Beruflich Dokumente
Kultur Dokumente
S,tructures morphologiques
Ir~ ~ at~~ ~~
d. IMS/'6J. ~t2t Je
de l'amazighe
et.
CIIL
~O~
i&ial-=
A c:Jk-~A~ift,ff-'U{UI3!j
Enterprlse'"
Nourddine AMROUS
Mohamed V University
Faculty of Education
ABSTRACT
The Tarifiyt variety of Moroccan Amazighe and its Tashlhit cogna te diverge with respect ta
the surface manifestation of the initial vowel of masculine and feminine no uns. This paper
daims that this divergence obtains only at the surface level and sa is not likely ta restrain
standardization. It will be argued that Tarifiyt ranks root faithfu/ness higher than affix
faithfulness while Tashlhit privileges affixes over roots. ONSET is shawn ta be higher
ranking in Tarifiyt while violated under duress in Tash/hit. Neither variety is shawn ta
violate MAX-RT (V) while bath behave idiosyncratically with regard ta other constraints.
The findings indicate that standardizing of this aspect is feasible, an initial-a system being
justifiably Jess problematic as a standard system.
1. INTRODUCTION:
Research on the sound system of Tarifiyt has been both seant and erra tic in
comparison with the research conducted on other varieties of Amazighe. Even
considering the existing studies on Tarifiyt, one would notice how balance has
tipped in favor of studies on consonants at the expense of vowels. This state of
affairs, coupled with the comparative complexity of the sound system of Tarifiyt in
The analysis presented in this paper has benefited from fruitful discussion with Karim
Bensoukas and ail other participants at the conference "Standardisation de l'amazighe",
organized by the IRCAM, Rabat, on 17-18 October 2004. Gerninate sounds are transcribed
with doubled letters. Emphasis is transcribed with a dot undemeath. Other sounds are
transcribed using IPA.
34
120
Structures morphologiques
de l'amazighe
general, has made a number of issues related to the vocalic system of Tarifiyt under
researched. One issue that sets Tarifiyt apart, but which is likely to provoke
discussion among Amazighists is the behavior of the initial vowel of sorne nouns,
an aspect the present paper purports to look into.
The behavior of initial vowel in Tarifiyt nouns is especial in the sense that
its deletion or retenti on is, at least as far as the existing accounts can tell,
unpredictable. ln Tashlhit, the situation does not seem to be as problematic, so long
as the vowel at stake remains intact in most instances. For example, to the Tashlhit
words 'afus' (hand), 'ajur' (moon) and 'anu' (weIl) correspond the Tarifiyt words
'fus', 'joa' and 'anu', respectively. This inconsistency calls for an explanatorily
adequate account of the manifestation of the initial vowel in the two varieties, an
account which the framework of Optimality Theory is justifiably able to handle.
It is proposed in the paper that the divergence between Tarifiyt and
Tashlhit with respect to the behavior of the initial vowel is the outcome of the
interaction of markedness and faithfulness constraints as in Optimality Theory.
Tarifiyt will be shown to rank M-REAL (a) over Affix Faithfulness. It will also be
shown that Tashlhit adopts reverse ranking of Affix Faithfulness over M-REAL (a)
(Affix Faithfulness>
M-REAL (a). Further ramifications of the issue will be
revealed as the analysis unfolds.
The paper will be organized as follows: section II presents the background of
the paper, for it presents a few facts that illustrate how the initial vowel of a
masculine noun in the two varieties oscillates between deletion and retention.
Section III presents the analysis of the problem in the light of the theoretical
framework of Optimality Theory. Section IV looks into the standardizability of the
initial vowel in Amazighe based on the results achieved. The paper then concludes
with a few remarks.
II. BACKGROUND:
1. Data:
Generally speaking, the initial vowel of masculine nouns is manifested
differently in Tarifiyt and Tashlhit, which seems to broaden the already existing
phonological gap between the two varieties of Amazighe. Where Tashlhit
consistently maintains the vowel in question in the free state, Tarifiyt deletes it in
sorne words but keeps it in others, a complex situation that has presumably left the
phenomenon hitherto unaccounted for. ln the following examples, abstraction
121
Structures morphologiques
de l'amazighe
being made of such minor phonetic details as spirantization and coda-r deletion",
the words in the two varieties carry the same meaning and are phonologically
identical, except with respect to the initial vowel.
(1)
Tarifiyt
fus
fud
la:
Tashelhiyt
afus
afud
alar
Gloss
'hand'
'knee'
'foot'
The focal point in the paper is the noun that is in the free state. ln the following
paragraphs are discussed the differences between the construct state and the free
state of nouns in order to make more explicit the difference between the two states.
ln the Amazighe language in general, nouns can occur in two major states:
they can be either in the construct state or in the free state. A noun is said to be in
the construct state when it is the subject postposed to the verb, the complement of
certain prepositions or as the second of the two conjoined NPs (Saib, 1982;
Guerssel, 1983; Tangi, 1991, among others). AH other environments represent the
free state. This pan-Amazighe aspect can account for the otherwise complex
morpho-phonological
phenomena that mark the language. It is therefore no
accident that the construct state and free state of nouns is used in the present paper
to explain the behavior of the initial vowel in the two varieties under investigation.
The following examples show the difference between the two states.
35Two
salient features of Tarifiyt are coda-r deletion and spirantization. The former is a
process deleting any Irl that is not a derivative of an III in the coda position (lazzi, 1991;
Tangi, 1991; Armous and Bensoukas, 2004; in prep; ms.). Spirantization consists in
transforming sorne stops into fricatives, so that sorne Tashlhit stops surface in Tarifiyt as a
fricative.
122
Structures morphologiques
de l'amazighe
(2)
Tashelhit
Tarifiyt
Free state
fus
fud
da:
Joa
anu
ataj
Construct state
Free state
ufus
ufud
uda:
ujoa
wanu
wataj
afus
afud
adar
ajjur
anu
ataj
Construct
state
ufus
wafud
udar
wajjur
wanu
wataj
'anu'=well;
'ataj'=tea
a) The Tarifiyt nouns in the free state with no initial vowe1 fail to acquire one
in the construct state.
b) The Tarifiyt nouns in the free state with an initial vowel keep this vowel in
the construct state.
c) For Tashlhit, the prefixal initial vowel appears both in the free and
construct states.
The phenomenon of initial vowe1 deletion in Tarifiyt affects feminine nouns as
well, ln Tashlhit, except in the construct state of feminine nouns, where the two
varieties un der scrutiny seem to converge, these nouns conserve their initial vowel
consistently. The following chart illustrates how the initial vowel is realized in
identical words in Tarifiyt and Tashlhit.
123
Structures morphologiques
de l'amazighe
(3)
Tashelhit
Tarifiyt
Free state
Bfust
Sfut
dda.t
SjoaS
SanuS
Construct state
Sfust
Sfut
dda.t
SjoaS
SanuS
Construct
state
Free state
tfustt
tfutt
t1artt
tajjurt
tanut
tafust
tafut
tadart
tajjurt
tanut
(fem);' SanuS'=well(fem).
ln the plural from, both Tarifiyt and Tashlhit keep sorne version of the initial
vowel. This shows how similar the Amazighe varieties are where the morphology
is concemed. The following examples indicate such striking similarities between
the two varieties.
(4)
Tarifiyt
SiDl?:ular
fus
fud
Joa
anu
ajjaw
Tashelhit
Plural
ifassen
ifadden
ijoan
anuSen
ajjawen
Sinaular
afus
afud
ajjur
anu
ajjaw
Plural
ifassn
ifaddn
lm
una
ajjawn
grandson.
124
Structures morphologiques
de l'amazighe
Free State
Gloss
Construct State
awar
adda:
aru3
wawar
wadda:
waru3
'speech'
'cliff
'porcupine'
8a33atS
8adda:8
8ajjawt
8a33atS
8adda:8
8ajjawt
'widow'
'house'
'granddaughter'
It should be pointed out that the examples in (5) above are attested both in
Tarifiyt and Tashlhit (see footnote 34). As is clear in the data, no initial vowel
deletion takes place in the construct state ..
2.2.Prefixal vowels:
Radical initial vowels in Tarifiyt nouns do not undergo deletion as do nonradical ones. The behavior of these vowels in the construct state is illustrative of
this state of affairs.
(6)
Free State
fus
Gloss
Construct State
'hand'
ufus
125
Structures morphologiques
yamm
8ma:8
de l'amazighe
'bamboo'
'beard'
uyamm
8ma:8
ln the analysis section below, it will be shown that this distinction, although it does
predict the retention of radical vowels, only partially predicts the process of
deleting the vowels that are prefixal. ln fact, assuming that the initial vowels that
are prefixal delete in Tarifiyt nouns is not enough of a generalization, since there
are many examples of prefixal vowels that resist deletion. Cases in point are the
following examples:
(7)
Free State
Construct State
Gloss
afunas
a3a:8ir
ada:yar
ufunas
u 3a:8ir
uda:yar
'bull'
'carpet'
'blind'
The data in (7) above show that, even if the initial vowel of sorne nouns is
prefixal, it does not elide. This further complicates the issue of initial vowel
deletion in Tarifiyt, more so than has as yet been assumed.
Further reca1citrant data to the generalization that prefixal initial vowels delete
cornes from the following examples:
(8)
Free State
Construct State
Gloss
asrm
arnnzu
aYJur
azwa:
usrm
urnnzu
'fish'
'eIder'
'donkey'
'vein'
uvjur
uzwa:
It can be inferred from (7) and (8) above that -if the initial vowel is deleted, the
outcome will be a word with a complex onset. ln data (7), the long vowel is the
outcome of a deleted 'r' sound, the consequence of which is a long vowel (Arnrous
and Bensoukas, 2004, in prep, ms.)
It is hypothesized in this paper that, due to the operation of Tarifiyt-exclusive
phonological processes, morphological
convergence among the dialects of
Moroccan Amazighe is blurred. The departure of Tarifiyt form its cognates is the
immediate consequence of adopting a different ranking of markendess and
faithfulness constraints.
126
Structures morphologiques
de l'amazighe
The paper attempts to explain the phenomenon in the light of the theoretical
framework of Optimality Theory (Prince and Smolensky, 1993; McCarthy and
Prince, 1993, 1995, 1999; Kager, 1999; McCarthy, 2000 and related works).
127
Structures morphologiques
de l'amazighe
(if>
/adrar/
a- adra:
b- adrar
*Coda-r
MAX-Seg
*
*!
ln (9) above, 'adra:' is the winner in the hierarchy, while 'adrar' loses as it
violates *Coda-r, a higher ranked constraint than MAX-Seg. It is in this sense that
'adra:' is accepted as optimal, while 'adrar' is rejected as suboptimal.
The rest of the paper is organized as follows: in the following section, an
analysis of the data presented in (1 through 9) is attempted. Then, a section follows
which deals with the standardizability ofthis aspect of Moroccan Amazighe.
III. ANALYSIS
Within the framework of Optimality Theory, outlined above, the
constraints that are relevant to explain insertion or deletion phenomena are
subsumed under correspondence constraints. ln particular, deletion processes
amount to violations of the faithfulness constraint Max-Seg. This constraint is itself
general as it militates against any input form that does not have an output
correspondent. Therefore, any kind of deletion is a violation of sorne version of
Max-Seg. This constraint is at the center of the analysis in the present paper
inasmuch as the locus of interest is the behavior of the initial vowel in Tarifiyt and
Tashlhit, which oscillates between deletion and retention.
This being so, and given that Tarifiyt realizes the initial vowel of nouns
differently from Tashlhit as is shown in the previous section, one can already
surmise that a similar set of constraints is ranked differently in the two varieties
under investigation. These constraints are Max-Seg, which militates against any
form of deletion and ONSET, which requires that any syllable have an onset. Based
on (Arnrous 2004), the present paper assumes that ranking these two constraints
can account for the mismatch between Tarifiyt and Tashlhit where the initial vowel
is concemed. The two conflicting constraints at stake are listed in (10) below:
128
Structures morphologiques
de l'amazighe
(10)
REALIZE
It should be pointed out that the constraint M.REAL will bifurcate into two
constraints: M-REAL (SING) and M-REAL (PL). This distinction will be relevant
later in the paper.
The constraints in (10) above are ranked differently in Tashlhit and Tarifiyt. Since
Tashlhit realizes initial prefixal vowels phonetically, which Tarifiyt generally does
not, it follows that the constraints requiring the masculine morpheme to be
phonetically realized must take precedence over that requiring syllables to have
onsets. This is a strong argument why Tashlhit involves no initial vowel deletion in
masculine nouns. Tarifiyt seems to prioritize ONSET, a constraint that bans
onsetless syllables, the reason why initial vowels are not permitted in sorne
masculine nouns. Consider the following ranking in Tarifiyt and Tashlhit
respectively, a ranking that is subject to adjustments as the analysis unfolds.
(11)
Tarifiyt ranking: ONSET and M-REAL (SING)
ONSET M-REAL (SING)
Tashelhiyt ranking: ONSET and M-REAL (SING)
M-REAL (SING) ONSET
The following tableaux make even more explicit the ranking Tarifiyt and
Tashlhit adopt for the two constraints in (12).
(12)
Tarifiyt ranking: ONSET and M-REAL (SING)
r:r
/a+fus/
i- afus
ii- fus
ONSET
M-REAUSING)
*!
*
129
Structures morphologiques
<Jr
/a+fus/
i- afus
ii- fus
de l'amazighe
M-REAL(SING)
*!
ONSET
*
MAX-ROOT
VOWEL (Max-RT(V))
<Jr
/anu/
t-anu
ll-anu
MAX-RT(V)
ONSET
*
*!
It can be inferred from the tableau above that, for all the differences that
hold between Tarifiyt and Tashlhit, the two varieties have as a commonality the
fact that root vowels must remain intact. This amounts to the c1aim that MAX-RT
(V) always ranks higher than ONSET in both Tarifiyt and Tashlhit. For Tarifiyt,
then, it is ONSET that outranks M-REAL but outranked by MAX-RT (V). For
Tashlhit, MAX-RT (V) dominates ONSET. Much, however, remains to be said
about why this should be the case. Recall that the overarching question in the
present paper relates to why sorne vowels delete and others do not.
Since the vowels that are part of the root are not deleted in the initial
position of nouns, it becomes c1ear that ONSET is dominated, so that input vowels
appear in the output. One hypothesis is that radical segments are different in their
status from affixal ones and so are to receive special treatment. ln fact, proposals in
the literature have been propounded that roots enjoy a privileged status over
130
Structures morphologiques
de l'arnazighe
affixes. McCarthy and Prince (1995) argue for this privileged status of radical over
affixal material, so that root faithfulness (RF) is favored over affix faithfulness
(AF): RF AF, a ranking corroborated by Beckman (1998). This amounts to the
claim that roots are less susceptible to phonological transformations than are
affixes. Ultimately, it seems that both Tarifiyt and Tashlhit privilege roots over
affixes, thereby conforming to the universal ranking of RF AF.
The third issue to be de aIt with in this section relates to the fact that, in
Tarifiyt, ONSET is violated under duress in sorne cases. ln fact, just as the
language chooses to sacrifie the morpheme in satisfaction of ONSET, so, too,
does it sacrifice ONSET in cases of syllable-initial CC clusters. ln other words, the
violation of ONSET is less serious than its satisfaction if the outcome would be a
complex onset, which the phonotactics of Tarifiyt ban. This amounts to the claim
that ONSET, which in Tarifiyt ranks higher than M-REAL as illustrated in (12)
above, is itself outranked by sorne other constraints which prevents the occurrence
of complex onsets. This constraint may be formulated as follows:
(15)
a- aYJUf
b- yjur
*COMPLEX
ONSET
*
*!
131
Structures morphologiques
de l'amazighe
adopts for the same constraints. Tashlhit prioritizes M-REAL over ONSET while
Tarifiyt adopts the reverse ranking. ln Tarifiyt, moreover, ONSET, which outranks
M-REAL, is govemed by a constraint banning complex onsets: *COMPLEX.
One remarkable fact about Tarifiyt is that, irrespective of whether the
initial vowel is radical or prefixal in nature, the plural form has to survive with an
initial vowel. This is also the case in Tashlhit, where no initial vowel deletion is
attested in the free state. The legitimate question then arises as to why the initial
vowel that is prefixal deletes in the singular form but is maintained in the plural
form. This further justifies the fact that the constraint M-REAL needs to split into
M-REAL (SING) and M-REAL (PL). This shows that realizing the plural
morpheme is more pressing, so to speak, than realizing the singular morpheme. ln
other words, while Tarifiyt may legitimately sacrifice initial vowels that are
prefixal in satisfaction of ONSET, it simply cannot do so for plural morphemes.
This state of affairs can be seen in altemations like 'fus' / 'ifassen' (hand). The
altemation 'fus'/ 'fassen' is attested in no single Tarifiyt variety, to the best of our
knowledge.
This said, it tums out that M-REAL (PL) outranks ONSET, a ranking that
appears in (18) below.
(18)
Tarifiyt ranking of M-REAL (PL) and ONSET
M-REAL (PL>ONSET
The ranking in (18) above is exemplified as follows:
(19)
rJir
/ i+fassen /
a- ifassen
b- fassen
M-REAL(PL)
ONSET
*
*!
It should be recalled that the ranking in (18) above applies to Tarifiyt only, since,
as has been stated earlier, Tashlhit involves no initial vowel deletion and, by
implication, no such asymmetry.
132
Structures morphologiques
de l'amazighe
So far, the analysis has centered on the behavior of masculine nouns in Tarifiyt
and Tashlhit. However, such an analysis, which is couched within Optimality
Theory, can as well shed light on the behavior of feminine nouns. Central to the
issue is the question as to why certain nouns in Tarifiyt survive with no initial
vowel and behave, instead, like their masculine counterparts. The following data is
illustrative of the phenomenon:
(20)
Gloss
Feminine
Masculine
'hand'
'moon'
'well'
'granddaughter'
Sfust
SjoaS
SanuS
Sajjawt
fus
Joa
anu
ajjaw
The phonetic form of the nouns in (20) above is heavily dependent on the
form of the masculine noun rather than on the input. ln other words, to the input
/t+a+fus+t1 corresponds the output 'Tfust' rather than sorne other form like
'Tafust'. ln this case we are in a situation where the output form of these feminine
nouns is not faithful to the input, but rather to the output, an interesting situation
indeed.
The framework being adopted for the analysis provides a satisfactory
answer to this problem. According to output output correspondence theory
(McCarthy, 1995), it is preferable to have uniform output paradigms than to have
outputs which are not uniform. ln other words output output constraints check the
identity of morphologically related output forms. ln a situation like the present,
there seems to be a need for a constraint that ensures that outputs are derived from
other outputs rather than inputs. This constraint can be expressed as follows:
(21)
Output
Output
Correspondence
(OO-COR)
(McCarthy,
1995):
133
Structures morphologiques
de l'amazighe
(22)
1tafustl
r:ir
OO-COR
a- Tfust
b- Tafust
*COMPLEX
*
*!
One further issue the paper brings up relates to the case of sorne masculine
nouns in the free state whose initial vowel does not elide, even if the vowel in
question is prefixal in nature. The following are illustrative examples, taken from
(Amrous, 2004). It should be recalled that the construct state is herein used to
determine the identity of the initial vowel.
.(23)
Free State
asa:dun
ava.da
ama:was
Construct State
Gloss
'horse'
'mouse'
'loan'
usa:dun
uva.da
uma:was
One observation that can be made about the data above is that aIl the words
involve the long vowel 'a:' deriving form an underlying sequence lar!. This process
is a feature that many Tarifiyt dialects share (Tangi, 1991; lazzi, 2001; Amrous and
Bensoukas, 2003; Amrous and Bensoukas, in prep). Suspiciously enough,
examples like these ones suggest that words like 'asa:dun' (horse), which would
otherwise surface as 'sa:dun', would involve two deletion processes, one deleting
the initial vowel by virtue of being a prefix and the other deleting coda-r, It is
hypothesized that two deletion processes may not be tolerated, the reason why only
those nouns in which no other vowel is deleted can incur prefixal initial vowel
deletion.
This is a phenomenon that can be accounted for using the OT concept of
Constraint Conjunction as laid down in Smolensky, (1993, 1995) and Padgett
(2002). According to a constraint conjunction analysis, violating two constraints
134
Structures morphologiques
de l'amazighe
separately is 1ess serious than violating them in combination. ln the case at hand,
there are two constraints that are 'ganged up' in sorne manner. One is a faithfulness
constraint that militates against deletion. This constraint is formulated as MAX-r.
ln sorne Tarifiyt dialects, MAX-r is violated, permitting words with an underlying
representation like 'asrdun', (horse) 'ayrda' (mou se) and 'arnrwas' (Joan) to surface
as 'asa:dun', 'aya:da' and' ama:was, respectively. The other constraint is M-REAL
(SING), formulated in (10) above, requiring that the singular morpheme be realized
phonetically.
ln order to account for the data in (23) above, therefore, it stands to reason
to invoke the concept of constraint conjunction, though the present paper does not
c1aim the analysis to be exhaustive. The conflict between 'sa:dun', which, to the
best of our knowledge, is attested in no single Tarifiyt dialect, and 'asa:dun', where
the prefixal initial vowe1 survives deletion, my be captured in terms of constraint
conjunction la Smolensky (1993, 1995). This state of affairs motivates the
following ranking:
(24)
MAX-r and M-REAL(SING>ONSET
The ranking in (24) above can be exemplified in the following tableau:
(25)
a+yrda
Cff>
a. aya.da
b. va.da
ONSET
*
*1
135
Structures morphologiques
de l'amazighe
ENTERPRIZE:
136
Structures morphologiques
de l'amazighe
Tashlhit, on the other hand, does not seem to involve the splitting of the M-REAL
constraint. The inconsistency of, say, 'fus'/ 'Tfust' called for the postulation 00COR, a constraint that reveals how an output form can be derived from another
output form rather than an input form. This constraint is not shown to be at stake in
Tashlhit, in which aIl initial vowels are manifested.
This said, it can be concluded that Tashlhit involves less complexity as to
the behavior of initial vowels, while the system of Tarifiyt does involve sorne
complexity. One recommendation in this regard is to take as standard the Tashlhit
version, which is more or less consistent. This is added to the fact that the initial
vowel is recoverable in Tarifiyt if it is deleted, which can be indicated in the plural
form where, for example, 'fus' corresponds to 'ifassen'. A further advantage in
standardizing an initial-'a'- system is that the feminine form always corresponds to
the masculine form, since aIl masculine nouns in Tashlhit have overt initial vowels
that survive in the feminine form. This is unlike Tarifiyt where the initial vowels
sometimes do not match in the masculine and the feminine. For pedagogical
purposes, then, it would seem more preferable to teach a system than involves more
regularity. A system with a surface initial vowel is better equipped to be
introduced.
FinaIly, it should be stressed that the present paper is by no means
exhaustive as there are many aspects of the phonology of the initial vowel that need
10 be researched. The phonology of Tarifiyt needs to receive due interest on the
part of researchers. Research on the linguistics of Amazighe should as weIl be
geared towards more comparative studies.
Bibliography
Arnrous, N. (1999). Phonological Pro cesses in Ait Oulichek Tarifiyt Berber.
Unpublished Memoir paper. Mohammed V University, Faculty of Letters
and Hurnan Sciences, Rabat.
Arnrous, N. (2004). "The Initial Vowel in Tarifiyt Nouns: An Optimality-Theoretic
Analysis". Paper presented at La langue Amazighe: Approche linguistique,
a conference held at the Faculty of Letters and Human Sciences, Ben
M'Sik, on March, 18,2004.
Arnrous, N. and K. Bensoukas. (2004). "Tarifiyt Long vowels and Diphthongs:
Independent Phonemes or Simple Phonetic Variants of the Basic Amazighe
Vowels?". Standardisation de /'amazighe. eds, M. Ameur & A. Boumalk.
Publications de l'Institut Royal de la Culture Amazighe, Rabat.
137
Structures morphologiques
de l'amazighe
138
Structures morphologiques
de l'amazighe
139