WikiDiscuss

WikiDiscuss


tags as connectives

posts: 2388
Use this thread to discuss the tags as connectives page.
posts: 2388

The issue of the order of sentences around connectives — relative to the order around the "corresponding selbri" (that there always is one is suspect in many cases) is as old as Loglan. The order for the causal/inferential connectives changed several times in printed Loglan text, mainly as the translation of the connectives moved between "because" (which has a certain naturalness from "cause") and "hence/therefore/thus" which seemed to fit the order better.

The argument always brought came back to the order of the underlying predicate, though I don't remember what it was most recently in Loglan. Of course, the predicates do not keep the corresponding orders completely, so that aargument may collapse a bit — and some are personal rather than formal, I seem to recall.
pc

posts: 1912


>
> Re:tags as connectives
>
> The issue of the order of sentences around connectives — relative to the
> order around the "corresponding selbri" (that there always is one is suspect
> in many cases) is as old as Loglan.

How can there not be one? We have two things that get connected,
so we can always define a new selbri 'brodrtaga' such that it represents
that relationship between the two things that the tag connects, i.e.
such that {lo nu bu'a cu brodrtaga lo nu bu'e} means {bu'a i <tag> bo bu'e}.

> The order for the causal/inferential
> connectives changed several times in printed Loglan text, mainly as the
> translation of the connectives moved between "because" (which has a certain
> naturalness from "cause") and "hence/therefore/thus" which seemed to fit the
> order better.
>
> The argument always brought came back to the order of the underlying
> predicate, though I don't remember what it was most recently in Loglan. Of
> course, the predicates do not keep the corresponding orders completely, so
> that aargument may collapse a bit — and some are personal rather than
> formal, I seem to recall.

The problem is not so much the underlying predicate, that is just
a convenient way to explain it. The problem is that:

{X i ba bo Y} corresponds to {ba gi Y gi X}

whereas

{X i ri'a bo Y} corresponds to {ri'a gi X gi Y}

So even if we don't mention any underlying predicates, there
is a discrepancy.

mu'o mi'e xorxes


__
Do you Yahoo!?
Free Pop-Up Blocker - Get it now
http://companion.yahoo.com/


posts: 2388

x

>pc:

>>"corresponding selbri" (that there always is one is

>>suspect in many cases)

>

>How can there not be one? We have two things that get connected,

>so we can always define a new selbri 'brodrtaga' such that it

>represents that relationship between the two things that the tag connects, >i.e. such that {lo nu bu'a cu brodrtaga lo nu bu'e} means {bu'a i <tag> bo

>bu'e}.



pc:

Sure, we can always fadge up a predicate, but that makes tags uninteresting – and a little suspect. What I meant was “natural predicate,” “existing predicate.” Maybe even “gismu.”



>> The order for the causal/inferential

>> connectives changed several times in printed Loglan text, mainly as

>>the translation of the connectives moved between "because" (which has a

>>certain naturalness from "cause") and "hence/therefore/thus" which seemed >>to fit the order better.

>>

>> The argument always came back to the order of the underlying

>> predicate, though I don't remember what it was most recently in

>>Loglan. Of course, the predicates do not keep the corresponding orders

>>completely, so that argument may collapse a bit — and some are personal >>rather than formal, I seem to recall.



>The problem is not so much the underlying predicate, that is just

>a convenient way to explain it. The problem is that:

>

> {X i ba bo Y} corresponds to {ba gi Y gi X}

>

>whereas

>

> {X i ri'a bo Y} corresponds to {ri'a gi X gi Y}

>

>So even if we don't mention any underlying predicates, there

>is a discrepancy.

pc:
I suspect that the discrepancy is connected with the predicates involved: with the typical PU, the following sumti is actually the second argument of the corresponding predicate, with BAI it is the first. The afterthought-connective use of the tags, since it clearly comes between the two, adopts the order of the predicate, no longer following the tag association of the other uses. Suppose {anai} were associated with “follows from”: then {X ijanai Y} would correspond to {ganai Y gi X} (though the principles involved are somewhat different).



posts: 1912


> >>"corresponding selbri" (that there always is one is
> >>suspect in many cases)
>
> Sure, we can always fadge up a predicate, but that makes tags uninteresting –
> and a little suspect. What I meant was “natural predicate,” “existing
> predicate.” Maybe even “gismu.”

In that case, it is almost certain that not every tag has an
associated gismu.

> > {X i ba bo Y} corresponds to {ba gi Y gi X}
> >whereas
> > {X i ri'a bo Y} corresponds to {ri'a gi X gi Y}
>
> pc:
> I suspect that the discrepancy is connected with the predicates involved:
> with the typical PU, the following sumti is actually the second argument of
> the corresponding predicate, with BAI it is the first.

That depends on how you define the "corresponding predicate". If you
define it coherently (ri'a = fi'o rinka, ba = fi'o selbalvi}, then
the predicate that "corresponds" to {ba} is {selbalvi}. The
correspondence of {ba} with {balvi} is etymological, but it is not
the fi'o correspondence, {balvi} is not turned to {ba} with the
cmavo that converts selbri to tags, {selbalvi} is.

> The
> afterthought-connective use of the tags, since it clearly comes between the
> two, adopts the order of the predicate, no longer following the tag
> association of the other uses. Suppose {anai} were associated with “follows
> from”: then {X ijanai Y} would correspond to {ganai Y gi X} (though the
> principles involved are somewhat different).

For some strange reason {X iju Y} corresponds to {gu X gi Y}, instead
of {gu Y gi X}.

mu'o mi'e xorxes


__
Do you Yahoo!?
Free Pop-Up Blocker - Get it now
http://companion.yahoo.com/


posts: 2388

x:
In that case, it is almost certain that not every tag has an
associated gismu.

pc:
Very true, but each should have a "natural" predicate, not just one made up to be a predicate for a tag.
> > {X i ba bo Y} corresponds to {ba gi Y gi X}
> >whereas
> > {X i ri'a bo Y} corresponds to {ri'a gi X gi Y}
>
> pc:
> I suspect that the discrepancy is connected with the predicates
involved:
> with the typical PU, the following sumti is actually the second
argument of
> the corresponding predicate, with BAI it is the first.
x:
That depends on how you define the "corresponding predicate". If you
define it coherently (ri'a = fi'o rinka, ba = fi'o selbalvi}, then
the predicate that "corresponds" to {ba} is {selbalvi}. The
correspondence of {ba} with {balvi} is etymological, but it is not
the fi'o correspondence, {balvi} is not turned to {ba} with the
cmavo that converts selbri to tags, {selbalvi} is.
pc:
Your claim is getting stranger and stranger. If "corresponding" is to have a meaning it needs to be fixed. Is it the case that we can just take a predicate and declare that some tag corresponds to some permutation of it, without any evidence in the tag? This again makes the whole notion suspect and uninteresting. Notice, for example, that {balvi} explains an apparent anomaly, {se balvi} does not. (Is {seba} even legit?)
On the other hand, the connection between predicate and tag is pretty loose: {bau} is "in a language" not "is a language." Here the outside sentence is not the second argument but the third (and this is true of {sebau} as well). But then {bau} is a bit hard to imagine as a connective (I expect an obvious example here, of course.) So maybe insisting on more than a mnemonic connection is absurd, but you seem to want to do it at all cost.

> The
> afterthought-connective use of the tags, since it clearly comes
between the
> two, adopts the order of the predicate, no longer following the tag
> association of the other uses. Suppose {anai} were associated with
“follows
> from”: then {X ijanai Y} would correspond to {ganai Y gi X} (though
the
> principles involved are somewhat different).

x:
For some strange reason {X iju Y} corresponds to {gu X gi Y}, instead
of {gu Y gi X}.

pc:
Sure? This is too clearly wrong for even CLL to have done, but there it is on p329. How in the Hell did it happen? The text even notes that it is odd. Apparently, the fact that {u} is not symmetric escaped notice — or it was thought that only asymmetries involving negation had to be treated peculiarly. However, it would give some analogy for the peculiarities of PU relative to their corresponding predicates.




posts: 2388

x:



<<the predicate that "corresponds" to {ba} is {selbalvi}. The

correspondence of {ba} with {balvi} is etymological, but it is not

the fi'o correspondence, {balvi} is not turned to {ba} with the

cmavo that converts selbri to tags, {selbalvi} is. >>



pc:

But {se balvi} is just {purci}, so perhaps we should exchange {ba} and {pu}, a suggestion that has been made – and rejected – before; it’s very Irish so not an irrational one. But it fails the “preserve past text” stricture. I think it is better just to recognize that tense PU behaves differently from BAI, which is hardly surprising.

<<x:

For some strange reason {X iju Y} corresponds to {gu X gi Y}, instead

of {gu Y gi X}.



pc:

Sure? This is too clearly wrong for even CLL to have done, but there it

is on p329. How in the Hell did it happen? The text even notes that

it is odd. Apparently, the fact that {u} is not symmetric escaped

notice — or it was thought that only asymmetries involving negation had to

be treated peculiarly. However, it would give some analogy for the

peculiarities of PU relative to their corresponding predicates.>>


On the other hand, how else would we do {gu}? Even if we swithced {gu} and {se gu}(?!) there would still be the same separation. The rule for A to GA seems to be to shift the pure form, leaving the negations attached as before, but preferring {nai} to {na} where possible (and front to center when relevant) but keeping the order of components the seame. The rules for tags seem to be less simple, but mainly because we are talking about a category “tag” which encompasses a number of items otherwise rather diverse.



posts: 1912


> pc:
> Your claim is getting stranger and stranger. If "corresponding" is to have a
> meaning it needs to be fixed.

By "corresponding predicate", I always meant a predicate broda such
that {fi'o broda} is equivalent to the tag in question.

> Is it the case that we can just take a
> predicate and declare that some tag corresponds to some permutation of it,
> without any evidence in the tag?

I'm sorry, I don't understand the question.

It is the case that for any given predicate, we can create
a tag corresponding to each of its arguments. So for example
from {klama} we can create 5 tags: {fi'o klama}, {fi'o se klama},
{fi'o te klama}, {fi'o ve klama} and {fi'o xe klama}.

For a number of predicates, those tags have short forms, so for
example the tag {ve ka'a} is the short form of the tag {fi'o ve klama}.

The tag {fi'o se nenri} has a short form equivalent: {ne'i}.
The tag {fi'o nenri} does not have a short form. In other words,
in {broda fi'o nenri ko'a}, you can't replace {fi'o nenri} with
a short tag and get the same meaning.

> This again makes the whole notion suspect
> and uninteresting. Notice, for example, that {balvi} explains an apparent
> anomaly, {se balvi} does not. (Is {seba} even legit?)

{seba} in ungrammatical. If {fi'o balvi} has an equivalent short form,
it is {pu}. In other words, the closest thing we have to
{mi klama fi'o balvi lo nu mi citka} is {mi klama pu lo nu mi citka}.
"I go, with my eating in the future."

> On the other hand, the connection between predicate and tag is pretty loose:
> {bau} is "in a language" not "is a language."

{bau} as sumti tcita tags a sumti which would be the x1 of {bangu},
i.e. the language. {bau} is {fi'o bangu}.

>Here the outside sentence is
> not the second argument but the third (and this is true of {sebau} as well).

Right. {fi'o} only selects one of the arguments from its selbri.
The other one has to be glorked.

> But then {bau} is a bit hard to imagine as a connective (I expect an obvious
> example here, of course.) So maybe insisting on more than a mnemonic
> connection is absurd, but you seem to want to do it at all cost.

I don't want to do anything with the fi'o-selbri. I just used it
to point out something which can be pointed out just as well without
recourse to the fi'o-selbri, namely that some tags switch the order
of their arguments in forethought and others don't.

mu'o mi'e xorxes


__
Do you Yahoo!?
Free Pop-Up Blocker - Get it now
http://companion.yahoo.com/


posts: 1912


> pc:
>
> But {se balvi} is just {purci}, so perhaps we should exchange {ba} and {pu},
> a suggestion that has been made – and rejected – before; it’s very Irish so
> not an irrational one. But it fails the “preserve past text” stricture. I
> think it is better just to recognize that tense PU behaves differently from
> BAI, which is hardly surprising.

It is surprizing for someone who expects Lojban to be regular.
You can't learn a single rule for the relationship between
{<tag> gi ... gi ...} and {... i <tag> bo ...}. You need to
learn different rules for different tags.

> <<x:
> For some strange reason {X iju Y} corresponds to {gu X gi Y}, instead
> of {gu Y gi X}.

pc:
> On the other hand, how else would we do {gu}? Even if we swithced {gu} and
> {se gu}(?!) there would still be the same separation.

How so?

{gu X gi Y} would correspond to {Y iju X}
{segu X gi Y} would correspond to {Y iseju X}.

Currently it is the other way around.

mu'o mi'e xorxes


__
Do you Yahoo!?
Free Pop-Up Blocker - Get it now
http://companion.yahoo.com/


posts: 2388

>
> > Is it the case that we can just take a
> > predicate and declare that some tag corresponds to
some permutation of it,
> > without any evidence in the tag?
>
> Im sorry, I dont understand the question.

Can we arbitrarily pick a predicate and say it
corresponds to a tag, even if there are no visible
similarities? This makes the notion that there is a
correspondence totally trivial and uninteresting. I
can always — as you note — make something up. So why
continue with this project? The claim you want to make
is pointless and the claim you appear to make is
false.
>
> It is the case that for any given predicate, we can
create
> a tag corresponding to each of its arguments. So for
example
> from {klama} we can create 5 tags: {fio klama}, {fio
se klama},
> {fio te klama}, {fio ve klama} and {fio xe klama}
..
This is, of course, built into the definition of
{fio}

> For a number of predicates, those tags have short
forms, so for
> example the tag {ve kaa} is the short form of the
tag {fio ve klama}.
>
> The tag {fio se nenri} has a short form equivalent:
{nei}.
> The tag {fio nenri} does not have a short form. In
other words,
> in {broda fio nenri koa}, you cant replace {fio
nenri} with
> a short tag and get the same meaning.

But this is just clearly wrong if the relation is to
the predicates in {fio} phrases. The connection of
{nei} is obviously with {nenri}, not {se nenri} --
unless the connection is arbitrary. In the latter
case, the whole {fio} discussion is irrelevant, the
connection is merely mnemonic and not reducible to
some other relation.
>
>
>
> {bau} as sumti tcita tags a sumti which would be the
x1 of {bangu},
> i.e. the language. {bau} is {fio bangu}.

Not in any obvious sense. {fio bangu} apparently
hooks on a reference to a language, without any
specification of how that is to be connected to the
rest of the sentence, {bau} is quite precise (if not
explicit) about what the connection is: the language
mentioned is the language in which some vocal activity
cited in the sentence is carried out (i.e. that
activity is te bangu).


> I dont want to do anything with the fio-selbri. I
just used it
> to point out something which can be pointed out just
as well without
> recourse to the fio-selbri, namely that some tags
switch the order
> of their arguments in forethought and others dont.

Yes, because some tags point to different kinds of
structures than others. I suspect that switch the
order of their arguments assumes that the same
structures are involved, whereas the fact that there
are different classes involves suggest that the
structures are also diffeeent in some interesting way.


I
> > think it is better just to recognize that tense PU
behaves differently from
> > BAI, which is hardly surprising.
>
> It is surprizing for someone who expects Lojban to
be regular.
> You cant learn a single rule for the relationship
between
> { gi ... gi ...} and {... i bo ...}. You need to >
learn different rules for different tags.

But the diferent tags belong to different categories,
so one expects that they behave differently. Look at
the distribution differences for the various regular
connectives — which have different sources. To be
sure, I would like to get rid of these differences --
maybe even with tag connectives — but I am not
surprised that the differences exist.
>
> > > For some strange reason {X iju Y} corresponds to
{gu X gi Y}, instead
> > of {gu Y gi X}.
>
> pc:
> > On the other hand, how else would we do {gu}? Even
if we swithced {gu} and
> > {se gu}(?!) there would still be the same
separation.
>
> How so?
>
> {gu X gi Y} would correspond to {Y iju X}
> {segu X gi Y} would correspond to {Y iseju X}.
>
> Currently it is the other way around.
Currently (since the other way around is ambiguous)
{gu X gi Y} corresponds to {X iju Y} and
{segu X gi Y} corresponds to {X segu Y}, which fits in
exactly with the rules for the other A connectives,
under the usual assumption that postposed {na} and
preposed {nai} belong with the sentence not, strictly,
the connective. That assumption — which is current
Lojban — is not a necessary one; the alternative has
been tried but resulted in even more confusion than
the present one does. The temptation to read X in
{ganai X gi Y} as the antecedent far outweighed the
symmetry of having this mean the same as {X anai Y}.
The option of changing orders does not arise in either
of these schemes. I expect that similar problems would
arise from mucking about with {ba} or {ria} to fit
them into one mold or the other. I also suspect that
the set up for the two different scheme depend upon
the very different natures of the two sorts of tags
involved.





posts: 1912


>
> Can we arbitrarily pick a predicate and say it
> corresponds to a tag, even if there are no visible
> similarities?

Can we arbitrarily pick a predicate for a given tag?
Obviously not.

Does every predicate have a corresponding tag?
Yes, the one obtained with {fi'o}.

Can we find a predicate for any given tag?
Yes, but obviously not one picked arbitrarily.

> This makes the notion that there is a
> correspondence totally trivial and uninteresting.

Yes, it is trivial. I don't understand the objection.

>I
> can always — as you note — make something up. So why
> continue with this project? The claim you want to make
> is pointless and the claim you appear to make is
> false.

Which project? I am simply pointing out an irregularity
in the way tags work in Lojban.


> > The tag {fio se nenri} has a short form equivalent:
> {nei}.
> > The tag {fio nenri} does not have a short form. In
> other words,
> > in {broda fio nenri koa}, you cant replace {fio
> nenri} with
> > a short tag and get the same meaning.
>
> But this is just clearly wrong if the relation is to
> the predicates in {fio} phrases. The connection of
> {nei} is obviously with {nenri}, not {se nenri} --
> unless the connection is arbitrary.

The sumti tagged by {ne'i} is the container, the se nenri.
So {ne'i} is {fi'o se nenri}.

> In the latter
> case, the whole {fio} discussion is irrelevant, the
> connection is merely mnemonic and not reducible to
> some other relation.

I am lost as to what your objection is.


> > {bau} as sumti tcita tags a sumti which would be the
> x1 of {bangu},
> > i.e. the language. {bau} is {fio bangu}.
>
> Not in any obvious sense. {fio bangu} apparently
> hooks on a reference to a language, without any
> specification of how that is to be connected to the
> rest of the sentence, {bau} is quite precise (if not
> explicit) about what the connection is: the language
> mentioned is the language in which some vocal activity
> cited in the sentence is carried out (i.e. that
> activity is te bangu).

According to CLL, the cmavo of selma'o BAI are convenient
abbreviations of the corresponding fi'o tags, so at least
in theory {bau} is fully equivalent to {fi'o bangu}.


> > It is surprizing for someone who expects Lojban to
> be regular.
> > You cant learn a single rule for the relationship
> between
> > { gi ... gi ...} and {... i bo ...}. You need to >
> learn different rules for different tags.
>
> But the diferent tags belong to different categories,
> so one expects that they behave differently. Look at
> the distribution differences for the various regular
> connectives — which have different sources. To be
> sure, I would like to get rid of these differences --
> maybe even with tag connectives — but I am not
> surprised that the differences exist.

Ok, so for you it is not surprizing. For me it was,
when I tried to make sense of it and couldn't.


> > {gu X gi Y} would correspond to {Y iju X}
> > {segu X gi Y} would correspond to {Y iseju X}.
> >
> > Currently it is the other way around.
> Currently (since the other way around is ambiguous)
> {gu X gi Y} corresponds to {X iju Y} and
> {segu X gi Y} corresponds to {X segu Y},

(Actually as it turns out not really ambiguous because
{X iju Y} happens to coincide with {Y iseju X}. So either
way you understand "the other way around" is correct.)

which fits in
> exactly with the rules for the other A connectives,

The other A connectives are symmetric, so either rule will
fit in exactly with them.


> I expect that similar problems would
> arise from mucking about with {ba} or {ria} to fit
> them into one mold or the other.

The most intuitive method would have been to make the
tags take the corresponding x2 as the argument, so
{pi'o X} for "using X" instead of {se pi'o X},
{ri'a X} for "causing X", {seri'a X} for "caused by X",
etc. I know why it wasn't done like that, but that's
the source of most of the confusion. That's the way {pu}
and {ba} work with respect to their mnemonic cognates,
but not with respect to their fi'o-counterparts, of course.
If fi'o had worked like that, then the mnemonic and fi'o
counterparts would match for them too.

> I also suspect that
> the set up for the two different scheme depend upon
> the very different natures of the two sorts of tags
> involved.

I don't see how their natures are so different.
{purci} is "x1 precedes x2 in time", {rinka} is
"x1 precedes x2 in causal link". Similarly for
{balvi} and {jalge} when x2 follows x1. The difference
is that their corresponding tags ended up in different
selma'o.

mu'o mi'e xorxes


__
Do you Yahoo!?
Free Pop-Up Blocker - Get it now
http://companion.yahoo.com/


posts: 2388

X

<<Can we find a predicate for any given tag?

Yes, but obviously not one picked arbitrarily.>>



That is, given the meaning of the tag, we can only use a predicate that means that same thing, pretty much. So the question is, given that we are in the general area of {broda} does the form of the tag affect the predicate form we can use — or, probably more accurately – does the form of the predicate affect the form of the tag. Specifically, for {se broda} could the tag be just {bo’a} or does it have to be {se bo’a}?



Part of the problem is that I am unsure what language this paper is about. This piece is clearly not about CLL Lojban nor any clearly approved variant on it, since in that langauge the answer is clearly “It has to be {se bo’a},” whereas you allow it might be {bo’a}. There are, of course (this is Lojban after all), exceptions to the rule above, motivated by convenience, I suppose, and I gather that your aim is to regularize them by having us learn for each tag what form of the underlying predicate underlies it, rather than just learning that some cases are anomolous. (In the end, I suppose it is going to be about the same amount of work, since the rule will still cover most cases.)



<< I am simply pointing out an irregularity

in the way tags work in Lojban.>>>>



But in the context of saying that every tag represents a {fi’o broda} phrase, though not always the one it appears to, for example





<<> The tag {fio se nenri} has a short form equivalent:

> {nei}.

> > The tag {fio nenri} does not have a short form. In

> other words,

> > in {broda fio nenri koa}, you cant replace {fio

> nenri} with

> > a short tag and get the same meaning.



<<The sumti tagged by {ne'i} is the container, the se nenri.

So {ne'i} is {fi'o se nenri}.>>>>



But that is flat against the rule, i.e., this is a exception made for convenence (and then not discussed or justified – not that the justification is not obvious).









<<> {bau} as sumti tcita tags a sumti which would be the

> x1 of {bangu},

> > i.e. the language. {bau} is {fio bangu}.

>

> Not in any obvious sense. {fio bangu} apparently

> hooks on a reference to a language, without any

> specification of how that is to be connected to the

> rest of the sentence, {bau} is quite precise (if not

> explicit) about what the connection is: the language

> mentioned is the language in which some vocal activity

> cited in the sentence is carried out (i.e. that

> activity is te bangu).



according to CLL, the cmavo of selma'o BAI are convenient

abbreviations of the corresponding fi'o tags, so at least

in theory {bau} is fully equivalent to {fi'o bangu}.>>



And match the predicate in ordering, etc. The claim is also in doubt, since there are many examples like {bau} where the {fi’o} requires glorking – or checking to see what the BAI (if there is one) says – but the BAI does not (cf. tanru and lujvo, I think)





<<> > It is surprizing for someone who expects Lojban to

> be regular.

> > You cant learn a single rule for the relationship

> between

> > { gi ... gi ...} and {... i bo ...}. You need to >

> learn different rules for different tags.

>

> But the diferent tags belong to different categories,

> so one expects that they behave differently. Look at

> the distribution differences for the various regular

> connectives — which have different sources. To be

> sure, I would like to get rid of these differences --

> maybe even with tag connectives — but I am not

> surprised that the differences exist.



Ok, so for you it is not surprizing. For me it was,

when I tried to make sense of it and couldn't.>>



Now this part is clearly about the current language, since it is a complaint (or at least a puzzlement) not a new stipulation. It is not what one would expect, coming at it from one way and just what one would expect coming from some others. There are several analogies possible here and they point in different directions. I suspect that at different times in the process different ones were uppermost and so we got various results. The BAI case (courtesy of its underlying predicates) is a bit more complex:

Predicate order {X rinka Y}

Tag order (Y ri’a X}

Adverb order {X i ri’a (la’e di’u) Y}

Logical connective transition order: A+B => +A,B



So, from the first part, what should the the afterthought connective be? Generally, the tag order wins this one over the adverbial and the predicative (and thus generates a potential confusion). {Y i ri’a bo X} But, for BAI, the the connective pattern dominates the tag in transition, giving {ri’a gi Y gi X} (the {iju}-{gu} pattern, with the potential problems it raises). For PU, the tag order wins out—probably a more satisfying result in the short run.





<<> > {gu X gi Y} would correspond to {Y iju X}

> > {segu X gi Y} would correspond to {Y iseju X}.

> >

> > Currently it is the other way around.

> Currently (since the other way around is ambiguous)

> {gu X gi Y} corresponds to {X iju Y} and

> {segu X gi Y} corresponds to {X segu Y},



(Actually as it turns out not really ambiguous because

{X iju Y} happens to coincide with {Y iseju X}. So either

way you understand "the other way around" is correct.)>>



In structure though not in result.



<<which fits in

> exactly with the rules for the other A connectives,



The other A connectives are symmetric, so either rule will

fit in exactly with them.>>



True, but the rule used is to keep order.



<<The most intuitive method would have been to make the

tags take the corresponding x2 as the argument, so

{pi'o X} for "using X" instead of {se pi'o X},

{ri'a X} for "causing X", {seri'a X} for "caused by X",

etc. I know why it wasn't done like that, but that's

the source of most of the confusion. That's the way {pu}

and {ba} work with respect to their mnemonic cognates,

but not with respect to their fi'o-counterparts, of course.

If fi'o had worked like that, then the mnemonic and fi'o

counterparts would match for them too.>>



I’m not sure why this is more intuitive – unless you mean “makes the stuff around the tag look more like the stuff around the predicate,” which it does. Practically, though, I don’t see that it imroves things any: some gismu will probably give preference to 2nd place ({balvi, nenri}) others to 1st ({bangu}) and others split about even ({rinka} and the causals generally). I don’t think one way has a clear Zipfean advantage, so the heuristic gain of having tag and predicate match form for form is probably a deciding factor.



<<> I also suspect that

> the set up for the two different scheme depend upon

> the very different natures of the two sorts of tags

> involved.



I don't see how their natures are so different.

{purci} is "x1 precedes x2 in time", {rinka} is

"x1 precedes x2 in causal link". Similarly for

{balvi} and {jalge} when x2 follows x1. The difference

is that their corresponding tags ended up in different

selma'o>>


By the time structure is inherently different from the causal: it has the double order of before-after and past-present-future, as well as a variety of physico-mathematical properties that causation lacks. But it is the interplay of the A and the B series in time that put {balvi} in the tense selma’o.



posts: 1912


> That is, given the meaning of the tag, we can only use a predicate that means
> that same thing, pretty much. So the question is, given that we are in the
> general area of {broda} does the form of the tag affect the predicate form we
> can use — or, probably more accurately – does the form of the predicate
> affect the form of the tag. Specifically, for {se broda} could the tag be
> just {bo’a} or does it have to be {se bo’a}?

The form of the tag, in principle, is irrelevant to the form of the
corresponding fi'o predicate. In practice, most tags were chosen based
on a gismu and thus with a related form, but this was not done
systematically. BAIs do have an officially associated gismu. But not
all tags need have a related gismu, and some may be related to a gismu
that has a totally different form, such as {re'o}, which is {fi'o lamji}.
(Or maybe {fi'o se lamji}, but in this case it makes little difference.)

The forms of the tag and related selbri may have mnemonic value but it
is irrelevant for this analysis.

> Part of the problem is that I am unsure what language this paper is about.
> This piece is clearly not about CLL Lojban nor any clearly approved variant
> on it, since in that langauge the answer is clearly “It has to be {se bo’a},”
> whereas you allow it might be {bo’a}.

I don't know what you mean. In Lojban, we have different
cases:

ri'a = fi'o rinka
ne'i = fi'o se nenri
re'o = fi'o (se) lamji

> There are, of course (this is Lojban
> after all), exceptions to the rule above, motivated by convenience, I
> suppose, and I gather that your aim is to regularize them by having us learn
> for each tag what form of the underlying predicate underlies it, rather than
> just learning that some cases are anomolous. (In the end, I suppose it is
> going to be about the same amount of work, since the rule will still cover
> most cases.)

I don't have any such aim. Whatever method you use to learn the meanings
of the tags, once you know what the tag means, you know at least roughly
the underlying predicate. Once you know what {ri'a} means, you know that
it is {fi'o broda} where the x1 of broda is a cause, and once
you know what {ne'i} means, you know that it is {fi'o broda} where
the x1 of broda is a container. You can then argue whether {ne'i} is
closer to {fi'o selnenri} or to {fi'o vasru}, or something else, but it
will never occur to you to say it is {fi'o nenri}.

All that has nothing to do with the irregularity that I am discussing.
It is simply a fact that every tag can be written as {fi'o broda} for
some suitably chosen broda.


> <<The sumti tagged by {ne'i} is the container, the se nenri.
> So {ne'i} is {fi'o se nenri}.>>>>
>
> But that is flat against the rule, i.e., this is a exception made for
> convenence (and then not discussed or justified – not that the justification
> is not obvious).

An exception to what rule? There is no rule for the forms of tags.


> <<according to CLL, the cmavo of selma'o BAI are convenient
> abbreviations of the corresponding fi'o tags, so at least
> in theory {bau} is fully equivalent to {fi'o bangu}.>>
>
> And match the predicate in ordering, etc. The claim is also in doubt, since
> there are many examples like {bau} where the {fi’o} requires glorking – or
> checking to see what the BAI (if there is one) says – but the BAI does not
> (cf. tanru and lujvo, I think)

I don't think {fi'o bangu} requires any more glorking than {bau}.
(Or {fi'o se bangu} than {se bau}, or {fi'o te bangu} than {te bau}.)


> The BAI case
> (courtesy of its underlying predicates) is a bit more complex:
>
> Predicate order {X rinka Y}
>
> Tag order (Y ri’a X}
>
> Adverb order {X i ri’a (la’e di’u) Y}
>
> Logical connective transition order: A+B => +A,B
>
> So, from the first part, what should the the afterthought connective be?
> Generally, the tag order wins this one over the adverbial and the predicative
> (and thus generates a potential confusion). {Y i ri’a bo X} But, for BAI,
> the the connective pattern dominates the tag in transition, giving {ri’a gi Y
> gi X} (the {iju}-{gu} pattern, with the potential problems it raises).

Actually, it's {ri'a gi X gi Y}, the opposite of the iju-gu pattern.

> For
> PU, the tag order wins out—probably a more satisfying result in the short
> run.

PU follows iju-gu: {X ibabo Y} = {ba gi X gi Y}

> <<The most intuitive method would have been to make the
> tags take the corresponding x2 as the argument, so
> {pi'o X} for "using X" instead of {se pi'o X},
> {ri'a X} for "causing X", {seri'a X} for "caused by X",
> etc. I know why it wasn't done like that, but that's
> the source of most of the confusion. That's the way {pu}
> and {ba} work with respect to their mnemonic cognates,
> but not with respect to their fi'o-counterparts, of course.
> If fi'o had worked like that, then the mnemonic and fi'o
> counterparts would match for them too.>>
>
> I’m not sure why this is more intuitive – unless you mean “makes the stuff
> around the tag look more like the stuff around the predicate,” which it does.

Yes. If I'm not mistaken, prepostions in many natlangs are
related to verbs, and their object is the direct object of the
verb.

> Practically, though, I don’t see that it imroves things any: some gismu will
> probably give preference to 2nd place ({balvi, nenri}) others to 1st
> ({bangu}) and others split about even ({rinka} and the causals generally). I
> don’t think one way has a clear Zipfean advantage, so the heuristic gain of
> having tag and predicate match form for form is probably a deciding factor.

I wasn't thinking so much of Zipfean advantage as of positional
consistency: the x2 argument typically falls right after the selbri.

The problem with that approach is that only x1 and x2 of the selbri
are easily accessible. To get x3 we would need something like
{fi'o se te se broda}, so that the x3 is in second position.

> <<> I also suspect that
> > the set up for the two different scheme depend upon
> > the very different natures of the two sorts of tags
> > involved.
>
> I don't see how their natures are so different.
> {purci} is "x1 precedes x2 in time", {rinka} is
> "x1 precedes x2 in causal link". Similarly for
> {balvi} and {jalge} when x2 follows x1. The difference
> is that their corresponding tags ended up in different
> selma'o>>
>
> By the time structure is inherently different from the causal: it has the
> double order of before-after and past-present-future, as well as a variety of
> physico-mathematical properties that causation lacks. But it is the
> interplay of the A and the B series in time that put {balvi} in the tense
> selma’o.

What are the A and B series?

mu'o mi'e xorxes


__
Do you Yahoo!?
Free Pop-Up Blocker - Get it now
http://companion.yahoo.com/


posts: 2388



<<The form of the tag, in principle, is irrelevant to the form of the

corresponding fi'o predicate.>>



Not what CLL says: page 195-6



<<In practice, most tags were chosen based

on a gismu and thus with a related form, but this was not done

systematically. BAIs do have an officially associated gismu. But not

all tags need have a related gismu, and some may be related to a gismu

that has a totally different form, such as {re'o}, which is {fi'o

lamji}.

(Or maybe {fi'o se lamji}, but in this case it makes little

difference.)



The forms of the tag and related selbri may have mnemonic value but it

is irrelevant for this analysis.>>



As I said, if you are going to insist on your (not CLL’s) version of {fi’o}. There are –this being Lojban – exceptions to the rule, but that does not mean that the rule is not there.



<<> Part of the problem is that I am unsure what language this paper is

about.

> This piece is clearly not about CLL Lojban nor any clearly approved

variant

> on it, since in that langauge the answer is clearly “It has to be {se

bo’a},”

> whereas you allow it might be {bo’a}.



I don't know what you mean. In Lojban, we have different

cases:



ri'a = fi'o rinka

ne'i = fi'o se nenri

re'o = fi'o (se) lamji



> There are, of course (this is Lojban

> after all), exceptions to the rule above, motivated by convenience, I

> suppose, and I gather that your aim is to regularize them by having

us learn

> for each tag what form of the underlying predicate underlies it,

rather than

> just learning that some cases are anomolous. (In the end, I suppose

it is

> going to be about the same amount of work, since the rule will still

cover

> most cases.)



I don't have any such aim. Whatever method you use to learn the

meanings

of the tags, once you know what the tag means, you know at least

roughly

the underlying predicate. Once you know what {ri'a} means, you know

that

it is {fi'o broda} where the x1 of broda is a cause, and once

you know what {ne'i} means, you know that it is {fi'o broda} where

the x1 of broda is a container. You can then argue whether {ne'i} is

closer to {fi'o selnenri} or to {fi'o vasru}, or something else, but it

will never occur to you to say it is {fi'o nenri}.>>



Which is why it is is an irregular formation.





<<> <<The sumti tagged by {ne'i} is the container, the se nenri.

> So {ne'i} is {fi'o se nenri}.>>>>

>

> But that is flat against the rule, i.e., this is a exception made for

> convenence (and then not discussed or justified – not that the

justification

> is not obvious).



An exception to what rule? There is no rule for the forms of tags.>>



It sure looks like one – or rather a set of examples and wording that indicates that they are the paradigm.





<<> <<according to CLL, the cmavo of selma'o BAI are convenient

> abbreviations of the corresponding fi'o tags, so at least

> in theory {bau} is fully equivalent to {fi'o bangu}.>>

>

> And match the predicate in ordering, etc. The claim is also in

doubt, since

> there are many examples like {bau} where the {fi’o} requires glorking

– or

> checking to see what the BAI (if there is one) says – but the BAI

does not

> (cf. tanru and lujvo, I think)



I don't think {fi'o bangu} requires any more glorking than {bau}.

(Or {fi'o se bangu} than {se bau}, or {fi'o te bangu} than {te bau}.)>>



Well, we disagree. I think it is merely that the range of options is very small for {bangu} – although {fi’o te bangu} is pretty obscure – but then so is {tebau}.





<<> The BAI case

> (courtesy of its underlying predicates) is a bit more complex:

>

> Predicate order {X rinka Y}

>

> Tag order (Y ri’a X}

>

> Adverb order {X i ri’a (la’e di’u) Y}

>

> Logical connective transition order: A+B => +A,B

>

> So, from the first part, what should the the afterthought connective

be?

> Generally, the tag order wins this one over the adverbial and the

predicative

> (and thus generates a potential confusion). {Y i ri’a bo X} But,

for BAI,

> the the connective pattern dominates the tag in transition, giving

{ri’a gi Y

> gi X} (the {iju}-{gu} pattern, with the potential problems it

raises).



Actually, it's {ri'a gi X gi Y}, the opposite of the iju-gu pattern.>>



In the pattern you gave (<<{X i ri'a bo Y} corresponds to {ri'a gi X gi Y}>>), the cause element remains in second place and the effect in first when the causal connective is fronted, analogous to the unchanged positions in {u}. CLL has it the opposite way(i.e. {ri’a gi X gi Y}, 8.2 v. 8.1, p.199), but I was dealing with your idea.

<<> For

> PU, the tag order wins out—probably a more satisfying result in the

short

> run.



PU follows iju-gu: {X ibabo Y} = {ba gi X gi Y}>>



Errh, you said <<{X i ba bo Y} corresponds to {ba gi Y gi X}>>. Your current version is that of CLL. Following you, I reversed the orders. The problem remains, though the exemplars shift. Actually, the more interesting problem is the one CLL covers with “As a result” at the bottom of p248, where {X ba Y} is shifted to {Y iba X}. Shifting all that back, it turns out the two cases are exactly parallel: effect ri’a cause => effect iri’abo cause => ri’a gi cause gi effect, and event ba axis => axis ibabo event => ba gi axis gi event, with only the one anomaly of axis ibabo event rather than event ibabo axis, wich is the result of some unmentioned – but referred to – factor.



<<> <<The most intuitive method would have been to make the

> tags take the corresponding x2 as the argument, so

> {pi'o X} for "using X" instead of {se pi'o X},

> {ri'a X} for "causing X", {seri'a X} for "caused by X",

> etc. I know why it wasn't done like that, but that's

> the source of most of the confusion. That's the way {pu}

> and {ba} work with respect to their mnemonic cognates,

> but not with respect to their fi'o-counterparts, of course.

> If fi'o had worked like that, then the mnemonic and fi'o

> counterparts would match for them too.>>

>

> I’m not sure why this is more intuitive – unless you mean “makes the

stuff

> around the tag look more like the stuff around the predicate,” which

it does.



Yes. If I'm not mistaken, prepostions in many natlangs are

related to verbs, and their object is the direct object of the

verb.>>



Some are, some aren’t and whether it is the direct object or the indirect or the subject varies from case to case. It is not a a principle that I would appeal to, since it does not hold to any very marked extent in any language we are likely to be familiar with (Chinese excepted, as usual).



<<> Practically, though, I don’t see that it imroves things any: some

gismu will

> probably give preference to 2nd place ({balvi, nenri}) others to 1st

> ({bangu}) and others split about even ({rinka} and the causals

generally). I

> don’t think one way has a clear Zipfean advantage, so the heuristic

gain of

> having tag and predicate match form for form is probably a deciding

factor.



I wasn't thinking so much of Zipfean advantage as of positional

consistency: the x2 argument typically falls right after the selbri.



The problem with that approach is that only x1 and x2 of the selbri

are easily accessible. To get x3 we would need something like

{fi'o se te se broda}, so that the x3 is in second position.>>



I am inclined to think that is good enough to drop the idea, even though the third and later places are rarely used – except for a few preds.

<<What are the A and B series?>>
Past-Present-Future and Before-After, both temporal and very similar at an abstract level but not mutually definable.



posts: 1912


> <<> The BAI case
> > (courtesy of its underlying predicates) is a bit more complex:
> >
> > Predicate order {X rinka Y}
> >
> > Tag order (Y ri’a X}
> >
> > Adverb order {X i ri’a (la’e di’u) Y}
> >
> > Logical connective transition order: A+B => +A,B
> >
> > So, from the first part, what should the the afterthought connective
> be?
> > Generally, the tag order wins this one over the adverbial and the
> predicative
> > (and thus generates a potential confusion). {Y i ri’a bo X} But,
> for BAI,
> > the the connective pattern dominates the tag in transition, giving
> {ri’a gi Y
> > gi X} (the {iju}-{gu} pattern, with the potential problems it
> raises).
>
> Actually, it's {ri'a gi X gi Y}, the opposite of the iju-gu pattern.>>
>
> In the pattern you gave (<<{X i ri'a bo Y} corresponds to {ri'a gi X gi
> Y}>>), the cause element remains in second place and the effect in first when
> the causal connective is fronted, analogous to the unchanged positions in
> {u}. CLL has it the opposite way(i.e. {ri’a gi X gi Y}, 8.2 v. 8.1, p.199),
> but I was dealing with your idea.

No, I'm perfectly happy with the way CLL does {ri'a}.
I think {ri'a gi X gi Y} = {Y iri'abo X} is the Right Thing,
and that it would have been better if {gu} had been defined
so that {gu X gi Y} was {Y iju X}.

> <<> For
> > PU, the tag order wins out—probably a more satisfying result in the
> short
> > run.
>
> PU follows iju-gu: {X ibabo Y} = {ba gi X gi Y}>>
>
> Errh, you said <<{X i ba bo Y} corresponds to {ba gi Y gi X}>>.

Where did I say that? If it did, there wouldn't be a discrepancy.


> Your current
> version is that of CLL. Following you, I reversed the orders. The problem
> remains, though the exemplars shift.

If I could reverse the order of something, I wouldn't reverse
the order of everything, that would be silly! Of course the problem
would remain. The problem is not with one particular order but rather
with the inconsistency that one rule is used for some tags and the
opposite rule for other tags.

> Actually, the more interesting problem
> is the one CLL covers with “As a result” at the bottom of p248, where {X ba
> Y} is shifted to {Y iba X}. Shifting all that back, it turns out the two
> cases are exactly parallel: effect ri’a cause => effect iri’abo cause =>
> ri’a gi cause gi effect, and event ba axis => axis ibabo event => ba gi axis
> gi event, with only the one anomaly of axis ibabo event rather than event
> ibabo axis, wich is the result of some unmentioned – but referred to –
> factor.

Yes, that's what I say on the wiki page. {ba} and {ri'a} are
parallel in everything except for {ibabo}/{iri'abo}.

mu'o mi'e xorxes



__
Do you Yahoo!?
Free Pop-Up Blocker - Get it now
http://companion.yahoo.com/


On Tue, 2003-11-25 at 05:56, Jorge Llamb=EDas wrote:
> Yes, that's what I say on the wiki page. {ba} and {ri'a} are=20
> parallel in everything except for {ibabo}/{iri'abo}.
>=20
> mu'o mi'e xorxes

FWIW, I always thought it odd that ibabo means "and then...". Always
seemed backwards to me. I guess I'll have to do a bit more research
before I understand why it isn't.

-- Ted



posts: 2388

<<> Errh, you said <<{X i ba bo Y} corresponds to {ba gi Y gi X}>>.



Where did I say that? If it did, there wouldn't be a discrepancy.





> Your current

> version is that of CLL. Following you, I reversed the orders. The

problem

> remains, though the exemplars shift.



If I could reverse the order of something, I wouldn't reverse

the order of everything, that would be silly>>



Your earlier contributions to this discussion, eg 11/22 @5:38pm. You have it right on the wiki page. I get those things turned around myself, so no harm done. Indeed some help, as it highlights the problem.



The problem arises because {i ba bo} already has an assigned meaning, the adeverbial one (as the gloss in CLL shows). Modals use {ku} for the same purpose and so there is no conflict. I suggest the easiest thing is to always use {ku} with {ba} as well (it is already legal, I think, and if not could easily be made so) and then bring PU into line with BAI (though that would ruin two pages of CLL which purpport – incorrectly –to explain the difference).




posts: 2388

<<The problem arises because {i ba bo} already has an assigned meaning,
the adeverbial one (as the gloss in CLL shows). Modals use {ku} for the
same purpose and so there is no conflict. I suggest the easiest thing
is to always use {ku} with {ba} as well (it is already legal, I think,
and if not could easily be made so) and then bring PU into line with
BAI (though that would ruin two pages of CLL which purpport – incorrectly
–to explain the difference).>>

On second thought, having two marginally different structures with completely opposite meanings looks to be a blueprint for disaster. Since the adverb use is so well embedded - from the form with a sumti — it might be better to use that form for the much less common (?) connective use, that is, have BAI follow PU rather than the other way.



posts: 1912


> On second thought, having two marginally different structures with completely
> opposite meanings looks to be a blueprint for disaster. Since the adverb use
> is so well embedded - from the form with a sumti — it might be better to
> use that form for the much less common (?) connective use, that is, have BAI
> follow PU rather than the other way.

Unfortunately, both {iseki'ubo} and {ibabo} have a rather entrenched
usage, so changing either would seem at this point to be out of the
question. We can only point out the irregularity. I assume, though
I'm not sure CLL says anything about it, that other non-BAI tags
follow the PU pattern.

mi'e xorxes


__
Do you Yahoo!?
Free Pop-Up Blocker - Get it now
http://companion.yahoo.com/


posts: 1

I can't keep track of who is saying what in response to what.

--And.

--- Original Message ---
From: John E Clifford <clifford-j@sbcglobal.net>
To: <wikidiscuss-list@lojban.org>
Sent: Tuesday, November 25, 2003 3:34 AM
Subject: WikiDiscuss Re: tags as connectives


>
>
> <<The form of the tag, in principle, is irrelevant to the form of the
>
> corresponding fi'o predicate.>>
>
>
>
> Not what CLL says: page 195-6
>
>
>
> <<In practice, most tags were chosen based
>
> on a gismu and thus with a related form, but this was not done
>
> systematically. BAIs do have an officially associated gismu. But not
>
> all tags need have a related gismu, and some may be related to a gismu
>
> that has a totally different form, such as {re'o}, which is {fi'o
>
> lamji}.
>
> (Or maybe {fi'o se lamji}, but in this case it makes little
>
> difference.)
>
>
>
> The forms of the tag and related selbri may have mnemonic value but it
>
> is irrelevant for this analysis.>>
>
>
>
> As I said, if you are going to insist on your (not CLL's) version of
{fi'o}. There are -this being Lojban - exceptions to the rule, but that
does not mean that the rule is not there.
>
>
>
> <<> Part of the problem is that I am unsure what language this paper is
>
> about.
>
> > This piece is clearly not about CLL Lojban nor any clearly approved
>
> variant
>
> > on it, since in that langauge the answer is clearly "It has to be {se
>
> bo'a},"
>
> > whereas you allow it might be {bo'a}.
>
>
>
> I don't know what you mean. In Lojban, we have different
>
> cases:
>
>
>
> ri'a = fi'o rinka
>
> ne'i = fi'o se nenri
>
> re'o = fi'o (se) lamji
>
>
>
> > There are, of course (this is Lojban
>
> > after all), exceptions to the rule above, motivated by convenience, I
>
> > suppose, and I gather that your aim is to regularize them by having
>
> us learn
>
> > for each tag what form of the underlying predicate underlies it,
>
> rather than
>
> > just learning that some cases are anomolous. (In the end, I suppose
>
> it is
>
> > going to be about the same amount of work, since the rule will still
>
> cover
>
> > most cases.)
>
>
>
> I don't have any such aim. Whatever method you use to learn the
>
> meanings
>
> of the tags, once you know what the tag means, you know at least
>
> roughly
>
> the underlying predicate. Once you know what {ri'a} means, you know
>
> that
>
> it is {fi'o broda} where the x1 of broda is a cause, and once
>
> you know what {ne'i} means, you know that it is {fi'o broda} where
>
> the x1 of broda is a container. You can then argue whether {ne'i} is
>
> closer to {fi'o selnenri} or to {fi'o vasru}, or something else, but it
>
> will never occur to you to say it is {fi'o nenri}.>>
>
>
>
> Which is why it is is an irregular formation.
>
>
>
>
>
> <<> <<The sumti tagged by {ne'i} is the container, the se nenri.
>
> > So {ne'i} is {fi'o se nenri}.>>>>
>
> >
>
> > But that is flat against the rule, i.e., this is a exception made for
>
> > convenence (and then not discussed or justified - not that the
>
> justification
>
> > is not obvious).
>
>
>
> An exception to what rule? There is no rule for the forms of tags.>>
>
>
>
> It sure looks like one - or rather a set of examples and wording that
indicates that they are the paradigm.
>
>
>
>
>
> <<> <<according to CLL, the cmavo of selma'o BAI are convenient
>
> > abbreviations of the corresponding fi'o tags, so at least
>
> > in theory {bau} is fully equivalent to {fi'o bangu}.>>
>
> >
>
> > And match the predicate in ordering, etc. The claim is also in
>
> doubt, since
>
> > there are many examples like {bau} where the {fi'o} requires glorking
>
> - or
>
> > checking to see what the BAI (if there is one) says - but the BAI
>
> does not
>
> > (cf. tanru and lujvo, I think)
>
>
>
> I don't think {fi'o bangu} requires any more glorking than {bau}.
>
> (Or {fi'o se bangu} than {se bau}, or {fi'o te bangu} than {te bau}.)>>
>
>
>
> Well, we disagree. I think it is merely that the range of options is very
small for {bangu} - although {fi'o te bangu} is pretty obscure - but then so
is {tebau}.
>
>
>
>
>
> <<> The BAI case
>
> > (courtesy of its underlying predicates) is a bit more complex:
>
> >
>
> > Predicate order {X rinka Y}
>
> >
>
> > Tag order (Y ri'a X}
>
> >
>
> > Adverb order {X i ri'a (la'e di'u) Y}
>
> >
>
> > Logical connective transition order: A+B => +A,B
>
> >
>
> > So, from the first part, what should the the afterthought connective
>
> be?
>
> > Generally, the tag order wins this one over the adverbial and the
>
> predicative
>
> > (and thus generates a potential confusion). {Y i ri'a bo X} But,
>
> for BAI,
>
> > the the connective pattern dominates the tag in transition, giving
>
> {ri'a gi Y
>
> > gi X} (the {iju}-{gu} pattern, with the potential problems it
>
> raises).
>
>
>
> Actually, it's {ri'a gi X gi Y}, the opposite of the iju-gu pattern.>>
>
>
>
> In the pattern you gave (<<{X i ri'a bo Y} corresponds to {ri'a gi X gi
Y}>>), the cause element remains in second place and the effect in first
when the causal connective is fronted, analogous to the unchanged positions
in {u}. CLL has it the opposite way(i.e. {ri'a gi X gi Y}, 8.2 v. 8.1,
p.199), but I was dealing with your idea.
>
> <<> For
>
> > PU, the tag order wins out-probably a more satisfying result in the
>
> short
>
> > run.
>
>
>
> PU follows iju-gu: {X ibabo Y} = {ba gi X gi Y}>>
>
>
>
> Errh, you said <<{X i ba bo Y} corresponds to {ba gi Y gi X}>>. Your
current version is that of CLL. Following you, I reversed the orders. The
problem remains, though the exemplars shift. Actually, the more interesting
problem is the one CLL covers with "As a result" at the bottom of p248,
where {X ba Y} is shifted to {Y iba X}. Shifting all that back, it turns
out the two cases are exactly parallel: effect ri'a cause => effect iri'abo
cause => ri'a gi cause gi effect, and event ba axis => axis ibabo event =>
ba gi axis gi event, with only the one anomaly of axis ibabo event rather
than event ibabo axis, wich is the result of some unmentioned - but referred
to - factor.
>
>
>
> <<> <<The most intuitive method would have been to make the
>
> > tags take the corresponding x2 as the argument, so
>
> > {pi'o X} for "using X" instead of {se pi'o X},
>
> > {ri'a X} for "causing X", {seri'a X} for "caused by X",
>
> > etc. I know why it wasn't done like that, but that's
>
> > the source of most of the confusion. That's the way {pu}
>
> > and {ba} work with respect to their mnemonic cognates,
>
> > but not with respect to their fi'o-counterparts, of course.
>
> > If fi'o had worked like that, then the mnemonic and fi'o
>
> > counterparts would match for them too.>>
>
> >
>
> > I'm not sure why this is more intuitive - unless you mean "makes the
>
> stuff
>
> > around the tag look more like the stuff around the predicate," which
>
> it does.
>
>
>
> Yes. If I'm not mistaken, prepostions in many natlangs are
>
> related to verbs, and their object is the direct object of the
>
> verb.>>
>
>
>
> Some are, some aren't and whether it is the direct object or the indirect
or the subject varies from case to case. It is not a a principle that I
would appeal to, since it does not hold to any very marked extent in any
language we are likely to be familiar with (Chinese excepted, as usual).
>
>
>
> <<> Practically, though, I don't see that it imroves things any: some
>
> gismu will
>
> > probably give preference to 2nd place ({balvi, nenri}) others to 1st
>
> > ({bangu}) and others split about even ({rinka} and the causals
>
> generally). I
>
> > don't think one way has a clear Zipfean advantage, so the heuristic
>
> gain of
>
> > having tag and predicate match form for form is probably a deciding
>
> factor.
>
>
>
> I wasn't thinking so much of Zipfean advantage as of positional
>
> consistency: the x2 argument typically falls right after the selbri.
>
>
>
> The problem with that approach is that only x1 and x2 of the selbri
>
> are easily accessible. To get x3 we would need something like
>
> {fi'o se te se broda}, so that the x3 is in second position.>>
>
>
>
> I am inclined to think that is good enough to drop the idea, even though
the third and later places are rarely used - except for a few preds.
>
> <<What are the A and B series?>>
> Past-Present-Future and Before-After, both temporal and very similar at an
abstract level but not mutually definable.
>
>
>
>



posts: 2388

&:
<<I can't keep track of who is saying what in response to what.>>

Sorry! One of the many flaws of my emailer de jour is that it does not carry over texts in replies, so I have to Copy them. Rather than go through line-by-line adding >, I just enclose the whole in continental quotes (as above). I see that I have been lax in adding attributions as well. But surely you know xores' and my fists well enough by now to sort us out with relative ease.

pc