LINGUIST List 14.1456

Wed May 21 2003

Sum: Languages with fricative-initial onset clusters

Editor for this issue: Karen Milligan <>


  1. Lisa Davidson, Languages with fricative-initial onset clusters

Message 1: Languages with fricative-initial onset clusters

Date: Sun, 18 May 2003 10:37:54 +0000
From: Lisa Davidson <>
Subject: Languages with fricative-initial onset clusters

A while back, I asked for information regarding fricative-initial,
word-initial consonant clusters (Linguist 13.3161). I got a number of
responses, and only now have I gotten around to putting them together
to post on LinguistList. At times, the ASCII of my email may prove
problematic, so I suggest contacting the original authors if you're
not sure what was meant. I have also collected more information from
other sources on Slavic languages, so if you are interested in this,
please let me know.

The text of my original message was:

	I'm trying to compile information on languages that allow
	two-member fricative-initial consonant clusters in
	word-initial position. Specifically, I'm interested in cases
	where the fricatives are something other than /s/ that combine 
	with nasals and obstruents.

Thanks to everyone who sent a message''they were very helpful!

Lisa Davidson

	From: Sylvia Moosm�ller

Albanian allows fS among others: e.g. ''fshati''. 

 From: Larry Trask

Basque permits the word-initial clusters /fl-/ and /fr-/. But I
stress that these clusters occur *only* in loan words from Romance,
and never in native words.

Some examples:

<flamenko> 'flamenco' <flauta> 'flute' <flota> 'fleet' <flotatu> '(to)

<fraide> 'friar' <frakak> 'trousers' <franko> ~ <frango> 'many, a lot
of' <frenu> 'brake' <fresku> 'fresh' <frijitu> '(to) fry' <froga>
'proof' <fruitu> 'fruit'

The Pyrenean dialects go a little further. First, I must explain that
Basque has two contrasting voiceless alveolar fricatives, a laminal,
notated <z>, resembling English /s/, and an apical, notated <s>,
resembling Castilian Spanish /s/ in the north of Spain. Most dialects
of Basque permit no initial clusters at all of the form <zC-> or
<sC->. But the Pyrenean varieties have undergone extensive syncope,
mostly in trisyllables but occasionally also in disyllables. This
syncope can bring about initial clusters which do not otherwise exist.
For example, the common finite verb-form <zara> 'you are' (singular)
appears in some Pyrenean varieties as <zra>, with a unique initial


	From: Daniel Collins

You'll probably hear from many Slavic specialists, as this is a common
type of cluster. For example, in Russian, where there are fricatives
f f' s s' z z' S Z x (plus fricativoid glides v v'),

xvala 'glory' xleb 'bread' xmuryj 'gloomy' xnykat' 'whine' xren
'horseradish' xto 'who (dialectal)'

zvuk 'sound' zdes' 'here' zloj 'evil' zmeja 'snake' znak 'sign' zrja
'in vain'

Zban 'jug' (voiced alveopalatal) Zvac^�ka 'cud' Zgu 'I burn'
Zdu 'I wait' Zmu 'I press' Znu 'I thresh' Zru 'I gobble'

Svabra 'mop' Sljapa 'hat' Smel' 'bumble-bee' Snur 'hose' Spaga 'sword'
Stat 'state' ScSi 'cabbage soup'

There are prefixes s- and v-, which following regressive devoicing
rules become z- and f-; moreover, following regressive palatality
assimilation rules, s- can become z^�- and s^�-. Thus
virtually every conceivable cluster can occur.

There was prehistorically a constraint on fricative + fricative
clusters (see H. Andersen, ''Lenition,'' Language 1973), but such
clusters now occur due to early medieval losses of schwas (jers): sxod
'descent,' vsjakij 'every' [fs-], vxod [fx] 'entry,' etc.

There are additional kinds of fricative clusters in Slavic languages
like Czech that have an additional voiced fricative [h] (from *g):

hve^�zda 'star' hlava 'head' hmyz 'bug' hn�zdo 'nest'
hra 'play'

 From: Michael Tjalve

In most European languages, the initial cluster /sf/ is kept in words
of Greek origin even if it's not found elsewhere, e.g. 'Sphinx' and

In Danish, you find a few words beginning with /fn/, e.g. 'fnadder'
(slush), 'fnat' (scabies), 'fnug' (speck of dust), 'fnyse' (snort),
and 'fnise' (giggle). As you can see, most of them give a somewhat
'unpleasant' association.

	From: Marc Picard 

Canadian French has all kinds of clusters of this type, mainly as a
result of the deletion of shwa in the first syllable of forms like
fen�tre 'window', venir 'to come', chenille 'caterpillar',
chemin 'road', etc., or with je 'I' and se 'oneself' in combination
with a following verb, e.g., je pense, je fais, je sais, je chante, or
se dire, se faire, se montrer, se gu�rir, etc.

	From: Michael Becker 

In Hebrew you have word-initial clusters that begin with s, z or
sh. Any non-guttural consonant can follow. Examples: stuma snai sgura
zkena zman shtut shmura f and v are missing from this position for
historical reasons (spirantization), and x and h initial clusters are
broken by epenthesis.

	From: Elizabeth J. Pyatt 

This may be a long shot from Welsh.

Welsh has /kn/ clusters, but after certain forms which cause the
''Aspirate Mutation'' (or ''Spirant Mutation''), the /kn/ clusters
become /xn/ clusters.

cnofil /knovil/ 'rodent' tri chnofil /tri xnovil/ '3 rodents'

The mutation is morphologically conditioned, not phonologically

ei chnovil /i xnovil/ 'her rodent' vs. ei gnofil /i gnofil/ 'his rodent'

I'm not a native so you may want to double-check, but I think that's
what would happen. Gareth King (1993) is a pretty good basic source,
and some of the books on Celtic languages, like the one by Martin J
Ball may have more.

	From: Theo Vennemann 

Modern Greek has f- and x-clusters. Older Germanic had h-cluster, and
Modern Icelandic still does. Older Germanic including Old English had
fn- but lost it; Proto-Germanic *fneusan 'to sneeze' became _niesen_
in German and _sneeze_ in English, the stronger s- here substituting
for the weaker f- in order to preserve the onomatopoeic effect. (On
the latter point, read Angelika Lutz's book on English phonotactics
and her articles on onset clusters, h, etc.)

	From: Bart Mathias 

This shouldn't really count, because /phonologically/ Japanese allows
no consonant clusters at all. But /phonetically/, Japanese allows two
allophones of /h/, a bilabial voiceless fricative (I don't know the
ASCII IPA; call it [F]) and a palatal voiceless fricative ([�])
before [t], [k], [p] (rare for certain reasons), [s]/[S], probably

The former occurs in words like /hutari/, /hukai/, /husuma/, /huhyo:/,
where the /u/ could be a high central unrounded voiceless vowel in
slow and/or careful speech, but vanishes in normal or fast speech
(theoretically lengthening the [F] in this mora-counting language).

The latter in words like /hito/, /hiku/, /hipocoNderi:/, /hisui/,
/hihaN/, where the /i/ is similarly swallowed into the [�(:)].

	From: Elisa Steinberg 

Try with Yiddish. As an example, sh ch a v e y : 'sorrel'. In
transliteration, sh followed by ch

	From: Torodd Kinn 

All varieties of Norwegian have at least four types of word-initial
consonant clusters beginning with /f/:

/fn/ as in ''fnyse'' (verb: 'snort') /fl/ as in ''flat'' (adjective:
'flat, plain') /fj/ as in ''fjern'' (adjective: 'distant') /fr/ as in
''fri'' (adjective: 'free') - /r/ is usually either an alveolar tap or
a uvular approximant or fricative (depending primarily on the dialect
of the speaker)

Norwegian also has initial /vr/ as in ''vri'' (verb: 'twist, wring') -
but /v/ is usually an approximant

and /sp/, /st/, /sk/, /sm/, /sn/, /sv/, /spl/, /spj/, /spr/, /stj/,
/str/, /skv/, /skl/, /skr/

Some dialects have more types of clusters.

	From: Madalena Cruz-Ferreira 

About your query to Linguist, in my variety of Portuguese (European,
Lisbon dialect) several of these clusters are allowed, like these
word-initial ones [vZ] in vegetal, vegetable [Zv] in esvaziar, to
empty [Sf] in esfera, sphere [dSpZ] in despejar, to throw out (the
word is reduced from 3 syllables to one) [dSt] in destilar, to
distillate [Zn] in gene'rico, generic [Zm] in esmeralda, emerald etc.
The clusters are due to vowel deletion, and their phonological
analysis is quite controversial, but this is what happens in real

	From: Dino Surendran

The site is a
good source of information on several Bantu languages. Lots of

For example, Shona has words starting with


is an online dictionary - albeit ones that only goes to
M. Shona has words beginning with zv- as well, but this is often
pronounced as the fricative /zh/ so isnt really a consonant
cluster. Check with your local Shona speaker though :)

In Swahili, there are consonant clusters
fy and fl.

	From: Michael R. Marlo

I was recently (just today!) looking at some data from Sipakapense
Maya that seem to be relevant for your project. Rusty Barrett
(, a professor here at the University of Michigan,
wrote a grammar of this language as his dissertation. It's
unpublished, but I'm sure he'd be happy to send you a PDF of the
relevant section if you want. The following Sipakapense words are in
SIL Doulos93 font, and there is some underlining, too. Hopefully your
email system can handle the Rich Text (HTML) format, and with any
luck, mine won't garble it up in the process of sending. (NOTE: My
email system did indeed have a problem, so if youre
interested in this language, please email the author.)

	From: Lameen Souag 

North African Arabic and northern Berber languages both allow any
possible two-consonant cluster at the beginning of a word, eg Algerian

xbeT 'he banged' 
xser 'he lost' 
fTen 'he woke up' 
{th}nin 'two (in compounds) 
xneq 'strangle' 
fna 'annihilation' (cf. the famous Moroccan Djemaa el-Fna) 
Hbel 'rope' 
hbel 'he went crazy' 
fqih 'religious scholar' 
Hmama 'pigeon'

But this may only be skin-deep; there is good reason to believe that
the first consonant of such clusters is in fact a floating
syllable-final (analogous to the ones at the end of words in French,
except that there is no deletion rule.) Specifically, a sequence
/C#CC/ will normally be realized as /.CeC.C/.

	From: Yehuda N. Falk 

Modern Hebrew allows such clusters. Other than the loan words, they
derive from an earlier form with some sort of schwa between the
consonants, but most speakers of modern Hebrew do not have an
intervening vowel. (S = voiceless alveolar fricative)

Modern Hebrew has the following fricatives: f v s z S Z 

Xbut /f, v, Z/ have limited distribution for historical reasons (/f,v/
were originally spirantized allophones of /p, b/, and /Z/ is only in
loanwords) and /x/ rarely clusters (because it used to be pharyngeal
and almost always is followed by /a/ when it's word-initial). I don't
think there's anything other than history keeping these from
clustering, but by far the best examples have /s, z, S/. And I don't
see any limits as to what they can cluster with.

Here are some examples:

zvuv 'fly' zvulun (a personal name) 
zxuxit (or zgugit) 'glass' 
zxut 'privilege' 
zminut 'availability'
znavot 'tails' (plural of zanav) 
zkifut 'erectness, uprightness' 
xlor 'chlorine' [I know /l/ isn't an obstruent or nasal :) ] 
Svil 'path' 
Svua 'oath' 
Svita 'strike' 
Sgia 'mistake' 
Smuel (a personal name) 
Smone 'eight' 
Smua 'rumor' 
Snayim 'two' 
Snicel 'schnitzel' 
Snaton 'university course catalog' 
Snati 'annual' 
Skia 'sunset' 
Skalim 'shekels' 
Stil 'sapling' 
Stiya 'drink' 
Stika 'silence'

	From: Georgios Tserdanelis 

Modern Greek is a language that allows many fricative/obstruent
clusters word initially. e.g.

xn- fth- xth- thn- fkj- ft- xt- xl-

/x/ being a velar fricative and /th/ an interdental fricative

I have written a paper on greek clusters and dissimilation where you can find more info about the language and its sounds. it's published in the working papers series here at OSU but you can download a word version of it from my website too.

	From: Susannah Davis 

Albanian has many fricative-initial consonant clusters, including
fricatives followed by nasals and obstruents.

/sh/ and /zh/ and the affricate /ts/ are found in lots and lots of
word-initial clusters in Albanian. /v/, /z/, and /s/ are also very
common in this position. 

	From:Nathan Sanders

German is one obvious case, where not only do they have esh+C clusters, but historically, these replaced s+C clusters.

Polish (and other Slavic languages) allows a wide variety of initial
fric+C clusters, including where the fricative is labiodental [f/v],
velar [x], post-alveolar [S/Z] (something between esh/yogh and
retroflex s/z), alveolo-palatal [s'/z'], and alveolar [s/z]. Some

ftem 'suddenly'
flet 'flute' 
vdova 'widow' 
vZask 'scream' 
vlets 'to drag' 
vnuk 'grandson' 
vwadza 'power' 
xts'e 'want (3sg, pres)' 
xSan 'hose-radish' 
xleb 'bread' 
xmura 'cloud' 
xwop 'peasant'
 Skoda 'damage' 
Sfed 'Swede' 
Slak 'route' 
Smer 'murmur'
Zvir 'gravel' 
Zmija 'viper' 
Zwobek 'nursery' 
s'pev 'song' 
s'lub 'wedding'
s'migwo 'propeller' 
s'fit 'dawn' 
z'dz'bwo 'stalk, blade'
 z'le 'badly'

Georgian is known for its outrageous clusters, so I'm sure it has good
clusters to look at.

The Semitic languages probably have some due to their morphology (root
consonants like [x] and [f] ending up next to other root consonants in
certain morphological templates, beginning with CC-).

I seem to remember that K'ichee' (Quiche) has [S]+C clusters, but I'm
not sure about that.

A quick look through a modern Greek dictionary shows a few possible
interesting clusters: vGazo 'to take off' ([G] = velar fric), Tlipsi
'grief' ([T] = interdental fric), fTonos 'envy', fterna 'heel of
foot', xTes 'yesterday', xlomos 'pale', xteni 'comb'. I'm not sure if
Greek phonology simplifies these or not.

	From: Paul Boersma 

well, coronal-peripheral is the usual generalization that you'll see
about Dutch phonology, i.e. allowing sp- and -pt. But if we disregard
s, the correct generalization is actually coronal-last. So you find
the same articulator sequence in pn- as in -mt, this observation has
occurred a couple of times in my papers. Initial fricative-initial
clusters are alright, therefore, as long as the second consonant is
coronal. We have the words fnuiken [f n oe y k schwa] and gnuiven [x n
oe y v schwa] (in an 1989 paper of mine it came out as crucial that
this x is voiceless even for speakers who allow a voiced fricative <g>
if followed by a vowel).

And of course you know that Greek allows fT- and xT-, again
coronal-last (well, there's one tm- word, I think). Yes, I would like
to see whether this coronal-last observation holds for non-IE
languages as well. It'll be probably be one of those implicational
universals, i.e. if xm- is allowed then so is xn- etcetera.

	From: R�my Viredaz

(1) In Modern Greek, in addition to s + various consonants, you have
ht- and ft-.

(2) In Slavic languages, various interesting initial groups.

(3) Georgian, in spite of its so many complex consonant groups, seems
not to have initial x + stop but only x + r, s, sh ; however, I have
no adequate information on Georgian at hand just now.

(4) In Rhaeto-Romance, you have not only frequent clusters of sh or zh
(spelt s; voice is distinctive before sonorant) + practically any
consonant (either from Latin s (or ex-, or dis-) + consonant, or,
rarely, from sch of various origins + schwa from unaccented e or i +
consonant), but also, at least in the Upper Engadine variety,
occasional groups of s or z (spelt s) and f or v (both spelt v for
etymological reasons) + various consonants; in addition, there is
initial ft in one village name in Lower Engadine. Skewness and gaps in
distribution are partly due to chance.

To have a better view of the distribution of these clusters, you can
use the following dictionaries:

Taggart, Gilbert. - Dicziunari rumantsch ladin-fran�ais,
Chur/Cuoira (Lia rumantscha), 1st ed. 1990. - Only Lower Engadine, but
with pronunciation indicated throughout.

Peer, Oscar, Dicziunari rumantsch ladin - tudais-sch, Lia Rumantscha,
1st ed. 1962. - Both Lower and Upper Engadine, the reference language
being German; with an introduction (in German) on pronunciation; voice
and stress indicated by a subscribed dot in the dictionary when not
deductible from the spelling.

Collective, Handw�rterbuch des R�toromanischen, 3 vol.,
Z�rich (Offizin) 1994. - A Romansh-German dictionary, including
all main varieties of Swiss Rhaeto-Romance, with brief etymological
notes for each entry-word.

All these books can easily be found at or through the Romansh League,
you can contact them at

	From: Mark Donohue 

Puare is just what youre looking for.

In summary, though Puare allows not only garden variety spV clusters,
etc., but fun things like lku :egg:, and dla :house:, wro :tree:. It:s
quite a fun language to speak. And the l is retroflex.

OK. The phonemes of the language are:

pw kw p t 
k mbw nggw mb nd 
ngg mw m n 
L y* 
r w sw 
s H

The L is a strongly (Dravidian or Australian-style) retroflex lateral,
and the y (their orthographic choice) is a palatalised fricative,
which is basically a coarticulated voiced dental fricative/lateral
fricative. It's quite a sound to hear, and say.

The clusters of the sorts you're asking about in Puare are:

sp sk

sm sn

yp ypw yk


Note that y cannot occur with a nasal, and that none of them occur
with a t, which is exceedingly rare in the language anyway
(historically *t > s in this subgroup).

	From: Saamira Farwaneh 

A quick response to your query: Many dialects of Arabic, e.g.,
Levantine, North African and Gulf dialects, allow all types of word
initial clusters including fricative initial ones. These clusters
derive from syncopating the nucleus of the initial syllable of the
verbal or nominal template; e.g., /shirib+t/ > [shribt] 'I drank',
/fihim+t/ > [fhimt] 'I understood'.

There is a chapter in my dissertation ''Directional Effects in Arabic
Dialect Syllable Structure'' in which I deal with initial clusters in
some dialects. All the Levantine dialects, North African dialects,
Gulf dialects (except Saudi) allow initial clusters
unrestrictedly. Below are some examples:

Fricative + Obstruent Verbs: 
fhimt ''I understood' 
shfiit 'I was cured' shirt 'I stayed up late' 
shkiit 'I complained' 

xbaar 'news' 
fTaam 'weaning' (T = emphatic)
 jbaal 'mountains' 
jdaad 'new-plural' 
Hbaal 'ropes' 
Hduud 'borders' (H = voiceless pharyngeal fricative)

Fricative + Sonorant Verbs: 
smi9t '' I heard' 
jriit 'I ran' hniit '' I was contented' 
shriit 'I bought' 
ghniit ''I became rich' 
ghDibit 'I became angry' (gh = voiced uvular fricative)

 jmaal 'camels' 
flaam 'films' 
fnuun 'arts' 
Hmaar 'donkey' 
shmaal 'left' 
xmuur 'wines' 
znuub 'sins' 
zlaam 'men'

Almost every Arabic dialect grammar has a page or two on initial
clusters; below are a few references; you may also check the Arabic
linguistics bibliography at:

Also available in PDF

	From: Saso Zivanovic 

Standard Slovene: h: hkrati (simultaneously) hvala (thanks) h�i

Spoken Slovene also includes: h: hteti (syncoped form of hoteti
(want)) hmal (kmalu > hmal) (soon)) v: vzeti (take) v�igalica
(match) vneti (inflame) f: vse [fse] (all) v�e�
[f�e�] (like) (In Standard Slovene, these v- and f-
forms are always pronounced as voiced and voiceless w respectively.
We think that v- forms are acceptable (at least in our dialect
(Styrian)), although most speakers would probably still pronounce them
as w- )

In the Prekmurje dialect of Slovenian, fC clusters are possible for
almost any consonant except voiced obstruents; in the latter case we
get vC cluster by (regressive) voicing assimilation. Examples:
fcipiti (to vaccinate), f�ela (bee), fkaniti (to cheat),
flai�ter (plaster), fpijaniti (to get sbd drunk), froc (child),
fsaki (every), f�tric (in parallel), ftopiti (to drown). There
are no #fm and #fn clusters, but we think this might be only a lexical

There is a glossary of this dialect available: Novak, Franc. Slovar
beltinskega prekmurskega govora. Murska Sobota: Pomurska
Zalo�ba 1996, 2nd Edition.

	From: William J Poser 

In response to your Linguist List query, Carrier, an Athabaskan
language of the central interior of British Columbia, allows onset
clusters beginning with the voiceless lateral fricative quite freely,
just as it does with /s/. Here are some examples, in the practical
orthography, with <lh> the voiceless lateral fricative and and <u> =
upside-down V as in English ''but''. <y> is the high front glide, IPA
<j>. The underscores mark lamino-dentals. These are in the
Stuart/Trembleur Lake dialect.

 lhba for each other 
lhch'az_t_se_t_selh double-headed axe 
lhghaninli meandering stream 
lhjutwhuch'a antiseptic 
lhk'einli fork in river 
lhlajun-ne choir 
lhnalh in one another's presence 
lhtulalhgusti Way of the Cross 
lhta'ooni accidentally 
lhtsun it is smelly 
lhts'us'al we (3+) are not eating 
lhyul it (generic) is white 
lhzitke female cross cousins

The fricatives in Carrier are:

 s _s sh x z _z gh lh

There is also <wh>, which can be analyzed as voiceless /w/ or as
/xw/. /f/ is not native to Carrier but occurs in the loanword lugafi
''coffee''. There is no /zh/ - where it is expected we find /y/
instead. /gh/ is the voiced velar fricative.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue