LINGUIST List 8.647

Fri May 2 1997

Sum: Maximal Onsets

Editor for this issue: Ann Dizdar <>


  1. Ivan A Derzhanski, Sum: Maximal Onsets

Message 1: Sum: Maximal Onsets

Date: Fri, 02 May 1997 11:33:34 -0700
From: Ivan A Derzhanski <>
Subject: Sum: Maximal Onsets

My belated but very cordial thanks to

 David Deterding <>
 John Coleman <>
 Mohamed Embarki <>
 Grover Hudson <>
 Richard Krause <>
 Alex Monaghan <alexCompApp.DCU.IE>
 Rob Pensalfini <rjpensalMIT.EDU>
 Marc Picard <>
 Marc Pierce <>
 John Reighard <>
 Jane Setter <>
 David Stampe <>
 Joe Stemberger <>
 Suzanne Wash <>

for their replies to my query on the Maximal Onset Principle (MOP) in
issue 8.391 of Linguist. My (very general) questions were:

> * Do I understand correctly that the Maximal Onset Principle implies
> that a VCCV sequence will always be syllabified as V-CCV if the CC
> cluster in question can occur word-initially in the same language?

Four respondents addressed this question specifically:

JSe: `Basically, yes.'
RK: `Yes, that is the usual interpretation.'
JSt: `Yes and no. If the CC in word-initial position is an onset,
 then yes, the Maximal Onset Principle implies that. However, there
 are instances where some phonologists claim that the first consonant
 of the CC is not part of the onset. This is especially common for
 clusters that don't fit the Sonority Sequencing Principle [...].'
JC: `No: the set of word onsets is a superset of the set of syllable
 onsets. The MOP refers to syllable onsets, not word onsets.'

>* Is the Maximal Onset Principle an absolute universal? If not, what
> are (some of) the exceptions?

RK wrote: `There are many exceptions to this "rule" in classical
Greek, classical Latin, Finnish, Estonian, English, Huichol, etc. as
indicated among other things by poetic meter and location of
accent/stress'. English was also brought up as a language which
violates the MOP by JSe, DD, GH, DS, JR, JSt and AM, and so were
Arrernte (Australia), Barbaren~o Chumash (California), French and
Gothic (by RP, SW, ME and MPie, respectively).

Here are some of the points that were made:

* `The first problem is deciding just what syllables are, and like
angels, not everybody believes in them.' (JR)

* The requirement for a maximal onset may clash with a requirement for
a minimal coda (that is, VCV may be syllabified as VC.V if the first V
may not occur in an open syllable). This is the case in English
(_better_ ['bEt.] and the like).

* A single intervocalic consonant may not be able to begin a word.
According to the MOP `'sinner' would be syllabified 'si-nner', but
'singer' as 'sing-er', which doesn't seem very elegant.' (DD)

* `[T]he MOP [...] does not hold after accented vowels in
"stress-timed" languages like English, where VCCV, if the second
syllable is unaccented, syllabifies as VCC.V, if CC is a possible
word-final cluster, e.g. in "pentagon", where n is elided and t is
flapped in N. American English or glottalized in non-RP British
English, exactly as in "pent up". Contrast e.g. pentagonal [...],
where t is aspirated, exactly as in "on time".' (DS)

* `[T]he word 'Wisconsin' is sometimes pronounced with an aspirated
[k] and sometimes not, evidencing both V-CCV and VC-CV
syllabifications, even though [sk] IS a possible word initial
sequence.' (GH)

* Morphology may come into the picture. `[A] word like 'discard' may
be pronounced with an aspirated [k], again showing syllabification as
VC-CV because, presumably, the dis- prefix is recognized, which
determines a syllable boundary.' (GH)

* `[T]he whole distributional approach is questionable since many
languages exhibit medial consonant sequences which cannot be
satisfactorily resolved into permissible word-initial and word-final
clusters, e.g. Finnish.' (RK)

* `Not to worry. It's just linguistics...' (JR)

And now for a short history of my query (followed by a more specific
question). It was prompted by a passage from Donca Steriade's article
`Syllables in Phonology' in _The International Encyclopaedia of Lx_:

 The MOP explains the difference between the syllabification of
 obstruent-sonorant clusters in IE and Semitic: such clusters are
 possible syllable-initials in IE, as indicated by their occurrence
 word-initially [...]; but they are not in Semitic, where only one
 consonant may begin a syllable. The MOP connects these differences
 in the phonotactics [...] with the difference in syllabic assignment
 of obstruent-sonorant clusters (e.g. _li.bra_ vs. _lib.ra_), and
 thus explains the latter.

Indeed, Classical Arabic and Latin have virtually identical rules for
stress placement (stress the penult if it is heavy and the antepenult
otherwise), but they differ in the two ways referred to in the quote,
in which a causal connexion seems to be implied: Latin stress follows
a `muta cum liquida' rule precisely because in Latin a word can begin
with an obstruent-sonorant cluster.

I know of a language (but not a natural one) which has the same stress
rules as Arabic and permits word-initial obstruent-sonorant clusters -
but does not have a `muta cum liquida' rule. It is Sindarin, one of
JRR Tolkien's languages.

So my next question is: Is there (or can there be) a natural language
which behaves in that (or a similar) way?

Thanks again!

"reH Sov yInej 'ej Dap yImuS, <dOstI bA mardom-e dAnA nEkO-st,
 jagh val qaq law' jup QIp qaq puS" do^sman-e dAnA beh az nAdAn dOst>
 (Sheikh Muslihuddin Abu Muhammad Abdullah Saadi Shirazi)
Ivan A Derzhanski <>
H: cplx Iztok bl 91, 1113 Sofia, Bulgaria <>;
W: Dept for Math Lx, Inst for Maths & CompSci, Bulg Acad of Sciences
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue