/hw-/ > /w-/ once more

[Larry Trask]

Well, I failed to notice that Mark Hale's posting, to which I replied
on this list, had been sent only to me.  My response must have
bewildered quite a few people; apologies.  Apparently it's now
impossible to post to the list by using a `reply' option.
 
> As far as I know the existence of a contrast between [kW] (a
> labiovelar) and sequences of the type /k/ + /w/ (similarly for the
> voiced and voiced aspirated series) is universally accepted by
> IEists. The ancestor of the /hw/ of 'what' is a labiovelar.
 
Thanks.  The recent discussions of PIE phonology I've seen have
concentrated on other and more exciting issues, and I haven't seen a
view expressed on this point for years.
 
[On my point that OE /hl-/ derives from PIE */kl-/, and so on]
 
> Etymological source is not probative for determining the synchronic
> phonological status of segments.
 
Agreed, of course, but then I wasn't suggesting that that the PIE
origin proved anything about Old English -- only that it shows that
cluster reduction must have occurred at some time.
 
>> Second, in OE alliterating poetry, /h-/ regularly and freely
>> alliterates with all of /hw-/, /hl-/, /hr-/, and /hn-/, suggesting
>> that, if anything, these items were perceived by speakers as
>> clusters.
 
> As the vowels show, alliteration is feature-driven, rather than
> segment-driven. The alliteration facts point to some phonetic
> similarity between the segments, but not to identity.
 
Well, I'm perfectly prepared to accept a featural element in
alliteration, if the evidence points that way.  But surely it is going
too far simply to declare that "alliteration is feature-driven", and
that's all there is to it.  As far as I can see, a segmental
interpretation of OE alliteration is a lot simpler and more successful
than a feature-based one.
 
I don't think the alliteration of vowels is a strong argument for a
feature-based view: it's just that zero onsets alliterated, and
unsurprisingly so.  How does it clarify matters to bring the vowels
into it?
 
Anyway, what would have been the featural basis of the alliteration of
the five onsets in question?  We have alliteration among /h-/ (a
voiceless vowel, probably, though conceivably a fricative), /hw-/
(perhaps a voiceless glide, probably a fricative), /hl-/ (a voiceless
lateral), /hr-/ (a voiceless rhotic), and /hn-/ (a voiceless nasal).
The only feature these obviously have in common is [- voice] -- but
other voiceless segments don't alliterate with any of these, or with
one another.
 
If you are happy to deny fricative status to the first two, then you
could point to [- voice, - obstruent] as the class in question.  But
this looks fishy to me.  First, I'm aware of no evidence that these
five items behaved as a natural class in any other respect in Old
English.  Second, none of the parallel groups [- voi, + obstr], [+
voi, + obstr], or [+ voi, - obstr] alliterate in Old English.
 
I therefore don't find it easy to agree, with Mark, that "The
alliteration facts point to some phonetic similarity between the
segments, but not to identity."  Are there other cases in OE of
alliteration between onsets which share some phonetic similarity but
not identity?
 
[on my intuition that my onset is phonologically /hw-/]
 
> Prof. Trask's "feelings" are, unfortunately, also non-probative, no
> matter when he started feeling them. They are non-probative for the
> Modern English dialect he speaks, and completely irrelevant to the
> question of the status of /hw/ in Old English. Even if Prof. Trask
> were *considerably* older than he appears.
 
I don't know how old I appear, but I think some of my younger students
have the vague impression I've been around since the Bronze Age :-)
Hell, even one of my British colleagues once asked me in all
seriousness what it was like watching Joe DiMaggio play.  Honestly.
But I digress.
 
Of course I agree that my intuitions don't prove a phonological
analysis -- but I don't think they're irrelevant, either.  My point is
that it's perfectly possible for speakers to perceive their speech in
phonological terms which are at odds with the phonetics -- the same
point made so famously by Sapir over 70 years ago.  Hence, even if we
had ironclad evidence that the onsets in question were phonetically
single segments in OE (and we don't, of course), such evidence would
not prove that they were not clusters phonologically.
 
>> The OE spelling is, of course, entirely consistent with the cluster
>> interpretation, but is hardly decisive.
 
> *hardly decisive* is rather an understatement: 'completely
> irrelevant' would be closer to the mark, in my view.
 
Well, no.  I'm afraid I can't agree that the orthography established
by native speakers is "completely irrelevant" to the phonological
facts of a language.  If that were true, orthographies would be
totally arbitrary, and they're clearly not.  Of course the orthography
doesn't prove anything, but, as Miguel Carrasquer Vidal has pointed
out, it's interesting that OE-speakers never showed any tendency to
represent these onsets with single letters.
 
> I agree with Prof. Trask that the matter is not completely
> determinable by the evidence. But we should distinguish between
> arguments which might in principle support a particular analysis and
> those which are in fact not relevant.
 
Yes, I am in full agreement with Mark here.  But I can't see that I've
raised any irrelevant arguments.  Once again, my points are these:
 
Etymology: three of the four onsets originated as clusters.
 
Alliteration: the alliteration facts are clearly more compatible with
a cluster analysis of all four onsets than with a single-segment
analysis.
 
Intuition: at least some modern speakers perceive /hw-/ as a cluster,
in defiance of the phonetics, and therefore purely phonetic facts
about OE, even if we had any, would not in principle count against the
cluster analysis.
 
Orthography: this is not inconsistent with either analysis, but is
more directly supportive of the cluster analysis.
 
I'll remind everybody that I was only trying to give a simple answer
to a question about the /hw-/ ~ /w-/ variation in mondern English, and
not proposing a serious analysis of Old English phonology.  But, if I
were to undertake this last task, I think I would probably conclude
that, while the evidence available is not decisive, a cluster analysis
looks more attractive than a single-segment analysis.
 
Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
England
 
larryt at cogs.susx.ac.uk