Re: [lojban] Centripetal-centrifugal, little-endian--big-endian, subsets-cont...

[pycyn@aol.com]

In a message dated 5/10/00 3:45:27 PM CST, phm@a2e.de writes:

<< > I am inclined to think that the fact that human languages fairly 
regularly 
> offer centrifugal constructions is itself evidence that the 
centripetal-only 
> thought pattern is not in fact the rule. 

The European languages are a tiny minority on the planet, but they have,
not through language design merits, marginalised most of the others.>>
Relevance? Surely you do not mean that all colonizing European languages are 
totally centrifugal and that none of the marginalized or other non-European 
languages have any centrifugal forms. I suspect that though a language may 
favor one or the other pattern, most languages have signifcant numbers of 
both. But, in any case, a few languages with heavy centrifugality would make 
my point, that, were human though thoroughly centripetal, such languages 
would have long since diosappeared (or, indeed, never evolved at all) as 
being unintelligible or at least to inefficient to be practical.

<<> And, of course, none of this decides the structure of dates, since it is 
> equally possible (and, to me, more natural) to take the year as the name 
of a 
> set and a month as specifying a subset within that set and the day as 
> specifying a unit subset within that and thus get dmy again but as a 
> centripetal structure.

You mean something like 

the year 2001 .
which year 20001 ?
the year 2001 of the 5th month .
of which month 5 ?
the year 2001 of the 5th month of the 20th day .

but I fail to assign this any meaning.>>
So, obviously, it is not what I meant: more like
2001
When in 2001?
month 5 in 2001
When in month 5 of 2001?
the 20th of month 5 in 2001

> Of course if you label the units (as the Chinese fairly regularly do) then 
> you can use any order, since there is no ambiguity about what the numbers 
> mean. 

the short form originates from the long one. One can only elide in a
meaningful way, when one has a well constructed long form.>>
Well, since we can elide in a meaningful way in the detri format, the long 
form must be well-constructed.

<<> One final item on centripetal -- left expanding -- strings. My memory 
of 
> Language Theory (admittedly often 40 years old) is that there was an 
> empirical law to the effect that there was an upper limit (probably the 
> classic 5+/- 2) on the length of left-branching structures that a human 
mind 
> could process. Insofar as centripetal structures are left-branching (and 
> that may vary with the kind of structure) this would mean that they could 
not 
> handle all situations, that some centrifugal forms are necessary. Does 
> anyone know whether that rule has been repudiated? Can anyone work out 
the 
> branching structure of various types of centripetal constructions?

I don't think so.
Chinese state agencies at all levels are always named with a centripetal
pattern based on leftward expansion, i.e. with the outermost / uppermost
hierarchy level on the left.
This is very easy to handle for the human mind.
Anything else wouldn't be, because languages usually don't offer enough
brace structures that could tell the listener when to switch back and
forth between the directions. >>
Yes, switching back and forth is hard and going in the same direction 
relatively easier, but there does not seem to be a preferred direction. The 
left-branching rules turns out, once I looked it up, to be something of a 
redherring, since it need not correspond to centripality, but rather to 
grouping with the string, ie.e it corresponds to right grouping. That is, 
the problem is getting a mess of unresolved modifiers in place before you 
know what is being modified. The two solutions are either to put the 
modified first and string the modifiers out after or resolve modifiers as you 
go along,so that at each step there is only one (possibly complex) modifier 
to deal with.
I gather that the latter is more or less what centripetality is about, but I 
am often not sure from the examples. vday(vmonth(year, month)monthv, 
day)dayv for your favorite case. And, of course, viewed from that point of 
view, the detri formula looks to be 
vday(day, vmonth(month, year)monthv)dayv, left branching and reversing to 
boot. But te detri formula might equally be vday(vday(day, month)dayv, 
year)dayv, a coherent right branching pattern. As far as I can understand 
all this at the moment.

from another messsage
<<> >In real-life, a certain day is a component of a month, just like a key is
> >a component of a keybox.
> >
> >However when I say "the 20th" I don't refer to a certain day,
> 
> When I say "the 20th" I am ALWAYS referring to a certain day. Absent the 
> context, it may not be clear to you what day that certain day is.

Yes, that is how it should be said.
In my mind I am referring to a certain day, but my utterance specifies a
large set of days, from which the listener must find what I am referring
to by some kind of restriction: either context based or based on a
container attribute which I specify.

> When I say Hartmut, I am referring to a certain person, one of several 
> possible people. There is a large (albeit not nearly infinite) number of 
> people I could be referring to, but context usually says which one. If I 
> need to add more information, I include a surname, which in all but a 
> couple of Oriental languages is added at the end in human language use (I 
> note that you reverse the order in your computer name, so at least you are 
> consistent).

Here only the Oriental language have a habit that is conformant to human
thinking: putting the important thing up-front. Contrary to what you say
below.
> This follows the convention of putting the most critical, relevant, or 
> interesting information up front. Additional clarifying information, if 
> needed, is added later.

The most critical information is that, which the *recipient* needs first
in order to narrow down the set of possible meant objects.

If there are several possible Hartmuts, then 'Pilch' is the critical part. 
If I am already in the Pilch family, then 'Pilch' is no part at all,
because people will call me by the given name.>>
Actually, the oriental system (i.e. Chinese) is not clearly the opposite of 
the European. Both start with the LEAST informative thing if looking for a 
particular person. There are only a few hundred family names in China 
(including Indo-China)
but a virtually infinite set of personal names (almost any word can be a 
personal name), while in European cultures there are only a thousand or so 
personal names (with any degree of currency) but almost any word can be a 
family name (plus, in many countries words from totally other languages as 
well). 

<<You can easily verify that by sorting a name list. The critical thing
comes first in sorting. How do you sort? By surname or by given name?>>
Works most places in the west and is used in Chinese lists, but with the 
result that actually finding the right person is harder. In Iceland, where 
the "family" name: --sen,-- datter, tends to be fairly limited but the 
personal names relatively free, the phone book is by personal name first.


<<Try sorting dates or addresses. Or imagine yourself to be a postman who
has to forward a letter. If you are at a USA Central Post Office, then
what is the critical information you ask for: 

- to which zip code area should I bring the letter?
or
- to which house number should I bring the letter? 

> Standard order tanru are unlike names because they can have a unitary 
> meaning quite distinct from the final term - not all tanru are restrictive 
> (see the many forms of tanru in the appropriate chapter of the reference 
> Book). In cases of restriction, Grician relevance supports putting the key 
> information up front.
> 
> >We have here the notions of subset vs component, which are easy to
> >confuse.
> 
> Not especially, but I don't see why either is relevant to dates which are 
> names for days. It is pure convention that makes days be labelled with 
> numbers or associated with month names.

1 It is a convention that evolves by abbreviation
2 The choice determines the speed and efficiency of thinking
namely the need or not of transposing in one's mind

> >It is good language design to expand tanru by prepending rather than by
> >appending, because in address constructions (including places, names,
> >dates etc) the subset-specifier is usually also a container, and it is a
> >necessity of human thinking to proceed from the container to the
> >contained.
> 
> Why would anyone think that? The norm of human thinking is to proceed from 
> the most relevant, adding less relevant information if needed for a more 
> complete or accurate picture.

The most relevant part is the outermost container that is still needed for
restricting the meaning. In order to find a point in space/time, I need
to narrow it down from some radius, which depends on where I stand.

The western centrifugal addressing is of course based on language habits,
but it also happens to be a somewhat egocentric (sorry for the moral
valuation) habit, because it scorns the need of the recipient and takes
the needs of the speaker as the criterion for deciding which information
is most important. It is a positioning from the perspective of the ego,
not the perspective of the person who needs to find something.>>
That is an interesting point: we are concerned with whom we wish the message 
to get to, so start with them and only secondarily how to get to them. 
Happily, the result is a pretty consistently reversible pattern, which the 
mailman simply reads backwards (probably ignoring the superfluities like city 
and state -- and street address as well in the longer forms). I suspect 
there are a number of these reversible conventions in use. Do Chinese 
letters really get addressed the opposite way? 
<<One cannot see anything without first knowing the big picture. The big
picture is the frame of reference.>>
Too broad and too narrow a definition. It will depend upon the situation 
what the relevant big picture is, but it is often bigger than the frame of 
reference and also often much smaller.


<<> >Apparently these considerations could create a conflict with the design
> >freeze. They show an inconsistency in the design of "detri". The removal
> >of which will probably have to wait until some official version upgrade of
> >the "Lojban Standard". Or is this not the way how Lojban is supposed to
> >evolve?
> 
> Lojban is intended to NOT evolve during the baseline period, other than the 
> grow in vocabulary. We will not even discuss possible changes to the 
> baseline while the baseline is in effect, which will be at least 5 years 
> from whenever.

This is what I meant. My question is whether Lojban will evolve by
"upgrades" in the way standards evolve, or rather by an anarchical process
as in "natural" languages. The fact that there is a design freeze seems
to suggest the former.>>
I'm still worried about "the inconsisttency in the design of detri" -- it 
is inconsistent with some idea (not too clear what, since it seems to have 
several bases) of Hr. Pilch. And even that is not obvious, since {detri} has 
the usual structure for a predicate, so apparently all that is at issue is 
the convention for specifying the date. But that convention fits at least 
half of the various things Hr. Pilch advocates, as manyas, or more than, the 
one he proposes, so the whole tempest seems hardly worth a teacup. Don't 
like, don't use: you may be misunderstood or have to be long-winded for 
simple things, but you will survive in lojban at least as well as all the 
other convention breakers who go on their way on this -- and associated -- 
lists.