posting to HISTLING[D[D[D

Prof. Dr. R. Hickey lan300 at
Wed Sep 18 18:28:15 UTC 1996

----------------------------Original message----------------------------
Email title:
/hw/ - /w/ and syllable structure
                                                Date: 17 September 1996
                                                Time: 00:36:21
To:      HistLing subscribers  via  owner-histling at VM.SC.EDU
From:    Hickey, Raymond            r.hickey at
Cluster analysis of single phonetic segments
Have I overlooked somebody's comments in the current discussion about
<hw> - <wh> or can it really be that no-one has considered structural
arguments here? To remind ourselves: we are dealing here with phonetic
reality and phonological postulations. Phonetic segments are there,
i.e. for Irish, Scottish and most American speakers of English in a
word like _which_ there IS a lack of voice during the approximant which
contrasts with the voice which IS present in a word like _witch_. The
next question in whether, in the abstract analysis of the place this
sound has in the system of English phonology, we assume two segments /h/
+ /w/ or just one /W/. On this level there is (as yet) no question of
proving existence. The ontological status of phonological postulates is
another issue which the linguist cannot address, given the inability to
link any of the analyses we make with any level of mental activity.
Other considerations determine the type of analysis we offer: symmetry
and balance in the system we construe, simplicity (parsimoniousness of
design, keeping to Occam's razor, etc.), maximum explanatory power of
single constructs (free-rides and the like). Sorry, all this is rather
platitudinous to those in the field, but it should nonetheless be borne
in mind.
        Now there is strong system-based evidence for the analysis of [W]
(voiceless labiovelar approximant), in those varieties of English which
still have it, as consisting of two segments (segment = a phonological
unit which does, if biuniqueness applies, correspond to a phonetically
identifiable sound, but does not have to).
        You might think to begin with that /w - W/ form a voiced -
voiceless pair in English like /s - z, t - d, p - b/, etc. However in
my opinion, the arguments for [W] as /h/ + /w/ are more compelling;
here they are.
1)      /hw/ and /h/ in general
The first segment in /hw/ which one can posit phonologically correlates
with /h/ word initially, that is to postulate /h/ + /w/ has additional
justification in the fact that /h/ occurs initially anyway (in
varieties with [W]). Conversely, to my knowledge no variety of English
which has /h/-dropping also has [W]. i.e. lack of /h-/ precludes the
cluster /hw-/.
2)  Position in syllable
It is a standard wisdom on syllable structure that there is in general
an increase of sonority from edge to centre. Analysing [W] as /hw/
means that one has an obstruent /h/, then a semi-vowel (a continuant
with very open articulation) /w/ and a following vowel which is in
keeping with this cline.
        Note that this same argument can be used to support an analysis of
[Cu:] (C = voiceless palatal fricative) in English _hue_ as /hju:/ :
obstruent, semi-vowel, vowel with increasing sonority from beginning of
onset to nucleus.
        I take it that I don't have to offer too much justification for
regarding /j/ and /w/ as semi-vowels, they are high front and high back
glides respectively and have been regularly posited as diphthong
end-points in American phonology since the early structuralists and are
seen PHONETICALLY as glides in hiatus, e.g. _seeing_ [si:jiN] and
_doing_ [du:wiN].
3)  Markedness considerations
Cross-linguistic observations hav lead in phonology to many valid
statements treated under the heading `markedness' (understood
statistically here). Thus voice is unmarked for vowels, semi-vowels and
sonorants just as voicelessness is for obstruents, i.e. there are
(far) more languages with voiced sonorants that with voiceless ones as
well and there are (slightly) more languages with voiceless fricatives
than with voiced ones as well.
        Now English of course does not have voiceless sonorants
(systemically) so that to posit [W] would mean that there would be an
unevenness in the distribution of voice, compare the following charts.
[W] = /hw/
   -----------------|                                        SONORITY
   voiceless        |--------------------------------           |
   voiced            voiced      voiced        voiced           |
   obstruents        sonorants   semi-vowels   vowels           v
   EDGE -------------------------------------> CENTRE
[W] = /W/
   -----------------|          |-------------------|          SONORITY
   voiceless        |----------| voiceless         |-------      |
   voiced            voiced      voiced              voiced      |
   obstruents        sonorants   semi-vowels         vowels      v
   EDGE -------------------------------------------> CENTRE
4)  Historical development
Notions of markedness can be applied historically as well; they can
render changes fathomable though they cannot, of course, predict them.
With regard to the issue at hand: English lost sequences of /h/ and
sonorant by early Middle English, /hr, hn, hl/ -> /r, n, l/ as the
voiceless sonorants which were their phonetic realisations were more
marked. Okay - I know - I am using `marked' in a different sense now.
Here I mean phonetically unnatural (or unusual for a less controversial
term). The reasoning is as follows: sonorants are characterised by
virtually no obstruction of air flow in the supra-glottal area, /r/ has
no contact (unless intermittently, if trilled), /l/ has free lateral
flow of air, /n, m, N/ have nasal flow. With free passage of air, the
vocal folds are more likely to vibrate (Bernoulli effect), i.e.
produce voice, so that voicelessness with these segments is a `marked'
        Stop, you say - if voiceless sonorants are unusual then voiceless
semi-vowels are even more so. True, the question is why has [W]
survived so long in so many varieties of English? There is no simple
answer to this, but I think one argument is acceptable, indeed
strengthened by the interpretation of [W] as /hw/. Note that this is
parallel to /hj-/ as in _hue_. And there are many sequences of /hj-/ in
English, guaranteed by the occurrence of /h/ before /ju:/. The point
here is that /hw/ could have been bolstered by the established position
of /hj-/ sequences, much as the voice distinction between interdental
fricatives owes its existence (cf. its tenuous functional load) not
least to the centrality of the voice-voiceless distinction among
obstruents in general in English.
5) A little support from outside
In case the above has not convinced the staunchest supporters of a
doggedly phonetic interpretation of [W] of the wrongness of their ways,
allow me in conclusion to cite some non-English evidence. In Irish
there are several clusters of /s/ followed by a sonorant: /sr-, sl-,
sn-/. When preceded by a grammatical element which demands lenition
(phonetically weakening of a word-initial segment) the /s/ is altered
to /h/ and what one obtains phonologically are the sequences /hr-, hl-,
hn/, e.g. _sro'n_ `nose', _a shro'n_ `his nose'. Now these sequences
are PHONETICALLY voiceless segments but it would be ludicrous to
postulate /R, L, N/ (voiceless sonorants) PHONOLOGICALLY for Irish.
        Returning to English: one could in fact take the phonetic
interpretation case to comic lengths to demonstrate its absurdity. If
one includes allegro phenomena in English, for instance, then one could
assume that English still has voiceless sonorants, after all
PHONETICALLY the sonorants in _come here_ [M], _clout_ [L], _snoose_
[N], _shrimp_ [R] are all at least partially voiceless.
        But maybe the matter at the end of the day boils down to one's
Weltanschauung - some people do not like abstract analyses and just
will not accept them no matter what amount of convincing evidence one
On the issues discussed here allow me, at the risk of blowing my own
trumpet (just a little puff), to offer a germane reference:
Hickey, Raymond. 1984. "Syllable onsets in Irish English", Word 35:
Btw: I would be a little wary of maintaining that naive speakers (or
colleagues like Larry Trask before they got infected by linguistics)
think of [W] as a _w_ sound with a _h_ before it: the validity of
phonological segments is not affected by speakers consciousness
awareness of them, or their lack of this. In the days of dark
ignorance before I started linguistics I did not realise that [W] could
be conceived of as /h/ + /w/ (although I have [W] for every _wh-_ in
English) and it might be reading too much into one's own early
biography to think so; however, this is not of immediate relevance to
the matter at hand (though speaker intuitions are of course important).
Ray Hickey
University Essen | Tel.  : +49 201 183 3580
FB 03            | Fax.  : +49 201 183 xxxx
-Anglistik-      | E-MAIL: r.hickey at
D-45117 Essen    | Germany

More information about the Histling mailing list