Skip to main content

Full text of "Saib - Schwa Insertion in Berber (1976)"

See other formats

Monographic Journals of the Near East 
General Editor: Giorgio Buccellati 


Robert Hetzron, Santa Barbara 

Advisory Board: 

Ariel Bloch, Berkeley 
John B. Callender, Los Angeles 
Talmy Givon, Los Angeles 
Thomas G. Penchoen, Los Angeles 
Stanislav Segert, Los Angeles 

Volume 3 
Issue 4 

September 1976 

Schwa Insertion in Berber: 
Un Probleme de Choix 


Jilali Saib 

Dndena Publications 

Malibu 1976 


AAL includes contributions in linguistics within the vast domain of Afroasiatic (Hamito-Semitic) languages. 
Articles of general, theoretical interest using Afroasiatic material, descriptive, historical and comparative 
studies are included. 

Editor: Robert Hetzron (1346 San Rafael, Santa Barbara, Ca. 93109, U.S.A.) 

Advisory Board: A. Bloch, J. B. Callender, T. Given, T. G. Penchoen, S. Segert. 


MJNE is a system of journals on the Near East, with each journal devoted to a specialized study area, and each 
issue consisting of a single article. Current journals in the system are Afroasiatic Linguistics and Assur. 

General Subscription 

For a prepayment of $1 2.50 the subscriber selects random issues from within the entire system as desired, up 
to a total of 200 pages. The subscriber is also entitled to ( 1) periodical lists of abstracts from all journals in the 
system, and (2) reservation to any journal within the system, whereby issues of a given journal are sent on 
approval immediately upon publication (and may be returned within two weeks). 

Library Subscription 

A prepayment of SI 2.50 for each journal in the system secures all issues of a single volume as soon as they are 
published. This subscription schedule does not allow the selection of random issues; in return, a discount is 
provided in the form of a greater number of pages for the basic price of $12.50 (since a volume will normally 
include more than 200 pages). 

Library subscriptions are available to both institutions and individual scholars. 

Individual issues are numb 'red sequentially within each volume. Each issue has its own pagination. A volume 
is closed when a total of between 200 and 250 pages is reached. 

A title page and a table of contents listing all issues within each volume are sent to all subscribers at the close 
of a volume. 

Periodicity in the order of appearance of issues is not predetermined. A volume, however, is generally 
completed within one year. 

Institutional and Professional discount of 2 07c on single subscriptions (higher on larger orders). Payment must 
accompany orders from individuals. A handling fee of 80d will be charged to Libraries if order is not prepaid. 
Order from: UNDENA PUBLICATIONS, P.O. Box 97, Malibu, California 90265, U.S.A. 

(C) 1976 by Undena Publications. 

All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any 
means, electronic or mechanical, including photo-copy, recording, or any information storage and retrieval 
system, without permission in writing from the publisher. 

MonogtiaplUc Jou/maLi o& the. MeaA EaAt 

A&tioaA-LcutLe. LCngcuAtCcA 3/4 (September 1976) 



Jilali Saib 

Department of Linguistics 
University of California, Los Angeles 

The question of whether schwas are inserted or present in underlying 
representations of CpC 2 and CqC 2 C 3 verb stems in Berber is examined 
here in detail. This examination consists of proposing, motivating, 
and evaluating a number of analyses possible within the generative 
framework. It also considers recent proposals concerning "abstract- 
ness" and "concreteness" of underlying representations, and alterna- 
tive orderings of phonological rules. It is concluded that a "total- 
ly concrete" analysis — i.e. with underlying schwas — is superior 
to other less concrete ones, chief among them the "totally abstract 
analysis" — with consonants only. The implications of this conclu- 
sion are discussed and suggestions for further research are made. 






2.1. The totally abstract analysis 3 

2.2. The Randomly Ordered Rules Abstract Analysis 7 

2.3. The totally concrete analysis 8 

2.4. The intermediate analyses 11 



This paper grew out of research begun in the Fall of 1972. In preparing it I have bene- 
fited from insightful discussions with Professors Victoria Fromkin and Larry Hyman; I am 
grateful for their help and their encouragements. I would also like to thank Professors 
Bedell, Givon, Penchoen, Johnson, and Mickey Noonan for their comments on an earlier draft 
of this paper. Any errors or omissions are my sole responsibility, however. An abridged 
version of this paper was presented at the 5th Annual Conference on African Linguistics, 
Stanford University (March 1974) . 

AAL 3, 71 


J. Saib 

[AAL 3/4 


One of the most intriguing problems in Berber phonology concerns the status and behavior of 
schwa. As seen in (1) , 

(1) i u 


most Berber languages, spoken throughout North Africa, have a simple underlying vowel system. 
However, on the surface, a fourth vowel, 2 schwa, occurs with great frequency. In fact, certain 
verbs, which are given in their citation form 3 in (2) have only schwa: 

(2) [axSsm] 'to work' [ansj 'to wear' 

LazSay] 'to dwell' [dar] 'to go down' 

This fourth vowel differs from the three vowels in (1) in that its occurrence is restricted. 

As exemplified in (3), the other vowels occur in both closed and open syllables, while schwa 
occurs only in closed syllables: 4 

(3) closed syllable open syllable 

[as] 'to you' [sa] 'something' 

[as] 'to give' *[sa] 

Moreover, the exact position of schwas is totally predictable. Returning, for example, to the 
phonetic verb forms in (2), we note, as indicated in (4), that schwa occurs either before a 
word- final consonant or before two consonants (or a geminate): 5 


The question of whether this schwa arises historically through vowel reduction as is the 
case in a great many languages (e.g. classical Arabic kataba, Moroccan Arabic [aktab] 'he 
wrote'), or is an underlying vowel in its own right is a very interesting question but one 
which I will not take up here. The problem of schwa and all the questions related to it 
will be taken up in detail in my forthcoming Doctoral Dissertation. 

3 In Berber the Citation Form which happens to be the Imperative Singular, is the least morph- 
ologically marked form. The data presented here are given in broad phonetic transcription. 

A dot under a consonant indicates that that consonant is pharyngeal i zed. 

4 The data presented in this paper from Tamaziyt (Ait Ndhir). One Berber language to which 
this statement will not apply is Twareg, which according to De Foucauld [1920] later con- 
firmed by Prasse [1960:58-60], does exhibit instances of CaCV (e.g. te.n’dye 'act of killing', 
cf. Tamaziyt OTwyf) , and of CdCd (e.g. efeabaA 'hill'). The only instances of CaCV in 
other Berber languages (e.g. Tamaziyt x-ammm+'dn -* [xanmaman] 'they reflected on'), are 
actually not counter-examples. I have a principled explanation for them: the second d 
of the stem, which should have been deleted, is maintained to keep the m's apart. In my 
dialect, deletion takes place in fast speech and the result is the absorption of the third 
m by the geminate, yielding [xanman] with a strong geminate m. 

5 This is in accordance with my paper on geminates (Saib [1973]) where it is argued that, at 
least in Berber, geminates are to be treated as a sequence of two identical segments. A 
convention whereby they will be redundantly specified as [+tense] is also proposed. 

AAL 3, 72 



/ C # 

/ CC 

Schwa I nAeJvtion -in Bcabca 


Since the occurrence and exact position of schwa can be predicted, it is possible to maintain 
that there are no schwas in underlying representations for the surface verb forms in ( 2 ) . 
Instead, they would consist simply of three and two consonants, as in (5): 

(5) /x3m/ 'to work' /ns/ 'to wear' 

/z3y/ 'to dwell' /dr/ 'to go down' 

It is to be noted that the underlying forms in (5) are similar to those posited for verb stems 
in Semitic languages (cf. the treatment of Arabic and Hebrew verbs, for example.) In dealing 
with stems like those in (5), generations of Berberists (Laoust [1918], Basset [1929] and 
[1952], Abdel Massih [1968], and Penchoen [1973] simply adopted the Semitists' approach and 
regarded the consonants to be the "truss" of the steins. As Basset [1952:11] put it: "Un 
groupement exclusif de consonnes constitue le radical et, partant, 1 ' armature semantique du 

As for the schwas, they are said to be inserted according to consonant clustering, which is in 
turn based on the nature of the consonants. Thus, in /C 1 C 2 / stems, for instance, the following 
clusterings are observed (cf. Laoust [1918]): if Cq is [sonorant] , and C 2 is [+sonorant], 
clustering is prohibited, and a [a] is inserted between Cl and C 2 (e.g. /dr/ 'to go down is 
realized as [dar] not *[adr]: and /gn/ 'to sleep' as [gan] not *[agn]. if on the other hand 
Cq is [+son] and C 2 is not, clustering is allowed (e.g. /Is/ 'to wear' is pronounced [als] 
not *llas]; and /nz/ 'to be sold' [anz] , not *[naz|. 

To discuss the inadequacies of this approach would take us too far afield. Suffice it to say 
that: (i) consonant clustering varies from one Berber language to another, from one speaker 
to another, and even within the speech of one speaker; (ii) if followed strictly, this ap- 
proach will force one to end up, as did Abdel Massih [1968:25-26], with incorrect or, at 
.best, very questionable forms like these: *[nas] 'to spend the night', *[hadam] 'to demolish', 
* [farah] 'to be happy'. 

Returning now to the main discussion, we find that, in addition to positing underlying forms 
like those in (5) , there would have to be a rule (or rules) of schwa insertion operating in 
the environments given earlier in (4) . In the remainder of this paper I shall first explore 
the theoretical implications of the analysis just outlined, which I will call the "totally 
abstract" analysis. After pointing out certain objections one might raise against this ap- 
proach, I shall then present a second analysis which avoids the shortcomings of the first. 

But before continuing, a word on what is meant by "abstract" and "concrete" is in order. These 
terms are used in this paper to indicate the distance from / / to [ ] . Since the underlying 
forms in (5) are far removed from their surface realizations, an analysis which posits them 
would be "totally abstract." In contrast, an analysis which posits, as underlying forms, forms 
like those in ( 2 ) would be a "totally concrete" one, since they are exactly identical to their 
phonetic shape or representation. 


2.1. The Totally Abstract Analysis 

In the "totally abstract" analysis a rule like the one in ( 6 ) is needed to take us from ab- 
stract underlying forms like those in (5) to the surface forms given in (2 ) : 6 

6 Rule ( 6 ) is written in this way so as to account for all the so-called vowel-less verb 
stems, be they /C/ (cf. [as] in (3)), /C 1 C 2 C 3 /, or /( 4 C 2 / (cf. (5) above), though an extra 

AAL 3, 73 


J. Saib 

[AAL 3/4 

(6) 0 -> a/ {#, C] C W [j } i} 

(Condition: Directionality 7 of application: right-to-left, counting segments). 

Notice two things. First, the subparts of the rule must be extrinsically ordered, and the 
directionality of their application must be right- to- left, counting segments. A sample deri- 
vation is given in (7) : 

(7) /#xS m#/ 'to work' 

x3am (by 6i) 

axSam (by 6ii) 


As seen in (8a-b) , failure to observe the two conditions just mentioned leads to incorrect 
surface forms: 

(8) a. Wrong ordering, correct directionality: 

/#x 3 m #/ 

xaS m (by 6ii) 

xaSam (by 6i) 

* [xsSom] 

b. Right ordering, wrong directionality: 

/# x a m#] 

ax a m (by 6ii, 6i inapplicable) 

axaBm (by 6ii) 

axaSam (by 6i) 

* [axaSam] 

In (8a) an incorrect output is obtained because the subparts of the rule are applied in the 
wrong order but with the correct directionality. In (8b) the subparts of the rules are 
applied in the right order but with the wrong directionality. Again the output is incorrect. 

The second fact about this analysis is that it requires that we distinguish between two dif- 
ferent word- internal grammatical boundaries (+ and #) for the suffixes. As seen in (9), the 
+ boundary has no effect on the application of rule (6) : 8 

rule of schwa movement (or schwa hopping) is needed in order to obtain the correct phonetic 
output of certain /C]^/ stems especially those containing sonorants. That this is so, 
one might note, should not be too alarming, as these stems have been observed to behave 
strangely in other languages with respect to epenthesis. Mickey Noonan, now at San Jose 
State University, informs me (personal communication) that this was the case in Pre- Indo- 
European, as reconstructed by Lehmann [1952:112-113] (i.e. with Obstruents, Resonants, 
Laryngeals and a "non-segmental" phone /a/, syllabicity) . Nevertheless, certain /CpC 2 / 
stems remain a distinct problem for a "totally abstract" analysis. 

7 For a detailed discussion of Directionality and the "Directional theory," see Johnson [1970] 
and Howard [1973] and the references cited there. 

8 The complete list of the personal subject affixes is the following: (to be compared to those 
in (19)) 

AAL 3, 74 


Scluva Imention In Reuben. 


(9) /#x3m+ x#/ 'I worked' 

x S m a x (by 6i) 

xa5m ax (by 6ii) 

Lxasm a x] 

While the correct form for the first person singular is obtained in (9) , an incorrect output 
is observed in most dialects for the second person singular, as seen in (10) : 

(10) /#9+x3m+ 3#/ 'you (sing.) worked' (wrong boundary) 

0 x 3 m a3 
0 xa3m a3 
a 8 xa3m a3 
* [a0 xa3m a3] 

In order to obtain the correct output, a stronger boundary is required before the suffix /3/ 
and an even stronger boundary, ##, after it. 

(11) /#0+ x3m#3 ##/ 'you (sing.) worked' (correct boundaries) 

0+ x3an#3## (by 6i) 

9 x3am#3## (by 6ii) 

[0 ax3am3] 

Other suffixes which, likewise, require an internal word boundary and the stronger boundary 
## are the second and third person feminine plural given in (12) and (13) : 

(12) /# 0+x 3m+ m#0##/ 9 'you (fern, plur.) worked' 

0+x 3m+am#0## 

0+xa3m a #0## 

0+xa3m a #0## 

[a0 xa3mam 0] 


x 1. n 

2. 0 

2m. 0 


3m. -i 

2f. 0 


3f . 0 



3f . 


They are listed here without boundaries so as not to prejudice the issues. It should be 
noted however, that the /0/ in the plural suffixes is the feminine marker (cf. also the 
feminine noun formation: e.g. masc. amaz-iy 'Berber man'; fern. Q-amaz-Ly-Q 'Berber woman.') 

9 1he strong boundary ## in (11) - (13) is needed to insure that rule (6) will not insert a 
2 before the simple suffix /5/ and the /0/ of the complex suffixes /-m9/ and /-n9/ which 
will result in incorrect surface forms. While some dialects allow the insertion of 2 be- 
fore /s/ , no dialect allows such an insertion before the /0/ of the complex suffixes. 

AAL 3, 75 


(13) /# x3m+ n#0##/ 

x 3 m+on #0## 
xaSm an# 6## 

J. Saib 

[ML 3/4 

[X9 3ma n 6] 

'They (fem. ) worked' 
(by 6i) 

(by 6ii) 

Again, if a + boundary is posited instead of an internal word boundary, the wrong output is 
obtained as seen in (14) and (15) : 

(14) /# 0 +x3m+m+ 0#/ 

8 +x 3 m+m+a 0 
0 +x3am+m+a0 
8a +x3a m+m+a 0 
* [6 a x3 a mma 8 ] 

(15) /# x 3 m+ n+ 8#/ 

x 3 m+ n a 0 
x3am+ n a 8 
a+x3am+ n a G 
*[ ax3amna0J 

'you (fem. plur.) worked' 
(by 6i) 

(by 6ii) 

(by 6ii) 

'They (fem.) worked' 

(by 6i) 

(by 6ii) 

(by 6ii) 

Since the problematic suffixes involve [3] and [0] (i.e. [+anterior, +coronal, +continuent, 
-strident] segments) , one might argue that one way of solving the problem — and hence salvage 
the "Totally Abstract Analysis" — would be by stating the environment of rule (6) so as to 
exclude these segments. It should be noted, however, that this restriction would further 
complicate the already complicated environment of the rule. Moreover, a further restriction 
making use of + would be required, viz., 

c f # 1 W 

"iF'sj (ii) 

as Berber has verb stems ending in /3/ and /9/ (e.g. [anya3] 'to become finer'; [awwa0] 'to 

But although this abstract analysis can be made to account for the data, it leaves one with 
a great deal of uneasiness. Indeed, several objections can be raised against it. For one 
thing, it posits unnatural and unpronounceable underlying forms, like those in (5), for the 
stems given in (2) . One might rightly question the necessity of positing such abstract under- 
lying forms since Berber has other verb stems with vowels (e.g. [fafa] 'to wake up', (-trans.) 
[a3u] 'to fold'). Thus, the underlying forms in (2) seem to break the pattern of Berber 
morphemes. 10 However, even if one could derive all vowels in all verb stems in the language, 

10 It might be argued, of course, that it is these stems which break the pattern. However, 
the statistics bear my statement out. According to De Foucauld, quoted in Basset's Intro- 
duction, [1929], out of a total of 1400 Twareg verbs, 400 stems are of the /CpC 2 C 3 / type, 
while 50 stems are of the /CpC 2 / type. This is including borrowings from Arabic, most of 
which are of the first type. Moreover, a great many verbs belonging to the basic core 
vocabularly have plain vowels. 

AA L 3, 76 


Schwa 1 nicotian in BcAbco. 


we might still question the motivation for setting up such abstract underlying forms. In 
addition, the assignment of different boundaries to certain affixes is in itself an incon- 
sistency which makes this analysis look very ad hoc. It is ad hoc, first, because of the 
arbitrariness of boundary assignment. There is no grammatical basis for assigning a + bound- 
ary in the case of certain personal affixes, but internal word boundaries for other personal 
affixes. In addition, this boundary assignment is circular: there is no other phonological 
process rule in the language which distinguishes between a + boundary and an internal word 
boundary #, as would the rule of schwa insertion in this "totally abstract" analysis, nor is 
there a rule calling for such a liberal use of boundaries. 

Moreover, the "totally abstract" analysis, as was pointed out above (cf. fn. 6) does not 
handle certain /C\C2/ stems without difficulty. If the correct phonetic output of verbs like 
[j-an] 'to sleep' and [say] 'to buy' can be obtained by the simple application of the first 
part of rule (6) to underlying /jn/ and /sy/ it is not so for others; viz. /nz / 'to be sold' -> 
[anzj , not * [naz] ; /rs/'to settle' -» Tars], not *[ras]; /z3/ 'to grind' -» [az5], not *[za3]. 

An extra rule of "schwa movement" or "schwa hopping" — the environment of which is highly 
restricted — is needed in order to get the correct surface forms for these verbs. Such a 
minor rule is clearly ad hoc since its sole function is to "fix up" incorrect outputs yielded 
by the application of (6i) to certain /C;^/ stems. Its ad hoc-ness adds to the cost of the 
"totally abstract" analysis. Alternatives to the addition of an extra rule, for example 
marking these stems with a [-Rule X] feature (in this case (6i)) in the lexicon, are not sat- 
isfactory either, as they are just as ad hoc and just as costly. 

Finally, even the extrinsic ordering constraint 11 can be dispensed with, as is discussed be- 
low, and can be replaced by a random sequential application of the subparts of rule (6). This 
is particularly interesting in that it suggests that, at least in some cases, the No-Extrinsic 
Ordering Hypothesis currently advocated by some phonologists (Vennemann [1972]; Hooper [1973], 
Koutsoudas et al [1974]), can be a viable alternative. 

Thus, there exists an equally abstract alternative to the "totally abstract analysis A," which 
I will call "The Randomly Ordered Rules Abstract Analysis." 

2.2. The Randomly Ordered Rules Abstract Analysis 

In this analysis the subparts of the rule of schwa insertion given in (6) can apply randomly 
(i.e. unordered) provided there is, in addition, a "corrective" rule of schwa deletion as 
stated in (16) : 

(16) a -* 0 / CV 

As seen in (17) , the same underlying form uniquely yields the correct output by application of 
rules (6) and (16) in any order and in any direction: 

(17) a. /#0+x3m-t m#9##/ 
0+x 3 m-tam #0## 
a0+x 3 m+am#0## 
[a0 xa 3 ma m 0] 

'You (fern, plur.) worked' 
(by 6i) 

(by 6ii) 

(by 6ii) 

“Talmy Givon (personal communication) suggested to me that another possible formal way of 
achieving this would be by giving the # in the right hand side of the environment the 
value of a consonant. What would follow from this suggestion is that the closed syllable 
would be CC or C#. I have not had the time to evaluate this suggestion but will in- 

vestigate its consequences in my dissertation. 

AAL 3, 77 

8 J. Saib [ AAL 3/4 

(In (17a.) the rules are applied first to the suffix, then to the prefix, then to the stem.) 

(17) b. /#0 +x3m+ m #0## 

0 +x3am+ m #0## 

(by 6ii) 

0a + x3am+ m #0## 

(by 6ii) 


(by 6i) 

0a+x S m+am#0## 

(by 16) 


(by 6ii) 

0 +xaSm+am#0## 

(by 16) 

a 6 +xaSm+am#0## 

(by 6ii) 

[aG xaS ma m 0J 

(In (17b.) the rules are applied to the stem, then to the prefix, then to the suffix.) 

It should be pointed out that the rule stated in (16) , though it was referred to as a "correc- 
tive rule," is actually not an ad-hoc one. It simply takes advantage of the general constraint 
on the restricted occurrence of schwas exemplified in (3) . It guarantees that no schwas will 
occur in an open syllable. 

But this alternative analysis, though it does not require extrinsic rule ordering, still suf- 
fers from the other pitfalls of the "totally abstract" analysis, namely the abstractness of 
the underlying forms and the ad hoc use of boundaries. This brings us to another alternative 
analysis, this time a "totally concrete" one. 

2.3. The Totally Concrete Analysis 

The "abstract" analyses A and B just presented can be said to be motivated by a principle which 
plays an important role in the S.P.E. type of phonology (Chomsky and Halle [1968]): the 
principle of eliminating, as much as is feasible, redundancy from the lexicon. That is, 
things which can be predictable by rule should not be included in the underlying fonns. How- 
ever, we have just seen certain difficulties which can result from the strict application of 
this principle to the Berber schwa situation. 

Now what about the "totally concrete analysis"? This analysis is, likewise, motivated by 
general principles. First, the phonetic shape of the least morphologically marked form tends 
to be chosen as the underlying form (cf. Greenberg [1966]; Vennemann [1972]). Second, the 
principle of minimization of the difference between underlying representation and phonetic 
representation (i.e. redundancy does not matter) (cf. Hale [1971], as reported in Kiparsky 
[1971] ; Vennemann [1972J and [1973]; Hooper [1973]). Third, the principle of No-Extrinsic 
Rule Ordering (cf. Koutsoudas et al [1971]; Vennemann [1972]; Hooper [1973]). Thus, according 
to this concrete approach, forms like the ones given in (2) will appear as given in (18): 

(18) /axSam/ 'to work' [x3am] 

/azSay/ 'to dwell' [azSay] 

/ans/ 'to wear' [ons] 

/dar/ 'to go down' [dar] 

(i.e. identical to their phonetic shape). 

As seen in (19), affixes will appear with vocalic support. (As for the reason why /-a/ does 
not appear with a a, cf. £n. 9). 

AAL 3, 78 


Schwa. I nAcrvUan In BeAbcn. 






+ 3X 

1. H3+ 


03+ + 3 

2m. 03+ 



X, + 

2f . 03+ 

+ 3 m+9 

3f . 



+3 n 

3f . 

+ 3W+0 

(Where m = masculine; and fa = 

feminine . ) 

It should be pointed out that by positing the forms in (19), I am not replacing boundaries 
with schwas so as to avoid problems. The positing of affixes with schwas is in accordance 
with the second general principle motivating the "totally concrete" analysis. Moreover, a 
look at the Berber data presented here as well as other data will indicate that, at the sur- 
face level, affixes — with the exception of /-a/ (cf. fn. 9 however) — appear with vocalic 
support . 

For the 

sake of reference, the full preterit conjugation of the verb 'to work' is given in (20) 












03+3X3 jm+ 3 




i+ 3X5 am 


[ ixSam] 

3f . 

03+ 3X3 3m 

3f . 




n 3+3XS jm 






[ aOxaBmam] 

2f . 


2f . 

[ a 0xa3mam0] 




[xa 3man] 

3f . 


3f . 


In (20) , we see that most forms require no more than the addition of the affixes and the low 
level phonetic rule of vowel deletion (or absorption) given in (21) : 12 

(21) 3 -* 0 / V 

Only a few forms require the application of the major rules given in (22) and (23) : 

(22) a -» 0 / — C V 

12 It should be pointed out that rule (12) is by no means an ad-hoc rule as it is found to be 
operative in other languages. It is similar, for instance, to the elision of a schwa in 
front of another vowel in French (e.g. As# aRbR/ -* [laRbR] 'the tree'.) 

ML 3, 79 


J. Saib 

[AAL 3/4 

(23) 0 -» a / --- C, C V 13 

(Again, geminates = C X C X , i.e. a sequence). 

Given the underlying forms in (18) and rules (21) , (22) and (23) , the derivation of the 
phonetic forms for the first person singular and the second person feminine plural would be 
that given in (24) : 



/#0a+ax3am+ m+0#/ 

ax S max 

(by 22) 

9a+ax a ma m 8 

(by 22) 


(by 23) 

0 ax S mam 0 

(by 21) 


(by 22) 

0 axaSmam 0 

(by 23) 


0 xaSma m 0 

(by 22) 

a0 xa3mam 0 

(by 23) 

[a0 xa3 ma m 0] 

It should be noted that: first, unlike analyses A and B, the "totally concrete" analysis does 
not require the use of morpheme boundaries for only certain affixes, a practice which we have 
indicated to be very questionable. Second, unlike analysis A, it does not require extrinsic 
rule ordering. Third, the rules it requires are both natural and needed in the grammar any- 
way (especially (23)). And fourth, unlike the other analyses, it does not run into any dif- 
ficulty with respect to the /CqC 2 / stems (cf. fn. 6 and the discussion on p. 7) : they are 
simply listed in the lexicon with their vowel in its appropriate place (cf. (2) and (18)). 

Two consequences follow from the "totally concrete" analysis. First, it requires redundant 
information in the lexical representation of formatives, since it has been shown that one 
can predict the phonetic occurrences of schwas. If this is the correct solution, it implies 
that speakers, in learning the grammar of their language, do not make all the possible gen- 
eralizations which the data would permit. That is, it implies that rather than constructing 
very complex rules the speaker will store formatives with underlying segments which could be 
predicted. It may also imply that, instead of storing complicated rules, native speakers 
learn patterns; which is what traditional grammarians have said all along. 11 * he do not at 
present have any ways of testing this hypothesis; nor does the proposed evaluation metric 
permit us to fully compare complexity of rules in relation to complexity in the lexicon. 

Second, this analysis shows that given less abstract representation extrinsic ordering is not 
necessary. The current debate regarding the No-Extrinsic Rule Ordering Hypothesis is far 
from settled (cf. Campbell [1973]). It is still not clear how one is to evaluate the conse- 
quences of constraining the ordering of rules in a grammar. It is clear, however, that where 
extrinsic ordering is disallowed, certain generalizations concerning predictable surface forms 
are lost, at least in this Berber case. 

The theoretical implications of the analyses of schwa in Berber are summarized in the chart 
given in (25) : 

1 3 Notice that since the environment is not stated as in (6) no directionality is needed. 

14 An attempt to test whether Berber speakers store formatives like those in (2) as given in 
(5) or as given in (18), or whether they learn patterns, is envisaged for my forthcoming 
Doctoral Dissertation. 

AAL 3, 80 


Sc-hm. 1m esttion 

in Be.tibe.Ji 


Analysis A 

Analysis B 

Analysis ( 

Lexical Redundancy 




Rule Redundancy 




Rule Ordering 





R to L** 



(Where: * Intrinsic in analysis B = random sequential ordering 
** R to L: right to left (extrinsic)) 

There are further problems, since, in addition to the "totally abstract" and "totally concrete" 
analyses there are demonstrably a myriad of "compromise" of "intermediate" analyses which lie 
logically in between the two extreme analyses. This is in itself a problem for the theory 
since one of its basic stated goals is to seek and select a unique analysis (or solution) . 

Once again, Chomsky and Halle's Evaluation Measure is put to task. And, as stated above, it 
does not provide us with an effective way of selecting the unique solution. 

2.4. The Intermediate Analyses 

The various factors can be permuted to yield intermediate analyses are listed in (26) : 

(26) a. limiting underlying schwas in one way or another, e.g. (i) /xSam/ instead of /x3m/; 

(ii) have schwas only in grammatical morphemes; (iii) have schwas in the stem but 
not in grammatical morphemes, etc. 

b. using versus not using boundaries of different kinds for grammatical morphemes 
(i.e. affixes here.). 

c. permitting extrinsic rule ordering versus permitting only intrinsic rule ordering. 

d. using directional rules versus non-use of such rules. 

An example of an intermediate analysis (or rather strategy) is given in (27) : 

(27) a. underlying form of the stem: /x3am/ 'to work'. 

b. underlying form of affixes: with schwas, except /a/ , etc. 

c. rules: (i) a -> 0 / C V 

(ii) 0 -» 2 / C C V 


/#xSom+8x/ -» x3 m^x -» [xoSmax] (by 27i and 27ii). 

Another example of an intermediate strategy is the one given in (28) :* 5 

1 5 This strategy was suggested to me by Ian Maddison of U.C.L.A. I am grateful to Ian and to 
the participants in Prof. Fromkin's seminar for providing me with needed feedback at the 
initial stages of the writing of this paper. 

AA1 3, 81 


J. Saib 

[Ml 3/4 


underlying form: 

/xaBam/ 'to work'. 


affixes with schwas, except /5/. 


rule: a -* 0 / 

C V — directionality 




/#xa3am+ax#/ -* 


right- to- left. 


In conclusion, we have seen that the problem of acIvm in Berber can be accounted for by a 
number of different analyses. This is not surprising and it clearly shows that merely ac- 
counting for the data, while necessary, is not a sufficient criterion for deciding between 
alternative solutions. Clearly one seeks some principled basis which must be determined by 
the theory. It should not be both a problem of accounting for schwa in Berber as well as 
"un probleme de choir.” Both the "totally abstract" and the "totally concrete" analyses 
were motivated by general principles. But we are still faced with the question as to which 
of these principles are to be accepted as theoretical constraints. 

Internal formal considerations are not enough to decide. Further work in Berber as well as 
other languages, and further comparison of such analyses seems to be required before we can 
agree on the theoretical constraints. In addition to formal comparisons it appears that we 
must seek external verifications of the theory which will depend on such "performance" factors 
as perceptual strategies, storage capacity of the brain, mental generalizing abilities, etc. 
Until we find such evidence the linguist is left with choosing the solution which appears to 
be the most intuitively satisfying. I find the concrete solution more satisfactory in this 
case because of the ad-hoc nautre of the abstract analysis. But using this criterion is not 
really satisfactory. Clearly, the theory is still deficient. Hopefully, descriptive work 
such as is outlined here will provide some of the necessary input required for revising and 
further constraining phonological theory. 


Abdel Massih, Ernest T. 1968. Tamazight Verb Structure: a generative approach.. Blooming- 
ton: Indiana University Publications. 

Basset, Andre. 1929. La longue berbere, morphologic, le Paris: Leroux. 

. 1952. La langue berbere. Oxford: International African Institute. 

Campbell, Lyle. 1973. "Extrinsic ordering lives," (Ms.) Bloomington: Indiana University 
Linguistics Club. 

Chomsky, N. and M. Halle. 1968. The Sound Pattern oft English. New York: Harper and Row. 

De Foucauld, (Rev.). 1920. Notes pour servir a un essai de grammaire touaregue. Algiers: 


Greenberg, Joseph. 1966. Language Universals . The Hague: Mouton. 

Hale, Kenneth. 1971. "Deep-surface canonical disparities in relation to analysis and change: 
an Australian example," (Ms.) To appear in Current Trends -in Linguistics , Vol. XL, 
ed. by T. Sebeok et al. The Hague: Mouton. 

Hooper, Joan. 1973. AApectA o< J Natural Generative Phonology. Unpublished doctoral dis- 
sertation, University of California, Los Angeles. 

Howard, Irwin. 1973. A Directional Theory o & Rule Application in Phonology. Bloomington: 
Indiana University Linguistics Club. 

AAL 3, 82 


Schwa Lniefvtlon In 


Johnson, Douglas. 1970. Eofumal Aspects o Phonological Dei cAA.ptA.on, in Pfioject on Linguistic 
Analysis, No. 11. UC Berkeley: California, published again in 1972, as Monographs 
on Linguistic Analysis, 3. The Hague: Mouton. 

Kiparsky, Paul. 1971. "Historical Linguistics" in A SuAvey Linguistic Science, ed. by 
W.O. Dingwall. College Park, Md. Linguistics Program, University of Maryland. 

Koutsoudas, Andreas, et alia. 1974. "The application of phonological rules," Language 50: 

Laoust, Emile. 1918. Etude sua le dlalecte del Ntlfia. Paris: Leroux. 

Lehmann, Winfred P. 1952. PAoto-Jndo-EuAopean Phonology. Austin: The University of Texas 
Press, and Linguistic Society of America. 

Penchoen, Thomas G. 1973. TamazLy-t 0 |J the A yt NdhlA. In Afroasiatic Dialects, Vol. 1, 
Berber, ed. by W. Leslau and T.G. Penchoen. Los Angeles: Undena Publications. 

Prasse, Karl. 1960. "Notes sur la langue touaregue." Acta Ofilentalla. Copenhagen. Vol. 
XXV: 43-111. 

Saib, Jilali. "The treatment of geminates: evidence from Berber," (Ms.) Paper read at the 
Annual Meeting of L.S.A. Dec. 1973. 

Vennemann, Theo. 1972. "Rule inversion," Lingua 29: 209-242. 

1973. "Phonological concreteness in natural generative grammar." (Ms.) 


AA L 3, 83 

Editors: Giorgio Buccellati, Marilyn Kelly Buccellati, Piotr Michalowski 

These two series make available original documents in English translation ( Sources ) and important studies by 
modern scholars {Monographs) as a contribution to the study of history, religion, literature, art and archaeology 
of the Ancient Near East. Inexpensive and flexible in format, they are meant to serve the specialist by bringing 
within easy reach basic publications often in updated versions, to provide imaginative educational outlets for 
undergraduate and graduate courses, and to reach the interested segments of the educated lay audience. 

General Subscription — For a prepayment of $10 the subscriber selects random issues from within the entire 
system as desired, up to a total of 200 pages. The subscriber is also entitled to (1) periodical lists of abstracts 
from both series, and (2) reservation to given categories to be specified by the subscriber (e.g. Assyriology or 

Library Subscription - the subscription price is $8 for Volume 1 of either the Sources or the Monographs. 

A volume will average 200 pages. Periodicity in the order of appearance of fascicles is not predetermined, but 
a volume will normally be completed within one year. 

Sources from the Ancient Near East 
Volume 1 

• 1. The Akkadian Namburbi Texts: an Introduction. By R. I. Caplice. 24 pp., $1.00 

• 2. Balag-Compositions: Sumerian Lamentation Liturgies of the Second and First Millennium B.C. 

By M. E. Cohen. 32 pp., $2.40 

In preparation: 

o The Epic of Erra. By L. Cagni. 

Monographs on the Ancient Near East 
Volume 1 

• 1. The Sumerian Temple City. By A. Falkenstein. Introduction and translation by M. Dej. Ellis. 21 pp., 95^ 

• 2. Three Essays on the Sumerians. By B. Landsberger. Introduction and translation by M. Dej. Ellis. 18 pp., 90^ 

• 3. Structure of Society and State in Early Dynastic Sumer. By I. B. Diakonoff. Introduction by M. Desrochers 

16 pp., 65^ 

• 4. The Conceptual Autonomy of the Babylonian World. By B. Landsberger. Translation by Th. Jacobsen, 

B. Foster and H. von Siebenthal. Introduction by Th. Jacobsen. 16 pp., $1.10 

In preparation: 

o Ebla in the Period of the Amorite Dynasty and the Dynasty of Agede: Recent Archaeological Discoveries 
of Tell Mardikh. By P. Matthiae. 

o Cuneiform Texts of the 3rd Millennium in Old-Canaanite from the 1974 Season at Tell Mardikh = Ebla. 

By G. Pettinato. 

Volume 2 

o Tales of the Old Gods and Other Essays. By T. Jacobsen. 


A series of grammars providing concise descriptions of individual languages within the Afroasiatic family, and 
directed to scholars and students in the given language areas as well as in linguistics. 

Editors: Wolf Leslau and Thomas G. Penchoen 

• Volume 1 (Berber). Tamaz ight of the Ayt Ndhir. By Thomas G. Penchoen. 124 pp. $8.50. 

• Volume 2 (Ancient Egyptian). Middle Egyptian. By John B. Callender. 150 pp. $10.00. 

In preparation: 

o Volume 3 (Semitic). Damascus Arabic. By Arne Ambros. 


Primary sources and interpretive analyses for the study of Mesopotamian civilization and its influences from late 
prehistory to the end of the cuneiform tradition. 

• Volume 1. Old Sumerian and Old Akkadian Texts in Philadelphia Chiefly from Nippur. 

Part 1. Literary and Lexical Texts and the Earliest Administrative Documents from Nippur. 

By Aage Westenholz. xii-210 pp., 3 plates. $18.50 (hardbound), $12 (softbound). 

• Volume 2. Babylonian Planetary Omens. Part 1. The Venus Tablet of Ammisaduqa. 

By Erica Reiner in collaboration with David Pingree. iv-60 pp. $6.75 (softbound). 

• Volume 3. Inscriptions from Al-Hiba-Lagash: The First and Second Seasons. 

By Robert D. Biggs, vi-45 pp., 2 plates. 

In preparation: 

o The Legacy of Sumer. Edited by D. Schmandt-Besserat. 

o The Economic Role of the Crown in the Old Babylonian Period. By Norman Yoffee. 
o Seals and Sealings in Ancient Mesopotamia. By McGuire Gibson and Robert D. Biggs. 

o Old Sumerian and Old Akkadian Texts in Philadelphia. By Aage Westenholz. 

Part 2. Late Sargonic Administrative Texts from Nippur: ‘Akkadian’ Texts, Legal Documents 
and the Onion Archive. 

Part 3. Late Sargonic Administrative Texts from Nippur: Ration Lists and Miscellaneous Accounts. 

Part 4. Presargonic and Sargonic Administrative Texts from Fara and Ur. 


Editors: Wolf Leslau (University of California, Los Angeles): Semitic and Cushitic 

Thomas G. Penchoen (University of California, Los Angeles): Berber 

Advisory Board: Giorgio Buccellati (University of California, Los Angeles): Akkadian 

John Callender (University of California, Los Angeles): Ancient Egyptian 
Russell G. Schuh (University of California, Los Angeles): Chadic 
Stanislav Segert (University of California, Los Angeles): Northwest-Semitic 

Afroasiatic Dialects ( AAD ) seeks to provide concise descriptions of individual languages which belong to the Afroasiatic 
language family. It is primarily directed toward an audience consisting, on the one hand, of students of one or several Afro- 
asiatic languages, and, on the other, of students of linguistics. In these volumes, both these groups should find succinct 
treatises such as to provide familiarity with the basic structure of the language in question in a comparative perspective. 
Each description will be comprehensive in scope and sufficiently detailed in exemplification. But at the same time the aim 
will be to cut through to the essentia] and to avoid specialized argumentation. The goal then is neither to publish a corpus of 
exhaustive reference grammars nor to provide a platform for the analytical defense of theoretical questions. In this sense the 
series is properly data-oriented. Though the authors will necessarily be of a variety of theoretical persuasions and each will 
have his own set of preferences for presentation, not the least important goal will be to achieve as high a degree as possible 
of uniformity in structure, and in the conventional signs and terminology used. This being accomplished, the reader should 
have no difficulty in finding points of resemblance and divergence amongst the languages which concern him with regard to 
some point of inquiry. The term ‘dialects’ in the series’ title refers not only to modem spoken vernaculars but to historically 
definable stages of any language of the various branches. Publication of studies of as many such dialects as possible would 
provide, we feel, both an encouragement to comparative work and a sound documentary base on which alone this work may 
fruitfully progress. 

AAD 1 - Berber: TAMAZIGHT OF THE AYT NDHIR by Thomas G. Penchoen. 1973, IV-124 pp., $8.50. 

The Ayt Ndhir dialect which is described belongs to one of the major Berber languages, Tamazight, spoken in the Middle 
Atlas Mountains of central Morocco. The description is based in the main on research undertaken with native speakers of 
the Ayt Ndhir territory surrounding El Hajeb. — While directed to the non-specialist, a number of points in the description 
proper will be of interest to the specialist as well: the presentation of noun and verb morphology points up a number of 
regularities which more often than not have been obscured in previous descriptions. Also, phonological rules are given which 
account for the major share of morphophonemic complexities. The reader will find in the appendices and ‘optional’ sections 
conjugation tables of typical verbs— including detailed observations on the placement of shwa in verbs—, a chart showing the 
main morphological patterns involved in verb derivation, a description of the phonological rules applying in complex 
sequences of morphemes of the verb group, the ‘basic’ vocabulary contained in several well-known lexicostatistic word lists, 
and a chart of the Tifina 7 alphabet used by the Tuareg. 

AAD 2 - Ancient Egyptian: MIDDLE EGYPTIAN by John Callender. 1975,150 pp., $10. 

This grammar deals with the literary language used in Egypt from ca. 2000 to 1200 B.C. and considered in even later times 
to be the classical written form of Egyptian. The book is directed toward the general linguist as well as the Egyptologist; 
examples are glossed and written in transcription and there is an index of grammatical terms and Egyptian morphemes. A 
comprehensive set of paradigms of both verbal and non-verbal predicate types is included as an appendix, together with an 
appendix on negation and one on the historical origin of certain constructions. — The grammar contains three main parts: 
phonology, morphology, and syntax, of which the last receives most emphasis. The section on phonology sketches the laws 
of sound change to the extent they can be discovered. The section on morphology stresses the paradigmatic character of 
verb tenses and their derivations. A distinction is made between truly paradigmatic tenses and tenses borrowed from Old 
Egyptian for quotations or special effect. Following Polotsky, the “emphatic forms” are treated as nominalizations under 
the rubric “manner nominalizations.” Unlike previous grammars of Egyptian, this grammar discusses syntax according to 
transformational categories. The process of “clefting” interrelates emphatic forms, the “participial statement” and 
constructions with pw+ relatives. The process character of negation is emphasized, and the implications of so considering it 
are developed in a special appendix. A sample text is also included, accompanied by a vocabulary and a translation. 

AAD 3 - Semitic: DAMASCUS ARABIC by Arne Ambros. In preparation. 

All prices are postpaid. Payment must accompany orders from individuals. A handling fee of 80tf will be charged to libraries 
if order is not prepaid. Discount of 20% on all orders received within one year of publication date. 

Order from: UNDENA PUBLICATIONS, P.O. Box 97, Malibu, California 90265, U.S.A. 


AAL includes contributions in linguistics within the vast domain of Afroasiatic (Hamito-Semitic) languages. 
Articles of general, theoretical interest using Afroasiatic material, descriptive, historical and comparative 
studies are included. 

Editor: Robert Hetzron (1346 San Rafael, Santa Barbara, Ca. 93109, U.S.A.) 

Advisory Board: A Bloch, J. B. Callender, T. Givon, T. G. Penchoen, S. Segert. 

Volume One 

1. P. Newman and R. G. Schuh, The Hausa Aspect System, 38 pp. 

2. J. L. Malone, The Development of the Anomalous Syriac Verb eskah ‘To Find’: 

A Case of Convergent Factors in Linguistic Change, 10 pp. 

3. R. Hetzron, Extrinsic Ordering in Classical Arabic, 25 pp. 

4. T. Givon, Verb Complements and Relative Clauses: 

A Diachronic Case Study in Biblical Hebrew, 22 pp. 

5. T. M. Johnstone, Die Modern South Arabian Languages, 29 pp. 

6. B. W. Andrzejewski, Indicator Particles in Somali, 69 pp. 

7. H. Minkoff, Graphemics and Diachrony: Some Evidence from Hebrew Cursive, 16 pp. 

Volume Two 

1. D. R. Cohen, Subject and Object in Biblical Aramaic: 

A Functional Approach Based on Form-Content Analysis, 23 pp. 

2. C. D. Johnson, Phonological Channels in Chaha, 13 pp. 

R. Hetzron, Die t-Converb in Western Gurage (Die Role of Analogy’ in Historical Morphology), 12 pp. 

3. A. Barnea, Reference to Time, Space and Other Types of Quantification 

in the City Dialect of Gaza, 10 pp. 

R. Nir, Die Survival of Obsolete Hebrew Words in Idiomatic Expressions, 1 pp. 

4. C. T. Hodge, Die Nominal Sentence in Semitic, 7 pp. 

G. Janssens, Die Semitic Verbal Tense System, 6 pp. 

5. S. Segert, Verbal Categories of some Northwest Semitic Languages: A Didactical Approach, 12 pp. 

6. J. B. Callender, Afroasiatic Cases and the Formation of Ancient Egyptian Constructions 

with Possessive Suffixes, 18 pp. 

7. J. L. Malone, Systematic vs. Autonomous Phonemics and the Hebrew Grapheme Dagesh, 17 pp. 

8. T. Givon, On the Role of Perceptual Clues in Hebrew Relativization, 17 pp. 

9. A. D. Corre, Waw and Digamma, 1 pp. 

Id., A Suprasegmental Feature of Length in Semitic, 6 pp. 

10. Bibliographical Bulletin, 61 pp. 

Volume Three 

1. R. G. Schuh, Die Giadic Verbal System and Its Afroasiatic Nature, 14 pp. 

2. G. Buccellati, On the Akkadian “ Attributive ” Genitive, 9 pp. 

Id., The Case Against the Alleged Akkadian Plural Morpheme -ami, 3 pp. 

3. R. Hetzron, Die Agaw Languages, 45 pp. 

4. J. Saib, Schwa Insertion in Berber: Un Probleme de Choix, 13 pp. 

Subscription to one volume of about 200 pp.: SI 2.50. 

Individual issues available separately. 

Descriptive flyers, with complete list of abstracts and prices for individual issues, are free on request. 
Write to: UNDENA PUBLICATIONS, P.O.B. 97, Malibu. California 90265, U.S.A.