LINGUIST List 3.584

Thu 16 Jul 1992

Disc: Dissimilation, assimilation

Editor for this issue: <>

Directory

1. Jennifer Cole, dissimilation
2. , 3.569 Phonology: Assimilation, Dissimilation, Natural

We are grateful to D.Stampe for sharing his views on dissimilation
with us [vol. 3.551]. In our summary of responses to our query on
dissimilation [vol. 3-546], we noted that we had not yet received any
productive cases of long-distance (beyond the syllable) dissimilation.
The only candidate was Latin Liquid Dissimilation, which has been
argued by Bernhard Hurch not to be productive as a long-distance
process. Stampe contends that the Latin case should nonetheless be
counted. Fortunately, the entire argument need not rest on the status
of Latin. As we reported in an update to our summary [vol. 3-560],
there are apparently two better examples of truly productive
long-distance dissimilation, recently brought to our attention by
David Odden and Wayles Brown: Meeussen's Rule and Georgian Liquid
Dissimilation.

However, we are puzzled by Stampe's arguments.

Citing Donegan (1978) and Jesperson, Stampe suggests that
dissimilation occurs between non-adjacent sounds when the intervening
sound is likely to assimilate to the flanking identical sounds. When
one of the flanking sounds dissimilates, it effectively blocks that
assimilation. Stampe says,

> If "distant" dissimilations of consonants protect the sounds they
> flank from complete assimilation, we wouldn't expect dissimilation
> either (1) if the flanked sounds are so many or so different from the
> flanking consonants as to resist assimilation, or (2) if there are no
> flanked sounds.

It seems to us that under this analysis long-distance dissimilation is
actually highly unlikely. How are we to reconcile Stampe's proposal
with the facts of Meuussen's Rule or Georgian Liquid Dissimilation?
In what sense are the flanked sounds (in caps) in Georgian
[phrANG-Uli] `French' (from /phrang-ur-i/) highly similar, so as to be
threatened by assimilation from the flanking r...r?

Stampe comments on Hurch's position that

> "productivity gradually decreases the
> more distant the [dissimilator] is from the [dissimilatee]". Good.
> But if adjacency favors dissimilation, why doesn't ll dissimilate in
> mellis (not *melris)?

Under his proposed account of dissimilation as a means of blocking
assimilation, Stampe argues that

>
> (2) we would NEVER expect dissimilation in mellis (*melris), where
> there is no flanked sound to protect from assimilation. (The only
> other reason for dissimilating two sounds is to keep them from
> assimilating to each other, but ll can't get any more similar!)

First of all, as we pointed out in a previous message, /mellis/ is not
a good example. It cannot undergo dissimilation simply because it does
not contain the suffix /-alis/. But leaving this aside,and leaving
geminate blockage aside (which may take place when in fact there is a
single long segment instead of two adjacent identical segments) there
certainly are cases where adjacent identical sounds dissimilate. A
couple of examples from Basque: the final /s/ of the negative particle
/es/ becomes [t] before another /s/; e.g.: /es-sara/ [etsara] 'you are
not'. In Biscayan Basque stem-final /a/ becomes [e] before the
singular determiner /a/: /alaba-a/ [alabea] 'the daughter'.
So, dissimilation involving adjacent identical sounds is attested.

We are unclear on another of Stampe's points. He states that,

> Another puzzle: Dissimilations, like other fortitions, apply in
> narrower prosodic domains only if they apply in wider domains.
> [...]
> This is the opposite of assimilations, which, like all lenitions,
> apply in wider prosodic domains only if they apply in narrower ones.

While this is certainly an interesting claim, it is not clear to us
how it is to be interpreted. What sort of ``prosodic domains'' does
Stampe refer to? From the examples it appears that a consequence of
the claim is the following: dissimilations affect unaccented nuclei if
they also affect accented ones, while the opposite is true for
assimilations. But if so, this claim has a number of clear
counterexamples; for instance Metaphony (the raising of a vowel under
the influence of a final high vowel- an assimilation) in many Romance
languages affects only stressed vowels. Thus, in the dialect of Lena
(Spain) /p'aSaru/ [p'eSaru] (cf/ plural [p'aSaros]) (S = prepalatal
voiceless fricative).

We would be grateful for clarification on any of these points.

Jose Ignacio Hualde (jihualdeux1.cso.uiuc.edu)
Jennifer Cole (colecogsci.uiuc.edu)

Univ. of Illinois, Urbana, IL 61801

Dr. Gladney response (3.569) to my posting on Polish voicing
assimilation (3.511) says that I was wrong to claim that ALL
voiced fricatives devoice after voiceless obstruents. However,
his example of dialectal /gzwo/ 'shirt', etymologically perhaps
from /kuzlo/ (with short /u/) is surely irrelevant, since this
form arose centuries ago, whereas my claims are about the modern
language (I could also quibble about the fact that, since this
belongs to some dialect which is quite foreign to me, it is
also irrelevant for that reason). On the other hand, the
evidence that I have (which involves Polish speakers mispronouncing
foreign words) does bear on the synchronic situation.

I might add yet another argument. Since Polish has a phonemic
voiceless velar /x/ (actually it is usually frictionless, so 'x'
is the wrong symbol, but that does not matter here) but lacks
(for most speakers) a corresponding voiced sound, a good test
for my hypothesis involves foreign words with clusters of
/x/ + a voiced fricative. According to an experiment I did
with several speakers, WITHIN the boundaries of a morpheme,
such clusters will be pronounced with progressive assimilation,
including orthographic (that's how forms were presented to
speakers) 'Achziw', rendered with /xs/.