Re: Phone frequencies

On Sat, 6 Sep 2008 17:46:14 -0400, Logan Kearsley <chronosurfer@...>
wrote:

>I used to have an IPA table that included the frequency of each phone
>among world languages- which phones occur in 90% of all languages,
>which phones occur in 80% of languages, which phones occur in only 5%
>of languages, etc. But I seem to have lost it, and I can't find
>anything like that on line. Anybody know where I could get a table or
>a list with frequencies for different phones among world languages?
Wouldn't you know it, I was _just_ looking for the very same thing.  UPSID
(the UCLA Phonological Segment Inventory Database) does nearly exactly this,
and there's an interface to it at
  http://web.phonetik.uni-frankfurt.de/upsid.html .
Use "find certain sounds and languages that have them", option #5; it gives
you a table with frequencies of each phone in the phonologies in its
database below the output.  Not sorted, but you can do that.

Below I excerpt from an offlist message on the Glossotechnia discussion
about this.

[excerpt begin]

For consonants it's got the irritating feature that dentals and
alveolars and unspecified dental/alveolars are all counted separately,
though.  I've corrected for that by taking the unspecified counts and
multiplying those by 14/5, and discarding the other two sorts -- this
is indefensibly hacky, when I could've done the summation, but it was
quick.  That gives the following top of the frequency list (warning,
monospace table ahead):

n   .9935     g   .5610     k_h .2284     dz) .1240
t   .9436     N   .5255     p_h .2239     G   .1220
m   .9424     ?   .4789     r*  .2234     c   .1197
k   .8936     tS) .4169     v   .2106     B   .1197
l   .8445     S   .4146     x   .2084     q   .1153
j   .8381     f   .3991     4   .1613   tS)_h .1131
s   .8381     r   .3167   ts)_h .1551     b_< .1086
p   .8315     J   .3126     t_> .1490     mb) .1064
w   .7361     t_h .3041     K   .1490     ts)_> .1056
b   .6364     ts) .2794     k_> .1397     nd) .1056
h   .6186     z   .2669     Z   .1353
d   .5650     dZ) .2506     k_w .1330

and no other sounds in more than one language in ten.  r* was glossed
in the list as "voiced dental/alveolar r-sound", whatever we make of
that.

For vowels the parallel irritation is that e.g. /e/ and /E/ and
indifferent /e/~/E/ are counted separately; I've corrected (slightly
less undefensibly) by dropping the indifferents and multiplying the
others by 11/7, but special-cased /@/ and left it alone.  This gives

i   .8714     I   .1641    a_": .0754
a_" .8692     U   .1463     e:  .0732
u   .8182     1   .1353     e~  .0627
E   .6481     E~  .1219
O   .5645     O~  .1116
o   .4565     o~  .0941
e   .4320     M   .0909
a_"~.1840     i:  .0887
i~  .1818     &   .0865
@   .1685     o:  .0836
u~  .1641     u:  .0798

and no other sounds in more than one language in sixteen.  My /a_"/
was just written /a/ in the UPSID but they called it unambiguously a
low central vowel (this is a hole in the IPA more than anything).

[exceprt end]

Alex

Replies