You are on page 1of 26

“Al-kalām mā kāna muktafiyan bi-nafsihī wa-huwa l-ğumla”: A Study in the history of

sentence-concept and the Sībawaihian legacy in Arabic grammar


Author(s): Rafael Talmon
Source: Zeitschrift der Deutschen Morgenländischen Gesellschaft , 1988, Vol. 138, No. 1
(1988), pp. 74-98
Published by: Harrassowitz Verlag

Stable URL: https://www.jstor.org/stable/43377735

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Harrassowitz Verlag is collaborating with JSTOR to digitize, preserve and extend access to
Zeitschrift der Deutschen Morgenländischen Gesellschaft

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
"Al-kalam mã kana muktafiyan bi^nafmhī
wa-huwa l-ģumla"

A Study in the history of sentence-concept and the


Sîbawaihian legacy in Arabic grammar*

By Rafael Talmon, Haifa

A. The problem - kalām and ģumla : their use by the later


Arab grammarians
The universal prestige Sibawaihi's Kitāb enjoyed in the Islamic world
since the medieval era until the present has rarely been challenged by
scholarly work in that civilization. §aťid al-Andalusī (d. 463/1071), in
describing this book as one of the only three compositions ever written
which containd the sum total of human knowledge (the other being Pto-
lemy's Almagest and Aristotle's Organon)1 , expressed the high esteem
enjoyed by this masterpiece among intellectuals.
The medieval Arabic grammatical literature was, in general, consist-
ent in its admiration for this book, which was, in fact, the earliest
extant work in the field.2 In fact, one can reasonably assume that the
remarkable uniformity in the theoretical model underlying the syntactic
descriptions of the medieval Arab grammarians originated from the
long-standing influence of the Kitab.
Nevertheless, modern scholarship has a long way to go until it can
produce a well-founded, non-impressionistic, evaluation of Sibawaihi's
legacy in the Arabic grammatical tradition. It is worth noting that not a
few studies on the syntactic terminology of medieval writers published

* I wish to express my gratitude to Prof. Gideon Goldenberg for his impor-


tant remarks on a former draft of this article. I also wish to thank Prof. A. Levin
from the Hebrew University of Jerusalem for our fruitful discussions about some
essential points in it.
1 ¡$aťid al-Andalusī: K. Tabaqãt al-umam. Ed. Cairo, p. 39 f.
2 Although the later a&òãr-literature preserved the names of some older
books in grammar, Sibawaihi's Kitãb seems to have always been the earliest
extant (and most influential) work. On alleged old books prior to Sibawaihi, see
GAS IX; Talmon BSOAS p. 234, idem ZAL (1985), p. 138f.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
A Study in the history of sentence-concept 75

during the last 15 years or so have demonstrated how specific points in


Sibawaihi's terminology and syntactic concepts did not survive in the
mainstream of later grammatical treatises.3
The purpose of this study is to trace the growth of terminology rele-
vant to the sentential framework in early Arabic grammatical works,
and to shed new light on their impact on later generations of indigenous
Arab grammarians. We shall focus on the history of the two main terms
kalām and ģumla with special emphasis on the early period of the
second and third Islamic centuries.
Gideon Goldenberg's study of the two terms, which constitutes
Section III of his article Subject and Predicate in Arabic Grammatical Tra-
dition 4 is convincing in its conclusions about the meanings and denota-
tions of kalām and ģumla in grammatical writings from Mubarraďs time
on.5 Although Goldenberg states clearly that "[t]he history of kalām
and ģumla as syntactical terms will not be studied ..." (sc. studied in
connection with the definition of the sentence as primarily consisting of
subject and predicate), his careful analysis of views concerning these
two terms developed by a large collection of later grammarians6 pre-
sented (in chronological order7) makes his general conclusions relevant
also from the historical perspective. Goldenberg concludes that
kalām and ģumla as syntactical terms overlap in their implications and
are comparable when denoting 'sentence'. In this sense they are some-
times interchangeable, and often confused (a point which will be dis-
cussed in what follows), but are not synonymous. He points out that
ģumla "ensemble, complex" is employed as a technical term meaning a
nexal construction, a structured sentence or clause of any form or type,
being opposed to mufrad "a single word".8 Kalām , on the other hand, is
closer to the notion of a sentence as complete in itself. Therefore, the
term kalām is never employed with reference to 'clause', where the term

3 Mainly Baalbaki 1982, Carter 1973, Levin 1979, 1981.


4 Cf. this volume p. 39-73.
5 For the details about source-material from al-Mubarrad on, see note 7.
The earliest grammarians are discussed in Goldenberg's study very brief-
ly indeed. See p. 54, where mention is made of Sibawaihi and Māzinī's views
(Māzinī died in 148/862).
7 This includes al-Mubarrad (d. 285/898), al-Zaģģāģ (d. 31 1/923), Ibn al-Sar-
rāģ (d. 316/928), al-Zaģgāģī (d. 337/949), Ibn Öinni (d. 392/1002), Ibn Fāris
(d. 395/1005) al-0urģānī (d. 471/1078), Ibn al-Anbārī (d. 577/1181) al-Zamah-
šarī (d. 538/1144), Ibn al-ljāģib (d. 646/1249) and his commentator al-Ra<Jī
al-Astarābādī (d. 686/1287) and Ibn Hišām (d. 761/1360).
8 Goldenberg, p. 53.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
76 Rafael Talmon

ģumla occurs. In this contex


kalām for ģumla is both un
According to these fundam
tradition as a whole maintai
ship between the two synta
aforementioned restrictions
could readily agree with Gol
designating a kernel senten
construction) forming a com
alone and making sense. We
of these terms could arise f
At this point it is importan
tion of the overlap of ģumla a
ing the historical context in
same denotation. I have noticed several indications that the two terms
have probably undergone a historical process at the end of which they
finally came to partially overlap. If it were possible to properly illumi-
nate this historical process, we would be able to advance our under-
standing of the difficulties encountered by medieval grammarians in
identifying exact semantic boundaries for each of the terms, and hence,
we would advance our study of the sentence concept in Arabic gram-
mar.

One fact which I consider as an evidence in favour of such a hi


process has already been studied by Goldenberg. He points
for (some?) early grammarians, not every clause was termed
al-Sarrāģ considered a verbal predicate whose pronoun resu
subject directly as fi% not ģumla , hence, yaqümu , the pre
(1) Zaidun yaqümu (HUWA)
is fi% contrary to other structures in which the pronoun re
subject indirectly, as in
(2) Zaidun <ļarabtu-HŪ
and
(3) Bakrun qãma yabü-HU
In these structures the predicates (ļarabtuhū and qāma yabūhu are
termed ģumla (the members in capital letters are the resumptive pro-
nouns). Al-Zaģģāģī also considered sentence (3) as fi€l wa-mã ttaçala
bihï minfãHl , although he elsewhere regarded verb + agent as a ģumla.11

9 Op. cit., p. 59.


Cf. p. 58 in his article.
Op. cit. p. 56.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
A Study in the history of sentence-concept 77

Another fact which seems to support the assumption of a historical


process is my own observation that the early grammarians used the
term ģumla mainly in reference to 'clause', roughly until the 4th/ 10th
century.12 My impression is that the use of ģumla for the denotation of
'sentence', independent in itself, was adopted mainly by later gramma-
rians.

The terminological status of kalām is also far from being clear. Gram-
marians continued referring to 'sentence' by kalām , but at the same
time used this word in the sense of 'speech', even within the description
of well-defined syntactical constructions.13
The recurring instances of confusion of ģumla with kalām in the medie-
val grammatical literature, which at times are reflected in the reactions,
even criticism, of scholars who were conscious of matters of terminolo-
gical precision14 evoke the following key question: Who were the autho-
rities whose determined distinction between the two terms guided
future generations in preserving some standard separation between the
two terms?

The answer to this question may be too complex for us to offer any
single work as the dominant model followed by later scholars. In fact,
we may say, for instance that each one of the grammarians who were
mentioned by Goldenberg and who maintained this distinction be-
tween kalām and ģumla could join his predecessors in exercising such
influence on his followers.
However, without refuting the relevance of this thesis, one must won-
der what role was played in this issue by the most influential book in
grammar, Sibawaihi's Kitāb. How far was Sibawaihi's teaching, in the
Kitãb itself as well as in its commentaries, responsible for the formula-
tion of the sentence concept and its characteristic terminology in Arabic
grammatical literature?
It is useful to consider the following passage in Ibn al-Anbārī's Asrār
al-'Arabïyatiy obtain some sense of Sibawaihi's influence on later gram-

12 öumla as 'clause' in al-Mubarrad's al-Muqta4ab is found in III: 130; IV: 78,


123, 124, 237. For its occurrences as 'sentence' in al-Mubarrad's writings, see
below. In K. al-Luma 1 by Ibn öinni in: 10, 14; 17, 5; 18, 5. In al-Zaģģāģī's K. al-
Gumal : 42, 48-9, 54, 77, 130, 153, 313, 330. Öumla as 'sentence' is rare. Some-
times the reference remains ambiguous due to uncertainty on my part whether
the writer meant 'sentence' or 'clause'.
13 See Goldenberg' s conclusions about Ibn ôinnî' s notion of Kalam in his
art. cit. p. 58, end of first paragraph.
14 Such criticism is expressed by Ibn Hišām, cf. Goldenberg p. 61.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:3Jan 1976 12:34:56 UTC
All use subject to https://about.jstor.org/terms
78 Rafael Talmon

marians on the issue of senten


marily to Slbawaihi's book itse
tor al-Sīrāfī (d. 368/979):
In the opening chapter of his
'sentence' by using the term k
nition. He introduced it mer
of kalim in the expression mā
It is actually an interpretatio
kalām , in the famous heading
an earlier interpretation, alr
same Ibn al-Anbārī uses the te
'clause'16 and 'sentence'17 an
book again.18
More significant for the illus
lowing excerpt taken from I
author seems to quote the A
1066) about the relationship
ment, they are synonymous
Sida (said): kalām is (an eq
known."]. Another view holds
meaning. The exponents of th
as grammatical terms: wa-qil
wa-huwa l-ģumla wal-qaul mā
l-ģuz* min al-ģumla. "Some
independent (lit. self contai
speech) which is not indep
(ģumla)." In order to support
qala Sibawaihi: ťlam anna qul
bihā mā kāna kalāman lā qau
said' (qultu) occurs in the lan
reproduction19 of kalām , n

15 See Kitãb (Ed. Bûlâq), p. 2 o


16 Cf. 31, 21; 33, 5; 57, 8; 78
17 46, 22 (the construction wi
18 After defining kalām as "sen
(i. e., parts of speech), Ibn al-Anb
next sentence he speaks of aqsã
19 See Troupeau (1976) s.v. I

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
A Study in the history of sentence-concept 79

The passage is quoted with slight modifications from Kitab i 50, 6. 20


There, it is part of chapter 31 in which Sïbawaihi deals with the subject
of i'māl vs. ilga* (i.e., the power of a certain class of words to effect
['amila] the i Võò -endings of other words vs. the absence of this power
in the same class of words under different conditions [hence they
become laģw , their ( amai is annulled]). According to Sïbawaihi' s teach-
ing, which is more or less repeated in I 419,17-420,8, qaul seems to re-
fer to a speech segment occupying the status of a single word following
the verb qãla, whereas kalām is ķikāya "oratio, a citation, reproduction
of a (whole) speech unit already heard as such". So, in
(4) qultu Zaidun muntaliqun (I 50,7)
the portion Zaidun muntaliqun could be an independent utterance (a-lã
tarā annahü yahsunu an taqüla Zaidun muntaliqun "don't you see that it
is appropriate to say 'Z. is going'", ibid.). Contrary to structure (4), one
may conclude, in
(5)* qãla Zaidan munfaliqan
the portion Zaidan munfaliqan is a qaul , which already al-Sīrāfi inter-
preted as Sïbawaihi' s way of saying that this portion takes the status of
qaul in
(6)* qãla qaulan. . .21
It is evident, then, that Sïbawaihi' s terminology was considered by a
later philologist (either Ibn Sïda or some other lexicographer, quoting
[wa-qila] an older source) to be the reliable source that could clarify the
exact meaning of the grammatical term kalām , especially when it stands
in contrast to qaul.
Ibn ôinnï, in the beginning of his Ķaņa*^, discussed exhaustively the
difference between qaul and kalām (and also ģumla). He also quoted the
same passage from the Kitab.22 His own formulation of the difference
seems to be greatly inspired by an interpretation of that passage, which
is similar to the interpretation attributed above to Ibn Sïda' s source.

20 The text in the Kitab says: waJlam anna qultu fi kalām al- Arab innamā
waqa'at ialā an yuhkâ bihā wa-innamã yuhkâ ba1 da l-qaul mā kāmi kalāman lã
qaulan ... In what follows, citations from the Kitab are taken from Deren-
bourg's edition, volume I (= i), unless otherwise marked.
21 Cf. Jahn's notes on chap. 31 (ad loc.): p. 118 note 18: Dahin rechnet Sir.
das Mandar qaul , welches im Acc. steht, z. B. qãla qaulan ķasanan , ferner kalām
[! - R.T.] weil es in der Bedeutung von qaul steht, Z.B. qäla kalāman hasanan

22 The discussion is found in pp. 17-20. The quotation from the Kitab is in p.
18 1. 18 f. It is noteworthy that neither Mehiri nor Rundgren (p. 135 if.) pay
any attention to the fact that Ibn Öinni based his arguments in favour of the dis-
tinction between the two notions on the passage in the Kitab.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
80 Rafael Talmon

While the exact genetic relatio


scure, they certainly both refle
many grammarians, that k
meant 'sentence'.
In the following section, B, we shall carefully examine Sïbawaihi's use
of the term kalām. We shall mainly consider whether this term in the
Kitāb only refers to 'sentence', or whether it also denotes this syntactic
entity.

B. Kalām and sentence concept in Sïbawaihi's Kitāb

1. Modern scholarship on the 'sentence-concept' in the


Kitāb :
Several scholars have already noted the absence of the term ģumla in
the Kitāb.23 The word ģumla meaning 'ensemble' occurs eight times in
the singular and once in the plural. In every case the word refers to the
organization of material or the summary of it.
Mediaeval Muslim scholars have, as far as I know, not remarked on
the above absence. The attribution of the term ģumla to Sïbawaihi must
accordingly be judged as anachronistic, or as a paraphrase.24
The discussion of a possible term for 'sentence' in the Kitāb has con-
centrated, so far, on the study of the meaning of kalām in that book.
Basically, we can distinguish the following two opposing approaches:
the one views kalām as meaning, inter alia, a syntactic unit; while the
other interprets it in the same occurrences as 'speech' in general.
Mosel and Beeston represent the first, Carter and Troupeau the
second. They all seem to agree that kalām has more than one meaning in
the Kitāb.25
Troupeau (1976, s.v. K-L-M) counts kalām as 'language' 666 times,
as word(s) (mot) - 141 times, and as "prose" (opposite of šilr) - 43
times. Mosel (pp. 17-18) too specifies such meanings (Sinne) as "lan-

23 See Carter (1972), 495 n. 23; Troupeau, 22.


24 Cf. e.g., Ibn al-Anbãri, 32, 7: fa-dahaba, Sïbawaihi wa-ģamāla min al -
nahwiyin ilā annahumā yu* addāni min al-ģumal (concerning a predicate which is
either %arf or a prepositional phrase). Further in al-Mubarrad: al-Kãmil (Ed. W.
Wright) I, 289, 14 where the authority of Sïbawaihi is consulted and he is quot-
ed as follows: (li-anna Sïbawaihi . . . qāla) 1-fāHl lã yakünu ģumla. Schaade (EI
s.v. FāHl) erroneously attributes this saying to al-Mubarrad. Actually, it should
be attributed to the smaller 4Ķfaš (see Introduction to al-Kāmil, p. IV, notice).
25 For Jahn's translation, see note 39, below.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
A Study in the history of sentence-concept 8 1

guage", "speech" and "prose", yet she also finds that Sïbawaihi desig-
nates (bezeichnet) sentences by kalām. However, Mosel refrains from
identifying kalām as 'sentence'. She explains (p. 18) that Sïbawaihi re-
fers to the protasis of conditional structures as al-kalām al-auwal (387,
1). Mosel seems to argue that Sïbawaihi could not conceive of condi-
tional structures as a combination of two independent sentences.
Rather, she synthesizes kalām with the two notions of (mā) yahsunu l-
sukūt ť alaihi and (mā) yastagnì lan (viz. ģairihi) to state that when Sïba-
waihi was concerned with speech units, he conceived of them in exactly
the same way as Harris did when he (Harris) classified them as
'utterances'.26
In a review of Mosel's dissertation, Beeston27 touches on the mean-
ing of kalām. He criticizes Mosel's interpretation of al-kalām al-auwal
in reference to the conditional protasis (in 387, 1 and 402, 11), and
draws our attention to the phrase auwal al-kalām , by which al-Ņalīl
elsewhere (377, 1) designates the first of two co-ordinated propositions.
How, then, can kalām be both the total stretch of utterance and each of
its portions (as in conditional structures)? Beeston's explanation,
which he attributes to Sïbawaihi, is that "conditional structures are
indeed two self-contained and independently intelligible propositions
linked by a nexus of a highly specialized kind". That is to say, Beeston
considers kalām as both a whole utterance and a potential utterance,
like each portion of the conditional structure. This point is later devel-
oped by Beeston from a different angle, when he follows Mosel in
synthesizing the notion of kalām and (mā) yahsunu l-suküt ť alaihi with
reference to the status of the dispensable parts (zawäHd) of the sen-
tence. Kalām , he concludes, "can be either potential, in the sense that
the speaker could if he wishes break off his discourse at a given point; or
actual, in the sense of including all the zawa*id which he in fact chooses
to add". For Beeston, then, kalām is roughly an equivalent of 'sen-
tence.'

Troupeau does not recognize in Sïbawaihi' s kalām the meaning of


sentence or utterance, as he states twice in his comparative study of
grammatical and logical terminology in Sïbawaihi' s time.28 He does,
however, identify the use of kalām to denote what one might occasional-
ly consider quasi-syntactical units which are: 'mot' (see above), 'dis-

26 Mosel, 18 and note 2.


27 Beeston (1976), 650.
28 Troupeau (1981), 43, 45. Also, Troupeau (1976) s.v.
6 ZDMG 138/1

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
82 Rafael Talmon

course'29 and 'énonciation' or


denotations of kalām as 'gene
Carter's ideas concerning t
rather indirectly. He emphas
Sibawaihi was interested inst
taqim) speech (kalām)?1 When
'utterance'32 he does not menti
two other expressions which
concept.33
It seems that none of these f
extent of the mediaeval scholar
of kalām in the Kitab. In the
made to reach a definite answ
this term.

2. The meaning of kalām in the Kitāb :


The various occurrences of kalām in the Kitab can be gathered into
two groups according to their reference: (a) speech, or the mode of
speech of the entire linguistic group; and (b) one utterance-unit (even a
specific one), instantiated by an individual member of the linguistic
group under discussion. In what follows, I attempt to prove that the
second referential 'group' is also an independent denotational group,
i. e., it retains the general meaning of "speech" and yet, being under cer-
tain contextual constraints it creates various denotations. Sibawaihi, I
claim, was conscious of these denotations and their common retention
of the meaning of "speech". This meaning is still retained even in those
loci where kalām refers to sentences or complete utterances.
29 Ibid.
30 Troupeau (1976) s.v. K-L-M. He counts it 277 times. The notion 'quasi
syntactical units' means that kalām refers to a fragment of a specific sequence
of speech or to the whole of it, and denotes it according to a given desirable
aspect. Of course, the specificity of the sequence of speech is context-bound.
31 Carter (1972), 495 n. 23.
32 Carter (1973), 150 Verso.
33 Other scholars seem to have either ignored or paid little attention to the
meaning of kalām. Levin (1981) 146 recto, translates lam yakun U-y alcuna hādā
kalãmanby . . It is impossible for this to become a complete sentence"; Rund-
gren, 124 explains the usage of kalim at the head of the first chapter instead of
kalām as follows: S. geht hier also nicht vom kalām im Sinne des einfachen
Hauptsatzes explizite aus . . . also Frank, 292, who translates Ķadafa l-kalām
by ". . . has abridged the sentence by ellipsis". Baalbaki seems to share the
hesitation as to the meaning of kalām and leaves it (p. 14) as "some kalām (i. e.,
musnad )". For Jahn's translation on some crucial points see n. 39, below.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
A Study in the history of sentence-concept 83

(i) Use of kalām in reference to the speech of the entire language com-
munity:
(a) kalām is "prose" (vs. "poetry"). "Prose" is "the everyday speech":
238, 11: qalilfi l-kalãm kalir fi 1-šiW ; 335, 1: yaýuzu fi l-ši*r . . . wa-lā
yaýuzu fi l-kalãm ; further: 387, 18-20 (on which see below, p. 88).
(b) Kalām is "everyday speech", with a normative syntactic structure
as opposed to matal lil- Arab . . . wa-qad ubtudi'a fi l-kalãm 'alā ģair dā
l-ma'nā. also 42, 13 (for yutakallamu bihi vs. tamtil ; 156, 8; 157, 11.
(c) kalām is either an actual "sequence of speech" or a segment of
such a sequence and it is contrasted with isolated segments or even
sequences of speech which are either theoretical or meaningless due to
a certain structural anomaly. These last are presented as lam yakun
kalāman , or lã yutakllamu/yataJcallamüna, bihi as is mubtada ' without
its complement and vice versa in 2 18, 1 1 , or the status of the 1st person
suffix pronoun/-!/ before it is attached to a certain noun: 274, 10: wa-li-
annahū lã yakünu kalāman hattā yaJcüna fi l-ism ka-mā anna l-tanwin idā
lam yakun fihi (viz. fi l-ism) lam yakun kalāman.
The appearance of a verb where a noun should occur, as in Huna ycuļ-
rìba yaHiriã creates a meaningless sequence ( lam yakun kalāman) , see 2 ,
15; further 23, 16 (for *mā Zaidun ť ãqilan lAmrun) . Similarly see 45, 14;
238, 10. Also 329, 18, where *atauni mā hāšā Zaidan "they came to me
except for Z." (in na$b) is said to be lam yakun kalāman , because hāšā,
being a particle cannot follow the relative noun mā. Only verbs can be
preceded by relative nouns in such "exceptive" constructions. There-
fore, . . . māķalā Zaidan is an acceptable structure, since halâis a (tran-
sitive) verb, while ķāšā is not. See further (lã yutakallamu/yatakalla-
mūna bihi) 304, 1,4; 335, 12. 343, 15.
In 223, 13 Sîbawaihi contrasts two incomplete structures: kāna Zai-
dun and (ļaraba Zaidun. He judges the first as lam yakun kalāman
whereas the other is said to be kāna kalāman.
One might conclude that in this passage kalām takes the meaning of
'sentence' or even 'a minimal combination of sentence mainparts'. So,
"Z. hit" includes a verb + agent, which are sentence main parts. "Z.
was" needs a predicate and therefore it is not a complete sentence.
I have found only one additional occurrence of the pair kāna kalāman/
lam yakun kalāman: 23, 15-16 concerning the structures mā Z. <āqilan
abūhu/*mā Z. ť ãqilan ' Amrun (already cited above). The conclusions
drawn above from other occurrences of lam yakun kalāman seem to sup-
port an interpretation of the same phrase in the case of kāna Zaidun as
"this sequence of speech is not actual, i.e. is not heard by (native)
speakers". Kāna kalāman applied to (ļaraba Zaidun may be, then: "this

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
84 Rafael Talmon

sequence of speech can be h


According to this interpretatio
Sibawaihi and referred to in
noted.
(d) Since kalām implies the performance of speech by the given lan-
guage community which uses it, we would expect Sibawaihi to refrain
from using the phrase kalām muķāl for structures which are unaccep-
table because they do not convey an intelligible message to the hearer.
A structure opening with aiy- is meaningful only as a question, not as an
affirmative, as Sibawaihi says in 353, 22: fa-huwa kalām fi l-istifhām
mvhāl fi l-iķbār ,35
(e) Kalām (i.e., "speech") is sometimes conceived in the Kitàb as a
framework for certain functions (mazuļi1), or for sentence positions
which are occupied by certain elements.36
(ii) Use of kalām in reference to concrete speech-units uttered by
individuals of the language community:
(a) Kalām refers to a whole utterance:
In chapter 26 Sibawaihi explains the na$b- case in lAmr or the struc-
ture /ra* aitu Zaidan wa-Amran kallamtuhü/ as due to its proximity to
the ism mançûb of the first portion. He prefers this variant to another
with raf (wa-Amrun), saying (35, 9): "It makes more sense to have the
kalām uniform" (Ja-kãna an yaküna l-kalām lalā waýh wahid . . . aqrabfî
l-mayķad)- Later, Sibawaihi explains the acceptability of the < Amrun
variant by (35, 2 1) : fa-içlã ģāza an yaküna fi l-mubtada* bi-hādihī l-man-
zïla ģāza an yaküna baina l-kalām "and if it is acceptable to have it in the
same status (i. e., raf- case) in an initial position, this is also acceptable
inside the kalām". Sibawaihi' s use of kalām in reference to coordinated
sentences is quite frequent. I have noticed two loci in which he marks
the two sentences combined into one utterance by kalāmāni, which can
mean either "two independent sentences" or "two speech-sequences
which can occur independently in living use". See 435, 18 (kalāmāni as
contrasted to 435, 11 (kalām wāķid)' also 433, 5.

34 Mosel 18 considers kalām here as 'sentence'.


35 The phrase kalām muhãl in Abboud, p. 60 is Abboud's own back-forma-
tion of the . . . (bab) al-istiqãma min al-kalãm wal-ihãla heading of the introduc-
tory sixth chapter. He rightly observes later that "muķāl is often set over against
kalām 'speech* ..." (ibid.). Sibawaihi's words kāna l-kalām muhālan (in 324, 20;
362, 3; 390, 18) should be taken as "(then,) the (actual) speech becomes
muķāl" as is expressed by ģ a1 alta in 353, 12.
36 See Carter (1972), p. 147 recto and esp., p. 148 recto. Also Versteegh
(1978), p. 263 and p. 272. The latter notices that mazuļi* differs from mauqi1.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
A Study in the history of sentence-concept 85

(b) Kalām refers to potentially independent units:


The structure /qad ť alimtu a-'Amrun ķairun min Zaidin/ includes a
verb without the 'amal-effect, which Sibawaihi interprets as due to the
previous lamal set in the structure before lalimtu was inserted. He uses
the expression li-annahū kalām qad ť amila bal(ļuhū fi bac<ļ (99, 16). This
expression is used by Sibawaihi several times.37 Interestingly, we also
find the expression (kuli) ŠAT lamila bal(ļuhū fi bal(ļ (ii 60, 8), applied
to a structure which is potentially independent. It is, however, impos-
sible to draw further conclusions from this single occurrence of šai ' in
this expression.
Jļattā does not stand in an initial position, but always follows a previ-
ous part of an utterance . . . calā l-kalām allodi qablahã (39, 3). Here
kalām refers to a potential utterance. Finally, Sibawaihi contrasts
kalām mu'allaq to kalām mubtada y (387, 1,8). Both segments under dis-
cussion are potentially independent utterances. Further, see his treat-
ment of idan in 365, 13; 366, 4, 12. Kalām in reference to clauses is
found, inter alia, in 259, 12; 411, 7 (in 412, 2 such kalām is defined as
ģair mustaģnin). Further, 439, 1.
(c) While the references indicated by kalām to utterances and poten-
tial utterances could support the 'utterance' (even ' sentence') inter-
pretation of this word in the Kitab , there are other loci in which kalām
clearly refers to speech segments which do not have the potential of an
independent occurrence:38
The verbal noun %ann can be devoid of any 'amaZ-effect only when it
follows another member. Thus, zanni Zaidun (ļāhibun / is ruled out
while /matā ?annaka Zaidun dãhibun/ is acceptable (51, 11): ... li-
anna qablahã kalāman "because it (sc. %ann) is preceded by some (seg-
ment of) speech".
Sibawaihi suggests deriving the superlative structure /auwal jāris/
from /auwal min al-fursān/ and explains the transformation by (232,
15) : wa-innamã yarādū min al-fursān fa-hadafü l-kalām istiļjfāfan "what
they meant was 'of the knights'. They omitted this (segment of) speech
out of quest for alleviation". The omission seems to include the whole
prepositional phrase.39

37 176, 7; 394, 15; 409, 20; 450, 1.


38 The identification of kalām with 'mot(s)' in Troupeau (1976) [see p. 81,
above] must be based, at least partially, on occurrences of kalām with the deno-
tation of the present category. Unfortunately, no specific references are given
there.
39 Jahn translates: . . der Ausdruck ist dann zur Erleichterung abgekürzt".
He certainly identifies a 'sentence' notion in Sibawaihi' s 'kalām1 and is fairly

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
86 Rafael Talmon

In the structure /hčidihī mV


said to be an attribute. Since th
Sibawaihi explains (236, 14) : f
ment of) speech becomes an
The structure /mā laka wa-
(129, 9): . . . ķamalta l-kalãm
ly) this (segment of) speech
In several other loci, kalām
l-kalãm *alã seems to refer s
have the potential of indepen
is the oAod-segment in lã mi
*mã ra* aitu min aķadin illã
acceptable.
In 255, 16, kalām in the same expression seems to refer to a sentence
structured by li (unexpressed) + labdãni (raj*). The same uncertainty
applies also to 423, 9.
To sum up, kalām in the expression ķamala l-kalãm lalã is ambiguous
in many of its occurrences. In three of the examined loci it may be refer-
ing to segments of speech smaller than a sentence but in two of these
same loci a different interpretation is also possible. According to this
interpretation kalām may refer to the whole sentence, not only to a
single segment of it.
In 237, 8-9 Sibawaihi analyzes the structure /fihã qãHman raýulun/
"in it (viz. in the house) there is a man, standing." QãHm is defined as

consistent in this identification (cf., e. g., Jahn to 23, 15, 16; 39, 3; 218, 11. Even
in 51, 11: li-anna qablahü kalāman [see above] he translates: . weil der Satz
vor Zannaka beginnt" [while al-Sirãfí, vol. I, 581, 4 - paraphrases kalām by
SaVX] but Jahn sometimes either refrains from translating ' kalām ' (e.g., his
translation to 129, 10 and 236, 14 which will be discussed below; further 387,
18) or translates it as 'Rede' (e.g., 377, 1 for li-anna auwal al-kalām ķabar ; fur-
ther 35, 9, 21; 274, 10).
40 I am inclined to reject the possibility of interpreting kalām here paradigma-
tically (sc. "and this class of words . . ."). I cannot find any occurrence of such a
denotation of kalām in the Kitab.

41 Note, however, that just before this statement (129, 8) Sibawaihi con-
sidered the whole structure (either the whole sentence or the relevant construc-
tion commencing with the preposition or its equivalent) as kalām : . . . li-qubķ al-
kalām idā humüa āķiruhū ť alā auwalihï.
42 In 308, 17 Sibawaihi defines the structure lā miļluhū raģulun as ķamaltahū
lalā l-maiuļi1 "you treated it (syntactically) according to its (basic) syntactical
position*. The pronoun in ķamaltahū seems to refer to the sentence mitluhū raģu-
lun.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
A Study in the history of sentence-concept 87

hāl and raģul in the part which joins fihā as being its complement (its
mabni talaihi ).43 Fihā is said to be al-kalãm al-auwal , which I under-
stand as "the former (segment of) speech."
Other occurrences of kalām in which it seems to refer to sentence
constituents rather than the whole sentence are 106, 18; 183, 12; 301,
5; 306, 8 and 415, 7. 44 In 258, 9 and 304, 16 the expression al-kalām al-
auwal implies that the following speech- segment, which does not have
the potential to occur independently, is considered by Sibawaihi as al-
katām al-āķir. In 304, 16 al-kalām al-auwal refers to the segment lã
ģulāmaini "no two servants . . ." which occupies the position of mub-
tada' Contrary to Be e ston (see p. 81 above), we can maintain that al-
kalām al-auwal simply meant for Sibawaihi "the former speech-seg-
ment" of structures in which he analysed the syntactical relationship
between "a later speech- segment" and its antecedent.45
These references by kalām to speech- segments which do not have the
potential to occur independent give ample support to the assertion that
in all its occurrences in reference to concrete speech-units, kalām main-
tains the basic meaning of "speech".
The very high degree of non-specificity of kalām in reference to con-
crete speech-units is restricted in its various ways by both syntactic
modifiers (such as the article, demonstrative pronouns, attributes, suf-
fix pronouns and i<ļāfa- relations) and the extra-linguistic context itself.
Its frequent participation in expressions such as $adr al-kalām , auwal
al-kalām and āhir al-kalām , its combination with the verbs (sc. ibtada*a
kalāmahū, inqaxjã l-kalām and the like, even yastagni l-kalām wa-yaktafi
in 345,23 and elsewhere) and its above-mentioned references to utter-
ances might have created the illusion that kalām is a syntactic term
which means "an utterance" or even "a sentence". Actually, Trou-
peau' s classification of the various denotations of kalām as "general" is
far more plausible.
I believe that even those scholars who still maintain that part of the
references of kalām to 'sentence4 in the Kitāb are actually denotations of
such a syntactical entity will agree, in light of the evidence presented
above, that the use of kalām by Sibawaihi as a syntactical term is far
from being necessarily connected with a systematic concept of sen-
tencehood.

43 On this structure see Levin (1985) p. 312 and n. 74. On a similar structure
see Carter (1973), p. 491 f.
44 423, 9 is too opaque for me to understand. 415, 7 clarifies 414, 18.
This is farther continued by 325, 4; also 159, 18 and 161, 12.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
88 Rafael Talmon

The fact that kalām in referen


unspecific causes Sibawaihi to u
with kalām , which denotes "p
wa-hādā kalām aktaruhū fi l-šťr
kind of) speech (-segment), which
prose", further 387, 18; also 12
occurs side by side with refer
We may summarize our discu
Kitab in the following few sent
quently in the Kitāb. It refers
by the entire language commun
by individuals of the group. Th
or 'sentence' (suggested by M
investigated by examination o
some of the occurrences of kalām
ones, there are others which r
(according to Sibawaihi's conce
that kalām has only one mean
vary according to the context
In conclusion, Sibawaihi did no
syntactic concepts of 'sentence'
ance concept by the notions ya
terms which both Carter an
81), and which amount to the
'minimal sentence structure'.
kalām , proposed by both Mose
the illusion among scholars tha
parlance. This illusion has been
medieval grammarians. Among
Hišām who used the expression y
of kalām as 'sentence'. The expr
in the definition of ģumla)4* and
kalām , see above p. 78 in the

46 Kalām al-1 Arab stands as a relat


fore, Sibawaihi finds no difficulty in
sequence of speech" in 50, 6-7.
47 Mosel, 217 considers harf as a
sion of harf rightly observes (366-7
"a word" ("reference" is my word -
to the 'Volkssprache'.
48 See Kitab al-Luma1 10, 14.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
A Study in the history of sentence-concept 89

ģumla) seem to reflect early attempts to locate a technical term which


means 'sentence' in Sibawaihi' s Kit ab. 49
We can now return to our initial interest in the meaning of kalām in
Sibawaihi' s technical vocabulary.
The ķikāya of kalām in Kitab 50, 6- is, according to my view, 'a repro-
duction of a (whole) speech unit', not 'a citation of 'sentence". The
dichotomy of kalām with qaul is not a dichotomy of 'sentence* and 'part
of a sentence' as Ibn öinni and Ibn Sida maintained in their interpreta-
tion of the passage in the Kitab but a distinction drawn between a
speech unit already heard from another person (viz. kalām), and then
literally quoted (viz. yuķkā), vs. a stretch of speech, serving as a mes-
sage which does not contain the quotational feature (viz. qaul). This
idea is brought foreward clearly in Kitab 420, 18 when Sibawaihi
explains that the conjunctional anna "that", unlike inna , turns an actual
speech unit into (mere) information: li-anna anna tuçayyiru l-kalāma ķa-
baran.50
The fact that Sibawaihi demonstrated the difference between quota-
tion and information by using a distributional device (viz. *qāla Zaidan
... ; Zaidan . . . = al-qavla) and the somewhat accidental occurrence of

49 In the Kitab , muktafin is synonymous to mustaģnin. The allegation of a


Kufan origin of this term vis-a-vis Basran roots of mustaģnin (see Weil, p. 31
n. 2) requires reconsideration.
50 Sibawaihi' s observation is concerned with the following contrastive pair of
sentences:
(7) intalaqa l-qaumu ķattā inna Zaidan la-munļaliqun "The people have left,
even Z. is leaving."
(8) *inļalaqa l-qaumu ķattā anna Zaidan la-muntaliqun.
Sentence (8) is deemed unacceptable ( muķāl , cf. 420, 17, 18), because the por
tion opened by anna is equivalent to the abstract noun intilaq "leaving". Its mes
sage "The people have left, even an information (concerning the leaving [of Z.])
is meaningless.
Elsewhere (420, 1), Sibawaihi uses the word ša*n (taýlalu l-kalāma šá*nan). I
is uncertain whether ša*n is an equivalent of ķabar (in 420, 18) or it is part of h
immediately following illustration by the sentence *qāla l-ša'na mutafãqima
"He said that the matter was serious." It is beyond the scope of the present stu
dy to elaborate on all of S.' s subtle observations concerning the difference b
tween the use of anna and inna. See further 418, 14, 16 (compare Q-$-$ in Trou
peau [1976]).
Another spectrum through which one would expect Sibawaihi to examine the
hikâya vs. non -Ķikāya expressions is that concerned with the contrastive qalü
salāmun/salāman Quranic occurrences, esp. in li, 25. Actually, Sibawaihi deals
only with qalü salāman in Qur. xxv, 63 ( 136, 3) , but the significant phrase 'mafūl
bīhā ' which occurs in his analysis is found only is ms. A. According to Jahn's
note 16 in p. 204 "sie stören die Constr. gänzlich".

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
90 Rafael Talmon

the abstract noun qaul must h


waihi that the Master consist
sentence, vis-a-vis qaul , whic
syntactic units that constitut
We shall now proceed to tra
term.

C. Al-Farrā' Malānī l-Qur*ãn and early use of the term


Gumla

The Qur'ân commentary of ai-Farra' (d. 204/820) is the second oldest


existing book (second only to the Kitãb), devoted extensively to the
grammatical structure of Arabic.
In Mamani l-Qur'ān we find the word ģumla four times, all of them in
reference to 'clause'. We also find expressions which include the partici-
pial past muģmal(a), a total of five times, three of which also refer to
clauses. All the occurrences will be discussed below.

a) Gumla:
II 195,2,5: Al-Farrā* analyses the syntactic structure of Qur. xx, 128
/a- fa-lam yahdi lahum kam ahlaknā qablahum min al-qurün/. He concen-
trates on the syntactic status of kam and says that it is governed by the
verb ahlaknā and is virtually in a position (mauýť) of 7ia$&-case.51
However, the entire portion /kam . . . al-qurün/ occupies the position of
an agent to the verb yahdi. Therefore it is said to be endowed "with the
intention of (placing a single noun in) the nominative" ([ß' ma'nā l-
raf). In one of the sentences cited as an example of the structure, Qur.
vii, 193: sawa*un talaikum a- da lautumuhum au antum çâmitûna , the por-
tion a- da lautumūhum . . . etc. is identified as al-ģumla. In another sen-
tence, a-wa-lam yubaiyan laka man ya<mal ķairan yuýza bihī , the portion
headed by man is called ģumlat al-kalām "the ensemble of speech (occu-
pying a position of a single noun)."
II 333,5: Qur'an xxxii, 26 (see note 51, above) includes the same struc-
ture as Qur. xx, 128, already discussed above.52 Here again, al-Farrā*
refers to the same «clause which occupies the agent position. In his own

51 Al-Farrā's presents the wa$6-case reading as the only possible reading (lã
yakünu ģairuhū), although in II 333,1- when he discusses the same structure
(Qur. xxxii, 26) he accepts also a raf-v ariani. See next note.
52 The only difference is that Qur. xxxii, 26 begins with a-wa-lam.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
A Study in the history of sentence-concept 9 1

example qad tabaiyana li a-qãma Zaidun am lAmrun, he explains that


the portion a-qãma etc. is fa-takūnu l-ģumla marļu1 atan ji l-malnã.

II 388,1* Similar to the above, ai-Farra* analyses a self-constructed


sentence (i.e. one not found in textual sources) qaraHu mina l-Quryãni :
l-ķamdu li-llāhi rabbi l- alamina , which paraphrases there Qur. xxxvii,
78 and says about the portion of al-hamdu . . . etc. : fa-yakūnu fi l-ģumla
fì ma'nā naçb tarfaluhã bil-kalãm "as an ensemble [viz. occupying a
position of a single noun] it occurs with the intention (of placing) a
(noun in) the m?6-case and you consider it (the opening word of the por-
tion, viz. al-ķamdu) as marļu ť in actual speech. "
To summarize, in the four occurrences of ģumla , al-Farra1 uses it to
denote 'clause', a complete sequence of speech which occupies the posi-
tion of a single noun.

b) Muģmal(a):

III 226: Qur. lxxvii, 35 includes an annexion construction ( i(ļāfa ), the


first member of which indicates time, the other being a clause: (hãdã)
yaumu lã yanfiqüna.53 Although Qur'än readers read the first member
only in the raj*- case (yaumu), ai-Farra* considers the na$b- case (yauma)
in this construction as possible. In the analysis of the na$b- variation al-
Farrā' refers to the clause four different times. He cautiously draws a
distinction between three distinct types of linguistic units which may
occupy this position:

(1) a perfect verb (fa'ala ), (2) an imperfect verb (yafalu), (3) kalima
muýmala , also kalām muģmal , which is exemplified by ātīka hïna l-ffaý-
ģāģu amirun "I will come to you at the time 0. is a chieftain". At the end
of the passage al-Farrā* mentions the tri-partite division again, now
paraphrasing the third component with al-ism al-mvķbar lanhū "the
noun about which (some) predication is made". The two expressions
kalima muýmala and kalām muģmal doubtless refer to nexal construc-
tions in which the first member is a noun about which something is pre-
dicated. The distinction maintained by ai-Farra* between this and
nexal-constructions opening with a verb (the fa'ala/yafalu types) has
already been pointed out by Goldenberg from the writings of Ibn al-
Sarrāģ and al-Zaģģāgī (see above p. 76). The denotation of kalima
muģmala (as well as of kalām muģmal ), seems to be "a nexal construc-
tion which occupies the position of a single noun and as being such a

53 This is the subject of Chapter 260 in the Kitãb (409,5-410,3).

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
92 Rafael Talmon

construction it does not carry


Farrā> adds this point to the
"it is not marked by the ge
second part of an annexion)
The expression fiH muģmal is
According to one interpretatio
the m?6-case in yaum refle
phrase, whence the verse is
tiqüna "this is the state of a
adverbial phrase is constructed
verb (+ agent) in noun-posit

II 69,2-3: Al-Farrā* uses the


non-grammatical sense when
word represents a general id
In Qur. ii, 49 we read yasüm
wa-yastahyüna nisa'akum. T
gathers the details ( ķabar ),
slaughter your sons and hu

To sum up, we have surveyed


tical term ģumla and the ex
and fťl muģmal. This termi
century of Islam, a few dec
The existence of synonymou
muģmal in a non-grammatic
the phrase ģumlat al-kalãm
ensemble" (II 383), all leave t
the very first stages of its
Al-Farra' used ģumla etc. t
occupying the status of a sing
of the agent (II 195, 333), t
second part of the annexion
speech are either of the nou
verb (+ agent) variety, as is

54 The rather disturbing fact th


to the nexal construction in clau
the same reference may be exp
itself, whereas kalima emphasize
single word position. Consider t
conditional sentence takes the sy

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
A Study in the history of sentence-concept 93

In all their occurrences, these terms clearly express the concept that
the clause takes a position of a single noun, and therefore it has the
same syntactic status as that noun should have in this position. Al-Far-
râ* emphasizes the syntactic status of the "intended" noun (the clause)
by stating that it receives unrealized case (II 195, 333; 388).
We can conclude now that the term ģ umla originated in an early stage
of Arabic grammar out of the need to define clauses, either as the mini-
mal dependent sentences of the noun + predicate type, or even as
enlarged meaningful stretches of speech including nuclear and extra
nuclear constituents.
It is noteworthy that al-Farrā> was aware of the existence of nexus
relationship between a verb and its unrealized, 3rd person pronoun
when they jointly occupy (without an adjunct) the position of a single
noun (III 226). Nevertheless, he preferred to consider it as fa'ala/yafalu
or fťl muģmal , whereas the minimal pair of noun + predicate seems to
have gained recognition as a more typical nexal-construction, probably
because of the fact that the verb is a single morphological unit.55

D. The development of sentence terminology: al-Mubar-


rad

I have stated that for Sibawaihi kalām was not the grammatical term
which denotes 'sentence', and that he did not employ any technical
term for the denotation of 'clause'. I have also traced the early use of
ģumla as 'clause' in Ma1 ānī l-Quryān. I shall now examine the develop-
ment of sentence terminology by studying its occurrences in al-Mubar-
rad's two major works, al-Muqta(ļab and al-Kāmil.S6
A striking fact observed in al-Muqta4ab is the relative infrequency
with which kalām occurs throughout the grammatical descriptions in
this work. It seems that the basic meaning of kalām = "speech", identi-
fied above in Sibawaihi' s Kitāb , was still retained, and that the distinc-

55 Other explanations could also be considered. Note that Sibawaihi, who did
not have any fixed 'clause' terminology often conceived of the verb alone (fill) as
the counterpart of mubtada 1 + ķabar (mabni ialaihi). Cf., notably, 410, 1 : . . . al-
zamān . . . utļīfa ilã l-fi'l wa-ilã l-ïbtida* wal-ķabar. However, I shall not try to
offer Any such explanation in the present study.
56 Grammatical and other philological works from the intermediate period
will not be exhausted here. Note that in Mcfānī l-Qur*ān of al-AJifaš (d. 2 15/830)
the syntactical term ģumla does not occur (ģumla in p. 155 is a free word) . I have
not examined the occurrences of the term kalām in this work as well as in ai-Far-
ra'^ closely enough. For later sources quoting al-Māzinī's use of ģumla , see Gol-
DENBERG.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
94 Rafael Talmon

tion between its reference to


ty and its reference to conc
the language community als
nology."
The second referential group of kalām used by al-Mubarrad is signifi-
cantly restricted with respect to the number of classes of linguistic units
to which it refers in comparison with Sibawaihi's use of the term. Gene-
rally, this group refers only to a whole utterance. Reference to potential-
ly independent units is expressed by the term ģ umla (see below), where-
as speech segments which do not have the potential for independent
occurrence (also referred to, sometimes, by kalām in the Kitāb ) have
disappeared completely.58
The contexts in which kalām refers to whole utterances are well
defined. They include al-Mubarrad's observations about the position o
certain formatives in the utterance, and more significantly, the identifi-
cation of meaningful speech units. The following are examples of thes
contexts:

(a) bai in affirmative utterances is meant to correct a mistake just


uttered in the same utterance by the same interlocuteur. It can also ini-
tiate a response of the listener to an interlocuteur who utters a mistake:
inna bal lā ta'tīfi l-wāģib fi kalām wãhid illā lil-i(ļrāb balda ģalaļ wa-
nisyãn . . . wa-taqülu : lindī cašaratun bal ķamsata lašara talā mitli hātļā.
fa- in atā ba1 da kalām qad sabaqa min ģairihī fal-ķaļa* innamā laķiqa
kalāma l-auwal "Bal of the affirmative utterance does not occur in that
same utterance (kalām), unless it is meant to correct59 a mistake or
negligence . . . You say: I have ten- not so, fifteen, for example. If it fol-
lows an utterance (kalām) uttered by another, the mistake belongs to
the utterance (kalām) of the first person."

(b) Further examples concerning the position of formatives are: III 92


(kaifa" how", aina "where"); 275 (kaifa); TV 126 ( mubtada '); 140 (rubba
"many a . . ."). In all these loci al-Mubarrad deals with auwal al-kalām
"the beginning of an utterance". Also III 174 where the case of temporal
adverbs is discussed (mahall min al-kalām).

57 I have noticed the dichotomy ši'r-kalām in al-Muqta<ļab III 213.


58 In the difficult sentence in Bab al-Ibtidã* (III 89) I read: fa-lladi lã yumta-
nalu mvnhū kalām yuķbaru * arihū l-batta and translate: "[the relative] '(the one)
who' is never prevented from having a sentence attached to it [as its $ila], of
which something is predicated".
For the meaning of iļrāb in the Kitab , see Troupeau (1976) s.v. ļ)-R-B.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
A Study in the history of sentence-concept 95

(c) Meaningful speech units (kalām) are discussed in III 95; 100; 191;
IV 126; 187.

In III 191 we find al-Mubarrad discussing a very unusual structure,


ra'aitu lladaini lladi qãmã ilaihi ahūka "I saw the two persons who, the
person for whom they stood up is your brother". This self-constructed
sentence (constructed without independent textual sources) was meant
to demonstrate the acceptability of the sentence which includes the
necessary resumptive pronouns. In al-Mubarraďs words: fa-tammat
çïlat alladaini wa-$ahha l-kalãm "the [clause following alladaini, termed
its] §ila has become complete [sc. by the introduction of the resumptive
pronoun] and the [whole] utterance is comprehensible".
In al-Muqta4ab we find the term ģumla used rather frequently. It
mainly denotes a nexus construction, mostly with reference to clauses.
See IV 123; 124, where circumstantial and attributive clauses are dis-
cussed; IV 347, 348 where ģumla constitutes the second part of an anne-
xion construction, the first part being the temporal noun id (also IV 177,
cited by Goldenberg, p. 54).
In IV 347 al-Mubarrad presents the chapter's heading as hādā bab
i<ļāfat al-azmina ilã l-ģumal "this is a chapter in which [presented is] an
annexion of nomina temporis to ģumal One might think that ģumal
denotes nexal constructions which are either free or bound (also IV,
348). In IV 123; 125 the impression is that for al-Mubarrad, ģumla is
'clause', not 'sentence'.
Gumla seems to denote the whole clause without specific reference to
the nexal construction in III 130 where the sentence ra* aitu lladi in
y à* tini ãtíhi "I saw the one, who if he comes to me I shall come to him" is
discussed. Al-Mubarrad considered the conditional clause as ģumla
(also IV 78 which will be discussed below).
In general, al-Mubarrad seems to follow the use of the term ģumla
(and its like) by al-Farrā' However, on two points he clearly widened
its denotation: 1) He considered the nexal construction of fill + fā'il as
ģumla , equivalent in all respects to mubtada ' + ķabar (cf. IV 123 [in IV
125 its equivalent is a "nominal" clause abuhu muntaliqun ]; IV 347).
2) In two passages (al-Kāmil I 184,20 and al-Muqta4ab III 127) he un-
doubtedly intended ģumla to denote sentences.
Al-Mubarrad, then, is the missing link between the old grammarians
Sïbawaihi and al-Farrā> and the grammarians of the fourth century of
Islam. In his writings we see an interesting symbiosis in the use of kalām
as "speech" and the introduction of the term ģumlaīov the denotation of
"clause".

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
96 Rafael Talmon

The need for exact terminolog


clause units seems to have grow
rad responded to this need rat
bulary used by his predecessor
clause conception.
A convincing example of this
ģumla' s denotation. I have also
tion of kalām is narrower than
ber of classes of language units
smaller than it is in the Kitāb.
In Sections A and B we realized that Ibn Sida (or some other philolo-
gist) and Ibn öinni attributes the meaning of "sentence" to Sibawaihťs
kalām on the basis of an interpretation of Kitāb I 50, 6-.
It is uncertain whether al-Mubarrad also interpreted Sibawaihi
wrongly on this passage. However, in his closely related description of
the same syntactic topic, al-Mubarrad substituted the Master's kalām
for ģumla : IV 78: . . . qultu innamā yaqa'u ba'dahā l-hikãya idā kānat
GUMLATAN nahwa l-ibtida* wal-habar wa-mã ašbaha dālika. "(the verb)
'I saw' is followed by a quotation, if (this quotation) is a complete sen-
tence, such as (the combination of) mubtada ' + ķabar etc."
In al-Mubarrad' s writings we already observe a tendency to combine
the two terms which became an established usage. Since ¿furnia originat-
ed as a term indicating 'clause' its later use largely preserved this deno-
tation as its basic meaning. Kalām , which was used (either virtually or
according to interpretations) in the Kitāb as a less binding syntactical
term, functions in later grammatical writings mainly as a denotation for
the general framework of the sentence, always in reference to indepen-
dent syntactical entities.

List of Abbreviations

Abboud = P. Abboud: Sibawayhi's notion of grammatieality. In: A


12 (1979), 58-67.
al-Aķfaš = Sa'īd b. Masťada al-Aķfaš (al-Ausat). Ma'ānī l-Qur^ãn. Beirut 1985.
Baalbaki (1980) = R. Baalbaki: Some aspects of harmony and hierarchy in
Sibawayhi's grammatical analysis. In: Zeitschrift fur arabische Linguistik 2
(1979), 7-22.
Baalbaki (1982) = R. Baalbaki: Tawahhum: an ambiguous concept in early
Arabic grammar. In: BSOAS 45 (1982), 233-44.
Beeston (1970) = A. F. L. Beeston: Arabic Language Today. London 1970.
Beeston (1976) = A. F. L. Beeston: Review of Ulrike Mosel: Die syntak-
tische Terminologie bei Sïbawaïh. Diss. phil. München 1975. In: BSOAS 39
(1976), 648-53.

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
A Study in the history of sentence-concept 97
Carter (1972) = M. G. Carter: Twenty dirhams in the Kitab of Sibawaihi. In:
BSOAS 35 (1972), 485-96.
Carter (1973) = An Arab grammarian of the eighth century A. D.: A contribution
to the history of linguistics. In: JAOS 93 (1973), 146-57.
al-Farrā* = Abū Žakariyā> Yahyã . . . al-Farrā': Ma'āni l-Qur^ãn. Ed. M. ťA. al-
Naööär et A. Y. Naôâtî. Cairo 1955.
Frank (1981) = R. M. Frank: Meanings are spoken in many ways: the
Arab grammarians. In: Le Museon 94 (1981), 259-319.
GAS = F. Sezgin: Geschichte des arabischen Schrifttums. Vol. 1 ff. Leiden 1967 ff.
Goldenberg = Gideon Goldenberg: Subject and predicate in Arab gram-
matical tradition. In: ZDMG 138 (1988), p. 39-73.
Ibn al-Anbãri = Abū 1-Barakat Ibn al-Anbārī: Asrar al-Arabiya. Ed. Ch. F. Sey-
bold. Leiden 1886.
Ibn öinni: Has&is = Abu 1-Fath Ibn öinni: K. al-Hastfis. Cairo 1952-56.
Ibn öinni, Luma ť = Ibn öinni: K. al-Lumalfi l-nahw. Ed. H. Kechrida. Up
1976.
Ibn al-IJaģib = Ibn al-IJaģib: K. al-Kãfiya fi l-nahw with the commentary of al-
Astarābādī. Beirut (n.d.)
Ibn Hišām = ťAbd al-Mālik Ibn Hišam: Muqni l-Labib . . . Cairo 1964.
Jahn = G. Jahn: Sîbawaihi's Buch liber die Grammatik. 2. Bde. Berlin 1895-
1900 [Reprinted Hildesheim 19691.
Levin (1979) = A. Levin; Sibawayhi's view of the syntactical structure of kana
wa-ahawātuhā. In: Jerusalem Studies in Arabic and Islam 1 (1979), 185-213.
Levin (1981) = A. Lewin: The grammatical terms al-musnad, al-musnad ' ilayhi
and al-isnād. In: JAOS 101 (1981), 145-65.
Levin (1985) = A. Lewin: The syntactic technical term al-mabniyy lalayhi. In:
Jerusalem Studies in Arabic and Islam 6 (1985), 299-352.
LA = Ibn Man?ūr: Lisan al-1 Arab. Beirouth 1955.
Mehiri (1973) = A. Mehiri: Les theories grammaticales d'IbnJinnī. Tunis 1973.
Mosel = U. Mosel: Die syntaktische Terminologie bei Sïbawaih. Diss. München
1975.
al-Mubarrad: Kāmil = Abū l-ťAbbas . . . al-Mubarrad: al-Kamil. Ed. W.
Wright. Leipzig 1864-92.
al-Mubarrad Muqta4ab = al-Mubarrad: al-Muqta<ļab. Ed. M. A. Uģaima. Cairo
1965-68.
Rundgren = F. Rundgren: Uber den griechischen Einfluß auf die arabische
Nationalgrammatik. In: Acta Societatis Linguisticae Upsaliensis N. S. 2,5
(1976), 119-44.
§aťid al-Andalusī = §a4id al-Andalusī: K. Tabaqat al-umam. Cairo n.d.
Sibawaihi (ed. Derenbourg) = (Amr ibn Bahr Sibawaihi: al-Kitab. Ed. H.
Derenbourg. Paris 1881-9 [Repr. Hildesheim 1970].
Sibawaihi (ed. Būlāa) = Sibawaihi: al-Kitab. Ed. Būlāa 1316 h.
Talmon BSOAS = R. Talmon: An eighth- century grammatical school in Medina:
the collection and evaluation of the available material. In: BSOAS 48 (1985) 224-
36.
Talmon ZAL (1985) = R. Talmon: Who was the first Arab grammarian? a mw
approach to an old problem. In: Zeitschrift fur arabische Linguistik 15 (1985),
128-45.
Troupeau (1976) = G. Troupeau: Lexique-Index du Kitab de Sïbawayhi.
Paris 1976.

7 ZDMG 138/1

This content downloaded from


78.191.249.252 on Sun, 14 Feb 2021 18:37:42 UTC
All use subject to https://about.jstor.org/terms
98 Rafael Talmon, A Study in the history of sentence-concept
Troupeau (1981) = G. Troupeau: La logique d'Ibn al-Muqaffa1 et les origines de
la grammaire arabe. In: Arabica 28 (1981), 242-50.
Versteegh (1978) = C. H. M. Versteegh: The Arabie terminology of syntactic
position. In: Arabica 25 (1978), 261-81.
Weil (1913) = G. Weil: Die grammatischen Schulen vonKufa und Basra. Leiden
1913.
Weiss (1910) = J. Weiss: Die arabische Nationalgrammatik und die Lateiner. In:
ZDMG 64 (1910), 349-90.
Zaģģāģī, öumal = Abū 1-Qãsim . . . al-Zaģģāģī: al-Gumal. Ed. M. Ben Cheneb.
Algier 1926/7.

This content downloaded from


78.ffff:ffff:ffff:ffff:ffff on Thu, 01 Jan 1976 12:34:56 UTC
All use subject to https://about.jstor.org/terms

You might also like