LINGUIST List 13.1815

Fri Jun 28 2002

Sum: Chain Shifts with Epenthesis or Deletion

Editor for this issue: Marie Klopfenstein <>


  1. Elliott Moreton, Chain shifts with epenthesis or deletion

Message 1: Chain shifts with epenthesis or deletion

Date: Wed, 26 Jun 2002 16:55:35 +0000
From: Elliott Moreton <>
Subject: Chain shifts with epenthesis or deletion

On April 15th, 2002, we posted the following query 

> We are searching for examples of synchronic or diachronic chain
> shifts (i.e., counterfeeding rule interactions, such that /A/->[B] while
> /B/->[C]) in which segments are inserted or deleted.

In the two months since, we have received one response. Michael Becker, 
of the Tel Aviv University Linguistics Department, offered one diachronic 
and one synchronic example from Arabic:

> In Egyptian Arabic, and in other dialects of Arabic,
> the glottal stops of Classical Arabic were deleted in
> the coda, with compensatory lengthening word
> internally:
> Classical Arabic > Egyption Arabic
> fa?r > fa:r 'mouse'
> bi?r > bi:r 'well'
> xubara:? > xubara 'experts'
> wuzara:? > wuzara 'ministers'
> (more on word-final vowels below)
> Now, /q/ went to /?/ across the board in Egyptian
> Arabic, so you have:
> faqr > fa?r 'poverty'
> saqf > sa?f 'roof'
> da:q > da:? 'he tasted'
> fura:q > fura:? 'separation'
> So you get fa?r > fa:r and farq > fa?r.
> Another possible case:
> Synchronically, in Egyptian arabic, word-final vowels
> are always short. However, word-final /h/s delete and
> leave a word-final long vowel behind:
> ?e: 'what?'
> le: 'why?'
> gato: 'cake' (from French), plural gato:ha:t
> The underlying /h/ of '?e:' shows up in the phrasal
> level:
> ?e:h ilkala:m da 'what is this?'
> In a rule based model, you will have:
> 1. V --> [-long] / _#
> 2. h --> 0 / _#

I.e., V:h# -> V:# -> V#. There is at least one segmental deletion, 
that of the /h/, and a second if the V: is analyzed as VV. 

We are interested in epenthesis and deletion in synchronic chain 
shifts because we have theoretical reasons to believe that chain 
shifts involving both epenthesis and deletion -- AxB -> AB -> AzB 
-- are not possible. We have examined 35 (purported) chain-shift 
cases without finding one. 

However, we also expect chain shifts with only deletion 
(AxyB->AyB->AB) or only epenthesis (AB->AyB->AyzB) to be 
possible. So far we have found 3 clear examples of the former 
(in Catalan, Chemehuevi, and Hidatsa), but none of the latter. 
Most chain shifts apparently do not involve epenthesis or 
deletion at all.

For a fuller discussion (including the 35 chain shifts), 
please see our paper

Moreton, E., and P. Smolensky (2002). Typological 
consequences of local constraint conjunction. To appear in: 
L. Mikkelsen and C. Potts (eds.), _Proceedings of the West 
Coast Conference on Formal Linguistics (WCCFL) 21_. 
Cambridge, MA:: Cascadilla Press.

A slightly expanded version is on the Rutgers Optimality Archive:

Many thanks,
Elliott Moreton
Paul Smolensky
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue