Hallo conlangers!

On Friday 16 September 2011 16:31:47, Basilius wrote:

> On Fri, 16 Sep 2011 09:21:02 +0200, Christophe Grandsire-Koevoets wrote:
> >On 16 September 2011 00:53, Basilius wrote:
> >> On Thu, 15 Sep 2011 21:32:00 +0200, Christophe Grandsire-Koevoets wrote:
> >> >Reconstructed languages are very much a model of a lost linguistic
> >> >reality indeed, not the real thing,and linguists don't consider them
> >> >to be
> >> 
> >> anything
> >> 
> >> >else.
> >> 
> >> There is a huge difference between modelling something real and
> >> modelling something imagined. The difference is very important.
> >
> >I disagree. What is the difference between, say, a model of the Statue of
> >Liberty, a model of the Colossus of Rhodes, and a model of the statue in
> >_Lost_? The fact that the first one represents something existing today,
> >the second one something we know existed but have little evidence of how
> >it looked like (although we can make strong guesses based on material
> >science and physics in general, and the little indirect evidence we have
> >for it), and the last one something fictional (but for which we know
> >exactly how it looks like) says nothing about the models *sui generis*.
> >The models will be as good as the builder made them, nothing less and
> >nothing more. The only difference is *how* they can be used, but that's a
> >separate issue entirely.

Yes.  They are all models.
 
> The difference is about the object, obviously. A model of a real object
> implies that it's real, and that the model can be used to understand or
> predict something about that real object (including something that can be
> checked on additional data etc.).

Sure.  Yet, the model is still a *model* and not the modelled
object itself - the map is not the territory.

>       A model of a fictional object implies
> that the object is fictional, and is not supposed to tell much about the
> real world, only about the plot of the fictional object's creator.

This is also true, and this is the reason why reconstructed
languages are not conlangs in the same way such languages as
Sindarin or Na'vi are.  The former are models of hypothesized
languages which were once real (or at least, assumed real by
the author, see for instance Proto-Nostratic), while the latter
are models of *imaginary* languages.  Yet, both are models of
languages rather than true living languages.

> A reconstruction of a protolanguage does not imply that the protolanguage
> was a conlang, and thus is not a conlang description.

Indeed not.  It implies that the protolanguage was a natlang,
and it is thus to be regarded a description of a natlang.
It may be wrong, though (for instance, there are two major
- and several minor - mutually incompatible reconstructions of
Proto-Nostratic on the market which cannot both be right).

But again, a "description of a conlang" is not the same as a
"conlang".  Don't confuse the object and the description!
*Nobody* has said that Proto-Indo-European (or any other such
reconstructed language) was a "description of a conlang"; it
is a model, or if you like that better, a description of a
hypothetical *natlang*.  The *model itself* can be considered
a conlang - that is Christophe's and my point - even if the
object language is not.

To stay in the model railroad metaphor (which is quite apt,
I think), a reconstructed language is like a model railroad
that is a scale model of a railway station of an earlier time
which has since then been demolished and rebuilt, built after
a collection of old photos and drawings.  A conlang proper is
a model railroad that merely embodies the fancies of its
builder, without reference to any real locale past or present.

>       It does not matter
> how formally similar it may prove to be with (a certain type of) conlang
> descriptions (which is how I understood your referring to "models sui
> generis"); the nature of the restrictions imposed on the model by the way
> it's supposed to be used is different, and the restrictions themselves are
> incomparably stronger (if we are speaking of reconstructions and conlangs
> in general).

Sure.  A scholar reconstructing a lost language is expected to
base *every detail* of his reconstruction on available evidence
from natlangs, and leave blank what he cannot reconstruct.
This, after all, is why I call Old Albic a "re-creation" rather
than a "reconstruction": it goes beyond the available evidence,
and I fill in the many irreconstructible blank spots with my
own fancies.
 
> (Note also that "model" is not my word in this discussion; I don't consider
> "modelling" a felicitous metaphor for what scientific research does, but
> that's a big and mostly unrelated issue by itself, and I won't be picky
> about this language, since it's common and I understand it.)

I think the metaphor is apt, and used by scholars of all trades.
Scientists often work with a representation of the studied object
rather than the object itself - and guess what is the word for
such a representation: _model_.
 
> >Where I can agree is that there is a difference between modelling
> >something *attested* vs. modelling something *unattested*. But that's
> >very different from the question of the reality of what is being
> >modelled.
> 
> And my point was that "attested" is not a simple concept; it does not
> exclude reconstruction; moreover, in a sense, we don't have anything
> attested *directly*. (I am not going to deny, though, that attestation of a
> language can be *more* direct than it is with reconstructed
> protolanguages.)

Surely, there is direct and indirect attestation.  It is, for
instance, said that some of the PIE laryngeals are "attested
in Hittite"; yet, we don't know the exact pronunciation of the
Hittite reflexes of the laryngeals, nor whether they were
pronounced the same way as in PIE, or have been changed.  So
the attestation of the PIE laryngeals in Hittite is an indirect
one.
 
> And, perhaps more relevant to this discussion, it's not attestation that
> makes a natlang different from a conlang.

The point of this thread is that the *reconstructions* used
in historical linguistics are in a wide sense conlangs, not
the languages which these reconstructions are meant to model.
 
> >> Alternatively, you can consider everything written in grammars (all
> >> kinds of
> >> grammars) as describing conlangs rather than natlangs, for grammars are
> >> just
> >> models, after all. But very few people will agree with you on that, I
> >> guess.
> >
> >Non-sequitur. This just does not follow from what I've said, so I'd by
> >happy if you wouldn't put words in my mouth that I didn't say.

Indeed, it doesn't follow.  Christophe is right here.
 
> I am sorry if my words sounded offensive for some reason, but I can't
> understand you without trying to (..ehm...) model your stance :) So, do you
> consider grammars (and language descriptions in general) to be a kind of
> models? If not, why not?

Of course, a language description is not a language!  It is a
*description* of a language - a kind of model indeed.
 
> >> And denying the reality of PIE (or protolanguages in general), despite
> >> all the evidence, is para-science.
> >
> >Reading comprehension fail! The OP wrote (and I repeated) that
> >reconstructed PIE is a model of a "lost linguistic reality". How can this
> >in any way be denying the existence of PIE? It's exactly the opposite!
> >Why do you keep arguing about things I didn't write?
> 
> Equalling reconstructions to conlangs (or conlang descriptions) amounts to
> denying the reality of protolanguages, for me. This is what I objected to,
> and tried to explicate. Maybe failed.

The *reconstruction* is a kind of conlang, not the language
it is meant to model!  (Unless of course the reconstruction
is one of a protolanguage which never really existed and is
based on faulty or misinterpreted data.  There are many
reconstructions around whose validity is doubted by most
scholars.)
 
> >> There was a real language from which some
> >> extant languages descended, and PIE reconstructions are about that real
> >> ancestral language;
> >
> >Of course! But they are just that: reconstructions. They are *not* that
> >real ancestral language, but a model of it, based on the indirect
> >evidence we have. It may be the best we have, but without direct evidence
> >(i.e. written texts we can decipher or a time-travel machine to allow us
> >to witness PIE people speaking), we will never know how real PIE was
> >like.
> 
> Heh. And you seem to rely too much on the idea that evidence can be
> "direct" ;)

Oh, now we are getting into epistemology ;)  What can we know
and what can we never know?  And can we trust our senses?
 
> >There is a reason why reconstructed PIE forms are always introduced with
> >the * sign in linguistic papers, and that it's the same sign used to mark
> >ungrammatical utterances: it's the sign for "unattested". We know PIE
> >existed, but everything we know about it is from *indirect* evidence. As
> >long as we have no *direct* evidence, we can never claim the
> >reconstruction *is* real PIE. What it is is a conlang meant to represent
> >PIE to the best of our knowledge, and we cannot claim it to be anything
> >else.

Exactly that is why the Indo-Europeanists (and their colleagues
in other branches of historical linguistics) carefully mark their
reconstructions with asterisks: to alert the reader that the forms
are not actually attested but reconstructed, and therefore to be
taken with a grain of salt.

And reconstructions are never the final word.  The reconstruction
of Proto-Indo-European, for instance, has been revised several
times during the last 200 years, and will continue to be revised
in the future.

> >But that doesn't make the reconstruction any less useful. It's just
> >that there is a difference of nature between knowledge gathered from
> >indirect evidence and knowledge gathered from direct evidence, and that's
> >the only distinction that truly counts in science.

Yes.
 
> My objections to this would look redundant; see above.
> 
> >If you cannot make the distinction between knowledge gathered from direct
> >evidence and knowledge gathered from indirect evidence, I hope you never
> >have to serve in a jury! :)
> 
> I hope not; I'm the kind of guy who will always question the idea of its
> being "direct" ;)
> 
> >> they'd have very little scientific value if they were
> >> just products of their proponents' creative imagination.
> >
> >True, but the term "conlang" does not mean "product of the proponents'
> >creative imagination". It means "constructed language", no more and no
> >less. It does include purely creative efforts, but it also includes
> >efforts to build a representation of something we only have indirect
> >evidence about, i.e. reconstructions (there *is* a reason why the concept
> >"construction" appears in both cases).
> 
> I see. But I uderstand the term "conlang" differently; for me, it does
> *not* mean "any model of language, no matter whether the language is
> invented by the author of the model or not".

Certainly, there is a difference between a reconstruction of a
lost ancient language, and a freely invented language.
 
> >Saying that the term "conlang" can only refer to purely imaginary products
> >is like saying the term "model train" can only refer to purely imaginary
> >train network models!

Amen!  See my example above.
 
> I disagree again. All known objects of the real world can have imagined
> equivalents, but the reverse is not true. It does not matter how real is
> the model, it's important whether it models something real. Just for the
> reason that there are tons of additional restrictions if it does.

Sure.  You can model things that never existed; we are doing it
all the time with our conlangs.  Yet, a model of a real object
is still a model and not the object itself.
 
> >> >Only as much as written French is, i.e. a consciously controlled
> >> >version
> >> 
> >> of
> >> 
> >> >a spoken language for literary use.
> >> 
> >> There is some difference: as you probably know, there has been (and
> >> there is) a lot of disagreement as for how exactly it was pronounced;
> >
> >No there isn't. We have actually a *very* good idea of how Classical Latin
> >was pronounced, both based on direct and indirect evidence. We even have a
> >good idea of the stress system, unlike that of Ancient Greek for instance.
> >The few disagreements there are are on par with the disagreement we've
> >seen on the list a few months ago about whether the _t_ in English
> >_still_ was better described as /t/ or as /d/. They are a long way from
> >the guesses we have to make as to the pronunciation of PIE!

Right.
 
> I'd agree with the last sentence, but "guesses" does not sound right in
> this context. And you seem to be too optimistic about Latin (your mention
> of "stress system" is kinda conspicuous).

How is it possible then that there is so much controversy about
details of Proto-Indo-European reconstruction?  Surely, the first
models of PIE (basically Sanskrit with sandhi and a few other
things reconstructed out) were less precise than Brugmann's model,
which was less precise than the model used today.  What regards
Classical Latin, our grammars are of course not the language itself,
and we'll never know exactly how it was pronounced, because there
are neither living speakers nor audio recordings from that time.
 
> >[...]
> >
> >> And a good deal of information comes not from texts but from grammars
> >> and lexica, i. e. *models*, most of which were composed when nobody
> >> spoke *Classical* Latin anymore.
> >
> >Actually no one ever *spoke* Classical Latin, except in reading speeches.
> >Classical Latin was always a literary version of the actual spoken Latin
> >language, for which we have *plenty* of direct evidence to choose from,
> >i.e. texts from the time it was used.
> 
> I won't agree on that either, unless a lot of reservations is added. Or
> better, just won't agree at all. But that's a completely different topic;
> OK, for simplicity's sake, let's say that by "Classical" I meant
> upper-class spoken standard (my actual stance being more like "Classical"
> being part of it, rather than a separate entity).
> 
> > [...] If you cannot see the difference between
> >
> >something attested and something unattested, and how it is orthogonal to
> >the difference between something real and something fictional, and for
> >the purposes of language reconstruction *more important* than that last
> >distinction, then we are doomed to disagree.
> 
> Amen. No sharp divide between "attested" and "reconstructed", "real vs.
> fictional" not exactly orthogonal, purposes certainly more important for
> they impose more restrictions, etc. Let's disagree :)

What is attested is of course also real; but there are many things
which are unattested but nevertheless real (unless you subscribe
to a subjectivist philosophy which claims that there is no reality
beyond our perception, but that is where any meaningful conversation
ends).  However, it is indeed difficult to know which unattested
phenomena are real and which are not.  The unattested language which
reconstructed PIE is meant to model is most likely real: the evidence
for its existence is overwhelming.  It is controversial whether the
unattested language modelled by your favourite reconstruction of
Proto-Nostratic (as I have already said, there are several) was real
or not; that is subject to heated debate among linguists (both
professional and amateur).  On the other end, we are pretty certain
that Okrand's Klingon grammar models something that never existed
outside the imagination of _Star Trek_ screenplay writers (and fans).
 
> >[...].
> >
> >> And note that there is no sharp
> >> divide between the two cases: if you don't like the one about Latin,
> >> what would you say about Old Chinese? Ancient Egyptian? Moabite?
> 
> [...]
> 
> >There *is* a sharp divide between something being attested and something
> >being unattested. Denying it is denying the distinction between direct and
> >indirect evidence. Both are valuable, but they definitely don't lead to
> >the same kind of knowledge.

Precisely.
 
> No, you don't seem to understand my examples, then. Moabite is *both*
> attested *and* reconstructed. Exaggerating a bit: we couldn't be sure that
> we read the inscriptions in the right direction, let alone understand them,
> without supposing that, and being able to check if, they're in a Semitic
> language akin to Hebrew, i. e. without a reconstruction.

There is a slim chance that the scholars are actually reading
the inscriptions in the wrong direction and with wrong sound
values of the letters, but it is so slim that nobody takes it
seriously.  I mean, how likely is it that two languages evolve
which are simple ciphers of each others?  Very close to zero.

Ray Brown certainly has something to say about such matters.
One of his interests are poorly attested and undeciphered
ancient languages.

>       The language of
> the Phaestos Disc (if it's a language) is attested, but its proposed
> interpretations are *much* worse than PIE reconstruction.

Sure.  We don't even know that it is a language; I have heard
of other interpretations.

>       "Direct"
> attestation is not always better or more important than huge amounts of
> "indirect" evidence.

Indeed not; the value of attestation depends on both quantity
and quality.  For instance, it is a time-honoured principle in
historical linguistics that the earliest available stages of a
language are the most valuable for reconstruction.  The _Rigveda_
is more valuable than the _Mahabharata_, and Homer's epics are
more valuable than the New Testament.  And in some ways, the
Mycenean inscriptions are more valuable than Homer's epics, for
instance, they attest the labiovelar stops that could otherwise
only be reconstructed for early Greek.  But the Mycenean material
is scarce, and the Linear B syllabary misses many details and
even entire segments, which relativizes the value of Mycenean.

>       And there is no "attestation" that excludes completely
> this component, of interpretation = *reconstruction of the system* behind a
> small sample of something that just cannot reflect the system "directly",
> without information losses and distortions.

True; we can never completely know an extinct language attested
in a written corpus because we cannot go and ask a native speaker.
 
> And, as I said, it's not attestation that makes a natlang different from a
> conlang. And it was the idea that reconstructions are conlangs that I
> objected to.

That is just a matter of how to define the word "conlang".
The term "re-conlang" is rather "tongue in cheek"; one should
not take it too seriously.  It is perhaps better to use two
different terms, a wider one which encompasses all models of
languages, and a narrower one which designates those models
which do not model a real language.

--
... brought to you by the Weeping Elf
http://www.joerg-rhiemeier.de/Conlang/index.html
"Bêsel asa Êm, a Êm atha cvanthal a cvanth atha Êmel." - SiM 1:1