LINGUIST List 5.129

Thu 03 Feb 1994

Sum: Distant Assimilation

Editor for this issue: <>

Directory

1. Brian D Joseph, SUMMARY -- Distant Assimilation

About 6 weeks ago, Rex Wallace of UMass and I posted a
query asking LINGUIST readers for clear examples of
distant (i.e. nonadjacent) assimilations. Though we weren't
as specific in our query as we should have been, our
interest was in distant consonantal assimilations, and more
particularly in whether there were any such instances
involving manner assimilation. The reason for our interest
in this question stemmed from a paper we were preparing
for the LSA meeting in January in which we examined how
Gamkrelidze & Ivanov's Glottalic Theory would account
for the development of Proto-Indo-European voiced
aspirates (e.g., the traditionally reconstructed *dh, or *d[h]
in the Gamkrelidze-Ivanov system, where *d[h] means [d]
with phonetically present but phonemically irrelevant
aspiration) in Italic. Basically, in order to do away with the
need to posit Grassmann's Law (dissimilatory loss of
aspiration on the first of two successive aspirated
consonants) in Greek and Sanskrit (and Tocharian, though
they don't discuss that) independently, Gamkrelidze &
Ivanov posit allophonic variation between [d] and [dh] for
PIE, governed by a constraint that only one aspirate was
allowed to surface in a root that had two potentially
aspirated consonants underlyingly. Thus, for them, PIE */
d[h]eyg[h]-/ would surface as [deygh-] or as [dheyg-] but
not as [dheygh-], in PIE itself, and thus there should be no
reflexes of diaspirate forms like [dheygh-] in any of the
daughter languages.

 It turns out though that in Italic there are reflexes of such
"diaspirate" forms, actually a whole bunch -- we eventually
found 8, though some are of lesser probative value than
others for philological reasons, problems with textual
transmission, obscurity of etymology, etc. One of the best
examples is Oscan < feihuss > 'walls', pretty clearly from
the *dheygh- 'fashion, shape' root. Gamkrelidze & Ivanov
would predic that **feig-(uss) would occur, since the
Oscan medial -h- is what would be expected as the reflex of
the aspirate *gh allophone and the initial f- is what is
expected from the aspirate *dh, in violation of their
putative "one aspirate at most per root:" constraint for
Proto-Indo-European. In their 1984 book (Indoevropejskij
jazyk i indoevropejci) they explicitly claim that feih- for
their expected feig- is the result of an assimilation
(presumably, therefore, a distant manner assimilation;
presumably they would also employ a similar strategy for
some of the other 7 examples we found). Hence, our
interest in the matter, for if f...g --> f...h is an "unnatural" or
unlikely sound change, their account of the Oscan form is
weakened.

All of the distant consonantal assimilations we were aware
of (and interestingly, there were not many examples in the
standard textbooks on historical linguistics that we had
ready access to) seemed to involve place assimilation (e.g.
the well-known case of p...kw --> kw...kw in Italic, (kw =
labiovelar here, by the way)), or nasal assimilation
(Medieval Greek mespilon 'loquat' --> later musmulon), or
aspiration (e.g. under one interpretation, Ancient Greek
thuphlos 'blind' (dialectal) versus tuphlos (in most dialects),
where the unaspirated initial may be the older form), but we
weren't aware of any involving manner assimilation alone
(e.g. of the sort f...g --> f...h); we were aware of cases
involving manner assimilation only incidentally, e.g. where
there was a total assimilation that ended up yielding a stop
out of a fricative in the context of a nonadjacent stop (as in
the case of Latin barba 'beard' from expected *farba (PIE
*bhardha:) or Italian pipistrello 'bat' that Leslie Morgan
mentions below), but ones involving manner alone not
ultimately yielding a total assimilation eluded us.

Below are some of the observations and examples that were
forwarded to us (I am excluding mention of vowel
assimilations (umlaut, vowel harmony, and the like), which
many people noted generally nonadjacent material
assimilating; it is interesting that nonadjacent assimilations
for vowels are so prevalent but not for consonants). Our
thanks to everyone who responded (Lloyd Anderson,
Lowell Bouma, Aaron Broadwell, Richard de Armond,
Hannele Dufva, Jill Hart, Ellen Kaisse, Patrick McConvell,
Leslie Morgan, Steve Seegmiller). As you can see, good
examples of assimilation in manner alone that are not total
assimilations are hard to find and perhaps do not exist; there may be
phonological reasons for this (as suggested by Ellen Kaisse) or
physiological/phonetic reasons (as suggested by Lloyd Anderson).

--FROM Patrick McConvell (mcconvell_puncl04.ntu.edu.au)

"Something I first treated as dissimilation I am now treating
as assimilation - specifically spreading of [-nasal]. It
concerns Gurindji and a number of other Australian
Aboriginal languages. In the environment of a preceding
nasal-oral stop cluster, nasal codas denasalise and delete
altogether if they the first part of homorganic clusterse.g.

kanka-yin -> kanka-yit
nyampa-wu-warla-yi-nta -> nyampa-wu-warla-yi-ta

As you see this acts at long distance, over any number of
liquids and glides; nasal and oral stops usually block and
here is where the story gets complex."

--FROM Ellen Kaisse <kaisseu.washington.edu>:

"Couldn't a large part of your problem be that manner
rarely assimilates, even in adjacent segments, whereas
place, nasality and laryngeal features typically do spread?
(by manner I mean stricture features like continuant
and maybe sonorant and maybe consonantal)

--FROM Hannele Dufva <DUFVAjylk.jyu.fi>

"How about changes in the slips of the tongue? Tendencies
that I would describe as assimilatory are quite common.

PS I'm no specialist in Finnish (morpho)phonology but I do
have a large collection of slips of the tongue (Finnish), and
I could pick up a couple of examples if you're interested."

--FROM Lloyd Anderson (<ECOLINGAppleLink.Apple.COM
 (Ecological Linguistics,Anderson,PRT)

"I would not expect manner assimilation of stop to fricative
in the case of the noisier fricatives at least, for both
articulatory markedness reasons and also for perceptual
reasons. Studies of child language have shown a "masking"
effect such that if there are two fricatives, one will go the
opposite direction, lose its fricative character."

--FROM Jill Hart <G.R.Hartdurham.ac.uk>

"A couple of examples from Hittite:

(1) eshahru "tear" where the consonant before -r- was
originally k' but has apparently been assimilated to the -h-
at the beginning of the second syllable - though that -h- is
itself a problem.

(2) A more doubtful one, though some people accept it:
harsar "head" from *k'(V?)rHsr, with change of *k' > h
under the influence of following H [here = a PIE
"laryngeal" consonant] assumed from cognates but lost in
Hittite (by subsequent dissimilation?)"

--FROM Leslie Z. Morgan <MORGANLOYOLA.EDU>

"I work with Italian; the example of interest is "pipistrello"
for Old Italian "vipistrello" (= "bat"). If you're interested in
other examples, many of which involve nasality, look at
Rohlfs, Gerhard. _Grammatica storica della lingua italiana
e dei suoi dialetti_. Vol 1: _Fonetica_. The edition I have
is a paperback, 1966 from Torino: Einaudi. It was
originally printed in German, 1949 in Bern by A. Francke,
AG.