Re: The map of typological features

First of all, please excuse me the delay. Answers to some of the posts
here take some thinking and searching. Besides, as I have already said,
work is above all, so I don't always have a chance to answer on time.

mb wrote:
Darkstar wrote:
mb wrote:
Darkstar wrote:
mb wrote:

correcting for non-genetic factors (which would be the
confounding factors of this particular investigation) with a
well-designed statistical treatment.

I don't know. If we have a properly reconstructed proto-state, it's
good by defintion.

1. A reconstruction is not a fact. There _are_ no proto-states but
convenient working hypotheses.

Then the whole comparative linguistics is not a fact. (?)

2. Divergence of features among known daughters of a well-documented
ancestor do exist (otherwise they wouldn't have evolved); a fortiori
among known/unknown daughters of a hypothetical ancestor.

These divergences must also be quantified, because it is the only
available control measure for the unknown genetic relationships of the
entire group, and the unknown general non-genetic factors that
determine distances and proximities.

As caricatural examples of control sets, one would include Latin and
the other also all the Romances. Or, if you were doing a treatment of
the Balkan Sprachbund features, fact set would be the South Slavic
dialects, Albanian, Mod. Grk, Vlach, Roumanian, etc., while the control
set would add OCS, Koine and latish Latin in order to do a proper
multifactorial analysis

You probably mean to say that the typology of PIE, for example, has
never been sufficiently reconstructed to be usable in my case. Well, I
believe it has. Again, I'm using sources that are labeled as "generally
accepted" to look for typological features. I'm not supposed to do my
own typological reconstructions of Latin, Turkic languages, or
whatever. It's all been done long before.

I still don't see how late Latin can be a control set for Romance
languages, since late Latin is just what we're reconstructing in this
case. What kind of control sets were you referring to?

You are working with the hypothesis that there are genetic and
non-genetic factors at play which may or may not have resulted in a
given geographic distribution.

So far, I don't speculate as to which factors may be at play. As for
now, I'm just trying to prove the existence of the geographical
distribution itself.
....
The result is a circular argument. All the work will be dismissible in
principle, as begging the question.

I still don't see how my argumentation is circular.

Because introducing any hypothesis like "IE features" instead of each
established language by itself posits at the start what you want to
prove by analysis.

"IE features" are not a hypothesis, they are real and well-established
facts.
Apparently, what you also might mean is that there's a risk of losing
some important typological parameters that contradict my thesis just
because I'm biased in favor of picking the parameters I like. But!
Since the parameters under consideration concern the most BASIC
linguistic features that are at the core of the language STRUCTURE
including the most important features of phonetics, prosody, morphology
and syntax (and this is why I called them 'structural' initially, not
just 'typological' which implies an arbitrary choice), I would hardly
lose anything. It's like describing a home by its roof, walls and
windows. If I miss a chair, that's not so important. Any other features
are just not of the same relevance and significance to the typology and
language structure. Consequently, my parameters must have the highest
contribution to an overall lineal, parametric function.

What I'm trying to say, it's rather heuristics than head-on statistics.
But that's a separate issue.

Can't see why it's separate. Heuristics, if used as a term for a group
of statistically-based methods, would have to work up from attested
facts (one set with oldest *attested and well-known* languages, and one
set with all daughter languages), not backward from faith-based theory.

Because methods of the Urheimat or the proto-state recinstruction are a
big separate issue. I mention them only in a few special cases.

You don't even need to mention any of that if you only feed established
fact and proper measurements to your analysis.

Okay, in some cases I do have to introduce additional comment,
including educated guesses and the like. But they are NOT supposed to
play a leading role in the research.

.

Relevant Pages

  • Re: Long-range structural similarities across the Old World.
    ... useless for determining genetic relationships among languages. ... of linguistic typology; the first editor is Martin Haspelmath. ... features are generally found in just a small part of the languages of a ... I have no idea what either clicks or incorporation have to do with the ...
    (sci.lang)
  • Re: Long-range structural similarities across the Old World.
    ... useless for determining genetic relationships among languages. ... of linguistic typology; the first editor is Martin Haspelmath. ... features are generally found in just a small part of the languages of a ...
    (sci.lang)
  • Re: The map of typological features
    ... No, I don't know of these languages, though I don't ... features, and which features should be split into multiple different ... Maybe you're just forgetting about the extensive use of reduplications ... not being a tone language). ...
    (sci.lang)
  • Re: Why not add namespace feature into standard C?
    ... features, I can't understand why not add this feature into standard C. ... macro in a namespace safe way. ... because other languages are solving real problems in programming that C ...
    (comp.lang.c)
  • Re: Long-range structural similarities across the Old World.
    ... You're confusing areal and typological features with relatedness again. ... Numerous IE and Dravidian languages, Burushaski, Ket. ... Most SE Aisia is SVO, for example, Munda seems to have acquired SOV ...
    (sci.lang)