The Word in Arabic 9004201432, 9789004201439


192 98 2MB

English Pages 276 Year 2011

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Contents......Page 6
Acknowledgements......Page 8
Introduction......Page 10
Section I The Word in the Arabic Linguistic Tradition......Page 24
The Concept of kalima in Old Arabic Grammar......Page 26
What is a kalima? 'Astarabadi's Answer......Page 42
Defining the Word within the Arabic Grammatical Tradition: 'Astarabadi's Predicament......Page 58
Ellipsis in the Arabic Linguistic Thinking (8th–10th century)......Page 78
Section II The Word in the Arabic Rhetoric Tradition......Page 92
‘Word’ in the linguistic thinking of 'Abd al-Qahir al-Jurjani......Page 94
On lafd and ma'na Again: Some Aspects of Their Relationship According to the balagiyyun......Page 118
Section III The Arabic Word in Contemporary Linguistic Theory......Page 154
Levels of Analysis of the Word in Arabic......Page 156
Automatic Extraction of Prepositions in a Corpus of Modern Standard Arabic Written Texts......Page 204
Section IV The Arabic Word in Contact......Page 222
Heavy and Light Borrowing of Arabic Verbs......Page 224
When Arabic Resonates in the Words of an African Language: Some Morphological and Semantic Features of Arabic Loanwords and Calques in Bambara......Page 238
Index of Names......Page 260
Subject Index......Page 265
Index of Arabic Words......Page 270
Language Index......Page 272
List of Contributors......Page 274
Recommend Papers

The Word in Arabic
 9004201432, 9789004201439

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

The Word in Arabic

Studies in Semitic Languages and Linguistics Editorial board

T. Muraoka, A.D. Rubin and C.H.M. Versteegh

VOLUME 62

The titles published in this series are listed at brill.nl/ssl

The Word in Arabic Edited by

Giuliano Lancioni and Lidia Bettini

LEIDEN • BOSTON 2011

This book is printed on acid-free paper. Library of Congress Cataloging-in-Publication Data The word in Arabic / edited by Giuliano Lancioni and Lidia Bettini.   p. cm. — (Studies in Semitic Languages and Linguistics ; 62)  The present volume is a selection of papers presented at the First Colloquium of Arabic Linguistics (CAL-01), held at Roma Tre University, Rome, March 1–3, 2007.  Includes bibliographical references and index.  ISBN 978-90-04-20143-9 (alk. paper)  1. Arabic language—Morphology. 2. Arabic language—Grammar. I. Lancioni, Giuliano. II. Bettini, Lidia. III. Title. IV. Series.  PJ6131.W67 2011  492.7’59—dc23 2011020327

This publication has been typeset in the multilingual “Brill” typeface. With over 5,100 characters covering Latin, IPA, Greek, and Cyrillic, this typeface is especially suitable for use in the humanities. For more information, please see www.brill.nl/brill-typeface. ISSN 0081-8461 ISBN 978 90 04 20143 9 Copyright 2011 by Koninklijke Brill NV, Leiden, The Netherlands. Koninklijke Brill NV incorporates the imprints Brill, Global Oriental, Hotei Publishing, IDC Publishers, Martinus Nijhofff Publishers and VSP. All rights reserved. No part of this publication may be reproduced, translated, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission from the publisher. Authorization to photocopy items for internal or personal use is granted by Koninklijke Brill NV provided that the appropriate fees are paid directly to The Copyright Clearance Center, 222 Rosewood Drive, Suite 910, Danvers, MA 01923, USA. Fees are subject to change.

CONTENTS

Acknowledgements  ...........................................................................................

vii

Introduction  ........................................................................................................ Giuliano Lancioni and Lidia Bettini

1

Section I The Word in the Arabic Linguistic Tradition The Concept of kalima in Old Arabic Grammar ...................................... Aryeh Levin What is a kalima? ʾAstarābāḏī’s Answer  ..................................................... Pierre Larcher Defijining the Word within the Arabic Grammatical Tradition: ʾAstarābāḏī’s Predicament  .......................................................................... Jean-Patrick Guillaume Ellipsis in the Arabic Linguistic Thinking (8th–10th century) ............. Cristina Solimando

17 33

49 69

Section II The Word in the Arabic Rhetoric Tradition ‘Word’ in the linguistic thinking of ʿAbd al-Qāhir al-Jurjānī  ................ 85 Antonella Ghersetti On lafḏ ̣ and maʿnā Again: Some Aspects of Their Relationship According to the balāġiyyūn  ..................................................................... 109 Lidia Bettini

vi

contents Section III The Arabic Word in Contemporary Linguistic Theory

Levels of Analysis of the Word in Arabic  ................................................... 147 Georges Bohas Automatic Extraction of Prepositions in a Corpus of Modern Standard Arabic Written Texts  ................................................................. 195 Giuliano Lancioni

Section IV The Arabic Word in Contact Heavy and Light Borrowing of Arabic Verbs ............................................. 215 Kees Versteegh When Arabic Resonates in the Words of an African Language: Some Morphological and Semantic Features of Arabic Loanwords and Calques in Bambara ............................................................................. 229 Francesco Zappa Index of Names ................................................................................................... Subject Index ....................................................................................................... Index of Arabic Words  ..................................................................................... Language Index  ..................................................................................................

251 256 261 263

List of Contributors  ........................................................................................... 265

ACKNOWLEDGEMENTS

The present volume could not be brought together without the colloquium which originated it. The authors would like to thank the Department of Linguistics of Roma Tre University for helping to organize the meeting and welcoming the participants at the Department venues; attendants to the colloquium have to be thanked too for their active, stimulating participation and for the questions and remarks which helped authors to focus their contributions and to partially reformulate the texts. The colloquium and subsequent editorial work were partly make possible by a grant from the Italian Ministry of University and Scientifijic Research for the Project of Relevant National Interest (prin) “Computer Analysis of the Hierarchical Structure of Arabic Lexicon: the Verbal System”. Fine-tuning work on the manuscript was partly put forth by Giuliano Lancioni during a stay as an invited professor at the Ecole Normale Supérieure—Lettres et Sciences Humaines, Lyon, and during Georges Bohas’s stay at Roma Tre University as an invited professor in the framework of the bilateral ENS–Roma Tre academic exchange. We would like to thank Cristina Solimando and Francesca Romana Romani for kindly reading the proofs of the book and making a number of valuable observations. Jeanna Malokos, Liesbeth Kanis, Franca de Kort and Jasmin Lange at Brill have followed the project during its diffferent stages; they all contributed to make the editing of this book enjoyable. A special thanks needs to be given to Renee Otto, who has patiently followed the crucial stage of proof editing and succeeded in turning a usually dreadful task into a gratifying experience.

INTRODUCTION Giuliano Lancioni and Lidia Bettini

The present volume is a selection of papers presented at the First Colloquium of Arabic Linguistics (CAL-01), held at Roma Tre University, Rome, March 1–3, 2007. The main theme of the colloquium was “The Word in Arabic”; this is a relatively original view of morphology, lexicography and syntax of Arabic, and some of the contributions presented are likely to shed new light on some highly-debated questions in both Arabic and general linguistics. Eight of the contributions derive from papers that were actually delivered at the colloquium (out of 15 papers presented); one more contribution was made by Aryeh Levin, who could not actually participate in the meeting, although he was an invited speaker. The papers included in the present volume are organized in four sections, according to a diachronic-to-synchronic development. The basic idea behind this organization is to start by a discussion on how the Arabic linguistic and rhetoric tradition in the Middle Ages analyzed words at the level of both defijinition and analytical practice, and then to go on with current revisions of the status of word and its analysis in contemporary linguistics, within and outside the Arabic-speaking world. The fijirst section, The Word in the Arabic Linguistic Tradition, includes articles about the status and the analysis of the word in the Arabic linguistic thinking and its relation to contemporary linguistic theories. The key term in this discussion is kalima, the closest Arabic equivalent to ‘word’, but the linguistic reality covered by this term is far from easily determined. The puzzling issue in the use of kalima in the Arabic linguistic tradition is that the term, while often fijitting quite well with our intuitive concept of ‘word’, sometimes refers to smaller units, which seem closer to the idea of ‘morpheme’ in contemporary linguistics; on the other hand, not anything that could conceivably be identifijied as a morpheme is denoted by kalima: other terms (e.g., ḥarf or zāʾida) are concurrently used, while the terminological borders between them are not always clear. The starting point for discussion is the seminal work by Levin (1986), who analyzed correspondences and diffferences between the use of kalima and the modern linguistic concept of morpheme in a wide range of early

2

giuliano lancioni and lidia bettini

and late Arab grammarians. The basic results of Levin’s analysis, in particular the distinction between kalima and zāʾida (‘augment’), are fairly summarized by Guillaume (this volume: 50). Aryeh Levin’s contribution to the present volume (“The Concept of kalima in Old Arabic Grammar”) refijines and extends the fijindings of Levin (1986, 2007) by proposing a fuller taxonomy of the diffferent terms employed and their contexts of use. Since most grammarians, including Sībawayhi, do not explicitly defijine most technical terms they use, Levin chooses to determine the meaning of terms involved by a thourough examination of their occurrences in the grammatical works (as attested by the copious footnote apparatus). The terms relevant for the discussions are kalima and zāʾida (pl. zawāʾid). While kalima generically corresponds to the idea of ‘word’, a “special sense” of the term is rather closer to the modern concept of ‘morpheme’, in the rather neutral defijinition of “smallest meaningful element” of the language. On the other hand, many items that would be understood as morphemes according to this defijinition are considered by Arab grammarians as zawāʾid instead: this term has a basically morphological origin, since it applies to the sounds added to the radical consonants of a word. While the term zāʾida often applies to meaningless elements (e.g., long vowels present in some word patterns), it can also apply to bounded meaningful elements, i.e. bounded morphemes. Levin chooses a very interesting methodological approach to determine in which cases zawāʾid are meaningful elements, by detecting explicit support in statements found in the texts themselves (pp. 25–27). The main part of Levin’s article is devoted to an inventory of cases where a morpheme (or a bundle of morphemes) is labeled as a kalima (pp. 21–23) and cases where it is labeled as a zāʾida (pp. 23–25). This distinction derives from considerations which are mostly internal to the machinery of the core principles of grammar as shared by linguists working in the Arabic tradition (with some paradoxical attribution, e.g. the kalima label attached to verb person-marking sufffijix in the perfect forms against imperfect prefijixes with the same function, which are regarded as zawāʾid instead), and overlaps in complex ways with the contemporary distinctions between words and morphemes. As a result of the thouroughness of analysis and the painstaking reference to the relevant loci in the literature, Aryeh Levin’s paper is likely to become the new standard of reference for the identifijication of “meaningful elements” in mainstream Arabic grammatical tradition just as his 1986

introduction

3

article (together with the résumé in his 2007 encyclopedia entry) has been ever since. While Levin’s discussion includes many of the most important Classical and Late Arab grammarians (Sībawayhi, Mubarrad, Ibn Sarrāj and Ibn Yaʿīš), many authors deliberately remain out of the picture. One such author is ʾAstarābāḏī, whose work has attracted a considerable scholarly interest in the last few years. Both Pierre Larcher (“What is a kalima? ʾAstarābāḏī Answer”) and Jean-Patrick Guillaume (“Defijining the Word within the Arabic Grammatical Tradition: ʾAstarābāḏī’s Predicament”) devote their contributions to a discussion of the doubtless innovative analysis of words by ʾAstarābāḏī. Signifijicantly enough for the richness and the complexity of the Arabic linguistic thinking, while the two papers largely overlap in textual coverage and reconstruction of ʾAstarābāḏī’s analysis, they difffer in a no-nonsense way in the evaluation of ʾAstarābāḏī’s role in the history of the Arabic grammatical tradition: while Larcher, by admitting a strong personal commitment to ʾAstarābāḏī’s work, stresses the revolutionary character of his theories and their powerful, as-yet not fully explored implications, Guillaume tends to play down this feature, by explicitly stating that “we could describe him as an enlightened conservative, not a radical reformist” (p. 61). What allows so divergent an evaluation and makes it compatible with textual data is the fact that ʾAstarābāḏī, while introducing many obvious innovations in his analysis of some of the more controversial cases in the linguistic data set of the classical Arabic tradition, does not introduce a new theory that renews the principles of the tradition itself: a move which would perhaps have been unthinkable, or at least extremely diffijicult to conceive, in the intellectual climate of late Arabic grammar, but at the same time a move whose lack limits the revolutionary character of ʾAstarābāḏī’s work. Thus, focusing on innovation or on conservation leads researchers to stress the relatively revolutionary, or conservative, character of his work. Another important point where the two authors’ attitudes clearly diverge is the relation between defijinitions and concrete analytical practice: while Larcher basically follows standard assumptions that analysis is based on defijinitions (assumptions that probably reach their acme in Owens’s 1990 method to promote implicit statements to the status of “quasi-defijinitions” in early Arabic grammar), and pivots his article on the efffort to explain away ʾAstarābāḏī’s “contradictions” with his own’s

4

giuliano lancioni and lidia bettini

defijinitions (pp. 35 fff.), Guillaume argues at length against giving too great an importance to defijinitions, which often are there to pay homage to tradition rather than to institute new theoretical stances, which often surface rather in empirical analytical work: in his opinion, they are often to be regarded as “a formal exercise”, introduced as an “outward manifestation of compliance with the dominant conception of what science should be”, with “limited, if any, practical implications”, at most useful to know “the kind of issues these resources served to discuss” (pp. 51–52). A fijinal point which epitomizes Larcher’s and Guillaume’s diffferent approaches to ʾAstarābāḏī’s text is the analysis of the latter’s treatment of discontinuous morphemes. In coping with the non-concatenative nature of many Arabic morphemes, ʾAstarābāḏī, while acknowledging the semantic contribution of these discontinous elements to the meaning of the whole, does not recognize them as kalima (as he would do for continuous bound morphemes in general, an innovation on the Arabic linguistic tradition standards) by limiting the defijinition of the latter to a (bound) morpheme “which immediately follows (mutaʿaqqib) the other” (Larcher, p. 38).1 Now, while even in this case Larcher stresses ʾAstarābāḏī’s originality and innovation within the Arabic tradition, Guillaume devotes the concluding section of his paper (“A killing objection: root and pattern”, pp. 66–68) to showing the contradictions of ʾAstarābāḏī’s analysis, claiming that the grammarian is compelled to fijind a way out “in adding an ad hoc stipulation to the defijinition of the complex word” (p. 67). The possibility and legitimacy of so strikingly diffferent evaluations of the same textual data clearly shows the complex, subtly interwoven character of the Arabic grammatical thinking as acknowledged by contemporary research, against the older vision of a rigid, monolithic and repetitive tradition as stressed by scholars such as Fleisch. While other articles in the fijirst section of the book discuss the analysis of words and its implications in the Arabic linguistic tradition, Cristina Solimando’s paper—“Ellipsis in the Arabic Linguistic Thinking (8th–10th century)”—examines the opposite side of the coin, ‘lacking words’ or ellipsis. The treatment of ellipsis in the Arabic linguistic thinking has not been much investigated by researchers—and in general ellipsis as a structural topic gets often a marginal status in linguistic theory and analysis, being

1  See the wide discussion of ʾAstarābāḏī ’s analysis in Larcher’s article (§ 2.1. Morphemes belonging to the non-concatenative morphology, pp. 36–38).

introduction

5

usually confijined to performance or discourse issues—despite its doubtless relevance for linguistic analysis (most real-world texts are of course elliptical). The main technical terms concerned with ellipsis in the early stages of the Arabic grammatical tradition are ʾiḍmār and ḥaḏf. Both words are ambiguous: ʾiḍmār, a syntactical concept, can be applied to “partial” ellipsis, e.g. the use of pronominal reference; ḥaḏf, primarily a phonological concept, is relative to the ʿelision’ of an element, whatever its status. In her article, Solimando reconstructs, by a thorough analysis of grammatical and tafsīr (Koranic comment) texts, a signifijicant segment of the history of these terms, by showing that ʾiḍmār starts as an exegetic, rather than a syntactic, term, which is applied to any missing element in early tafsīr. The passage to the syntactic concept of the term, clearly attested in Sībawayhi’s Kitāb, is shown in Maʿānī l-Qurʾān by Farrāʾ. The main part of the article is devoted to a detailed analyses of some of the passages in the Kitāb where ellipsis is analysed. A very interesting result of this analysis is the evidence for a terminological distinction of ʾiḍmār and ḥarf, which is clearly shown in passages where the two terms are used together; the distinction seems to be linked to the dychotomy ‘supposition’ vs ‘deletion’, with the fijirst term linked to a syntactico-semantic, the second one to a morphologico-phonetic, operation. This distinction gradually fades away: it is already declining in such an early treatise as Mubarrad’s Muqtaḍab, and the two terms become virtually synonymous in the 4th/10th century with Ibn Jinnī, who inserted ellipsis in a theoretical framework in which “the drop of a phoneme, the deletion of a part of the discourse and its supposition are assimilated and conceived as a single operation” (p. 81). In the aforementioned discussion on ʾAstarābāḏī’s vision of words and morphemes, Guillaume (p. 61), while refusing the label of a “radical reformist” to the late grammarian, thinks this tag could be more aptly attributed to Jurjānī instead. This leads us into the second section of the book, The Word in the Arabic Rhetoric Tradition. Another important tradition, parallel but quite distinct from grammar, within Classical Arabic linguistic analysis is balāġa ‘rhetoric’ (though the Arabic term covers a somewhat wider domain than its Western equivalents do); this tradition, especially the work of ʿAbd al-Qāhir al-Jurjānī, perhaps the most renown author in the fijield, is targeted in papers by Lidia Bettini (“On lafḏ̣ and maʿnā Again: Some Aspects of Their Relationship According to the balāġiyyūn”) and Antonella Ghersetti (“ ‘Word’ in the Linguistic

6

giuliano lancioni and lidia bettini

Thinking of ʿAbd al-Qāhir al-Jurjānī”); whereas Ghersetti investigates what concept of ‘word’ can be deducted from Jurjānī’s writings, Bettini’s contribution is mainly devoted to the dichotomy lafḏ̣/maʿnā, which she examines in a number of works of balāġa scholars. In her own contribution, Ghersetti endeavours to determine the originality of Jurjānī’s approach to grammar, analysing—alongside his more widely known rhetoric works—also some of the lesser known grammatical works. In the relevant passages, Jurjānī—while not excluding the classical “formal” elements for the identifijication of grammatical categories—constantly focuses on the semantic component of grammatical categorization. Thus, in tackling what is probably one of the most formal concepts of classical Arabic grammatical theory, ʾiʿrāb or ‘case-marking’, Jurjānī, after listing its formal properties, stresses its role in the ‘elucidation of meanings’ (ʾīḍāḥ al-maʿānī: p. 91), rather than considering its primary function as a word boundary marker. Perhaps the most striking aspect of this “semantically oriented” approach is the reduction, if not the refusal, of the concept of lexical meaning: according to Jurjānī, words have a meaning only in context, to the point that he argues that in case of lexical items expressed through construct state, such as ibn ʿirs ‘weasel’, we have two words formally, but one word only at the semantic level. As a consequence, Jurjānī’s analysis proceeds in a top-down, rather in a botton-up, fashion, by dividing the utterance in smaller units, rather than by mechanically adding the meanings of the single words to build a whole (p. 102). According to Ghersetti’s analysis, Jurjānī’s attitude can be summarized as follows: “The semantic, or conceptual, component is by far the most signifijicant since it assures the identifijication of ‘word’ as a discrete and coherent unit” (p. 106). Bettini’s contribution is devoted to the dychotomy lafḏ̣/maʿnā in the balāġī tradition. In this respect, three main subject matters are examined: the respective defijinition of the two terms in the rhetorical metalanguage, the relation between maʿnā and lexical meaning (especially in the analysis of verses), and how the latter term is implicitly delimited in the description of ornaments of style. The fijirst two sections (pp. 110–117) analyze the relationship between lafḏ̣ and maʿnā in the works of balāġiyyūn. Authors in this tradition agree upon the conceptual distinction (often stated in metaphorical terms) between the raw matter of the rhetorical craft and the form the craft gives to this matter. The table in p. 112 summarizes some of the ways authors, and even

introduction

7

the same author in diffferent works, identify this two aspects; in particular, the ‘form’ of poetic expression is sometimes expressed by maʿnā—which in fact gathers the meanings of ‘semantic content of an utterance’ and ‘image, idea’ expressed by a verse,—sometimes by lafḏ̣ (or kalām). Sections 3 through 5 of Bettini’s contribution (pp. 117–132) examine the relations between the concept of maʿnā and lexical meaning. A central concern in Arabic rhetoric works is the identifijication of diffferent ways (‘forms’, one of the meanings of lafḏ̣) to express the same concept, in order to proceed to a critical evaluation of the relative poetical and stylistic value of each alternative (e.g., by preferring terseness, or ʾījāz, see pp. 129–132). At the utterance level, this is obtained by paraphrasing or the use of one of several stylistic ornaments or fijigures; at the word level, while each word has its own denotation (its maʿnā in the narrower sense), the relation of the individual word with its contexts determines diffferent connotations and, as a result, diffferent meanings. This is linked to the top-down procedure of analysis advocated by Jurjānī and examined in Ghersetti’s paper. The last part of Bettini’s contribution (pp. 132–139) analyzes the relations between lafḏ̣ and maʿnā in the context of ornaments of style and borrowings (sariqāt). Contemporary linguistic analysis aims to put forward, or even to counteract, theories proposed by Medieval Arab grammarians and rhetoricians, by exploiting new approaches, both general and language-specifijic, offfered by current research on natural language. The two papers in section 3, The Arabic Word in Contemporary Linguistic Theory, present, respectively, a theoretical and an applied exercise in this domain: the contribution by Georges Bohas (“Levels of Analysis of the Word in Arabic”) focuses on the reinterpretation of the structure of the Arabic word in terms of contemporary linguistic research, especially within the author’s own theory of matrices and etymons; Giuliano Lancioni’s paper (“Automatic Extraction of Prepositions in a Corpus of Modern Standard Arabic Written Texts”) applies methods from computational linguistics to the task of tagging word classes automatically in untreated Arabic written texts from paradigmatic, context-free knowledge only, by capitalizing on diagnostic techniques inspired by Ibn Jinnī’s proposals. The paper by Bohas epitomizes the general theoretical framework and some recent developments of the theory he has been developing with his associates for the last decade: the Theory of Matrices and Etymons (tme). According to this theory, the traditional analysis based on the (mainly

8

giuliano lancioni and lidia bettini

triconsonantal) root is insufffijicient to account for the lexical relationships within the Arabic (and, in general, Semitic) lexicon in a satisfactory way, and it is necessary to make recourse to a diffferent theory, which appeals to two novel levels of analysis, which Bohas calls, respectively, “etymon” and “matrix”. In tme, the etymon is defijined as “a non linearly-ordered bi-consonantal base made up of two phonemes” (p. 153), that is a couple of consonants with possible order inversion. This accounts for a large series of cases, noticed in comparative Semitic studies at least since the 19th century, where roots that share two radicals have a, fully or partially, common meaning (a “notional invariant” in Bohas’s terminology; see a sample list in Paradigm 2, p. 148), a phenomenon that was previously explained within theories of (original) biconsonantism in Semitic languages. On its turn, the matrix is “a non linearly-ordered combination of a pair of phonetic feature vectors” (p. 153), that is a couple of unordered (bundle of) phonetic features. This concept tries to capture the presence in the Arabic lexicon of many cases (well-known to scholars working with Classical texts) where roots with some phonetic afffijinities, but with diffferent consonants, share a common meaning (see examples in Paradigm 3, pp. 149–150). Bohas conceives his theory as a three-layer hierarchical system matrix → etymon → radical, where the matrix is the most general category (a “virtual structure”) encoding a “notional invariant”, which in its turn is realized as several etymons (obtained by instantiating one of the several phonemes sharing the bundle of phonetic features of each element of the matrix). The radicals are obtained from etymons by distinct processes: difffusion, incrementation of a sonorant or guttural, prefijixation, blending (fusion of two etymons), incrementation of a glide (or transformation of a vowel into a glide), incrementation of the fijinal consonant, reduplication (see discussion in pp. 153–157). The possibility that an etymon is derived from diffferent matrices, and a radical from several etymons, gives the system a great flexibility and helps account for the high degree of ambiguity in the Arabic lexicon (e.g., roots which include two or more series of words with irreducibly diffferent meanings). In this respect, the article in the present volume presents the widest treatment of homonymy in Arabic within tme, and one of the widest discussion of homonymy in the Arabic lexicon in general, which will doubtless be of interest for many readers (see pp. 157–165). The inventory of matrices is limited by their very nature, since they correspond to the most general categories for the organization of the lexicon:

introduction

9

10 matrices have been identifijied so far by Bohas and coworkers, while an eleventh matrix, still to be fully identifijied, is tentatively proposed for the fijirst time in this volume (pp. 178 fff.). Bohas’s contribution is thus the most recent, widest presentation of tme, and supplements or replaces in many ways preceding formulations, including the two monographs, Bohas 1997 and Bohas & Dat 2007. A continuously growing fijield in Arabic studies is the design of computational linguistics applications to Arabic texts. Most such applications are currently adaptations of models devised for the computational treatment of Western languages, above all English, without taking into account the long tradition of linguistic study of Arabic both in the Arabic linguistic tradition and in Orientalist grammars. Giuliano Lancioni’s contributions proposes instead a reinterpretation of diagnostics introduced by Ibn Jinnī in Kitāb al-Lumaʿ to detect grammatical categories. Ibn Jinnī’s proposal, one of the most ancient examples of diagnostic methods in linguistics, chooses typical morphemes that can cooccur with nominal or verbal items only in order to assign a categorical label, leaving ḥurūf ‘particles’ as a last resort category for words that do not pass tests of verbality or nominality. Basically, Lancioni’s approach is to mark as candidate verbs words that can follow the aspectual marker qad, and as candidate nominals (an umbrella category which includes nouns, adjectives, pronouns and most adverbials) words that can follow the article (or determination morpheme) al-. The original proposal by Ibn Jinnī, who regarded as an ism “what is matched by a particle governing the genitive case” (leaving aside the other possibility, namely “what denotes a person”, for the difffijiculties in fijinding an automatical way to detect such a denotation), is introduced in a second step, where candidate nominals are confijirmed as such when following a preposition. In order to apply such a method to a large lexicon, Lancioni proposes a fully automated process which capitalizes on the paradigmatic properties of Arabic words. The input data for the system are represented by a large corpus of newspaper articles automatically segmented in words (for an overall sample of over 1 million words). Apart from the inspiration from Ibn Jinnī’s methodology, an innovative aspect of the system is the reliance on paradigmatic regularities only, without taking into account any grammatical information outside the corpus or any contextual information within the corpus. This approach has the advantage to tag unambiguously a large amount of words with

10

giuliano lancioni and lidia bettini

virtually no encoding efffort, in order to feed further analytical tools with plenty of linguistic information. In fact, it is surprising how such a general and “blind” approach is able to label correctly, and unambiguously, a relatively signifijicant portion of the Arabic lexicon in a virtually untreated corpus. More complex and linguistically more informed analytical tools would thus be able to concentrate on tagging only the words that were not labeled because of ambiguity or fuzziness of data. The process is divided in 6 cascading subtasks—marking unambiguously classifijiable particles, identifying candidate nominals, distinguishing preposition + nominal aggregates from non-aggregate words, distinguishing ambiguous conjunction + word aggregates, adding words after separately written prepositions, identifying candidate verbs—which progressively increase the number of lexical items unambiguously marked and reduce the amount of untagged items. Another point of view is of primary relevance in contemporary research on the Arabic word, namely the ways interaction with diffferent varieties within and outside the Arabic Sprachraum afffects the word, its structure and its status within the linguistic system. The role of Arabic as a classical and culture language, especially in connection with the difffusion of Islam, lead to a rich and varied network of influence from Arabic into other languages. Section 4 of the book, The Arabic Word in Contact, includes two papers about contact phenomena between Arabic and other languages. Kees Versteegh’s contribution (“Heavy and Light Borrowing of Arabic Verbs”) draws a general overview on how Arabic words are borrowed in a variety of languages as verbs, both directly and with the help of light verbs, while Francesco Zappa (“When Arabic Resonates in the Words of an African Language”) focuses on a single language, Bambara, which underwent a long, deep process of cultural and linguistic exchange with Arabic, by analysing some of the diffferent ways Arabic words adapted to a very diffferent linguistic structure. The study of verb borrowing is particularly interesting, since in general loans tend to be nominal elements in most cases; moreover, the peculiar structure of the Arabic morphological system makes it particularly diffijicult a direct borrowing of Arabic verbs into non-Semitic languages (the case of semitic Languages is obviously diffferent—see examples of Arabic loanverbs in Western Neo-Aramaic, p. 218—and even in non-Semitic Afroasiatic languages such as Berber, similarities in the morphological structure make direct borrowings relatively easier).

introduction

11

However, a signifijicant number of cases of Arabic verbal loanwords can be found in some Islamic languages. Versteegh detects two main strategies in this borrowing process: direct adaptation of an Arabic verb (“heavy borrowing”) and addition of an indigenous light verb to an Arabic nominal element (“light borrowing”), a familiar process from best-known Islamic languages, like Persian and Turkish. The simplest form of heavy borrowing of Arabic verbs can be found in Arabic pidgins and creoles, and testifijies the prevailing spoken character of the interaction which gave rise to the borrowing: the verb is borrowed in its imperative form, as a result of simple speech acts (commands, wishes). Similar cases can be found in Swahili, where the form of many verbal loanwords suggests the imperative as the most likely source (with many complex cases, however), or in West African languages such as Fulfulde. The use of light verbs, in what Versteegh calls the ‘DO-construction’ (named after the prototypical representative of the class), seems to be a more widespread strategy. Versteegh discutes several examples from such diffferent languages as Persian, Dravidian languages, or Nubian. The choice between the two strategies seems to be linked to sociolinguistic factors, beside structural considerations. Thus, while Hausa spoken in Western Sudan has a relatively small amount of verbal loanwoards from Arabic, the situation is very diffferent in Eastern Sudan, where Hausa is spoken in a situation of bilingualism and code-switching with Arabic, and Arabic verbs are often morphologically integrated. In general, the pattern of verb borrowing can be reconstructed in four steps, corresponding to four stages of progressively more intense language contact: fijirst, pidginization, which involves the adoption of bare (or imperative) verbs; second, extensive bilingualism, where DO-constructions appear; third, what Versteegh calls “full bilingualism”, with morphological integration of Arabic verbs; fourth, borrowing of inflected Arabic verbs, which is typical of a context of language shift (e.g., in Cypriot Arabic). Francesco Zappa’s contribution is devoted to the case of Bambara, a language which is very distanct from Arabic from a structural and typological point of view, but which underwent a strong process of linguistic contact for cultural and religious reasons, and therefore includes a large stock of Arabic loanwords in the lexicon. For these reasons, Bambara is an ideal subject for a case study of borrowings from Arabic into an Islamic language. Some of the cultural specifijicities of linguistic contact between Arabic and local languages within the process of Islamization can be described according to what Zappa calls “learned orality”, the practice of providing

12

giuliano lancioni and lidia bettini

oral periphrastic translations in the local languages of learned Arabic texts. Interestingly enough, the lack of a direct exposition to spoken Arabic in Bambara-speaking areas lead to borrowing from classical Arabic in the framework of Koranic pedagogy: thus, Arab-Islamic anthroponyms are often borrowed with case endings (e.g., Amadu from ʾAḥmad, see p. 232) or with tāʾ marbūṭa, which would not be the case if these words were borrowed from spoken Arabic. Variants found in borrowings should therefore be referred to adaptation to the phonology of the target language rather than to diffference in the Arabic sources. The main part of Zappa’s contribution (“Morphological and semantic aspects of borrowings and calques”, pp. 235–247) analyses some revant features in the adaptation of Arabic words in Bambara: integration and structural interference in the processes of inflection, derivation and compounding, cases of borrowing of elements as unanalyzed wholes, and phenomena of semantic shifts in the borrowing process. In order to make the text more accessible to linguists without an Arabist formation, the editors decided to consistently transliterate all Arabic words and quotations, and to ensure that translations and/or glosses are thoroughly provided. To ease the identifijication of proper names, Arabic names are given in short form without the article—e.g., ʾAstarābāḏī and not al-ʾAstarābāḏī,— while articles are reproduced when the full form of the name is employed. The fuller form of names, together with (approximate) date of death, is provided in the Name Index; both pieces of information have been uniformed, when possible, according to entries in the second edition of Encylopédie de l’Islam. The transliteration adopted complies with the standard of Brill’s Encyclopedia of Arabic Language and Linguistics; the only point where this transliteration style is likely to puzzle some readers accustomed to the Arabistic tradition is the use of the (phonologically more accurate) ḏ̣ instead than the more usual ẓ, which, however, is used in proper names.

References Bohas, Georges. 1997. Matrices, étymons, racines, éléments d’une théorie lexicologique du vocabulaire arabe. Paris and Louvain: Peeters. —— & Mihai Dat. 2007. Une théorie de l’organisation du lexique des langues sémitiques: matrices et étymons. Lyon: ENS.

introduction

13

Levin, Aryeh. 1986. “The Medieval Arabic Term kalima and the Modern Linguistic Term Morpheme: Similarities and Diffferences”. Studies in Islamic History and Civilization in Honour of Professor David Ayalon, ed. M. Sharon. Jerusalem: Cana & Leiden: Brill, 423-446. ——. 2007. “Kalima”. Encyclopaedia of Arabic Language and Linguistics. Leiden: Brill, 2.545–548. Owens, Jonathan. 1990. Early Arabic grammatical theory: heterogeneity and standardization. Amsterdam & Philadelphia: Benjamins.

SECTION I

THE WORD IN THE ARABIC LINGUISTIC TRADITION

THE CONCEPT OF KALIMA IN OLD ARABIC GRAMMAR Aryeh Levin

1. Introduction The Arab grammarians usually do not include any systematic discussion of their main theories and grammatical notions in their treatises. Nor do they explain the principles of their main theories. It is therefore not surprising that Sībawayhi and some other of the old grammarians do not say anything about their concept of ‘word’ (= kalima), nor do they defijine the term kalima. However, Sībawayhi’s concept of kalima can be inferred from data to be found in various places in the Kitāb. One of the most important texts for understanding Sībawayhi’s view of kalima is chapter 5081 of the Kitāb, which deals with the number of sounds composing Arabic words. Further important aspects can be inferred from Sībawayhi’s discussions of the following topics: the division of words into parts of speech (chapter 1),2 the categories of sounds composing a word (chapters 509 and 510),3 the phenomenon of ʾiʿrāb and the tanwīn. (chapter 2).4 In the works of Sībawayhi, Mubarrad, Ibn Sarrāj and Ibn Yaʿīš, the term kalima sometimes occurs in a sense corresponding to that of the modern linguistic term morpheme. This special sense of kalima does not occur, as far as I know, in the works of other grammarians.5 Sībawayhi’s concept of the occurrence of linguistic units which are the smallest meaningful elements in the Arabic language, basically accords with that of the modern concept of morpheme, although he divides such units into two groups: kalim and zawāʾid (see below). This paper proposes to discuss and explain the technical terms that in Sībawayhi are denoted by the word kalima, as well as the terminology he uses with reference to what we call morphemes. It also proposes to

1

 Chapter 508 = Kitāb 2.330.15–2.339.19.  Chapter 1 = Kitāb 1.1.1–8. 3  Chapter 509 = Kitāb 2.339.20–2.340.21; chapter 510 = Kitāb 2.340.22–2.342.24. 4  Chapter 2 = Kitāb 1.1.9–1.6.9. 5  However, see in this regard the discussion by Guillaume (this volume) and Larcher (this volume) about the use of kalima by ʾAstarābāḏī, which might in part correspond to the modern concept of morpheme. 2

18

aryeh levin

discuss the points of resemblance and diffference between the old grammarians’ view of the smallest meaningful elements in the Arabic language and the modern view of morpheme. The paper includes some new notions and conclusions, thus difffering in some respects from my previous studies on kalima.6

2. Sībawayhi’s Division of Words into Parts of Speech Sībawayhi divides words (= kalim) into three main parts of speech: ism (= noun), fijiʿl (= verb) and ḥarf jāʾa li-maʿnan (= a particle denoting one specifijic sense).7 These are divided into many sub-categories.8 It is inferred from the sources that every kalima must belong to one of the above three parts of speech.

3. The Sounds Composing a kalima 3.1 The Division of the Sounds Composing a kalima into ḥurūf and ḥarakāt The smallest phonetic units composing a kalima are called ḥurūf (singular ḥarf) and ḥarakāt (singular ḥaraka). The term ḥarf as a phonetic unit should not be confused with the term ḥarf as a particle.9 As a phonetic term, the form ḥarf denotes a sound, but it refers only to a sound which is represented in Arabic orthography by a letter, or, in phonetic terms, only to a sound occurring in both juncture and pause.10 Irrespective of this, it is attested by the text of the Kitāb that the distinction between a sound, 6

 See Levin (1986; 2007).  I believe that this is the correct explanation of this term. For arguments supporting it, and for other interpretations of the term by some old Arab grammarians and some modern scholars, see Levin (2000). 8  See, for example, Zamaxšarī (Mufaṣṣal: 51.3–58.5). 9  For the various meanings of technical terms denoted by the form ḥarf see Levin (2000: 25.33–26.7). 10  The fact that the term ḥurūf refers only to sounds represented in the Arabic alphabet by a letter is not accidental: as Ibn Jinnī (Sirr 2.491.2–2.492.15) says, the Arabic script is based on pause forms. Hence, only sounds occurring in pause, i.e., the consonants, the semi-vowels and the long vowels are represented in the alphabet. The short vowels and the tanwīn are dropped in pause; as a result they are not represented in the script, and are not called ḥurūf. Note that Nöldeke, independently of Ibn Jinnī, held the view that the Arabic script is based on pause forms (see Nöldeke 1904: 7; I thank my friend Simon Hopkins for this reference). The term ḥurūf also refers to sounds pronounced in a way 7

the concept of kalima in old arabic grammar

19

as a phonetic unit, and a letter of the alphabet was clear to Xalīl and to Sībawayhi.11 In referring to the ḥurūf the grammarians used the names of the letters of the alphabet, but they distinguished between the name of the ḥarf and its phonetic value.12 3.2 The Division of Sounds Composing Nouns and Verbs into ḥurūf ʾuṣūl, ḥurūf az-zawāʾid and ḥurūf al-badal In the grammarians’ view, most nouns and verbs are composed of two types of sounds: the radical consonants of the root from which the word is derived (= ḥurūf ʾuṣūl), and ḥurūf az-zawāʾid, i.e., sounds added to the radical consonants of the word.13 Some of the sounds added to the radical consonants, i.e., the short vowels and the tanwīn are called zawāʾid, since they do not occur in pause, and therefore they are not represented by a letter, and are not conceived of as ḥurūf.14 The above division of sounds may be illustrated by the classifijication of the sounds contained in the active participle form ḍāribun: this form contains the ʾuṣūl, the radicals ḍ, r, b; the ā, which is a ḥarf zāʾid; and the short vowels i, u and the tanwīn, which are zawāʾid. Some nouns and verbs include a sound belonging to the category of ḥurūf al-badal, i.e., the sounds substituted for one of the radical consonants of the word, or for one of its ḥurūf az-zawāʾid. For example: the ʾalif substituted for the radical w in qāla ‘he said’ (Kitāb 2.341.3), and the d substituted for the ḥarf zāʾid t in the verb izdajara15 ‘he was driven away’ (Kitāb 2.341.21) both belong to the category of ḥurūf al-badal.16 All the ḥurūf that belong to the categories of al-ʾuṣūl and al-badal are meaningless elements. Some of the sounds that belong to the category of ḥurūf az-zawāʾid, like the ā in ḥimār ‘a donkey’, are also meaningless

which in some dialects deviates from the standard occurring in most dialects (see Levin 1986: 425, fn. 13). 11  See Kitāb: 2.56.17–2.57.12. European scholars have sometimes incorrectly translated ḥurūf as ‘letters’ instead of ‘sounds’. See, for example, my remarks on Lane’s defijinitions of ḥurūf az-ziyāda in Levin (1986: 431, fn. 58), and on Weiss’s translation of ḥurūf as ‘Buchstaben’ in Levin (2000: 41.5–22). 12  See Kitāb: 2.56.17–2.57.12. 13  See Levin (2007: 2.546A.20–23); Levin (1986: 431.9–18, especially fn. 56); Levin (1978: 182.7–14). 14   See, Levin (2007: 2.546A.34–39); Levin (1986: 431.17–18). 15  In the example izdajara the d is substituted for the t which is characteristic of the verbal pattern iftaʿala. 16  For ḥurūf al-badal see Levin (2007: 546A.40–546B.3); Levin (1978: 15–21).

20

aryeh levin

elements. In contrast, some of the ḥurūf az-zawāʾid are meaningful elements.17 The above classifijication of sounds composing nouns and verbs does not refer to those sounds that are contained in words belonging to the category of the particle (= ḥarf) and to the sub-category of nouns which do not take a case-ending,18 including the pronominal sufffijixes determined by Sībawayhi as kalim (see below 4.2.1).

4. Sībawayhi’s View of the Smallest Meaningful Elements: kalima and zāʾida 4.1 Sībawayhi was aware of the occurrence of morphemes, i.e., of the occurrence of the smallest meaningful elements in the utterances of the Arabic language.19 However, unlike modern linguistics which classifijies all the smallest meaningful elements as morphemes, Sībawayhi divides morphemes into two categories: kalim (singular kalima) and ḥurūf zawāʾid (singular ḥarf zaʾid). Some sounds, like the short vowels and the tanwīn, which are not perceived as ḥurūf, are called zawāʾid (singular zāʾida) (see above § 3.2). The forms zawāʾid, zāʾida and ziyāda also occur as contractions referring to ḥurūf az-zawāʾid.20 Analysis of the data gathered from the Kitāb shows that Sībawayhi’s distinction between morphemes classifijied as kalim and morphemes classifijied as zawāʾid derives mainly from morphological considerations, and from the notion that etymologically, the essence of the sounds composing morphemes classifijied as kalim difffers from that of the sounds composing morphemes classifijied as zawāʾid: the sounds composing morphemes classifijied as zawāʾid belong to the category of sounds called az-zawāʾid or ḥurūf az-zawāʾid, while those composing morphemes classifijied as kalim do not belong to any specifijic category of sounds, i.e., they do not belong to any of the categories of sounds composing nouns and verbs, known by the terms ḥurūf ʾuṣūl, ḥurūf az-zawāʾid and ḥurūf al-badal.

17

 Ibn Yaʿīš (ŠMufaṣṣal Jahn: 2.1337.7–14) = Ibn Yaʿīš (ŠMufaṣṣal Cairo: 9.143.10–9.144.6).  See Ibn Jinnī (Munṣif: 1.7.1–1.8.16). 19  Hockett (1960: 123.10–11) defijines morphemes as follows: “Morphemes are the smallest individually meaningful elements in the utterances of a language”. 20  See, for example, Kitāb: 1.3.12; 1.3.20; 1.4.15 (ziyādatāni); 1.3.17 (az-zaʾida). 18

the concept of kalima in old arabic grammar

21

4.2 The Inventory of Morphemes 4.2.1 The Inventory of Morphemes Classifijied as kalim A detailed, although incomplete inventory of morphemes classifijied as kalim occurs in Chapter 508 of the Kitāb (2.330.15–2.339.19), in Mubarrad (Muqtaḍab: 1.36–1.52) and in Ibn Sarrāj (ʾUṣūl: 3.171.1–3.179.5). The subject of their discussions there is how many sounds, known to them as ḥurūf, may occur in a combination of sounds forming a kalima in the Arabic language. Chapter 508 of the Kitāb, which deals with this topic is titled: hāḏā babu ʿiddati ma yakūnu ʿalayhi l-kalimu ‘this is the chapter dealing with the number [of sounds known as ḥurūf] of which [Arabic] words (= kalim) are composed’ (Kitāb: 2.330.15).21 Immediately following this title Sībawayhi says: fa-ʾaqallu mā takūnu ʿalayhi l-kalimatu ḥarfun wāḥidun ‘The minimal [number of ḥurūf ] of which an [Arabic] kalima is composed is one’ (Kitāb 2.330.15).22 It is inferred from this statement that a linguistic unit which does not include a ḥarf, like the short vowels denoting the casemarkers, cannot be determined as a kalima. Another important point is that Sībawayhi’s inventory of kalim consisting of one ḥarf includes only pronominal sufffijixes and particles occurring as prefijixes. These kalim are bound forms which in speech are always afffijixed to given bases or words, and in writing never appear separately. Sībawayhi says that nouns belonging to the category of ism muḏ̣har and verbs cannot be composed of one ḥarf.23 Only some masculine singular forms of the imperative of doubly weak verbs, like ʿi ‘pay attention!’, consist of one ḥarf.24 Some of the pronominal sufffijixes composed of one ḥarf, which are kalim according to Sībawayhi and Mubarrad, are mentioned in Chapter 508 of the Kitāb: a. In Sībawayhi’s view, the following nominative pronoun sufffijixes of the perfect are kalim: 1. The fijirst person singular sufffijix -tu, as in faʿaltu ‘I did’ and ḏahabtu ‘I went away’ (Kitāb 2.331.12).25 21

 For Mubarrad’s similar title see Mubarrad (Muqtaḍab: 1.36.1–2).  The same text is found in Mubarrad (Muqtaḍab: 1.36.3). 23  An ism muḏ̣har is a noun which is not a pronoun, like farasun ‘a mare’. See Kitāb: 2.331.17–2.332.4. An explanation of this fijinding occurs in this text. 24  ʿI is the imperative of waʿā ‘he paid attention’. See Kitāb: 2.331.22–2.332.4. 25  See Kitāb: 2.331.11–12; cf. Mubarrad (Muqtaḍab: 1.36.10–11) who gives the example qumtu ‘I stood’. 22

22

b.

c.

d. e.

aryeh levin 2. The third person singular feminine sufffijix -t, as in faʿalat fulānatu ‘Such and such (fem.) did’ (Kitāb 2.331.14). In Sībawayhi’s view, this sufffijix is a feminine marker, and not a pronoun.26 Sībawayhi, or Mubarrad, classify the following accusative pronoun suffijixes as kalim: 1. The second person singular masculine sufffijix -ka, as in raʾaytuka ‘I saw you’ (Kitāb: 2.331.11).27 2. The second person singular feminine sufffijix -ki.28 The following genitive pronoun sufffijixes: 1. The second person singular masculine sufffijix -ka, as in ġulāmuka ‘your servant’ (Kitāb: 2.331.11). 2. The third person singular masculine sufffijix -hi as in ʿalayhi ‘on him’ (Kitāb 2.331.12). For the -ta in ʾanta, the -ka in ḏalika, the mā in ʾinnamā and the lā in lawlā as a kalima—see Levin (1986: 427.8–21). Each of the following morphemes, occurring as bound forms prefijixed to nouns and verbs, is a kalima, according to Sībawayhi: 1. The conjunction wa-, as in the example marartu bi-Zaydin wa-ʿAmrin ‘I passed Zayd and ʿAmr’ (Kitāb 2.330.17).29 2. The particle -ka ‘like, as’: ʾanta ka-Zaydin ‘You are like Zayd’ (Kitāb 2.331.2).30 3. The particle bi- ‘with, by’: xarajtu bi-Zaydin ‘I went out with Zayd’ (Kitāb 2.331.6).31 4. The oath particle wa-, known as wāw al-qasam: wa-llāhi lā ʾaf ʿalu ‘By God, I won’t do [it]!’ (Kitāb 2.331.8).32 5. The oath particle ta-: ta-llāhi lā ʾaf ʿalu ‘By God, I won’t do [it]!’ (Kitāb 2.331.9).33 6. The particle la-, preposed to the verb in the imperfect, after expressions denoting an oath, as in the example wa-llāhi la-ʾaf ʿalu ‘By God, I will do [it]!’ (Kitāb 2.331.10).34

26  See Kitāb: 2.331.10–14; 1.201.12–1.202.3. For this view held by other grammarians see the references in Levin (1986: 426, fn. 19). Since in Sībawayhi’s view this -t is not a pronoun it is not mentioned in his discussion of the pronouns in Kitāb: 1.188.1–10. 27  See Kitāb: 2.331.10–11. Cf. Mubarrad (Muqtaḍab: 1.36.13). 28  See Mubarrad (Muqtaḍab: 1.36.13). 29  See Kitāb: 2.330.15–18. 30  See Kitāb: 2.331.2. Cf. Mubarrad (Muqtaḍab: 1.39.4–6). 31  See Kitāb: 2.331.6–7; Mubarrad (Muqtaḍab: 1.39.11–13). 32  See Kitāb: 2.331.8. Cf. Mubarrad (Muqtaḍab: 1.40.1–2). 33  See Kitāb: 2–331.8–9. 34  See Kitāb: 2.331.10. Sībawayhi, who calls this particle lām al-yamīn does not mention in his example the expression of oath-taking which should precede this la-.

the concept of kalima in old arabic grammar

23

7. The interrogative particle ʾa-.35 8. The future tense marker sa-, as in the example sa-yaf ʿalu ‘He will do [it]’ (Kitāb 2.331.9).36 9. According to Sībawayhi and Ibn Yaʿīš the defijinite article al- is a kalima.37 The examples given by Sībawayhi are: al-qawmu ‘the people’ and ar-rajulu ‘the man’ (Kitāb: 2.335.5). It should be noted that Sībawayhi mentions Xalīl’s view that al- is a particle (= ḥarf ), denoting that the noun to which it is prefijixed is defijinite, and as such it forms a unit separate from the noun defijined by it.38 10. For the classifijication of the particles fa- and li- as a kalima see Levin (1986: 428.6–13). 4.2.2 The Inventory of Morphemes Classifijied as zawāʾid Sībawayhi classifijies the following morphemes, contained in verbal and nominal forms occurring in certain morphological patterns, as zawāʾid, and not as kalim: a. The prefijixes of the imperfect verb ʾa-, ta-, ya- and na-, known to the Arab grammarians as az-zawāʾid al-ʾarbaʿ ‘the four zawāʾid ’.39 For example: ʾaf ʿalu ‘I shall do’; taf ʿalu ‘you shall do’ (Kitāb: 1.2.5);40 yaḍribu ‘he will hit’ (Kitāb: 2.340.5);41 naf·alu ‘we shall do’ (Kitāb: 2.340.12).42 b. Sībawayhi classifijies the following sufffijixes, added to nominal bases, as zawaʾid: 1. The tanwīn which is the fijinal -n sufffijixed to nouns and adjectives, as in the proper noun Zaydun (Kitāb: 1.2.7).43 The Arab grammarians believe that the tanwīn is a marker denoting that the noun which receives it belongs to the category of al-ʾasmāʾ al-mutamakkina.44

35

 See Kitāb: 2.331.10.  See Kitāb: 2.331.9–10. 37  See Kitāb: 2.335.5; Ibn Yaʿīš (ŠMufaṣṣal Jahn: 1.21.11–14) = Ibn Yaʿīš (ŠMufaṣṣal Cairo: 1.19, 6–9). 38  See Kitāb: 2.59.1–14. 39  See Kitāb: 1.2.4–6; Mubarrad (Muqtaḍab: 2.1.8–14); Ibn Jinnī (Lumaʿ: 50.6–7); Zamaxšarī (Mufaṣṣal: 108.15–17); Ibn Yaʿīš (ŠMufaṣṣal Jahn: 2.915.21–23). 40  See also Kitāb: 2.340.14–15; cf. Mubarrad (Muqtaḍab: 1.60.8–9). 41  See Kitāb: 2.340.3–5. 42  See Kitāb: 2–340.8–12; cf. Mubarrad (Muqtaḍab: 1.59.5). 43  For the tanwīn as one of the zawāʾid see Kitāb: 2.340.8–10; Kitāb: 1.79.5–7; Mubarrad (Muqtaḍab: 1.59.12); Ibn Jinnī (Lumaʿ: 4.3–4; Munṣif: 15.5–13). 44  It may be inferred from data in the text of the Kitāb that the term al-ʾasmāʾ al-mutamakkina refers to nouns which have the following qualities: (1) these nouns can occur in many syntactic constructions; (2) they can have various syntactic functions; (3) they can 36

24

aryeh levin

2. The case-endings -u, -a, -i, as in the examples ad-dāru, ad-dāra and ad-dāri ‘the wandering territory of the tribe’.45 3. The feminine sufffijix -at-, as in ṭalḥatun ‘an acacia tree’ (Kitāb: 2.340.13).46 4. The feminine plural sufffijix -āt-, as in the example munṭaliqātun ‘going away (plural feminine)’ (Kitāb: 2.340.13).47 5. The ending -iyy- denoting relation when sufffijixed to nouns,48 as in the examples haniyyun (Kitāb: 2.340.7)49 and tamīmiyyun ‘Tamimite, belonging to the tribe of Tamīm” (Mubarrad Muqtaḍab: 1.57.10).50 6. The masculine plural endings -ūna/īna, as in the examples al-muslimūna/al-muslimīna ‘the Muslims’ (Kitāb: 1.3.23–1.4.1).51 7. The dual endings of the noun -āni/ayni, as in the examples ar-rajulāni/ar-rajulayni ‘the two men’ (Kitāb: 1.3.19).52 c. Sībawayhi classifijies the following sufffijixes, added to imperfect verb forms, as zawāʾid: 1. The mood-endings -u and -a, as in the examples ʾaf ʿalu (indicative) ‘I shall do’ (Kitāb: 1.2.5) and lan yaf ʿala (subjunctive) ‘He would never do’ (Kitāb: 1.2.8).53

take the case-endings and most of them can take the tanwīn; (4) they can occur either as indefijinite or defijinite nouns. In contrast to this, the term al-ʾasmāʾ ġayr al-mutamakkina refers to nouns which can occur only in a restricted number of syntactic constructions and functions, and which cannot take the case-endings and the tanwīn. These nouns must be either indefijinite or defijinite (see, for example, Kitāb 2.40.22–41.10; 33.12–14). In Levin (1986) I accepted Wright’s explanation of this term, which is based on some later grammarians’ views (see Levin (1986: 432, § 2, and especially fn. 68). The revised explanation of al-ʾasmāʾ al-mutamakkina suggested in the present note is preferable , since it is based on Sībawayhi’s view of this term. 45  Sībawayhi’s classifijication of the case-endings as zawāʾid is inferred from his view that all the short vowels are zawāʾid (see Kitāb: 2.342.21–24). These endings cannot be determined as kalim since a kalima must include at least one ḥarf (see above § 4.2.1). 46  See Kitāb: 2.340.12–13; Mubarrad (Muqtaḍab: 1.60.3). 47  See Kitāb: 2.340.12–13; Mubarrad (Muqtaḍab: 1.60.4). 48  This ending is known to Sībawayhi as yāʾay al-iḍāfa ‘the two yāʾs denoting the relation’ (see, for example, Kitāb: 2.64.10–12). For this term see Levin (1986: 437, fn. 98). 49  This form seems to be the nisba of hanun ‘a thing’ (see Levin 1986: 437, fn. 99). 50  For this ending as a zāʾida see Kitāb: 2.340.3–7; Mubarrad (Muqtaḍab: 1.57.10). 51  See Kitāb: 1.3.19–1.4.1; Mubarrad (Muqtaḍab: 1.59.9–11). Sībawayhi refers to the sounds contained in these endings as ziyadatāni ‘two ḥurūf belonging to the category of ḥurūf az-zawāʾid’ (Kitāb: 1.3.19–20). For the -na contained in these endings as a zāʾida see Kitāb: 2.340.8–11. 52  See Kitāb: 1.3.12–19; Mubarrad (Muqtaḍab: 1.59.9–11). The sounds contained in these endings are also referred to by Sībawayhi as ziyādatāni (Kitāb: 1.3.12). 53  Sībawayhi’s classifijication of the mood-endings as zawāʾid is inferred from his view that all the short vowels are zawāʾid (see Kitāb: 2.342.21–24). These endings cannot be determined as kalim since a kalima must include at least one ḥarf (see above §4.2.1).

the concept of kalima in old arabic grammar

25

2. The endings -īna (indicative) / -ī (subjunctive and jussive) denoting the second person feminine singular as in the examples ʾanti taf ʿalīna ‘you shall do’ (Kitāb: 1.4.20), lan taf ʿalī ‘you would never do’ (Kitāb: 1.4.20–21), lam taf ʿalī ‘you did not do’ (Kitāb 1.4.20).54 3. The endings -ūna (indicative) / -ū (subjunctive and jussive), suffijixed as plural endings to the second and third person forms of the imperfect verb, as in the examples hum yaf ʿalūna ‘they will do’; lan yaf ʿalū ‘they would never do’; lam yaf ʿalū ‘they did not do’ (Kitāb 1.4.18).55 4. The endings -āni (indicative) / -ā (subjunctive and jussive), sufffijixed as the dual endings to the second and third person forms of the imperfect verb, as in the examples humā yaf ʿalāni ‘both of them would do’ (Kitāb: 1.4.14); lan yaf ʿalā ‘neither would ever do’, lit. ‘both of them would never do’; lam yaf ʿalā ‘neither did’ lit. ‘both of them did not do’ (Kitāb: 1.4.15).56 5. The plural ending -na contained in the third person feminine plural form yaf ʿalna ‘they (feminine.) will do’ (Kitāb: 2.340.11).57 d. The plural ending -na contained in the third person feminine plural form of the perfect, as in the example faʿalna ‘they (feminine) did’ (Kitāb: 2.340.11).58 4.2.3 The zawāʾid as Meaningful Elements The data found in the Kitāb confijirms that Sībawayhi conceived of the morphemes that he classifijied as zawaʾid as meaningful elements. His view is attested by the following points:

54  See Kitāb: 1.4.18–21 (referring to the -ī as a ziyāda); Kitāb: 2.340.8–11 (referring to the -na as one of ḥurūf az-zawāʾid). 55  See Kitāb: 1.4.15–18. Sībawayhi here refers to the ending -ūna as zāʾidatāni (see Kitāb: 1.4.15). He does not here give examples of these endings sufffijixed to the second person. 56  See Kitāb: 1.4.3–15. Sībawayhi does not say explicitly that the ending -āni contains two ḥurūf zawāʾid, but it is inferred from Kitāb: 1.4.15 that this dual ending, like the plural ending -ūna sufffijixed to the imperfect verb, contains zāʾidatāni. He does not give an example of such an ending sufffijixed to the second person. 57  See Kitāb: 1.340.8–11. Sībawayhi does not give an example of this ending sufffijixed to the second person plural feminine. The ending is discussed in Kitāb: 1.4.21–1.5.8, but is not there referred to as a zāʾida. 58  See Kitāb: 2.340.8–11.

26

aryeh levin

1. Sībawayhi believes that the case-endings of the noun are meaningful elements, expressing a grammatical sense. The grammatical sense of these endings is liable to change according to their declension.59 2. Sībawayhi refers to the imperfect prefijix ta- as ḥarfijin60 jāʾa li-maʿnā l-muxāṭabati wat-taʾnīṯi ‘a ḥarf (= sound) which occurs in order to [denote] the sense of the second person61 and [the sense of ] the feminine’62 (Kitāb: 2.476.13).63 3. The terms referring to some of the endings testify that in Sībawayhi’s view these endings belong to the category of the smallest meaningful elements: a. The feminine ending -at- is al-hāʾ lit-taʾnīṯ ‘the hāʾ [occurring] in order [to denote] the feminine’.64 b. The feminine plural ending -āt- is called at-tāʾ lil-jamʿ ‘the tāʾ [occurring] in order to [denote] the plural’.65 c. The ending -iyy- denoting the relation is called yāʾay al-ʾiḍāfa—‘the two yāʾ’s denoting relation’.66 4. Sībawayhi refers to the masculine plural sufffijix -ūna which is added to nouns as al-zāʾidatāni lil-jamʿ ‘the two zāʾida’s [which occur] in order [to denote] the plural’.67 5. The masculine plural sufffijix -ūna which is sufffijixed to the second and the third person of the imperfect verb is according to Sībawayhi ʿalāma lil-jamʿ ‘a plural marker’.68 6. The text of the Kitāb attests that Sībawayhi conceived of the following endings as well as meaningful elements:

59  This notion which is inferred from Kitāb: 2.336.17–20, is discussed in detail in Levin (2000: 30.28–31.18). 60  Ḥarfijin here takes the genitive since it occurs as a muḍāf ʾilayhi (see Kitāb: 2.476.13). 61  As, for example, in forms such as taf ʿalu, taf ʿalīna, taf ʿalāni and taf ʿalūna. In this context Sībawayhi mentions the form tataḏakkarūna ‘you (plural masculine) will remember’ (Kitāb: 2.476.10). 62  As, for example, in the third person feminine form taf ʿalu. In this context Sībawayhi mentions the form tatanazzalu (Kitāb: 2.476.2), a feminine form referring to a subject in the plural meaning ‘they will descend’. 63  For the context where this expression occurs see Kitāb: 2.475.25–2.476.16. For a similar expression see Kitāb: 2.475.15–16. 64  See Kitāb: 2.196.11–12. Cf. Kitāb: 2.12.16. For hāʾāt at-taʾnīṯ ‘the hāʾ’s [denoting] the feminine’, see ibid. This ending is called al-hāʾ since its form in pause is -ah. 65  See Kitāb: 2.83.14. Cf. Kitāb: 2.340.12–13. 66  See Kitāb: 2.64.9–12. The sense of iḍāfa included in this term is the same as that of nisba ‘relation’ (see, for example, the title of Chapter 318 of the Kitāb: 2.64.9). 67  See Kitāb: 2.83.5. 68  See Kitāb: 1.4.15–18.

the concept of kalima in old arabic grammar

27

a. The tanwīn.69 b. The dual endings -āni / -ayni which are sufffijixed to nouns.70 c. The ending -īna contained in the second person feminine singular form of the imperfect verb.71 d. The ending -na denoting the feminine plural in imperfect forms such as yaf ʿalna ‘they (plural feminine) will do’.72 4.2.4 The Division of Morphemes Classifijied as zawāʾid It is inferred that the morphemes classifijied as zawaʾid are divided into two main groups: (1) morphemes conceived of as an integral part of the pattern of the word where they are included; (2) morphemes occurring as sufffijixes of nouns and imperfect verbs. The data gathered from the text of the Kitāb attest that fijive types of morphemes are classifijied as zawāʾid, and not as kalim. These morphemes are: 1. The case-endings of the noun and some nominal sufffijixes that Sībawayhi conceived of as corresponding to the case-endings. 2. The tanwīn and some endings conceived as corresponding to it. 3. Some sufffijixes preceding a case-ending. 4. The prefijixes of the imperfect verb. 5. The mood-endings of the imperfect verb. 4.2.4.1 The Morphemes Conceived of as an Integral Part of Noun and Verb Patterns Some sufffijixes and prefijixes attached to nouns and verbs are conceived of as an integral part of their patterns: 1. The tanwīn In Sībawayhi’s view, the tanwīn is the fijinal limit of the noun,73 and it is the part of the noun that makes it complete.74 These notions show that he believes that the tanwīn is an integral part of the kalima. 2. The case- and the mood-markers

69  In Sībawayhi’s view, the tanwīn is ʿalāmat at-tamakkun (see §4.2.2 and fn. 44 above). 70  See Kitāb: 2.89.21–22. 71  See Kitāb: 1.4.18–21. 72  See Kitāb: 1.4.21–1.5.1. 73  See Levin (1986: 436–437, §6.3.1). 74  See Kitāb: 1.274.8–10.

28

aryeh levin

Sībawayhi also conceived of the case- and mood-markers as an integral part of the patterns of nouns and verbs respectively. This point is inferred from his main discussion of the case- and mood-markers, where he refers to them as majārī ʾawāxiri l-kalim, lit. ‘the ways of the fijinal part of the words’.75 4.2.4.2 Sufffijixes Classifijied as zawāʾid although they are not Part of the Pattern of the Noun or Verb Some sufffijixes which are not part of the pattern of the noun or verb, are also classifijied as zawāʾid, because of Sībawayhi’s concept of the fijinal limit of nouns and verbs: 1. Since in Sībawayhi’s view the tanwīn is the fijinal limit of the noun, he believes that all the sufffijixes preceding it are an integral part of the kalima, and hence he classifijies them as zawāʾid. Hence, the vowels denoting the case-endings, and the sufffijixes preceding them, such as -at-, denoting the feminine in the example ṭalḥatun ‘an accacia tree’, the sufffijix -āt-, denoting the feminine plural in munṭaliqātun ‘going away’, and the sufffijix -iyy denoting relation, as in tamīmiyyun ‘belonging to the tribe of Tamīm’ are determined as zawāʾid.76 2. In Sībawayhi’s view, the dual endings -āni/ayni, as in ar-rajulāni and ar-rajulayni, and the plural sufffijixes -ūna/-īna, as in al-muslimūna/ al-muslimīna, correspond to the combination of the case-ending and the tanwīn in nouns like Zaydun; the -ā/-ay included in the dual endings and the -ū/-ī in the plural sufffijixes correspond to a case-ending, while the fijinal -ni and -na correspond to the tanwīn.77 Hence these endings are classifijied as zawāʾid. 3. Sībawayhi believes that in the sufffijixes of the imperfect verb, like the dual endings -āni/-ā in yaf ʿalāni and yaf ʿalā, the plural endings -ūna/-ū in yaf ʿalūna and yaf ʿalū, and the feminine ending -īna in taf ʿalīna, the fijinal -ni and -na correspond to the mood marker -u denoting the indicative (see §4.2.2 above). Since in Sībawayhi’s view the mood markers -u and -a are zawāʾid which form the fijinal limit of the imperfect verb, the endings -ni and -na corresponding to them also form the fijinal limit of this category of verbs, and hence are classifijied as zawāʾid. It is clear

75

 See the title of chapter 2 (Kitāb: 1.1.9).  See Kitāb: 2.340.12–13; Mubarrad (Muqtaḍab: 1.57.10–1.60.3–4). 77  See Kitāb: 1.3.12–1.4.1. 76

the concept of kalima in old arabic grammar

29

that the long vowels preceding these endings in -īna, -āni and -ūna occur before the fijinal limit of the above verb forms, and hence are also classifijied as zawāʾid. It should be noted that Sībawayhi conceives of the long vowels in the above endings as pronouns, irrespective of their classifijication as zawāʾid.78 This classfijication attests that the morphological consideration in the classifijication of morphemes is decisive: although pronouns are usually classifijied as nouns belonging to the category of kalim, in the above endings they are conceived of as zawāʾid, since they precede the mood markers, which are conceived of as the fijinal limit of the imperfect verb. 4. For the considerations that led Sībawayhi to classify as a zāʾida the third person feminine plural pronoun -na in the verbal forms faʿalna ‘they did’ and yaf ʿalna ‘They will do’ see Levin (1986: 441–442, § 6.3.5). This pronoun is also classifijied as a ḥarf zaʾid. 4.3 Kalima, zāʾida and Morpheme The Arab grammarians’ concept of the individually smallest meaningful elements in the Arabic language basically accords with the modern concept of morpheme, although they divide such units into two groups: kalim and zawāʾid. This division of morphemes derives from one main diffference between the criterion for the determination of a morpheme, and the criteria for the determination of a kalima or a zāʾida: the criterion for the determination of a given meaningful form as a morpheme is the fact that it cannot be divided into smaller meaningful forms, i.e., the fact that this meaningful form is one of the smallest individually meaningful elements in the utterances of the language. This criterion, which is a semantic one, is also the main criterion for the determination of a given form as a kalima or a zāʾida, as is evidenced by the classifijication of the defijinite article aland the pronoun sufffijix -tu in faʿaltu as kalim, and the classifijication of the imperfect prefijixes ʾa-, ta-, ya- and na-, and the case-endings of the noun as zawāʾid.79 However, it should be emphasized that in contrast to the determination of morphemes, the determination of kalim and zawāʾid

78

 For these long vowels as pronouns see Kitāb: 1.4.3–21; 1.330.12–14.  This respect is also evidenced by the inventory of kalim and zawāʾid in § 4.2.1; 4.2.2 above. 79

30

aryeh levin

is restricted by certain morphological and etymological considerations which led Sībawayhi to divide the morphemes into kalim and zawāʾid. The following example illustrates how morphological and etymological considerations interfere with the criteria for the determination of a kalima or a zāʾida: The pronoun sufffijix -tu in faʿaltu ‘I did’ which denotes the fijirst person singular is classifijied as a kalima, while the prefijix ʾa- in ʾaf ʿalu ‘I shall do’ which denotes the same, is conceived of as a zāʾida. Semantically, there is no diffference between these two morphemes, as both of them denote the same person. The fact that the sufffijix -tu is classifijied as a kalima, while the prefijix ʾa- is classifijied as a zāʾida, derives from the view that the pronominal sufffijixes of the perfect, which are composed of sounds which do not belong to ḥurūf az-zawāʾid, do not form an integral part of the perfect patterns, in contrast to the imperfect prefijixes, which are conceived of as an integral part of the imperfect patterns.

5. Conclusions It is inferred from the inventory of the morphemes discussed in the Kitāb (see above §4.2.1; 4.2.2), and from other data gathered from this text, that in Sībawayhi’s terminology, the form kalima denotes two types of linguistic units: 1. A kalima occurring on a certain morphological pattern. The characteristics of this type of kalima are: 1. The kalim belonging to this type are nouns and verbs, like farasun ‘a mare’, yaf ʿalu ‘he will do’, ḏahaba ‘he went away’ and kul ‘eat!’. 2. These kalim may contain one morpheme, like ḏahaba and kul, or more than one morpheme, like farasun and yaf ʿalu: the morphemes contained in farasun are faras-, -u and -n, and the form yaf ʿalu contains the morphemes ya-, -f ʿal- and -u. 3. The sounds composing these kalim belong to the categories of ḥurūf ʾuṣūl and zawāʾid. For example: in the form farasun the radicals f-r-s are the ʾuṣūl, and all the other sounds, which are the short vowels a-a-u and the tanwīn, are zawāʾid. 4. The grammatical meaningful elements contained in these kalim are expressed by zawāʾid which form an integral part of their patterns. For example: in the imperfect verb yaf ʿalūna, the third person is expressed by the zāʾida ya-, the plural pronoun by the -ū- , and the fijinal -na is an ʾiʿrāb marker.

the concept of kalima in old arabic grammar

31

5. Some of these kalim contain sufffijixes classifijied as zawāʾid because of considerations deriving from the theory of ʾiʿrāb, and the concept that the tanwīn is the fijinal limit of the kalima. 2. The second type of kalima consists of one morpheme. The characteristics of these kalim are: 1. The linguistic units belonging to this type of kalima are particles, indeclinable nouns and pronominal sufffijixes added to nouns, verbs and particles. For example: the defijinite article al-, the future marker sa-, the indeclinable noun man ‘who’, and the pronoun sufffijix -tu contained in ḏahabtu ‘I went’. 2. The kalim belonging to this category do not occur on any morphological pattern. 3. The sounds composing these kalim do not belong to any category of sounds: they are not classifijied either as hurūf ʾuṣūl or as zawāʾid. It seems that in the process of the classifijication of morphemes by Sībawayhi and his teachers, all the morphemes conceived of as an integral part of the pattern were classifijied as zawāʾid. Morphemes which could not be integrated in a certain morphological pattern, were classifijied as kalim. Apart from this, some sufffijixes which do not form an integral part of the pattern were classifijied as zawāʾid, because of considerations deriving from the theory of ʾiʿrāb, and the concept that the tanwīn or the case- or mood-markers form the fijinal limit of the kalima. I have pointed out that in Sībawayhi’s view the essence of the sounds composing the sufffijixes and prefijixes classifijied as kalim difffers from that of the sounds composing the nominal and verbal bases to which they are added: while the nominal and verbal bases are composed of ḥurūf ʾuṣūl and zawāʾid, their sufffijixes and prefijixes have no radical consonants and no zawāʾid. Since the sounds composing these morphemes are not zawāʾid, it is impossible to conceive of them as sounds which form an integral part of the pattern of the nominal and verbal bases to which they are added. Hence they are perceived as linguistic units which are morphologically separate from the units composing their bases. Since these separate forms are meaningful elements which cannot be an integral part of the nouns and verbs to which they are attached, they are classifijied as kalim. The sufffijixes and prefijixes thus classifijied as kalim can be detached from their bases, since they are not an integral part of the pattern of these bases. The detachment does not semantically and morphologically afffect the base, although sometimes it creates some slight morpho-phonemic changes. The combination of the base+sufffijix or prefijix+base in this case is

32

aryeh levin

a combination of two linguistic units, classifijied as two kalimas. The status of each of these linguistic units is that of a separate kalima. In contrast, the detachment of a zāʾida from a kalima occurring on a certain morphological pattern, distorts the kalima, since the zāʾida is an integral part of its pattern, and hence an indispensable part of the kalima. For example: if the zāʾida ya- is dropped from the imperfect verb yaf ʿalu ‘he will do’, this verb cannot exist, since the theoretical form *f ʿalu is meaningless. But if the kalima sa- is dropped from the form sa-yaf ʿalu, the result is the form yaf ʿalu, which is a meaningful form found in the Arabic language.

References 1. Primary Sources Ibn Jinnī. Lumaʿ. Kitāb al-Lumaʿ Fi-n-Naḥw (Manuel de Grammaire Arabe), ed. Hadi M. Kechrida. Uppsala, 1976. ——. Munṣif. Al-Munṣif, Šarḥ al-Imām ʾAbī l-Fatḥ ʿUṯmān Ibn Jinnī li-Kitāb at-Taṣrīf lilImām ʾAbī ʿUṯmān al-Māzinī, ed. ʾIbrahīm Muṣṭafā and Abd Allah ʾAmīn, 3 vols. [n.p.], 1373h=1954. ——. Sirr. Sirr ṣināʿat al-ʾiʿrāb, ed. Ḥasan Hindāwī, 2 vols. Dimašq, 1405h=1985. Ibn Sarrāj. ʾUsūl. Kitāb al-ʾUṣūl fī n-naḥw, ed. ʿAbd al-Ḥusayn al-Fatlī, 3 vols. Bayrūt, 1408h=19883. Ibn Yaʿīš. ŠMufaṣṣal Cairo. Šarḥ al-Mufaṣṣal, 10 parts. Al-Qāhira, [n.d.]. ——. ŠMufaṣṣal Jahn. Commentar zu Zamachśarîʾs Mufaṣṣal, ed. G. Jahn, 2 vols. Leipzig, 1882–1888. Mubarrad. Muqtaḍab. Al-Muqtaḍab, ed. ʿAbd al-Xāliq ʿUḍayma, 4 vols. Al-Qāhira, 1385– 1388h. Sībawayhi. Kitāb. Le livre de Sîbawaihi. Traité de grammaire arabe, texte arabe publié par Hartwig Derenbourg, 2 vols. Paris, 1881–1889. Zamaxšarī. Mufaṣṣal. Al-Mufaṣṣal, ed. J.P. Broch. Christianiae, 18792.

2. Secondary Literature Hockett, Charles Francis. 1960. A Course in Modern Linguistics. New York. Levin, Aryeh. 1978. “The ʿImāla of ʾAlif Fāʿil in Old Arabic”. Israel Oriental Studies 8, 174– 203. ——. 1986. “The Medieval Arabic Term kalima and the Modern Linguistic Term Morpheme: Similarities and Diffferences”. Studies in Islamic History and Civilization in Honour of Professor David Ayalon, ed. M. Sharon. Jerusalem: Cana & Leiden: Brill, 423-446. ——. 2000. “The Meaning of Ḥarf Ǧāʾa li-Maʿnan in Sībawayhi’s al-Kitāb”. Jerusalem Studies in Arabic and Islam 24, 22–45. ——. 2007. “Kalima”. Encyclopaedia of Arabic Language and Linguistics. Leiden: Brill, 2.545–548. Nöldeke, Th. 1904. Beiträge zur semitischen Sprachwissenschaft. Strassburg. Wright, W. 1951. A Grammar of the Arabic Language, 2 vols. Cambridge.

WHAT IS A KALIMA? ʾASTARĀBĀḎĪ’S ANSWER* Pierre Larcher

Introduction The choice of the present subject undoubtedly derives from the questions aroused in the occasion of this fijirst colloquium of Arabic linguistics. However, the question is not new to me. In fact, it dates back to my days as a PhD student, when I acquainted myself with the writings of a currently well known, at that time less if not ill-known, Arab grammarian. I am obviously referring to Raḍī ad-dīn al-ʾAstarābāḏī the author of the Šarḥ al-Kāfijiya and of the Šarḥ aš-Šāfijiya, two great commentaries on two compendiums written by another author, Ibn Ḥājib, namely al-Kāfijiya fī n-naḥw and aš-Šāfijiya fī ṣ-ṣarf. The fijirst deals with syntax and the second with morphology. The fijirst constitutes the core of the present study. I will quote it from the Istanbul edition of 1310H, for a long time the only one available (there exist at least three recent editions today). I will not translate kalima, but limit myself to reminding that this word is both the singulative (ism al-waḥda) of the collective kalim and the singular of the plural kalimāt.

1. The kalima: Neither a Word, Nor a Morpheme At the beginning of the Šarḥ al-Kāfijiya, Raḍī d-dīn al-ʾAstarābāḏī begins with the defijinition of the kalām given by Ibn Ḥājib in the Kāfijiya: An utterance is everything that contains two kalima as a result of the predicative connection. This only occurs in the case of two nouns or in the case of a verb and a noun.1

which he comments in these terms:

* I kindly thank Professor Giuliano Lancioni for his invitation, which obliged me to resume an old work. 1  ʾAstarābāḏī (ŠKāfijiya: 1.7): al-kalām mā taḍammana kalimatayn bil-ʾisnād wa-lā yataʾattā ḏālika ʾillā fī smayni ʾaw fī fijiʿl wa-ism.

34

pierre larcher By “which contains two kalima”, we mean that the utterance is composed by these two or that these constitute its two parts (or elements).2

by adding (I will only quote what is relevant to the scope of this paper): [. . .] and the two parts of the utterance can be either explicit as in Zaydun qāʾimun and qāma Zaydun, or implicit, [. . .] or one of the two can be implicit but not the other. The implicit one can be either the verb [. . .] or the subject, as in Zaydun qāma.3

It is clear that for Raḍī d-dīn al-ʾAstarābāḏī both: (1) Zaydun qāʾimun Zayd

standing

‘Zayd is standing’ and: (2) qāma stood

Zaydun Zayd

‘Zayd stood’ are utterances (kalām), each of them constituted of two kalima, i.e. two nouns in (1), a verb and a noun in (2). However, at the beginning of the third and last part of his work, devoted to the particle (ḥarf: this constitutes the third of the three kinds of kalima recognised by the Arabic grammatical tradition, after the noun and the verb) he gives as an example: (3) = (1) Zaydun qāʾimun and: (4) qad qāma Zaydun

2  In ʾAstarābāḏī (ŠKāfijiya: 1.7): wa-naʿnī bi-taḍammunihi l-kalimatayn tarakubbahu minhumā ʾaw kawnahumā juzʾayhi. 3  Wa-juzʾā l-kalām yakūnāni malfūḏ̣ayn ka-Zaydun qāʾimun wa-qāma Zaydun wamuqadarrayn [. . .] ʾaw ʾaḥaduhumā muqaddar dūna l-ʾāxar wa-huwa ʾimmā fijiʿl [. . .] ʾaw fāʿil kamā fī Zaydun qāma.

what is a kalima?

35

indicating that “each of the two abovementioned utterances is composed of four kalimāt”,4 i.e. the tanwīn (-n) which is added to Zayd and qāʾim in (3), the particle qad which comes before the verb, and the tanwīn which is added to Zayd in (4). At this point, a fijirst conclusion can be drawn: a kalima is not a word. There would actually be two words in (1) and in (3), two in (2) but three in (4). A kalima is not even a morpheme. If by morpheme we mean a minimal meaningful unit, we would at least count six morphemes in (1) and (3), i.e.: (1’) = (3’) Zayd-u-n qāʾim-u-n and at least four in (2), i.e.: (2’) qāma Zayd-u-n and fijive in (4), i.e.: (4’) qad qāma Zayd-u-n where, besides the tanwīn, we would count as a morpheme the vowel -uwhich is the mark of the nominative case. As for Western scholars of Arabic, most of them would see in qāʾimun and qāma the association of two morphemes: a lexical one (to which correspond the “consonantal root” √qwm and the meaning ‘to stand up’), and a grammatical one (to which correspond the patterns fāʿil and faʿala and the meanings ‘active participle’ and ‘3rd person m.s. of the perfect’). This would bring the total number of morphemes to seven in (1) and (3), fijive in (2) and six in (4) . . . If a kalima is neither a word nor a morpheme, what is it then? The answer to this question is uncertain, because we are faced with two possibilities: 1. either Raḍī d-dīn al-ʾAstarābāḏī contradicts himself 2. or he does not contradict himself, in which case we have to fijind an interpretation of kalima compatible with the double counting of kalimāt in (1) that the author does, respectively two and four.

4

 2.319: kull wāḥid min al-kalāmayn al-maḏkūrayn murakkab min ʾarbaʿ kalimāt.

36

pierre larcher 2. What If, Despite Everything, the Kalima Was a Morpheme? 2.1. Morphemes Belonging to the Non-Concatenative Morphology

If a kalima is not a word, it might be a morpheme. This comes directly from the defijinition of kalima given by Ibn Ḥājib as “an expression instituted for a single meaning”.5 One expression, one meaning: what we have here seems to be a kalima as a minimal meaningful unit, i.e. what modern Western—mainly American—linguistics calls a morpheme. Unfortunately, ʾAstarābāḏī’s commentary of the terms “instituted for a single meaning” does not confijirm this interpretation: He means by that the meaning, of which one part is not signifijied by any part of the expression, whether this meaning has a part (like the meaning of ḍaraba which means the maṣdar and the tense) or whether it does not, like the meaning of ḍarb.6

In other words, an expression can be single, as ḍaraba and ḍarb, each counting for one and only one kalima, but its meaning can be either single, as in ḍarb, or multiple, as in ḍaraba, which adds to the meaning of ḍarb that of the past tense . . . Nevertheless, at this point, one could perfectly maintain that kalima is a morpheme, in the sense of the American distributionalist linguistics, which “conceives the meaningful unit as a segment of the speech chain, vehicle of a meaning which is exterior to it” (Ducrot & Schaefffer 1995: 360). In such a conception, it is possible that “an unanalysable phonic element can bear many clearly distinct meanings”. This very fact leads to distinguish between morpheme as an abstract unit, and morph as a concrete segment and consequently, to introduce the concept of allomorph, when to the same abstract morpheme correspond several concrete morphs, and of portmanteau morph, when to the one and same morph correspond several morphemes. This would be the case of ḍaraba, counted as one kalima, even if we recognise two meanings in it. Is ḍaraba really an “unanalysable phonic element”? ʾAstarābāḏī goes back to the verb in the past tense. He writes:

 In ʾAstarābāḏī (ŠKāfijiya: 1.2): lafḏ̣ mawḍūʿ li-maʿnā mufrad.  ʾAstarābāḏī (ŠKāfijiya: 1.3): yaʿnī bihi al-maʿnā allaḏī lā yadullu juzʾ lafḏ̣ihi ʿalā juzʾihi sawāʾ kāna li-ḏālika al-maʿnā juzʾ naḥw maʿnā ḍaraba ad-dāll ʿalā l-maṣdar waz-zaman ʾaw lā juzʾ lahu ka-maʿnā ḍarb. 5

6

what is a kalima?

37

As for the verb in the past tense, like ḍaraba, this is an interesting point: in fact, it is undoubtedly a kalima, though the process is signifijied by its ordered consonants and the assertion of this process in the past is signifijied by its form, which afffects its consonants. The form is a part of the expression, since it represents the number of the consonants, including the whole set of the short vowels and the lack of short vowels; this set has been the object of a specifijic institution. Vowels are a part of what is uttered. Thus the verb in the past tense is a kalima made up of two parts, each indicating a part of its meaning.7

It is then clear that for ʾAstarābāḏī, in ḍaraba there are not only two meanings, but to each of these meanings corresponds a part of the expression: the “ordered consonants”, i.e. the consonantal root √ḍrb for ḍarb and the whole represented by the number of vocalised or non-vocalised consonants, i.e. the pattern (wazn) R1aR2aR3 for the past tense.8 ʾAstarābāḏī adds that what is true for ḍaraba is also true for the plural ʾusud ‘some lions’ as opposed to ʾasad ‘a lion’ and for the diminutive [ʾusayd ‘a lion cub’],9 for the plurals masājid ‘some mosques’ and rijāl ‘some men’, for ḍārib ‘striking’, maḍrūb ‘striken’ and miḍrab ‘a striking tool’: because the diminutive, the plural, the agent, the patient and the tool in the abovementioned examples are signifijied by the short vowels occurring with the added consonant and/or the long vowel.10

He concludes: It is not correct to claim here that the incidental form is a kalima which, as a result of the combination, has become as a part of another kalima [. . .]. These kalim consitute therefore a relevant objection, unless one limits the explanation of “complex expression” and says “it is something a part of the expression of which indicates a part of its meaning”. In such an expression,

7  ʾAstarābāḏī (ŠKāfijiya: 1.5–6): ʾammā l-fijiʿl al-māḍī naḥw ḍaraba fa-fīhi naḏ̣ar li-ʾannahu kalima bi-lā xilāf maʿa ʾanna al-ḥadaṯ madlūl ḥurūfijihi al-murattaba wal-ʾixbār ʿan ḥuṣūl ḏālika l-ḥadaṯ fī z-zaman al-māḍī madlūl waznihi aṭ-ṭāriʾ ʿalā ḥurūfijihi wal-wazn juzʾ al-lafḏ̣ ʾiḏ huwa ʿibāra ʿan ʿadad al-ḥurūf maʿa majmūʿ as-sakanāt al-mawḍūʿa waḍʿan muʿayyanan wal-ḥarakāt mimmā yutalafffaḏ̣ bihi fa-huwa ʾiḏan kalima murakkaba min juzʾayn yadullu kullu wāḥid minhumā ʿalā juzʾ maʿanāhu. 8  At fijirst sight, ʾAstarābāḏī’s analysis is not so diffferent from that of the Western scholars of Arabic. Actually, it deviates from it on a crucial point: the lexical meaning ‘to strike’ is not directly related to the consonantal root √ḍrb. It is only indirectly attached to it since it represents, in the derived word, (here the perfect ḍaraba), the base it derives from, i.e. the maṣdar ḍarb. 9  ʾAstarābāḏī does not quote the word itself. 10  ʾAstarābāḏī (ŠKāfijiya: 1.6): li-ʾanna d-dāll ʿalā maʿnā at-taṣġīr wal-jamʿ wal-fāʿil walmafʿūl wal-ʾāla fī l-ʾamṯila al-maḏkūra al-ḥarakāt aṭ-ṭāriʾa maʿa al-ḥarf az-zāʾid.

38

pierre larcher one of these two parts immediately follows the other, whereas in the abovementioned kalima, the two parts are heard together.11

In other words, ʾAstarābāḏī recognises that a word like ḍaraba, just like other words of the same kind, is not only semantically, but also morphologically complex. According to modern Western linguistics, one can recognise in it not only two morphemes, but also two morphs; according to the Arabic grammatical tradition, one should—given the defijinition of the kalima—see in it two kalima and not one. However, the Arabic grammatical tradition has always seen in it only one kalima. In order to avoid the contradiction, ʾAstarābāḏī proposes to consider as complex only the expression a part of which immediately follows (mutaʿaqqib) the other. This confijirms that a kalima, though not being a word, a morpheme or a morph, is a segment of the utterance. The criterion of the consecution allows us to exclude from the defijinition of kalima the bound morphemes, belonging to what today is called non-concatenative morphology. It is already clear that this criterion does not allow to exclude the morphemes belonging to the concatenative morphology, i.e. those which ʾAstarābāḏī refers to in the passage above in brackets ([. . .]) that we quote here: as we claimed concerning the previous kalimāt and as it would be correct to claim concerning the inflectional vowels.12

2.2. The morphemes belonging to concatenative morphology Here is what the “abovementioned kalimāt” were: If one says: when you say muslimāni, muslimūna, baṣrī and all the imperfect verbs, a part of the expression of each of them indicates a part of their meaning; the wāw marks the plural, the ʾalif marks the dual, the yāʾ marks the nisba and the prefijixes of the imperfect mark at the same time the imperfect and the person. The same can be said for the tāʾ of feminisation in qāʾima, the tanwīn and the defijinite article, and the two ʾalif of feminisation. The expression of each of them [i.e. of these kalimāt], just like their meaning, is therefore necessarily complex. There are then two kalima, not one.13

11  Wa-lā yaṣiḥḥu ʾan naddaʿiya hāhunā ʾanna l-wazn aṭ-ṭāriʾ kalima ṣārat bit-tarkīb ka-juzʾ kalima [. . .] fal-iʿtirāḍ bi-hāḏihi l-kalim iʿtirāḍ wārid ʾillā ʾan nuqayyida tafsīr al-lafḏ̣ al-murakkab fa-naqūla huwa mā yadullu juzʾuhu ʿalā juzʾ maʿnāhu wāḥid al-juzʾayni mutaʿaqqib lil-ʾāxar wa-fī hāḏihi l-kalima al-maḏkūra al-juzʾāni masmūʿāni maʿan 12  Kamā ddaʿaynā fī l-kalim al-mutaqaddima wa-kamā yaṣiḥḥu ʾan yuddaʿā fī l-ḥarakāt al-ʾiʿrābiyya. 13  ʾAstarābāḏī (ŠKāfijiya: 1.5): ʾin qīla ʾinna fī qawlika muslimāni wa-muslimūna wa-baṣriyy wa-jamīʿ al-ʾafʿāl al-muḍāriʿa juzʾ lafḏ̣ kull wāḥid minhā yadullu ʿalā juzʾ maʿnāhu ʾiḏ al-wāw

what is a kalima?

39

All the expressions mentioned by ʾAstarābāḏī constitute a sequence of two segments. We shall refer here to his own examples and when he does not give any, we shall refer to those given by his commentator ʿAlī b. Muḥammad, known as as-Sayyid aš-Šarīf al-Jurjānī, in the “margin” (Ḥāšiya): 1. muslimāni ‘two Muslims’, analysable as muslim + ā(ni), where ā is the mark of the dual, but also a mark of declension, in this case the nominative, as opposed to muslimayni, accusative-genitive; 2. muslimūna ‘Muslims’, analysable as muslim + ū(na), where ū is the mark of the masculine plural, but also a mark of declension, in this case the nominative, as opposed to muslimīna, accusative-genitive ; 3. baṣriyy, analysable as baṣr (stem of al-Baṣra) + iyy, which marks the ‘relation’ (nisba), i.e. ‘from Basra, Basrian’; 4. the imperfect prefijixes: ʾAstarābāḏī does not give any example, but his commentator gives ʾaḍribu ‘I (will) strike’. It is clear that ʾaḍribu, taḍribu ‘you (will) strike’, yaḍribu ‘he (will) strike’ . . . can be analysed as prefijixes ʾa-, ta-, ya- added to a stem ḍrib14 and that these prefijixes, as mentioned by ʾAstarābāḏī, mean both the imperfect and the ‘state of the agent’ (ḥāl al-fāʿil). His commentator is more detailed, for he speaks on the one hand of ‘future or present’ (al-ʾistiqbāl ʾaw l-ḥāl), on the other of ‘speaker, addressee, masculine, for example’ (at-takallum wal-xiṭāb wat-taḏkīr maṯalan); 5. qāʾima ‘standing’ analysable as qāʾim + at-, where at- is a mark of the feminine; 6. the two tāʾ of “feminisation” are exemplifijied by his commentator (Šarīf Jurjānī Ḥāšiya in ŠKāfijiya:, 1.istidrakāt 5) through ḥublā ‘pregnant’ and ḥamrāʾ ‘red’. It is worthwhile noting that the segmentation of ḥublā into ḥubl + ā and of ḥamrāʾ into ḥamr + āʾ leaves a segment that coincides neither with a word, nor with a stem: the word is ḥabal ‘pregnancy’ in the fijirst case and ḥumra ‘redness’ in the second. We could see here some intermediary (in synchrony) or transitional (in diachrony)

tadullu ʿalā l-jamʿiyya wal-ʾalif ʿalā t-taṯniya wal-yāʾ ʿalā n-nisba wa-ḥurūf al-muḍāraʿa ʿalā maʿnā fī l-muḍāriʿ wa-ʿalā ḥāl al-fāʿil ʾayḍan wa-kaḏā tāʾ at-taʾnīṯ fī qāʾima wat-tanwīn wa-lām at-taʿrīf wa-ʾalifā at-taʾnīṯ fa-yajibu ʾan yakūna lafḏ̣ kull wāḥid minhā murakkaban wa-kaḏā l-maʿnā fa-lā yakūnu kalima bal kalimatayn. 14  Since the vowels of both the prefijix and the stem can be subject to apophony, for example the passive yuḍrab ‘he is /will be stricken’, the prefijix is exactly ʾ(V), t(V), y(V) and the stem is ḍrVb . . .

40

pierre larcher forms between concatenative morphology and non-concatenative morphology.

We have obviously noted that ʾAstarābāḏī includes in this list the defijinite article and the tanwīn. The defijinite article is prefijixed to the noun, e.g. ar-rajul ‘the man’, and the tanwīn is sufffijixed to the noun, which bears the declension vowel: Zaydu-n, qāʾimu-n etc. ʾAstarābāḏī concludes: the answer is that all I have mentioned constitutes two kalima which have become, as a result of the strong amalgam, like one kalima: the compound expression has been given the desinential inflection of the kalima, and this because the afffijixes are not independent; the same can be said of the inflectional vowels.15

In his answer, ʾAstarābāḏī, as a good logician, draws the conclusion necessarily resulting from Ibn Ḥājib’s defijinition of the kalima as a minimal meaningful unit. Such a defijinition leads to recognise as a kalima all morphemes belonging to the concatenative morphology, i.e. not only those he generically labels as ḥurūf muttaṣila, but also the declension vowels. ʾAstarābāḏī’s originality clearly appears when comparing his text to Šarīf Jurjānī’s critical commentary in margin of the ŠKāfijiya: It is clear that the tanwīn and the article belong to the particles and that both have been counted amongst them: each of the two is therefore immediately one kalima. Thus, an expression such as ar-rajul is made of two—not one—kalima.16

Further on he adds: as for the ʾalif of dualisation, the wāw of the plural, the yāʾ of the nisba, the tāʾ of vocalised feminisation and the two ʾalif of feminisation, we have already said that they belong to the formatives, which are added to the kalim.17

Šarīf Jurjānī’s commentary shows that ʾAstarābāḏī groups together under the name of ḥurūf muttaṣila (translated here as ‘afffijixes’) what Šarīf Jurjānī himself distinguishes as ḥurūf al-maʿānī on one side and ḥurūf al-mabānī

15  Fal-jawāb ʾanna jamīʿ mā ḏakartu kalimatāni ṣāratā min šiddat al-ʾimtizāj kakalima wāḥida fa-ʾuʿriba al-murakkab ʾiʿrāb al-kalima wa-ḏālika li-ʿadam istiqlāl al-ḥurūf al-muttaṣila fī l-kalim al-maḏkūra wa-ka-ḏālika al-ḥarakāt al-ʾiʿrābiyya. 16  Lā xafāʾ fī ʾanna t-tanwīn wa-lām at-taʿrīf min ḥurūf al-maʿānī wa-qad ʿaddūhumā fīhā fa-kull wāḥida minhumā kalima ʿalā ḥiyālihā [sic: read ʿalā ḥālihā] fa-naḥw ar-rajul kalimatāni lā kalima wāḥida. 17  ʾAmmā ʾalif at-taṯniya wa-wāw al-jamʿ wa-yāʾ an-nisba wa-tāʾ at-taʾnīṯ al-mutaḥarrika fa-qad qīla ʾinnahā min ḥurūf al-mabānī zīdat fī l-kalim.

what is a kalima?

41

on the other. The fijirst term is the plural of ḥarf maʿnā (litt. ‘semantic articulation’) which, in the Arabic grammatical terminology, designates the particle, as opposed to ḥarf lafḏ̣, which designates the ‘phonic articulation’ (consonants and long vowels). The second term is the plural of ḥarf mabnā, which designates the elements taking part in word formation (for this reason we have translated it as ‘formative’). If ʾAstarābāḏī can operate such a grouping, it is clearly because the ḥurūf al-mabānī are as meaningful as the ḥurūf al-maʿānī. Therefore, ʾAstarābāḏī introduces a double conceptual and terminological innovation. Conceptually, he fijinds the idea of morpheme in its most general sense: this concept groups together the morphemes belonging to non-concatenative morphology (‘root’ and ‘pattern’) as well as those belonging to concatenative morphology. In this last case, the concept groups together the morphemes belonging to derivation as well as inflection, and, in the case of inflection, those belonging to purely morphological inflection as well as those belonging to desinential inflection (at the intersection between morphology and syntax). Terminologically, he has a term for “afffijix”, even if this term is less general than his concept of morpheme: in fact, it excludes the inflectional vowels, though they are recognised as morphemes afffijixed to the word.18 This is explained by a terminological reason, i.e. the fact that in the Arabic grammatical terminology ḥarf and ḥaraka are in opposition. This blurring of the boundary between ḥurūf al-maʿānī and ḥurūf al-mabānī is confijirmed by the fact that ʾAstarābāḏī transfers the defijinition of the ḥurūf al-maʿānī as ‘expression signifying [not in itself but] in another’ (lafḏ̣ dāll ʿalā maʿnā fī ġayrihi) to the ḥurūf al-mabānī. As he writes: “the ḥarf creates its meaning in an expression other than itself, which either precedes it, as in baṣriyy, or follows it, as in ar-rajul”.19 Nevertheless, we should not attribute to ʾAstarābāḏī what he does not say. He simply reacts as a logician to Ibn Ḥājib’s defijinition of the kalima. This defijinition, of an intensional type, has undesirable efffects in extension: literally, it would end up considering all morphemes as kalima. However,

18  Cf. Šarīf Jurjānī’s commentary to the sentence “the same can be said of inflectional vowels” (ʾAstarābāḏī ŠKāfijiya: 1.istidrakāt on p. 5.6–7): “he means that they are also kalimāt in themselves, belonging to the category of ḥurūf  ”(yaʿnī ʾannahā ʾayḍan kalimāt bi-raʾsihā min qabīl al-ḥurūf ): in this context, Šarīf Jurjānī is obviously referring to ʾAstarābāḏī’s ḥurūf mutaṣṣila, all the more so as he adds: “but as a result of the strong amalgam between inflectional vowel and something else, he counted the whole as one kalima” (lākin li-šiddat imtizāj al-ḥaraka bi-ġayrihā ʿadda l-majmūʿ kalima wāḥida). 19  ʾAstarābāḏī (ŠKāfijiya: 1.9): fal-ḥarf mūjid li-maʿnāhu fī lafḏ̣ ġayrihi ʾimmā muqaddam ʿalayhi kamā fī naḥw baṣriyy ʾaw muʾaxxar ʿanhu kamā fī ar-rajul.

42

pierre larcher

not every morpheme is a kalima; moreover, some sets of two morphemes are considered as one kalima . . . In order to avoid these undesirable efffects, there is no other solution in logic, than to restrict the defijinition (qayd, pl. quyūd). A fijirst restriction, the non-consecution of two elements of a set, allows him to exclude from the category of kalima the bound morphemes belonging to the non-concatenative morphology. A second restriction, i.e. the incidence of desinential inflection, allows him then to exclude from the category of kalima the bound morphemes belonging to the concatenative morphology. However, in his commentary to the sentence “the compound expression has been given the desinential inflection of the kalima”, Šarīf Jurjānī distinguishes between four cases: this is evident in expressions such as baṣriyy, qāʾima, ḥublā and ḥamrāʾ, because the desinential inflection, at the end of the compound expression, afffects a part which originally has no right to it. As for the noun afffected by the tanwīn, this tanwīn occurs after the inflectional vowel afffecting the fijirst part of the noun. In the case of the dual and the plural, if we take the marks of each of them as the desinential inflection itself, since they replace the inflectional vowels, then there is no desinential inflection afffecting the compound expression, but only the fijirst part of it. The desinential inflection, in expressions such as ar-rajulu and ʾaḍribu, actually afffects the second part of the expression, which has right to it, not the set constituted by the second and the fijirst part: think about it!20

Šarīf Jurjānī’s commentary has a double merit. It shows the double function of the inflectional vowels in ʾAstarābāḏī: if they are not kalimāt, but parts of kalimāt, they are at the same time boundaries of kalimāt. Moreover, it stresses the inadequacy of ʾAstarābāḏī’s criterion: if it allows considering the tanwīn as a kalima, it no longer allows considering the defijinite article as a kalima! This pernicious efffect bears also a paradox: the Arab grammatical tradition has always considered the article as a kalima. On the contrary, as Levin (1986; 2007) reminds us this is not the case for the tanwīn: a grammarian such as Sībawayhi (d. 177/793?) does not consider

20  ʾAstarābāḏī (ŠKāfijiya, 1.istadrākāt 5): haḏā fī naḥw baṣriyy wa-qāʾima wa-ḥublā wa-ḥamrāʾ ḏ̣āhir li-ʾanna l-ʾiʿrāb fī ʾāxir al-murakkab ʿalā juzʾ lā yastaḥiqquhu ʾaṣlan wa-ʾammā l-munawwan fat-tanwīn fīhi baʿd ḥarakat al-ʾiʿrāb ʿalā l-juzʾ al-ʾawwal wa-fī l-muṯannā wal-majmūʿ ʾin juʿila l-ʿalāma nafs al-ʾiʿrāb qāʾima maqām al-ḥarakāt fa-lā ʾiʿrāb lil-murakkab bal lil-juzʾ al-ʾawwal wal-ʾiʿrāb fī naḥw ar-rajulu wa-ʾaḍribu ʾinnamā huwa liljuzʾ aṯ-ṯānī allaḏī yastaḥiqquhu lā lil-majmūʿ al-murakkab minhu wa-min al-juzʾ al-ʾawwal fa-taʾammal.

what is a kalima?

43

it as a kalima, but as a part of kalima, to be included among the zawāʾid ‘added elements’.21 So we are now back to the starting point. In the sequence ar-rajulu as well as in the sequence Zaydun, ʾAstarābāḏī counts the same number of morphemes, i.e. three: al + rajul + u; Zayd + u + n. In the two sequences, he alternatively counts one kalima (al-rajulu; Zaydun) or two (al + rajul; Zayd + n), where the inflectional vowel is never counted as a kalima, but only as a part of a kalima: al + rajul-u; Zayd-u + n. This confijirms the negative response that we have given to the initial question: kalima is neither a word, nor a morpheme. In both cases, there is only one word, in which case the Arab grammarians do not use the term kalima but lafḏ̣a, as Šarīf Jurjānī reminds us in margin of the ŠKāfijiya: “but, because of the amalgam between the two [i.e. the article and the noun], we apply to them the name of lafḏ̣a”.22 In any case, this allows us to provide a positive answer to the same question.

3. The Kalima, a Constituent of the Utterance There is only one possible interpretation of the kalima compatible with this double counting. It is not a semantic one, even if the kalima is a meaningful unit. It is not a morphological one, even if the kalima is an “expression”, i.e. a concrete segment of the utterance, with a signifijicant exception, that of the “hidden pronoun”. It is then a syntactic one: a kalima is a “part of the constructed discourse”, i.e. a constituent of the utterance, which in its turn can be divided into constituents. The example ar-rajulu, compared to rajulun ‘a man’, shows the principle of segmentation of the utterance into constituents and of its constituents into constituents: this cannot but be distribution.23 If ar-rajulu

21

 See also Levin (this volume)  Wa-lākin li-šiddat al-ʾimtizāj baynahumā yuṭlaq ʿalayhimā al-lafḏ̣a. Cf. ʾAstarābāḏī himself, (ŠKāfijiya: 1.5.3–4): “an expression such as qālā [‘they both said’] and qālū [‘they said’] is, just like arṭā [a kind of tree] and burquʿ [‘veil, burkah’], one and only one word. The same applies to all that is uttered in one and only one time, though each of the fijirst two is made of two kalima, as opposed to the last two” (miṯl qawlika qālā wa-qālū ka-arṭā wa-burquʿ lafḏ̣a wāḥida wa-kaḏā kull mā yutalafffaḏ̣ bihi marra wāḥida maʿa anna kull wāḥid min al-awwalayn kalimatān bi-xilāf aṯ-ṯāniyayn). 23  On the syntactic and, more specifijically, distributional basis of word classifijication, see Owens (1989). 22

44

pierre larcher

and rajulun count each as one kalima, it is because they have the same distribution. Both can be the subject of a verbal sentence such as: (5) qāma r-rajulu stood

art-man.n

‘the man stood’ (6) qāma rajulu-n man.n-tanwīn

‘a man stood’ On this point ʾAstarābāḏī does not innovate: I refer the reader to Zamaxšarī’s Mufaṣṣal in which he defijines the kalām as “what is composed of two kalima, one of which is predicated of the other” adding that “this occurs only in the case of two nouns [. . .] or of a verb and a noun”.24 Yet, he gives as an example of the fijirst case Zaydun ʾaxūka ‘Zayd is your brother’ and Bišrun ṣāḥibuka ‘Bišr is your companion’. Zamaxšarī counts indeed two kalimāt, which are nouns, in this utterance, but the second one is divided in its turn into two nouns, a substantive and a sufffijixed pronoun. Conclusion: ism is used here not as a noun in the sense of word class, but in the sense of a noun phrase. If ar-rajulu and rajulun count each as two kalimāt, it is because the defijinite article and the tanwīn are in complementary distribution. Because of its mutual exclusion with the article, we can see in the tanwīn one of the determiners of the noun. However, there is some uncertainty amongst grammarians, on the meaning of this determiner, whereas there is none on the meaning of the article (“particle of defijinition”). According to Zamaxšarī (Mufaṣṣal: 328), the tanwīn of rajulun is the same as that of Zaydun. Since Zayd is a proper noun, this cannot be a mark of indefijinition: it is only a mark of “declinability” (makāna or tamakkun). According to ʾAstarābāḏī (ŠKāfijiya: 1.13), on the contrary, “the tanwīn, in rajulun, is a mark of indefijinition, too” (at-tanwīn fī rajulun yufīdu t-tankīr ʾayḍan). This uncertainty is easily explainable if we think that the tanwīn can be pronounced only if the inflectional vowel is pronounced, too. Yet, in the orthoepy of Classical Arabic, there is a case in which the tanwīn is not pronounced: the pause. The noun determination system changes whether the distribution is funded on the noun’s inflected (ar-rajulu vs rajulun) or

24  Zamaxšarī (Mufaṣṣal 6): al-kalām huwa l-murakkab min kalimatayn ʾusnidat ʾiḥdāhumā ʾilā l-ʾuxrā, and wa-ḏāka lā yataʾattā ʾillā fī smayn [. . .] ʾaw fī fijiʿl wa-ism.

what is a kalima?

45

pausal (ar-rajul vs rajul) forms: in the fijirst case, we can oppose a mark of defijinition (al-) to a mark of indefijinition (-n); in the second, we can oppose only a positive mark of defijinition to a negative mark of indefijinition (al- vs -∅). Actually, since the inflected and pausal forms of the noun coexist, the tanwīn of rajulun can hardly be anything else than a purely redundant mark of indefijinition. For this reason, the status of the tanwīn hesitates between kalima and part of kalima. I am of course well aware that what has been said has to do with a theory of the sentence that, from the 4th/10th century on, has been strongly influenced by the Greek logic: at that time, the term ʾisnād acquired the meaning of ‘predication’. Each utterance (kalām) is therefore defijined as a sentence ( jumla), i.e. a set (literal meaning of the term) of two elements ( juzʾ) in a predicative connection. One can easily imagine that at times some recalcitrant structures are introduced by force into the model . . . In any case, the number and the hierarchy of the kalimāt will be further justifijied with reference to this model: the noun is what can be either subject or predicate in a sentence; the verb is what can be predicate but not subject in a sentence; the particle is what can be neither subject nor predicate in a sentence.25 Another influence of the logic can be seen in the distinction, among the kalimāt, between the noun and the verb as “expression signifying [each one] in itself ” (lafḏ̣ dāll ʿalā maʿnā fī nafsihi), as opposed to the particle which is “an expression signifying in an [expression] other than itself ”.26 This distinction immediately reminds us of the diffference done by ancient logic between categorematic and syncategorematic terms, as presented by Priscian (Institutiones grammaticae 2.4.15): partes orationis sunt duae nomen et verbum, quia haec solae et jam per se conjunctae plenam faciunt orationem; alias autem partes syncategoremata, hoc est consignifijicantia, appellant.

Lalande’s Vocabulaire technique et critique de la philosophie (1972) made this quotation in order to support his defijinition of the term categorematic as “when speaking of words, those which have a meaning in themselves”. Actually, this simply repeats in a semantic way what categorematic does in a syntactic way, i.e. which can constitute by itself the predicate (gr. katēgorēma) in a clause. 25  Cf. ʾAstarābāḏī (6, ll. 8–12; 8, ll. 31–32). For a critical commentary, cf. Guillaume (1988, though he does not quote ʾAstarābāḏī). Cf. also Owens (1989). 26  Cf. Guillaume (1988) for more references.

46

pierre larcher

The same applies here: the logical or semantic vocabulary must not hide the syntactic reality. How to understand that the sufffijixes of the perfect are categorised as kalima, but not the prefijixes of the imperfect, when we know that they are both recognised as bound morphemes playing the same semantic role (they mark fijirstly the person, secondly the gender and/or the number)? How to understand that some sufffijixes of the perfect and the imperfect, recognised as marks of gender and/or number, as classifijied as personal pronouns (ḍamāʾir) and so as kalima, but not the -t of ḍaraba?27 Finally, how to understand the existence of the implicit kalima which is represented by the “hidden pronoun” (aḍ-ḍamīr al-mustatir)? There is no answer to these questions unless one introduces in one way or another the concept of distribution. This is evident in the case of the perfect: 1st and 2nd person sufffijixes have the same distribution as the 3rd person lexically full subject: (7) ḍarab-tu ḍarab-ta

ḍarab(a) Zaydun

stroke-I

stroke-you

stroke Zayd

‘I stroke’

‘you stroke’

‘Zayd stroke’

On the contrary, the fact that in the 3rd person of the imperfect the prefijix and the lexically full subject occur together leads grammarians not to see in the 1st and 2nd persons some SV structures, but rather to rebuild V[S] structures by analogy with the perfect (considered as a basic structure), i.e.: (8) yaḍribu Zaydun he-strike Zayd

taḍribu [ʾanta]

ʾaḍribu [ʾanā]

you-strike [you]

I-strike [I]

‘Zayd strikes (will strike)’ ‘you strike/will strike’ ‘I strike/will strike’ Similarly, what makes necessary to see in Zaydun qāma, not a bound sentence subject/verb, but a segmented sentence topic/comment,28 is the fact

27  Called “unvowelled tāʾof feminisation” (tāʾ at-taʾnīṯ as-sākina) and counted by Zamaxšarī (Mufaṣṣal, 328) as a ḥarf “the occurrence of which aims at indicating immediately that the subject is feminine” (duxūluhā lil-ʾīḏān min ʾawwal al-amr ʾanna l-fāʿil muʾannaṯ). The ḥarf is undoubtedly a kalima, but it is not a major constituent. 28  We use here the particularly accurate terminology employed by the Swiss linguist Charles Bally (1865–1947: Bally 1965).

what is a kalima?

47

that we fijind in the same context for example, Zaydun qāma ʾabūhu. This allows again for an analogical reconstruction, i.e.: (9) Zaydun qāma [huwa] Zayd stood [he]

Zaydun qāma ʾabūhu Zayd stood father-he

‘as for Zayd, he stood’ ‘as for Zayd, his father stood’

Conclusion This explains why ʾAstarābāḏī’s writings have delighted me for over thirty years. Who else offfers us a double analysis of an expression such as ʾaḍribu, once as a word (lafḏ̣a) and once as a sentence ( jumla)? As a word, ʾaḍribu can be analysed as a sequence of three segments ʾa + ḍrib + u, to which correspond at least than fijive meanings: 1st person singular (marked by the prefijix [ʾ]) of the imperfect (marked by the prefijixation itself to the stem of the verb) indicative (marked by the vowel -u) of the active voice (marked by the vocalisation of the prefijix and of the stem). However, as a sentence, ʾaḍribu must be analysed as a sequence of two constituents (kalima), one explicit (malfūḏ̣ bihā), the verb, and one implicit (muqaddara), the “hidden pronoun”, subject of the verb.29 This double analysis allows us to understand not only what a kalima is not, but also what it is.

References 1. Primary Sources ʾAstarābāḏī. ŠKāfijiya. Šarḥ Kāfijiyat Ibn al-Ḥājib, 2 vols. Istanbul: Maṭbaʿat aš-šarika aṣ-ṣiḥāfijiyya al-ʿuṯmāniyya, 1275h1 and 1310h2. [Reprint Bayrūt: Dār al-kutub al-ʿilmiyya, n.d.]. Šarīf Jurjānī. Ḥāšiya. Al-Ḥāšiya ʿalā šarḥ al-Kāfijiya, cf. ʾAstarābāḏī. Zamaxšarī. Mufaṣṣal. Al-Mufaṣṣal fī ʿilm al-ʿarabiyya. Bayrūt: Dār al-Jīl, n.d.

29  Pointed out by Šarīf Jurjānī in margin of the ŠKāfijiya (1.5.below): “the same applies to the prefijixes of the imperfect: the hamza of ʾaḍribu is not a constituent, but is in fact, together with what follows it, one and only one constituent, and the hidden pronoun is another constituent” (wa-kaḏā fî ḥurūf al-muḍāraʿa fal-hamza fī ʾaḍribu laysat kalima bal hiya maʿa mā baʿda-hā kalima wāḥida haqīqatan waḍ-ḍamīr al-mustatir kalima ʾuxrā).

48

pierre larcher 2. Secondary Literature

Bally, Charles. 1965. Linguistique générale et linguistique française, 4th ed., reviewed and corrected. Berne: Francke. Ducrot, Oswald and Jean-Marie Schaefffer. 1995. Nouveau dictionnaire encyclopédique des sciences du langage. Paris: Seuil. Guillaume, Jean-Patrick. 1988. “Le discours tout entier est nom, verbe et particule. Elaboration et constitution de la théorie des parties du discours dans la tradition grammaticale arabe”. Langages 92, 25–36. Lalande, André. 1972. Vocabulaire technique et critique de la philosophie, 11th ed. Paris: Presses Universitaires de France. Levin, Aryeh. 1986. “The medieval Arabic term kalima and the modern linguistic term morpheme: similarities and diffferences”. Studies in Islamic History and Civilization in Honour of Professor David Ayalon, ed. M. Sharon. Jerusalem: Cana & Leiden: Brill, 423–446. ——. 2007. “Kalima”. Encyclopedia of Arabic Language and Linguistics. Leiden: Brill. 2.545–548. Owens, Jonathan. 1989. “The syntactic basis of Arabic word classifijication”. Arabica 36, 211–234.

DEFINING THE WORD WITHIN THE ARABIC GRAMMATICAL TRADITION: ʾASTARĀBĀḎĪ’S PREDICAMENT Jean-Patrick Guillaume

Preliminaries: Kalima in the Arabic Grammatical Tradition Although nearly all Arabic grammatical treatises begin with a chapter devoted to the parts of speech, which, from the early 4th/10th century onwards, usually contains a formal defijinition (ḥadd) of the noun, the verb, and the particle, it is only at a later period that a similar attention was given to the word (kalima). The fijirst grammarian to do so was quite probably Zamaxšarī, and his defijinition, “an expression instituted for a simple meaning” (lafḏ̣atun ḍāllatun ʿalā maʿnan mufrad),1 seems to have quickly become standard, since it appears in many later treatises,2 usually accompanied by a short commentary whose salient point is that “a simple meaning” should be understood as “a meaning such as none of its parts is denoted by a part of the [corresponding] expression” (al-maʿnā llaḏī lā yadullu juzʾu lafḏ̣ihi ʿalā juzʾihi). As Levin (1986; 2007) remarks, this defijinition, which is substantially equivalent to that of the morpheme in modern linguistics, does not exactly correspond to the actual use of kalima in grammatical texts; this discrepancy, however, does not seem to have ever been acknowledged, except by ʾAstarābāḏī, whose long and careful discussion of the standard defijinition (ŠKāfijiya: 1.2–6) is certainly the most lucid and signifijicant contribution to the question. Before we turn our attention to this text, which, obviously enough, is the main object of the present paper, it is necessary to examine its historical background, in order to assess the rather complex way in which ʾAstarābāḏī relates to his predecessors. Levin’s two aforementioned papers provide us with a rough sketch of the situation in the early grammatical treatises, mainly Sībawayhi’s Kitāb and Mubarrad’s Muqtaḍab. According to his analysis, they use the term

1

 See Zamaxšarī (Mufaṣṣal: 6) and Ibn Yaʿīs (ŠMufaṣṣal: 1.18–20).  Beside Ibn Ḥājib, which we will discuss presently, we can mention, among others, Ibn Hišām (Šuḏūr: 11–12; Qaṭr: 13–14) and Ibn ʿAqīl (ŠAlfijiyya: 1. 15). 2

50

jean-patrick guillaume

kalima to refer to some classes of morphemes, but also to more complex units; on the other hand, other classes of morphemes are described not as kalim (pl. of kalima) but as mere ‘augments’, zawāʾid (sing. zāʾida). The following table, based upon the data gathered by Levin, can give an idea of this distribution: Table 1: the distinction between kalima and zā’ida according to Levin (1986) Zāʾida

Bound morphemes I: prefijixes of the imperfective (ʾa-f ʿalu, ‘I do’); tanwīn (sufffijixed -n of indefijinite nouns); marker of the feminine gender in nouns (-at); sufffijix -iyy of the relative adjective; sufffijixes of the dual and the external plural

Kalima

A. Clitics and bound morphemes II: sufffijixed pronouns (raʾaytu-ka, ‘I saw you’), personal sufffijixes of the perfect (ḏahab-tu, ḏahaba-t ‘I went away, she went away’), wa- and fa- (conjunctions of coordination), defijinite article (al-qawmu ‘the people’), sa-yaf ʿalu (preverbal particle marking the future), some prepositions (xarajtu bi-Zaydin, ‘I got out with Zayd’) B. Independent morphemes: huwa ‘he’, kul ‘eat!’, xuḏ ‘take!’ C. Independent lexical units consisting in one or several bound morphemes of zāʾida status: faras-u-n ‘horse-nominative-indefinite’, ya-ḍrib-u ‘he-hits-indicative’ D. ‘Graphic words’ comprising one or several clitics or bound morphemes of kalima status: ʾaʿṭay-tu ‘I gave’, at-taṣdīru ‘the breast girth of a camel’

It should be noted that this classifijication reflects Levin’s purpose, which is a comparison between the notions of kalima in Arabic grammar and of morpheme in modern linguistics. Such an approach has, from my point of view, the advantage of offfering a rough sketch of the kind of data that will be dealt with in the following pages. However, it should not be taken as a wholly adequate representation of the Arabic grammarians’ conceptions and analyses. A most typical case is that of the sufffijixes of the perfect, which they consider not as sufffijixes at all, but as subject pronouns, on the grounds that in ḏahab-tu ‘I went away’ for instance, -tu occupies the same place, and consequently has the same status, as Zayd in ḏahaba Zaydun ‘Zayd went away’. This analysis, however, does not concern all those sufffijixes: in ḏahaba-t ‘she went away’, the sufffijix -t is considered not as a pronoun, but as a marker of the feminine; as for the pronoun of the 3rd person, they consider it to be a void element, the so-called ‘masked

defining the word within grammatical tradition

51

pronoun’ (aḍ-ḍamīr al-mustatir), so that the underlying representation of our example should be something like ḏahaba-∅-t, with ∅ standing for the pronoun and -t being a feminine marker. Moreover, this classifijication does not take into account the historical evolution of the grammatical doctrine: according to later grammarians, kul and xuḏ are derived from the underlying forms li-ta-ʾkul and li-ta-ʾxuḏ (respectively ‘let you eat’ and ‘let you take’), the elision of the injunctive particle li- and of the personal prefijix -ta- being explained by ‘frequency of usage’ (katrat al-ʾistiʿmāl). According to this analysis, then, xuḏ and kul consist, at the underlying level at least, in three elements, and to consider them as ‘independent morphemes’ does not really reflect those grammarians’ conceptions. Last but not least, the distinction between kalima and zāʾida seems not to be as stable and clear-cut as we could think, since the tanwīn and the -t of the 3rd pers. fem. of the perfect, which, according to Levin, are considered as ‘augments’ by Sībawayhi and Mubarrad, are classifijied by some later grammarians, among them Zamaxšarī, among the particles (ḥurūf al-maʿānī: see below). The above remarks are not intended as a criticism of Levin’s analysis, which is after all quite consistent with his own premises and based on sound textual data. They only aim at showing the complexity of the data subsumed by Arabic grammarians under the appellation of kalima, and also the fact that their conception of what was a word and what was not seems to have evolved in the course of the centuries. It is the causes and the dynamics of this evolution that I shall try to bring into light in the following pages.

Zamaxšarī’s Definition: Its Historical Background and Implications In the history of the Arabic grammatical tradition, the emergence of a formal standard defijinition for a given technical term does not usually entail an immediate and drastic change in the apprehension of the concept it denotes. On the contrary, these defijinitions were normally formulated so as to conform to the already established technical use of the term in question, or at least so as not to contradict it too openly. Actually, they frequently strike us as a formal exercise, or even as a merely outward manifestation of compliance with the dominant conception of what a

52

jean-patrick guillaume

science should be, and seem to have limited, if any, practical implications.3 What these defijinitions do inform us upon, however, is the kind of theoretical and epistemological resources available to the grammarians at a given period, and to the kind of issues these resources served to discuss. If we approach Zamaxšarī’s defijinition in this spirit, it becomes clear that its most relevant aspect is not the connection it suggests between kalima and morpheme, striking as it is, but the fact that it is clearly derived from Aristotle’s defijinitions of the noun and the verb (De int.: 16a 17sq; 16b 6sq; Poet.: 57a 10–17), and especially from their last clause: “none of [their] parts signify by [them]selves”, which is basically the same idea that Zamaxšarī expresses by saying that a word denotes a “simple meaning”. Taken in itself, such a borrowing has of course nothing remarkable: at Zamaxšarī’s time, “mixing grammar with logic”—a most heinous offfence in the eyes of the 4th/10th century grammarians—was not frowned upon anymore, and many aspects of Aristotelianism, notably logic, had become thoroughly acclimated in Arabic high culture, so that they were no more considered as foreign. What is more important is that this defijinition, transplanted in the context of Arabic grammar, was to take wholly new implications, and give rise to questions which were totally foreign to Aristotle’s approach, and sometimes explicitly excluded by it. Aristotle’s basic purpose was to provide a theoretical framework for the analysis of the simple assertive proposition,4 consisting in a subject, which is normally a noun, and a predicate, which is ordinarily a verb: it is probably for this reason that he considers nouns and verbs as the only ‘signifijicant’ parts of speech, dismissing arthron and sundesmos5 (which apparently refer to classes of grammatical words which can be neither subject nor predicate) as ‘non-signifying’ (asēmantikos). On the same basis, his stipulation that no part of a word has meaning in itself aims at keeping a clear-cut distinction between ‘signifying’ words (i.e. nouns and verbs), and complex expressions (logoi) consisting in two words or more. This principle is exemplifijied by the proper noun Kalippos: although it is obviously made

3  For example, it is impossible to deduce from the standard defijinition of the noun (“a word which expresses a meaning in itself, not related to a specifijic time”) that the Arabic grammarians consider ʿinda ‘at, near’ or ʾayḍan ‘also’ as nouns: their reasons for taking this decision are totally independent from the criteria referred to in the defijinition. 4  See Dupont-Roc & Lallot (1980: 314–39). The errors or over-simplifijications are, of course, mine own. 5  Arthron and sundesmos are usually translated as ‘article’ and ‘conjunction’, but Aristotle clearly uses them in a diffferent meaning than the later Greek grammarians. See Dupont-Roc & Lallot (1980: 321–28).

defining the word within grammatical tradition

53

up of two distinct elements, kalos ‘beautiful’ and hippos ‘horse’, such an analysis, according to Aristotle, is strictly pointless and indeed illegitimate, since Kalippos refers not to a beautiful horse, but to somebody who bears that name. Aristotle brings this line of reasoning still further when he states that in epaktrokelēs ‘pirate ship’, kelēs ‘ship’ has no meaning in itself, although he admits that it contributes to the meaning of the whole. Such an assertion can of course seem questionable, but, as Dupont-Roc and Lallot (1980: 330) remark: En somme, c’est bien une question de légitimité6 qui est tranchée ici [. . .] : aucun nom, fût-il composé de parties signifijiantes identifijiables, ne doit être pris pour un énoncé. [. . .] Ce qui est afffijirmé, en fijin de compte, plutôt que démontré, c’est que le composé se caractérise par des efffets sui generis d’unifijication sémantique qui en interdisent l’analyse.

In substance, then, we could say that Aristotle seems to admit that some words at least can be broken down into smaller units which can have something to do with the meaning of the whole, but that such a breakingdown is altogether pointless and tends, moreover, to cloud the distinction between the level of the word and the level of the utterance. But that, of course, reflects Aristotle’s own approach and preoccupations, which had very little in common with those of the Arabic grammarians. It is small wonder, then, that when they adopted his defijinition, they put it to use in quite diffferent contexts, and drew totally new conclusions from it. In order to begin to clarify this point, we can turn to Zamaxšarī’s commentator, Ibn Yaʿīš, who usually gives a fairly clear picture of the standard grammatical doctrine in his time. This is how he explains what Zamaxšarī means by “simple meaning”: ‘Simple’ is a second diffference by which he [i.e. Zamaxšarī] distinguishes [the word] from the complex [expression, murakkab] such as ar-rajul ‘the man’ and al-ġulām ‘the servant’ and suchlike [nouns] which are defijinite through the [article] al-, for these [expressions] signify two meanings, defijiniteness and what is made defijinite [i.e. ‘man’ or ‘servant’ in the preceding examples]. As regards the pronunciation [nuṭq] they are a [single] speech unit [lafḏ̣a] but [they are] two diffferent words, since they consist in the [article] al- which signifijies defijiniteness, and which is a word, as it is a particle, and of the defijinite [noun, which is] is another word. The pertinent criterion is that the whole expression denotes a meaning and that none of its parts denote a part of this meaning [. . .]. It goes the same with ḍarabā ‘they-dual-hit’ and ḍarabū ‘they hit’ and suchlike [i.e. the person marks of the perfect]:

6

 All the italics in this paragraph are the author’s.

54

jean-patrick guillaume each one of those is a [single] speech unit but consists technically of two words [kullu wāḥidin min ḏālika lafḏ̣atun wa-fī l-ḥukmi kalimatān], because -ā and -ū are nouns, since they express the subject [tufīdu l-musnada ʾilayhi] (ŠMufaṣṣal: 1.19.6–15)

Now, what is remarkable in this text is that it puts the old Aristotelian theorem that “no part of a word is signifijicant in itself ” to a wholly new use. For Aristotle, it meant, basically, that any attempt a breaking down the word into smaller signifying units is both pointless and illegitimate; which presupposes that the word is, at this stage, clearly identifijied as such, that his status as a word is already settled. Ibn Yaʿīš, on the other hand, uses the same theorem as a tool to break down into words a more complex kind of unit, lafḏ̣a which seems to occupy an intermediate position between the utterance (kalām) and the word. Although he does not state it expressly, and appears to consider it as existing only at the level of “pronunciation” (nuṭq), we can assume from the data he discusses that lafḏ̣a, in this context, refers to what we would call ‘graphic words’, corresponding to a chain of characters between two blank spaces, and frequently including a number of cliticized particles and/or pronouns together with a kernel word.7 In a way, these two diffferent approaches reflect the specifijicities of the scripts of the two languages. In ancient Greek, where words were not graphically separated, the word had to be recognized through a ‘bottom-up’ procedure, by grouping letters into syllables and syllables into words. Hence the importance of identifying the word correctly, i.e. deciding whether the sequence of letters such as hippos or kelēs should be considered as an independent word or simply as a part of Kalippos or epaktrokelēs. That is probably one of the reasons why Aristotle insists that the word, once it has been correctly identifijied (or, more exactly, constructed), should not be broken down again into smaller components, signifying or otherwise: such a reversal of the process would be pointless, if not actually disruptive. In Arabic, on the other hand, the reader is confronted with discrete, but often complex, graphic units, so that individual words must be recognized through an analytical, top-down process. This implies a wholly diffferent use of the principle that “no part of a word is signifijicant by itself ”: it is no longer a way to protect the word, once constructed, from any attempt at breaking it down again into smaller compo-

7  It seems in any case difffijicult to consider the word (graphic or otherwise) as a relevant prosodic unit (see Frolov 2000: 61–68).

defining the word within grammatical tradition

55

nents, but on the reverse a means to break down complex graphic units into simple words, the principle being that if a sequence of characters can be associated with a meaning, then there is grounds to consider it as an independent word. Admittedly, Ibn Yaʿīš’s use of this new device in the fragment quoted above remains rather limited. There is nothing new, within the grammatical tradition, in saying that clitics and personal endings of the perfect verb should be considered as independent words: as we have seen above, this idea was already taken for granted by the earliest grammarians. Obviously, his purpose is just to clarify and illustrate Zamaxšarī’s defijinition by applying it to rather simple and straightforward examples, and not to examine its deepest implications; in this, as in other matters, he appears as a competent if somewhat conservative exponent of the standard theory of his time. At this stage, then, we could be tempted to say that the borrowing of Aristotle’s defijinition did not lead to a substantial change in the traditional grammatical doctrine in this regard, and that its main advantage was to provide more self-consistent and theoretically founded ways in which to express old ideas and analyses. Such a statement could nevertheless be qualifijied. As I have noted above, the third part of Zamaxšarī’s Mufaṣṣal, devoted to the particles (ḥurūf al-maʿānī), mentions among them the tanwīn and the -t mark of the 3rd pers. fem. of the perfect. Since Zamaxšarī states explicitly—as a matter of fact in agreement with everybody else—that particles are a subspecies of the word, it should follow that he considers them not as augments, but as words, of a status comparable to the person marks of the perfect. In any case, this would be in accordance with his defijinition, since they both express a ‘simple meaning’.8 On the other hand, one could argue that these two alleged ‘particles’ appear near the end of the third part of the book, and that they are immediately followed by other elements whose status as actual particles is quite dubious, such as šīn al-waqf (i.e. the š-sound which follows the afffijixed pronoun of the 2nd pers. fem. sing. in the old tribal dialect of Tamīm, e.g. ʾakramtu-kiš instead of ʾakramtu-ki, ‘I honored you’) or ḥarf at-taḏkīr (i.e. the lengthening the fijinal vowel of a word while the speaker tries to remember what he wants to say next;

8  It should be noted, however, that for Zamaxšarī as for a majority of Arabic grammarians, the tanwīn is not a mark of indefijiniteness, but of the fact that the noun it afffects is “fully nominal” (mutamakkin; see Kouloughli 2007: 26–28). But then, such an abstract property, even if we would not call it semantic, would correspond to a frequent usage of maʿnā.

56

jean-patrick guillaume

e.g. min al-ʿāmī . . . kaḏā instead of min al-ʿāmi kaḏa, ‘since the year, er, such-and-such’), and some other whose presence in this section seems due to the fact that Zamaxšarī probably did not see where else he could put them. If we could nevertheless take the presence of the tanwīn and of the feminine marker -t in this section as proof that Zamaxšarī considers them as particles, this would indeed be the sign of a certain evolution within the grammatical theory, although a rather inconsistent one: why, one could ask, should the feminine marker -t of the perfect be considered as a particle, and not the sufffijix -at, which is its counterpart in nouns? At best, we could say that these hesitations and inconsistencies seem to indicate an incipient, if as yet rather dim, awareness that this new defijinition leads to classify as words many elements which were, up to now, not considered as such, and, as a matter of fact, do not correspond to the most common idea of what a word should look like. As we shall see presently, ʾAstarābāḏī was to be the fijirst, and actually the only, grammarian to clearly confront this question, and to give it at least a partial solution. ʾAstarābāḏī’s Discussion of Ibn Ḥājib’s Definition We can now turn our attention to ʾAstarābāḏī’s text, which consists in a long and thorough discussion of the standard defijinition, or more exactly of Ibn Ḥājib’s version of it: The word is an expression instituted for a simple meaning (1.2.13).9

The fijirst part of ʾAstarābāḏī’s commentary is devoted to a word-for-word explanation of this sentence, focused on its formulation; in spite of its intrinsic interest, it does not bear on our subject, except on two points, which illustrate the way ʾAstarābāḏī relates to his predecessors. The fijirst point concerns Ibn Ḥājib’s substitution of the term lafḏ̣ for lafḏ̣a, which appeared in Zamaxšarī’s original defijinition. Slight as it may appear, this diffference is probably not fortuitous, as Ibn Ḥājib’s treatise relies heavily on the Mufaṣṣal and usually keeps rather close to its formulations. ʾAstarābāḏī, however, does not explicitly pick out the discrepancy between the two texts, although he could hardly have been unaware of it, but raises the question in an impersonal way:

9

 Al-kalimatu lafḏ̣un wuḍiʿa li-maʿnan mufrad.

defining the word within grammatical tradition

57

If one said: He [Ibn Ḥājib] should have said lafḏ̣a [‘speech unit’] in order to exclude [the case of] two or several words, since there would be then two or several speech units. I would answer: The [insertion] of -at [marking] the singulative does not exclude this, because qālā ‘they said-dual’ and qālū ‘they said’ are single speech units just like ʾarṭā ‘a species of tree’ and barquʿ ‘veil’ and anything else which is uttered all at one time [kullu mā yutalafffaḏ̣u bihi marratan wāḥida], and yet the fijirst two [consist of ] two words, unlike the two last.

The discussion, in fact, bears on a fijine point of logical technique, a fijield in which Astarabādī is remarkably profijicient, and usually quite uncompromising: should the genus of kalima be identifijied as lafḏ̣ (‘articulated sound’ in general) or as lafḏ̣a (‘speech unit’, i.e. a discreet sequence of articulated sound), as Zamaxšarī suggests? Apparently, Zamaxšarī’s solution seems more adequate, since lafḏ̣a is, in a way, more specifijic than lafḏ̣. But, points ʾAstarābāḏī, such a gain of specifijicity is actually irrelevant, as lafḏ̣a does not exclude anything more than lafḏ̣ does. A ‘speech unit’ is simply a sequence of articulated sound ‘uttered all at one time’, which can contain one or more words, or no word at all, as for instance qak or ṣaṣaṣa10 (mentioned by Ibn Yaʿīš ŠMufaṣṣal: loc. cit.); the fact that it is ‘uttered all at one time’ does not bring it nearer to the word, since the diffferentia which identifijies the word is not a matter of the length of the sound sequence or its discreet character, but of its capacity to signify a ‘simple meaning’. The second aspect of the standard defijinition which ʾAstarābāḏī discusses is, precisely, the expression maʿnā mufrad ‘simple meaning’, which, taking once again the logician’s stance, he fijinds rather infelicitous: The established usage in the technical terminology of the logicians [al-mašhūr fī ṣṭilāḥ ʾahl al-manṭiq], is to consider ‘simple’ and ‘complex’ as properties of the expression, and to talk of ‘simple expression’ and ‘complex expression’ [al-lafḏ̣ al-mufrad wal-lafḏ̣ al-murakkab]. It is not appropriate to make innovations [lā yanbaġī ʾan yuxtaraʿ] in the defijinitions of [technical] terms, but on the contrary it is necessary to use what is already accepted and established, since the aim of a defijinition is to make [the defijined object] clear and distinct [li-ʾanna l-ḥadda lil-bayān]. And [Ibn Ḥājib] can not defend himself by saying “I meant by ‘simple meaning’ the meaning which is not complex”, because all verbs would then fall out of the defijinition of the word. If he had said “The word is a simple instituted [i.e. conventional] expression [or

10  These sequences are not chosen at random: both violate quite elementary and wellknown constraints on the structure of the root. They are not only inexistent as words, but unattestable as well.

58

jean-patrick guillaume ‘sequence of sound’]” [al-kalima lafḏ̣un mufradun mawḍūʿ], he would have escaped from this [objection]. (1.4.1–5)

Although this criticism is apparently directed at Ibn Ḥājib alone, its real target is obviously, once more, Zamaxšarī himself, the ultimate responsible for the ‘innovation’ ʾAstarābāḏī frowns upon. Here again, the point of the discussion is, at fijirst glance, mainly formal: ʾAstarābāḏī’s argument is that, since the logicians already had a satisfactory defijinition of the word referring to the notion of ‘simple expression’ (lafḏ̣ mufrad), replacing it with ‘simple meaning’ is both useless and potentially confusing. It is useless, because what Zamaxšarī wants to say by ‘simple meaning’ (“a meaning such as none of its parts is signifijied by a part of the expression”) is exactly the same thing that the logicians meant by ‘simple expression’ (“an expression such as none of its parts signify a part of its meaning”), so that there is no advantage in substituting one for the other. It is, moreover, potentially confusing, because the natural interpretation of ‘simple meaning’ would be ‘a meaning which is not complex’; but then, since verbs denote both a process and a time, it would follow that they have a complex meaning11 and so are not words, which is clearly absurd. Of course, that is not what Ibn Ḥājib or Zamaxšarī intended to say, but, by forcing an arbitrary, unnatural interpretation on the term “simple meaning”, they have only managed, in ʾAstarābāḏī’s opinion, to make the question more difffijicult to grasp. In a way, this discussion could strike us as merely formal and academic. But, at another level, ʾAstarābāḏī’s criticism seems motivated by more substantial considerations. The main reason why Zamaxšarī’s (to his mind) infelicitous wording should be rejected is that it risks to cloud a major theoretical principle he has stated some lines before; it appears in his commentary on the standard defijinition: “Instituted for a simple meaning”: He [Ibn Ḥājib] means by that a meaning such as none of its parts is signifijied by a part of the [corresponding] expression [yaʿnī bihi l-maʿnā llaḏī lā yadullu juzʾu lafḏ̣i-hi ʿalā juzʾihi]. [And this] regardless of whether this meaning has parts, such as the meaning of ḍaraba

11  Actually, Ibn Sarrāj uses the terme ‘simple meaning’ exactly in this sense, in order to distinguish the noun from the verb (Uṣūl: 1.38). It is difffijicult to decide whether Astārābādī had this passage in mind when criticizing Zamaxšarī’s defijinition, but there is no doubt that Ibn as-Sarrāj’s interpretation is more natural and straightforward.

defining the word within grammatical tradition

59

‘[he] hit’12 which signifijies [both] the maṣdar [i.e. the process] and the time, or not, such as the meaning of ḍarb ‘the fact of hitting’ and naṣr ‘victory’. According to this principle, a complex meaning is such as each of its parts is signifijied by a part of the [corresponding] expression, as in ḍaraba Zaydun ‘Zayd hit’ or ʿabdu llāh ‘servant of God’, when they are not used as proper nouns. When they are used as proper nouns, though, their meaning is simple, and so is their expression, since a simple expression is such as none of its parts signifijies a part of its meaning, and a complex expression is such that each of its parts signifijies a part if its meaning. (1.3.24–1.4.1)

Although the interpretation of the standard formula is basically the same as in Ibn Yaʿīš’s fragment quoted above, and indeed in all the postZamaxšarian tradition, it is expressed here in a particularly precise and forceful way: a meaning (or, conversely, an expression) can not be considered as simple or complex in itself, but only by referring to the corresponding expression (or, conversely, meaning). In other terms, deciding whether a meaning or an expression is simple is not a matter of common sense or of basic linguistic feeling, but of principle: although we can immediately discern that the proper noun ʿAbd Allāh consists, formally at last, in two elements, it should nevertheless be considered as a ‘simple’ expression (a single word), since none of these parts corresponds to equivalent divisions of its meaning. At fijirst glance, this fragment appears strongly reminiscent of Aristotle’s analysis of the compound words. But there is also something new in the manner ʾAstarābāḏī stresses the inseparability and parallelism of expression and meaning. What his reasoning amounts to, actually, is that, in order to break down an utterance into minimal relevant units, you have to look simultaneously at its form (lafḏ̣) and its meaning (maʿnā), and that the units you identify, as a consequence, have both a formal and a semantic aspect, a signifijier and a signifijied. Actually, what is most striking here is not so much the Aristotelian undertones of this idea than its analogy with the Saussurian metaphor of the sheet of paper: La langue est encore comparable à une feuille de papier: la pensée est le recto, et le son le verso; on ne peut découper le recto sans découper en même temps le verso; de même, dans la langue, on ne saurait isoler ni le son de la pensée, ni la pensée du son. (Quoted from Benveniste 1966: 52)

12

 In this example, ḍaraba should clearly taken as representing simply the verb stem, devoid of any person mark, or else it could in no way be considered as ‘simple’, since it would contain a pronoun.

60

jean-patrick guillaume

This conception of the relation between sound (lafḏ̣) and meaning (maʿnā) is quite diffferent from that of the early Arabic grammarians and rhetoricians, who, on the opposite, postulated the basic autonomy of these two dimensions of language (Kouloughli 1983). It was ʿAbd al-Qāhir al-Jurjānī who fijirst elaborated—and defended with an uncommon polemical zeal—the principle of “a strict functional correlation between the form of an utterance and its semantic content” (Kouloughli 2008), and some of his ideas found their way into the mainstream grammatical tradition, principally through Zamaxšarī. In this regard, Zamaxšarī’s defijinition of the word, and, consequently, Astarabādī’s commentary on this defijinition, can be considered as a by-products of Jurjānī’s theoretical discovery. ʾAstarābāḏī, then, appears until now as a particularly lucid and articulate exponent of a set of ideas which were already current within the grammatical tradition. At the same time, he has brought the matter to such a degree of clarity and precision that it has become impossible not to raise questions that his predecessors had managed to avoid.

The Status of Morphological Markers The fijirst question concerns the status of the morphological markers, and more specifijically those which are classifijied by the early grammarians as “augments” (zawāʾid): Question: in muslimāni ‘two Muslims’, muslimūna ‘Muslims’, baṣriyyun ‘Basrian, of Basrah’, and all the [forms of the] imperfect verb, in all this a part of the signifijier13 does signify a part of the signifijied, since -ū- [in muslimūna] signifijies plurality, -ā- [in muslimānī] duality, -iyy [in baṣriyyun] relation and the marks [i.e. personal prefijixes] a meaning of the imperfect [i.e. either present or future] and also the state of the subject [i.e. the person] And the same thing goes for the [marker] -at of the feminine in qāʾimatun ‘standing up [feminine]’, the tanwīn, the defijinite article al-: in all this the signifijier is necessarily complex, and also the signifijied, so they are not one word but two. (1.5.14–18)

ʾAstarābāḏī’s line of reasoning is quite clear: since each of these “augments” is a minimal signifijicant unit or, according to the standard wording “signifijies a simple meaning”, it implies that they should be considered

13  For simplicity’s sake, I shall henceforward translate lafḏ̣ and maʿnā by ‘signifijier’ and ‘signifijied’; as I have tried to show above, this translation is adequate in this precise context, although it can be seriously misleading in most cases.

defining the word within grammatical tradition

61

as words. But, on the other hand, this notion clearly runs against the basic linguistic feeling (corresponding, by and large, to the established opinion and practice of the grammarians) that qāʾimatun or baṣriyyun, for example, are words, while -at or -iyy are not. Reasoning from a modern point of view, we should naturally be tempted to conclude that the defijinition is inadequate, or more exactly that it defijines not the word but a smaller type of linguistic units, which we are used to call morpheme, and that ʾAstarābāḏī, in spite (or because) of the keenness and the clarity of his analysis, has simply achieved to make more glaring the confusion between the level of the word and the level of the morpheme, which was inherent to the early grammarians’ approach. ʾAstarābāḏī, however, follows a quite diffferent logic: while being uncommonly perceptive of the inconsistencies of his predecessors—and quite outspoken in pointing them out—he nevertheless keeps a strong sense of commitment to the tradition as a whole, which prevents him from pushing his criticism to the point where it would endanger the very fundaments of the theory. This attitude, which appears in many discussions throughout his treatise, leads him to fijind his solutions within the tradition itself, by rearranging its theoretical and technical resources in a way that is both compatible with the conceptual framework of the theory and more selfconsistent and/or empirically accurate than its standard formulations; we could describe him as an enlightened conservative, not a radical reformist (which distinguishes him, by the way, from Jurjānī). Such is the background against which we should examine his answer: Answer: in all the cases you have mentioned, [we have] two words which are so tightly merged that they have become as one [kalimatāni ṣārat min šiddati l-ʾimtizāji ka-kalimatin wāḥida], so that this compound [murakkab] receives its case mark as if it were a [single] word. This is due to the lack of autonomy of the phonological segments which are bound [wa-ḏālika li-ʿadami stiqlāli l-ḥurūfiji l-muttaṣila] within the aforementioned words. The same thing could be said of the case vowels. (1.5.18–20)

Here again, the argumentation, at fijirst glance, seems quite clear. On the one hand, these morphological markers, since they fall within the fijield of the defijinition, should be considered as words: ʾAstarābāḏī clearly considers that the canonical defijinition is axiomatic, and should not be challenged. But, on the other hand, they form a special, atypical class of words, owing to their “lack of autonomy”, which prevents them to stand as independent units such as muslim or Baṣra, and makes it necessary for them to be combined with one or several others. In this way, muslimatun or Baṣriyyun, although they could appear to the basic linguistic feeling as

62

jean-patrick guillaume

‘single’ words, are actually ‘complex’, made of several words “which have become as one”. One important aspect of this solution is that, although it seems quite new in the context of the Arabic tradition (to my best knowledge, no Arabic grammarian before ʾAstarābāḏī has ever made this use of the notion of ‘compounded word’),14 it does not imply any drastic changes in its basic principles. As we have already seen, all grammarians agreed that the forms of the perfect verb such as ḍarabtu ‘I hit’ should be considered, technically at last, not as a single word, but as two, since it integrated a verb and a personal pronoun; as we shall see presently, they also admitted that this compound behaved, in some ways, like a single word. From that point of view, then, we could say that ʾAstarābāḏī’s explanation consists simply in an extension of this analysis, which was usually referred to in the context of the personal sufffijixes of the perfect, to the morphological markers as a whole. The same thing could be said the ‘lack of autonomy’ which accounts for the fact that these markers, in spite of their being words (theoretically a least), must be ‘combined’ with another word: although ʾAstarābāḏī does not expatiate on this point, it is clear that he refers to a basic principle of Arabic phonographematics, saying that one-consonant (or ‘one-letter’, ḥarf ) words cannot stand alone as independent graphic units, and must be attached to the following and preceding word. Admittedly, this principle was mainly invoked, before ʾAstarābāḏī, in the case of clitic particles or pronouns; but then, its generalization to bound morphemes, although unusual, could be felt as altogether acceptable within the framework of the standard grammatical theory. There are two points, however, on which ʾAstarābāḏī clearly departs from the principles laid out by his predecessors. The fijirst one concerns the old distinction between kalim and zawāʾid, which, for obvious reasons, becomes irrelevant. But then, as we have seen, this distinction, which in any case does not seem to have played a major part in the theory, never was very clear to begin with, and had already begun to give way in post-Jurjānian grammatical treatises; in this regard, we could say that

14  This expression is commonly used to refer either to a class of proper nouns, of Arabic or foreign origins, such as Taʾabbaṭa Šarran, Maʿdīkarib or ʿAmrawayhi (see Zamaxšarī Mufaṣṣal: 6–7; 48), or indeclinable complex locutions such as xamsata ʿašara, ‘fijifteen’ or bayna bayna, ‘[something] standing in the middle, neither good nor bad’ (176–179). Its use by ʾAstarābāḏī could arguably be a borrowing from Aristotelian logic, although Aristotle expressly excludes morphology from his analysis of nouns and verbs (De Int. 16a–b; Poet. 57a 18–23; see also Dupont-Roc & Lallot 1980: 332–337).

defining the word within grammatical tradition

63

ʾAstarābāḏī, by letting it quietly disappear, simply gives the fijinal touch to a process which had begun, at least, in Zamaxšarī’s Mufaṣṣal. The second point concerns the status of case vowels: in the last sentence of the passage, ʾAstarābāḏī clearly implies that they also should be considered as ‘words’. Admittedly this conception is quite consistent with his reasoning: since it was commonly agreed, in his time, that every case mark had a specifijic, intrinsic meaning,15 it logically followed that it should be considered as a word, on a par with the markers of the feminine or of the dual. But then, the problem is that, for the Arabic tradition as a whole, short vowels are consider as units smaller than the “segment” (ḥarf, i.e. consonant or glide);16 so then, in this case, we would have words consisting in less than one segment. Such an idea would probably strike a contemporary reader as rather unusual. The rather peculiar way in which ʾAstarābāḏī expresses it, as a kind of afterthought, could be interpreted as a subtle strategy to make it acceptable, the underlying reasoning being something like this: if morphological markers consisting in one segment, owing to their lack of autonomy, must be combined with other words, then the same thing holds a fortiori for the case vowels which are even smaller. In this way, the potentially contentious assertion that case vowels should be considered as words is presented as a self-evident truth, which can be taken for granted and does not need further demonstration. ʾAstarābāḏī further insists that these ‘compounded’ or ‘complex’ words are more that the mere juxtaposition of minimal semantic units, and that, at a certain level they behave as if they were a single word, or more exactly as “two words that have become as one”. In the passage quoted above, he gives an instance of this behavior when he notes that “they receive their case mark as if their were a single word”. This idea reflects the fact that the case marks (together with the tanwīn) have always had a rather special status within the Arabic grammatical theory: besides the part they play in the expression of the meaning, they are also considered as marking the fijinal boundary of the words (ʾawāxir al-kalim).17 The fact that,

15  This idea was fijirst expressed by Jurjānī, and found its way into the standard doctrine through Zamaxšarī. See Guillaume (1998). 16  It is commonly expressed by saying that the short vowels are ‘parts’ of the corresponding glides. See Bohas and Guillaume (1984: 244, fn. 2). 17  This conception is prominent, for instance, in the 9th chapter of Zajjājī’s ʾĪḍāḥ (72–75; see also Versteegh 1995: 118–120), titled “Why do the case marks occur at the end of the word instead of at the beginning or at the middle?” It is also the basis of Quṭrub’s famous claim, which amounts to saying that the primary function of the case marks is not syntactico-semantic, but prosodic (Zajjājī ʾĪḍāḥ: 69–71; see also Versteegh 1995: 101–108).

64

jean-patrick guillaume

in muslim-at-u-n, for instance, the -u- of the nominative and the tanwīn occur after the -at- of the feminine is here adduced as a proof that muslim– and -at-, although they are, according to ʾAstarābāḏī’s theory, two diffferent words, their behavior, as far as case marking is concerned, is that of a single word. In the lines immediately following this fragment, ʾAstarābāḏī mentions other instances of this principle: The fact that they [i.e. these ‘compounded’ words] are treated as a [single] word, is the reason why the vowel following the fijirst element [i.e. the fijirst consonant of the root] in the imperfect verb is deleted, and why [the radical of ] the relative adjective is modifijied in cases such as numarī ‘from the tribe of Numayr’, ʿalawī ‘ʿAlid, descendant of ʿAlī’ or wašawī ‘multicolored’ [from wašā ‘multicolored efffect’]. In this way, the structure of the noun of which the relative adjective is derived and that of the imperfect verb are changed by [the presence of ] these two particles [i.e. the sufffijix -ī and the personal prefijixes, respectively], which become a part of the overall structure of the word.

The formation of the imperfect of the simple form of the verb, or more exactly its analysis by the Arabic grammarians, has been treated in detail by Bohas (Bohas and Guillaume 1984: 91–94). I will limit myself, accordingly, to evoke briefly its relevant aspects. According to the Arabic grammatical theory, the imperfect is derived from the perfect by a sequence of morphological and phonological processes which can be represented as follows: Perfect Introduction of the personal prefijixes and modifijication  of the fijinal vowel Ablaut Deletion of the surplus vowel

ḍaraba ya-ḍarab-u ya-ḍarib-u ya-ḍ∅rib-u

This last process is, for the Arabic grammarians, the result of a general constraint on the structure of the words, which Ibn Yaʿīš expresses in the following way. There is no word in Arabic consisting in four consonants followed by vowels [i.e. with the structure CVCVCVCV] (ŠMulūkī: 48)18

Some pages later he explicitly states that this constraint is responsible for the deletion of the ‘surplus’ vowel in ya-ḍarib-u →yaḍribu:

18

 Laysa fī l-ʿarabiyya kalimatun tatawallā fī-hā ʾarbaʿu mutaḥarrikāt.

defining the word within grammatical tradition

65

And the vowel following the [consonant] after the prefijix of the imperfect is necessarily deleted in the three-consonant verbs [. . .] it is deleted in order to avoid the succession of four consonants followed by vowels in the morphological base [ʾarbaʿ mutaḥarrikāt lawāzim] (ŠMulūkī: 62).

It should be noted that the same “*4CV constraint”, as Bohas calls it, also plays a part in the formation of the perfect. As we have seen above, the Arabic tradition considers such forms as ḍarabtu ‘I hit’, not as consisting in a stem (ḍarab-) and a sufffijix of the 1st pers. sing. (-tu), but as two distinct words, a verb (ḍaraba) and a pronoun (-tu). The combination of the two would give a theoretical form *ḍarabatu: the deletion of the vowel is accounted for by the application of the *4CV constraint. But then, this analysis raises a problem: since this constraint functions only within the boundaries of the word, how can it apply to *ḍarabatu, which consists in two distinct words, a verb and a subject pronoun? Moreover, how is it possible to account for the fact that is does not apply when the verb is followed by an object afffijixed pronoun, as in ḍarabaka ‘he hit you’, which exhibits the same CVCVCVCV structure? This is how Ibn Yaʿīš answers these two questions: Although the subject is [as a matter of principle] distinct from the verb, [the subject pronoun] is treated technically as if it were one of its elements [ʾujriya mujrā baʿḍi ḥurūfijihi ḥukman]. Don’t you see that [the speakers] delete the fijinal vowel of the verb when it is bound to the subject pronoun, as in ḍarabtu ‘I hit’ and katabtu ‘I wrote’, in order to avoid having four consonants followed by vowels in the same word, but that they do not [delete the vowel] when [the verb] is bound to the object pronoun, as in ḍaraba-ka ‘He hit you’ or šatama-ka ‘He abused you’. (ŠMufaṣṣal: 7.7)

Incidentally, this passage shows clearly that, although the Arabic grammarians’ terminology do not seem to distinguish systematically between bound morphemes (as we call them today) and clitics, they were nevertheless quite conscious of the diffference between them, at least in the cases where this diffference plays a crucial part. But perhaps the most important aspect here is the fact that Ibn Yaʿīš’s analysis of the perfect (which reflects the position of the Arabic tradition as a whole) corresponds exactly in substance, if not in wording, to ʾAstarābāḏī’s defijinition of the “compounded word”: ḍarabtu consists in two words which “have become as one”, so that it is subjected to a constraint which can apply only within words. This remark can help us in clarifying ʾAstarābāḏī’s approach, and the way he relates to the preceding tradition. As I have noted earlier, he is probably the fijirst (and I should add the only) grammarian to make that

66

jean-patrick guillaume

particular usage of the notion of compounded word; but then, he did not invent it wholesale, and, if it cannot be excluded that he was influenced by Aristotelism, the peculiar way he uses it shows that it is not a simple case of borrowing. As most of his innovations, this one is fijirmly grounded in the standard theory: all he did, actually, was to generalize the commonly accepted analysis of the personal sufffijixes of the perfect to all the bound morphemes. Naturally, this implied abolishing the distinction between kalim and zawāʾid, but, as we have noted, it had never had a really important status in the theory.

A Killing Objection: Root and Pattern Up to this point, ʾAstarābāḏī’s discussion was concerned only with what we call today external inflexion: as we have seen, he has achieved a solution of this problem which enables him to maintain the standard defijinition of the word, while remaining reasonably self-consistent and in agreement with the basic theoretical principles of the tradition. But then, external inflexion is only one half in the problem, especially in Arabic: at this point, ʾAstarābāḏī cannot very well avoid treating the other half, and this last step will prove practically fatal for his theory: As for the perfect verb, such as ḍaraba ‘[he] hit’, the question is not settled [ fīhi naḏ̣ar], as there is no discussion that it is one word [i.e. it is not compounded], although the process is signifijied by the succession of the consonants [of the root, i.e. √ḍrb] and the assertion that this process happened in the past is signifijied by the pattern which afffects these consonants [i.e. CaCaCa]. Now, the pattern is a part of the signifijier, since it expresses [the specifijic disposition of the vowels which are associated to the root consonants, this disposition having been instituted in order to express a specifijic semantic value—free translation], and since the vowels are a part of what is pronounced [i.e. of the signifijier]. It would then be a word compounded of two parts, each referring to a part of the signifijier. The same analysis would apply, for instance, to ʾusud, the plural of [ʾasad ‘lion’], to all diminutives [e.g. rujayl ‘little man’, from rajul ‘man’], of [broken plurals] such as rijāl ‘men’ or masājid ‘mosques’, or of [deverbative nouns] such as ḍārib ‘hitting’, maḍrūb ‘hit’ and maḍrib ‘the cutting edge of a saber’ [lit. ‘what one uses to hit’], since, in all these examples, the notions of diminution, plurality, agent, patient or instrument, are signifijied by the pattern.

The problem, then, is the following: if one admits that in ḍaraba, for instance, the process (‘hitting’) is signifijied by the root consonants √ḍrb, and the past tense by the pattern (which is a totally commonplace idea in Arabic grammar), it logically follows that ḍaraba falls within the defiji-

defining the word within grammatical tradition

67

nition of the compounded word as given above, since both root and pattern can be considered as parts of the signifijier, and express a specifijic, identifijiable part of the signifijied. The same analysis applies, in an even more limpid way, to broken plurals: the pattern CiCāC in rijāl expresses plurality exactly in the same way as the -ū- in muslimūna, and there is no logical reason in giving a diffferent treatment to each. Admittedly, such an analysis goes against the accepted view that ḍaraba is a simple, non-compounded word: as we have seen earlier, ʾAstarābāḏī uses it as an example of a word which must be considered as “simple”, since no part of its signifijier refers to a part of its signifijied. But, adds ʾAstarābāḏī, there is a much more serious objection to this analysis: Now, there would be no sense in saying here that the pattern is a word which has been combined [with another] to become the part of a single word, as we did for the words we saw earlier [i.e. muslimatun, Baṣrī and so on], or as we could do for the case vowels. So then, the objection grounded on these words [of which we are now speaking] is valid . . . (ŠKāfijiya: 1.5–6).

According to ʾAstarābāḏī’s defijinition, a compounded or complex word consists in “two or several words which are so tightly merged that they become as one”. As we have seen, it is not absurd, in the framework of Arabic grammar, to consider bound morphemes such as the prefijixes of the imperfect or the -at- of the feminine as words of a kind; it is even possible, by stretching a point, to include the case vowels in this analysis. But the same can hardly be said of either root or pattern, since the former is simply a sequence of vowel-less consonants and the latter represents the overall vocalic melody of the word: in other terms, they stand at such a level of abstraction that they cannot qualify any longer as words. So then, ʾAstarābāḏī is confronted to a logic contradiction: on the one hand, it is impossible to say that these words are such that no part of their signifijier denotes a part of their signifijied, and on the other hand, it is impossible to consider them as “compounded”. The only possible way he sees out of this predicament consists in adding an ad hoc stipulation to the defijinition of the complex word: Except, possibly [ʾallahumma ʾillā] if we added a constraint to the defijinition of the complex word, and said that it is such that each of its parts denotes a part of its signifijied when those parts follow each other: in the words we are discussing, those two parts are heard all at one time (ŠKāfijiya: 1.6)

Such a constraint would ensure that all the words in question would be considered as simple, since, as we have seen earlier, the condition for a word to be complex is not that its signifijied is complex, but that each part of its signifijier denotes a part of its signifijied. But ʾAstarābāḏī is clearly aware

68

jean-patrick guillaume

that it is a mere device to get rid of a problem which, in the framework of the Arabic grammatical tradition, was probably impossible to solve.

References 1. Primary Sources Aristotle. De int. De l’interprétation, tr. J. Tricot. Paris: Vrin, 1984. Aristotle. Poet. See Dupont-Roc and Lallot (1980). ʾAstarābāḏī. ŠKāfijiya. Šarḥ al-Kāfijiya fī n-naḥw, 2 vols. Bayrūt: Dār al-Kutub al-ʿIlmiyya, n. d. [Reprint of the Istanbul edition, 1275/1884]. Ibn ʿAqīl. ŠAlfijiyya. Šarḥ Alfijiyyat Ibn Mālik, ed. M.M. ʿAbd al-Ḥamīd, 2 vols. Bayrūt: Dār al-Fikr, n.d. Ibn Hišām. Qaṭr. Šarḥ Qaṭr an-nadā wa-ball aṣ-ṣadā, ed. M.M. ʿAbd al-Ḥamīd. Bayrūt: Dār al-Fikr, n.d. ——. Šuḏūr. Šarḥ Šuḏūr aḏ-ḏahab, ed. M.M. ʿAbd al-Ḥamīd. Bayrūt: Dār al-Fikr, n.d. Ibn Sarrāj. ʾUṣūl. Kitāb al-Uṣūl fī n-naḥw, ed. A.H. al-Fatlī, 3 vols. Bayrūt: Muʾassasat ar-Risāla, 1985. Ibn Yaʿīš. ŠMufaṣṣal. Šarḥ al-Mufaṣṣal. 10 vols. Bayrūt: ʿĀlam al-Kutub & al-Qāhira: Maktabat al-Mutanabbī, n.d. Zajjājī. ʾĪḍāḥ. Kitāb al-ʾĪḍāḥ fī ʿilali n-naḥw, ed. M. al-Mubārak. Bayrūt: Dār an-Nafāʾis, 1970. Zamaxšarī. Mufaṣṣal. Al-Mufaṣṣal fī ʿilm al-ʿarabiyya, ed. M.B. an-Naʿsānī al-Ḥalabī, 2nd ed. Bayrūt: Dār al-Jīl, n.d. 2. Secondary Literature Benveniste, Emile. 1966. Problèmes de linguistique générale. Paris: Gallimard. Bohas, Georges & Jean-Patrick Guillaume. 1984. Etude des théories des grammairiens arabes: I Morphologie et phonologie. Damas: Institut Français d’Etudes Arabes. Dupont-Roc, Roselyne & Jean Lallot. 1980. Aristote: La Poétique, texte, traduction et notes. Paris: Seuil. Frolov, Dmitry. 2000. Classical Arabic Verse: History and Theory of the ʿArūḍ. Leiden: Brill. Guillaume, Jean-Patrick. 1998. “Les discussions des grammairiens arabes à propos des marques d’iʿrāb”. Histoire, Epistémologie, Langage 20.2, 43–62. Kouloughli, Djamel. 1983. “A propos de lafẓ et de maʿnā”. Bulletin d’Etudes Orientales 35, 43–63. ——. 2007. Le résumé de la grammaire arabe par Zamakšarī. Lyon: ENS Editions. ——. 2008. “Maʿnā”. Encyclopedia of Arabic Language and Linguistics, ed. K. Versteegh, Leiden: Brill: 3. Levin, Aryeh. 1986. “The Medieval Arabic term kalima and the modern linguistic term morpheme: Similarities and diffferences”. Studies in Islamic history and civilization in honour of Professor David Ayalon, ed. M. Sharon, Jerusalem: Cana & Leiden: Brill, 423–446. ——. 2007. “Kalima”. Encyclopedia of Arabic Language and Linguistics, ed. K. Versteegh. Leiden: Brill, 2.545–548. Versteegh, Kees. 1995. The Explanation of Linguistic Causes: Az-Zajjājī’s Theory of Grammar. Amsterdam: Benjamins.

ELLIPSIS IN THE ARABIC LINGUISTIC THINKING (8TH–10TH CENTURY) Cristina Solimando

Ellipsis is a universal property of any natural language, but its scope and means of realization difffer substantially from one language to another. Working on ellipsis requires reference to syntax, prosody, lexical semantics and stylistics; it represents a useful instrument for the descriptional and functional analysis of a language.1 The study of a sentence in terms of the word’s or sentence’s non-expression plays a central role in the syntactic analysis: identifying the omitted part of the discourse and its following reconstruction allow linguists to point out the structural properties and grammatical characteristics of any language. Although syntactic ellipsis— which consists in the non-expression of a syntactical category whose referent can be recovered by syntactic rules or discourse cues—is a universal property of a language, still the specifijic permitted types and their restrictions difffer form one language to another. Contemporary linguistics identifijies two diffferent kinds of ellipsis: the semantic and the syntactic ellipsis. The semantic ellipsis is referred to as omitted information—which is not syntactically relevant for the sentence structure but fundamental for a complete semantic representation. The syntactic ellipsis consists of the non-expression of an obligatory syntactic category2 recoverable through syntactic rules and internal cues. The aim of the reconstruction of a syntactic ellipsis is the establishment of a textual relation, whereas the reconstruction of the semantic ellipsis needs an extra-textual reference in order to recover the omitted information. The sentence grammar analizes the syntactic aspects of the emergence of the ellipsis and its recoverability with exclusive attention to the syntactic and formal proprieties of the sentence. The discourse grammar—which represents a more comprehensive approach towards the study of the ellipsis—is more focused on highlighting the communicative relevance of

1

 See McShane (2005).  As “syntactically obligatory categories” we intend the verb, its arguments and some elements belonging to other parts of the discours like the conjuctions. 2

70

cristina solimando

the extra-linguistic context, the phrase intonation and all the non-formal aspects, determinant for what is said and what is understood in a comunicative process, are linked to the cognitive-linguistic universe of the speakers. Functional grammar acknowledges the language as a means of interaction among human beings3 and considers the exclusive syntactical analysis insufffijicient to identify the diffferent nuances of the comunicative act. Scholars such as Lambrecht (1994) and Kuno (1987) strive to converge these two approaches in the study of language: “there are numerous conflicts in the outlooks of pure and functional syntacticians with respect to how to analyse a given linguistic phenomenon. Pure syntacticians tend to give characterizations to linguistic phenomena which are controlled by non-syntactic factors, or they label them as non-syntactic phenomena and brush them aside, without attempting to fijind out what kind of nonsyntactic factors are in control.”4 Despite the difffijiculty of codifying the mental system which allows us to produce and understand a sentence, the analysis of the “non-syntactic factors” is fundamental for the identifijication of the limits of the elliptical mechanism and the recovery of the deleted element. Chao (1988: 1) identifijies two typologies of elliptical constructions distinguished according to the absence/presence of the main syntactical constituents (phrasal heads): where syntactical constituents are not expressed (absence), the interpretation of the linguistic material is strictly syntactical. However, when they appear under an anaphoric form, their reconstruction is based not only on syntactical rules but also on the discourse representations. This analysis includes all the aspects of a complete description of a communicative act from the syntactical-linguistic aspects to the semantic-contextual vis-à-vis the cognitive background of the speaker. For syntactic ellipsis to be possible the language must license ellipsis in that given confijiguration and more signifijicantly the content of the elided category must be recoverable and understandable.5 In classical or contemporary Arabic linguistics, ellipsis never acquire a well-defijined theoretical status, and it is worth saying that is has never been subject to a specifijic and thorough study. This does not imply that Arabic grammarians did not explore such phenomenon to clarify the sentence

3

 Dik (1997: 3).  Kuno (1987: 2). 5  Cf. McShane (2005: 16). 4

ellipsis in arabic linguistic thinking

71

structure and defijine the relationship between morphology, syntax, and rhetoric in the Medieval Arabic grammatical theories. This absence of a theoretical status together with some practical difffijiculties—mainly the dispersion of passages related to the ʾiḍmār and often the necessity of collecting texts that do not belong to linguistic tradition, such as Koranic commentaries (tafsīr)—determined the scanty studies on this topic. Arabic grammarians have always analyzed the syntax of a sentence using reconstruction operations; this consisted of formulating virtual propositions in which they indicate the presence of an unexpressed element in the initial phrase. The use of such a linguistic mechanism is based on exemplary representations of a sentence in which an element was elided in order to highlight the system of syntactical norms that regulate the behaviour of that constituent. Most common strategies to license the ellipsis of a particular constituent or lexical category were: the elided category must justify the syntactic structure of the sentence, which has to respect the grammatical norms related to the ʾiʿrāb analysis; the deleted element needs to be easily recoverable in order to avoid cases of ambiguity between the two interlocutors. This objective is pursued by trying to elide the least number of elements possible and by preferring to elide words that are derived from the same origin of the element that appear in the same phonic sequence.6 Moreover, cases in the colloquial speech are not to be excluded, as we will see in the case of Kitāb di Sībawayhi, the non-expressed material is recoverable due to the background information known to the interlocutors. Terms that are used in the Arabic Linguistics and that indicate the ommision of one part of the discourse and its restitution are ʾiḍmār— ḥaḏf and tamṯīl—taqdīr, respectively. ʾIḍmār, verbal noun of ʾaḍmara ‘pronouning’ but also ‘hiding,’ ‘smothering,’ can indicate anaphoric elements like pronouns, as well as elements without any phonic value but has a determining role in defijining the syntactic structure of the sentence.7 In contemporary grammar such term seems to disappear and to be replaced by the term ḥaḏf used to indicate the omission of one part of the discourse. As a verbal noun of ḥaḏafa ‘to delete, suppress, cut,’ this was used mainly to indicate, in phonologic terms, phenomena of eliding of a phoneme within a word. Therefore, the present work proposes to analyze the 6

 Bohas (2001).  Its contrary ʾiḏ̣ hār, verbal noun of ʾaḏ̣ hara ‘to show, to make visible’ and thus ‘to express’ indicates the operation of making explicit one element in the utterance; see Ayoub (1990). 7

72

cristina solimando

theoretical path which determined this terminological change, in light of theoretical-linguistic mechanisms which were able to determine more relevant mutations in the ellipsis theoretical framework. To this end, I have focused my attention on a specifijic period in the Arabic linguistic tradition, from the 2nd/8th century to the 4th/10th century, a period in which the initial phase of the construction of an epistemological and normative frame of reference took place. The particularity of the ellipsis is born in a context outside the grammatical tradition: it has its roots in the exegetic activity of the fijirst mufassirūn. Although grammar owes a lot to the early tafsīr, the group of rules that were adopted as guidelines for the interpretation of the Quran were still far from constituting a defijined and stable terminological asset.8 Moreover, commentators did not focus only on the studying language but also the message it portrayed. In this perspective, research on the “missing word” did not use to form a syntactic instrument of research, as it will be the case in the linguistic tradition which will be inaugurated by Sībawayhi, by the interpretation key of the divine message. Below are some sample passages of the comment of Muqātil (Tafsīr): Cor. iv 160 Tafsīr: 1.422.1 Cor. xii 42 Tafsīr: 2.335.10

Wa-bi-ṣaddihim ʿan sabīli llāhi kaṯīran ‘they hindered many from Allah’s Way’ ∅ dīn al-ʾIslām wa-ʿan Muḥammad ‘from Islam and Muḥammad’ li-llāḏī ḏ̣ anna ʾannahu nājin minhumā ‘to that one whom he consider about to be saved’ ∅ min al-qatl ‘from the killing’

In these two examples the commentator is further precising the meaning of the two paragraphs which might not be so clear. The ʾiḍmār, in this case, does not have any syntactical signifijicance: the concept of ʿamal ‘operation’ is absent and the sentence structure is not taken in consideration. In the panorama of Koranic comments, the passage from an exegetic exception to the linguistics of the concept of ʾiḍmār is mediated by Maʿānī l-Qurʾān by Farrāʾ, a text which reinterprets, in theoretical-linguistical sense, the results of the exegetic experience. Without getting into the rhetoric of whether to place this comment in the genre of tafsīr or naḥw,9 it is enough to reveal that the aim of Farrāʾ is to use grammar in

8

 See Versteegh (1994).  Cf. Dévényi (1987: 102–103).

9

ellipsis in arabic linguistic thinking

73

the fijield of exegesis as he mentions at the beginning of the text (Farrāʾ Maʿānī: 13): tafsīr muškil ʾiʿrāb al-Qurʾān wa-maʿānīhi ‘explanation of the desinential vocals of the word and its meanings.’ Therefore, we can afffijirm that the origins of Maʿānī date back to a phase in which the borderline between linguistics and exegesis was quasi inexisting. Albeit the fact that Farrāʾ understood that determining the ʾiḍmār is related to theologicalinterpretative implications, he expands the confijines of his research to a more syntactical type. Dévényi (2008) has recently analyzed the use of the term ʾiḍmār in the Maʿānī: in her study, Dévényi notes that the omission of one element from the sentence should be justifijied, according to Farrāʾ, by specifijic considerations such as the syntactical coherence, another occurence in the Holy Text, an extra-textual linguistic source (kalām al-ʿArab and a poem), the grammatical rules and a qirāʾa. The two considerations which can be considered efffectively relevant in Farrāʾ’s linguistic analysis are the phrasal syntactical coherence and the presence of an occurence in the Holy Text in which the ʾiḍmār appears. At the fijirst place, the syntactical relation must be clear and coherent: wa-ʾinna-mā yaḥsunu al-ʾiḍmār fī l-kalām allāḏī yajtamiʿu wa-yadullu ʾawwaluhu ʿalā ʾaxirihi ‘the ʾiḍmār is good in the utterance in which the fijirst is a sign of the last’; another condition is the presence of other Koranic passages in which the omitted element is expressed. In this paragraph, the Kūfan grammarian analyzes the reading of a Koranic verse (Cor. ii 7) attributed to a ʿĀsim b. ʾĀbī an-Najūd (sentence b; Farrāʾ, Maʿānī: 1.22; 1.102): 1. Xatama llāh ʿalā qulūbihim wa-ʿalā samʿihim wa-  ʿalā ʾabṣārihim  ġišāwat-un veil-nom.indef

‘God has set a seal on their hearts and on their hearing, and on their eyes is a veil’ a) Xatama llāh ʿalā samʿihim wa-ʿalā qalbihim ʿalā baṣarihim ġišāwat-an b) Xatama llāh ʿalā samʿihim wa-ʿalā qalbihim wa-jaʿala ʿalā baṣarihim ġišāwat-an and-make.pf.3ms

‘God has set a seal on their hearts and on their hearing, and He made a veil on their eyes’ 2. a) Yasʾalūna-ka ʿan aš-šahri ask-you

on the month

b) Yasʾalūna-ka ʿan aš-šahri

al-ḥarāmi

qitāl-in

fī-hi

prohibited

fijighting-gen.indef

in-it

qitāl-in

fī-hi

al-ḥarāmi ʿan concerning-prep.

‘They ask thee concerning fijighting in the Prohibited Month’

74

cristina solimando

Farrāʾ legitimizes the proposition structure (1b) resorting to the ellipsis of the verb jaʿala. According to the Kūfan grammarian, both sentences are correct as they give the same information without any discrepancy in their meaning. Still, we are at an early phase, distant from the formulation and theorization of the concept of ʿamal, where the question of “case marking” was not treated openly although the diffference with early commentators is evident. Farrāʾ often refers to the omission of concrete elements that give a very precise and structurally-inscribed meaning of the phrase. Although strictly syntactic concerns are not at the center of the Farrāʾ’s observations, he makes grammar as his main instrument for a thorough analysis of the Holy Text; this renders Maʿānī l-Qurʾān a fijirst step towards the grammaticalization of the concept of ʾiḍmār and of the establishment of a theory of ellipsis which will later acquire its complete form with Sībawayhi’s Kitāb. Later in this paper, unlike what Dévényi (2008: 61) reckons on ʾiḍmār as being the obligatory supposition of an element, I will argue that grammarians use ellipsis to explain the syntax of the utterances when exceptional elements need to be explained (such as the absence of the ʿāmil). Moreover, in ellipsis-referred terminology, there is no explicit evidence that could lead to Dévényi’s conclusion: the ʾiḍmār can be jāʾiz ‘licit’ but never wājib ‘obligatory’. Obviously, ʾiḍmār and ḥaḏf refer to a diffferent linguistic operation but do not necessarily draw the diffference which Farrāʾ seems not to have grammaticalized yet. To this end, and in order to understand further the emergence of a linguistic theory of ellipsis, the Sībawayhi’s Kitāb is meticulously analyzed herein. Sībawayhi dedicates an entire section of his Kitāb to issues related to the legitimacy of the operation of supposition of one part of the discourse. Many pages were dedicated for a precise, and at times explicit, theoretical classifijication of cases of ellipsis (verbal above all). In this part, the author sets the ʾiḍmār on an intermediate level of “interface” among various levels of syntactical and semantic analysis of the sentence, where the act of communication among speakers represents one of the fundamental aspects. The expression muḍmar fī n-niyya ‘intentionally omitted’ highlights the role of the speaker10 who consciously omits a constituent of the sentence. The operation of recovering the implicit element through the tamṯīl is an operation of an exclusively syntactical nature which aim is to reconfijigure the syntactical relations and to explain the sentence structure. A very interesting aspect which Sībawayhi afffijirms repeatedly in his study,

10

 Cf. Ayoub (1990: 3).

ellipsis in arabic linguistic thinking

75

is the importance given to the communication process between the two speakers: ellipsis is considered legitimate when it does not lead to any ambiguity between the two speakers. The attention given to the reciprocal comprehension constitutes an important theoretical parameter of reference and of legitimization together with grammar, demarkating the birth of the concept of linguistics of ellipsis in an original and modern manner. The fijirst category comprises imperative verbs: This chapter dwells upon the order and the prohibition presented as an elided form of a verb whose application in an explicit form could be possible.11

I quote here some examples (Sībawayhi Kitāb: 1.128): (3) a.

al-ʾasad-a def-lion-acc

b. lā taqrab

‘the lion, the lion!’ al-ʾasad-a

Do not get closer.imp.2ms

‘do not get closer to the lion!’ (4) a.

al-jidār-a def-wall-acc

b. iḥ d̠ar

‘the wall!’ al-jidār-a

be careful .imp.2ms

‘be careful with the wall!’ In these two examples the extralinguistic concept permits the speaker to understand what the speaker wishes to say, i.e. to mind the lion, in the fijirst case, and the wall, in the second. The linguistic context allows, instead, the grammarian to return back to the missing lexical material. The fact that the relative section of the ʾiḍmār begins with the imperative verb as implicit constituent, makes us suppose that Sībawayhi is thinking about a concrete communicative act and is aware of how much the ellipsis is not a peripherical language phenomenon that one can do without. Rendering implicit an element of the phonic chain means to render the

11  Sībawayhi (Kitāb: 128): hāḏā bābu mā jarā min al-ʾamri wan-nahyi ʿalā ʾiḍmāri l-fijiʿli l-mustaʿmali ʾiḏ̣ hāruhu.

76

cristina solimando

sentence more efffective (as shown in examples 3 and 4) and consequently less heavy. The understanding between the two speakers remains an important element of study in the ellypsis even when the elided verb is not in its imperative form: This chapter dwells upon those sentences in which a verb is elided but could still be expressed even if the sentences do not comprise an order or a prohibition.12

Below are some examples (Sībawayhi Kitāb: 1.128): (5) a.

Zayd-an Zayd-acc.indef

b. li-

yaḍrib

ʿAmr-u-n

Zayd! Zayd-an

That hit.condit.3ms ʿAmr–nom-indef

‘Let Amr hit Zayd!’ (6) a.

Zayd-un Zayd-nom.indef

b. li-

yudrab

Zayd! Zayd-u-n

That hit.subj.pass.imperf.3ms

‘Let Zayd be hit’ After having reconstructed the implicit elements li-yaḍrib ʿAmrun in (5) and li-yuḍrab in (6), Sībawayhi stresses that where order is not imparted to the speaker this might result in ambiguity and thus is not permitted. The following sentence is particularly interesting as the extra-linguistic context represents the most relevant criterion for constructing the elided verb (Sībawayhi Kitāb: 1.130): (7) al-qirṭās-a def-target-acc

wa-llāh-i Oh-God-gen

‘The target, I swear by God!’

12  Sībawayhi (Kitāb: 1.129): hāḏā bābu mā yuḍmaru fīhi l-fijiʿlu l-mustaʿmalu ʾiḏ̣ hāruhu fī ġayri l-ʾamri wan-nahyi.

ellipsis in arabic linguistic thinking

77

Sībawayhi hypothesizes the verb examining two diffferent background scenarios: a. yu-ṣīb-u 3ms hit.imperf-ind

al-qirṭās-a

wa-llāh-i

def-target-acc

Oh-God-gen

‘It will hit the target, I swear by God!’ b. aṣāb-a al-qirṭās-a wa-llāh-i hit.perf-3ms

def-target-acc

Oh-God-gen

‘It hit the target, I swear by God!’ The choice of the tense of the verb is brought back to the extra-linguistic context: if the sound of the arrow that hits the target is heard then the interlocutor understands that the verb ‘hit’ was used in its past tense; otherwise the speaker is expressing an action that still needs to take place and thus the verb is in muḍāriʿ. It is interesting to note that the recoverability of the referent, as a required condition for the use of ellipsis, is not a sufffijicient condition in the case of the ellipsis of the particle, such as in the following sentence: (8) a.

aṭ-ṭarīq-i det-street-gen

*

b. tanaḥ ḥ a Move away.imp.2ms

ʿan

‘the street’ aṭ-ṭarīq-i

from

‘move away from the street’ The particle is easily reconstructible due to the extra-linguistic context and to the linguistic structure, as the verb needs the particle ʿan and ṭarīq is in genitive. Nonetheless, according to Sībawayhi, it constitutes a linguistic unity that is inseparable from the term from which it is held up, and for that reason it cannot be elided. In line with the theoretical route of the term ʾiḍmār it is necessary to examine the terminological facet. In the Kitāb the diffference between syntactic, morphological and phonetic levels is precised where terminology reflects such coherence: using the term ʾiḍmār to refer to the cases of ellipsis is almost constant, while ḥaḏf remains bound to phonological issues. However, there are passages that pose some difffijiculties: at the beginning of my research I was intrigued by a chapter entitled hāḏā bābun yuḥdafu minhu l-fijiʿlu li-kaṯratihi fī kalāmihim ḥattā ṣāra bi-manzilati l-maṯal, ‘a chapter on verbs that are deleted in the used expressions very often to

78

cristina solimando

become proverbs’ where Sībawayhi uses the term yuḥḏaf, ‘was omitted’. in the long section dedicated to the morpho-phonologic issues the grammarian uses yuḥḏaf to refer to the occurrence of a phenomenon which is then used in all the treaty. It is interesting to note how from a theoretical point of view the analysis of the ellipsis remains unchanged, as could be noted in the following sentence (Sībawayhi Kitāb: 1.144): ʾamr-a-n

(9) a. Yā fulān-u Oh fellow-nom

qāṣid-a-n

thing-acc-indef pleasant-acc-indef

‘Oh man, a pleasant thing!’ And they continue to be studied according to the same principle: b. intahi do.imp.2ms

yā fulān-u

ʾamr-a-n

qāṣid-a-n

oh fellow-nom

thing-acc-indef

pleasant-acc-indef

‘Do, oh man, a pleasant thing!’ While the concept of ellipsis remains unchanged in relation to the previous chapters of the section in which such notion is studied, the terminology used seems to undergo a radical change where the grammarian had given a theoretical status that is very precise and coherent to this operation. Thus, the interrogative is the reason why this change occurs: examining the entire chapter we note that the term ḥaḏf is always associated with the expression li-kaṯratihi fī kalāmihim. It is possible to hypothesize that the frequent use of a sentence provokes a sort of “attrition” leading to dropping one of its components in a completely independent manner by the free will of the speaker, whom we have seen being a prerogative of the ʾiḍmār. Surely ḥaḏf is not used randomly, as the distinction between the two terms for Sībawyhi is very clear. Examining the mechanisms of the verbal ellipsis, the grammarian uses, though very rarely, a very interesting expression: yajūz or lā yajūz ḥaḏf al-fijiʿl wa-ʾiḍmāruhu. This means that ʾiḍmār and ḥaḏf cannot be used in a synonymic way as a precise and distinct operation corresponds to each of the two terms: supposition in the fijirst case and deletion in the second. The meaning of ‘deletion’ seems to be confijirmed by the fact that Sībawayhi stresses that the omission of one part of the discourse corresponds to the needs to lighten the sentence: istixfāfan or taxfīfan ‘lightening,’ a formulation that is usually used to indicate phonetic lightness but in this case it indicates the necessity of slimness in the frequently used expressions which subsequently turn into proverbs.

ellipsis in arabic linguistic thinking

79

Although the Kitāb does not present neither a solution that is exclusively dedicated to the ellipsis (except for the verbal ellipsis which results more compact) nor a theoretical exposition regarding the numerous examples that prove the the importance that this operation occupies among the linguistico-theorical mechanisms. At this point, one might wonder if Sībawayhi was anticipating the future terminological use by abandoning the term ʾiḍmār and using the term ḥaḏf to indicate the omission of a certain constituent within an utterance. An overview on the developement of the grammatical theory of the ellipsis requires the analysis of Mubarrad’s Muqtaḍab which played a central role in the transmission of Sībawayhi’s linguistic thinking. Although in this text there isn’t any section that is dedicated to the ellipsis, it still represents a fundamental instrument of linguistic analysis. It does not fijind the collocation it had in the Kitāb. Mubarrad uses the term ʾiḍmār to refer to the verbal ellipsis and the “pronominalization process” as it appears in the chapters ʾiḍmār jamʿ al-muḏakkar ‘pronominalization of the masculin plural’, or ʾiḍmār al-mutakallim ‘pronominalization of the speaker’. The terms ḥaḏf and ʾiḍmār, used to refer to the elliptical mechanism, are used without reference to the theoretical distinction we found in the Kitāb. (10) ʾA-

ʿišr-ūna

interr twenty-nom.indef

ʾA-

ʿišrūna

ġulām-a-n

ġilmān-u-

-ka

boy.pl-nom.det

-2ms

ġilmānu-

-ka

boy-acc-indef

‘Are these twenty boy your boys?’ (11) Bi-

kam

rajul-i-n

For

How many

man-gen-indef

Bi-

kam

min

rajul-in

of

‘How many men?’ In both cases (Mubarrad Muqtaḍab: 3.56–57) the constituents of the fijirst sentence represent the instrument which permits to recover the elided lexical material: mā ʾaḏ̣ harta dalīlun ʿalā mā ḥaḏafta ‘what you expressed is the sign of what you elided’. Mubarrad breaks the dicotomy ʾiḍmār— ʾiḏ̣ hār using ḥaḏafa as opposed to ʾaḏ̣ hara. Here the dalīl ‘sign’ can be morpho-semantic, as in the fijirst case, or syntactical as in (11) . In reference to the use of the preposition kam Mubarrad (Muqtaḍab: 3.56–57) says:

80

cristina solimando they made of the particle governing the genitive tat operates on kam [i.e., the particle bi-] the sign of min and they deleted it. Without this particle, it is beyond dispute that the ʾiḍmār could not be licit.13

From a theoretical point of view, we could observe through the work of Mubarrad how ellipsis moves from a partially semantic concept to a morphonological one. This change is linked to the gradual terminological change form ʾiḍmār to ḥaḏf which use was limited to the phonological level. The rules that legitimize the elliptical process are the same: the reciprocal understanding and recoverability obtained through the analysis of the explicit constituents are dalīl ʿalā mā ḥuḏifa ‘sign of what has been deleted’. With philosophers of language of the 4th/10th century, and in particular with Ibn Jinnī, we fijind a terminological stability and ʾiḍmār- ḥaḏf will be use as synonyms. Mehiri pointed out the originality of Ibn Jinnī’s treatment of ellipsis which consisted of the particular attention given to the extra-linguistic context (ḥāl). As shown before, Sībawayhi considered the extra-linguistic context as a necessary condition for the deletion process. The innovation in Ibn Jinnī’s thought is rather the operation of terminological stabilization and the renewed methodological strictness: in the chapters dedicated to the analysis of the taqdīr the grammarian uses the term ḥaḏf to refer to all cases of supposition that Sībawayhi would have defijined as ʾiḍmār. An interesting aspect to highlight is the rare occurrence of the term ʾiḍmār: 8 times in the Xaṣāʾiṣ, against 215 in the Kitāb; moreover, in none of these 8 cases the ellipsis is the main object of study. The evidence of this new approach is in the chapter Bāb fī sajāʿat al-ʿarabiyya ‘chapter on the courage of the Arabic language’ (Ibn Jinnī Xaṣāʾiṣ: 2.360): Know that most of this consists of the ellipsis, in the addition, the transposition, the “constructio ad sensum”, and the alteration. The Arabs can delete a sentence, a single word, a letter and a vowel. Nothing of that can occur without a sign, otherwise it would be just a conjecture. The ellipsis of a single word can take place under three forms: noun, verb and particle.14

13  Yajʿalūna mā daxala ʿalā kam min ḥurūfiji l-xafḍi dalīlan ʿalā min wa-yaḥḏafūnahā. Fa-ʾiḏā lam yadxalhā ḥarfu al-xafḍ fa-lā xtilāfa fī ʾannahu lā yajūzu l-ʾiḍmār. 14  Iʿlam ʾanna muʿḏ̣ am ḏālika ʾinnamā huwa l-ḥaḏf waz-ziyāda wat-taqdīm wat-taʾxīr wal-ḥaml ʿalā l-maʿnā wat-taḥrīf. Al-ḥaḏf qad ḥaḏafat al-ʿArab al-jumla wal-mufrad walḥarf wal-ḥaraka. Wa-laysa šayʾun min ḏālika ʾillā ʿan dalīl ʿalayhi. Wa-ʾillā kāna fīhi ḍarb min taklīf ʿilm al-ġayb fī maʿrifatihi. Wa-ʾammā ḥaḏfu l-mufrad fa-ʿalā ṯalāṯati ʾaḍrāb: ism wa-fijiʿl wa-ḥarf.

ellipsis in arabic linguistic thinking

81

Ḥaḏf refers to any operation of phonological or morpho-syntactical deletion, and the distinction between supposition and deletion that is vey evident in the Kitāb seems to disappear. Sībawayhi used the term ʾiḍmār to refer to the verbal ellipsis (for nouns and particles the terminology is much more variable) and he uses ḥaḏf only in case of verbal supposition in proverbs or with ʾiḍmār to indicate a diffferent linguistic operation, as highlighted above. Ibn Jinnī establishes the synonymity between the two terms and ʾiḍmār remained in use as the usual term to indicate “deletion,” as a synonym of ḥaḏf also in later traditions. He can be considered the grammarian who, after Sībawayhi, gave a major contribution to the edifijication of the theory of the ellipsis: he renewed the approach to this linguistic mechanism through its application to all parts of the discourse. He arranged all cases of ellipsis under the term ḥaḏf assigning to it the value of ʾiḍmār. It is signifijicant that the ellipsis is considered a characteristic of the Arabic language insomuch as being included, like the taqdīm and the taʾxīr, among the linguistic mechanisms that make out of the arabic language a dynamic and flexible one. In conclusion, we can afffijirm that the principles of applicability of the ellipsis as they were theorized by Sībawayhi remained unchanged for two centuries. The extra-linguistic context and the syntactical recoverability of the sentence structure are the fundamental conditions of the supposition of the lexical material. The reciprocal understanding between the two interlocuters, as prerogative of a communicative act, is one of the most important parameters of legitimacy for the supposition of one or more constituents in a sentence. Sībawayhi had clarifijied the distinction between the two operations of deletion and supposition; with Ibn Jinnī the ellipsis enters a coherent and fijixed theoretical system together with a theorical generalization in which the drop of a phoneme, the deletion of a part of the discourse and its supposition are assimilated and conceived as a single operation.

References 1. Primary Sources Farrāʾ. Maʿānī. Maʿānī l-Qurʼān. Bayrūt: Dār al-Kutub al- ʿilmiyya, n.d. Ibn Jinnī. Xaṣāʾiṣ. Al-Xaṣāʾiṣ, 3 vols. Bayrūt: Dār al-Kitāb al-ʿArabī, 1952–56. Mubarrad. Muqṭaḍab. Al-Muqṭaḍab, 4 vols. Αl-Qāhira: Dār at-Taḥrīr, 1965–68.

82

cristina solimando

Muqātil. Tafsīr. At-Tafsīr, 4 vols. Al-Qāhira: al-Ḥayʾa al-Miṣriyya al-ʿĀmma lil-kitāb, 1980–1987. Sībawayhi. Kitāb. Kitāb, 2 vols. Būlāq: al-Maktaba al-Kubrā al-ʾAmīriyya, 1316h. 2. Secondary Literature Ayoub, Georgine. 1990. “De ce qui ‘ne se dit pas’ dans le Livre de Sībawayhi: la notion de tamṯīl”. Studies in the History of Arabic Grammar II, ed. K. Versteegh and Michael G. Carter. Amsterdam: John Benjamins, 1–15. Bohas, Georges. 2001. “Le rasoir d’Occam et la tradition grammaticale arabe”. Arabica 48, 1–19. Chao, W. 1988. On Ellipsis. New York: Garland. Dévényi, Kinga. 1987. “Farrāʾ’s linguistic methods.” Studies in the History of Arabic Grammar, vol. 2, Proceedings of the second Symposium on the History of Arabic Grammar, Nijmegen, 27 april–1 may 1987, 101–110. ——. 2008. “ʾIḍmār in the Maʿānī of al-Farrāʾ”. Approaches to Arabic linguistics presented to Kees Versteegh on the occasion of his sixtieth birthday, ed. Everhard Ditters and Harald Motzki. Leiden: Brill, 45–65. Dik, Simon C. 1997. The Theory of Functional Grammar. Berlin: de Gruyter. Kuno, Susumu. 1987. Functional Syntax: Anaphora, Discourse and Empathy. Chicago: University of Chicago Press. Lambrecht, Knud. 1994. Information Structure and Sentence Form: Topic, Focus and the Mental Representation of Discourse Referents. Cambridge: Cambridge University Press. McShane, Marjorie J. 2005. A Theory of Ellipsis. Oxford and New York: Oxford University Press. Versteegh, Kees. 1994. “The notion of ‘underlying levels’ in the Arabic grammatical tradition”. Historiographia Linguistica 21, 271–296.

SECTION II

THE WORD IN THE ARABIC RHETORIC TRADITION

‘WORD’ IN THE LINGUISTIC THINKING OF ʿABD AL-QĀHIR AL-JURJĀNĪ Antonella Ghersetti

Le mot, c’est beaucoup plus et c’est beaucoup moins que la phrase. (Paul Ricoeur 1969: 92)

It is well known that the notion of ‘word’, whose most immediate correspondent in Arabic is kalima, is almost impossible to defijine in linguistics, even if it is a unit very clear to the consciousness of the speakers. If in traditional linguistics a plurality of defijinitions has been put forward,1 structural linguistics denies the possibility of attaining a satisfactory and unequivocal defijinition of such an intuitive notion. The Arab grammarians of the classical period, at least the “ancients” (al-mutaqaddimūn) did not feel the need to advance a rigorous defijinition of this term.2 In any case, the intuitive notion is so widely spread and the strictly linguistic meaning of this term is so difffijicult to establish that, for instance, the Encyclopaedia of Islam, the reference work for Arabists, deserves to the entry kalima an article completely revolving about the philosophical and theological sense of kalima, completely omitting the linguistic treatment.3 Incidentally, this also confijirms the impression that Arabic linguistics, nonetheless so basic in the fijield of Arabic studies, is often neglected in this work. On the contrary, the entry kalima in the recently published (and most welcome to scholars interested in Arabic linguistics) Encyclopaedia of Arabic Language and Linguistics offfers a detailed description of the theoretical

1

 ‘Word’ has been variously defijined as “the union of a particular meaning with a particular complex of sounds capable of a particular grammatical employment”; “a unit separated by blanks in writing”; “a particular group of morphemes with a formal individuality”; “a meaningful linguistic unit composed by one or more phonemes, keeping totally or partially its form in the diffferent syntagmatic uses”, or, in the classical defijinition of Bloomfijield “a minimal free form” (for a discussion of the notion of ‘word’ see Lyons 1968: 194–206). Nonetheless, some basic features remain at the core of this problematic defijinitions, such as the notion of “cohesion” (i.e. ‘position mobility’ and ‘impossibility to break up’), “internal stability” and “meaningfulness”. 2  On the contrary, they were basically interested in categorizing the words and identifying the partes orationis. 3  MacDonald & Gardet (1978).

86

antonella ghersetti

questions and of the criteria of classifijication implied in the defijinition of this term, as well as of its two basic meanings (‘word’ and ‘morpheme’).4 We can safely state that in the Arabic grammatical tradition almost two points are clearly established: the fijirst is that the notion of kalima can be situated between the two notions of ‘word’ and ‘morpheme’ and the second that the classifijication of words (as well as the defijinitions of its boundaries) rests almostly exclusively on syntactic criteria.5 But are these points true also in the case of a grammarian/rhetorician as ʿAbd al-Qāhir al-Jurjānī “the founder of grammatical semantics”6 whose approach is as original as it is sometimes conflicting with that of his colleagues? Was the “double meaning” of kalima (‘morpheme’ and ‘word’) taken into consideration by him? Have syntactic criteria the same signifijicance to him they have in the more traditional analysis of his colleagues? Or, is there any other criterion important, and perhaps more important, in his way of defijining and treating the notion of ‘word’? And, fijinally, was this notion really basic in his theoretical system? We intend to investigate in Jurjānī’s linguistic thinking to see how ‘word’7 was conceived, if this concept was major to him and if there is a diffferent approach to the matter in his grammatical and in his rhetorical works.8 In particular, the hypothesis to be tested is that of his greater attention to semantic criteria in this connection, the syntactic and morphological criteria remaining in the background, thus revealing what could be characterized as a “revolutionary” approach to linguistic analysis.9 Among the grammar books written by Jurjānī there are two concise syntactic treatises explicitely conceived as having a didactic aim in view: al-ʿAwāmil al-miʾa (Jurjānī ʿAwāmil) and al-Jumal, each concerning the functioning of government (ʿamal). The fijirst of the two, that is nowadays 4

 Levin (2007).  Levin (2007) and Owens (1989). 6  Kouloughli (2007: 627). 7  We must specify that in our corpus ‘word’ normally corresponds to kalima; but sometimes this notion is expressed by other lexemes such as lafḏ̣ , or has no direct correspondent and must be deduced from elliptical expressions. 8  See Baalbaki (1983: 10): “In fact that Jurjānī’s approach in naḥw difffers from his approach in balāġa, in spite of the apparent similarity between a number of the issues these two subjects treat, is further proof that those who exaggerate the influence of grammarians on rhetoricians actually misrepresent the special relation between the two subjects”, even if it must be underlined that “il n’existe pas entre les disciplines de cloison étanche et qu’il est vain de les étudier séparément ou, pis, d’élargir subrepticement le domaine de l’une au détriment des autres” (Larcher 1993: 250). 9  We hereby recall the terms “révolution jurjanienne” put forward by Kouloughli (1983). 5

‘word’ in the linguistic thinking of jurjānī

87

almost unobtainable in spite of the renown it had in the past,10 consists of a list of the governing operators (ʿawāmil). They are arranged following a double criterion: fijirst they are divided into formal (lafḏ̣ iyya) or abstract (maʿnawiyya), and then those pertaining to this second category are divided into analogic, or productive (qiyāsiyya), and attested, or nonproductive (samāʿiyya). At fijirst glance, the global reading of the text permitted us to see that ‘word’ (kalima) does appear only once and in a non-signifijicative context, while its hyponyms (the specifijic terms identifying the parts of speech) are rather used in the course of the exposition. The second treatise, al-Jumal,11 is devoted to the same subject, but with a diffferent arrangement of the matter. This follows the traditional categorization of the three parts of speech in Arabic grammar: fijiʿl, ḥarf and ism are treated in this very same order which corresponds to their qualifijication to hold the fonction of governing operators. The verb is the fijirst one, since it is the foundation (ʾaṣl) of the government, the governing operator par excellence. The fijirst chapter, which contains the preliminary elements (muqaddimāt) and the last one, devoted to “various” elements, complete this work. Being the declension (ʾiʿrāb) the main subject of this short book, we must not expect to fijind in it a structured defijinition of ‘word’, nor a discussion of the notion. But if we take into consideration the use Jurjānī makes of this term and the context where he uses it, we can sense what ‘word’ (kalima) is to our author. The fijirst occurrence of kalima, in the plural (kalimāt), is found at the very beginning of the treatise that starts—in the traditional way—by introducing the three parts of speech. Here, instead of the canonical sentence al-kalāmu smun wa-fijiʿlun wa-ḥarf mentioned by Mubarrad, Ibn Sarrāj and Zajjājī among others,12 we fijind

10  Actually a modern edition does exist, but it is not so much the work by Jurjānī as a commentary that incorporates it without making any diffference between the annotated text and the exegetic part. On this edition see Gilliot (1983, for the fijirst print) and Larcher (1993, for the second print); on the history of al-Jumal see also Troupeau (1963). 11  Jurjānī ( Jumal). 12  Zajjājī (ʾĪdāḥ: 41) mentions this expression as an object of all the grammarians’ general agreement (ʾijmāʿu n-naḥwiyyīn): fa-ʾawwalu mā naḏkuru min ḏālika ʿalā ʾanna l-kalāma smun wa-fijiʿlun wa-ḥarf. Ibn Sarrāj (ʾUṣūl: 1.36), one of the closest pupils of Mubarrad, also uses the term kalām: al-kalāmu yaʾtalifu min ṯalāṯati ʾašyāʾ . . ., even if he stresses that kalām is composed of the three parts of speech. In fact kalām is the word Mubarrad had recourse to in Muqtaḍab (2.141). The dictionary of technical terms by Xuwārazmī (Mafātīḥ: 29) makes an intermediate choice between the canonical defijinition and the version of Ibn Sarrāj, (al-kalāmu ṯalāṯatu ʾašyāʾ . . . ) and the word yaʾtalifu used by Ibn Sarrāj is erased. Zajjājī mentions the word kalām when speaking of the followers of Sībawayhi (ʾaṣḥāb Sībawayhi). But in another passage (ʾĪḍāḥ: 42) he quotes exactly the passage of the Kitāb where the term kalim appears instead and attributes to the caliph ʿAlī b. ʾAbī Ṭālib the

88

antonella ghersetti

iʿlam ʾanna l-kalimāti ṯalātun: ismun wa-fijiʿlun wa-ḥarf ( Jumal: 5) where the plural kalimāt is used instead of the term kalām. This formal distinction between kalām and kalima, which had already been drawn explicitely by Ibn Jinnī, in fact goes back to Sībawayhi, who uses the term kalim rather than kalām.13 Jurjānī intentionally follows Sībawayhi’s terminological choice, and this is motivated by his particular concept of kalima vs kalām. To him in fact ‘kalām’ is what results from a process of combination (iʾtilāf ) of those discrete units (kalimāt) that the grammarians usually divide into three categories.14 This is not accidental: his choice on the contrary is a signifijicant hint at what Versteegh (1992: 117) defijines as “attention to the terminology”. As a consequence, the other occurrences of the term kalima in al-Jumal refer to an intuitive notion of word as a discrete unit from the morphological and syntactic point of view, since semantics is not taken into consideration in this very concise treatise.15 This is very clear when he ( Jumal: 6) gives the canonical defijinition of ʾiʿrāb which insists on the fijinal position of the ʾiʿrāb markers (ḥaddu l-ʾiʿrābi ʾan taxtalifa ʾāxiru l-kalimati bi-xtilāfiji l-ʿawāmil): here the term kalima is clearly referred to as a discrete unit whose end is afffected by the variations depending on its syntactic status. Sometimes, nevertheless, this variation can afffect the totality of the word, as in the case of the ‘implicit’ declension (al-ʾiʿrāb ġayr aṣ-ṣarīḥ), that implies a completely diffferent form of the word for diffferent syntactic positions. This is the case of the second singular person

choice of the term kalām (ʾĪḍāḥ: 43). This sentence is the one mentioned in the anecdotes concerning the origins of grammar in connection with the caliph ʿAlī: see e.g. Marzubānī (Nūr: 3, al-kalāmu kulluhu smun wa-fijiʿlun wa-ḥarf ) and Ibn ʾAnbārī (Nuzha: 4, al-kalāmu kulluhu smun wa-fijiʿlun wa-ḥarf ). 13  In commenting the text of Sībawayhi, Ibn Jinnī (Xaṣāʾiṣ: 1.25) precises that the plural kalim refers to countable units (. . . ṯalātati ʾašyāʾa maxṣūṣatin wa-hiya l-ismu wal-fijiʿlu wal-ḥarf ) while kalām rather refers to the overall totality of words (al-jamʿ). That’s why Sībawayhi preferred to use kalim referring to the three parts of speech. 14  This seems to herald the principle, openly stated in Dalāʾil al-ʾiʿjāz, that the individual words as concrete units of analysis are not conceivable outside the frame of the utterance. See e.g. the following statement: “words (i.e. noun, verb and particle) individually taken are not an utterance nor a poem if they are not afffected by the textual organization, which is ultimately the implementation of their syntactic functions and status” (lā takūnu l-kalimu l-mufradatu llatī hiya ʾasmāʾun wa-ʾafʿālun wa-ḥurūfun kalāman wa-šiʿran min ġayri ʾan yuḥdaṯa fīhā n-naḏ̣ mu llaḏī ḥaqīqatuhu tawaxxī māʿānī n-naḥwi wa-ʾaḥkāmihi; Jurjānī Dalāʾil: 425). 15  Which is very typical of the Arabic traditional grammar considering that, as Kouloughli (2007: 627) puts it “it is true, however, that the permanent concern of Arab grammarians has been to account for the mechanism of case assignement and that, in the course of this endeavour, the recourse to meaning has always remained essentially accessory”.

‘word’ in the linguistic thinking of jurjānī

89

pronoun where ʾanta stands for the subject case and ʾiyyāka for the object,16 a case where there is no (total or partial) formal similarity in its syntagmatic uses. In this reasoning we can recognize the professional bias of the grammarian, whose concern is to explain the diffferent linguistic phenomena on the basis of the declension i.e. ultimately on morphosyntax, which implies a somehow forced identifijication between two linguistic concrete units (ʾanta and ʾiyyāka) that could hardly be defijined as “the same word” if considered in a purely formal (phonological or orthographic) perspective. To get to a more explicit defijinition of the concept of word (kalima), a very concise one indeed, the reader has to await the end of the treatise where the generalities are presented. Here Jurjānī afffijirms that “each of these categories: noun, verb and particle, is called kalima and when two of these elements combine and are so informative as in the case of xaraja Zaydun, they are called kalām and also jumla”.17 In this context kalima appears as a discrete unit that encompasses the three parts of speech and contributes to the meaning of a wider unit, the sentence. Beyond these occurrences, no passage is devoted to the discussion or the defijinition of this concept. It is in an other grammatical treatise that we can best see what ‘word’ represents to our author: al-Muqtaṣid, the “medium” version of the commentary Jurjānī composed on the Kitāb al-ʾĪḍāḥ by ʾAbū ʿAlī al-Fārisī. This title of ʾAbū ʿAlī enjoyed a great celebrity, which is witnessed by the great number of manuscripts preserved and by the considerable quantity of commentaries devoted to it. ʾAbū ʿAlī al-Fārisī, one of the most renown grammarians of the 4th/10th century, was the pupil of Ibn Sarrāj and of Zajjāj. He also was, even if in an indirect way, the master of Jurjānī, who studied with his nephew ʾAbū l-Ḥusayn. ʾAbū ʿAlī, a very ‘canonical’ grammarian, played an important role in the difffusion of the Kitāb of Sībawayhi: actually his collation stands at the basis of the “oriental branch”, represented by Zamaxšarī’s recension, in the transmission of the Kitāb. As Humbert (1995: 77) explains, “ʾAbū ʿAlī s’inscrit [. . .] de façon privilégiée dans une chaîne de transmission dont al-Mubarrad est le point de départ”. The interest of Jurjānī’s commentary of al-ʾĪḍāḥ

16  Al-ʾiʿrābu ġayru ṣ-ṣarīḥi ʾan takūna l-kalimatu mawḍūʿatan ʿalā wajhin maxṣūṣin mina l-ʾiʿrābi wa-ḏālika fī l-muḍmari naḥwa ʾanta fa-ʾinnahu wuḍiʿa lil-marfūʿi wa-ʾiyyāka lilmanṣūb ( Jumal: 38). 17  Iʿlam ʾanna l-wāḥida mina l-ismi wal-fijiʿli wal-ḥarfiji yusammā kalimatan fa-ʾiḏā ʾtalafa minhā ṯnāni fa-ʾafādā naḥwa xaraja Zaydun summiya kalāman wa-summiya jumla ( Jumal: 40).

90

antonella ghersetti

derives from its “intermediate” position or, as Larcher (1993: 250) puts it, from the fact that it is one of the works that “font le lien entre «Anciens» (al-mutaqaddimūn) et “Modernes” (al-mutaʾaxxirūn) . . . ou entre période classique et post-classique . . .”.18 But this treatise arouses our interest for another, and perhaps more important, reason: although al-Muqtaṣid is very orthodox in the grammatical analysis, it reveals its author’s peculiar approach to linguistic analysis, an approach which is fully developped in Jurjānī’s two books on rhetoric, Dalāʾil al-ʾiʿjāz and ʾAsrār al-balāġa. In this connection, we would like to recall the statement of Versteegh (1992: 130): “even as a grammarian he operated within a diffferent theoretical framework from his fellow grammarians”. The notion of word (kalima) is outlined just at the beginning of the commentary, where Jurjānī discusses the sentence of ʾAbū ʿAlī al-kalām yaʾtalifu min ṯalāṯati ʾašyāʾa smin wa-fijiʿlin wa-ḥarf, who repeats the very same expression used by Ibn Sarrāj (ʾUṣūl: 1.36). Jurjānī openly criticizes the habit of many “Ancients” (al-kaṯīr min al-mutaqaddimīn), who used to establish a sort of equivalence between the three parts of speech and the term kalām by using the terms as synonymous. On the contrary, he appreciates the lexical choice of ʾAbū ʿAlī al-Fārisī, who stresses that kalām is the result of a combination (he says al-kalām yaʾtalifu min . . . ) and of Sībawayhi, who has recourse to kalim instead of kalām. Jurjānī precises that the term kalām is suitable only for a linguistic unit based on the combination (iʾtilāf ) of single discrete units called ‘parts of speech’ or ‘words’ (ʾalfāḏ̣ al-kalām ʾaw ʾajzāʾ al-kalām . . . ʾawi l-kalim) and that these, taken separately, do not communicate information (ʾifāda). In this context kalima seems to be used simply as the equivalent of each of the three canonical partes orationis.19 Nothing new is said so far: the relation of hypernymy that links the three terms identifying the parts of speech to kalima is repeated. What is perhaps more interesting from this viewpoint is the statement Jurjānī made concerning the problematic description of ḥarf of Sībawayhi, that he quotes as mā jāʾa li-maʿnan laysa ġayru referring to the current defijinition given by grammarians.20 Considering the criteria upon which the diffference between nouns and particles is made, he says that nouns

18

 For a general evaluation of this work see also Versteegh (1992: 114–115; 117–118).  Wa-qawluka Zaydun wa-xaraja min ġayri smin ġayru mufīdin wa-kaḏā kullu juzʾini nfarada kāna ʿāriyan min al-ʾifāda, where juzʾ is used as a synonym of kalima (wal-kalimatu taqaʿu ʿalā kulli juzʾin ḥarfan kāna ʾawi sman ʾaw fijiʿl, Muqtaṣid: 1.68–69). 20  Muqtaṣid: 1.84; Sībawayhi says jāʾa li-maʿnan laysa bi-smin wa-lā-fijiʿl (Kitāb: 1.4) which is literally quoted by Abū ʿAlī al-Fārisī. 19

‘word’ in the linguistic thinking of jurjānī

91

would have a single meaning (maʿnā mufrad) just as the particles have,21 but they also have an additional meaning: that of the declension (maʿnā t-taṣarruf ), i.e. the syntactic function of the word whose formal marker is the ʾiʿrāb. All this contributes to make the meaning of the nouns ‘declinable’ (maʿnā munṣarif ). It is just this feature, i.e. the additional meaning, that makes the diffference, considering that the particles do not have but a single meaning.22 Jurjānī then mentions the existence of a declinable meaning (maʿnā munṣarif ) whose marker is the declension, and of a nondeclinable meaning (maʿnā ġayr munṣarif ). On this semantic criterion rest two of the three conditions he considers necessary in the defijinition of ḥarf. They are 1) a non-declinable meaning; 2) the lack of ʾiʿrāb being the formal marker of n. 1; and 3) the lack of the notion of time, which is typical of verbs; the three must coexist.23 Furthermore, ʾiʿrāb, which is peculiar to nouns, is here considered as informative and as a function of the communicative end and of the global meaning of the utterance. Jurjānī is rather explicit in this connection and states that the aims and meanings vary in the word ‘Zayd’ due to the variations afffecting the end of this word.24 Thus he links the formal expression of syntactic functions directly to the sphere of semantics and communicative intentions, and sets the analysis in a communicative perspective. Basing his discussion on the etymology of the term, Jurjānī goes on afffijirming that ʾiʿrāb corresponds to ‘elucidation of meanings’ (ʾīḍāḥ al-maʿānī) and (Muqtaṣid: 1.97) goes so far as to state that when this fijinal variation in the words (taġayyur ʾāxir al-kalima) is lacking, the sentence remains obscure and ambiguous.25 Anyway, this

21  The verb is not relevant in this discussion, since its meaning has already been defijined as the result of the combination between a single meaning (maʿnā mufrad) and the notion of time. 22  ʾA-lā tarā ʾanna Zaydan war-rajula wal-farasa yajiʾu kullu wāḥidin min ḏālika li-maʿnan mufradin? Fal-jawābu bi-ʾanna maqṣūdahum fī ḏālika li-maʿnan ġayri munṣarifijin wa-qawluhum laysa ġayru yadullu ʿalā ḏālika. Wa-maʿnā t-taṣarrufiji ʾan yakūna fāʿilan wa-mafʿūlan wa-muḍāfan ʾilayhi taqūlu ḍaraba Zaydun, wa-ḍarabtu Zaydan, wa-jāʾanī ġulāmu Zaydin fa-taxtalifu l-maqāṣidu wal-maʿānī fī Zaydin bi-xtilāfiji ʾāxirihi wa-lā yakūnu hāḏā fī l-ḥarfiji li-ʾanna qawlaka hal yadullu ʿalā l-istifhāmi wa-bal ʿalā l-istidrāk . . . wa-lā yakūnu fīhā šayʾun mina t-taṣarruf (Muqtaṣid: 1.84–85). 23  Wal-ḥarfu mā dalla ʿalā maʾnan ġayri mutaṣarrifijin wa-lam yakun lahu ʾiʿrābun biwajhin wa-lā yataḍammanu z-zamān (Muqtaṣid: 1.85). 24  Fa-taxtalifu l-maqāṣidu wal-māʿānī fī Zaydin bi-xtilāfiji ʾāxirihi (Muqtaṣid: 1.85). 25  ʾAʿraba, in the most common acceptation, means ‘to speak clearly, to express clearly, to manifest’. It is in fact presented as the IV form of the verb ʿaruba ‘to be upset, unwell, troubled [of the stomach]’ with a privative value due to the prefijixation of hamzat as-salb and in this sense the verb ʾaʿraba would acquire the meaning of ‘to put an end to corruption [in the speech]’ (ʾizālat al-fasād, Muqtaṣid: 1.98).

92

antonella ghersetti

is an assertion that is often found in the grammatical treatises, when the authors feel the need to justify the study of grammar, and declension in particular, which was seen as a burden and a petty detail by common people.26 Now, this statement remains questionable, since it is based on the principle that declension is a distinctive feature of Arabic (which is a matter open to question) and reflects more a dogmatic position than a linguistic reality. Nonetheless, it hints at the importance of the semantic component in the thinking of Jurjānī who has recourse to the sphere of meaning to justify and explain ʾiʿrāb. He considers it as pertaining to meaning and not to form,27 even if by common consent the declension is normally considered as the marker of the boundaries of the word, and consequently as a constituent of the form.28 Jurjānī actually does not object to the canonical defijinition of ʾiʿrāb as a formal variation, but strongly emphasizes its semantic value. In Jurjānī’s opinion, thus, the overall meaning of the declinable word is the result of the lexical meaning plus the grammatical meaning. This concept announces, even if implicitely, the basic principle expressed in Dalāʾil al-ʾiʿjāz: the pure lexical meaning is mere abstraction since each word carries its proper meaning plus an additional meaning that is the syntactic function expressed, in the case of nouns and imperfect tense verbs, by the ʾiʿrāb. This is tantamount to saying that word, as a discrete unit, is nothing else than mere abstraction: words in fact are always used in a context and in the scope of communication. The relevance of the semantic component in the defijinition of ‘word’ is stressed in the course of a digression where some words having a problematic classifijication are presented. It is the case of the passage about

26  See e.g. Zajjājī (ʾĪḍāḥ: 96), who claims that, even if in ordinary speech people speak without having recourse to declension, this is nevertheless necessary in order to speak correctly and to clarify ambiguous meanings; pp. 69 fff. are also devoted to accounting for the necessity of the declension to avoid ambiguity. These are hints, very rare indeed in traditional grammatical treatises, at everyday speech. On that and on the hints at freedom in word order, which cast dome doubts on the inherent character of declension, see Versteegh (1995: 167, fn. 11; 105, n. 2, respectively). 27  See e.g. statements as al-ʾiʿrābu fī l-ḥaqīqati maʿnan wa-lā lafḏ̣ (Muqtaṣid: 1.98) and qad ʾaṯbatnā ʾanna l-ʾiʿrāba ʿibāratun ʿan maʿnan yaḥṣulu bil-ḥarakāti wal-ḥurūf (Muqtaṣid: 1.101). The correlation between case endings and semantic functions had also been stressed by earlier grammarians (see e.g. Zajjājī ʾĪḍāḥ: 69) but was progressively rejected by later grammarians whose analysis was characterized by an increasing formalistic approach. See also Zamaxšarī (Mufaṣṣal: 10), on the diffferent kinds of ʾiʿrāb, where he afffijirms that each case is a ‘marker of meaning’ (ʿalam ʿalā maʿnan), raf ʿ being for instance the marker of the subject etc. 28  One of the sine qua non conditions for the defijinition of ʾiʿrāb is its occurrence at the end of the word, the others being the variation and the fact that this variation is functional.

‘word’ in the linguistic thinking of jurjānī

93

ruwayda, which is a noun but governs as a verb (ismun yaʿmalu ʿamala l-fijiʿl), where Jurjānī deals with the canonical classifijication of the parts of speech. Here (Muqtaṣid: 1.152–153) ism/fijiʿl/ḥarf, previously given as the hyponyms of kalima, are defijined as “phonetic forms provided with meaning” (ʾalfāḏ̣ allatī lahā dalālatun), which entails that ‘word’ (kalima) also refers to a phonetic form provided with meaning. Now, the limitation of the parts of speech to three is due to the existence of three kinds of meanings, and no more than three, each diffferent from the others. The fijirst is the designation of a specifijic entity in itself, in order to distinguish it from others (ʾan yakūna š-šayʾa nafsahu ḥattā yatamayyaza bil-ʿalāmati l-manṣūbati lahu ʿan ġayrihi); the second is the designation of an action and a time conjointly (ʾiḏā qulta ḍarabtu dalalta ʿalā ḍarbin wa-zamānin); the third is the designation of a meaning intervening only in the combination of the fijirst two categories (maʿnan iʿturiḍa ʿalā l-fijiʿli wal-ismi baʿda tilāfijihimā). Up to now we can easily recognize the defijinition of the three parts of speech of philosophical origin which are usual in the traditional grammar starting with Ibn Sarrāj, but with a stronger bias on the semantic component. Jurjānī goes further in his reasoning: he does not limit himself to state that a noun refers to a specifijic entity, but he also states that the noun is “a sign indicating to the addressee that you intend this entity and not that one” and furthermore that “each of these words indicates to the addressee that the speaker intends this kind [of thing], this entity and this meaning, all the others excluded”.29 An analogous but more concise reasoning is put forward for verbs and particles in the following passage. Nouns, and in general words, are therefore signs referring to concepts that difffer from each other: we cannot refrain from thinking to the concept of linguistic sign of Saussure and to the pair ‘signifijiant/signifijié’. The similarity with the concept of ‘signifijié’, that on the paradigmatic level is delimited by the meanings of other linguistic terms which defijine it and are in contrast with it, is striking. Moreover, in Jurjānī’s analysis linguistic signs are conceived in the frame of a communicative event where they are used by the speaker to express his communicative intention: this perspective reflects no doubt a pragmatic approach, and foretells the conceptual apparatus fully expressed in Dalāʾil al-ʾiʿjāz.

29  A noun is ʿalāmatun tadullu s-sāmiʿa ʿalā ʾannaka qaṣadta hāḏihi ḏ-ḏāta wa-lā tilka and kullu wāḥidin min hāḏihi l-kalimi yadullu l-muxāṭaba ʿalā qaṣdi l-mutakallimi ʾilā hāḏā l-jinsi wa-hāḏā š-šayʾi wa-hāḏa l-maʿnā dūna ġayrihi (Muqtaṣid: 1.152; 153 respectively).

94

antonella ghersetti

An interesting case for the discussion of the notion of ‘word’ is presented in the part of the commentary (Muqtaṣid: 2.1035–1041) devoted to the compound nouns i.e. two distinct nouns that coalesced into one (bāb al-ismayni allaḏayni yujʿalāni sman wāḥidan). The question here is, as usual, the determination of their declensional status, but the passage presents some interest from a theoretical point of view, considering that it touches the question of the boundaries of the word. In the course of the discussion, Jurjānī (Muqtaṣid: 2.1035–1036) puts forward some points pertaining to morphology and establishes an analogy between the compound noun and the feminine form with tāʾ at-taʾnīṯ.30 For the noun, he says, the unmarked case (ʾaṣl) is the singleness (al-ʾifrād), while the marked case is its being compound; likewise, the masculine noun is unmarked compared to the feminine, marked. Both the feminine noun and the compound nouns are then ‘a derivation’ i.e. a marked case ( farʿiyya). It is starting from this analogy and from the assimilation with the feminine that Jurjānī explains the declension of ‘Ḥaḍramawt’, following the path traced by Sībawayhi. The second part ‘mawt’ is assimilated with the marker of the femine (tāʾ) of the word qāʾimatun and the rāʾ of the fijirst part ‘Ḥaḍr’ to the mīm of qāʾimatun. This analogy permits to derive two conclusions: a) the fijirst part of the compound noun ‘Ḥaḍr’ must necessarily have a fatḥa at the end, which is exactly the same as the fatḥa preceeding the feminine marker (tāʾ at-taʾnīṯ) and b) the declension marker must occur at the end of the second part just as it occurs with tāʾ at-taʾnīṯ. In other terms, in compound nouns such as Ḥaḍramawt the phoneme/letter receiving the declension marker must be the one at the end of the last part of the noun, just as in qāʾimatun the ḥarf al-ʾiʿrāb is the feminine marker tāʾ, since the declension marker cannot occur in the middle of the word

30  It is the principle of analysis established by Sībawayhi (who speaks rather of ‘h’ than explicitly of tāʾ at-taʾnīṯ), in accordance with the teaching of his master Xalīl who afffijirmed that in the case of two nouns that coalesced into one he recognized to the second word (kalima) the status of hāʾ. This afffijirmation is proved by virtues of morphological arguments, namely the formation of the diminutive, which afffects only the fijirst part of the noun (Kitāb: 2.267, see also Kitāb: 3.476–6, where a whole chapter is devoted to this item). Sībawayhi discusses the matter and concludes by this statement: wa-haḏā yadullu ʿalā ʾanna l-hāʾa tuḍammu l-isma kamā tuḍammu l-ismu l-ʾāxiru ʾilā l-ʾawwal” (Kitāb: 2.268). In another passage, always quoting Xalīl, the examples of proper names as Ṭalḥa and Ḥamza are given to stress the analogy: tulqī l-ʾāxira minhumā kamā tulqī l-hāʾa min Ḥamzata wa-Ṭalḥata li-ʾanna Ṭalḥata bi-manzilati Ḥaḍramawt (Kitāb: 3.373). The second part of ‘Ḥaḍra-mawt’, it is stated, has thus the same status as the fijinal ‘h’ of these names and, as a consequence, the fijirst part of it has no right to declension and must have a fatḥa in the end, just as it occurs in nouns ending with tāʾ of the feminine.

‘word’ in the linguistic thinking of jurjānī

95

( fī ḥašwi l-kalimati).31 From all this we can deduce that if, ad absurdum, the declension marker would occurr in the middle of the word, i.e. at the end of the fijirst part of the word ‘Ḥaḍr’, this would no longer be a single noun (then—we gloss—a single word). The criteria used to defijine a lexical unit as a single word rest in this case on a morphosyntactic logic that identifijies the declension marker with the element fijixing the boundaries of the word, since -by defijinition- it can occur only at the end and not in the middle (which is, in a sense, a tautology). If the case of ‘Ḥaḍramawt’ is unambiguous, since no alternative possibility is admitted,32 the topic of compound nouns presents also more ambiguous cases, such as those of ‘Maʿdīkarib’ and ‘Baʿlabakk’ which can both have a ‘fijinal’ and an ‘intermediate’ declension marker. Such cases are rather accounted for on the basis of an analogy with annexation, which implies that these nouns can also be considered as two separate words. In particular, dealing with the declension of ‘Maʿdīkarib’,33 Jurjānī advances three possibilities (ʾawjah), strictly adhering to Sībawayhi’s analysis. The fijirst one is the “fijinal declension” where ‘b’ is ḥarf al-ʾiʿrāb, which implies that we have to deal with a single and indivisible noun. The second and the third are the existence of an “intermediate declension”, with a diptotic or triptotic declension of the second part of the noun respectively. These two possibilities are based on the interpretation of the compound noun as an annexation structure,

31  Incidentally, Ibn Jinnī seems to admit such a possibility when he states that the declension marker shifts into the middle of the word in case of pause ( fī baʿḍi l-waqfiji . . . tantaqilu ḥarakatu l-iʿrāb ʾilā ḥašwi l-kalimati, Xaṣāʾiṣ: 2.331). Why the marker of the declension cannot be at the beginning or the middle of the word is a topic treated by Zajjājī (ʾĪḍāḥ: 76): a whole chapter is devoted to the subject (for more details on the position of other grammarians see Versteegh 1995: 120, fn. 1). Curiously, in the course of the explanation the end of the word (ʾāxir al-ism) is also hinted at as “what is afffected by the pause” (al-waqf yudrikuhu). The operating concept of “pause” is also referred to in modern linguistics in which one of the several defijinitions of ‘word’ (“any segment of a sentence bounded by successive points at which pausing is possible”, Lyons 1968: 199) is hinging on the idea of a “potential pause”. 32  Fa-baqiya fī Ḥaḍramawta sababun wāḥidun wa-huwa jaʿlu l-ismayni sman wāḥidan (Muqtaṣid: 2.1036). 33  The declension of this noun has been debated since the time of Sībawayhi, who afffijirms that this is a case admitting diffferent possibilities ( fīhi luġāt): some speakers realize it as an annexation, some realize it as an annexation but without declension and make it as a feminine (i.e. with diptotic declension), some realize it as a single noun (ism wāḥid): waminhum man yaqūlu Maʿdīkaribin fa-yuḍīfu, wa-minhum man yaqūlu Maʿdīkariba fa-yuḍīfu wa-lā yaṣrifu yajʿalu Kariba sman muʾannaṯan wa-minhum man yaqūlu Maʿdīkaribu fa-yajʿaluhu sman wāḥidan (Kitāb: 3.296–297). From this it can be inferred that the three cases are respectively the full (i.e. triptotic) declension of both parts of the noun, the “intermediate declension” and the diptotic declension of the second part, and the lacking of “intermediate declension” with diptotic declension of the second part.

96

antonella ghersetti

where the two parts are considered as discrete units on a morphosyntactic level.34 Nevertheless, a little farther, the morphosyntactic reasoning is left aside to the advantage of a semantic corrective. This happens when Jurjānī goes through the hypothesis of analogy with annexation (ʾiḍāfa), the means the grammarians have recourse to in order to explain the “intermediate” declension of compound nouns such as ‘Maʿdī-karib’, ‘Qālī-qalā’, ‘Bādī-badā’ to account for the possibility of putting the declension marker at the end of the fijirst noun.35 This solution is not agreed upon by Jurjānī, who considers the analogy a partial and purely formal correspondence, because in this case the annexation is purely formal and completely devoid of semantic grounds.36 He emphasizes that in such cases the annexation has not the same status as it has in constructions such as ġulām Zayd. He rejects the analogy for two reasons: fijirst, in ġulām Zayd there is no identity between Zayd and gulām (or, more precisely, between the concepts they express) since these are two separate entities; second, in this case an additional meaning (that of li-) does exist, besides those of these two terms.37 On the contrary, in the cases of ‘Baʿlabakk’ or ‘Qālīqalā’, there is actually no referent corresponding to ‘Bakk’ or ‘Qalā’ individually taken when ‘Baʿl’ or ‘Qālī’ are annexed to them. Therefore ‘Bakk’ has the same status the fijinal rāʾ has in relation with the other phonems in the proper name ‘Jaʿfar’. The whole ensemble, i.e. the word ‘Baʿlabakk’, refers to one single and indivisible entity that is the place so named, while ġulām Zayd refers to two separate entities linked by a relation of ownership. This noun must therefore be intended as one and a single noun (or, we gloss, ‘word’) denoting a place since “there is no separation in the meaning between the annexation [i.e. the annexation postulated at the morphosyntactic level] and the fact of making a single noun from two nouns, because it

34  The same reasoning is held for Baʿlabakk, which however has only two possibilities (Muqtaṣid: 2.1037). 35  In this case Sībawayhi (Kitāb: 3.296) accounts for two possibilities attested in the speech of the Arabs: the fijirst is the annexation of the second noun (Bakk) to the fijirst (Baʿl), which does not make it a compound noun (ism wāḥid). The second case it that of the constitution of a compound noun (i.e. a single word). This afffijirmation implies that the unmarked case for compound nouns is the lacking of the “intermediate declension”. 36   Fa-hāḏihi ʾiḍāfatun lafḏ̣ iyyatun wa-lā maʿnawiyya (Muqtaṣid: 2.1041). 37  In traditional grammar the presence of an underlying li-, or min, elided in the surface structure explains the indirect case of the second term as the result of the government of the elided preposition.

‘word’ in the linguistic thinking of jurjānī

97

is an annexation that concerns the form but not the meaning”.38 Farther, he (Muqtaṣid: 2.1041) seems to reject the possibility of the “intermediate” declension, so automatically identifying compound nouns as ‘Baʿlabakk’ as a single and individual word (ism wāḥid) just as ‘Ḥaḍramawt’ is. Semantics then plays an important role in his way of analyzing the problematic declension of compound nouns. It must be noted that a similar reasoning was made by Sībawayhi, who had also taken into consideration the existence of two separate entities, the second not deriving from or being part of the fijirst. But his reasoning is purely morphological: he does not take into consideration the semantic aspect and thus completely neglects the scope of meaning.39 Then, if we are not mistaken, what really guarantees the unity of the word is the semantic facet, that is the impossibility to split the word into two parts indepent from the semantic point of view, even if—from the morphosyntactic point of view—this would be acceptable and is in fact admitted by some. The extension of this semantic criterion then leads Jurjānī to propose an equivalence between the compound nouns like ‘Baʿlabakk’ and syntagms like ibn ʿirs40 ‘weasel’, where the two components could be considered as a single word. As a matter of fact, he notices, ʿirs and ibn have no separate referents, and the two nouns are an unique sign (ʿalam) referring to a well known entity. Here too (Muqtaṣid: 2.1041) the annexation is purely formal, and this structure is alike to that of two nouns that coalesced into one.41 It is obvious that here the boundaries of the word tend to vanish and the individuation of the word as a textual unit (since as a lexicographical unit it is the matter of another science, ʿilm al-luġa) starts to be problematic. That single words, taken in isolation, have—strictly speaking—no meaning, is a principle the rhetorical works of Jurjānī are pervaded with. Actually, in Dalāʾil al-ʾiʿjāz and still more in ʾAsrār al-balāġa, it is not a matter of grammar (morphophonology and syntax) but rather of sentence

38

 Wa-ʾinnamā l-jamīʿu smu mawḍiʿin fa-lā faṣla fī l-maʿnā bayna l-ʾiḍāfati wa-jaʿli l-ismayni sman wāḥidan, fa-hāḏihi ʾiḍāfatun lafḏ̣ iyyatun lā maʿnawiyya, and ʾiḏā qulta ‘hāḏā Baʿlabakku’ jaʿalta fī l-ismi l-ʾaxīri ḥarfa l-ʾiʿrābi bi-manzilati juzʾin mina l-ʾawwali karrāʾi min Jaʿfarin lafḏ̣ an wa-maʿnan. ʾAmmā l-lafḏ̣ u fa-ḏ̣ āhirun wa-ʾammā l-maʿnā fa-huwa ʾannahu lā yadullu ʿalā šayʾin ʾāxara kamā yadullu Zaydun ʿalā ġayri mā yadullu l-ġulāmu fī qawlika ġulāmu Zaydin (Muqtaṣid: 2.1041). 39  Li-ʾannahumā kānā bāʾinayni wuṣila ʾaḥaduhumā bil-ʾāxari fal-ʾāxiru bi-manzilati l-muḍāfiji ʾilayhi fī ʾannahu laysa mina l-ʾawwali wa-lā fīhi wa-humā mina l-iʿrābi ka-smin wāḥidin lam yakun ʾāxiruhu bāʾinan mina l-ʾawwal (Kitāb: .268). 40  The editor reads ibn ʿurs (Muqtaṣid: 2.1041). 41  ʾA-lā tarā ʾanna ʿirsan laysa bi-smin ġayri l-ibni wa-ʾinnamā l-ismāni ʿalamun lil-jinsi l-maʿrūfiji fal-ʾiḍāfatu lafḏ̣ iyyatun wa-ḥukmuhumā ḥukmu jaʿli l-ismayni sman wāḥidan.

98

antonella ghersetti

and across-sentence analysis or, more precisely, of textual analysis. The importance of a holistic approach to the sentence, mostly at the semantic level, is not stranger to the classical grammatical tradition but in Jurjānī “this holistic perspective takes precedence over the meaning of individual words” to the point that “Jurjânî suggests in places that there is no such thing as lexical meaning”.42 Actually, Jurjānī seems to go as far as to deny the existence of a lexical meaning, considering that, for instance, in his enunciative act the speaker derives from the ensemble of words a single sentence meaning (mafhūm wāḥid), and by no means a series of separate lexical meanings (ʿiddatu maʿānin).43 If this idea was openly expressed already in al-Muqtaṣid,44 in Dalāʾil al-ʾiʿjāz it becomes one of the pivots of the analysis. In this treatise, in fact, the individual word is not taken into consideration but in a context, where ‘context’ refers to the direct textual context but also to the situational context.45 Hence the importance attached to the concept of naḏ̣ m, the very hinge of this work46 on which is grounded, e.g., the impossibility of giving an aesthetic judgement upon the words individually taken.47 If words are considered separately, before they combine (qabla duxulihā fī taʾlīf ) and can form an utterance thanks to their combination (ḍamm al-kalima ʾilā l-kalima), it is impossible to 42

 Owens (1988: 249).  ʾIḏa qulta ḍaraba Zaydun ʿAmran yawma l-jumʿati ḍarban šadīdan taʾdīban lahu fa-ʾinnaka taḥṣulu min majmūʿi hāḏihi l-kalimi kullihā ʿalā mafhūmin huwa maʿnan wāḥidun lā ʿiddatu maʿānin . . . li-ʾannaka lam taʾti bi-hāḏihi l-kalimi li-tufīda ʾanfusa maʿānīhā wa-ʾinnamā jiʾta bi-hā li-tufīda wujūha t-taʿalluqi llatī bayna l-fijiʿli llaḏī huwa ḍarbun wabayna mā ʿamila fīhi . . . (Dalāʾil: 370–371). Farther this phrasal meaning is described in terms of a speech act as ʾiṯbāt ‘assertion’. The examples are numerous: see e.g., p. 470 where ‘what is conceivable’ (the meaning of the sentence) is a relation (nisba) deriving from the overall ensemble of the words (majmūʿi jumlati fijiʿlin wa-smin . . .ʾawi smin wasmin) and not from the single meanings, even after they combined. Incidentally, Ibn Jinnī already stated that the meaning of a sentence does depend on its overall organization more than it depends on the meaning of the individual words: wa-ḏālika ʾanna l-kalāma wuḍiʿa lil-fāʾidati wal-fāʾidatu lā tujnā mina l-kalimati l-wāḥidati wa-ʾinnamā tujnā min aljumali wa-madāriji l-qawl (Xaṣāʾiṣ: 2.331). 44  Fa yaʾtalifu l-fijiʿlu maʿa l-ismi wa-yaḥṣulu bi-tilāfijihimā fāʾidatun lā naʿqiluhā min kulli wāḥidin minhumā ʿalā l-infijirād (Muqtaṣid: 1.153). 45  From this stems the importance attached to the enunciative context (ḥāl) which is a peculiarity of the canonical (i.e. post-Qazwīnī) balāġa (see Ghersetti 1998: 64 fff.). 46  The defijinition of naḏ̣ m (textual organization) is repeated so often in the text that we’ll confijine ourselves to a few quotations: naḏ̣ m is tawaxxī maʿānī n-naḥwi wa-ʾaḥkāmihi fīmā bayna l-kalim (pp. 357, 359), tawaxxī maʿānī n-naḥwi wa-ʾaḥkāmihi fīmā bayna maʿānī l-kalim (p. 365); tawaxxī maʿānī n-naḥwi wa-ʾaḥkāmihi wa-wujūhihi wa-furūqihi fīmā bayna maʿānī l-kalim (p. 459); tawaxxī maʿānī n-naḥwi fī maʿānī l-kalim (p. 336). 47  E.g., ʾinna l-faṣāḥata lā taḏ̣ haru fī ʾafrādi l-kalimāti wa-ʾinnamā taḏ̣ haru biḍ-ḍammi ʿalā ṭarīqatin maxṣūṣatin (Dalāʾil: 409) and ʾanna l-faṣāḥata lā takūnu fī ʾafrādi l-kalimāti wa-ʾannahā ʾinnamā takūnu fīhā ʾiḏā ḍumma baʿḍuhā ʾilā baʿḍin (358–359). 43

‘word’ in the linguistic thinking of jurjānī

99

recognize their excellence one over the other. Τhen it will be impossible to state that rajul fijits better its meaning than its equivalent in Persian, or to state that a certain word is aesthetically better than another, if we do not take into consideration the mutual links connecting the words, (irtibāṭ haḏihi l-kalimi baʿḍihā bi-baʿḍ) or, in other terms, the context in which the word occurs. Some words are not to be considered better than others if individually taken (kalima mufrada), and this is true when we realize that one and the same word can be both beautiful or ugly depending on its context.48 When speaking of tropes, Jurjānī afffijirms that no fijigure of speech can be applied to the word individually considered, disregarding the mutual links with other words and the functional values afffecting it.49 But, once the sentence and across-sentence/inter-utterance links have duly been stressed, what remains of the concept of ‘word’ as a discrete unit? The following pregnant passage comes to our rescue: “know that there is a principle [. . .]: isolated forms (ʾalfāḏ̣ mufrada), which have been instituted by convention (ʾawḍāʿ) in the language, have not been established to know their meanings in themselves, but rather to join them mutually as to draw information thanks to their connection”.50 We are facing once more the statement of the quasi non-existence of the semantic value of the isolated words; but in the fijirst part of this passage the concept of ‘word’ appears in all its components. These are namely the formal, or phonetic, component (it is in this sense that we interpret the term ʾalfāḏ̣ ), the conventional institution referred to by the term ʾawḍāʿ and the semantic component referred to by the phrase maʿānīhā fī ʾanfusihā.51 Now, Jurjānī recognizes in places that the relation between the phonetic facet, that is the chain of sounds (tawālī l-ḥurūf fī n-nuṭq) and the semantic one is the result of an arbitrary association, since in the words there is no necessary relation between the phonetic chains and the meanings.52 Let’s now go to the second part of the passage mentioned above, where it is

48

 Dalāʾil: 90–92.  Lā yutaṣawwaru ʾan yadxula šaʾyun minhā [i.e. the tropes] fī l-kalimi wa-hiya ʾafrādun lam yatawaxxā fīmā baynahā ḥukmun min ʾaḥkāmi n-naḥw (Dalāʾil: 358). 50  Iʿlam ʾanna hāhunā ʾaṣlan . . . ʾanna l-ʾalfāḏ̣ a l-mufradata llatī hiya ʾawḍāʿu l-luġati lam tūḍaʿ li-tuʿrafa maʿānīhā fī ʾanfusihā wa-lākin li-ʾan yuḍamma baʿḍuhā ʾilā baʿḍin fa-yuʿrafa fīmā baynahā fawāʾid (Dalāʾil: 469). 51  The idea of a discrete unit with a semantic facet connected by convention to a formal one is also openly stated elsewhere: see e.g. Dalāʾil: 354, maʿānī l-kalimi l-mufradati llatī hiya lahā bi-waḍʿi l-luġa. The conventional or divine character of language was debated for over a century: on this see Loucel (1963–4). 52  He says “If the institutor of speech would have said rabaḍa rather than ḍaraba, no corruption would have been derived from this”: naḏ̣ mu l-ḥurūfiji huwa tawālīhā fī n-nuṭqi 49

100

antonella ghersetti

said that to the lexical meaning of the word a second meaning must be added, carried by the functional value that words acquire in their mutual combination (ḍamm baʿḍihā ʾilā baʿḍ). It is precisely this that transmits the information ( fawāʾid) or, in other terms, an overall meaning which is not the result of the meanings of the single words, but to which the individual words contribute. The semantic component and its psychological reality are in a sense a primary in the concept of ‘word’ as well as in the process of convention. The institution (waḍʿ) of a certain phonetic form (lafḏ̣ ) cannot be realized but to refer to a meaning (corresponding to ‘concept’). The institution and the use of each of the three parts of speech (ism, fijiʿl and ḥarf ) cannot precede the concept; it is rather the opposite, since the convention as association between form and meaning (muwāḍaʿa) cannot exist but for concepts already known and present to the mind of the speaker. In this, the convention is really similar to deixis (ʾišāra) because both refer to an extralinguistic referent which is not selected in itself ( fī nafsihi) but rather among the elements of a wider gamut (min bayn sāʾir al-ʾašyāʾ).53 The semantic facet of the word, whose formal part is a sign (sima) standing for a concept,54 is then the most essential component of it and, in a sense, its point of departure and its foundation. It is precisely on this principle that the analysis of Jurjanī rests. Actually, it is by virtue of a semantic basis that both the textual organization of words that combine to form an utterance and their order are possible.55 Now, this organization and this order (naḏ̣ m wa-tartīb) do not rest on the pronunciation of a phonetic chain or a series of sounds, (an-nuṭq bil-lafḏ̣ a baʿda l-lafḏ̣ a) but rather on the mutual coherence of the meanings (taʿlīq maʿānīhā baʿḍihā bi-baʿḍin).56 Should it be the contrary, we would be in a posi-

faqaṭ wa-laysa naḏ̣ muhā bi-muqtaḍin ʿan maʿnan . . . fa-law ʾanna wāḍiʿa l-luġati kāna qad qāla rabaḍa makāna ḍaraba la-mā kāna fī ḏālika mā yuʾaddī ʾilā fasād (Dalāʾil: 94). 53  Wa-kayfa wal-muwāḍaʿatu lā takūnu wa-lā tutaṣawwaru ʾillā ʿalā maʿlūmin fa-muḥālun ʾan yūḍaʿa smun ʾaw ġayru smin li-ġayri maʿlūmin li-ʾanna l-muwāḍaʿa kal-ʾišārati fa-kamā ʾanna ʾiḏā qulta ʿxuḏ ḏāk’ lam takun hāḏihi l-ʾišāratu li-tuʿrifa s-sāmiʿa al-mušāra ʾilayhi fī nafsihi wa-lākin li-yuʿlama ʾannahu l-maqṣūdu min bayni sāʾiri l-ʾašyāʾi wa-kaḏālika ḥukmu l-lafḏ̣ i maʿa mā wuḍiʿa lahu (Dalāʾil: 470). 54   See e.g. Dalāʾil: 470, where the signifijicants are nothing but signs standing for a certain meaning (ʾalfāḏ̣ al-luġāt simāt li-ḏālika l-maʿnā). 55  Wan-naḏ̣ mu wat-tartību fī l-kalāmi kamā bayyannā ʿamalun yaʿmaluhu muʾallifu l-kalāmi fī maʿānī l-kalimi lā fī ʾalfāḏ̣ ihā (Dalāʾil: 334). 56  See e.g. ʾanna l-faṣāḥata lā takūnu fī l-kalimi ʾafrādan wa-ʾannahā ʾinnamā takūnu ʾiḏā ḍumma baʿḍuhā ʾilā baʿḍin wa-kāna yakūnu l-murādu bi ‘ḍammi baʿḍihā ʾilā baʿḍin’ taʿlīqa maʿānīhā baʿḍin bi-baʿḍ (Dalāʾil: 410, also 409).

‘word’ in the linguistic thinking of jurjānī

101

tion to utter, and to utter with eloquence ( faṣāḥa), sentences like ḍaḥika xaraja,57 which is obviously absurd. We must deduce that the privation of the semantic component makes the words “pure sounds (ʾaṣwāt) and like echoes of the phonemes (ʾaṣdāʾ al-ḥurūf )”.58 On a diffferent level, Jurjānī advances this same principle in connection with the notion of authorship. The author of a poem is such because he created and organized concepts, and not because he uttered the phonetic forms that come out from his mouth: if we would attribute to Imruʾ al-Qays the qualifijication of ‘author’ only because he uttered some words and some phonetic forms, it could be plausible to describe his rhapsodists as authors as well, which is obviously absurd.59 In Dalāʾil al-ʾiʿjāz a ‘word’ is therefore conceived as a discrete linguistic unit provided with a phonetic form (lafḏ̣ ) and carrying a meaning (maʿnā), the two associated by virtue of convention (waḍʿ). The most basic component appears to be meaning, which in this case is an entity consisting not only of a lexical meaning (nafs al-kalima, ḏāt al-kalima), but also of a component deriving from the syntactic function of the word and from the status so acquired in the utterance. Actually, in Jurjānī’s conceptual framework, the meaning of the word virtually does not exist in isolation,60 and is almost unconceivable61 if considered separately and devoid of its syntactic values (maʿānī n-naḥw).62 This is not as denying a priori the lexical meaning of each word, but rather

57

 Dalāʾil: 359.  See e.g. ʾannahā [al-kalimu] law xalat min maʿānīhā ḥattā tatajarradu ʾaṣwātan wa-ʾaṣdāʾa ḥurūfijin, la-mā waqaʿa fī ḍamīrin wa-lā hajasa fī xāṭirin ʾan yajiba fīhā tartībun wa-naḏ̣ mun (Dalāʾil: 98). 59  Yanbaġī ʾan lā yaštabiha ʾanna š-šiʿra lā yaxtaṣṣu bi-qāʾilihi min jihati ʾanfusi l-kalimi wa-ʾawḍāʿi l-luġa . . . fa-ʾin zaʿamta ʾanna-ka jaʿalta-hu [Imraʾa l-qays] qāʾilan lahu min ḥayṯu ʾinnahu naṭaqa bi-kalimi wa-sumiʿat ʾalfāḏ̣ uhu min fīhi . . . fa-jʿal rāwiya aš-šiʿri qāʾilan lahu (Dalāʾil: 336–337). 60  Even if from a diffferent perspective this was also stated by others; see e.g. Zajjājī: according to him the noun, individually taken, is not informative or, even worse, is like a sound devoid of meaning: al-ismu . . . lā tuḥṣalu minhu fāʾidatun mufradan ḥattā naqtarinu bi-smin ʾāxar . . . wa-ʾillā kāna ḏikruka lahu laġwan wa-haḏaran ġayra mufīd (ʾĪḍāḥ: 49). 61  Dalāʾil: 368–9: lā tutaṣawwaru ʾan-yataʿallaqa l-fijikru bi-maʿānī l-kalimi ʾafrādan wamujarradatan min maʿānī n-naḥw. 62  What is meant with these terms is the syntactic function, corresponding in fact to a logical function. Jurjānī advances a description that is formally based on the government theory and on the canonical terminology of the Arabic grammar. In Dalāʾil: 369, after having mentioned maʿānī n-naḥw, he afffijirms that “it is unconceivable to think of the meaning of a verb without thinking in the meanwhile of its government on the noun, nor to the meaning of a noun without conceiving the government of the verb making it subject ( fāʿil) or object (maf ʿūl)”. Now, if the syntactical functions are nothing other than the formal expression of the logical relations that link the words, which is unanimously accepted by the grammarians (hāḏā qawlu jamīʿi n-naḥwiyyīn, Zajjājī ʾĪḍāḥ: 69), the insistence 58

102

antonella ghersetti

is tantamount to stress that this is always conceived in connection with the semantic values that the syntactic relations add to it. It is therefore possible to conceive the individual meanings of words in themselves (maʿānī l-kalimi l-mufradati ʾaṣlan), but—as concrete units of analysis— not deprived of their syntactic function (mujarrada min maʿānī n-naḥw).63 The lexical meaning, we would venture, is nothing but a partial feature of the overall meaning constituting the core of the concept of ‘word’ in this work. In this perspective, the single word nearly represents a unit deduced by abstraction. Actually, in his speech act the speaker does not think of the lexical units separately ( fardan) but rather selects them inside a paradigmatic axis only after having conceived his communicative intention (qaṣd), where by qaṣd we must intend the transmission of the information (ʾan tuʿlima s-sāmiʿa bihā šayʾan lā yaʿlam). This information is grounded on a mutual connection of words, whose purely semantic character is stressed (taʿlīq maʿnā kalimatin bi-maʿnā kalimatin ʾuxrā).64 This approach echoes the Saussurian approach in defijining the linguistic unit (which does not correspond to ‘word’, that de Saussure defijines as incompatible with the ‘concrete unit of analysis’).65 That is the exact opposite of the “cellular” notion of meaning widespread in the canonical tradition of Arabic grammar, where sentence meaning is nothing more than the result of a process of semantic agglutination.66 In principle, in Jurjānī’s mind, the linguistic analysis starts from the utterance and proceeds, by abstraction, dividing it into minor discrete units.67 Therefore the word would be rather

on the syntactic component as an inherent part of the meaning seems peculiar to Jurjānī’s thinking. 63  Dalāʾil: 369. It is a basic principle also, and mostly, in ʾAsrār al-balāġa; see e.g. p. 3: al-ʾalfāḏ̣ lā tufīdu ḥattā tuʾallafu ḍarban xāṣṣan min at-taʾlīf. 64  See e.g. kunta qad fakkarta fī maʿānī ʾanfusi l-kalimi ʾillā ʾanna fijikraka ḏālika lam yakun ʾillā min baʿdi ʾan tawaxxayta fīhā min maʿānī n-naḥwi . . . wa-lam tajiʾ ʾilā fijiʿlin ʾawi smin fa-fakkarta fīhi fardan wa-min ġayri ʾan kāna laka qaṣdun ʾan tajʿalahu xabaran ʾaw ġayra xabar (Dalāʾil: 368–9). 65  “Incompatible avec notre notion d’unité concrète” (Saussure Cours: 147). In this connection see Lucidi (1966: 69) who remarks that the linguistic sign does not consist of the individual word and when a speech act (in itself a sign) consists of a single word, this ceases being such (French translation in Cours: 459, fn. 214). 66  See e.g. Sīrāfī, Šarḥ al-Kitāb, apud Frank (1981: 271–2, fn. 34): al-kalāmu yūḍaʿu kullu kalimatin minhu tadullu ʿalā maʿnan mā, ṯumma yurakkabu fa-yaqtarinu baʿḍuhā bi-baʿḍin fa-taqaʿu bihā l-fawāʾidu l-mustafādatu bi-qtirānihā, wa-ʾin kānat kullu wāḥidatin minhā qad dallat ʿalā maʿnan bi-ʿaynihi. 67  Rundgren (1982–3: 115, fn. 10), in the very same perspective as the approach of De Saussure, proposes a similar process for Arabic texts: linguistic units much wider than the single words are the real object of the linguistic analysis. Rundgren considers the tex-

‘word’ in the linguistic thinking of jurjānī

103

“an entity by abstraction” which does not exist in the utterance but for the enunciative process.68 This enunciative perspective is precisely the one characterizing the beginnings of the Arabic grammatical tradition and namely the Kitāb of Sībawayhi. Actually, in the arrangement of his Kitāb, he does nothing more than following “the natural order in which the data would present itself to the observer”.69 Syntax, morphology and phonology follow one upon the other, exactly in the same way as the linguistic analysis moves from the level of the utterance corresponding to the communicative event to the following levels, where discrete and abstract units (words and phonemes) are taken into consideration. The same approach is typical of the second rhetorical work of our author, ʾAsrār al-balāġa. In this text the existence of a perfect cohesion between meanings, as psychological realities, and their formal expression is repeatedly emphasized. Furthermore, the order of the formal expressions follows, or—better—reproduces the order of concepts; as a consequence, each variation in the sequence of the form afffects the meaning.70 Here we are obviously dealing with textual analysis, where the individual word is not considered but as the component of a wider ensemble trasmitting an overall meaning. In this framework, the notion of ‘word’ (kalima) as a discrete unit does not emerge with a particular evidence. Nevertheless, in the last pages we can fijind some passages which are revealing in this connection. When the defijinition of majāz is discussed we can clearly see that the notion of ‘word’ rests mostly on semantic and (less) on syntactic criteria, while the morphological facet remains completely marginal. Here meaning (maʿnā) and the syntactic status afffecting the word by virtue of its function in the syntagm are acknowledged as inherent components of the word (kalima). When dealing with the correct defijinition of majāz,

tual unit (texteme)—and not the single lexemes—as the starting point of the analysis. He therefore criticizes those whose analysis proceeds the other way round, e.g.: “It is obvious that Ricoeur, passing from the phoneme to the word and from the word to the sentence, has not observed the “condition of the beginning” [. . .] On the contrary, as a part of the sentence the word is no longer a word but a phrase or a part of a phrase”. 68  Rundgren (1982–3: 96; see also p. 100): “In a case like Faras! ‘Look, a horse!’ we have no difffijiculty to cope with, the textual minimality of this utterance being obvious. For in this case the text, as an entity by derivation, is materially identical with the word faras considered to be an entity by abstraction”. 69  Carter (2004: 38). 70  See e.g. al-ixtiṣāṣu fī t-tartībi yaqaʿu fī l-ʾalfāḏ̣ i murattaban ʿalā l-maʿānī l-murattabati fī n-nafsi l-muntaḏ̣ imati fīhā ʿalā qaḍiyyati l-ʿaql (ʾAsrār: 4). See also Dalāʾil, 94: wa-ʾammā naḏ̣ mu l-kalimi fa-laysa l-ʾamru fīhi kaḏālika li-ʾannaka taqtaḍī fī naḏ̣ mihā ʾāṯāra l-maʿānī wa-turattibuhā ʿalā ḥasbi tartībi l-maʿānī fī n-nafs. On this idea, unfamiliar in traditional Arabic grammar, see Kouloughli (1983).

104

antonella ghersetti

Jurjānī (ʾAsrār: 383–385) afffijirms that this phenomenon can afffect a word (kalima) both in modifying its meaning (maʿnā) or afffecting its syntactic status. To explain the second possibility, i.e. the variation of the syntactic status, he puts forward an example familiar to grammarians: wa-sʾali l-qaryata (Cor. xii 82). Here the reconstruction of the underlying form (ʾaṣl) is wa-sʾal ʾahla l-qaryati: the syntactic status pertaining to al-qarya, in the abstract representation as well as in its literal meaning, is the indirect case. Therefore the direct case attributed to this word in this context is considered as a case of majāz. Jurjānī goes on stating that a word (kalima) is conceivable—at least in the context of this analysis—only if it is provided with a meaning (dalāla), because it is impossible to deprive a word of its meaning without giving it another one. Thus the pertinent features of ‘word’ are a meaning and a syntactic function, to which a third element must be added: the conventional nature of kalima. The fijigurative sense can only exist if the “word is used for some other aim than that for which it has been instituted, or if something which is not inherent to it is added or erased”.71 Therefore it is a matter of fact explicitely recognized that words are provided with such features subsequent to convention (waḍʿ) that establishes the zero degree of linguistic usage. The concept of ‘word’ in the rhetorical analysis of Jurjānī seems to fijit rather well the canonical defijinition of kalima in later Arabic grammatical tradition, where kalima is defijined as al-lafḏ̣ atu d-dāllatu ʿalā maʿnan mufradin bil-waḍʿ. This was introduced, as far as we know, by Zamaxšarī (Mufaṣṣal: 4), then taken up by other grammarians.72 It is also the defijinition that was later on included in some reference works such as the technical dictionaries by Šarīf Jurjānī and Tahānawī.73 This defijinition hinges on three facets: the phonetic aspect (lafḏ̣ a),74 the semantic aspect (maʿnā) and convention (waḍʿ). The notion of ‘word’ Jurjānī hints at, even without formulating a proper defijinition, is not very far from the ‘standard concept’ found in later sources. But still there is a great diffference in approach,

71  Li-ʾanna l-majāza ʾan yurāda bil-kalimati ġayru mā wuḍiʿat lahu fī l-ʾaṣli ʾaw yuzādu fīhā ʾaw yūhamu šayʾun laysa min šaʾnihā (ʾAsrār: 384). 72  E.g., Ibn Ḥājib, Kāfijiya: al-kalimatu lafḏ̣ un wuḍiʿa li-maʿnan mufrad. We quote from the commentary of Raḍī d-Dīn al-ʾAstarābāḏī ŠKāfijiya: 1.21. 73  Šarīf Jurjānī (Taʿrīfāt: 236) and Tahānawī (Kaššāf: 2.1247) respectively, where it is said al-kalimatu huwa l-lafḏ̣ u l-mawḍūʿu li-maʿnan mufrad. 74  Al-lafḏ̣ u ḥāṣṣun bi-mā yaxruju mina l-fam, ʾAstarābāḏī (ŠKāfijiya: 1.22). It is precisely in the sense of ‘phonetic form’ that Jurjānī uses the term to refer to lexical units (for a discussion of the possible meanings of lafḏ̣ see Owens 1988: 91 fff.).

‘word’ in the linguistic thinking of jurjānī

105

since our author lays a greater emphasis on the semantic facet, the syntactic status being an inherent component of it. As a consequence, the need of a holistic approach to the utterance is strongly stressed. Another point deserving to be taken into consideration is the question of the morphematic value of kalima that can be found in the linguistic usage of Sībawayhi and Mubarrad, and that had been for a while a subject of discussion in the grammatical tradition. Kalima, in this sense, is not found in Jurjānī’s works: this term is not used, as far as we can see, to refer to the concept of morpheme. This point is usually extensively treated by late grammarians, such as Ibn Yaʿīš and the two ʾAstarābāḏī, Raḍī d-Dīn and Rukn ad-Dīn,75 in connection with feminine markers as tāʾ at-taʾnīṯ such as in qāʾima, or plural and dual markers such as in muslimūna and muslimāni. The passages of al-Muqtaṣid (1.190 fff.) devoted to these items bear no trace of the acceptation of kalima as morpheme, nor Jurjānī openly tackles the matter. He simply afffijirms, following in this the usage of Sībawayhi and Mubarrad, that ʾalif in muslimāni is the marker of the declension and furthermore that “it is part of the substance (ʾaṣl) of the word exactly as the mīm is”.76 Likewise, the markers of sound plural ū and nūn in muslimūna are considered ziyādatāni, just as it is tāʾ at-taʾnīṯ.77 Neither in this passage nor in other places Jurjāni has recourse to the term kalima to defijine these or other morphemes. Nevertheless, the question of the defijinition of the morpheme as a discrete unit, of its relation with a discrete unit of a higher level as the word and—ultimately—the question of the boundaries and of the cohesion/indivisibility of the word is not out of the mind of Jurjānī. A passage of ʾAsrār al-balāġa (386–387) is illuminating in this connection. In this passage, devoted to the defijinition of majāz, Jurjānī offfers a thorough explanation of elision (ḥaḏf ) and addition (ziyāda). Answering the question whether they are to be considered as forms of majāz, he comes to the conclusion that they are such since they imply a variation in the syntactic status of the word. First of all he precises that these two phenomena are pertaining to utterances globally considered ( jumlati l-kalāmi) and not to words individually considered (lā ʾilā l-kalimati). To support his statement he advances the following reasoning: to afffijirm that in the sentence laysa ka-miṯlihi šayʾun (Cor. xlii 9), the kāf is an addition (ziyāda) referring to one and a single  Respectively Ibn Yaʿīš: ŠMufaṣṣal: 1.19; ʾAstarābāḏī: ŠKāfijiya: 1.25–26 and Rukn ad-Dīn al-ʾAstarābāḏī: Wāfijiya: 2–3. See Levin (1986: 430–431) and Owens (1988: 109–112). 76  Al-ʾalifu ḥarfu ʾiʿrābin wa-kāʾinun min ʾaṣli l-kalimati kal-mīm (Muqtaṣid: 1.193). 77  See Muqtaṣid: 1.193. 75

106

antonella ghersetti

word (i.e. miṯl) rather than to the overall ensemble is a glaring error. On the contrary, this afffijirmation would be well grounded in the case of the insertion of a letter (ḥarf ) inside the paradigm of a noun or of a verb, although this letter, individually considered (ʿalā l-infijirād), has no meaning and cannot be defijined as a kalima. This is the case of yāʾ lit-taṣġīr in the diminutive rujayl and of tāʾ at-tāʾnīṯ in the feminine ḍāribatun.78 The text is defijinitely explicit in stating that the term kalima must not be used to refer to morphemes, that is to units that have no lexical independence and cannot be perceived as discrete meaningful units. In Jurjānī’s opinion the two additional letters under discussion, yāʾ and tāʾ, do have a precise meaning since they transmit respectively the (grammatical, not referential) meanings of ‘diminutive’ and of ‘feminine’: but, as our author duly emphasizes, not in isolation but rather in the frame of a morphological paradigm (ṣīġa) they are an inherent part of. And even in this case, whenever taken individually, these two letters do not deserve the defijinition of kalima since the defijinition of this term (would it correspond to ‘word’ or ‘morpheme’) necessarily requires (Asrār: 387) the crossing of the morphological criterion, as in the analysis of Ibn Yaʿīš, or the phonological one as in Rāḍī d-Dīn al-ʾAstarābāḏī, with the semantic criterion. Once more, this is tantamount to insist on the facet of meaning as a primary in the concept of ‘word’ (kalima). To sum up, we can afffijirm that in Jurjānī’s thinking ‘word’ corresponds to a discrete linguistic unit that rests, in terms typical of the Saussurian approach to the concept of “linguistic sign”, on the arbitrary and conventional association between a formal facet (lafḏ̣ ) and a conceptual facet (maʿnā), the last one being determined by diffference and opposition in a gamut of concepts present to the speaker’s mind. The semantic, or conceptual, component is by far the most signifijicant since it assures the identifijication of ‘word’ as a discrete and coherent unit. We must also admit that the concept of ‘word’ as a discrete linguistic unit often vanishes (and this happens mostly in the rhetorical works) to the advantage of a holistic approach to utterances and textual units: in this context ‘word’ corresponds more to the result of a process of abstraction than to the basic unit of linguistic analysis. This all is expressed in detail, and in a somehow redundant way, in Jurjānī’s rhetorical works; but it remains true that the  [. . .] li-ʾanna hāḏihi l-ʿibārata ʾinnamā taṣluḥu ḥayṯu yurādu ʾanna ḥarfan zīda fī ṣīġati smin ʾaw fijiʿlin ʿalā ʾan lā yakūna li-ḏālika l-ḥarfiji ʿalā l-infijirādi maʿnan lā tuʿaddu waḥdahu kalimatan ka-qawlika zīdat yāʾun lit-taṣġīri fī rujayli wa-tāʾu t-tāʾnīṯi fī ḍāriba (Asrār: 387). This analysis is supported by the reasoning Jurjānī put forward in Muqtaṣid: 1.190. 78

‘word’ in the linguistic thinking of jurjānī

107

very same principles and the same perspective, if not the same defijinitions, are also found in the grammatical treatises of this author, whose intellectual consistency is no doubt admirable.

References 1. Primary Sources ʾAstarābāḏī. ŠKāfijiya. Šarḥ al-Kāfijiya, ed. I.B. Yaʿqūb, 5 vols. Bayrūt, n.d. Jurjānī. ʿAwāmil. Al-ʿAwāmil al-miʾa an-naḥwiyya fī ʾuṣūl ʿilm al-ʿarabiyya šarḥ aš-šayx Xālid al-Azharī al-Jirjāwī, ed. Badr Rāwī Zahrān. Al-Qāhira, 19882 (fijirst printing 1983). ——. Jumal. Kitāb al-Jumal, ed. ʿA. Ḥaydar. Dimašq, 1392/1972. ——. Muqtaṣid. Kitāb al-Muqtaṣid li-šarḥ al-ʾĪḍāḥ, ed. K.B. al-Murjān. Baġdād 1982, 2 vols. ——. ʾAsrār. ʾAsrār al-balāġa, repr. ed. H. Ritter (Istanbul 1954). Al-Qāhira, 1399/1979. ——. Dalāʾil. Dalāʾil al-ʾiʿjāz, ed. M.R. ad-Dāya, F. ad-Dāya. Dimašq 1407/19872. Ibn ʾAnbārī. Nuzha. Nuzhat al-alibbāʾ fī ṭabaqāt al-ʾudabāʾ, ed. A. Amer. Stockholm-Göteborg-Uppsala, n.d. Ibn Jinnī. Xaṣāʾiṣ. Al-Xaṣāʾiṣ, ed. M.ʾA. an-Najjār, 3 vols. Bayrūt: Dār al-kitāb al-ʿarabī, n.d. Ibn Sarrāj. ʾUṣūl. Al-ʾUṣūl fī n-naḥw, ed. ʿA.Ġ. al-Fatlī, 2 vols. Bayrūt, 1408/1988. Ibn Yaʿīš. ŠMufaṣṣal. Šarḥ al-Mufaṣṣal. Al-Qāhira-Bayrūt, n.d. Marzubānī. Nūr. Nūr al-qabas al-muxtaṣar min al-Muqtabas fī ʾaxbār an-nuḥāt wal-ʾudabāʾ waš-šuʿarāʾ wal-ʿulamāʾ, ixtiṣār ʾAbī l-Maḥāsin Yūsuf b. ʾAḥmad b. Maḥmūd al-Yaḥmawī, ed. R. Sellheim. Wiesbaden-Beirut: Harrassowitz, 1964 (“Bibliotheca Islamica”, 23a). Mubarrad. Muqtaḍab. Kitāb al-Muqtaḍab, ed. Muhammad ʿAbd al-Xāliq ʿUdayma, 4 vols. Al-Qāhira, 1415/1994. Rukn ad-Dīn al-ʾAstarābāḏī. Wāfijiya. Al-Wāfijiya fī šarḥ al-Kāfijiya, ed. ʿAbd al-Ḥafīẓ Šalabī. [Masqaṭ,] 1404/1983. Šarīf Jurjānī. Taʿrīfāt. At-Taʿrīfāt, ed. ʿA.R. ʿUmayra. Bayrūt, 1407/1987. Sībawayhi. Kitāb. Kitāb Sībawayhi, ed. ʿA. S.M. Hārūn. 5 vols. Bayrūt, 1411/1991. Tahānawī. Kaššāf. Kitāb Kaššāf iṣtilāḥāt al-funūn, repr. Tihrān 1967, 2 vols. Xuwārazmī. Mafātīḥ. Mafātīḥ al-ʿulūm. Al-Qāhira, 1968 (reprint of the Van Vloten edition, Leiden: Brill, 1895). Zajjājī. ʾĪdāḥ. Al-ʾĪdāḥ fī ʿilal an-naḥw, ed. M. al-Mubārak. Bayrūt, 1416/19962. Zamaxšarī. Mufaṣṣal. Al-Mufaṣṣal, ed. J.P. Broch. Christianiae, 1859. 2. Secondary Literature Baalbaki, Ramzi. 1983. “The relation between naḥw and balāġa: a comparative study of the methods of Sibawayhi and Gurgani”. Zeitschrift für Arabische Linguistik 11, 7–23. Carter, Michael G. 2004. Sībawayhi. London–New York: Tauris. Frank, Richard M. 1981. “Meanings are spoken of in many ways: the earlier Arab grammarians”. Muséon 94, 259–319. Ghersetti, Antonella. 1998. “Quelques notes sur la defijinition canonique de balāġa”. Philosophy and arts in the islamic world. Actes du XVIII Colloque de l’U.E.A.I. (Louvain, 3–10 septembre 1996), eds. U. Vermeulen, D. De Smet. Louvain: Peeters, 57–72. Gilliot, Claude. 1989. “Textes arabes anciens édités en Egypte”, (n. 7). Mideo 19, 278–280. Humbert, Geneviève. 1995. Les voies de la transmission du Kitāb de Sībawayhi, Leiden: Brill, (“Studies in Semitic languages and linguistics”, 20). Kouloughli, Djamel Eddine. 1983. “À propos de lafẓ et maʿnā”. Bulletin d’Etudes Orientales 35, 43–63.

108

antonella ghersetti

——. 2007. “Lafẓ”. Encyclopaedia of Arabic Language and Linguistics, ed. K. Versteegh et al., Leiden: Brill, 2.623–628. Larcher, Pierre. 1993. “Un grammairien «retrouvé»: ʿAbd al-Qāhir al-Jurjānī. Note sur quatre éditions récentes de ses ouvrages grammaticaux”. Arabica 40, 248–253. Levin, Aryeh. 1986. “The medieval Arabic term kalima and the modern linguistic term morpheme: similarities and diffferences”. Studies in Islamic history and civilization in honour of Professor David Ayalon, ed. M. Sharon. Jerusalem: Cana and Leiden: Brill, 423–446. ——. 2007. “Kalima”. Encyclopaedia of Arabic Language and Linguistics, ed. K. Versteegh et al., Leiden: Brill, 2.545–548. Loucel, Henri. 1963–4. “L’origine du langage d’après les grammairiens arabes”, Arabica 10, 188–208, 253–281; 11, 57–72, 151–187. Lucidi, Mario. 1966. Saggi linguistici. Napoli: Istituto Universitario Orientale. Lyons, John. 1968. Introduction to Theoretical Linguistics. London: Cambridge University Press. MacDonald, Douglas B. and Gardet, Louis. 1978. “Kalima”. Encyclopédie de l’Islam, n.é., Leiden: Brill, 4.530–531. Owens, Jonathan. 1988. The foundations of grammar: an introduction to medieval Arabic grammatical theory, Amsterdam: Benjamins (“Amsterdam Studies in the Theory and History of Linguistic Science”). ——. 1989. “The syntactic basis of Arabic word classifijication”. Arabica 36, 211–234. Ricoeur, Paul. 1969. Le conflit des interprétations, Paris: Seuil. Rundgren, Frithiof. 1982–3. “On the representation of morphemic categories in Arabic: a contribution to the fundamentals of literary hermeneutics”. Orientalia Suecana 31–32, 93–115. Saussure. Ferdinand de. Cours. Cours de linguistique générale, publié par Ch. Bally et A. Séchehaye [. . .]. ed. T. de Mauro. Paris: Payot, 1995. Troupeau, Gérard. 1963. “Deux traités grammaticaux arabes traduits en latin”, Arabica 10, 225–236. Versteegh, Kees. 1992. “Grammar and Rhetoric. al-Jurjānī on the Verbs of Admiration”. JSAI 15, 113–133. ——. 1995. The Explanation of Linguistic Causes. Al-Zajjājī’s Theory of Grammar. Introduction, translation, commentary, Amsterdam: Benjamins (“Amsterdam Studies in the Theory and History of Linguistic Science” 75).

ON LAFḎ̣ AND MAʿNĀ AGAIN: SOME ASPECTS OF THEIR RELATIONSHIP ACCORDING TO THE BALĀĠIYYŪN Lidia Bettini

Introduction In the traditional Arabic sciences, lafḏ̣ and maʿnā historically had several technical meanings and the kind of relationship which ties them varies with the science considered. As far as the balāġa1 is concerned, we shall examine here this relationship from three points of view: fijirst the defijinitions of lafḏ̣ and maʿnā in correlation2 taken from the metaphorical metalanguage3 the rhetoricians themselves use in describing their fijield, then the relation between maʿnā and linguistic meaning in the case of the separate word and within the framework of the syntactic analysis (ʾiʿrāb) of the ʾabyāt al-maʿānī and fijinally the delimitation of the maʿnā one can fijind from the description of some ornaments of style.

1  For the history of the balāġa “eloquence, rhetoric” as an independent subject, see Sanni (2003), who translates into English the short section on the balāġa of Tawḥīdī’s Risāla fī l-ʿulūm. The earliest extant works are Mubarrad’s al-Balāġa and Ibn Mudabbir’s (d. 279) ar-Risāla l-ʿaḏrāʾ. 2  Von Grunebaum (1952: 327), renders the couple maʿnā—lafḏ̣ with reference to the framework of poetics, as res ‘objects’ and verba ‘wording’; Heinrichs (1969: 69), in the chapter “Inhalt und Form”, as ‘gedanklicher Gegenstand’ and ‘sprachlicher Ausdruck’, see also the renderings proposed by Sadan (1991). The very fact that maʿnā, even apart from the philosophical context (for which see particularly Frank 1978 and Peters 1976) or the rhetorical one, is only seldom used as ‘meaning’ in the linguistic meaning, but covers a much wider range of values, has been pointed out many times. Carter (1980: 25), concerning the Kitāb by Sībawayhi, talks about “a purely social defijinition of meaning”, Frank (1981: 314– 319) in his careful examination of the diffferent uses of the term maʿnā among the Arab grammarians until the 4th/10th century, sums them up as follows: 1. meaning as the intent of the sentence, i.e. of the speaker; 2. the meaning of a noun or a verb as its referent; 3. meaning as semiotic equivalent of a word, phrase or sentence; 4. meaning as the content or conceptual signifijicate of a word, phrase or sentence. 3  For a general idea, see Kilito (1979) and Sadan (1991: 64–65).

110

lidia bettini 1. Lafḏ̣ and maʿnā: In Which of Them Does the Matter of the balāġa Lie?

The balāġiyyūn4 often employ an image to describe the fijield of their theoretical reflections, by comparing the literary creative work to a craft or an art, such as the carpenter’s or the goldsmith’s craft. According to such a comparison, the craftsmen have at their disposal a raw material, such as wood or silver or gold, which serves as the matter of their craft. It is, however, the skill of the craftsman that lies at the heart of his craft: what makes a ring a ring is not the silver alone but the workmanship that goes into his creation. Similarly, like every other craft, poetry and prose need a matter (mawḍūʿ, Xafājī Sirr: 83, madda, Qudāma Naqd: 4), that receives the ‘impression’ (taʾṯīr), the ‘moulding’ (taṣwīr) by the craftsman. But what is the element that we must consider as the raw material which, after it has been fijinely worked, becomes a fijinished artefact? According to Jurjānī (Dalāʾil: 196–197), the maʿnā, which is to be understood here as ‘content’ in general, or ‘theme, motif, image developed within a line’ is the object of the artisan’s craft, and the quality of the kalām cannot lie in the maʿnā only; Qudāma b. Jaʿfar too in the Naqd aš-šiʿr holds that the maʿānī are the matter of poetry and that poetry itself is the ‘form’ (ṣūra) the matter took. Xafājī quotes in the Sirr (84) this opinion of Qudāma, but he adds that the selfsame Qudāma in the Kitāb al-xarāj5 4  Of course by using this term we do not mean to denote a generic and monolithic entity; although the balāġiyyūn naturally shared many attitudes and convictions, some of the diffferences among them will appear as this paper proceeds. Jurjānī, in particular, established on entirely new grounds the relation between lafḏ̣ and maʿnā: in his theory lafḏ̣ indicates only the separate word with its lexical meaning, instead of the lafḏ̣ of the other rhetoricians he uses kalām, that means for him the result of the ordering of the separate words with their lexical meaning according to carefully chosen syntactical constructions; kalām indicates both signifijier and signifijiant; maʿnā indicates the linguistic meaning of an utterance, whereas he calls maʿnā l-maʿnā the image or idea a (literary) utterance conveys. Of course this is only a broad outline of Jurjānī’s thought and his very use of this terminology is somewhat more flexible. He is closely identifijied with these ideas; one of the aims of this paper is to show that this position was not unknown to others balāġiyyūn. 5  Qudāma b. Jaʿfar (Kitāb al-xarāj wa-ṣināʿat al-kitāba). This work has not reached us in its entirety; M.Ḥ. az-Zabīdī, who edited (under this title) the part that has survived (Baġdād 1981), states in the introduction (10) that Qudāma would have organized the work in 8 manāzil, (or 9, according to someone, like Yāqūt ʾUdabāʾ: 17.12–13) each of them devoted to an independent subject. The fijirst four manāzil, as far as we know, are lost. The fijifth manzila has been edited separately by Ṭ.J. Rifāʿī, Al-manzila l-xāmisa min Kitāb al-xarāj wa-ṣanʿat al-kitāba, Makka 1987; in it (183), Qudāma, in the chapter devoted to dīwān ar-rasāʾil, states that he told, in the third manzila, min ʾamr al-balāġa wa-wujūh taʿullumihā wa-taʿrīf al-wujūh al-maḥmūda fīhā wal-wujūh al-maḏmūma minhā (ed. Zabīdī 37). See also Bonebakker (1956: 44–45; 57).

on lafḏ̣ and maʿnā again

111

states that in the art of speech, it is the luġa which acts as the matter. He continues: it is evident that these two statements are contradictory and it is the second that is correct. In fact, Xafājī says, those who claim that the maʿānī are the matter on which the craftsman-poet practises his craft must explain what is the place of the ʾalfāḏ̣ , that the craftsman-poet takes and assembles and that weigh heavy with the quality of the literary composition. ʾĀmidī (Muwāzana: 1.402–404) also compares the art of speech to the other arts, which need four elements in order to be carried out perfectly: the instrument (ʾāla), such as the wood of the carpenter, the silver of the silversmith and the ʾalfāḏ̣ of the poet or the prose writer; this is the material (al-ʿilla l-hayūlāniyya) that has been already mentioned as the basis (ʾaṣl); second, the carrying out of the artisan’s aim (al-ʿilla ṣ-ṣūriyya); third, the checking that there are no defects or imperfections (al-ʿilla l-fāʿila); and, fijinally, the checking that there is nothing superfluous and nothing lacking (min ġayri naqṣin minhā wa-lā ziyādatin ʿalayhā wa-hiya l-ʿilla t-tamāmiyya). These paragraphs agree in considering the ʾalfāḏ̣ as the “matter” of the craft, but a closer look at the texts to which they belong shows that the two authors do not refer, in using these terms, to the same reality. In fact, ʾĀmidī continues: if the craftsman, after he fulfijils all these conditions, succeeds in producing a beautiful and unusual maʿnā, this enhances the value of his craft, nevertheless even without this, the craft does already exist, in itself. Since a line or a poem, which are the fijinished product of the art, doubtless have for ʾĀmidī a ‘meaning’ in the linguistic sense, it follows that in this context ʾĀmidī means by ʾalfāḏ̣ the whole utterance in its components, signifijier and signifijied, and by šiʿr, both the ‘content’ and the ‘form’. On the other hand, Xafājī’s attitude is diffferent, since he maintains that the ʾalfāḏ̣ must be considered merely sounds. These authors share the same idea of poetry as ‘fashioning’ (taṣwīr), but the textual scrutiny of their statements shows that they do not relate the main technical terms (maʿnā, ʾalfāḏ̣ , kalām, etc.) to the same reality. In the bipolarization formcontent, the part assigned to the “form”, which from a linguistic point of view includes signifijier and signifijied together, can be assigned sometimes to the maʿnā/maʿānī sometimes to the kalām/ʾalfāḏ̣ .

112 Jurjānī maʿnā kalām

lidia bettini

(‘content’) (‘form’, signifijier and signifijied)

matter result, product of the craft

(‘content’) (‘form’, signifijier and signifijied)

matter result, product of the craft

(the language)

matter

(merely sounds) (signifijier and signifijied)

matter result, product of the craft

matter

maʿnā

(signifijier and signifijied) ‘content’

šiʿr

(poetry)

Qudāma (Naqd) maʿānī šiʿr Qudāma (Xarāj) luġa Xafājī ʾalfāḏ̣ maʿānī ʾĀmidī ʾalfāḏ̣

an element whose originality it is not essential to the achievement of the craft result, product of the craft.

In other words, those who regard the maʿnā (in the meaning of ‘content’) as the matter of the craft, conceive its linguistic formulation (which can be called šiʿr or kalām) as signifijier and signifijied; whereas those who, like Xafājī, regard the ʾalfāḏ̣ as the matter, use maʿnā to mean the whole utterance. The fact that not only lafḏ̣ , as we have seen, but also maʿnā is employed to denotate both the image to be conveyed and its verbal wording and that maʿnā can be used by the balāġiyyūn to refer to the whole utterance, is, as many examples show, not rare. For example, Qudāma (Naqd: 23–24) when he talks about the qualities of the maʿānī (‘images, ideas’) the poets express, states at fijirst that in poetry there are two tendencies: one develops the maʿnā as far as possible (al-ġuluww fī l-maʿnā), while the other is more moderate (al-ʾiqtiṣār ʿalā l-ḥadd al-ʾawsaṭ). Amongst the examples Qudāma cites this line of Ḥassān b. Ṯābit:

on lafḏ̣ and maʿnā again

113

lana l-jafanātu l-ġurru yalmaʿna biḍ-ḍuḥā wa-ʾasyāfunā yaqṭurna min najdatin damā ‘Ours are the gleaming bowls that glitter in the forenoon and our swords let fall blood in drops because of our quickness in giving relief’.

All the criticisms of this line, which is discussed as being a maʿnā, regard the lexical choices of the poet, who ought to have employed the word bīḍ ‘white’ instead of ġurr ‘which have a blaze or a white mark on the forehead’, because ġurra denotes a small quantity of whiteness within a diffferent colour which is preponderant, or yajrīna ‘they flow (with blood)’ instead of yaqṭurna ‘they let fall in drops’.6 By maʿnā, then, it is necessary to understand here not the content, nor the image or idea, but rather the whole utterance. Similarly ʿAskarī (Ṣināʿatayn: 222), introducing the literary technique known as ḥall al-manḏ̣ ūm wa-naḏ̣ m al-maḥlūl ‘setting a verse into prose and a passage of prose into verse’, asserts that in this case the maʿānī are ready to be used: one just has to add or to take ‘something’ (šayʾan) away and he obtains a poetry line or a prose passage. He goes on to describe the various methods that can be adopted, such as, for example, inserting a word (bi-ʾidxāli lafḏ̣ atin bayna ʾalfāḏ̣ ih), or changing the position of another, etc. In these instances too we must understand by maʿānī not only the ‘content’ or the ‘images, ideas’ expressed, but the whole utterance. ʿAskarī himself, when introduces the examples, employs ʾalfāḏ̣ in the same context in which he had shortly before used maʿānī (wa-lam tazid fī ʾalfāḏ̣ ihī šayʾan: 223). When he describes the last type, which consists for the poet in “dressing the prose maʿānī (‘images, ideas’) with his own ʾalfāḏ̣ (‘wording’)”, the two terms correlated again take on the meaning ‘content’ and ‘form’.

6  This refers to the dripping in the bowls presented to guests. This well-known episode in which Nābiġa Ḏubyānī reproaches Ḥassān for his choice of some separate words (such as ġurr instead of bīḍ) or even for his choice of the forms of plural (like the plural of paucity jafanāt instead of jifān), because in his opinion the poet would not throughly express generosity and bravery ( fa-qāla lahū ʾanta šāʿir wa-lākinnaka ʾaqlalta jifānaka wa-ʾasyāfaka) is variously related by Marzubānī (Muwaššaḥ: 54–55), Muẓafffar Ḥusaynī (Naḍra: 228–229).

114

lidia bettini 2. Lafḏ̣ and maʿnā in the riwāya

More general statements about the relationship between these two fundamental ideas such as it is conceived among the balāġiyyūn are to be found in their considerations about the riwāya ‘oral transmission’ (of poetry, in our case). It is an established fact that writing was known in the pre-Islamic and early Islamic Arab world, even nomadic, and that it was used, to a certain extent at least, to record poetry.7 Nevertheless, the fact remains that the oral transmission holds a particularly high place in Arab tradition, and not only because written records did not replace memorizing at all. Xalīl for instance says: I never heard something without setting it down, I never set it down without memorizing it, I never memorized it without profijiting by it.8 Ṣūlī relates this anecdote: the poet Ḏū Rumma says to the grammarian ʿĪsā b. ʿUmar: set my poems down, because the (bedouin) Arabs forget the word that I spent a whole night searching for, they replace it by another with the same metre and then they recite (the line) in this way; the written word does not forget and does not replace.9 In other words, even if the advantages of written records were evident and well-known, oral transmission continued to thrive. In fact, the

7  In his beautiful book on the sources of pre-Islamic poetry, ʾAsad (1962) collects and quotes several direct examples of this; we will not deal here with the question of the use of writing for difffusing or “publishing” a work; on this question see Schoeler (2002) and for a recent updating Arazi (2004), Cheikh-Moussa (2006). 8  Baġdādī (Taqyīd: 114–115). 9  Jāḥiẓ (Ḥayawān: 1.41, with a slightly diffferent riwāya), Baġdādī (Taqyīd: 119). According to most of the traditions quoted by these two authors and by Bayhaqī (Maḥāsin: 1–16), writing was seen much more as a way of keeping and transmitting the already composed texts, than a way of helping to compose. The book, in the praises that Jāḥiẓ (Ḥayawān: 1.38–42, 50–54) and Bayhaqī (Maḥāsin: 1–16) sing of it, and that Baġdādī (Taqyīd: 117–126) repeats, is presented as a more accurate and faithful form of the voice. The book is a jalīs (‘a companion with whom one sits’) a ʾanīs (‘a cheerful companion’), a rafīq (‘companion’), a nadīm (‘a boon companion’) who, unlike human “companions”, does not lie, is not quarrelsome, does not divulge your secrets, keeps you company only when you agree, is a teacher from which you learn in one month more than in a lifetime from the voice of human beings, those human beings who, moreover, seek only profijit and regard themselves as being superior even if they are inferior to you (Jāhiẓ Ḥayawān: 1.51, Baġdādī Taqyīd: 120–122). The book brings back what memory lost (Taqyīd: 124), keeps what has been put in its care better than those placed in charge of it (Ḥayawān: 1.40, Taqyīd: 122); reading a book is more efffective and better than meeting the author himself (Taqyīd: 118). It is interesting to compare these statements with those of the upholders of the opposite view, namely that the only sound way for the transmission of the knowledge is the oral one, from the teacher to the disciple (samāʿ), and with their scornful judgements about the ṣuḥūfī ‘the one who gets his knowledge from the cahiers’, see Schoeler (2002: 121); on the praise for the book see the recent publication by Günther (2006).

on lafḏ̣ and maʿnā again

115

importance of the riwāya does not consist (only) in its preventing the memorized text from disappearing, it consists above all in the interiorizing of a tradition. Jāḥiẓ in his Risāla on the teachers10 distinguishes between looking at (qaraʾa, taṣafffaḥa) great authors’ works to get out of them given elegant expressions for immediate usage (li-yastafīda l-ʾalfāḏ̣ ) and looking at them to interiorize them (li-yastafīda l-maʿānī). Jāḥiẓ describes this using one of his similes: this second way is something that makes a nest for itself (ʿaššaša) in the heart, then it lays eggs, which hatch chicken which eventually learn to fly. There is a great diffference between this process during which the mind lets itself be permeated and that in which the mind picks (here and there, al-xāṭir muxtār) and wording is strained and arbitrary.11 Ibn Ṭabāṭabā (ʿIyār: 14) uses instead the image of an ingot cast from various metals or a river made up of various streams, when he advises the muwallad poet to persist in perusal (naḏ̣ ar) of the poetry he chose in his work, “so that the maʿānī and their very foundations become deeply rooted and close to his mind, and a raw material for his natural disposition”. In so doing the poet avoids making a direct borrowing of a maʿnā, which it would be pointless to try to veil by replacing a metre with another or a word with another. Ṣūlī (ʾAxbār: 38), while defending ʾAbū Tammām against attacks due in his opinion to ignorance, wonders how someone who not only is unable to compose an acceptable line, but does not even know by heart ten amongst the tens of thousands of maʿānī expressed by the poets, can dare to criticize. As we can see from these examples, the diffferent authors intend by maʿānī not only the conceptual content of the lines in question, but the whole linguistic form that conveys it; the traditional poetical heritage is composed in a inseparable manner from the images and their wordings. As for lafḏ̣ , it indicates in these contexts a linguistic form complete with signifijier and signifijied, but incomplete and partial from the poetical viewpoint, because it seems to refer in this case, as far as it seems to be understood, only to some given occurrences of a maʿnā.12

10

 Kitāb al-muʿallimīn, in Jāḥiẓ (Rasāʾil: 3.40–42).  The content of this Risāla is summarized by Hirschfeld (1922); we do not always agree with his translation. 12  See Jāḥiẓ (Kitāb al-muʿallimīn, in Rasāʾil: 3.42): the wrong way of memorizing is when one learns by heart “given utterances from a given book” (ʾalfāḏ̣ an bi-ʿaynihā min kitābin bi-ʿaynihī) “and then he wants to get ready for them their portion of ideas, or images” 11

116

lidia bettini

Let us consider what Ibn Rašīq says on this subject, when he maintains that the sariqa ‘literary theft’ occurs “in the unusual ornaments of style, namely in the utterances, that are the ʾalfāḏ̣ ” ( fī l-badīʿ an-nādir wa-ḏālika fī l-ʿibārāt allatī hiya l-ʾalfāḏ̣ ). The ordinary maʿānī (‘images, ideas’) as such are in fact easily accessible for everybody; it is difffijicult to make up something new and in this case the sariqa becomes evident.13 Ibn Rašīq provides the example of a line by Buḥturī describing a spear: ḥamalat ḥamāʾiluhu l-qadīmatu baqlatan min ʿahdi ʿādin ġaḍḍatan lam taḏbulī ‘His old shoulder-belts bore a green plant since the time of ʿĀd, fresh, not withered’

and a line by Ibn Muʿtazz: wa-yahuzzūna kulla ʾaxḍara kal-baq-lati māḍin ʿala l-qulūbi rasūbī ‘And they shake all the (swords) green like a plant, deeply penetrating into the hearts’.14

Another example of the borrowing of a maʿnā in the framework of the verbal embellishments (here mubālaġa):15 Imruʾ al-Qays describes the brightness of a woman’s face: yuḍīʿu l-fijirāša wajhuhā li-ḍajīʿihā ka-miṣbāḥi zaytin fī qanādīli ḏubbālī ‘Her face illuminates the bed to her bedfellow like an oil lamp in candlesticks the wicks of which are lighted’;

(ṯumma yurīdu ʾan yuʿidda li-tilka l-ʾalfāḏ̣ i qismahā min al-maʿānī), namely he starts from a given lafḏ̣ (which obviously conveys a given image, or idea, maʿnā) but which has also a given linguistic meaning which does not necessarily fijit for other possible occurrences of this maʿnā. 13  Jurjānī (Dalāʾil: 390) expresses the same idea in his own way, when he says that in order to veil the borrowing of a maʿnā, the later author must present it in a picture, a shape (ṣūra) diffferent from that which it had (in his forerunner’s poem). Jurjānī calls ṣūra that which Ibn Rašīq calls ʾalfāḏ̣ , because for him lafḏ̣ means only the separate word. He says: if he who borrows a maʿnā does no more than substitute a lafḏ̣ (a word) with another, it is impossible for the borrowing to remain undiscovered. Later on Ibn Rašīq (Qurāḍa: 55) expresses the same idea more clearly: what is to be considered a sariqa ‘theft’ in fact, is not the borrowing of the maʿānī which allegedly have been “invented” or expressed for the fijirst time by a poet, but rather the borrowing of their organisation (tartīb) and of the ways to achieve them. 14  Ibn Rašīq (Qurāda: 20). The common image is that of a fresh green plant to which the spear is compared because of its pliableness. 15  Ibn Rašīq (Qurāda: 31–32).

on lafḏ̣ and maʿnā again

117

this maʿnā, always in the form of a hyperbole, is borrowed by Ibn Muʿtazz: ʾalṯimuhū fiji d-dujā wa-barqu ṯanā-yāhu yurīnī mawāḍiʿa l-laṯmī ‘I kiss his mouth in the darkness and the lightening of his teeth shows me where to kiss’.

3. The μaʿnā and the Linguistic Meaning Nevertheless, the balāġiyyūn, and not only Jurjānī, are quite able to distinguish the maʿnā as the ‘semantic content of an utterance’ from the maʿnā as ‘image, idea’. Qudāma16 enumerates amongst the kinds of iʾtilāf ‘combination’ between lafḏ̣ and maʿnā, the ʾirdāf, the defijinition of which runs: a poet who wants to express a maʿnā (an ‘idea’, such as the length of the neck, ṭūl al-jīd, of a woman or her ease of life, tarafffuh), does not use the terms relating distinctively to this maʿnā (‘meaning’), but he uses another wording which has another maʿnā (‘meaning’), from which the former follows (lafḏ̣ un yadullu ʿalā maʿnan huwa ridfuhū wa-tābiʿun lahū; lafḏ̣ un yadullu ʿalayhi wa-laysa huwa l-mawḍūʿa lahū).17 This formulation, evidently, is not far removed from that used by Jurjānī to introduce his notion of maʿnā l-maʿnā (the image or idea called up by the linguistic meaning of an utterance): yadulluka l-lafḏ̣ u ʿalā maʿnāhu llaḏī yaqtaḍīhi mawḍūʿuhū fī l-luġa ṯumma tajidu li-ḏālika l-maʿnā dalālatan ṯāniyatan taṣilu bihā ʾilā l-ġaraḍ.18 In the same way, our authors are aware of the fact that each change in the lafḏ̣ involves a change in the maʿnā. This is constantly true in at least two fijields: the analysis of a line according to the ʾiʿrāb and the lexical fijield.19 The linguistic meaning of a given line is scrutinized with meticulous attention in works such as the Kitāb aš-šiʿr ʾaw šarḥ al-ʾabyāt al-muškilat al-ʾiʿrāb by ʾAbū ʿAlī al-Fārisī. This book is founded on the syntactic analysis it is necessary to carry out on some grammatically difffijicult verses, in order to

16  Qudāma (Naqd: 88), token up by Bāqillānī (ʾIʿjāz: 71), ʿAskarī (Ṣināʿatayn: 360), Xafājī (Sirr: 221). 17  Xafājī (Sirr: 221); e.g. the poet says: baʿīdat mahwā l-qurṭi ‘the one whose earring hangs far away’, or naʾūm aḍ-ḍuḥā ‘the one who sleeps until forenoon’, according to two wellknown ancient metonymies. 18  Dalāʾil: 202–203; Jurjānī (Dalāʾil: 227) too employs the word ridf. 19  See Rāzī (Nihāya: 90).

118

lidia bettini

assign them correctly a maʿnā (‘meaning’), or, if several analyses are possible, to notice that there are more than one maʿnā. We might consider, for instance, this line of Farazdaq: (chapter on the elision of the xabar of a mubtadaʾ: 275–276): wa-baytāni baytu llāhi naḥnu wulātuhū wa-baytun bi-ʾaʿlā ʾīlyāʾa mušarrafū ‘(We have) two abodes: of the house of God we are the wardens and of an abode high up on the holy city, eminent (we are the wardens)’.20

ʾAbū ʿAlī says: we must not regard baytu llāhi as a xabar of an elided mubtadaʾ; this would be possible and the meaning of the line would be: ‘(we have) two abodes: (one of them) is the house of God, of which we are the wardens; the other is an abode, etc.’ But the poet boasts that they are put in charge of the two abodes, hence we must consider naḥnu wulātuhū as a xabar of baytu llāhi, so that we can imply it as virtually existing in the second phrase with the same function. The sentence thus becomes analogous to: Zaydun ḍarabtu ʾabāhu wa-ʿAmrun, by which is meant: Zaydun ḍarabtu ʾabāhu wa-ʿAmrun [ḍarabtu ʾabāhu]. Another example, a line by ʿAlqama: wa-qad ʾuṣāḥibu fijityānan šarābuhumū xuḍru l-mazādi wa-laḥmun fīhī tanšīmū ‘Often I keep company with strong young men, whose drink is the greenish (water) of a leathern water-bag (and whose food) is a malodorous meat’.

According to ʾAbū ʿAlī, laḥmun fīhi tanšīmu must be considered as a xabar of an elided mubtadaʾ. Otherwise the line would mean: ‘their drink is the greenish (water) of a leathern water-bag and a malodorous meat’.21 Jurjānī, in his commentary on a grammatical treatise of ʾAbū ʿAlī himself, the ʾĪḍāḥ,22 offfers other examples of the decisive part the grammatical analysis plays in establishing the meaning unquestionably. We might consider these two verses of Imruʾ al-Qays:23

 ʾĪlyāʾ: madīnatu bayti l-maqdisi, see Lisān: s.v. ʾyl.  Abū ʿAlī (Šiʿr: 22). 22  Jurjānī (Muqtaṣid) see Larcher (1993: 250–251). 23  Jurjānī (Muqtaṣid: 342–344). The verses are quoted in several philological and poetical works, see for instance Marzubānī (Muwaššaḥ: 25), Ibn Rašīq (Umda: 2.37); see Jurjānī (Muqtaṣid: 342, fn. 213). 20 21

on lafḏ̣ and maʿnā again

119

fa-law ʾanna mā ʾasʿā li-ʾaḏnā maʿīšatin kafānī wa-lam ʾaṭlub qalīlan/un min al-mālī wa-lākinnamā ʾasʿā li-majdin muʾaṯṯalin wa-qad yudriku l-majda l-muʾaṯṯala ʾamṯālī ‘Strove I to earn the minimal subsistence a little money would be sufffijicient for me and I would not ask for (anything else) but I strive to earn a fijirmly established glory and a fijirmly established glory (is something that) he who is like me can attain.’

Establishing the exact maʿnā (‘meaning’) is very important in this case, because several critics regard these lines as being among the most vainglorious verses of the Arabs.24 The question is whether qalīl is governed by lamʾaṭlub (and its case will be accusative), or by kafānī (and its case will be nominative). Jurjānī proceeds as follows: the way to elucidate (bayān) that must be followed is by carefully checking the meaning of law. The meaning of a sentence governed by this conjunction is that something is impossible because something else is impossible; in other words the apodosis of law is impossible because the protasis is impossible. Therefore when the poet says law saʿaytu kafānī, this means that he is not content and that he does not strive. Moreover, if the apodosis consists in a negative sentence (wa-lam ʾaṭlub), this is in fact an afffijirmative statement: “after law the afffijirmative is negative and the negative is afffijirmative” ( fal-muṯbatu baʿda law manfijiyyun wal-manfijiyyu muṯbatun: 343). Therefore, if the search for a minimal way of subsistence is negated, the contentment too is negated and what is afffijirmed is that the poet asks for (something). If at this point we consider qalīl as governed by lam ʾaṭlub, we obtain a meaning that is self-contradictory from two viewpoints: if the contentment is negated (because it is the jawāb law), the fact of asking for a small amount of money (which is in fact afffijirmed, being lam ʾaṭlub a jawāb law in a negative form) is openly self-contradictory, because the sentence would signify that the minimal subsistence and a small amount of money is not enough for him and at the same time that he asks for a small amount of money. The other point of view is that this reading contradicts with the meaning conveyed by the second line, where the poet asserts that he asks for a fijirmly established glory. Therefore the contentment with a small amount

24

 See for instance Marzubānī (Muwaššaḥ: 25–26), Muẓafffar Ḥusaynī (Naḍra: 406).

120

lidia bettini

of money is negated and the search, not for a small amount of money but for power (mulk), is afffijirmed. In conclusion, Jurjānī succeeds in determining the meaning which meets the requirements of poetry, namely that a great poet and personage such as Imruʾ al-Qays cannot state that he easily satisfijied, through the analysis of the whole sentence and of its syntactic construction. Another case shows the subtlety and rigour of Jurjānī’s linguistic thinking, even if other critics reached the same meaning using other methods: ʾastaġfijiru llāha ḏanban lastu muḥṣiyahū rabba l-ʿibādi ʾilayhi l-wajhu wal-ʿamalū ‘I beg God’s forgiveness for (my) faults of which I do not know the number, the Lord of (His) servants, to Him the prayer and the worship (are directed)’.

Jurjānī proceeds as follows:25 all the authorities maintain that this verb governs a preposition (istaġfartu min ḏanbin) and that the occurrence of this line, for instance, is explained by means of an elision of the preposition. But Jurjānī says that he thinks that the matter is not as they maintain. This derived form of the verbs, when it has the meaning of a request (as it has in this case istaġfartu llāha ‘I asked God to forgive’) and when the radical verb is transitive, governs two direct complements. In our case the verb ġafara is transitive: we say ġafara llāhu ḏanban ‘God forgave for a fault’ and not min ḏanbin, as is proved by the explanations ad hoc that the Qurʾānic passage yaġfijir lakum min ḏunūbikum (xlvi 31 and lxxi 4) requires. Therefore we must consider that istaġfara correctly governs two direct complements and that when this verb governs a preposition (min), this is a particular case ( farʿ) which applies when the verb is used with the meaning of ‘I repented for a fault’ (tubtu min ḏanbin). In another case, Jurjānī says, about a nominal sentence introduced by kāna, that if there are a defijinite and an indefijinite noun, the ism kāna must be the defijinite one and its xabar, the indefijinite one: the discordant occurrences that can be found in poetry must be considered as instances of poetic licence. The line he (Muqtaṣid: 403–404) quotes is the following: ka-ʾanna sabīʾatan min bayti raʾsin yakūnu mizājahā ʿasalun wa-māʾū

25

 Jurjānī (Muqtaṣid: 614–617); the line is quoted by Sībawayhi without attribution.

on lafḏ̣ and maʿnā again

121

‘As if (on her mouth) there was a wine from Bayt Raʾs mixed with water and honey’.26

Jurjānī relates that some critics tried to provide a diffferent explanation for this syntactic construction, rather than accepting it as a instance of a poetic licence. They argued that ʿasalun does not indicate a given variety of honey nor does al-ʿasalu, but that it is question of a generic noun for ‘honey’ (yurādu bihī l-jins); therefore there is no diffference between the defijinite and the indefijinite form. Jurjānī asserts that a defijinite noun cannot have the same meaning as an indefijinite one, no matter whether it denotes a generic or a particular sense of a noun ( jinsan kāna ʾaw ʿahdan) and he sets against these arguments a grammatical one: if we say ʾaxadtu min al-ʿasali l-ḥulwi mizzahū ‘I took the best part of the sweet honey’, it would not be possible to make the attribute agree with the noun as if it were an indefijinite noun and to say: *min al-ʿasali ḥulwin, so as we say min ʿasalin ḥulwin ‘of a sweet honey’. Jurjānī therefore concludes that the line must be explained as poetic licence and that there is not a great diffference between the construction of this line and the incorrect sentence *yakūnu Zaydan muntaliqun.27 ʾAbū ʿAlī was a grammarian and it was his pupil28 and commentator, the balāġī Jurjānī, who drew stylistic conclusions from this attitude. In fact, it is not a question of choosing the case endings to be set, nor of using the grammatical analysis to distinguish “a correct utterance from an incorrect one”, according to the defijinition of the function of the ʾiʿrāb given by the muʿtazilite grammarian ʿAlī b. ʿĪsā ar-Rummānī. The question is rather of choosing, between two equally proper analyses of a line, the one which enables us to give it the maʿnā the poet means, and so of taking into account those phenomena that Jurjānī systematised under the name of maʿānī n-naḥw.29

26

 This line is by Ḥassān b. Ṯābit and it is widely quoted in the philological works, see the references in Jurjānī (Muqtaṣid: 404, fn. 45). The insertion ‘on her mouth’ (ʿalā ʾanyābihā) is drawn from the beginning of the following line; the line is quoted because the ism kāna would be a defijinite noun and it is instead an indefijinite one, the xabar would be an indefijinite noun and it is a defijinite one. 27  Jurjānī adds that his šayx has got another riwāya, received from Abū ʿUṯmān al-Māzinī: yakūnu mizājuhā ʿasalan wa-māʾū (ka-annahū qāl: wa-hunāka māʾun). 28  He was, to be precise, actually the pupil of Abū ʿAlī’s pupil and nephew ʾAbū l-Ḥusayn Muḥammad al-Fārisī, whom he often calls šayxunā. 29  In short, how the choice of one or another syntactical structure determines every nuance of meaning and then every poetical efffect. See, for instance, the syntactic analysis of a line of Abū Tammām carried out by Jurjānī (Dalāʾil: 283–284): it is similar to those

122

lidia bettini

It is possible to assert that the Arab theorists were aware of the relationship between lafḏ̣ and maʿnā in the linguistic meaning of ‘signifijier’ and ‘signifijied’. But this awareness appears, on reading the texts, to be discontinuous, since lafḏ̣ and maʿnā in their correlation have got a very large and variable range of uses, as we have seen; in particular it has to be stressed that a strict correlation between signifijier and signifijied does not represent the viewpoint from which the reflections of the balāġiyyūn develop. In fact grammatical analysis can ascertain whether two utterances can be accepted as equivalent on the linguistic level, but the balāġa is a science separated from the other sciences of the language. This is proved by the diffferent attitude of the same author towards the texts, depending on which fijield of these sciences he is acting in. Rummānī, in his commentary to the Kitāb of Sībawayhi, emphasises the relationship which ties the casual endings to the meaning and more generally the fact that if the lafḏ̣ varies, so does the maʿnā.30 Rummānī gives this sentence as an example: marartu bi-rajulin ḥasan ʾabūhu. The three possible meanings of this utterance require three diffferent possibilities in the determination of the case ending of the noun ḥasan. First, ḥasan can be a proper noun (the sentence would then mean ‘I passed by a man whose father is called Ḥasan’); here the case ending can be only the raf ʿ. Second, ḥasan can be an attribute “par excellence” (‘I passed by a man whose father is called “the Handsome One”’), and here the case ending can be the raf ʿ or the jarr. Finally, ḥasan can be a simple qualifijier (and in which case the sentence would mean: ‘I passed by a man whose father is handsome’); the case ending here will be the jarr.31 Rummānī concludes by saying that these diffferent possibilities show that we cannot assign correctly the ʾiʿrāb before we understand the meaning (lā yastaqīmu l-ʾiʿrābu ʾillā baʿda fahmi l-maʿnā).

considered until now, but Jurjānī inserts it in his general theory of the maʿānī n-naḥw, see Bettini (1987–1988: 97). 30  Wa-xtilāfu l-lafḏ̣ i li-xtilāfiji l-maʿnā wa-ttifāqu l-lafḏ̣ i li-ttifāqi l-maʿnā wa-mā xaraja min ḏālika fa-ʿalā jihati l-ʿāriḍ, see Dumayrī (1988: 2.120), see also Mubārak (1995: 252). The Šarḥ by Rummānī to the Kitāb has not yet been edited in its entirety; Ambros (1979) edited seven chapters (on hamz, numerals, broken plurals, some questions about the gender of the nouns), Dumayrī (1988) edited the chapters on ṣarf, Mubārak (1995) edited the chapters on istiṯnāʾ and on mā ʾaf ʿalahū. 31  Quoted in Mubārak (1995: 253). The attribute “par excellence” is called by Rummānī aṣ-ṣifa l-ġāliba; Mubārak (1995: 253 fn. 2), illustrates the meaning of this expression and states that he never met so subtle a diffferentiation in the works of the other grammarians.

on lafḏ̣ and maʿnā again

123

In another work the selfsame Rummānī32 compares the utterance al-qatlu ʾanfā lil-qatl ‘killing is the best way to prevent killing’33 to the Qurʾānic passage wa-lakum fī l-qiṣāṣi ḥayātun ‘and for you there is a life in the talion’ (ii 179). According to Rummānī, the preeminence of the Qurʾānic passage from the point of view of the balāġa, emerges thanks to four elements: it is richer in meaning (ʾakṯar fī l-fāʾida), more compact (ʾawjaz fī l-ʿibāra), not unwieldy because of the repetition (ʾabʿad min alkulfa bi-takrīr al-jumla), and more harmonious in the combination of the consonants (ʾaḥsan taʾlīfan bil-ḥurūf al-mutalāʾima). In fact the Qurʾānic passage presents a great many positive notions (maʿānin ḥasana), such as the mention of justice, thanks to the use of the juridical technical term qiṣāṣ, of life, which is the purpose one aims at, and so on;34 as for terseness, Rummānī says, what is equivalent, in the Qurʾānic utterance, to al-qatlu ʾanfā lil-qatl is al-qiṣāṣu ḥayātun ‘the talion is life’ and when we count the consonants which make up each utterance, we see that the former numbers fourteen consonant and the latter ten.35 The comparison between the two utterances that share the same maʿnā (‘content, idea’) is made at the balāġiyyūn level: there is feeble interest in the binary correspondence between signifijier and signifijied, to the point of modifying the letter itself of the Qurʾānic text, and on the other hand there is much concern for the respect of the established conventions (brevity, harmony of sound), from the point of view of the form and the content as well.

32  Rummānī (Nukat: 71–72). In Rummānī’s opinion, grammar (naḥw) is an independent craft (ṣināʿa), which must not be mixed with what does not belong to it: for instance he criticizes Sībawayhi for giving lexical explications in his grammatical treatise (Mubārak 1995: 251–252). 33  These words are ascribed to the Sasanid king ʾArdašīr b. Bābak (the son of Pāpak) as advice to his successors, see Mubarrad (Balāġa: 92) who quotes the utterance as al-qatl ʾaqall lil-qatl, ʿAskarī (Ṣināʿatayn: 181), Xafājī (Sirr: 200–201), Ṯaʿālibī (ʾĪjāz: 5), who repeats some of the views of Rummānī; on the possibility of comparison between the Qurʾānic passage and other similar utterances, see Jurjānī (Dalāʾil: 202). 34  The allusion to such implicit contents by calling them maʿānī can be found in chapters dealing with ʾījāz, see below fn. 54. Jurjānī too (Dalāʾil: 224), quotes this Qurʾānic passage, but only to dwell upon the meaning of ḥayātun as an indefijinite noun; this fijits the context, because the talion can be a source of life in given circumstances only, had ḥayāt instead been a defijinite noun, this would have meant that talion is the source of life. 35  For similar quantitative remarks by other authors, see Simon (1993: 270 fn. 536), Xafājī (Sirr: 201).

124

lidia bettini 4. The Separate Word

What can be ranked as having the same maʿnā at the utterance level represents a very loose and varied set, as is shown, in another fijield, by the lines quoted to prove that one poet has borrowed from another. The different utterances which constitute these verses are deemed to share the maʿnā in question, although each of them has, of course, its own linguistic signifijication. The separate word can instead be given an unambiguous defijinition, since it was set up by the waḍʿ (‘the invention of the language’) to indicate only one thing or entity.36 Rāzī (Nihāya: 90–91) asserts that each word has one meaning; he who knows it understands (the utterance) quite well, he who does not know it, does not understand anything at all. Unlike the maʿnā (‘content’) of the sentence, the maʿnā (‘meaning’) of the separate word cannot be expressed in either a shorter or longer form; it can only be replaced by a synonym. Moreover, the same conditions apply to this synonym. As far as poetics is concerned, the balāġiyyūn are aware of this binary relationship between the separate lafḏ̣ and its maʿnā; it is even possible to state, in my opinion, that the basic feature of our theorists’ thinking in this respect, is the meticulous search for the gap or the consistency or the relation between sentence meaning and poetical meaning of an utterance. ʾAšjaʿ Sulamī, for instance, says:37 wa-ʿalā ʿaduwwika ya-bna ʿammi Muḥammadin raṣadāni ḍawʾu ṣ-ṣubḥi wal-ʾiḏ̣ lāmū fa-ʾiḏā tanabbaha ruʿtahū wa-ʾiḏā ġafā sallat ʿalayhī suyūfaka l-ʾaḥlāmū

36  The diffference in the way of meaning of the separate word and the sentence, according to the thought of the balāġiyyūn, is expressed clearly by Jurjānī (Dalāʾil: 388), in his commentary on the section in which he compares several couplets from verses supposed to share the same maʿnā. Jurjānī says (Dalāʾil: 388): when the ʿulamāʾ say, about two verses, that the maʿnā of the one is the maʿnā of the other, they do not mean that the two are exactly the same nor that their status is like that of two nouns which indicate the same thing have (like layṯ and ʾasad ‘lion’), rather they mean two “things” that share the same kind ( jins), although they have diffferent features, like the various kinds of jewels. 37  ʿAmīdī (Ibāna: 51). The verses were composed for Hārūn Rašīd at Raqqa, see Ṣūlī (ʾAwrāq: 76).

on lafḏ̣ and maʿnā again

125

‘Your enemy, o Muḥammad’s nephew, has two (enemies) lying in wait, (in the) morning light and (in the) darkness; when he is awake you frighten him and when he sleeps the dreams draw your swords forth against him.’

Mutanabbī: yarā fiji n-nawmi rumḥaka fī kulāhu wa-yaxṣā ʾan yarāhu fiji s-suhādī ‘He sees in his sleep your spear upon his kidneys and he fears to see it in his sleeplessness.’

ʿAmīdī is highly critical of this line because by suhād the poet means the waking state (al-yaqaḏ̣ a) which makes a pair with the sleep (al-muṭābiqat an-nawm), but he spoiled the maʿnā (‘image’) since the word he uses actually means insomnia (intihāʾ al-karā laylan). Jurjānī too offfers some examples of a commentary on the proper or improper use of a separate word: ʿAbbās b. al-ʾAḥnaf:38 sa-ʾaṭlubu buʿda d-dāri ʿankum li- taqrubu wa-taskubu ʿaynāya d-dumūʿa li-tajmudā ‘I will search for going away from you so that you come near and my eyes will shed tears so that they get dry’.

In both hemistichs the poet states the same idea, namely that he will embrace the pain of the separation now, in order to enjoy the delight of the reunion in the future; Jurjānī in his comments on the second hemistich says that the lexical choice is suitable as far as the fijirst part is concerned, because shedding tears is a sign of sadness; in the second part, on the other hand, the verb li-tajmudā referring to eyes does not fijit the context, since it must be used when eyes are dry even if the situation demands tears (maʿa ʾanna l-ḥāla ḥālu l-bukāʾi: 208). Two other famous examples: Muslim b. al-Walīd says: ʿāṣā š-šabāba fa-rāḥa ġayra mufannadi wa-ʾaqāma bayna ʿazīmatin wa-tajalludī 38  Jurjānī (Dalāʾil: 207), see Mubarrad (Balāġa: 62), ʿAskarī (Ṣināʿatayn: 225), Qāḍī Jurjānī (Wasāṭa: 234, within a group of verses considered as having the same maʿnā), ʾĀmidī (Muwāzana: 1.71.14).

126

lidia bettini ‘He disobeyed youthfulness and journeyed, not blamed, and he remained between resolution and endurance’.

This line is criticized because the poet employs the verbs rāḥa ‘to go (from one place to another)’ and ʾaqāma ‘to remain’ together and so he lets the man he eulogizes move and stay at the same time.39 Kuṯayyir ʿAzza says: ʾa-lā ʾinnamā laylā ʿaṣā xayzurānatin ʾiḏā ġamazūhā bil-ʾakufffiji talīnū ‘Undoubtely Layla is a stick of cane when one puts a hand on her she bends’.

The poet Baššār reproaches Kuṯayyir for making Layla coarse and harsh by the choice of word xayzurāna, used moreover with ʿaṣā ‘stick’.40 This kind of analysis depends on the close examination of the lexical meaning of the words chosen by the poet.41 In order to judge the suitability of a word in a line, the critic needs not only a somewhat punctilious lexicographical sensibility, but also some extralinguistic parameters such as the detailed knowledge of the poetical conventions, namely the collective set of rules which govern the exchange between the poet and his audience and require a given meaning in a given context.42 Jurjānī in the example quoted, to show how the lexical fijield of the jumūd is not suitable when we talk about eyes which are dry because circumstances are happy (like in this case the lovers’ meeting), examines at length the use of this lexical fijield in poetry, which applies to eyes only when they are “stingy” with their tears in situations which would require sadness; stinginess is by no means a praiseworthy quality in the Arabic poetical tradition and in this case eyes (and the person to which they belong) are the object of blame and disapproval. The process which enables the reader to get to the meaning of an utterance through the knowledge of the poetical conventions is described in some details by Jurjānī in connection with the kināya ‘metonymy’, which

39  See Ibn Qutayba (Šiʿr: 806), Marzubānī (Muwaššaḥ: 246), Ibn Rašīq (ʿUmda: 2.245– 246). 40  ʿAskarī (Ṣināʿatayn: 219). 41  Many other examples can be found in Marzubānī (Muwaššaḥ: e.g. pp. 40, 41, 69 and on p. 54–55 the line of Ḥassān b. Ṯābit already quoted above). 42  Jurjānī’s analysis of the psychological efffects of istiʿāra in ʾAsrār al-balāġa and his distinction between metaphor and simile are mainly based on the collective knowledge of the rules of the exchange between the poet and his audience; this knowledge is required for a proper understanding of the poetry.

on lafḏ̣ and maʿnā again

127

he regards as a quality concerning the lafḏ̣ (in the case he cites an obligatory construction like the ʾiḍāfa) and not the naḏ̣ m ‘(syntactic) construction of the separate words’, namely the sentence.43 Jurjānī says: when we fijind in poetry a phrase like kaṯīr ramād al-qidr ‘having a lot of ashes on his kettle’, used to eulogize someone, we must assume that having a lot of ashes on the kettle is not in itself a motive of praise, and therefore we must also assume that in their allusion to the ashes the poets wish to suggest that the eulogized one cooks a great deal and that he is therefore very generous and hospitable. In order to reach this interpretation, the lexical meaning alone does not sufffijice, we must clearly proceed by means of reasoning (maʿqūl). A particular kind of relationship between lafḏ̣ and maʿnā with regard to a separate word are the diffferent occurrences in which the lexical meaning of the word considered implies in the context another meaning. The fijirst instance is the use of metaphors such as raʾaytu ʾasadan44 ‘I saw a lion’. Jurjānī dwells upon the diffference between this sentence and Zaydun ʾasadun ‘Zayd is a lion’ in which the separate word ʾasad is also employed. In both cases the use of this word is obviously based upon a simile, but in the latter the two terms of the simile are mentioned, whereas in the former this is not the case and the purpose of using a simile remains “harboured in the soul, concealed in the heart”; the process of fusion between the two terms of the comparison this involves is such that the sentence seems to refer to the man and the lion at the same time.45 The lexical meaning of the word remains the same both in its proper and metaphorical usage46 and there is no transfer of a noun from its basic meaning to denote something else, as the current defijinitions of the istiʿāra maintain.47 Therefore, a separate word too is likely to connote more than its linguistic meaning does, even if its lexical meaning does not change and even if this takes place by means of a rather subtle mental or psychological process and not by means of defijinable syntactic choices as happens 43

 Dalāʾil: 329–331, see Abu Deeb (1979: 76–77; 164–165).  Or ʿannat lanā ḏ̣ abya ‘a gazelle presented itself to us’ (Jurjānī ʾAsrār: 296). 45  Jurjānī (ʾAsrār: 298–299), see Abu Deeb (1979: 179–180; 187). A later scholar, Sakkākī (Miftāḥ: 358; 371–373), holds himself aloof from this position by introducing the notion of taʾwīl ‘interpretation’ in order to avoid the undesirable complete identifijication of the man with the lion, including, for example, the physical aspect of the animal. 46  Jurjānī (Dalāʾil: 280, 334). 47  Jurjānī (Dalāʾil: 280 and passim). For an inquiry into the history of the meanings of the term istiʿāra see Heinrichs (1977; 1984). A description of Jurjānī’s theory of poetical imagery goes beyond the aims of this paper and has in any case been undertaken many times, see e.g. Abu Deeb (1979), Ṣammūd (1994). 44

128

lidia bettini

in the sentence. There is, moreover, the particular phenomenon called by Rummānī taḍmīn48 ‘implication’, according to which qātil ‘killer’ from both a formal and a semantic point of view is a word which by its schema (binya, namely of an active participle) implies qatl ‘killing’ and maqtūl ‘killed’, “without the need to mention any adjective or noun which may be considered to express that idea”.49 In another chapter of the same work, Rummānī goes back over the question of the meaning implied by the separate word by means of its ištiqāq.50 The author of an early anonymous commentary on Nukat51 does not agree with the way Rummānī presents this question. For him nouns, verbs and so on imply no more than their lexical meaning, unless their user intends them to; taḍmīn ‘implication’ is, thus, “the act of he who implies” (at-taḍmīn fijiʿl al-muḍammin), namely it is the result of a conscious choice of his own for given expressive purposes and not a quality of words in themselves. He nevertheless points out that the language (kalām) of most great authors implies much more than it expresses or utters and that there are many kinds of taḍmīn, but he does not want to deal with the matter since he limits himself to a commentary on the text of the Risāla.52

48  Rummānī (Nukat: 102). Carter (1984: 228–230) deals in detail with the way in which Rummānī employs taḍmīn (which “has nothing to do with the normal rhetorical fijigure of the same name”: 228, fn. 26) not only in Nukat but also in his al-Ḥudūd fī n-naḥw. According to Carter (1984: 228), who underlines the muʿtazilite side of Rummānī’s thinking, “with taḍmīn we come to the most original and potentially the most fruitful [ . . . ] concept in Nukat, by means of which al-Rummānī evidently intended to create a systematic justifijication for taʾwīl on a grand scale and which, moreover, he proceeded to put into practice in his Qurʾān commentary”. 49  Rummānī (Nukat: 102), translation by Carter (1984: 228). 50  Rummānī (Nukat: 107); the author says that the bayān fī l-kalām (to be understood here as ‘to express oneself through the language’, without literary connotations), can be made up only of a noun or an attribute (ism ʾaw ṣifa) or as the result of a “junction” which does not contain a noun or an attribute having that meaning (taʾlīf min ġayri smin lil-maʿnā ʾaw ṣifa), for instance ġulām Zayd ‘the servant of Zayd’ denotes the “ownership”, without resort to a noun or ṣifa expressing it directly. Similar to taʾlīf according to Rummānī is ištiqāq, here the formation of the nominal schemas by derivation, for this reason qātil denotes also the killed one and the act of killing. 51  This commentary, which according to the editor Z. Saʿīd ʿAlī dates back to the 5th century, is not complete; in particular neither the title nor the author of the explained work are quoted anywhere. The editor has easily identifijied the Risāla the commentary deals with as Rummānī’s Nukat; as for the author of the commentary, he thinks that it is likely to be (ka-ʾannahū) an early work of Jurjānī. 52  Šarḥ Rummānī: 124–125.

on lafḏ̣ and maʿnā again

129

5. The ʾījāz Another similar way53 of conveying “many ideas using a small number of words (lafḏ̣ )” is called ʾījāz ‘terseness’, according to the defijinition many authors give.54 We have seen an example of this in Rummānī’s reflections on the maʿānin ḥasana of the Qurʾānic passage wa-lakum fī l-qiṣāṣi ḥayātun (ii 179);55 also as far as ʾījāz is concerned, for a proper understanding of a very concise utterance it is necessary to place it in its cultural context.56 53  On the connection between taḍmīn and ʾījāz see Rummānī (Nukat: 103) and Šarḥ Rummānī: 127–129. 54  ʾIṯbāt al-maʿānī l-kaṯīra bil-lafḏ̣ al-qalīl, for instance Ibn Zamlakānī (Tibyān: 110), who says that ʾījāz is also called ʾišāra; the examples he quotes are examples of implicit notions that can be drawn from some Qurʾānic passages. For instance the passage fa-ṣdaʿ bi-mā tuʾmaru, ‘come out openly with what was ordered to you’ (xv 94), sums up the meaning of the Message, the passage xuḏ al-ʿafwa wa-ʾmur bil-ʿurfiji wa-ʾaʿriḍ ʿan al-jāhilīn ‘tax the superfluous, order what is correct and avoid the ignorant’ (vii 199), sums up the noble characteristics (makārim al-ʾaxlāq), for more details see also ʿAskarī (Ṣināʿatayn: 182–183). Jurjānī (Dalāʾil: 356–357) comments on the current defijinition of ʾījāz by saying that when we talk about “a great deal of meanings in a small number of words” we must bear in mind that the meanings the words have in those terse occurrences did not change with respect to their lexical meaning, and then this defijinition does not make any sense unless we understand it in this way: by means of the meaning implicit in the sentence meaning of the utterance, we attain notions ( fawāʾid) which would require a number of words if we wanted to express them explicitly. The use of the word fawāʾid, which is as vague as the maʿānin ḥasana of Rummānī, reveals in my opinion that Jurjānī shares with the other rhetoricians he criticizes the same appreciation for that which is unspoken and implicit in a literary utterance and therefore difffijicult to describe except vaguely (see e.g. Ibn Rašīq ʿUmda: 2.251, when he says that, in the ellipsis, the mind of the hearer expands in thinking and supposing, since what is known is easy but limited) but that he expresses it in another way, given the rigorous meaning he attributes to the terms lafḏ̣ and maʿnā. See, for instance, his statements about ḥaḏf ‘ellipsis’ (Jurjānī Dalāʾil: 112). 55  Rummānī (Nukat: 76) regards ʾījāz as one of the parts of the balāġa and states that when a maʿnā may be expressed by means of a long wording or a short, the short one (al-ʾalfāḏ̣ al-qalīla) is the ʾījāz. 56  It is important to distinguish between ‘terseness’ without ellipsis (qiṣar, by which one can choose between equivalent utterances the terser one) and the kind of terseness the Arab critics call ḥaḏf ‘ellipsis’ (Rummānī Nukat: 76–77, ʿAskarī Ṣināʿatayn: 181, Ibn ʾAṯīr Maṯal: 2.264), or sometime xaṣr or ixtiṣār (see e.g. Jāḥiẓ Ḥayawān: 1.91, Ibn Qutayba Taʾwīl: 210) ‘abridgement’, in which a kontextueller Indikator (Simon 1993: 268, fn. 529) clarifijies the elided part. In fact, this device (often exemplifijied by the Qurʾānic passage fa-sʾal al-qarya ‘ask the town’ [xii 82], to be understood as isʾal ahla l-qarya ‘ask the people of the town’), requires a reconstitution and must be put on a diffferent plane from that of two or more semantically equivalent utterances. Jurjānī too (Dalāʾil: 233–234) quotes this Qurʾānic passage to distinguish between instances in which an elision is somewhat foreseeable and mechanically reconstituted like this, and other poetical instances in which the reconstitution of the implied elision would destroy the poetical image. See the line of Xansāʾ he cites, where a she-camel (in fact the poetess herself) is described: fa-ʾinnamā hiya ʾiqbālun wa-ʾidbārū ‘truly she is a “going back and forth” ’. The she-camel, owing to her continuous going back and forth, became as if she was this movement embodied and

130

lidia bettini

Muẓafffar Ḥusaynī (Naḍra: 21) quotes this line of Ṭarafa as an example of a line in which “the maʿnā is more than the lafḏ̣ ”: wa-lastu bi-ḥullāli t-tilāʿi maxāfatan wa-lākin matā yastarfijidi l-qawmu ʾarfijidī ‘I am not one of those taking their abode in the bottom of valleys for fear, but when people seeks (for gifts) I give’.

Muẓafffar Ḥusaynī says that the poet means that he does not camp in hidden places for fear that he will be obliged to provide hospitality (maxāfata al-qirā), but on the contrary he camps in quite visible places, which visitors cannot miss, and when he is asked to provide hospitality he does not avoid it (wa-ʾiḏā stuqrītu qaraytu). But, Muẓafffar Ḥusaynī goes on, the poet’s words mean only that he does not settle in the bottom of valleys for fear, and only in the second hemistich, when ‘giving’ (ar-rifd) is mentioned, does one understands that ‘fear’ is the “fear of providing hospitality” (al-qirā). Nevertheless, the poet did not express this idea by means of the relevant utterance, but is satisfijied with the fact that the hearer knows and with the general content of his words (wa-lam yuqābil al-lafḏ̣ a [. . .] fa-ktafā bi-maʿrifati s-sāmiʿi wa-bi-mā dalla l-kalāmu ʿalayhi: 21). Thus, when Muẓafffar Ḥusaynī says that the maʿnā is more than the lafḏ̣ , he means that the competence needed for decoding the message is not only literary but also requires the knowledge of a social code of behaviour the values of which are extolled by the poetical conventions. We must distinguish this kind of ʾījāz from another one which deals with a point relevant to the topic of this paper, namely the equivalence of two or more utterances as far as maʿnā is concerned; this represents, from the viewpoint of the poetics, an important item in the description of the ʾījāz.57 Even before being ranked among the embellishments of the style and receiving its defijinitive technical noun ʾījāz, ‘terseness’ had been regarded since the earliest authors as the essential component, sometimes

if we reconstitute the elided words and say: ‘she is one who goes back and forth’, the poetical efffect would be destroyed. For a résumé of the later authors’ opinions, see Simon (1993: fn. 529). 57  See for instance Rāzī (Nihāya: 90–91), who states that the ʾījāz, like its opposite the ʾiṭnāb, may occur only within the ʾifāda maʿnawiyya, ‘conveying a meaning by a mental way, (i.e. by means of given syntactic or stylistic devices chosen by the poet)’ which allows the same maʿnā (‘content, idea’) to be expressed in diffferent, more or less complete, ways, a process that the dalālāt waḍʿiyya ‘lexical denotations’ do not allow because in Rāzī’s opinion only in the fijield of the dalāla waḍʿiyya does each change of lafḏ̣ necessarily involve a change of maʿnā ‘meaning’.

on lafḏ̣ and maʿnā again

131

the very defijinition, of the balāġa.58 The short epistle by Mubarrad on the balāġa which does not deal with ʾījāz, nor does it mention it, is mostly based upon the criterion that the more concise an utterance is, the more it is worthy and the closer it comes to perfection. The same kalām ‘act of speech’ expressed in one line instead of in two, is regarded as ʾajmaʿ wa-ʾaxṣar ‘more compact and terse’, for example:59 wa-tabrudu barda ridāʾi l-ʿarū-si raqraqta biṣ-ṣayfiji fīhi l-ʿabīrā wa-tasxunu laylata lā tastaṭīʿu nubāḥan biha l-kalbu ʾillā harīrā ‘(Even) in summer she feels cool like the chill of the girdle of the bride, on which perfume mixed with safffron has been sprinkled and she is warm in a night (that is so cold) that the dog cannot bark, but only whine’.

The poet ʾAʿšā expresses this in two lines; Ṭarafa is deemed as having done better than him, since he expresses it in one line only: taṭrudu l-barda bi-ḥarrin sāxinin wa-ʿakīka l-qayḏ̣ i ʾin jāʾa bi-qurrī ‘She chases away cold by her intense warmth and the strong heat of summer, when it comes, by coolness’.

Another example: ʿAntara says:60 fa-ʾiḏā šaribtu fa-ʾinnanī mustahlikun mā lī wa-ʿirḍī wāfijirun lam yuklamī wa-ʾiḏā ṣaḥawtu fa-mā ʾuqaṣṣiru ʿan nadan wa-kamā ʿalimti šamāʾilī wa-takarrumī ‘Thus when I drink I am ruining my fortune, while my honor remains in good standing unhurt; and when I am sober I am not remiss in generosity my qualities and my noble attitude are as you know (them to be)’.

The verses are praised, but Imruʾ al-Qays expressed the same idea in one line: samāḥata ḏā wa-birra ḏā wa-wafāʾa ḏā wa-nāʾila ḏā ʾiḏā ṣaḥā wa-ʾiḏā sakir

 See for instance Jāḥiẓ (Bayān: 1.96), ʿAskarī (Ṣināʿatayn: 38), Rummānī (Nukat: 76).  Mubarrad (Balāġa: 82–83), Marzubānī (Muwaššaḥ: 50). 60  Mubarrad (Balāġa: 84). 58

59

132

lidia bettini ‘This generosity, this benefijicence, this loyalty and this bounty, when he is sober and when he is drunk’.61

Later on Ibn ʾAṯīr talks expressly about two utterances having the same maʿnā;62 Baššār says: man rāqaba n-nāsa lam yaḏ̣ far bi-ḥājatihī wa-fāza biṭ-ṭayyibāti l-fātiku l-lahiju ‘He who is too mindful of people does not gain what he desires and he who is bold and eager attains the good things’.

His pupil Salm Xāsir63 says: man rāqaba n-nāsa māta ġamman wa-fāza bil-laḏḏati l-jasūrū ‘He who is too mindful of people dies in grief and he who is bold attains enjoyment’.

In the Risāla by Mubarrad we can fijind one of the most ancient occurrences of a Qurʾānic quotation which will become a standing one in the history of the ʾījāz: wa-lakum fī l-qiṣāṣi ḥayātun (ii 179). Mubarrad does not use the arguments of Rummānī, but rather he (Balāġa: 92) states that the Qurʾānic passage is clear whereas the other is ambiguous, since sometimes a murder, far from preventing other murders, is their cause.

6. The Ornaments of the Style We still need to look at another point of the relationship between lafḏ̣ and maʿnā: the delimitation of a maʿnā and the degree regarded as acceptable of the equivalences among utterances within the ornaments of the style

61  The translation of the last three examples is by Grunebaum (1941: 380–381), slightly modifijied. 62  Ibn ʾAṯīr (Maṯal: 3.258), in the chapter that deals with a type of borrowing considered an excellent one, namely that in which the poet formulates a borrowed maʿnā more tersely than his forerunner; so he proves his ability and skill in the speech art, see also ʿAskarī (Ṣināʿatayn: 220). Other examples of two verses having the same maʿnā and compared on the basis of their respective terseness can be found e.g. in Xafājī (Sirr: 206–207), who, moreover, takes into account the number of the words and, when it is the same, the perspicuity of the utterances. 63  Born in Baṣra, he lived in Baghdad and eulogized the caliphs Mahdī, Hādī and Hārūn.

on lafḏ̣ and maʿnā again

133

such as the ʾīġāl ‘epiphrasys’, the ḥašw ‘padding’ and in the sariqa ‘literary theft’. The delimitation of a maʿnā is dealt with by the balāġiyyūn in the framework of the difffijiculties a poet can face in putting three fundamental elements, the metre, the rhyme and the maʿnā (the ‘image’ or the ‘idea’) together within the limits of a line. According to Qudāma (Naqd: 96), it is necessary that the maʿānī (‘images, ideas’) are expressed wholly and perfectly in the line, without the poet having to reduce them or to insert a padding, because of the requirements of the metre and the rhyme. Therefore, the consistency of metre, rhyme and maʿnā is analysed to understand what stands below or beyond the maʿnā, what makes it unfijinished or redundant. Statements about this question can be found when the balāġiyyūn deal with ʾīġāl (or tablīġ), or with ḥašw; it is in this framework that the conformity between the extent of the line and that of the maʿnā can be considered. ʾĪġāl is a fijigure of style thanks to which a poet who fijinished the maʿnā before fijinishing the line, succeeds in doing it and in adding to its clearness and glamour. ʾĪġāl, even if not under this name, was well-known from the earliest time. In a well-known passage, ʾAṣmaʿī says that the best poet is the one who, having concluded what he wanted to express within a line before the rhyme, succeeds in producing it and adding something to its content.64 For example, this line of Imruʾ al-Qays is very often quoted: ka-ʾanna ʿuyūna l-waḥši ḥawla xibāʾinā wa-ʾarḥulina l-jazʿu llaḏī lam yuṯaqqabī ‘As if the eyes of wild animals around our tents and our dwellings were beads that have not been pierced’,

in particular by Ḥātimī, who quotes it twice, in the framework of the ḥašw and in the framework of the ʾīġāl. In the fijirst instance, it is the metre which is not complete, whereas the maʿnā would be so (ka-ʾanna ʿuyūna l-waḥši l-jazʿu llaḏī lam yuṯaqqab: the simile is tāmm), and the poet adds ḥawla xibāʾinā wa-ʾarḥulinā because of the metre, but in this way he adds to the maʿnā because the reference to the dead animals around his tent shows that he succeeded in hunting.65

64  Yanqaḍī kalāmuhū qabla l-qāfijiyati wa-ʾiḏā ḥtāja ʾilayhā ʾafāda bihā maʿnan, Qudāma (Naqd: 99), Ḥātimī (Ḥilya: 156), Ibn Rašīq (ʿUmda: 2.57), ʿAskarī (Ṣināʿatayn: 395). 65  Ḥātimī (Ḥilya: 191).

134

lidia bettini

In the second, it is of course the fijinal part which is considered as being beyond the maʿnā, and allaḏī lam yuṯaqqab adds to the maʿnā as well, because the comparison is more pertinent and better established.66 It results that the maʿnā is: ka-ʾanna ʿuyūna l-waḥši l-jazʿu ‘as if the eyes of wild animals were beads’. Another famous example is the line of Xansāʾ: ʾaġarru ʾablaju yaʾtammu l-hudātu bihī ka-ʾannahū ʿalamun fī raʾsihī nārū ‘Noble and splendid, the leaders follow his guidance as if he was a mountain on the top of which there is a fijire’.

Fī raʾsihī nāru is seen as a completion (takmīl) of an already concluded maʿnā.67

7. The sariqa Works devoted to sariqāt show the extent of the training. It is not simply a matter of borrowing an idea or a sentence, but of an intertextual dialogue68 made up by verbal borrowings, echoes, turns of phrases, as is shown by utterances considered by the theoricians as examples of sariqa, even if apparently neither lafḏ̣ nor maʿnā are involved. The technical terms the critics use show this: iʿtamada ‘he relied on’, ištaqqa ‘he derived’,69 lamaḥa ‘he glanced’.70 According to Ḥātimī, Mutanabbī stated that no poet, either pre-Islamic or Islamic, is faultless as far as borrowing is concerned and in order to prove this, he quotes several verses in which even the most famous amongst the ancient poets, Imruʾ al-Qays could be considered guilty of borrowing:

66  Ḥātimī (Ḥilya: 155–156), see also Qudāma (Naqd: 97), ʿAskarī (Ṣināʿatayn: 396), Bāqillānī (ʾIʿjāz: 92), Ibn Rašīq (ʿUmda: 2.58), who quote several other examples. 67  ʿAskarī (Ṣināʿatayn: 406); the line is quoted also in other contexts, see Ibn Rašīq (ʿUmda: 2.141). 68  Heinrichs (2007). 69  Ḥātimī (Mūḍiḥa: 57; 74; 143; 144 and passim). For the technical terminology of the sariqa see e.g. Bonebakker (1997) and Heinrichs (2007). 70  ʿAmīdī (Ibāna: 103; 109; 110; 111; 132).

on lafḏ̣ and maʿnā again

135

mikarrin mifarrin muqbilin mudbirin maʿan ka-julmūdi ṣaxrin ḥaṭṭahu s-saylu min ʿalī 71 ‘He (= the poet’s horse) attacks and flees, he goes forward and back together like a mass of rock which the torrent has caused to descend from above’;

in his opinion Imruʾ al-Qays relied on (iʿtamada) this line of ʾAbū Duʾād:72 minfaḥun miṭraḥun miʿannun mifannun mixlaṭun mizyalun jamūḥun xarūjū ‘Quickly he enters into the crowd, performs various forms of running, interferes vehemently in afffairs, overcomes his rider, outstrips others in the race’.

Statements about sariqa, of course, also touch the question of the limits of a maʿnā, because the core of this question is the comparison between two lines or more, by which a poet is caught out because the maʿnā he produced is not of his own, but it has been “stolen” by him from another earlier poet; the aim of the critics in such works, in particular in the works of malevolent critics, is to reveal the sariqa, no matter how the poet sought to conceal it.73 What is meant by maʿnā in these cases, and what are the limits in which two verses or more are considered as having the same maʿnā? We can take as an example the following verses describing the desert from ʿAmīdī (ʾIbāna: 26–27): al-ʿAlawī al-Kūfī, known as Ḥimmānī:74 tayhāʾu lā yataxaṭṭāha d-dalīlu bihā ʾillā wa-nāḏ̣ iruhū bin-najmi maʿqūdū ‘A desert in which no guide walks, unless the pupil of his eye is fijirmly fijixed on the stars’

71  ʿAskarī (Ṣināʿatayn: 466) quotes this line as an example of a poet who succeeds in producing a rhyme though he has a very small number of syllables at his disposal in order to complete the line. 72  Ḥātimī (Mūḍiḥa: 143–144). 73  We do not enter into the question of the sariqāt mamdūḥa (‘the praiseworthy literary theft’), namely the cases in which the second poet succeeds better than the fijirst in rendering the idea and so to say thus obtains rights to it, because this does not touch on the side of the problem which concerns us here, that is the limits of a maʿnā. For a general view of the sariqa, see Haddāra (1981) and Heinrichs (2007). 74  The only author named al-Ḥimmānī found by the editor of ʿAmīdī (Ibāna: 26, fn. 2), in biographical dictionaries is a muḥaddiṯ of Kufa dead in 228h, see Ziriklī (ʾAʿlām: s.v.).

136

lidia bettini

Mutanabbī: ʿaqadtu bin-najmi ṭarfī fī mafāwizihī wa-ḥurra wajhī bi-ḥarri š-šamsi ʾiḏ ʾafalā ‘I fijixed on the stars my eye in its waterless deserts and when the stars set (I fijixed) the ball of my cheek to the sun’.

Diʿbil: wa-dawiyyatin ʾanḍaytu fīhā maṭiyyatī wajīfan wa-ṭarfī bis-samāʾi muwakkalū ‘And in how many unhealthy lands I exhausted my steed, riding it quickly with my eye which trusted in the sky’.

According to the commentary of ʿAmīdī, who seldom gives any judgement, but limits himself to quoting two or more verses having in his opinion the same maʿnā,75 Diʿbil is the poet who best conveyed the maʿnā, because he was more concise and eloquent and added something (zāda) to those who preceded him.76 He does not think that the jinās of Mutanabbī enlivens the line, nor his mention of the sun, but in comparing the three verses he considers probably that the general mention of the sky encompasses in a concise way the stars and the sun altogether; moreover, within the line the poet inserts the theme of the steed exhausted by the journey. Another example:77 Naṣr b. Sayyār:78 wa-la-rubbamā nafaʿa l-ʿaduwwu bi-ʿaqlihī wa-la-rubbamā ḍarra ṣ-ṣadīqu l-jāhilū ‘Often the enemy is useful by means of his intelligence and often the stupid friend harms’.

This idea is borrowed, in ʿAmīdī’s opinion, by Mutanabbī in the following line: wa-ʾiḏā ʾatatka maḏammatī min nāqiṣin fa-hiya š-šahādatu lī bi-ʾannī kāmilū

75  See Heinrichs (1973: 58), who arranged “ʿAmīdī’s material according to whether or not the author comments on the cases of plagiarism discovered by him”. 76  ʿAmīdī (Ibāna: 27). 77  ʿAmīdī (Ibāna: 50). 78  The last governor of Khurasan under the Umayyads, dead in 131/748, see EI2: s.v.

on lafḏ̣ and maʿnā again

137

‘When he who defames me to you is a defijicient, this bears witness to my being perfect’.

Another example: Ṣāḥib al-ʿAlawī:79 ʾanā fī janābi siwāka fī marʿan nadin wa-ʾuqīmu ʿindaka fī janābin mujdibī ʾin kunta ḏā baṣarin fa-mayyiz faḍla mā bayna l-farāʾi wa-bayna ṣaydi l-ʾarnabī ‘In the country which is not next to yours, I live as in a moist pastureland, and next to you I live in a barren land; if you have good eyesight, distinguish the value (in the hunting) of wild ass from a hare’.

According to ʿAmīdī, Mutanabbī borrowed the maʿnā without the lafḏ̣ , since he turned the image (maʿnā) of the wild ass into a grey falcon and the hare into a vulture, and he comments on the line by saying that nobody is more clever at taking and changing the maʿnā: wa-šarru mā qabaḍathu rāḥatī qanaṣun šuhbu l-buzāti sawāʾun fīhī war-raxamū ‘The worst game my hand seized (is that) where a grey falcon and a vulture are equal’.

Another example:80 Qudāma b. Mūsā al-Jumaḥī,81 who describes a qalam: wa-ʾaṣfara yaxfā šaxṣuhū min nuḥūlihī yudāwī kamā yudwī wa-yaqḍī wa-yafṣilū yaxabbu bi-lā rijlin wa-yasṭū bi-lā yadin wa-yabkī bi-lā ʿaynin wa-yadrī wa-yajhalū ‘A yellow thing in which a person is concealed because of its leanness it cures just as it renders diseased and it judges and decides; it leaps without hind legs, and it paces without fore legs it cries without eyes and it knows and it ignores’.

Mutanabbī transferred (naqala) this line to describe the death: wa-mā l-mawtu ʾillā sāriqun daqqa šaxṣuhū yaṣūlu bi-lā kafffijin wa-yasʿā bi-lā rijlī

 According to ʿAmīdī (Ibāna: 125), he was dāʿī in Tabaristan.  ʿAmīdī (Ibāna: 81). 81  A Meccan muḥaddiṯ, d. 153/770, Ziriklī (ʾAʿlām: s.v.). 79

80

138

lidia bettini ‘What is death if not a thief with a tiny fijigure who springs without hands and runs without feet’.

What is meant by lafḏ̣ is hinted at in those passages in which authors, like ʿAmīdī, assert that a poet borrowed both the lafḏ̣ and the maʿnā. We can, for instance, compare the verses of Baššār b. Burd:82 ḥušāšatī waddaʿatnī yawma baynihimū wa-šayyaʿathum wa-xallatnī wa-ʾaḥzānī wa-qad ʾašārū bi-taslīmin ʿalā ḥaḏarin min ar-raqībi bi-ʾaṭrāfijin wa-ʾajfānī ‘The last remains of my breath bade me farewell on the day of their departure, they accompanied them and left me with my sorrows, after they had hinted a greeting with glances and winks of the guard to a spy’

and of Mutanabbī: ḥušāšatu nafsī waddaʿat yawma waddaʿū fa-lam ʾadri ʾayya ḏ̣ -ḏ̣ āʿinayni ʾušayyiʿū ʾašārū bi-taslīmin fa-judnā bi-ʾanfusin tasīlu min al-ʾāmāqi was-simu ʾadmuʿū ‘The last remains of my spirit bade farewell on the day they bade farewell and I did not know which of the two leaving groups I escort; they hinted a greeting and we shed abundantly our souls which flowed from the inner corners of the eyes, even if their name is tears’.83

Marwān b. Saʿīd:84 ʾinna l-jiyāda ʿarafna mahda dārihā fa-ṣahalna bākiyatan ʿalā sukkānihā ‘The excellent horses knew the place of their abode and they neighed weeping onto its dwellers’.

82

 ʿAmīdī (Ibāna: 151–152), see also Badīʿī (Ṣubḥ: 154).  Ibn Wakīʿ (Munṣif: 172–174) for each of the two verses of Mutanabbī quotes previous poets who expressed the image Ibn Wakīʿ considers relevant in the line, namely for the former, that the poet does not know which leaving group he escorts and for the latter, that which flows on the cheeks is not tears, in spite of appearances, but the soul (rūḥ, nafs) itself, see Heinrichs (1973: 61–62) and the editor of ʿAmīdī (Ibāna: 152, fn. 1). The point of view of ʿAmīdī as far as this sariqa of Mutanabbī is concerned, is not the same, since he seems rather to aim at the verbal similarities. 84  Muhallabī, poet and ġulām of al-Xalīl b. ʾAḥmad, died about 190/805, Ziriklī (ʾAʿlām: s.v.), ʿAmīdī (Ibāna: 74; 153–154). 83

on lafḏ̣ and maʿnā again

139

Mutanabbī: marartu ʿalā dāri l-ḥabībi fa-ḥamḥamat jawādī wa-hal tašju l-jiyāda l-maʿāhidū ‘I passed by the dwelling of my beloved and my excellent horse neighed, do abodes cause horses to mourn?’

Muḥammad b. ʾAbī ʿUyayna al-Muhallabī:85 wal-mawtu ʾaṭyabu fī famī ʿinda l-hawāni min al-mudāmī ‘Death is sweeter to my mouth than wine, compared with ignominy’.

Mutanabbī: wa-ʿindahā laḏḏa ṭaʿma l-mawti šāribuhū ʾinna l-maniyyata ʿinda ḏ-ḏulli qindīdū ‘In such circumstances he who tastes death fijinds it delicious truly death is as sweet as wine compared with shame’

Ibn Ṭabāṭabā l-ʿAlawī:86 qarmun jawādun yaʿummu l-ʾarḍa nāʾiluhū fa-laysa yafraḥu ʾillā bi-llaḏī yahabū lahū ʾiḏā jiʾtahū fī kulli muškilatin ar-raʾyu wal-jūdu wal-ʾafḍālu wal-ʾadabū ‘A generous lord whose liberality includes the whole extent of the earth and does not rejoice except at what he gives; if you turn to him in every difffijiculty he has judgement, generosity, merits and refijinement’.

Mutanabbī: ʾa-fiji r-raʾyi yušbahu ʾam fiji s-saxā-ʾi ʾam fiji š-šajāʿati ʾam fiji l-ʾadab ʾiḏā ḥāza mālan fa-qad ḥāzahū fatan lā yusarru bi-mā lā yahab ‘Perhaps in judgement can he be compared (to someone) or in his munifijicence, or in bravery or in refijinement? when he obtains wealth, it is got by a hero who does not rejoice at what he does not give’.

 Son of the well-known poet of the 2nd/8th century, see EI2: s.v.; ʿAmīdī (Ibāna: 165).  ʿAmīdī (Ibāna: 174).

85

86

140

lidia bettini 8. Conclusions

We have already examined, in the paragraphs 3 and 5, the kind of awareness the balāġiyyūn had of the relationship between lafḏ̣ and maʿnā in the linguistic meaning of ‘signifijier’ and ‘signifijied’, as far as both the phrase and the word are concerned. From the poetical point of view, we have quoted on purpose a large sample of utterances, mostly in poetry, that are expressly spoken about as having the same maʿnā, in order to have some elements for understanding what the balāġiyyūn mean with this term and with lafḏ̣ : generally speaking, it is evident that these utterances do not share the same linguistic form nor the very same images. As for maʿnā, in what does the common maʿnā consist? Not in the sentence meaning of the verses, even if the way in which a maʿnā is expressed may be closely examined, as we have seen, nor in the whole image that the line develops. So maʿnā may indicate: 1. the idea in its more general and broader terms, that is where tawārud ‘casual meeting (of two poets on the same image)’ occurs (like, as brave as a lion, as handsome as the full moon etc.); 2. a particular given image, which can provide a later poet with a starting point or an inspiration and which can have in the two verses a more or less developed common point (for example the image of the sword as a green plant or of a being, the death or the pen, which walks without feet); 3. the line in both its form and content. As for lafḏ̣ , when authors speak about the borrowing of a lafḏ̣ they mostly quote a given word or some given words; it follows that lafḏ̣ is not employed to indicate the linguistic form of the utterances nor their sentence meanings. Lafḏ̣ , particularly in the context of the sariqa, seems to have a narrower range of uses and it seems to refer to: 1. an isolated word or phrase; 2. formal stylistic devices by which a maʿnā can be expressed (such as the ornaments of the style), which are called ʾalfāḏ̣ . From a more general point of view, as we have seen, the binary correlation between signifijier and signifijied, of which the balāġiyyūn were aware, does not constitute the point of departure nor the focus of their reflections and

on lafḏ̣ and maʿnā again

141

it shows the independence of the balāġa from the other sciences of the language. In the context of the relationship between lafḏ̣ and maʿnā, two aspects would deserve, in my opinion, particular attention: an analysis of the literary borrowings seen as “intertextual dialogue” and an analysis of the reflexions of the balāġiyyūn on the “unspoken”.

References 1. Primary Sources Šarḥ Rummānī. Anonymous (Jurjānī?). Šarḥ risālat ar-Rummānī fī ʾiʿjāz al-Qurʾān, ed. Z. Saʿīd ʿAlī. Al-Qāhira, 1997. Abū ʿAlī. Šiʿr. Kitāb aš-šiʿr ʾaw šarḥ al-ʾabyāt al-muškilat al-ʾiʿrāb, ed. M.M. aṭ-Ṭanāḥī. Al-Qāhira, 1988. ʾĀmidī. Muwāzana. Al-Muwāzana bayna šiʿr ʾAbī Tammām wal-Buḥturī, ed. A. Ṣaqr. Al-Qāhira, 1965. ʿAmīdī. ʾIbāna. Al-ʾIbāna ʿan sariqāt al-Mutanabbī, ed. ʾIbrāhīm ad-Dasūqī al-Bisāṭī. Al-Qāhira, 1961. ʿAskarī. Ṣināʿatayn. Kitāb aṣ-Ṣināʿatayn al-kitāba waš-šiʿr, ed. ʿA.M. al-Bijāwī and M.A. ʾIbrāhīm. Al-Qāhira, 19712. Badīʿī. Ṣubḥ. Ṣubḥ al-munabbī ʿan ḥayṯiyyat al-Mutanabbī, ed. M.Y. ʿArafa. Dimašq, 1350h. Baġdādī. Taqyīd. Taqyīd al-ʿilm, ed. Y. al-ʿIšš. Dimašq, 1949. Bāqillānī. ʾIʿjāz. ʾIʿjāz al-Qurʾān, ed. A. Ṣaqr. Al-Qāhira, 19774. Bayhaqī. Maḥāsin. Kitāb al-Maḥāsin wal-masāwī, ed. F. Schwally, 2 vols. Giessen, 1900– 1901. Ḥātimī. Ḥilya. Ḥilyat al-muḥāḍara fī ṣināʿat aš-šiʿr, ed. J. al-Kitānī, 2 vols. Baġdād, 1979. ——. Mūḍiḥa. Ar-Risāla al-Mūḍiḥa fī ḏikr sariqāt ʾAbī ṭ-Ṭayyib al-Mutanabbī wa-sāqiṭ šiʿrih, ed. M.Y. Najm. Bayrūt, 1965. Ibn ʾAṯīr. Maṯal. Al-Maṯal as-sāʾir fī ʾadab al-kātib waš-šāʿir, ed. A. al-Ḥūfī and B. Ṭabāna, 4 vols. Al-Qāhira, 1959–1963. Ibn Qutayba. Šiʿr. Kitāb aš-Šiʿr waš-šuʿarāʾ, ed. A.M. Šākir. Al-Qāhira, 19672. ——. Tawʾīl. Tawʾīl muškil al-Qurʾān, ed. A. Ṣaqr. Al-Qāhira, 19813. Ibn Rašīq. ʿUmda. Al-ʿUmda fī maḥāsin aš-šiʿr wa-ʾādābihi wa-naqdihi, ed. M.M. Abd al-Ḥamīd. Bayrūt, 1972. ——. Qurāḍa. Qurāḍat aḏ-ḏahab fī naqd ʾašʿār al-ʿarab, ed. aš-Šaḏlī Bū Yaḥyā. Tūnis 1972. Ibn Ṭabāṭabā. ʿIyār.ʿIyār aš-šiʿr, ed. ʿA. b. Nāṣir al-Māniʿ. Ar-Riyāḍ, 1985. Ibn Wakīʿ. Munṣif. Al-Munṣif fī naqd aš-šiʿr wa-bayān sariqāt al-Mutanabbī, ed. M.R. ad-Dāya. Dimašq, 1982. Ibn Zamlakānī. Tibyān. At-Tibyān fī ʿilm al-bayān, ed. A. Maṭlūb and X. al-Ḥadīṯī. Baġdād, 1964. Jāḥiẓ. Bayān. Kitāb al-bayān wat-tabyīn, ed. ʿA. Hārūn, 4 vols. Al-Qāhira, 1975. ——. Ḥayawān. Kitāb al-Ḥayawān, ed. ʿA. Hārūn, 8 vols. Al-Qāhira, 1965–1969. ——. Rasāʾil. Rasāʾil al-Jāḥiẓ, ed. ʿA. Hārūn. 4 vols. Al-Qāhira, 1979. Jurjānī. ʾAsrār. ʾAsrār al-balāġa, ed. H. Ritter. N.p., n.d. ——. Dalāʾil. Dalāʾil al-ʾiʿjāz. N.p., n.d. ——. Muqtaṣid. Kitāb al-Muqtaṣid fī šarḥ al-ʾĪḍāḥ, ed. Kāẓim Baḥr al-Mirjān. Baġdād, 1982. Lisān. Ibn Manẓūr, Lisān al-ʿarab. N.p., n.d. Marzubāni. Muwaššaḥ. Al-Muwaššaḥ fī maʾāxiḏ al-ʿulamāʾ ʿalā š-šuʿarāʾ, ed. M. Ibn al-Xaṭīb. Al-Qāhira. 19653.

142

lidia bettini

Mubarrad. Balāġa. Al-Balāġa, ed. R. ʿAbd at-Tawwāb. Al-Qāhira, 1965. Qāḍī Jurjānī. Wasāṭa. Al-Wasāṭa bayna l-Mutababbī wa-xuṣūmihi. Al-Qāhira, 1370/1951. Qudāma. Naqd. Kitāb Naqd aš-šiʿr, ed. S.A. Bonebakker. Leiden, 1956. Rāzī. Nihāya. Nihāyat al-ʾījāz fī dirāyat al-ʾiʿjāz, ed. B.Š. ʾAmīn. Bayrūt, 1985. Rummānī. Nukat. “An-Nukat fī ʾiʿjāz al-Qurʿān”, in Ṯalāṯ rasāʾil fī ʾiʿjāz al-Qurʿān, ed. M. Xalaf Allāh and Zaġlūl Salām. Al-Qāhira, 1968, 73–158. Sakkākī. Miftāḥ. Miftāḥ al-ʿulūm, ed. N. Zarzūr. Bayrūt, 19872. Ṣūlī. ʾAxbār. ʾAxbār ʾAbī Tammām, ed. Xalīl Maḥmūd ʿAsākir et al. Bayrūt, 19803. ——. ʾAwrāq. Kitāb al-ʾAwrāq (qism ʾaxbār aš-šuʿarāʾ), ed. J.H. Dunne. London, 1934. Ṯaʿālibī. ʾĪjāz. “Al-ʾījāz wal-ʾiʿjāz”, in Xams rasāʾil. An-Najaf, 1301h. Xafājī. Sirr. Sirr al-faṣāḥa, ed. ʿA. aṣ-Ṣaʿīdī. Al-Qāhira, 1969. Yāqūt. ʾUdabāʾ. Muʿjam al-ʾUdabāʾ (ʾIršād al-ʾarīb ʾilā maʿrifat al-ʾadīb), ed. I. ʿAbbās. Bayrūt, 1993. 2. Secondary Literature Abu Deeb, Kamal. 1979. Al-Jurjānī’s Theory of Poetic Imagery. Warminster: Aris & Phillips. Ambros, Edith. 1979. Sieben Kapitel des Šarḥ Kitāb Sībawaihi von ar-Rummānī in Edition und Übersetzung, Wien: Verl. des Verb. der wiss. Ges. Österreichs. Arazi, Albert. 2004. “Périodisation, oralité et authenticité de la poésie arabe préislamique”. Jerusalem Studies in Arabic and Islam 29, 377–412. ʾAsad, Nāsir ad-dīn al-. 1962. Maṣādir aš-šiʿr al-jāhilī. Al-Qāhira. Bettini, Lidia. 1987–88, “Langue et rhétorique au Vème siècle”. Quaderni di Studi Arabi 5–6, 91–104. Bonebakker, Seeger Adrianus. 1956. “Introduction” to Qudāma, Šiʿr. ——. 1997. “Ancient Arabic poetry and plagiarism: a terminological labyrinth”. Quaderni di Studi Arabi 15, 65–92. Carter, Michael G. 1980. “Sibawayh and the modern linguistics”. Histoire Epistémologie Langage 2, 21–26. ——. 1984. “Linguistic science and orthodoxy in conflict—the case of al-Rummānī”. Zeitschrift für Geschischte der Arabisch-Islamischen Wissenschaften 1, 212–232. Cheikh-Moussa, Abdallah. 2006. “Considérations sur la littérature d’adab. Présence et efffets de la voix et autres problèmes connexes.” Al-Qantara 27, 25–62. Dumayrī, al-Mutawallī Ramaḍān Aḥmad ad-. 1988. Šarḥ Kitāb Sībawayhi li-ʾAbī l-Ḥasan ar-Rummānī (296–384), I ad-dirāsa, II at-taḥqīq. Al-Qāhira. Frank, Richard M. 1978. Beings and their Attributes. Albany. ——. 1981. “Meanings are spoken of in many ways: the early Arab grammarians”. Le Muséon 94, 259–319. Grunebaum, Gustav Edmund von. 1941. “Al-Mubarradʼs Epistle on Poetry and Prose”. Orientalia N.S. 10, 372–382. ——. 1952. “The aesthetic foundation of Arabic literature”. Comparative Literature 4, 323–340. Günther, Sebastian. 2006. “Praise to the book! Al-Jāḥiẓ and Ibn Qutayba on the excellence of the written word in medieval Islam”. Jerusalem Studies in Arabic and Islam 32, 125–143. Haddāra, Muḥammad Muṣṭafā. 1981. Muškilat as-sariqāt fī n-naqd al-ʿarabī, 3rd ed. Bayrūt and Dimašq. Heinrichs, Wolfhart. 1969. Arabische Dichtung und griechische Poetik. Beirut: OrientInstitut der deutschen Morgenländischen Gesellschaft. ——. 1973. “Literary theory: the problem of its efffijiciency”. Arabic poetry—Theory and development, ed. G.E. von Grunebaum. Wiesbaden: Harrassowitz, 19–69. ——. 1977. The hand of the Northwind—Opinions on metaphor and early meaning of istiʿāra in Arabic Poetics. Wiesbaden: Deutsche morgenländische Gesellschaft.

on lafḏ̣ and maʿnā again

143

——. 1984. “Istiʿārah and badīʿ and their teminological relationship in early Arabic literary criticism”. Zeitschrift für Geschichte der Arabisch-Islamischen Wissenschaften 1, 180–211. ——. 2007. “Sariḳa”. Encyclopédie de l’Islam, 2ème éd., suppléments, s.v. Hirschfeld, Hartwig. 1922. “A volume of essays by al-Jāḥiẓ”. A volume of Oriental studies presented to Edward G. Browne on his 60th birthday (7 february 1922), ed. T.W. Arnold, and Reynold A. Nicholson. Cambridge: Cambridge University Press, 200–209. Kilito, Abdelfattah. 1979. “Sur le métalangage métaphorique des poéticiens arabes”. Poétique 38, 162–174. Larcher, Pierre. 1993. “Un grammairien ‘retrouvé’: ʿAbd al-Qāhir al-Jurjānī”. Arabica 40, 248–253. Mubārak, Māzin al-. 1995. Ar-Rummānī an-naḥwī fī ḍawʾ šarḥih li-Kitāb Sībawayhi, 2nd ed. Dimašq: Dār al-Fikr. [1963. 1st ed.] Peters, Johannes Reinier Theodorus Maria. 1976. God’s created speech. Leiden: Brill. Sadan, Joseph. 1991. “Maidens’ hair and starry skies”. Israel Oriental Studies 11, 57–88. Ṣammūd, Ḥammādī. 1994. At-Tafkīr al-balāġī ʿinda l-ʿarab, 2nd ed. Tūnis. Sanni, Amidu Olalekan. 2003. “At-Tawḥīdīʿs “Epistle on the sciences”: An epistemological study of the section on the balāgha (rhetoric)”. Journal of Oriental and African Studies 12, 39–51. Schoeler, Gregor. 2002. Écrire et transmettre dans les débuts de l’islam. Paris: PUF. Simon, Udo Gerald. 1993. Mittelalterliche arabische Sprachbetrachtung zwischen Grammatik un Rhetorik—ʿilm al-maʿānī bei as-Sakkākī. Heidelberg: Heidelberger Orientverlag. Ziriklī, Xayr ad-dīn az-. ʾAʿlām. Al-ʾAʿlām, 8 vols. Bayrūt, 1989.

SECTION III

THE ARABIC WORD IN CONTEMPORARY LINGUISTIC THEORY

LEVELS OF ANALYSIS OF THE WORD IN ARABIC1 Georges Bohas

A. Levels of Representation of a Word in the Theory of Matrices and Etymons When addressing the problems posed by the analysis of the word in Arabic, one typically thinks of the root and the pattern, grammatical concepts developed by the Arab grammatical tradition and adopted by Arabists and linguists of all sorts (comparativists, structuralists, generativists, etc.). Most of these linguists remain unaware that their conception of Arabic grammar is completely diffferent from that of their Arab counterparts, a topic I broach in Bohas & Guillaume (1984). In order to better understand how the Arab root functions, let us consider Paradigm 1:

batara batira ʾabtara inbatara bātirun ʾabtaru ʾabātirun

‘to cut the tail of an animal’ ‘to cut, to remove by cutting’ ‘to have a tail cut offf’ ‘to cut the tail of an animal / to keep someone from having children (said of God)’ ‘to be cut’ ‘which cuts, sharp (sabre)’ ‘cropped, an animal with its tail cut offf’ ‘someone who does not have children / who abandons and gives up his family’ Paradigm 1

Phonetically, each of these words contains three identical consonants ordered linearly in the same manner -btr-, highlighted in bold face in the paradigm. Semantically, we can see that all of these words have something to do with the idea of “cutting”, an action that can be specifijic, as in ‘to cut

1  This article builds on Bohas (2006) Bohas and Dat (2007), Bohas and Saguer (2006), Bohas and Saguer (2007), Bohas (forthcoming), Bohas and Saguer (2007a). Translated from the French by Alison Terry, in collaboration with Nigel Briggs, CEREL, ENS LSH, Lyon, revised by Dennis Philps, CAS-IRPALL, UTM.

148

georges bohas

the tail’, and which may be subjected to various metaphorical interpretations: When God deprives someone of children, it is as though he has cut offf their sexual organ; when someone abandons their family, they are essentially cut offf from them, just as in French: “il s’est coupé des siens, de sa famille, etc.” Having identifijied both the phonetic and semantic aspects, we obtain the root √btr. To form a word, this root combines with a grammatical pattern, which we transcribe using the symbols of the Arabic grammarians: f ʿ l, where f=C1, ʿ =C2, and l=C3. Hence faʿala represents the pattern of the perfect. Incorporating √btr into this model yields batara. To obtain this composite, we replace the consonants f ʿl by those of the root.2 fa ʿa la = batara

b

t

r

One can therefore postulate that the sense of ‘he cut’ (perfect, 3rd person masculine) emerges from the combination of the root √btr ‘to cut’ with the faʿala pattern. The logic behind the extraction of the root in paradigm 1 consists in identifying the phonetic and semantic constants: √btr / “to cut”. We shall apply the same logic to Paradigm 2:

batta batara inbataʿa bataka batala balata barata sabata

‘to cut, to remove by cutting’ ‘to cut the tail of an animal’ ‘to cut, to remove by cutting, to remove’ ‘to be separated, isolated, removed from a whole or from other parts’ ‘to cut, to remove’ ‘to separate a part from its whole’ ‘to cut, to remove’ ‘to separate a part from its whole’ ‘to cut, to remove, to separate, to divide by cutting’ ‘to cut’ ‘to cut, to remove by cutting’ ‘to shave (a head)’ Paradigm 2

2  For a detailed explanation of the idea of the root and its signifijicance to Arabic grammarians, see Bohas (1979).

levels of analysis of the word in arabic

149

Just as it proved possible to extract the common base √btr from paradigm 1, so it is possible to extract a common base from this example: all of these verbs include a b and a t, and they all express the action of “cutting”. We would like to insist on the fact that we followed exactly the same logic as when determining the root in paradigm 1, and that this logic—identifying the common phonetic and semantic properties as well as the conclusion we draw—has not been called into question. We shall call this base bt an etymon, symbolized as ∈bt. In realizing the etymon ∈bt in the verbs in Paradigm 2, we see that this binary composite was developed by reduplication of the last consonant, as in batta, by adding a fijinal consonant: batara, a medial consonant: balata, or by adding an initial one: sabata, without afffecting the fundamental common meaning, to use the terminology of Brockelmann (1910). These diffferent processes have been the subject of many publications and theses which may be consulted by the reader. The etymon reveals that all these words have something in common, both on a phonetic level: the phonemes bt and on a semantic level: the sense of “cutting”, a fact that traditional dictionaries organized into triconsonantal roots cannot highlight. Let us now consider Paradigm 3:

batta batara batala barata balata badaḥa bad̠aḥa bazzun bazala baḍaʿa baṭṭa baṭara tabba ḥad̠afa d̠ubāb šabara ʿaḍiba hadaba faʾasa faʾsun fatta

‘to cut, to remove by cutting’ ‘to cut the tail of an animal, to dock it’ ‘to cut, to remove’ ‘to cut’ ‘to cut, to remove, to separate, to divide by cutting’ ‘to splinter, to tear’ ‘to cleave (the tongue of a camel)’ ‘a sword’ ‘to splinter’ ‘to splinter, to cut, to remove’ ‘to open an ulcer’ ‘to splinter, to lance (an ulcer)’ ‘to cut, to remove by cutting’ ‘to remove’ ‘the sharp edge or point of a sword’ ‘to tear or cut lengthwise (a material)’ ‘to have a cleft ear’ ‘to cut, to fell something with a sharp instrument’ ‘to strike someone with an axe’ ‘an axe’ ‘to split (stones)’

150

georges bohas

farat̠a farasa faraṣa faraḍa fasaʾa faṭara faṭama ṣafaḥa šafratun ṣafīḥatun sāfa sayfun

‘to pierce, to puncture, to empty’ ‘to tear into (prey)’ ‘to cut, to split in half’ ‘to make notches or slashes in a piece of wood’ ‘to tear, to lacerate’ ‘to split, to set about splitting something in two’ ‘to cut by making an incision’ ‘to hit someone with the flat part of a sabre or another instrument’ ‘knife, cutlass, a cobbler’s tool’ ‘a sabre with a big blade’ ‘to strike with a sabre’ ‘sabre’ Paradigm 3

On the semantic level, all these words have something in common: they all involve the notional invariant “to cut”. However, they cannot be grouped around a common root or a common etymon. At this stage, we must move on to another level of analysis, that of the distinctive features3 characterizing phonemes. To do this, let us return to Paradigm 3 and consider the phonemes in bold face. On the one hand, we have: b, f

and on the other: t, d, ḏ, ṯ, s, z, š, ṣ, ḍ, ṭ.

B, f form a group characterized by the feature [labial],4 which refers to sounds produced by a constriction of the lips. T, d, ḏ, ṯ, s, z, š, ṣ, ḍ, ṭ form a class characterized by the feature [coronal], which refers to sounds produced with a constriction formed by the front of the tongue that is located between the upper incisors and the hard palate (dental, alveolar, palato-alveolar). We have now demonstrated the phonetic and semantic properties of the terms in paradigm 3, which is called the matrix: Phonetic: [labial] × [coronal] Notional invariant: ‘to deliver a blow’ characterization: ‘with a sharp object’

3

 See the table of features in Appendix 1.  See the table of features in Appendix 1.

4

levels of analysis of the word in arabic

151

We consider matrices to be primitive in terms of lexical organization, as they group together large numbers of terms on the basis of their common properties, both semantic and phonetic. Clearly, a simple root cannot accomplish this task. We have just demonstrated three levels of analysis for Arabic words: the radical, the etymon, and the matrix. One might ask: Is the entire Arabic lexicon organized in this manner? The answer is “Yes”, and we will demonstrate this by exposing yet another case, leaving the reader free to carry our investigation further. We now tackle another example:

matta matā mataʾa mataʿa matana

‘to extend something lengthwise (a rope)’ ‘to extend lengthwise (a rope)’ ‘to increase tension, to extend a rope lengthwise’ ‘to lengthen, to extend lengthwise’ ‘to increase tension, to extend and lengthen something’ Paradigm 2b

Once again, it is impossible not to see that each verb contains the sequence mt and that all imply the idea of “stretching” to some degree. If the lexicon is organized by setting up the triconsonantal root as a primitive base, these observations appear to be purely coincidental, and it proves impossible to draw meaningful conclusions because all the verbs are linked to diffferent roots: √mty, √mtʾ, √mtʿ, and √mtn. In an attempt to resolve these discrepancies, we have proposed (Bohas 1997; 2000; and here) the notion of the “etymon”, a binary composite of non linearlyordered5 phonemes—in this case mt—of which the terms in Paradigm 2b are products. It is clear that in order to extract an etymon, we proceed exactly as if we were extracting a root, fijirst fijinding evidence of the common invariant phonetic and semantic properties, and then arriving at a conclusion. Let’s develop this paradigm further:

5

 See Bohas & Darfouf (1993).

152

georges bohas

matta matā mataʾa mataʿa matana madda maṭṭa maṭala maṭā šanaʾa masara masana masā mašaqa maṣaxa nataxa natara nataša [nt] × [nš]6 natafa nāša naḍḍa

F.VIII F.V

‘to extend something lengthwise (a rope)’ ‘to extend lengthwise (a rope)’ ‘to increase tension, to extend a rope lengthwise’ ‘to lengthen, to extend lengthwise’ ‘to increase tension, to extend and lengthen something’ ‘to stretch out, like a rug’ ‘to increase tension and lengthen something by pulling on it with force’ ‘to stretch out a cord’ ‘to lay out a path for someone’ ‘to pull, to extract something’ ‘to pull, to extract something from the place where it is usually found’ ‘to pull, to extract something from something else’ ‘to pull, to extract something by pulling it toward you’ ‘to pull something in order to extend it’ ‘to pull, to extract something from something else’ ‘to pull, to extract, to pull out (a tooth, a hair)’ ‘to pull the string of a bow toward oneself with force’ ‘to pull, to extract something, to pull out (a thorn from the foot, a hair)’ ‘to pull out, to pull (hair, feathers, wool fijibres)’ ‘to pull, to extract’ ‘to pull something toward you’ Paradigm 3b

Continuing with the previous analysis, we fijind that in these words, either m or n, i.e. an element of the [nasal] class, combine with t, d, ḏ, ṯ, s, z, š, ṣ, ḍ, ṭ, i.e. an element of the [coronal] class. Clearly, these words all encompass some form of the idea of “traction”. The phonetic invariant is therefore the combination [nasal] × [coronal] and the notional invariant “traction”, a more detailed study of which is provided in Saguer (2003). In order to explain the common phonetic characteristics of these forms, we must reason in terms of phonetic features, that is, at matrix level. From now on, we will refer to the “notional invariant” or the “generic notion” to designate the common semantic content of all the realizations of the matrix. It is obvious that if the triconsonantal root is taken as a basis for analysis, then these phonosemantic generalizations, no matter how self-evident, can never be expressed. We must therefore 6

 Indicates an analysis of this etymon in terms of hybridization, see Bohas (2000).

levels of analysis of the word in arabic

153

return to the evidence: the concept of the root can never be the primary organizing element of the Arabic lexicon because it is not situated at the pertinent phonetic level. The initial constituents of the lexicon must be composed of matrices of features, and not phonemes, if one intends to highlight an organizational structure that takes into account the various phonosemantic relationships that appear in the words. Working from the examples we have examined up to now, we propose a three-tiered structure for organizing the Arabic lexicon: The matrix: a non linearly-ordered combination of a pair of phonetic feature vectors (or binary combination) linked to a “generic notion”, a “notional invariant”, is postulated as being the lexicogenic source of large notional fijields. The matrix is a virtual structure, the level at which referents are transformed into simple phonetic features. The etymon: a non linearly-ordered bi-consonantal base made up of two phonemes taken from a given matrix that exhibits not only the features of this matrix, but also its notional invariant. The radical (nominal or verbal): a semantically autonomous item composed of an etymon that has been realised by difffusion of the last consonant or by incrementation, or that results from the fusion of two etymons. Its semantic content is similar to the semantic value of the etymon, which the radical can specify semantically, thus contributing to the creation of extensive fijields of semantic associations, and whose hyperonym is the semantic value assigned to the matrix. The radical is the level at which diverse morphological and apophonic processes take place. This organization is neither synchronic nor diachronic, but rather achronic: it was valid 1,500 years ago, and is still valid today. The data upon which I base my analyses come from Kazimirski (1860), and have been cross-checked in Lisān and/or Qāmūs.

B. From the Etymon to the Radical Here I will fijinish my introduction to the tme by showing how the etymon develops to form a radical. 1. Development by diffusion of the last consonant over two slots in the pattern furnished by the morphology in: batta

‘to cut, to remove by cutting’

154

georges bohas

We can make the same observation regarding Paradigm 2b: matta

‘to extend something lengthwise (a rope)’

2. Development by incrementation of a sonorant (r, l, m, n) or a guttural (ʿ, ḥ, h,ʾ) in:7 batara inbataʿa batala balata bartun/burtun

‘to cut the tail of an animal’ ‘to cut, to remove by cutting’ ‘to be separated, isolated, removed from the whole or from other parts’ ‘to cut, to remove’ ‘to separate a part from its whole’ ‘to cut, to remove, to separate, to divide by cutting’ ‘an axe’

This incrementation can be carried out at the beginning, at the end, or between the two elements of the etymon. A number of examples can be found in Bohas (1997: 36 fff.). 3. Development by prefixation Hurwitz (1913: 55–60) identifijied the status of “prefijix” as far back as the early twentieth century: The preformatives are thus seen to possess a fairly defijinite, though remote, relationship to each other. The sibilants and gutturals are to be traced to causative stems; the dental t and liquid n to reflexive stems; the liquids m, l, and r are to be connected etymologically with the reflexive n, and preformative y may be considered to be a denominative stem.

Consider the verb nakafa ‘to distance from, to move away from, to shy away from something’. It has a reflexive sense. M. Cohen (1929=1955: 240) defijines ‘reflexive’ as being a process in which the agent and the patient are identical. Let’s look at an example of two opposing phrases: Pierre washes his car and Pierre washes himself (Pottier 1974: 117).

7  The most die-hard triliteralists have noted that “certain series indicate that two fundamental consonants can receive, in order to make up synonymous roots or quasi-synonyms, the addition of liquids or even of other consonants . . .” (Cohen 1947: 58). This did not lead them to question the foundations of their approach: “One could wonder: Would we have found this biliteral level of analysis through Chamito-Semitic comparison, had we gone back in time. This is defijinetly not the case.” And more explicitly: “From the outset, the decision was taken to undertake the comparison of whole roots, without dissecting them” (Cohen 1947: 59).

levels of analysis of the word in arabic √nkf

nakafa

155

‘to distance from, to move away from, to shy away from something’.

This form is semantically linked to the following: kafffa

[kf ] F.VII kafaʾa [kf ]ʾ kafaḥa [kf ]ḥ

‘to send away, to push someone away’ ‘to withdraw’ ‘to be repulsed and put to flight’ ‘to chase away, send away’

We can conclude that nakafa is a form that incorporates the etymon [kf ] with the addition of an initial n that functions as a prefijix, giving a reflexive sense to the word. Note that the reflexive sense expressed in the initial form is identical to the meaning expressed by F.VII derived from kafffa: inkafffa, which shows that reflexiveness can be expressed either lexically or morphologically. Consider the relationship of saxala ‘to remove, to take away’ with the following: xalla xalaʿa laxaṣa salaxa naxala xalaba xalasa

F.II

‘to remove food particles from the teeth with a toothpick’ ‘to pull out, to remove gently from something’ ‘to remove, to pull, to extract the best or purest part of something’ ‘to skin, to remove the skin of a sheep, etc.’ ‘to pull, to separate the bad parts and remove the good parts of something’ ‘to remove, to capture in claws (said about prey)’ ‘to take away, to remove, to ravish unknowingly with the blink of an eye’

This paradigm allows the extraction of the etymon {l,x} and isolates the initial consonants n and s, so that an analysis of saxala would be [s[xl]]. 4. Development by blending Let us study the case posed by: bataka ‘to cut, to remove’ ‘to separate a part from its whole’

We notice that there are in fact two etymons that can be seen in the examples below: batta takka

‘to cut, to remove by cutting’ ‘to cut, to remove’

156

georges bohas

bataka can therefore be analyzed as the combination of these two etymons: bt × tk = bataka

The two identical elements (t) are reduced to a single element by applying the Obligatory Contour Principle (ocp), which prohibits identical adjacent elements at this level.8 There exists three processes of blending, which we will expand on at a later stage. 5. Development by incrementation of a glide This phenomenon is manifest in the following cases: ∈{b,d} badda wabida

F.IV

∈{b,x} baxxa bāxa baxā

‘to return to a state of calm after being angry’ /bawaxa/ ‘to calm oneself, to extinguish (fijire, warmth, anger)’ /baxawa/ ‘to return to a state of calm after being angry’

‘to separate, to move away’ ‘to separate, to isolate’

6. Development by incrementation of a final consonant Here are two examples: ∈{x,š} xašša xašafa

‘to enter’ ‘to enter inside’

∈{d,ʿ} da‘ʿa ‘to push violently’ da‘aba ‘to send away, to push away’

An issue that has already been addressed by Lambert (1897) is whether or not this consonant is derived from a sonorant (here, m) or if it is inserted just as it is.

8  For the defijinition and more examples of the application of this principle to Semitic languages in a diffferent framework to our own, see McCarthy (1986) and fn. 18.

levels of analysis of the word in arabic

157

7. Development by reduplication This type of development yields a quadriconsonantal radical. Given the large number of examples of this sort,9 we will limit ourselves to the following: baxxa baxbaxa

‘to return to a state of calm after being angry’ ‘to soften oneself, to lose one’s intensity (warmth)’

bazza bazbaza

‘to remove, to take away’ ‘to remove, to take away’

bašša ‘to give someone a gracious welcome’ bašbaša -F.II ‘to give someone a warm welcome’ t̠abba t̠abṯaba

‘to sit down and establish a solid presence’ ‘to sit down and establish a solid presence’

We have established, then, that etymons are developed in the following ways: • by difffusion; • by incrementation of a sonorant or guttural; • by prefijixation; • by blending; • by incrementation of a glide or transformation of a vowel into a glide; • by incrementation of the fijinal consonant; • by reduplication. As stated by Renan (1855), these processes “only nuance the principal meanings”; however, it also happens that they do not change the original meaning at all.

C. Explanation of Homonymy within the Framework of the tme10 At this point, one may be thinking that although this looks fijine, how useful is it all? Clearly, when one groups words into matrices and etymons, one is not doing it simply for pleasure, but also to explain the facts. Indeed, the

  9

 Bohas (1997: ch. 3).  All of the analyses contained in this section are developed in Bohas & Saguer (2006; 2007a). 10

158

georges bohas

task of the linguist is to pinpoint facts concerning language or the use of language, to formulate generalizations about their subject and to try and explain these within the framework of a theory. This is what linguistics scholars do every day; in other words, it is not enough to simply make a list of lexical items: one must also explain the synonymy, polysemy, homonymy, and antithetical polysemy (or enantiosemy)11 that appear in the items of the lexicon. In this paper, we shall deal with the phenomenon of homonymy, but fijirst, we must defijine some key terms. Polysemy can be explained as “a word which brings together several meanings between which users can recognize a link” (Nyckees 1998: 194). “Homonymy is distinct from polysemy in that, in the case of homonymy, it seems impossible to re-establish a plausible semantic relationship” (Nyckees 1998: 194) between the diffferent meanings. The following examples are taken from French: louer ‘to give praises’ and louer ‘to rent’; avocat: ‘an avocado’ and avocat: ‘a barrister’. In English, one fijinds: flies ‘winged insects’ and flies ‘the opening at the front of a pair of trousers’, or sound (v.) ‘to make a noise’ and sound (v.) ‘to measure the depth of water’—diffferent, unrelated meanings. We explain cases of homonymy either by phonetic evolution or by direct borrowing. So, for the word louer, the two terms locāre and laudāre would have been combined into a simple form through phonetic evolution. laudāre S112 ‘to give praises’

locāre S2 ‘to rent’ louer S1 ‘to give praises’ S2 ‘to rent’

To explain the evolution of a word in terms of borrowing, consider avocat. It is clear that the meaning of ‘barrister’ is the ‘original’ meaning and that avocat, the fruit, is borrowed from a Caribbean language via the Spanish word avocado, with a phonetic adaptation.

11

 We use the term antithetical polysemy or enantiosemy for words that mean something and its opposite, such as “to rent”, which means both “to take temporary possession for the payment of a fee” and “to give temporary possession for the receipt of a fee”. 12  S = sense.

levels of analysis of the word in arabic avocat S1 ‘barrister’

159

avocado S2 ‘fruit’ avocat S1 ‘barrister’ S2 ‘fruit’

The Arabic lexicon contains many examples of homonymy: naʾaja

S1: S2: S3: S4: S5:

‘to go away and disappear far inland’ ‘to moan’ ‘to let out a doleful cry (owl)’ ‘to bellow (bull)’ ‘to cry out in distress’

Senses 2–5 seem to be related polysemically: they are similar, because in all cases a mufffled noise is emitted; and yet they are also diffferent because of the diffferentiation between mooing and moaning, or a moan and a cry. However, S1 is homonymically related to the four other meanings because there is no possible link between ‘to go away and disappear far inland’ and ‘producing a mufffled noise’ The verb labaxa encompasses eight meanings: S1: S2: S3: S4: S5: S6: S7: S8:

F.III F.V

‘to be plump (in reference to the body)’ ‘to beat, to hit someone’ ‘to kill someone’ ‘to take something from someone, from someone’s hand’ ‘to extract information from someone by tricking them’ ‘to slap someone in the face’ ‘to perfume oneself with musk’ ‘to insult someone’

S2, S3, and S6 all seem to share a semantic relationship: to do something harmful to someone else. However, these three defijinitions also share a homonymic relationship with all of the others. According to the traditional linguistic framework, in which the root is the most basic unit of analysis, what can we say about homonymy? We can easily pinpoint cases of borrowing, as with the word avocat. In this way, with the root √brd, we may observe the appearance of meanings that have to do with ‘freshness’, but also, with the word barīd, the meaning of ‘mail, post’. Since there is no plausible semantic relationship between the two, we can formulate the hypothesis that barīd ‘mail, post’ has a foreign origin. Obviously, once we have recognized that the word

160

georges bohas

is of foreign origin, we must search for this source within a neighbouring language. In this case, we consulted a good Latin dictionary to fijind ueredus, and advanced the hypothesis that the word was borrowed by the Arabic language from the original Latin, which was actually linked to Gaulish (according to Ernout & Meillet 1932). But the problem is not that simple, in that Lane (1863: s.v.) has put forward another possible explanation: barīd could also come from the Persian baridah dum ‘which has the end cut offf, a distinctive sign of mail-mules’.13 Therefore, assigning an origin to a word identifijied as having a non-Arabic source is not always straightforward. In any case, if one focuses primarily on the roots and patterns, barīd is an ordinary faʿīl, and the fact that it has a foreign origin does not influence its usage. Studies of the foreign origins of Arabic words are not without interest, but they do not provide any new insights into the way the lexicon functions. At best, they can show us how phonetic adjustments occur, e.g. how ueredus took on the form barīd, in the same way that avocado became avocat. Now let’s look at the verb ġaraza, which has two strictly homonymic meanings: S1:

‘to prick something with a needle; to drive in, to penetrate (with a pointed instrument); to plunge its tail into the earth to lay eggs (said of locusts)’ S2: ‘to give only a small amount of milk (she-camel)’

The same semantic meaning is found with ġārizun: S1:

‘someone who drives in, who plunges a pointed instrument, a needle, into something; who plunges its tail in the earth to lay eggs (locusts)’ S2: ‘which only produces a small amount of milk (she-camel)’

Since we can not establish a plausible semantic link between these two meanings, we are obviously dealing with a typical case of homonymy. However, we may want to look at the following two words: ġārra14 ġirārun muġārrun

13

‘found in a small quantity (said of the milk of a female)’ ‘a small quantity in general, such as the small amount of milk produced by a female’ ‘a female with a small amount of milk in its udders (she-camel)’

 Kazimirski (1860: s.v.) also considers this to be Persian in origin.  The segments that make up the etymon are indicated in bold.

14

levels of analysis of the word in arabic

161

Phonetically, the analysis in terms of the etymon of these words can only point to {ġ,r} because their radical does not contain another consonant,15 and it is obvious that they are referring to the meaning S2. The verb razza signifijies ‘to plunge its tail into the ground in order to lay (of locusts); to stick, plunge, and solidly fijix one object in another of in the ground’. It thus clearly displays sense S1, and can be analysed at etymon level as {r,z}.16 The explanation is thus that ġaraza includes S1 and S2 because it results from the blending of the two etymons {ġ,r} and {r,z}. This blending occurs as sketched out in model A:17 CC

CC

j i

i k

Si

Sj

C C C 18 j

i

k

Si + SJ

Or, more clearly: {ġ,r} ‘lack of milk’

{r,z} ×19

Si

‘to push a sharp object into’ Sj

ġaraz Si ‘lack of milk’  +  Sj ‘to push a sharp object into’

From now on, we shall use the term “blending” for such cases.

15

 This is what we call non-ambiguous radicals.  Non-ambiguous radical. 17  See Bohas (1997: 175 fff.; 2000: 49). 18  The Obligatory Contour Principle explains the fusion of the two Ci’s into a single segment. For the defijinition and more examples of the application of this principle that prohibits identical adjacent elements on the same level to Semitic languages, see McCarthy (1986). Since then, the ocp has given rise to a large number of studies that it would be superfluous to list here. 19  We use × to indicate blending. 16

162

georges bohas

If one were to unearth only one case of this type, one could perhaps attribute it to chance. But if one should fijind two or three hundred examples (which is actually the case, cf. Laïla Khatef ’s 2003 thesis), one would be in a position to say that a step forward had been made in the explanation of homonymy in the Arabic lexicon. Now let’s look at a more elaborate solution to this problem within the τμε framework. There are ten matrices that we have identifijied and organized up to this point:

Matrix 1 Notional invariant: example Matrix 2 Notional invariant: example

Matrix 3 Notional invariant: example Matrix 4 Notional invariant: example

Matrix 5 Notional invariant: example Matrix 6 Notional invariant: example

20

 See Diab-Duranton (2005).

{[labial], [coronal]} “to strike a blow” ḍaraba: ‘to hit, to beat’ batta ‘to cut’ + paradigm 2 and 3 {[labial], [+continuant]} “movement of air” habba ‘to breathe’ fāxa ‘to pervade (odour), to whistle (wind); to break wind (man)’ {[labial], [pharyngeal]} “(a) tightening” ḥabaka ‘to link, to tighten’ ʿaffa ‘to abstain oneself from something’ {[coronal], [pharyngeal] [~dorsal] [-voiced]} “a mufffled voice, a muted sound, hoarse” ʾanna ‘to moan’ ḥanna ‘to moan, to let out a sound of afffection (she-camel showing love for its calf)’ {[coronal], [dorsal]} “to strike a blow”20 qatta ‘to cut’ jazza ‘to cut fur, grains, dates’ {[labial], [dorsal]} “curvature” kuʿb: mammal qabā: to curl, to bend jubbun: wells

levels of analysis of the word in arabic Matrix 7 Notional invariant: example Matrix 8 Notional invariant:

163

{[dorsal], [pharyngeal]} “animal cries” zaqaʿa, saqaʿa, ṣaqaʿa ‘to sing (said of a rooster)’ {[+approximant] [+continuant]} [+lateral] “the tongue”

There follows an explanation of the ramifijications of the notional invariant or matrices 8 and 9, insofar as they are often cited: 21 1. The tongue and its characteristics lisān ‘tongue’ 1.1. The tongue and the physical movements it makes 1.2. To perform an operation on the tongue lasana ‘to take hold of someone’s tongue with your hand, to take the tongue of a person you are kissing in your mouth (speaking of romantic gestures between a man and a woman)’ 1.3. To grab, to pull with your tongue, to pull with your tongue in order to eat or drink >22 to taste walaġa and waliġa ‘to lap up, to drink by putting the tongue into a bowl or in a liquid and then removing the tongue back into the mouth; (to drink as dogs do); to taste’ 1.4. To lick lassa ‘to lick (a bowl, a pot, etc.)’ lasiba ‘to lick (honey)’ laḥafa ‘to lick something’ 1.4.1. Consequence (1): to moisten and glue, to stick together lasiqa ‘to stick together, to attach to . . . ’ lazza ‘to stick together and join’ lazaba ‘to be strongly attached, to glue to something else’ 1.4.2. Consequence(2): to be slick or smooth, polished23 ʾamlasu ‘joined, polished, slick/smooth, shaved, shorthaired’ saḥala F.VII ‘to be smooth, polished, flat’ dalaṣa ‘to be polished and shiny (said about a metal plaque, of armour)’ 1.5. To sample, to taste laṣama ‘to taste, to sample something’

21

 See Bohas (2006).  >: this symbol signifijies the existence of a semantic relationship. 23  This development is also seen in Berber: alessas ‘soft to the touch’, see Biarnay (1908). 22

164

georges bohas

2. The tongue as an instrument of language lasana ‘to use harsh words; to bad-mouth someone, to tear someone to shreds to have the upper hand over someone with words, to outwit them with words’ lasaʿa ‘to hurt someone with mean words, with sarcasm’ ‘to bad-mouth someone’ saliṭa ‘to be absolute and unyielding in leadership style’ sulṭān ‘power; prince, sultan’ lad̠aʿa ‘to hurt someone with one’s tongue; by a cutting remark’ laġā ‘to speak, in general’ ‘to say stupid things, to give meaningless, frivolous, inconsiderate discourse’ ‘to be mistaken when speaking, to make an error’ 3. The tip of the tongue: pointed, to be sharp, to form a point and, by extension, to sting, to be sharp with one’s remarks lasab ‘to prick someone (said about a snake)’ d̠alaqa ‘to hone, to sharpen (a knife, etc.)’ ‘to be well sharpened (said fijiguratively about the tongue of a man who expresses himself with ease, or who can strike back in response to insults)’ 4. Tongue of fijire: to blaze, to flame, to burn lasana F.V  ‘to burst into flames, to flame (said about fijires)’ lad̠aʿa ‘to burn something (said about fijire, about a fijire that burns a body)’ fijig. ‘to burn (said about love that consumes the heart)’

Matrix 9 Notional invariant:

{[+ nasal] [+continuant]} “the nose”

The phonetic material of this matrix is constituted on the one hand by the two nasals –m and n– and on the other by various fricatives. 1. The nose 1.1. The organ itself ʾanfun ‘nose’ manšaqun ‘organ of smell; nose or nostrils’ 1.2. Specifijication of parts (the top, the sides) manxar, minxarun, munxarun ‘nostril’ nuxaratun ‘tip of the muzzle, tip of the snout; nostril’ 1.3. To be pointed>prominent>to precede ʾānifun ‘that which precedes, which is on the front line’

levels of analysis of the word in arabic

165

2. Specifijications of the organ (big, small . . .) šamamun ‘a well-shaped nose, which is protuberant and thin, whose upper part is straight, and which sticks out at the end, angled downwards’ xinnābun ‘someone who has a large nose’ 3. To lift the nose: a gesture of pride or distaste ʾanfānun ‘someone who holds their nose in the air; full of pride’ ʾašammu ‘someone who has a well-shaped nose, skinny, straight, a bit pointed at the end, and angled downward’. From that > ‘full of pride, someone who holds their head high and is very conscious of their personal honour and rights’ 4. The nose and the air: to breathe in, to exhale, to sense odours, to snifff out naxafa ‘to exhale through the nose, as though one is blowing one’s nose to inhale through the nose’ F.IV to snifff’ našaʾa F.X ‘to sense an odour by smelling it’ nušūʾun ‘a nice scent’ 5. The influence of the nose on the voice: a nasal-sounding voice; animal cries resembling buzzing or growling xunnatun ‘A nasal-sounding voice, to speak through the nose loudly and annoyingly’ xanīnun ‘laughs or tears accompanied by snorts or snivelling’ ġunnatun ‘a nasal sound emitted through the nose; the buzzing of insects’ ʾaġannu ‘someone who speaks through their nose making a nasal sound; having a nasal voice’ 6. Diverse secretions—such as mucus—that pass through the nose xaniba ‘to have a congested nose’ xanabun ‘mucus’ naxaṭa ‘to remove or blow mucus through nose into a tissue; to blow one’s nose’ d̠amma ‘to let mucus run; to expel . . . (said about the nose)’

Matrix 10 Notional invariant: example

{[+nasal], [coronal]} “traction”24 matta ‘to extend’ natara ‘to pull, to extract, to uproot’ + paradigm 2b and 3b

24  For matrices 1 to 6 see Bohas (2000) and Dat (2002); for a more detailed study of matrix 6, see Serhane (2003) and Bohas & Serhane (2003); for matrix 7, see Bohas & Dat (2005); for matrices 8 and 9, see Bohas (forthcoming), and for matrix 10, see Saguer (2003).

166

georges bohas D. Exploring the Methods of Explanation 1. Homonymy by the Blending of Two Etymons

We have just seen the fijirst case—the blending of etymons—with ġaraza, which encompasses two strictly homonymic meanings:25 S1: ‘to prick something with a needle; to drive in, to penetrate (with a pointed instrument); to plunge a stalk into the earth to lay eggs (locusts).’ S2: ‘to give only a small amount of milk (pregnant camel)’

Let us elaborate a little more on the model for blending: etymon ġr

etymon rz ×26

‘lack of milk’

‘to push a sharp object into’

Si

Sj

ġaraz Si ‘lack of milk’  +  Sj ‘to push a sharp object into’

2. Homonymy by Realization of Multiple Matrices in One Etymon Then we have the verb mataʾa, meaning: S1: S2:

‘to hit someone with a stick’ ‘to increase tension, to extend a rope lengthwise’.

The m is [labial], the t is [coronal]: the etymon {m,t} may thus be a realisation of Matrix 1 Notional invariant:

{[labial], [coronal]} “to strike a blow”

And, as such, it takes on the meaning S1:

‘to hit someone with a stick’.

25

 See Bohas et Saguer (2007b) for further details.  We use × to indicate blending.

26

levels of analysis of the word in arabic

167

But the m is also [nasal] and the etymon {m,t} can also serve as a realisation of Matrix 10 Notional invariant:

{[+nasal], [coronal]} “traction”

And, as such, it takes on the meaning of ‘to increase tension, to extend a rope lengthwise’. Figure: M1

M2

{[ labial], [ coronal]}

{[+nasal], [coronal]}

“to strike”

“traction” mt ‘to hit’ ‘to pull’

3. Homonymy Due to Two Diffferent Analyses: n[XY] , [nX]Y. The verb natara includes the following meanings: S1: F.I F.II F.V F.VI

“to disperse” ‘to scatter, to disperse, to disseminate’ ‘to scatter a lot, in great quantity: intensive of F.I’ ‘to be scattered, dispersed, to separate’ ‘to be scattered, dispersed, to disseminate, to be spread’

S2: natura F.II

“nasal functions” ‘to blow one’s nose’ ‘to blow one’s nose’ ‘to aspirate water through the nostrils’ ‘to blow one’s nose’ ‘to draw water through the nose and blow it out through the nostrils’

F.VIII S3: F.I S4: F.IV

“to pull, to dig up” ‘to remove, to take offf someone else’s clothes, to steal them’ “to strike someone with a pointed object”27 ‘to pierce someone with a pointed instrument and draw blood’

27  We do not have an explanation for this meaning (we have not yet explored all of the Arabic matrices); the notional invariant “to be sharp” is clearly important, but for the moment, we do not know much about it.

168

georges bohas

The fijirst hypothesis that we formulate is that naṯara is a development of the etymon {n,t} and that to this extent, it is a manifestation of matrix 9:28 Matrix 9 Notional invariant:

{[+ nasal] [+continuant]} “the nose”

We notice that the meanings of S2: “to blow one’s nose”, “to draw water through the nostrils”, occur in rubrics 6 and 4, p. 165. However, since n is [+nasal] and t [coronal], {n,t} could also be an example of matrix 10: Matrix 10 Notional invariant:

{[+nasal], [coronal]} “traction”

For that matter, we fijind another manifestation of this etymon {n,t} in: natala

[nt]l

‘to take offf a vestment or armour’

Meaning S3 “to pull, to pull up” of F.I ‘to take offf, to remove a vestment from someone else’s body, to steal it from them’ can therefore be linked to this matrix. We can therefore understand why the {n,t} etymon is homonymic: if we take into account the feature [coronal] in t, then it may be construed as a realization of matrix 10 “traction”, while if we take into account the feature [continuant], then it may be taken to be a realization of matrix 9 “the nose”. In this case we shall talk of ambiguity originating in the fact that an etymon is the realization of several matrices. Let us pursue the analysis of natara. We have seen that it also manifests the sense ‘to disperse, to scatter’. It is true that we can establish both a phonetic and a semantic relation in: tarra

[tr]r

‘to disperse, disseminate

the etymon of which can only be {ṯ,r},29 and with: tāra ṯ[w]r tartaratun [ṯr]ṯr farat̠a f [rṯ]

28

‘to be lifted and scatter in the air’ ‘dispersion, dissemination’ ‘to be dispersed, disseminated (of a tribe)’

 For a detailed study of this matrix, see Bohas (forthcoming).  No ambiguity.

29

levels of analysis of the word in arabic

169

This leads us to analyse naṯara by positing that {ṯ,r} is the etymon and n an initial increment. The homonymy therefore stems from the fact that in A and B, the form is analysed as [nt]r, which may be linked to two matrices, and in C, as n[tr]. We shall say in the latter case that homonymy arises because several analyses are possible at etymon level. To summarize: the homonymy of a radical can be caused by: • the fact that it is the result of a blending process: it displays the meanings of the two “parent” etymons • the fact that the etymon is the embodiment of multiple matrices: it encompasses the meaning of each of these matrices • the fact that there are two possible analyses at etymon level, as in [nX]Y and n[XY]. One might ask: does this really work, or is it just an isolated case? I will now provide further examples of this analysis to convince those who are still dubitative.

E. A Case Study 1. The Verb naḥaṭa can mean30 S1: S2:

“to breathe in air” “to have difffijiculty breathing, to have laboured breathing (said of a man charged with a heavy burden)” “to reject someone harshly”

S3:

S1 and S2 share a common meaning: “inhaling (modality: with efffort)”. The n is the feature [+nasal] and the feature [continuant]: the etymon is an example of matrix 9 and the meaning appears in the rubric: 4. The nose and the air: to breathe in, to exhale, to sense odours, to snifff out.

We see that among the derivatives of this root, Kazimirski (1860: s.v.) identifijies: naḥḥāṭun

30

‘full of pride, proud’

  For verbs beginning with n, see the French original in Bohas & Saguer (2006).

170

georges bohas

which may be placed in rubric 3 of the organization of the notional invariant of this matrix: 3. To lift the nose: a gesture of pride or distaste

In this case, the analysis of naḥaṭa is: [nḥ]ṭ, the ṭ being incremental, and hence without any semantic value. As for S3: “to reject someone harshly”, the analysis requires us to take into account the following: waṭaḥa ṭaḥ ā ṭaḥ ara ṭaraḥ a ṭāḥ a ṭaḥ aṯa

‘to reject someone violently’ ‘to agitate, to provoke someone; to throw someone to the ground (cause-efffect relationship)’ ‘to push forward, to chase’ ‘to throw, to push away from oneself forcefully’ ‘to fall (cause-efffect relationship)’ ‘to push, to push someone back with one’s hand’

Each of these meanings contains the sequence ṭḥ and a striking semantic unity, which allows us to identify the etymon {ṭ,ḥ} in cases where the reverse realization furnishes the same sense: ḥ aṭā

‘to throw, to heave something’

This motivates the analysis n[ḥṭ], for naḥ aṭa ‘to push someone away harshly’. We see that the diversity of etymon-level analyses does in fact account for this homonymy. In one case, we identifijied the matrix, while in the other we could not go any further than the level of the etymon. This leads us to discuss two levels of explanation: identifijication of the matrix (optimal), and identifijication of the etymon. 2. Simple cases 2.1. laḥana31 S1: S2: S3: S4:

“to take a liking to someone, to fall in love with someone” “to speak Arabic poorly” “to speak with someone in slang so as not to be understood by others” “to hear a word or expression, to understand (link between speaking and comprehension)”

31

 For verbs beginning with l, see Bohas & Saguer (2007b) for a slightly diffferent analysis.

levels of analysis of the word in arabic

171

We begin by studying the fijirst meaning, S1: “to take a liking to someone”. We can easily establish a phonetic and semantic relationship with: ḥanna

[ḥn]n

nāḥa ḥanḥana

n[w]ḥ [ḥn]ḥn

ḥanā

[ḥn]w

ḥaniba

[ḥn]b and F.II

saḥana

s[ḥn] F.III

‘to moan, to let out a tender cry (said of certain animals, for example a she-camel expressing afffection for its calf )’ ‘to coo (said of pigeons)’ ‘to have and receive afffection, compassion, of both emotion and worry’ ‘to have great afffection for someone (said of a mother who, out of love for her children, does not want to remarry)’ ‘to experience a sentiment of pity or compassion for someone’ ‘to treat someone with kindness’

From this relationship, we can deduce that laḥana should be analyzed as a form incorporating the etymon {ḥ,n} by initial incrementation of an l that plays the role of a prefijix marking the middle voice. In the semantic organization of matrices, the point of departure is always a concrete meaning. We turn once again to Hurwitz (1913: 72): “It must also be borne in mind that primitive ideas are generally concrete, and that an abstract idea is secondary, in that it is often based on some objective aspect involved in the expression of the abstract idea, as when anger is denoted by ‘a reddening of the face’, displeasure, by ‘a falling of the countenance’ etc.” The example that we are studying here is particularly illuminating. The word ḥanna has an abstract meaning of ‘being emotionally moved’, ‘to have compassion for someone’, ‘to experience a great tenderness for someone’. In matrix terms, ḥanna means ‘to moan, to emit a tender cry (speaking of a she-camel)’, just like ʾanna ‘to moan’ which comes from the same matrix: Matrix 4 Notional invariant:

{[coronal], [pharyngeal] [~dorsal] [-voiced]} “a mufffled voice, a muted sound, hoarse”

From a physical moan, as a noise, we move on to why the noise is produced, and from there, to what it expresses. In ḥanna, the two meanings are both present. However, in laḥana, only the abstract sense appears. We have now identifijied the matrix of which laḥana ‘to take a liking to someone’ is a realization. We shall now move on to meanings 2 and 3, which are visibly linked: S2 “to speak Arabic poorly”, S3 “to speak with someone in slang so as not to be understood by others”. For this sense, the

172

georges bohas

analysis [lḥ]n is implemented. In other words, the etymon is {l,ḥ}, which is itself a realization of matrix 8 {[+lateral], [+continuant]}, linked to the notional invariant “the tongue”. We have also seen (p. 164) that “to speak” is one of the developments of the notional invariant “the tongue”: 2. The tongue as an instrument of language: to speak, to speak in many ways, to be talkative, to curse, to argue with someone, to hurt with words, to hurt with insulting words; to speak with authority > to command. 32

We fijind the same meaning in: laxxa

[lx]x

laxlaxāniyyun

[lx]lx

laxiya laġā

[lx]y [lġ]w

‘to be incomprehensible, to speak (especially Arabic) in an incomprehensible way’ ‘someone who has difffijiculty speaking or who only speaks with angry words’ ‘to be very talkative and to speak around the main idea’ ‘to say stupid things, to speak purposelessly or frivolously’

S4 “To hear a word or expression, to understand”, a meaning we meet again in laḥina ‘to be intelligent’, also seems to fijit into the same analysis. The cause and efffect relationship that we assume—‘to speak’ > ‘to understand’ > ‘to be intelligent’—is explicit in Syriac, where mlīla means not only ‘(someone) who is speaking’, but ‘(someone) who is intelligent’, or in Greek, where logikós covers two types of meaning: I ‘that which concerns speaking’, II ‘that which concerns reasoning’.33 Thus, we can draw the lexical tree of the word: M8

M4

{[+lateral], [+continuant]}

{[+coronal], [pharyngeal]}

“the tongue”

“a mufffled voice”

{lḥ}

{ḥn}

[lḥ]n

l[ḥn] laḥan

32 33

  See the organization of the notional invariant, pp. 163–164.   See Bailly (1950).

levels of analysis of the word in arabic

173

2.2. lasaba and lasiba S1: S2: S3:

lasaba lasaba lasiba

“to puncture someone (said of serpents)” “to give someone a lashing” “to lick, to attach and glue to something”

Let’s begin with S2. The fact that this form is semantically and phonetically linked to: sabaʾa

[sb]ʾ

‘to lash at someone with a whip so as to draw blood’

leads us to deduce that lasaba must be analyzed as a form that incorporates the etymon [sb] by initial incrementation of a prefijix l, devoid of semantic value. The etymon [sb] is originally from the matrix {[labial], coronal]} the notional invariant of which is: “to strike”, here with the specifijication of method (“with a whip”), and other specifijications in: rabasa safaʿa

‘to hit with the hand’ ‘to hit, to punch (said of birds, who give heavy hits with their wings when fijighting)’ sāfa ‘to hit someone with a sabre’ nasama ‘to kick the soil with your foot’

The other two meanings emerge from the matrix {[+lateral], [+continuant]} “the tongue”. The fijirst fijits into the rubric: 3. The tip of the tongue: pointed, to be sharp, to form a point and, by extension, to sting, to be sharp with one’s remarks

as in: lasaʿa lasana salama F.IV passive ʾuslima

‘to prick someone, to sting (said of a scorpion or a snake)’ ‘to form a point (of a shoe, etc.)’ ‘to sting (said of a scorpion)’ ‘to prick someone, to bite (said of a snake)’ ‘to be bitten by a snake’

and the second in the rubric 1.4 and 1.4.1 (see p. 163), as: lassa lasada ladasa lahasa

‘to lick (a bowl, a pot, etc.)’ ‘to lick (a bowl)’ ‘to lick’ ‘to lick’

Here it is simple to generate the lexical tree:

174

georges bohas M8

M1

{[+lateral], [+continuant]}

{[labial], [coronal]}

“the tongue”

“to strike a blow”

{ls}

{sb}

[ls]b

l[sb] lasab

2.3. One fijinal example The verb masana also includes two homonymic meanings: S1: “to pull, to extract one thing from another” S2: “to lash at someone with a whip so forcefully that they fall to the floor”

This homonymy can be explained by the fact that the same etymon {m,s} is the result of two diffferent matrices, where in both cases the analysis is [ms]n. The fijirst is once again matrix 10: {[+nasal], [coronal]} which has the notional invariant: “traction”. This etymon is also present in the following words: masā masara masala F.VIII masaxa F.VIII masaḥa F.VIII

‘to pull, to extract by pulling towards you’ ‘to pull, to extract something from where it is originally found’ ‘to pull from the wrapper’ ‘to unsheathe, to pull from a casing (a sabre)’ ‘to unsheathe, to pull from a casing (a sabre)’

But the m is also [labial] and the s [coronal]. The etymon {m,s} can also be a realisation of matrix 1: {[labial], [coronal]} which is a development of the notional invariant “to strike”, as is the case with the following words: masaṭa maṣaʿa lasaba sabaʾa

[ms]ṭ [mṣ]ʿ l[sb] [sb]ʾ

‘to lash someone with a whip all over their body’ ‘to lash someone with a whip’ ‘to give someone a lashing with a whip’ ‘to lash someone with a whip to the point of drawing blood’

with both the m and the b being [labial].

levels of analysis of the word in arabic

175

In this case, following the analysis proposed in Bohas (2000), the meaning S2 is treated as the notional invariant: “to strike”; characterization of the instrument: “with a whip”. The lexical tree looks like this: M1

M10

“to strike a blow”

“traction”

{[labial], [coronal]}

{[+nasal], [coronal]} {m,s}

[[ms]n] masan

3. More Complex Cases 3.1. The Verb labaxa Encompasses Eight Meanings34 S1: “to be plump (said of the body)” S2: “to beat, to hit someone” S3: “to kill someone” S4: “to take something to someone, from someone’s hand” S5: “to draw information out of someone by tricking them” S6: F.III: “to slap someone in the face” S7: F.V: “to perfume oneself with musk” S8: “to insult someone, an insult”

We can see that many of these meanings have a polysemous relationship. In other words, we can establish a probable semantic relationship: • S2, S6 and S3 are realizations of the notional invariant “to strike a blow”. Killing is linked to this concept by the cause > efffect relationship and S6 specifijies the modality of the action. • S4 and S5 are both derived from the notional invariant: “to pull, to bring something toward oneself ”.

34

  For a slightly diffferent analysis see Bohas & Saguer (2007b).

176

georges bohas

Which leaves us with fijive homonymic meanings: A. S1: B. S2+S6+S3:

C. S4+S5:

D. S7 F.V: E. S8:

“to be plump (referring to the body)” “to strike a blow” S2: “to fijight, to hit someone” S6 F.III: “to slap someone in the face” S3: “to kill someone” “to pull, to bring toward oneself” S4: “to take something from someone, from someone’s hand” S5: “to extract information from someone by trickery” “to perfume oneself with musk” “to insult someone, an insult”

We will now attribute a source matrix to each of these homonymic meanings. Starting with meaning A.: S1: labaxa “to be plump (referring to the body)”, it is easy to see that this form is semantically linked to: rabīxun habayyaxun ʾanbaxun muxabxabatun

r[bx] h[bx] n[bx] [xb]xb

‘fat; a heavy, flabby body’ ‘a chubby young person’ ‘fat, heavy, hard’ ‘beautiful and fat (camels)’

This leads us to conclude that labaxa should be analyzed as a form incorporating the etymon {b,x} with initial incrementation of an l which has the role of an incremental prefijix designating the meaning that we called “stative”, as defijined by Joüon (1923: 95). This etymon {b,x} is a realization of the matrix {[labial], [dorsal]}, which has the notional invariant “curvature”. One of the fijirst manifestations of curvature in a convex form is precisely the concept of fatness:35 ‘fat, fatty, robust, something that looks bloated: ∩, as in: bajja bājilun fajiʾa hijaffun

F.VII

‘to be fatty, to have round haunches (said of animals at pasture who tend to gain weight)’ ‘fat, chubby’ ‘to have a fat stomach’ ‘someone who has a big stomach, a big paunch’

Thus, meaning A.: labaxa “to be fat (referring to the body)” can be analysed as l[bx], and is therefore a product of the matrix {[labial], [dorsal]} of which the notional invariant is “curvature”.

35

  See Bohas (2000: 108).

levels of analysis of the word in arabic

177

Meaning B. = S2+S6+S3: to strike: S2: “to fijight, to hit someone”; S6: F.III: “to slap someone in the face”; S3: “to kill someone”. Here it is easy to show the relationship with other manifestations of the etymon [lb] such as: wabala labana

w[bl] [lb]n

laḥaba

l[ḥ]b

‘to hit someone with a stick’ ‘to hit someone violently, to knock someone out by hitting them with a stick’ ‘to strike someone with a sabre’

and for the implied sense: habila

h[bl]

to lose one’s son, a son in death

This enables us to conclude that labaxa includes the etymon {l,b}, which is itself a product of the matrix {[labial], [coronal]}, of which the notional invariant is: “to strike a blow”. But we observe that a triplet also exists: laxxa laxaba laxama

[lx]x ‘to slap someone in the face’ [lx]b ‘to slap someone in the face’ [lx]m ‘to strike someone on the face’

which allows us to extract the etymon {l,x}, also connected to the meaning “to strike”. The segment l is analyzed as [coronal} and x as [dorsal]. Thus this etymon is a realization of the matrix {[coronal], [dorsal]} and we can conclude that labaxa in meaning B results from the blending of two etymons {l,b} and {l,x}, a type B combination (see Bohas 2000: 50). For meaning C. = S4+S5: “to pull, to bring toward oneself”: S4: “to take something to someone, from someone’s hand”; S5: “to extract information from someone by trickery” we will propose an analysis l[b]x at etymonlevel, that is, an etymon {l,x} with an increment b. We can now establish a phonosemantic relationship with: (in the order xl): xalaja xalā xalxala saxala xalasa F.III F.V. F.V.

(xly)

‘to pull toward oneself, to attract toward oneself’ ‘to pull up’ ‘to remove all of the skin from a bone’ ‘to take offf, to remove, to rob with a trick’ ‘to take, to remove, to ravish secretively in the blink of an eye, to pull up’ ‘to take something from someone else; to seize; to take over something’ ‘to remove, to take’ ‘to extract reciprocally, each taking something for their side’

178

georges bohas

F.VIII.

‘to pull something toward oneself hurriedly’ (same meaning as F.I. except adding the element of celerity)

(in the order lx:) malaxa

‘to pull something toward oneself with force by grabbing it with hands or teeth’ ‘to pull, to extract, to pull out (a tooth, an eye); to pull from a sheath (a sabre, etc.)’

F.VIII

The etymon {l,x} therefore encompasses two homonymic meanings “to strike”, as a manifestation of the matrix {[labial], [coronal]}, and “to bring something toward oneself”, as the manifestation of another matrix, which we have not yet studied, although we are fairly certain that it would include the following features: {[+approximant],36 [+continuant]} {[coronal]

for the notional invariant: “to pull something toward oneself ” and which is shown in many more words than those we list hereunder: 37 (etymons with l) ḥalaʾa ḥalata salla salaba šalaḥa ṣalaba laḥā laḥata halaba halata halaḍa

F.II

‘to remove the flesh of a dead animal’ ‘to dig up, to remove in flakes’ ‘to pull, to softly remove one object from another’ ‘to violently snatch something from someone’ ‘to skin, to remove someone’s clothes’ ‘to pull, to extract the marrow from bones’ ‘to remove the bark of a tree’ ‘to remove the bark from a piece of wood, to peel it offf’ ‘to pull at one’s hair or whiskers’ ‘to peel something by removing the skin or the bark’ ‘to pull, to extract, to dig up something from its place’

[+approximant]  This composition [coronal]   seems very complex in our attempt to categorize the class r, l. This is due to the fact that according to Yeou & Maeda (1994), pharyngeals and uvulars in Arabic are also characterized by the feature [+approximant]. In fact, the case study that follows will allow us to discuss this point. If the gutturals can occur in this matrix, and if it were formulated simply as {[+approximant], [+continuant]}, then this would constitute proof that the gutturals and r, l are both part of the same class: [+approximant]. 37  For further data see Bohas & Saguer (2007b). 36

levels of analysis of the word in arabic

179

(etymons with r) ḥaraba ḥarasa saraqa sarā ʿaraza ʿarʿara

‘to skin someone, to plunder (a caravan, a tribe)’ ‘to steal something’ ‘to steal’ ‘to remove, to move away, to distance something from someone’ ‘to pull with force’ ‘to wiggle the cork of a bottle in order to remove it; to remove, to take out the cork, to pull, to pull up, to puncture the eye’ ʿarama ‘to eat the meat that is attached to the bone’ ʿariya ‘to be naked, without clothes (cause-efffect relationship)’

Although we have yet to organize the notional framework for all of these examples, they still sufffijice, collectively, to justify the existence of the matrix (M 11) {[+approximant], [+continuant]} {[coronal] “to bring something toward oneself”

and the analysis where labaxa has a meaning C. = S4+S5 “to pull, to bring toward oneself”: S4: “to take something to someone, from someone’s hand”; S5: “to extract information from someone by trickery” is a manifestation of this matrix. I am often asked: how do you extrapolate a matrix from these analyses? And indeed, I have just explained the method used. Now we must scrutinize the lexicon to see if [+approx] should be characterized by the feature [cor], or if the other approximants—pharyngeals—can also be incorporated. We will have thus completed the examination of the phonetic constituents; the organization of the notional invariant still needs to be completed. As for labaxa , D. S7 F.V: “to perfume oneself with musk” it is analyzed as l [bx], in which we fijind the etymon {b,x}, as in: baxara baxira ʾabxaru baxxa

[bx]r

‘to perfume someone or something with incense’ ‘to smell bad’ ‘someone with bad, fetid breath’ ‘to snore while sleeping’

This etymon is a product of matrix 2. {[labial],

[+continuant]}

which combines the labials b and f with non-voiced fricatives.

180

georges bohas

The ramifijications of the notional invariant are simple: • movement of the air: wind, breath • exhaling of air by man or animal • >38 consequences (various odours) Below are some realizations of this matrix with f: ∈{f,ḥ} faḥḥa faḥfaḥa faḥā F.II fāḥa/fawaḥa lafaḥa nafaḥa

‘to hiss (snake); to whistle while sleeping’ to be hoarse ‘to season dishes with spices’ ‘to emit a smell; to smell good or bad’ ‘to blow (referring to a warm breeze)’ ‘to emit an odour; to blow (referring to a cold wind)’

∈{f,x} faxxa fāxa/fawaxa nafaxa

‘to snore and whistle, referring to someone who is sleeping; an aroma that pervades an area’ ‘to spread (referring to an odour); to whistle (wind); to break wind (referring to a human)’ ‘to breathe through ones mouth, to fart’

We still have to look at the fijinal meaning: E. S8: “to insult someone, an insult”. We notice that labaxa begins with an l [+lateral] and ends with an x which is a segment [+continuant]. Therefore the analysis is l[b]x, from the etymon ∈{l, x}, which also appears in: laxiya xaṭila

‘to be very talkative and to talk around a subject’ ‘to use many words to say very little’

In this case, the etymon is a product of matrix 8, Matrix 8 Notional invariant:

{[+approximant] [+continuant]} [+lateral] “the tongue”39

38  Remember that we use the following sign to mean that there is a semantic relationship: cause > efffect. 39  The ramifijications of the notional invariant are explained on pp. 163–164.

levels of analysis of the word in arabic

181

exactly like: ∈{l,ġ} ‘to speak, in general’ ‘to say stupid things, to give a discourse without purpose, to speak without sincerity’ ‘to contradict oneself while speaking, to make an error’ ‘to make a mistake while speaking, to make an error’

laġā

laġiya

All of the meanings encompassed by this radical are explained in the following diagram, which not only acts as a lexical tree schematizing the phonosemantic composition, but also accounts for the homonymy that we have also identifijied: Matrix level M6

M2

M1

M5

M8

M11

{[lab], [dor]} {[lab], [cont]} {[lab, cor]} {[cor], [dor]} {[+lateral], [+cont]}

{[+approx], [+cont]}

“curvature”

“air>odour”

“a blow”

“a blow”

“the tongue”

“to bring towards oneself”

A

D

B

B

E

C

{b,x}2

{l,b}

{l,x}1

{l,x}2

{l,x}3

Etymon level {b,x}1

× l[bx]

lbx

Radical level labax

l[b]x

182

georges bohas

A. S1: B. S2+S6+S3:

C. S4+S5: D. E. S8:

“to be plump (referring to the body)” “to strike a blow” S2: “to fijight, to hit someone” S6 F.III: “to slap someone in the face” S3: “to kill someone” “to pull, to bring toward oneself” S4: “to take something to someone, from someone’s hand” S5: “to extract information from someone by trickery” S7: F.V: “to perfume oneself with musk” “to insult someone, an insult”

When the tree is complete, the analysis reaches its end. Let us explain this: for meaning B, there is a blending of the two etymons: {l,b} and {l,x}. The etymon {b,x} is the product of two matrices because the x can represent either the feature [dorsal] = matrix 6 or the feature [+continuant] = matrix 2: M6

M2

{[labial], [dorsal]}

{[labial], [+continuant]}

“curvature”

“air > odour”

A

D {b,x}

The etymon {l,x} may result from three matrices because the l could be analysed as [coronal] = matrix 5, [+lateral] = matrix 8, or [+approximant] = matrix 11: M5

M8

M11

{[coronal], [dorsal]}

{[+lateral], [+continuant]}

{[+approx], [+continuant]}

“to strike a blow”

“the tongue”

“to bring towards oneself”

B

E

C

{l,x}

levels of analysis of the word in arabic

183

Following the analysis, the l could either be an element of the etymon, or incremental. The same is also true of b; x is always a component of the etymon. In the present case, we have succeeded in matching every meaning to a matrix: this is the optimal level of explanation. However, it is not always possible to do so. 3.2. laṣafa and laṣifa This item encompasses the following meanings: S1:

“to adjust by placing side by side one on top of the other (e.g. stones when raising a wall, when building it)” S2 laṣifa “to be shrivelled up and glued to the bones (referring to the skin on an emaciated body)” S3: “to wrap the bottom of an arrow with a strap of sinew” S4: “to shine, to gleam”

These four meanings seem to have a homonymic relationship—we cannot establish a plausible semantic link between them. For S1: “to adjust by placing side by side one on top of the other (e.g. stones when raising a wall, when building it)”, the connection with: ṣaffa xaṣafa ṣannafa

[ṣf ] f x[ṣf ] ṣ[n] f

‘to put in order’ ‘to adjust and join solidly’ ‘to compose, to make (a work of art, a book)’

allows us to identify a common etymon {ṣ,f } “to tidy, to arrange”. For S2, we look back at matrix 8: 1.4.1. Consequence (1): to moisten and glue, as in: laṣṣa laṣiqa laṣaġa

[lṣ]ṣ [lṣ]q [lṣ]ġ

F.VIII

‘to attach to, to glue fijirmly to’ ‘to be glued’ ‘to be shrivelled and glued to the bones (referring to the skin on an emaciated body)’

For S3: “to wrap the base of an arrow with a strap of sinew”, the taking into account of: raṣafa

r[ṣf ]

ʿafaṣa

ʿ[ fṣ]

ʿaṣaba

ʿ[ṣb]

‘to wrap he top of an arrow with a strap of flattened sinew to ensure the arrowhead is locked securely’ ‘to encircle the opening of a bottle with an ʿifāṣ [the skin with which the head of a flask, or bottle, is covered]’ ‘to bandage, to wrap with a bandage; to dress with a bandage (the head, a limb)’

184

georges bohas

allows us to extract the etymon {ṣ,f } which is itself a derivative of matrix 6, f as a [labial] and ṣ [dorsal]40 of which the notional invariant is curvature and of which “to encircle”, “to surround” is a consequence. 41 Finally laṣafa in terms of S4: “to shine, to gleam” can be related to: walafa w[lf ] jafala

j[ fl]

‘to be illuminated with repeated bursts of light that follow each other without interruption (said of lightning)’ ‘to shine’

where we identify the etymon {l,f } “to shine”. As for the matrix to which it belongs, it seems to include the feature [approximant], as in: balaqa F. VIII baraqa lamaḥa lamaʿa ramaḥa

‘to shine, to gleam’ ‘to shine, to gleam, to glitter’ ‘to shine’ ‘to shine’ ‘to shine (said of lightning)’

In each of these cases, as in walafa and laṣafa, we observe the combination {[approximant], [labial]}42 linked to the invariant “to shine”, but it requires a very complex matrix that we have not yet worked on. For the moment, we can say that the identifijication of the etymon is certain, but that of the matrix is not (yet). However, we are still able to provide a lexical tree: M8

M?

M6

{[+approximant]?}

{[labial], [dorsal]}

“tongue”

“brilliance”

“curvature”

{l,ṣ}

{l,f } to shine

{ṣ,f }1  {ṣ,f }2 to arrange

[lṣ] f

l[ṣ] f

{[+lateral], [+continuant]}

l[ṣf ]

l[ṣf ]

laṣaf

40

 The emphatics are characterized by the features [dorsal], [pharyngeal], [coronal].  See Bohas (2000: 102–106). 42  Here we make the same remark as in note 36. It appears that gutturals are a part of this matrix, so that they can be linked to the class of approximants. If not, we must limit [+approximant] by adding [coronal]. 41

levels of analysis of the word in arabic

185

3.3. Kazimirski (1860: s.v.) gives eight meanings of the verb mataḥa for F.I. We can explain three of them fairly easily: S1: S2: S3:

“to draw water (from a well)” “to hit someone” “to fart, to break wind”

The fact that mataḥa is semantically linked to: matta mataxa F.VIII matasa mataha matara

F.VI

‘to draw water from a well with a rope and without the help of a pulley’ ‘to uproot something from its place’ ‘to take away, to dig up’ ‘to tug and twist from all sides trying to pull something up’ ‘to pull, to extract water from a well with the help of a rope’ ‘to pull, each person from their own side, to tug’

allows us to identify the etymon {m,t}, which is a product of matrix 10. Matrix 10

{[+nasal], [coronal]} Notional invariant: “traction”

In this case, the analysis is: [mt]ḥ. The fact that mataḥa is semantically linked to: mataʾa ‘to hit someone with a stick’ matana ‘to hit someone with force’ latama ‘to hit someone’

once again allows us to identify the etymon {m,t}, but this time as a product of matrix 1. Matrix 1

{[labial], [coronal]} Notional invariant: “to strike a blow”

Finally, the fact that mataḥa is semantically linked to: xaḍama ‘to break wind’ maxratun ‘a fart’ maxana ‘to blow’

allows us to identify the etymon {m,x}, a manifestation of matrix 2 Matrix 2

{[labial], [+continuant] } Notional invariant: “movement of air”, one of which is the expulsion of air from a human or an animal, otherwise known as a ‘fart’.

186

georges bohas

There follows a number of analogous embodiments of this matrix: tamiha tahima xabaja nafaxa fāḥa/fawaḥa/

‘to be spoiled and foul-smelling (referring to milk, butter, or food)’ ‘to smell badly, to stink (said about the mouth, breath)’ ‘to break wind’ ‘to breathe through the mouth; to fart’ ‘to spread its smell; to smell good or bad’

The lexical tree can be created at the optimal level for these three meanings:

M10

M1

M2

{[+nasal], [coronal]} “traction”

{[labial], [coronal]} “to strike”

{[labial], [+continuant]} “movement of air”

[mt]ḥ ‘to pull water from a well’

[mt]ḥ ‘to hit someone’

m[t]ḥ ‘to fart’

As for the other meanings that Kazimirski (1860: s.v.) gives, most can be grouped with those that we have already identifijied: • “to leave droppings” is metonymically related (in terms of continuity or cause-efffect) to the meaning “to fart” S3. • “to knock someone to the ground” in the analysis of the matrix 1 {[labial), [coronal]}, “to knock to the ground”, fijits into the rubric “to precipitate, to cause to fall, to throw on the ground” (Bohas 2000: 70). • “to pull from the ground or separate a part from its whole, by extracting it from its base” is also a product of the matrix 1 {[labial), [coronal]}, rubric A133: “to cut, to separate a part from its whole”, see Bohas (2000: 69). • “to be advanced (said of days when the sun is already high in the sky)” could be inserted into the matrix for “traction” rubric B5. “to extend a period of time”, see Saguer (2003).43 We can therefore extend this verb’s lexical tree by adding the derived meanings:

43  It is not so much a question of solid or liquid, but of time, as in the French “tirer en longueur ou tirer à sa fijin”.

levels of analysis of the word in arabic

187

M10

M1

M2

{[+nasal], [coronal]} “traction”

{[labial], [coronal]} “to strike”

{[labial], [+continuant]} “movement of air”

[mt]ḥ ‘to pull [draw] water from a well’

[mt]ḥ ‘to hit someone’

m[t]ḥ ‘to fart’

‘to be advanced (day)’ ‘to throw to the ground’ ‘to leave droppings’ ‘to pull or separate from its whole’

Regarding ‘to lay eggs by pushing its tail into the ground (referring to locusts)’, the etymon {t,ḥ} may be a variant of the etymon {t,x}, which is itself a realization of the matrix {[coronal], [dorsal]} “to strike a blow”, with specifijication of the instrument added: “to strike (with a pointed object)”, as we saw in our study of mataxa. This leaves us with F.V ‘to move one’s feet alternately when walking, that is, fijirst using the left foot, and then the right foot one in front of the other (said of camels)’, about which we prefer to say quite simply that we have no insight at the present time, rather than invent far-fetched hypotheses.

Conclusion Our study has proven very diffferent from those which centre on the triconsonantal root, the Arabic doxa widely accepted by most scholars of linguistics. For the latter, it sufffijices to identify the three consonants involved in order to complete the analysis, even if incongruities or semantic incompatibilities are found. In this way, they accept their results without question. For example, they might choose to look at the word mataʾa: S1: “to hit someone with a stick” or S2: “to tender, to stretch out along a rope” and the word masana: S1: “to pull, to remove one thing from another” or S2: “to punch someone so forcefully that they fall to the floor”. In these cases, the two meanings of each word have nothing to do with each other, and now we know why. If one remains within the framework of the traditional tri- or quadri-consonantal root, it is impossible to explain phenomena like homonymy and antithetical polysemy which are omnipresent in the Arabic lexicon, and are there for everyone to observe. However, for all those

188

georges bohas

who make their living searching for roots, it is more than a little difffijicult to accept the organization I am proposing. This situation is described quite accurately by Retsö (1994): “On the whole, it is striking how difffijicult it is for western Arabists to disentangle themselves from the models and explanations developed by traditional Arabic grammar”. It should be noted that all of our studies have been carried out taking the lexicon of Arabic as a single, achronic whole. As we have often repeated, when studying the lexicon, it is vain to go back to an earlier, biliteral stage in order to diachronically derive a triliteral stage. In other words, the old debate—biliteral or triliteral—is pointless. We fijind it diffijicult to understand why certain readers―hasty readers, more inclined to reduce the new to the old in order to deny innovation―should wish to see in the tme a revival or rephrasing of this opposition. Binary composites (matrices and etymons) and ternary composites (radicals) exist for all to see. They have not been disguised or destroyed with the passing of time. In fact, they are visible the second you open a good dictionary. Every Arabic word is binary or ternary, according to the level of analysis. Here, I have been able to show you how the tme explains the homonymy that appears in Arabic words. It also explains antithetical polysemy, but that’s another story . . .

levels of analysis of the word in arabic

189

Appendix 1. Table of Phonetic Symbols

m b f ṯ ḏ t d s z š j ṭ ḍ ḏ̣ ṣ l n r k g q G x ġ ḥ ʿ

ʾ h

[±consonantal]

+ + + + + + + + + + + + + + + + + + + + + + + + + +

[±sonant]

+ - - - - - - - - - - - - - - + + + - - - - + + + +

[±approximant]

- - - - - - - - - - - - - - - + - + - - - - (+)(+)(+)(+)

[±voiced]

+ + - - + - + - + - + - + + - + + + - + - + - + - + - -

[±continuant]

+ - + + + - - + + + + - - + + (-) + + - - - - + + + + - +

[labial]

+ + +

[coronal]

+ + + + + + + + + + + + + + +

[dorsal]

+ + + +

[pharyngeal]

+ + + +

[±anterior]

+ + + + + + - - + + + + + + +

[±distributed]

+ + - - - - + + - - + - - - -

[±lateral] [±nasal]

(+)

+ + + + + + + + + + + + + +

- - - - - - - - - + - - + - + - - - - - - - - - - - - - - - + - - - - - - - - - - -

Table 144 [±consonantal]

[±sonorant]

“The feature [+consonantal] denotes sounds with a radical constriction in the supralaryngeal cavity” (Kenstowicz 1994: 36); [-consonantal] denotes sounds produced without such a constriction. Since the feature [consonantal] does not encompass consonants that are produced in the larynx, the two glottals (h and ʾ) are excluded from this classifijication. “The [+sonorant] sounds are produced with a vocal tract cavity confijiguration in which spontaneous voicing is possible” (Chomsky & Halle 1968: 302). The [-sonorant] sounds require a constriction that reduces the flow of glottal air and makes voicing more difffijicult. “Thus the natural state for sonorants is [+voiced] and for nonsonorants (termed obstruents) is [-voiced].” (Kenstowicz 1994: 36).

44  In the table, the brackets ( ) indicate that the specifijication in question has yet to be settled.

190 [±approximant]

[±voiced]

[±continuant]

[±nasal]

[±lateral]

[labial]

georges bohas In terms of articulation, the Encyclopaedia Britannica defijines the approximant as “a sound that is produced by bringing one articulator in the vocal tract close to another without, however, causing audible friction”, and the defijinition of Malmkjær (1991) adds that “any speech sounds so articulated as to be just below friction limit, that is, just short of producing audible friction between two speech organs”. The most common examples of approximants are liquids and high glides. Ladefoged (1975: 55–56) adds the h. Yeou & Maeda (1994) describe Arabic pharyngeals and uvulars as [approximant]. The fact that gutturals may be characterized as [approximant] is therefore a hypothesis that seems to be fairly well established. “The vocal folds can assume a number of confijigurations that are linguistically signifijicant. Here we distinguish two. First, the folds may be brought together (adducted) along their entire length in such a way as to be set in vibration when air passes between them. Sounds produced with this laryngeal confijiguration are called voiced. They are opposed to voiceless sounds, which lack vibration due to an instruction to either separate the folds or increases their tension. For example, the initial [s] and [z] sounds of sip and zip are opposed as voiceless to voiced.” (Kenstowicz 1994: 14). “Continuants are produced by impeding, but not completely blocking, the flow of air throught the glottis, or the pharynx or through the centre of the oral tract; noncontinuants are made by completely blocking the flow of air through the centre of vocal tract” (Katamba 1989: 50). In Chomsky & Halle (1968: 318), the defijinition of l poses a problem: “the characterization of the liquid [l] in terms of the continuant-noncontinuant scale is even more complicated. If the defijining characteristic of the stop is taken as total blockage of airflow, then [l] must be viewed as a continuant . . . If, on the other hand, the defijining characteristic of stops is taken to be blockage of air flow past the primary stricture, then [l] must be included among the stops.” “Nasal sounds are produced with a lowered velum which allows the air to escape through the nose; nonnasal sound are produced with a raised velum so that the air from the lungs can escape only through the mouth.” (Chomsky & Halle 1968: 316) A [+lateral] sound “is produced by making a constriction with the central portion of the tongue but lowering one or both margins so that air flows out the side of the mouth.” (Kenstowicz 1994: 35). characterizes the sounds produced with a constriction made by the lips.

levels of analysis of the word in arabic [coronal]

[±distributed]

[±anterior]

[dorsal]

[pharyngeal]

191

“Coronal sounds are produced with the blade of the tongue raised from its neutral position; non-coronal sounds are produced with the blade of the tongue in the neutral position. The so-called dental, alveolar, and palato-alveolar consonants are coronal, as are the liquids articulated with the blade of the tongue” (Chomsky & Halle 1968: 304). “Distributed sounds are made with an obstruction extending over a considerable area along the middle-line of the oral tract; there is a large area of contact between the articulators. In nondistributed sounds, there is a smaller area of contact. This feature is primarily used to distinguish apical sounds form laminal sounds . . .” (Katamba 1989: 44–45). “In the production of anterior sounds, the main obstruction of the airstream is at a point no farther back in the mouth than the alveolar ridge; for nonanterior sounds the main obstruction is at a place farther back than the alveolar ridge” (Katamba 1989: 44). characterizes the sounds produced with a constriction formed by the back of the tongue and situated between the soft palate and the uvula (velar and uvular consonants). The arrow above the table indicates that, lexically, the jīm is a dorsal (the voicing of k and q). characterizes the segments called gutturals in Arabic tradition, notably: ʾ, h, ʿ, ḥ, ġ, x and q. For more on the problems that this characterization poses, see Kenstowicz (1994: 456fff).

192

georges bohas References

Bailly, Anatole. 1950. Dictionnaire grec-français. Paris: Hachette. Biarnay, Samuel. 1908. “Etude sur le dialecte berbère de Ouargla”. Paris: Ernest Leroux. Bohas, Georges. 1979. Contribution à l’étude de la méthode des grammairiens arabes en morphologie et en phonologie, d’après des grammairiens arabes tardifs. Doctoral thesis, Paris 3 (Lille: Atelier national de reproduction des thèses, 1982). ——. 1997. Matrices, étymons, racines, éléments d’une théorie lexicologique du vocabulaire arabe. Paris and Louvain: Peeters. ——. 2000. Matrices et étymons, développements de la théorie. Lausanne: Editions du Zèbre. ——. 2006. “De la motivation corporelle de certains signes de la langue arabe et de ses implications”. Cahiers de linguistique analogique 3, 311–41. ——. forthcoming. L’émergence du sens dans le lexique de l’arabe. 1st International Conference “La fabrique du signe”, Toulouse, october 2006. —— and J.-P. Guillaume. 1984. Étude des théories des Grammairiens arabes, 1. Morphologie et phonologie. Damascus: Publications de l’Institut français de Damas. —— and N. Darfouf. 1993. “Contribution à la réorganisation du lexique de l’arabe, les étymons non-ordonnés”. Linguistica Communicatio 5, 1–2, 55–103. —— and Mihai Dat. 2005. “La matrice acoustique {[dorsal], [pharyngal]} en arabe classique et en hébreu biblique, première esquisse”. Regards croisés sur le Moyen Âge arabe, Mélanges à la mémoire de Louis Pouzet s.j. Mélanges de l’Université Saint-Joseph 58, 125–143. —— and Abderrahim Saguer. 2006. “Sur un point de vue heuristique concernant l’homonymie dans le lexique de l’arabe”. Grammar as a Window onto Arabic Humanism, A Collection of Articles in Honour of Michael G. Carter, ed. L. Edzard and J. Watson. Wiesbaden: Harrassowitz, 130–154. —— and ——. 2007a. “The explanation of homonymy in the lexicon of Arabic”. Approaches to Arabic Linguistics, ed. H. Motzki and E. Ditters. Leiden: Brill, 255–289. —— and ——. 2007b. “La TME et l’explication de l’homonymie en arabe”. Langues et Littératures du monde Arabe 6, 53–96. —— and Rachida Serhane. 2003. “Conséquences de la décomposition du phonème en traits”. Phonologie Champs et perspectives, ed. J.-P. Angoujard and S. WauquierGravelines. Lyon: ENS éditions, 131–155. Brockelmann, Carl. 1910. Précis de linguistique sémitique, translated from the German by W. Marçais and M. Cohen. Paris: Geuthner. Chomsky, Noam and Morris Halle. 1968. The Sound Pattern of English. New York and London: Evanston and Harper & Row. Cohen, Marcel. 1947. Essai comparatif sur le vocabulaire et la phonétique du chamitosémitique. Paris: Champion. ——. 1955. Cinquante années de recherches linguistiques, ethnographiques, sociologiques, critiques et pédagogiques, bibliographie complète. Paris: Imprimerie nationale and Klincksieck. Dat, Mihai. 2002. Matrices et étymons. Mimophonie lexicale en hébreu biblique. Doctoral thesis, Ecole Normale Supérieure Lettres et Sciences Humaines, Lyon. Diab-Duranton, Salem. 2005. La matrice {[coronal], [dorsal]}, Les étymons impliquant le jîm, Master’s degree research dissertation, Lyon, Ecole Normale Supérieure Lettres et Sciences Humaines, Lyon. Ernout, Alfred and Antoine Meillet. 1932 [1959, 4th edition]. Dictionnaire étymologique de la langue latine, histoire des mots. Paris: Klincksieck. Hurwitz, Solomon. 1913 [1966]. Root-Determinatives in Semitic Speech, a Contribution to Semitic Philology. New York: Columbia University Press. Joüon, Paul. 1923. Grammaire de l’hébreu biblique. Rome: Institut Biblique Pontifijical.

levels of analysis of the word in arabic

193

Katamba, Francis. 1989. An Introduction to Phonology. London and New York: Longman. Kazimirski, Albert de Biberstein. 1860. Dictionnaire arabe-français. Paris: Maisonneuve et Cie. Kenstowicz, Michael J. 1994. Phonology in Generative Grammar. Cambridge MA and Oxford UK: Blackwell. Khatef, Laïla. 2003. Statut de la troisième radicale en arabe: le croisement des étymons. Doctoral thesis, University Paris 8. Ladefoged, Peter Nielsen. 1975. A Course in Phonetics. New York: Harcourt Brace Jovanovich. Lambert, Mayer. 1897. “De la formation des racines trilitères fortes”. Semitic Studies in Memory of Rev. Dr. Alexander Kohut. Berlin, 354–362. Lane, Edward William. 1863–1893. Arabic-English Lexicon. Londres-Edinburgh: Williams & Norgate. Lisān. Ibn Manẓūr, Lisān al-ʿarab, undated critical edition by ʿAbd Allāh ʿAlī al-Kabīr, Muḥammad ʾAḥmad Ḥasab Allāh, Hāšim Muḥammad aš-Šāḏilī. Al-Qāhira, Dār al-Maʿārif. McCarthy, John J. 1986. “OCP Efffects: Gemination and Antigemination”. Linguistic Inquiry 17/2, 207–263. Malmkjær, Kirsten, ed. 1991, The Linguistics Encyclopedia. London: Routledge. Nyckees, Vincent. 1998. La sémantique. Paris: Belin. Pottier, Bernard. 1974. Linguistique générale. Théorie et description. Paris: Klincksieck. Qāmūs. Fīrūzābādī. Al-Qāmūs al-Muḥīṭ. Bayrūt: Muʾassasat ar-Risāla. Renan, Ernest. 1855. Histoire générale et système comparé des langues sémitiques. Paris: Imprimerie Impériale. 1958 edition, Paris: Calmann-Lévy. Retsö, Jan. 1994. “Iʿrāb in the Forebears of Modern Arabic Dialects”. Actes des premières journées internationales de dialectologie arabe de Paris, ed. D. Caubet and M. Vanhove. Paris: INALCO, 333–342. Saguer, Abderrahim. 2003. “La matrice {[+nasal), [coronal)}, ‘traction’ en arabe. première esquisse”. Langues et littératures du monde arabe 4, 138–183. Serhane, Rachida. 2003. Etude de la matrice {[labial], [dorsal]} en arabe. Doctoral thesis, University Paris 8. Yeou, Mohamed and Shinji Maeda. 1994. “Pharyngales et uvulaires arabes sont des approximantes: caractérisation acoustique”. 20ème Journées d’Études sur la Parole, Trégastel, Centre national d’étude des télécommunications, groupe francophone de la communication parlée, 409–414.

AUTOMATIC EXTRACTION OF PREPOSITIONS IN A CORPUS OF MODERN STANDARD ARABIC WRITTEN TEXTS Giuliano Lancioni

Introduction An experiment in Natural Language Processing [nlp] of Modern Standard Arabic [msa] written texts is presented, which aims to uniquely assign broad grammatical categories to words in a corpus by exploiting paradigmatic regularities in the lexicon, without any previous morphological or syntactic knowledge. As usual in Arabic, the texts included in the corpus are unvocalized (i.e., basically, only consonants and long vowels are marked). Of course, this makes automatic reading and understanding much more difffijicult. The approach taken capitalizes on a set of simple formal tests, inspired by strategies proposed in Classical Arabic linguistic thinking, which help to detect nominal or verbal forms according to the possibility to place some grammatical elements before them. The task is subdivided in fijive subtasks; the core subtask (and the most difffijicult one) is the identifijication of simple one-letter prepositions, which in Arabic script are written together with the following word. The idea behind the identifijication of simple inseparable prepositions is to determine if the rest of the graphical word (i.e., the aggregate of preposition and following nominal) can be regarded as a “candidate nominal”, while the whole form cannot.1 The system succeeds in assigning a grammatical category to a large subset (> 85%) of both data sets (training and testing, with a small degradation in performance in the latter case). Comparing the results with (quasi-)native speakers’ judgments shows a satisfactory level of accuracy (> 60% for recognized word types).

1  The category “nominal” includes both nouns and adjectives, which are not easily distinguishable in msa. For a discussion of categories in the Arabic linguistic tradition, see § 4.2 below.

196

giuliano lancioni 1. The Problem

In many nlp systems, the extraction of prepositions from written texts is a relatively easy task. In a language such as English, prepositions usually have the status of whole written words, and identifijication problems only come from homographs (for instance, after as a preposition and after as an adverb). On the other hand, most simple prepositions in msa are not graphically distinct from the fijirst word of the following noun phrase: they are monosyllabic morphemes attached, both in writing and in pronunciation, to the following word: for instance, bi- ‘at, with’, li- ‘to’, ka- ‘as, like’. In the writing system of Arabic, these morphemes appear as a single consonant grapheme attached to the next word. This feature of graphic (and phonological) coalescence is mirrored by the identical term used by the Arabic grammatical tradition to refer to ‘(consonant) grapheme’ and ‘preposition’: ḥarf, literally ‘segment’. This article is the report of an experiment of automatic extraction of prepositions from a signifijicantly large subset of a corpus of written msa texts taken from the Internet edition of Asharq Alawsat (aš-Šarq al-ʾAwsaṭ). The sample used in the experiment includes 32 daily issues of the newspaper, with about 1,100,000 words. The main strategy adopted in the experiment is based upon the paradigmatic selection of word types with and without a preposition. This strategy has already been successfully employed in the identifijication of nominals—which are not easily distinguished from verbs in an unvocalized Arabic text—and has the advantage of allowing a high degree of accuracy without any human intervention.2 Though simpler, the philosophy of this experiment is analogous to statistical part-of-speech taggers, like Tnt (Brants 2000) and mxpost (Ratnaparkhi 1996).

2  A parallel but diffferent strategy has revealed itself very successful in the identifijication of specialized lexicon: by exploiting the classifijication of articles in the diffferent sections of the newspaper (see Table 3 below), the ratio between occurrences of an item in a given section (say, economy or sports) and the total occurrences of the same item in the corpus allows one to mark items as (relatively) specialized or general. Thus, an item such as kaʾs ‘cup’, which appears 200 times in the sports section of the newspaper out of 208 total occurrences (or 96,2%), shows up as a specialized, sport-related word. This is an interesting result, since kaʾs is recorded in dictionaries as a general word, with a wider meaning of ‘holder for liquids, glass, cup’, but it has clearly become a predominantly sport-related word in msa newspaper texts.

automatic extraction of prepositions in msa

197

The extracted prepositions with their context have been checked with the judgments of native speakers of Arabic,3 in order to evaluate the accuracy of identifijication. The tests indicate a good degree of success for the extraction proposed strategy: an overall identifijication ratio of 63% compares reasonably well with context-free judgments of native speakers. The classifijication results could conceivably feed a higher-level system, based upon a less shallow, if not a truly deep, syntactic component.

2. The Arabic Writing System The writing system of Arabic usually represents consonantal graphemes only.4 While long vowels are marked by means of semi-consonants— called matres lectionis by Semitists,—short vowels are left unmarked in most texts and need to be added by the reader. Generally speaking, adding vowels of a unvocalized Arabic text is no trivial task: two or more choices may often look plausible or even equivalent, although syntactic and pragmatic constraints may reduce ambiguity in many cases.5 Diacritical marks are used to indicate short vowels and doubling of consonants: the use of these symbols varies across text types, but—except for the Koran and poetry—it rarely goes beyond marking some vowels in ambiguous contexts. The inherently defective nature of the Arabic script offfers no certainty in the absence of diacritical marks.6 Newspaper texts seldom, if ever, mark short vowels. The relative plausibility of text frames usually helps reduce false readings, but even skilled readers often have doubts about the exact rendition of a given text, for both the interference of spoken varieties—which often difffer dramatically from msa in the pronunciation of short vowels—and the presence of textual ambiguity. Thus, the computational burden is tremendous for a system that has not syntactic or encyclopedic information at its disposal. 3  The diglossic nature of the Arabic linguistic system implies that virtually every Arab is a native speaker of one of the varieties of spoken Arabic, so-called Arabic dialects. Nobody can be properly regarded as a native speaker of msa, since this form of the language is mostly written and learned through education. However, Arabic dialects and msa occupy distinct positions within the same linguistic continuum, and learned Arabs can be considered as quasi-native speakers of msa. 4  This section is primarily written for non-Arabists, in order to defijine and explain the most important concepts covered in the text. However, Arabists are recommended not to skip it entirely, but rather to read it cursorily, since it contains the precise defijinition of some of the key terms used elsewhere in this article. 5  For a concise yet accurate account of the Arabic writing system see Cowan (1964). 6  A special sign, called sukūn ‘quiescence’, is used to indicate specifijically that no vowel should be read.

198

giuliano lancioni

Some examples will illustrate how msa writes words in vocalized and unvocalized script. The main line in an Arabic text is a sequence of consonants optionally followed by (a combination of) reduplication mark or šadda, short vowels and sukūn (see Table 1).7 Table 1 IPA transcription Phonemic transcription Fully vocalized version Unvocalized version

[ˈkataba] kataba

[kaˈtabtu] katabtu

‘he wrote’

‘I wrote’

Unvocalized forms are inherently ambiguous: besides kataba, can be read kattaba ‘he made (sb.) write’, or even kutub ‘books’. As said above, contextual, syntactic and pragmatic constraints help choose the right form, but structurally ambiguous contexts are easy to build, and are actually found in texts. Long vowels are represented by a sequence . Since only three vowels are phonemically distinct in msa, the allowed sequences are for ā, for ī, for ū.8 In unvocalized texts, the semi-consonant only suggests the presence of long vowels, but alternative readings (e.g. diphthongs ay and aw) are always possible. A couple of ambiguous cases are shown in Table 2:9 Table 2 IPA transcription Phonemic transcription Fully vocalized version Unvocalized version

[kiˈta:b] [kutˈta:b] kitāb kuttāb

‘book’

‘writers’

7  Phonemic transcriptions, in italics, follow the (more or less) standard practice in Arabic studies, while graphemic transcriptions, in angle brackets, establish a one-to-one correspondence in transliterating written Arabic. In the latter case, sukūn is transcribed as and šadda as . 8  In traditional Arabic grammar, the mater lectionis for ā () is a virtual consonant without a phonetic value. 9  An interesting illustration of (some of) the problems involved in automatically treating unvocalized Arabic texts is in Al-Onaizan and Knight (2002).

automatic extraction of prepositions in msa

199

3. The Corpus As we said before, the corpus on which this study is based includes 3,208 articles from the online edition of the London-based, Saudi-owned, pan-Arabic newspaper Asharq Alawsat [aš-Šarq al-ʾAwsaṭ], from May 1 through June 2, 2005. The newspaper is organized in the following major sections:10 Table 3 section

[Arabic]

news economy local (Saudi) news sport opinions readers’ letters fijirst page diplomatic events (‘Arabs and non-Arabs’) fijirst page II science cultural forum publishing and books arts and television tastes real estate arts (‘the sixth string’) weekly digest media IT health business cars op-ed (‘idea and article’) tourism dialogue religion culture archeology and architecture

ʾaxbār iqtiṣād maḥalliyyāt (suʿūdiyya) riyāḍa ar-raʾy barīd al-qurrāʾ al-ʾūlā ʿArab wa-ʿAjam

10

number of articles 1019 585 303 232 216 207 120 106

ʾūlā 2 ʿulūm al-muntadā aṯ-ṯaqāfī ʿālam an-našr wal-kitāba funūn wa-tilifijizyūn ʾaḏwāq ʿaqārāt al-watar as-sādis ḥaṣād al-ʾusbūʿ ʾiʿlām tiqniyyat al-maʿlūmāt ṣiḥḥa ʿālam al-aʿmāl as-sayyārāt fijikra wa-maqāl as-siyāḥa ḥiwār dīn ṯaqāfa ʾāṯār wa-ʿimāra

 See fn. 2 for a possible way to exploit this broad thematic classifijication.

57 36 35 34 30 28 24 23 22 21 17 15 14 11 10 9 6 6 6 5

200

giuliano lancioni

The text was automatically segmented into words according to a very simple defijinition: a word is anything between spaces or between a space and a punctuation mark. This way, 1,236,508 words were selected. After excluding words composed of punctuation marks only (a frequent case, since punctuation marks in Arabic are often separated by a space for readability’s sake), the count came down to 1,156,043. In order to perform reliable tests, articles of the last edition (June 2, 2005) were removed from the training set and reserved for testing. The resulting testing set included 80,465 words.

4. Specification of the Subtasks The overall project is divided in several sequentially executed subtasks: 1. Marking unambiguously classifijiable particles written as independent words; 2. Identifying candidate nominals; 3. Distinguishing ambiguous preposition + nominal aggregates (candidates nominals) from non-aggregate words; 4. Distinguishing ambiguous conjunction + word aggregates; 5. Adding words after separately written prepositions; 6. Identifying candidate verbs. 4.1. Unambiguously Classifijiable Particles In Arabic, as in any language, prepositions and conjunctions form a relatively small, closed class. In order to ease the task for the following recognition steps, a list was prepared by selecting some of those written as independent words (i.e., words composed of at least 2 consonantal signs, including perhaps one mater lectionis, e.g. fī ). Separately written words that are ambiguous are excluded from this list, while the word Allāh ‘God’ is included, because it is always written with the article (somewhat like Italian Iddio from il + Dio, lit. ‘the God’) and is thus difffijicult to detect by standard strategies, yet it occurs frequently in the corpus; relative pronouns were also added, since in Arabic they are deictic elements preceded by the article, e.g. al-laḏī ‘which (singular masculine)’. An example of separately written words that are ambiguous—and cannot be included in the closed-class list—is min ‘from (mainly partitive)’,

automatic extraction of prepositions in msa

201

which in unvocalized script coincides with man ‘who (interrogative or relative)’, since both words are written . Cases of homophony such as mā ‘what’ (interrogative or relative pronoun) or ‘not’ (negation of the perfect tense), are also excluded. Only 22 word types (e.g., ḥattā ‘until, even’ and qabla ‘before’) have been included in the closed-class list at this stage, by selecting the most frequent such items. Their contribution to the overall recognition result is signifijicant, however, for both their high frequency (131,271 tokens in the training set, i.e. 11.79%) and their contribution to marking other words as nominals or verbs (see below). 4.2. Candidate Nominals Distinguishing nominals from other parts of discourse in unvocalized Arabic is no easy task. Case markings, which for singular and most plural nominals are represented by a single short vowel, are almost entirely discarded, while in most cases plural forms (so-called broken plurals) are internal and relatively idiosyncratic. The marker of indefijiniteness, tanwīn (-n sufffijixed to the case marker), is represented in the Arabic writing system by the doubling of the fijinal vowel, a feature usually left out in unvocalized script.11 Since the third person singular masculine perfect (which is the citation form of verbs in Arabic and—together with the third person singular feminine—the most frequent verb form) has an -a marker, also discarded in the unvocalized script, simple nominals and third person singular masculine perfect of the fijirst verb form, which both consist of the consonants of the root only, are indistinguishable outside the context. The difffijiculty of distinguishing word classes from one another led classical Arab grammarians to devise formal tests to determine the membership of a word. Ibn Jinnī opens his introductory treatise Kitāb al-lumaʿ fī n-naḥw (‘The book of splendor, on syntax’; Ibn Jinnī Lumaʿ: 1) by introducing an ingenious test to defijine membership in word classes:

11  In unvocalized script, the tanwīn is explicitly marked only in the accusative, where an orthographic ʾalif is required in most cases.

202

giuliano lancioni

A noun is what is matched by a particle governing the genitive case or what denotes a person. Examples of the [case of a] particle governing the genitive are min Zaydin ‘from Zayd’, ʾilā ʿAmrin ‘to ʿAmr’; examples of [the case of] denoting a person are, hāḏā rajulun ‘this is a man’, hāḏihi mraʾatun ‘this is a woman’.

Fal-ism mā ḥasuna fīhi ḥarf min ḥurūf al-jarr ʾaw kāna ʿibāratan ʿan šaxṣ. Fa-ḥarf al-jarr naḥwa qawlika min Zaydin wa-ʾilā ʿAmrin wa-kawnuhu ʿibāratan ʿan šaxṣ naḥwa qawlika hāḏā rajulun wa-hāḏihi mraʾatun.

A verb is what is matched by [the particle] qad ‘already, about to’ or what is a command [=imperative]. Examples of [the case with] qad are qad qāma ‘he stood up’, qad qaʿada ‘he sitted’ qād yaqūmu ‘he is about to stand up’, qad yaqʿudu ‘he is about to sit’. Examples of commands are qum ‘stand up’, iqʿud ‘sit’.

Wal-fijiʿl mā ḥasuna fīhi qad ʾaw kāna ʾamran. Fa-ʾamma qad fa-naḥwa qawlika qad qāma wa-qad qaʿada wa-qād yaqūmu, wa-qad yaqʿudu. Wa-kawnuhu ʾamran naḥwa qum wa-qʿud.

A particle is what is not matched by any of the markers of the noun or the verb, but only adds meaning to another word: hal [interrogative particle], bal ‘yet’. You could not say min hal, nor qad hal, and you could not make a command with it.

Wal-ḥarf mā lam taḥsun fīhi ʿalāma min ʿalāmāt al-ʾasmāʾ wa-lā ʿalāmāt al-ʾafʿāl wa-ʾinnamā jāʾa li-maʿnā fī ġayrihi naḥwa hal wa-bal. Wa-qad lā taqūlu min hal wa-lā qad hal wa-lā taʾmuru bihi.

‘Noun’ (ism) is an umbrella category, which includes also adjectives (comprising participles) and most adverbs; the traditional Arabic grammar has more fijine-grained subcategories, but it prefers to consider them as members of a single category because of opacity of distinction. In most cases, an adjective or an adverb is morphologically indistinguishable from a noun, and also syntactic disambiguation is not always clear or even possible—two or more readings of one and the same sentence can be perfectly plausible. Since our task is to determine if a form could be analyzed as a preposition + noun aggregate, Ibn Jinnī’s nominality test will be replaced by a similar one, based upon the defijinition that a noun is what can be preceded by the article al-. Designation of the defijinite article al- as the preferred marker to test nominality is a purposeful choice: since case and indefijiniteness markers are mostly missing in unvocalized script and most plurals can only be identifijied lexically, the presence or absence of the article is the most rel-

automatic extraction of prepositions in msa

203

evant marker of nominality. Since the article is written attached to the following nominal and its allophones—produced by elision of a wordinitial a after a word that ends in a vowel, addition of a glottal stop at the beginning of a phonological clause, assimilation of l to a following ‘sun letter’ (i.e. a coronal consonant)—are not recorded in unvocalized script, the presence or absence of the article can be recognized as the most efffective nominality test. Subtask 1 capitalizes on this feature by considering as “candidate nominals” words that occur in the corpus both with and without the article. Since there exist nominals starting with the digraph (e.g. iltiqāʾ ‘meeting’, written ), overrecognition is prevented through the further requirement that a candidate nominal never occurs with a duplicated digraph . Formally, a candidate nominal can be defijined as follows: Defijinition 1 A corpus C is a list of n forms [C1, C2, . . . Cn] A corpus-set CS is a set of m forms {CS1, CS2, . . . CSn} without duplication, with frequency { FS1, FS2, . . . FSn} A form α is a “candidate nominal” (α ∈ CN) in a corpus C with corpus-set CS if: a) α ∈ CS; b) ( + α) ∈ CS; c) ( + α) ∉ CS. Where ‘+’ is the standard string concatenation operator. Contrary to what may seem, nominals graphically beginning with , such as iltiqāʾ above, are not excluded by Defijinition 1, since clause a) does not require that α does not itself begin with . The only possible underrecognition occurs in the relatively rare cases where both α and + α do exist, so that Defijinition 1 would improperly exclude α (but not + α). Both forms α and + α are marked as nominals; with the total amounting to 19,774 word types.

204

giuliano lancioni 4.3. Ambiguous Preposition + Nominal Aggregates

The main subtask in the experiment aims to detect single-letter prepositions that are attached to the following nominal. This is an ambiguous case, since the three prepositions under consideration—bi- ‘by (instrumental or locative)’, ka- ‘as (comparative)’, li- ‘to (benefactive or possessive)’— could be interpreted as the fijirst consonant of a single word. The strategy adopted in this subtask is to consider a word as an aggregate if the word without the fijirst consonant is a candidate nominal, but the whole word is not. Formally: Defijinition 2 A form α is an “aggregate preposition + candidate nominal” in a corpus C with corpus-set CS if: a) b) c) d) e)

α ∈ CS; α ∉ CN; β ∈ NP (set of “non-separable prepositions”); α = β + γ; γ ∈ CN.

Defijinition 2 above must be slightly adapted in the case of li-, since the orthographic rules of Arabic script prescribe the graphic elision of the initial after the aggregation. Therefore, condition d) becomes: d) if β = and left(γ,2) = then α = β + mid(γ,2) else α = β + γ;

where “left(a,b)” is an operator which returns the fijirst b characters from a and “mid(a,b)” is an operator which returns the substring of a beginning with the b-th character. 4.4. Ambiguous Single-Letter Conjunctions Together with ambiguous prepositions, there are a couple of frequent single-letter conjunctions (wa- and fa-, both translatable as ‘and’, the latter

automatic extraction of prepositions in msa

205

marking a more distinct clause break) which can cause similar detection problems. The main diffference is that the resulting word in these cases is not necessarily a nominal—on the contrary, it can be everything. The scope of this subtask was limited to detecting Conjunction + word aggregates, without trying to analyze the isolated word any further than this. A recursive application of the subtask could easily obtain the identifijication of some more nominal, but such a move was not considered useful enough to justify the additional computational burden required. For testing sake, only the conjunction wa- was detected, and the resulting aggregate was labeled as “wa- + word”. The formal specifijication of the task is as follows: Defijinition 3 A form α is a “conjunction + word” aggregate if: a) b) c) d) e)

α ∈ CS; β ∈ NC (set of “non-separable conjunctions”); γ ∈ CS; α = β + γ; (β + α) ∉ CN.12

4.5. Words that Follow Prepositions Written as Separate Words If a word follows one of the prepositions written as separate words that were identifijied in the subtask 1 above, it can be regarded as a nominal. The few cases in which such a preposition is followed by a conjunction that introduces a dependent clauses (e.g., fī ʾan ‘in that’) were fijiltered out in the same subtask, since such conjunctions have already been marked as single-class words.13

12  The defijinition is slightly simplifijied, since what should be checked is the presence of an accumulation of co-ordinate conjunctions, which is not allowed (i.e., not only wa-waor fa-fa, but also wa-fa- and fa-wa-). In order to avoid overgeneration problems, only the sequence wa-wa- was included in the simulation. 13  Some of these conjunctions (e.g., ʾanna ‘that’) must be followed by a noun in the accusative or by a sufffijixed accusative pronoun. Since these pronouns are written together with the conjunction, they would not be fijiltered out by the strategy mentioned in the text. Two solutions are possible: adding the conjunction + sufffijix pronoun aggregate to the closed-class word list or adding a specifijic subtask for the detection of the sufffijix pronoun. The latter solution would perform a remarkable achievement, since the same endings are sufffijixed to verbs and nouns (in the latter case, to mark genitive case in a possessive

206

giuliano lancioni

The detection strategy can be formalized as follows: Defijinition 4 A form (α = Cl ) ∈ CN if (β = Cl-1 ) ∈ CC (the set of “closed-class words”). 11,934 forms in the corpus are recognized as nominals by this strategy (the most frequent being al-miʾa ‘one hundred’). Of course, many nominals had already been recognized by previous subtasks, so the independent contribution of this strategy to the overall recognition task is clearly smaller. 4.6. A Test of Verbality Ibn Jinnī’s testing strategy suggests a second test along the lines developed in Defijinition 1, albeit simpler: a word is recognized as a verb if it is preceded by the particle qad, written . This task is easier because qad, unlike the article al- and many simple prepositions, is written as a separate word in standard Arabic orthography. The idea of extracting information from co-occurrence and ordering restrictions is not new: Widdows & Dorow (2005) suggest a way to fijind idioms by selecting patterns that show ordering asymmetries without a syntactical trigger (e.g. hue and cry does occur in a corpus, while cry and hue does not: therefore, we can conclude that the former is an idiom). An approach closer to mine is taken by Sharofff (2004), who extracts prepositional idioms from untagged Russian corpora by simple statistical algorithms. The defijinition is similar to Defijinition 4 above: Defijinition 5 A form (α = Cl ) ∈ CV (the set of “candidate verbs”) if (β = Cl-1 ) = .

construction). Because of intrinsic difffijiculties in disambiguation—some of these pronominal forms are represented by a single consonant—I chose the fijirst alternative, even if it can be regarded as unduly ad hoc solution.

automatic extraction of prepositions in msa

207

5. Testing the Results The subtasks are applied sequentially in the order specifijied, so that a subtask starts only if the previous one failed to recognize the current word: this is a rather strong limitation, since the parallel application of subtasks could achieve a better performance. Another huge limitation of the current version of the model is the impossibility to attach two or more labels to the same item, so that intrinsically ambiguous words—e.g., words that be interpreted as either a nominal or a verb according to context—obtain a unique, perhaps wrong interpretation. Notwithstanding these limitations, the system achieves satisfactory results. The overall classifijication results in the training set are the following: Table 4 Assigned Class

Word Types

%

Tokens

Unrecognized bi- + Noun ka- + Noun li- + Noun Noun Verb wa- + Word

49,873 4,133 582 2,154 19,774 7,103 21,205

47.58% 3.94% 0.56% 2.05% 18.86% 6.78% 20.23%

168,878 24,572 5,665 15,431 539,801 311,104 90,592

Total

104,824

% 14.61% 2.13% 0.49% 1.33% 46.69% 26.91% 7.84%

1,156,043

The most important fijigure in Table 4 is the percentage of unrecognized tokens: 14.61% is a remarkably low ratio for a system that uses statistical, mostly paradigmatic, information only, without previously tagging or preparing the corpus. The number of unrecognized word types is higher (49,873, i.e., 47.58%), but it remains low enough, since—as usual in large raw corpora—most word types occur very rarely. In the corpus at hand, 51,406 words are singletons, and 16,189 occur only twice; it is a great achievement that the system manages to classify even words occurring once in the whole corpus.

208

giuliano lancioni

With the testing set, we have the following data (Table 5): Table 5 Assigned Class

Word Types

%

Tokens

%

Unrecognized bi- + Noun ka- + Noun li- + Noun Noun Verb wa- + Word

9,192 665 43 390 7,128 2,068 2,459

41.89% 3.03% 0.20% 1.78% 32.48% 9.42% 11.21%

14,727 1,525 366 933 35,974 21,510 5,430

18.30% 1.90% 0.45% 1.16% 44.71% 26.73% 6.75%

Total

21,945

80,465

The system still performs well on the testing set. The ratio of unrecognized tokens increases only to 18.30%, and the recognition task yields in general only a slight degradation of the performance even while working on texts unseen before.14 That all the texts come from the same source clearly makes the task easier. Applying the system to data from other sources could yield worse results, yet—in the lack of reliable information—we might argue that texts from similar sources (e.g., other newspaper articles) could be analyzed with a reasonable degree of success. Cross-assignment of membership to the same grammatical categories by (quasi-)native speakers of msa for a subset (ca. 10,000 words) of the testing set shows a fair degree of accuracy of the system: in cases where the model provides a grammatical category assignment, the decision of the system agrees with the speakers in 63% of the cases.15

14  It is interesting to note that 13,756 words (or 62,68%) in the test set are singletons. Of course, most of these words have a higher frequency in the training set, but this result is nevertheless encouraging. 15  Even if these fijigures might appear unsatisfactory to some readers, they are quite remarkable if we take into consideration the designed limits of the experiment: the latter aims to tag only the subset of words that can unambiguously be labeled as verbs or as nouns. Of course, this task does not exhaust the analysis of the text, but does help further analyzing it. The same proviso applies to the judgments of native speakers. For readers that are not familiar with the methods and the goals of natural language processing, a very readable introduction is in Bird, Klein & Loper (2009); tagging is especially discussed in Chapter 5 (“Categorizing and Tagging Words”).

automatic extraction of prepositions in msa

209

A fijirst screening of the cross-assignments shows that most errors are caused by under- and, to a lesser degree, overrecognition. Underrecognition results from the failure of the system to fijind minimal couples in case of relatively rare words (e.g., only one form, with or without the article, can be found, so Defijinition 1 fails to apply). Overrecognition results mostly from spurious minimal couples caused by misprints (see the discussion below). These problems are linked to the issue of specifijicity vs robustness that is intrinsic in automatic recognition systems.16

6. Conclusions and Further Work The procedure described in this article succeeds in tagging a large corpus of msa unvocalized written texts. Recognition results are reasonably good in both the training and the testing sets. Moreover, the tagging of the recognized words shows a fair degree of consistency with the judgments of (quasi-)native speakers. These promising results could be optimized in several diffferent ways. A fijirst possibility is to extend and refijine the coverage of the system by adding more stringent rules to reduce the impact of under- and overrecognition. An idea worth developing is introducing a fuzzy categorization system. For instance, the words categorized as nominals by Defijinition 1 vary widely in the ratio of tokens with and without the article, from 831:1 to 1:39,109. Clearly, the assignment of a word’s membership to nominals is clearer when the ratio is close to 1:1 and increasingly doubtful as it increases or decreases. In extreme cases, the single occurrence of one of the two variants is probably due to a misprint, which in the current system assigns the item to a wrong category. A simple solution would be to set an arbitrary threshold to the ratio in Defijinition 1. A more interesting possibility would be to introduce an “index of nominality”, for instance:

(

)

max(a,b) min(a,b)

–1

16  See Zhang & Kordoni (2006) for a discussion of problems related to over- and undergeneration of natural language processing systems.

210

giuliano lancioni

Where a and b are, respectively, the number of tokens with and without the article, and “max” and “min” are functions which return, respectively, the greater and the lesser of two numbers. This way, some word types could be classifijied as fully nominal, some other words as, say, 80% nominal and 20% verbal. The right label to assign to a word would then be determined by contextual, coincident restrictions. Another research idea would be to feed higher, deeper syntactic analyzers with the shallow information given by the present system. Higher-level systems may thus benefijit from the simple yet wide coverage of the lexicon assured by statistical methods to help alleviate the well-known problem of parsing failures due to items missing from the lexicon.17 In particular, information about simple prepositions could help determine the valency frameworks of verbs, a task that is far from easy in a VSO language such as msa.18

References Al-Onaizan, Yaser and Kevin Knight. 2002. “Machine Transliteration of Names in Arabic Text”. Proceedings of ACL Workshop on Computational Approaches to Semitic Languages. Philadelphia: Association for Computational Linguistics. Baldwin, Timothy, Emily M. Bender, Dan Flickinger, Ara Kim and Stephan Oepen. 2004. “Road-testing the English Resource Grammar over the British National Corpus”. Proceedings of the Fourth International Conference on Language Resources and Evaluation (LREC 2004). Lisbon, Portugal: LREC. Bird, Steven, Ewan Klein and Edward Loper. 2009. Natural Language Processing with Python. Sebastopol, Ca.: O’Reilly. Brants, Thorsten. 2000. “TnT—a statistical part-of-speech tagger”. Proceedings of the Sixth Conference on Applied Natural Language Processing ANLP-2000. Seattle, WA: ANLP. Cowan, David. 1964. Modern Literary Arabic. Cambridge: Cambridge University Press. Diab, Mona, Kadri Hacioglu and Daniel Jurafsky. 2004. “Automatic Tagging of Arabic Text: From Raw Text to Base Phrase Chunks”. Proceedings of HLT-NAACL 2004, Boston. Fujita, Sanae and Francis Bond. 2002. “A Method of Adding New Entries to a Valency Dictionary by Exploiting Existing Lexical Resources”. Proceedings of the 9th International Conference on Theoretical and Methodological Issues in Machine Translation (TMI2002). Keihanna, Japan, 45–52. Habash, Nizar and Owen Rambow. 2005. “Arabic Tokenization, Part-of-Speech Tagging and Morphological Disambiguation in One Fell Swoop”. Proceedings of the 43rd Annual

17

 See Baldwin et al. (2004) for discussion.  See Fujita & Bond (2002) for an application aimed to add entries to a valency dictionary by exploiting lexical resources. As to alternative tagging and tokenization strategies, see Diab, Hacioglu & Jurafsky (2004), Habash & Rambow (2005), Stetina & Nagao (1997). 18

automatic extraction of prepositions in msa

211

Meeting of the Association for Computational Linguistics (ACL’05). Ann Arbor, Michigan: Association for Computational Linguistics, 573–580. Ibn Jinnī. Lumaʿ. Kitab al-Lumaʿ fī-n-Naḥw, manuel de grammaire arabe, edité et annoté par Hadi M. Kechrida. Uppsala: Uppsala Universitet. 1976. Ratnaparkhi, Adwait. 1996. “A maximum entropy model for part-of-speech tagging”. Proceedings of the Conference on Empirical Methods in Natural Language Processing, ed. Eric Brill and Kenneth Church. Somerset, Philadelphia: University of Pennsylvania, 133–142. Sharofff, Serge. 2004. “What is at stake: a case study of Russian expressions starting with a preposition”. Second ACL Workshop on Multiword Expressions: Integrating Processing (Barcelona, Spain, July 2004), ed. John Goldsmith and Richard Wicentowski. East Stroudsburg, PA: Association for Computational Linguistics, 17–23. Stetina, Jiri and Makoto Nagao. 1997. “Corpus based PP attachment ambiguity resolution with a semantic dictionary”. Proc. of the 5th Annual Workshop on Very Large Corpora, ed. Joe Zhou and Kenneth Church. Tsinghua and Hong Kong, 66–80. Widdows, Dominic and Beate Dorow. 2005. “Automatic Extraction of Idioms using Graph Analysis and Asymmetric Lexicosyntactic Patterns”. Proceedings of the ACL-SIGLEX 2005 Workshop on Deep Lexical Acquisition, ed. Timothy Baldwin, Anna Korhonen, and Aline Villavicencio. Ann Arbor: University of Michigan, 48–56. Zhang, Yi and Valia Kordoni. 2006. “Automated Deep Lexical Acquisition for Robust Open Texts Processing”. Proceedings of LREC–2006, 275–280.

SECTION IV

THE ARABIC WORD IN CONTACT

HEAVY AND LIGHT BORROWING OF ARABIC VERBS Kees Versteegh

The presence of a Latin/Greek framework, not only in the lexicon of the Romance languages, but in that of all European languages, has created a common semantic structure in these languages. This common structure is independent of the genealogical position of the languages involved. Basque and Hungarian, for instance, are just as much indebted to this framework as the Romance or the Germanic languages are. A special characteristic of this dependence is the layering of the borrowings. Most European languages exhibit a double layer of loanwords from Latin/Greek. In English, for instance, old loanwords like cellar (< Latin cellarium) coexist with neologisms like telescope. In Romance linguistics, this layering has become known as the diffference between formation populaire and formation savante, as in French cheval and cavalerie. Latin is not the only global language that has exerted such an influence in a vast variety of languages. In this paper, I shall be dealing with the example of Arabic and its influence in languages all over Africa and Asia. Surprisingly little attention has been devoted to Arabic loanwords in the Islamic languages, either by scholars of Arabic, or by specialists in other Islamic languages. One has to search extensively to fijind studies on Arabic loanwords in Ottoman Turkish, in Farsi, in Indonesian, in Urdu, to mention but the most important ones. The importance of the study of loanwords should not be underestimated. The structure and layering of loanwords may help us elucidate the history of relations between speakers of Arabic and those of other languages, the spread of Islam, and sometimes even the history of Arabic itself. As I have attempted to show elsewhere (Versteegh 1999), the reflexes of Arabic phonemes in Arabic loanwords in Spanish and Malay tell us something, not only about the process of Arabization, but also about the phonemic inventory of early Arabic (cf. Brown 2007). Arabic loanwords in other languages exhibit a layering that is quite similar to that of Latin loanwords in European languages. In many African languages, there is an older layer of Arabic loanwords borrowed at an early stage of contact. Very often, these are no longer recognizable as

216

kees versteegh

loanwords from Arabic, since they have been fully integrated in the lexicon, both phonologically and morphologically. An example is the Fulfulde word hinjiiru ‘pig’, whose plural kinjiiji efffectively hides its Arabic origin (< xinzīr). At a later stage, when Islamic education reached these communities, and more people became acquainted with Arabic as a learned language, some of the original loanwords were re-Arabized, a process that might be called ‘denativization’ (Boumans 2007: 308). Nowadays, Fulfuldespeaking Islamic scholars say zamaan ‘time’, for instance, rather than using the old loanword jamanu, jamanuuru. In some cases, re-borrowing is a better description of what took place. The most obvious example is that of Indonesian perlu ‘ought to’, which contrasts with modern fardu ‘religious duty’, both from Arabic farḍ. In spite of their thorough knowledge of Arabic, not even Islamic scholars in Indonesia are aware of the Arabic origin of the former, whereas the latter is commonly recognized as an Arabic word. A similar layering may be assumed to exist in all Islamic languages, even in Persian and Turkish (in Persian, the complicated issue of the -e and -at reflexes of the Arabic feminine ending may be an example of layering, cf. Perry 1991). It seems to be generally true of all contact processes that the overwhelming majority of loans is nominal in nature. In the case of Arabic, the structure of the triradical root makes it even more difffijicult to borrow verbs from this language than from others. Yet, some Islamic languages exhibit verbal forms that are based on borrowed elements from Arabic. Basically, two strategies appear to be involved in this borrowing process. The fijirst strategy consists in borrowing an Arabic verbal form and integrating it in the structure of the receiving language. This is what has taken place in Swahili, and to some extent in Hausa, e.g. Hausa sàllamàa ‘to greet’ (< Arabic sallama). The second strategy consists in using a light verb or dummy verb, usually one meaning ‘to do’, and combining this with a nominal form borrowed from Arabic, so that a compound or complex verb emerges. This is what has taken place in Persian, Turkish, and Urdu, e.g. Persian taʿlim kardan ‘to teach’ (< Arabic taʿlīm ‘teaching’ + Persian kardan ‘to do’). According to Muysken’s (2000) typology, the two strategies represent two diffferent types of code-mixing (a third type, congruent lexicalization is less relevant in this connection). The fijirst strategy, using a foreign verb directly, with or without special afffijixes, is a form of insertion. The second strategy, using a light verb—or as Muysken calls it, a helping verb— belongs to an alternation mode of code-mixing, for which he reserves the term code-switching.

heavy and light borrowing of arabic verbs

217

Gardner-Chloros & Edwards (2004) follow Muysken’s distinction between insertion and alternation as diffferent modes of bilingual speech behavior. They emphasize that these diffferent modes belong to diffferent social contexts, and exhibit diffferent patterns of switching and also diffferent constraints (Gardner-Chloros & Edwards 2004: 1444–1445): i. in insertion, foreign material is inserted within a matrix belonging to one language that is primarily activated; this type of code-mixing is more related to borrowing and takes place when there is a clear asymmetry between the speakers’ profijiciency in the two languages ii. in alternation, the elements before and after the switch are not structurally related; in alternation two diffferent languages may be said to be activated alternately; this happens in stable bilingual communities

Gardner-Chloros & Edwards (2004: 1445) state that “[i]nter-generational language shift may be reflected in a change in the direction of the insertion”. This points to a diachronic development in the borrowing procedure. Within the alternation mode of borrowing verbs, Muysken (2000: 211) suggests a diachronic development as well. He believes that there are three stages in the formation of verbal compounds with a light verb: at fijirst, foreign nouns are inserted in constructions with helping verbs, then the same is done with nominalized verbal forms or gerunds, and fijinally, verbs are adjoined to the helping verbs. In this last stage, the government relation between the noun or the nominalized verbal form and the helping verb is cut offf and the relation becomes one of modifijication. The helping verb no longer governs the adjoined word. The question to be asked here is: what afffects the choice between the two strategies? Is the choice connected to the structure of the language(s) involved? Are the two strategies steps in a gradual development? Which factors determines the morphological form of the borrowed element? In improvised situations of linguistic contact with speakers of Arabic, verbs are apparently borrowed freely, presumably because the speakers are exposed to the simplest speech acts, such as wishes and commands, for which only one verbal form sufffijices, usually the imperative. This is indeed the form that seems to be at the basis of the verbal system in Arabic pidgins and creoles, such as Ki-Nubi (Wellens 2005) or Juba Arabic; Ki-Nubi verbs like rua ‘to go’ probably derive from an Arabic imperative form, in this case rūḥ ‘go!’. Later, when this primitive form of communication is expanded, more fijine-grained distinctions become important in linguistic interactions, including narrative strategies. It becomes important, for instance, to make a distinction between telling someone that you want something done and telling them that something has been done. When

218

kees versteegh Table 1: Arabic Loanverbs in Western Neo-Aramaic

Arabic loan

Western Neo-Aramaic

III. VI. VII. VIII. X.

šāraṭ tarāfaḳ infajar iftaham istaqbal

stem stem stem stem stem

> šōreṭ > črōfeḳ > inəfžar > if əčham > sčaḳbel

‘to bet’ ‘to accompany’ ‘to explode’ ‘to understand’ ‘to accept’

the target language is still available, speakers use diffferent strategies to reanalyze relevant forms from the language they are exposed to in order to develop linguistic markers for such distinctions. At this stage, structural constraints may kick in: it is easier to borrow verbs from some languages than from others, and conversely, it is easier to integrate foreign verbs in some languages than in others. In the case of languages with the same non-concatenative morphological structure as Arabic, for instance Neo-Aramaic, Arabic verbs with diffferent verbal patterns can be integrated with relative ease, as in the examples in Table 1, taken from Arnold (2007). In Berber, like in Neo-Aramaic, integration of Arabic verbs was relatively uncomplicated, because the language has a similar triradical structure, although not completely identical with the Arabic system. Not many Arabic verbs exist in the language but those that have been borrowed exhibit the same morphological pattern as Berber verbs, e.g. fṛəḥ (< Arabic fariḥa) ‘to rejoice’, which has the same structure as Berber mġər ‘to grow’ (Aguadé & Behnstedt 2006). The integration of these verbs was no doubt facilitated by the fact that in many dialects of Maghrebi Arabic, all verbs have the same perfect pattern, C1C2əC3, e.g., ktəb ‘to write’, unlike verbs in Eastern Arabic dialects, which usually have one of two or three patterns. It may be added here that this verbal pattern in Maghrebi Arabic may well have originated as a result of a stress shift under Berber influence. Some loan verbs in Berber may be derived from an Arabic imperative, e.g. dub (< Arabic ḏāba/yaḏūbu) ‘to melt’. In languages without a non-concatenative structure, borrowing verbs from Arabic is a more complicated matter, although apparently not entirely impossible. In the case of Swahili, Tucker (1946–1947) assumes that Swahili verbs borrowed from Arabic are based on the Arabic imperative, which would imply that the borrowing context for Swahili-speakers was similar to those of Arabic pidgins and creoles. An important piece of evidence for this thesis is the form in which Arabic weak verbs were

heavy and light borrowing of arabic verbs

219

borrowed in Swahili (Schwarz 2004: 74): dhuku (< Arabic ḏāqa, imperative ḏuq) ‘to taste’; duru (< Arabic dāra, imperative dur) ‘to go round’; ghibu (< Arabic ġāba, imperative ġib) ‘to get lost’. In the strong verbs, the situation is less clear, but here too the Swahili verb often contains the imperfect/imperative vowel of the Arabic verb, rather than the perfect vowel, e.g. hutubu (< Arabic xaṭaba, imperative uxṭub) ‘to preach’; miliki (< Arabic malaka, imperative imlik) ‘to take possession of, to possess’; hukumu (< Arabic ḥakama, imperative uḥkum) ‘to sentence’. The correlation is far from exact, however, and for some of the verbs an imperatival origin is not very plausible for semantic reasons. The situation is complicated by the fact that at a later stage, Swahili also borrowed from Arabic through written transmission in treatises by Islamic scholars who had studied the language. In Arabic makeshift pidgins, such as Gulf Pidgin Arabic, both prefijixed and non-prefijixed forms seem to occur, the latter being more frequent than the former (Naess 2008:83–93). I intend to deal elsewhere with the issue of the imperatival origin of verbs in Arabic pidgins. In West African languages, such as Fulfulde, Arabic verbs were borrowed freely, too, and integrated in Fulfulde morphology (Theil 2007); it is not entirely clear, however, which form of the Arabic verb served as the basis of derivation. Verbs like juuroo (< Arabic zāra/yazūru) ‘to visit’ and faama (< Arabic fahima/yafhamu) could be derived from the imperfect or from the imperative of these verbs which share the same vowel as against the vowel of the perfect. According to Theil (2007), it is more likely that the underlying form is the Arabic imperfect, but in view of the parallel examples in Swahili (and Ki-Nubi), it is perhaps preferable to regard them as derived from the Arabic imperative. Provisionally, we may conclude that during initial contact between speakers of Arabic and speakers of other languages, verbs were perfectly borrowable, often in the imperative form. In the case of Arabic pidgins/ creoles, such forms remained without inflection, as in Ki-Nubi, but in those cases where the indigenous language remained in use, the borrowed verbs were integrated in the morphological system of that language, as in Fulfulde. These are all instances of insertion, the fijirst strategy of borrowing verbs mentioned above. The second strategy consists in the use of a light verb (helping verb, dummy verb), usually a verb meaning ‘to do’. This construction will be called here the ‘DO-construction’. There may be more than one helping verb in the borrowing language, in which case one of them means ‘to do’ and is used with agentive verbs, whereas the other usually means ‘to be’ and is used with stative verbs. In a language like English,

220

kees versteegh

the DO-construction, which is used in negation, emphasis, and interrogation, shifts the inflectional burden onto a dummy verb, and leaves the original verb without inflection. According to Muysken (2000), however, the two constructions represent diffferent phenomena, because in English, there is a government relation between the helping verb and the adjoined verbal element, whereas in some of the bilingual contexts in which the construction occurs, this is not the case. In Sarnami/Dutch code-mixing, the fijirst element of the verbal compound is very often a form that would be a fijinite verb in Dutch, e.g. leg uit kare ‘to explain’ (< Dutch uitleggen [infijinitive]), verplaats kare ‘to transfer’ (< Dutch verplaatsen [infijinitive]), but also uitleg kare (< Dutch uitleg ‘explanation’) (Muysken 2000: 201– 202). There is no government relation between the foreign verb and the helping verb, but one of adjunction, the foreign verb modifying the helping verb. If the relation between the two would be one of government, the resulting complex verb would be intransitive because the transitivity of the helping verb would be absorbed by the governed foreign verb. In fact, the complex verb can have its own direct object (Muysken 2000: 205). If the DO-verb were an inflection carrier, it would always have the subcategorization pattern of the foreign verb. Yet, Muysken acknowledges that in some instances the DO-verb does have the function of an inflection carrier, namely when it functions as the auxiliary of an inserted infijinitive. This function of the light verb as carrier of inflection is emphasized by Matras (2002: 135), who analyzes the structure of loan verbs in Romani: “For active bilinguals who have full access to the fijinite-verb inflection systems of both languages, the switch at the point at which the predication is initiated creates ambiguity as to the matrix language [. . .] of the clause, disturbing the overall grammatical plan of the utterance”. Shifting the role of predication initiator to a dummy verb (native carrier) that carries all inflectional endings, solves this problem. In the case of borrowing from Arabic, the DO-construction is found in Islamic languages all over Central, South, and Southeast Asia, and in SubSaharan Africa. In modern Iran, Persian is both the offfijicial language and the mother tongue of the majority of the inhabitants, while Arabic is used exclusively in a religious context, and even there its function is restricted, and the average believer can neither speak nor read the language. The Persian lexicon is replete with Arabic loanwords, but there seem to be no examples at all of Arabic verbal forms, neither with, nor without a light verb. All verbal expressions with Arabic roots are based on Arabic nominals, e.g. mokātabe kardan ‘to correspond’ (< Arabic mukātaba ‘correspondence’), harakat kardan ‘to move’ (< Arabic ḥaraka ‘movement’).

heavy and light borrowing of arabic verbs

221

A similar situation obtains in other Islamic languages that received most of their Arabic heritage through Persian, such as Turkish (and other Turkic languages), e.g. refakat etmek ‘to accompany’ (< Arabic rafāqa ‘companionship’), tebdil etmek ‘to change’ (< Arabic tabdīl ‘change, alteration’), and Hindustani/Urdu, e.g. tamir karnā ‘to build [monument]’ (< Arabic taʿmīr ‘building [verbal noun]’), qabūl karnā ‘to accept’ (< Arabic qabūl ‘acceptance’) (Désoulières 2008). Arabic verbal forms are never integrated in the Dravidian languages (Telugu, Malayalam and Tamil), either. These languages acted as intermediaries for the transmission of Islam (and Arabic) to Southeast Asia; contact with speakers of Arabic took place through trading in the Indian Ocean, but this contact did not lead to the introduction of Arabic verbal forms. In Tamil, for instance, all verbal expressions with Arabic material are combinations of borrowed nouns with native verbs, e.g. hālir ā- ‘to appear’ (< Arabic ḥāḍir ‘present’) (Tschacher 2008). In Indonesian/Malay, there are no traces of the DO-construction and no Arabic verbs seem ever to have been borrowed directly. There may be an older layer of loanwords (cf. Versteegh 2003), which was introduced before the spreading of Islam by Persian-speaking missionaries, but this layer does not contain any loan verbs, either. Indonesian verbs containing Arabic material are always derived from nominal forms, e.g., menghukum-kan ‘to pronounce judgement’ from hukum ‘judgment’ (< Arabic ḥukm). There is only one Arabic loanword that may appear as a simple verb without prefijixes, pikir ‘to think’, but this word is probably derived from the Arabic noun fijikra ‘thought’. Because of the morphological structure of Indonesian/Malay, it was very easy to derive verbs from any borrowed noun, just like indigenous nouns can always be transformed into verbs by prefijixes and sufffijixes. In some African languages that were or are in contact with Arabic, the DO-construction is found as well. In Nubian, for instance, all verbal expressions with foreign material are formed with a light verb (Jakobi 2007). In Dongolawi Nubian, the light verb does not mean ‘to do’, but ‘to say’, -e, -ee, e.g., harb-ee (< Arabic xarab) ‘to destroy’, gaabil-ee (< Arabic gābal) ‘to meet. In view of the vocalic melody of the Arabic word, the underlying Arabic form may have been the imperative, so that this might very well be interpreted as a direct borrowing of a verbal form, rather than a DO-construction. In fact, the light verb ‘to say’ may be interpreted as a quotative form: ‘to perform the action they call [in Arabic] xarab’. Backus (1996: 224–225) points out that in some languages the two strategies—direct borrowing of a verbal form or using a DO-construction— are used for borrowing material from diffferent languages. In the case of

222

kees versteegh

Spanish, for instance, English verbs may be incorporated both morphologically and with the DO-construction. In some cases, the choice seems to depend on the source language. In Tamil, for instance, English verbs are incorporated with the DO-construction, as are those from Arabic, whereas verbs from Kannada, another Dravidian language, are integrated morphologically. In the case of Persian, one strategy, the DO-construction, is used for the incorporation of foreign verbs, both in the old loanwords from Arabic, and in a contemporary code-switching context from Swedish. In Lotfabbadi’s corpus of Persian/Swedish code-switching, there is not a single instance of a Persian verb with Swedish inflection. This is to be expected because, as she explains (2002: 160), the Swedish inflection would leave out essential information for a Persian verb (subject-verb agreement and aspect). The absence of Swedish verbs with Persian inflection from the corpus is more puzzling, however, since Swedish nouns and adjectives freely take Persian sufffijixes (plural, case endings, comparative etc.). Lotfabbadi (2002: 162) adduces a number of explanations, none of them conclusive. But the crucial point here is that almost all her examples combine Swedish bare infijinitives with the light verb (2002: 109), whereas Persian complex verbs with Arabic elements are always combinations of verbal nouns with kardan. Sometimes, as Gardner-Chloros & Edwards (2004: 1446–1447) point out, the same pairs of code-mixing languages use diffferent strategies, depending on the context. German/English code-mixing in the United Kingdom, for instance, is diffferent from that in Australia, even to the point where German word order constraints apply to the former, but not to the latter. This means that the sociolinguistic circumstances determine to a far higher degree than the linguistic constraints the nature of the processes involved. Arabic itself is a case in point: French verbs in Moroccan Arabic in Morocco are fully integrated, whereas speakers of Moroccan Arabic in the Netherlands use a light verb when borrowing Dutch verbal concepts (see Boumans 1998, 2007; Versteegh 2009). The case of Arabic verbs in Hausa may help to single out the factors that determine the choice between the two alternative strategies. Language contacts between speakers of Hausa and speakers of Arabic took place in two widely difffering contexts. The resulting borrowing pattern is quite diffferent, too. In the Western Sudan, Hausa is the lingua franca for a large number of speakers, for whom contact with Arabic is limited to religious contexts and the recitation of the Qurʾān. In Western Sudanic Hausa, loan verbs are not entirely unknown; these must have been adopted before the

heavy and light borrowing of arabic verbs

223

emigration of the Hausa speakers to the Eastern Sudan (e.g. sawwataa ‘to vote’, siddaa ‘to pay the dowry’, etc.), but they constitute a minority in the loanword lexicon. According to Greenberg (1947), Hausa also formed compound verbs with the help of the verb yi ‘to do’, e.g. yi kàráatuu ‘to read’ (< Arabic qirāʾa ‘reading’). The examples quoted by Greenberg for DO-constructions belong to the context of the Western Sudan, and possibly go back to a stage of limited Hausa/Arabic bilingualism that preceded the present situation in West Africa. In the Eastern Sudan, however, speakers of Hausa live in a situation of bilingualism and code-switching. In Sudanese Hausa, Abu-Manga (1999: 137–161) observes a very high percentage of loan verbs (19.3 percent of all loanwords), and from his description it becomes clear that virtually all Arabic verbs can be integrated in the Hausa lexicon. In fact, the situation as he describes it looks more like a fully developed code-switching context, with Sudanese Arabic as the dominant language. Examples of borrowed verbs abound: jahhazaa ‘to prepare’, assasaa ‘to establish’, wallaʾaa ‘to set on fijire, light’, zuuraa ‘to visit’, dummaa ‘to attach to’, gaabalaa ‘to meet’, darrabaa ‘to train’, and káràntáa ‘to read’ (contrasting with West Sudanic Hausa yi kàráatuu). The form of some of these loan verbs suggests that they are derived from imperatives (e.g. suugaa ‘to drive’ < Sudanese Arabic sūg ‘drive!’), although derivation from an Arabic imperfect form, as posited by Abu-Manga (1999: 140), remains a possibility. These fully integrated examples of Arabic verbs in the Eastern Sudan reflect a situation of full bilingualism, in which the speakers go beyond the DO-construction and start incorporating verbs from the dominant language directly into their speech. A third strategy for incorporating Arabic loan verbs in Hausa, with the help of a verbalizing sufffijix -ta, e.g. Arabic ḥukm > hukunta ‘to pass judgment’ (Abu-Manga 2007) is more difffijicult to explain for the moment. At diffferent times, and in diffferent situations, Hausa speakers have, therefore, used diffferent strategies to incorporate Arabic verbal material in their language: during initial contact, they used Arabic imperatives; when the contact became more intensive, they switched to a DO-construction, and when full bilingualism obtained, they freely integrated Arabic verbs in their own language. There may be many more examples of a diachronic development of diffferent options, which have become lost in history. The present-day situation in Iran with its low degree of Persian/Arabic bilingualism has obtained since the beginning of the 16th century, but it must have been preceded by a period of extensive bilingualism after the initial conquest of Iran by the Arabs in the 7th century. Such a situation with

224

kees versteegh

extensive bilingualism and code-switching must also have existed at one time in Egypt with speakers of Coptic, and in North Africa with speakers of Berber. About Arabic loanwords in Coptic nothing much is known (Richter 2006), but it seems to be the case that for the borrowing of Arabic verbs a method was chosen that had already become conventionalized for the (extensive) borrowing of Greek verbs; these were always borrowed in one form, the infijinitive, which functioned in the same way as the Coptic infijinitive, because infijinitives in both languages can be used as a command and as a predicate. The few examples of Arabic verbs that exist stem from alchemistic treatises in which imperatives abound, but it is not quite clear which form of the Arabic verb was selected in the process of borrowing, possibly the imperative. If we attempt to correlate the borrowing strategies mentioned here with the borrowing context, four diffferent options for languages to integrate verbal loans from Arabic appear to be available theoretically: i. ii. iii. iv.

bare verbs: pidginization DO-constructions with verbal elements: extensive bilingualism morphological integration: full bilingualism inflected verbs: language shift

The fijirst option is the one chosen in initial contact situations, for instance by speakers of Arabic pidgins/creoles, and speakers of Fulfulde. The second option is that of using light verbs in a DO-construction. This is the option chosen for Arabic loan verbs in West Sudanic Hausa, Persian, Turkish, and Urdu. The third option is the full morphological integration of Arabic verbs that we have seen above in Eastern Sudanese Hausa and Neo-Aramaic. The fourth option, borrowing inflected verbs, may be selected during the fijinal stage of linguistic contact and may be indicative of an ongoing language shift, as in the case of Greek inflected verbs borrowed by Cypriot Arabic (cf. Borg 1985; Versteegh 2009). This option does not seem to occur in the case of Arabic loanwords in other languages. One borrowing process has not been included in these four options. There may, indeed, be a fijifth option for the incorporation of Arabic verbs, which is selected when there is no contact at all between the speakers of the borrowing language and the speakers of Arabic. This is the case when borrowing takes place through written transmission, by scholars whose knowledge of Arabic is derived from books, in a process that may be compared to the formation savante of Latin loanwords in the European languages. This is what happened in Persian, Turkish, Urdu/Hindu-

heavy and light borrowing of arabic verbs

225

stani, Indonesian/Malay, and, to some extent, in Swahili and Hausa. In Kitap-Malay (the name sometimes given to a variety of written Malay in which Arabic texts were translated almost word-by-word), for instance, the initiators of the borrowing were scholars with a high command of both languages. When they integrated foreign elements—often religious terms they needed in their exposés of Islamic theology—, they used Arabic nouns and derived from them verbal forms with the afffijixes in the morphological system of Malay. The same procedure was followed by Swahili scholars for most of the later loans from Arabic, which were introduced much later than the initial period of contacts with Arab traders. Those Arabic loan verbs in Hausa that are formed with the verbalizing sufffijix -ta (mentioned above) can also be explained by assuming that they were derived in a learned context. In some contexts of Islamic learning, DO-constructions are used in combination with Arabic words, but these are diffferent from the ones that are common in a truly bilingual oral context. In Kitap-Afrikaans (i.e., the Afrikaans written in Arabic script by Muslims, most of whom had been brought to the Cape Colony from the East Indies), verbs are usually integrated with the help of a DO-verb māk ‘to do’ (i.e. Afrikaans maak < Dutch maken), in combination with an Arabic verbal noun, e.g., māk ʿibāda ‘to serve’, māk sujūd ‘to kneel in prayer’, ṣalā māk ‘to pray’ (Hoedemaekers 2006: 64). These seem to be independent innovations, because in Malay there are no DO-constructions, but borrowed nouns are verbalized by indigenous derivational processes. The situation in Kitap-Afrikaans is complicated by the fact that there are a few examples of Arabic nouns being used as verbs without the help of a dummy verb, e.g., om teṣalāt ‘in order to pray’ (< Arabic ṣalāt ‘prayer’). This construction is paralleled by the occurrence of Malay verbs in Afrikaans used in the same way, e.g., gebacha ‘read [past participle]’ and tebacha ‘to read’ (< Malay (mem)baca ‘to read’) (Kähler 1971: 47–64; Hoedemaekers 2006: 80). In written transmission, therefore, nouns are the central element in the borrowing process, and the linguistic category of the borrowed elements may therefore serve as a diagnostic criterion for the type of bilingual context in which the borrowing took place. Boumans’ (2007) conclusion that the use of DO-constructions is always connected with intensive language contact between the speakers of a community language and those of a superimposed language, for instance in a migration context, should therefore be modifijied. DO-constructions are in fact used outside a context of intensive language contact, but the resulting complex verbs difffer. In Persian/Swedish code-switching, nouns are never used in DO-constructions,

226

kees versteegh

but only bare infijinitives. In Arabic loanwords in Persian, on the other hand, only verbal nouns are used in the DO-construction. Park (1995) analyzes similar constructions in Korean and Japanese. In Korean, the light verb ha is used chiefly with verbal nouns borrowed from Chinese, as in john-ka saken-ul cosa-(ul) ha-ass-ta ‘John investigated the afffair’

Although he does not explicitly say so, these borrowings must therefore be the result of written transmission within a learned context. When loanwords from English are used in the same construction, Park (1995: 351) notes that they can never be nominal, but only verbal, as in john-ka nonmun-ul revise-(ul) ha-ass-ta John-Nom paper-acc do-Past ‘John revised the paper’ *john-ka nonmun-ul revision-ha-ass-ta

This may explain why Arabic loanwords in complex verbs in Persian are always verbal nouns, whereas in Swedish/Persian code-mixing only verbs can be used in these constructions. According to Park, the only function of the light verb is to support inflection, which means that all other verbal functions are performed by the verbal noun. Accordingly, in Persian only those Arabic words can be used which have a highly verbal character, such as the verbal nouns. One thing is certain: Romaine’s claim (1989: 120) that complex verbs are an areal feature, which she bases on the case of Punjabi code-switching, cannot possibly be correct, because DO-constructions are found elsewhere as well, as one step in the incorporation of foreign material in contact situations. All areal claims about the occurrence of either construction are refuted efffectively by the examples of speakers using diffferent strategies in diffferent situations, as in the case of Hausa. The determining factor in all instances of borrowing is the context in which transmission takes place, as well as the mode of transmission.

heavy and light borrowing of arabic verbs

227

References Abu-Manga, Al-Amin. 1999. Hausa in the Sudan: Process of adaptation to Arabic. Cologne: R. Köppe. ——. 2007. “Hausa”. Encyclopedia of Arabic Language and Linguistics II, 250–256. Leiden: E.J. Brill. Aguadé, Jordi and Peter Behnstedt. 2006. “Berber”. Encyclopedia of Arabic Language and Linguistics, I, 289–293. Leiden: E.J. Brill. Arnold, Werner. 2007. “Neo-Aramaic”. Encyclopedia of Arabic Language and Linguistics, III, 370–373. Leiden: E.J. Brill. Backus, Ad. 1996. Two in one: Bilingual speech of Turkish immigrants in The Netherlands. Ph.D. diss. University of Tilburg. (Tilburg: Tilburg University Press.) Borg, Alexander. 1985. Cypriot Arabic: A historical and comparative investigation into the phonology and morphology of the Arabic vernacular spoken by the Maronites of Kormakiti village in the Kyrenia district of North-Western Cyprus. Wiesbaden: F. Steiner. Boumans, Louis. 1998. The syntax of codeswitching: Analysing Moroccan Arabic/Dutch conversation. Ph.D. diss., University of Nijmegen. (Tilburg: Tilburg University Press.) ——. 2007. “The periphrastic bilingual verb construction as a marker of intense language contact: Evidence from Greek, Portuguese and Maghribian Arabic”. Approaches to Arabic linguistics presented to Kees Versteegh on the occasion of his sixtieth birthday, ed. Everhard Ditters and Harald Motzki, 291–311. Leiden: E.J. Brill. Brown, Jonathan A. 2007. “New data on the delateralization of ḍād and its merger with ẓāʾ in Classical Arabic: Contributions from Old South Arabic and the earliest Islamic texts on ḍ/ẓ minimal pairs”. Journal of Semitic Languages 52, 335–368. Désoulières, Alain. 2008. “Urdu/Hindustani”. Encyclopedia of Arabic Language and Linguistics IV, 595–603. Leiden: E.J. Brill. Gardner-Chloros, Penelope and Malcolm Edwards. 2004. “Touching base: The relevance of grammatical models to code-switching”. Actas del II Simposio Internacional Bilinguïsmo Vigo Spain, 1423–1452. Available at: http://webs.uvigo.es/ssl/actas2002/07/05.%20 Penelope%20Gardner.pdf. Greenberg, Joseph. 1949. “Arabic loan-words in Hausa”. Word 3, 85–97. Hoedemaekers, Iris. 2006. ‘Die kitāb wat sal prāt op die rules fan die naḥw . . .’: Het ArabischAfrikaans. M.A. thesis, University of Nijmegen. Jakobi, Angelika. 2007. “Nubian”. Encyclopedia of Arabic Language and Linguistics III, 435–439. Leiden: E.J. Brill. Kähler, Hans. 1971. Studien über die Kultur, die Sprache und die arabisch-afrikaanse Literatur der Kap-Maleien. Berlin: Reimer. Lotfabbadi, Leyla Naseh. 2002. Disagreement in agreement: A study of grammatical aspects of codeswitching in Swedish/Persian bilingual speech. Ph.D. diss., University of Stockholm. Available at: http://www.ling.su.se/forskning/avh/disagreement_in_agreement .pdf. Matras, Yaron. 2002. Romani: A linguistic introduction. Cambridge: Cambridge University Press. Muysken, Pieter. 2000. Bilingual speech: A typology of code-mixing. Cambridge: Cambridge University Press. Naess, Unn Gyda. 2008. ‘Gulf Pidgin Arabic’: Individual strategies or a structured variety? A study of some features of the linguistic behaviour of Asian migrants in the Gulf countries. M.A. thesis, University of Oslo. Park, Kabyong. 1995. “Verbal nouns and Do-insertion”. Studies in Generative Grammar 5, 319–357. Perry, John R. 1991. Form and meaning in Persian vocabulary: The Arabic feminine ending. Costa Mesa, Cal. & New York: Mazda Publishers. Richter, Tonio Sebastian. 2006. “Coptic”. Encyclopedia of Arabic Language and Linguistics I, 495–501. Leiden: E.J. Brill.

228

kees versteegh

Romaine, Suzanne. 1989. Bilingualism. Oxford: Blackwell. Schwarz, Nadja. 2004. Arabic loan verbs in Swahili: A study of the history of language contact on the East African Coast and an analysis of Swahili verbs of Arabic origin. M.A. thesis, University of Nijmegen. Theil, Rolf. 2007. “Fulfulde”. Encyclopedia of Arabic Language and Linguistics II, 137–142. Leiden: E.J. Brill. Tschacher, Torsten. 2008. “Tamil”. Encyclopedia of Arabic Language and Linguistics, IV, 433–436. Leiden: E.J. Brill. Tucker, A.N. 1946–1947. “Foreign sounds in Swahili”. Bulletin of the School of Oriental and African Studies 11, 854–871; 12, 214–232. Versteegh, Kees. 1999. “Loanwords from Arabic and the merger of ḍ/ḏ”. Israel Oriental Studies 19, 273–286. ——. 2003. “The Arabic component of the Indonesian lexicon”. Rintisan kajian leksikologi dan leksikografiji, ed. Bambang Kaswanti Purwo, 1–26. Jakarta: Universitas Atma Jaya. ——. 2009. “Loan verbs in Arabic and the DO-construction”. Arabic dialectology: In honour of Clive Holes on the occasion of his sixtieth birthday, ed. by Enam Al-Wer and Rudolf de Jong, 187–200. Leiden: E.J. Brill. Wellens, Ineke. 2005. The Nubi language of Uganda: An Arabic creole in Africa. Leiden: E.J. Brill.

WHEN ARABIC RESONATES IN THE WORDS OF AN AFRICAN LANGUAGE: SOME MORPHOLOGICAL AND SEMANTIC FEATURES OF ARABIC LOANWORDS AND CALQUES IN BAMBARA Francesco Zappa

This article aims to highlight some morphological and syntactic features of the impact of (mainly Classical) Arabic upon the lexicon of one of the most widespread and rapidly expanding ‘indigenous’ African languages: Bamanankan (better known through the heteronym “Bambara”), a language which is quite distant from Arabic in geographical, genetic and typological terms. Within the borders of the modern state of Mali, though not enjoying offfijicial status, Bambara is presently standing out with a growing momentum as a vehicular language on a national scale, at least in oral communication,1 thus achieving a longstanding process which probably dates back as early as the age of medieval Manding empires (especially the empire of Mali, 13th to 15th century), and was subsequently further fostered by the dynamics of colonial and postcolonial history, often beyond the intentions and stated policies of political authorities. If one considers the Manding language continuum as a whole (a large group of fairly mutually intelligible varieties, of which Bambara is, alongside Maninkakan/ Malinké and Julakan/Dioula, one of the most widely spoken), the number of speakers involved rises to over fijifteen millions stretching over a much wider region of West Africa.2 The Bambara- (and, more generally, Manding-)speaking world is thus separated from the areas where Arabic is mostly practiced as a native language by natural barriers (the Sahara desert) as well as by the spread of

1

 See Dumestre (1994) and Canut (1996: 60–63).  See Dumestre (2003: 7) and, for more generous estimates, Vydrine (1999: 8–9). Quantitative data about the number of speakers vary dramatically from source to source, depending also on a number of criteria: distribution of speakers among the diffferent Manding varieties (whose delimitation, in turn, is far from unproblematic); distinction between L1 and L2 speakers (while it is still more difffijicult to assess what are the minimum profijiciency standards and the minimum frequency of language use for being considered an L2 speaker); inclusion or exclusion of speakers with only passive—however relatively high— profijiciency in a (usually vehicular) Manding variety. 2

230

francesco zappa

other languages, which have acted as intermediaries—as we shall see— for the transmission of Arabic loanwords. Genetically speaking, Bambara, along with the other Manding varieties, belongs to the Mande branch of the Niger-Congo family which, as is well known, shares no common origins with Semitic languages.3 In typological terms, quite a number of features diffferentiate it from Classical Arabic. To make just a few obvious examples, the basic word order in Bambara is SOV (Subject-Object-Verb) as opposed to VSO (VerbSubject-Object) that is standard in Arabic, especially Classical. And while the latter is a typically synthetic and fusional or inflecting language, Bambara is basically “an isolating language with some agglutinative elements” (Vydrine 2007: 1916): in other terms, an extreme scarcity of inflectional morphemes and a very limited flexibility in the word order of this language are compensated by widespread resort to compounding and derivation, through a wide range of afffijixes, which enriches its vocabulary with some remarkably lengthy words. Moreover, determinative sequences and phrases follow a regressive construction: determiners precede determined elements,4 according to an order which is opposite to Arabic construct state. Finally, it is a tone language as far as suprasegmental features are regarded.5 However, all these diffferences, though remarkable, did not prevent Arabic from being massively present in the lexicon of Bambara, as well as of other Manding languages. Already in the 1920s, Maurice Delafosse evaluated the share of roots borrowed from (or passed through) Arabic in the lexical thesaurus common to diffferent Manding varieties between 15 and 20 percent,6 even though newer dictionaries of Bambara tend to record a lower percentage.7 In fact, not all Arabisms used in Malinké and Dioula can be found in Bambara (and the other way round), and on the other hand the inventory of forms attested in Bambara in far from complete: in printed works or texts of oral literature I happened quite often to fijind borrowings or calques from Arabic which are not listed in dictionaries, mostly coming from the Islamic religious lexicon. I think this delay in lexicography 3  For more details on the classifijication of Manding languages and their respective position within the Mande family, see Vydrine (1999: 9–11). 4  See Dumestre (2003: 101–103; 109–114). 5  Tonological features of integration of borrowings from Arabic are still to be explored and will not be taken into account in the following pages. 6  See Delafosse (1955: i–iii, posthumous work); see also Delafosse (1929: 136). 7  See for instance Bailleul (2007), which has been the most complete and reliable dictionary since the publication of the fijirst edition (1981).

when arabic resonates in an african language

231

can be explained by a tendency—more noticeable in newer Manding linguists than it is in Delafosse (who had an orientalist background)—to understate, or in any case to neglect, the impact of Islam on languages and cultures of Manding-speaking peoples. On the contrary, the richness of the Manding Islamic lexical stock and its employ in several genres of verbal arts have been fijirst acknowledged by the anthropologist and Islamicist Tal Tamari (1996; 2001; 2002; 2005a; 2005b; 2006).

Language Contact between Arabic and Bambara: Contexts and Specificities So deep an influence was made possible by a linguistic contact of a particular kind, which arose essentially in the lack of a direct, massive oral interaction between native speakers. This kind of contact evokes the main lines of linguistic Islamization as sketched by the Islamicist Alessandro Bausani (1967; 1981) some decades ago in his comparative reflections about what he called ‘Islamic languages’. According to Bausani, the impact Arabic, as the educated language of Islam par excellence, had on the languages spoken by culturally Islamic peoples usually stemmed from writing practices of non-Arabic speaking local Muslim elites, which nevertheless mastered written Arabic quite well. However, the elitist nature of this “bookish influence”—to borrow Bausani’s own terms8—was reduced by what Bausani called “top-down osmosis”, that is the capacity of the language and the culture of Muslim learned layers to penetrate popular layers, including illiterate people. In our context, however, the lack of a writing tradition before the impact of Islam limited the use of Arabic writing to write Bambara (even more so than other languages in the region) to entirely sparse and subsidiary employs.9 Therefore, the influence of Arabic passed mainly through what I called elsewhere a “learned orality”,10 that is the systematic recourse to oral periphrastic translations into Bambara (as well as in other West African languages) of written Arabic texts: a

 8  Bausani’s original phrase was “influsso libresco”, which in Italian conveys an even stronger sense of dusty old paper.  9  See Tamari (1994) and Vydrine (1998); for an overview of the use of ʿajamī (i.e. Arabic script applied to a language other than Arabic) in West African languages, see Hassane (2008). 10  See Zappa (2009a: 170).

232

francesco zappa

practice which is widespread in the advanced levels of the traditional Islamic educational system.11 A fijirst clue of this decisive influence of traditional Koranic pedagogy is given by the addition of Arabic endings in the pronunciation of anthroponyms of Arab-Islamic origin, as well as of a large number of common nouns, especially masculine (in Arabic, since there is no grammatical gender in Bambara): take Amadu (from ʾAḥmad) among proper nouns and baliku (‘grown up’ from bāliġ) among common nouns for example. It is true that the Bambara syllable structure, which does not allows for close syllables (CVC), encourages the insertion of a word-fijinal vowel, and that this vowel is /u/ might at fijirst sight appear casual. Nevertheless, phenomena like the integration of tāʾ marbūṭa followed by ḍamma in such anthroponyms as Ayisatu (from Arabic ʿĀʾiša)12 or the addition of endings in person-names structured as construct states, such as Abudulayi can only be explained by the influence of the entirely vocalized reading of Arabic which is taught in Koranic schools, where pupils learn to pronounce also these morphemes with a purely grammatical value, which are customarily omitted—as everyone knows—in the current pronunciation of spoken Arabic, even in the classical variety.13 That does not prevent speakers from any social origin (including illiterate and non-Muslim people) from adopting this pronunciation, through a typical “osmosis” process. Besides, it is no chance that such pronunciations can be found in many West-African languages. Even if an Arabic proper noun can give rise to several diffferent outcomes, which correspond to as many diffferent degrees of alteration owing to spontaneous transmission, it holds true that the variants which are the least close to the Arabic “school” model show no tendency to come closer to the pronunciation of Spoken, or dialectal, Arabic, but rather to the phonological features of the target language: one

11  See Tamari’s works cited above. It should not be overlooked that a large part of the scholars devoted to this kind of translations are multilingual; many of them are accustomed to teach in several diffferent native languages, and Bambara is often not their native language, which makes the issue of linguistic interference more complex. 12  Not surprisingly, most of the nouns incorporating ḍamma are masculine (in Arabic; Bambara has no grammatical gender): while an epenthetic vowel is required in most masculine Arabic nouns, which end by consonant, it is not necessary to incorporate tāʾ marbūṭa followed by ḍamma to have a vowel ending in most feminine loanwords. 13  This phenomenon was already noted by Delafosse (1910: 259–260), without explicitly linking it to the influence of Koranic education, while Lagarde (1988: 38) mentions in this regard that “c’est surtout ce type d’arabe que les Bambara ont connu et pratiqué tout au long de leur histoire”.

when arabic resonates in an african language

233

has only to consider the (doubtless incomplete) series of variants of the Arabic fijirst name ʿĀʾiša in Bambara: Ayisatu, Ayisata, Asetu, Setu, Sata, Asitan, Sitan. Besides learned Arabisms, the vocabulary of Bambara includes a certain number of Arabic borrowings entered through the channel of commercial contacts. Even in this case the contact was often not direct, since it passed through the mediation of geographically close languages (such as Soninke, Songhay and, to a lesser extent, Fulfulde), whose speakers had been Islamicized for longer and were in more direct contact with Arab traders.14 As Gérard Dumestre (1983: 18–20) observed, unlike Arabisms entered through bookish influence, Arabisms with a commercial origin are distinguished by their semantic fijields, as well as by a more marked phonological alteration compared with the original, in the absence of a periodical comparison with their model. They are distinguished also by the date of their entry into the lexicon of Bambara, which is almost always before the colonial age: in fact, since this age terms relative to concrete objects or to modernity, in a way or another, are borrowed almost exclusively from French.15 On the other hand, the process of adoption and creation of learned neologisms, though an old one, is still well underway, and implies, together with borrowing, a non-negligible amount of calques: it is exactly today’s socio-cultural context of speakers, marked by a more and more expanding and spreading Islamicization, that explains the continuing openness of this inventory. In fact, if the hypothesis holds that Arabic had been used, at least to some extent, as a spoken languages by Manding-speaking learned elites before the collapse of the Mali Empire, at least according to Tal Tamari (2002: 98), it is a fact that the recent flourishing of modern Koranic schools, where Arabic is the only teaching language,16 in addition to the growing number of Malian students attending University courses in Arab countries, promoted an unprecedented spread in the mastering of spoken Arabic among Malians.17

14

 See Dumestre (1983: 15–17).  See Dumestre (1983: 14). 16  This is true at least in theory; recent fijieldwork carried out by Tamari (2009) has shown that in many such schools, in spite of what directors offfijicially claim, Bambara (or, for that matter, other local vernaculars) still retain an important role in oral discussion and explanation. 17  See Bouwman (2005) and Brenner (2001). 15

234

francesco zappa

In its turn, the role played by Bambara in the Islamic religious fijields is far from being marginalized by this wave of modern education in Arabic; it is, rather, strengthened by the emergence of new, more and more widespread, forms of religious communication, such as radio-broadcast or tape-recorded sermons, or new Islamic books in Bambara.18 All these phenomena paved the way for the creation of new Arabisms, especially in the semantic fijields more connoted by Muslim faith, as well as for the broadening of the user base of already existing ones, well beyond the community of traditional Muslim scholars. In such a dynamically changing context, I consider as useless or at least questionable any attempt to draw a sharp distinction between “loanwords”, well integrated at the phonological, morphological and semantic levels, and “quotations”, weakly integrated and employed by some speakers in some registers only.19 It is no less difffijicult to make a clear distinction in the learned lexicon between newly created and older neologisms, the latter having long been unnoticed as specifijic to specialized languages. Even more so because some little explored genres of traditional oral literature offfer some big surprises about the familiarity of uneducated people with learned words.20 Likewise, the opposition between loanwords within and outside the religious fijield, such as proposed by Bouwman (2006: 257), is, in my opinion, open to question: in fact, we end up in classifying as “nonreligious” a number of words which are strictly speaking outside religion, but are indirectly linked to it in some way ore another, and entered the language through Islamic education, such as words denoting moral properties, feelings, time expressions, objects (and activities) linked to writing practices. A distinction between learned, “Islamic” Arabisms (whether borrowings or calques) and those entered through commercial contacts seems to me more convincing. It is exactly words belonging to the former category that are the subject of the following observations.

18

 See respectively Schulz (2003) and Zappa (2004; 2009a).  See for instance Baldé (1980: 196, inspired by Pierre Francis Lacroix’s work), also cited by Bouwman (2006: 256). 20  See Zappa (2004: 87–108 ; 2009b). 19

when arabic resonates in an african language

235

Morphological and Semantic Aspects of Borrowings and Calques On the morphosyntactic level, the borrowing of lexemes from Arabic into Bambara does not seem to involve wide-ranging phenomena of integration (or “nativization”) or, even less, of structural interference.21 The high degree of morphological integration of most Arabisms derives fijirst of all from their adaptation to Bambara verbal and nominal flexion, derivation and composition patterns. On the other hand, Bambara verbal and nominal flexion employs a very limited set of afffijixes, and involves no type of internal flexion. Integration and Structural Interference through Inflection, Derivation and Compounding Borrowing rarely starts from an Arabic inflected form rather than from a lemma; even in the former case, we cannot speak of morphological interference, since it is not the flectional mechanism of Arabic as such that is borrowed: for example, Arabic broken plurals are lexicalized as singulars. I was able to note two new cases of nominal borrowing from an Arabic broken plural only; moreover, they are quite rare words relative to the magico-religious fijield, which I found in an oral epic text with a hagiographic subject, and are not recorded in dictionaries. The fijirst example is the word kawatimu, derived from Arabic xawātim (pl. of xātam or xātim), in the talisman-related meaning.22 The second one is the word lasirari, derived from Arabic al-ʾasrār (pl. of sirr, ‘secret’), in the meaning of ‘occult properties’ of a given Koranic verse or passage.23 Another example, cited 21

 Phenomena directly deriving from phonological integration and phonological interference will not be take into account in this paper, since they have already been thoroughly dealt with by Lagarde (1988: 26–39) and Bouwman (2006: 256–257). 22  For further details see Hamès (1997: 212 n. 27); on the other hand, in this meaning the Arabic word is frequently used in its plural form (see Hamès 1997: 42 n. 29). 23  As I had occasion to note while making inquiries in Bambara-speaking Marabout milieus, the fijirst word is as frequently attested, with roughly the same meaning, as katimu, directly derived from the Arabic singular form and also absent from dictionaries. The second word, on the other hand, is well known in the form quoted in the text (lasirari), which is directly derived from the Arabic broken plural (besides, the plural form of this word is more frequent in Arabic esoteric texts); however, we might ask if the Bambara word siri, with the meaning of ‘spell’, could be derive in its turn from the Arabic singular form of the same word, even if its homophony with a verb meaning ‘to tie’ drove investigators to link its meaning to the latter, since in Bambara society many forms of spell involve

236

francesco zappa

by Bouwman (2006: 257) and recorded by Bailleul’s (2007: 115) dictionary, is the verb durusi ‘memorize’, which derives from the Arabic broken plural noun durūs (pl. of dars, ‘lesson’), with change of lexical category (from an Arabic noun to a Bambara verb), even if the hypothesis of a derivation from an inflected form of the imperative (udrus) should, in my opinion, be considered as well. On the other hand, I could identify no example of borrowing of Arabic sound plural endings, unlike what was observed in other languages.24 At least two Bambara verbs seem to be derived from the form of Arabic imperfect (muḍāriʿ marfūʿ): the fijirst one is the verb yamaru ‘to allow, to prohibit’, which would derive, according to Lagarde, from the imperfect yaʾmuru of the Arabic verb ʾamara ‘to order, to command’, with a slight semantic shift; the other example is the Bambara verb yafa ‘to pardon’, which would derive from Arabic yaġfijiru, imperfect of ġafara (same meaning), even if in my opinion the possibility of a derivation from the Arabic synonym ʿafā should not be excluded. In fact, for both verbs it is possible to claim a regular derivation from a lemma (that is, the Arabic perfect form) obscured by the semivowel outcome (in palatal voiced approximant yāʾ) of an initial phoneme which is relatively instable in the passage from a phonological system to another one: the glottal plosive hamza in the case of ʾamara, the voiced pharyngeal fricative ʿayn in the case de ʿafā; in any case, it would be a rather exceptional outcome.25 Some cases should also be noticed where the verb borrowing probably started from an infijinitive form (the Arabic maṣdar), or perhaps even from an inflected form of the imperative. At least, this seems to me the explanation most consistent with the form of Bambara verb lemmas such as juura or joora (‘to visit the tomb of a saint’, perhaps from the imperative zur of the verb zāra: among the infijinitive forms of the latter, ziyāra is phonologically too far removed,

the use of knots- an allegedly ‘pagan’ practice which may possibly have drawn inspiration, in turn, from a mention in the Qurʾān (cxiii 4) of ‘women who blow [incantations] over knots (see Soares 2005: 131). A further borrowing of the same noun from the Arabic singular, retaining the inflectional morpheme (ḍamma), is also attested by Soares (2005: xii) under the form siru. 24  See Versteegh (2001: 479; 497). 25  Other examples of verb borrowing from the imperfect not recorded in dictionaries were identifijied by Tamari (2005b: 438) in the jargon of Islamic learning. Interestingly enough, even in these cases, the lemma starts with the voiceless glottal plosive hamza: see for example yakaro (‘to delay, to be in delay’ < Ar. yuʾaxxiru, imperfective form of ʾaxxara, same meaning).

when arabic resonates in an african language

237

while zawra has not the same meaning); jiidi (‘to increase, to multiply’, probably from the imperative form zid of the verb zāda, for lack of similarity to the infijinitive form ziyāda); faamu (‘to understand’; here, both hypotheses, respectively from the maṣdar fahm or the imperative ifham, seem plausible).26 A signifijicant amount of derived words are formed from borrowings from Arabic; among Bambara derivational afffijixes (which, unlike flexional afffijixes, are quite numerous), the sufffijix –ya is among the most productive in this respect. It can be added to a nominal, adjectival or verbal, simple or complex base, and can work as an abstraction (or “state”) sufffijix, or to form verbs starting from adjectives, nouns or (more rarely) even other verbs, in the latter case without any change in meaning. Its near-homophony with the Arabic derivational sufffijix –iyy followed by the singular feminine ending ( fatḥa + tāʾ marbūṭa) is much more remarkable since they share the value of abstraction or state, so much so that one could ask if it is not a morpheme borrowed from Arabic: it would be the only case of borrowing of a grammatical morpheme. Anyway, I never happened to fijind cases of morphological reanalysis which are based on this similarity of sound and meaning. It is rather the converse that holds in the case (from the same epic-hagiographic text cited above) of a word such as wahabiyaya, which denotes the Wahhābiyya in an abstract sense, and which is formed by gluing together the two homophone sufffijixes, the fijirst Arabic and the second Bambara. It is a well-known fact that in Arabic the singular feminine flexional suffix ( fatḥa + tāʾ marbūṭa) can be added to the derivational sufffijix –iyy of the relation nouns-adjectives not only to form the corresponding abstract noun, but also to form a plural, especially when one wants to denote the set of the members of a sect, a dynasty or a school. Therefore, the word wahhābiyya can represent at the same time the set of disciples and followers of Muḥammad b. ʿAbd al-Wahhāb and his doctrine, or the membership in his school. Since this word entered Bambara by integrating in its phonology through a simplifijication of geminates (wahabiya) as an unanalysable whole and as a collective noun only, without being felt as an abstract noun as well, the necessity was felt to add the Bambara

26

 The hypothesis of a derivation from an imperative form has never been proposed for these verbs (nor for other Bambara verbs), at the best of my knowledge; yet it seems to be the most plausible one, and it is confijirmed by similar examples in other languages influenced by Arabic (see Versteegh 2001: 486–7, and the same author’s contribution to the present volume).

238

francesco zappa

abstraction sufffijix –ya in order to form the correspondent abstract noun wahabiyaya. Besides, the same sufffijix can be added to borrowed words without a corresponding Arabic derived abstract noun formed through the sufffijix –iyya. Thus, while, for example, in Arabic the word šayx is matched by the abstract nouns šiyāxa, mašyaxa and šayxūxa, but not *šayxiyya, in Bambara the derived abstract noun sekuya is derived from the (phonologically well integrated) borrowing seku. Similarly, the Arabic words bidʿa (‘heretic innovation’) and ʾīmān ‘faith’, which have already an abstract meaning, entered Bambara exclusively with the addition of the abstraction sufffijix and created, respectively, the forms bidaya and limaniya. The latter lexeme, limaniya, also works as a verb with the meaning ‘to believe’ (since the Bambara sufffijix –ya also forms verbs, as we have seen). Adding the punctual agent sufffijix –baga produces the derived noun limaniyabaga, whose meaning ‘believer’ corresponds to Arabic muʾmin, active participle of the same verb of which ʾīmān is the infijinitive form. The word muʾmin on its turn was also borrowed into Bambara as the much rarer (and unrecorded in dictionaries) mumini (or muminu). The latter word, which takes up the ameliorative meaning of ‘pious, devout’, produces the abstract word muminiya (in the sense of ‘devotion’) through the addition of the abstraction sufffijix –ya. Then, we have to do with two almost synonymous couples (with a certain degree of semantic specialization) of borrowings from two diffferent forms of the same Arabic verb: the two abstract nouns limaniya and muminiya on the one hand, and the two concrete nouns limaniyabaga and mumini on the other hand. The derivational sufffijix –ya can also be added to an Arabic determinative phrase borrowed as an unanalysed whole: this is the case of the Sufiji title quṭb az-zamān ‘pole of [his] time’, a lexicalized construct state which entered Bambara as kutubujamaani (with incorporation of the Arabic flexional endings), which on its turn produces, by adding the usual abstraction sufffijix –ya, the abstract derived noun kutubujamaaniya ‘status of a quṭb az-zamān’, attested in an oral hagiographical text. Besides derivation, lexemes borrowed from Arabic are integrated morphologically also through composition, by combining with other borrowings or with native lexemes. There are also cases of borrowing from Arabic which enter into compound nouns without ever appearing as autonomous lexemes. For instance, the morpheme kara, borrowed from the Arabic qaraʾa ‘to read’, never occurs in isolation, but is always embedded in the compound noun karamɔgɔ (lit. ‘man of reading’, ‘master, learned man’) and in its derived forms.

when arabic resonates in an african language

239

In the formation of compound nouns which follow the schema of determinative phrases, the word order of Bambara, which is regressive and the opposite of the Arabic construct state, is almost always respected. Thus, for example, the lexeme misiri, derived from masjid ‘mosque’, enters into composition with the native lexeme wele ‘call’ to form the compound noun misiriwele (lit. ‘mosque call’, hence ʾāḏān), which respects the regular word order of Bambara determinative phrases, where the determining element always precedes the determined one. One intriguing example are the Bambara terms for the Arabic scholarly phrases ʿilm al-ẓāhir (lit. ‘science of the exterior’, hence ‘exoteric science’) and ʿilm al-bāṭin (lit. ‘science of the interior’, hence ‘esoteric science’), which translate, respectively, as kɛnɛmanadɔnni and sokɔnɔnadɔnni: the compound’s head, dɔnni (the standard indigenous verbal noun for ‘knowledge’) is postponed to the modifijiers, kɛnɛmana and sokɔnɔna, which are also made of exclusively indigenous morphemes.27 Almost all the extremely exceptional cases of true syntactic calques derive from specialized “marabout” lexicons. Let us take the compound kabakolo larusi, which corresponds to the Arabic falak al-ʿarš (‘Heaven of the Throne’, one of the celestial spheres in the Sufiji cosmology): its two elements (kabakolo, which translates the Arabic falak, and larusi, borrowing from the Arabic al-ʿarš) are a calque of the syntactic order of the construct state they translate, rather than following that of the Bambara determinative phrases. Moreover, the formation of compound or derived nouns as calques is especially developed in the framework of translations of the Islamic technical lexicon. The least problematic cases are those which translate Arabic qualifying phrases, since the word order in Bambara determinative phrases is the same. In such cases, an Arabic adjective may be translated by a Bambara sufffijix, which does not change at all the order of morphemes. We may cite in this regard the names of the two main holidays in the Muslim year: seliba, which corresponds to al-ʿīd al-kabīr, and selidennin, which corresponds to al-ʿīd aṣ-ṣaġīr. Both are derived nouns, formed from 27  Interestingly enough, while both modifijiers (though seldom employed in a metaphorical sense in current speech) translate quite literally the meaning of ‘exterior’ and ‘interior’, their structure and etymology is much unlike those of the Arabic terms they translate. Kɛnɛmana is a derivative noun from kɛnɛ (‘clearing’, ‘open space’) + ma (a postposition indicating direction or contact) + –la/-na (a locative derivational sufffijix); sokɔnɔna has the same structure, from so (‘house’, ‘inhabited space’) + kɔnɔ (a postposition meaning ‘inside’) + –la/-na (locative derivational sufffijix, cf. above).

240

francesco zappa

the borrowing seli (derived from ṣalāt, which in Arabic has only the meaning of ‘canonical prayer’, while in Bambara it takes, among other things, also the meaning of ‘religious holiday’) with the addition of, respectively, the augmentative sufffijix –ba and the diminutive sufffijix –dennin,28 which correspond almost literally to the Arabic adjectives kabīr and ṣaġīr. A richer and more complex inventory, where the phenomena of derivation and composition are often combined, can be found in translations from Arabic of the 99 “beautiful names” of God (al-ʾasmāʾ al-ḥusnā), which—although they were elaborated in the elitist milieu of advanced Islamic education—are so ubiquitous in devotion and in magico-religious practices of Bambara Muslims that they penetrated even into the ritual lexicon of non-Muslims.29 For instance, the divine name Māliku l-Mulki (lit. ‘Master of the Kingdom’) is literally translated as Masayamarabaga, a noun compounded of two derived nouns, formed from masa (‘king’), -ya (abstraction sufffijix), mara (‘to govern’) -baga (agent sufffijix), according to the model of Bambara determinative phrases. Similarly, ʿĀlimu l-ġaybi (lit. ‘Knower of the unknown’, a divine appellation not included in the canonical list) is literally translated as Gundodogolendɔnbaga, a determinative phrase compounded of a qualifying phrase and a derived noun, from gundo (‘secret’), dogo (‘to hide’), -len (past participle sufffijix), dɔn (‘to know’) and -baga (agent sufffijix): none of these elements is borrowed from Arabic. There are also calques where the three phenomena of composition, derivation, and morphological reanalysis, are combined. Thus, the compound noun adamaden (‘son of Adam’, hence ‘human being’),30 formed from the fijirst name Ādam (clearly entered into Bambara from the Koran) and the Bambara noun den (‘child’) has the doublet bunadamaden, where the fijirst element, bun, is the most frequent phonological outcome of the Arabic ibn (‘child, son’) in the adaptation to Bambara phonology of Arabic genealogical chains. Seemingly, the Arabic construct state ibn ʾĀdam was 28  -dennin is in fact a juxtaposition of two sufffijixes: -den and -nin. Although –den in itself is a noun, at least originally, Bailleul (2007: 10; 94) regards it, in some contexts, as a true sufffijix whose function is, among others, “concrétiser un mot de sens plutôt abstrait” or to denote the part of a whole. 29  See Tamari (2001: 99–105). 30  This noun (together with its derived forms) is attested in several variants among which is hadamaden, where the phoneme h looks like a hypercorrection, due to its frequency in borrowings from Arabic. As notes Dumestre (2004: 53), who elsewhere (1983: 13) suggests that the phoneme itself is a borrowing from Arabic: “une énergique frication de cette consonne semble être comme une façon de signaler l’appartenance au monde pieu”.

when arabic resonates in an african language

241

borrowed as a whole, and the Bambara doublet den was added later on to its element bun, whose meaning is no more transparent to speakers. From this couple of compound nouns the (also synonymous) abstract nouns, respectively, adamadenya and bunadamadenya were formed with the meaning ‘humanity, solidarity, belonging to the human kind’. It is worth mentioning in passing that all such terms are at least as current in Bambara as is ibn ʾĀdam in spoken Arabic (and, in my experience, even more than is the case in classical Arabic). Instances of Borrowing of an Unanalyzed Unit We have already seen some examples, including the last one, where a borrowing includes several lexical or grammatical morphemes which are integrated as an unanalyzed whole. This is most often a morpheme (an article, a flectional or, exceptionally, a derivational morpheme) which is reanalyzed as an integral part of the borrowed lexeme (as in the case of tāʾ marbūṭa above in proper names such as Ayisatu, or even in common nouns such as suratu ‘sura’). The most frequent cases of this phenomenon are occurrences with an Arabic article embedded in the nominal borrowing, even if most borrowings from Arabic entered into Bambara without an article. However, whether the article is embedded or not cannot be linked to a constant of a phonological, morphological or semantic nature, or to the ‘bookish’ or ‘spoken’ historical and social context where the borrowing occurred. A very clear example is given by the names of weekdays, most of which do not embed the article (cf. ntɛnɛn, from [al-] ʾiṯnayn, araba < [al-]ʾarbiʿāʾ, tarata < [al-]ṯulaṯāʾ, juma < [al-]jumʿa, sibiri < [al-]sabt), with only two exceptions: alamisa, from al-xamīs and the rarer alaadi or lahad, from al-ʾaḥad, which is most often replaced by the native word kari,31 or by the Gallicism dimansi. As can be seen, the belonging of all these lexemes to a same lexical series, besides the lack of phonological specifijicities in the case of both exceptions discourage any attempt to explain the phenomenon. In this regard, it can be useful to recall a comparative remark by Sarah Thomason (2007: 669):

31  Vydrine (1996: 71) argues that even this term might derive from Arabic al-ʾaḥad, through the following phonological shifts: ḥ > k and d > r. Should this hypothesis be confijirmed, it would show even further how unpredictable is the incorporation of the article or lack of it in an Arabic loanword, since we would have two doublets of a same Arabic noun, one retaining the article, and the other leaving it aside.

242

francesco zappa The frequent pattern in which an Arabic noun is borrowed with the Arabic defijinite article al- (or one of its allomorphs) attached is not evidence of Arabic morphological interference in the receiving language. The reason is that the Arabic morpheme and the following nominal are borrowed as an unanalysable whole. In none of the languages that have borrowed from Arabic is any productive use of the Arabic article mentioned as a separate morphosyntactic element; instead, it is a mere phonological part of the noun in the borrowing language.

In my opinion, similar, even if not identical, considerations can be made about the form in which the Arabic article was embedded, even if variations in this phenomenon in other languages gave rise to a lively debate among scholars. For instance, as to borrowings from Arabic into Hausa, Greenberg (1947) claimed that lexemes embedding the whole article alcame directly from Classical Arabic through the malamai (Hausa ʿulamāʾ) class, while lexemes embedding the lateral phoneme l- only (perhaps with the epenthesis of a vowel –i-) came from Maghrebi spoken Arabic through oral contacts with North African traders. This claim was later strongly refuted by Hiskett (1965), but was presented again recently by Abu-Manga (1999: 56–59) in his study on spoken Sudanese Hausa. The same variation can be found in borrowings from Arabic into Bambara, although in my opinion it can be explained by phonological, rather than historical and social, reasons. In fact, in most cases where the whole article al- is embedded (normally with the epenthesis of a vowel –i-), the Arabic lexeme begins almost always with a plosive, while in cases where only the lateral l- is embedded, the Arabic lexeme begins with a phoneme which tends to be elided in its passage into Bambara: voiced pharyngeal fricative ʿayn, voiceless pharyngeal fricative ḥāʾ, voiceless glottal plosive hamza, voiceless glottal fricative hāʾ. Besides, it would be difffijicult to claim that such words as lahara (‘the other life’, < Ar. al-ʾāxira) are popular borrowings from a Maghrebi dialect. While borrowings from a set of unanalyzed morphemes are relatively frequent, I do not think we can fijind occurrences of true morphological reanalysis which are productive in new contexts. The most puzzling case I have ever seen is rather closer to false etymology: it is the name for a ritual mask, associated to Bambara “traditional religion”, which represents a sacred cow. This name, bakaramisi, was interpreted by informants of the ethnologists who studied it as a juxtaposition of the roots ba (‘mother’), kara (‘perfect circle’, a symbol of the Creator Spirit of the traditional religion, on its turn a possible borrowing from the Arabic kura) and misi (‘cow’); according to Tamari, it is more likely an a posteriori

when arabic resonates in an african language

243

reinterpretation of the juxtaposition of the Arabic word baqara (‘cow’, which doubtless entered by suggestion of the homonymous Koranic sura) and its Bambara equivalent misi.32 More signifijicant cases of unanalyzed en bloc borrowings involve whole Islamic formulas (formed by phrases, or even by true propositions) which give rise to interjections in Bambara, not without some semantic shifts. Thus, ʾastaġfijiru Llāh (‘I seek forgiveness from God’, also used in Arabic as a formula of modesty) give rises to safurulayi (‘pardon!’, but also ‘may God save me from that!’, which corresponds rather to the Arabic ʾaʿūḏu bi-Llāh), subḥān Allāh (‘Glorious is God!’), to subahanalayi (exclamation of amazement, but also of reprobation)33 and even the fijirst part of the Muslim profession of faith, lā ʾilāha ʾillā Llāh (‘there is no divinity but God’), gives rise to the interjection of amazement laayilaa or laahawulaa. There exist even Bambara verbs which derive from Islamic pious formulas, such as the formula bi-smi Llāh (‘in the name of God’), which turned into a reflexive verb, i bisimila, with the meaning of ‘to settle down, to make oneself comfortable’, or a transitive one, bisimila, with the meaning of ‘to welcome’. Grammatical and Semantic Shift of Loanwords, Semantic and Phraseological Calques, Periphrastic Re-Translations While these examples, no matter how extravagant they may sound, remind us of similar instances of lexicalized catachresis that are current in some Arabic dialects,34 cases of lexical category change are in fact quite frequent in the passage of lexemes from Arabic into Bambara, thanks to the capacity of Bambara free roots to perform easily several grammatical functions, which, on the other hand, are unambiguously marked by the position of elements within the clause. Thus, for example, a verb such as daamu (‘to enjoy something for long’), which according to Lagarde would derive from the Arabic dāma ‘to last’, has a nominal homophone, unsurprisingly with the meaning of ‘lasting pleasure’. Similarly, both the

32  See Tamari (2001: 101–102), who, moreover, detected traces of an Islamic influence in some oral commentaries which are associated with as many esoteric symbols carved on this mask. 33  The noun subahana, which means something extraordinary (though with a pejorative nuance), is also coined from this phrase, probably through abbreviation rather than reanalysis. 34  I am indebted to the anonymous reviewers of this article for drawing my attention to this parallel.

244

francesco zappa

Bambara noun waajibi ‘obligation’ and the homophone verb waajibi or its derivative waajibiya (‘to force’, the sufffijix –ya when added to a verbal base does not change the meaning) are derived from the Arabic active participle wājib ‘compulsory’. From the Arabic noun wird (‘litany of the Sufiji brotherhoods’) are derived the Bambara noun wurudi (same meaning, but also ‘Islamic rosary’ to accompany the recitation of wird) and a homophone verb (with the meaning of ‘to recite the wird’ as well as of ‘to fijinger a rosary’). The Bambara word lakika, which works as a noun (‘truth’) and as an adjective (‘authentic’) at the same time, 35 derives from al-ḥaqīqa ‘reality, truth’, which is a true noun in Arabic. It is doubtless thanks to this polyfunctionality of Bambara words that verbal borrowings from Arabic can occur according to all the three main schemata identifijied by Versteegh:36 borrowing of an Arabic verb, borrowing of a (verbal or non-verbal) noun which is later integrated into the morphosyntax of Bambara verbs, borrowing of a verbal noun which is combined with a dummy verb, in particular kɛ, which on the other hand often performs this function in Bambara. And it is because of this polyfunctionality that the original meaning of a borrowed word might not exist anymore, as happens in the case of the Arabic noun ʿabd (‘servant’, but also ‘worshipper’), which is most probably the source of the Bambara verb bato ‘to worship, to venerate’.37 The latter gave rise to a homophone noun which has, however, the abstract meaning of ‘worship’, even though it is not derived from the corresponding Arabic derived noun (ʿibāda). Besides, passing from the category of concrete nouns to the category of abstract nouns (or the other way round) may happen even if the borrowing stays within the same grammatical category. Thus, the Arabic word ʾislām gives rise to the Bambara silamɛ, not in the meaning of ‘Islam’, but rather in the sense of ‘Muslim’; to translate the concept of ‘Islam’ into Bambara it is needed to add the abstraction sufffijix –ya, yielding the derived noun silamɛya.

35  In the latter meaning, it belongs, more precisely, to a lexical category classifijied among adjectives by Vydrine (1999: 23) and among “noms dépendants” by Dumestre (2003: 54). 36  See Versteegh, this volume. The two other schemes outlined in that article (i.e. incorporation of “bare verbs” and of “inflected verbs”), as instances of rather extreme phenomena such as, respectively, pidginization and language shift, are of course not found in the Bambara texts I examined. 37  This might also derive directly from the verb ʿabada, but the fijinal vowel of bato makes a derivation from the noun ʿabd-u (incorporating, as usual, the inflectional desinence –u) more likely.

when arabic resonates in an african language

245

A particularly interesting case, always within the category of nouns, is the Arabic tamr ‘date-palm’, which appears in Bambara in the pre-nasalized form ntamaro. Since Bambara words which begin with a pre-nasalized consonant are mostly plant names, which lead some linguists to see in it the remnants of a nominal class system, this might be an instance of a borrowing integrated within the remnants of such a classifijication system.38 A diffferent case is that of the Arabic noun sabab ‘cause’, which—after entering into Bambara with the two synonym forms sabu and sababu— yielded the phrase sabu la, with a causal value, which was later transformed, through the fall of the postposition la and a grammaticalization process, into a simple causal conjunction (sabu, in the meaning of ‘because’). This example of grammaticalization shows how far Arabic words can reach in their integration into the morphosyntactic structure of Bambara; on the other hand, the interference of Arabic can even extend to grammatical words. This happens most often in learned contexts such as the translation of Arabic texts, especially if they are scriptural texts, which tend to be translated word for word. Thus, in translating the Koranic expression al-maġḍūb ʿalay-him, al-Ḥājj Modibo Diarra, who is currently publishing in installments the fijirst written Bambara translation of the Koran, uses the expression funu bɛ minnu kan, in which the postposition kan (lit. ‘on, upon’) is calqued mechanically on the Arabic ʿalā, while the corresponding Bambara verb funu normally governs the postposition kɔrɔ.39 Beside the mechanical substitution of grammatical words, other instances of phraseological calques can be detected, even in everyday language: it is the case of phrases such as baasi tɛ (‘not bad’), which calques the Arabic phrase lā baʾs by setting a borrowing from Arabic (baʾs > baasi) into a syntactic context which corresponds, except for word order, to the original expressions. On the other hand, learned translations sometimes involve also semantic calques. A clear example can be found in the tendency to literally calque the repertoire of “medical” metaphors under which ʾaḥādīṯ are classifijied in Arabic according to the degree of authenticity assigned to them: thus, 38  Without explictly talking about nominal classifijication, Dumestre (2004: 51–52) considers prenasalization as a linguistic feature typical of “l’univers de la brousse” or, more precisely, of what he calls “le pôle bámanan” (opposed to “le pôle móri”), and observes that it is no chance that “aucun emprunt à l’arabe (en dehors de ntɛnɛn, ‘lundi’) ne connaît de nasalité initiale”. It should also be recalled that in Mali, dates are perceived as quintessentially Islamic food. 39  See Diarra (n.d.: 20); an alternative translation, using the standard postposition kɔrɔ, was however adopted by Diarra in a later version (2009: 2).

246

francesco zappa

metaphoric Arabic expressions, such as ṣaḥīḥ (‘sound’, hence in this context ‘authentic’) or ḍaʿīf (‘weak’, hence ‘unreliable’), are translated by Bambara expressions which have the same literal meaning (specifijically, kɛnɛ and lafulen) and which therefore take also the metaphoric meaning of their models.40 In other cases of neology, back-translation of the new concept is preferred to calque or borrowing: an example of that is the word misiriwele above (lit. ‘mosque call’), which translates the Arabic ʾāḏān. Analogously, the month of ramaḍān is called in Bambara sunkalo, lit. ‘month of fasting’, a noun compounded from sun (derived from Arabic ṣawm) and kalo (an indigenous word which means ‘moon’ and, by extension, ‘month’). Through a further passage, the month of šaʿbān, which precedes ramaḍān, is called sunkalomakɔnɔ (lit. ‘waiting for the month of fasting’). Sometimes the backtranslation is in sharp contrast with the etymology of the Arabic reference word. Thus, the word rakʿa, ʻunit of measureʼ of the canonical prayer, which denotes the set of positions assumed by the person praying, and which derives from the same root of rukūʿ (‘inclination of the body’ during the said prayer), is translated into Bambara with jɔko, a noun compounded from the root jɔ ‘to rise, to stand up’): the original word and its translation take then as reference point two diffferent postures among those assumed during a rakʿa. It happens also, for instance in some translations of divine names, that some exegetical elements are added to the borrowing, or are back-translated. It is the case of translations of ar-Raḥmān and ar-Raḥīm by, respectively, hinɛlabɔlentigi and hinɛkɛrɛnkɛrɛnnentigi, that is ‘Lord of the overwhelming grace’ and ‘Lord or the reserved grace’ (the word hinɛ is here a borrowing from Arabic ḥinna used instead than its Arabic synonym raḥma): these terms evoke an interpretation, relatively widespread among classical exegetes, according to which the name ar-Raḥmān hints at a grace granted to all men without exception in this world, while ar-Raḥīm refers instead to a grace reserved to Muslims in the hereafter.41 Lack of space prevents me from discussing in detail diffferent types of shift of meaning, which—as we saw in some examples—happen in the passing of lexemes from a language into the other one. In this regard, let it sufffijice here to point out that while in the case of calques created in 40

 See Tamari (1996: 54) and Zappa (2004: 75–76). Moreover, according to Tamari (2005b: 440 n. 16) the etymology of laafulen (past participle of laafu) is directly linked to the Arabic ḍaʿīf. 41  For a detailed explanation, see Tamari (1996: 55–56; 68; 70 fff.).

when arabic resonates in an african language

247

a learned context the prevailing tendency is to attribute a metaphoric, abstract meaning to words with a concrete meaning (as we have just seen in the case of technical terminology about ʾaḥādīṯ), in the case of borrowings integrated into everyday language an inverse tendency prevails to secularize the meaning of words from the religious semantic fijield. Thus, for instance, barika, from the Arabic baraka, takes—in addition to the original meanings of the Arabic word—the more mundane, but more usual, meaning of ‘physical force’. Sariya, derived from šarīʿa, took the generic meaning of ‘law, rule’, not necessarily Islamic. Hukumu, from the Arabic ḥukm, lost in current speech its religious and juridical meaning and took the meanings of ‘authority’ and ‘aegis’. Such examples could easily be multiplied.

References Abu-Manga, Al-Amin. 1999. Hausa in the Sudan. Process of Adaptation to Arabic. Köln: Rüdiger Köppe Verlag. Bailleul, Père Charles. 2007. Dictionnaire bambara-français. Troisième édition corrigée. Bamako: Éditions Donniya. Baldé, Abdoulaye. 1980. “Contribution à l’étude de l’impact culturel des emprunts arabes dans les langues manding du Sénégal”. Itinérances . . . en pays peul et ailleurs. Mélanges réunis par les chercheurs de l’ERA 246 du CNRS à la mémoire de Pierre Francis Lacroix. I: Langues, 195–209. Paris: Société des Africanistes. Bausani, Alessandro. 1967. “Per una letteratura comparata delle lingue islamiche”. Atti del terzo congresso di studi arabi e islamici. Ravello, 1966, 145–156. Naples: Istituto Universitario Orientale. ——. 1981. “Le lingue islamiche: interazioni e acculturazioni”. Il mondo islamico tra interazione e acculturazione, ed. Alessandro Bausani and Biancamaria Scarcia Amoretti, 3–19. Rome: Istituto di Studi Islamici. Bouwman, Dinie. 2005. Throwing Stones at the Moon: The Role of Arabic in Contemporary Mali. Ph.D. dissertation, Leiden University. ——. 2006. “Bambara”. Encyclopedia of Arabic Language and Linguistics, ed. Kees Versteegh. I: A-Ed, 255–257. Leiden: Brill. Brenner, Louis. 2001. Controlling Knowledge. Religion, Power and Schooling in a West African Muslim Society. Bloomington: Indiana University Press. Canut, Cécile. 1996. Dynamiques linguistiques au Mali. Paris: CIRELFA—Agence de la francophonie, difffusion Didier érudition. Delafosse, Maurice. 1910. “Les noms des Noirs musulmans du Soudan Occidental”. Revue du Monde Musulman 12/10: 257–261. ——. 1929. La langue mandingue et ses dialectes (Malinké, Bambara, Dioula). Tome 1: Introduction, grammaire, lexique français-mandingue. Paris: Librairie orientaliste Paul Geuthner. ——. 1955. La langue mandingue et ses dialectes (Malinké, Bambara, Dioula). Tome 2: Dictionnaire mandingue-français. Paris: Librairie orientaliste Paul Geuthner. Diarra, El-Hadji Modibo. N.d. Kuranɛ kɔrɔ bamanankan na. Suran 17. Bamako. ——. 2009. Kuranɛ kɔrɔ bayɛlɛmanen bamanankan na Rub’u Yaasiin. Bamako.

248

francesco zappa

Dumestre, Gérard. 1983. “Note à propos des termes bambara empruntés à l’arabe”. Langue arabe et langues africaines. Mémoire spécial du Centre d’Études sur les Relations entre le Monde Arabe et l’Afrique et du Centre d’Études sur l’Océan Indien Occidental, 13–20. Paris: Conseil International de la Langue Française. ——. 1994. “Introduction. La dynamique des langues au Mali: le trinôme langues régionales—bambara—français”. Stratégies communicatives au Mali: langues régionales, bambara, français, ed. Gérard Dumestre, 3–11. Paris: Agence de Coopération Culturelle et Technique (ACTT), difffusion Didier érudition. ——. 2003. Grammaire fondamentale du bambara. Paris: Karthala. ——. 2004. “Note sur un tropisme móri / bámanan en bambara”. Mandenkan. Bulletin d’études linguistiques mandé 39: 47–54. Greenberg, Joseph H. 1947. “Arabic loan-words in Hausa”. Word 3: 85–97. Hamès, Constant. 1997. L’art talismanique en Islam d’Afrique Occidentale. Analyse anthropologique et islamologique d’un corpus de talismans à écritures. Thèse de Doctorat, Paris: École Pratique des Hautes Études. Hassane, Moulaye. 2008. “Ajami in Africa: the use of Arabic Script in the transcription of African Languages”. The Meanings of Timbuktu, ed. Souleymane Bachir Diagne & Shamil Jeppie, 109–122. Cape Town: Human Sciences Research Council Press. Hiskett, Mervyn. 1965. “The historical background to the naturalization of Arabic loanwords in Hausa”. African Languages Studies 6: 18–26. Lagarde, Michel. 1988. Influences de l’arabe sur la langue bambara. Falajè (Mali): Centre d’Étude de la Langue Bambara. Schulz, Dorothea E. 2003. “‘Charisma and Brotherhood’ Revisited: Mass-Mediated Forms of Spirituality in Urban Mali”. Journal of Religion in Africa 33/2: 146-171. Soares, Benjamin F. 2005. Islam and the prayer economy. History and authority in a Malian town. Edinburgh: Edinburgh University Press. Tamari, Tal. 1994. “Cinq textes bambara en caractères arabes: présentation, traduction, analyse du système graphique”. Islam et Sociétés au Sud du Sahara 8: 97–121. ——. 1996. “L’exégèse coranique (tafsîr) en milieu mandingue. Rapport préliminaire sur une recherche en cours.” Islam et Sociétés au Sud du Sahara 10: 45–79. ——. 2001. “Notes sur les représentations cosmogoniques dogon, bambara et malinké et leurs parallèles avec la pensée antique et islamique”. Journal des Africanistes 71/1, Numéro spécial: Les empreintes du renard pâle: Pour Germaine Dieterlen, ed. MarcHenri Piault et Joëlle Hauzer, 93–111. ——. 2002. “Islamic Higher Education in West Africa: Some Examples from Mali”. Islam in Africa, ed. Thomas Bierschenk and Georg Stauth, vol. 4: Yearbook of the Sociology of Islam, ed. Helmut Buchholt and Georg Stauth, 91–128. Münster: Lit Verlag. ——. 2005a. “À la confluence des traditions orale et islamique: «Soumba et Lansiné»”. Oralité africaine et création. Actes du colloque de l’International Society for Oral Literature in Africa (10–12 juillet 2002), ed. Anne-Marie Dauphin-Tinturier and Jean Derive, 235–266. Paris: Karthala. ——. 2005b. “La prose littéraire arabe en traduction bambara: une maqāma d’al-Harīrī”. Paroles nomades. Écrits d’ethnolinguistique africaine en hommage à Christiane Seydou, ed. Ursula Baumgardt and Jean Derive, 431–463. Paris: Karthala. ——. 2006. “L’enseignement islamique traditionnel de niveau avancé: cursus, pédagogie, implications culturelles et perspectives comparatives”. Mande Studies 8: 39–62. Thomason, Sarah G. 2007. “Language Contact”. Encyclopedia of Arabic Language and Linguistics, ed. Kees Versteegh. II: Eg-Lan, 255–257. Leiden: Brill. Versteegh, Kees. 2001. “Linguistic Contacts between Arabic and Other Languages”. Arabica 48/4: 470–508. ——. “Heavy and light borrowing of Arabic verbs”. This volume, p. 215–228. Vydrine, Valentin. 1996. “Sur le «dictionnaire Nko»”. Mandenkan. Bulletin d’études linguistiques mandé 31, 59–75.

when arabic resonates in an african language

249

——. 1998. “Sur l’écriture mandingue et mandé en caractères arabe (mandinka, bambara, soussou, mogofijin)”. Mandenkan. Bulletin d’études linguistiques mandé 33: 1–87. ——. 1999. Manding-English Dictionary (Maninka, Bamana). Vol. 1: A, B, D-DAD, supplemented by entries from subsequent volumes. St. Petersburg: Dimitry Bulanin Publishing House. ——. 2007. “Reciprocal and Sociative Constructions in Bambara”. Reciprocal Construction. 4: 1912–1940, ed. Vladimir P. Nedjalkov. Amsterdam: Benjamins, 1915–1940. Zappa, Francesco. 2004. L’islamizzazione della lingua bambara in Mali. Tra pubblicistica scritta e letteratura orale. Supplement n. 2 of Rivista degli Studi Orientali 77. ——. 2009a. “Écrire l’islam en bambara. Lieux, réseaux et enjeux de l’entreprise d’al-Hâjj Modibo Diarra.” Archives de sciences sociales des religions 147: 167–186. ——. 2009b. “Popularizing Islamic knowledge through oral epic: a Malian bard in a media age.” Die Welt des Islams 49/3–4: 367–397.

INDEX OF NAMES

ʿAbbās b. al-ʾAḥnaf, ʾAbū l-Faḍl al-ʿAbbās b. al-ʾAḥnaf (d. after 193/808) 125 ʾAbū ʿAlī, al-Ḥasan b. ʿAlī al-Fārisī (d. 377/ 987) 89–90, 90 n. 20, 117–118, 121, 121 n. 28, 141 Abu Deeb, Kamal 127 nn. 43, 45, 47, 142 ʾAbū Duʾād Juwayra b. al-Ḥajjāj al-ʾIyādī (6th c.) 135 ʾAbū l-Ḥusayn, al-Fārisī (4th/10th c.) 89 Abu-Manga, Al-Amin 223, 227, 242, 247 ʾAbū Tammām Ḥabīb b. ʾAws (d. 231– 2/845–6) 115, 121 n. 29, 141–142 Aguadé, Jordi 218, 227 ʿAlī b. ʾAbī Ṭālib (d. 40/660), 64, 87 n. 12, 88 Al-Onaizan, Yaser 198 n. 9, 210 ʿAlqama b. ʿAbada t-Tamīmī (6th century) 118 Ambros, Edith 122 n. 30, 142 ʾĀmidī, ʾAbū l-Qāsim al-Ḥasan b. Bišr al- (d. 370/981) 112, 125 n. 38, 141 ʿAmīdī, ʾAbū Saʿd Muḥammad b. ʾAḥmad al- (d. 433/1042) 124 n. 37, 125, 134 n. 70, 135, 135 n. 74, 136, 136 nn. 75–77, 137, 137 nn. 79–80, 138, 138 nn. 82–84, 139 nn. 85–86, 141 Arazi, Albert 114 n. 7, 142 Ardašīr I b. Bābak, Sasanid king (d. ca. 224) 123 n. 33 Aristotle 52, 52 n. 5, 53–55, 59, 68 Arnold, Werner 218, 227 ʾAʿšā, Maymūn b. Qays (d. after 625) 131 ʾAsad, Nāsir ad-dīn al- 142 ʾAšjaʿ Sulamī, al-ʾAšjaʿ b. ʿAmr as-Sulamī ʾAbū l-Walīd (2th/8th century) 124 ʿAskarī, ʾAbū Hilāl b. Sahl al- (d. 395/1005)  113, 117 n. 16, 123 n. 33, 125 n. 38, 126, 129 nn. 54, 56, 131 n. 58, 132 n. 62, 133 n. 64, 134 nn. 66–67, 141 ʾAṣmaʿī , ʾAbū Saʿīd ʿAbd al-Malik b. Qurayb al- (d. 213/828) 133 ʾAstarābāḏī, Raḍī d-dīn Muḥammad b. al-Ḥasan al- (d. 686–8/1287–9?) 3–4, 4 n. 1, 5, 12, 17 n. 5, 33–34, 34 n. 2, 35–36, 36 n. 6, 37, 37 nn. 8, 10, 38, 38 n. 13, 39–41, 41 nn. 18–19, 42, 42 n. 20, 43, 43 n. 22, 44, 45 n. 25, 47, 49, 56–62, 62 n. 14, 63–68, 74, 105, 105 n. 75, 106–107

ʾAstarābāḏī, Rukn ad-Dīn. See Rukn ad-Dīn al-ʾAstarābāḏī Ayoub, Georgine 71 n. 7, 74 n. 14, 82 Baalbaki, Ramzi 86 n. 8, 107 Backus, Ad 221, 227 Badīʿī, Yūsuf al- (d. 1073/1662–3) 138 n. 82, 141 Baġdādī, ʾAbū Bakr ʾAḥmad b. ʿAlī al-Xaṭīb al- (d. 463/1071) 114 nn. 8–9, 141 Bailleul, Père Charles 230 n. 7, 236, 240 n. 28, 247 Bailly, Anatole 172 n. 33, 192 Baldé, Abdoulaye 234 n. 19, 247 Baldwin, Timothy 210, 210 n. 17, 211 Bally, Charles 46 n. 28, 48 Bāqillānī, ʾAbū Bakr Muḥammad b. aṭ-Ṭayyib (Ibn) al- (d. 403/1013) 117 n. 16, 134 n. 66, 141 Baššār b. Burd, ʾAbū Muʿāḏ (2th/8th c.) 126, 132, 138 Bausani, Alessandro 231, 247 Bayhaqī, ʾIbrāhīm b. Muḥammad al(4th/10th c.) 114 n. 9, 141 Behnstedt, Peter 218, 227 Bender, Emily M. 210 Benveniste, Emile 59, 68 Bettini, Lidia 1, 5–7, 109, 122 n. 29, 142 Biarnay, Samuel 163 n. 23, 192 Bird, Steven 208 n. 15, 210 Bohas, Georges 7–9, 12, 63 n. 16, 64–65, 71 n. 6, 82, 147, 147 n. 1, 148 n. 2, 151, 151 n. 5, 154, 157 nn. 9–10, 161 n. 17, 163 n. 21, 166 n. 25, 168 n. 28, 169 n. 30, 170 n. 31, 175, 175 n. 34, 176 n. 35, 177, 178 n. 37, 184 n. 41, 186, 192 Bond, Francis 210, 210 n. 18 Bonebakker, Seeger Adrianus 110 n. 5, 134 n. 69, 142 Borg, Alexander 224, 227 Boumans, Louis 216, 222, 225, 227 Bouwman, Dinie 233 n. 17, 234, 234 n. 19, 235 n. 21, 236, 247 Brants, Thorsten 196, 210 Brenner, Louis 233 n. 17, 247 Brockelmann, Carl  149, 192 Brown, Jonathan A. 215, 227

252

index of names

Buḥturī, ʾAbū ʿUbāda l-Walīd b. ʿUbayd al- (d. 284/897) 116, 141 Canut, Cécile 229 n. 1, 247 Carter, Michael G. 82, 103, 107, 109 n. 2, 128 nn. 48–49, 142 Chao, W. 70, 82 Cheikh-Moussa, Abdallah 114 n. 7, 142 Chomsky, Noam 189–192 Cohen, Marcel 154, 154 n. 7, 192 Cowan, David 197 n. 5, 210 Darfouf, N. 151 n. 5, 192 Dat, Mihai 9, 147 n. 1, 192 Delafosse, Maurice 230, 230 n. 6, 231, 232 n. 13, 247 Désoulières, Alain 221, 227 Dévényi, Kinga 72 n. 11, 73, 74, 82 Diab-Duranton, Salem 192 Diab, Mona 210, 210 n. 18 Diarra, El-Hadji Modibo 247 Diʿbil, ʾAbū ʿAlī Muḥammad b. ʿAlī (d. 246/860) 136 Dik, Simon C. 70 n. 3, 82 Dorow, Beate 206, 211 Ducrot, Oswald 36, 48 Dumayrī, al-Mutawallī Ramaḍān Aḥmad ad- 122 n. 30, 142 Dumestre, Gérard 229 nn. 1–2, 230 n. 4, 233, 233 nn. 14–15, 240 n. 30, 244 n. 35, 245 n. 38, 248 Dupont-Roc, Roselyne 52 nn. 4–5, 53, 62 n. 14, 68 Edwards, Malcolm 217, 222, 227 Ernout, Alfred 160, 192 Farazdaq, Tammām b. Ġālib al- (d. 110–2/ 728–30) 118 Fārisī, ʾAbū ʿAlī al-. See ʾAbū ʿAlī Farrāʾ, ʾAbū Zakariyyāʾ Yaḥyā b. Ziyād al(d. 207/822) 72–74, 82 Fīrūzābādī, Abū ṭ-Ṭāhir Majd ad-dīn Muḥammad b. Yaʿqūb al- (d. 711/1311–12), author of al-Qāmūs al-muḥīṭ 153, 193 Fleisch, Henri 4 Flickinger, Dan 210 Frank, Richard M. 102 n. 66, 107, 109 n. 2, 142 Fujita, Sanae 210, 210 n. 18 Gardet, Louis, 85 n. 3, 108 Gardner-Chloros, Penelope 217, 222, 227

Ghersetti, Antonella  5–7, 85, 98 n. 45, 107 Gilliot, Claude 87 n. 10, 107 Greenberg, Joseph 223, 227, 242, 248 Grunebaum, Gustav Edmund von 109 n. 2, 132 n. 61, 142 Guillaume, Jean-Patrick 2–5, 17 n. 5, 45 nn. 25–26, 48–49, 63 nn. 15–16, 64, 68, 147, 192 Günther, Sebastian 114 n. 9, 142 Habash, Nizar 210, 210 n. 18 Hacioglu, Kadri  210, 210 n. 18 Haddāra, Muḥammad Muṣṭafā 135 n. 73, 142 Hādī, Mūsā al- Hādī ʾilā l-Ḥaqq, Abbasid caliph (d. 170/786) 132 n. 63 Halle, Morris 189–192 Hamès Constant 235 n. 22, 248 Hārūn Rašīd, Hārūn b. Muḥammad b. ʿAbd Allāh ar-Rašīd, Abbasid caliph (d. 193/809), 124 n. 37, 132 n. 63 Ḥassān b. Ṯābit b. al-Munḏir (d. 40–54/ 659–673?) 112, 113 n. 6, 121 n. 26, 126 n. 41 Hassane, Moulaye 248 Ḥātimī, ʾAbū ʿAlī Muḥammad b. al-Ḥasan al- (d. 388/998) 133, 133 nn. 64, 69, 134, 134 n. 66, 135 n. 72, 141 Heinrichs, Wolfhart 109 n. 2, 127 n. 47, 134 nn. 68–69, 135 n. 73, 136 n. 75, 138 n. 83, 142, 143 Ḥimmānī, al-ʿAlawī l-Kūfī l- (d. 228/ 842–3?) 135, 135 n. 74 Hirschfeld, Hartwig 115 n. 9, 143 Hiskett, Mervyn 242, 248 Hockett, Charles Francis 32 Hoedemaekers, Iris 225, 227 Hopkins, Simon 18 n. 10 Humbert, Geneviève 89, 107 Hurwitz, Solomon 154, 171, 192 Ibn ʾAnbārī, ʾAbū l-Barākāt Ibn al-ʾAnbārī (d. 577/1181) 107 Ibn ʿAqīl, ʿAbd Allāh b. ʿAbd ar-Raḥmān (d. 769/1367) 49 n. 2, 68 Ibn ʿAṯīr, Ḍiyāʾ ad-dīn ʾAbū l-Fatḥ Ibn al-ʿAṯīr, (d. 637/1239) 129 n. 56 , 132 n. 62, 141 Ibn Ḥājib, Jamāl ad-dīn ʾAbū ʿAmr Ibn al-Ḥājib (d. 646/1249) 33, 36, 40–41, 47, 49 n. 2, 56–58, 104 n. 72 Ibn Hišām, Jamāl ad-dīn ʾAbū Muḥammad (d. 761/1360) 49 n. 2, 68

index of names Ibn Jinnī, ʾAbū l-Fatḥ ʿUṯmān (d. 392/1002) 5, 7, 9, 18 n. 10, 20 n. 18, 23 nn. 39, 43, 32, 80–81, 88, 88 n. 13, 95 n. 31, 98, 107, 201–202, 206, 211 Ibn Manẓūr, Muḥammad b. Mukarram Jamāl ad-dīn ʾAbū l-Faḍl (d. 711/1311–12), author of Lisān al-ʿarab 118 n. 20, 141, 153, 193 Ibn Muʿtazz, ʾAbū l-ʿAbbās ʿAbd Allāh Ibn al-Muʿtazz (d. 284/897) 116–117 Ibn Qutayba, ʿAbd Allāh b. Muslim (d. 276/889) 126 n. 38, 129 n. 56, 141–142 Ibn Rašīq, ʾAbū ʿAlī Ḥasan (d. 456–63/ 1063–71?) 116, 116 nn. 13–15, 118 n. 23, 126 n. 38, 129 n. 54, 133 n. 64, 134 nn. 66–67, 141 Ibn Sarrāj, ʾAbū Bakr Muḥammad Ibn as-Sarrāj (d. 316/929) 3, 17, 21, 32, 58 n. 11, 68, 87, 87 n. 12, 89–90, 93, 107 Ibn Ṭabāṭabā, ʾAbū l-Ḥasan Muḥammad (d. 322/934) 115, 141 Ibn Ṭabāṭabā l-ʿAlawī, Muḥammad b. ʾIbrāhīm (d. 199/815) 139 Ibn Wakīʿ, ʾAbū Muḥammad al-Ḥasan b. Wakīʿ at-Tinnīsī (d. 393/1002–3) 138 n. 83, 141 Ibn Yaʿīš, Muwafffaq ad-Dīn ʾAbū l-Baqāʾ (d. 643/1245) 3, 20 n. 17, 23, 23 nn. 37, 32, 39, 49 n. 1, 53–55, 57, 59, 64–65, 68, 105, 105 n. 75, 106–107 Ibn Zamlakānī, Kamāl ad-dīn (d. 651/1253) Ibn az-Zamlakānī 129 n. 54, 141 Imruʾ al-Qays b. Ḥujr (d. ca. 550) 101, 116, 118, 120, 131, 133–135 Jāḥiẓ, ʾAbū ʿUṯmān ʿAmr b. Baḥr al- (d. 255/ 869) 114 n. 9, 115, 115 n. 12, 129 n. 56, 131 n. 58, 141–143 Jakobi, Angelika 221, 227 Joüon, Paul 176, 192 Jurafsky, Daniel 210 Jurjānī, ʾAbū Bakr ʿAbd al-Qāhir al- (d. 471/ 1078) 5–7, 60–61, 63 n. 15, 85–86, 86 n. 8, 87, 87 nn. 10–11, 88, 88 n. 14, 89–99, 101, 101 n. 62, 102, 104, 104 n. 74, 105–106, 106 n. 78, 107–108, 110, 110 n. 4, 112, 116 n. 13, 117, 117 n. 18, 118, 118 nn. 22–23, 119–120, 120 n. 25, 121, 121 nn. 26–27, 29, 122, 123 n. 33–34, 124 n. 36, 125, 125 n. 38, 126, 126 n. 42, 127, 127 nn. 44–47, 128 n. 51, 129 nn. 54, 56, 141–143

253

Jurjānī, Qāḍī. See Qāḍī Jurjānī Jurjānī, Šarīf. See Šarīf Jurjānī Kähler, Hans 225, 227 Katamba, Francis  190–191, 193 Kazimirski, Albert de Biberstein 153, 160, 169, 185–186, 193 Kenstowicz, Michael J. 189–191, 193 Khatef, Laïla  162, 193 Kilito, Abdelfattah 109 n. 3, 143 Kim, Ara 210 Klein, Ewan  208 n. 15, 210 Knight, Kevin 198 n. 9, 210 Kordoni, Valia 209 n. 16, 211 Kouloughli, Djamel 55 n. 8, 60, 68, 86 nn. 6, 9, 88 n. 15, 103 n. 70, 107–108 Kuno, Susumu 70, 70 n. 4, 82 Kuṯayyir ʿAzza b. ʿAbd ar-Raḥmān (d. 105/ 723) 126 Ladefoged, Peter Nielsen 190, 193 Lagarde, Michel 232 n. 13, 235 n. 21, 236, 243, 248 Lalande, André 45, 48 Lallot, Jean 53, 68, 52 nn. 4–5, 62 n. 14 Lambert, Mayer 156, 193 Lambrecht, Knud 70, 82 Lancioni, Giuliano 7, 9, 195 Lane, Edward William 19 n. 11, 193 Larcher, Pierre  3–4, 4 n. 1, 33, 87 n. 10, 90, 108, 118 n. 22, 143 Levin, Aryeh 1–3, 13, 17, 18 nn. 7, 9, 19 nn. 10–11, 13–14, 16, 22, 22 n. 26, 23, 24 nn. 44, 48–49, 26 n. 59, 27 n. 73, 29, 32, 42, 43 n. 21, 48–51, 68, 86 nn. 4–5, 105 n. 75, 108 Loper, Edward 208 n. 15, 210 Lotfabbadi, Leyla Naseh 222, 227 Loucel, Henri 99 n. 51, 108 Lucidi, Mario 102 n. 65, 108 Lyons, John 85 n. 1, 95 n. 31, 108 MacDonald, Douglas B. 85 n. 3, 108 Maeda, Shinji 178 n. 36, 190, 193 Mahdī, ʾAbū ʿAbd Allāh Muḥammad al-, Abbasid caliph (d. 169/785) 132 n. 63 Malmkjær, Kirsten 190, 193 Marzubānī, ʾAbū ʿUbayd Allāh Muḥammad al- (d. 384/994) 88 n. 12, 107, 113 n. 6, 118 n. 23, 119 n. 24, 126 nn. 38, 41, 131 n. 59, 141 Matras, Yaron 220, 227 McCarthy, John J. 156 n. 8, 193 McShane, Marjorie J. 69 n. 1, 70 n. 5, 82

254

index of names

Meillet, Antoine 160, 192 Mubārak, Māzin al- 68, 107, 122 nn. 30–31, 123 n. 32, 143 Mubarrad, ʾAbū l-ʿAbbās Muḥammad b. Yazīd al- (d. 286/900) 3, 5, 17, 21, 21 nn. 21–22, 22 nn. 27–28, 31–32, 23, 23 nn. 40, 42–43, 30, 25, 24, 24 nn. 46–47, 50–52, 28 n. 76, 32, 49, 51, 79–81, 87, 87 n. 12, 89, 105, 107, 109 n. 1, 123 n. 33, 125 n. 38, 131, 131 nn. 59–60, 132, 142 Muhallabī, Marwān b. Saʿīd al-Ḥusayn al(d. ca. 190/805) 138, 138 n. 84 Muḥammad Fārisī, Muḥammad ʾAbū ʿAlī, al-Ḥasan b. ʿAlī al-Fārisī (5th/11th century) 121 n. 28 Muqātil, Ibn Sulaymān b. Bašīr al-ʾAzdī (d. 150/767) 72, 82 Muslim b. al-Walīd al-ʾAnṣārī (d. 208/823) 125 Mutanabbī, ʾAbū ṭ-Ṭayyib ʾAḥmad b. al-Ḥusayn al- (d. 354/955) 125, 134, 136–139, 141 Muysken, Pieter 216–217, 220, 227 Muẓafffar Ḥusaynī, al-Muẓafffar b. al-Faḍl al-Ḥusaynī (d. 642/1244) 113 n. 6, 119 n. 24, 130, 142 Nābiġa Ḏubyānī, an-Nābiġa ḏ-Ḏubyānī (6th c.) 113 n. 6 Naess, Unn Gyda 219, 227 Nagao, Makoto 210 n. 18, 211 Naṣr b. Sayyār al-Layṯī l-Kinānī (d. 131/748) 136, 136 n. 78 Nöldeke, Theodor 18 n. 10, 32 Nyckees, Vincent 158, 193 Oepen, Stephan 210 Owens, Jonathan 3, 13, 43 n. 23, 45 n. 25, 48, 86 n. 5, 98 n. 42, 104 n. 74, 105 n. 75, 108 Park, Kabyong 226–227 Perry, John R. 216, 227 Peters, Johannes Reinier Theodorus Maria 109 n. 2, 143 Pottier, Bernard 154, 193 Qāḍī Jurjānī, ʿAlī b. ʿAbd al-ʿAzīz al-Qāḍī l-Jurjānī (d. 392/1001) 125 n. 38, 142 Qudāma b. Jaʿfar ʾAbū l-Faraj (d. 320–337/ 932–948?) 110, 110 n. 5, 112, 117, 133, 142 Qudāma b. Mūsā al-Jumaḥī (d. 153/770) 137, 137 n. 81

Quṭrub, ʾAbū ʿAlī Muḥammad b. alMustanīr (d. 206/821) 63 n. 17 Rambow, Owen 210, 210 n. 18 Ratnaparkhi, Adwait 196, 211 Rāzī, Faxr ad-dīn Muḥammad b. ʿUmar ar- (d. 606/1209) 117 n. 19, 124, 130 n. 57, 142 Renan, Ernest 157, 193 Retsö, Jan 188, 193 Richter, Tonio Sebastian 224, 227 Ricoeur, Paul  85, 103 n. 67, 108 Rifāʿī, Ṭalāl Jamīl, 110 n. 5 Romaine, Suzanne 226, 228 Rukn ad-Dīn al-ʾAstarābāḏī, al-Ḥasan b. Muḥammad (d. 715–8/1315–9?), 105 n. 75 Rummānī, ʾAbū l-Ḥasan ʿAlī b. ʿĪsā r-, (d. 384/994) 121–122, 122 nn. 30–31, 123, 123 nn. 32–33, 128, 128 nn. 48–52, 129, 129 nn. 53–56, 131 n. 58, 132, 141–143 Rundgren, Frithiof 102 n. 67, 103, 108 Sadan, Joseph 109 nn. 2–3, 143 Saguer, Abderrahim  152, 157 n. 10, 166 n. 25, 170 n. 31, 175 n. 34, 178 n. 37, 186, 192, 193 Sakkākī, ʾAbū Yaʿqūb Yūsuf b. ʾAbī Bakr as- (d. 626/1229) 127 n. 45, 142–143 Salm Xāsir, Salm b. ʿAmr al-Xāsir (d. 186/ 802) 132 Ṣammūd, Ḥammādī 127 n. 47, 143 Sanni, Amidu Olalekan 109 n. 1, 143 Šarīf Jurjānī, ʿAlī b. Muḥammad as-Sayyid aš- (d. 816/1413) 39–40, 41 n. 18, 42–43, 47, 47 n. 29, 104, 104 n. 73, 107 Saussure, Ferdinand de 93, 102, 102 nn. 65, 67, 106, 108 Schaefffer, Jean-Marie 36, 48 Schoeler, Gregor 114 nn. 7, 9, 143 Schulz, Dorothea E. 234 n. 18, 248 Schwarz, Nadja 219, 228 Serhane, Rachida 192–193 Sharofff, Serge 206, 211 Sībawayhi, ʾAbū Bišr ʿAmr b. ʿUṯmān b. Qanbar (d. 180/796?) 2–3, 5, 17–22, 22 nn. 26, 34, 23–24, 24 nn. 44, 48, 51–53, 25, 25 nn. 55–56, 26, 26 nn. 61, 62, 27, 27 n. 69, 28–32, 42, 49, 51, 71–72, 74–76, 76 n. 9, 77–82, 87 n. 12, 88, 88 n. 13, 89–90, 90 n. 20, 94, 94 n. 30, 95, 95 n. 33, 96 n. 35, 97, 103, 105, 107, 109 n. 2, 120 n. 25, 122, 123 n. 32, 142–143

index of names Simon, Udo Gerald 123 n. 35, 129 n. 56, 130, 143 Soares, Benjamin F. 248 Solimando, Cristina  4, 5, 69 Stetina, Jiri 210 n. 18, 211 Ṣūlī, ʾAbū Bakr Muḥammad b. Yaḥyā ṣ(d. 335/947) 114–115, 124 n. 37, 142 Ṯaʿālibī, ʾAbū Manṣūr ʿAbd al-Malik aṯ- (d. 429/1038) 123 n. 33, 142 Tahānawī, Muḥammad ʾAʿlā at- (12th/ 18th c.) 104, 104 n. 73, 107 Tamari, Tal 231, 231 n. 9, 232 n. 11, 233, 236 n. 25, 240 n. 29, 242, 243 n. 32, 246 nn. 40–41, 248 Ṭarafa ʿAmr b. al-ʿAbd b. Sufyān (5th c.?) 130–131 Theil, Rolf 219, 228 Thomason, Sarah G. 241, 248 Troupeau, Gérard 87 n. 10, 108 Tschacher, Torsten 221, 228 Tucker, A.N. 218, 228 Versteegh, Kees  10–11, 63 n. 17, 68, 72 n. 9, 82, 88, 90, 90 n. 18, 92 n. 26, 95 n. 31, 108, 215, 221–222, 224, 227–228, 236 n. 24, 237, 244, 244 n. 34, 247–248 Vydrine, Valentin 229 n. 2, 230, 230 n. 3, 231 n. 9, 241 n. 31, 244 n. 35, 248–249 Wellens, Ineke 217, 228 Widdows, Dominic 206, 211

255

Wright, William, 24 n. 44, 32 Xafājī, ʾAbū Muḥammad b. Sinān al(d. 466/1073) 110–112, 117 nn. 16–17, 142, 123 nn. 33, 35, 132 n. 62 Xalīl, al-Xalīl b. ʾAḥmad b. Tamīm al-Farāhidī (d. 160–175/776–791?) 19, 23, 94 n. 30, 114, 138 n. 84 Xansāʾ, Tumāḍir bint ʿAmr al(6th–7th c.) 129 n. 56, 134 Xuwārazmī, ʾAbū ʿAbd Allāh Muḥammad al- (4th/10th c.) 87, 107 Yāqūt ar-Rūmī (d. 626/1229) 110 n. 5, 142 Yeou, Mohamed 178 n. 36, 190, 193 Zajjāj, ʾAbū ʾIsḥāq ʾIbrāhīm b. as-Sarī az- (d. 311/923) 89 Zajjājī, ʾAbū l-Qāsim ʿAbd ar-Raḥmān b. Isḥāq az- (d. 337–40/948–50)  63 n. 17, 68, 87, 87 n. 12, 92 nn. 26, 27, 95 n. 31, 101 nn. 60, 62, 107–108 Zamaxšarī, ʾAbū l-Qāsim Maḥmud b. ʿUmar az- (d. 538/1144) 18 n. 8, 23 n. 39, 32, 44, 44 n. 24, 46 n. 27, 47, 49, 49 n. 1, 51–53, 55–58, 58 n. 11, 60, 62 n. 14, 63, 68, 89, 92, 104, 107 Zappa, Francesco 10–12, 229, 231 n. 10, 234 nn. 18, 20, 246 n. 40, 249 Zhang, Yi 209 n. 16, 211 Ziriklī, Xayr ad-dīn az- 135 n. 74, 137 n. 81, 138 n. 84, 143

SUBJECT INDEX

abstraction 67, 92, 102–103, 103 n. 67, 106 abstraction of state 237 abstraction sufffijix 237–238, 240, 244 adjective 9, 23, 128, 195 n. 1, 202, 222, 237, 239–240, 244, 244 n. 35 relative adjective 50, 64 adverb 9, 196, 202 afffijix 21, 40–41, 55, 65, 216, 225, 230, 235 derivational afffijix 237 flexional afffijix 237 African Languages 215, 221, 229 aggregate 10, 195, 200, 202, 204–205, 205 n. 13 allomorph 36 annexation 95, 95 n. 33, 96, 96 n. 35, 97 anthroponym 12, 232 Arabic Arabic grammatical terminology 41, 101 n. 62 Arabic grammatical tradition 2, 6, 34, 38, 49, 51, 63–64, 68, 71, 86, 102–103, 103 n. 70, 104, 196, 198 n. 8, 202 Arabic linguistic thinking 1, 3–4, 9, 69, 72, 195, 195 n. 1 Arabisms 230, 233–235 Arabization 215 article 9, 12, 40, 42–44, 52 n. 5, 200, 202–203, 206, 209–210, 241, 241 n. 31, 242 defijinite article 23, 29, 38, 40, 42, 44, 50, 60, 202, 242 articulated sound 57 articulation semantic articulation 41 back-translation 246 bilingualism extensive bilingualism 11, 223–224 full bilingualism 11, 223–224 Hausa/Arabic bilingualism 11, 223 Persian/Arabic bilingualism 223 borrowing 7, 10–12, 215–220, 224, 230, 230 n. 5, 233–236, 236 n. 25, 237–239, 240, 240 n. 30, 241–242, 244–246 borrowing context 218, 224 borrowing pattern 222 borrowing process 224

borrowing strategies 224 direct borrowing 10, 158, 221 from logic into grammar 52, 55, 62 n. 14, 66 heavy 215 heavy borrowing 11 in poetry 115–116, 116 n. 13, 132 n. 62, 134, 140–141, 158–159 integrated 245, 247 light 215, 222 light borrowing 11 of a grammatical morpheme 237 of an unanalyzed unit 241–243 re-borrowing 216 through commercial contacts 233 through written transmission 224 variants in borrowings 12 calque 229–230, 233–235, 239–240 phraseological and semantic 243–247 syntactic 239 case marker 21 case marking 6, 64, 74, 201 catachresis 243 category lexical category 71, 236, 243, 244 n. 35 categorematic terms 45 coalescence 94, 94 n. 30, 97, 196 code-mixing 216–217, 220, 222 German/English code-mixing 222 Swedish/Persian code-mixing 226 code-switching 216 Arabic/Coptic code-switching 224 Berber/Arabic code-switching 224 Hausa/Arabic code-switching 11, 223 Persian/Swedish code-switching 222, 225 Punjabi code-switching 226 composition composition pattern 238 phonosemantic composition 181 composition pattern 235, 239–240 compound, compound expression, compound(ed) word 40, 42, 59, 61–67, 94–96, 96 n. 35, 97, 216–217, 220, 223, 238–240, 246 compounding 12, 230, 235–243

subject index consonantal root pattern 8, 35, 37, 37 n. 8, 149, 151–153, 157, 187 constituent 43, 47, 70–71, 74–75, 79–81, 92, 153, 179 constraint contextual constraints 198, 210 pragmatic constraints 198 syntactic constraints 198 construct state 6, 230, 232, 238–240 construction DO-construction 11, 219–226, 228 possessive construction 206 n. 13 regressive construction 230 syntactic construction 127 context 7, 87, 92, 98–99 bilingual context 220, 225 extra-linguistic context 70, 76–77, 80 historical-social context 241 language contact context 225 learned context 225–226, 245, 247 linguistic context 75 migration context 225 religious context 220, 222 situational context 98 socio-cultural context 233 syntactic context 245 textual context 98 convention 99, 99 n. 51, 100–101, 104, 123 poetical convention 126, 130 criterion ʾAstarābāḏī’s criterion 42 consecution criterion 38 diagnostic criterion 225 for the determination of membership 29 morphological criterion 106 phonological criterion 106 semantic criterion 29, 91, 97, 106 declension 26, 87, 89, 91–92, 92 n. 26, 94, 94 n. 30, 95 n. 33, 97 declension marker, mark of declension 39, 94, 95, 95 n. 31, 96, 105 declension vowel 40 declensional status 94 diptotic declension 95, 95 n. 33 fijinal declension 95 implicit declension 88, 95 n. 33, 96 n. 35 intermediate declension 95–97 meaning of declension 91 triptotic declension, full declension 95, 95 n. 33 deixis 100

257

denativization 216 derivation 12, 41, 94, 103 n. 67, 128 n. 50, 219, 223, 230, 235–236, 238, 240, 244 n. 37 derivational process 225 derivational morpheme 241 derivational sufffijix 237–238, 239 n. 27 derivative 169, 184, 239 n. 27 determinative phrase 238–240 dictionary dictionaries of Bambara 230 technical dictionaries 104 traditional dictionaries 149 discrete unit 88–90, 92, 96, 99, 99 n. 51, 102, 105 elision 5, 51, 105, 118, 120, 129 n. 56, 203–204 eloquence 101, 109 n. 1 enantiosemy (antithetical polysemy) 158, 158 n. 11, 187–188 etymology 91, 239 n. 27, 246, 246 n. 40 false etymology 242 etymon 7–8, 147–174, 176–180, 182–185, 187–188 exegesis 73 feminine form 26 n. 62, 94 formal marker 91 formation populaire 215 formation savante 215 formative 40–41 government 86–87, 96 n. 37, 101 n. 62, 217, 220 grammar 5, 52, 71, 88 n. 15, 92–93, 95 n. 31, 97, 123 n. 32, 147, 202 discourse grammar 69 early Arabic grammar 17–18, 55, 60–61, 85, 92 n. 27, 109 n. 2 functional grammar 70 late Arabic grammar 3, 24 n. 44, 51–52, 89, 92 n. 27, 105 Orientalist grammar 9 origins of Arabic grammar 88 n. 12 principles of grammar 2 sentence grammar 69 grammarians Arab grammarians 2–3, 7, 17, 18 n. 7, 19, 23, 29, 43, 50–52, 52 nn. 3, 5, 53, 55 n. 8, 61–62, 64–66, 70, 86, 87 n. 12, 88, 90, 96 n. 37, 101 n. 62, 102, 103 n. 70, 104, 148, 148 n. 2, 188, 198, 201 Greek grammarians, 52 n. 5 grammaticalization 74, 245

258

subject index

head of a compound 239 phrasal head 70 homonymy 158–160, 162, 166–170, 174, 181, 187–188 homophone 237, 243–244 hypercorrection 240 n. 30 hypernymy 90 image (poetical) 7, 110, 110 n. 4, 112–113, 115, 115 n. 12, 116, 116 n. 14, 117, 125, 127 n. 47, 129 n. 56, 133, 137, 138 n. 83, 140 implication 128 inflection 12, 41, 219–220, 222, 226, 235 desinential inflection 40–42 inflection carrier 220 inflection ending 220 inflection system 220 morphological inflection 41 inflectional burden 220 inflectional desinence, 244 n. 37 inflectional morpheme 230, 236 n. 23 inflectional vowel 38, 40–41, 41 n. 18, 42–44 influence bookish influence 231, 233, 241 information 74, 90, 99–100, 102, 199, 206, 210 background information 71 contextual information 9 encyclopedic information 197 grammatical information 9 linguistic information 10 omitted information 69 shallow information 210 statistical information 207 syntactic information 197 integration morphological integration 11–12, 218, 224, 230 n. 5, 232, 235, 245 phonological integration 235 n. 21 interference 197, 245 morphological interference 235, 242 phonological interference 235 n. 21 structural interference 12, 235 interjection 243 Islam 10, 215, 221, 231 Islam(ic)ization 11, 233 linguistic Islamization 231 Islamic Islamic education 216, 232, 234, 240 Islamic formulas 243 Islamic learning 225, 236 n. 25 Islamic lexical stock 231

Islamic peoples 231 Islamic religious fijields 234 Islamic religious lexicon 230 Islamic scholars 216, 219 Islamic technical lexicon 239 Islamic theology 225 language contact 11, 222, 225, 231 language shift 11, 217, 224, 244 n. 34 language type agglutinative 230 fusional or inflecting 230 isolating 230 synthetic 230 lemma 235–236, 236 n. 25 lexeme, lexicographical unit 86 n. 7, 97, 103 n. 67, 235, 238–239, 241–243, 246 lexicalization 216 lexicography 1, 230 lexicon 9 Arabic lexicon 8, 10, 153, 158, 162, 187–188 Arabic loanwords 11 Bambara lexicon 229–230, 233 Hausa lexicon 223 learned lexicon 234 lexicon of non-Muslims 240 lexicon of the Romance languages 215 loanwords 216, 223 marabout lexicon 239 organization of the lexicon 8, 151, 153 Persian lexicon 220 regularities in the lexicon 195 specialized lexicon 196 n. 2 loanword 215, 223, 234 Arabic loanwords 11, 215–216, 220–224, 226, 229–230 English loanwords 226 fully integrated loanwords 234 grammatical and semantic shift 243 Latin/Greek loanwords 215, 224 re-Arabized loanwords 216 Mali 229 Mali Empire 229, 233 matrix 7–9, 150–153, 157, 163–164, 166–174, 176–188 meaning declinable meaning 91 grammatical meaning 92 lexical meaning 6–7, 92, 98, 100–102, 110 n. 4, 126–128, 129 n. 54 non-declinable meaning 91 poetical meaning 124

subject index sentence meaning 98, 102, 109 n. 2, 124, 124 n. 36, 129 n. 54, 140 single meaning 36, 91, 91 n. 21, 98 n. 43 word meaning 124 n. 36 meaningful element 2, 17–18, 20, 20 n. 19, 25–27, 29–31 meaningless element 2, 19–20 modifijier 239, 239 n. 27 morph 36 portmanteau morph 36 morpheme 1–2, 4–5, 9, 17, 17 n. 5, 18, 20, 20 n. 19, 29, 31, 33, 35–36, 49–50, 86, 105–106 bound, continuous morphemes 2, 4, 50, 196 bundle of morphemes 2 determination morpheme 9 discontinuous morphemes 4, 36–43 grammatical morpheme 237, 241 independent morphemes 50–51 lexical morpheme 241 morphemes classifijied as kalim 20–23 morphemes classifijied as zawāʾid 20, 23–27 division 27–29 morphology 1, 41, 62 n. 14, 71, 94, 103, 153 concatenative 38, 40–42 Fulfulde morphology 219 non-concatenative 36, 38, 40–41 morphophonology 97 morphosyntax 89, 244 nativization 235 neologism 215 learned neologisms 233 newly created neologisms 234 older neologisms 234 neology 246 nominal class 245 notional invariant 8, 150, 152–153, 163, 170, 172, 172 n. 32, 173–180, 184 noun 9, 18–24, 26–31, 33–34, 40, 42–45, 49–50, 52, 54, 56, 64, 80–81, 89–97, 106, 121, 128, 202, 207, 217, 222, 225, 236–237, 240, 242, 244–245 abstract nouns 237–238, 241, 244 borrowed nouns 221, 225 collective nouns 237 common nouns 232, 241 compound nouns 94–97, 238–241, 246 concrete nouns 238, 244 defijinite noun 24 n. 44, 121, 121 n. 26, 123 n. 34 derived nouns 238–240, 244 deverbative nouns 66 feminine noun 94

259

foreign nouns 217 indeclinable nouns 31 indefijinite noun 24 n. 44, 50, 120–121, 121 n. 26, 123 n. 34 masculine noun 94 noun determination system 44 noun phrase 44, 196 proper noun 23, 44 proper nouns 59, 122, 232 single noun 96 true nouns 244 verbal nouns 71, 222, 225–226, 239, 244 oral oral communication 229, 242 oral context 225 oral epic 235 oral interaction 231 oral literature 230, 234, 238 oral transmission 114 orality learned orality 11, 231 partes orationis 85 n. 2, 90 parts of speech 17–18, 49, 52, 87, 89–90 pattern 2, 23, 27–32, 37, 41, 66–68, 147–148, 153, 160, 206, 217–218, 220 phoneme 8, 71, 81, 85 n. 1, 94, 101, 103, 103 n. 67, 149–151, 153, 215, 236, 240 n. 30, 242 phonetic phonetic adaptation 158 phonetic adjustment 160 phonetic afffijinities 8 phonetic aspect 104, 148 phonetic chain 100 phonetic chains 99 phonetic component 99 phonetic constants 148 phonetic evolution 158 phonetic feature vector 153 phonetic features 8, 152–153, 189–191 phonetic form 93, 100–101, 104 n. 74 phonetic invariant 152 phonetic level 153 phonetic lightness 78 phonetic properties 149–151 phonetic relation 168, 171 phonetic unit 18–19 phonetic features 8 phonographematics 62 phonology 12, 103, 237, 240 pidginization 11, 224, 244 n. 34 plural broken plurals 66–67, 201, 235–236

260

subject index

sound plurals 105, 236 polysemy 158 antithetical polysemy. See enantiosemy postposition 245 pre-Islamic poetry 114 n. 7 pronoun 9, 21, 21 n. 23, 22, 22 n. 26, 29, 29 n. 78, 30, 46, 50–51, 54, 59 n. 12, 62, 65, 89, 205 n. 13 hidden pronoun 43, 46–47, 47 n. 29, 51 interrogative pronouns 201 pronoun sufffijix 22, 29–31, 44, 50, 55, 65, 205 n. 13 relative pronouns 200–201 radical 2, 8, 19, 30–31, 151, 153–157, 161, 161 nn. 15–16, 169, 181, 188 reanalysis morphological 237, 240, 242, 243 n. 33 rhetoric 1, 5–7, 60, 71, 86, 86 n. 8, 90, 97, 103–104, 106, 109, 109 nn. 1–2, 110 n. 4, 129 n. 54 semantic semantic constants 148 semantic properties 149–151 semantic relation 168, 171 semantic shift 12, 236, 243 semantics 88, 91, 97 grammar semantics 86 lexical semantics 69 sentence 45, 47, 69, 103 n. 67 bound sentence 46 nominal sentence 120 segmented sentence 46 sentence structure 69–72, 74, 81 theory of the sentence 45 verbal sentence 44–45 singulative 33, 57 speech unit 53–54, 57 stem 39, 39 n. 14, 47, 59 n. 12, 65, 154, 218 synonymy 158 syncategorematic terms 45 syntactic function 23 n. 44, 88 n. 14, 91–92, 101, 101 n. 62, 102, 104 syntax 1, 33, 41, 69, 71, 74, 97, 103 terseness 7, 123, 129, 129 n. 56, 130, 132 n. 62 texteme. See textual unit textual organization 88 n. 14, 98 n. 46, 100 textual unit 97, 103 n. 67, 106 Theory of Matrices and Etymons (tme) 7–9, 147–153, 157, 188

underlying form 104, 219 utterance 6–7, 20 n. 19, 33–34, 38, 43–47, 53–54, 59–60, 71 n. 7, 73–74, 79, 91, 100–103, 105–106, 110 n. 4, 111–113, 116–117, 121–122, 124, 126, 129 n. 54, 130, 132, 134, 140, 220 vehicular language 229 verb 2, 9–11, 18–38, 44–47, 49, 52, 55, 57–58, 60, 62, 64–66, 74–77, 80, 87, 89, 91–93, 106, 120, 128, 149, 151, 200–202, 205–208, 210, 215–226, 230, 236–238, 243–245 agentive verbs 219 bare verbs 11, 222, 224, 226, 244 n. 34 complex verbs 216, 220, 223, 225–226 doubly weak verbs 21 DO-verbs 220, 225 fijinite verbs 220 inflected verbs 11, 224, 244 n. 34 light (dummy, helping) verbs 10–11, 216–217, 219–222, 224–226, 244 reflexive verbs 243 simple verbs 221 stative verbs 219 strong verbs 219 weak verbs 218 word 1–10, 12, 17–21, 27–28, 33, 35–39, 41, 43, 45, 47, 49, 52, 52 n. 3, 53–54, 54 n. 7, 55–57, 57 n. 10, 58–67, 69, 71, 73, 80, 85, 85 n. 1, 86–88, 88 n. 14, 89–92, 92 n. 28, 93–94, 94 n. 30, 95, 95 n. 31, 96–106, 109 n. 2, 116 n. 13, 140, 147–153, 158–160, 187–188, 195–196, 200–201, 204–207, 216, 220, 230, 242–243 derived word 37 n. 8 grammatical word 52, 245 graphic word 50, 196 kernel word 54 missing word 72 non-aggregate words 10, 200 word boundary 6, 63 word class, class of words 7, 44, 61, 201 word formation 41, 94 n. 30, 128 n. 50, 217, 224, 239 word type 195–196, 201, 203, 207, 210 word order 230, 245 Bambara word order 230, 239 basic word order 230 freedom in word order 92 n. 26 German word order 222 SOV word order 230 VSO word order 230

INDEX OF ARABIC WORDS

ʿajamī 231 n. 9 ʿamal 72, 74, 86 ʿāmil (pl. ʿawāmil) 87 ʾaṣl 87, 94, 104–105, 111 ḥarf ʾaṣl (pl. ḥurūf ʾuṣūl) 19–20, 30–31 badal ḥarf al-badal (pl. ḥurūf al-badal) 19–20 balāġa 5–6, 86 n. 8, 98 n. 45, 109, 109 n. 1, 110–113, 122–123, 129 n. 55, 131, 141 balāġī (pl. balāġiyyūn) 6, 109–110, 110 n. 4, 112, 114, 117, 121–124, 124 n. 36, 133, 140–141 bayt bayt al-maʿnā (pl. ʾabyāt al-maʿānī) 109 dalāla 104 dalāla waḍʿiyya 130 n. 57 ḏ̣ ihār 71 n. 7, 79 faṣāḥa 101 fijiʿl 18, 87, 93, 100 ḥaḏf 5, 71, 74, 77–79, 80–81, 105, 129 nn. 54, 56 ḥaraka (pl. ḥarakāt) 18, 41, 220 ḥarf (pl. ḥurūf ) 1, 5, 18, 18 n. 9, 19–21, 23, 24 nn. 45, 53, 26, 34, 41, 46 n. 27, 62–63, 87, 90–91, 93, 100, 106, 196 ḥarf al-badal (pl. ḥurūf al-badal). See under badal ḥarf al-ʾiʿrāb. See ʾiʿrāb, ḥarf alharf ʾaṣl (pl. ḥurūf  ʾuṣūl). See under ʾaṣl ḥarf az-zāʾida (pl. ḥurūf az-zawāʾid). See under zāʾida ḥarf lafḏ̣ . See under lafḏ̣ ḥarf mabnā (pl. ḥurūf al-mabānī). See under mabnā ḥarf maʿnā (pl. ḥurūf al-maʿānī). See under maʿnā harf zāʾid. See under zāʾid ḥurūf az-ziyāda (pl. of ḥarf ziyāda). See under ziyāda ḥašw 133 ʾiḍāfa 26, 96, 127 ʾiḏ̣ hār 71 n. 7, 79 ʾiḍmār 5, 71–75, 77–81

ʾiḍrāb 71 ʾifāda 90 ʾifāda maʿnawiyya 130 n. 57 ʾifrād 94 ʾīġāl 133 ʾījāz 7, 123 n. 34, 129, 129 nn. 53–55, 130, 130 n. 57, 131–132 ʿilm ʿilm al-luġa 97 ʾiʿrāb 6, 17, 30–31, 87–88, 91–92, 92 nn. 27–28, 109, 117, 121–122 ḥarf al-ʾiʿrāb 94–95 ʾiʿrāb marker 88 lack of ʾiʿrāb 91 ʾirdāf 117 ʾišāra 100, 129 n. 54 ism 9, 18, 44, 87, 93, 100, 202 ism al-waḥda 33 ism kāna 120 ism muḏ̣ har 21, 21 n. 23 ism wāḥid 97 istiʿāra 126 n. 42, 127, 127 n. 47, 143 ištiqāq 128 , 128 n. 50 iʾtilāf 88, 90, 117 ʾiṭnāb 130 n. 57 jāʾiz 74 jumla 45, 47, 89 juzʾ 45, 90 n. 19 kalām 7, 33–34, 44–45, 54, 87 n. 12, 88, 88 n. 13, 89, 90, 110, 110 n. 4, 111–112, 128, 131 kalima, kalim 1–2, 4, 13, 17, 17 n. 5, 18, 20–23, 24 nn. 45, 53, 27–29, 29 n. 79, 30–38, 40–41, 41 n. 18, 42–43, 43 n. 22, 44–46, 46 n. 27, 47–52, 57, 62, 66, 68, 85–86, 86 n. 7, 87, 87 n. 12, 88, 88 n. 13, 89–90, 90 n. 19, 93, 94 n. 30, 103–106, 108 kalima mufrada 99 lafḏ̣ (pl. ʾalfāḏ̣ ) 6–7, 56–57, 59–60, 100–101, 104 n. 74, 106, 109–110, 110 n. 4, 112, 114–115, 116 nn. 12–13, 117, 122, 124, 127, 129, 129 n. 54, 130, 130 n. 57, 132, 137–138, 140–141 ḥarf lafḏ̣  41 lafḏ̣ mufrad (pl. ʾalfāḏ̣ mufrada) 58, 99

262

index of arabic words

lafḏ̣ a 43, 47, 53–54, 56–57, 104 lafḏ̣ iyya (f. of lafḏ̣ ī) 87 luġa 111, 112 ʿilm al-luġa. See under ʿilm

qiṣar 129 n. 56 qiyāsiyya (f. of qiyāsī) 87

mabnā ḥarf mabnā (pl. ḥurūf al-mabānī)  40–41 maʿdar 59 majāz 103–105 maʿnā (pl. maʿānī) 6–7, 55 n. 8, 59–60, 68, 109 n. 2, 101, 103–104, 106–107, 109–110, 110 n. 4, 111–116, 116 n. 12–13, 117–119, 121–123, 123 n. 34, 124, 124 n. 36, 125, 125 n. 38, 127, 129 nn. 54–55, 130, 130 n. 57, 132, 132 n. 62, 133–135, 135 n. 73, 136–138, 140–141 ḥarf maʿnā (pl. ḥurūf al-maʿānī) 40–41, 55 maʿānī n-naḥw (pl. of maʿnā n-naḥw) 101, 101 n. 62, 121, 122 n. 29 maʿnā ġayr munṣarif 91 maʿnā l-maʿnā 117 maʿnā mufrad 57, 91, 91 n. 21 maʿnā munṣarif 91 maʿnā t-taṣarruf 91 maʿnawiyya (f. of maʿnawī) 87 ʾifāda maʿnawiyya. See under ʾifāda maṣdar 36, 37 n. 8, 236–237 mubtadaʾ 118 mufrad kalima mufrada. See under kalima lafḏ̣ mufrad (pl. ʾalfāḏ̣ mufrada). See under lafḏ̣ maʿnā mufrad. See under maʿnā munṣarif maʿnā munṣarif, maʿnā ġayr munṣarif. See under maʿnā murakkab 53, 61

samāʿ 114 samāʿiyya (f. of samāʿī) 87 sariqa (pl. sariqāt) 7, 116, 116 n. 13, 133–140 ṣuḥūfī 114 n. 9

naḏ̣ m 98, 98 n. 46, 127 naḥw 72, 86 n. 8, 123 n. 32 maʿānī n-naḥw (pl. of maʿnā n-naḥw). See under maʿnā

riwāya 114–117

tablīġ 133 taḍmīn 128, 128 n. 48, 129 n. 53 tafsīr 5, 71–72 tamṯīl 71, 74 taʾnīṯ tāʾ at-taʾnīṯ 26, 26 n. 64, 46 n. 27, 94, 94 n. 30, 105 tanwīn 17, 18 n. 10, 19–20, 23, 23 n. 43, 24 n. 44, 27, 27 n. 69, 28, 30–31, 35, 38, 40, 42, 44–45, 50–51, 55, 55 n. 8, 56, 60, 63–64, 201, 201 n. 11 taqdīr 71, 80 taṣarruf maʿnā t-taṣarruf. See under maʿnā waḍʿ 100–101, 104, 124 wājib 74, 244 xabar 118, 121 n. 26 xabar kāna 120 zāʾid ḥarf zāʾid 19 zāʾida (pl. zawāʾid) 1–217, 19, , 23 n. 43, 24, 24 nn. 50, 53, 25, 25 n. 57, 26–32, 43, 50–51, 60, 62, 66 ḥarf az-zāʾida (pl. ḥurūf az-zawāʾid)  19–20, 24 n. 51, 25 nn. 54, 56, 30 ziyāda 20, 25 n. 54, 105 ḥurūf az-ziyāda (pl. of ḥarf ziyāda)  19 n. 11

LANGUAGE INDEX

African Languages 215, 221, 229 Afrikaans Kitap-Afrikaans 225 Arabic 1, 4, 8–12, 17–18, 20–21, 29, 32, 35, 54, 64, 66, 85, 92, 147, 160, 190, 195–198, 200–201, 215–226, 229–247 Arabic creoles 11, 217–219, 224 Arabic pidgins 11, 217–219, 224 Arabic Sprachraum 10 Classical Arabic 5, 10, 12, 44, 230, 241–242 Cypriot Arabic 11, 224 early Arabic 215 Eastern Arabic 218 Gulf Pidgin Arabic 219 Juba Arabic 217 Maghrebi Arabic 218, 242 Modern Standard Arabic 195 Moroccan Arabic 222 spoken Arabic 12, 232, 241, 243 Sudanese Arabic 223 Bambara/Bamanankan 10–13, 229–231, 232 nn. 11, 233–235, 235 n. 23, 236–237, 237 n. 26. 238–244, 244 n. 34, 245–246 Basque 215 Berber 10, 163 n. 23, 218, 224 Caribbean languages 158 Coptic 224 Dioula (Julakan) 229–230 Dravidian languages 11, 221–222 Dutch 220, 222, 225 English 9, 158, 196, 215, 219–220, 222, 226 French 148, 158, 169 n. 30, 186 n. 43, 215, 222, 233 Fulfulde 11, 216, 219, 224, 233 German 222 Germanic languages 215

Greek 224 Ancient Greek 54, 172, 215, 224 Hausa 216, 222–223, 225–226, 242 Eastern Sudanic Hausa 11, 223–224 Western Sudanic Hausa 11, 222–224 Hindi (Hindustani) 221, 225 Hungarian 215 Indonesian 215–216, 221, 225 Islamic Islamic languages 11, 215–216, 220–221, 231 Kannada 222 Ki-Nubi 217, 219 Latin 160, 215, 224 Malay 215, 221, 225 Kitap-Malay 225 Malayalam 221 Malinké (Maninkakan) 229–230 Mande languages 230 Manding languages 229, 229 n. 2, 230, 230 n. 3, 231, 233 Neo-Aramaic 218, 224 Western Neo-Aramaic 10, 218 Niger-Congo languages 230 Nubian 11, 221 Dongolawi Nubian 221 Persian (Farsi) 11, 99, 160, 215–216, 220–226 Romance languages 215 Sarnami 220 Songhay 233 Soninke 233 Spanish 158, 215, 222 Swahili 11, 216, 218–219, 225 Swedish 222, 225–226

264 Tamil 221–222 Telugu 221 Turkic languages 221 Turkish 11, 216, 221, 224 Ottoman Turkish 215

language index Urdu 215–216, 221, 224 West African Languages 11, 219, 231, 231 n. 9

LIST OF CONTRIBUTORS

Lidia Bettini is Professor of Arabic language and literature at the University of Florence. Her research interests include Arabic medieval linguistic thinking and modern Arabic Bedouin dialects. Her recent publications include the volume Contes féminins de la Haute Jézireh syrienne— Matériaux ethno—linguistiques d’un parler nomade oriental, Firenze 2006 and the volume Min sawālif al-ḥarīm fī l-Jazīra al-sūriyya al-‘ulyā, published by the Ministry of Culture of the Syrian Arab Republic, Damascus 2009. Georges Bohas, PhD in Linguistics and Docteur d’Etat, member of the Institut Universitaire de France, is Professor at the Ecole Normale Supérieure de Lyon. He has been working for about twenty years on the restructuration of the Theory of Matrices and Etymons (TME). His last published monograph (in collaboration with M. Dat) is Une théorie de l’organisation du lexique des langues sémitiques : matrices et étymons, Lyon, ENS Editions, 2007. Antonella Ghersetti is Professor of Arabic Languange and Literature, Università Ca’ Foscari, Venice. She holds a PhD in Semitic Studies (Semitic Linguistics) from the University of Florence with a thesis on Dalāʾil al-ʾiʿǧāz of ʿAbd a-Qāhir al-Jurjānī. She is member of IPOCAN, Istituto per l’Oriente “Carlo Alfonso Nallino”, Rome and of UEAI (Union Européenne des Arabisants et Islamisants). She is member of the editorial board of Journal of Arabic and Islamic Studies, Quaderni di Studi Arabi. N.s.; she is also member of the advisory board of al-Masāq (University of Leeds), of Annali della Facoltà di Lingue e Letterature Straniere, University of Sassari and of the international advisory board of Library of Arabic Literature (LAL) of New York University Abu Dhabi and NYU Press. Her main fijields of research are themes and techniques of medieval Arabic narrative, the Arabic linguistic tradition, physiognomics in the medieval Arabic tradition. Jean-Patrick Guillaume, a former Pensionnaire scientifijique at the French Institute of Damascus, is Professor at the Université de la Sorbonne nouvelle and a member of the Laboratoire d’Histoire des théories linguistiques (CNRS). He has published a number of studies on the history

266

list of contributors

of Arabic grammar, among them The Arabic Grammatical Tradition (with G. Bohas and D. E. Kouloughli), London, Routledge, 1990. Giuliano Lancioni, PhD in Linguistics (1995), Roma Tre University, is Professor of Arabic language and literature at the Roma Tre University. He has published a number of studies on the syntax of Arabic (Ordini lineari marcati in arabo, supplement to the Rivista degli Studi Orientali, Rome, 1996) and the history of the Arabic linguistic tradition. He is currently working on issues involved in Arabic computational and corpus linguistics. Pierre Larcher, PhD (1980) and Docteur d’Etat (1996), University of Paris – Sorbonne Nouvelle, is Professor of Arabic Linguistics at the University of Aix-en-Provence (France) and researcher at the IREMAM (CNRS). He lived in the Arab World (Syria, Libya, Morocco) between 1971 and 1982, holding several research and teaching positions. He has published extensively in diffferent fijields of Arabic and Semitic Linguistics (Le Système verbal de l’arabe classique, 2003; (co-edited with P. Cassuto), La Sémitologie, aujourd’hui, 2000 and La formation des mots dans les langues sémitiques, 2007), and translated Pre-Islamic Arabic Poetry into French (Les Mu‘allaqât, 2000; Le Guetteur de Mirages. Cinq poèmes préislamiques, 2004). Aryeh Levin was born in Israel in 1937. He is Professor Emeritus of Arabic at the Hebrew University of Jerusalem. His main fijields of research are: Arabic medieval grammatical thought and terminology, history of Arabic language, and modern Arabic dialects. He was the Head of the Institute of Asian and African Studies of The Hebrew University of Jerusalem, 1992— 1998. In 2010, he was awarded the most prestigious “Israel Prize in general linguistics” for his achievements. Cristina Solimando, PhD in Arabic and Islamic Studies (2006), “La Sapienza” University, Rome, is Associate Professor of Arabic Language and Literature at the Roma Tre University. Her researches are focused on the history of Arabic medieval linguistic thinking (ʽIlal al-taṯniya: le cause del duale in Ibn Ǧinnī, Le origini della grammatica araba). She is currently working on Corpus Linguistics, especially on blogs.

list of contributors

267

Kees Versteegh (1947) is Professor of Arabic and Islam at the University of Nijmegen (The Netherlands). He graduated in Classical and Semitic languages and wrote his PhD on Greek elements in Arabic linguistic thinking (University of Nijmegen, 1977). His fijield of research is historical linguistics and the history of linguistics, focusing on processes of language change and language contact. His books include Pidginization and creolization: The case of Arabic (Amsterdam, 1984), Arabic grammar and Qurʾanic exegesis in early Islam (Leiden, 1993), The Arabic linguistic tradition (London, 1997), and The Arabic language (Edinburgh, 1997). He co-edited the Handbuch für die Geschichte der Sprach- und Kommunikationswissenschaft (Berlin, 2000–2005) and is the editor-in-chief of the Encyclopedia of Arabic language and linguistics (Leiden, 2005–2008) and of the Arabic/ Dutch – Dutch/Arabic dictionary (Amsterdam, 2003). Francesco Zappa PhD in Islamic Studies, “La Sapienza” University, Rome, is Associate Professor of Islamic Studies at the University of Aix-enProvence (France). His research interests bridge ‘classical’ Islamic Studies, (Linguistic) Anthropology, Oral Literature and Literacy Studies, focusing on the role of Bambara (a Mande language spoken in West Africa) as a channel for Islamic expression and transmission of religious knowledge in Mali. He has carried out fijieldwork in Mali since 2000, and contributed articles, journal supplements and book chapters to peer reviewed academic journals and edited volumes, including Die Welt des Islams, Archives de Sciences Sociales des Religions and Rivista degli Studi Orientali.