Re: Let dogs have their day too

From: Richard Wordingham
Message: 16099
Date: 2002-10-09

--- In cybalist@..., Piotr Gasiorowski <piotr.gasiorowski@...>
wrote:
>
> ----- Original Message -----
> From: Richard Wordingham
> To: cybalist@...
> Sent: Tuesday, October 08, 2002 10:37 PM
> Subject: Re: [tied] Let dogs have their day too
>> The initial consonant is /k/ (or very similar) in 23% of the
cases. If one adds in /h/ and the affricates, for affricates often
derive from /k/, one obtains 13/31, or 12/29 if one discounts /ps/,
between 42% and 41%. In fact, of the sample, we know
that 'hond', 'sje' and 'seta' derive from words beginning with
gutturals, so words now or once beginning with gutturals make up at
least 9/31 = 29%, or 8/29 = 27% when double-counting of words derived
from PIE k^wo:n and Spanish perro is removed.

> Adding /h/ is a highly questionable idea, since /h/ can derive from
practically any other onset (e.g. *p or *s in Armenian, *s or *j in
Greek, *p or *d in Seimat, ex nihilo in many languages). We happen to
know the origin of /h/ in <hond>, but what about more exotic
languages? Affricates can't be added as a matter of fact either,
since they frequently derive from coronal stops or reinforced glides
(not to mention rarer sources).

I included /h/ because it occurred to me that whatever made /k/
popular might also make /h/ popular, rather than because I was
thinking /k/ > /h/. The affricates I encountered were voiceless, so
I do not think reinforced glides are a frequent origin. Coronal
stops are indeed a possible source. What does Miguel reckon for
Basque txakur?

> With just 31 examples, the margin of uncertainty is very large.
Each item affects the total figures by ca. 3,2%. I chose 10 more
languages more or less at random, honest to God, just trying not to
chose members of the same family twice:

Well, 9 more languages - Hmong Daw was one of my sample - and 8 new
words - Kannada (another in my sample) and Tamil both have na:y.

<snip>
> Swahili (Bantu) --------------------------------- mbwa
<snip>
> Warlpiri (Pama-Nyungan) ------------------------- maligi
> Hmong Daw (Hmong-Mien) -------------------------- dav
<snip>
> Tamil (Dravidian) ------------------------------- na:y
<snip>
> Paiwanese (Austronesian) ------------------------ vatu
> Yau (Trans-New Guinea) -------------------------- sap
> Mopán Maya -------------------------------------- pec'

> Where are all the dorsals gone? Here, the initials include 50% (!)
of labials, 30% of coronals and 20% of velars.

I could claim there was one lurking in Warlpiri 'maligi', rather like
the one lurking in 'mitakeu' of Lavukaleve (East Papuan):) Actually,
I saw no tendency in non-initial consonants. I also noticed a lot of
labial initials, but no clustering within them.

I was very disappointed by how small a sample I got. I though that
by taking one sample per family of a moderately conservative word
(near the middle of the Swadesh one hundred word list) I would get a
large number of independent samples, but even then Arawak and Carib
gave me a Spanish loanword. (The origin of Finish koira was not an
issue, for the sample contains no related words.) If I use more
languages, I will have a problem with cognates. I took a gamble on
the Creole 'family', and got 'sje' from Haitian. Perhaps if I picked
Farsi I should allow 'sag' to be added to the list, though in this
case I know that the initials are cognate. To exclude cognates could
be difficult - I do not know whether Swahili mbwa and Baka-Pigmy
bóló are cognate or not. Whatever I do will make the statistical
analysis difficult. The patchiness of the word lists is also a
potential problem - languages of the Pacific are generally well-
provided with them, but the Americas seem badly underrepresented.

Actually, I may have overlooked another derivative of PIE k^wo:n!
Thai su22nak55 is spelt <sunakh>, but Sanskrit <s> and <s'> are, so I
am told, often confused in Thai spelling. (Is this a Pali feature?)
Is there a Sanskrit or Pali word such as s'unakh- or s'unakha? The
fault may be mine in other ways; I pondered whether to override the
dictionaries and enter the native alternative maa24 when I drew the
list up. (That's an interesting word in itself - I have a vague
recollection that one of the non-Tai members of the Tai-Kadai family
(Sui?) has qma: as a cognate for maa24. The Thai word is spelt
<h><m><a:>. However, unless someone else produces a Tai-Kadai
Swadesh word list, this word is statistically invalid.)

If we dig deep enough, will we discover that most ancient words once
began with a guttural? The present-day tendency to prefix initial
resonants in PIE with laryngeals makes it look that way!

The statistical significance of the results is beginning to look
dodgy. To gauge some initial phoneme frequencies, I looked at
initial consonants in Classical Greek, discarding words beginning
with a vowel or 'h'. (I just counted pages in my dictionary.) 14%
for /k/ looked promising, but 4% for /kh/ and another 2% for /g/
start to make the results look insignificant. Putting some flesh on
these results seems highly desirable. I noticed that 31% of the
unexcluded initals were labial plosives. Is this untypically high?
(Dental plosives got 19%.) Possibly I need to think about excluding
words beginning with prepositions - they *might* be distorting the
proportions.

There were some interesting patterns, but I well remember my
statistics lecturer warning that every set of data was peculiar in
some way. Some of the strange features were:

/k/ much, much commoner than /g/. However, this may just reflect a
frequent lack of /g/ in languages.

Voiceless affricates occur, but not voiced. Again, this may just
reflect the frequent lack of their voiced correlates.

/p/ and /b/ seem equally common.

/d/ occurs, but /t/ (unaffricated) does not.

> At the very least, the claims that (1) _most_ 'dog' words begin
with /k/, and that (2) the shape /kan/ is "typical" are not supported
by such preliminary surveys.

There is an alternative claim, that most languages have a dog word
beginning with /k/. One of our readers remarked:

> My point is that some languages have more than one work for dog,
bitch, cur, hound, mongrel, puppy, pooch etc.

> With multiple words for dog there may be more instances
of /k/ words than first appears.

If we reckon on 3 independent words a language, I believe that most
languages will have a 'dog' word beginning with /k/. However,
gutturals are not rare initial consonants. For any concept with four
independent words a language, I believe we should expect most
languages to have a word beginning with a guttural.

Richard.