Treatise on consonant clusters

Here's some ramblings on consonant clusters from the viewpoint of sonority.
I'll be only covering 2-consonant clusters, since longer ones can be broken
down to combinations (<str-> = <st-> + <tr->, etc.) Questions can be found
near the end.

---

In ascending order, the sonority hierarchy of consonants goes something like
this:

--Obstruents (O)--
P = voiceless plosivs
B = voiced plosivs (blosivs?)
F = voiceless fricativs
V = voiced fricativs (vricativs?)

N = nasals

--Approximants (A)--
L = liquids
W = semivowels


The theory says that clusters within a syllabe tend to have a sonority that
rises until some peak (usually a vowel) and then declines. IME the most
abundant type of clusters is O+A (or, at the coda, A+O, but I'll be stating
the types in initial form.) I can't think of much interesting to say about
them.

N+A clusters aren't rare either, but seem to prefer codas a bit more. This
is probably related to the fact that nasal codas are more common than
obstruent codas - if language only allows one type of coda consonants,
chances are they're the nasals.

O+N isn't rare either, but also seems to prefer codas; this might be because
such clusters offen act as unit "prenasalized" phonemes. English is a good
example: there's coda -mp, -mt, -md, -nt, -nd, -Nk, -Nd, -mz, -ns, -nz, -Nz,
-mT, -nT, -NT but initially only sm-, sn-; marginally also Sm-.
Initial plosiv + nasal clusters do occur too, of course, abundant in eg.
Khmer; likely various related languages, too. Within IE langs, I know of
Slavic having dn-, Greek pn-, and Germanic kn-, gn-.

P+F or B+V clusters generally occur initially in the form of affricates;
finally I've observed a bit more variation, a la "rift". P+V seems to be
exceedingly rare; I'm only aware of instances in Germanic languages where V
= /v/ deriving from former /w/. B + F is nonexistant AFAIK.

I've seen some L+A around, too (eg. "Rwanda") but it doesn't appear to be
particularily common either - except maybe with difthongs interpreted as
vowel + semivowel.

---

The supposedly rarer same-level P+P, B+B, F+F, V+V, N+N and L+L clusters all
occur in English already ("act", "ribbed", "fifth", "lives", "damn",
"pearl"); but only F+F initially ("sphere"). Greek has mn- & pt-; the name
of the German city of Gdansk (lang?) has initial B+B. I know of no initial
L+L or V+V occurances anywhere.
Double-glide A+A clusters I've heard occure in Romanian and some
Sino-Tibetan languages, but those are probably extremely rare.

Are English and other IE langs a freak exception or can these kind of
clusters really be all that common? How about in your conlangs?

---

As far as the supposedly rare counter-sonorous clusters go, Indo-European
has F+P (some branches also V+B) flying all over the place; "stags" etc.
Another common type is N+P/B, found initially in several language families,
and when homorganical, offen considered unit phonemes. N+F/V is similar but
less common.

I don't know if S+B occurs anywhere - does anyone else? It seems plausible
enuff.

Slavic seemingly has some V+F (mostly vs-), but I suspect the V is actually
syllabic here. Can anyone confirm? Actually, to be more general, does anyone
know of languages that can switch voice off and back on within a syllabe?

I think other attempted N/A+O generally decay into syllabic N/A plus O;
English is again a good example with words such as "prism" or "cattle".

Swedish seems to have a few cases of W+L; /arj/, "angry" for example; but is
this really a /j/ and not /j\/ or /I/? Likewise, did the English words
beginning with wr- ever have that phonetic structure?
Another suspect for these might be the languages that allow syllabic
liquids... can such syllabes have a glide onset? Eg /jl=/?

---

I have little idea how glottals generally fit in... my instinct would be to
group ejectivs and /?/ with P, implosivs with B, and /h/ with L or F, but is
this correct? I know some languages treat ejectivs/implosivs as if they were
/?/+plosiv... any other odd patterns?

Finally, if anyone has any odd clustering patterns in their conlangs (or
knows good ANADEWisms) - do tell.

And the obligatory AFMCL:
uwjge has only an assortment of P/B+L's plus almost all sorts of C+W's
resulting from vowel breaking. All of these are initial. There's also a
nigh-unitary F+F which can also occur finally.
Zöng has a boringly regular O/N+A initial set.
A recent new project of mine has a more varied initial selection; besides
O/N+A, there's plenty of P+N and P+F.
Another, tonal new project has initially only prenasalization and
labiovelars, but allows final N+F, V+F, A+F and A+N.
I also want to some day create a language with Georgian-level clusters such
as <sktfn->, but that's little more than an idea by now.

John Vertical

Reply