LINGUIST List 13.1596

Tue Jun 4 2002

Sum: Phonetics: s/z/h Alteration

Editor for this issue: Marie Klopfenstein <>


  1. Remy Viredaz, General phonetics : z > h ?

Message 1: General phonetics : z > h ?

Date: Tue, 4 Jun 2002 11:43:29 -0700 (PDT)
From: Remy Viredaz <>
Subject: General phonetics : z > h ?

On 10 Jan 2002, I asked the Linguist List
<> :

> Various languages in different parts of the world
> undergone a conditioned change s > h. But :
> (a) are there languages where there has been a
> unconditioned change s > h ?
> (b) are there languages where z (as voiced
> of s) has changed to voiced h or any similar sound,
> under what conditions ?
> (c) among the languages in (b), is voiceless s 
> maintained or is it changed to h ?

Within a few days, I got responses by :
Suzanne Boyce <> (invalid
Lance Eccles <>
Johannes Heinecke
Mark Jones <>
John E Koontz <John.KoontzColorado.EDU>
Johanna Laakso <>
Tadhg � hIfearn�in <>
Mikael Parkvall <>

In addition, I was given the names of
Louis-Jacques Dorais
Michael Fortescue <>
Willem J. de Reuse <>
who kindly accepted to answer specific questions.

To all of them I express my heartfelt thanks. 
I also apologize for not having had time to complete
this summary earlier.

The text below consists partly of summaries of
information given by colleagues, followed by their
names in round brackets, partly of additions by
myself, marked off by square brackets. Of course, I am
sole responsible for possible errors or
misunderstandings in either the summaries or the

Question (a):

Are there languages with a general, unconditioned
change s > h ?

Yes, there are. [However, they seem all to have had s
 only in a restricted set of possible environments.]

Unconditioned change has occurred:

- In Eastern Polynesian languages (Tongan, Hawaiian,
Tahitian, Maori, and others), where Proto-Polynesian
*s > h (or �). There is no s in any of these
languages. See Krupa 1982, 4-5, 18-19. (Eccles)
[There were no consonant clusters, so that the
possible environments were only #sV and VsV. - This
change is not a case of "push chain", i.e. there was
no other phoneme that might have been about to become
s and would have "pushed" original s to another

- In a number of Tup�-Guaran� languages. See Jensen
1999, 137-139. (Parkvall) 
[Proto-TG *ts appears variously as ts, s, h, or zero,
in the daughter languages. *tS (*c^) yields the same
results, except in one subgroup, where it is retained.
- There were no consonant clusters, so that the
possible environments were only #sV and VsV (and Vs#
?). - No *s or *h is reconstructed for proto-TG,
Jensen 134.]

- [In Arapaho-Atsina. Though Picard 1994, 89
recognizes only *VsV > *VhV explicitly, various
examples in pp. 93-173 involve *'s > *'h (' = glottal
stop) and *sk > *hk as well. In addition, the strange
change Proto-Algonquian *s > Arapaho n in word-initial
position (Pentland 1998) implies *s > *h in the
environment #_V as well, then *hV > *yV (an unusual
change in itself), then regularly *y > *l > n
(Pentland 317-318).] [Proto-Algonquian, as
reconstructed, had only few consonant clusters (which
consisted of only two consonants, not counting w or
y). *s occurred only in the above-mentioned
environments and in *hs > *s; not even in (+)*sp, (+)*st.]

- In parts of Eskimo. A general change of s to h is
(or has been) in progress in Northern Greenland, parts
of Western Canada and Alaska Eskimo; namely: Kivalliq,
Nattilingmiut, Inuinnaqtun dialects of West Canadian
Inuit; Uummarmiut and Nunamiut varieties of North
Slope dialect of Alaska Inupiaq; Inughuit dialect of
NW Greenland or "Thule". (Dorais)
The relevant environments are #sV and VsV; in
addition, ps, ts, ks, qs > ph, th, kh, qh or ff/ss,
tt/ts/tch, xx, XX, according to dialects. s did not
occur in other positions. [The instability of s might
be due to its palatal origins, cf. Dorais 1996, 92; no
*s is recon-structed for Proto-Eskimo; s as
mentioned above is from proto-Eskimo *c.] See Dorais &
Lowe 1982 [128, 131-2]; Fortescue 1983 [see one of the
maps; relevant pages not known to me]; Dorais 1986
[23, 25, 27-29, 38, 46-47]; Fortescue 1991 [2-3];
Fortescue-Jacobson-Kaplan 1994 [xvi � A 4]; Dorais
1996, 92-94; Fortescue 1998, 28 (map). (Dorais,
[In North Greenlandic at least, the occasional voiced
realization of /s/ is maintained, too, e.g. /kisiisa/
[kiziiha] 'they, only'. However, these instances of
retention are not very significant, because the change
s > h is only recent in North Greenlandic, where [Xs]
> [XX] is not entirely completed with the old
generation. See Jacobsen 1991, 53-54.]
In the Alaskan and Canadian area at least, many young
people tend to replace local h by the s of the Baffin
dialect which they learn in school (Dorais).

[The change of the 'chuintante' s^ to a velar
fricative x, presumably after a retroflex stage, has
occurred in various languages, too, but s > h and s^ >
x, though similar, are better considered as different

Question b:

Are there languages where z > voiced h or a
similar sound, and under what conditions ?

The answer is uncertain.

- [In Burushaski, initial *s changes to h in the same
context (unaccented syllable) where voiceless stops
undergo voicing (Berger 1959, 33; I didn't have access
to Lorimer 1935-1936 and haven't seen Berger 1998
yet). This may mean, either that *s > *z > h, or that
the fricative s has been weakened in another way
than the stops (perhaps at a different time). The
former interpretation is more likely (especially if
Bur. h is a former voiced sound as in Indo-Aryan).
However, that *z would not have been a phoneme, but
the weak allophone of *s .]

- In the Baltic Finnic languages (Finnish, Estonian
and some smaller, closely related languages), there
are cases of morphophonological alternation between s
 and (intervocalic, word-internal) h , e.g. (Finnish)
mies 'man' - miehe-n; l�he-ll� 'near' (an
adverbial form containing the adessive case suffix) -
l�s-n� 'present' (the same adverb stem with a
different local case suffix); taivas 'sky' :
taivaha-n (surviving in some dialects) > taivaan. The
change -s- > -h- in intervocalic positions goes mainly
for *suffixal* -s-, i.e. after non-first syllable
(l�he- and perhaps miehe- are exceptions). Many
scholars have preferred to reconstruct a *z as the
"missing link" between /s/ and /h/. This would allow
for the reconstruction of a "gradation" s : *z similar
to the Proto-Finnic consonant gradation, where
stem-internal voiceless stops alternated with
(according to classical handbook view) homorganic
voiced spirants. However, the conditions for the s : h
alternation seem to be different, more simple: h
appears in all intervocalic positions, not just in the
beginning of closed syllables, as the weak grade in
the consonant gradation did. (Laakso)
[Here again, one hesitates between *s > *[z] > h, with
non-phonemic *[z], and a weakening *s > h that would
not be parallel to the lenition of plosives (*p > v
etc.). The latter is quite possible, as the
conditioning context is not the same anyway, and s > h
is probably later than the lenition of plosives. (at
least, weak s > /h/ must be later that *s^ > *x > h
which is the main source of Finnish h ).]

- [In Algerian Tuareg of the Hoggar (Tahaggart),
Berber *z > h (whereas zz and emphatic z are
maintained). However, various reasons point rather to
a surprising evolution z > z^ > s^ > h. See Prasse
1969; Prasse 1972, 45-46. By the way, one example of
these dialectal variations is the name of the language
itself: Tamazight, Tamasheq, Tamahaq.]

- In Breton, Celtic /t/ became under certain
conditions */th/ (that is 'thorn', 'theta'), as it is
still in Welsh, and then became /z/ in the KLT
dialects of Breton (Cornouaille, L�on, Tr�guier), /h/
in Vannetais (hence the compromise spelling zh in
Modern Standard Breton). The question is whether in
Vannetais *th > *z > h or *th > h directly. (Heinecke,
O' hIfearnain)

[ (A) I'll try to summarize what Jackson 1967 has to
say on the question. < If you're in a hurry, go to B >
The main origins of this diaphoneme are Celtic *aht
(*axt) > *ait (where a = any vowel) and Celtic *tt >
*t. These new instances of postvocalic *t [being
preaspirated?] were lenited to spirant *th. (Unlike
Jackson 1967, 317-319, I cannot see the Brittonic
phenomenon as a direct change from geminate voiceless
stop to spirant.) Besides the voiceless spirant *th,
Brittonic had a voiced spirant *dh ('eth', 'delta'),
maintained in Welsh (spelt dd).
As to the question of the diachronical "path" that
these phonemes followed in Breton, see Jackson 1967,
516-534 (with a summary p. 516) and 643-710 (esp.
683-684). The question is rather complex. 
The modern products of the voiceless spirant Britt.
*th are: 
(a) in internal or final position, z in L�on,
Tr�guier, and most of Cornouaille, voiced h in
Vannetais, and nil or a hiatus-filling y in varying
parts of South-West and South Cornouaille; 
(b) in word-initial position, generally z (including
Vannetais), but s in those areas (in Tr�guier) where
the "new lenition" did not occur (516-534). 
The modern products of the voiced spirant Britt. *dh
(from Celtic *d after a vowel as well as *y after *i
as in *nowi(y)os 'new' > W. newydd, Br. nevez) are,
roughly, z in L�on, nil in the rest of Breton (or h as
hiatus-filler). In initial position (as the result of
initial lenition), the result of Britt. *dh is z
everywhere.However, instead of z from either Britt.
*th or *dh, a special sound (Le Roux's underdotted z),
phonemically different of z, is attested in parts of
Vannetais and Belle-Ile as well as on the Ile de Sein.
French speaking linguists have tended to describe it
as dh (the voiced English th), though it is in fact a 
z with the tip of the tongue approaching the teeth.
See Jackson 1967, 504-5, 661-3, 696-8. 
Basing on the above and on the study of earlier
spellings (which seems impossible to summarize),
Jackson reaches the conclusion that the change to h 
(from Proto-Brittonic *th) or to nil (from
Proto-Brittonic *dh) occurred at a stage when the
spirant was no longer thorn or eth and not yet
ordinary, apical z, but still an intermediate sound
which he notes as sigma or zeta respectively (the
latter being probably identical with or similar to
just-mentioned underdotted z, 697). (See Jackson
1967, 517-523,645-8, 667-684, 686-7; summary: 687-8 ;
see also 732-3, 753-4).) "Jackson is pretty good, but
much more work has been done since" (O' hIfernain).
However, it would have been too long for me to study
the Breton case any further.

(B) To summarize the Breton case, the relevant points
for our purpose are the following, if Jackson is
- it is not *s and *z that became h and nil
respectively, but sounds that were phonetically
intermediate between *th, *dh and *s, *z (and
contrasting phonemically with *s, *z)
- the voiced spirant seems to have been lost directly,
with no intermediate stage such as voiced h (where the
end result is h it is a hiatus filler). ]

[In conclusion, it would seem that there are simply no
languages where /z/ as a phoneme has been changed to
(voiced) h, whereas there is at least one possible
example (Burushaski) for the change of *[z] as an
allophone of */s/ to (voiced) h ; this interpretation,
however, involves linguistic (internal)
reconstruction, which makes it hypothetical.]

Question (c) 

[My former question c thus vanishes, but another
question can be asked:]

What is the articulatory phonetic interpretation of
this sound change ? (Boyce)

If we accept the phonetic explanation that glottal
abduction for the fricative gesture 'takes over' the
entire segment, whether in response to or as a
catalyst for gestural reduction, then it seems clear
that a z > h change is unlikely without an intervening
[s] phase. The glottal abduction for frication, which
already makes voiced fricatives aerodynamically
troublesome and often leads to their devoicing, would,
if increased, devoice /z/ on the way to /h/. Cf.
Widdison 1997. (Jones)

[If Mark Jones' suggestion is correct, it would
explain why no examples of /z/ > (voiced) h have been

In the traditional (?) interpretation, the cause of s
> h is a weakening of the buccal gesture, with the
consequence that the breath inherent to s is the only
feature remaining and is reinterpreted (and
refashioned) as /h/.Widdison remarks that, as the
widening of the glottis (which is required for the
high rate of airflow involved in sibilants) is
anticipated, the preceding vowel gets a breathy voice
quality which largely mimics [h]; he further remarks,
on the basis of experimental tests, that this effect
plays a role in the recognition of /s/ and may even
(with Spanish-speaking hearers) be perceived itself as
an /s/. However, he doesn't address the case of s > h
in initial or even interconsonantal position (Greek
*septm > hepta, *ekstos > ekhthos). Widdison's
observations do not point to an "active" role of the
inherent breath in the weakening of the buccal gesture
for s. They are compatible with the traditional
interpretation (weakening of the principal
articulation, hence increased importance of previously
secondary clues).

Another point in Widdison's paper is the only
partially voiced realization of phonemically voiced
sibilants, due to the inherent conflict between robust
voicing, which requires low oral air pressure, and of
noisy frication, which requires the opposite condition
(Widdison 1997, 260, quoting Haggard and Ohala). (In
e.g. English, the s : z contrast relies largely on a
difference in phonetic length. However, this is true
of stops as well.) Thus, my assumption that a change
of /z/ to a voiced h would be parallel to the
well-known change /s/ > /h/ is questionable because it
did not take account of the phonetic detail.]

(about A(rapaho), Be(rber), Br(eton), Bu(rushaski),
E(skimo), P(olynesian), Sp(anish), T(upi-Guarani),

(Bu) Berger, Hermann, "Die Burushaski-Lehnwoerter in
der Zigeunersprache �, IIJ 3, 1959, 17-43.

(Bu) Id., Die Burushaski-Sprache von Hunza und Nager.
Wiesbaden: Harrassowitz, 1998, 3 vols., one of whom
is: Grammatik.

(E) Dorais, Louis-Jacques, "Inuktitut Surface
Phonology: A Trans-dialectal Survey", International
Journal of American Linguistics, 52, 1986, 20-53).

(E) Id., La Parole inuit. Langue, culture et soci�t�
dans l'Arctique nord-am�ricain. Paris: Peeters, 1996.

(E) Id. & Ronald Lowe, "Les dialectes de l'Arctique de
l'ouest", �tudes/Inuit/Studies 6(2), 1982, 127-133. 

(E) Fortescue, Michael D., A Comparative Manual of
Affixes for the Inuit Dialects of Greenland, Canada
and Alaska (Copenhague, Meddelelser om Gr�nland),

(E) Id., Inuktun. An Introduction to the Language of
Qaanaaq, Thule. Copenhague: K�benhavns Universitet,
Institut for Eskimologi, 1991.

(E) Id., Language relations across Bering Strait.
Cassell Academic, 1998.

(E) Id., Steven Jacobson, & Lawrence Kaplan,
Comparative Eskimo dictionary with Aleut cognates.
Fairbanks, AK : Alaska Native Language Center,
University of Alaska, Fairbanks (POB 757680,
Fairbanks, AK 99775-7680), 1994. 614p. 

(Br) Jackson, Kenneth L., A Historical Phonology of
Breton. Dublin Institute for Advanced Studies, 1967.

(E) Jacobsen, Brigitte, "Recent phonetic changes in
the Polar Eskimo dialect", �tudes/Inuit/Studies 15(1),
1991, 51-73.

(T) Cheryl Jensen, "Tup�-Guaran�", in Robert M. W.
Dixon & Alexandra Y. Aikhenvald, The Amazonian
Languages. Cambridge University Press, 1999, pp.

(P) Viktor Krupa. The Polynesian Languages: A Guide
(Languages of Asia and Africa, vol 4). London:
Routledge & Kegan Paul, 1982.

(A) David H. Pentland, "Initial *s > n in
Arapaho-Atsina", Diachronica 15, 1998, 309-321.

(A) Marc Picard, Principles and Methods in Historical
Phonology, From Proto-Algonkian to Arapaho. Mc Gill
University Press, 1994.

(Be) Karl-G. Prasse, A propos de l'origine de h
touareg (tahaggart) 1969, Copenhagen, 96 p.

(Be) Id., Manuel de grammaire touar�gue (tahaggart), 3
vol. (274, 440, 294 p.), 1972-1974 ; Copenhagen
University ress; vol. 1 = Phon�tique, Ecriture,

(Sp) Kirk A. Widdison, "Phonetic explanations for
sibilant patterns in Spanish", Lingua 102, 1997,

Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue