Case and proto-Arabic, Part I

Abstract

It is a fundamental point of comparative and historical linguistics that genealogical affiliation can only be established on the basis of concrete linguistic features, the more central the feature the more important for classificatory purposes. While there is no absolute consensus about how a central linguistic feature be identified, it can be taken as axiomatic that long-term reconstruction and classification rests most fundamentally on phonological and morphological criteria. Of these two, Hetzron (1976b) has argued that it is the morphological which is the most important because morphology represents the level of grammar that is both more complex and more arbitrary in the sense that the sound-meaning dyad has no natural basis. Precisely this arbitrariness ensures that morphological correspondences are relatively unlikely to be due to chance.

Full text

Bulletin of the School of Oriental
and African Studies
http://journals.cambridge.org/BSO
Additional services for Bulletin of the School of
Oriental and African Studies:
Email alerts: Click here
Subscriptions: Click here
Commercial reprints: Click here
Terms of use : Click here
Case and proto-Arabic, Part I
Jonathan Owens
Bulletin of the School of Oriental and African Studies / Volume 61 / Issue 01 / February 1998,
pp 51 - 73
DOI: 10.1017/S0041977X00015755, Published online: 05 February 2009
Link to this article: http://journals.cambridge.org/abstract_S0041977X00015755
How to cite this article:
Jonathan Owens (1998). Case and proto-Arabic, Part I. Bulletin of the School of
Oriental and African Studies, 61, pp 51-73 doi:10.1017/S0041977X00015755
Request Permissions : Click here
Downloaded from http://journals.cambridge.org/BSO, IP address: 132.180.10.132 on 21 Dec 2015

Case and proto-Arabic, Part I1
JONATHAN OWENS
Bayreuth University
It is a fundamental point of comparative and historical linguistics that genealo-
gical affiliation can only be established on the basis of concrete linguistic
features, the more central the feature the more important for classificatory
purposes. While there is no absolute consensus about how a central linguistic
feature be identified, it can be taken as axiomatic that long-term reconstruction
and classification rests most fundamentally on phonological and morphological
criteria. Of these two, Hetzron (1976b) has argued that it is the morphological
which is the most important because morphology represents the level of gram-
mar that is both more complex and more arbitrary in the sense that the sound-
meaning dyad has no natural basis. Precisely this arbitrariness ensures that
morphological correspondences are relatively unlikely to be due to chance.
Though Hetzron's principles of genetic classification are surely coloured
by his experience in comparative Semitic and Afroasiatic, where there are
striking morphological correspondences to be found between languages widely
separated both geographically and diachronically (see 2 below), his principle
of morphological precedence, as it can be dubbed, may be taken as a general
working hypothesis. Unquestionably, morphological case belongs potentially
to the basic morphological elements of a language. Whether morphological
case belongs to the basic elements of a language family is, of
course,
a question
requiring the application of the comparative method. In Niger-Congo, for
example, case apparently does not belong to the proto-language,
2
whereas in
Indo-European it is a key element of the proto-language (Antilla, 1972: 366).
In Afroasiatic, to which Arabic belongs, the status of case in the proto-
language is, as yet, undecided. None the less, the assumption of a case system
within at least some branches of the language family, Semitic in particular
(Moscati et
ah,
1980), has had consequences both for the conceptualization of
relations within Semitic and for the reconstruction of the proto-language for
the entire family. Given the as yet uncertain status of morphological case at
the phylum level, I believe a critical appraisal of its status at all genealogical
levels to be appropriate. Within this perspective I seek here to elucidate the
interplay between case conceptualization and the reconstruction of one
proto-variety, namely Arabic. Given the importance of Arabic within Semitic,
conclusions reached regarding this language will have consequences for the
sub-family and beyond, as I will attempt to show.
The paper consists of five parts. Sections 1-3 appear in Part I, sections 4,
5 in Part II. In section 1, I set the stage by reviewing the basic genealogical
concepts which have been employed to describe the evolution of Arabic—Old
Arabic, Altarabisch, Neuarabisch and the like. In section 2, I briefly review
the status of case in the various branches of Afroasiatic. Here it will be seen
that a case system is not self-evidently a property of the entire phylum. In
section 3,1 turn to case in Classical Arabic,
inter alia
considering the descriptive
work of Sibawaih, who, I believe, was instrumental in defining the nature of
1
1 would like to thank Mauro Tosco for his stimulating comments of an earlier version of
this paper, and for being an agreeable critic to disagree with.
2
In an overview of Niger-Congo languages edited by Bendor-Samuel of the ten or so Niger-
Congo families, only the Ijoid languages appear to have some case marking (1990: 115). My
colleagues at Bayreuth, Gudrun Miehe and Carl Hoffman, both with long experience in Niger-
Congo, inform me that it is very unlikely that case belongs to the proto-language.
© School of Oriental and African Studies, University of London 1998

52 JONATHAN OWENS
Classical Arabic. Finally, in sections 4 and 5 I consider the evidence for case
in the modern Arabic dialects, addressing in particular the question of whether
the dialects should be seen as being the offspring of a case-bearing variety,
and if not, whether caseless varieties are innovative or go back to a caseless
form of proto-Semitic.
1.
Proto-Arabic and
its problems
Proto-Arabic has yet to be adequately conceptualized, either by Arabicists or
by Semiticists. A basic explanation for this, I believe, is the failure to draw the
modern dialects into any meaningful attempt to reconstruct proto-Arabic.
Three reasons play a role in this failure. The first has to do with the relationship
between Classical Arabic and the modern dialects, in particular, the fact that
the modern dialects have no official legitimization in the Arabic world. To set
the two on an equivalent basis, which is what a dispassionate comparativist
account must do (see Part II, sections 4, 5), could be interpreted as calling into
question the asymmetric diglossic relationship (Ferguson, 1959a) between the
high Classical (or Modern Standard) variety and the low dialect.
A second reason, I think, is based on the fact that the oldest detailed
accounts of Classical Arabic are undeniably older, by a range of some 1,000
years,
than any detailed accounts of the dialects. Coupled with this, to anticip-
ate the third point, is the assumed greater complexity of the Classical language
relative to the dialects (see e.g. Ferguson 1959a).
3
Linking these two perspect-
ives,
it is a relatively easy step to interpret the modern dialects as the simplified
or even bastardized offspring of an older, more perfect Classical variety (Mahdi,
1984:
37).
4
A third reason, related to the second, I believe is simply one of convenience.
The Classical language offers a ready-made starting point for the summary of
the history of Arabic. Fuck's prestigious Arabiya offers a history of Arabic
(subtitle, Untersuchungen zur arabischen Sprach- und Stilgeschichte [my
emphasis]) which starts with the literary language and makes little serious
attempt to incorporate dialect material (however denned, see Spitaler's review
of 1953).
Taking these three perspectives in order critically and beginning with the
first, while comparing the dialects with the Classical language on an equal
footing would be relatively uncontroversial among most (perhaps not all)
linguists, the comparison is likely to be misunderstood by many for cultural
reasons, where it may well be assumed that a declaration of linguist equivalence
(as it were) is tantamount to a statement of cultural and political equality
between the dialects and the classical language. Logically, however, linguistic
reconstruction is independent of cultural and political considerations.
Turning to the second point, it is clear that the relative time of diachronic
reconstruction is what comparativists work with, rather than the absolute time
of the Gregorian calendar. In terms of absolute time the earliest, extensive
3 Sections 7 and 9 of Ferguson's classic article on diglossia, where he seeks to prove the greater
grammatical and phonological complexity of the high as opposed to the low varieties is surely
the weakest part of
his
characterization of
diglossia.
In Arabic there can be found instances where
modern dialects have equally or more complex structures than the classical language.
4 Mahdi admonishes us to study the dialects in order that the negative influences (sicknesses,
amrad) of the dialect on the standard language (fusha) be eradicated. That is, one studies the
dialects not to shed light on the Arabic language as a whole, but, rather to purify the standard.
Where Chejne's (1969) 'The Arabic language' mentions the dialects it is often in a derogatory
context (e.g., p. 84), though the discussion of language policy in the final chapters is somewhat
balanced.

CASE AND PROTO-ARABIC, PART I 53
sources
of
Classical Arabic date from
the
seventh
or
eighth centuries.
5
These
are
far
younger than
the
earliest sources
for
Akkadian, dating from about
2,500
B.C.
No
Semiticist, however, would argue that Akkadian therefore must
be assumed
to
represent
the
earliest state
of the
Semitic languages. What
is
criterial
is the
relative time scale that emerges from
the
application
of the
comparative method, which shows, inter alia, that
the
younger
(in
absolute
terms) Arabic language contains
an
older inventory
of
phonological elements
(e.g.
a
complete
set of
differentiated emphatic correspondences, Moscati
et al.,
1980:
24) than does the older Akkadian. Relative
to
the ur-Semitic phonological
inventory
one can say
that Arabic
is '
older'
(in the
sense that
it has
preserved
older traits) than Akkadian.
An
analogous argument applies
in
principle
to
any comparison between
the
modern dialects
and the
classical language:
a
priori
(i.e. prior
to the
application
of
the comparative method and/or internal
reconstruction)
one
does
not
know whether
a
given trait
in a
dialect
is '
older'
(in relative terms), 'younger'
or
'equivalent'
to a
comparable trait
in the
classical language.
Finally, considering the third point, convenience is
no
substitute
for
consist-
ent application
of a
well-tested methodology. Just
why few
serious attempts
have been made
to
reconstruct proto-Arabic
is a
question
for the
historian
of
Arabic
and
Semitic studies.
6
One
reason,
I
suspect,
is the
(understandable)
preference among philologists
for the
written word (Classical Arabic) over
the
spoken (dialects). What
is not
written
is not
fully legitimate. Linguistically-
orientated comparative studies
of
Arabic,
and
more generally oriental
and
Islamic studies,
for the
first half
of
this century have tended
to be
dominated
by philologists.
7
In
recent years, since
c. 1960,
there
has
been
a
remarkable
growth
of
interest
in
modern dialectal
and
sociolinguistic aspects
of
Arabic
(see
e.g.
part
2,
1987
of
the journal
Al-'Arabiyya).
These, however, have been
largely restricted
to
descriptive
and
so-called theoretical linguistics,
to the
exclusion
of
comparative perspectives.
With these preliminary points
in
mind,
it is
time
to
turn
to a
summary
of
the concepts which have been used
to
characterize Arabic
in
diachronic terms.
I.I.Arabic,
Old and New
The most frequently used terms
to
differentiate
the
main varieties
of
Arabic
are
Old
Arabic
(OA) and
Neo-Arabic (NA).
The
terminology
is
particularly
widespread
in
German scholarship (Altarabisch
vs.
Neuarabisch).
The
values
5
1 would leave open the question of whether one should place the earliest attestations of
Classical Arabic in the seventh or eighth century. One potential source, the eighth-century papyri,
is probably too fragmentary for use to reconstruct a complete grammar, while a second, the
Quran, besides being stylistically unique, is associated with a range of problems (absolute dating
of oldest extant text, the variant readings) which renders its inclusion in the present study
impractical. Barring full-scale studies of these sources, I believe it legitimate for two reasons,
when referring to Classical Arabic, to concentrate on the variety described by Sibawaih. First,
Sibawaih is unequivocally datable to the second half of the eighth century, and secondly, the
variety he describes easily surpasses in detail any other early sources, effectively establishing a
standard by which other varieties, both earlier and later, are measurable. None the less, given the
open questions related to pre-Sibawaihan Arabic, I will continue to refer to Classical Arabic (or,
as defined in Part II, section 5, Old Arabic) as the variety attested in the seventh and
eighth centuries.
5
One of the few explicit attempts is Cowan (1960). As this concentrates exclusively on
phonology it is not directly relevant to the present study.
7
1 think this prejudice is Ii betrayed, for instance, in Brockelmann's (1982 [1908]: 6) contention
that Akkadian is the oldest independent Semitic language. It is undeniably the oldest one attested
in writing. There is, however, no way of proving that it is older than proto-Arabic or proto-
Ethiopic (as opposed to an undifferentiated proto-West Semitic such as Brockelmann assumes).
On comparative grounds, Diakonoff (1988: 24) assumes a dialectal differentiation of Semitic as
early as 4,000-5,000 B.C., which would give adequate time for a differentiated ancestor of Arabic
to have arisen, parallel to Akkadian.

54 JONATHAN OWENS
which these terms take vary from scholar to scholar and context to context.
Old Arabic, for instance, may designate the old pre-Islamic languages/dialects
of the Arabian peninsula known through epigraphic inscriptions (Miiller, 1982).
More commonly, Old Arabic designates a variety attested in old texts and
opposed to Neo-Arabic, which minimally includes the modern dialects. This
opposition, however, is developed in different
ways.
Brockelmann (1982 [1908])
appears to restrict Neuarabisch to actually attested contemporary dialects. On
the other hand, his Altarabisch encompasses three distinct sub-varieties, the
classical language of
poetry,
quranic Arabic, and the old Arabic dialects (1982:
23).
'Alt' and 'neu' for Brockelmann are thus largely objects datable by
absolute time. Given his suspicion of proto-Semitic as a concrete entity (1982:
4),
it is not surprising that Brockelmann made no attempt to link the two
varieties in a single systematic entity.
For Blau
(1981,
1988), Old and Neo-Arabic are given a more concrete
linguistic characterization, designating linguistic types.
8
Blau suggests a number
of contrasts between the two, the most important probably being the presence
of case in Old Arabic and its lack in Neo-Arabic (see also Fuck, 1950: 2; see
sections 2, 4). On the basis of the linguistic differences, Blau takes an important
methodological step in reading an historical dimension into the typological
differences. Elements of the neo-Arabic type can be discerned in various types
of early writing, dating from as early as the eighth century. On the basis of
these older texts Blau develops a three-fold model for the development of
Arabic, Old Arabic, more or less coterminous with the Classical language,
Middle Arabic, the language of
the
older texts which deviates from Old Arabic,
and Neo-Arabic, the final phase in the development where, as it were, the
mixed nature of Middle Arabic (some old elements, some new) gives way fully
to the situation found in today's dialects. Middle and Neo-Arabic developed
linearly out of Old Arabic because of the loss of classical elements under the
pressure of the rapid expansion and urbanization of Arab culture in the Islamic
period (Fuck, 1950: 5 ff.).
Both Brockelmann and Blau anchor their models in Old Arabic, a variety
whose features are fixed and stable and which provides the starting point for
the development of their analysis of
Arabic.
A very different approach is taken
by Corriente (1976).
9
For him Classical Arabic is itself the endpoint of a
development within the complex of varieties of
Old
Arabic is itself the endpoint
of a development within the complex of varieties of Old Arabic (also Rabin,
1955).
Its crystallization in the late eighth century was determined '... by native
grammarians whose main concern was to set up a standardized, socio-
linguistically biased type of Arabic for formal register purposes, i.e. Al-
'Arabiyya' (1976: 62). Although one can take issue with the mechanistic role
of the grammarians in this characterization (see 3.2)—their work after all
reflected trends in the society as a whole—this summary does I believe correctly
describe one of the important effects of the grammarians' work. Corriente
adduces many examples supporting the contention that Classical Arabic, the
fusha, is itself the result of a filtering process whereby variation is classified,
pared and brought within manageable bounds. If Classical Arabic is a well-
defined variety, it has its origins in varieties which are not. Since my own
8 Blau's concept of Old and Neo-Arabic is taken over more or less intact by Fischer
(1982:
83
ff.), see Part II, 4.2.
9 A diachronic, yet neutral usage of the term Old Arabic can be noted where Blanc
(1964:
183,
n.8) uses the term to refer to the putative ancestor of the modern dialects. With Blanc the term
is,
so to speak, an article of faith, as he does not attempt to give it any linguistic content.

CASE AND PROTO-ARABIC, PART I 55
position expands on Corriente's, I will leave documentation of Corriente till
section 3.1, where Classical Arabic is treated in greater detail.
As a corollary to Corriente's ideas, I would add that it is no contradiction
to see the modern dialects as developing out of the same material as did the
classical language.
1.2.
Related ideas
A number of scholars have contributed to the discussion of the relationship
between Old and Neo-Arabic without referring specifically to it in such terms.
Ferguson's (1959b) contribution in fact antedates the work of Blau and
Corriente, and may well have served as a methodological model for Blau's
work in particular. Ferguson, like Blau, taking Classical Arabic as an anchor,
argued that the modern dialects formed a homogeneous contrast to the
Classical language. The modern dialects were said to have arisen in the form
of
a
koine that took shape in the military camps of the original Arab conquerers
in the period of Islamic expansion. This elegantly simple idea ran into problems
as more work was done on Arabic dialects, and it became clear that the dialects
themselves differ probably as much amongst each other as they do from the
Classical language (see e.g. Kaye, 1976 and Part II, section 4 of this paper).10
A work which took a philological perspective a diachronic step deeper is
Diem's (1973) study of case endings in the Arabic words found in the Aramaic
inscriptions of the Arabs of Nabataea in southern Jordan, dating from about
100 B.C.11 Diem shows that Arabic personal names found in the inscriptions
did not show traces of a living case system. If Diem's interpretation of the
data is correct, and Blau himself does not explicitly refute it (1988: 11 ff.), it
would mean that the oldest written evidence of Arabic is characterized by a
linguistic trait, the lack of case endings, which is otherwise said to be a
characteristic par excellence of Afeo-Arabic (see 2.2).
I will make reference to these two perspectives in the following, Ferguson
implicitly for the methodology in section 4 and Diem for the purported
antiquity of Arabic case inflection.
1.3. An
impasse
Summing up this section, by and large there are two opposed conceptualizations
of the status of Old and Neo-Arabic. The one sees the classical language as
either identical with Old Arabic, or closely related to it—for example, one of
its representatives. Neo-Arabic developed out of Old Arabic via a process of
simplification (see 4.2). A second sees the classical language as itself
the
product
of evolutionary forces within Old Arabic, and developing further in ways I
will specify in sections 3-5.
An obvious problem with the first perspective is the undeniable fact that
Classical Arabic in its most detailed single account, that of Sibawaih, was
characterized by a good deal of internal variation (see section 3). I hesitate to
speak of Old Arabic dialects here {pace Rabin, 1951), as this prejudges and
pre-categorizes the many attested variant forms in ways not necessarily inten-
ded by Sibawaih and other early grammarians (e.g. Farra').12 In any case, Blau
10
A more recent work which uses the classical language as the first anchor is Versteegh's
(1984) attempt to derive the modern dialects from the classical language, via a pidginization stage
(see Owens, 1989, for criticisms). A number of writers who examine the linguistic history of
Arabic assume positions which I criticize here. Garbell (1958), for example, follows Fuck in using
Classical Arabic as the initial anchor in Arabic diachrony.
11
The Nabataean Arabs used Aramaic as their literary variety.
12
This is not to gainsay the invaluable work of Kofler (1940), Rabin (1951) and others. None
the less, I know of no critical studies which link the many attested linguistic variants noted by
the Arabic grammarians to the conceptual terminology by which these same grammarians
understood these forms. A variant form in Sibawaih most likely has a different status from a

56 JONATHAN OWENS
tends to assume that there was a single
'Arabiyya
which was largely coterminous
with the language of the Bedouins. It is not completely clear how Blau would
integrate this internal variation into his developmental model. Fischer (1995)
provides a more explicit model than Blau. He draws a contrast between Old
Arabic, by which Classical Arabic is understood, and Old Arabic dialects. The
modern dialects are the offspring of Old Arabic dialects, not of Classical
Arabic. The difficulty with this formulation is that it fails to explain the
overwhelming similarities which are found between Classical Arabic and the
modern dialects (and by implication, between Classical Arabic and the Old
Arabic dialects). Deriving the modern dialects from the Classical language,
pace Blau, avoids the anomaly produced in Fischer's model (see 4.2).
Probably the most prominent difference between Old Arabic, however
defined in the past literature, and modern dialects, the one at the head of the
differentiating list (Fuck, 1950: 2; Blau, 1988: 2; Fischer, 1982: 83), is the case
and mode inflectional system. It is thus time to turn to a detailed consideration
of this phenomenon in the debate about the relationship between Old Arabic,
proto-Arabic and the modern dialects. I will concentrate exclusively on the
central feature of Arabic case marking, short-vowel nominal inflection.
2.
Case in the
Afroasiatic phylum
The phylic unity of Afroasiatic rests on striking correspondences found in all
branches of the phylum within the verbal and pronominal systems. Personal
markers (in pronouns and/or verbs), for instance, in -k ' second person' (sg.
and pi.),
-n
V'
lp.'
(m in Chadic), s ~ s ' third person', a second and third person
plural formed from the consonantal person marker
+ u +
(nasal) (nasal lacking
in Chadic) are found in all branches. Verb conjugations differentiated according
to prefixing and suffixing classes (in various distributions) are found in three
of the branches, Semitic, Cushitic and Berber, while the other two, Egyptian
(suffix only) and Chadic (prefix only) have conjugations with clear correspond-
ences to one or the other (see e.g. Rossler, 1950; Voigt, 1987; Diakonoff, 1988).
It is therefore equally striking that only two of the five
13
branches, Semitic
and Cushitic, have languages with morphological case systems. Even within
these two branches there is a good deal of variation among individual lan-
guages, and the question of the extent to which Cushitic case corresponds to
Semitic is far from clear. In 2.1 I briefly summarize the situation for Cushitic,
and in 2.2 that for Semitic.
2.1.
Case in Cushitic
While many of the Cushitic languages have case systems, it is by no means
clear that they derive from a proto-Cushitic case system. Such a proposal has
variant form in Suyuti, who lived 700 years after Sibawaih, and both of
these,
in turn, may differ
from the modern concept of 'dialect'.
131 assume a very traditional Afroasiatic family tree, well aware that there are classificatory
questions at all levels. I don't think, at this point, that such questions bear crucially on the present
treatment of
case,
however.
Berber and Chadic do not have case. Since Egyptian orthography did not mark short vowels,
whether or not ancient Egyptian had a case system is difficult to know. Even if Calender's (1975)
attempt to reconstruct ancient Egyptian cases on the basis of the functional behaviour of verbal
forms is on the right track, his attribution of formal values to them (e.g. nominative
-u,
accusative
-a) is speculative at best, at worst, no more than the filling in of an Egyptian consonantal text
with short vocalic values taken over from Classical Arabic. Petracek
(1988:
40) does not reconstruct
a case system for ancient Egyptian.

CASE AND PROTO-ARABIC, PART I 57
been put forward by Sasse (1984) where a proto-Cushitic case system with
nominative opposed to accusative is postulated.
14
Against this Tosco
(1993,
in the spirit of Castellino, 1978: 40) has argued
that the origin of many Cushitic nominative markers lies in a focus morpheme.
Tosco's argumentation is based on both universal and formal considerations,
the main features of which are as follows.
First he notes (as have a number of scholars before him) that the nominat-
ive-absolutive (roughly = accusative) distinction in Cushitic is typologically odd
since it is the nominative which is the marked form by a number of criteria.
It is the nominative noun, for example, which is morphologically marked (see
(1) vs. (2)),
15
e.g., Oromo
(1) namicc-i ni-dufe
man-nom. prev.-came
'The man came'.
(2) namiccd arke
man(abs.) saw
' He saw the man'.
The unmarked absolutive serves as the basis for further inflections, cf. genitive
(ka)
namicc-da'
of the man'. Furthermore, the nominative has a more restricted
distribution, limited only to the subject, and is far less frequent in texts. In
these points the Cushitic nominative has close affinities to grammaticalized
topics.
Reviewing the literature on Cushitic languages, Tosco shows further that
a suffix -/ throughout the branch, for example, Highland East Cushitic (e.g.
Sidamo min-i' house-i'), Central Cushitic (Awngi -ki), is found which marks
not only subjects but other topicalized constituents as well. Where -i has been
grammaticalized as a subject marker other markers develop as topicalizers.
An -n is particularly common in this function, as perhaps exhibited in the
Harar Oromo.
(3)
namiccd-n
arke
man-topic saw
'
I
saw the man'.
Relating the Cushitic data to Semitic, it is furthermore noteworthy that
Sasse's reconstruction, nominative
*-//-«,
absolutive
*-a
does not self-evidently
correspond to the three-valued Semitic system. In fact, as Tosco shows, the
only widespread nominative-like (Tosco's topicalizer) inflection on full nouns
is -Z.
16
It is true that a 'nominative' u is found throughout Cushitic, though
as part of the article or demonstrative, not as a nominal affix (e.g., Oromo
14 Sasse leaves open the possibility that other cases might be reconstructive. Diakonoff (1988:
60) proposes an ' abstract' proto-Afroasiatic case system, characterized above all, apparently, by
its abstractness. Formally there was an opposition between i~u
vs.
-a~0, though how this system
worked functionally at the proto-Afroasiatic stage is not spelt out in detail. Diakonoffs
reconstruction rests largely on data from Semitic and Cushitic and hence is open to all the
criticisms of postulating a proto-Cushitic case system contained in this section. Furthermore, his
entire reconstruction of Afroasiatic case is based on the dubious assumption that the proto
language was an ergative one. His claim (1988: 59) that Oromo (and similarly, I suspect, his
claims for Beja, Sidamo and Ometo) is an ergative language (or has traces of an ergative system?)
is mysterious (see Owens, 1985).
15 Greenberg (1978: 95, universal 38) notes that if there is a case system the subject of the
intransitive verb will be marked by the least marked case (also Croft, 1990: 104).
16 In a number of Highland East Cushitic languages (Hudson, 1976: 253 if.) the nominative
(or topic) form is phonologically determined relative to the absolute form: if the absolute ends in
a front vowel or -a it is -i, if in -o it is -u.

58 JONATHAN OWENS
kuni 'this-nom.', vs. kana 'this-abs.').
17
Moreover, Cushitic case marking is,
unlike Semitic, overwhelmingly phrase final (Central Cushitic Awngi [Hetzron,
1976a: 37] and Oromo are exceptional). In Somali, for instance, where the
nominative subject is generally shown by lowering a tone from H to L, one
has in absolutive case,
nin
' man '->nom
nin,
but nom. nin-ku ' the man', where
the determiner assumes the low tone, allowing nin to re-assume its unmarked
absolutive form with high tone (Saeed, 1987: 133). Any attempt to link this
Cushitic data to Semitic case would have to account for significant structural
mismatches. To these problems can be added that of the Cushitic genitive,
which neither Sasse nor Tosco integrates into his model, both being cognizant
of the special problems accompanying the task.
To summarize this section, while it is certainly correct, paraphrasing and
changing Sasse's formulation slightly, to speak of certain Cushitic endings as
bearing ' a striking resemblance to certain formatives ... in other Afroasiatic
branches' (1984: 111), it does not appear possible, at this point at least, to
link these directly to Semitic case markers. Even assuming a link to be possible,
it would not automatically follow that it would be made in terms of case.
Indeed, given that it is only the Semitic branch (following Tosco for Cushitic)
which unequivocally has a proto-case system, it would not be surprising if
such a system developed at the proto-stage of Semitic out of markers of
another type.
2.2.
Semitic
case
It is not my purpose here to review the literature on case in Semitic. The
situation in Arabic will be reviewed in detail in any case in the next two
sections. For present purposes two basic points need to be made.
First, although a three-valued case system (nominative -u, accusative -a,
genitive -/) has to be reconstructed for proto-Semitic, only a minority of the
well-attested Semitic languages have it. Moreover assuming Moscati et al.'s
tri-partite classification of Semitic into North-east, North-west and South-west
sub-branches, caseless languages (or dialects) are attested in each sub-branch.
The earlier stages of Akkadian (North-east) had it, though after 1,000
B.C.
the
case system showed clear signs of breaking down.
18
Most of the North-west
Semitic languages did not have case (Hebrew, Aramaic, Phoenician), only
Ugaritic
19
and Eblaitic (probably) possessing it. The situation with Ugaritic in
17 Paradoxically, the Semitic article/demonstrative system is neither particularly unified, nor
do most of the Semitic languages (including Classical Arabic and most stages of Akkadian, Von
Soden, 1969: 47) show case differentiation in it (Moscati et
al.,
1980: 110).
18 The outside observer may be slightly disquieted by Von Soden's comment (1969: 80) that
even in Old Assyrian and Old Babylonian exceptions to the expected case-marking system occur.
He attributes these to orthographic errors or to ' bad pronunciation' (?). A closer study of such
' errors' would be interesting.
19 Even Semiticists are not united on which Semitic languages demonstrably have case systems.
Rabin (1969: 161), for example, cautions that there is not enough data to reconstruct case syntax
in Ugaritic and hence does not include Ugaritic among the attested case-bearing Semitic languages.
He has a two-valued case system for Ga'az (as opposed to Weninger's 1993 one), but would
apparently rather identify the Ga'az 0 ' nominative' etymologically with the Akkadian absolutive
(i.e.
lack of morphological case) than with Classical Arabic
-u
(1969:
196). Some modern Ethiopian
Semitic languages (e.g., Amharic
-ri)
have secondarily developed an object
case,
sometimes sensitive
to definiteness features. Barth's (1898: 594) assumption that proto-Hebrew had case is based
crucially on the
assumption
that proto-Semitic had case and only case (i.e. no parallel caseless
variety, see section 5). Reading between the lines of
his
article, however, it is clear that one could
no more justify the reconstruction of a case system from internal Biblical Hebrew sources than
one can an Arabic case system from the modern Arabic dialects.
There are no attempts to my knowledge to explain how the assumed proto-Semitic case system
disappeared so completely throughout the family. Moscati (1958), for example, is not so much an
attempt to justify the assumption of a family-wide proto-case system as to determine the length
of the assumed endings. Given the significant counter-evidence that some Semitic languages/
varieties never had a case system, it may minimally be expected that the assumed 'disappearance'
of case from these varieties be given a unified explanation.

CASE
AND
PROTO-ARABIC,
PART
I 59
this respect is not very satisfying, as the only direct evidence for case endings
conies from the word-final symbol for the glottal stop. In Gordon's (1965)
lexicon, these amount to barely ten noun lexemes from which the entire case
system must be constructed. It is noteworthy that neither Rabin (1969, see n.
19) nor Petracek (1988: 39) list Ugaritic among the case-bearing Semitic lan-
guages. Among the South-west Semitic languages Classical Arabic has it,
though Ga'az (in absolute terms some 350-500 years older) probably did not,
at least not in a way which self-evidently corresponds with the three-valued
proto-Semitic system (see n. 19). The modern Ethiopic Semitic languages do
not have it (see n. 19), nor do the modern Arabic dialects. The modern South
Arabian languages do not have it, while the situation for epigraphic South
Arabic is unclear due to the script.
From a distributional perspective one can approach the problem in two
ways.
First, it can be assumed that the cases are original and lost in those
varieties where not attested. This, of course, is the approach taken by most
Semiticists (e.g. Moscati et al., 1980: 94) and could be said to be supported
indirectly at least by the situation in Akkadian where the breakdown of a case
system is diachronically attested. A second approach would be to view the
caseless situation as original, the Akkadian and Classical Arabic system as
innovative. This is problematic in view of the fact that Akkadian is the oldest
of the languages in absolute terms and that the case system in the two languages
is in general terms comparable. It is unlikely that the two innovated in the
same way independently of each other, and if they did not, a common origin
pushes the case system back into the proto stage. A third solution is that the
proto-language had two systems (two dialects as it were), one with case, one
without. I will be developing this perspective in the rest of this paper. For the
moment it suffices to note that postulating a caseless variety at the proto-
Semitic stage is supported by family-internal distributional facts, namely the
broad range of Semitic languages which do not have case systems (to turn the
argument introduced in the previous point on its head), and the arguments of
the preceding section, where it was seen that Semitic case, within Afroasiatic,
is probably innovative.
Secondly, it can be noted that in the Semitic languages with a case system
there are contexts where, in synchronic terms, the system is neutralized. In the
Akkadian genitive relation, the possessed noun does not bear case (or appears
in the so-called absolute form) before a nominal possessor, and before a
pronominal possessor generally only when the possessed noun ends in a vowel
(Von Soden, 1969: 82 ff., 189 ff.).
(3) bel-0 bTt-i-m assas-su
master-0 house-gen.-m.wife-his
' master (0 =
"
absolute " case) of the house'
In Classical Arabic the neutralization, at least in traditional accounts (see
3.2.3), occurs in pausal position. Besides raising questions of the functional
centrality of case in Semitic (see 3.1 below), the presence of these 'caseless'
contexts suggests that even those Semitic languages with morphological case
systems possessed traces of the caseless variety. I will touch on further case-
related comparative aspects of Semitic later (3.3, 4.2, 5).
Brief though the remarks in the present section are, they are consistent
enough to underscore Petracek's conclusion (1988: 41, see also Rabin, 1969:
191),
based on comparative Afroasiatic data, that 'Die pragnant gebildete
Struktur des Kasussytems im Semitischen
{-u,
-i, -a) diirfen wir als eine semi-
tische Innovation ansehen.' If this point is accepted, however, there emerges

60 JONATHAN OWENS
a further Semitic-internal issue, namely at what point Semitic itself developed
a case system and whether this development represented the ancestor of all
Semitic languages or only some of
them.
In the rest of
this
paper I will attempt
to show that a detailed consideration of the issue for proto-Arabic will provide
one important component in answering the question.
3.
Classical Arabic
It should by now be becoming evident that the assumption that there is a clear
distinction between those Semitic languages with case systems and those with-
out, the latter possessing, in this respect, an older trait than the former, is
perhaps not so problematic as is represented in the textbooks (e.g. Moscati
et al., 1980; Fischer, 1982a). In part 3 I examine the status of
case
in Classical
Arabic in greater detail, using two sources. In 3.1 I summarize the work of
Corriente, which is not adequately integrated into the debate about Old Arabic,
perhaps because his views about case (and other matters) in Old Arabic are
somewhat iconoclastic. In 3.2 I turn to the grammarian who, if not the
' founder' of Classical Arabic, doubtlessly played a more pivotal role in expli-
citly denning its form than any other individual, namely the eighth-century
grammarian Sibawaih, in order to gain a more precise insight into the nature
of
the
Classical Arabic which he defined. This account will initiate the compar-
ison between the classical language and the modern dialects, a necessary step
in the discussion of
Blau's
theory deriving the modern dialects from the classical
language.
3.1.
Corriente
In a series of articles
(1971,
1973, 1975, 1976) Corriente argued that Classical
Arabic stood at the end of a development and that its crystalization in a more
or less
fixed
form was due in large part to the efforts of the Arabic grammarians.
Many of the points he makes relate to the case system. These include evidence
of
two
main sorts: linguistic internal interpretations and an examination of the
philological record.
The first perspective is prominent in his 1971 article where he showed that
the functional yield of Classical Arabic cases—roughly those contexts where
a difference of meaning can be affected by a change of
case
alone—is extremely
low. While one may agree with Blau (1988: 268) that case systems generally
have a high degree of redundancy, the fact remains that Arabic case is function-
ally not deeply integrated into the grammar. The case forms, furthermore, are
not well integrated into the morphology
(1971:
47). They are marked by a lack
of allomorphy, being without exception tacked on to the end of the word, with
little morpho-phonological interaction with either the stems to which they are
suffixed or the items which may be suffixed to them. To this can be added the
fact that, unlike in many languages, they are not subject to variational rules
based on animacy and/or definiteness.
In his 1975 article Corriente cites deviations from the classical norms found
in various verses of the Kitdb
al-Aghani,
including an inflectionally invariable .
dual (1975: 52; cf. Rabin, 1951: 173, and also of course Q. 20: 63), the mixing
up of cases
(1975:
57), or their complete absence
(1975:
60).
Corriente's explana-
tion for these phenomena, and for the development of Middle Arabic out of
Old Arabic, was to postulate a caseless form of Arabic formed in pre-Islamic
times along the NW Arabic borderland in Nabataea (1976: 88; expanding on
Diem, 1973, see 1.2 above). Associated with commercial centres, this variety
of Arabic would have quickly acquired prestige status and in the aftermath of
the early Islamic Arabic diaspora served as a model for the development of

CASE AND PROTO-ARABIC, PART I 61
caseless Middle Arabic in urban contexts. Note that he does not break with
Blau completely, in that he sees the caseless varieties arising out of the border
contacts.
The present study agrees with Corriente on the need to recognize a caseless
form of Arabic existing contemporaneously with case varieties. Where issue
can be taken with his account is the readiness to postulate a simple link
between one variety of Old Arabic (Nabataean Arabic) and the modern dia-
lects.
The difficulties in drawing such a simple linkage will become evident in
the discussion in part II, sections 4 and 5. As a general introductory remark,
however, it may be noted that integrating the modern dialects into the recon-
struction of proto-Arabic will yield results which do not self-evidently replicate
the linguistic entities denned by the Arabic grammarians or by the epigraphic
record (e.g. caseless Nabataean Arabic). This follows from the different
methods and goals of the comparative approach and that of the Arabic
grammarians, to whom I turn presently, as well as from the very fragmentary
nature of the epigraphic record. A full reconstruction of proto-Arabic requires
an independent and detailed definition of each of these components before
they are put together into a larger picture.
3.2. Sibawaih
Without a full-scale study of all the material to be found in all grammatical
treatises, the reasons for concentrating on Sibawaih in the first instance are
self-evident. His Kitab is, even by modern standards, a paragon of detail and
completeness (if not necessarily of organization and clarity). More importantly,
it is arguably the only work where a large body of directly observed linguistic
usage has been systematically recorded. This is not to deny that interesting
material is to be found in the later treatises. None the less, later grammarians
were dependent to a large degree on Sibawaih for the simple reason that the
'Arabiyya came to consist of a more or less closed set of data by the tenth
century,
20
rendering observations on the contemporary spoken language super-
fluous (see below). Nor does it mean that sources other than the grammatical
treatises should not be invoked. However, as Corriente (1975: 43) points out,
non-classicisms tend to be edited out of even less formal poetic corpora (see
n. 5 for other sources).
There are two aspects of Sibawaih's work which may be kept in mind when
interpreting his observations on Arabic. The first is that Sibawaih was con-
fronted by a mass of variant forms which he evaluated in his own inimitable
style,
to which I shall turn presently. By and large, in the later grammars,
written after the end of the ninth century, the variant forms were either
excluded altogether or, in the more detailed grammars (e.g., Ibn al-Sarraj, Ibn
Ya'ish),
treated in addenda to the general rules. If the rare, new material was
added, it was nearly always by reference to forms recorded from persons or
tribes contemporaneous with Sibawaih or earlier.
21
The second point relates to Sibawaih's attitude towards the linguistic data
he described. Carter
(1973:
146; see Ditters, 1990: 131 for criticisms) has
contrasted Sibawaih's 'descriptivist' approach with the ' prescriptivism' of
later grammarians. This contrast of
styles,
as noted in the preceding paragraph,
20 The work of the lexicographers, for which there are relatively few critical modern studies,
may have to be excepted here. Unlike grammar, lexicography deals with an open-ended system.
Adding a word to the lexicon rarely changes the lexicon in the way that adding a rule to the
grammar potentially changes the entire grammar. When the lexicon of the 'Arabiyya became a
closed set is as yet an open question.
21 In this respect the analogy with the history of Arabic linguistic thinking breaks down; there
were conceptual breakthroughs after the tenth century, though in areas of thinking ' adjacent' to
the core areas of morphophonology and syntax, such as pragmatics and semantics.

62 JONATHAN OWENS
reflects a general reorientation in the definition of what sort of data are
allowable in the definition of the 'Arabiyya. It is not clear, however, what is
to be understood by descriptivism, on which Carter does not elaborate.
Understanding Sibawaih on this point is important to understanding and
defining his theoretical linguistic thinking, the nature of the raw linguistic data
which his detailed observations make accessible, and the motivations for and
mechanisms by which grammars generally are developed. Baalbaki (1990: 18)
has made the important point that the grammarians, confronted by a mass of
linguistic data, did not 'content themselves with a purely descriptive exposition
of linguistic material, but attempted to present this material within a
coherent
system ... ' (my emphasis). Sibawaih, he observes, was the key figure in this
process. In the next three subsections I discuss aspects of this system, with
special reference to case-related problems.
First, however, it is relevant to mention that the very terminology of case
marking among the grammarians may bear on the question of the existence
of case in Old Arabic. The two oldest grammatical works, one (Sibawaih's)
definitely from the end of the eighth century, the other equally old or only
slightly younger, utilized case form as a central formal criteria for organizing
their exposition of syntactic structures. Grammars and case marking go hand
in hand in the history of Arabic grammatical theory. Khalaf al-Ahmar's
Muqaddimafi
'l-Nahw
is a short, practical grammar (see Owens, 1990, ch. ix),
whereas Sibawaih's Kitab is one of the most detailed grammars of Arabic ever
written. Sibawaih, in particular, goes to considerable pains at the very beginning
of his grammar
(i,
2.1 ff., see Baalbaki, 1990) to distinguish lexically determined
from syntactically determined short vowels functionally, the latter of course
being the case markers. Sibawaih's terminology is as follows:
(4a) Short
vowel terminology
in
Sibawaih
lexical morpho-syntactic phonetic value
damma raf u
fatha nasb a
kasra jarr i
It appears that the
'
discovery' of functionally differentiated vowels was pre-
ceded by a time when the same terminology was used undifferentiated for
vowels of both types. Such a system is still in evidence in Farra"s terminology
(Owens, 1990: 159).
(4b) Short
vowel terminology
in Farm'
lexical morpho-syntactic phonetic value
damma ~ raf raf u
fatha ~ nasb nasb a
kasra ~ khafd khafd i
In Farra' the morpho-syntactic values also are used to describe the phonetic
values of lexical vowels, so that the vowel in umm (i: 6) is described as
raf.
22
This supposition finds support in Versteegh's study on early Arabic gramma-
22 That Farra"s terminology should be the ' older', though he lived a generation after Sibawaih
(n.b.
in absolute time), may be explained by Tahnon's theory (e.g. 1990) that Farra' represented
an older grammatical tradition than did Sibawaih. There may be other explanations, too, but this
is not the place to consider Talmon's ideas in detail.
I would tend to accept Farra"s explanation for the variant -i of
al-hamd-i
li
lldhi
(i:
3) among
some Bedouins, namely that the nominative is assimilated to the following -i of li- within the
compound-like unit that has arisen due to ' frequency of
use'.
None the
less,
the example illustrates
(1) the convenience of not having a distinctive terminology for case vs. lexical vowels, and (2) the
non-case functional value of final nominal vowels among at least some groups of speakers.

CASE
AND
PROTO-ARABIC, PART
I 63
tical theory. He shows (1993b: 125) that the quranic exegete Muhammad
al-Kalbi (d. 763), who lived a generation before Sibawaih (d. 793), used the
term damm for an u vowel ' within a word', as an ' ending' (Versteegh does
not specify what sort here) and for a nunated noun. Nasb is used for a lexical
vowel a, an ending and for nunation, and similarly for the other terms in the
lists in (4).
Rather than consider the data in terms of their implications for an under-
standing of the development of Arabic linguistic theory, which has been the
main focus of interest in such data to date, they may be interpreted in terms
of the present question of the status of case endings in the history of the
language. It would appear that Sibawaih made explicit two aspects of vocalic
variation, one lexically the other morpho-syntactically determined, which
existed in the language he described.
One may ask here whether the variation and imprecise distinction between
lexical and morpho-syntactic vowels found in Farra' and other early linguists
and commentators do not originate in the fact that there existed varieties of
Arabic, studied by Farra', in which vocalic variation at the end of words did
not represent case endings, i.e. caseless varieties of Arabic. In such circum-
stances a consistently differentiating terminology would, of course, have been
unnecessary. Farra"s 'imprecise' terminology would thus reflect not a less
differentiated grammatical thinking than Sibawaih's, but rather its application
to a different data base. This perspective is admittedly speculative, though the
idea of relating early terminological problems to actual language forms should
be pursued further.
23
3.2.1.
An
example
The following example will serve as an introduction both to Sibawaih's treat-
ment of case in Classical Arabic and to the way in which he processed linguistic
data. From ch. 24 on (i, 31 ff), Sibawaih considers some fairly complex data
in which he is concerned to define the case form of a topic and/or agent noun
(see Khan,
1988:
25
ff.
for
discussion).
A basic contrast is shown in (5a)
vs.
(5b).
(5a) zayd-un laqi-tu akh-a-hu (i: 32.16)
Zayd-nom. met-I brother-acc.-his
'As for Zayd, I found his brother'.
(5b) zayd-an laqi-tu
akh-a-hu
(i: 32.20)
zayd-acc.
'As for Zayd, I found his brother'.
'Zayd' can equally appear in nominative or accusative form here. Were
Sibawaih a simple descriptivist he would presumably have been content to
note that a topic can appear in nominative or accusative form, in free variation.
Such a statement appears to account for most of the topicalization structures
which Sibawaih discusses in this and the following chapters. Such an approach
is quite foreign to his methodology, however. For Sibawaih no variation is
simply' free' because every variant which he catalogues implies its own concep-
tual interpretation. In the present example, both nominative and accusative
cases are justified by a series of analogies with other, simpler structures and
paraphrases, with examples which allow a regularization of apparently anomal-
ous structural elements.
23
One thinks,
for
example,
of the
meaning
of the
designations
for
linguistic varieties/entities,
kalam,
lugha, qawl (Versteegh, 1993: 91,
99
ff.)
in the
development
of
the notion
of an
'Arabiyya.

64 JONATHAN OWENS
At
the
beginning
of ch. 31 (i, 31.17 ff.)
Sibawaih explains
the
following
pairs
of
examples (where
(6a)
corresponds
to (5a), (6b) to
(5b)).
(6a) zayd-un darab-tu-hu
Zayd-nom. hit-I-him
'As
for
Zayd,
I hit him'.
(6b) zayd-an darab-tu-hu
zayd-acc.
' Zayd,
I hit him'
(6a) contains
a
nominative noun
in the
function
of
topic
(mubtada'),
with
the
verb structurally
set
against
the
topic
as its
comment
24
{mabniyy
'aid
'l-mubtada',
see
Levin, 1985). Here
the
nominative
in
Zayd
is
explained
by the
general property that topics
are
nominatively marked.
In (6b) the
problem
is
to explain
the
accusative
in
Zayd, which Sibawaih does
by
assuming
an
implicit
verb
(idmar
al-fi'l,
i, 32.1) of the
same value
as the
main verb which governs
the accusative
in
Zayd
(=
(darabtu)
zaydan
darabtuhu).
Sibawaih then proceeds
to
more complex examples where
the
co-referential
pronoun
is
detached from
the
verb, first where
the
direct object
is
marked
by
a preposition
(zayd-unjan
marartu bihi
'As for
Zayd,
I
passed him'), then
to
the
set in (5). (5a) is
explained analogously
to (6a), as a
topic
(nominative)
+
comment structure. That
in (5b) is
more problematic because
in contrast
to (6b)
there does
not
appear
to be a
direct semantic link between
the topicalized noun
and the
main verb. There
is no
obvious sense
in
which
the action
of '
hitting'
can be
directly related
to the
topic, Zayd.
To
explain
these structures Sibawaih invokes
a new
principle, namely that
' if the
action
falls
on the
object containing
a
co-eferential pronoun
it is as if
the action falls
on
the
object itself (32.17). This semantic equivalence carries over
to a
morpho-syntactic one, where
the
topicalized noun assumes
the
same case form
as
the
item
it is
linked with.
In (5b), for
instance,
the
co-referential pronoun
-hu links
akhd-
to
zaydan. Since
akhd-
is
accusative,
so, too, can
zayd be.
25
This principle
is
invoked
a
number
of
times
to
explain ever more complex
structures
in the
succeeding pages (particularly
i, 42).
Examples such
as
these from Sibawaih
are
abundant.
I
will come back
to
them presently,
for the
moment noting only that what
is
important here
is
that
an example cited
by
Sibawaih always
has to be
integrated into
his
linguistic
thinking. Sibawaih cites
(5a)
with minimal comment because
for
Sibawaih
its
structure
is
clear. When
he
moves
to (5b) he is
confronted
by a new
structural
state of affairs which requires new principles, new explanations. What happens,
however, when
he
meets structures which
he
finds clearly wrong?
In
fact this
happens relatively infrequently.
26
It is
true,
as
Carter (1973) points
out,
that
Sibawaih does have
an
evaluative vocabulary which allows
him to
rank
the
acceptability
of
one structure against another. When
he
uses
it,
however,
it is
usually
to
recommend
a
over
b,
without rejecting
b
altogether.
In the
present
24
The
term 'comment'
is a
more appropriate translation
of the
later 'khabar'. Levin (1985:
302) translates Sibawaih's term
as 'the
part which makes
the
sentence complete'.
For
brevity's
sake
I use the
shorter term 'comment'.
25
Sibawaih may not be entirely convinced of this explanation
himself,
for he adds two further
(somewhat strained) examples where a noun is grammatically affected by a verb, without bearing
a direct semantic relation to the action represented in it. Thus one might say, ' I honoured him
as you honoured his brother' (i, 32.17) where the honouring is equal in both actions, though in
the first it is ' him' who is honoured, whereas in the second it is not' him' but his brother.
26
Admittedly 'infrequently' is an impressionistic evaluation. It can in principle be fairly
strictly measured, however, taking all the examples completely disallowed by Sibawaih divided by
all his examples discussed. The percentage would be quite low.

CASE
AND
PROTO-ARABIC,
PART
I 65
example, for instance, he says that the nominative (6a) is 'better' (i, 32.22).
Clearly, however, he makes this judgement on the basis of the grammatical
merits of each structure, i.e. in terms of the rules by which he evaluates them
in the first place.
None the less, it is important for present purposes to know if Sibawaih
sets limits to the acceptable. One positive answer to this question can be
illustrated by his discussion of pausal forms, which will be referred to further
in 3.2.3. Sibawaih (n, 309, ch. 495) notes that it may happen in -CC#-final
nouns (particularly with sonorants as the final C apparently) that in pausal
position the genitive or nominative case markers are not deleted but rather,
by a process of what may be termed 'case epenthesis', form a final CVC
syllable. Thus ' some Arabs', instead of
saying
bakr(u) or bakr(i), have
bakurft
'Bakr-nom.',
bakirft
'Bakr-gen.'. He adds, however, that this is possible only
so long as the resulting structure meets acceptable word structure constraints.
Udl-'
equal', may undergo genitive case epenthesis,
'idilft,
but in the nominative
this is impossible, *'idul, because 'they [Bedouins] have no words of structure
fi'uV (309.20). Here the non-occurrence of particular forms is apparently
confirmed not in terms of what Sibawaih observed or tested, but rather in
terms of the violation of his own general rule.
What should by now be clear is that there is no pure
'
data' to be found
in Sibawaih. Everything he observes and writes about is filtered through his
own grammatical thinking. One salutory effect of this is that it was this very
systematization of linguistic facts which helped him to produce a work of
extraordinary detail. In examples (5-6) discussed above, Sibawaih starts with
basic
N-V-obj.
structures, moves to N V prep.-obj. structures, and finally to
N V Obj.-possessor-pro, structures, with each step tackling a slightly more
complex case. His description is partly carried along and expanded by the very
logic of his grammatical thinking.
This is not to say, however, that Sibawaih had no regard for the linguistic
facts provided to him from his various sources. I think examples like (5) and
(6) can be understood in the following way. Sibawaih was presented with raw
data, and this was that the topic noun varies freely between nominative and
accusative case. He accepted both forms, but on terms of his own theoretical
making. It was Sibawaih's achievement to integrate these ' facts' into a more
or less coherent whole (the definitive interpretation of Sibawaih remains to be
written), in this case through such concepts as 'topic', mabniyy 'ala
al-mubtada',
co-referentiality, and so on.
At the same time one has to assume that there are many elements of
'Arabic' which were outside the scope of Sibawaih's cognizance. Some of
these, of course, are due to the mundane fact that Sibawaih was mortal, the
amount of observations he could make finite. Other elements, however, would
have escaped Sibawaih's notice because they could not be fitted into his
linguistic thinking (Baalbaki 1990: 22). This is a necessary corollary of the
system-driven nature of
his
methodology. As seen above, Sibawaih on principle
rules out forms like
*'idul.
It is therefore unlikely, if such forms did exist, that
they would have been observed by
him.
Caseless forms of Arabic could similarly
have been outside his purview.
3.2.2. Stable
cases,
free variation
While it is improper to speak of free variation of case within a Sibawaihian
analytic framework, the fact remains that this effectively is what he documents
in many instances. Looking beyond the topic construction, there are many
examples of what amount to free variation in case form discussed in the Kitab.

66 JONATHAN OWENS
In fact, the discussion above around (5-6) is typical of Sibawaih's exposition,
intimately concerned to define the proper case forms. Full proof of this is a
task beyond the confines of the present exposition. What can be offered here
is a brief overview of the type of case variation Sibawaih dealt with, based on
a review of the first 100 pages of the
Kitdb,
just under a quarter of book i. In
these 100 pages, roughly the following topics are dealt with (initial pages of
topics are given): general concepts (p. 1), transitivity (p. 10), negative in laysa
and ma (p. 18), left noun dislocation (tandzu', p. 28), extraposition (p. 31)
arranged according to type of predicate and predicational type, extraposition
in inalienable-like constructions (p.
64),
governance of participles, verbal nouns
and adjectives (p. 70), extension of function {ittisd') and
(ishtighal,
90; Owens,
1990:
251 ff.). In the following I will excerpt a representative example, summar-
izing Sibawaih's comment on each example.
(7a) ma 'abdu
llahi akh-d-ka
~ akh-u-ka
'Abdullahi is not your brother'. (21.20)
Usual=nominative, accusative = dialectal usage, md
al-hijdzi
(7b) darab-tu wa
daraba-rii zayd-un
~ zayd-an
hit-me and hit-me Zayd-nom. ~
ace.
'I hit (Zayd) and Zayd hit me'. (28.18)
Nominative is better because of proximity to second verb, which logically
requires a nominative agent. Accusative also allowable.
(7c) a
zayd-an
~zayd-un anta
ddrib-u-hu
Q Zayd-a you hit-nom. him
'As for Zayd, are you going to~have you hit him'? (45.30)
Accusative correlates with verb-like imperfect meaning of active participle,
nominative with nominal-like perfective meaning. This example follows the
much more detailed and complicated instances of extraposition with verbal
predicates, as in (5,6).
(7d)
'abd-a
~
'abd-u ulldhifa-drib-hu
Abdallah-acc. ~
nom.
so-hit-him
'Abdallah, so hit him'. (58.12)
Accusative is preferred, since marked modal sentences (imperatives, condi-
tionals, questions) imply a verbal predicate (which governs the accusative).
None the less, contexts can be found (as here) allowing nominative as well.
(7e) duriba 'abdu llahi
zahr-u-hu~
zahr-a-hu
hit Abdallah back-nom.~acc.-his
'Abdullah was hit on his back'. (68.9),
Nominative, as badal or
tawkid,
accusative as nominal complement brought
into direct governance of the verb, with implied preposition
{'aid
zahrihi)
like
dakhaltu 'l-bayt-a~fi
'l-bayt-i.
(7f) 'ajib-tu min darbi zayd-in wa
'amr-in/an
(81.1)
surprised-I from hitting Zayd-gen. and Amr-gen. ~acc.
'I was surprised by the beating of Zayd and Amr'. (81.1)
Genitive in Amr by agreement with Zayd, accusative by virtue of an understood
verb
(daraba)
'amran.

CASE
AND
PROTO-ARABIC,
PART
I 67
(7g) duriba bihi darb-un da'if-un ~
darb-an
da'
Tf-an
hit by it hitting-nom. weak-nom. ~ hitting-acc.
'A weak blow was hit with it~It was hit a weak blow with'. (97.2)
In passives without an expressed ' underlying' direct object, the choice is free
as to which of a range of further complements can be promoted to agent. In
this case, either the verbal noun is promoted (nom.), or no complement is
promoted (ace).
As the brief expositions make clear, there is no single explanation for the
observed variation. It may be due to dialect variation (7a), though more
frequently (7b, d-g) it is embedded within the logic of Sibawaih's own gramma-
tical formulations. In some instances Sibawaih ranks the alternatives by some
measure of relative appropriateness, while in others both variants are of equal
value. In one case it may be objected that the example is not an example of
free variation at all, since in (7c) the use of accusative or nominative in zayd
presumably correlates with a difference in meaning. While the point of this
section does not stand or fall on such examples, it is relevant here to draw a
distinction between what Sibawaih said and what one may read between the
lines of his pronouncements. In particular, given his predilection for (and task
of) systematizing the language, one may at certain points (though certainly
not in general) question whether what is systematized is really a part of the
Arabic spoken in the eighth century, as opposed to the language as idealized
by Sibawaih and other grammarians, who probably were more ready than the
population at large to concretize subtle distinctions among competing variants
whose origin was not necessarily of a purely linguistic (as opposed to stylistic,
sociolinguistic or dialectal) nature. In any case, taken as a whole, ' free vari-
ation ' is an adequate characterization of the
product
of Sibawaih's observations
in (5-7). This does not mean, of
course,
that the variation would be conceived
of in such terms by Sibawaih
himself.
To the contrary, as the explanatory
notes are intended to make clear, each variant for Sibawaih is associated with
its own structural logic.
It does not appear that the variation in the case system points to an
impending breakdown. Sibawaih is too specific about which forms are uniquely
correct in many contexts and too specific about the implications of choosing
one variant or another to lend such speculation any weight. Taken as a whole,
however, the variation does point to a system with an inner dynamic and
flexibility, that is, a variation which grew out of various historical develop-
ments. It could even have evolved out of a non-case system (see 3.2.3).
It may be noted here, that there is a clear structural tendency in the
variation, namely that most case variation involves the accusative as one of
the two alternatives. Expanding on this observation, it is fair to say that the
accusative is the unmarked term relative to frequency of functional occurrence.
The only positions which are unequivocally not accusative are objects of
prepositions and possessors (=
genitive),
comments, topics when the comment
is not a verb and agents in VS (verbal) sentences
(
=
nominative).
Otherwise
(the various objects, tamyiz, hal, even subjects after the
inna
class of comple-
mentizers) sentence constituents take accusative, or vary freely in accusative
with another case form (as in examples above).
3.2.3.
Pausal and context forms
Probably the greatest degree of variation (of any type) associated with a single
functional position is that relating to pausal forms. Sibawaih devotes most of
the 28 pages between n, 302-30 to its explicit description, and there are various
references to it in other parts of his work. It is clear that for Sibawaih pausal

68 JONATHAN OWENS
context
is not
simply
a
nominal stem minus
the
indefinite
and
case suffixes,
but rather
a
position engendering phonological changes
of
various sorts.
27
The
topic
is
potentially important, because
it has
been assumed
by
many scholars
(Brockelmann,
1982
[1908]:
462;
Birkeland,
1952: 9;
Fleisch,
1974: 23;
Blau,
1981:
3;
Diem, 1991:
303)
that
the
modern caseless dialects derive from
the
Classical Arabic pausal forms. Concentrating here
on
those chapters which
explicate pausal (waqf) forms,
28
it
emerges that much
of
what
he
describes
for
pausal phenomena
is not
immediately relatable
to the
modern dialects.
The
following typology, without answering definitively
the
question
of the
extent
to which there
is a
direct link between pausal forms
in
Sibawaih
and the
modern dialects,
at
least defines where
the
problems lie.
The
typology consists
of two main parameters.
One
relates
to
Sibawaih's description
of a
particular
phenomenon
as
being
a
property
of
context
or
pausal position
(or
both),
the
other
to the
range
of
distribution
of the
phenomenon, both
in
Sibawaih
and
in
the
modern dialects.
I
will illustrate these points here
by
means
of an
informal scale,
at
whose initial point
no
obvious connection between
the
dialects
and
Sibawaih's description exists
and at
whose
end
point
a
fairly
plausible relation
may be
postulated.
At
one
extreme there
are
many parts
of
Sibawaih's description which have
no relevance
to the
present question because they have
no
obvious reflexes
in
the modern dialects. Perhaps
the
clearest example
of
this sort
is the
fate
of
the
case vowels themselves.
In
pausal position they
do not
simply disappear.
Rather,
the
pausal position
at
which they occur
may
take
on
four different
values
(ch.
494).
It is
unnecessary
to go
into details here,
the
important point
being that since
the
case vowels
do not
occur
in the
modern dialects
it is
impossible
to
draw connections between Sibawaih's description
and
their
reflexes
in the
modern dialects.
Moving
up the
scale,
a
second type pertains
to
word-final
a,
apparently
when written with
the
alif
maqsura.
Sibawaih notes
(n,
314.8) that although
most Arabs pronounce
it in
pause
as -a, the
Qays change
it to y, as in
hubla-*hublay
' pregnant'. Among modern dialects, Blanc
(1964:
50) notes that
a final feminine
-a
irregularly undergoes imala
in
Jewish Baghdadi Arabic,
heble 'pregnant' being among
the
lexical items where this happens.
The
-a-+-ay/-e
change
is
frequent neither
for
Sibawaih,
nor in the
modern dialects.
While there
may be a
connection between
the
Qays pausal form
and the
Baghdadi example, definitive proof is highly unlikely.
A third instance
has
already been mentioned above, where Sibawaih notes
that 'some Arabs' employ case epenthesis
in
pausal position,
e.g.
bakru#->bakur§
'
Bakr'. This case
is
more interesting than
the
previous
one in
two directions.
On the one
hand,
for the
classical language, Sibawaih does
not
appear
to
place such severe restrictions
on the
Arabs
who use the
form.
On
the other,
for the
modern dialects,
as
will
be
seen
in 4.2 (11),
under certain
interpretations
it can be
related
to a
fairly widespread contemporary phenom-
enon. Very briefly, many modern dialects have
a
rule inserting
an
epenthetic
vowel before
a
sonorant (e.g.
-r, -I)
consonant. Interestingly,
all of
Sibawaih's
examples, admittedly only seven
in
all, have
a
sonorant, -r,
-I
or
-m
as the
final
27
As
with many
of his
concepts, Sibawaih does
not
define what
he
means
by
pause
and
context. The fact that he includes topics among the ' pausal' chapters (see n. 25) which are not
obviously descriptions of pausal phenomenon, e.g. the Assad and Tamimi realization of -shi for -
ki' 2fsg object suffix' (ch. 504), means that a closer look at these concepts in Sibawaih would be
appropriate.
28
There are 14 chapters in the page range cited above (chs. 490-504, 507) which deal
exclusively or extensively with pausal forms. There is probably nowhere to be found a more
detailed description of pausal phenomena in Classical Arabic than in Sibawaih.

CASE
AND
PROTO-ARABIC,
PART
I 69
consonant. Uncharacteristically, if sonority is indeed a conditioning factor,
Sibawaih does not state a phonological environment in respect of the final
consonant, though he does explicitly note (n, 310.5) that the process does not
occur when a semi-vowel occurs as C
2
(e.g.
zayd,
'awn). Even if the sonority
condition plays a role in Sibawaih, a difference exists with the modern dialects
where, as will be seen, the rule applies anywhere in a word, not only finally as
in Sibawaih. Certainly the present example potentially represents a more gen-
eral correspondence between the modern dialects and Sibawaih's treatment of
pausal phenomena than the case discussed in the preceding paragraph. The
correspondence is not complete, however, so there will always be a risk in
drawing definitive conclusions.
In a fourth set of cases correspondences can be drawn between modern
dialects and a variety of pausal alternatives, or even with context forms.
Sibawaih, for example, (ch. 500) says that the pausal form of nominative and
genitive nominals of the form faliy may be
ramijf,
rdmT§
or rdmft' has thrown \
29
The modern dialects have rami here, or perhaps
ramT—the
choice between the
short -i or long
-T
being one of phonological theory—
30
but not the pausal
ram.
On the other hand, the definite context forms also have
-Ty,
al-rdmi,
so
they could also have been a ' source' for the modern dialectal forms. In this
case correspondences between some of Sibawaih's morphological alternatives
and modern dialects are close to perfect, but still too ambiguous to decide on
a definite correspondence. In this category can be cited instances where it is
Sibawaih's context form which provides the clearest link to the modern dialects.
Such a case is found among the Tamim (n, 314.14), who in the f. near
demonstrative have
hddhih
in pause,
31
but
hddhTin
context. In modern dialects
hddhi(y) 'this f.' is very common.
Sibawaih's
fine-grained
descriptions certainly deserve more detailed discus-
sion than there is space for here. I think that the examples are representative
of a general predicament, however; namely that in only rare cases can an
unequivocal connection be drawn between Sibawaih's description of pausal
forms generally, i.e. not only those relating to the treatment of the final case
vowel, and comparable forms in modern dialects. Even when such connections
exist, it is rarely so that they would explain anything but a part of modern
dialectal forms. Similarly, one would like to know why often only certain
Arabs (' some of them', the Qays, Tamim, etc.) have forms analogous to the
modern dialectal ones. Until these problems have been given more serious
attention, I think it over selective to argue that the modern dialects arose from
pausal forms, when the main piece of evidence supporting this position in
Sibawaih would appear to be that one only of four possible ways of pronoun-
cing case vowels in pausal position is by deleting the vowel altogether (jazm,
ch. 394).
Besides interpretive problems of the above kind, there is a more unequivocal
argument against assuming that Classical Arabic pausal forms were the fore-
runners of the dialectal caseless forms. According to Sibawaih pausal forms
should occur only before
pause.
He mentions at various points in his discussion
that the peculiarities which he describes for them do not apply to forms in
connected speech
{wasl,
e.g. n, 302.8, 306.5, 313.18). Dealing as we are with
29
Carter (1990) suggests that
the
last form
is
rare.
30
The V-final suffixes
in the
dialects generally have
two
forms, long before
a
further suffix,
otherwise short. Which,
if
either, variant
is
assumed
to be
basic
is as
much
a
matter
of the
linguistic theory
one
assumes
as the
historical linguistic perspective
one
adopts.
See
also Part
II,
section 4.2.
31
Though later
(n,
322.15), Sibawaih reports from
a
reliable source that some Arabs have
hddhih
as a
context form.

70 JONATHAN OWENS
written texts, there is no way to measure where precisely pauses were placed
in the Classical language, at least not in non-poetic style, which certainly must
be assumed to be the purported model of the modern dialect ancestor. To
arrive at an idea of how frequent pauses actually are in spoken language I
used a corpus of texts which I have collected over the last seven years from
the spoken Arabic of NE Nigeria. The texts are transcribed with a basic
phonetic alphabet and pauses are explicitly marked wherever they occur. Since
this material is computerized it is an easy matter to calculate how many pauses
there are in each text. Table
1
gives basic information from five texts about
the number of pauses relative to the total number of words.32
Text
no.
1
2
3
4
5
Total
Number
of
words
2460
6287
5329
7152
3455
TABLE
1.
Number of pauses
598
956
1264
1325
689
word/pause
ratio
4.11
6.57
4.21
5.39
5.01
5.09
The ratio of 5:1 means that, on average, only one word in five occurs before
a pause. Four words in five do not. Assuming this ratio to be generally
representative of spoken Arabic—lacking further statistics, the present ones
must do—it is clear that most words do not occur in pausal contexts, and by
extrapolation, that in Classical Arabic the non-pausal forms in normal speech
would have considerably outnumbered the pausal. To argue that the modern
dialects grew out of the pausal forms of the Classical language is to say that
forms which are a relatively small minority became the standard for the further
development of the language. This, I think, is a priori unlikely. A popular
refinement on the pausal origin hypothesis is problematic and speculative. This
would have it (e.g. Blau, 1988: 9; Corriente, 1976: 84) that under the influence
of foreign-language learners, even in the Classical language the pausal forms
began to be used for the non-pausal. Strictly speaking the idea is unverifiable;
Sibawaih gives no intimation of such a process, and there are no modern
analogies, in Arabic at least, by which to be guided. In trying to reconstruct
the presumed process, lack of motivation is a stumbling block. Judging by the
complex morphology of modern Arabic dialects, it appears that non-Arabs
learned complex Arabic morphology and phonology and made it into their
native language, apparently with little problem. Why should they have had
such a problem with the cases? Moreover, what was really dropped was short
final
vowels,
among which were found the cases. Even in the unlikely situation
that the cases were too difficult for non-Arabs to learn, conceptual difficulty
can certainly not be invoked to explain the disappearance of, say the -a from
cayna ' where'.
The hypothesis which I am developing avoids these mental gymnastics,
since it is (roughly, see Part II, section 5 for a more refined discussion) claimed
that the dialects descend from a variety which never did have case endings.
Before moving on to the modern dialects, I would like to mention one
well-known characterization of Sibawaih, which he treated inter alia among
the pausal forms. This is the opposition between the high vowels i, u vs. the
low vowel -a, in particular the relative stability of the latter against the former.
32 All five texts
are
informal conversations recorded between Nigerian Arabs.

CASE AND PROTO-ARABIC, PART I 71
Thus the high vowels are deleted in open syllables in CaCi/uC-V forms, both
nouns and verbs, e.g.
kabid-un—>kabd-un
' liver',' adud-un^' addun ' upper arm'
(n, 317.17, 320.6), ''alima-*''alma 'he knew', 'usira->'usra 'it was squeezed'
(n, 277.22),
vs.
jamal-un (*jamlun). Similarly, as is well known, when indefinite,
whereas the high vowel case markers, -u (nom.) and -/ (gen.) are deleted in
pause, the low vowel -an (ace.) is lengthened to -a. It is precisely in this lack
of symmetry that one might search for the origins of the Arabic case system
(proceeding on the assumption that case in Semitic, where it exists, is innovat-
ive).
This pausal alteration may represent an older state of affairs where an
-a(a) suffix (as seen above, representing the unmarked case in Arabic) was
opposed to a bare nominal stem (0). The nominative and genitive vowels may
then have developed out of epenthetic vowels which were inserted in particular
contexts.33 One can cite the Ga'az opposition -a-0 (genitive-0), or even the
Berber construct-independent (unmarked, u-0) state contrasts for analogous
morpho-syntactic dualities in related languages.
Having developed the thesis that a caseless variety of Arabic is as old or
older than one possessing case, on the basis of the comparative and philological
record, it will be the main task of the second part of the paper to bring evidence
to bear on the question from the modern dialects. This will lead to the
development of a general model defining the genesis of case in Arabic.
REFERENCES
Al-'Arabiyya, Vol. 20, 1987.
Antilla, Raimo. 1972. An introduction to historical and comparative linguistics. New York:
MacMillan.
Baalbaki, Ramzi. 1990. ' Trdb and
Bind',
from linguistic reality to grammatical theory', in Kees
Versteegh and Michael Carter (1990),
17-33.
Barth, J. 1898. 'Die Casusreste im Hebraischen', ZDMG, 53: 593-9.
Behnstedt, Peter. 1985. Die nordjeminitischen Dialekte, Teil r. Atlas. Wiesbaden: Reichert.
Bendor-Samuel, John. 1990. The Niger-Congo languages. Lanham: University Press of America.
Birkeland, Harris. 1952. 'Growth and structure of the Egyptian Arabic dialect', Avhandlinger
utgitt av det Norske Videnskaps-Akademi, 1: 1-57.
Blanc, Haim. 1964. Communal dialects in Baghdad. Cambridge Mass.: Center for Middle
Eastern Studies.
Blau, Joshua. 1969. ' Some problems of the formation of the Old Semitic languages in the light
of Arabic dialects', Proceedings of the International Conference on Semitic Studies.
Jerusalem: Israel Academy of Sciences and Humanities, 38-44.
Blau, Joshua. 1976. A grammar of Biblical Hebrew. Wiesbaden: Harrassowitz.
Blau, Joshua. 1981 (2nded.). The emergence and linguistic background of Judaeo-Arabic. Jerusalem:
Ben Zvi Institute.
Blau, Joshua. 1988. Studies in Middle Arabic. Jerusalem: Magnes Press.
Braukamper, Ulrich. 1993. 'Notes on the origin of Baggara Arab culture with special reference
to the Shuwa', Sprache und Geschichte in Afrika, 14: 13-46.
Brockelmann, Carl. 1982 [1908]. Grundriss der vergleichenden Grammatik der semitischen Sprachen.
Hildesheim: Georg Olms.
33 This suggestion is speculative, and it should be emphasized is not crucial to the argument.
Its overall plausibility may be heightened by the following (again speculative) considerations.
Mauro Tosco (p.c.) points out that it is relatively unlikely for non-morphemic material to be
reinterpreted with a morphemic value, as the present suggestion entails. To reply to this, there are
situations where it is very common for non-morphemic material in one language to acquire
morphemic value in another, namely via language contact. A standard example is Arabic kitab,
where non-morphemic ki- becomes in the Swahili loanword ki-tabu identified with the Swahili ki-
noun class marker, hence pi. vi-tabu. The process is quite regular and is not restricted to this one
class of nouns (see e.g. Krumm, 1940: 52 ff.). It cannot be ruled out (though proof is equally
elusive) that a similar process did not occur, introducing case into Arabic. The oldest mention of
the 'Arabs' is from an Akkadian text of 853
B.C.
Actual contact between speakers of Akkadian
and speakers of proto-Arabic very likely occurred. It could therefore also be that Akkadian
speakers with a functional case system shifting into Arabic (or Arabic-Akkadian bilinguals)
reinterpreted the original Arabic epenthetic vowels as case vowels. This process may have been
helped if Arabic already had a suffix -a (perhaps adverbial, as in Hebrew). (See section 5 on
co-existence of case and caseless varieties, and 4.2 (13) for parallels between the distribution of
case suffixes and epenthetic vowels.)

72 JONATHAN OWENS
Carbou, Henri. 1913. Methode pratique pour I'etude de Varabe parle au Oudayy et a Vest du
Tchad.
Paris: Geuthner.
Carter, Michael. 1973. 'An Arabic grammarian of the eighth century
A.D.:
a contribution to the
history of linguistics', JAOS, 93: 146-57.
Carter, Michael. 1990. 'Qadi, qddi, gad: which is the odd man out?', in Versteegh and Carter
(1990), 73-90.
Castellino, G. 1978. 'The case system of Cushitic in relation to Semitic', in Pelio Fronzaroli (ed.),
Atti del secondo Congresso Internazionale di Linguistica Camito-Semitica. Florence:
Instituto di Linguistica e di Lingue Orientali.
Callender, John. 1975. 'Afroasiatic cases and the formation of Ancient Egyptian constructions
with possessive suffixes', Afroasiatic Linguistics, 2: 95-112.
Chejne, A. 1969. The Arabic
language.
Minneapolis: University of Minnesota Press.
Corriente, Frederico. 1971. 'On the functional yield of some synthetic devices in Arabic and
Semitic morphology', Jewish Quarterly Review, 62: 20-50.
Corriente, Frederico. 1973. 'Again on the functional yield in some synthetic devices in Arabic and
Semitic morphology', Jewish Quarterly Review, 63: 151-63.
Corriente, Frederico. 1975. 'Marginalia on Arabic diglossia and evidence thereof in the Kitab
al-Agani', Journal of Semitic Studies, 20:
38-61.
Corriente, Frederico. 1976. 'From Old Arabic to Classical Arabic through the pre-Islamic koine:
some notes on the native gramarians' sources, attitudes and goals', Journal of Semitic
Studies, 21: 62-98.
Corriente, Frederico. 1977. A grammatical sketch of the Spanish Arabic dialect bundle. Madrid:
Instituto Hispano-Arabe de Cultura.
Cowan, William. 1960. 'A reconstruction of Proto-Colloquial Arabic' Ph.D. thesis, Cornell
University.
Croft, William. 1990. Typology and
universals.
Cambridge: Cambridge University Press.
Diakonoff,
I. 1988. Afrasian
languages.
Moscow: Izdatel'stvo Nauka.
Diem, Werner. 1973. 'Die nabataischen Inschriften und die Frage der Kasusflexion im
Altarabischen', ZDMG, 123: 227-37.
Diem, Werner. 1984.' Philologisches zu den arabischen Aphrodito-Papyri', Der Islam,
61:
251-75.
Diem, Werner. 1991. ' Vom Altarabischen zum Neuarabischen. Ein Neuer Ansatz', in Alan Kaye
(ed.),
Semitic studies in honor of Wolf
Leslau.
Wiesbaden: Harrassowitz, 297-308.
Ditters, Everhard. 1990. 'Arabic corpus linguistics in past and present', in Versteegh and Carter
(1990),
129-41.
Doss, Madiha. 1995. ' Some remarks on the oral factor in Arabic linguistics', in Harviainen et al.
(ed.),
49-62.
Eksell, Kerstin. 1995. 'Complexity of linguistic change as reflected in Arabic dialects', in
Harviainen et al. (1995), 63-74.
Al-Farra', Abu Zakariya. Ma'aniy al-Qur'dn. Beirut: 'Alam al-Kutub, 1983.
Ferguson, Charles. 1959a. 'Diglossia',
Word,
15: 325-40.
Ferguson, Charles. 1959b. 'The Arabic Koine', Language, 35: 616-30.
Fischer, Wolfdietrich. 1961. 'Die Sprache der arabischen Sprachinsel in Uzbekistan', Der Islam,
36:
232-263.
Fischer, Wolfdietrich. (ed.) 1982a.
Grundriss
der Arabischen
Philologie,
Bd. I: Spachwissenschaft.
Wiesbaden: Reichert.
Fischer, Wolfdietrich. 1982b. 'Fruhe Zeugnisse des Neuarabischen', in Fischer (1982a), 83-94.
Fischer, Wolfdietrich. 1995. 'Zum Verhaltnis der neuarabischen Dialekte zum Klassisch-
Arabischen', in Harviainen et al. (1995), 75-86.
Fischer, Wolfdietrich and Otto Jastrow. 1980. Handbuch der arabischen Dialekte. Wiesbaden:
Harrassowitz.
Fleisch, Henri. 1974 (2nd ed.). Etudes d'arabe dialectal. Beirut: Dar el-Machreq.
Fuck, Johann. 1950. Arabiya. Berlin: Akadamie Verlag.
Garbell, Irene. 1958. 'Remarks on the historical phonology of an East Mediterranean Arabic
dialect',
Word,
14: 303-37.
Greenberg, Joseph. 1978 (2nd ed.). 'Some universals of grammar with particular reference to the
order of meaningful elements', in Joseph Greenberg (ed.), Universals of language,
Cambridge: MIT Press, 73-113.
Gordon, Cyrus. 1965. A Ugaritic textbook. Rome: Pontificum Institutum Biblicum.
Harviainen, Tapani et al. (ed.). 1995. Dialectologia Arabica. Helsinki: Finnish Oriental Society.
Hetzron, Robert. 1976a. 'The Agaw languages', Afroasiatic Linguistics, 3: 31-75.
Hetzron, Robert. 1976b. 'Two principles of genetic reconstruction', Lingua, 38: 89-108.
Hudson, Grover. 1976. 'Highland East Cushitic', in Lionel Bender (ed.), The non-Semitic
languages of Ethiopia, East Lansing: Michigan State University, 232-77.
Ingham, Bruce. 1994a. Najdi Arabic. Amsterdam: John Benjamins.
Ingham, Bruce. 1994b. 'The effect of language contact on the Arabic dialect of Afghanistan' in
Jordi Aguade, Federico Corriente and Marina Marugan (ed.), Adas del Congreso
Internacional
sobre
Interferencias Linguisticas Arabo-Romances y
Paralelos
Extra-Iberos,
Zaragoza: Navarro e Navarro, 105-18.
Ingham, Bruce. 1995. 'Texts in the dialect of the Rwalah of northern Arabia', in Harviainen
et al. (1995), 121-140.
Kaye, Alan. 1976. Chadian and Sudanese Arabic in the light of
comparative
Arabic dialectology.
The Hague: Mouton.
Khan, Geoffrey. 1988. Studies in Semitic Syntax. Oxford: Oxford University Press.

CASE AND PROTO-ARABIC, PART I 73
Kofler, Hans. 1940. 'Reste altarabischer Dialekte', Wiener Zeitschrift fur die Kunde des
Morgenlandes, 47: 61-130.
Krumm, J. 1940. Worter
orientalischen Ursprungs
in Suaheli. Berlin.
Levin, Aryeh. 1985. 'The syntactic technical term al-mabniyy
'alayhV,
Jerusalem Studies in Arabic
and Islam, 6: 299-352.
Levin, Aryeh. 1989. 'What is meant by 'Akalum l-Baraghithu', Jerusalem Studies in Arabic and
Islam,
12: 40-65.
Macuch, Rudolph. 1982. Grammatik des samaritanischen Aramdisch. Berlin: Mouton de Gruyter.
Mahdi, Muhsin (ed). 1984. One Thousand and One Nights. Leiden: Brill.
Moscati, Sabatino. 1958. 'On Semitic case endings', Journal of Near Eastern Studies, 17: 142-4.
Moscati, Sabatino, Anton Spitaler, Edward Ullendorff and Wolfram Von Soden. 1980 (2nd ed.).
An introduction to the comparative grammar of the Semitic languages. Wiesbaden:
Harrassowitz.
Mflller, Walter. 1982. 'Das Altarabische der Inschriften aus vorislamischer Zeit', in Fischer
(1982a), 30-36.
Noldeke, Theodor. 1910. Neue Beitrdge zur semitischen Sprachwissenschaft. Strassbourg: Triibner.
Owens, Jonathan. 1985. A grammar of Harar Oromo. Hamburg: Buske.
Owens, Jonathan. 1989. 'Zur Pidginisierung und Kreolisierung im Arabischen', Afrika und
Ubersee, 72: 91-107.
Owens, Jonathan. 1990. Early Arabic grammatical theory: heterogeneity and standardization.
Amsterdam: Benjamins.
Owens, Jonathan. 1993a. 'Nigerian Arabic in comparative perspective', Sprache und Geschichte
in Afrika, 14: 85-175.
Owens, Jonathan. 1993b. A grammar of
Nigerian
Arabic. Wiesbaden: Harrassowitz.
Owens, Jonathan. 1995. 'Minority languages and Urban Norms', Linguistics, 33: 305-58.
Petracek, Karel. 1988. Altdgyptisch, Hamitosemitisch. Prague: Univerzita Karlova.
Rabin, Chaim. 1951. Ancient West Arabian. London: Taylor's Foreign Press.
Rabin, Chaim. 1955. 'The beginnings of Classical Arabic', Studia Islamica, 4: 19-37.
Rabin, Chaim. 1969. 'The structure of the Semitic system of case endings', in Proceedings of the
International Conference on Semitic Studies. Jerusalem: Israel Academy of Sciences and
Humanities, 190-204.
Reichmuth, Stefan. 1983. Der arabische Dialekt der Shukriyya im Ostsudan. Hildesheim: Georg
Olms.
Retso, Jan. 1994. "I'rab in the forebears of modern Arabic dialects', in Dominique Caubet and
Martine Vanhove (ed.), Actes
des
premieres journees internationales de dialectologie arabe
de Paris. Paris: INALCO, 333-42.
Retso, Jan. 1995. 'Pronominal state in Colloquial Arabic: a diachronic attempt', in Harviainen
et al. (1995), 183-92.
Rossler, Otto. 1950. ' Verbalbau und Verbalflexion in den semito-hamitischen Sprachen', ZDMG,
100:
461-514.
Saeed, John. 1987. Somali reference grammar. Wheaton: Dunwoody Press.
Sasse, Hans-Jurgen. 1984. 'Case in Cushitic, Semitic and Berber', in James Bynon (ed.), Current
progress in Afroasiatic linguistics. Amsterdam: Benjamins, 111-26.
Sibawaih, Ibn 'Uthman. Al-Kitab, ed. M. Derenbourg. Hildesheim: Georg Olms.
Spitaler, Anton. 1953. Review of Fuck (1950), Bibliotheca Orientalis, 10: 144-8.
Talmon, Rafael. 1990. 'The philosophizing Farra': an interpretation of an obscure saying
attributed to the grammarian Tha'lab', in Versteegh and Carter (1990), 265-80.
Tosco, Mauro. 1993. 'On case marking in the Ethiopian language area', in Atti delta 7" Giornata
di Studi Camito-Semitici e Indoeuropei. Milan.
Versteegh, Kees. 1984. Pidginization and
creolization:
the case of
Arabic.
Amsterdam: Benjamins.
Versteegh, Kees. 1993a. 'Leveling in the Sudan: from Arabic Creoles to Arabic dialect',
International Journal of the Sociology of
Linguistics,
99: 65-80.
Versteegh, Kees. 1993b. Arabic grammar and Qur'dnic exegesis in early Islam. Leiden: Brill.
Versteegh, Kees and Michael Carter (ed.). 1990. Studies in the history of Arabic grammar, II.
Amsterdam: Benjamins.
Voigt, Rainer. 1987. 'The two prefix conjugations in East Cushitic, East Semitic and Chadic',
BSOAS, 50/2, 33O^t5.
Vollers, Karl. 1981 (2nd ed.). Volkssprache und Schriftsprache im altern Arabien. Amsterdam:
APN Oriental Press.
Von Soden, Wolfram. 1969. Grundriss der Akkadischen Grammatik. Rome: Pontificum
Institutum Biblicum.
Weninger, Stefan. 1993. Ga'sz. Munich: Lincom Europa.
Woidich, Manfred and Peter Behnstedt. 1985. Die dgyptisch-arabischen Dialekte. Wiesbaden:
Reichert.
Zaborski, Andrzej. 1995. 'First person pronouns in Arabic in the light of Arabic and Hamito-
Semitic dialectology', in Harviainen et al. (ed.), 289-94.