Re: The map of typological features

Nathan Sanders wrote:
In article <1148951639.177783.87980@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
"Darkstar" <darkstar100@xxxxxxxx> wrote:

Nathan Sanders wrote:
In article <1148741436.842961.72040@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
"Darkstar" <darkstar100@xxxxxxxx> wrote:

"Ergative/acusative phenomena".
These are language groups that explicitly mark ergative or accusative
(nominative) case or differentiate them in any given way. For instance,
Chinese has no case-marking, so it's neither ergative nor accusative.
While Georgian is both.

Isn't this just a subset of your "well-developed cases" category? Or
have you come across languages with "well-developed cases", but no
ergative or accusatives?

Exactly. The idea was that some analytic languages may still use
rudimentary cases. Polynesian and Tibeto-Burman do, for instance.

Since when is "rudimentary" equivalent to "well-developed"?

It doesn't seem possible for a language to have well-developed cases
and *not* also have some case that corresponds to either accusative or
ergative.

Polynesian languages are pretty analyitc so they use pre- and
postpositional complexes which (for nouns) include articles,
directional words, and other markers.

Pre- and postpositions are not the same thing as case.

Consequently, east Polynesian are
accusative, whereas Tonga and some Samoan are ergative.

So what are examples of languages with well-developed cases that do
*not* have accusative or ergative case?

Excuse me, there was a lot of typing for me, I didn't read you well
from the start. No, I don't know of these languages, though I don't
have enough information on some (e.g. Yukaghir). In theory, some
languges might use only locative/directional cases, but omit relational
cases. Dunno.

How about voiceless-voiced-prenasalized systems, as in many African
languages? Surely this is a "complex" plosive system.

Or Austronesian voiceless-(semi)voiced-prenasalized. Nasals are
sonorants formed in the front part of the mouth cavity. While the third
row should probably be a back consonant - laryngeal, aspirated, etc.

You've lumped a 3+ phonation contrast for plosives with emphatic
(pharyngealized) consonants and with ejectives, both of which are
completely separate phenomena, but not with prenasalization or
implosives. Why? Simply because it makes your groupings work out
nicer?

Exactly. Conducting a statistical research, you do not necessarily have
to provide a causal mechanism.
Nasals are sonorants, while I'm clueless as to why Type B tends to use
implosives, as well. Probably because they sound *less plosive*, that's
why. Type B just dislikes well-formed, noisy explosives. Many Type A
consonants (pharyngeals, aspirated, geminates) are just clear-cut
plosives that require certain energy. Maybe the energy of the sound
wave and the distinctiveness play a role here.
But I do not *have* to account for that, do I?

A more objective approach would be to look at each of these properties
separately.


If they're independent, then I should put them into separate lines,
okay.

Other signs of this complexity should also be present (consonant
clusters, some affricates), syllables must be closed, vocalism should
not be particularly developed.

So this feature only counts when combined with other features?! No
wonder you get the groupings you want: "Feature X only counts when
feature Y is present, and lo and behold, languages with feature X
happen to have feature Y, too!"

I'm just looking for a stack of phonological features. There's nothing
wrong with that.

The idea is that Type A sounds something
like phethshkhtenchdkghtdchtegkpbh (Arabic, Caucasian, PIE), whereas
Type B like nanangimororrolakalaumoluronge (almost all African,
Oceanic, Japanese, Finnish). That's not easy to explain, but you know
it when you hear it.

Real comparative and typological linguistics don't rely on "knowing it
when you hear it". They use data in the form of regular,
well-defined, verifiable patterns.

Well, as a real linguist, you might help me identify the patterns of
the first and the second series more precisely. :-)

* What does it means for nasals to be "widespread"? Again, is this
widespread within the set of phonemes, within possible words, actual
words, or by frequency? And what is the numerical requirement for
being "widespread"?

By the frequency within an average spoken text. To get an exact number,
I'd have to find weights for particular phonemes in a given typical
speech sample.

So you haven't done this yet for the languages you claim do and do not
have "widespread nasals"?

It's all described in books, encyclopedias and other sources. I don't
have to do my own field research or go to Micronesia to find out how
they talk over there. Any kind of work is based on other people's work.
In this particular example, some encyclopedia article cited nasals to
be "widespread" or something to that extent and I took this opinion to
be "generally accepted". In some cases, I did go to Micronesia, though,
by downloading sound files to check certain languages. Anyway, there
are no standard algorithms in linguistics that reduce fuzzy sets to a
number. Linguistics is still one of the humanities. And even if I start
inventing my own algorithms for counting phonemes, you'll probably say
I'm wrong again. So if a respectable source says "widespread", I take
that for granted.

This happens automatically in your mind when you hear a
foreign language for the first time (or a child does so when it's
born)

Humans are notoriously bad at distinguishing real patterns from
perceived patterns, and at determining statistical significance.

But in most cases, it's much easier than that. Apparently, there were
no nasals in PIE (except /m/ and /n/), there are very few in Mongolic.
Whereas there are many nasalized sonorants in some African and
Austronesian languages.

And yet many IE languages developed more nasals (Romance, Slavic,
Hindi, etc.)... why is it not possible that the African and
Austronesian languages also similarly developed their numerous nasals
from original m and n?

So maybe they did. But not all at the same time. Again, I don't not
claim any kind of genetic relation, so far. I'm just looking for
long-range patterns.

* How often must reduplication occur be to be "frequent"? Are you
counting by number of possible grammatical uses, or actual frequency
in conversation? What's the numerical cutoff point?

They should play a significant role in word formation or grammar. For

You've replaced "frequent" with "significant", and still haven't given
a proper, well-defined delineation between languages that would have
this feature and those that wouldn't.

Well, I'll have to repeat: if reduplications are frequently used in
grammar and word formation, they belong to Type B.
Give me at least some examples from any IE, FU or Alt languages of the
north that extensively use reduplications to replace grammatical
categories (plural, pronouns, etc) in formal speech. Maybe I'll desist,
then.

How would another researcher be
able to independently verify your conclusions by applying your
features to languages, if you don't provide something concrete,
objective, and repeatable? Without rigor, what you've created is a
pretty picture of your opinions, not a scientific theory.

Okay, okay, I agree that my picture lacks sufficient rigor.

instance, in IE languages there are special words to say "every".
But in Bamana, you simply say "don o don" (every day), "den o
den" (every child). In this case, reduplications infiltrated into the
pronominal paradigm. In IE languages, you also might occasionally say
"day after day" which is much more rare than "for many days"
but you don't say "child after child". In Austronesian they form
plurals, etc.

In English, reduplication is frequently used for emphasis. It really
really is. Does this count?

No, in formal grammar.

Anyway, you and mb have made me aware of the fuzziness and hazziness of
my superficial research - maybe I can find a way to make it more
rigorous.

That should be your next step, yes. An informal description of your
observations and opinions is one thing, but trying to present it as a
coherent theory requires rigor that is completely missing from your
current map.

I just hope that someone would pick up serious holes before I procede
with the formalization. I just think if there are any serious errors,
people would see them "just like that".

* What kind of "complex verb morphology" doesn't cover the complex
verb morphology of African languages?

Gerundial forms with many affixes as in Altaic. The African languages I
am familiar with are largely analytic and do not use sufficiently
complex verb morphology. Maybe I'm underestimating something.

So it only counts if its inflectional/fusional? Why shouldn't complex
analytic verbal morphology count as complex verbal morphology? Again,
you are requiring two features to be present in order to assign one of
them to a language, artificially increasing the amount of feature
co-occurrence.


I'd be happy if someone could find at least some counter-examples
(analythic, synthetic, whatever). My sources cited "complex verbal
morphology" for Saharan, so I put it honestly on the map.
As for the others: Benue-Kongo -- "participles are not characteristic".
Gur, Kwa - may have a complex system, Mande -- no, primitive
morphology, Atlantic Congo -- do have something. Altogether, maybe I
should correct this point. I'll look up more. Thanks.

Although, the original idea was to look for something analogous to the
Altaic gerunds, Afro-Asiatic base stems or IE tenses. These were
thought to be good examples of the "complex verb morphology".

You need to define your features independently from each other.
Otherwise, you're essentially building a classification system
designed from the outset to achieve the results you want it to
achieve, rather than measuring a reality that may be messier than
you'd like.

Okay, but I don't know what results I want to achieve---
If these features are independent, they may indicate genetic
relationship. If they are dependent they prove that all language groups
of Type A/B form an unexplicable long-range geographic pattern across
several continents. What you're probably forgetting is that *the
language groups* are also thought to be *independent* from the start
(!). For this reason, even if I use sets of partially dependent
features in typology, I can still expect a purely statistical,
non-patternized A/B distribution across all of the territory. Yet, (of
course, if there are no typological errors involved), A and B seem to
line up well forming transcontinental patterns.

In fact, we're considering a matrix of partially dependent groups and
partially dependent typological features---

* More broadly, languages can evolve from having a feature to not
having it (e.g., well-developed cases and three-way plosives from PIE
to English), or from not having it to having it (e.g., tone in
Chinese, Ancient Greek, Proto-Slavic, etc.).

No, I'm not buying in this anymore. These are all bad examples! English
and French are unique in many ways. They're basically semi-creoles

There is no such thing as a semi-creole. There are regular languages,
pidgins, and creoles. English is a regular language, not a creole.

Creoles by definition arise from pidgins, and English was never a
pidgin.

Don't give me
creoles as examples of anything - they are just melting pots with
very complex linguistic histories.

That is not how creoles are defined.

The natural interaction that occurs between languages in contact is a
completely different phenomenon from creolization, which requires
particular social/cultural interactions, as well as a distinct pidgin
phase.

Never mind. It's a separate discussion.

Chinese is a border-line case
influenced by both Type A and B.

You can't just pick and choose which languages you want to exclude
because of some perceived over-abundance of language contact!
Especially when we don't have full historical records for every
language. Some features you want to dismiss may in fact be native,
and others that you want to accept may in fact be borrowed.

So what other example of unexplicable major typological changes do we
have, BTW?

Ancient Greek and Proto-Slavic could
have been tonal from the start, because they just *might* have been
non-some phonological tones or musical intonation in PIE.

Either PIE was a tone langauge, or it wasn't (you've classified it as
not being a tone language). Since Ancient Greek and Proto-Slavic were
tonal, then by your classification, they must have changed from
non-tonal PIE to tonal, ergo, the feature of tone is not stable (and
this is already known).

*Non-phonological* is the key word here. IE languages have never been
truely tonal. Their tones are rudimentary and do not play a significant
role within a phrase. In Latvian no one will probably say a thing if
you speak evenly. However, in Moore (Gur), a friend of mine was
constantly correcting me, because I made no distinction between the low
and high tones. In Mandarin there are thousands of minimal pairs, the
whole dictionary is just one big minimal pair, while in Serbo-Croatian
people were able rake up only 2 or 5.

Even ignoring Proto-Slavic and Ancient Greek (which would be silly),
Swedish and Norwegian show that tone can develop from a non-tonal
language.

Also, from the history of Slavic and Greek, we know that tone need not
persist, either. It's an accident of history that Chinese still has
its tones, while Greek and most Slavic languages do not.

The point here is that languages don't change just like that, there
must be a number of good reasons for it.

But languages do change just like that! Have you ever taken a course
in historical linguistics? Languages can change in any of the ways in
which they differ from each other: phonetics, phonology, morphology,
syntax, semantics, etc. Every aspect that makes two languages
different is vulnerable to language change. Where there were once
different, they may evolve to be the same, and vice versa.

First of all, let's not make any abstract arguments. Do or do not, have
or have not -- it's all just the same'o'sam'o. Give me something
specific and concrete, then I'll be able to give you something specific
and concrete in return. For example, the part with the African verb
morphology may turn out to be pretty good.

What examples of unexplicable drastic large-scale changes do we have?

And! I'm not saying things can't change -- everything changes, nothing
stays the same, that's just a trivial fact of life. My thesis was
probably that *the FRAMEWORK of partially interdependent typological
features might recieve high additional stability.*

Since typological features can arise and disappear naturally over
time, how far back do you go? For example, many researchers believe
that Chinese tones are derived from consonants, so depending on how
far back you go, Chinese either is or isn't a tone language (I also
think there are arguments that Chinese once had a four-way plosive
distinction...).

I do not specifically set up a goal of going back to the Jurassic
Period.

And indeed, it doesn't seem you have not set up a goal of going back
to any consistent time period at all. The (proto-)languages you are
comparing date back to different times. In order to accurately
measure any type of systematic shared typology, you need to compare
languages at the same point in time.

It would like comparing the cultures of 1980s USA, 1990s Mexico, and
1780s Canada, and concluding that the USA and Mexico are more cultural
similar than either is to Canada.

That's rather interesting. But I don't think this interferes logically
with my thesis directly. I still have the distribution pattern (if we
suppose for a moment that I have it). It's still there, even if it now
goes back to different times. It hasn't gone, just because it became
diachronic. You still get that "unique, unexplicable coincidence", even
though it's scattered through different periods of time. Or what do you
think? --

Besides, I don't think the periods are altogether so different.
Generally, the map is supposed to reflect a one-time distribution c.
2000-6000 BC.
And if my logical opponents told me, we don't know where, say, Basque
was at the time, I would answer that there's no evidence it was in
central Africa. In other words, there's no evidence that any of the
language groups ever crossed the line during this period.
Anyway, you've mentioned a good point -- firstly, the exact location of
Urheimats is not always well-known. Although again, we suppose that
Type B languages have always been somewhere in the south, and Type A in
the north, and it's all we have to know, for all practical purposes.
Secondly, as to the temporal distinctions, what evidence do we have
that, say, Niger-Congo was non-tonal and inflected at the time?
Probably none, right? And even if it *were*, then again we get the
aforementioned "unique coincidence" across the contintents, since all
groups just happen to be in the right state at *certain* (even though
different) periods of time. This reasoning leads to believe the
distribution is still non-coincidential even it's partly diachronic. I
hope you get what I mean.

Again, the last Ice Age was only c 15-12000 BC ago, that's not too
far away.

It's about twice as far as we can reliably reconstruct most
proto-languages (especially for languages lacking the large written
record of IE).

(Not to mention the fact that, without outside evidence, we have
absolutely no idea just how far back our reconstruction goes, since
languages change at different rates.)

It might help to look further into the past beyond 8000 BC,
although that's just a humble hope, and it's too early to speak
about that now.

I don't think it's too early to say that it won't provide anything
scientifically sound.

Well, I don't have all the necessary resources for it, consider it just
a game.

Without more actual data, we simply cannot go
back any further. And since you're excluding data, your results are
going to be skewed.

Maybe I could elaborate a little more. Although, I still believe if
there are any mistakes, there's already enough data to spot them now.

Nathan

--
Nathan Sanders
Linguistics Program
Williams College
http://wso.williams.edu/~nsanders/

.

Relevant Pages

  • Re: A beautiful Gaelic song by Wiliam Ross
    ... languages which have reached the outer limit of linguistic ... When speaking English the only rule is that to be understood, ... The Chinese languages, of which there are many, with variations spoken ... Get the tone wrong and you ...
    (soc.culture.scottish)
  • Re: The map of typological features
    ... No, I don't know of these languages, though I don't ... So this feature only counts when combined with other features?! ... Only for those who don't know what linguistics is. ... not being a tone language). ...
    (sci.lang)
  • Re: SQL
    ... That is a language-specific feature. ... we are talking about features of languages. ... Relational only cares that the "expression engine" follow a minimum set ... linear programming and other optimization problems, ...
    (comp.object)
  • Re: Payngo = Fist & Austronesian Numerals
    ... Cruz Reef languages that some call East Papuan, ... as cognate with *qiNaS? ... the feature which occurs between them. ... whole vast Austronesian area as a dialect continuum is a different ...
    (sci.lang)
  • Re: What does the "--" mean when placed before a function/formula?
    ... Excel, and usually would not work in other languages. ... I think it is unique feature of Excel that ... You may find the following two implementations of computer math interesting. ...
    (microsoft.public.excel.worksheet.functions)