Professional Documents
Culture Documents
R. Kirk Belnap
Abstract
This paper examines Variation in marking agreement in Arabic. In particular, the investigation
focuses on the Variation between plural and feminine singular agreement with plural head nouns.
Tape-recorded naturalistic Speech from sociolinguistic Interviews conducted in Cairo constitute
the data for the study of agreement in New Arabic. These results are compared to agreement
patterns found in a corpus of Old Arabic texts. Some have suggested that the agreement Variation
fbund in New and Old Arabic varieties is random and meaningless. However, multivariate analy-
sis of the Cairene and the Old Arabic patterns indicate both are systematic and that the two are
similar in many respects. The agreement patterns in question appear to be a resource which
Speakers exploit to classify referents.
It is generally agreed that the language contact Situation resulting from the spread of Islam
had a profound effect on the development of vernacular varieties of Arabic. Some have argued
that the process of language shift to Arabic was rapid, resulting in deep-reaching changes in
spoken varieties of Arabic. On the other hand, the formal variety of Arabic that came to be Clas-
sical Arabic is touted äs having changed little. This study suggests that Classical Arabic, too,
appears to be the result of some contact induced change and that the agreement System of Cairene
is, in some ways, closer to that of early Old Arabic than is that of its standardized cousin, Modern
Standard Arabic. From the standpoint of agreement, it would appear that varieties such äs Cairene
have changed less and that Classical Arabic changed more than one might suppose. These find-
ings suggest that a re-examination of the history of Arabic is in order.
agreement with nouns referring to human beings show the greatest degree of
symmetry: masculine Singular forms occur with masculine Singular modifiers,
feminine Singular forms with feminine singular modifiers, masculine plural
forms with masculine plural modifiers, and feminine plural forms with feminine
plural modifiers, äs in (1) and (2).
(1) a. mudiir saabiq b. mudiir-uun saabiq-uun
director.M.so former.M.SG director-M.PL former-M.PL
'a former director' 'former directors'
(2) a. mudiir-a saabiq-a b. mudiir-aat saabiq-aat
director-F.SG former-F.SG director-F.PL former-F.PL
'a former director (f.)' 'former directors (f.)'
There are two basic types of Arabic plural forms. Plural nouns and adjectives
are both traditionally referred to äs either sound plurals or broken plurals.
"Sound plural" refers to the formation of the plural form by suffixation, äs in
mudiir 'director', mudür-uun 'directors'. "Broken plural" refers to a plural
formed by alteration of the stem, äs in rajul 'man', rijaal 'men'. Tfypically, a
given singular form has either a corresponding broken plural or a sound plural,
though occasionally both. While sound plurals could be said to be "regulär",
broken plurals are more frequent, i.e. the plural of most nouns is of the broken
type. A plural head (broken or sound) may occur with either an accompanying
broken plural or sound plural modifier or both, äs in (3) where the head and first
adjective are broken plurals but the second adjective is a sound plural.
(3) wuzaraa? judud ka6iir-uun
ministers.PL new.PL many-M.PL
'many new ministers'
With reference to Classical Arabic, Wright observed that broken plurals "differ
entirely from the sound plurals; for the latter denote several distinct individuals
of a genus, the former a number of individuals viewed collectively, the idea of
individuality being wholly suppressed" (1974:i, 233). In other words, histori-
cally, there appears to have been a semantic distinction between broken and
sound plurals. The degree to which this distinction has survived in MSA is a
matter in need of further investigation.
A distinctive syntactic feature of many varieties of Arabic is the use of
feminine singular forms to agree with some types of plural heads. In MSA, ad-
jectives preserve the masculine and feminine grammatical gender distinction
when agreeing with singular, nonhuman head nouns, äs in (4a) and (5a); how-
ever, adjectives modifying plural nonhuman head nouns show no corresponding
differentiation, instead, only feminine singular forms are used, whether the head
is a broken plural (4b) or a sound plural (Sb).1 In other words, the gender dis-
tinction found in the singular is maintained in plural agreement with human
heads, äs in (1) and (2), but not with non-human heads, äs in (4) and (5). In
Brought to you by | University of Arizona
Authenticated
Download Date | 7/8/15 4:12 PM
171
MSA, feminine Singular agreement is essentially categorical for all forms (at-
tributive and predicative adjectives, demonstrative pronouns, verbs, and ana-
phoric pronouns) occurring with nonhuman plural heads, regardless of the gen-
der of the corresponding Singular of the head noun.
(4) a. kitaab jadiid b. kutub jadiid-a
book.M.SG new.M.so books.PL new-F.so
*a new book' *new books'
(5) a. sayyaar-a jadiid-a b. sayyaar-aat jadiid-a
car-F.so new-F.SG car-F.PL new-F.SG
4
a new car' 'new cars'
Unlike MSA, most other varieties of Arabic exhibit considerable Variation in
their patterns of agreement. This is true of other varieties of Old Arabic (e.g.,
the Pre-Islamic poetic koine of the 6th-7th Century and the language of the
Qur'an, also 7th Century) äs well äs New Arabic (e.g. Cairene, Damascene, and
Tunisian). (See examples (6), (7), and (8) from Cairene, and note that it and
many other varieties of New Arabic exhibit no gender distinction in marking
agreement with plural heads, that is, the categories of agreement are: masculine
Singular, feminine Singular, and common plural.) In many vernacular varieties
of Arabic, nonhuman plural heads usually occur with feminine Singular agree-
ment, though plural agreement is not uncommon, äs in (6). Human head nouns,
on the other hand, generally occur with forms exhibiting plural agreement, but
feminine singular agreement also occurs, especially with broken plurals, äs in
(7). If a head noun (human or non-) occurs with a numerical quantifier, plural
agreement is nearly categorical, äs in (8) (Belnap 1993:104-106).
(6) biyuut kabiir-a (kubaar)
houses.PL large-F.SG large.PL
'large houses'
(7) riggaala kuwayyis-iin (kuwayyis-a)
men.PL nice-PL nice-F.sc
'nice men'
(8) talat sitt-aat kuwayyis-iin
three women-F.PL nice-PL
'three nice women'
The complexities of this agreement Variation have been the topic of a number of
studies or have figured prominently in the discussion of specific topics, par-
ticularly in recent years the history of Arabic. A representative sample might
include: Belnap (1993), Blanc (1970), Ferguson (1989), Mitchell (1973), Owens
& Bani-Yasin (1987). Such studies represent a variety of perspectives or orien-
tations. Some have proposed ' Solutions'; others offered largely descriptive ac-
counts, sometimes extolling the richness of the complexity. This paper examines
patterns of agreement Variation in a modern vernacular variety, Cairene Arabic
Brought to you by | University of Arizona
Authenticated
Download Date | 7/8/15 4:12 PM
172
(Section 2). These patterns are compared with those in a corpus of Old Arabic
texts (Section 3). We then discuss the implications of this study for the history
of Arabic (Section 4).
for the variable rule analysis is less than the total number of tokens in the sam-
ple, These excluded tokens are set aside only for the purpose of the variable rule
analysis; they are crucial to the total analysis but due to the constraints of the
variable rule program they must be dealt with separately. (See Belnap (1991:57-
61, 157-165) for more Information on the interviewees and the interview proc-
ess and for a more detailed account of the coding procedures and the statistical
analysis.)
Table l also indicates that broken plurals are more likely to occur with feminine
singular agreement loci than are sound plurals. In the sample only one human
sound plural head (sittaat 'women') occurred with feminine singular agreement
(two feminine singular verbs).4 Eight sound plural heads that refer to men (or to
men and women) occurred in the sample, all with plural agreement. The litera-
ture leads one to believe that these human sound plural heads categorically re-
quire plural agreement. This is not so, however. There were a few instances in
the Interviews of such human sound plurals occurring with feminine singular
agreement; however, these examples did not occur in the sample used for the
variable rule analysis. Three Speakers used feminine singular verbal agreement
with fallahiin 'farmers', yarbiyyiin 'Westerners', and mufattisiin 'inspectors'
(see example (11)). The feminine sound plural sittaat 'women' was much more
likely to occur with feminine singular agreement. Ten Speakers used feminine
singular agreement with sittaat 'women.'
(11) ?il-fallah-iin kaan-it hilw-a
the-farmer-PL were-F.sc sweet-F.sc
'The farmers were great.' (Kawthar, f., 55, maid)
In a general treatment of agreement, including a broad survey of agreement
phenomena in a diverse sample of languages of the world, Barlow (1988) con-
Brought to you by | University of Arizona
Authenticated
Download Date | 7/8/15 4:12 PM
175
form immediately precedes its subject head and the plural pronoun -hum imme-
diately follows.
(12) siwayyit hag-aat ... mis bi-tartiib ?ahammiyyit-ha
few thing-PL not by-rank importance-their.F.SG (6)
wi-baideen ni-§uuf ?iza kun-na ni-rattib-hum
and-afterward we-see if were-we we-order-them.PL(ll)
'... a few things ... not in order of their importance ... and afterward we'll
see if we were ordering them.' (Riham, f., 21, Student)
(13) yoom masal-an t-iigi 1-iiyaal ?a-?ul-lu-hum
day example-ADV F.SG-come the-children I-say-to-them.PL
One day, for example, the children come and I teil them...'
(Magdi, m., 30, skilled construction worker)
Table 2 presents the factored effect of distance on agreement. Negative numbers
indicate an agreement locus occurring in prenominal position, such äs in a verb-
subject construction. For example, the verb preceding its subject head in (13) is
counted äs -1. Positive numbers refer to the distance of a given locus from the
head it follows, äs in (12).
Table 2: Distance of agreement locus from head
The effect of distance is clear. The closer the locus of agreement is to the head
the greater the frequency of feminine Singular agreement. The pre-nominal po-
sitions favor feminine Singular agreement most.5 Although functional explana-
tions often fall short (Labov 1994:547-568), it would appear there may be a
functional basis to the hierarchy discussed here, äs concerns the recoverability
of Information. The nearer an agreement locus is to its head the more immediate
is the association between the two: feminine Singular agreement is far less likely
to interfere with the interlocutor's perceiving the grammatical relationship be-
tween the head and locus. The more distant the locus is from its head, the
greater the potential for confusing the relationship between the head and the
locus. Related discourse processing constraints appear to influence the type of
agreement that obtains in Tsez (Polinsky & Comrie, this volume; see also Bar-
low's discussion of the discourse function of agreement in this volume.).
The results reported here and elsewhere for Cairene indicate the non-
random nature of its agreement Variation. They indicate that a high correlation
obtains between the morphology and animacy of a head noun and the type of
agreement that obtains. Further research has demonstrated that plural agreement
is used with heads that are perceptually more salient or perceived äs distributed
plurals and that feminine Singular agreement is used with heads that would ap-
pear to be less salient or perceived äs non-distributed plurals. Speakers appear to
be able to exploit this agreement Variation to classify referents. The effect of
distance between the head and its locus of agreement underscores the reality of
the human discourse processing factor.
The sample from the Qur'an included no plural heads referring to animals.6 As a
result, and due to the fact that the agreement Variation with human plural heads
in the Qur'an sample was confmed almost entirely to instances of prenominal
verbs, only the Pre-Islamic poetry tokens were included in the variable rule sta-
tistical analysis. The results for animacy/head-type of the variable rule run are
found in Table 5.7
Table 5: Pre-Islamic Texts: Anitnacy, head type and feminine Singular agreement
4. Implications
The parallels between the Cairene and the early Old Arabic data sets are strik-
ing. This is particularly so in view of the fact that at least a millennium has
elapsed since New Arabic split off from Old Arabic. The similarities between
the two are also all the more significant given the differences in the Cairene and
the Pre-Islamic corpora (informal spoken conversation versus poetry). The dis-
tance effect in Old Arabic suggests that the Variation is not the result of the ran-
dom demands of poetic meter, but rather indicative that the corpus exhibits the
tendencies of natural language. The Old Arabic data, particularly the patterning
according to animacy and morphological constraints, suggest there is good rea-
son to believe the New Arabic Variation derives from the old, that it is a case of
stable Variation. This might appear so obvious äs to preclude further discussion.
Brought to you by | University of Arizona
Authenticated
Download Date | 7/8/15 4:12 PM
181
However, when we consider that MSA, the only surviving variety of Old Ara-
bic, does not exhibit similar patterns of agreement Variation, then what appeared
obvious raises questions. Why should a New Arabic variety such äs Cairene
preserve such patterns while its supposedly more conservative cousin MSA
does not?
It is generally agreed that the language contact Situation resulting from the
spread of Islam had a profound effect on the development of vernacular varie-
ties of Arabic (Fischer & Jastrow 1980:15-19). By the ninth Century, imperfect
learning of Arabic on the part of new converts (owing to this language contact
Situation) was perceived äs a threat to the correct transmission and future com-
prehension of the Qur'an. As a result, efforts were undertaken to preserve the
Qur'an and codify Classical Arabic. There is some debate äs to whether the
efforts of the early Arab grammarians were merely descriptive or whether they
were normative in nature. If they were normative then the language contact
Situation of the early Islamic era not only led to changes in spoken Arabic but
also, to some degree, to changes in what is generally referred to äs Classical
Arabic.
The most appealing solution to the puzzle äs to how Classical Arabic
agreement patterns came to be simplified would appear to be that this Variation
was essentially eradicated from the formal written language when Old Arabic
was codified and apparently standardized. Variable agreement patterns are well
attested in the Qur'an, the theoretical ideal for Arabic style and usage, but only
vestiges of this Variation remain in general written usage a few centuries later.
Standardization at the hands of the early Arab grammarians would account for
the facts of Old Arabic, but we have found no evidence that they ever addressed
the issues of agreement under consideration here. Given only the facts of Pre-
Islamic, Classical, and Modern Standard Arabic, one would be tempted to argue
that there was a change in progress (a tendency to use increasingly more femi-
nine singular agreement with non-human heads) that eventually resulted in an
essentially categorical rule soon after the spread of Islam outside the Arabian
Peninsula.9 The large numbers of nonnative Speakers and writers who came to
use Arabic during this period could easily have accelerated a process which was
already underway, a familiär phenomenon in language contact situations. Again,
this seems a straight-forward solution, were it not for the robust survival of such
patterns in many, if not most, varieties of spoken Arabic. Nonnative Speakers of
Arabic learned both the written and the spoken language. However, the simpli-
fication of the agreement patterns took place in the written language only.
Belnap & Gee (1994) argue that this apparently spontaneous Standardization
or regularization in the written language but not in the spoken language is the
result of L2 learning strategies, but with a twist. They argue that the new rule of
categorical feminine singular agreement with nonhuman heads was indeed the
result of learner overgeneralization. Research on probability matching in the
Brought to you by | University of Arizona
Authenticated
Download Date | 7/8/15 4:12 PM
182
Notes
In order to lay the groundwork for this paper, Section 2 contains some material previously
published äs Belnap (1993). This material has, however, beer» updated and includes some
additional data, äs well äs a re-analysis of the data set.
The interaction of animacy with agreement marking is found in a number of languages. In
this volume, see Barlow's listing of types of feature discord, Corbett's discussion of Miya,
Polinsky & Comrie's treatment of Tsez, and van den Berg's analysis of Akusha Dargi. Al-
though not directly based on animacy, verbal agreement with neuter nouns is singular in
Classical Greek — äs well äs in Koine Greek, but not strictly so. For discussion of the possi-
ble influence of Greek on Arabic agreement patterns, see Belnap & Gee (1994:137-140).
Killean (1968) observed that the phenomenon of feminine singular agreement with non-
human plurals is similar to that of neuter gender found in other languages. One argument
against this analysis, which Killean mentions, is the fact that cardinal numbers and bidf
'several' do not agree with non-human plurals äs if they were feminine singular, but rather
show reverse agreement with the gender of the singular form of the noun, äs discussed
above. The following example shows feminine singular verbal and adjectival agreement
while the number agrees (or rather, disagrees) with the singular form of the head, which is
feminine in the singular:
(i) ?ittafaq-at 9alaa6 duwal ?arabiyy-a ?ala 1-masruuT
agreed-3.F.SG three.M nation.F.PL Arab-F.SG on DEF-project
'Three Arab nations agreed on the project.'
Abdel-Massih et al. noted that "some Speakers of E[gyptian] A[rabic] use the latter structure
[kutub kibiira 'big books' — a broken plural head followed by a feminine singular adjective]
even when the modified noun is human plural" and give äs examples ?a\vlaad kitiira 'many
boys' and banaat kitiira 'many girls' (1979:22). In the Interviews, 73% (19/26) of the inter-
viewees used feminine singular agreement at least once with human heads other than naas
'people* (with which all Speakers used feminine singular agreement to some degree).
Clearly, if 19 of 26 Speakers used scattered instances of feminine singular agreement with
human heads in the space of 90 minutes or less, it is certain that a longer sample would yield
a higher percentage of Speakers using feminine singular agreement. In other words, it is not
"some Speakers" who use feminine singular agreement with human heads, äs Abdel-Massih
et al. (1979) suggested, it is the majority — and probably all — who do so.
Sallam observed that the occurrence of feminine singular agreement with human plurals
is restricted to "some collocations" (1979:49). Tomiche (1964:179) suggested feminine sin-
gular agreement is rare and restricted to collectives (naas 'people') and feminine plurals
(banaat 'girls')· Again, the corpus clearly demonstrates that the Variation is not restricted to
these head types.
The probabilities in Table l differ from the percentages in that they correct for skewing in
the data set. For example, the probabilities given for animal broken plurals and for naas
'people' are much lower than the percentages listed. This is due to the fact that a dispropor-
tionate number of tokens in these two categories occurred nine or more words away from the
head, a position almost categorically favoring plural agreement (See Section 2.3.). As a re-
sult, the percentages are somewhat inflated.
Human sound plurals were excluded from the variable rule analysis due to the fact that they
present special complications and there were too few types and tokens in the sample to allow
an analysis which would take into account the differences between masculine sound plurals
and feminine sound plurals.
The monotonic decrease in the probability of feminine singular agreement äs one proceeds
from the farthest prenominal position to the most distant postnominal position in Table 2
suggests the possibility that the farther left the locus of agreement, the higher the probability
of feminine singular agreement. One may confidently conclude that this is the case for post-
nominal agreement loci; however the case of prenominal agreement loci warrants further in-
vestigation. The results for the left-most prenominal position are based on only six agree-
ment loci; while the results for this position are suggestive they are hardly conclusive.
6 The Qur'an sample contained one head construction containing a singular collective refer-
ring to animals:
(i) fa-xud ?arba?atan min aHayri fa-sur-hunna ?ilay-ka
CONJ-take four of DEF-bird.coLL CONJ-incline-them.F.PL. to-you
Take four birds and make them come to you ...' (Qur'an, II, 260)
Note that plural agreement obtains, likely due to the presence of the numerical quantifier
which cancels the grouped nature of the collective and results instead in a distributed read-
ing.
7 A skewed distribution accounts for the higher percentage of feminine singular agreement
with inanimate sound plurals in Table 5. No agreement tokens for this category occur farther
than five words away from the head; 70% (7/11) occur to the immediate right of the head,
where feminine singular agreement is more likely to obtain (See Section 3.3.).
8 The low percentage for position l in Table 6 is due, again, to skewing of the data. Position l
is dominated by adjectives. Adjectives are far less likely to show feminine singular agree-
ment than verbs and pronouns. The frequencies of feminine singular agreement by locus
type for position l are: adjectives 10% (6/61), verbs 69% (22/32), pronouns 85% (17/20).
The locus type factor group proved to be statistically significant but is not discussed here äs
it falls outside the scope of this paper. The results in Table 6 appear to generally support
Corbett's agreement hierarchy (1983). However, when we hold distance constant, at least for
position l, the distribution of the frequencies according to syntactic category are the reverse
of what Corbett's agreement hierarchy would predict (1983). This suggests that the Variation
between feminine singular agreement and plural agreement may be more complex than sim-
ply a matter of syntactic versus semantic agreement. A more detailed analysis is needed. For
further discussion see Belnap (1991:87-89,123-128).
9 Some variable agreement patterns continued to be used after the spread of Islam, mostly in
literary works and much less so in expository prose (Belnap & Gee 1994:126-133). One relic
pattern in particular can be found in MS A prose of a recherche style (Belnap 1991:128-129).
References
Abdel-Massih, Ernest T., Zaki N. Abdel-Malek & El-Said M. Badawi. 1979. A Reference Gram-
mär of Egyptian Arabic: A Preliminary Edition, A Comprehensive Study of Egyptian Ara-
bic, 3. Ann Arbor: Center for Near Eastern and North African Studies, University of Michi-
gan.
Barlow, Michael. 1988. A Situated Theory of Agreement. Ph.D. Dissertation, Stanford Univer-
sity.
Barlow, Michael. "Agreement äs a Discourse Phenomenon." This volume, 187-210.
Beeston, Alfred RL. 1975. "Some Features of Modern Standard Arabic." Journal of Semitic
Studies 20,62-68.
Belnap, R. Kirk. 1991. Grammatical Agreement Variation in Cairene Arabic. Ph.D. Dissertation,
University of Pennsylvania.
Belnap, R. Kirk. 1993. "The Meaning of Agreement Variation in Cairene Arabic." In: Mushira
Eid & Clive Holes (eds.), Perspectives on Arabic Linguistics V. Amsterdam - Philadelphia:
Benjamins, 97-117.
Belnap, R. Kirk & John Gee. 1994. "Classical Arabic in Contact: The Transition to Near-
Categorical Agreement Patterns." In: Mushira Eid, Vicente Cantarino & Keith Walters
(eds.), Perspectives on Arabic Linguistics VI. Amsterdam - Philadelphia: Benjamins, 121-
149.
Blanc, Haim. 1970. "Dual and Pseudo-Dual in the Arabic Dialects." Language 46,42-57.
Corbett, Greville G. 1983. Hierarchies, Targets and Controllers: Agreement Patterns in Slavic.
London: Croom Helm.
Corbett, Greville G. (this volume). "The Place of Agreement Features in a Specification of Possi-
ble Agreement Systems."
Ferguson, Charles A. 1989. "Grammatical Agreement in Classical Arabic and the Modern Dia-
lects: A Response to Versteegh's Pidginization Hypothesis." Al-cArabiyya 22, 5-17.
Brought to you by | University of Arizona
Authenticated
Download Date | 7/8/15 4:12 PM
185
Fischer, Wolfdietrich & Otto Jastrow. 1980. Handbuch der Arabischen Dialekte. Wiesbaden:
Harrassowitz.
Ibrahim, Muhammad H. J 970. A Study of Gender. Ph.D. Dissertation, Princeton University.
Killean, Carolyn G. 1968. "Interesting Features of Gender-Number Concord in Modern Literary
Arabic." In: BJ. Darden, C.-J. N. Bailey & A. Davison (eds.), Papers from the Fourth Re-
gional Meeting, Chicago Linguistic Society, 40-49.
Labov, William. 1994. Principles of Linguistic Change: Volume 1: Internal Factors. Cambridge,
Mass. - Oxford, UK: Blackwell.
Mitchell, Terry F. 1973. "Aspects of Concord Revisited, with Special Reference to Sindhi and
Cairene Arabic," Archivum Linguisticum 4, 27-50.
Owens, Jonathan & Raslan Bani-Yasin. 1987. "The Lexical Basis of Variation in Jordanian Ara-
bic." Linguistics 25, 705-738.
Polinsky, Maria & Bernard Comrie. (this volume). Agreement in Tsez.
Sallam, A.M. 1979. "Concordial Relations within the Noun Phrase in Educated Spoken Arabic,
ES A." Archivum Linguisticum 10, 20-56.
Sankoff, David & William Labov. 1979. "On the Uses of Variable Rules." Language in Society 8,
189-222.
Tarallo, Fernando. 1986. "Functional and Structural Properties in a Variable Syntax." In D.
Sankoff (ed.), Diversity and Diachrony. Amsterdam - Philadelphia: Benjamins, 249-260.
Tomiche, Nada. 1964. Le Parier Arab du Caire. Paris: Mouton.
Van den Berg, Helma. "Gender and Person Agreement in Akusha Dargi." This volume, 153-168.
Wright, William. 1974. A Grammar of the Arabic Language, Translated from the German of
Caspari, and Edited with Numerous Additions and Corrections, vols. I & II, 3rd ed. Beirut:
Librairie du Liban.