mean what I expected it to mean, based on my experience of English.".

>>1063

> no da zo'u ge do prami da gi mi prami da
> CLL: do e mi prami no da
> You: ?

no da se prami do .e mi

Surface order reflects scope order: what comes later is within the scope of what came first.

> I don't see any practical objection to using the CLL interpretation other than "It doesn't mean what I expected it to mean, based on my experience of English.".

That would be a bad objection, I agree. My objection is different: "It doesn't mean what I expect it to mean, based on the internal working of the Lojban syntax."

So, for you, "no da se prami do .e mi" doesn't mean "ge no da se prami do gi no da se prami mi". A lot of the CLL rules you don't agree with help allow any logical connective to be expanded to a logical connective between bridi. It's an important property of the language, since it makes it easier to find the predicate logic equivalent of some Lojban expressions.

> Surface order reflects scope order: what comes later is within the scope of what came first.

Why? I don't see any reason not to allow operators which have scope over previous things (e.g. infix operators), or syntactic constructs which have a semantic scope larger than their syntactic scope, or semantic operators which have a backwards-extending scope larger than their syntatic scope.

> That would be a bad objection, I agree. My objection is different: "It doesn't mean what I expect it to mean, based on the internal working of the Lojban syntax."

Why should the semantics of Lojban be restricted by the context-free grammar used for the syntax of Lojban? For example, setting a sticky tense gives that tense a semantic scope extending well beyond its syntactic scope.

I guess what I'm saying is that objections in the form "It doesn't mean what I expected based on the syntax." are meaningless. It's convenient when syntax and semantics line up, but it isn't always possible, or even a good idea, and syntax ultimately has no semantic content.

>>1068

>So, for you, "no da se prami do .e mi" doesn't mean "ge no da se prami do gi no da se prami mi".

Right. CLL and I both agree on that one, because the quantifier appears before the connective.

> A lot of the CLL rules you don't agree with help allow any logical connective to be expanded to a logical connective between bridi.

Any logical connective can be expanded to a logical connective between bridi, with either CLL or my rules. Logical connectives always connect two bridi, "na"/"naku" always negates a bridi, and quantifiers always quantify a bridi. The difference is in the order in which they do it.

> It's an important property of the language, since it makes it easier to find the predicate logic equivalent of some Lojban expressions.

My rules make it easier to find the predicate logic equivalent. CLL rules make it harder, because they are more arbitrary, so one never knows for sure what the rules are supposed to be. There isn't one simple rule like "every operator has scope over all following operators acting on the same bridi".

How do you know, for example, that:

ro da poi prami ro de cu censa
= ro da zo'u ganai ro de zo'u da prami de gi da censa

and not:

ro da poi prami ro de cu censa
= ro da zo'u ro de zo'u ganai da prami de gi da censa

(Are we even sure that's the CLL rule?)

After all, CLL tells us that:

ro da zo'u ganai da prami ro de gi da censa
=roda rode zo'u ganai da prami de gi da censa

How do we know, if we don't pay attention to the syntax, whether the "ro" inside the relative clause has scope restricted to the relative clause bridi and not wider?

>> Surface order reflects scope order: what comes later is within the scope of what came first.

> Why?

Because that way you never run into inconsistencies. You don't have to introduce new rules at every step to cover cases that were not covered before.

> I don't see any reason not to allow operators which have scope over previous things (e.g. infix operators),

Infix (afterthought) operators have scope over their arguments, of course. They are always strictly equivalent to their forethought form.

> or syntactic constructs which have a semantic scope larger than their syntactic scope, or semantic operators which have a backwards-extending scope larger than their syntatic scope.

The only reason is simplicity. You know what to expect. You don't need to learn a new rule for every new case.

For example, what happens with quantifiers inside abstractions? Do you know what the CLL rule about them is? You will probably have to check, because in principle it could be anything. There is no general rule that you can trust.

> Why should the semantics of Lojban be restricted by the context-free grammar used for the syntax of Lojban?

It isn't always, in fact. The question should rather be: why go against the syntax when there doesn't seem to be any reason given for doing so? If there was a good reason for playing around with the scopes of negation and quantifiers, then I might go along with it. But there isn't any reason given. It just happened that someone who didn't understand very well what he was doing came up with some crazy rules just because.

> For example, setting a sticky tense gives that tense a semantic scope extending well beyond its syntactic scope.

Yes. (That has its own problems, but I don't object to that. There is a reason for having it.)

> I guess what I'm saying is that objections in the form "It doesn't mean what I expected based on the syntax." are meaningless. It's convenient when syntax and semantics line up, but it isn't always possible,

In the case under consideration, it is possible.

> or even a good idea,

In the case under consideration, it is a good idea.

> and syntax ultimately has no semantic content.

We'll leave that one for another debate. :)

>> So, for you, "no da se prami do .e mi" doesn't mean "ge no da se prami do gi no da se prami mi".

> Right. CLL and I both agree on that one, because the quantifier appears before the connective.

Actuall, I think the CLL says those two are equivalent. "ge no da se prami do gi no da se prami mi" is not the same as "ge no da se prami do gi no de se prami mi".

http://jbotcan.org/cllc/c16/s14.html

> It just happened that someone who didn't understand very well what he was doing came up with some crazy rules just because.

That's pretty prejudicial. I think the rules were constructed to make it easy to turn common constructs in second-order logic into compact Lojban. E.g., it's standard to put all quantifiers at the beginning of their expression in 2nd-order logic, so making that the default in Lojban makes sense as well.

>>1088

>>> So, for you, "no da se prami do .e mi" doesn't mean "ge no da se prami do gi no da se prami mi".
>> Right. CLL and I both agree on that one, because the quantifier appears before the connective.
>Actuall, I think the CLL says those two are equivalent.

Wait, don't the CLL rules say that:
no da se prami do .e mi
= noda zo'u ge da se prami do gi da se prami mi

?

>"ge no da se prami do gi no da se prami mi" is not the same as "ge no da se prami do gi no de se prami mi".

>http://jbotcan.org/cllc/c16/s14.html

That doesn't make sense in predicate logic (variables can't be re-bound like that). But even if we accept that strange use of variables, that would still not be equivalent to "noda zo'u ge da se prami do gi da se prami mi".

"ge no da se prami do gi no de se prami mi" would say "you don't love anyone, and of those that you don't love, I don't love any", which is not the same as "there is nobody loved by both of us". If you only love one person and I only love someone different the first is false but the second true.

>> It just happened that someone who didn't understand very well what he was doing came up with some crazy rules just because.
>That's pretty prejudicial.

The main culprit (i.e. lojbab) is proud that he flunked logic in college. He will be the first to admit he doesn't understand this stuff very well. That doesn't mean he hasn't done a great job with Lojban, but these rules are just crazy.

>I think the rules were constructed to make it easy to turn common constructs in second-order logic into compact Lojban. E.g., it's standard to put all quantifiers at the beginning of their expression in 2nd-order logic, so making that the default in Lojban makes sense as well.

I don't see how that is accomplished any better with CLL rules than with something more standard.

Here is another "interesting" effect of the CLL rules:

"noda mulno prane ije node mulno tolprane"

That seems a simple conjunction "nothing is completely perfect and nothing is completely imperfect".

But let's see what CLL rules say it means:

noda node zo'u ge da mulno prane gi de mulno tolprane
=roda su'ode zo'u ge da mulno prane gi de mulno tolprane
"For every perfect thing, there is some imperfect thing."

Who would have guessed?

Another problem: Consider the equivalence relationships:

roda = noda naku = naku su'oda naku = naku me'ida

and all other similar ones. With my rules those relationships always hold. With CLL rules, they hold if the terms are in the prenex, but if they are in the matrix they may or may not hold, depending on the context.

>>1055

> It's identical in meaning to "do e mi prami no da", which seems like a good idea to me. I guess you think it should mean "You love nothing and I love something.", which would be "do prami noda ije mi prami de" in CLL Lojban.

ta'a ta'o

Can we be certain "no da" means "nothing"? What about "zi'o"? How different are these bridi:

do prami no da

do prami zi'o

>>1091

> Can we be certain "no da" means "nothing"?

Yes.

> What about "zi'o"?

What "zi'o" does is eliminate a place from the place structure. It essentially converts the selbri into a new selbri with fewer arguments. If the argument was essential to the meaning of the relationship, the result can be hard to interpret.

> How different are these bridi:

> do prami no da

"You love nobody."
"You don't love anyone."

> do prami zi'o

"You are a lover." (where "being a lover" does not, for some reason, necessarily involve any object of love.)

>>1090

> Wait, don't the CLL rules say that: no da se prami do .e mi = noda zo'u ge da se prami do gi da se prami mi?

Yes.

> "ge no da se prami do gi no de se prami mi" would say "you don't love anyone, and of those that you don't love, I don't love any",

I assume you meant the version with two uses of the same variable.

Alright, I screwed up. Doing logic on empty categories can be tricky. So, go back to >>1068 and change it to "So, for you, no da se prami do e mi doesn't mean do e mi prami no da, in exactly the same way that the structurally equivalent English versions don't mean the same thing.". Whereas the Lojban versions are logically equivalent.

> That seems a simple conjunction "nothing is completely perfect and nothing is completely imperfect".

Only if you think the Lojban should map perfectly to the English. It's a different language; you can't expect everything that has the same structure to mean the same thing.

As stated earlier, reasoning with empty categories can be tricky, and saying "There's no X there's no Y" does imply "For all X there's a Y". That's just a consequence of logic, not a problem with Lojban. You should have no problem constructing an equally superficially astonishing example of that identity in your version of Lojban.

> With CLL rules, they hold if the terms are in the prenex, but if they are in the matrix they may or may not hold, depending on the context.

Incorrect. The equivalences always hold. Mapping the quantifiers to their position in the prenex is a little more complicated, but on the other hand, under your rules, quantifiers have the minimum possible scope.

By CLL rules "ge ro da e su'o de prami do gi de xebni mi" means "For all X, there exists Y, such that X and Y love you and Y hates me."

Correct me if I'm wrong, but for you that means "For all X, X loves you, and there exists Y such that Y loves you, and Y hates you.", where the second "de" could be replaced with "di" (or the second "Y" with "Z") without changing the meaning. Who would have guessed? (Either that, or you can only determine your quantifiers' scopes depending on context.)

>>1100

>Only if you think the Lojban should map perfectly to the English. It's a different language; you can't expect everything that has the same structure to mean the same thing.

Forget English. English is not even my native language. I expect "bridi1 ije bridi2" to be the conjunction of bridi1 and bridi2, i.e. to be true if both bridi1 is true and bridi2 is true. I don't expect quantifiers to jump out of them and mess things up.

>As stated earlier, reasoning with empty categories can be tricky, and saying "There's no X there's no Y" does imply "For all X there's a Y". That's just a consequence of logic, not a problem with Lojban.

I have no problem with "no da no de" meaning "ro da su'o de". My problem is with the "no da" from one bridi combining with the "no de" of another conjoined bridi in that way.

>You should have no problem constructing an equally superficially astonishing example of that identity in your version of Lojban.

There's nothing very astonishing in the transformation from "for no x is there no y such that ..." to "for every x there is some y such that ...". Or from "no person has no mother" to "every person has some mother".

>> With CLL rules, they hold if the terms are in the prenex, but if they are in the matrix they may or may not hold, depending on the context.

>Incorrect. The equivalences always hold.

No, not in the matrix, only in the prenex.

"ge noda blabi gi da xekri" will not be equivalent to "ge roda naku blabi gi da xekri".

The first is, according to CLL:

no da zo'u ge da blabi gi da xekri

The second is:

ro da zo'u ge da na blabi gi da xekri

The first one says that nothing is both black and white. The second one says everything is black and is not white.

So, by CLL, "noda" is not equivalent to "roda naku" in the matrix of the bridi.

>Mapping the quantifiers to their position in the prenex is a little more complicated, but on the other hand, under your rules, quantifiers have the minimum possible scope.

Not "minimum possible". Just the scope corresponding to their position. In "noda blabi gi'e xekri" minimum scope would give "nothing is white and nothing is black", but my rule gves "nothig is both black and white". That's maximum scope in this case.

> By CLL rules "ge ro da e su'o de prami do gi de xebni mi" means "For all X, there exists Y, such that X and Y love you and Y hates me."

Right.

> Correct me if I'm wrong, but for you that means "For all X, X loves you, and there exists Y such that Y loves you, and Y hates you.", where the second "de" could be replaced with "di" (or the second "Y" with "Z") without changing the meaning.

Correct.

> Who would have guessed?

You did! :)

(To get the first meaning I just say: "ge ro da prami do gi su'o de xebni mi", which is also valid by CLL and more clear.)

>>1093

>> Can we be certain "no da" means "nothing"?
> Yes.

How does "no da" negates the possibility of there being "de" or "di" as the object of prami? If I say "do prami no da .i do prami de", doesn't that mean "You don't love something, but you do love something else" rather than "You love nothing, but you love something"?

>>1103

>How does "no da" negates the possibility of there being "de" or "di" as the object of prami?

"do prami no da": "the number of things/people you love is exactly zero", "you love noone".

> If I say "do prami no da .i do prami de", doesn't that mean "You don't love something, but you do love something else" rather than "You love nothing, but you love something"?

No. "You don't love something" is "su'o da zo'u do na prami da", i.e. "there is at least one thing such that you don't love that thing".

>>1104

>> If I say "do prami no da .i do prami de", doesn't that mean "You don't love something, but you do love something else" rather than "You love nothing, but you love something"?

> No. "You don't love something" is "su'o da zo'u do na prami da", i.e. "there is at least one thing such that you don't love that thing".

So, "do prami no da .i do prami de" are contradictory claims?

>>1106

>So, "do prami no da .i do prami de" are contradictory claims?

Exactly.

>>1102

I was getting a little irritable in my last post. I apologize. Let me see if I can summarize without getting snarky again. You're concerned about the fact that the syntactic scope of quantifiers does not match their semantic scope. ("I don't expect quantifiers to jump out of them [bridi] and mess things up.") You want semantic relationships to be exactly mirrored in syntactic relationships (e.g., "noda", the text, when transformed to "roda naku", the text, should have the same semantic effect no matter where the text appears.

I see the value of the principle (that syntax should mirror semantics). However, I think the rule that a quantifier in the outermost prenex can be attached to the first instance of the quantified variable in the expression is good, because it allows one to avoid repeating the name of the variable. Since you can rewrite any expression with all the quantifiers moved to the outer prenex, this rule can potentially save a lot of characters.

You already accept quantifiers that jump from being attached to sumti, to being in a prenex with scope over one bridi, why not to being in a prenex with scope over the whole expression?

> (To get the first meaning I just say: "ge ro da prami do gi su'o de xebni mi", which is also valid by CLL and more clear.)

My attempted point there was that, as far as I can tell, the only way your version of Lojban can use the same variable in different parts of a logical connective is to explicitly quantify the variable in a prenex with scope over the whole conjunction.

Here's another attempt, where I avoid having "roda prami do" allow you to get rid of "su'ode prami do":

roda prami su'ode ijo de xebni mi
== roda su'ode zo'u go da prami de gi de xebni mi
== "For all X, there exists Y such that X loves Y if and only if Y hates me."

(Changed the connective from ije to ijo to prevent you from rewriting it as "roda prami su'ode poi xebni mi".)

>>1109

> I was getting a little irritable in my last post. I apologize. Let me see if I can summarize without getting snarky again.

No need to apologize, feel free to get a little snarky from time to time (I know I do). I'm happy to discuss these issues because I think they need discussing, even though they are mostly theoretical because they don't affect most of what Lojban is actually used for.

> Since you can rewrite any expression with all the quantifiers moved to the outer prenex, this rule can potentially save a lot of characters.

Let's compare:

With my rules: "noda xamgu prane ije node xlali prane"

With CLL rules: "roda naku xamgu prane ije rode naku xlali prane"

CLL rules take more characters than mine in this particular case. And saving characters cannot in any case be the ultimate goal at the expense of everything else, even clarity.

BTW, would you say the KUs are required by CLL there? Which rule has priority, the CLL selbri-na-rule or the CLL quantifier-over-connective-rule, and why?

> You already accept quantifiers that jump from being attached to sumti, to being in a prenex with scope over one bridi, why not to being in a prenex with scope over the whole expression?

Because in the first case they are not jumping over any scope boundary, and in the second case they are. That's what makes for harder to parse expressions.

> Here's another attempt, where I avoid having "roda prami do" allow you to get rid of "su'ode prami do":

> roda prami su'ode ijo de xebni mi
> == roda su'ode zo'u go da prami de gi de xebni mi
> == "For all X, there exists Y such that X loves Y if and only if Y hates me."

Of course there will be expressions that can be written more succinctly with CLL rules, that's indisputable. The question is: does the frequency of these expressions and the succinctness this rule permits in those cases, compensate the loss of succinctness in other expressions and the general loss of clarity that the rules introduce?

The default option should be to go with the syntax. To go against the syntax a case needs to be made not just by presenting one example or a few examples where this might be convenient, but by showing how the advantages compensate for all the disadvantages.

>>1113

> BTW, would you say the KUs are required by CLL there? Which rule has priority, the CLL selbri-na-rule or the CLL quantifier-over-connective-rule, and why?

I would say the KUs are not required there. Quantifiers by default belong to the outermost enclosing prenex, but NA terms belong to the innermost enclosing prenex. Why? Because logicians traditionally write all their quantifiers in a single prenex with scope over the entire expression, but NOT operators get used with many different scopes, so the rules were chosen to make that tradition easy to continue. Why is the tradition of putting quantifiers in the outermost scope a good thing, which we want to continue? See below.

> The default option should be to go with the syntax. To go against the syntax a case needs to be made not just by presenting one example or a few examples where this might be convenient, but by showing how the advantages compensate for all the disadvantages.

So I have to give evidence that the CLL standard is better, but because your way goes with the syntax, you don't? I would say the default option should be to go with the standard, and if any proposal for change is made, evidence for the new proposal and against the standard would be needed before counter-arguments would be required.

You've obviously passed that threshold already, but I'm fairly certain your proposal hasn't been accepted as the standard yet, so you still have to defend your proposal.

As for examples vs general problems:

> With my rules: "noda xamgu prane ije node xlali prane"
> With CLL rules: "roda naku xamgu prane ije rode naku xlali prane"

Poor example, since the natural form for that in CLL would be "noda xamgu prane ija da xlali prane", which is shorter, and would usually be reduced to "noda xamgu prane gi'a xlali prane". However, "noda xamgu prane i node xlali prane" would also work, if you insist on using two different variables. And that's shorter than yours too.

In your version of Lojban, if you want to use the same variable in different parts of a logical connective, you have to explicitly put it in a prenex with scope over the entire connective. Expressions in CLL standard Lojban can always be rewritten to get rid of all explicit prenexes.

There. Is that a clear enough statement of a general problem with your proposal for you?

>>1121

> I would say the KUs are not required there. Quantifiers by default belong to the outermost enclosing prenex, but NA terms belong to the innermost enclosing prenex.

Sounds like a reasonable choice (given the CLL constraints). But we don't know for sure whether that's the official definition. In fact, there probably isn't an official definition about that. That's a problem with having ad-hoc rules, every new case needs a new ad-hoc rule.

> Why? Because logicians traditionally write all their quantifiers in a single prenex with scope over the entire expression, but NOT operators get used with many different scopes, so the rules were chosen to make that tradition easy to continue.

Do you have a reference for that claim? It doesn't ring true at all.

Consider for example these identities: http://en.wikipedia.org/wiki/First-order_logic#Provable_identities
where the quantifiers are not all in a single prenex. Is that somehow against tradition?

> So I have to give evidence that the CLL standard is better, but because your way goes with the syntax, you don't?

Neither you nor I need give evidence for anything. All I'm saying is that if the rule was to follow the syntax, no further justification would be required for having that rule. If the rule is something else, then it is reasonable to expect to be told why. It is an unexpected rule.

> I would say the default option should be to go with the standard, and if any proposal for change is made, evidence for the new proposal and against the standard would be needed before counter-arguments would be required.

Yes, from to point of view of the reformer that would be right. I'm taking the point of view of the learner: someone new to Lojban but familiar with predicate logic. They are told that the rule for interpreting quantifiers is not what the syntax suggests but something else. Naturally they would want to know why.

> You've obviously passed that threshold already, but I'm fairly certain your proposal hasn't been accepted as the standard yet, so you still have to defend your proposal.

Well... I'm fairly certain a large majority of Lojbanists, including the author of CLL, would interpret "noda xamgu prane ije node xlali prane" in the way I suggest, CLL notwithstanding. That's just my impression, of course, and we would have to make the experiment to see if I'm right or not. Our discussion here is, fortunately or unfortunately depending on how we consider it, mostly academic. It will likely have no impact on the actual use of Lojban. This is based on many similar discussions I have had in the past. These issues don't really come up in actual usage, and if by chance they do, nobody is very sure about what's what.

Very few Lojbanists, if anyone, are actually aware of the implications of that rule. In fact I had forgotten that example was there in CLL, and the first time I read it I assumed it was just an erratum. The rule is more likely to be invoked when there is a shared variable that appears in both bridi, but I'm pretty certain the case where the two bridi didn't have a shared variable was not even considered at the time that was written.

> In your version of Lojban, if you want to use the same variable in different parts of a logical connective, you have to explicitly put it in a prenex with scope over the entire connective.

Yes, exactly.

> Expressions in CLL standard Lojban can always be rewritten to get rid of all explicit prenexes.

Not quite always. For example, you can't get rid of the prenex in:

su'o da zo'u mi djica lo nu mi ponse da

But you were probably referring to expressions without subordinate bridi.

> There. Is that a clear enough statement of a general problem with your proposal for you?

But why is it a problem that some expressions require a prenex? What is so problematic about prenexes? Are they so problematic that we want to sacrifice clarity in exchange for not having to use them?

>>1126

>> In your version of Lojban, if you want to use the same variable in different parts of a logical connective, you have to explicitly put it in a prenex with scope over the entire connective.

> Yes, exactly.

Is it the same story for the "ko'a/fo'a" series? Both "da" and "ko'a" are of the selma'o KOhA. "do prami no ko'a ije do prami ko'e" is a contradictory claim too?

>>1126

> Sounds like a reasonable choice (given the CLL constraints). But we don't know for sure whether that's the official definition. In fact, there probably isn't an official definition about that. That's a problem with having ad-hoc rules, every new case needs a new ad-hoc rule.

It's there, among the examples of De Morgan's Law: http://jbotcan.org/cllc/c16/s12.html (near the bottom)

I should note that your rules are rather poorly documented too, and saying "follow the syntax" or "do what the syntax suggests" isn't getting the message across.

> Do you have a reference for that claim? It doesn't ring true at all.

Granted, I said it's a tradition because I've observed many examples of complicated expressions where all the quantifiers are on the left, not because I read somewhere that one should always attempt to move all the quantifiers to the left. Consider it a rhetorical device which can safely be ignored.

> All I'm saying is that if the rule was to follow the syntax, no further justification would be required for having that rule. If the rule is something else, then it is reasonable to expect to be told why. It is an unexpected rule.

And all I'm saying is that "It is an unexpected rule." is under the scope of the 'then' clause of "if the rule was to follow the syntax". Along with everything between. :) And that's not the rule.

> Well... I'm fairly certain a large majority of Lojbanists, including the author of CLL, would interpret "noda xamgu prane ije node xlali prane" in the way I suggest, CLL notwithstanding.

That's fine. The CLL is the theory of the language, but people make mistakes in practice. In theory, Lojban is a logical language, even if people have trouble getting logic right in practice. But I wouldn't bet that nobody would interpret that sentence correctly.

> su'o da zo'u mi djica lo nu mi ponse da

That's logically identical to "mi djica lo nu mi ponse su'o da" under CLL, but I'll concede that an example can be constructed using abstractions which is difficult (but not impossible) to rewrite without a prenex.

Just to make sure we understand each other:
Ex Ey nu(x, ponse(mi, y)) <=> Ex nu(x, Ey ponse(mi, y))
And in both cases x is an event of possessing, correct?

ro da su'o de zo'u ganai da nu mi ponse de gi mi djica da
ro da zo'u ganai da nu mi ponse su'o de gi mi djica da
ro da poi nu mi ponse su'o de zo'u mi djica da
mi djica ro nu mi ponse su'o de

su'o da ro de zo'u ganai de nu mi ponse da gi mi djica de
ganai nu mi ponse su'o da kei fa ro de gi mi djica de
Ick! Not impossible, but neither short nor elegant. The shortest version which uses a prenex is:
su'o da ro de poi nu mi ponse da zo'u mi djica de
Neither variable can be removed from the prenex while the order of the variables in the main expression would be different. (http://jbotcan.org/cllc/c16/s5.html)

> But why is it a problem that some expressions require a prenex? What is so problematic about prenexes? Are they so problematic that we want to sacrifice clarity in exchange for not having to use them?

In most cases (but not all, as you forced me to concede), getting rid of the prenex creates a shorter expression. Should we sacrifice some clarity for some brevity? Ah, but we aren't sacrificing clarity. The CLL rules are clear and well documented, even if you and I occasionally forget a rule. The only trade-off is in getting brevity in return for a small increase in the complexity of the rules (some things go to the outermost enclosing prenex and others don't, instead of everything going in the innermost enclosing prenex). And, quite frankly, since languages are "learn once, use many times", one eventually gains more from brevity no matter how much more complex the rules have to be; and in this case the increase in complexity is small.

>>1132

> Is it the same story for the "ko'a/fo'a" series? Both "da" and "ko'a" are of the selma'o KOhA.

They are different kind of variables. ko'a/fo'a are constants (assignable variables), they are not bound by quantifiers. ko'a/fo'a are assigned a value (or values) and then every time they are used they refer to that value/tose values. da/de/di don't have any value assigned, they are always bound by a quantifier.

> "do prami no ko'a ije do prami ko'e" is a contradictory claim too?

No. It says you don't love any of the referents of ko'a and you do love the referent(s) of ko'e. They would be contradictory only if ko'a and ko'e had the same referents assigned, which is unlikely (why assign the same referents to two different variables?)

"no ko'a" can be written in terms of bound variables as "no da poi me ko'a". "ko'a" is not a variable bound by "no".

>It's there, among the examples of De Morgan's Law: http://jbotcan.org/cllc/c16/s12.html (near the bottom)

None of those examples have quantifiers though.

> I should note that your rules are rather poorly documented too, and saying "follow the syntax" or "do what the syntax suggests" isn't getting the message across.

Which case do you find dubious?

> In theory, Lojban is a logical language, even if people have trouble getting logic right in practice.

The CLL rules make it especially hard to get the logic right. It's as if they were designed to trap you.

>> su'o da zo'u mi djica lo nu mi ponse da

> That's logically identical to "mi djica lo nu mi ponse su'o da" under CLL,

I don't think so. But if it was, then change the example to:

"mi djica lo nu su'o da zo'u mi ponse da"

One of the two needs the prenex.

(I'm pretty sure quantifies don't escape abstractions, I don't recall anyone ever suggesting that they do, but I can't find an example in CLL one way or the other. Yet another problem with ad-hoc rules.)

> Just to make sure we understand each other:
> Ex Ey nu(x, ponse(mi, y)) <=> Ex nu(x, Ey ponse(mi, y))
> And in both cases x is an event of possessing, correct?

I think you must have miswritten something. Where is "djica"?

"There is something such that I want to have it" is not the same as "I want that there be something that I have".

> Ah, but we aren't sacrificing clarity. The CLL rules are clear and well documented, even if you and I occasionally forget a rule.

Suppose the rule was "a quantifier will bind the second variable after its appearance, not the variable it forms a syntactic term with". That's a rule that is clear but goes against the syntax. The result would be a complete mess to understand. Clear rules don't necessarily result in clear parsing for humans. The more you deviate from the syntax, the harder it is to follow the meaning.

> The only trade-off is in getting brevity in return for a small increase in the complexity of the rules

I could leave with the complexity of the rules. It is the complexity of the resulting expressions that bothers me. Operators crossing scope boundaries are hard to parse.

> And, quite frankly, since languages are "learn once, use many times", one eventually gains more from brevity no matter how much more complex the rules have to be; and in this case the increase in complexity is small.

I think knowing that "noda" is always equivalent to "naku su'oda" (and similar transformation rules) is infinitely more valuable than being able to save a couple of syllables here and here. Being able to interpret any bridi without the need to know where it is embedded is much more valuable than saving on a "zo'u".

>>1136

> Which case do you find dubious?

The case where you claim "follow the syntax" is the only rule needed. I often don't see a way to follow the syntax. For example, strictly following the syntax of "roda broda" would give broda(ro(da)), not ro(da, broda(da)). (Treating "broda" as a function from a constant to a boolean, "da" as a variable, and "ro" as a function from a variable name and an expression to an expression.)

>> Just to make sure we understand each other:
>> Ex Ey nu(x, ponse(mi, y)) <=> Ex nu(x, Ey ponse(mi, y))
>> And in both cases x is an event of possessing, correct?

> "There is something such that I want to have it" is not the same as "I want that there be something that I have".

Your response shows that we do not have the same understanding of abstractions.

"pa nu su'o da zo'u mi ponse da" is an event of ownership, not an event of something existing. Specifically, it's "one event of me owning at least one thing", not "one event of there being something that I have." It's equivalent to a version of "pa nu mi ponse ko'a a ko'e a ko'i" with all the constants in the universe OR-ed together.

su'o da zo'u mi djica pa nu mi ponse da
<=> mi djica pa nu mi ponse ko'a a ko'e a ...
<=> mi djica pa nu su'o da zo'u mi ponse da

Understand? Disagree?

> I often don't see a way to follow the syntax. For example, strictly following the syntax of "roda broda" would give broda(ro(da)), not ro(da, broda(da)).

Right. And similarly "ko'a .e ko'e broda" would give "broda(ko'a & ko'e)" and not "broda(ko'a) & broda(ko'e)", and "naku ko'a broda" would give "broda(naku, ko'a)" instead of "~ broda(ko'a)".

We know that "broda(ro(da))", "broda(ko'a & ko'e)", "broda(naku, ko'a) are meaningless in standard predicate logic notation, and so we need some rule of interpretation for these odd forms that Lojban permits. I propose a rule that treats all operators in the same way ("they operate on the bridi in which they appear as a pseudo-argument, in the order in which they appear as pseudo-arguments). The other rule "connectives operate on the bridi in which they are pseudo-arguments, quantifiers operate on the outermost possible one, for negation, it depends on what other operators are around" is more complicated, and it also has the consequence that a syntactic bridi cannot be interpreted by itself, it may mean different things in different contexts.

> Your response shows that we do not have the same understanding of abstractions.

Yes, I think I eventually figured out what you meant. There are two important differences: One is that my "lo" was not meant to be "su'o lo". I forgot to make that explicit, but you somehow corrected for that. You re-interpreted generic as universal. The other, probably more important difference, is that we indeed have a different understanding of abstractions.

> su'o da zo'u mi djica pa nu mi ponse da
> <=> mi djica pa nu mi ponse ko'a a ko'e a ...
> <=> mi djica pa nu su'o da zo'u mi ponse da

> Understand? Disagree?

I understand and disagree. An abstraction forms an intensional context and you can't move quantifiers (or connectives, it's the same thing) out of them.

(You also have an additional problem in crossing "su'o" with "pa" there, "su'o" and "pa" don't generally commute. That's not the essential point though. I assume you are taking pa=ro here in which case they do commute.)

Consider another example, perhaps more clear. Suppose a bag contains three balls, one blue, one red and one green. You take one out without showing me which color it is. Then:

(1) pa da poi bolci zo'u mi djuno lo du'u do pu jgari da
(2) mi djuno lo du'u pa da poi bolci zo'u do pu jgari da

(1) is false. It is false for the blue ball, it is false for the red ball, it is false for the green ball. For none of them do I know that you took it. (2) is true. I know that there is exactly one ball such that you took it.

The fact that I know is: "lo du'u pa da poi bolci zo'u do pu jgari da", which is a quantified assertion.

>>1141

> The other rule "connectives operate on the bridi in which they are pseudo-arguments, quantifiers operate on the outermost possible one, for negation, it depends on what other operators are around" is more complicated, and it also has the consequence that a syntactic bridi cannot be interpreted by itself, it may mean different things in different contexts.

I think the rules are:
Quantifiers operate on the outermost enclosing bridi, in the order in which they appear.
Logical connectives get transformed to logical connectives between bridi.
All other operators (tags?) operate on the innermost enclosing bridi.
Operators attached to selbri are evaluated first, then operators which appear as terms/tags, in the order they appear.

There's no special rule for NA. The meaning of NA doesn't depend on context. I think the main thing bothering you is that the quantifiers operate on the outermost bridi, and possibly the bit about operators attached to selbri being evaluated first.

> (1) pa da poi bolci zo'u mi djuno lo du'u do pu jgari da
> (2) mi djuno lo du'u pa da poi bolci zo'u do pu jgari da

> (1) is false. It is false for the blue ball, it is false for the red ball, it is false for the green ball. For none of them do I know that you took it. (2) is true. I know that there is exactly one ball such that you took it.

We agree there. But nu is different from du'u. "mi djuno lo du'u ..." effectively says you think the contents of the du'u bridi evaluate to 'true'. But "mi zgana lo nu ..." says you observed a physical event described by the contents of the nu abstraction.

I'm going to simplify the example even further: the bag contains one ball and one cube.

nu pa da zo'u do jgari da
== nu gonai do jgari le bolci gi do jgari le kubli

pa da zo'u nu do jgari da
== gonai nu do jgari le bolci gi do jgari le kubli

I figure that observing an event of the ball xor the cube being grasped implies observing an event of the ball being grasped xor observing an event of the cube being grasped, and vice-versa.

However, I still concede that using abstractions can make it effectively impossible to eliminate some prenexes. The point that the CLL rules allow one to translate logic into more compact Lojban more often still stands.

>>1147

> I think the rules are:
> Quantifiers operate on the outermost enclosing bridi, in the order in which they appear.
> Logical connectives get transformed to logical connectives between bridi.
> All other operators (tags?) operate on the innermost enclosing bridi.

You still need to say something about what happens when the outermost and innermost bridi is the same bridi. The relative order of naku and quantifiers is respected in that case, but not in the case when the outermost bridi is different from the innermost.

You also need to say something about the order of negation and connectives: "naku ko'a .e ko'e broda" and "ko'a .e ko'e naku broda" mean different things. You need to say that the relative order of connectives and naku is always respected.

> Operators attached to selbri are evaluated first, then operators which appear as terms/tags, in the order they appear.

Except when they are not. In "su'oda na blabi gi'e xekri", "su'o" has scope over "na". In "su'oda na blabi" it's the other way around.

> There's no special rule for NA. The meaning of NA doesn't depend on context.

The meaning of na/naku doesn't depend on context, it always negates a bridi. It is its scope ("which bridi?") that depends on context. This happens both with CLL or my rules, and for all operators. The difference is that I give one rule that applies to all operators equally, CLL gives a different rule for each kind of operator.

> I think the main thing bothering you is that the quantifiers operate on the outermost bridi, and possibly the bit about operators attached to selbri being evaluated first.

Yes, because I see no reason to make rules for quantifiers different to rules for conectives different to rules for negators/unary operators. It can be done, of course, and the result is a mess. It is that mess that bothers me.

> But nu is different from du'u.

They are different, yes, but not, I think, in a relevant way with respect to scopes of quantifiers.

> "mi djuno lo du'u ..." effectively says you think the contents of the du'u bridi evaluate

to 'true'.

Right.

> But "mi zgana lo nu ..." says you observed a physical event described by the contents of the nu abstraction.

Right. An event for which the contents of the bridi evaluate to true.

> I'm going to simplify the example even further: the bag contains one ball and one cube.

> nu pa da zo'u do jgari da
> == nu gonai do jgari le bolci gi do jgari le kubli

Right.

> pa da zo'u nu do jgari da

== gonai nu do jgari le bolci gi do jgari le kubli

(You missed a "nu":)

== gonai nu do jgari le bolci gi nu do jgari le kubli

Yes.

> I figure that observing an event of the ball xor the cube being grasped implies observing an event of the ball being grasped xor observing an event of the cube being grasped, and vice-versa.

Not necessarily. You may be able to distinguish that exactly one of the two objects was grabbed without being able to distinguish which one, in which case the first is true and the second false.

> However, I still concede that using abstractions can make it effectively impossible to eliminate some prenexes.

OK.

> The point that the CLL rules allow one to translate logic into more compact Lojban more often still stands.

The claim is still hard to justify. How much "more often"? There are an infinite number of cases in which they can and another infinite number of cases in which they can't. How do you determine "more often"?

And how much "more compact"? Reducing a 100-word text to a 98-word text is not all that impressive, from a practical point of view. Is the idea that with these rules the average length of a Lojban text will be considerably reduced?

>>1150

You've raised some good points. Let me try again:

1: Convert all afterthought connectives into forethought connectives. (E.g., ko'a na blabi gi'e xekri -> ko'a ge na blabi gi xekri)
2: Move all selbri to the front of their bridi or bridi-tail. (E.g., ko'a na klama -> na klama fa ko'a)
3: Proceeding from left to right, when you encounter:
3a, a logical connective: expand it to a forethought logical connective between bridi.
3b, a quantifier: move it to the right-hand end of the outermost enclosing prenex.
3c, any other tag/term which can be moved to a prenex (including those attached to selbri): move it to the right-hand end of the innermost enclosing prenex.

Better?

>> I figure that observing an event of the ball xor the cube being grasped implies observing an event of the ball being grasped xor observing an event of the cube being grasped, and vice-versa.
> Not necessarily. You may be able to distinguish that exactly one of the two objects was grabbed without being able to distinguish which one, in which case the first is true and the second false.

I interpret the second as meaning that one of those two events was observed, and that it's true even if the observation wasn't clear enough to distinguish between the two.

> The claim is still hard to justify. How much "more often"? There are an infinite number of cases in which they can and another infinite number of cases in which they can't. How do you determine "more often"?

I don't know how to quantify the number of cases where a prenex can be eliminated under CLL rules but not under yours. (If you do, I'd be happy to learn something new.) The best I can do is point out that any expression which doesn't require a prenex under your rules, also doesn't require a prenex under CLL rules, but some CLL expressions which don't require a prenex, do under your rules. Therefore, the number of cases where prenexes are required under your rules is strictly greater than the number of cases where prenexes are required under CLL rules.

> And how much "more compact"? Reducing a 100-word text to a 98-word text is not all that impressive, from a practical point of view. Is the idea that with these rules the average length of a Lojban text will be considerably reduced?

Eliminating a prenex saves, at a minimum, a DA word, a "zo'u", the space between them, and the space after "zo'u". So, a minimum of 8 characters saved per prenex eliminated, and more for each additional variable in the prenex, especially when "ku'o" terminators are required between them.

I suspect the most affected texts are statements translated directly from logic into Lojban. Other than that, I don't have enough data to guess how "the average length of a Lojban text" (which one?) would be affected. But I think it's reasonable for Lojban to have rules which make it easy to translate statements from formal logic into compact Lojban, if only to encourage the use of logic in Lojban.

>>1156

> 1: Convert all afterthought connectives into forethought connectives. (E.g., ko'a na blabi gi'e xekri -> ko'a ge na blabi gi xekri)

That's a good first step for my rules too.

> 2: Move all selbri to the front of their bridi or bridi-tail. (E.g., ko'a na klama -> na klama fa ko'a)

I interpret this as: "Move all selbri to the front of their bridi if there is no bridi-tail connection, else move it to the front of the bridi-tail."

> 3: Proceeding from left to right, when you encounter:
> 3a, a logical connective: expand it to a forethought logical connective between bridi.
> 3b, a quantifier: move it to the right-hand end of the outermost enclosing prenex.
> 3c, any other tag/term which can be moved to a prenex (including those attached to selbri): move it to the right-hand end of the innermost enclosing prenex.

OK. You also need to specify how to do the expansion of logical quantifiers in the presence of connectives: {ga ko'a gi ko'e prami ro da} -> {ga ko'a prami ro da gi ko'e prami da}, and not {ga ko'a prami ro da gi ko'e prami ro da/de} which is how I would do it. (The change of variable is not actually required, but just to emphasize that they are independently bound variables.)

BTW, do you treat "roroi" as a quantifier or as a tag, given that it is both?

> I interpret the second as meaning that one of those two events was observed, and that it's true even if the observation wasn't clear enough to distinguish between the two.

How, then, would you distinguish between "for exactly one of the objects I observed that it was grabbed" vs. "I observed that exactly one of the objects was grabbed (but not which one)".

> I don't know how to quantify the number of cases where a prenex can be eliminated under CLL rules but not under yours. (If you do, I'd be happy to learn something new.)

Easy, it's infinite. :)

> The best I can do is point out that any expression which doesn't require a prenex under your rules, also doesn't require a prenex under CLL rules,

We already found one that did: "mi djuno lo du'u do pu jgari pa bolci" does not require a prenex with my rules, but you do need a prenex to express the same thing with yours.

> but some CLL expressions which don't require a prenex, do under your rules. Therefore, the number of cases where prenexes are required under your rules is strictly greater than the number of cases where prenexes are required under CLL rules.

Both numbers are infinite, so it doesn't really make sense to say one is greater than the other. I think you mean that the set of cases that require a prenex under CLL is a proper subset of the set of cases that require a prenex under my rules. That's not true under your interpretation above for quantifiers inside du'u. But even if that were true, that still would not necessarily make the expressions more compact.

Consider this example:

mi pendo su'o prenu poi se bangu no fange

Under my interpretation, this means:
"I am friends with someone who speaks no foreign language."

Under your interpretation, I'm not very sure what it means. Let's see:

mi pendo su'o prenu poi se bangu no fange
==su'o da poi prenu zo'u ge mi pendo da gi da se bangu no fange
==su'o da poi prenu ku'o no de poi fange zo'u ge mi pendo da gi da se bangu de
==su'o da poi prenu ku'o ro de poi fange naku zo'u ge mi pendo da gi da se bangu de
==su'o da poi prenu ku'o ro de poi fange zo'u ga mi na pendo da gi da na se bangu de

But that is true as long as some person is not my friend, even if all my friends speak some foreign language! Not what "mi pendo su'o prenu poi se bangu no fange" means with my rules.

So how do we get that meaning with your rules? The wanted prenex form is:

su'o da poi prenu zo'u ge mi pendo da gi no de poi fange zo'u da se bangu de
==su'o da poi prenu zo'u ge mi pendo da gi ro de poi fange zo'u da na se bangu de
==su'o da poi prenu ku'o ro de poi fange zo'u ge mi pendo da gi da na se bangu de
==su'o da poi prenu ku'o ro de poi fange zo'u ge mi pendo da gi da na se bangu de
==ge mi pendo su'o da poi prenu gi da na se bangu ro de poi fange
== mi pendo su'o prenu poi na se bangu ro fange

So your prenex-less expression is in this case (slightly) longer than mine.

> Eliminating a prenex saves, at a minimum, a DA word, a "zo'u", the space between them, and the space after "zo'u". So, a minimum of 8 characters saved per prenex eliminated, and more for each additional variable in the prenex, especially when "ku'o" terminators are required between them.

Only if the variables in the matrix appear in the right order. If they don't, you may need to add characters to re-order the variables before eliminating the prenex.

> I suspect the most affected texts are statements translated directly from logic into Lojban. Other than that, I don't have enough data to guess how "the average length of a Lojban text" (which one?) would be affected.

A generic Lojban text. A specific text won't have a relevant "average length". My guess is that the answer is "very little" or "insignificantly". (You can pick any random Lojban text around and check. It probably won't be affected at all by any of this.)

>But I think it's reasonable for Lojban to have rules which make it easy to translate statements from formal logic into compact Lojban, if only to encourage the use of logic in Lojban.

Totally! That's why I object to CLL rules, which make that task complicated instead of simple.

>>1157

> I interpret this as: "Move all selbri to the front of their bridi if there is no bridi-tail connection, else move it to the front of the bridi-tail."

Correct. "ko'a ge ko'e broda gi ko'i brode" becomes "ko'a ge broda ko'e gi brode ko'i", not "ge broda ko'e gi brode ko'i vau fa ko'a". Don't treat two connected selbri as a selbri itself, for the purposes of rule #2, and only move selbri to the beginning of their smallest enclosing bridi or bridi-tail.

> You also need to specify how to do the expansion of logical quantifiers in the presence of connectives

3a, a logical connective: expand it to a forethought logical connective between bridi. If an outer quantifier is duplicated by this operation, all but the left-most duplicate should be deleted.

> BTW, do you treat "roroi" as a quantifier or as a tag, given that it is both?

I treat "ro" as a number, which can act as a quantifier in certain contexts, and "roi" as an operator which converts the preceeding number into a tag.

> How, then, would you distinguish between "for exactly one of the objects I observed that it was grabbed" vs. "I observed that exactly one of the objects was grabbed (but not which one)".

"nu jgari pa da" vs "nu jgari pa zo'e", where I think the second means the same as "nu jgari pa da poi me zo'e".

> mi pendo su'o prenu poi se bangu no fange

You aren't applying the rules robotically enough. :-)

Rule #1 doesn't apply. We can skip rule #2, because in this case it has no effect on meaning (but you should normally apply it). By rule #3, we deal with "su'o prenu" first, under rule #3b:
-> su'o da poi prenu zi'e poi se bangu no fange zo'u mi pendo da

And we're done. There's a hole in the rules that leaves the scope of quantifiers inside relative clauses undefined. The same hole applies to quantifiers inside abstractions and inner quantifiers. After paging through the CLL, I haven't found any examples or rules that make it impossible for quantifiers in relative clauses and abstractions to have short scope, the way you want (though I may have missed something). Thoughts?

You haven't explicitly laid out your own algorithm for translating Lojban into prenex form. Would you? Don't worry if it isn't complete, neither is the CLL algorithm. I just want to compare your version with my interpretation of the CLL algorithm.

> Correct. "ko'a ge ko'e broda gi ko'i brode" becomes "ko'a ge broda ko'e gi brode ko'i", not "ge broda ko'e gi brode ko'i vau fa ko'a".

(Minor digression: for me, "ko'a", "ko'e", "ko'i" are constant terms, and therefore the order in which they appear is irrelevant for scope purposes. But I understand you are using them to represent quantified terms there.)

> Don't treat two connected selbri as a selbri itself, for the purposes of rule #2, and only move selbri to the beginning of their smallest enclosing bridi or bridi-tail.

What happens with "su'o da na ge blabi gi xekri", then? Does "na" have scope over "su'o" or not?

> I treat "ro" as a number, which can act as a quantifier in certain contexts, and "roi" as an operator which converts the preceeding number into a tag.

That's good, in "romei" for example "ro" does not act as a quantifier. But in "roroi" it does, it quantifies a bridi. It is (roughly) equivalent to "ca ro da poi temci mokca", and similarly for other number ROI.

> After paging through the CLL, I haven't found any examples or rules that make it impossible for quantifiers in relative clauses and abstractions to have short scope, the way you want (though I may have missed something). Thoughts?

I'm petty sure the idea was for them to have short scope there.

But that means you lose some useful identities. With my rules we have:

ro da poi broda ... cu brode ... == ro da zo'u ganai da broda ... gi da brode ...

su'o da poi broda ... cu brode ... == su'o da zo'u ge da broda ... gi da brode ...

Those identities will break down for you in the presence of quantified terms.

> You haven't explicitly laid out your own algorithm for translating Lojban into prenex form. Would you? Don't worry if it isn't complete, neither is the CLL algorithm. I just want to compare your version with my interpretation of the CLL algorithm.

1: Convert all afterthought connectives into forethought connectives.
2: Apply each operator, from left to right, to the immediate bridi in which it appears.

That's it. The possible operators are quantifiers, connectives, and negation/other unary operators.

(With my rules, all the usual logical manpulations, like DeMorgan's laws, expansions of "poi", etc, will automatically remain valid, even when done in "compact" notation.)

>>1164

I just noticed this:
"By default, a variable always behaves as if it is bound in the prenex which (notionally) is attached to the smallest enclosing bridi, and its scope does not extend beyond that bridi."
http://jbotcan.org/cllc/c16/s8.html

>>1166

> What happens with "su'o da na ge blabi gi xekri", then? Does "na" have scope over "su'o" or not?

Good point. Do treat connected selbri/bridi-tails as selbri, moving them to the beginning of their enclosing bridi or bridi-tail.

> Those identities will break down for you in the presence of quantified terms.

It all works if you're careful.

mi pendo su'o prenu poi se bangu no fange
-> su'o da poi prenu zo'u ge no de poi fange zo'u da se bangu de gi mi pendo da
-> su'o da poi prenu ku'o ro da poi fange zo'u ge da na se bangu de gi mi pendo da
"There exists person X such that for every foreign thing Y, X doesn't speak language Y and I am a friend of X"

Or, if you really want to go crazy:
-> su'o da ro de zo'u da na se bangu de ije mi pendo da ijanai de fange ije da prenu
"There exists X such that for all Y, X doesn't speak language Y and I am a friend of X is implied by Y is a foreign thing and X is a person."

> 1: Convert all afterthought connectives into forethought connectives.
> 2: Apply each operator, from left to right, to the immediate bridi in which it appears.

Omitting that much, I could condense rule #3 to:

3: Apply each operator from left to right, to the outermost bridi if it's a quantifier, or to the innermost bridi for all other operators.

The details of how to apply each type of operator should be explicit. Also, note that I view my rule #2 as the same sort of operation as rule #1. I could have merged those two rules into "1: convert all infix operators to prefix form."

>>1168

Ah. So it was there and I did miss it.

>>1169

> Good point. Do treat connected selbri/bridi-tails as selbri, moving them to the beginning of their enclosing bridi or bridi-tail.

But now what do we do with "su'o da ge prami ro de gi xebni ro di". Does "su'o da" NOT have scope over the "ro"s?

> > Those identities will break down for you in the presence of quantified terms.
> It all works if you're careful.

With your rules, "noda" cannot always be changed to "naku su'oda". It can only be done inside a prenex, or in a non-connected bridi, but it fails inside a connected bridi. That's because the negation inside "noda" will have scope over the connective, whereas the "naku" negation will not.

But your rule about quantifiers now seems to be in contradiction with: "By default, a variable always behaves as if it is bound in the prenex which (notionally) is attached to the smallest enclosing bridi, and its scope does not extend beyond that bridi."

> The details of how to apply each type of operator should be explicit.

These are the details:

... naku ... -> naku zo'u ......

... na ... -> naku zo'u ......

... roda (poi ...) ... -> roda (poi ...) zo'u ......

... (ge ... gi ...) ... -> ge ......... gi .........

> Also, note that I view my rule #2 as the same sort of operation as rule #1. I could have merged those two rules into "1: convert all infix operators to prefix form."

Your rule #2 is still unclear for cases where the bridi tail contains tail terms.

>>1174

Correction:

... roda (poi ...) ... -> roda (poi ...) zo'u ... da ...

>>1174

> But now what do we do with "su'o da ge prami ro de gi xebni ro di". Does "su'o da" NOT have scope over the "ro"s?

Technically, that should become:
2: -> ge prami ro de gi xebni ro di vau fa su'o da
3a: -> ge prami fa su'o da ro de gi xebni fa da ro di
But the rules for expanding logical connectives get unacceptably convoluted that way. So, a new version of rule #2 that avoids the problems of moving connected selbri:
2: Convert all operators attached to selbri or bridi tails into terms at the front of their innermost enclosing bridi.

> But your rule about quantifiers now seems to be in contradiction with: "By default, a variable always behaves as if it is bound in the prenex which (notionally) is attached to the smallest enclosing bridi, and its scope does not extend beyond that bridi."

http://jbotcan.org/cllc/c16/s14.html appears to contradict that. To resolve the contradiction, I take "smallest enclosing bridi" in the above to refer to abstractions, since that's the example being discussed.

Note that examples 8.1 to 8.3 and the text contradict your "ro da poi broda ... cu brode ... == ro da zo'u ganai da broda ... gi da brode ..." identity. http://jbotcan.org/cllc/c16/s8.html

> These are the details:

They are not detailed enough. :-) For example, "... na ... -> naku zo'u ......" makes it look as though you intend that "ro da zo'u da na broda" be replaced with "naku zo'u ro da zo'u da broda". Similar problems concerning what, exactly, belongs in the "..." portions, exist in all your rules. I have a good idea of what you intend, after all these discussions, so I can fill in the blanks, but few others will be able to. I was hoping for a fully explicit version so the next newbie like me who comes along can learn your rules without having to repeat all these discussions.

>>1182

>>"By default, a variable always behaves as if it is bound in the prenex which (notionally) is attached to the smallest enclosing bridi, and its scope does not extend beyond that bridi."
>http://jbotcan.org/cllc/c16/s14.html appears to contradict that.

That section is hard to understand, as it doesn't correspond with anything in standard predicate logic. For example, is "ci da poi mlatu na blabi .ije re da cu barda" at all meaningful, according to those rules? "It is not the case that there are exactly three white cats. And two of them are big." Two of which??? Two cats? Two of the other-than-three cats that are white? Two of those that are not white? This is the problem with introducing these non-standard uses of quantifiers. One example may sort of make sense, but the general rule is not given, and we can't rely on standard predicate logic because they don't correspond to anything there.

http://jbotcan.org/cllc/c16/s10.html is I think a slighlty better argument for your rule. (Example 10.4, with two "ro"s, is still odd, but we could take that as a typo.)

> To resolve the contradiction, I take "smallest enclosing bridi" in the above to refer to abstractions, since that's the example being discussed.

Or you could take it as "smallest bridi that encloses all appearances of the bound variable", for example. In any case, it requires some arbitrary interpretation.

> That's one option. The other option is to

Note that examples 8.1 to 8.3 and the text contradict your "ro da poi broda ... cu brode ... == ro da zo'u ganai da broda ... gi da brode ..." identity. http://jbotcan.org/cllc/c16/s8.html

True, but then CLL again contradicts itself by saying that "As explained in Section 9, when a prenex negation boundary expressed by ``naku'' moves past a quantifier, the quantifier has to be inverted." http://jbotcan.org/cllc/c16/s11.html
For that rule to work (for example in going from 11.6 to 11.7) you need that identity to hold. Without that identity, you can't move "naku" past a restricted quantifier by just inverting it.

And that identity is the usual interpretation of restricted quantifiers in predicate logic. The CLL rule about "ro da poi" entailing "su'o da poi" is just another idiosyncratic and unnecessary complication.

>> These are the details:
>They are not detailed enough. :-) For example, "... na ... -> naku zo'u ......" makes it look as though you intend that "ro da zo'u da na broda" be replaced with "naku zo'u ro da zo'u da broda".

But in my rule I said "the immediate bridi". The immediate bridi there is "da na broda", not "ro da zo'u da na broda".

> Similar problems concerning what, exactly, belongs in the "..." portions, exist in all your rules.

They always represent the immediate bridi containing the operator. There is always only one immediate bridi. And since operators will be applied from left to right, there is no other operator in front of the one we are currently dealing with.

> I have a good idea of what you intend, after all these discussions, so I can fill in the blanks, but few others will be able to.

Explanation can go on indefinitely, depending on how much one assumes to be understood. Someone who doesn't know what a bridi is will need more explanation than someone who does. Someone who doesn't understand that "naku" negates a bridi will need more explanation than someone who does, and so on. I'm sure there are more clear ways of stating the rule. My question for you is this: Do you have the slightest doubt about how my rule is applied in any case? I'm not asking whether I have stated the rule in the most clear way possible, but whether in your mind there is still any doubt about what my rule is. In my mind, I still don't know what the CLL rules, or your interpretation of them, are exactly supposed to be. CLL contradicts itself. It presents some examples but doesn't give complete rules. And any rules we try to derive from the examples get more and more convoluted as we examine new cases. With my rule I don't need to keep introducing exceptions to deal with special cases, I always know what the rule says, and not because it's my rule but because it is by construction complete and self-consistent.

> I was hoping for a fully explicit version so the next newbie like me who comes along can learn your rules without having to repeat all these discussions.

The negation of a bridi is always "naku zo'u <bridi>". Which bridi is negated by "naku"? The most immediate one that contains it as an argument (after you have dealt with preceding operators). The same applies to quantifiers. The same applies to connectives, but here is a more explicit version:

<head-shared-part> ge <first-connectand-part> gi <second-connectand-part> <tail-shared-part>
--> ge <head-shared-part> <first-connectand-part> <tail-shared-part> gi <head-shared-part> <second-connectand-part> <tail-shared-part>

>>1183

> "It is not the case that there are exactly three white cats. And two of them are big."

No no, it's "Three cats X exist such that X is not white and two of them are big."

> For that rule to work (for example in going from 11.6 to 11.7) you need that identity to hold.

The identity still holds. A "no empty sets" rule doesn't interfere. (Which doesn't mean I think the rule is a good idea. Just that your stated reason for ignoring it is invalid.)

> In my mind, I still don't know what the CLL rules, or your interpretation of them, are exactly supposed to be.

I'm working under the assumption that there is an underlying set of rules, which the CLL attempts to explain to a non-technical audience. I'm still trying to figure out what, exactly, the underlying rules are. However, I think that there can be a set of rules which is "by construction complete and self-consistent", and matches the CLL better than your rules do. If my attempts have holes in them, it's only because I made a mistake. (Thanks for helping me debug!)

The latest version of my attempt to explain the CLL is:
1: Convert all afterthought connectives into forethought connectives.
2: Convert all operators attached to selbri or bridi tails into terms at the front of their innermost enclosing bridi.
3: Proceeding from left to right, when you encounter:
3a, a logical connective: expand it to a forethought logical connective between bridi.
3b, any other tag/term which can be moved to a prenex: move it to the end of the innermost enclosing prenex.

I recently had the idea to refactor this to:
1: Proceeding from left to right, move all operators attached to selbri to the end of their innermost enclosing prenex
2: Proceeding from left to right, move all remaining operators to the end of their innermost enclosing prenex
3: Proceeding from left to right, expand logical connectives into connectives between bridi

An order of operations system seems to explain the CLL fairly well. There are a few places where it's inconsistent with something the CLL says, but considerably fewer than with your rules.

>>1187

>> For that rule to work (for example in going from 11.6 to 11.7) you need that identity to hold.
>The identity still holds. A "no empty sets" rule doesn't interfere.

roda poi klama cu cadzu
== naku naku roda poi klama zo'u da cadzu
== naku su'oda poi klama ku'o naku zo'u da cadzu
== naku su'oda poi klama zo'u naku da cadzu
== naku su'oda poi klama zo'u ge da klama gi naku da cadzu
== roda poi klama ku'o naku zo'u ge da klama gi naku da cadzu
== roda poi klama zo'u ganai da klama ginai naku da cadzu
== roda poi klama zo'u ganai da klama gi da cadzu

Which step fails?

> The latest version of my attempt to explain the CLL is:
> 1: Convert all afterthought connectives into forethought connectives.
> 2: Convert all operators attached to selbri or bridi tails into terms at the front of their innermost enclosing bridi.

What is the "innernost enclosing bridi" for "na" in "mi ge ti na lebna gi klama". Do you take "ti lebna" as a bridi there, even though "ti" is in the x2 of lebna?

> 3: Proceeding from left to right, when you encounter:
> 3a, a logical connective: expand it to a forethought logical connective between bridi.
> 3b, any other tag/term which can be moved to a prenex: move it to the end of the innermost enclosing prenex.

Wait: Quantifiers don't scope over connectives now?

> I recently had the idea to refactor this to:
> 1: Proceeding from left to right, move all operators attached to selbri to the end of their innermost enclosing prenex
> 2: Proceeding from left to right, move all remaining operators to the end of their innermost enclosing prenex
> 3: Proceeding from left to right, expand logical connectives into connectives between bridi

Logical connectives are (binary) operators. So presumably rule 2 is limited to unary operators, like negation and quantifiers.

How much do the connectives expand? Does "su'o da zo'u ge ko'a gi ko'e broda da" expand to "su'o da zo'u ge ko'a broda da gi ko'e broda da", or to "ge su'o da zo'u ko'a broda da gi su'o da zo'u ko'e broda da"? Same for "naku zo'u".

>>1191

> Which step fails?

None of them? Didn't I just say that the identity holds? uanai

> The latest version of my attempt to explain the CLL is:

... and I pasted the wrong version in. Sorry about that. The correct version (I hope):

1: Convert all afterthought connectives into forethought connectives.
2: Move all operators attached to selbri or bridi tails to the front of their innermost enclosing prenex.
3: Proceeding from left to right, when you encounter:
3a, a logical connective: expand it to a forethought logical connective between bridi. If an outer quantifier is duplicated by this operation, delete all but the left-most duplicate.
3b, a quantifier: move it to the end of the outermost enclosing prenex.
3c, any other tag/term which can be moved to a prenex: move it to the end of the innermost enclosing prenex.

> Logical connectives are (binary) operators. So presumably rule 2 is limited to unary operators, like negation and quantifiers.

Yes. For some reason I couldn't think of a word for "non logical-connective operator".

> How much do the connectives expand? Does "su'o da zo'u ge ko'a gi ko'e broda da" expand to "su'o da zo'u ge ko'a broda da gi ko'e broda da", or to "ge su'o da zo'u ko'a broda da gi su'o da zo'u ko'e broda da"? Same for "naku zo'u".

The prenex isn't considered part of the bridi, for the purposes of these rules. The same applies to your rules, I presume.

Ambiguities like this are one reason it's a good idea to include some examples of the intended effect along with any rule written in anything other than debugged computer code. (Even then, some examples in the comments might be helpful. Or, better yet, a suite of test cases.)

>>1196

> Didn't I just say that the identity holds? uanai

You wrote, in >>1183:

>Note that examples 8.1 to 8.3 and the text contradict your "ro da poi broda ... cu brode ... == ro da zo'u ganai da broda ... gi da brode ..." identity. http://jbotcan.org/cllc/c16/s8.html

But now you are saying, in agreement with me and against CLL, that that identity does in fact hold. If so, great, we are in agreement on that point.

> The correct version (I hope):

OK. That still leaves the problem of quantifiers inside abstractions and relative clauses. Do they too move out of the clause to the outermost enclosing prenex?

> The prenex isn't considered part of the bridi, for the purposes of these rules. The same applies to your rules, I presume.

My rules only mention the "immediate bridi" where the operator appears as a pseudo-argument. The prenex is never part of the immediate bridi because the non-prenex part (the matrix) will always necessarily be more immediate.

In general, a prenex is part of a bridi. In fact the prenex is the part that contains the operator which takes one bridi and outputs another bridi.

>>1201

We, and the CLL, agree that "roda poi klama cu cadzu" implies "roda poi klama zo'u ganai da klama gi da cadzu". The CLL claims that neither of those imply "roda zo'u ganai da klama gi da cadzu", because "roda poi klama" is a way of saying "For all X from the (non-empty) set of things which are goers". You see why I was confused?

> OK. That still leaves the problem of quantifiers inside abstractions and relative clauses. Do they too move out of the clause to the outermost enclosing prenex?

Not covered by these rules, since relative clauses and most abstractions get moved to the prenex, and thus don't get processed as we move from left-to-right through the expression. Your rules, if they work in the same way, leave the same things unstated.

> My rules only mention the "immediate bridi" where the operator appears as a pseudo-argument. The prenex is never part of the immediate bridi because the non-prenex part (the matrix) will always necessarily be more immediate.

Clever. I think I'll adopt that language for the rule about expanding logical connectives.

>>1209

> We, and the CLL, agree that "roda poi klama cu cadzu" implies "roda poi klama zo'u ganai da klama gi da cadzu". The CLL claims that neither of those imply "roda zo'u ganai da klama gi da cadzu", because "roda poi klama" is a way of saying "For all X from the (non-empty) set of things which are goers". You see why I was confused?

But the CLL also says that you can move a negation past a restricted quantifier by inverting the quantifier. Do you agree that that requires "roda poi klama cu cadzu" and "roda zo'u ganai da klama gi da cadzu" to be equivalent?

>> That still leaves the problem of quantifiers inside abstractions and relative clauses. Do they too move out of the clause to the outermost enclosing prenex?
> Not covered by these rules, since relative clauses and most abstractions get moved to the prenex, and thus don't get processed as we move from left-to-right through the expression. Your rules, if they work in the same way, leave the same things unstated.

I don't make any exceptions for relative clauses and abstractions, they follow the same single rule: all operators apply from left to right, always to the immediate bridi in which they appear as an argument.

We haven't discussed things like "lo prami be roda", but the same rule will apply there: "roda" appears as an argument of "prami", so we expand that as a bridi ("zo'e noi ke'a prami roda" in my case, and probably "su'o de poi ke'a prami roda" in your case), and then we apply the same rule. Also things like "lu'i roda" will need to be expanded since "roda" there is an argument in a hidden bridi: "lu'i roda" -> "lo selcmi be roda" -> "zo'e noi ke'a selcmi roda". Presumably these don't need special mention in your rules either, because they would first be expanded to full bridi-form and then treated accordingly.

But you do need to say what to do with relative clauses. If "ro da poi <bridi> cu broda" is to imply "ro da poi <bridi> zo'u ganai <bridi> gi broda", what will happen with any quantifiers inside <bridi>? Will they have scope over "ganai ... gi ..."?