Re: NATLANG: Phonotactics

Perfect timing Eldin! I'm working on reanalyzing the phonology and
phonotactics of Taruven right now.

* Eldin Raigmore said on 2008-11-24 17:42:35 +0100
> (2) Around what fraction of natlangs have syllable-onsets and around what
> fraction don't?
Supposedly there is an australian language (Arrernte?) which only allows
vowels *to start a word*. Whether the remaining syllables were CV or VC
I don't recall. With that possible exception, all languages have
CV-syllables.

> ObConLang; Whose and which conlangs have what syllable-structures?
>
> AFMCL: I have (C)(C)V(V)(C)(C)
Taruven: (C)(C)V(V)(C) as well as (C)C: and C:(C). I know of no Taruven
word which is CC:C.

The possible onset-clusters are very limited, and medial clusters
(between vowels) are never longer than two consonants, so CC+C must
simplify to C+C. The long/geminate consonants complicate matters though.

> Is that all true?
Define "all true" (or as my professor in logic/philosophy used to say:
"truth" and "provability" exists only in maths/logic). Universals come
in two types:

- statistical (the rule holds for x% of languages)
- provable ("all languages have at least one vowel")

The latter are few, and with a single counter-example become the former
type... Greenberg didn't much like the former type IIRC.

> (b) How common is it for languages to have clusters that never occur except
> across a morpheme-boundary?
Very, I'd say. In all languages that allow CVC-syllables for instance.

> He also says, that _usually_, if C1C2 occurs as an onset, then _usually_
> _both_ C1 _and_ C2 occur as single-consonant onsets; and if C1C2C3 occurs
> as an onset, then _usually_ _both_ C1C2 _and_ C2C3 occur as two-consonant
> onset-clusters; and so on, and the same for codas.
>
> What, exactly, does "usually" mean here? "The overwhelming majority of
> languages with such clusters"? "The overwhelming majority of clusters in that
> language"?
I'm guessing this. Clusters of three C's or more are rare (in the set of
all languages currently existing), so easy to count and check. (Some
languages have lots of course, but they're statistical anomalies.)

> (f) Is it true that, in almost every language that has both onsets and codas,
> almost every consonant both can be an onset and can be a coda?
No. In CVC-languages, the sets of codas are usually much smaller than
the sets of onsets. Often, the codas are a true subset of the onsets.

There area lot fewer languages with a more complicated syllable than
CVC. I'm not sure the actual numbers have been tallied up.

> (f') What consonants are likeliest to be allowed in onsets but not in codas?
> What consonants are likeliest to be allowed in codas but not in onsets?
It's not either or, since codas often are more limited than onsets, it's
more "in both onset and coda" vs. "in just onset". But, English for instance,
doesn't allow /N/ {ng} to start a syllable.


t.

Reply