Approaches to Arabic Linguistics: Presented to Kees Versteegh on the Occasion of His Sixtieth Birthday 9004160159, 9789004160156, 9789047422136

For a lifetime, Kees Versteegh played a leading role in Arabic linguistics, dialects (diglossia, creolization, pidginiza

166 74 4MB

English Pages 757 [795] Year 2007

Report DMCA / Copyright


Polecaj historie

Approaches to Arabic Linguistics: Presented to Kees Versteegh on the Occasion of His Sixtieth Birthday
 9004160159, 9789004160156, 9789047422136

Citation preview

Approaches to Arabic Linguistics

Studies in Semitic Languages and Linguistics Editorial board

T. Muraoka


Kees Versteegh

Approaches to Arabic Linguistics Presented to Kees Versteegh on the Occasion of his Sixtieth Birthday

Edited by

Everhard Ditters and Harald Motzki


This book is printed on acid-free paper. Library of Congress Cataloging-in-Publication Data A C.I.P. record for this book is available from the Library of Congress.

ISBN 0081-8461 ISBN 978 90 04 16015 6 Copyright 2007 by Koninklijke Brill NV, Leiden, The Netherlands. Koninklijke Brill NV incorporates the imprints Brill, Hotei Publishing, IDC Publishers, Martinus Nijhoff 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. printed in the netherlands


Preface ..................................................................................................


Bibliography Kees Versteegh ............................................................


HISTORY Inside the Speaker’s Mind: Speaker’s Awareness as Arbiter of Usage in Arabic Grammatical Theory ......................................... Ramzi Baalbaki


Pragmatics and Contractual Language in Early Arabic Grammar and Legal Theory ......................................................... Michael Carter


Idmār in the Maānī of al-Farrā: A Grammatical Approach between Description and Explanation ......................................... Kinga Dévényi


Arabic alladī as a Conjunction: An old Problem and a New Approach ......................................................................................... Werner Diem


Les origines de la grammaire arabe, selon la tradition: description, interprétation, discussion .......................................... Pierre Larcher


Sībawayhi’s View of the zarf as an āmil ......................................... Aryeh Levin


Problems in the Medieval Arabic Theory of Sentence Types ......... Yishai Peled




Arabic avant la lettre. Divine, Prophetic, and Heroic Arabic ....... Stefan Wild Inflection and Government in Arabic According to Spanish Missionary Grammarians from Damascus (XVIIIth Century): Grammars at the Crossroads of Two Systems ............................. Otto Zwartjes



LINGUISTICS The Linguistic Analysis and Rules of Pause in Arabic ................... Salman H. Al-Ani


The Explanation of Homonymy in the Lexicon of Arabic ............. Georges Bohas and Abderrahim Saguer


The Periphrastic Bilingual Verb Construction as a Marker of Intense Language Contact. Evidence from Greek, Portuguese and Maghribian Arabic ................................................................. Louis Boumans


Faula, faila, faala: dispersion et régularités sémantiques dans les trois schèmes simples du verbe arabe ..................................... Joseph Dichy


Featuring as a Disambiguation Tool in Arabic Natural Language Processing ....................................................................... Everhard Ditters


Arabic on the Media: Hybridity and Styles ..................................... Mushira Eid The Use of Morphological Patterns in Arabic Grammars of Turkic ............................................................................................... Robert Ermers Lexical Gaps in Arabic: Evidence from Dictionaries ...................... Jan Hoogland




contents Masdar Formation .............................................................................. Joost Kremers Méthodologie linguistique: organisation de la langue arabe. Organisation générale des langues ............................................... André Roman

ix 475


DIALECTS How to be KOOL in Arabic Writing: Linguistic Observations from the Side Line .......................................................................... Gert Borg “Hello, I say, and welcome! Where from, these riding men?” Arabic Popular Poetry and Political Satire: a Study in Intertextuality from Jordan ........................................................... Clive Holes Notes on the Dialects of the ‘Lēgāt and H amāda h of Southern Sinai ................................................................................................. Rudolf de Jong Classical and Colloquial Arabic Archaisms ..................................... Alan S. Kaye †





Do They Speak the Same Language? Language Use in Juba Local Courts .................................................................................... Catherine Miller


Paradigmatic Stability and Final Laryngeals in Nigerian Arabic: Why History Repeats itself, without Actually Doing so ............. Jonathan Owens


Some Aspects of Diglossia as Reflected in the Vocabulary of Literary and Colloquial Arabic ..................................................... Judith Rosenhouse




Everything you Always Wanted to Know about āl, yiūl ‘to say’ in Egyptian Arabic ......................................................................... Manfred Woidich


Index ....................................................................................................



From when we began to compile a Festschrift for Kees Versteegh on the occasion of his sixtieth birthday, the first problem we had to tackle was: Who should we invite to contribute? The broad range of his scholarly interests, his expertise in different fields, his academic contribution to them, as well as his network of global contacts made it clear that the circle of his colleagues and friends is too large to invite all to contribute. So we decided to confine this volume to Kees Versteegh’s core contribution: Arabic linguistics. This decision will surely be regretted by his colleagues working in related fields of scholarship in which Kees Versteegh is engaged as well, but we had to make a choice. Moreover, our time schedule forced us to only select the first set of early contributions brought in. In the end, what does it matter? Someone will have to prepare another for his sixty-fifth! Even with this limitation, the present Festschrift has become voluminous, since the eagerness of scholars in the field of Arabic linguistics to contribute has been overwhelming. It clearly demonstrates: Kees Versteegh has been widely considered to be one of the most eminent scholars in this discipline. He also is appreciated as a tirelessly working editor and co-editor of renowned book series, collective volumes and encyclopaedias. Moreover, he has won many friends by his kindness and dependability. With the list of Kees Versteegh’s publications at hand, one realises the number, diversity, and depth of his scholarly interests, from Hellenistic elements in Arabic linguistics and other fields of Islamic culture, to the history of Arabic grammar as a scholarly discipline from classical to modern times, grammatical and linguistic phenomena described by scholars of Arabic grammar, changes in written and spoken Arabic through the ages, as well as early Qurānic exegesis as source for the beginnings of Arabic grammar. In addition, he ranslated and commented on Arabic texts, grammatical treatises and other genres including a novel. His books Greek Elements in Arabic Linguistic Thinking; Pidginization and Creolization: the Case of Arabic; Arabic Grammar and Qurānic Exegesis in Early Islam; and The Arabic Language are handbooks. Every single scholar and student of Arabic linguistics can not bypass them.



Several of his publications have been translated into Arabic, and others will follow. At Nijmegen University, where Kees Versteegh is working since 1972, first as lecturer and later as full professor, he organized several workshops about the ‘History of Arabic Grammar’ (1984 and 1987) and the ‘Model of Arabic Grammar in other Languages’ (1997). With intense pleasure the participants of these conferences remember these fruitful academic gatherings occurring in a very informal and convivial atmosphere. These gatherings resulted in two collective volumes he edited in cooperation with one of his many colleagues. Thanks to his co-editorship, international publication projects such as the History of Language Sciences and the Encyclopaedia of Arabic Language and Linguistics, Arabic linguistics acquired a firm place in the field of linguistics in general. Generations of Dutch students, translators and interpreters of Arabic will know his name from the two volumes of the dictionary Woordenboek Arabisch-Nederlands/Nederlands-Arabisch of which he is one of the editors. This Festschrift will be a monument for Kees Versteegh in Arabic studies, as a homage to his scholarly oeuvre. Several contributions camd from former students who wrote their Ph.D. theses under his supervision. As a matter of fact, all contributors studied with him. We divided the collected articles into three chapters reflecting the foci of his scholarly oeuvre: history of Arabic grammar, Arabic linguistics, and Arabic dialectology. We have put the history of Arabic grammar first, since it is Kees Versteegh’s true domain. He wrote his Ph.D. in this field, dedicated most of his publications to it, and returns to it. He translated and commented upon al-Zajjājī’s Kitāb al-Īdāh , a theoretical treatise on Arabic grammar and one of the classical works on this issue. Since the 1980s, his research and publications also have been directed towards other historical and thematical topics in linguistics, and he certainly will recognize many of his own ideas in the contributions in all three sections. His studies on pidginization and creolization in Arabic and his interest in Arabic dialects express his feeling for the historical development of the Arabic language. Kees Versteegh is a polyglot. When invited to lecture outside the Netherlands he enjoys lecturing in the language of is audience, in English, French, German and Arabic, as well as Spanish or Czech. From the start of his academic career—he first studied Greek and Latin—he cherished a fondness for these languages. In his research this became



apparent in his search for Hellenistic elements in Arabic linguistics. Together with a colleague of Classical Studies, he also teaches regular classes on linguistics. His interest in—well, let us say—‘exotic’ languages is exceptional. His eyes sparkle when he tells how he learned Inuktitut, the language of the Inuit, from an indigenous scholar in Alaska or how he mastered Hottentot clicks in South Africa. Many colleagues enjoyed Kees Versteegh’s hospitality in his ‘sexton’s house’ in picturesque Batenburg. They will never forget the tasty meals he prepares himself as the evening’s conversation quickly passes from scholarly issues to personal and private ones, as for example his experiences as director of the Netherlands-Flemish Institute in Cairo or the birds he suddenly noticed by a Zamalek window. One may be engaged with Kees in a really basic discussion of Arabic linguistics, only to be interrupted by his quick sprint outdoors to observe an ornithological event, afterwards resuming the discussion exactly at the comma or full stop of his last intervention. Bird watching remains his hobby and on a walk along the Maas (or any other river), he never forgets his binoculars. Besides linguistics, a guest learns a lot about the birds of the region, their names in Dutch and other languages, including Arabic. Guests get free access to his rich private library with its rare books on Arabic linguistics and even the privilege of borrowing without any check for solvability. The study of Arabic has been Arabic-speaking scholars’ domain for many centuries. Kees Versteegh has unearthed their understanding of the language and ‘translated’ it for Western scholarship. Western scholars’ linguistic study of Arabic, according to current standards, started only at the beginning of the 19th century. So while the western linguistic approach is much younger than that of those native scholars, Western Arabic linguistics are on par with even native intellectual efforts, and Kees Versteegh has contributed to this state of affairs. Had classic scholars of the Arabic language like Sībawayhi or al-Zajjājī the change to witness the linguistic achievements of their modern Western colleagues like Kees Versteegh, they would have been impressed. They might have cited the verses of Ibn H amdīs (Sicily, 447–527/1056–1133): alā šadawāti tuyūrin fisāh in lahunna aārīdu inda l-halīli turajjiu fīhā durūba l-luh ūni

alā anna afsah ahā ajamu muhammalatu l-wazni lā tulamu fa-tutribunā wa-hya lā tufhamu

On the tunes of birds that speak true Arabic, but the most eloquent of them are foreigners


preface who use metres that are neglected by al-Khalīl and unknown in which they quaver notes of various melodies, delighting us although they are unintelligible.

This Festschrift would not have seen the light without several helping hands. Greetje Heemskerk compiled the bibliography of his publications. Marjolein van der Heul and Ine Smeets produced the raw version of the index. Elizabeth Bishop copy-edited the English contributions. Joed Elich of Brill has been willing to publish the book. Ingrid Heijckers lead us safely through the whole of the production process and beyond. Renee Otto supervised the publication process. We are very grateful for their commitment. Everhard Ditters and Harald Motzki


1977 Greek Elements in Arabic Linguistic Thinking. (= Studies in Semitic Languages and Linguistics, 7). Leiden: E. J. Brill. Also published as PhD thesis University of Nijmegen. [Arabic translation by Mahmud Kanākrī, 2000: Anāsir yūnāniyya fī l-fikr al-luġawī al-arabī. Amman: Jamiyyat ummāl al-matābi at-taāwuniyya]. 1978 ‘The Arabic Terminology of Syntactic Position.’ Arabica 25, 261–281. 1979 ‘Die Mission des Kyrillos im Lichte der arabo-byzantinischen Beziehungen.’ Zeitschrift der deutschen Morgenländischen Gesellschaft 129, 233–262. Reviews Ephrem Hunayn Festival Baghdād 4–7/2/1974. 1974. Baghdad: Mat būāt Majma al-Luġa al-Suryāniyya, Bibliotheca Orientalis 36, 96–97. Rundgren, Frithiof. 1976. Über den griechischen Einfluß auf die arabische Nationalgrammatik. Uppsala: Acta Universitatis Upsaliensis. N.S. 2(5), 119–144. Bibliotheca Orientalis 36, 235–236. 1980 ‘Hellenistic Education and the Origin of Arabic Grammar.’ Progress in Linguistic Historiography: Papers from the International Conference on the History of the Language Sciences, Ottawa, 28–31 August 1978. Edited by Konrad Koerner. Amsterdam: J. Benjamins. 333–344. ‘Logique et grammaire au dixième siècle.’ Histoire, Épistémologie, Langage 2. 39–52. ‘Notice bibliographique.’ Histoire, Épistémologie, Langage 2. 67–75. ‘The International Project Onomasticon Arabicum.’ Bibliotheca Orientalis 37, 291–294. ‘The Origin of the Term qiyās in Arabic Grammar.’ Zeitschrift für arabische Linguistik 4, 7–30. ‘The Stoic Verbal System.’ Hermes 108, 338–357.


bibliography kees versteegh

Reviews Ambros, Edith. 1979. Sieben Kapitel des Šarh Kitāb Sībawaihī von arRummānī in Edition und Übersetzung. Wien: Verlag des Verbandes der wissenschaftlichen Gesellschaften Österreichs. Bibliotheca Orientalis 37, 361–362. Malti Douglas, Fedwa and Geneviève Fourcade. 1976. The Treatment by Computer of Medieval Arabic Biographical Data: An introduction and guide to the “Onomasticum Arabicum”. Paris: Centre national de la recherche scientifique. Bibliotheca Orientalis 37, 362. Traini, Renato. 1977. Sources biographiques des Zaïdites (années 122–1200 h.): Letters alif-ha’. Paris: Centre national de la recherche scientifique. Bibliotheca Orientalis 37, 362–363. 1981 ‘A Dissenting Grammarian: Qutrub on declension.’ Historiographia Linguistica 8, 403–429. [See also The History of Linguistics in the Near East. Edited by Kees Versteegh, Konrad Koerner and Hans-Josef Niederehe. 1983. Amsterdam: J. Benjamins. 167–193]. ‘De taalsituatie in de Arabische wereld.’ De taal van de Islam: Opstellen over Arabische, Turkse en Afghaanse cultuur. Nijmegen: Nederlandse Vereniging voor de Studie van het Midden-Oosten en de Islam. 19–38. ‘La conception des temps du verbe chez les grammairiens arabes.’ Analyses, théorie 3, 47–68. Reviews Klein-Franke, Felix. 1980. Die klassische Antike in der Tradition des Islam. Darmstadt: Wissenschaftliche Buchgesellschaft. Bibliotheca Orientalis 38, 734–737. 1982 ‘Progress and Change in the History of Arabic Grammar.’ Linguistics in the Netherlands 1982. Edited by Saskia Daalder and Marinel Gerritsen. Amsterdam: North-Holland. 39–50. ‘Structural Change and Pidginization in the History of the Arabic Language.’ Papers from the 5th International Conference on Historical Linguistics, Galway, April 6–10 1981. Edited by Anders Ahlqvist. Amsterdam: J. Benjamins. 362–373. Vertalingen voor 1ste jaarsstudenten Arabisch (with Gert Borg). Nijmegen: Katholieke Universiteit Nijmegen.

bibliography kees versteegh


Reviews Dagorn, René. 1981. La geste d’Ismaël d’après l’onomastique et la tradition arabes, Genève: Droz; Paris: Champion. Bibliotheca Orientalis 39, 720–726. Rowson, Everett and Seeger Bonebakker. 1980. A Computerized Listing of Biographical Data from the Yatīmat al-Dahr by al-Thaālibī. Malibu: Undena Publications. Bibliotheca Orientalis 39, 727–729. 1983 ‘A Dissenting Grammarian: Qutrub on declension.’ The History of Linguistics in the Near East. Konrad Koerner, Hans-Josef Niederehe, and Kees Versteegh (Eds.). Amsterdam: J. Benjamins. 167–193. [See also Historiographia Linguistica 1981. 8, 403–429]. ‘Arabic Grammar and the Corruption of Speech.’ Arab Language and Culture. Edited by Ramzi Baalbaki. Beirut: American University of Beirut. [= al-Abhāth 31]. 139–160. ‘Current Bibliography on the History of Arabic Grammar.’ Zeitschrift für arabische Linguistik 10, 86–89. ‘Current Bibliography on the History of Arabic Grammar.’ Zeitschrift für arabische Linguistik 11, 84–86. ‘History of Eastern Linguistics in the Soviet Union.’ Historiographia Linguistica 10, 289–307. The History of Linguistics in the Near East. (= Studies in the History of Linguistics, 28). Editor (with Konrad Koerner and Hans-Josef Niederehe). Amsterdam: J. Benjamins. 1984 ‘Arab Grammatical Studies before Sībawayh.’ Matériaux pour une histoire des théories linguistiques = Essays toward a history of linguistic theories = Materialien zu einer Geschichte der sprachwissenschaftlichen Theorien. Edited by Sylvain Auroux, Michel Glatigny, André Jolly, Anne Nicolas and Irène Rosier. Lille: Université de Lille III. 227–238. ‘Current Bibliography on the History of Arabic Grammar.’ Zeitschrift für arabische Linguistik 12, 86–89. ‘Piginigado, kreoligado kaj Esperanto.’ Hungara Vivo 4, 127–129. Pidginization and Creolization: The case of Arabic. (= Current Issues in Linguistic Theory, 33). Amsterdam: J. Benjamins.


bibliography kees versteegh

Reviews Bakalla, Muhammad. 1983. Arabic linguistics. London: Mansell. Bibliotheca Orientalis 41, 751–754. Carter, Michael. 1982. Arab Linguistics: An introductory classical text with translation and notes. Amsterdam: J. Benjamins. Bibliotheca Orientalis 41, 225–230. Nebes, Norbert. 1982. Funktionsanalyse von kāna yafalu. Hildesheim: Olms. Bibliotheca Orientalis 41, 754–757. 1985 ‘Current Bibliography on the History of Arabic Grammar.’ Zeitschrift für arabische Linguistik 14, 79–81. ‘La ‘Grande étymologie’ d’Ibn Jinnī.’ La linguistique fantastique. Edited by Sylvain Auroux, Jean-Claude Chevalier, Nicole Jacques-Chaquin and Christiane Marchello-Nizia. Paris: Denoël. 44–50. Studies in the History of Arabic Grammar. Papers Presented during a Workshop held at the University of Nijmegen, April 16–19, 1984. Editor (with Hartmut Bobzin). Wiesbaden: O. Harrassowitz. (= Zeitschrift für arabische Linguistik 15). ‘Survey of Journals.’ Arab Journal of Language Studies / al-Majalla al-arabiyya li-d-dirāsāt al-luġawiyya 2, 189–197. ‘The Development of Argumentation in Arabic Grammar: The declension of the dual and the plural.’ Studies in the History of Arabic Grammar. Edited by Hartmut Bobzin and Kees Versteegh. Wiesbaden: Harrassowitz. 1985. 152–173. [= Zeitschrift für arabische Linguistik 15]. Translation Salih, Tayyib. Seizoen van de trek naar het noorden. Uit het Arabisch vertaald en van een nawoord voorzien. Amsterdam: Meulenhoff. [Dutch translation of Mawsim al-hijra ilā š-šamāl]. Reviews Sezgin, Fuat. 1982, 1984. Geschichte des arabischen Schrifttums. VIII. Lexikographie. IX. Grammatik. Leiden: E. J. Brill. Historiographia Linguistica 12, 452–461. 1986 History of Arabic Grammar. Nijmegen: Instituut voor Talen en Culturen van het Midden Oosten, Katholieke Universiteit.

bibliography kees versteegh


‘Latinitas, Hellènismós, Arabiyya.’ The History of Linguistics in the Classical Period. Edited by Daniel J. Taylor. (= Historiographia Linguistica 13, 425–448). Amsterdam: J. Benjamins, 1987. 251–274.]. ‘The Origin of the Romance Languages and the Arabic Dialects.’ Islão e arabismo na península ibérica: Actas do XI. congresso da União Europeia de Arabistas e Islamólogos, Evora, Faro, Silves, 29 set.–6 out. 1982. Edited by Adel Y. Sidarus. Évora: Universidade de Évora. 337–352. ‘Word Order in Uzbekistan Arabic and Universal Grammar.’ On the Dignity of Man: Oriental and classical studies in honour of Frithiof Rundgren. Edited by Tryggve Kronholm and Eva Riad. (= Orientalia Suecana 33–34). Uppsala: Almqvist and Wiksell. 443–453. 1987 ‘al-Arab fī jibāl al-Alp.’ al-Azmina 7, 26–30. ‘Current Bibliography on the History of Arabic Grammar.’ Zeitschrift für arabische Linguistik 16, 130–133. ‘Die arabische Sprachwissenschaft.’ Grundriß der arabischen Philologie, II. Edited by Helmut Gätje, Wiesbaden: L. Reichert. 148–176. Het Arabisch: Norm en realiteit (with Arie Schippers). Muiderberg: D. Coutinho. ‘Marginality in the Arab Grammatical Tradition.’ Papers in the History of Linguistics: Proceedings of the Third International Conference on the History of the Language Sciences (ICHoLS III), Princeton, 19–23 August 1984. Edited by Hans Aarsleff, Louis G. Kelly and Hans-Josef Niederehe. Amsterdam: J. Benjamins. 87–96. ‘Nahwiyyūna wa-luġawiyyūna wa-mawqif Dozy izā at-turāt  an-nahwī al-arabī.’ Fī l-mujamiyya al-arabiyya al-muāsira. Edited by Ahmed El-Ayed and Ibrahim Ben Mrad. Tunis: Dār al-Ġarb al-Islāmī. 401–413. Reviews Sublet, Jacqueline. 1985. Cahiers d’onomastique Arabe, 1982–1984. Paris: Centre national de la recherche scientifique. Bibliotheca Orientalis 44, 130–133. 1988 ‘De ontwikkeling van de technische woordenschat in Modern Standaard Arabisch: Naar aanleiding van een recente publicatie.’ Sharqiyyât 1, 80–85.


bibliography kees versteegh

‘Geschiedschrijving in de klassieke Arabisch-Islamitische samenleving.’ Tussen traditie en wetenschap: Geschiedbeoefening in niet-westerse culturen. Edited by R.B. van de Weijer, P.G.B. Thissen and R. Schönberger. Nijmegen: Katholieke Universiteit Nijmegen. 133–145. ‘Pourquoi étudier la tradition grammaticale?’ Le Maroc et la Hollande: Études sur l’histoire, la migration, la linguistique et la sémiologie de la culture. Edited by Abdelmajid Kaddouri, Jilali Saïb and Abdelmajid Zeggaf. Rabat: Université Mohammed V. 207–217. 1989 ‘A Sociological View of the Arab Grammatical Tradition: Grammarians and their professions.’ Studia linguistica et orientalia memoriae Haim Blanc dedicate. Edited by Paul Wexler, Alexander Borg and Sasson Somekh. Wiesbaden: Harrassowitz. 289–302. ‘‘Early’ and ‘late’ Grammarians in the Arab Tradition: The ‘morphonology’ of the hollow verbs.’ Zeitschrift für arabische Linguistik 20, 9–22. ‘La tradition arabe. I : Le langage, la religion et la raison.’ Histoire des idées linguistiques, I : La naissance des métalangages en Orient et en Occident. Edited by Sylvain Auroux. Liège: P. Mardaga. 243–259. ‘The Definition of Philosophy in a Tenth-century Grammarian.’ Jerusalem Studies in Arabic and Islam 12, 66–92. ‘Vulgair Latijn en Koine-Grieks: De verhouding tussen standaardtaal en volkstaal.’ Lampas 22, 74–91. Reviews Bergter, Annette. 1989. Das Kapitel inna wa-ahawātuhā aus dem ‘Manhaj as-sālik’ des Grammatikers Abū H ayyān al-Ġarnātī (1256–1344). Hildesheim: G. Olms. Historiographia Linguistica 16, 180–184. 1990 ‘Are Linguists Ridiculous? Notes on a heavenly discussion between grammarians in the 11th century.’ History and Historiography of Linguistics: Proceedings of the Fourth International Conference on the History of the Language Sciences (ICHoLS IV), Trier, 24–28 August 1987, I: Antiquitity–17th Century. Edited by Konrad Koerner and Hans-Josef Niederehe. Amsterdam: J. Benjamins. 147–155. ‘Borrowing and Influence: Greek grammar as a model.’ Le langage dans l’antiquité. Edited by Pierre Swiggers and Alphons Wouters. Leuven: Peeters. 197–212.

bibliography kees versteegh


‘Freedom of the Speaker? The term ittisā and related notions in Arabic grammar.’ Studies in the History of Arabic Grammar, II. Edited by Michael G. Carter and Kees Versteegh. Amsterdam: J. Benjamins, 1990. 281–293 [Arabic translation by Bouchaib Barramou: ‘H urriyyat al-mutakallim? Mussa lah al-ittisā wa-l-mafāhīm al-murtabiha bihi fī n-nahw al-arabī.’ Fikr wa-naqd 3:24 (1999). 99–110]. ‘Grammar and Exegesis: The origins of Kufan grammar and the Tafsīr Muqātil.’ Der Islam 67, 206–242. [Reprint in The Early Islamic Grammatical Tradition. Edited by Ramzi Baalbaki. Aldershot [etc.]: Ashgate/Variorum. 2007. 37–73]. Over taal en verandering. Nijmegen: Katholieke Universiteit. Inaugural lecture, University of Nijmegen. Studies in the History of Arabic Grammar, II. Proceedings of the 2nd Symposium on the History of Arabic Grammar, Nijmegen, 27 April–1 May, 1987. Editor (with Michael G. Carter). Amsterdam: J. Benjamins. ‘The Arab Presence in France and Switzerland in the 10th Century.’ Arabica 37, 359–388. ‘The Earliest Commentary on the Qurān: Muqātil’s Tafsīr.’ Makalahmakalah yang disampaikan dalam rangka kunjungan menteri Agama R.I.H. Munawir Sjadzali, M.A. ke Negeri Belanda (31 Oktober– 7 November 1988). Edited by Wim Stokhof and Nico Kaptein. Jakarta: INIS. 213–219. [with Indonesian translation: “Tafsir Quran paling awal: Tafsir Muqatil”]. Reviews Owens, Jonathan. 1988. The Foundations of Grammar: An introduction to medieval Arabic grammatical theory. Amsterdam: J. Benjamins. Die Welt des Islams 30, 248–250. 1991 ‘Arabic Language Teaching and the Status of Standard Arabic.’ Proceedings of the Symposium on Differentiation in LSP, Learning and Teaching, Leuven, 7–10 November 1990. Edited by Serge Verlinde. Leuven: Instituut voor levende talen. 95–102. ‘Greek Translations of the Qurān in Christian Polemics (9th century A.D.).’ Zeitschrift der Deutschen Morgenländischen Gesellschaft 141, 52–68. ‘Manā. 1. In grammar.’ The Encyclopaedia of Islam: New edition. Edited by C. E. Bosworth, E. van Donzel, B. Lewis and Ch. Pellat, assisted by F. Th. Dijkema and S. Nurit. Leiden: E. J. Brill. VI, 346.


bibliography kees versteegh

‘New approaches towards the history of Arabic.’ Taqaddum al-lisāniyyāt fi l-aqt ār al-arabiyya [Progrès de la linguistique dans les états arabes]. Edited by Abdelkader Fassi Fehri. Beirut: Dār al-Ġarb al-Islāmī. 199–216. ‘The substratum debate in creole linguistics.’ Diachronica 8, 59–80. ‘Two Conceptions of Irreality in Arabic Grammar: Ibn Hišām and Ibn al-ājib on the particle law.’ De la grammaire de l’arabe aux grammaires des arabes. Edited by Pierre Larcher. Damas: Institut Français de Damas. 77–92 [= Bulletin d’Études Orientales, 43]. Reviews Holes, Clive. 1990. Gulf Arabic. London: Routledge. Bibliotheca Orientalis 48, 293–296. Wouters, Alfons. 1988. The Chester Beatty Codex Ac. 1499: A GraecoLatin lexicon on the Pauline Epistles and a Greek grammar. Leuven: Peeters. Bibliotheca Orientalis 48, 523–527. 1992 ‘De val van de Barmakiden volgens Tabarī.’ Elf wijzen van interpreteren: Essays over het lezen van teksten uit het islamitisch cultuurgebied. Edited by Ed de Moor. Nijmegen: Instituut voor Talen en Culturen van het Midden-Oosten. 117–126. ‘Grammar and Rhetoric: Jurjānī on the verbs of admiration.’ Studies in Semitic Linguistics in Honor of Joshua Blau. Edited by M. Bar-Asher [et al.]. Jerusalem: Hebrew University of Jerusalem. 113–133. [= Jerusalem Studies in Arabic and Islam 15]. ‘The Debate about Latin and Early Romance.’ Diachronica 9, 259–285. The History of Linguistics in the Low Countries. (= Studies in the History of the Language Sciences 64). Editor (with Jan Noordegraaf and Konrad Koerner). Amsterdam: J. Benjamins. [Published in 1988 as Historiographia Linguistica 15, 1+2]. Reviews Malina, Renate. 1987. Zum schriftlichen Gebrauch des kairenischen Dialekts anhand ausgewählter Texte von Sadaddīn Wahba. Berlin: Schwarz. Bibliotheca Orientalis 49, 526–528. Mitchell, T. F. 1990. Pronouncing Arabic, I. Oxford: Clarendon. Bibliotheca Orientalis 49, 523–526. Sluiter, Ineke. 1990. Ancient Grammar in Context, Amsterdam: VU University Press. Historiographia Linguistica 19, 401–408.

bibliography kees versteegh


1993 Arabic Grammar and Qurānic Exegesis in Early Islam. Leiden: E. J. Brill. ‘Esperanto as a First Language: Language acquisition with a restricted input.’ Linguistics 31, 539–555. ‘Leveling in the Sudan: From Arabic creole to Arabic dialects.’ International Journal of the Sociology of Language 99, 65–79. ‘Three is a Crowd: lawyers and linguists on Qurān 4/11.’ Wolfdietrich Fischer zum 65. Geburtstag. Edited by Hartmut Bobzin and Otto Jastrow. Wiesbaden: Harrassowitz. 302–315 [= Zeitschrift für arabische Linguistik 25]. Reviews Jastrow, Otto. 1990. Der arabische Dialekt der Juden von Aqra und Arbīl. Wiesbaden: O. Harrassowitz. Bibliotheca Orientalis 50, 858–862. Johnstone, Barbara. 1991. Repetition in Arabic Discourse: Paradigms, syntagms, and the ecology of Language. Amsterdam: J. Benjamins. Bibliotheca Orientalis 50, 855–858. 1994 Arabic Outside the Arab World. Editor. New Delhi: Bahri Publications. [= Indian Journal of Applied Linguistics, 20]. ‘The Arabic Linguistic Tradition in a Comparative Perspective.’ Arab Journal of the Humanities 12 (no. 46), 282–297. ‘The Notion of ‘Underlying Levels’ in the Arabic Grammatical Tradition.’ Historiographia Linguistica 21, 271–296. [Arabic translation by Bouchaïb Barramou: ‘Mafhūm al-mustawayāt at-tahtiyya fī t-turāt an-nahwī al-arabī.’ Recherches Linguistiques 6:2 (2001), 81–126]. Reviews Åkesson, Joyce. 1990. Ahmad b. Alī b. Masūd on Arabic morphology = Marāh al-arwāh, edited with translation, commentary and introduction. Part 1: The strong verb as-sahīh. Leiden: E. J. Brill. Bibliotheca Orientalis 51, 717–719. Broselow, Ellen, Mushira Eid and John McCarthy (Eds.). 1992. Perspectives on Arabic Linguistics, IV. Amsterdam: J. Benjamins. Bibliotheca Orientalis 51, 176–179. Comrie, Bernard and Mushira Eid (Eds.). 1991. Perspectives on Arabic Linguistics, III. Amsterdam: J. Benjamins. Journal of the American Oriental Society 114, 107–109.


bibliography kees versteegh

Robins, Robert H. 1993. The Byzantine Grammarians: Their place in history. Berlin: Mouton de Gruyter. Bibliotheca Orientalis 51, 719–723. 1995 The Explanation of Linguistic Causes: Az-Zajjājī’s Theory of grammar, introduction, translation, and commentary. Amsterdam: J. Benjamins. (Translation of Abū l-Qāsim Abd ar-Rahmān b. Ishāq az-Zajjājī, Kitāb al-Īdāh fī ilal an-nahw). Arabic Outside the Arab World. (= Indian Journal of Applied Linguistics, 20). (Co-editor). New Delhi: Bahri Publications. 1996 ‘Arab Grammarians: Abū l-Aswad ad-Dualī, 3 i-ii; Abū H ayyān, 4 i-5 ii; Abū Ubayd, 5 ii; Abū Ubayda, 6 i; al-Axfaš al-Asġar, 11 ii; al-Axfaš al-Awsat, 11 ii-12, i; al-Asmaī, 49 ii-50 i; al-Astarābādī, 50 i-51 i; al-Azharī, 59 i-ii; al-Batalyūsī, 70 ii–71 i; al-Fārābī, 288 ii-290 i; al-Fārisī, 290 i-ii; al-Farrā, 291 i-292 i; al-Fīrūzābādī, 300 i-ii; al-Jarmī, 326 ii-327 i; al-Jawharī, 330 i–ii; al-Jurjānī, 380 i-381 ii; Xalaf al-Ahmar, 384 i-ii; Xalīl, 385 i–386 i; Ibn Abī Ishāq, 438 i; Ibn Ājurrūm, 438 i-ii; Ibn al-Anbārī (Abū l-Bakr), 438 ii; Ibn al-Anbārī (Abū l-Barakāt), 438 ii-439 ii; Ibn Aqīl, 439 i-440 i; Ibn Durayd, 441 i-ii; Ibn Fāris, 442 ii-443 i; Ibn Jinnī, 445 ii-447 ii; Ibn al-H ājib, 447 i-ii; Ibn Xālawayh, 447 ii-448 i; Ibn Hišām, 448 i-ii; Ibn Kaysān, 448 ii-449 i; Ibn Madā, 449 ii-450 ii; Ibn Mālik, 450 ii-451 i; Ibn Manzūr, 451 i-ii; Ibn Qutayba, 452 i-ii; Ibn as-Sarrāj, 452 ii-453 i; Ibn Sīdah, 453 ii-454 i; Ibn as-Sikkīt, 454 i-ii; Ibn Šuqayr, 454 ii-455 i; Ibn Usfūr, 455 ii-456 i; Ibn Wallād, 456 i-ii; Ibn Yaīš, 456 i-ii; Ixwān as-Safā, 458 i-ii; Īsā ibn Umar at-Taqafī, 461 ii-462 i; al-Kisāī, al-Māzinī, 617 ii- 618 i; al-Mubarrad, 651 ii-657 ii; an-Nahhās, 666 ii- 667 i; al-Qālī, 766 i-ii; Qutrub, 767 ii- 768 i; ar-Rummānī, 808 i-ii; Sībawayhi, 856 i-859 i; as-Sīrāfī, 863 i-864 i; as-Suyūtī, 895 ii-896 i; at-Taālibī, 904 i-ii; Talab, 905 i-ii; Ukbarī, 946 i-ii; Yahyā ibn Adī, 1028 i-ii; az-Zabīdī, 1037 i-ii; az-Zajjāj, 1038 i-ii; az-Zajjājī, 1038 ii-1039 ii; az-Zamaxšarī, 1039 ii-1040 i; az-Zubaydī, 1045 i-ii.’ Lexicon Grammaticorum. Edited by Harro Stammerjohann. Tübingen: Niemeyer. ‘Jahilia.’ Het woordenboek van De duivelsverzen. Edited by Maarten Asscher [et al.]. Amsterdam: Contact. 52–53. ‘Linguistic Attitudes and the Origin of Speech in the Arab World.’ Understanding Arabic: Essays in contemporary Arabic linguistics in

bibliography kees versteegh


honor of El-Said Badawi. Edited by Alaa Elgibali. Cairo: American University in Cairo Press. 15–31. ‘The Linguistic Introduction to Rāzī’s Tafsīr.’ Studies in Near Eastern Languages and Literatures. Memorial volume of Karel Petráček. Edited by Petr Zemánek. Prague: Academy of Sciences of the Czech Republic. 589–603. ‘Western Studies on the History of Arabic Grammar, 1969–1994.’ Proceedings of the Colloquium on Arabic Linguistics, Bucharest August 29– September 2. 1994. Edited by Nadia Anghelescu and Andrei Avram. Bucharest: University of Bucharest. I, 9–27. Reviews Fernández, Mauro. 1993. Diglossia: A comprehensive bibliography 1960–1990, Amsterdam: J. Benjamins. Word 47, 288–291. Peled, Yishai. 1992. Conditional Structures in Classical Arabic. Wiesbaden: O. Harrassowitz. Bibliotheca Orientalis 53, 557–560. Wagner, Daniel A. 1993. Literacy, Culture, and Development: Becoming literate in Morocco. Cambridge: Cambridge University Press. International Journal of the Sociology of Language 119, 129–134. 1997 ‘Bedouin and Townspeople: Ibn Khaldūn’s view on the development of Arabic.’ The Diversity of Development: Essays in honour of Jan Kleinpenning. Edited by Ton van Naerssen, Marcel Rutten and Annelies Zoomers. Assen: Van Gorcum. 332–340. Landmarks in Linguistic Thought, III: The Arabic linguistic tradition. London: Routledge. ‘Sarf.’ The Encyclopaedia of Islam: New edition. Edited by C. E. Bosworth, E. van Donzel, W. P. Heinrichs and G. Lecomte, assisted by P. J. Bearman and S. Nurit. Leiden: E. J. Brill. IX, 53–54. ‘al-Shaybānī, Abū Amr Ishāk b. Mirār.’ The Encyclopaedia of Islam: New edition. Edited by C. E. Bosworth, E. van Donzel, W. P. Heinrichs and G. Lecomte, assisted by P. J. Bearman and S. Nurit. Leiden: E. J. Brill. IX, 394–395. The Arabic Language. Edinburgh: Edinburgh University Press. [2nd revised edition 2001] [Arabic translation by Muhammad aš-Šarqāwī: Al-luġa al-arabiyya. Cairo: Al-Majlis al-alā li-t -ta qāfa, 2003]. ‘The Arabic Tradition.’ The Emergence of Semantics in Four Linguistic Traditions: Hebrew, Sanskrit, Greek, Arabic. Edited by Wout van


bibliography kees versteegh

Bekkum, Jan Houben, Ineke Sluiter and Kees Versteegh. Amsterdam: J. Benjamins. 225–284. The Emergence of Semantics in Four Linguistic Traditions: Hebrew, Sanskrit, Greek, Arabic. (with Wout van Bekkum, Jan Houben and Ineke Sluiter). Amsterdam: J. Benjamins. ‘Wetenschap in de Islamitische samenleving.’ In het huis van de Islam: Geografie, geschiedenis, geloofsleer, cultuur, economie, politiek. Edited by Henk Driessen. Nijmegen: SUN. 278–295. Reviews Bohas, Georges (Ed.). 1993. Développements récents en linguistique arabe et sémitique. Damas: Institut Français de Damas. Bibliotheca Orientalis 54, 788–793. Dialectologia Arabica: A collection of articles in honour of the sixtieth birthday of Professor Heikki Palva. Helsinki: Finnish Oriental Society. 1995. Bibliotheca Orientalis 54, 800–802. Durand, Olivier. 1995. Introduzione ai dialetti arabi. Milan: Centro Studi Camito-Semitici. Bibliotheca Orientalis 54, 799–800. Gille, Christiane. 1995. Das Kapitel al-Mausūl (‘Das Relativum’) aus dem Manhaj as-sālik des Grammatikers Abū H ayān al-Ġarnāt ī (1256–1344). Hildesheim: G. Olms. Bibliotheca Orientalis 54, 793–795. Khan, Mohammad-Nauman. 1994. Die exegetischen Teile des Kitāb alAyn: Zur ältesten philologischen Koranexegese. Berlin: K. Schwarz. Bibliotheca Orientalis 54, 806–809. Watson, Janet C. E. 1993. A Syntax of Sanānī Arabic. Wiesbaden: O. Harrassowitz. Bibliotheca Orientalis 54, 795–799. 1998 Reviews Endress, Gerhard and Dimitri Gutas. 1994, 1995. A Greek and Arabic Lexicon: Materials for a dictionary of the mediaeval translations from Greek into Arabic, fasc. 2 and 3. Leiden: E. J. Brill. Journal of the American Oriental Society, 118, 108–109. Ingham, Bruce. 1994. Najdi Arabic: Central Arabian. Amsterdam: J. Benjamins. Bibliotheca Orientalis 55, 294–297. Marais, Jean (Ed.). 1996. Quebec’s Aboriginal Languages: History, planning, development. Clevedon: Multilingual Matters. Applied Linguistics 19, 131–135.

bibliography kees versteegh


1999 ‘Loanwords from Arabic and the Merger of d /d.’ Compilation and Creation in adab and luġa: Studies in memory of Naphtali Kinberg (1948–1997). Edited by Albert Arazi, Joseph Sadan and David J. Wasserstein. Winona Lake, IN: Eisenbrauns. 273–286. [= Israel Oriental Studies 19]. ‘The Dutch/Arabic—Arabic/Dutch Dictionary Project: Aims and methods’ (with Jan Hoogland). Revue de la lexicologie 14–15, 13–26. ‘Zayd ibn Alī’s Commentary on the Qurān.’ Arabic Grammar and Linguistics. Edited by Yasir Suleiman. London: Curzon Press. 9–29. Reviews Rizk, Attia. 1992. Ibn aš-Šajarī. Mā ttafaqa lafzuhū wa-xtalafa manāhu. Stuttgart: F. Steiner. Der Islam 76.184–186. 2000 ‘Ed de Moor (1936–2000).’ Sharqiyyāt 12, iii-vi. ‘Editors’ Foreword.’ (With Sylvain Auroux, Konrad Koerner and HansJosef Niederehe). History of the Language Sciences / Geschichte der Sprachwissenschaften / Histoire des sciences du langage. I. Berlin: W. de Gruyter. Volume I, HSK 18.1. XXV–XXXV. History of the Language Sciences: An international handbook on the evolution of the study of language from the beginning to the present = Geschichte der Sprachwissenschaften: Ein internationales Handbuch zur Entwicklung der Sprachforschung von den Anfängen bis zur Gegenwart = Histoire des sciences du langage: Manuel international sur l’évolution de l’étude du langage des origines à nos jours, I. Editor (with Sylvain Auroux, Konrad Koerner and Hans-Josef Niederehe). Berlin: W. de Gruyter. Volume I, HSK 18:1. ‘Language and logic in the Arabic grammatical tradition.’ History of the Language Sciences / Geschichte der Sprachwissenschaften / Histoire des sciences du langage, I. Edited by Sylvain Auroux, Konrad Koerner, Hans-Josef Niederehe and Kees Versteegh. Berlin: W. de Gruyter. Volume I, HSK 18.1. 300–306. ‘Tasrīf.’ The Encyclopaedia of Islam: New edition. Edited by P. J. Bearman, Th. Banquis, C. E. Bosworth, E. van Donzel and W. P. Heinrichs. Leiden: E. J. Brill. X. 360–361.


bibliography kees versteegh

Reviews Enk, Gerrit J. van, and Lourens de Vries. The Korowai of Irian Jaya: Their language in its cultural context. Oxford: Oxford University Press. Word 51, 134–137. Lallot, Jean. Apollonius Dyscole, De la construction (syntaxe). I. Introduction, texte et traduction. II. Notes et index. Paris: J. Vrin. Word 51, 304–308. 2001 ‘Arabic in Europe: From language of science to language of minority.’ Lingua e stile 36, 335–346. ‘Arabic in Madagascar.’ Bulletin of the School of Oriental and African Studies 64, 177–187. History of the Language Sciences / Geschichte der Sprachwissenschaften / Histoire des sciences du langage, II. Editor (with Sylvain Auroux, Konrad Koerner and Hans-Josef Niederehe). Berlin W. de Gruyter. Volume II, HSK 18:2. ‘Linguistic Contacts between Arabic and other Languages.’ Arabica 48, 470–508. Naphtali Kinberg. Studies in the Linguistic Structure of Classical Arabic. Editor (with Leah Kinberg). Leiden: E. J. Brill. Reviews Kropp Dakubu, M. E. Korle meets the sea: A sociolinguistic history of Accra. Oxford: Oxford University Press. Word 52, 163–166. Owens, Jonathan. 1998. Neighborhood and Ancestry: Variation in the spoken Arabic of Maiduguri, Nigeria, Amsterdam: J. Benjamins. Journal of African Languages and Linguistics 22, 101–105. 2002 ‘Dead or Alive? The Status of the Standard Language.’ Bilingualism in Ancient Society: Language contact and the written word. Edited by J.N. Adams, Mark Janse and Simon Swain. Oxford: Oxford University Press. 52–74. ‘al-Zadjdjādj, Abū Ishāk Ibrāhīm b. al-Sarī.’ The Encyclopaedia of Islam: New edition. Edited by P. J. Bearman, Th. Banquis, C. E. Bosworth, E. van Donzel and W. P. Heinrichs. Leiden: E. J. Brill. XI, 377–378. ‘al-Zadjdjādjī, Abu ’l-K āsim Abd al-Rah mān b. Ish āk.’ The Encyclopaedia of Islam: New edition. Edited by P. J. Bearman, Th. Banquis,

bibliography kees versteegh


C. E. Bosworth, E. van Donzel and W. P. Heinrichs. Leiden: E. J. Brill. XI, 378–379. ‘al-Zamakhsharī, Abu ’l-K āsim Mahmūd b. Umar.’ The Encyclopaedia of Islam: New edition. Edited by P. J. Bearman, Th. Banquis, C. E. Bosworth, E. van Donzel and W. P. Heinrichs. Leiden: E. J. Brill. XI, 432–434. Translations ‘Al Jahiz: Traktaat over de lof van de zwarten boven de blanken.’ De taal der engelen: 1250 jaar klassiek Arabisch proza. Edited by Arnoud Vrolijk. Amsterdam: Contact. 135–139. ‘Hunayn ibn Ishaq: De behandeling van staar.’ De taal der engelen: 1250 jaar klassiek Arabisch proza. Edited by Arnoud Vrolijk. Amsterdam: Contact. 279–282. Reviews Abu-Manga, Al-Amin. 1999. Hausa in the Sudan: Process of adaptation to Arabic. Cologne: R. Köppe. Word 53, 292–296. 2003 ‘Arab tradition.’ International Encyclopedia of Linguistics (2nd ed.). Edited by William J. Frawley. Oxford: Oxford University Press. 202–204. ‘De Arabische taalkunde in Nederland.’ Sharqiyyāt 15, 29–42. ‘The Arabic component of the Indonesian lexicon.’ Rintisan kajian leksikologi dan leksikografi. Edited by Lilie Suratminto and Munawar Holil. Jakarta: Fakultas Ilmu Pengetahuan Budaya Universitas Indonesia. 216–229. Woordenboek Arabisch-Nederlands. Amsterdam: Bulaaq. Editor (with Jan Hoogland and Manfred Woidich). Woordenboek Nederlands-Arabisch. Amsterdam: Bulaaq. Editor (with Jan Hoogland and Manfred Woidich). 2004 Approaches to Arabic Dialects: A collection of articles presented to Manfred Woidich on the occasion of his sixtieth birthday. Editor (with Martine Haak and Rudolf de Jong). Leiden: E. J. Brill. ‘From Classical Arabic to the Modern Vernaculars.’ Morphologie: Ein internationales Handbuch zur Flexion und Wortbildung = Morphology: an international handbook on inflection and word-formation. Edited


bibliography kees versteegh

by Geert Booij, Christian Lehmann, Joachim Mugdan and Stavros Skopeteas in collaboration with Wolfgang Kesselheim. Berlin: W. de Gruyter. II, 1740–1754. ‘Grammar and Lexicography..’ Encyclopaedia of Islam and the Muslim World. Editor in chief: Richard C. Martin. USA: MacMillan Reference USA. I, 279–281. ‘History of the Arabic Language.’ Encyclopaedia of Islam and the Muslim World. Editor in chief: Richard C. Martin. USA: Macmillan Reference USA. 58–62. ‘Meanings of Speech: The category of sentential mood in Arabic grammar.’ Le voyage et la langue: Mélanges en l’honneur d’Anouar Louca et d’André Roman. Edited by Joseph Dichy and Hassan Hamzé. Damascus: Institut Français du Proche-Orient. 269–287. ‘One Letter does not make a Word: The doubly weak imperatives in Arabic.’ Words, Texts and Concepts Cruising the Mediterranean Sea: Studies on the sources, contents and influences of Islamic philosophy and Arabic philosophy and science dedicated to Gerhard Endress on his sixty-fifth birthday. Edited by R. Arnzen and J. Thielmann. Leuven: Peeters. 131–141. ‘Phonological Constraints in Arabic Grammatical Theory: The Iltiqā as-sākinayn.’ Romano-Arabica 3, 221–235. ‘Pidginization and Creolization Revisited: The case of Arabic.’ Approaches to Arabic Dialects: A collection of articles presented to Manfred Woidich on the occasion of his sixtieth birthday. Edited by Martine Haak, Rudolph de Jong and Kees Versteegh. Leiden: E. J. Brill. 343–357. ‘Taalkunde en letterkunde in de kleine letteren.’ Gij letterdames en letterheren: Nieuwe mogelijkheden voor taalkundig en letterkundig onderzoek in Nederland. Edited by Joost R. Swanborn. Amsterdam: Koninklijke Nederlandse Akademie van Wetenschappen. 189–203. ‘The Future of Creolistics.’ Linguistics Today: Facing a greater challenge. Edited by Piet van Sterkenburg. Amsterdam: J. Benjamins. 305–319. ‘What’s Where and How’s What? Interrogatives in Arabic dialects.’ Estudios de dialectología norteafricana y andalusí 8, 239–251. 2005 ‘Breaking the Rules Without Wanting to: Hypercorrection in Middle Arabic texts.’ Investigating Arabic: Current parameters in analysis and learning. Edited by Alaa Elgibali. Leiden: E. J. Brill. 3–18.

bibliography kees versteegh


Reviews Talmon, Rafael. 2003. Eighth-century Iraqi Grammar: A critical exploration of pre-H alīlian Arabic Linguistics. Winona Lake, IN: Eisenbrauns. Jerusalem Studies in Arabic and Islam 30, 528–535. 2006 ‘A new treatise about the ilal an-nah w : Ibn al-Warrāq on inna wa-axawātuhā.’ Grammar as a Window onto Arabic Humanism: A collection of articles in honour of Michael G. Carter. Edited by Lutz Edzard and Janet Watson. Wiesbaden: O. Harrassowitz. 51–65. ‘Arabic Linguistic Tradition.’ Encyclopedia of Language and Linguistics: Second edition. Editor in chief: Keith Brown. Oxford: Elsevier. I, 434–440. ‘D ād.’ Encyclopedia of Arabic Language and Linguistics. Leiden: E. J. Brill. I, A-Ed. 544–545. Encyclopedia of Arabic Language and Linguistics. Leiden: E. J. Brill. General editor (with Mushira Eid, Alaa Elgibali, Manfred Woidich and Andrzej Zaborski). I, A-Ed. ‘Introduction.’ (With Mushira Eid, Alaa Elgibal, Manfred Woidich and Andrzej Zaborski. Eds.). Encyclopedia of Arabic Language and Linguistics. Leiden: E. J. Brill. I, A-Ed. v-x. ‘History of Arabic language teaching.’ Handbook for Arabic Language Professionals in the 21st Century. Edited by Kassem M. Wahba, Zeinab A. Taha and Liz England. Mahwah, NJ: Lawrence Erlbaum. 3–12. History of the Language Sciences / Geschichte der Sprachwissenschaften / Histoire des sciences du langage. Editor (with Sylvain Auroux, Konrad Koerner and Hans-Josef Niederehe). Berlin: W. de Gruyter. Volume III, HSK 18.3. Leerboek modern geschreven Arabisch. Grammatica, woorden, idioom en oefeningen. (with Everhard Ditters and Lieke de Jong). Nijmegen: Radboud Universiteit. 1, 2, 3. ‘The Study of the non-Western Linguistic Traditions.’ History of the Language Sciences / Geschichte der Sprachwissenschaften / Histoire des sciences du langage. Edited by Sylvain Auroux, E. F. K. Koerner, Hans-Josef Niederehe and Kees Versteegh. Berlin: W. de Gruyter. Volume III, 18.3. 2791–2802.


bibliography kees versteegh

Reviews Crowley, Terry. 2004. Bislama Reference Grammar. Honolulu. University of Hawai‘i Press. Bijdragen tot de taal-, land- en volkenkunde 162, 572–575. 2007 Encyclopedia of Arabic Language and Linguistics. General editor (with Mushira Eid, Alaa Elgibali, Manfred Woidich and Andrzej Zaborski). Leiden: E. J. Brill. II: Eg-Lan. Encyclopedia of Arabic Language and Linguistics. General editor (with Mushira Eid, Alaa Elgibali, Manfred Woidich and Andrzej Zaborski). Leiden: E. J. Brill. III: Lan-Ser. ‘H araka.’ Encyclopedia of Arabic Language and Linguistics. Leiden: E. J. Brill. II: Eg-Lan. 232–236. ‘ Idāfa.’ (With Karin C. Ryding). Encyclopedia of Arabic Language and Linguistics. Leiden: E. J. Brill. II: Eg-Lan. 294–298. ‘Idmār.’ (With Michael C. Carter). Encyclopedia of Arabic Language and Linguistics. Leiden: E. J. Brill. II: Eg-Lan. 300–302. ‘ Ilġā.’ Encyclopedia of Arabic Language and Linguistics. Leiden: E. J. Brill. II: Eg-Lan. 307–308. ‘Illa.’ Encyclopedia of Arabic Language and Linguistics. Leiden: E. J. Brill. II: Eg-Lan. 308–311. ‘Inna wa-axawātuhā.’ Encyclopedia of Arabic Language and Linguistics. Leiden: E. J. Brill. II: Eg-Lan. 354–358. ‘Interrogative Pronoun.’ Encyclopedia of Arabic Language and Linguistics. Leiden: E. J. Brill. II: Eg-Lan. 387–389. ‘ Isnād.’ Encyclopedia of Arabic Language and Linguistics. Leiden: E. J. Brill. II: Eg-Lan. 434–437.



1. Introduction The Arab grammatical theory as represented by the later grammarians is generally characterized by its focus on the formal aspects of the utterance within well-defined, albeit complex and often controversial rules of usage. Although it is very difficult to represent linearly the shift which took place during the evolution of the theory towards formal aspects at the expense of meaning since there have indeed been some attempts to restore to meaning its primary position in linguistic analysis, it may be safely argued that this shift started immediately after the first major grammatical work, namely Sībawayhi’s (d. 180/796) Kitāb. It is, of course, true that at the levels of morpho-phonology and morpho-syntax Sībawayhi’s formal considerations were adopted almost in their entirety by subsequent grammarians. But Sībawayhi’s method of probing the relationship, at the syntactico-semantic level, between form and meaning was continuously eroded by the grammarians’ attempt to codify rules, systematize usage, and analyze structure largely on the basis of formal considerations which govern its constituent elements. Generally speaking, the continuous shift from Sībawayhi’s method of syntactico-semantic analysis culminates in works from the seventh century A.H. onward, such as Alfiyya commentaries and the extensive sources (mutawwalāt), where pedantic formulae and rigid rules almost fully replace the vivid and dynamic nature of Sībawayhi’s analysis which takes into account both the formal and the semantic aspects of citations and utterances.1 1 In his Muqaddima (1081–1084), Ibn Xaldūn (d. 808/1406) praises Sībawayhi’s Kitāb on the grounds that its author did not confine it to the formal rules related to irāb and that it is replete with proverbs and citations from poetry and speech. Those who study the Kitāb are therefore likely to enhance their malaka (natural linguis-


ramzi baalbaki

One of the most significant aspects of Sībawayhi’s method of grammatical analysis is the pragmatic role he assigns to the speaker (mutakallim), and by extension to the listener (muxātab), as part of the social interaction which language represents for him. This aspect which lies at the core of Sībawayhi’s understanding of language as a form of social behavior and which embodies his originality, and perhaps the essence of his value in the history of linguistic ideas, has unfortunately been severely diminished, if not totally annulled, by later authors. Although this development is obvious in the works which immediately follow the Kitāb—such as Mubarrad’s (d.285/898) Muqtadab and Ibn as-Sarrāj’s (d. 316/929) Usūl—the shift of focus in linguistic analysis from social interaction and context of situation to formal considerations becomes more dramatic in later stages of the history of grammar. Equally unfortunate is that modern scholarship has only recently paid any meaningful attention to the role which Sībawayhi assigns to the speaker, the listener, and the context in which speech takes place. Carter (2004, 56–57) observes that “one of the most striking features of Sībawayhi’s analysis is that it concerns itself almost exclusively with language as behavior: speech is a set of actions, each named according to its intention, e.g. istifhām ‘asking a question’, tatniya ‘making something dual’, tanbīh ‘drawing attention to something”. He further observes (p. 57) that “every utterance takes place in a context of a ‘speaker’ . . . and listener” and that this approach “places great emphasis on the pragmatic roles of speaker, listener and context” and “invites the analyst to propose psychological explanations of linguistic phenomena”.2 Bohas et al. (1990, 38) convincingly argues that, from a typological perspective, “grammatical and linguistic systems can be divided into two rough classes: on the one hand, those which analyze utterances in terms of formal relationships between their components; on the other hand, those which analyze them in terms

tic ability), although some of them end up mastering grammar as a sināa (craft), but not as a malaka. Contrarily, the books of the later authors (kutub al-mutaaxxirīn) are void of poetry and the speech of the Arabs, and contain nothing but grammatical rules (al-qawānīn an-nahwiyya). Readers of such works, according to Ibn Xaldūn, can hardly be expected to enhance their malaka and can only master the craft. See also Zakariyyā (1986, 23 ff.). 2 See also Carter (2004, 95–98) for further discussion of the speaker’s role and its significance to Sībawayhi’s reasoning. It would be particularly interesting to examine in more detail the effect of the speaker’s choice (96) and the speaker’s intention (97) on utterances in Sībawayhi’s analysis of speech.

speaker’s awareness as arbiter of usage


of operations performed by the speaker in order to achieve a specific effect on the allocutee”. Based on this distinction, they maintain that “Sībawayhi’s approach basically belongs to the latter category, while the classical grammarians’, typically belongs to the former”. It must be stressed that what we shall call the linguistic awareness of the speaker is, according to Sībawayhi, an essential component of the competence which is demonstrated in successful interaction with the listener. By linguistic awareness we refer to what he and a few other authors perceive as the speaker’s alertness to the various tools which language places at his disposal and his ability to use them to decide what form best expresses the meaning which he intends to convey to the listener. Within this context, the purpose of the present paper is twofold. First, we shall try to show that although Sībawayhi’s method, which assigns a central role to the meaning and to the speaker’s awareness, was generally not followed by subsequent authors, some of them have indeed attempted to restore that central role, albeit from different perspectives. In a cursory look at the basic elements of the most important of these attempts, it may be possible to establish that what they have in common with Sībawayhi is a genuine concern for meaning and for the role of the speaker. The latter part of the paper looks into how Sībawayhi’s assumption of the speaker’s awareness is practically applied in his analysis of particles whose etymology is essential in determining usage. For this, we shall examine his analysis of those compound particles which may be used as single words or split into their constituent elements.

2. Meaning and speaker’s awareness The relationship between meaning and the speaker’s awareness in Sībawayhi’s analysis cannot be overemphasized. As a general rule, whenever Sībawayhi highlights the role of the speaker in the utterance or his competence in producing correct speech, the reader should expect meaning to be at the center of the author’s argument. The Kitāb abounds with examples which support this conclusion, and it is certainly beyond the scope of this paper to investigate the larger number of formal, semantic and contextual elements which contribute to the link between meaning and the speaker’s awareness in Sībawayhi’s analysis. This notwithstanding, it is important to establish the link between meaning and one of the most basic concepts in the Kitāb, namely taqdīr


ramzi baalbaki

(suppletive insertion of elided elements), mainly because, as a grammatical tool, taqdīr embraces the formal, semantic and contextual elements that are at the core of correct speech according to Sībawayhi. Another compelling reason for establishing this link is that it can help us understand the sharp contrast between Sībawayhi’s approach and that of the later grammarians, for whereas they have largely adopted the formal aspects of his taqdīr, they have, more often than not, ignored the semantic and contextual dimensions of the process of supplying missing elements to structure. We shall not, however, make any detailed comparison between Sībawayhi and the later grammarians in this respect as this would require an independent study. Rather, we shall point out two of the most illuminating principles of the relationship between meaning and taqdīr in Sībawayhi’s system of grammatical analysis.3 The first principle is that in the process of taqdīr, the proposed construction should not contradict the meaning of the original construction, i.e. before the suppletion of the elements which are judged to be elided. In an earlier study of the harmony which Sībawayhi tries to establish, through taqdīr, in several types of constructions (Baalbaki 1979, 7–14), I discussed his method of breaking up one sentence into two, both of which share a common feature and thus demonstrate an underlying harmony. For example, the two sentences Zaydan darabtuhu and a-Abdullāhi daraba axūhu Zaydan are interpreted, at the level of “deep structure”, as *darabtu Zaydan darabtuhu and *a-daraba Abdullāhi daraba axūhu Zaydan respectively (Kitāb I:81, 102). If a nominal sentence is conjoined to a verbal sentence, as in raaytu Zaydan wa-Amran kallamtuhu, he intervenes to restore the harmony by supplying a verb to the nominal sentence, hence the proposed construction *raaytu Zaydan wa-kallamtu Amran kallamtuhu (Kitāb I:88). In defending the restoration of a verb to produce a verbal sentence parallel to the first one, he argues that the introduction of the verb causes no contradiction in meaning (lā yanqud manā; Kitāb I:88–89). Closely related to this argument is his assertion that in utterances which express amr (command) or nahy (prohibition), the verb may be uniformly elided, as in (idrib) Zaydan, (lā taqrabi) l-asada, and (xalli) t-tarīqa (Kitāb I:253–254). Sībawayhi’s discussion of such constructions reveals that the element of

3 Carter (1991, 127–128) notes that although Sībawayhi uses the term taqdīr exceedingly sparingly, only 24 times in fact (see Troupeau 1976, 167), he does give plenty of advice on reconstruction without calling it taqdīr.

speaker’s awareness as arbiter of usage


meaning which justifies the elision of the verb is present in the context of situation since the listener would assume the virtual existence of a verb of which he is the agent. Consequently, the grammarian’s intervention to restore this verb would only elucidate, rather than contradict meaning, based on his understanding of what the speaker has in mind. The second principle which governs the relationship between taqdīr and meaning represents a further stage in Sībawayhi’s analysis of constructions. While he remains faithful to the rule that taqdīr should not contradict meaning, he explores the niyya4 (intention) of the speaker to explain why he may well utter a certain part of the construction although he has another usage in mind. An example of this is his belief that each of the two constructions lam ātika and lā ātīka has the status (manzila) of a noun at the level of the speaker’s niyya, and can thus be interpreted as lam yakun ityānun (Kitāb III:28–29). The grammatical implication of this niyya is only fully revealed when the uttered part is virtually replaced by what the speaker intended to say. Thus, in Farazdaq’s line: mašāīmu laysū muslih īna ašīratan * wa-lā nāibin illā bi-baynin ġurābuhā, nāibin is in the genitive although at the level of the actual utterance it is conjoined to muslih īna, which, being the predicate of laysa, is in the accusative. According to Sībawayhi, the recurrent use of the preposition bi- with the predicate of laysa (e.g., laysū bi-muslih īna) reveals the true intention of the speaker, and hence nāibin is in reality conjoined to a genitive noun which does not feature in speech, but is as valid as an uttered noun. In other words, the preposition, which the speaker has in mind, is syntactically valid and operational, and it determines usage as if it were actually uttered (h attā kaannahum qad takallamū bihā fī l-awwal; Kitāb III:29). As far as meaning is concerned, Sībawayhi asserts that the assumption of the preposition bi- in bi-muslih īna does not alter the meaning5 (lā yuġayyir al-manā) because the preposition is indeed frequently used with the predicate of laysa. Similarly, Sībawayhi intervenes in constructions of the type marartu bihi fa-idā lahu sawtun sawta h imārin/surāxun surāxa t-taklā to supply a verb (i.e. yusawwitu,

4 According to Troupeau (1976, 208), the terms nawā and niyya occur 13 and 27 times respectively in the Kitāb. The concept of “intention,” however, is often expressed by much more frequent terms—including anā (136 times), manā (891 times), arāda (1361 times), etc. (Troupeau 1976, 150; 102)—or by expressions such as kaannahum qālū, tawahhamū, ixtīra, etc. 5 The significance of preserving the meaning of the construction in this line as well as the meaning of other constructions within the context of Sībawayhi’s analysis of the taqdīr of an after fā is discussed by Baalbaki (2001, 186–209, esp. 188).


ramzi baalbaki

yasruxu) before the accusative noun (Kitāb I:355f.). The intention of the speaker is expressed here by the term tawahhama,6 and the actually uttered words lahu sawtun are said to have the same status as yusawwitu. Syntactically, the outcome of taqdīr is the use of the accusative in sawta and surāxa, whereas at the semantic level, it is implicit that the introduction of yusawwitu and yasruxu does not contradict the meaning because these verbs share the same root with the nouns they govern. The study of the relationship between meaning and taqdīr in the Kitāb strongly indicates that in his analysis of constructions, Sībawayhi transcends the levels of grammatical correctness and the effect of the operants on case-endings to examine the speaker’s thinking and the mental processes involved in the choices he makes. This linguistic awareness on the part of the speaker becomes a real arbiter of usage and allows him, for example, to use nāibin where nāiban is expected, or to use sawta h imārin in spite of the absence of a verb from the utterance. It is this feature in Sībawayhi’s analysis that one so much misses in the work of the later grammarians. In spite of that, some authors were indeed interested in the role of the speaker and the effect of his awareness on the speech he produces. From this perspective we shall briefly examine the contribution of two leading figures, Ibn Jinnī (d. 392/1002) and Jurjānī (d. 471/1098), and show how their priorities are largely consistent with those of Sībawayhi’s. In the light of the speaker’s awareness and his internal thinking, we shall then examine the analysis proposed by Sībawayhi for constructions in which the etymology and word-class of certain particles are crucial for the speaker to achieve correct speech and for the listener to comprehend what is meant.

3. Post-Sībawayhi authors The first post-Sībawayhi author whose work reflects serious concern for the speaker’s awareness is certainly Ibn Jinnī. Groomed in the grammatical tradition and himself author of several works which are in full conformity with the general grammatical theory—most notably Sirr sināat al-irāb, al-Luma fī l-Arabiyya, and at-Tasrīf al-mulūkī—Ibn 6 For the various senses in which the term tawahhum is used in the Kitāb, see Baalbaki (1982, 234–237). In this particular case of Farazdaq’s line, tawahhum refers to the speaker’s mental restoration of elided parts in the utterance, resulting in their government of parts actually uttered.

speaker’s awareness as arbiter of usage


Jinnī stands out as a unique scholar whose innovative approach in his most impressive work, al-Xasāis, represents a major step in the direction of determining the theoretical bases of language usage. The traditional phonetic, morphologic and syntactic data becomes in al-Xasāis the subject of study from a methodological and epistemological perspective. Within this framework, the speaker takes a central role in Ibn Jinnī’s interpretation of linguistic activity. His point of departure for this is a firm belief in the intuitiveness7 of native speakers of Arabic as well as in their mental abilities, which are manifest in their production of speech and subtle analysis of a host of linguistic phenomena. A particularly illuminating example is that in which he tries to examine the linguistic awareness of a Tamīmī informer of his. When asked why he would say darabtu axāka, and not *darabtu axūka, but would still use axūka in darabanī axūka, the Tamīmī expresses his astonishment and wittily comments that each of the two expressions has a different perspective (ixtalafat jihatā l-kalām; Xasāis I:76; cp. 1:250). Ibn Jinnī confidently concludes that the Arab native speakers scrutinize the syntactic positions of speech elements (adall šay alā taammulihim mawāqi al-kalām) and that they knowingly and consciously (an mīza wa-alā basīra) assign to each element the position and case-ending it merits. Similarly, he describes vowel mutation which results from the speakers’ sense of lightness (istixfāf ) and heaviness (istitqāl ) as proof of keen insight and fineness of perception and observation (li-quwwat nazarihim wa-lutf istišfāfihim wa-tasaffuh ihim; I:78). Ibn Jinnī’s interest in the intuitiveness of Arab native speakers and in their mental awareness of the various phonetic, morphologic, and syntactic processes involved in speech is best understood as part of his attempt to reveal the h ikma which underlies Arabic and which constitutes its intellectual basis. Whether the discussion relates to the reasons (ilal ), purposes (aġrād) and intention (qasd ) associated with speech (I:237, 245), or to the change which forms, etc. undergo due to recurrent usage (II:31), or to the temporal precedence of one part of speech

7 Ibn Jinnī often expresses the notion of intuitiveness by derivatives of the root t-b- (e.g., sun al-bārī subh ānahu fī an tabaa n-nās alā hādā; tahjum bihim tibāuhum alā mā yantiqūna bihi; a-turāhu lā yuh sin bi-tabihi . . . hādā l-qadr; Xasāis II:117; III:273, 275 respectively). Cf. also the terms salīqiyya and najr (I:76) for intuitiveness, and the expressions min lutf al-h iss wa-safāihi wa-nasāat jawhar al-fikr wa-naqāihi; quwwat nafsihi wa-lutf h issihi; I:239; III:75). See also the comments of Suleiman (1999, 64–65) on the intuition of native speakers and the rationality of Arabic within the more general framework of Ibn Jinnī’s study of talīl (causation).


ramzi baalbaki

over others (II:33), or to the onomatopoeic nature of phonemes within words (II:162, 164), or to the occurrence of two dialects side by side in speech (III:317), the term h ikma is used as an inalienable trait of the original wādi (creator) of Arabic and of his logic in deciding on what is to be used or not and how it should be used. It can thus be argued that the abstract notion of the wādi finds its practical dimension in the role of the speaker (or, more generally, the speech community) since Ibn Jinnī ascribes to him an awareness of the linguistic processes that are the result of the original h ikma. In this respect, Ibn Jinnī’s ideas are very much in line with Sībawayhi’s not only because both of them place the speaker at the center of their linguistic analysis, but also due to their mutual interest in the communication of meaning as the ultimate aim of successful speech. One of the most striking results of Ibn Jinnī’s focus on the speaker’s role and of his firm belief that, at all levels of analysis, the linguistic phenomena of Arabic are essentially rational is his discussion of grammatical awāmil or operants and their relatedness to meaning. Although he adopts the traditional division of these operants into two types, lafzī (formal; expressed) and manawī (abstract), his interpretation of the awāmil in the light of the speaker’s role redresses the imbalance between the two types in the tradition. For example, Jurjānī in his al-awāmil al-mia n-nahwiyya (85–86; 312), labels ninety-eight of the awāmil as lafzī and only two as manawī.8 This imbalance, of course, predates Jurjānī, and it is interesting to note that Ibn Jinnī’s own master, Abū Alī al-Fārisī (d. 377/987), is reported to have authored a book entitled al-awāmil al-mia (Sezgin 1984, 107). By insisting that all types of amal (rection)— i.e., raf , nasb, jarr and jazm, which cover the three nominal types (nominative, accusative and genitive) and the three verbal types (indicative, subjunctive and jussive)—are in reality ( fī l-h aqīqa) produced by the speaker (Xasāis I:109–110), Ibn Jinnī effectively reduces the traditional divide between lafzī and manawī operants merely into a didactic technique that tries to distinguish between amal which is accompanied by an uttered operant and amal which lacks such accompaniment (I:109). In other words, whether an operant is actually uttered or not, rection is in all cases the result of the speaker’s internal thinking and 8 Note also the assertion of some grammarians that no āmil may be classified as manawī unless it cannot possibly be explained as lafzī (lā yudal ilā jal al-āmil manawiyyan illā inda taadd  u  r al-lafī as-sālih ; Suyūtī, Ham I:159).

speaker’s awareness as arbiter of usage


a reflection of the meaning he intends. Based on this and on the similarly untraditional view that qiyās lafzī (formal analogical extension) is not devoid of meaning, Ibn Jinnī confidently formulates the general principle that one cannot but bestow a manawī dimension on what is lafzī, whereas what is manawī may well do without a lafzī dimension (I:111). It may be useful at this point, at the risk of disrupting the historical sequence, to bring into the discussion the sixth-century Z āhirite scholar, Ibn Madā (d. 592/1196). Ironically, Ibn Jinnī’s attribution of amal to the speaker is enthusiastically received by Ibn Madā. He quotes Ibn Jinnī’s statement fa-l-amal . . . innamā huwa li-l-mutakallim nafsihi lā li-šay ġayrihi and highlights his use of nafsihi as a corroborative to emphasize al-mutakallim, followed by the assertion that rection is attributable to nothing other than the speaker (lā li-šay ġayrihi; Radd, 77). Ibn Madā quickly recognizes his odd position as a Z āhirite embracing a Mutazilite view, and thus hastens to resolve the situation by explaining that contrary to the Mutazilites, the doctrine of ahl-al-h aqq (i.e., the Z āhirites) stipulates that case-endings (here, aswāt) are in reality produced by God (innamā hiya min fil Allāh taālā) but metaphorically attributed to man. Irrespective, however, of this modification and of Ibn Madā’s argument that neither the uttered forms of the operants nor their meanings cause rection, his merciless criticism of the grammarians’ focus on amal and taqdīr rests in part on their disregard for both the speaker and the intended meaning. In fact, he often refers to the speaker’s intention (cf. yanwī, 89; yurīd, 93) and assesses the relationship between meaning and taqdīr (80, 109). These two aspects of his theory obviously form his best defense against traditional grammar and firmly place him, albeit from the different perspective of his Z āhirite doctrine, with the few authors who challenged its shortcomings. After Ibn Jinnī, the most important attempt to reinstate a primary role to meaning and the speaker who intends it is undeniably that of Jurjānī. Like his predecessor, Jurjānī followed in the footsteps of the traditional grammarians in some of his works. In addition to al-Awāmil al-mia n-nahwiyya mentioned earlier, such works include al-Jumal, most of which is a didactic summary of awāmil (chapters 2 to 4), and a commentary on Fārisī’s Īdāh entitled al-muqtasid fī šarh al-Īdāh . However, Jurjānī’s two major works in the field of stylistics, primarily Dalāil al-ijāz but also Asrār al-balāġa, represent a major shift from the traditional syntactical analysis of the grammarians. In his Dalāil, he


ramzi baalbaki

makes a zealous appeal for placing meaning at the center stage of grammatical analysis and repeatedly argues that nazm—whose narrow sense corresponds to word order, but which more generally refers to the complex relations among the constituents of a structure—is nothing other than the proper adherence to the discipline of grammar (laysa n-nazm illā an tada kalāmaka l-wad alladī yaqtadīhi ilm an-nahw, 64). As such, nazm should aim at what he calls the meanings of grammar (an-nazm huwa tawaxxī maānī n-nahw, 276, 282, 310, 403–404; cf. Asrār, 65). This means that syntactical rules, which govern the relationships among the various parts of any utterance, can express the exact meaning intended by the speaker since speech formation begins in one’s mind ( fī n-nafs) and only then are words arranged to formulate the outcome of this mental process according to a set of syntactical relationships (43f.). Based on the conviction that the arrangement of meanings (at-tartīb fī l-maānī) is prior to the arrangement of words, and that form is subsidiary to meaning (44–45), it follows that any change in syntax is necessarily accompanied by a change in meaning (86f.) The speaker’s awareness of the intricacies of syntactic relations hence acts as a virtual arbiter in his choice of the nazm, which best expresses the intended meaning. Jurjānī’s theory of nazm is to a large extent a reaction against traditional grammar in which formal aspects acquired greater prominence at the expense of meaning. Even Sībawayhi, it has been suggested (Baalbaki 1983, 12f.), may have been the target of some of Jurjānī’s critical comments in which he accuses the grammarians of giving too little attention to meaning. This notwithstanding, Jurjānī is surely much closer in spirit to Sībawayhi than to the later grammarians. Both authors strive to investigate the internal thinking of the speaker and examine its influence on actual utterances. On a wider scale, Sībawayhi, Ibn Jinnī and Jurjānī, the three most original authors in the related fields of nahw (grammar), philology (ilm al-luġa), and stylistics (balāġa) respectively, share the view that meaning should be the main focus of linguistic analysis. A variety of concepts—such as the speaker’s intuitiveness, competence, intention and awareness of the tools at his disposal—feature in the works of the three authors as part of their study of meaning and the mental processes to which it is related. Unfortunately, however, the three have one more thing in common, for although their focus on meaning and the speaker’s awareness represents the most significant and original aspect of their contribution, that focus gave way in later writings to an ever-growing shift towards formal considerations and pedantic formulae which rel-

speaker’s awareness as arbiter of usage


egate meaning and those concepts that are related to it to a much lower position in their scale of linguistic analysis.

4. The importance of the particles The importance which Sībawayhi attaches to the speaker’s competence and his alertness to the means of successful communication of meaning feature distinctly in his discussion of some particles.9 More specifically, Sībawayhi is interested in proving that the speaker can express the intended meaning only if he is aware of certain peculiarities of these particles. The rest of this paper shall deal with etymology of particles as one such peculiarity, the awareness of which by the speaker, according to Sībawayhi, is crucial for correct speech. It should be noted here that the later grammarians largely adopt Sībawayhi’s analysis of particles and their usage, but this does not at all mean that they have preserved his method of syntactical analysis or embraced the psychological and contextual explanations which he emphasizes so much. Rather, they seem to dwell on the formal side of his analysis, or simply repeat his views without incorporating them into their own methods of analysis or developing them in any meaningful way. Initially, it is essential to introduce the concept of h ikāya, which Sībawayhi applies to those particles which he discusses from the etymological perspective. The original sense of the root h -k-y, “to report; to imitate”, is preserved in Sībawayhi’s use of h ikāya for ‘direct speech’. For example, in the construction qāla Zaydun Amran xayru n-nāsi (III:142), Zayd’s words are reported verbatim and therefore qāla does not govern Amrun, which remains in the nominative as in the original utterance. Derived from this sense is the use of the term to refer to elements that are intentionally not integrated into the syntactical build of the construction. Commenting on the sentence marartu bi-Zaydin, for example, one may respond by saying man Zaydin, in the genitive (II:413). This strategy in dialogue can result in the creation of forms that are only used in

9 The term “particle” is used here in a general sense which includes not only what the traditional grammarians classify as particles proper (h urūf ), but also what they consider to be verbs or nouns. We shall therefore refer, for example, to ammā, interrogative mā, and the halumma as particles although they are traditionally classified as particle, noun, and verb respectively.


ramzi baalbaki

h ikāya, as in the use of ayyayni/ayyīna or manayni/manīna in response to raaytu rajulayni/rijālan (II:407–408). Further removed from the original sense of the term h ikāya is its use to describe proper nouns that are syntactically whole sentences or phrases and are uttered verbatim regardless of their grammatical position. These include attested proper nouns such as taabbata šarran (II:269; III:326), baraqa nah ruhu (II:269; III:326), šāba qarnāhā (II:85; III:207, 326), ibn jalā (III:207), and forms such as xayran minka, dāriban rajulan and min Zaydin (III:328–329), which are hypothetically proposed as proper nouns and serve as a testing device for the full potential of h ikāya. Sībawayhi’s use of the term h ikāya in connection with proper nouns that are made of more than one element is most probably what facilitated his generalization of the term to apply to compound particles, that is, particles which he believes are etymologically made up of more than one element. As a preliminary example, we can consider the particle h aytumā which Sībawayhi describes as h ikāya. This sense of the term is clearly distinguished in the Kitāb from neighboring concepts, most notably ism wāh id and laġw. The latter term, laġw, refers to elements of speech which are otiose or redundant. The negative particle lā, for example, is considered to be laġw when it does not signify negation, as in liallā yalama ahlu l-kitābi (Q 57:29), which is interpreted as li-an yalama in the affirmative (IV:222).10 As for the term ism wāh id, Sībawayhi recurrently uses it in connection with the place name H adramawta, which he cites as an example of two distinct nouns that coalesced into one (e.g., the four chapter titles in II:267; III:296, 374, 475). Although at times he refers to particles which exemplify h ikāya as ism wāh id (II:417–418), h arf wāh id (II:418), or kalima wāh ida (III:115), Sībawayhi clearly demarcates h ikāya, as in h aytumā, from ism wāh id in the sense which H adramawta embodies. He achieves this by looking into the function of -mā in h aytumā and similar particles (see below) and comparing it with the corresponding element -mawta in H adramawta. The difference lies in the relationship which each of -mā and -mawta has with the preceding element (III:331). The introduction of -mā to h aytu-, he observes, does not result in the retention of the final damma in h aytu- since h aytamā, with a fath a, is also attested. Moreover, and more importantly, the introduction of -mā causes a semantic change since to the adverbial sense of

10 Another example of laġw is mā in mahmā (interpreted as mā- + -mā), and after conditional in, adverbial id, and pronominal ayy (III:59–60).

speaker’s awareness as arbiter of usage


h aytu- is added to the conditional sense which h aytumā expresses. Similarly, -mā is introduced to an- in ammā (ammā h aytu- + -mā; innamā> inna- + -mā),11 and the element which is identified as being introduced—either as a prefix, such as lain laalla, or suffix, such as -mā in h aytumā—is not laġw. But there is another dimension to this issue since there are particles, such as exceptive illā, which may seem to be examples of h ikāya according to the above criteria, but which Sībawayhi interprets as non-compound particles. Since Sībawayhi does not use a special term for such particles, we shall refer to them as non-h ikāya or non-compound particles. The differentiation between the two types is crucial for our understanding of the role which Sībawayhi assigns for the speaker in using certain particles. Following is an illustrative list of h ikāya particles, with the most essential characteristics which Sībawayhi ascribes to each. These will be analyzed and then compared to non-h ikāya particles. The h ikāya particles12 comprise the following:


Only rarely does h ikāya involve more than two constituent elements, as in dālika which Sībawayhi mentions side by side with other h ikāya particles that are made up of two elements (III:332), and laallamā in which -mā is added to laalla- (IV:221), itself a compound particle (la- + -alla; III:332). 12 Sībawayhi also uses the term h ikāya in connection with personal and demonstrative pronouns (e.g., anta, hādā, hāulāi, dāka, dālika, etc.; III:332; IV:218), but these are not relevant to our discussion. Also beyond our scope are particles like lan which may well be the result of merging two elements (III:5) but which speakers do not normally recognize as compound particles.


ramzi baalbaki

1. innamā: h ikāya (III:331); h arf wāh id (II:418; III:57); laysat -mā bilaġw (III:57); -mā changes inna- by imparting a new meaning to it (IV:222; cf. II:138); -mā changes (i.e., annuls) the rection of inna(IV:222). 2. annamā: h ikāya (III:331). 3. h aytumā: h ikāya (III:331); -mā does not prevent the final vowel of h aytu- from being either -u- or -a- (III:331); -mā imparts to h aytuthe meaning of conditional (III:331; cf. III:59, 518; IV:222); -mā is muġayyira (III:331; cf. III:59), is unlike -mawta in H adramawta (III:331), and is not laġw (III:331); h arf wāh id (II:418; cf. III:57); has the status of in, immā (III:59) and ayna (IV:221). 4. immā: h ikāya (III:331); -mā is attached (madmūma) to in- and may be elided (III:331–332; cf. I:266; III:141; IV:222); has the status of ammā in ammā anta muntaliqan intalaqtu maaka (III:332); see also h aytumā. 5. ammā (in ammā anta, as in “4” above); introduction of -mā prevents an- from governing the subjunctive and -mā is hence muġayyira (III:331, 332). 6. halumma: h ikāya in both H ijāzī and Tamīmī dialects (III:332; cf. I:252; III:529, 534). 7. idmā: -mā imparts to id- the meaning of conditional (III:56); has the status of innamā and kaannamā (III:57); laysat -mā bi-laġw (III:57); h arf wāh id (III:57). 8. lawmā and lawlā: h ikāya (III:333); -mā and -lā impart a new meaning to law- (III:115; IV:222); -mā is muġayyira (IV:222–223); h arf wāh id (III:115); kalima wāh ida (II:180); as excitative (tah dīd) particles, they precede only verbs (III:115). 9. lammā (which governs the jussive): -mā is muġayyira (i.e., it changes the syntactic properties of lam-; IV:223). 10. kamā: h arf wāh id (III:116). 11. kadā: h ikāya (III:332; cf. III:151); šay wāh id (III:171). 12. kaayyin: h ikāya (III:332; cf. II:171; III:151). 13. kaanna: h ikāya (III:332; cf. III:151, 164); šay wāh id (II:171). 14. kaannamā: h ikāya (III:331); h arf wāh id (II:418; III:57); laysat -mā bi-laġw (III:57); -mā changes the rection of kaanna- (IV:221; cf. II:138). 15. laalla: h ikāya (III:332). 16. laallamā: -mā changes the rection of laalla- (IV:221; cf. II:138).

speaker’s awareness as arbiter of usage


17. rubbamā: kalima wāh ida, like qallamā and other similar particles (ašbāhuhumā; III:115); precedes verbs, unlike rubba- (III:116). 18. h abbadā: kalima wāh ida (II:180). 19. alā and amā (as interrogative particles): h ikāya (III:332). 20. allā and hallā: h arf wāh id (III:5, 115); -lā imparts a new meaning to hal- (IV:222). 21. illā (as a conditional particle): h ikāya (III:332). 22. immā (as a conditional particle): h ikāya (III:332); see also h aytumā. 23. mādā and badamā: see below. Sībawayhi spares no effort to demonstrate the coherency of the above group through cross-references in various and often distant parts of the Kitāb and by using a largely unified terminology to describe the properties of the particles within the group. Furthermore, this coherency is supported by the characteristics, which its members share particularly at the level of meaning (nos. 1, 3, 7, 8, 20), syntax (8, 9, 17) and rection (1, 5, 14, 16). This notwithstanding, the particles within this group considerably vary in their clarity, from the perspective of the speaker, as to whether they are compound in nature or not. For example, one can safely assume that idmā, kadā and laallamā are much more easily recognizable by the speaker as compound particles than, say, halumma or ammā which are far removed from their supposed origin and which can even be the object of disagreement among grammarians concerning what that origin really is (cf. the case of halumma in Suyūtī , Ham II:106–107). The most revealing examples, as far as speaker’s awareness is concerned, are those in which the particle may be either broken down into two elements or used as a single entity. To use the late grammatical term lamh al-asl13 (lit. recognition of origin), one can say that the speaker who splits such particles into two elements is aware of their etymological origin as compound words, unlike the speaker


The term lamh al-asl is used by the later grammarians mainly to refer to the speaker’s recognition of word class as reflected in usage. One example is that proper nouns such as H ārit and H asan, contrary to the norm, may be prefixed with the definite article al- since they are originally adjectives. The article is hence called al-allatī li-lamh al-asl (Suyūtī , Ham I:174–175), and the speaker’s recognition of the adjectival origin (lamh as-sifa) of such proper nouns justifies its prefixation to them (Ibn Aqīl, Šarh 91; Ušmūnī, Šarh I:85–86). Due to the obvious similarity between the recognition of an original word class and the recognition of a particle’s etymology, the term lamh al-asl is perfectly applicable to mā dā, man dā, etc. when they are split into two units.


ramzi baalbaki

who treats them as single-word particles. Sībawayhi specifically deals with this issue in a chapter which he designates for mādā and mandā. Depending on whether dā enjoys the independent status of the relative pronoun alladī or is integrated with mā into a single unit, which Sībawayhi typically describes as ism wāh id (II:417), the speaker respectively uses either the nominative or the accusative in the noun which follows mādā. In Labīd’s line alā tasalāni l-mara mā dā yuh āwilu * a-nah bun fa-yuqdā am dalālun wa-bātilu, dā is given an independent status equivalent to alladī and mā by itself acts as interrogative. The corroborative noun nah bun thus acquires the nominative because it modifies mā, which has the grammatical function of the subject of a nominal sentence (i.e., mubtada). Conversely, mādā could have been treated as a single interrogative particle which grammatically serves as the direct object of yuh āwilu. In this case, corroborative nah ban would be, like mādā which it modifies, in the accusative. It is clear in Sībawayhi’s discussion that the speaker has the choice of using mādā as a single-word particle or splitting it into its constituent elements. The mental process which leads up to the speaker’s decision is not restricted to the deconstruction of the particle into its elements or its retention as one unit. It also has to do with the syntactical ramifications of the speaker’s choice since the case-endings of the noun which follows the particle have to be consistent with that choice.14 The speaker’s competence is thus demonstrable at two levels, namely his awareness of the nature of mādā as a compound particle which admits two possibilities of usage, and his confirmation of this awareness by observing the syntactical implications of each possibility. Furthermore, Sībawayhi takes into consideration the listener’s interpretation of the speaker’s use of the particle in order to prove that successful communication also depends on the listener’s awareness of how the speaker’s choice between two possibilities of usage affects his own response. The correct response to mā dā raayta, for example, would be matāun h asanun in the nominative since mā itself is mubtada and hence nominative. On the other hand, mādā raayta is equivalent to mā raayta since the verb governs 14 Note that Sībawayhi does not refer to the fact that, in actual speech, stress may be an essential factor in differentiating mādā from mā dā. As a single word, mādā would normally receive stress on its first syllable. Conversely, the separation of the two elements would be indicated by a stronger stress on dā than mā, perhaps to underline the likely demonstrative function of dā. In short, the difference in meaning between the two options can be best demonstrated by translating mādā faalta and mā dā faalta as “What have you done?” and “What is this that you have done?” respectively.

speaker’s awareness as arbiter of usage


the single-word interrogative particle which precedes it in both cases, and the answer should thus be in the accusative as in mādā anzala rabbukum qālū xayran (Q 16:30).15 As part of his discussion of mādā/mā dā and mandā/man dā, Sībawayhi mentions two other particles which the speaker can use as a single word or as two separate elements. These are kaannamā and h aytumā (II:418; cf. also badamā II:139). Sībawayhi often examines the function of the suffixed -mā in these and similar particles (see detailed list above), including annulment of rection (cf. innamā) and shift in the part of speech which the particle precedes. For instance, rubba and qalla can only precede nouns—as one cannot say *rubba yaqūlu and *qalla yaqūlu—but the suffixation of -mā causes the new particle to precede verbs only (axlasūhumā li-l-fil; III:115). In such cases, correct speech is contingent on the speaker’s reanalysis16 of such compound particles with respect to their constituents as this would ensure proper distinction between utterances like qallamā yaf alu, innamā yaf alu, kaannamā yaf alu, etc. and their counterparts with relative mā; i.e., qalla mā yaf alu, inna mā yaf alu, kaanna mā yaf alu, etc.17 For his part, the listener is obviously expected to differentiate between the two types for successful communication to take place. As we pointed out earlier, there is a group of particles which may seem to be examples of h ikāya but which Sībawayhi interprets as simple or non-compound (i.e., non-h ikāya) particles. In order to highlight the contrast between these particles and the h ikāya particles, Sībawayhi discusses both types side by side in one of his chapters (III:331–332). It is important, however, to determine at which level this contrast exists. By classifying, for example, the exceptive particle illā with non-h ikāya particles, Sībawayhi does not want to specifically deny that it is historically the result of a merger between in- and -lā, in contrast with the

15 Sībawayhi cites the reverse usage, i.e., the accusative after mā dā and the nominative after mādā. But although this is grammatically explicable, he asserts that to use the nominative after mā dā and the accusative after mādā is the most appropriate manner of response (wajh; aqrab ilā an taxud bihi; II:418–419). 16 Cf. the role of reanalysis in the use of tālamā as discussed by Anghelescu (2004, 115–116). 17 To illustrate this distinction in the case of innamā, for example, we can replace it by inna mā in some Qurānic verses where this is syntactically possible. The resulting constructions are grammatically sound, but they obviously differ in meaning from the original constructions. Cf. innamā/inna mā ūtītuhu alā ilmin indī (Q, 28:78); waman yabxal fa-innamā/fa-inna mā yabxalu an nafsihi (Q, 47, 38); innamā/inna mā tuxzawna mā kuntum tamalūna (Q, 66:7).


ramzi baalbaki

conditional particle illā (see no. 21 in the list of h ikāya particles). But irrespective of whether or not he actually differentiates between the two types of illā at the etymological level, it is obvious that he intends to establish the contrast between h ikāya and non-h ikāya particles at the level of the speaker’s awareness of their compound versus non-compound nature. In other words, the fact that linguistic analysis may well prove the compound nature of exceptive illā and similar particles is of no interest here to Sībawayhi since it has little to do with the speaker’s awareness of their etymology. Based on the fact that proper usage and comprehension of pairs of utterances which have two apparently identical particles is dependent on the distinction between the simple versus the compound nature of each particle in a pair (e.g., exceptive versus conditional illā), we can interpret Sībawayhi’s interest in such pairs as part of his overriding interest in the competence of the speaker in correctly communicating the intended meaning to his listener. As for why the particles which he describes as non-h ikāya are considerably fewer in number than the h ikāya particles, two reasons may be suggested. The first is Sībawayhi’s faith in the speaker’s ability to break down compound particles into their components—although some of them are more easily recognizable as compounds than others—and thus arrive at what the later authors call lamh al-asl. The second reason may have to do with the grammarians’ own failure, due to their largely synchronic and non-comparative approach, to ascertain the compound nature of a large number of particles (e.g., lam, layta, lāta, laysa, kayfa, ayna, etc.). Within the above confines, the particles which Sībawayhi explicitly describes as non-h ikāya (III:332) are the following: 1. illā: As an exceptive particle, it has the status of diflā (oleander)—i.e., it is a simple or non-h ikāya particle—and stands in contrast to the conditional particle illā (see h ikāya particles, no. 21). 2. h attā: Like illā, it has the status of diflā. It should be noted that h attā is the only member of Sībawayhi’s group of non-h ikāya particles which is apparently not made up of two elements and which has no counterpart in the h ikāya group. We list it here, nevertheless, for the sake of completeness. 3. ammā: As an inceptive particle in constructions like ammā Zaydun fa-muntaliqun, it is a non-h ikāya particle which has the status of šarwā (the like of a thing) and stands in contrast to ammā as in ammā anta muntaliqan intalaqtu maaka (see h iyāka particles, no. 5).

speaker’s awareness as arbiter of usage


4. alā: When used inceptively, it is a non-h ikāya particle which has the status of words like qafā (back) and rah ā (quern), in contrast to interrogative alā which is a h ikāya particle (no. 19 above). 5. amā: Like alā, it is a non-h ikāya particle when used inceptively, in contrast to interrogative amā (no. 19 above). The distinction between non-h ikāya particles and their h ikāya counterparts in context is, of course, dependent on the speaker’s competence in using the appropriate syntax, stress intonation, etc. But proper communication—as is implied in Sībawayhi’s text—is also to a large extent a function of the listener’s ability to identify each of two identical particles one of which is simple and the other is compound. This “etymological” distinction is to be assumed in the listener’s comprehension of sentences in which these particles appear. Cf., for example: lā tatīnā fa-tuh additanā illā izdadnā fīka raġbatan (III:32) and wa-illā taġfir lī wa-tarh amnī akun min al-xāsirīna (Q 11:74; note that the choice of these two examples is ours, since Sībawayhi does not provide contrasting examples for illā); ammā Zaydun fa-muntaliqun (III:332), and ammā anta muntaliqan intalaqtu maaka (III:332); and alā innahu dāhibun (IV:235) and alā rajula immā Zaydun wa-immā Amrun (I:289). The fact that the distinction between formally identical particles in the above sentences on the basis of their simple or compound nature is accompanied by certain syntactic peculiarities is not without parallels in the grammatical corpus. The mandatory use of fā after simple or non-h ikāya ammā—e.g., ammā anta fa-muntaliqun versus ammā anta muntaliqan intalaqtu maaka—is strikingly similar to the mandatory use of lām in constructions like in kāna la-sālih an (cf. III:104). Although Sībawayhi does not use a special term for this lām, it acquired later the name of al-lām al-fāriqa because its presence indicates that in is not a negative particle but the lightened form of inna, known as in al-muxaffafa. In other words, lām contributes to the distinction of two formally identical particles both of which, unlike the case of ammā, are not compound. In Sībawayhi’s own words, this lām is mandatory (alzamahā l-lām) in order that the lightened form of inna not be confused with in which has the status of mā (li-allā taltabis bi-in allatī hiya bi-manzilat mā allatī tanfī bihā; II:139).


ramzi baalbaki 5. Conclusion

Sībawayhi’s interest in the speaker’s awareness of the etymology of particles and its effect on his usage and on the listener’s response is yet another proof of his method of grammatical analysis which aims at examining the mental operations which the speaker performs and at determining the formal and semantic effects of these operations. As the study of mādā and mandā versus mā dā and man dā in the Kitāb shows, successful communication between the speaker and the listener (who in turn assumes the role of speaker) is measured by their ability to associate each of the two possibilities of usage (i.e., simple versus compound particle) with the formal and semantic aspects which pertain to it. In this particular case, what the grammarians refer to as “etymology” and is intuitively discerned both by the speaker and the listener represents the crucial factor upon which successful communication hinges. The close association between form, meaning, and speaker’s awareness—as exemplified in Sībawayhi’s analysis of h ikāya and non-h ikāya particles—has certainly been degraded, and at times even totally obliterated by subsequent grammarians. On a wider scale, Sībawayhi’s model of linguistic analysis which, like that of Ibn Jinnī’s and Jurjānī’s, largely rests on exploring the “dialectics” or “interplay”, so to speak, between form and meaning, gave way to an alternative model which is heavily tipped in favor of formal considerations. Proponents of this model consequently failed to delve, as did Sībawayhi, into the mind of the speaker in order to pursue the complex processes which result in his choice of the form that most appropriately expresses the intended meaning and is expected to have the desired effect and elicit the “correct” response from the listener.

6. References 6.1 Primary sources Ibn Aqīl, Šarh = Bahā ad-Dīn Abdallāh Ibn Aqīl, Šarh Ibn Aqīl alā Alfiyyat Ibn Mālik. Ed. by Ramzī Munīr Baalbakī. Beirut: Dār al-Ilm lil-Malāyīn, 1992. Ibn Jinnī, Xasāis = Abū l-Fath Utm  ān Ibn Jinnī, al-Xasāis. Ed. by Muhammad Alī anNajjār. Cairo: Dār al-Kutub al-Misriyya, 1952–56. Ibn Madā, Radd = Abū l-Abbās Ahmad b. Abd ar-Rahmān Ibn Madā al-Laxmī, arRadd alā n-nuh āt. Ed. by Šawqī D ayf. 3rd ed. Cairo: Dār al-Maārif, 1988. Ibn as-Sarrāj, Usūl = Abū Bakr Muhammad b. Sahl Ibn as-Sarrāj, al-Usūl fī n-nahw. Ed. by Abd al-H usayn al-Fatlī. Beirut: Muassasat ar-Risāla, 1985.

speaker’s awareness as arbiter of usage


Ibn Xaldūn, Muqaddima = Waliyy ad-Dīn Abū Zayd Abd ar-Rahmān b. Muhammad Ibn Xaldūn, al-Muqaddima. Beirut: Dār al-Kitāb al-Lubnānī, 1956. Jurjānī, Asrār = Abū Bakr Abd al-Qāhir b. Abd ar-Rahmān al-Jurjānī, Asrār al-balāġa. Ed. By Helmut Ritter. Istanbul: Government Press, 1954. Jurjānī, Awāmil = Abū Bakr Abd al-Qāhir b. Abd ar-Rahmān al-Jurjānī, al-Awāmil al-mia n-nahwiyya fī usūl ilm al-Arabiyya, bi-šarh Xālid al-Azharī. Ed. by al-Badrāwī Zahrān. 2nd ed. Cairo: Dār al-Maārif, 1988. Jurjānī, Dalāil = Abū Bakr Abd al-Qāhir b. Abd ar-Rahmān al-Jurjānī, Dalāil al-ijāz. Ed. by Muhammad Rašīd Ridā. Repr. from the Cairo edition, Beirut: Dār al-Marifa, 1981. Jurjānī, Jumal = Abū Bakr Abd al-Qāhir b. Abd ar-Rahmān al-Jurjānī, al-Jumal. Ed. by Alī H aydar. Damascus: Dār al-H ikma, 1972. Mubarrad, Muqtadab = Abū l-Abbās Muhammad b. Yazīd al-Mubarrad, al-Muqtadab. Ed. by Muhammad Abd al-Xāliq Udayma. Cairo: Dār at-Tahrīr, 1965–68. Sībawayhi, Kitāb = Abū Bišr Amr b. Utm  ān Sībawayhi: al-Kitāb. Ed. by Abd as-Salām Muhammad Hārūn. Cairo: al-Haya l-Misriyya l-Āmma, 1977. Suyūtī , Ham = Jalāl ad-Dīn Abū l-Fadl Abd ar-Rahmān b. Abī Bakr as-Suyūtī , Ham al-hawāmi šarh jam al-jawāmi fī ilm al-Arabiyya. Cairo: Matb aat as-Saāda, 1327 A.H. Ušmūnī, Šarh = Abū l-H asan Alī b. Muhammad al-Ušmūnī, Šarh al-Ušmūnī alā Alfiyyat Ibn Mālik al-musammā Manhaj as-sālik ilā Alfiyyat Ibn Mālik. Ed. by Muhammad Muhyī d-Dīn Abd al-H amīd. Cairo: Dār al-Kitāb al-Arabī, 1955. 6.2 Secondary sources Anghelescu, Nadia. 2004. La langue arabe dans une perspective typologique. Bucharest: University of Bucharest. Baalbaki, Ramzi. 1979. “Some Aspects of Harmony and Hierarchy in Sībawayhi’s Grammatical Analysis.” Zeitschrift für arabische Linguistik 2, 7–22. ——. 1982. “Tawahhum: An Ambiguous Concept in Early Arabic Grammar.” Bulletin of the School of Oriental and African Studies 45, pt. 2. 223–44. ——. 1983. “The Relation between nahw and balāġa: A Comparative Study of the Methods of Sībawayhi and Jurjānī.” Zeitschrift für arabische Linguistik 11. 7–23. ——. 2001. “Bāb al-fā [ fā + Subjunctive] in Arabic Grammatical Sources.” Arabica 48. 186–209. Bohas, Georges, J. P. Guillaume and D. E. Kouloughli. 1990. The Arabic Linguistic Tradition. London: Routledge. Carter, Michael, G. 1991. “Elision.” Proceedings of the Colloquium on Arabic Grammar, Budapest 1–7 September 1991, Kinga Dévényi and Tamás Iványi: eds. Budapest: Eötvös Loránd University and Csoma de Kőrös Society. 121–33. ——. 2004. Sibawayhi. London and New York: I.B. Tauris and Oxford University Press. Sezgin, Fuat. 1984. Geschichte des arabischen Schrifttums. IX. Grammatik. Leiden: E.J. Brill. Suleiman, Yasir. 1999. The Arabic Grammatical Tradition: A Study in talīl. Edinburgh: Edinburgh University Press. Troupeau, Gérard. 1976. Lexique-index du Kitāb de Sībawayhi. Paris: Klincksieck. Zakariyyā, Michel. 1986. al-Malaka l-lisāniyya fī Muqaddimat Ibn Xaldūn. Beirut: alMuassasa l-jāmiiyya li-d-dirāsāt wa-n-našr wa-t-tawzī.


It is well known that Arabic grammar and Islamic law enjoy a peculiarly close relationship. The two sciences are united by a common purpose, to control linguistic and general behaviour respectively, and they share a common methodology, namely the inductive derivation of rules from a linguistic corpus and the deductive application of these universal rules to particular acts of the Muslim. Where they differ is in their sources. Grammar relies on the natural, worldly speech of a select range of human speakers (Bedouin), law on the inspired texts of the Qurān and the H adīt, which are supernatural in origin and holy in status. All legal systems are linguistic codes of one sort or another, spoken or written, but the total dependence of Islamic law on a finite body of revealed and prophetic language is unique. Its modern secular analogue is the type of law which is derived from a written constitution, and here too, the law has to be discovered by an essentially linguistic process, whereby there is often disagreement over the presumed intentions of those (invariably dead) who framed the document. The development of a method for interpreting the language of the Qurān and H adīt took several centuries, and at risk of oversimplification it can be said that the two sciences of grammar and law, aided by imported Aristotelian logic, leap-frogged each other in an evolutionary series, where the advances of one made further progress possible in the other. This paper will review the general similarities in the approach to language in early grammar and law, especially Sībawayhi’s intuitive pragmatism (here in the non-technical sense) and his awareness of the legal implications of grammatical form. There follows a brief account of some grammatical/legal problems discussed in an intermediate phase in the 3rd–4th/9th–10th centuries, and the paper concludes by listing a number of features of legal methodology which can be linked with ideas first noted in Sībawayhi, but which only acquired their fully developed form after the maturing of usū l al-fiqh as a discipline.


michael carter

Islamic legal hermeneutics proceeds from the axiom that, regardless of its supernatural origin, the language of the holy texts conforms entirely to the principles of human discourse. Two short quotations, one from Sībawayhi (d. ca 180/796) and one from aš-Šāfiī (d. 204/820, and probably born about the same time as Sībawayhi) will testify that this notion was well defined at the very birth of the sciences of grammar and law : God’s servants were spoken to in their own speech and the Qurān came down in their language and according to what they mean ibād[u] llāhi kullimū bi-kalāmihim wa-jāa l-Qurānu alā luġatihim wa- alā mā yanūna (Kitāb Der. I:139/Būl. I:167).

In almost identical wording from aš-Šāfiī (the original Arabic could not be checked: this is from Khadduri 1987, 94, and looks very like a quotation from or paraphrase of the Kitāb): God has addressed his book to the Arabs in their tongue in accordance with the meanings known to them.

The context in aš-Šāfiī’s case was the dispute over the possibility of foreign words in the Qurān, while for Sībawayhi it was a syntactical issue of indefinite expressions such as salāmun alayka; although a definite as-salāmu would be expected, the indefinite is an old-established Arab usage that must be accepted, especially when it appears in the Qurān. Aš-Šāfiī is regarded as the first legal theorist to give proper weight to the linguistic aspects of the law, since which time both grammarians and lawyers have shown themselves to be remarkably strict and uncompromising in subordinating the language of God to the linguistic conventions of ordinary Arabic. The latter, for Sībawayhi, was a dialogue between speaker and listener, both being required to conform to what are basically ethical criteria to speak ‘well’ (h asan) and ‘rightly’ (mustaqīm). For the lawyers there could be no dialogue with God, only the contemplation of the written record of what he and his Prophet said, but over time they evolved a system of interpretation in which they played the role of silent listeners to a speaker of their own tongue, under the same conditions as natural speech. For this they constructed an elaborate hermeneutical mechanism (usūl al-fiqh) which, as documented by Ali (2000), exhibits an impressive congruence in many details with the modern branch of linguistics

pragmatics and contractual language in early arabic


known as Pragmatics (henceforth with capital P as a school of thought). The thesis of the present article is that a form of Pragmatics can be discerned in Sībawayhi’s analysis of speech (kalām) long before the the usūl al-fiqh were codified, and, further, that this approach to language passed through at least two stages before it grew into the explicit type of Pragmatics seen in Ali’s sources. For the purposes of this article usūl al-fiqh will be taken in the broad sense of the principles of legal argumentation in deriving law from the texts. As for Pragmatics, here is a recent definition which so closely reflects the usūlīs’ position that it might have been written by one of them, or even by Sībawayhi: The study of language from the point of view of the users, especially of the choices they make, the constraints they encounter in using language in a social interaction, and the effects their use of language has on the other participants in the conversation [. . .] including aspects of deixis, implicatures, presuppositions, speech acts and discourse structure. [. . .] It has been characterized as the study of the principles and practice of conversational performance—this including all aspects of language usage, understanding and appropriateness (Crystal 2000, s.v. ‘Pragmatics’).

Readers of the Kitāb will find all these notions very familiar, and some will be illustrated below. But first it is necessary to introduce the essential elements of Pragmatics as laid down by Grice. Grice (1989) treats speech (writing, significantly, does not fit easily into his model, nor into Sībawayhi’s, see below) as a cooperative activity that is both purposeful and rational, in which the participants understand each other by a logically structured process of ‘conversational implicature.’ Speech is a ‘quasi-contractual matter’, governed by four ‘maxims’ (Grice 1989, 26f): The maxim of Quantity : 1. Make your contribution as informative as is required (for the current purposes of the exchange). 2. Do not make your contribution more informative than is required. The maxim of Quality : 1. Do not say what you believe to be false. 2. Do not say that for which you lack adequate evidence. The maxim of Relation: Be relevant.


michael carter

The maxim of Manner: 1. Avoid obscurity of expression. 2. Avoid ambiguity. 3. Be brief (avoid unnecessary prolixity). 4. Be orderly. All these maxims can be deliberately ignored or flouted, sometimes for personal reasons, e.g. a desire to mislead or deceive, sometimes for rhetorical purposes, e.g. exaggeration, irony, implying what one is reluctant to say explicitly, intentional ambiguity etc. There can be and often is a difference between the words actually spoken (‘sentence meaning’) and what is really meant (‘utterance meaning’). In Larcher (1998) an earlier form of Pragmatics is applied to the category of inšā, i.e. performative utterances of the type qabiltu hādā n-nikāh a “I hereby accept this marriage [proposal].” This is the Pragmatics of J. Searle and J. L. Austin of “How to do things with words” fame (bibliographical details in Larcher), out of which Grice’s maxims were later elaborated. However, Grice has been invoked at least twice over the past few years in articles on Arabic linguistics. Moutaouakil, who quotes the above maxims in full (1990, 233), deals only with the Pragmatics of the relatively late scholastic author as-Sakkākī (d. 626/1229) and will not detain us further, except to remark, following Simon (1993, 15) (where further sources), that it was not until grammar and usūl al-fiqh had themselves achieved systematic perfection that an independent science of rhetoric could emerge, effectively the last of the ‘Islamic sciences’ to appear. The other is a direct comparison between Grice and Sībawayhi made by Buburuzan. She interprets the constructions nima l-rajulu abdullāhi ‘What a fine man Abdullāh is!’ and abdullāhi nima l-rajulu ‘Abdullāh, what a fine man [he is]!’ (1993, 424, Kitāb Der. I:259f/Būl. I:300f) as compound expressions where the second part presupposes a question from the listener. The first part, she says, has violated Grice’s maxim of quantity and requires completion in answer to the question ‘who?’ or ‘what about him?’1

1 It is relevant here to recall a similar analysis of the syntax of the zaydun jāa abūhu structure, “Zayd, his father came”, which Bravmann (1953, 1–36) explained as deriving from a self-answered question, “What about Zayd? His father came”; Sībawayhi loc. cit. actually compares the nima construction to this same type, viz. abdullāhi dahaba axūhu “ ῾Abdullāh, his brother has gone.” Bravmann’s Isolated Natural Subject is not

pragmatics and contractual language in early arabic


Sībawayhi makes more or less the same statement about the syntax of marartu bi-rajulayni muslimin wa-kāfirin/muslimun wa-kāfirun “I passed by two men, one Muslim, one pagan,” where the speaker will choose between the oblique case ( jarr) in adjectival agreement or the independent (raf ) case as if muslimun wa-kāfirun were predicates of elided subjects, because, as Sībawayhi observes (Der. I:182/Būl. I:214, and see Der. I:215/Būl. I:252 for another, similar case), the speaker tries to anticipate the question the listener might pose, either, “what sort of men?” or, “who were these two men?” Like Grice, Sībawayhi is very concerned with the listener’s role in conversation, and there are many linguistic events in which the listener influences the speaker’s choices. In what would be a ‘neither [. . .] nor [. . .]’ construction in English, it is qabīh . i.e. structurally incorrect, to say marartu bi-rajulin lā fārisin “I passed by a man who was neither a knight” without completing it with wa-lā šujā῾in “nor a valiant person” or the like, because “it is an answer to someone who asked you—or whom you have put in the status of having asked—whether you passed by a knight or a valiant man” (Der. I:313/Būl. I:358, translator’s italics). By the same token a listener who answers ‘no’ to the disjunctive question “Is it Zayd who is with you or Bišr?” when one of them is known to be there, has broken the communicative contract so gravely that his answer is classified as muh āl “[morally] wrong, [semantically] absurd,” i.e. an utterance which is self-contradictory and therefore meaningless (Der. I:432/Būl. I:483). As if to reinforce the importance of the listener, Sībawayhi comments that in talking to oneself, e.g. hallā af alu “why don’t I do this,” “you are like the listener” (Der. I:114/Būl. I:136, kunta fīhi ka-l-muxātabi). There is even a discussion of what looks like body-language when Sībawayhi describes how, on seeing the figure of an unknown person, some ‘sign’ (āya, the same word as for the verses of the Qurān!) appears by which you identify him, so you exclaim, “Abdullāh! Good Lord!” Not only that, the same elliptical exclamation (that is, a predicate without a subject) can be uttered when the ‘sign’ by which you identify a person is his voice or perfume, or simply what you hear said about him (Der. I:240/Būl. I:279). intrinsically Gricean, but it accords well with the eminently Pragmatic principle stated by Sībawayhi (Kitāb Der. I:346/Būl. I:394, a notion he acquired from his teacher alXalīl b. Ahmad), that every subject must have a predicate because the listener is expecting it.


michael carter

It will be apparent that as well as the speaker and listener, the physical context is also linguistically relevant. Both Grice and Sībawayhi like to present their data in the form of utterances set in a described situation. In Grice it is undoubtedly fictitious and often involves broken-down cars or sherry parties, in Sībawayhi it probably reflects actual observation, e.g.: an example of the suppression of the verb which could be expressed in normal usage is when you see a man who has just returned from a journey and you say, ‘the best of returns’ xayra maqdamin, [. . .] where the dependent (nasb) form is as if [the speaker] had [syntactically] constructed it on the basis of having said ‘may you return’, viz. qadimta xayra maqdamin, and even though he was not heard to say this expression, the arrival of the other person and the sight of him have the same [linguistic] status as the speaker’s saying qadimta (Der. I:114f./Būl. I:136f.).

The similarity with Gricean Pragmatics is unmistakable here: not only does the speaker engage in a cooperative activity with the listener in a real context, but that context itself can become an active constituent in the grammatical form of the utterance (cf. Carter 2002, 7, where the above item is discussed). In the same way the vocative particle can be dispensed with when the listener is standing right in front of the speaker (Der. I:104, 274/ Būl. I:125, 316); the object of a blessing does not have to be indicated if the intended recipient is obvious from the context (Der. I:131/Būl. I:157)—the speaker however remains free to name the recipient for purposes of emphasis). In a rather violent scenario (Der. I:107/Būl. I:128) the speaker can dispense with the verb and merely shout the person’s name if he sees someone about to be killed or being abused. This is common in warnings, e.g. al-jidār(a) ‘[mind] the wall!’, al-asad(a) ‘[don’t go near] the lion!’, and the cry of at-tarīq(a), at-tārīq(a) ‘[get out of] the way! [Get out of] the way!’2 A striking feature of the Kitāb is the sheer quantity of commercial and contractual talk, much of it admittedly trivial, though it does tell us how preoccupied the Bedouin in the Mirbad were with the price of sheep, camels and wheat.3 But many items are strictly legal in form and con-

2 These are always printed with dependent (nasb) case endings, however, in the circumstances they are bound to be in pausal form. 3 The Mirbad was not only a market but also a place where H adīt scholars came to check their vocabulary (EI 2, art. “Mirbad” by C. Pellat), and this may be one of the reasons why Sībawayhi came to Basra to study ātār or H adīt. As it happens mirbad

pragmatics and contractual language in early arabic


tent, involving contracts, debts, sale of goods and property, testimony, deposits, theft and the activity of fiqh itself. The following list appeared in Carter 1972, 90ff: inna fī alfi dirhamin la-madrabun (Der. I:99/Būl. I:119); baya l-malatā lā ahda wa-lā aqda (Der. I:115/Būl. I:137); tabīu d-dāra h addun minhā kadā wa-h addun minhā kadā (Der. I:137/Būl. I:165: note that the sale of undefined property was not legal, at least among the H anafīs, see Hamilton 1870, 257), lahu ilmun ilmu l-fuqahāi (Der. I:151/Būl. I:181); lahu alayya alfu dirhamin urfan (Der. I:160/Būl. I:190); bitu š-šāa šātan wa-dirhaman, qāmartuhu dirhaman fī dirhamin, bituhu dārī dirāan bi-dirhamin, etc. (Der. I:165/Būl. I:196); hādā dirhamun waznan and others (Der. I:235/ Būl. I:275); inna alfan fī darāhimika bīdun (Der. I:245/Būl. I:285); kam minkum šāhidun alā fulānin (Der. I:256/Būl. I:297); alayhi šaru kalbayni daynan (Der. I:257/B I:298); al-wadīatu ayyuhā l-bāiu (Der. I:284/Būl. I:326); qadiyyatun wa-lā Abā H asanayni = the caliph ‘Alī, (Der. I:310/Būl. I:355); āti l-amīra lā yaqtau l-lissa (Der. I:402/Būl. I:435); ammā juhda rayī (Der. I:418/Būl. I:470). The pre-emptive in šāa llāhu to avoid being bound by an oath may also be mentioned here: it is called istitnā both by Sībawayhi and in later legal terminology (Der. I:399/Būl. I:448, see also Carter 1972, 90 n. 2).

Additional legal or commercial material which has come to light since includes (and is not exhaustive: marginal items have been ignored such as “have you barley, wheat or dates?”, though this may well reflect a mercantile context, Der. I:434f/Būl. I:485f): Placement of merchandise (Der. I:64/Būl. I:76: even in those days it was better to buy from the top of the pile rather than the bottom!); expressions of time (important for contracts, Der. I:90–93, 176/Būl. I:110–114, 208); litigation (Der. I:94/Būl. I:114); price variations (Der. I:122, 170/Būl. I:147, 200); profit, sadaqa and zakāh, rendering accounts (Der. I:165f/Būl. I:196f); weights and measures (Der. I:141f, 183/Būl. I:216, 292f); substitution of goods (Der. I:245/Būl. I:285); giving testimony (Der. I:421/ Būl. I:473); default masc. for mixed genders, i.e. slaves (Der. II:180/Būl. II:174).

One specifically contractual type of utterance will be discussed here in some detail, involving the way prices are stated. In a long analysis (over four chapters, 92–95, Der. I:165–168/Būl. I:195–8) Sībawayhi explores how deals are struck, beginning significantly with the grammar of kallamtuhu fāhu ilā fiyya “I spoke to him face to face” and bāyatuhu

is mentioned once in the Kitāb (Der. II:265/Būl. II:248) but only as an example of the mif al pattern.


michael carter

yadan bi-yadin “I traded with him hand in hand” (i.e. for cash), where the dependent (nasb) forms are obligatory, since they are not literal, but simply mean “immediately, on the spot”, no matter how physically close the listener might be. In other words the legal (utterance) meaning is different from the overt (sentence) meaning: by saying these words in this form a legal obligation is created regardless of their literal meaning. This leads to a whole string of commercial expressions, bitu š-šāa šātan wa-dirhaman, qāmartuhu dirhaman fī dirhamin, bituhu dārī dirāan bi-dirhamin, bitu l-burra qafīzayni bi-dirhamin, axadtu zakāta mālihi dirhaman li-kulli arbaīna dirhaman, bayyantu lahu h isābahu bāban bāban, in all of which the phrase indicating the unit and price (scil. šātan wa-dirhaman “one sheep and one dirham”) must be stated in its entirety, otherwise “the meaning will not be valid” lā yasih h u l-manā, i.e. legally. By this Sībawayhi means only in the case of a contractual intent, since the shortened expressions omitting the price are still meaningful but not in any contractual sense, e.g. bitu šāī šātan šātan “I hereby sell my [collective] sheep, sheep by sheep”, bituhu dārī dirāan “I hereby sell him my house, one cubit”, but this would lead the listener to believe that you were selling your sheep one at a time or that your house was only one cubit in size, and so on. Nevertheless, as he observes, the price or the unit are frequently omitted in ordinary speech, and people will say kāna l-burru qafīzayni “the wheat was [for sale] at two bushels’, omitting the price, or al-burru bi-sittīna ‘the wheat is for sixty [dirhams]’, omitting the unit of quantity. They do this, Sībawayhi says, because in the first instance they know in their hearts ( fī sudūrihim) that bi-dirhamin is meant and that the dirham is the standard price unit (alladī yusaaru alayhi), so it is as if they were answering the question, “How much you get for a dirham?”, while in the second they and the listener both know what they mean, as if someone had asked “What is the price of a load?” and received the answer “The load is [for sale] at sixty [dirhams]” (Der. I:166/Būl. I:196). Sībawayhi advises us to follow the practice of the Arabs in this, though al-Xalīl complicates the picture somewhat by pointing out alternative formulations. Sībawayhi’s Pragmatic approach is self-evident here: he puts the conversation in a real-life setting, which assumes all the Gricean maxims: he distinguishes between utterance and sentence meaning, and he accounts for the grammatical features of the expressions in terms of the extralinguistic situation and the intentions of the participants. If a statement such as yajūzu an taqūla bitu d-dāra dirāun bi-dirhamin (Der.

pragmatics and contractual language in early arabic


I:166/Būl. I:197) were taken out of context there would be no way of knowing whether it was from a legal work, scil. ‘it is [legally] permitted4 to utter [the binding contractual formula] “I hereby sell etc.”, or a grammatical work, scil. it is allowed [by the rules of grammar] to say, “I hereby sell etc.” ’5 Only when Sībawayhi goes on to discuss the possibility of the independent (raf ) case in dirāun here (as advocated by al-Xalīl) do we learn that the topic is syntax. The goal, however, is to determine the legal consequences of the syntactic options. Why else would he raise the matter at all? In the 3rd/9th century a new weapon enters the methodological armory, the concept, borrowed from Aristotle, that science has to be logically structured and its enquiries carried out along systematic, that is, logical lines. For about a century and a half there was a great deal of experimentation before the sciences reached the point where they could be defined and classified, notably in such works as the Mafātīh al-ulūm of al-Xwārazmī, written between 366/976 and 387/997. Two debates from that period will be mentioned here as evidence of the transitional stage in the development of grammar and law. Ibn Wallād (d. 332/943) reports a dispute which originated with Mubarrad (d. 285/898) over the semantic status of commands and prohibitions, whether they were logically complementary, i.e. whether a command is equivalent to a prohibition from doing the opposite and a prohibition is equivalent to a positive command to do the opposite. Ibn Wallād’s wording (Intisār 42) is: kullu amrin amarta bihi fa-anta fī l-manā nāhin an xilāfihi [. . .] fa-idā nahayta amarta bi-xilāfihi hence the positive exclamation h adaraka “beware!” has the negative meaning, “do not approach”. The probable Stoic origins of this controversy (cf. Versteegh 1977, 181) are not the main point of interest here:6 what is important is to observe how the intellectual environment stimulated such arguments while the usūl al-fiqh were still in gestation. Al-Mubarrad is among the earliest of the grammarians to reveal the direct influence


The old definition of jāiz by Bergsträßer (1935, 32) is still the most informative: zuläßig im moralisch-religiösen und zugleich rechtlichen Sinne, und daher rechtsgültig, rechtswirksam. Replace ‘legal’ by ‘linguistic’ to see what the term meant to Sībawayhi: “permissible in a religious-moral and at the same time linguistic sense, hence linguistically valid, linguistically effective.” 5 Sībawayhi’s examples in this section are not unambigously performatives, and they can be, and have been translated elsewhere as literal statements “I have sold” etc. 6 The Stoic term pragmata for the “things done” which are represented by words is the basis of our linguistic “Pragmatics” in the sense of “doing things with words”.


michael carter

of Greek ideas before there is any reliable documentation, suggesting that the ideas circulated informally, perhaps from Christian to Muslim scholars. Ibn Wallād himself is very sceptical, and he flatly asserts that there is nothing intrinsically positive or negative about commands: only by inference (istidlāl) can we determine whether an imperative verb is ordering us to do something or prohibiting us from doing the opposite. The language and style of this argument are exactly what we encounter when the usūlīs take up the theme later in the century. In al-Bāqillānī (d. 403/1013, and a strong supporter of the notion of complementarity), the debate is given a thorough airing (Taqrīb II:198–207), and by this time both the grammatical and the logical techniques are well advanced, and the level of argumentation shows that the Aristotelian dialectic had been fully absorbed. There is a good deal of sophistry from al-Bāqillānī’s opponents, who claim, amongst other things, that imperative and prohibitive verbs have different forms, therefore cannot be complementary to each other. He matches their sophistry with a reductio ad absurdum, putting it to them that by their own criteria a positive act would have to be defined as “not refraining from doing the opposite” ġayru tarki diddihi and he concludes with the practical argument that the meanings of commands or prohibitions are understood naturally and grasped immediately from the words themselves, without the need for any kind of logical inference (while this may not look like Pragmatics, it is really a question of differing opinions on conversational implicature: the methods of inference were a very contentious issue among usūlīs, see Ali 2000, especially ch. 5). There is thus a clear line of progression from Sībawayhi’s casual recognition of amr and nahy as grammatical categories (e.g. Der. I:105/Būl. I:126–7), through Ibn Wallād’s early dialectical treatment to the fully structured arguments of al-Bāqillānī and his fellow usūlīs. In other words, there is now a theory where before there was only data. Under these circumstances there is a universal tendency for theory to triumph over data, and a very illuminating example is the treatment of the word kadā “such and such [an amount]”. For Sībawayhi, it was no more than a dummy numeral (kināya, as he calls it, an “allusion” to any number, just as fulān “so-and-so” is an allusion to any person), and it behaves like interrogative kam, with a dependent (nasb) complement, kadā wa-kadā dirhaman “so and so many dirhams” (Der. I:256/Būl. I:297). By the end of the 4th/10th century a more complicated system makes its appearance, when ar-Rummānī (d. 384/994, Šarh fol. 161r) proposes that kadā by itself has the value of 11–20 (because these numbers are

pragmatics and contractual language in early arabic


regarded as single words) while kadā wa-kādā stands for compound numbers from 21 upwards.7 Ar-Rummānī was famous for the exaggerated logical rigor of his experiments with language, but he did not go as far as he could have done. Others did go further, and by the 7th/13th century we find the fully elaborated system in Ibn Mutī (d. 628/1231, Fusūl 23, further references in 244–5), exactly mimicking the syntax of the numerals: kadā wa-kadā darāhima = from 3 to 10 dirhams. kadā wa-kadā dirhaman = from 11 to 99 dirhams. kadā wa-kadā dirhamin = from 100 dirhams upwards.8

This exquisitely artificial and almost certainly unattested scheme is connected with the H anafīs and (which is not much different in this context) with the Kūfans, and it survived at least until the 19th century in the work of Nāsīf al-Yāzijī (d. 1871, in Nār al-Qirā, see Fleischer I:568 for references). What it shows is that as the linguistic and legal sciences evolved they became increasingly abstract, passing beyond the limits of actual usage. It does not seem likely that Abū H anīfa himself explicitly correlated the syntax of kadā with that of the numerals in this manner, but it fits the reputation of the H anafīs for artificial and over-systematic reasoning, and can be seen as a fine specimen of what can happen when a legal judgement (scil. how much is meant by kadā in such and such a case?) has to depend more on the methodology than on actual speech. Classical Arabic was no longer a living language at this time, and the usūlīs could only consult the rules of the dead language. In doing so they nevertheless assumed that their one-way communication with the texts was a natural use of language. Here follow some illustrations of the inherent similarities between Sībawayhi’s concept of speech and the lawyer’s approach to consultations with God. It has already been said above that God’s speech had to follow the formal rules of Arabic grammar. Furthermore, as the usūlīs were well aware, in order to make himself understood, God had to obey the conventions of human communication. For this reason the usūlīs, exactly like Sībawayhi, strove to account for meaning in terms of the motives of the speaker (man or God) and the real-life context of the utterance. A good specimen from Sībawayhi of the dependence of meaning on 7 Through a scribal error the single kadā (scil. kadā dirhaman) and paired kadā (scil. kadā wa-kadā dirhaman) are not distinguished in the manuscript. 8 Further references in Carter 2003a, endnote 27 referring to p. 180.


michael carter

motive and context is the statement mā atānī rajulun: it can mean “no man [at all] came to me”, “not one man came to me[but several]”, “no [real] man came to me [but a weakling]”, or “no man came to me [but a woman did]” (Der. I:20/Būl. I:27). There are no formal linguistic clues whatsoever as to which meaning the speaker has in mind: the clues lie in the extra-linguistic context, in this case, what question might the speaker be answering, and this is why Sībawayhi spends so much time offering psychological explanations for the speaker’s choices. Some centuries later the lawyers had to do the same for God, trying to penetrate his words to divine his purpose, which they could only do by assuming that he spoke to his servants in the same way as a rational human being would under normal circumstances. Sībawayhi took this for granted. The muxātabūna in this next quotation are the same people who are the muxātabūna of everyday conversation, but this time it is God speaking to them: When [God] said ‘your mothers are prohibited to you’ and so on to the end of his speech (kalām), those being addressed (al-muxātabūna) knew that this was proscribed for them, and affirmed, but God went on to say kitāba llāhi ‘by written decree’ to add emphasis, tawkīdan (Der. I:160/Būl. I:191, on Q 4/23–24).

Striking here is the attempt to explain away the mention of the written decree: writing is ill-suited to Pragmatic analysis for the simple reason that the recipient of writing (or in this case the sender!) is usually absent. There can be no conversation, still less any conversational implicature, without the presence of both participants. Writing is mentioned only occasionally in the Kitāb, and is completely marginal and secondary. For the usūlīs, too, it was not part of normal communication: it was a special case, a kind of act of faith in the future, like farming, lending or borrowing, and other actions which require a presumption of continuity, istish āb al-h āl, i.e. that the recipient would still be alive when the letter was delivered (see Ali 2000, 80). We have to admire the usūlīs for their commitment to the belief that the spoken words of God and Muhammad can be directly experienced through their written record: the paradox was to a large extent resolved by the device of learning the texts by heart so that, once implanted in the memory, they ceased to be a document (mush af, kitāb) and became a virtual oral event (Qurān, h adīt), neurologically the same as a remembered discourse. God as a speaker also has the same privilege as humans of presuming knowledge in his listeners: in Q 3/180 (Der. I:347/Būl. I:395) he says “let

pragmatics and contractual language in early arabic


not those who are miserly with the bounty that has come to them reckon [being miserly] is better for them . . .”, and Sībawayhi explains that God here omits the word al-buxla “being miserly” which would be required as the first direct object of “reckon”, because the listener, al-muxātab, will know from the verb yabxalūna “they are miserly” that miserliness is meant. Anyone who has consulted a tafsīr will be familiar with this method of filling gaps in the Qurān, and the usūlī sources are no exception in the use of this procedure. A second point of similarity between Sībawayhi and the usūlīs is the presumption of sanity. Sībawayhi assumes, without spelling it out, that a speaker will be mentally capable of formulating an idea and conveying it successfully—this is implicit in his criterion of mustaqīm “right”, used for utterances which are fully understood in their intended meaning, and in the term murād “what is intended” for the meaning of speech acts. The lawyers turned this into an overt legal principle. They had to decide who was allowed to speak, and to eliminate those who were not legal persons, and therefore had no voice, such as infants and the insane, and they produced the following conditions of legally valid speech (here paraphrased from Ali 2000, 42 based on al-Āmidī, d. 631/1233): 1. It must be uttered intentionally. 2. It must be intended for a particular listener. 3. The listener must be rational and understand it.

The first condition presupposes sanity, because only a sane person can form an intention at all (legal or otherwise), or indeed be a Muslim for that matter (sanity later formed part of the definition of a Muslim); the second criterion excludes soliloquy, and the third anchors speech in a sane society, as well as giving us a hint as to how the usūlīs saw themselves in the dialogue. To recall the notion of leap-frogging introduced above, it should be noted that these new legal definitions of speech found their way back into grammar, where the speech of the insane, or of those talking in their sleep, or even of birds imitating humans, were excluded for the sole reason that their speech could not be intentional (e.g. ašŠirbīnī d. 977/1570, Nūr 10, though it appeared before his time). God himself comes under the same constraints: in order to communicate with humanity he must speak rationally. His attributes allow for this: he has an intellect, a will and the power of speech, and his language is that of the people he is addressing (there is some literature, which cannot be looked at here, on the requirement that God address his prophets in the language of their own people otherwise the revelation will be in


michael carter

vain: you cannot punish sinners for disobeying laws they are unable to understand). For the usūlīs God’s mode of communication (bayān) is inevitably a major theme in their treatises: they either begin their works with a detailed review of the nature of Arabic and of legal semantics, e.g. the introductory chapters of Abū l-H usayn al-Basrī (d. 436/1044), Mutamad, or else the subject is raised after the epistemological topics have been covered, e.g. al-Bāqillānī (d. 403/1013), Taqrīb, from I:316. For Sībawayhi, and later the usūlīs, lexical meaning is arbitrary. Definition by synonyms only leads to infinite regression (Der. II:339/Būl. II:312), and meaning is nothing more than intention, hence the verb arāda ‘to want’ and its derivatives are among the commonest terms in the Kitāb (1,362 times, plus 20 in the passive, according to Troupeau (1976), s.v., and there are also synonyms). It is clear, too, that Sībawayhi fully aware of the distinction between ‘utterance meaning’ and ‘sentence meaning’: he refers more than once to manā l-kalām and manā l-h adīt “the [integral] meaning of the utterance” i.e. not simply the sum of its lexical parts, and manā itself is almost exclusively used to denote the meaning of speech acts, not of words, such as the acts of expressing surprise, asking a question, giving an order etc., e.g. “the meaning of swearing an oath” manā l-qasam, “the meaning of calling” manā n-nidā, and even of grammatical categories, “the meaning of the dependent form” manā n-nasb, “the meaning of tanwīn”, etc. A significant similarity between Sībawayhi and the lawyers is that they both define the meanings of the particles (h urūf ) in terms of their discourse functions: thus wa- “and” is used to “to bring one thing together with another and join them without indication of order”, and fa- “and [then]” is the same except that “you leave some scope for one to be after the other” (Der. II:330/Būl. II:304), cf. Abū l-H usayn, Mutamad I:20, very concisely, wa- is li-l-jam “for joining” while fa- is li-l-taqīb “for arranging consecutively”. Sībawayhi’s definition of naam “yes” is interesting: instead of the expected “agreement” or “consent” he gives us a rather legalistic definition: naam indicates “promise and belief ”, ida wa-tasdīq (Der. II:339/Būl. II:312), the former implying some kind of contractual commitment (“yes, I promise do it”), the latter indicating assent to a proposition (“yes, I believe what you say”), which in our context could mean believing the seller’s description of the goods or the terms of a contract. Sībawayhi never even asks where meaning originates, but the usūlīs were obliged to agree on an answer before they could proceed to the

pragmatics and contractual language in early arabic


derivation of law from the texts. Here is a case where the debates of the intervening century and a half carried the lawyers a long way from Sībawayhi’s agnostic position. They had to reconcile the potentially infinite backward extension of meaning with the historical fact that Arabic is not the oldest human language. Part of the solution, which will not be discussed here, was to attribute to Adam the bridging role connecting the supernatural Arabic he spoke in Heaven with the temporal world he inhabited after the Fall. He himself did not speak Arabic on earth which appeared only later, evolving naturally until it reached perfection in the time of Adam’s prophetic heir and descendant, Muhammad.9 After Sībawayhi there was considerable discussion of the origins of language, which has been investigated for the grammarians by Loucel (1963–64), with the general conclusion that the choice lay between divine ordination (tawqīf) or human convention (tawādu). The lacuna which Loucel pointed out, that there was no comparable study of the origins of language in the legal sciences, remains unfilled and will not be dealt with here. For our purposes it must suffice to note that there was overall preference for the view that language is in some way conventionally “imposed” (by wad) but the identity of the “imposer”, wādi, is left obscure, perhaps deliberately: it may be God, it may be the first users of Arabic, it may be all users of Arabic who agree amongst themselves on the meaning of a word. The real dispute concerned whether meaning could be imposed independently of a word’s being used. At one end of the spectrum (Mutazilī), a word does have a meaning before it is used, and at the other (Ibn Taymiyya), a word cannot have a meaning until or indeed unless it is used. These issues are well described by Ali (2000): what is important for this paper is that none of these ideas, including the term wad in this sense, are found in the Kitāb. The concept must have emerged later, probably under the influence of the Platonic debate over whether words had meaning by their nature ( physis, cf. Arabic tabīa, replaced in Islām by the creating God) or by imposition (thesis, the same as the wad of the grammarians and lawyers). From the lawyers’ point of view it was important to detach meaning from prehistory: in spite of disagreement about the origins of language, lexical meaning was taken as given, either a priori or as recorded by lexicographical experts (ahl al-luġa), or synchronically by mere usage. It


See Carter (2003b).


michael carter

could then be treated as purely conventional, and in this way the lawyers, after a long debate in which Sībawayhi took no part, eventually returned to his Pragmatist position. Meaning, regardless of where it comes from, is what you do with the language, or, put another way, language is only meaningful for usūlīs when it has legal effects outside the actual utterance. This restriction of meaning to habit and community usage enables the law to control that community’s behaviour. The result is the same for both grammarians and lawyers: speakers are obliged to stay within the habitual codes, whether linguistic or social. We might say that Sībawayhi took a lawyer’s view of language and lawyers a Sībawayhian view. Consider his Pragmatist interpretation of such verbs as raā etc. “to see, consider, regard, be of the view that”, of which he says, “even a blind man can use raā ‘to see’ and say, ‘I regarded Zayd as the good man’ raaytu zaydan-i s-sālih a” (Der. I:13/Būl. I:18).10 The equivalent lawyer’s position is stated by aš-Šaybānī (b. 132/749, d. 189/804, and therefore a contemporary of Sībawayhi): a blind man who has to feel the goods for sale when making a purchase is in the same “place” (mawdi, i.e. legal situation) as a sighted man ( Jāmi 81). The primacy of usage and habit is asserted many times by Sībawayhi, and there is no better illustration (because one senses a tongue in cheek here) than his discussion of expressions of praise and blame. They are not unconstrained, he says, you must follow the speech habits of the Arabs, so you cannot, for example, praise someone for being a tailor or a seed-merchant, still less praise a person in terms normally used of God, e.g. al-h amdu li-zaydin “praise be to Zayd!”. Sībawayhi, or possibly a commentator, allows himself a pun here by saying “that would be a grave sin”, aīm, playing on taīm “magnification”, the name for this laudatory construction (Der. I:214f/Būl. I:251). And although it is correct to use such attested idioms as “he is as close to me as where my waist-cloth is tied” huwa minnī maqida l-izāri, you cannot say “he is as close to me as where the horse is tethered *huwa minnī marbita l-farasi (Der. I:174/Būl. I:206). Abū l-H usayn discusses the interdependence of meaning and use in a similar way (e.g. Mutamad I:17f, 22–28), and it is, of course the central problem of usūlī semantics, as Ali (2000) demonstrates in great detail.

10 The example is perhaps deliberately perverse, as the natural reading would be “I saw the good man Zayd”, but this is in a chapter on “verbs of the heart”, so raā must have the complete sentence “Zayd [is] the good man” as its direct object.

pragmatics and contractual language in early arabic


Finally Sībawayhi’s recognition of the performative and illocutionary uses of speech can be compared with the far more elaborate and selfconscious expressions of the same ideas in the legal works, written after the lawyers had had time to absorb Aristotle’s categories of ‘sentences’ (a post-Sībawayhian innovation), which include vocatives, requests, commands and entreaties. Thus when Sībawayhi defines the vocative noun in yā abdallāhi as “made dependent (nasb) through the suppression of the verb which [in this context] is not expressed” (Der. I:262/Būl. I:303, nasbun alā idmāri l-fili l-matrūki ihāruhu), he does leave it open for others to supply a verb such as unādī “I call out to”, though his choice of “the verb” here is significant, and in fact there are situations where a verb cannot be restored, e.g. subh āna llāhi “glory be to God”, for which there is no verb available to be suppressed. This is enough (especially when taken with Sībawayhi’s other observations on elliptical utterances, see above with xayra maqdamin) to permit the assumption than he was aware of the performative aspects of yā and other speech elements. In the usūlī treatment, those who did regard yā as representing an elided verb such as “I call” are roundly rebuked by Abū l-H usayn (Mutamad, I:20). He firmly rejects this on the grounds that merely to say the word yā and its noun conveys in itself the information that an act of calling has occurred. Here he makes explicit what is left implicit in Sībawayhi, as so often happens. By Abū l-H usayn’s time it had become a polemical issue, and he accuses those who would restore a verb in this situation of either oversimplifying for pedagogical purposes (how condescending!) or misinterpreting what Sībawayhi meant by h arf nidā a “particle of calling”, a performative element by nature, which needs no verb, as Abū l-H usayn observes. What this article has tried to show is that Sībawayhi, perhaps by his own genius, perhaps encouraged and inspired by the community of pioneering Islamic scholars around him, chose to treat language in a manner which shows a remarkable affinity with modern Pragmatics. The nine points which Ali (2000, 3f) regarded as “essential pragmatic insights” are taken from usūlī sources, but equivalents to most of them can easily be found in the Kitāb. Indeed it might have made a tidier version of this paper simply to match them seriatim with material from the Kitāb. These ideas were, so to speak, embryonic in Sībawayhi, but the marriage of Aristotelian logic and Arab linguistic studies in the 4th/10th century led to the birth of a much more self-conscious and program-matic version of the same attitude in legal science, reaching its peak of


michael carter

development in the 7th/13th century works which provide the bulk of the material for Ali (2000). Symbolically we may point to the term qānūn “law” as an indication of the change of direction, though the term has never carried much weight in medieval Islām. It would not have meant much to Sībawayhi, who had half a dozen terms for the correct use of Arabic, one of them sunna, nor is it prominent in medieval legal contexts, but perhaps because it was circulating in the 4th/10th century, e.g. in Fārābī (d. ca 339/950) Ih sā  10, it may have fostered the notion that a systematic code of behaviour along the lines of Greek ethics was desirable and possible. As with grammar from that point on, works on usūl al-fiqh display an impressive mastery not only of the large body of historical Islamic data but also the methods of logical enquiry and dialectical disputation. The two disciplines henceforth continue in parallel, grammar becoming increasingly legalistic to stop the language from changing, and jurisprudence becoming more and more a grammatical analysis of the unchangeable texts. One result is that it was possible to formulate a set of five principles of communication which, as set out by Ali (2000, 64), present an instructive analogue to Grice’s four maxims, here briefly paraphrased: 1. 2. 3. 4. 5.

The speaker’s disposition to make his intention manifest. The speaker’s truthfulness. That what is said should have its due effect. That what is said should be grasped immediately. That the existing conventions should be maintained.

With great diligence and subtlety the classical usūlīs applied these and other principles to the interpretation of God’s speech “in search of God’s law” as it has been put. This had not been Sībawayhi’s goal, but his exhaustive description of how the Arabic language works between speaker and listener prefigured the Pragmatics of the usūlīs and provided a basis (filtered through the subsequent grammatical tradition) for their scholarly exertions. In a sense there was always an implicit Pragmatics in Arabic grammar and Islamic law. On the macro-level Islām itself is a covenant with God, and Islamic society is a kind of social contract in which everything said between Muslims has a contractual dimension, exactly as in Grice’s perception of speech. That is one reason why Muslims are enjoined not to lie to each other, particularly when transmitting the religious knowledge on which the survival of their faith depends. Sībawayhi could afford to take truthfulness for granted, since it does not affect linguistic form, but the

pragmatics and contractual language in early arabic


usūlīs had to stipulate truthfulness as a condition of a valid legal utterance precisely because there was no way to tell from the words themselves whether the speaker was lying, i.e. breaking Grice’s unenforceable maxim of Quality, where lying is likewise formally undetectable. The last point to make is that all this legal matter found in the Kitāb goes to confirm two other aspects of Sībawayhi’s life and work which tie him closely to the lawyers. Firstly, from what little we know of his biography we can deduce that he associated not only with revered authorities on Arabic but also with early legal and religious scholars, and secondly it is evident from the Kitāb that these scholars were the inspiration for much of his technical vocabulary and methodology. While this does not solve the problem of the origins of grammar completely, it does make it more likely that the Kitāb is the creation of a single mind, an unprecedented description of Arabic in all its domains, religious, poetic, public and private, in a theoretical framework which drew deeply upon the principles of the nascent legal system and owes almost nothing to external traditions. This view may not meet with the approval of the dedicatee of the present volume, but it is a tribute to his belief in the right of dissent that such heresies can be published without fear of legal action.

References 2.1

Primary sources

Abū H usayn, Mutamad = Abū l-H usayn al-Basrī, Kitāb al-Mutamad fī usūl al-fiqh. M. H amīdullāh with M. Bekir and H . H anafī, eds. Damascus, 1964. al-Bāqillānī, Taqrīb = Abū Bakr Muhammad b. at-Tayyib al-Bāqillānī, at-Taqrīb wa-liršād as-saġīr. Abd al-H amīd b. Alī, Abū Zayd, ed. Beirut, 1998. al-Fārābī, Ih sā = Abū Nasr Muhammad b. Muhammad, Ih sā  al-ulūm. A. Gonzáles Palencia, ed. Madrid, 1932. Ibn Mutī , Fusūl = Abū l-H usayn Yahyā b. Abd an-Nūr, Ibn Mutī , al-Fusūl al-xamsūn. Mahmūd Muhammad at-Tanāhī, ed. Cairo, 1976. Ibn Wallād, Intisār = Ahmad b. Muhammad b. al-Walīd, Ibn al-Wallād, Kitāb al-Intisār. Monique Bernards, ed. Changing Traditions. Al-Mubarrad’s Refutation of Sībawayh and the Subsequent Reception of the Kitāb, Arabic pp. 1–212. Leiden, New York, Köln, 1997. ar-Rummānī, Šarh = Abū l-H asan Alī b. Īsā ar-Rummānī, Šarh Kitāb Sībawayhi. MS Feyzulla 1984. aš-Šaybānī, Jāmi = Muhammad b. al-H asan aš-Šaybānī, al-Jāmi as-saġīr fī l-fiqh. Margin of Abū Yūsuf, Kitāb al-xarāj. Būlāq, 1884. Sībawayhi, Kitāb = Abū Bišr Amr b. Utm  ān Sībawayhi, al-Kitāb. (1) Hartwig Derenbourg, ed. Le livre de Sibawaihi. Paris, 1881–1889. Repr. Hildesheim, 1970. (2) Kitāb Sībawayhi. Būlāq Press, 1898–1900. Repr. Baghdad, 1965. (3) Kitāb Sībawayhi. Abd as-Salām Muhammad Hārūn, ed. Cairo, 1968–77 (cross-paginated with the Būlāq edition). References to the Kitāb are usually in the form Der/Būl.


michael carter

as-Sīrāfī, Šarh = Abū Saīd as-Sīrāfī, Šarh Kitāb Sībawayhi. MS Atif Efendi 2548. aš-Širbīnī, Nūr = Muhammad aš-Širbīnī al-Xatīb, Arab Linguistics, an introductory classical text with translation and notes [Nūr al-sajiyya fī h all alfāz al-Ājurrūmiyya]. Michael G. Carter, ed. Amsterdam. 1981. 2.2 Secondary sources Ali Mohamed M. Yunis. 2000. Medieval Islamic Pragmatics. Sunni legal theorists’ models of textual communication. Richmond. Bergsträßer, G. 1935. G. Bergsträßer’s Grundzüge des islamischen Rechts: bearbeitet und herausgegeben von Joseph Schacht. Berlin, Leipzig. Bravmann, Meir. 1953. Studies in Arabic and General Syntax. Cairo. Buburuzan, Rodica. 1993. “Exclamation et actes de langage chez Sībawayhi.” Revue Roumaine de Linguistique 38, 421–437. Carter, Michael G. 2002. “Patterns of reasoning: Sibawayhi’s analysis of the h āl.” Proceedings of the 20th Congress of the Union of European Arabists and Islamicists, Part One, Linguistics, Literature, History [= The Arabist, vol. 24–25]. K. Dévényi, ed. Budapest. 3–15. ——. 2003a. “Legal Schools and Grammatical Theory.” Arabistikai islamoznanie. Tom 2. Studi po sluchai 60–godishnata na dots. d.f.n. Penka Samsareva, Simeon Evstatiev, ed. Sofia. 177–183. ——. 2003b. “Talking with and about God, Adam and the Arabic language.” Majāz, culture e contatti nell’area del Mediterraneo. It ruoli dell’ Islam (21st Congress of the Union of European Arabists and Islamicists, Palermo 2002) [= La Memoria vol. 15]. Antonino Pellitteri, ed. Palermo. 197–208. Crystal, David. 2000. A Dictionary of Linguistics and Phonetics. 4th ed. Oxford. Fleischer, Heinrich Leberecht. 1885–88. Kleinere Schriften, gesammelt durchgesehen und vermehrt von A. Huber, prof. Thorbecke, und F. Bühlau. Leipzig. Repr. Osnabrück. 1968. Grice, H. Paul. 1989. “Logic and Conversation.” Studies in the Way of Words, 22–40. Cambridge, Mass. and London. Khadduri, Majid. 1987. Al-Imām Muh ammad ibn Idrīs al-Shāfiī’s al-Risāla fī usūl alfiqh. Treatise on the foundations of Islamic jurisprudence. Translated with an introduction, notes and appendices. 2nd ed. Cambridge. Larcher, P. 1990. “Éléments pragmatiques dans la théorie grammaticale arabe post-classique.” Studies in the History of Arabic Grammar II. Kees Versteegh, Michael G. Carter, eds. Amsterdam. 193–214. ——. 1998. “Une pragmatique avant la pragmatique: ‘mediévale,’ ‘arabe’ et ‘islamique’.” Histoire, Epistémologie, Langage 20, 101–116. Loucel, Henri. 1963, 1964. “Les origines du langage d’après les grammairiens arabes.” Arabica 10, 188–208, 253–291; 11, 57–72, 151–187. al-Marġīnānī, Burhān ad-Dīn. 1870. The Hedaya or Guide: a commentary on Musulman laws, trans. by Charles Hamilton, 2nd edition. Standish G. Grady, ed. London (reference is to the reprint 1963). Moutaouakil, Ahmad. 1990. “La notion d’actes de langage dans la pensée linguistique arabe ancienne.” Studies in the History of Arabic Grammar II. Kees Versteegh, Michael G. Carter, eds. Amsterdam. 229–238. Simon, Udo. 1993. Mittlelalterliche arabische Sprachbetrachtung zwischen Grammatik und Rhetorik. Heidelberg. Troupeau, Gérard. 1976. Lexique-index du Kitāb de Sībawayhi. Paris. Versteegh, Kees [C.H.M.]. 1977. Greek Elements in Arabic Linguistic Thinking. Leiden: E.J. Brill.


1. Introduction Kees Versteegh stimulated the discussion on the history and development of Arab grammatical thinking in a number of his publications. In one of his books (Versteegh 1993, 150), he reflects upon my earlier analysis of al-Farrā’s linguistic methods in his Maānī (Dévényi 1990). He pointed out the insufficient analysis of idmār and its related terms in this author’s work. In another chapter of the same work (Versteegh 1993, chapter five), he assembled data in an attempt to present the interrelationship between grammarians, readers and commentators who worked in the 2nd/8th century. The present contribution would like to pick up these two threads and examine, on the one hand, in some detail the role of idmār in the Maānī l-Qurān of al-Farrā and, on the other hand, analyze the role and place of this grammatical commentary of the Qurān from the point of view of other grammars (mainly Sībawayhi’s Kitāb) and other exegetical works, like for example those of al-Axfaš and at -Tabarī. It is a well-known fact that al-Farrā (d. 207/822) held in great esteem Sībawayhi’s (d. 180/796) Kitāb, which more than twenty years predated his composition. Their starting points and approaches were, however, widely different. Versteegh (1993, 180) has already pointed out that the interests of the two authors lay elsewhere. To this, we can add that alFarrā and Sībawayhi, though working within the framework of practically one grammar—or one grammatical ideal—had widely different aims. While al-Farrā, in his Maānī, used his grammatical knowledge for the analysis of an existing corpus which he described from the point of view of the listener to this text, Sībawayhi aimed at creating, from the point of view of the speaker, a comprehensive grammar in which he used poetical and Qurānic excerpts only by way of illustration.


kinga dévényi 2. An overview of al-Farrā’s methods

In order to have a brief overview of the methods used by al-Farrā, a few examples from the beginning of the Maānī will be presented first. That the corpus al-Farrā is working on is the text of the revelation has special importance. On the one hand, al-Farrā—like grammarians and later rhetoricians—considers that irāb is a necessary prerequisite for the understanding of any text and so the text of the Qurān. On the other hand, since the Qurānic text is usually understood without relying on the irāb endings, there is a strong tendency to analyze different endings at a given place without entailing a change in the meaning.1 This method aims at eliminating the problems posed by the different qirāāt. Although in most of the cases the grammatical analysis only underlines and systematizes the interpretation given by the first exegetes of the Qurānic text, in several cases, however, it is the grammatical analysis which helps to disclose the meaning of the āya. 2.1

Maānī I:3 regarding Q 1:2 al-h amdu li-llāhi

The task here is to determine the vowel ending of the word al-h amd. Step (1): The examination of the readings (qirāāt): According to al-Farrā the readers are in total agreement concerning the raf  ending.2 It is also interesting to note, that al-Farrā only mentions this ending but does not present a grammatical explanation for it.3 Step (2): The elicitation of extra-textual linguistic source: The Bedouins (ahl al-badw) say three things: (a) al-h amda li-llāh, (b) al-h amdi li-llāh, (c) al-h amdu lu-llāh.

1 A notable exception, where different readings reflect a difference in meaning is e.g. Q 5:6. For the analysis of this āya, see Dévényi 1987–88 and Burton 1988. 2 Makram and Umar (1985, I:5) also list al-h amda and al-h amdi among the readings of this verse. 3 This is in contrast with al-Axfaš (Maānī I:9 ff.) whose analysis at this place is rather similar to that of al-Farrā, but who also provides a detailed grammatical analysis of the raf  ending.

 ĀR in the MAĀNī of al-farrĀ IDM


Step (3): The explanation of the variants: For (a), i.e. the nasb ending, al-Farrā (i) gives a grammatical rule according to which h amd is a masdar in place of which a verb could also have been used. This is an example that fixes the rule; (ii) supports it with other similar instances in the Qurān, like: Q 47:4: fa-idā laqītumu lladīna kafarū fa-darba r-riqābi—instead of which one could say in kalām: fa-dribū r-riqāba Q 12:79: maāda llāhi an naxuda . . .—which is the same as: naūdu bi-llāhi; (iii) and props it with Bedouin usage (qawl al-arab) where saqyan laka may be used instead of saqāka llāh. The (b) variant, i.e. al-h amdi li-llāh is treated as one word. As such, it is compared to ibil, where two i vowels follow each other. This is a descriptive explanation making reference to usage. Other parallels are presented as well, among them the (c) variant: al-h amdu lu-llāh. Al-Farrā approached this corpus from the point of view of the listener to the text. And it seems from the second step employed by al-Farrā that the listener could hear some variants. And because al-Farrā was interested not only in the text of the Qurān, but obviously placed it in the context of the Arabic language as a whole, he analyzed these versions. What is even more, he proceeded to explain these variants in detail though—according to his knowledge—these variations were not Qurānic readings. 2.2 Maānī I:7–8 regarding Q 1:7 [. . . anamta alayhim] ġayri l-maġdūbi [alayhim wa-lā d-dāllīna] The task here is to determine the vowel ending of the word ġayr. There are neither variant readings nor extra-textual variants. Step (1): Grammatical analysis: Al-Farrā explains the i in ġayri as a nat to alladīna, mentioning that it is definite (marifa) because of the following word (al-maġdūbi).4 He also

4 It is important to note in this respect, that while the analysis of al-Axfaš goes along the same lines (Maānī I:18) (with the usual difference in terminology), but with one basic difference, i.e. that he does not define negatively the ending of the word, in other words he does not say what it is not, rather contents himself with saying what it is. That is to say, he does not deal with the refutation of grammatically incorrect endings or with


kinga dévényi

categorically refuses to connect ġayri to the preceding word, alayhim. Doing this he seems to argue against the view of those among his contemporaries who carry the surface descriptive analysis or the analysis based upon proximity (itbā) to the extremes. Step (2): Semantic analysis: Turning to the semantic side of the explanation, he states that ġayri is connected to wa-lā d-dāllīna which follows it and its meaning is lā. He also adds that if it meant siwā, it could not have been followed by lā. So it can be established that al-Farrā connects the two types of analysis, the semantic and the grammatical. 2.3 Maānī I:11–12 regarding Q 2:2 [dālika l-kitābu lā rayba fīhi] hudan li-l-muttaqīna The question is whether hudan is in raf  or nasb. Though in the case of this particular word both endings are realized in the same surface form, the question should be decided both from the point of view of the semantic interpretation of the structure and for the sake of other similar structures in the Qurān where even the surface realization is different. In the case of similar phrases in the Qurān a number of qirāāt have been preserved with readings in both raf  and nasb, as e.g. Q 31:1–3 (ā lām tilka āyāt al-kitāb al-h akīm hudan wa-rah matun/an li-l-muh sinīn) and Q 11:72 (a alidu wa-hādā balī šayxun/an). Step (1): Grammatical statement: Both endings can be explained in different ways (wajh). If the interpreter of grammatical structures does not deviate from the intended meaning, he may freely choose between raf  and nasb or the different ways. This freedom is expressed by al-Farrā in the use of the 2nd person: idā aradta . . ., wa-in jaalta . . . rafata . . . (Maānī I:11, 12ff.). Step (2): The detailed grammatical analysis: (i) raf  (1) It can be the xabar of dālika l-kitāb which—in this case—is analyzed as the mubtada. A paraphrase is given: dālika hudan.

the refutation of grammatically not permissible analyses. This difference will remain characteristic throughout the two books.

 ĀR in the MAĀNī of al-farrĀ IDM


(2) It can be a nat (modifier) of the xabar: lā rayba fīhi. A Qurānic parallel is quoted: wa-hādā kitābun anzalnāhu mubārakun (Q 6:92, 6:155). (3) It can also be considered the beginning of a new structure which comes after a complete phrase (alā stināf li-tamām mā qablahu).5 A Qurānic parallel is quoted: wa-hādā balī šayxun (Q 11:72).6 (ii) nasb alā l-qat (1) It is either cut off from al-kitābu, which in this case would be analyzed as the xabar of dālika. (2) Or it is cut off from the -hi in lā rayba fīhi. A paraphrase is given as: ‘lā šakka fīhi hādiyan’. A ready-made grammatical rule is given as an explanation for both possibilities: li-anna hudan nakira ittasalat bi-marifa qad tamma xabaruhā fa-nasabtahā li-anna n-nakira lā takūnu dalīlan alā l-marifa. 2.4

Maānī I:14–15 regarding Q 2:16 fa-mā rabih at tijāratuhum

The task is to define the reason why certain structures are permissible in contrast to other seemingly similar structures that are not. Step (1): Parallels from kalām al-arab: hādā laylun nāimun Step (2): Parallels from the Qurān: Q 47:21 fa-idā azama l-amru Step (3): Semantic definition The permissibility of specific structures greatly depends upon their communicational value, i.e. they can only be permitted if they can be understood unambiguously (ulima manāhu I:14, 17). In these examples, the reference of the verbs and the adjective is unambiguously not to

5 In connection with this third analysis, attention should be called to the difficulties of grammatical and semantic analysis arising from the lack of punctuation. No wonder that writings on qat and istināf developed into a special branch of the Qurānic sciences. On the use of istināf and related terms in early grammar and exegesis in general, see Versteegh 1993, 132–136 and the literature cited there. 6 The reading in nasb can be found at this place in today’s printed editions, and this was al-Farrā ’s reading as well (Maānī I:12, 2). The raf  ending was read by Ibn Masūd and Ubayy among others (see Makram and Umar 1985, III:125).


kinga dévényi

the other word in the phrase which is a common noun (tijāra, layl, amr) but to the ‘people’ who act ‘behind these words’. Similar structures are not permitted in the case of possible ambiguity, e.g. *qad xasara abduka is not permitted because of the ambiguity of the word abd (he can both be trader and the object of trade), as al-Farrā puts it, its meaning cannot be known: fa-lā yulamu manāhu (Maānī I:15, 3). 2.5 Maānī I:16 regarding Q 2:17–18 wa-tarakahum fī ulumātin . . . summun bukmun umyun The task is to explain the endings of summ, bukm and umy. (i) The raf  ending is explained by istināf and the completeness of the preceding clause in itself (li-anna l-kalām tamma). An additional reason is that the whole phrase is split between two āyas, but this is not a prerequisite of istināf. It should, however, be mentioned that istināf is frequently found at the beginning of āyas. (ii) The nasb reading (summan bukman umyan) is explained in two ways: (1) according to the meaning (alā l-manā), i.e. by referring back to the verb in the preceding clause (tarakahum); (2) by their being expressions of blame (damm). A parallel Qur’ānic passage quoted by al-Farrā is Q 9:111–112: wa-man awfā bi-ahdihi mina llāhi . . . at-tāibūna l-ābidūna l-h āmidūna—explained as istināf or the reading in nasb . . . at-tāibīna l-ābidīna l-h āmidīna—explained as qat The use of the explanatory terms istināf and qat shows clearly that for al-Farrā the most important issue was to understand the text as a whole. And one of the first issues to be dealt with was the correct segmentation of the text. The term qat expresses more clearly the syntactic structure, whereas the term h āl can only refer to the meaning.7 2.6 Maānī I:17 regarding Q 2:19 yajalūna asābiahum . . . h adara l-mawti Meaning, however, can also play a decisive role in the explanation of an irāb ending and the determination of the syntactic structure, as it is the


For the difference in al-Farrā’s usage between hāl and qat, see Kinberg 1996, 194.

 ĀR in the MAĀNī of al-farrĀ IDM


case, e.g. in Q 2:19 (Maānī I:17, 1ff.): yajalūna asābiahum . . . h adara l-mawti. The word h adara is in nasb not because the verb yajalūna affects it (*yajalūnahā h adaran) but by way of specification (tafsīr). It is the same as e.g. in the phrase ataytuka xawfan where ‘fear’ is not given but something is given because of it.

3. The explanatory technique of idmār in the Maānī The above examples were presented to show the steps followed by al-Farrā in his explanations. One of the basic explanatory techniques used by al-Farrā in the Maānī is the term idmār (together with its related forms). In the course of the explanation of the text of the Qurān, al-Farrā is basically confronted with three types of problems: (i) The meaning of the āya needs clarification; (ii) Though the meaning is clear, there is a problematical irāb-ending; (iii) The qirāāt of similar structures at different places in the Qurān should be harmonized and explained. It will be seen that the technique of idmār features as a prominent solution in all of these cases. It is outside the scope of the present paper to examine all the occurrences of idmār and its related terms in the Maānī. In the following, however, we shall try to present a typology of the most significant types of occurrences of this term. 3.1 idmār is supported by another place in the Qurān where the suppressed element occurs in the text 3.1.1 Maānī I:13 regarding Q 2:7 xatama llāh alā qulūbihim wa-alā samihim wa-alā absārihim ġišāwatun This āya has a variant reading, ġišāwatan, which entails no difference in the meaning of the āya (wa-manāhumā wāh id).8 This reading is explained by others—says al-Farrā—by the idmār of jaala. This

8 It might be interesting to note that al-Axfaš did not mention the existence of different readings here (see al-Axfaš, Maānī I:34).


kinga dévényi

is supported by Q 45:23 where the phrase actually occurs as wa-jaala alā basarihi ġišāwatan. Al-Farrā accepts this explanation, i.e. the use of another Qurānic place to support the explanation by idmār, though he emphasizes that were it not for the Qurānic parallel, the idmār in this sentence could not have been accepted. This is because according to al-Farrā, idmār is acceptable only in those utterances (kalām) which are coherent (yajtami), i.e. where the beginning refers ( yadullu alā) to the end. The idmār is acceptable (yah sunu) if it is well known (urifa; or at other places malūm). A kalām example is brought in to illustrate what is meant by the term yadullu alā, and how it operates in structures. The example is as follows: qad asāba fulānun al-māla fa-banā d-dūra wa-l-abīda wa-limāa wa-l-libāsa l-h asana. It can be seen from this example, proceeds al-Farrā, that the notion of ‘building’ (binā) cannot be extended to slaves and clothes. By quoting this example from kalām, al-Farrā appeals to the listener’s linguistic insight and competence in the Qurānic text. He explains that the first part of the utterance (asāba) refers to the slaves, etc., i.e. qad asāba fulānun al-māla fa-banā d-dūra wa-[asāba] l-abīda wa-l-imāa wa-l-libāsa l-h asana. Without saying that it should be repeated, instead he uses the term dalla, yadullu alā. Sībawayhi would rather say explicitly that asāba should be repeated before al-abīda. Al-Farrā only hints at the idmār of asāba. In the following (Maānī I:14), Q 56:22 is mentioned as a similar example: yatūfu alayhim wildānun muxalladūna bi-akwābin . . . wah ūrin īnin.9 Al-Farrā accepts the -in reading on the basis of proximity (itbā) but mentions that those who put it in raf  (wa-h ūrun īnun) do this on the basis of the meaning, since the cups cannot be co-ordinated with the beautiful companions, and by the idmār of a word like indahum, fīhā, maa dālika. Again, as in the previous example from kalām, al-Farrā does not use the term idmār here, but only hints at it. The limits of idmār are imposed by understandability and coherence, or rather their lack. This is called by al-Farrā qillat al-ijtimā. So. e.g. the idmār of qataltu is not permissible in the following utterance: darabtu fulānan wa-fulānan [*qataltu]. The reason is that there is no reference

9 Mainly because he considered it the accepted reading. This was the reading of a great number of readers, among them al-Kisāī (cf. Makram and Umar 1985, VII:65), but in today’s printed mush af the -un reading can be found.

 ĀR in the MAĀNī of al-farrĀ IDM


(dalīl ) to the meaning here, the speaker’s intention cannot be known by the listener. At- Tabarī adds (Tafsīr I:87) that xatama cannot refer to the eyes because they are never described by xatama either in the Qurān or in the kalām. 3.1.2 Maānī I:35 regarding Q 2:72: wa-id qataltum nafsan fa-ddāratum fīhā The jawāb of id is present neither in this āya nor several others similar to it in the Qurān. Al-Farrā explains the structure by the idmār of udkurū before id, and supports this explanation by similar utterances in the Qurān where this word is present, as e.g. in Q 8:26 (wa-dkurū id antum qalīl . . .). Suppression of udkurū with id can be inferred from places like Q 8:26 and Q 7:86. al-Farrā adds that if udkurū was not mentioned in Q 7:86, you could still infer that this is what is intended because it occurs previously. A parallel case of idmār is also mentioned: Q 7:73 (wa-ilā Tamūda axāhum Sālih an) where the known meaning allows the explanation of this structure by the idmār of arsalnā, i.e. arsalnā Sālih an. This example in itself can be regarded as a sub-case of the acceptance of idmār on the basis of parallel Qurānic passages, since though the verb arsala is used in the Qurān, it is mainly the content of the āya which determines the selection of the suppressed element.10 3.1.3

Maānī I:166 regarding Q 2:246 wa-mā lanā allā nuqātila

Al-Farrā starts his analysis with comparing similar utterances from different parts of the Qurān, some of them with an and some other without it, like e.g.: Q 14:12: mā lanā allā natawakkala alā llāh Q 57:8: wa-mā lakum lā tuminūna bi-llāhi wa-r-rasūlu yadūkum lituminū bi-rabbikum Dropping (ilqā) an here cannot be considered deficiency (illa) according to al-Farrā because of its frequent usage in the arabiyya. The usage of an relies on the meaning of manaa. According to this explanation,


On the meaning of this term in the Maānī, see Kinberg 1996, 377.


kinga dévényi

the utterance mā laka lā tusallī fī jamāa means: mā yamnauka an tusalliya. The an may be inserted into the mā laka phrase because the latter’s meaning is identical with manaa. And a last evidence is put forward from the text of the Qurān itself: Q 7:12: mā manaaka allā tasjuda id amartuka runs parallel to Q 15:32: mā laka allā takūna maa s-sājidīna. idmār is supported basically by kalām


3.2.1 Maānī I:32 regarding Q 2:48 wa-ttaqū yawman lā tajzī nafsun  an nafsin šayan Al-Farrā explains this passage by the idmār of fīhi after tajzī, insisting on the necessity of distinguishing by reference to the meaning adverbial structures allowing idmār from verbal structures with sifa (prepositional phrase) complement, where idmār is not possible. For example, anta lladī takallamtu fīhi should be distinguished from anta lladī takallamtu. In doing this he argues with al-Kisāī and others who understand too rigidly the formal description and identify the two, formally similar but semantically different types. On the basis of this identification al-Kisā’ī refutes both structures with idmār while others permit both of them. The same holds true, says al-Farrā, for the fīhi—hu alternation within adverbial structures of time and place but not in verbal prepositional phrases. The phrase ātīka yawma l-xamīs is interchangeable with ātīka fī yawmi l-xamīs, since the sifa and hā here agree in meaning (muttafaq manāhumā). But when the meaning differs it is not allowed to suppress fī in place of hā and vice versa: yuh ibbuhā does not equal with yuh ibbu fīhā. at-Tabarī, who quotes extensively (Tafsīr I:203, 11ff.) al-Farrā’s interpretation, adds that the idmār in this āya is possible because it is wellknown (malūm), this being the generally accepted explanation for idmār by the exegetes (ash āb at-tawīl). 3.2.2

Maānī I:113, 9, Q 2:185: wa-li-tukmilū l-idda

The li- + verb cannot be šart11 (cause) of a previous verb because the wāw blocks its impact. Al-Farrā discusses in many places that some particles,


On the meaning of this term in the Maānī, see Kinberg 1996, 377.

 ĀR in the MAĀNī of al-farrĀ IDM


especially the wāw, do not allow us to suppose a grammatical link with a preceding regent. This phenomenon is explained on the basis of kalām examples: jituka li-tuh sina ilayya cannot be transformed to jituka wa-li-tuh sina ilayya only if you mean ( jituka) wa-li-tuh sina ilayya jituka. This second occurrence of the same verbal phrase is, naturally, suppressed (al-arab tudxiluhā fī kalāmihim alā idmār fil badahā). This rule stated, al-Farrā cites many Qurānic places where the same causal structure—wa-li and the idmār of a verbal phrase—can be observed signifying the result of this cause, e.g. wa-kadālika nurī Ibrāhīma malakūta s-samawāti wa-l-ardi wa-li-yakūna min al-mūqinīna (Q 6:75) where the suppressed (mudmar) phrase is a repetition of the verbal phrase before wāw: araynāhu. At-Tabarī (Tafsīr II:88, 3ff.) quotes two opinions. One is a verbatim quotation of al-Farrā’s view without mentioning his name (qāla bad nahwiyyī l-Kūfa). The other is the view of bad ahl al-arabiyya which was rejected by al-Farrā. This supports our analysis that whenever al-Farrā rejects a seemingly only theoretical possibility of analysis, he rejects the opinion of a certain group of people without attributing their view to anybody. 3.2.3 Maānī I:271 regarding Q 4:46 mina lladīna hādū yuh arrifūna l-kalim The meaning is: man yuh arrifūna l-kalim. This is supported by kalām examples where man is suppressed in the beginning of the utterance (mubtada al-kalām): minnā [man] yaqūlu or minnā [man] lā yaqūlu. This is possible according to al-Farrā, because min, i.e. the first part (awwal) of the utterance refers to the meaning of what has been left out, being its part (bad). Al-Farrā considers it important to note that man cannot be left out (idmār) as a rule, but only in those cases where the prepositional phrase refers to it. This latter sometimes may be fī too: fīhā sālih ūna wa-fīhā dūna dālika. 3.3 idmār is supported by the immediately preceding context (the first part of the utterance or the previous utterance): Maānī I:141–142 regarding Q 2:220: wa-in tuxālitūhum fa-ixwānukum With idmār of the rāfi, i.e. the nominal subject: hum, or if it is nasb, with idmār of the second occurrence of the verbal phrase: fa-ixwānakum tuxālitūna. In both cases the suppressed element is given in the first


kinga dévényi

part of the utterance or, sometimes, in a previous utterance, as in Q 2:239: fa-in xiftum fa-rijālan. This is called ijtimā al-kalām. Here, says al-Farrā, it would not be correct to suppose huwa because it is not an (everlasting) state (dāim) but an action. The previous āya (h āfiū alā s-salawāti . . . wa-qūmū li-llāhi . . .) gives a clue for the explanation. Hence the meaning can be restored as: in xiftum an tusallū qiyāman fa-sallū rijālan. At-Tabarī (Tafsīr II:208) follows al-Farrā’s argument, and refers to the same Qurānic passage, etc. but everything is more detailed. It is as if he worked from a fuller version of the Maānī. 3.4 idmār is supported primarily by the readings of Abdallāh and Ubayy: Maānī I:156 regarding Q 2:240: wa-lladīna yutawaffawna minkum wa-yadarūna azwājan wasiyyatan Here al-Farrā presents two readings, nasb (wasiyyatan) and raf  (wasiyyatun), and both of them are explained on the basis of idmār. The first idmār relies on the rule of the absolute object, the suppressed verb being a jussive of the same root (li-yūsū . . . wasiyyatan). He refuses, however, to accept the supposition that the previous verb in the yadarūna azwājan phrase could have put wasiyya into nasb. This may have been held by the supporters of the proximity-theory who always preferred the explanation on the basis of the neighboring word. The second, raf, reading (wasiyyatun) may also be explained by idmār. The source of this idmār is the reading of Ibn Masūd or Ubayy:12 fa-matāun li-azwājikum or kutiba alaykum al-wasiyyatu li-azwājikum, both containing raf , the first also a lexical variation.

12 From among the 64 Qurān readers whom al-Farrā quotes by name, Ibn Masūd is by far the most frequently quoted, with 411 references. Ubayy is the seventh most frequently quoted reader, with 92 occurrences. Cf. Dévényi 1991, 160–161.

 ĀR in the MAĀNī of al-farrĀ IDM


3.5 The use of major, meaning-triggered grammatical rules in supposing idmār 3.5.1

h āl Maānī I:24 regarding Q 2:28: kayfa takfirūna bi-llāhi wa-kuntum amwātan The meaning is established as ‘qad kuntum’, since the second verb is a h āl to the first one and it refers back to a state prior to the past action denoted by the first verb. After the definition of the meaning, al-Farrā formulates the following rule: this type of h āl necessitates qad, either overtly (ihār) or in a suppressed way (idmār). If this condition is not fulfilled, i.e. the first verb does not refer to the past (like e.g. in the case of kāda, asā) then the second cannot contain qad in either way. This rule is reinforced by another type of proof, a parallel place in the Qurān where qad is being suppressed and where it is generally understood to be necessary for the correct meaning (Q 12:27: in kāna qamīsuhu qudda min duburin fa-kadibat). At-Tabarī (Tafsīr I:146, 18ff.) follows al-Farrā’s arguments but he adds that the explanation for the suppression of qad lies in the fact that if the verbal form faala takes the place of h āl it is self-evident (malūm) that it requires qad. It is to be emphasized that al-Farrā also considers that only those elements can be suppressed which are well-known to the speakers. Maānī I:372 regarding Q 7:4: fa-jāahā basunā bayātan aw hum qāilūna The meaning is established: [wa]-hum qāilūna is a h āl, parallel to bayātan. Its grammatical rule is: wa-huwa fāilun, accordingly the idmār of wa- is compulsory. The wa- should be present either overtly or in a suppressed way. All this is supported by an example from kalām where both ways (ihār and idmār) are correct: ataytanī wāliyan aw wa-anā mazūl or awanā mazūl.13

13 Az-Zajjāj (Maānī II:317) refutes this explanation without mentioning al-Farrā and says that it is not necessary to suppose wa- in h āl in general.


kinga dévényi


The delimitation and rules of idmār in quotation (hikāya) Maānī I:38 regarding Q 2:58 wa-qūlū h itta tun Considering h itta tun as a h ikāya entails the idmār of the rāfi (i.e. the nominal subject) which is supposed to be mā umirtum bihi or simply hiya. Though not called idmār in this place, it is analyzed as such, as will be seen below. It is a quotation (h ikāya) which presupposes that it should be correct (saluh a) if the rāfi/nāsib or xāfid is suppressed. Al-Farrā exemplifies this rule as follows: qultu lā ilāha illā llāh fa-yaqūl al-qāil: qultu kalimatan sālih atan. But kalimatan sālih atan cannot stand alone without a nāsib. To decide whether an utterance is quotation or indirect speech is very important for the Qurānic text where change(s) of speaker and hearer can occur even within the same āya. Maānī I:40 regarding Q 11:69: qālū salāman qāla salāmun The word salām is used here in two meanings (alā manāyayni). In the first step the two endings (nasb and raf ) are explained. The explanation of the endings relies on the grammatical rule differentiating indirect speech from quotation (h ikāya). Having fixed this al-Farrā adds that in the case of h ikāya the idmār of alaykum should be supposed. In a third step he demonstrates that this idmār is possible because salām frequently occurs in kalām alone, i.e. without alaykum. Maānī I:93 regarding Q 2:154: wa-lā taqūlū li-man yuqtalu fī sabīli llāh amwātun It is another instance of h ikāya where the raf  ending is explained by the idmār of hum.14 The irāb ending and the difference between quotation and indirect speech is demonstrated by substitutional analysis (huwa bimanzilat . . .). He says that the nasb ending is not permitted here because amwāt is an ism and not a qawl, i.e. it is not a nominalized phrase. He illustrates the possible and impossible usages as follows: qultu laka xayran, i.e. kalāman h asanan. qultu laka xayrun, is similar to: qultu laka mālun.


Az-Zajjāj (Maānī I:229) also accepts this analysis.

 ĀR in the MAĀNī of al-farrĀ IDM


In the case of xayr both structures are correct whereas their substitutes cannot be used in the other structure. Since there are no different readings here it might be supposed that by illustrating these structures in detail al-Farrā may possibly refute analyzing techniques that take into account only the surface structure. Maānī I:296, 7 regarding Q 4:171: wa-lā taqūlū talātatun This and similar examples show that it is a general grammatical rule that necessitates the use of idmār, scil. if there is a marfū after the qawl (as a quotation after the verb qāla), there must be a rāfi (subject) as well, in the above example: wa-lā taqūlū hum talātatun General rule: if something is in raf  after qāla (yaqūlu, qawl), there should be a rāfi either overtly or in a suppressed way (idmār). Here alFarrā gives a formal explanation of an irāb ending. 3.5.3 The āid in the relative clause: Maānī I:157 regarding Q 2:246 ibat lanā malikan nuqātil fī sabīli llāhi The verb may not be nuqātilu (in raf ) referring to the preceding noun as a sila, because there is no reference back to malikan. However, if one accepts the yuqātil reading, it can be explained as either sila ( yuqātilu) or jazā ( yuqātil ) after an imperative (amr), as in the case of nuqātil. Al-Farrā shows that the form yuqātilu can also be used when there is no antecedent, with idmār: ibat lanā lladī yuqātilu. Then he presents a kalām example: allimnī ilman antafiu bihi and allimnī lladī antafiu bihi. But if bihi is dropped, then only the jazm is correct, since there is no reference back (āid). 3.5.4 jazā: Maānī I:178, 4, Q 2:265: fa-in lam yusibhā wābilun fa-tallun Al-Farrā here refers to the grammatical rule of jazā that requires mādī forms, thus he supposes the suppression of kāna before tallun. The interesting point of his analysis is, however, when he says: udmirat kāna fa-saluh a l-kalām, because it sheds light on the meaning of idmār: It is not simply suppression or deletion but the supposition of a suppressed element which corrects the utterance and makes it fit the grammatical rules.


kinga dévényi

3.5.5 ammā . . . fa-: Maānī I:228 regarding Q 3:106: fa-ammā lladīna swaddat wujūhuhum a-kafartum This is a good example of a well-known grammatical rule (the obligatory use of fa- after ammā) which makes the grammarian suppose the idmār of fa-. At the same time the meaning of the āya also requires the supposition of a suppressed phrase yuqāl. After dropping yuqāl, however, the fa- is also dropped. In reconstructing the meaning one needs to reconstruct the utterance as well: fa-ammā lladīna swaddat wujūhuhum fa-yuqāl a-kafartum. In the above five examples idmār is necessitated by some basic grammatical rules considered by al-Farrā to be generally accepted and evident to such an extent that he does not even try to explain them with the exception of the last example. 3.6 The impossibility of idmār: Maānī I:195–197 regarding Q 3:15: qul a-unabbiukum bi-xayrin min dālikum li-lladīna ttaqū inda rabbihim jannātun Jannāt is in raf  because of the influence of lām. This intervening lām prevents referring back to the beginning of the utterance (awwal al-kalām). He then draws the conclusion: the xāfid (here bi-) cannot be suppressed (lam yudmar), therefore one cannot suppose that the bi- (in bixayrin) is suppressed before jannāt causing it to be in xafd. This also means that the irāb-ending xafd cannot be used independently. The same principle is stated in connection with Q 6:96 (Maānī I:346) where a h āl stands in the way of the effect of a preceding verbal noun (jāil): wa-jāilu l-layli sakanan wa-š-šamsa wa-l-qamara h usbānan. Since aš-šamsa and al-qamara cannot be in xafd, therefore in al-Farrā’s view they take nasb ending according to their meaning.

4. Concluding remarks 4.1

A summary of idmār types

To sum up the method of idmār, it can be established that al-Farrā derives the explanations of specific Qur’ānic passages from the following main types of sources:

 ĀR in the MAĀNī of al-farrĀ IDM


(i) the end of an utterance is related to its beginning, the preceding context (sometimes on the basis of the different segmentations of the text), where—as al-Farrā (Maānī I:13, 14–15) puts it—wainnamā yah sunu l-idmāru fī l-kalāmi lladī yajtamiu wa-yadullu awwaluhu alā āxirihi, i.e. the context is coherent. It is contrasted with inqataa, where the context is incoherent. (ii) another place in the Qurān; (iii) an extra-textual linguistic source: kalām al-arab/badw or sometimes a poem, though it should be pointed out that poems do not have their central significance they have in the Kitāb; (iv) a grammatical rule often quoted without example or proof. Its main types are: (a) only the rule is mentioned; (b) a rule-like example is given (kamā taqūl . . .); (c) a grammatically incorrect example is used to support his reasoning (lā yajūz . . .); (v) a qirāa; (vi) the codex of Abdallāh b. Masūd or, less frequently, that of Ubayy b. Kab or the codices of their followers which sometimes contain non-canonical variations. 4.2

The meaning of idmār

The meaning of idmār is not simply h adf (deletion), but rather the obligatory supposition of an element. Thus when al-Farrā writes that lā budda min idmār kāna li-anna l-kalām jazā (Maānī I:178, 4) it means just the opposite of obligatory h adf (the obligatory deletion); it means that it is obligatory to suppose the existence of an element either present in the utterance or suppressed. The use of the terms uskitat and ulqiyat also sheds light on the main characteristic of idmār, i.e. its obligatory nature. For example at Maānī I:163 regarding Q 2:246 there are two options to express the same meaning: either with allā or only lā. The dropping of an, however, cannot be considered idmār, since both variants are equally correct Arabic (alā wajh al-arabiyya), the structure with only lā not being brought back to the one with an.

62 4.3

kinga dévényi Two kinds of traditions in the Maānī

It can be established that al-Farrā relied on two kinds of traditions in his Maānī. I

The grammatical tradition

A) al-Farrā and Sībawayhi Disregarding the terminological differences it can be stated that on the whole he followed the same grammatical tradition that became exemplified by Sībawayhi’s Kitāb. a) He takes sides with a Sībawayhi-like grammatical analysis on the basis of the āmil theory, without, however, using the same terminology.15 b) Frequent reference can be found to grammatical rules considered self-evident and thus in no need of further explanations. c) We can see the main difference between the Maānī of al-Farrā and Sībawayhi’s Kitāb not in the method of grammatical analysis but in their approach. Al-Farrā, starting from a complete text, always has in mind the text as a whole, as a series of utterances, and he proceeds accordingly. Whereas Sībawayhi cites only specific examples to illustrate the grammatical rules he wishes to present. B) al-Farrā and al-Axfaš The difference between al-Farrā and al-Axfaš lies in the fact that the latter basically writes about grammar, while the former deals essentially with Qurānic exegesis on a grammatical basis. This may account for al-Axfaš’s lack of interest in eliciting all the known readings at a given place whereas al-Farrā seems keen on mentioning whatever can be heard (see, e.g. al-h amdu/al-h amda/al-h amdi above), whereas does not seem to attach great importance to the explanation of structures or endings that he considers self-evident.

15 According to Talmon (2003, 309–312) Sībawayhi’s main concern in syntax is irāb carried out by amal effect. Talmon also postulated that in the Kitāb al-h udūd, al-Farrā seemed to focus in his syntactic description on sentence-types and the determination of syntactic relations. We can also experience in the Maānī that while al-Farrā dealt with irāb-endings in a somewhat flexible way, he did not make allowances in the case of syntactic structures.

 ĀR in the MAĀNī of al-farrĀ IDM


C) al-Farrā’s opinion on the linear interpretation of the text a) At least half of al-Farrā’s analysis of idmār deals with the interpretation of different irāb endings which he considers fundamental in clarifying the meaning of the Qurānic text. Therefore he is in constant struggle with those views according to which the short vowel endings are merely phonetic or morphological phenomena and not irāb endings. The most frequent of these solutions is the choice of the final vowel on the basis of proximity, itbā.16 b) He also introduced some formal descriptive interpretations in which he might have followed Kūfan exegetical methods. He accepts these explanations if they fit into his concept of the linear descriptive text analysis. The best examples to demonstrate this are those places where he insists that certain particles, such as wa-or the prepositions, may block the impact of the previous grammatical structure (e.g. Q 2:185 as analyzed in 3.2.2 above). c) al-Farrā, however, refuses to accept superficial formal analyses which run contrary to the meaningful interpretation (e.g. Q 2:48 when he refuses identification of fī in temporal adverbs with fī in prepositional verbs). In such cases he does not say whose opinion, whose analysis is rejected by him. Sometimes he uses the passive (yuqāl), some other times it can only be supposed that he is at variance with some people who only looked upon the text as a linear string of utterances and derived the explanations quasi mechanically from the preceding string. II

The exegetical tradition

Al-Farrā’s grammatical activities in the field of Qurānic exegesis cannot be overestimated. It becomes especially evident if we compare his explanations to the relevant passages in at- Tabarī’s Tafsīr. It goes without saying that the scope of at-Tabarī’s commentary is much wider than that of al-Farrā’s work. But if we limit our comparison to the linguistic exegesis, we find that al-Farrā’s explanations are most of the time taken over verbatim without his name being mentioned. One can also suppose that 16 Al-Farrā accepts these variant forms in case of different existing samā in the collected corpus, as e.g. in the case of al-h amda/i/u (Q 1:2). If, however, the difference appears not in the given vowel ending but in the explanation, then he always advocates the explanation based on imāl, as e.g. in Q 1:7 where he does not accept the explanation according to which the ending of ġayri would be determined by the preceding alayhim.


kinga dévényi

at-Tabarī had access to a fuller version of the Maānī, it is, however, also possible that al-Farrā was only one prominent exponent of a common thinking about grammatical issues in the Qurān and the analyses he dictated from his memory belonged to a common stock of knowledge, a long line of grammatical exegetical tradition, with which at -Tabarī was still familiar. It is interesting to note, however, that the two other Maānī works, Abū Ubayda’s Majāz and al-Axfaš’s Maānī do not seem to have been incorporated into what became the ‘definitive commentary’ of the Qurān for centuries. So it might be concluded that while the grammatical tradition as it was shaped in Basra outshone the Kūfan, its trace in the grammatical analysis of the Qurān is not significant.17

5. References 5.1 Primary sources Abū Ubayda, Majāz = Abū Ubayda Mamar b. al-Muta nnā at-Taymī, Majāz al-Qurān. Muhammad Fuād Sazgīn [Fuat Sezgin], ed. 2 vols. Cairo: al-Xānjī, n.d. al-Axfaš, Maānī = Abū l-H asan Saīd b. Masada al-Mujāšiī al-Axfaš al-Awsat , Maānī l-Qurān. Fāiz Fāris, ed. 2 vols. Amman: Dār al-Bašīr, 1981. Farrā, Maānī = Abū Zakariyyā Yahyā b. Ziyād al-Farrā, Maānī l-Qurān. Ahmad Yūsuf Najātī and Muhammad Alī an-Najjār, eds. 3 vols. Cairo: al-Haya al-Misriyya l-Āmma li-l-Kitāb, 19802. Sībawayhi, Kitāb = Abū Bišr Amr b. Utm  ān Sībawayhi, al-Kitāb. Abd as-Salām Muhammad Hārūn, ed. 5 vols. Beirut: Ālam al-Kutub, 1966–1977. n.d. at-Tabarī, Tafsīr = Abū Jafar Muhammad b. Jarīr at -Tabarī, Jāmi al-bayān fī tafsīr al-Qurān. 30 vols. Cairo: al-Mat baa al-Maymaniyya, n.d. az-Zajjāj, Maānī = Abū Ishāq Ibrāhīm b. as-Sarī az-Zajjāj, Maānī l-Qurān wa-irābuhu. Abdaljalīl Abduh Šalabī, ed. 5 vols. Beirut: Ālam al-Kutub, 1988.

17 An interesting feature of al-Farrā’s explanatory method is his usage—at certain places—of the 2nd person singular in the case of irāb endings and other grammatical interpretations. The use of 2nd person singular is on the one hand the usual practice in Sībawayhi’s Kitāb, but Sībawayhi presents his linguistic analysis from the point of view of the speaker, the producer of different utterances, whereas al-Farrā deals with a concrete text, and what is even more, the text of the revelation. The use of the 2nd person singular in this case might indicate that al-Farrā considered that the sacred text was the one without the short vowel endings or that it had been revealed according to what is termed as sabat ah ruf which leaves the reader of the text some freedom in the vocalic realization. It should, however, be noted that al-Farrā’—similarly to other commentators, or in fact Sībawayhi when he deals with the Qurān (e.g. Kitāb II:155, 10 ad Q 5:69)—does use the 3rd person singular or the passive when he deals with different irāb endings in the Qurān.

 ĀR in the MAĀNī of al-farrĀ IDM 5.2


Secondary sources

Burton, John. 1988. “The Qurān and the Islamic Practice of wudū.” BSOAS 51:1:21–58. Dévényi, Kinga. 1987–88. “Mujāwara: A Crack in the Building of irāb.” Quaderni di Studi Arabi 5–6, 196–207. ——. 1990. “On Farrā’s linguistic methods in his work Maānī l-Qurān.” Studies in the History of Arabic Grammar II. Proceedings of the 2nd Symposium on the History of Arabic Grammar, Nijmegen, 27 April–1 May 1987, Michael G. Carter and Kees Versteegh, eds. Amsterdam: J. Benjamins. ——. 1991. “Al-Farrā and al-Kisāī: References to Grammarians and Qurān Readers in the Maānī l-Qurān of al-Farrā.” The Arabist. Budapest Studies in Arabic 3–4, 159– 176. Kinberg, Naphtali. 1996. A Lexicon of al-Farrā’s Terminology in his Qurān Commentary. Leiden: E.J. Brill. Makram, Abd al-Āl Sālim and Ahmad Muxtār Umar. 1985. Mujam al-qirāāt alQurāniyya. 8 vols. Kuwait: Jāmiat al-Kuwayt. Talmon, Rafael. 2003. Eighth-century Iraqi Grammar. A Critical Exploration of PreXalīlian Arabic Linguistics. Winona Lake, Indiana: Eisenbrauns. Versteegh, C.H.M. 1993. Arabic Grammar and Qurānic Exegesis in Early Islam. Leiden: E.J. Brill.


Sunt aliquot quoque res, quarum unam dicere causam non satis est, verum pluris, unde una tamen sit. Lucretius: De rerum natura VI 703f.

1. Introduction This paper is devoted to the Arabic sub-standard phenomenon of the relative pronoun alladī in the function of a conjunction predominantly meaning “that; because”. The historical interpretation of this phenomenon, which had been noted by Arab purists for the formula al-h amdu li-llāhi lladī “Praise be to God that” as early as the 9th century C.E., has been a topic for Arabists during the last fifty years. Considering the intensity with which the historical dimension of alladī “that; because” has been discussed, it may seem superfluous to want to take it up yet again. This, however, would be a rash conclusion as no communis opinio concerning the origin of the conjunctional alladī has ever been agreed upon. The aim of this paper is threefold. First, I shall discuss the theories as to the diachronic aspects of alladī as a conjunction in chronological order. Then I shall present additional early evidence from documentary sources dating from the 11th–12th centuries C.E. and later. After that, I shall present my own theoretical approach combining important insights of my predecessors with hitherto neglected aspects. Finally, I shall deal with the origin of the formula al-h amdu li-llāhi lladī. A point that I shall not consider in this paper is the general function of the H amdalah in formulaic expressions used by today’s Arab Muslims. The important role of this formula, simple and expanded, in “everyday Arabic speech” is borne out by numerous examples in Piamenta (1979, 247f., Index) and (1983, 209, Index), among them examples of the


werner diem

al-h amdu li-llāhi lladī type with alladī as a conjunction meaning “that; because” (Piamenta 1979, 89, 172f., 176).

2. Theoretical approaches 2.1

Meir M. Bravman (1953)

In the chapter “The development of the ‘psychological’ (‘logical’) subjectpredicate relation” in his Studies in Arabic and General Syntax (1953), Meir M. Bravmann, after stating that in constructions of the al-h amdu li-llāhi lladī type and similar expressions alladī can be interpreted as “that”, maintains that the construction is already found in Classical Arabic, for which he adduces a verse by Imra’ al-Qays. This verse need not be discussed here as it allows of other interpretations. More important is Bravmann’s general theory about the origin of alladī as a conjunction (1953, 41): [. . .] we have to assume [. . .] relative clauses which do not link up with the immediately preceding expression (as al-h amdu li-llāhi etc.), but are parts of an independent new sentence or, more exactly, predicates whose subject—known from the preceding sentence or from the situation—is mentally supplied but not linguistically expressed [. . .]. [. . .] the sentence al-h amdu li-llāhi lladī lam ’amut would mean: “Thank God. [I am one] who has not died” etc.

Spitaler (1963) rejected Bravmann’s theory (see below 2.3), and it was not discussed by other scholars. I tend toward Spitaler’s rejection of this theory as it does not explain satisfactorily the phenomenon in question, and therefore I shall disregard it. 2.2

Joshua Blau (1961)

In his grammar of Judaeo-Arabic, Joshua Blau treats the conjunctional alladī in § 346, which begins with the words “alladī introducing noun clauses”.1 The paragraph falls into five sections according to the syntactic status of the alladī clauses: (a) attributive clauses, (b) subject clauses, (c) predicate clauses, (d) object clauses, (e) clauses in which alladī has a causative function. Most of the examples refer to past events, but there


This and the following translations from the Hebrew of Blau are mine.

arabic allad ī as a conjunction


are also some referring to future events with the heads of the clauses expressing prevention and necessity. As to the origin of the conjunctional alladī, Blau begins section (a) with the following words (1961, 226): Relatively usual are attributive clauses (apposition), which represent the transition from relative clauses,

a statement which is continued in a note with the remark (words in brackets are my additions): However, this does not mean that alladī introducing noun clauses has originated from attributive clauses only. It is, for example, possible that from sentences such as al-h amdu li-llāhi lladī ’a‘ānanī “Praise be to God who helped me” alladī = kī, še- [“that; because”] originated, marking object clauses or mipne še [“because”].

These are quite general statements, which hint to the direction where, according to Blau, a historical interpretation should be looked for. In his opinion, as can be grasped from both remarks cited above, the origin of alladī as a conjunction has to be seen (a) in attributive relative clauses and (b) in the al-h amdu li-llāhi lladī type. 2.3

Anton Spitaler (1963)

Spitaler’s approach differs fundamentally from Bravmann’s and Blau’s as it is monocausal. He starts from the al-h amdu li-llāhi lladī type and assumes that this construction was first re-interpreted (Spitaler does not use this word) as meaning “gottlob, daß”, after which it was generalized. The following quotations will illustrate Spitaler’s approach: al-h amdu li-llāhi lladī ist der Anfang eines normalen, von dem Wort Allāh abhängigen Relativsatzes, aber aus seinem eigentlichen Zusammenhang gelöst und sekundär in eine neue Konstruktion übertragen. (1963, 101) Es ist nun ganz eindeutig, dass das stereotyp wiederkehrende al-h amdu li-llāhi im Lauf der Zeit einer Funktionsschwächung unterlegen ist [. . .]. Dadurch verlor das nachfolgende alladī zwangsläufig den lebendigen Zusammenhang mit seinem Beziehungsnomen und der ganze Ausdruck wurde zu einem starren, wenn auch mit einem ganz bestimmten Affektgehalt geladenen Syntagma, bei dem alladī nur mehr als überleitendes Element, als verbindende Partikel, letztenendes eben als Konjunktion empfunden wurde. Und nunmehr war es natürlich gleichgültig, welche syntaktische Form der anschließende Satz hatte. (1963, 235)


werner diem

After presenting some examples semantically corresponding to al-h amdu li-llāhi lladī “praise be to God that” which, however, contain finite forms of h md I and škr I instead of the nominal al-h amdu li- (1963, 105), and after rejecting Bravmann’s above-mentioned theory (2.2), Spitaler continues with the words: Mit der Herauslösung aus seinem ursprünglichen legitimen Relativsatzgefüge und der Übertragung in eine ganz neue Konstruktion hat nun aber unser Typus seine Entwicklung nicht abgeschlossen; es zeigt sich nämlich, dass die Formel al-h amdu li-llāhi als solche zurücktritt und durch andere Ausdrücke ersetzt wird, die ebenfalls der Äusserung des Dankes, der Freude, frohen Überraschung, Befriedigung usw. dienen. [. . .] Die dabei gegebenen phraseologischen Möglichkeiten sind sehr mannigfaltig [. . .]. Sie werden aber noch dadurch bereichert, dass die Konstruktion nunmehr [. . .] auch für die damit kontrastierenden Gefühle des Bedauerns, der Reue usw. verwendet wird. (1963, 109)

There follow in Spitaler’s article numerous examples of these latter constructions both from post-classical writings and modern dialects, all of them referring to events in the past. Additionally, Spitaler presents some examples of alladī “that” after verbs from other semantic fields, among them some where the alladī-clause, as in some of Blau’s examples, refers to future events expressed by the Arabic imperfect. While Spitaler’s approach to the problem is definitely a monocausal one, he was well aware of the fact that a multicausal approach would also in principle have been possible, all the more in view of Hebr. ’ašer (on its origin see now Rubin 2005, 49f.) and Aram. dī-, which, besides their original function as relative particles, developed the meaning “that; because” (1963, 106f.). In his final remarks, Spitaler emphatically defended his own theory on the grounds: (a) that the Arab grammarians noted the conjunctional alladī only for the H amdalah, which therefore seems to be central, (b) that the al-h amdu li-llāhi lladī type possesses the chronological priority within the evidence of the conjunctional alladī, and (c) that it would be implausible to assume two different starting points of the conjunctional alladī converging into one phenomenon (1963, 111f.). There is no reference to Blau’s Diqduq (1961) in Spitaler’s article, probably due to the fact that he was unaware of it when he wrote the article. In fact, Blau in some respect anticipated Spitaler’s theory about the central role of al-h amdu li-llāhi lladī.

arabic allad ī as a conjunction 2.4


Joshua Blau (1965)

In his Emergence and Linguistic Background of Judaeo-Arabic (1965), which appeared four years after Diqduq, Blau is essentially of the same opinion as in Diqduq (1961) about the origin of alladī “that”, and even the wording reminds one of that used in Diqduq although it is more explicit: Since a relative clause may often be mistaken for a substantive clause, the transition of ’alladī into a conjunction introducing substantive clauses was easily accomplished. (1965, 109)

After this general statement, Blau gives two examples: wa-kamā ‘alimta min h adīti bni Dwrdy’ jamī‘a mā fa‘alahu lladī ‘abida s-sanama and lillāhi l-h amdu lladī ’aslamahu fī ’aydīkum, in Blau’s translation “and likewise, you know the story of B. D., everything that he had done, that he had served the idols” and “praise be to God that he has delivered him into your hands”. According to Blau, it is in both cases only the word order that indicates that alladī introduces a substantive clause and not a relative clause, i.e. that alladī is a conjunction and not a relative pronoun. While the second example indeed illustrates this transition, as alladī can be interpreted both as a relative particle and a conjunction (“praise be to God, who delivered him into your hands” and “praise be to God that He delivered him into your hands”), this, in my opinion, is not possible with the first example as it admits of the interpretation of alladī as a conjunction only. From a semantic point of view, the alladī clause in this sentence is an apposition to mā fa‘alahu, which means that it cannot be interpreted at the same time as an attribute to Ibn Dwrdy’. Thus, this example does not represent the starting point of the re-interpretation of alladī but alladī in its new conjunctional function. Spitaler’s theory is not taken into consideration by Blau, but his article is mentioned in a footnote (1965, 109). 2.5

Joshua Blau (1967)

In contrast with Emergence (1965), where Spitaler’s article (1963) is mentioned in a footnote only, having no impact on Blau’s theoretical approach, it is fully considered in the third volume of Blau’s monumental Grammar of Christian Arabic, which appeared in 1967. Blau writes there:


werner diem ’alladhî opens substantive clauses. Beginnings of this construction already appear in CA2 and it becomes rather frequent in MA. A. Spitaler [. . .] has brought out in full relief the early and very interesting history of the type ’al-h amdu li-llâhi-lladhî ‘thank God that’ and its development. The use of ’alladhî for introducing substantive clauses in ASP, insofar this is not due to literal translation of Aramaic d-, is mainly from this type and its developments. This is all the more remarkable since in JA other substantive clauses (especially apposition clauses) introduced by ’alladhî are quite common. It stands to reason that, in accord with Spitaler’s findings, this use of ’alladhî first developed in the type al-h amdu li-llâhi. Early ASP exhibits mainly this stage, whereas in later JA other kinds of ’alladhî introducing substantive clauses have arisen. (1967, 526f.)

Among Blau’s examples there is none of the nominal al-h amdu li-llāhi lladī type. As to his examples of the corresponding verbal type that contains a reference to God, such as h amida llāhi lladī or šakara llāhi lladī, it must be stated that none of them unequivocally represent the new type with alladī as a conjunction because each can also be interpreted as a relative clause referring to Allāh. Of Blau’s remaining examples, many are likewise syntactically ambiguous, while some do represent the new type. In several cases, Blau himself draws attention to the alternative interpretation as relative clauses. Apart from that, as Blau remarks himself, some examples might be calques on the Aramaic d-, which, as has been mentioned above, is both a relative particle and a conjunction meaning “that; because”. Thus unequivocal evidence of the new type is less for Ancient South Palestinian than might be inferred from the number of Blau’s examples at first glance. 2.6

Manfred Woidich (1980)

Manfred Woidich’s article “illi als Konjunktion im Kairenischen” (1980) is an investigation into illi, which goes back to alladī, as a conjunction in modern Cairene Arabic. In his masterly analysis, which shows an interest and a competence in syntax not common for Arabic dialectology, Woidich gives a thorough picture of illi as a conjunction from a synchronic perspective. Woidich’s examples are mostly drawn from written sources, but he also asked native speakers of what is called “Educated Cairene Arabic”. It would be beyond the scope of this paper to comment

2 CA = Classical Arabic, MA = Middle Arabic, ASP = Ancient South Palestinian, JA = Judaeo-Arabic.

arabic allad ī as a conjunction


extensively on what Woidich found out about illi as a conjunction in Cairene Arabic. Therefore I shall concentrate on some points that are also important from a historical perspective. Woidich begins with the statement that the “üA” (übergeordnete Ausdrücke), i.e. the heads on which the illi clauses as “uS” (untergeordnete Sätze) depend, are: Äußerungen des Dankes, der Freude, der Befriedigung, des Bedauerns usw. [. . .] Der Sprecher nimmt jeweils emotional Stellung zu einem im untergeordneten Satz (uS) geschilderten Sachverhalt und wertet ihn in diesem Sinne. [. . .] Wir teilen das gesammelte Material nach der Art der üAA ein, von denen sich zwei Typen unterscheiden lassen. Die einen (Gruppe A) werten den Sachverhalt des uS direkt, indem sie [. . .] emotional Stellung nehmen, die anderen (Gruppe B) werten auf eine indirekte Weise, indem sie angeben, wie eine Person sich oder andere wertet aufgrund des Sachverhalts des uS, oder zu welcher affektisch bestimmten Handlung eine Person durch diesen veranlaßt wird. (1980, 225f.)

Examples of type A are il-h amdu li-llāh illi ultaha b-nafsak “Gott sei Dank, daß du es selbst gesagt hast!” and kuwayyis illi wiit fi gēbak inta “Gut, daß es in deine Tasche gefallen ist!” Characteristic of this type is according to Woidich the absence of resumptive pronouns: Eine Verknüpfung des üA mit dem uS durch pronominale Rück- und Verweise kann hier nicht stattfinden, da im üA keine Pronomen auftreten. (1980, 226).

Though this statement is corroborated by the above-cited examples, the last two examples of Woidich’s type A do have a pronominal concatenation, among them ġayizni lli dayya‘t il-bazburt-i btāi “Es ärgert mich, daß ich meinen Paß verloren habe”. Woidich considers examples of this type “einen Übergang zur Gruppe B” (1980, 227). In my opinion, such examples do not belong to group A but rather to group B, as expressions like ġayiz + object pronoun are a kind of pseudo-verbs, the pronoun being the logical subject, and there are similar examples in group B. As for type B, examples are d-ana batnaddim illi gēt “Ich bereue, daß ich gekommen bin” and inta karihni lli baūl il-h a “Du kannst mich nicht leiden, weil ich die Wahrheit sage”. Concerning the syntactic characteristics of this type B, Woidich says: Logisches Subjekt des üA ist [. . .] eine Person, die im uS als Subjekt auftritt. Dadurch kommt eine Verknüpfung der beiden Teilsätze durch pronominalen Verweis zustande, der ein Charakteristikum für diese Gruppe darstellt. (1980, 227)


werner diem

Contrary to this statement, the person in the üA (i.e., the head) to whom the illi clause refers is not always the logical subject, as is, e.g., borne out by the above-mentioned example inta karihni lli baūl il-h a, where the “Person, die im uS als Subjekt auftritt” is the (logical) object, not the (logical) subject of the sentence. Furthermore, there are examples of type B declared by Woidich to be marginal where there is no pronominal concatenation between the illi clause and an element of the head at all, e.g. ana mabsūt illi ma-gāš “I am glad he did not come” (1980, 229). Considering this evidence, it seems to me that the feature of coreferentiality is questionable for Woidich’s type B in Cairene Arabic, and it is evident from Blau’s, Spitaler’s and my evidence that it cannot be sustained in a general diachronic and diatopic perspective. A certain degree of coreferentiality is per se bound to exist for the simple reason that when in the heads of sentences emotions are spoken of the causes of these emotions as expressed in substantival clauses depending on those heads are in most cases connected with the persons mentioned in the heads of the sentences as having these emotions. Important insights in Woidich’s article are that in type B the element of the head to which the illi clause refers must be human (1980, 230), that the predicates are “faktiv”, that is, refer to real facts, mostly past events (1980, 231), and that this kind of illi can always be replaced by inn- “that”, with the difference that illi is considered by the informants as being “ stärker” than inn-, that is, more affective (1980, 234). 2.7

Manfred Woidich (1989)

While Woidich’s article of 1980 is essentially descriptive, his contribution “illi ‘dass’, illi ‘weil’ und zayy illi ‘als ob’: zur Reinterpretation von Relativsatzgefügen im Kairenischen,” which was published in 1989, is diachronically oriented. As announced in the title, Woidich distinguishes between three different kinds of illi: illi “that”, illi “because” and zayy illi “as if ”. (a) illi “that” The type illi “that” corresponds to type A of Woidich (1980), e.g., il-h amdu li-llāh illi sabitak “gottlob, daß sie dich verlassen hat” and kuwayyis illi gēt “gut, daß du gekommen bist” (1989, 110f.). As for the origin of this type, Woidich follows Spitaler (1963) in saying that il-h amdu li-llāh in

arabic allad ī as a conjunction


sentences like il-h amdu li-llāh illi waaik fiyya “Lob sei Gott, der dich mit mir zusammengebracht hat”: unterlag einer Funktionsschwäche [. . .]. Als “starres Syntagma” wurde diese Formel nicht mehr als analysierbar aufgefaßt, was zur Folge hatte, daß insbesondere N = allāh nicht mehr als Nomen gesehen wurde, dem ein syndetischer Relativsatz angeschlossen werden konnte. (1989, 111)

While in the assumption of “Funktionsschwäche” Woidich explicitly follows Spitaler, he differs from him in assuming that the al-h amdu li-llāhi lladī formula was re-interpreted because there are similar sentences in Cairene Arabic having the same marked structure of “Rhema-Thema”, that is, comment-topic, such as h ilwa di “prima ist die!” or ‘ēb ikkalām da “eine Schande sind solche Worte!” He sums this up by saying daß die Reinterpretation von illi als Relativpronomen zu illi “daß” ausgelöst wurde durch den Umstand, daß die ursprüngliche syntaktische Struktur mit dem eingebetteten Relativsatz nicht der thematisch-kommunikativen Funktion dieser Sätze entsprach. Sie war markiert und wurde durch Reinterpretation der bei diesen Sätzen üblichen funktionalen Satzstruktur Präd.—Subjekt angeglichen. Dadurch konnte dem illi die Funktion einer Konjunktion zugeordnet werden, die Subjektssätze einleitet. (1989, 115)

Here we have to ask what Woidich intends to demonstrate from a general Arabic perspective. Does he want to demonstrate a historical development valid for Cairene Arabic only, or a general development in Neo-Arabic the results of which are palpable in Cairene Arabic also? A reconstruction of the first kind would be flawed by the fact that the Cairene conjunctional alladī cannot be detached from the common history of Neo-Arabic, and a reconstruction of the second kind by the fact that it would be problematic to reconstruct a common Neo-Arabic development on the basis of the specific evidence of one modern dialect, while disregarding other evidence, older and newer. Since we are concerned here with the second perspective only, which is tantamount to a general reconstruction of the conjunction alladī in Neo-Arabic, we have to ask what Woidich’s theory means for the history of the conjunctional alladī in early Neo-Arabic. In Neo-Arabic the equivalents of Woidich’s above-mentioned two Cairene examples would probably be something like h ilwa hādī and ‘ayb hādā l-kalām. Comparing these sentences with a sentence like al-h amdu li-llāhi lladī jita, it seems very improbable that the two types of sentences should have been mentally connected by any speaker of Arabic,


werner diem

and thus Woidich’s theory seems to me to be farfetched. Nevertheless, his drawing attention to the Rhema-Thema (or comment-topic) structure, which syntactically corresponds to a predicate-subject structure, is an important insight, to which I shall come back below when discussing my own theory (4.3). Goldenberg comments on Woidich’s theory with the words “This form is structurally parallel to the classical Arabic constructions of cleft sentences with mā having the same order of constituents” (1994, 16/261), but I doubt the validity of this alleged parallel, as the cleft sentences mentioned by Goldenberg, apart from their Rhema-Thema structure, differ fundamentally from the sentences mentioned by Woidich. (b) illi “because” illi “because” corresponds to type B of Woidich (1980), and examples are ana farh ān illi šuftak “I am happy because (that) I saw you” (1989, 116) or itnaddimit illi gat maāya “sie bereute es, daß sie mit mir gekommen war” (1989, 117). As for the meaning of illi in the second example, it remains to be shown that in Cairene Arabic illi in such sentences can be replaced by the causal conjunctions ašān and li-ann-. Interestingly, no transformations of sentences containing expressions of this type are among those adduced by Woidich (1980, 235) for the replacement of illi by the causal conjunctions ašān and li-ann-. In view of the semantic difference between the two kinds of illi in Cairene Arabic as seen by him and also in view of the feature “Referenzidentität” (that is, “coreferentiality”) in the second type, which he thinks is essential, Woidich assumes for the second type (type B in Woidich 1980) an origin independent of the al-h amdu li-llāhi type (type A in Woidich 1980). This origin he finds in sentences such as ana h mār illi dafat il-h isāb “ich bin ein Esel, der ich die Rechnung bezahlt habe!”, which contains a direct relative clause depending on ana. Since in sentences of this kind there exists a causal connection between the relative clause and the head, Woidich assumes that illi could be re-interpreted as a causal conjunction, “ich bin ein Esel, daß ich die Rechnung bezahlt habe!”, adding that this re-interpretation is also true of “alle anderen Sätze dieser Struktur” (1989, 118). It goes without saying that if we are to assume this kind of re-interpretation for early Neo-Arabic in general (and not only for Cairene Arabic), we would have to assume that alladī had been generalized as a relative particle by then, an assumption which poses no problem.

arabic allad ī as a conjunction


The fact that Woidich declares his example (13) itnaddimit illi gat maāya “sie bereute es, daß sie mit mir gekommen war” not to belong to the original type, confirms that, in his opinion, it is in nominal sentences that this type originated (1989, 119). Against this it can be argued (a) that verbal sentences (such as itnaddamit illi) seem to prevail over nominal sentences (such as ana h mār illi), and (b) that it seems highly improbable to me that speakers should have used sentences like ana h imāruni lladī addaytu d-danānīra “I am an ass, (I) who have paid the dinars”, anā farh ānuni lladī najawtu “I am glad, (I) who have escaped” or anā mutaassifuni lladī taaxxartu “I am sorry, (I) who was late” at all. For this theory to be accepted, it would be prerequisite to find unambiguous relative clauses of this kind in Classical Arabic. I have checked more than a thousand items with alladī in Tradition, many of them in dialogues, without finding even one example of such constructions, and I dare say that it is very improbable that they occurred at all. To sum up, as long as the syntactic type ana h imār illi dafat il-h isāb, which Woidich presumes to be the starting point of this type, is not shown to have been a normal construction in pre-Neo-Arabic, I consider sentences of this kind in Neo-Arabic the result of a specific development, which should be explained otherwise, rather than the origin thereof. (c) zayy illi “as if ” In interpreting this third type, Woidich starts with cases such as ir-rāgil firih bīna zayy illi laa līya “der Mann freute sich über uns, wie einer, der einen Fund gemacht hat”, in which the syntagma zayy illi “wie jemand, der; like somebody who” could be interpreted as meaning “als ob; as if ”. In cases representing unequivocally the result of this re-interpretation, only the interpretation of zayy illi as “as if ” is possible, such as ma-kanš-i fī tagāwub, zayy illi kunt-i bakallim h agar “es gab keine Resonanz, es war, als ob ich mit einem Stein spräche”. As for earlier evidence of this type, the ka-lladī mentioned in grammars, which structurally corresponds to zayy illi, has a different function, namely that of a conjunction of comparison with real facts, such as xudtum ka-lladī xādū “You have plunged as they have plunged” Qurān 9:69, already adduced by Reckendorf (1921, § 192, 4). For more evidence of this ka-lladī, see Blau (1967, 527f.), Hopkins (1984, 238) and Goldenberg (1994, 27f./276).


werner diem

An example of ka-lladī that functionally corresponds to zayy illi as a particle of comparison with unreal or hypothetical facts, as described by Woidich, was adduced by Spitaler (1962, 108) in a footnote: fa-lā yazālu kadālika ka-lladī yuh ibbu an yasxara min sāh ibihī 3 “and he continues to behave in this manner like one (or as if) intending to mock his friend”, sc. the tomcat (as-sinnawr) who plays with the mouse (al-farah) after having caught her (al-Jāhiz, H ayawān V, 202). Spitaler’s second example concerns the well-known other ka-lladī type: fa-sallaw ka-lladī kānū yaf alūna “and they prayed as they always had”. Another example of ka-lladī as a particle of comparison with hypothetical facts is wa-man axadahu bi-išrāfi nafsin lam yubārak lahu fīhi ka-lladī yakulu wa-lā yašbau “and who takes it (sc. certain money) with haughtiness of mind will not be blessed regarding it, like one (or as if) eating without getting replete” (al-Buxārī, Sah īh , Kitāb az-zakāh, No 1379), and more evidence of this kind can be found in Tradition. Since in examples of this kind alladī can still be interpreted as a relative pronoun, they represent kalladī in the meaning “as if ” in statu nascendi. Obviously, Spitaler did not realize that his two examples belong to two fundamentally different types. The first ka-lladī is a conjunction corresponding to kamā (accordingly it should perhaps be transcribed as kalladī), while the second ka-lladī consists of the particle ka- and the relative pronoun of the 3rd masc. sing. In Cairene Arabic, this second type was grammaticalized as zayy illi, with zayy standing for ka-, which is unusual in Cairene Arabic. Future research should pay attention to the existence of two different types of ka-lladī. It is, then, evident that ka-lladī has to be separated from alladī “that”, which is the topic of this paper. Therefore, these remarks must suffice, and I shall not return to this special type. 2.8 Gideon Goldenberg (1994) In his article “alladī al-masdariyyah in Arab grammatical tradition” (1994), which was reprinted in 1998, Gideon Goldenberg begins with general observations, then mentions Spitaler’s article (1962) and Woidich’s second article (1989). Goldenberg describes the aim of his paper as follows: 3 There follow four further an-clauses dependent on ka-lladī yuh ibbu. Spitaler’s quotation has the variant bi-sāh ibihī.

arabic allad ī as a conjunction


About the distribution of “infinitival” alladī in Arabic writings in general I have nothing to add; in the present paper I just wish to adduce some further evidence for a fuller understanding of the relevant structures as described and treated in Arabic grammatical literature. Such examination is important, because syntactical constructions that are considered grammatical by the great masters of Arabic grammar cannot easily be discarded as inadmissible or non-Classical. (1994, 16f., 262)

Some of the examples that the Arab grammarians commented on are from the Qurān, while other examples were made up by the grammarians themselves. A survey of the examples in the “Concluding remarks” shows that the examples belong to disparate types. In some of them, alladī is combined with a preposition, such as ka-lladī “as”, alā lladī “for that”, bada lladī “after”, in which alladī stands for the more usual mā. In other cases, it is again either mā, or an, which alladī stands for. It might even be doubted that some of the sentences on which the grammarians dwelt in length ever occurred in normal speech, such as alladī mararta mamarrun h asanun “It is a good passing that you passed”, apart from the fact that the syntactic interpretation of alladī in this sentence is far from being unequivocal. As far as I can see, Goldenberg makes no attempt to draw historical conclusions from his examples within the general discourse of conjunctional alladī, nor can I do this myself. So, in spite of Goldenberg’s thorough and learned approach, the historical dimension of his examples has yet to be demonstrated.

3. Additional early documentary evidence In the following, I shall give documentary evidence of alladī as a conjunction, which is the result of a wide and systematic reading of premodern Arabic documents. The major part of the evidence comes from Judaeo-Arabic documents found in the Genizah in al-Fustāt  (Old Cairo). Considering that I likewise have read Arabic documents written by Muslims as well as Christians, the preponderance of the Judaeo-Arabic evidence is noteworthy, being due to the more sub-standard character of Judaeo-Arabic writings as compared with Muslim and Christian ones. The value of documentary examples lies in that they are chronologically well-defined, which is not necessarily the case with literary examples, especially when these occur in sub-standard texts which have come down to us through late manuscripts.


werner diem

Meanwhile, many of the Judaeo-Arabic documents which I had read in their editiones principes have been re-edited by Moshe Gil in his two comprehensive collections of Genizah documents. Where this is the case I shall cite both editions, partially in order to give credit to the original editors and partially because there are occasionally differences between Gil’s re-editions and the original versions. While some interesting or difficult cases of dialectal divergences from the literary language will be commented upon in footnotes, substandard elements as such are not systematically marked. Some of my examples have already been cited by Blau (1961), as became clear to me while preparing this article. I have retained these examples for the reason that they belong to an exclusively documentary corpus, whereas Blau draws his examples from both literary and documentary sources, with the majority being literary. My examples will be presented according to a linguistic typology which mirrors the historical development of alladī as I see it. My division owes much to Woidich’s division of his Cairene material into his types A and B (see above 2.6 and 2.7), even if his division and mine do not fully overlap from a historical perspective as I assume more subtypes to have existed than he does. Many examples contain vulgarisms, which I shall not mark systematically. A. The head of the sentence expresses praise of or gratitude to God (a) Nominal type (1) li-llāhi l-h amdu lladī kānati l-āqibatu li-xayrin4 “Praise be to God that the result (of the affair) was good” Goitein, “Arkiyon,” No 39r, 21 = Gil, Texts, No 215r, 20 (letter to al-Fustā t, early 11th c. C.E.) (2) fa-li-llāhi l-h amdu wa-l-minnatu llatī5 kānati l-aqībatu h amīdatan “So to God be praise and gratitude that the result (of the 6 affair) was praiseworthy”


Goitein reads [ilā l-]xayri, Gil li-xayrin. Gil does not mention Goitein’s reading. allatī, which seemingly refers to al-minnah, is due to a kind of hypercorrection. 6 Another example with allatī “that,” or, more exactly, li-llatī “in order that,” is lianna qawm mina l-maġāribati qad sārū yusallū indanā li-llatī (!7%%) yutībahum in Goitein, “Kneset,” No 4v, 4f. Indeed, Goitein rendered !7%% as the Hebrew final conjunction *.)%. However, Gil’s re-edition in Texts, No 328, has %% instead of !7%%. So what we have here is the simple prayer Allāhu yutībhum “May God reward them!”—sc. the Maghrebis for praying in the synagogue of the Jerusalemites in Old Cairo. 5

arabic allad ī as a conjunction


Diem, Geschäftsbriefe Wien, No 45r, 10 (letter to a merchant in Egypt, 12th–13th c.s C.E.) (b) Verbal type (3) fa-h amidnā llāha alā dālika lladī nazara ilaynā wa-lam yušmit binā “and we praised God for this (happy outcome),7 namely that He cared for us and did not have us mocked at”8 Goitein, “Iggeret” r, 16 = Gil, Texts, No 616r, 17 (letter of al-Mahdīyah to al-Fustā t, 11th c. C.E.) (4) wa-h amidtu llāha]9 anā wa-axī lladī masaynā10 lah iqnā alā xayr wa-āfiyah li-Māsār11 “and we praised God], I and my brother, that we (finally) reached Egypt on foot well and in good health” Toledano, “Teudot”, No 2, 1 (letter from Egypt, 1540 C.E.) B. The head of the sentence does not express praise of or gratitude to God (a) The head of the sentence contains an abstract noun expressing a non-personal emotional evaluation of the contents of the alladī clause (5) wa-mā wajadnā azāan li-qulūbinā ġayra annahu l-waylu lanā nah nu lladī išnā li-hādihi l-masāibi wa-našrabū hādihi l-akwāsa l-murrata “We did not find consolation for our hearts but (all we can say is) that woe is us that we have lived (long enough to go through) these disasters und (that we) have to drink these bitter cups!” Gottheil and Worrell, Fragments, No 27r, 7–9 = Gil, Texts, No 501r, 7–9 (letter from Jerusalem to al-Fustā t, 1065 C.E.)

7 The details of the happy outcome of the affair in question are imparted immediately before the passage cited. 8 Gil in his translation considers dālika an antecedens of alladī in the sense of the German “darüber, daß.” However, the alladī—clause is an apposition to dālika, which in its turn refers to details mentioned before. In other words, the sentence is an expansion of the usual fa-h amidnā llāha alā dālika. 9 The addition is mine. Other additions are possible, but praise of God is the most probable one. 10 Dialectal form for mašaynā.—The language of the letter is substandard and exhibits features of Moroccan Jewish Arabic. 11 The grapheme renders Masar < Masr < Misr.


werner diem (b) The head of the sentence contains a verb or a participle expressing the emotion of a person caused by the contents of the alladī clause Joy (6) wa-anā bi-h amdi llāhi muġtabitun bi-lladī ttasaltu ilayhim “I am, thank God, rejoicing that I joined them12 (in marriage)”13 Ashtor, “Documentos,” No 2r, 11 = Gil, Documents, No 457r, 12 (letter from Jerusalem to Toledo, 1057 C.E.)14 (7) wa-qad sarranī lladī anfadta lahu rah lahu “It pleased me that you sent him his merchandise” Gottheil and Worrell, Fragments, No 9v, margin, 11–13 (letter from al-Fustā t  to Aden, probably 12th c. C.E.) (8) fa-qad radītu laka bi-lladī uxidtum fa-hāulāi nāsun muhtašimīn wa-llāhi z-zabbālīn mā radū bi-lladī uxidtum “I was content concerning you that you (both)15 were punished for these are decent people. Only the street-sweepers16 were, by God, not content that you (two) were punished” Diem, Geschäftsbriefe Wien, No 10, 8 (letter of a jealous wife to her husband, 12th c. C.E.) Wonder (9) wa-tumma innī ajabu minka lladī lam tusīb man yaktubu laka kitāb illā daf atan “Furthermore, I am astonished at you that you (allegedly) found only once somebody writing a letter down for you”17 Goitein, “Saloniqi,” No 1r, 37f. (letter from Saloniki to alFustāt, 11th c. C.E.)


Sc. the male relatives of the bride. Ashtor conceives of alladī as standing for alladīna, translating “Estoy contento de aquellos con que [he emparentado por matrimonio].” 14 According to Ashtor, the date is 1053. The difference consists in the reading of the last letter in the date according to the Jewish era on the margin (ttyd vs. ttyh ) 15 Sc. the addressee and his mistress, the plural standing for the dual in vulgar language. 16 That is, the mob. 17 Sc. a letter to be sent to the writer of the letter, who is the addressee’s father. 13

arabic allad ī as a conjunction


Anger (10) wa-anā ġadbānu alayka yā-axī katīr alladī wasal[ta il]ā Misra wa-lam tasil ilā Adana “I am very angry about you, O brother, that you came to Egypt (or Cairo), whereas you did not come to Aden”18 Braslavski, “Mishar,” r 12 (Letter from Aden to al-Mahdīyah, c. 1149 C.E.) Reproach19 (11) mā baqiya alaynā šayun illā lladī lam tuarrifnā kayfa kānat wasīyatu xālika naxuduhu (!) lladī lam tuarrifnī in kāna wasala laka šayun mina l-kutubi “Nothing remains for us (to say) except that (until now) you have not informed us as to how your uncle’s last will was. What we reproach him (!)20 with is that you did not inform me as to whether you received any of our letters” Gottheil and Worrell, Fragments, No 9v, 40f. (letter from al-Fustā t to Aden, probably 12th c. C.E.) Grief (12) wa-azza alaynā dālika katīr alladī lam yakūn ah adun minnā indaka yuāwinuka fī-mā jarā alayka fī taabika fī mā yaxussu amra xālika “and we were very much grieved by that,21 (namely) that none of us was with you to assist you in that which you had to endure in your concerns regarding the affair of (the illness and death of) your uncle” Gottheil and Worrell, Fragments, No 9r, 12f. (letter from al-Fustā t to Aden, probably 12th c. C.E.) (13) wa-dāqa sadrunā katīr alladī lam yakūn laka maahu kitābun yutamminunā “and we were very much distressed that he had no letter of yours with him setting our minds at rest (with regard to you)” 18 The writer intends to say that continuing the travel from Egypt to Aden would have been easy for the addressee, his brother. 19 Only the second alladī in (11) is an example of alladī following a verb expressing reproach. For the first alladī, see (17). 20 Scribal error for “you.” 21 Reference to the illness and death of the addressee’s uncle, the details of which are recapitulated by the writer before.


werner diem Gottheil and Worrell, Fragments, No 9r, 28f. (letter from al-Fustāt to Aden, probably 12th c. C.E.) (14) nulimukum annī bi-xayrin fī āfiyatin wa-qad azza alayya katīr alladī lam takūn h ādir h attā tuxallisa laka rah laka “I inform you (herewith) that I am well (and) in good health, but (that) I am much grieved that you are not present so that you might free your luggage (from the authorities)” Amari, Diplomi, 53, 2f. (letter from Tunis to Pisa, 12th– 13th c.s. C.E.) (c) The head of the sentence contains a verb or noun not expressing an emotion (15) wa-qad šakartu tafaddulahumā—h arasahumā llāhu—lladī qad dakarūnī fī kitābihimā bi-s-salāmi wa-bi-fili l-jamīli fī bābī “I am also grateful for their22 kindness—may God protect them—(consisting in) that in their letter they gave greetings to me and (also consisting) in performing good deeds to me” Assaf, Meqorot, 51, line 20f. = Gil, Documents, No 298r, 20f. (letter from Jerusalem to al-Qayrawān, 1039 C.E.) (16) wa-jāb lanā kitābaka wa-nah nu laysa indanā xabarun h attā jābahu r-rajulu l-warrāqu—jazāhu llāhu annā xayr alladī nalamu “and he brought us your letter while we (still) had no news of you until that man, the book-seller, brought it—may God requite him in our stead with good that we (now) know (how you are)” Gottheil and Worrell, Fragments, No 9v, 36–8 (letter from al-Fustāt to Aden, probably 12th c. C.E.) (17) See (11), first alladī.

I am the first to admit that 17 examples of alladī “that” are a somewhat meagre result of a decade-long reading of texts, but this result simply shows how rare this phenomenon is in texts whose writers intended literary Arabic. Surprisingly scarce is especially evidence of the type A(a) al-h amdu li-llāhi lladī. 22

Reference to two addressees, whose letter the writer answers.

arabic allad ī as a conjunction


Asking whether we can conclude from the scarcity of this type in documents, especially letters, that it was likewise marginal in the spoken language, we must allow for some reservations. The H amdalah occurs frequently in letters, mostly at the beginning but also in other parts. Usually the H amdalah is mentioned in the context of news which are deemed praiseworthy, as, for example, the writer’s or other persons’ good health or a good outcome of a difficult situation. However, this kind of the H amdalah is usually preceded by the report of the fact to which it refers, as, e.g., katabtu ilayka . . . wa-anā wa-man qibalī bixayrin wa-āfiyatin wa-l-h amdu li-llāhi (alā dālika) “I am writing you . . ., while I and my family are in good health—Praise be to God (for this)”. This conventional structure prevents the al-h amdu li-llāhi lladī type from frequently occurring in letters, which in its turn must also lead to lower frequency of the re-interpreted al-h amdu li-llāhi lladī in letters than may have been the case in the spoken language. Documentary texts where the al-h amdu li-llāhi lladī type usually does occur are, e.g., waqf documents, appointments of high officials and marriage contracts between persons belonging to the upper class. In these kinds of texts, an introductory al-h amdu li-llāhi lladī is expanded into long complicated passages praising God for bounties related to the topic of the text, but they are of so elaborated a style that they lack any deviation from the literary language. To sum up, the rare occurrence of the re-interpreted al-h amdu li-llāhi lladī in documents is of no relevance for the question of its potential frequency in the spoken substandard language. Scarce is also my evidence of the type B(a), where an abstract noun expresses a non-personal evaluation of the contents of the alladī clause, especially when compared with the abundant evidence of this type in Cairene Arabic as collected by Woidich (1980 and 1989). Again it is possible that the scarcity of early evidence of this type in documents is due to specific circumstances. Expressions of this type are mostly exclamatory and thus have a clear “Kundgabefunktion,” which makes it more likely for them to be used in the spoken language than in writing.

4. Historical typology 4.1

Introductory remark

Since the question of whether, and if so, how the different subtypes of alladī “that” are connected with each other is still open at this stage


werner diem

of this investigation (even if the answer to this question is admittedly anticipated by the arrangement of my own material in the preceding paragraph), I shall deal with them one by one. In doing so I shall consider whether there is reason to assume that the subtype in question had an origin of its own independent of the other subtypes or whatever other origin there might be. The sequence of the subtypes corresponds to decreasing syntactic and semantic markedness and specificity. 4.2 Types A(a) al-h amdu li-llāhi lladī “Praise be to God that” and type A(b) h amidtu llāha lladī “I praised God that” There seems to be nobody so far denying the validity of Spitaler’s theory that in al-h amdu li-llāhi lladī, to which my type A(a) corresponds, the relative particle came to be re-interpreted as a conjunction meaning “daß; that”. In Spitaler’s view, this development is due to a “Funktionsschwächung” of alladī, whereby “das nachfolgende alladī zwangsläufig den lebendigen Zusammenhang mit seinem Beziehungsnomen [verlor]” (for details see above 2.3), and this opinion is whole-heartedly shared by Woidich (see above 2.7). In my opinion, this approach is flawed by the fact that, besides the nominal type al-h amdu li-llāhi lladī, attested in the Qurān and elsewhere, there existed a corresponding verbal type, as e.g., in the saying ascribed to Muhammad, yā-mašara l-arabi h madū llāha lladī rafaa ankumu l-ušūra “O company of the Arabs, praise God, who took the tithes from you!” Ibn H anbal, Musnad, Musnadu l-ašarati l-mubašširīna bi-l-jannati, No 1566. This verbal type is likewise attested in the form of the re-interpreted verbal type A(b), along with the re-interpreted nominal type A(a). The existence of the re-interpreted verbal type A(b), whose verb varies according to the syntactic context, clearly proves that when the re-interpretation of al-h amdu li-llāhi took place its constituents, in spite of its formulaic character, must have been present in the minds of the speakers and cannot have been a de-etymologized complex as, e.g., addēš “how much?” (< *qadr ayyi šayin) in Syrian Arabic or izzayy “how?” (< *ēš zayy) in Cairene Arabic. Both the nominal al-h amdu li-llāhi and the verbal h amidtu llāha are equally attested as early as in pre-Islamic poetry,23 which proves that they had existed side by side from the first. 23 An example of al-h amdu li-llāhi occurs in the Muallaqah of Imra al-Qays, verse 124, in Ahlwardt’s edition, and for examples of the verbal type see Brockelmann (1922,

arabic allad ī as a conjunction


Rather, the re-interpretation took place because the implicit logical structure of many items of the Qurānic al-h amdu li-llāhi lladī is causal. A typical example is al-h amdu li-llāhi lladī najjānā mina l-qawmi z-zālimīna Q 23:28, in Arthur J. Arberry’s translation: “Praise belongs to God, who has delivered us from the people of the evildoers.” This literal translation of the relative clause is of course correct. However, when considering the propositional structure of the Arabic sentence we can easily discern that there is between the relative clause and the head of the sentence an intrinsic connection consisting in the implication that God is to be praised because He delivered the believers from the evildoers.24 It should be noted that this causal structure does not automatically exist in all al-h amdu li-llāhi lladī sentences but, it seems, only or mainly in those which refer to certain individual bounties of God. In this respect, it may, e.g., be doubted that the relation in al-h amdu li-llāhi lladī xalaqa s-samāwāti wa-l-arda “Praise be to God, who created the heavens and the earth” Qurān 6:1 or al-h amdu li-llāhi lladī wasia samuhu l-aswāta “Praise be to God, whose ear comprises all sounds” an-Nasāī, Sunan, Kitāb at-talāq, No 3406 and Ibn H anbal, Musnad, Bāqī musnad al-Ansār, No 23064, is as causal as the first example or causal at all. But the cases of al-h amdu li-llāhi lladī in the Qurān and still more those in Tradition refer mostly to individual deeds of God and thus express also an inherent causal relation. A combination of a non-causal relation and a causal relation is, e.g., found in the following passage in Ibn Abī d-Dam aš-Šāfiīs (d. 1244 C.E.) Kitāb adab al-qadā: al-h amdu li-llāhi lladī šahidati l-uqūlu bi-qidamihi wa-wah dānīyatih / wa-waqafat dūna idrāki jalālihi waazamatih || alladī btadaa l-maxlūqāti bi-badīi h ikmatih / wa-sawwara

116). Additional verbal examples occur in Labīd’s Dīwān No 6, 1 (h amidtu llāha wallāhu l-h amīdu etc.) and No 5, 1 (wa-llāhu rabbī mājidun mah mūdun). For examples of the related bi-h amdi llāhi, see Brockelmann (1922), and an additional example occurs in the Muallaqah of an-Nābiġah, verse 12, likewise in Ahlwardt’s edition. 24 Such causal function of relative clauses is, it seems, not dealt with by Lehmann 1989 in his otherwise comprehensive morphological and semantic analysis of the relative clause in a great number of languages, including Arabic. Nor is his chapter “Vom Relativpronomen zur Konjunktion” (1989, 389–393) of much help for the problem of alladī as a conjunction, as he considers any relative pronouns that have lost their inflection to be “Konjunktionen” if there is no resumptive pronoun in the subordinate clause. How Lehmann thinks the transition from uninflected relative pronouns to “true” conjunctions meaning “that” or “because” to have been remains unclear, in spite of some final remarks (1989, 391).


werner diem

l-kāināti bi-lutfi sanatih etc.25 “Praise be to God to whose pre-existence and uniqueness all minds testify and whose magnificence and power they are unable to grasp, who made creation with His wonderful and singular wisdom and designed the existing world with His kind work, etc.” The first alladī clause is non-causal as it describes attributes of God, and thus it is not possible to paraphrase it with a causal clause such as *“Praise be to God because all minds testify to His priority and uniqueness and are unable to grasp His magnificence and power.” In contrast to the first alladī clause, the second one is inherently causal, which is evident from the fact that it can easily be re-formulated as a causal clause: “Praise be to God because He made creation with His wonderful and singular wisdom and designed the existing world with His kind work, etc.” The causal character of this alladī clause is due to the fact that men own their existence to God having created them and thus have to thank Him for His benevolence. It is also interesting that the content of the first alladī clause is of a more inherent nature than the second in that it describes inherent, inalienable attributes of God, while the second refers to a deed that God performed of His own will. This hierarchy of the two alladī clauses conforms with a rule of Arabic syntax which says that attributes follow the noun they refer to in order of decreasing inherence. Considering that there is an inherent causal relation in many items of the Qurānic al-h amdu li-llāhi lladī between the head of the sentence and the relative clause, it comes as no surprise that the alladī that connects the two parts of the sentence and thus holds the position which a conjunction explicitly expressing the causal relation would have, should have been re-interpreted as a causal conjunction. This re-interpretation was in all probability triggered by parallel constructions in which the same causal relation is expressed explicitly. These constructions are as follows: (a) al-h amdu li-llāhi alā + noun/pronoun “praise be to God for . . .” and h amidtu llāha alā + noun/pronoun “I praised God for . . .” In both constructions, the alā phrase indicates the reason why God shall be praised. They are so frequent, particularly the former, that it is superfluous to give examples.


Adab al-qadā I, p. 247.

arabic allad ī as a conjunction


(b) al-h amdu li-llāhi (alā) an(na) “praise be to God (for) that . . .” and h amidtu llāha (alā) an(na) “I praised God (for) that . . .” The preposition alā can be elided as can any preposition preceding an(na). Sentences of this kind seem absolutely normal, but I can adduce only few examples of the verbal type, one from pre-Islamic times and the other examples later: h amidtu llāha an amsā Rubayun * bi-dāri l-hūni malh īyan muqāmā (Wāfir) “I praised God that Rubay got * in the house of disgrace, placed there in a shameful way” (Maqil b. Xuwaylid) Dīwān al-Hudalīyīn I, No 14, 3; istahalla yah madu rabbahu an lā yakūna asābahu dū h aqqin fī l-fayi “and he began to praise his Lord that none who had a right to the booty had injured him” ad-Dārimī, Sunan, Kitāb al-muqaddimah, No 91; allāha ah madu alā an jaalanī min ulamāi l-arabīyati “God I praise that he made me belong to the scholars of the Arabic language” az-Zamaxšarī, Mufassa l, 2, 2. Another example of the verbal type with God as the subject of an anna-clause is cited by Blau for Christian Arabic for škr I: šakara llāha annahu lam yuxallīhi yusīu ilā abdihi “He thanked God that He had not let (or made) him act unjustly to His servant” (1966–1967, 526). With subordinate clauses whose subject is coreferential with Allāh (second and third examples), this construction is optional instead of (a), whereas it is the only possible construction for subordinate clauses whose subject is not coreferential with Allāh (first example). It cannot be excluded that this construction is more usual in or even restricted to the verbal type, which is less formulaic than the nominal type. More examples would be desirable; their being so scarce so far is perhaps due to my (and my predecessors’) not noting them simply because they are so normal and therefore do not seem to deserve attention. (c) ah maduhu (alā) an(na) “I praise Him (for) that . . .” Whenever Allāh is replaced by the personal pronoun, only this construction is possible. This construction is especially frequent as a continuation of the relative clause type in the religious arengas (xutbahs) of Mamluk documents: al-h amdu li-llāhi lladī . . . nah maduhu alā annahu . . . This construction would also be the only one possible for the nominal head lahu l-h amdu but I have no examples. (d) h amidtu llāha id/fa- “I praised God because.” In this type, the subordinate clause is not dependent on the verb of the head as in the preceding types but forms a subordinate causal clause. Examples of each of the two conjunctions are the pre-Islamic verse h amidtu


werner diem ilāhī bada Urwata id najā * Xirāšun wa-badu š-šarri ahwanu min 26 badin (Tawīl) “I praised my God after (the death of my brother) Urwah (in battle) because (my son) Xirāš was (at least) saved, * and some evil is easier to bear than another” (Abū Xirāš) Dīwān al-Hudalīyīn III, No 14, 1, and, in Tradition, yā-Āišata h madī llāha fa-qad barraaki llāhu “O Ā’išah, praise God because God has exculpated you!” al-Buxārī, Sah īh , Kitāb aš-šahādāt, No 2467.

These functional and syntactic correspondences are set off in the following table, in which all sentences express the notion of praise being due to God for the salvation of the speakers, which is directly or indirectly ascribed to Him. Of the types marked with ?, no evidence has been adduced so far, but they would be normal Arabic from a syntactic point of view. (a) Nominal type Head




al-h amdu li-llāhi lladī najjānā al-h amdu li-llāhi alā najātinā ?al-h amdu li-llāhi (alā) an najjānā

Praise be to God who saved us Praise be to God for our salvation Praise be to God (for) that He saved us ?al-h amdu li-llāhi (alā) an najawnā Praise be to God (for) that we were saved (b) Verbal type Head




h amidtu llāha h amidtu llāha h amidtu llāha

lladī najjānā alā najātinā (alā) an najjānā

I praised God I praised God I praised God

h amidtu llāha

(alā) an najawnā I praised God

h amidtuhu

(alā) an najjānā

I Praised Him

h amidtu llāha h amidtu llāha

id/fa-najjānā id/fa-najawnā

I praised God I praised God

who saved us for our salvation (for) that He saved us (for) that we were saved (for) that He saved us because He saved us because we were saved


badī in the rhyme.

arabic allad ī as a conjunction


It is from this specific constellation that the re-interpretation of alladī in al-h amdu li-llāhi lladī and h amidtu llāha lladī as a causal conjunction meaning “(for) that” started. Since the relative clause type shared with the other types both the overall construction and the causal meaning of the subordinate clause, alladī could by analogy be interpreted as a particle expressing this causal relation. It is in sentences such as al-h amdu li-llāhi lladī najawnā “Praise be to God (for) that we were saved” and h amidtu llāha lladī najawnā “I praised God (for) that we were saved,” with the subject of the alladī no longer being coreferential with Allāh of the head, that the re-interpretation surfaced. It should be stressed that this re-interpretation of the relative clause as a causal subordinate clause was only possible because of the three following characteristics of al-h amdu li-llāhi lladī (and the corresponding h amidtu llāha lladī), of which the first is syntactic, the second semantic and the third morphological. (a) The relative clause of al-h amdu li-llāhi lladī is non-restrictive, which means that it does not serve the purpose of identifying Allāh, who is the object of praise. It is the loose connection typical of non-restrictive relative clauses that permitted the syntactic shift of the subordinate clause from its status as a relative clause depending on the antecedent Allāh to its new status as a causal subordinate clause which no longer depends on Allāh but on the head of the sentence as a whole. (b) Verbs such as “to praise” or “to thank” have, due to their specific semantics, an inherent causal complement indicating the grounds for praise or gratitude. Even if in the case of an isolated H amdalah in a religious context the reason is not explicitly expressed it is nevertheless there, consisting in God’s general bounties, the knowledge of which exists in the religious subtext. Thus whenever something positive is mentioned in the syntactic context of a H amdalah it is in principle subject to being interpreted as being the reason for it. Exactly this happened in the case of the relative clause of the H amdalah, but it happened only there because it is only the non-restrictive clause that, due to its structure, has the syntactic potential of such a subordinate clause, while this is not the case with, e.g., nominal attributes referring to Allāh. (c) The re-interpretation of alladī may have been helped by the fact that at the time when it happened the relative pronouns had probably already been reduced in the spoken language to alladī, which etymologically is the pronoun of the 3rd masc. sing. As a result of this development, the congruence between the antecedent and the relative pronoun in terms of gender, number and, in the case of the dual, case had


werner diem

disappeared, and thus the syntactic connection between the antecedent and the relative clause had been weakened. However, behind this development there is yet another factor, which is of an ontological nature. Of the constructions mentioned above sub (a) and (b), those where the subject of the subordinate clause is not coreferential with Allāh are as a rule not or less possible with other syntactic objects of “to praise”, unless the specific reason of praise is mentioned before or otherwise known to the person addressed. Thus one could say in English (and correspondingly in German and Arabic and many other languages), “I would like to thank you that you helped me” (coreferential subject of the subordinate clause), but usually not “I would like to thank you that we were helped” (non-coreferential subject of the subordinate clause), whereas the second, non-coreferential construction is a normal construction with God as the syntactic object of “to thank, to praise”: “(I) thank God that we were helped.” The difference lies here in that whenever something (positive) happens it can be attributed by the believer to God as the one who with His power has caused it. This is why the H amdalah (nominal or verbal) could be followed from the beginning by non-coreferential causal clauses (alā an, an, id) as mentioned above, and this is also why in al-h amdu li-llāhi lladī the coreferentiality of the subject of the relative clause with its antecedent Allāh was not an inherently indispensable syntactic feature, a fact which made it possible for coreferentiality to be given up in the course of the re-interpretation of the relative clause. As for the question of chronological priority of the re-interpreted nominal and verbal types, Spitaler seems to have assumed the verbal type to have been secondary (1962, 105). In support of this assumption it may be argued, as Spitaler has, that there is evidence of the nominal type from as early as the 9th century C.E. whereas the verbal type is attested somewhat later, but this does not necessarily mean that the verbal type had not existed before the earliest written evidence of it. More convincing is the argument that it is only the nominal type al-h amdu li-llāhi lladī that is part of the Qurān and thus a formula the Muslims had been familiar with from the beginning of Islam. Accordingly, we find in Tradition numerous examples of (the correct) al-h amdu li-llāhi lladī + a coreferential verb and many examples of al-h amdu li-llāhi alā + a noun, but few examples of the verbal type. We could even say that al-h amdu li-llāhi unexpanded or expanded has been and still is part of everyday Muslim speech, while this is not the case with the verbal type to the same

arabic allad ī as a conjunction


extent. On the other hand, it is the verbal type who seems to exhibit greater variability, and thus it may have been this type, or perhaps rather the existence of this type in its various subtypes, that may have triggered the re-interpretation of the more formulaic nominal type. Besides the examples of h md, there are examples of the same basic structures containing synonymous expressions, which shows a first semantic generalization of the re-interpreted alladī. There is my example (2) fa-li-llāhi l-h amdu wa-l-minnatu llatī (!) kānati l-aqībatu h amīdatan, and among Spitaler’s examples there is uškuri r-Rah īma r-Rah mān / alladī lam tajid al-malik Qays fī hādā l-makān “danke dem Allbarmherzigen, dass du den König Qais nicht an diesem Ort gefunden hast” from the Antar novel, where additionally Allāh is replaced by its synonym ar-Rah mān ar-Rah īm, reversed there because of the rhyme. Blau adduces further examples of synonymous verbs (sbh II, škr I, mjd II), which, however, are syntactically ambiguous (1967). We have reason to assume that the synonymous roots had already been in use, besides h md, as alternatives of h md from the first, as we have nominal examples such as wa-li-llāhi l-h amdu wa-š-šukru “To God praise and thanks are due!” (Dietrich, Briefe Hamburg, No 42a, 3 [c. 916/7 C.E.], and likewise in Tradition fa-laka l-h amdu wa-laka š-šukru Abū Dāwūd, Sunan, Kitāb al-adab, No 4411); wa-š-šukru li-llāhi “Thanks be to God!” (Anawati and Jomier, “Papyrus chrétien,” line 3 [9th c. C.E.]); wa-li-llāhi š-šukru katīran kamā huwa ahluhu wa-mustah iqquhu “To God repeated thanks are due as he is entitled to and worthy of it” (Rāġib, “Lettres” II, No 17r, 5 [9th c. C.E.]). Furthermore, there are the Christian formulae as-subh u li-llāhi “Praise be to God!” and al-majdu li-llāhi “Glory be to God!,” for which the reader is referred to my Briefe Heidelberg, 24 and 25 respectively. An example of šukr and a relative clause is fa-aqimi š-šukra li-llāhi lladī sānaka bi-dālika “Therefore extend in the right way your thanks to God, who preserved you thereby” (Rāġib, “Lettres” I, No 5, 22 [9th c. C.E.]). When al-h amdu li-llāhi lladī and h amidtu llāha lladī were re-interpreted, this development comprised the synonyms of h md also. A final point to be made regards the translation of the re-interpreted alladī as “that, daß.” Asking why alladī can be translated with simple “that” “daß,” while, according to what I have tried to point out above, its re-interpretation is based on an implicit causal relation, we should be aware of a hitherto neglected fact: the syntactic opacity of “that,” “daß.” As with Arabic an and anna, prepositions preceding “that,” “daß” in English and German can or must be elided, which leads to the basic


werner diem

logical relations, as expressed by the prepositions, becoming blurred on the surface. But the prepositions can also be retained on the surface, probably more so in German than in English. Thus, instead of the syntactically opaque “Lob sei Gott, daß wir gerettet worden sind” the more explicit “Lob sei Gott dafür, daß wir gerettet worden sind” is also possible. 4.3

Type B(a) malīh alladī “(it is) good that”

Half a dozen examples of this type were adduced by Spitaler (1962, 109f.) for modern dialects, intermingled however with examples of the farihtu lladī type (for this latter type, see below 4.4). It was Woidich who distinguished between the two groups, of which he characterized the first one as “satzäquivalente Ausdrücke” (1980, 226). Examples of his type A for Cairene Arabic are, apart from il-h amdu li-llāhi illi “Gott sei Dank, daß”, e.g., baraka illi “ein Segen, daß”, ya-xsāra lli “schade, daß”, and kuwayyis illi “gut, daß”. Corresponding examples were given by Spitaler for other modern dialects, whereas he had, according to himself, not come across examples in literary texts (1962, 110). Earlier evidence of this kind is my example (5) from the 11th century C.E. containing al-wayl lanā nah nu lladī “woe is us that”.27 Two other examples from that time quoted by Blau are yā-baxtik alladī “what good fortune for you (fem. sing.) that” and yā-h ayf alā abūkī lladī “what a pity for your (fem. sing.) father that” (1961, 227 sub ). These three early examples contain references to persons in their heads, but this is also the case with two modern Cairene examples cited by Woidich for this type A, and therefore this should not lead us to assume historically different types. As Woidich (1980) begins his Cairene evidence of this type with two examples of the re-interpreted il-h amdu li-llāh illi it is clear that he considers this type to be somehow the prototype of the whole group. This conclusion is confirmed by Woidich (1989), where he derives this group from the al-h amdu li-llāhi lladī expression, assuming as a first step “Funktionsschwäche” and as a second step a re-interpretation according to the “Satzstruktur Präd.—Subjekt” as in h ilwa di “prima ist die!” or ēb ikkalām da “eine Schande sind solche Worte!” As I have remarked above (2.7), sentences of this kind are not likely to have been connected by native speakers with al-h amdu li-llāhi lladī.


alladī might also be interpreted here as a relative pronoun standing for alladīna.

arabic allad ī as a conjunction


In my opinion the solution to this problem should not and need not be looked for in Cairene examples such as h ilwa di or their early NeoArabic equivalents if any existed. But before going into details, I think it appropriate to recapitulate some important points: (a) That the type A(a) al-h amdu li-llāhi lladī is likely to be the starting point of the type B(a) malīh un alladī may be assumed for the simple fact that it is only al-h amdu li-llāhi lladī where the transition of the relative particle alladī to a conjunction can be explained, whereas this is not feasible for malīh un alladī and the other expressions of this type. (b) The sentences of my type B(a) have a comment-topic structure. (c) When asking for a link between the type A(a) al-h amdu li-llāhi lladī and the type B(a) malīh un alladī, one can rightly assume, as Woidich did, that this link lies in that al-h amdu li-llāhi lladī had (finally) been reduced to being a (positive) evaluation of the fact mentioned in the alladī-clause. This reduced meaning of al-h amdu li-llāhi lladī would have been equivalent to ‘good that; nice that.’ However, there remains the question of the structure of the type B(a) malīh un alladī. In this respect, my theory differs from Woidich’s. In my opinion, we should not start from sentences such as Cairene h ilwa di or ēb ik-kalām da nor their Neo-Arabic equivalents in order to explain malīh un alladī and similar expressions. Asking rather whether there are other examples of the structure “comment + that + topic” in Classical Arabic (and Neo-Arabic), we can answer in the affirmative: there is an absolutely normal type consisting of a noun + an(na) clause, as for example malūmun anna “it is known that,” h asanun anna “it is good that,” barakatun anna “it is a blessing that” and so on.28 This is why we can transpose all examples of my type B(a) into “normal” Arabic by replacing alladī by an(na), with potential small changes due to anna demanding an accusative. It should be noted that the transformation is always possible from alladī to an(na) but not always from an(na) to alladī. Thus the transformation of malūmun an(naka) ji’ta into ? malūm alladī ji’ta would be very questionable, and for Cairene Arabic ? malūm illi gēt would probably not be acceptable to most speakers. This

28 In modern literary Arabic, the type mina l-malūmi anna would be more likely to correspond.


werner diem

is due to the restriction of this alladī type to specific heads referring to emotions. Starting from sentences evaluating events in an emotional manner of the type barakatun (h asanun) an(na) + perfect, which were structurally and semantically equivalent to al-h amdu li-llāhi lladī + perfect, the next step of the development, as I see it, was the replacement of an(na) by alladī in analogy with al-h amdu li-llāhi lladī. This development was made possible by two factors: (a) A semantic-syntactic connection between the type al-h amdu li-llāhi lladī + perfect and the type barakatun (h asanun) an(na) + perfect was mentally established, due to “Funktionsschwäche” of the H amdalah. (b) The alladī of al-h amdu li-llāhi lladī was mentally connected with an(na), because it overlapped with it in function after it had been re-interpreted as a conjunction. The generalization of alladī led to an opposition between an(na) and alladī, with alladī as the marked element being subject to certain restrictions as compared with the unmarked an(na). These restrictions, which the type B(a) had inherited from the primary type al-h amdu li-llāhi lladī “Praise be to God who” via the secondary type A(a), are as follows: (a) The structure of the type B(a) is always “comment + alladī + topic” and cannot be reversed. Thus malīh alladī ji’ta “It is nice that you came” is possible, whereas *?alladī ji’ta malīh is not. Woidich remarked this for his type B, with which my type B(b) partially overlaps, but this restriction is valid of his type A and my type B(a) also (1980, 235). (b) The head of the sentence consists of an “evaluating” expression. Thus malīh alladī jita “It is nice that you came” is possible, whereas *?malūm alladī jita “It is known that you came” is usually not. (c) The logical subject of the alladī-clause is human. Woidich remarked this for his type B for the syntactic subject of the alladī clause, but the same holds true for this type (1980, 230). Thus a sentence such as malīh lladī wasala is possible in the meaning “It is nice that he (sc. a certain person) arrived,” whereas the meaning “It is nice that it (sc. a letter or a parcel) arrived” is excluded or questionable. What, however, is attested in both the old and new evidence are alladī clauses with a non-human syntactic subject which however refer to a logical human subject. Thus, a sentence such as malīh alladī wasalanī “It is nice that it (sc. the letter) reached me” or, more freely,

arabic allad ī as a conjunction


“It is nice that I received it” would probably be possible, though perhaps not accepted by all speakers of individual dialects. (d) The subject of the alladī clause is almost exclusively pronominal and as such mostly implied in a finite verb. Thus again a sentence such as malīh alladī wasala “It is nice that he arrived” is possible, whereas malīh alladī wasala Ah mad “It is nice that Ahmad arrived” is with all probability either not possible at all or at least doubtful in most dialects. (e) The alladī clause refers to actual, mostly past events. Woidich found out this feature of “Faktivität” for his type B, but this is also true of his type A and my type B(a) (1980, 231). Thus malīh alladī jita “It is nice that you came” is possible, whereas malīh alladī tajī “It would be nice for you to come” is not in most dialects. These restrictions of alladī in type B(a) as compared with an(na) corroborate the assumption of its being the result of a generalization of the original al-h amdu li-llāhi lladī “Praise be to God who,” whereby the head was replaced with similar expressions while the other features of the construction were retained: (a) al-h amdu li-llāhi lladī najjānā has the structure “al-h amdu li-llāhi + alladī clause,” which is not reversible. (b) The head is evaluating. (c) The subject of the alladī clause, which is Allāh, is, in a anthropomorphistic perspective, “human.” (d) The subject of the alladī clause is pronominal, being implied in the verb that refers to Allāh. (e) The relative clause refers to past events only due to the semantics of “to praise” in the meaning of “to thank.” 4.4 Type B(b) farihtu alladī “I was glad that” Turning now to the type farih a alladī “he was happy that,” that is, alladī after verbs expressing emotions, the function of alladī as a conjunction is clearly borne out by cases where it follows a preposition as in (6) and (8) of my examples which have bi-lladī “by that.” While Spitaler holds that this type goes directly back to the al-h amdu li-llāhi lladī type by way of generalization (1962; see above 2.3), Woidich


werner diem

declares it to have had a separate origin in relative clauses where the subject of the head and the subject of the alladī clause are coreferential, as in Cairene ana h mār illi dafat il-h isāb (1989). In my opinion, this interpretation is flawed in several respects, for which the reader is referred to paragraph 2.7. Even if I do not find Woidich’s historical explanation of this type convincing, it is worthwhile to mention that the sentence which, according to him, was the starting point shares with al-h amdu li-llāhi lladī, in spite of their different structures, the three characteristics mentioned above: (a) The relative clause is non-restrictive, (b) there is an underlying causal relation between the head and the relative clause (“I am an ass! Why?”), (c) illi goes back to the uninflected alladī, to which the relative pronouns had been reduced. Before continuing, I think it appropriate and useful to draw attention to the fact that this type B(b) shares with type B(a) the restrictions (b)–(e) mentioned above, that is, all restrictions with the exception of the one concerning specific heads. As in the case of type B(a), a historical theory must take account of these restrictions and explain why they exist. The assumption that this type had developed from coreferential relative clauses as claimed by Woidich would explain the restrictions (a)–(c) but not (d) nor (e), as some emotions can concern both past and future events. Therefore, it is plausible to assume that this type B(b) developed from the verbal type A(b) h amidtu llāha lladī in the same way as the type B(a) developed from the nominal type A(a) al-h amdu li-llāhi lladī. In the first case, features of an impersonal expression were generalized, and in the other case, features of a personal expression. This development, as I see it, was as follows: The inherited conjunction in Arabic sentences of the type “I was glad that he came” was in Classical Arabic, and still is in most dialects, an(na), e.g. farihtu an (or annahu) jāa in the classical language. In a certain sense, praising God for an agreeable event that has happened is an expression of an emotion. Saying, for example, in German “Ich danke Gott, daß ich davon verschont geblieben bin,” is equivalent to “Ich bin froh, daß etc.” This semantic overlap made it possible for the alladī of h amidtu llāha lladī “I praised (thanked) God that” to be generalized to verbs expressing likewise the positive emotions of joy, contentment etc. Subsequently, as Spitaler assumed, this generalization was extended, by “Kontrastanalogie,” to verbs expressing negative emotions such as anger, reproach, grief and so on (1962, 109).

arabic allad ī as a conjunction


As a result of this development, the replacement of the marked alladī in B(b) by the unmarked an(na) is always possible, whereas the replacement of an(na) by alladī is only possible where the an(na) clause fulfills the above-mentioned restrictions. 4.5

Type B(c) alamtuhu lladī “I informed him that”

In Spitaler’s, Blau’s and my material examples remain where alladī “that” is preceded by verbs not expressing emotions. As I have mentioned above (2.3), this was a major problem for Spitaler. In my opinion, the problem is less serious than Spitaler thought it to be. It is true that the heads of these sentences do not contain expressions of emotion, but the contents of the alladī-clauses are something calling for positive or negative comments. Take, for example, the two following examples cited by Spitaler: wa-axbarūhu lladī qatalahu Antar “und sie teilten ihm mit, dass Antar ihn getötet hatte” (1962, 111) and, quoting Michel Feghali (1928, 313) for Lebanese Arabic, min hayk illi twaffiqti b-jāztik “c’est à cause de cela que tu as été heureuse dans ton mariage!,” with Feghali adding: “avec ironie.” Supporting evidence for this are also my example (11) with its first alladī and my examples (15)–(16). In sentences of this type, alladī has the function of expressing a certain emotion or the empathy of the speaker or narrator, the kind of which can be inferred from the context only. Thus, alladī in axbarūhu lladī qatalahu Antar indicates a certain emotion on the part of the narrator, which, depending on the specific context, can either be positive (joy) or negative (grief). It should be stressed that this kind of affective alladī is not a result of a secondary development but something inherent in alladī “that” from the beginning. From the moment when in al-h amdu li-llāhi lladī and h amidtu llāha lladī the relative pronoun was re-interpreted as a conjunction it became the marked counterpart of an(na), which through this development became unmarked. The difference between them lay in that an(na) continued to express the simple “that,” while alladī as its marked counterpart expressed a certain emotional involvement of the speaker. Should one try to express the specific affective content of alladī in sentences of this type B(c), one could do this, for example, by means of adverbs. Consequently, the first of the two sentences could be rendered, depending on the context, alternatively with “and they informed


werner diem

him that Antar had fortunately/regrettably killed him”. Strictly speaking, such explicit translation would be appropriate for any case of the conjunction alladī as long as it is restricted to this and the aforementioned types, translating, e.g., Cairene ana mutaassif illi taxxart with “I am so sorry that I was late.” 4.6

Remaining cases

There remain some examples in Spitaler’s corpus and elsewhere where the clause following alladī (illi) does not refer to past, but to future or possible events. It is interesting that two of Spitaler’s three examples are from Tunisian Arabic, among them ammin elli tūsil l-martek w-ulādek “sei sicher, dass du zu deiner Frau und deinen Kindern kommen wirst.” Here we have the final point of the development of alladī. As HansRudolf Singer pointed out for the dialect of Tunis, illi is used there in every context (1984, 669), apparently in the same sense and distribution as would be an(na) in Classical Arabic and inn- in many modern dialects including Cairene Arabic, e.g. andna z-zhar illi lqīnāhum29 “wir hatten Glück, daß wir sie trafen” on the one hand (referring to a past event) and tāh il-ahid illi ma yšūfhāš h atta . . . “er gab ihm das Versprechen, daß er sie bis [zum Hochzeitstag] nicht anschauen würde” (future 30 event). In this dialect, illi “that” is unmarked, and its marked “emotional” counterpart for expressing “that” is kīf (< kayfa “how”), for which this function is not an isolated phenomenon in modern dialects nor in the pre-modern substandard language. This remark about kīf typical of the dialect of Tunis must suffice here as the hitherto unwritten story of kayfa as a conjunction is too long to be told here in a few words. The dialect of Tunis is not the only one to have generalized the usage of alladī as a conjunction. The same usage of illi/li “that” is found in Maltese Arabic (Schabert 1976, 216 and Aquilina 1987, I 566), which is no coincidence as “La langue maltaise a pour origine un dialecte arabe, vraisemblablement proche des vieux dialectes citadins de Tunisie” (Vanhove 1993, 1). The Jewish dialect of Tripolis in Libya is yet another dialect using ëlli/li “that” (Yoda 2005, 278). In addition, similar extreme examples of illi “that” are known from other dialects but we lack compre-


I have simplified Singer’s complicated transcription. See also Woidich (1980, 224) for the same function of illi in the Judaeo-Arabic dialect of Tunis. 30

arabic allad ī as a conjunction


hensive studies of their exact distribution. For Cairene Arabic, one can state on the basis of Woidich’s two studies that examples as cited above for Tunisian Arabic would most certainly be declared to be ungrammatical or unusual by native speakers. As for the early history of Cairene Arabic, one should be cautious about taking the language of the Jewish-Arabic documents of the Cairo Genizah from the 11th century C.E. and later as early evidence of Cairene Arabic, because many of the writers of those documents hailed from the Maghreb, especially from what is today Tunisia, and there remained strong bonds between the Jewish traders who had settled in al-Fustāt and elsewhere in the Islamic East and their relatives and partners in the Maghreb. This means that deviations in those documents in the usage of alladī as a conjunction as compared with today’s Cairene Arabic might be ascribed to the Maghrebine background of their writers. Therefore “extreme” examples of the conjunction alladī in Judaeo-Arabic documents are to be connected to modern Tunisian Arabic rather than to modern Cairene Arabic. 4.7

alladī vs. an(na)

In describing the development of alladī as a conjunction, I have not dealt in detail with the various functions which it can assume. Some additional remarks might therefore be helpful. According to the development of alladī as I see it, the relative pronoun alladī in al-h amdu li-llāhi lladī and h amidtu llāha lladī was first re-interpreted in the sense of “that”, which is normally expressed by an(na), and then, by a gradual generalization, it replaced an(na) in specific syntactic and semantic contexts. We can express this by saying that alladī took over part of the functions of an(na) step by step, a process that led to an(na) and alladī becoming the unmarked and marked members of an opposition in many, if not all, dialects. There is a well-known rule in Arabic that any preposition preceding an(na) can be deleted. When alladī replaced an(na) in specific contexts, which have been described above, alladī consequently was felt to be submitted to the deletion rule as its counterpart alladī was. This meant that where, according to the syntactic context, a preposition was to be expected before alladī in analogy with an(na), this preposition could be considered to exist in the deep structure of the construction. On the other hand, since a preposition could be retained before an(na) without being deleted, alladī could likewise be preceded by prepositions, again in


werner diem

analogy with an(na). To put it another way, the existence of cases such as (6) wa-anā . . . muġtabitun bi-lladī ttasaltu ilayhim “I am . . . rejoicing that I joined them”, where alladī is preceded by the preposition bi- typical of the verb ġbt VIII on which it depends, proves that the process of the re-interpretation of alladī as a conjunction meaning “that” had taken place to its full extent. In the modern dialects, the combination of illi with prepositions does not seem possible; at least Woidich does not mention it for Cairene Arabic (1980 and 1989), and I do not have examples thereof either. Let us now have a look at some of my examples and see what syntactic status alladī has and which prepositions are possibly missing. I shall insert the prepositions missing in the surface structure but existing in the depth structure in brackets, and also give a translation of these prepositions: (1)

li-llāhi l-h amdu {alā} lladī kānati l-āqibatu li-xayrin “Praise be to God for the result having been good” Remark: The verb is h md I fulānan alā “to praise s.o. for” (15) wa-qad šakartu tafaddulahumā . . . {bi-}lladī qad dakarūnī fī kitābihimā bi-s-salāmi wa-bi-fili l-jamīli fī bābī “I am also grateful for their kindness . . . of giving greetings to me and of performing good deeds to me” Remark: The verb is fdl V bi- “to be so kind as to do” (13) wa-dāqa sadrunā katīr {li-}alladī lam yakūn laka maahu kitābun yutamminunā “and we were very much distressed due to his not having a letter of yours with him setting our minds at rest” Remark: (li-)lladī indicates the cause as (li-)an(na) does. On the other hand, there do exist many sentences where alladī introduces a subject or an object clause so that no preposition can be supplemented in the deep structure: (7)

wa-qad sarranī lladī anfadta lahu rah lahu “It pleased me that you sent him his merchandise” Remark: the alladī-clause is the subject of the sentence.

There is also the adnominal usage of alladī, forming, as is frequent with an(na), a syntagma that seems to be an apposition to a noun. It would

arabic allad ī as a conjunction


be difficult to elucidate the true syntactic status of alladī here. However, this is not necessary; it suffices to say that the status of alladī is here analogous to the status of an(na), by which it can be replaced. (9) innī ajabu minka lladī lam tusīb man yaktubu laka kitāb illā daf atan “I am astonished at you that you (allegedly) found only once somebody writing a letter down for you”

5. Summary In the preceding paragraphs, I have endeavored to sketch a picture of the development of alladī as a conjunction valid of all varieties of NeoArabic as far as they are known. The stages of the development are set off in the following table: (a) Original expressions containing relative clauses: al-h amdu li-llāhi lladī h amidtu llāha lladī “Praise be to God who” “I praised God who” (b) Re-interpretation of alladī as a causal conjunction on the pattern of parallel constructions with semantically explicit syntactic means, this process being due to (a) the relative clause being non-restrictive, (b) verbs such as “to praise” having an inherent complement indicating the cause of praise, (c) the relative pronouns having been reduced to alladī, whereby the connection of the relative clause to the head was weakened: 31 al-h amdu li-llāhi lladī h amidtu llāha lladī “Praise be to God that” “I praised God that” (c) Generalization of alladī for heads expressing positive emotions, by (optional) replacement of an(na) with alladī, while the syntactic and semantic restrictions typical of (a)–(b) were retained: malīh alladī (for malīh anna) farihtu lladī (for farihtu anna) “It is nice that” “I was glad that” (d) Generalization of alladī for heads expressing negative emotions ceteris paribus as in (c):


And by analogy semantically related verbs such as škr I, sbh II and mjd II.


werner diem

wayl laka lladī (for wayl laka taassaftu lladī (for taassaftu anna) anna) “Woe is you that” “I regretted that” (e) Generalization of alladī for heads not expressing emotions, with alladī still having an affective value and retaining its original restrictions as in (a)–(d): alamtuhu lladī “I informed him (of the pleasant/regrettable fact) that” (f) Generalization of alladī as an unmarked conjunction “that” without syntactic or semantic restrictions: tāh il-ahid illi ma yšūfhāš “He made a pledge that he would not see her” (Tunisian Arabic) Stage (f), which is found in the dialect of Tunis and the dialect of the Jews of Tripolis (Libya), is the final point of the development of alladī as a conjunction. The development of alladī (illi) in Cairene Arabic, which is the best-known of all Arabic dialects, reached stage (d) only.

6. Appendix: Notes on the origin of al-h amdu li-llāhi lladī As demonstrated in paragraph 4.2, there were, besides the relative clause, other, more explicit means of expanding the H amdalah. This raises the question, which has not yet been given the attention it deserves, why in the case of al-h amdu li-llāhi a causal relation between a subordinate clause and the head is expressed by a relative clause if the subject of the relative clause is coreferential with Allāh of the head, and not more explicitly. Indeed, in terms of “normal” Arabic one is in some respect entitled to say that al-h amdu li-llāhi lladī expressing causal relations is a special case, which is, however, so innate in the religious language of Islam that Muslims are not likely to be aware of this fact. On the other hand, the very fact that the relative pronoun was re-interpreted, at some time in the history of Arabic, as a conjunction demonstrates that this relative clause was something peculiar not wholly in line with the normal syntax of Arabic. That Western Arabists do not seem to have felt al-h amdu li-llāhi lladī to be something special either, must be ascribed to the existence of corresponding relative clauses in the religious style of the Western languages. For German, I can say by way of introspection that this kind of causal relative clauses, apart from the religious language, e.g. “Lob

arabic allad ī as a conjunction


sei Gott dem Herrn, der etc.,” is restricted to a high stylistic level and is possible for verbs such as “danken, preisen, loben” only, e.g. “Es ist mir ein Anliegen, an dieser Stelle Herrn N.N. zu danken, der dieses Projekt hochherzig gefördert hat.” A corresponding formulation in less formal language is, e.g., “Ich möchte Herrn N.N. dafür danken, daß er dieses Projekt unterstützt hat.” So the fact that al-h amdu li-llāhi lladī seems familiar to us must not prevent us from enquiring as to the reason for this construction. To this question the following lines will be dedicated. Considering that al-h amdu li-llāhi was in use in pre-Islamic times already (see above 5.2), we can assume that its expanded form al-h amdu li-llāhi lladī is likewise pre-Islamic, although there is no explicit evidence of it so far, at least none that I am aware of. The basic formula itself is probably a calque on the Syriac šubh ā l-alāhā, as remarked by Theodor Nöldeke and Friedrich Schwally (1909, 112, footnote 1) for the Fātih ah of the Qurān: “  entspricht genau syrischem šubh ā l-alāhā32 bezw. tešbohtā l-alāhā und neutestamentlichem δόξα τῷ θεῷ.” In this context, Nöldeke and Schwally also mention the so-called Berākā in the Old Testament and the Christian liturgy as a parallel to the Syriac and Greek formulae, without, however, connecting it to the Qurān. Anton Baumstark’s article “Jüdischer und christlicher Gebetstypus im Koran” (1922) is in a certain sense a comment on the short remarks of Nöldeke and Schwally, although he does not refer to them. In the following, I shall first sum up some important points of Baumstark’s article and then add some deliberations of my own: (a) In the Old Testament, there is the so-called Berākā (“blessing”) of the structure bārūk Yahwē “Blessed is (or be) Yahwe,” to which a nominal attribute or a relative clause can be added, e.g. bārūk Yahwē ašer hissī l etke m miy-yad Misrayim ū-miy-yad Parō “Blessed be Yahwe, who delivered you from the hand of the Egyptians and the hand of Pharaoh” Exodus 18:10.33 A corresponding Qurānic expression is tabāraka in the subtypes tabāraka llāhu rabbu l-ālamīna “Blessed be God, the Lord of the whole world” Q 7:54 and tabāraka lladī bi-yadihī l-mulku “Blessed be the One in whose hand is the kingdom” Q 67:1, as well as the exceptional būrika in būrika man fī n-nāri 32

In the original text in Syriac script. This passage has already been cited by Spitaler (1962, 107) but in another context, where he asks whether this expression might be a parallel to the development of al-h amdu li-llāhi lladī. Instead of Spitaler’s h issī l, read hissī l. 33


werner diem

wa-man h awlahā “Blessed be He Who is in the fire and around it” Q 27:8. Baumstark tends to assume that the Berākā was transmitted to Muhammad via Christian formulae in the New Testament and elsewhere, in which the Jewish Berākā lives on, of the type Εὐλογητός ὁ Θεός, a last echo of this development being Benedictus dominus (Baumstark 1922, 231ff.). A fact not mentioned by Baumstark is that the Qurānic formula is syntactically different both from the Biblical Berākā and the H amdalah, as it lacks a subtype tabāraka llāhu lladī “Blessed be God who.” (b) The type which in the Christian liturgy, especially of the East, has become the prevailing one is the so-called doxology, which has probably developed from the Hebrew Berākā, e.g. Σoὶ ἡ δόξα, Gloria tibi domine, Laus tibi Christe, and so forth, the doxology being “[eine] possessive Form der Prädikation, durch welche Herrlichkeit, Ruhm, Lob, Ehre oder wie immer man das schillernde griechische δόξα wiedergeben will, als der Gottheit eigen oder gebührend bezeichnet wird” (Baumstark 1922, 234). The dative of the Greek formula can be expanded by “eine partizipiale Apposition oder Anrede,” and the reason for God’s praise can also be given in the form of a ὅτι clause. This Christian doxology is likewise found in the Qurān, in two forms: as subh āna + a pronoun or a genitive (which does not interest us in this context) and as al-h amdu li- (Baumstark 1922, 234ff.). However, as Baumstark remarks: Der frühchristlichen partizipialen Ergänzung einer Doxologie entspricht diejenige eines     fast immer vielmehr durch einen Relativsatz [. . .] (1922, 237)

After dealing with details of subh āna and al-h amdu li- in the Qurān and comparing them to Christian liturgical formulae they are likely to be derived from, Baumstark finally poses the question of how these formulae may have passed to Muhammad. He supposes that they were brought to the Arabs by Nestorian missionaries, who translated their Syriac religious texts into Arabic, which he says is typical of the Nestorians: Die nestorianische Kirche hat aber immer wieder eine auffallende Geneigtheit bekundet, ihre angestammte syrische Kultsprache beim Betreten neuer Missionsgebiete der Volkssprache derselben zu opfern (1922, 247f.)

So far my summary of Baumstark’s article. As we have seen, Baumstark is aware of the fact that the Qurānic al-h amdu li-llāhi is expanded by a

arabic allad ī as a conjunction


relative clause while there are participles and ὅτι clauses in the Greek doxologies, but he does not pay further attention to this difference. In my opinion, the solution to this problem is found in the Syriac context, through which the doxology was probably transmitted to the Arabs. Baumstark (1922, 235) mentions himself that the Syriac doxology is šubh ā l-, without however further dwelling upon it, while Nöldeke and Schwally (1909), as mentioned above, were of the opinion that al-h amdu li-llāhi goes back to šubh ā l-alāhā. It is also interesting in this context that for Barhebræus, as cited by S. Payne Smith (1879 II, 4026), the šubh ā l- formula must have been closely connected to the H amdalah, as he renders in his dictionary the Syriac šubh ā l-mrayymā l-ālmīn “Praise be to the Elevated One to all eternity” with the Arabic al-h amdu li-l-ālī ilā d-dahri. In the Syriac New Testament, the simple doxology is part of a sentence in w-kulleh ammā da-h zā yab-wā šubh ā l-alāhā “Et omnis plebs ut vidit, dedit laudem Deo” Lk 18:43. Syntactically independent simple doxologies are šubh ā l-alāhā “Praise be to God” (Payne Smith 1903, 563a) and šubh ā l-abā w-la-brā wa-l-rūh ā qaddīšā “Praise be to the Father, the Son and the Holy Spirit” (Payne Smith 1879 II, 4026). As for expanded doxologies, we find evidence of them in the Breviarium Chaldaicum, e.g. šubh ā l-tābā da-b-yad h ubbeh glā tešbohtā la-bnaynāšā “Praise be to the Good One, who through His love revealed glory to men” (Breviarium I, kh = 28) or šubh ā l-māryā da-b-yad tuqpeh hg am la-trūnā wa-p raq l-abdaw “Praise be to the Lord, who through His power overthrew the tyrant and saved His servants” (Breviarium III, 333). Examples of expanded doxologies in colophons of books are šubh ā l-abā w-la-brā wa-l-rūh ā qaddīšā d-h ayyel wa-dar w-sayya “Praise be to the Father, the Son and the Holy Spirit, who strengthened, assisted and supported (respectively)” and šubh ā l-abā d-h ayyel w-la-brā d-sayya wa-l-rūh ā d-qudšā d-šamlī “Praise be to the Father, who strengthened, the Son, who assisted, and the Spirit of Holiness, who accomplished” (Payne Smith 1879 II, 4026). We can safely infer from these examples that the expanded Arabic doxology al-h amdu li-llāhi lladī is nothing but a rendition of the Syriac šubh ā l-alāhā d- in the same way as the simple Arabic doxology al-h amdu li-llāhi is a rendition of the Syriac šubh ā l-alāhā. This means that the Arabic relative pronoun alladī corresponds to the Syriac d-, which in its chief function is also a relative pronoun. It is true that d- is also a conjunction of wide and vague meaning so that the d-clauses of the


werner diem

above-mentioned Syriac doxologies might be translated as causal clauses as well, e.g., šubh ā l-abā d-h ayyel w-la-brā d-sayya wa-l-rūh ā d-qudšā d-šamlī as “Praise be to the Father because He strengthened, the Son because he assisted, and the Spirit of Holiness because he accomplished.” On the other side, it may be argued that the d-clauses of Syriac doxologies, at least to my knowledge, are all coreferential with the nouns of the l-phrases and thus likely to have been conceived of as relative clauses rather than as causal clauses. Although the relative pronoun of the Qurānic al-h amdu li-llāhi lladī is thus accounted for, there remains the question why the Syriac doxology is expanded by d-clauses while the older Greek formulae have participles and ὅτι clauses. The solution to this problem lies probably in the difference of the verbal systems. The Greek verbal system, which per se is more complex than the West Semitic one, is still more so with regard to the active participle, as there are different Greek participles derived from the corresponding tenses, whereas the Semitic languages have but one active participle, whose temporal value depends on various factors. Therefore, when Greek doxologies with active participial attributes were translated into Syriac, the problems connected with this disparity seem to have been avoided by translating the Greek participle, if the specific context demanded this, with a relative clause containing a finite verb in the (Syriac) perfect. As the Syriac relative pronoun d- is also a causal conjunction, Greek expansions of the doxology by means of the conjunction ὅτι “that; because” could likewise be rendered with d-. In this process of rendering Greek attributive participles and ὅτι clauses with Syriac d- clauses, the translators may also have been influenced by the Berākā of the Old Testament with its ašer, which is also both a relative particle and a causal conjunction, and which in the Syriac Old Testament is rendered with d-. Thus the Hebrew bārūk Yahwē ašer hissī l etke m miy-yad Misrayim ū-miy-yad Parō Exodus 18:10 is rendered as the Syriac brīk-ū māryā d-p assī kōn men īdā d-Misrāyē w-men īdeh d-P erōn, with d- exactly corresponding to ašer. Even if there exists, as I have tried to demonstrate, a historical connection between the Hebrew bārūk Yahwē ašer, the Syriac šubh ā l-alāhā d- and the Arabic al-h amdu li-llāhi lladī, this does not fully explain why relative clauses were chosen from the beginning and not explicit causal conjunctions as the Hebrew kī, the Syriac mettu l d- or the Arabic alā an or li-an. The reason is probably that the implicit indication of the cause is more in line with the religious attitude deemed appropriate

arabic allad ī as a conjunction


towards God. If an explicit causal conjunction were used the praise of God would be limited to a specific deed of His as mentioned in the subordinate clause, that is, God would be praised for this specific deed and for nothing else. But this would not be adequate as God, according to Jewish, Christian and Islamic thinking, is entitled to praise in general, as is clearly borne out by the unexpanded Berākā and the unexpanded doxology. Thus by not explicitly expressing an (implicitly existent) causal relation the relative clause is much more in line with a general, unconditioned praise of God than an explicit causal clause would be. Accordingly, the Arabic H amdalahs, nominal and verbal ones, where the subordinate clause is introduced by an explicit conjunction (including the re-interpreted alladī followed by non-coreferential clauses) do not imply a general praise of God but are limited to praise of God for specific facts. In this respect, sentences of this kind belong to the language of everyday life and not the language of religious worship, a fact which is demonstrated by the contexts in which such sentences occur. A final remark concerns the Arabic verbal type h amidtu llāha lladī/ an. There is the Syriac verb šabbah “to praise,” and this verb is likely to have been used with God as the object. However, the dictionaries do not contain evidence of a formula such as šabbh et l-alāhā d- “I praised God who (or that)” used for the expression of gratitude to God for individual bounties as the Arabic h amidtu llāha lladī/an is. Furthermore, the verbal forms in both languages are different. Consequently, the verbal h amidtu llāha lladī/an is the result of a specific Arabic development rather than a calque on a Syriac expression. This in turn would suggest that the doxology had a history of its own in Arabic after it had been taken over in its formulaic nominal form from Syriac. Now, as Hartmut Bobzin (2004, 59) has stressed, it can be concluded from certain passages of the Qurān that even in pre-Islamic times Allāh had been considered a “Schöpfer- und Rettergott,” a designation of which the notion of “Rettergott” (saviour God) is significant in this context. For if Allāh was considered a “Rettergott” he was especially qualified for the H amdalah as in every day life the H amdalah, nominal or verbal, is mostly used for the happy outcome of an affair. It seems, therefore, that in the Arabic H amdalah the Christian doxology might well have melted in a syncretistic way with pre-existing pagan notions of the role of Allāh.


werner diem 7. References

7.1 Primary sources Abū Dāwūd, Sunan = Abū Dāwūd Sulaymān b. al-Ašat  as-Sijistānī, Kitāb as-Sunan. See Mawsūat al-hadīt. Ahlwardt, Wilhelm see Arazi and Masalha Amari, Diplomi = Michele Amari, I diplomi arabi del R. Archivio fiorentino. Florence: Tipografia di Felice le Monnier, 1863. Anawati and Jomier, “Papyrus chrétien” = PP. Anawati and Jomier, “Un papyrus chrétien en arabe (Égypte, IXe siècle Ap. J.-C.).” Mélanges islamologiques 2, 1954, 91–102. Arazi and Masalha, Six Early Arab Poets = Albert Arazi and Salman Masalha, Six Early Arab Poets. New Edition and Concordance Based on W. Ahlwardt’s The Divans of the Six Ancient Arabic Poets. Jerusalem: The Hebrew University, 1999. Ashtor, “Documentos” = Eli Ashtor, “Documentos españoles de la Genizah.” Sefarad 24, 1964, 41–80. Assaf, Meqorot = Simha Assaf, Meqorot u-meh qarim be-toledot Yisrael. Jerusalem, 1946. Braslavski, “Mishar” = Josef Braslavski, “Al ham-mishar hay-yehudi ben hay-yam hattikon wa-Hoddu bam-meah ha-12.” Zion 7, 1941–1942, 135–139. Breviarium Chaldaicum = Breviarium iuxta ritum Syrorum orientalium id est Chaldaeorum. Slawwātā qānūnāyātā d-kāhnē. Rome, 2002. (Reprint of Eugène Tisserant’s edition, Rome 1938, itself a reprint of the original edition, Rome 1886.) al-Buxārī, Sah īh = Muhammad b. Ismāīl al-Buxārī, Kitāb al-Jāmi as-sah īh. See Mawsūat al-hadīt. ad-Dārimī, Sunan = Abd Allāh b. Abd ar-Rahmān ad-Dārimī, Kitāb as-sunan. See Mawsūat al-hadīt. Diem, Briefe Heidelberg = Werner Diem, Arabische Briefe auf Papyrus und Papier aus der Heidelberger Papyrus-Sammlung. Wiesbaden: O. Harrassowitz, 1991. Diem, Geschäftsbriefe Wien = Werner Diem, Arabische Geschäftsbriefe des 10. bis 14. Jahrhunderts aus der Österreichischen Nationalbibliothek in Wien. Wiesbaden: Harrassowitz, 1995. Dietrich, Briefe Hamburg = Albert Dietrich, Arabische Briefe aus der Papyrussammlung der Hamburger Staats- und Universitäts-Bibliothek. Hamburg: J. H. Augustin, 1955. Dīwān al-Hudalīyīn = Kitāb šarh ašār al-Hudalīyīn sanat Abī Saīd al-H asan ibn al-H usayn as-Sukkarī. Ed. by Abd as-Sattār Ahmad Farrāj and Mahmūd Muhammad Šākir. Cairo: Dār al-Urūbah, 1965. Gil, Documents = Moshe Gil, Palestine during the First Muslim Period (634–1099). Part I. Studies. Part II. Cairo Geniza Documents. Part III. Cairo Geniza Documents. Tel Aviv: Tel Aviv University etc., 1983. Gil, Texts = Moshe Gil, In the Kingdom of Ishmael. Volume I. Studies in Jewish History in Islamic Lands in the Early Middle Ages. Volume II. Texts from the Cairo Geniza. The Jews of Iraq and Persia (nos. 1–101). Letters of Jewish Merchants (nos. 102–303). Volume III. Texts from the Cairo Geniza. Letters of Jewish Merchants (nos. 304–607). Volume IV. Texts from the Cairo Geniza. Letters of Jewish Merchants (nos. 608–846). Indexes. Jerusalem: Tel Aviv University etc., 1997. Goitein, “Arkiyon” = Shelomo Dov Goitein, “Olelot me-arkiyono šel Yūsuf ibn Awkal.” Tarbiz 38, 1968, 18–42. Goitein, “Iggeret” = Shelomo Dov Goitein, “Iggeret Labrat  ben Moše ben Siġmār dayyan ha-ir al-Mahdīyah.” Tarbiz 36, 1967, 59–72. Goitein, “Kneset” = Shelomo Dov Goitein, “Bet hak-kneset we-siyyudo lepi kitbe hagGenizah.” Erez Israel 7, 1964, 81–97. Goitein, “Saloniqi” = Shelomo Dov Goitein, “Eduyot qedumot min hag-Genizah al qehillat Saloniqi.” Sefunot, 11, 1970, 11–33.

arabic allad ī as a conjunction


Gottheil and Worrell, Fragments = Richard Gottheil and William H. Worrell, Fragments from the Cairo Genizah in the Freer Collection. New York: Macmillan, 1927. Ibn Abī d-Dam aš-Šāfiī, Adab al-qadā = Šihāb ad-Dīn Abū Ishāq Ibrāhīm b. Abd Allāh al-Hamdānī al-H amawī al-marūf bi-bn Abī d-Dam aš-Šāfiī, Kitāb adab al-qadā. Ed. by Muhyī Hilāl as-Sarhān. 2 volumes. Bagdad: Matb aat al-iršād, 1984. Ibn H anbal, Musnad = Ahmad b. H anbal, Kitāb al-Musnad. See Mawsūat al-hadīt. al-Jāhiz, H ayawān = Abū Utm  ān Amr b. Bahr al-Jāhiz, Kitāb al-h ayawān. Ed. by Abd as-Salām Muhammad Hārūn. Cairo: Mat baat Lajnat at-talīf wa-t-tarjamah wa-nnašr, 1969. Labīd, Dīwān = Šarh Dīwān Labīd b. Rabīah al-Āmirī. Ed. by Ihsān Abbās. Kuwait: Governmental Press, 1962. Mawsūat al-hadīt  = Mawsūat al-hadīt  aš-šarīf. al-Kutub at-tisah. CD-Rom. 1st edition. Šarikat Saxr li-barāmij al-hāsib (1991–1996). an-Nasāī, Sunan = Abū Abd ar-Rahmān an-Nasāī, Kitāb as-Sunan. See Mawsūat al-hadīt. Rāġib, “Lettres” I–II = Yūsuf Rāġib, “Lettres arabes.” Annales islamologiques 14, 1978, 15–35; 16, 1980, 1–29. Toledano, “Teudot” = Jacob Moses Toledano, “Teudot mik-kitbe yad.” Hebrew Union College Annual 4, 1927, 449–467. az-Zamaxšarī, Mufassa l = Abū l-Qāsim Mahmūd b. Umar az-Zamaxšarī, Kitāb al-Mufassa l fī n-nahw. Ed. by J. P. Broch. Christiania etc.: P. T. Mallingius, 1879. 7.2 Secondary sources Aquilina, Joseph. 1987. Maltese-English Dictionary. Malta. Baumstark, Anton. 1927. “Jüdischer und christlicher Gebetstypus im Koran.” Der Islam 16, 229–248. Blau, Joshua. 1961. Diqduq ha-arabit ha-yehudit šel yeme hab-benayim. Jerusalem: Magnes Press, The Hebrew University. ——. 1965. The Emergence and Linguistic Background of Judaeo-Arabic. A Study of the Origins of Middle Arabic. Oxford: Oxford University Press. ——. 1966–1967. A Grammar of Christian Arabic Based Mainly on South-Palestinian Texts from the First Millennium. 1–3. Louvain: Secrétariat du Corpus SCO. Bobzin, Hartmut. 2004. Der Koran. Eine Einführung. München: C.H. Beck, 5th edition. Bravmann, Meir M. 1953. Studies in Arabic and General Syntax. Cairo: Imprimerie de l’Institut français d’archéologie orientale. Brockelmann, Carl. 1922. “Allah und die Götzen, der Ursprung des islamischen Montheismus.” Archiv für Religionswissenschaft 21, 99–121. Feghali, Michel. 1928. Syntaxe des parlers arabes actuels du Liban. Paris: P. Geuthner. Goldenberg, Gideon. 1994. “alladī al-masdariyyah in Arab grammatical tradition.” Zeitschrift für Arabische Linguistik 28, 7–35 = Gideon Goldenberg, Studies in Semitic Linguistics. Selected Writings, 250–285. Jerusalem: Magnes Press, The Hebrew University, 1998. Hopkins, Simon. 1984. Studies in the Grammar of Early Arabic Based upon Papyri Datable to before 300 A.H./912 A.D. Oxford: Oxford University Press. Lehmann, Christian.1984. Der Relativsatz. Typologie seiner Strukturen—Theorie seiner Funktionen—Kompendium seiner Grammatik. Tübingen: G. Narr. Nöldeke, Theodor. 1909. Geschichte des Qorāns. Bearbeitet von Friedrich Schwally. Erster Teil. Über den Ursprung des Qorāns. Leipzig. Reprint Hildesheim: G. Olms, 1961. Payne Smith, J. 1903. A Compendious Syriac Dictionary founded upon the Thesaurus Syriacus of R. Payne Smith, DD. Ed. by J. Payne Smith (Mrs. Margoliouth). Oxford: Oxford University Press.


werner diem

Payne Smith, R. 1879. Thesaurus Syriacus. Oxford: Clarendon. Piamenta, Moshe. 1979. Islam in Everyday Arabic Speech. Leiden: E.J. Brill. ——. 1983. The Muslim Conception of God and Human Welfare as Reflected in Everyday Arabic Speech. Leiden: E.J. Brill. Reckendorf, Hermann. 1921. Arabische Syntax. Heidelberg: C. Winter. Rubin, Aaron D. 2005. Studies in Semitic Grammaticalization. Winona Lake, Indiana: Eisenbrauns. Schabert, Peter. 1976. Laut- und Formenlehre des Maltesischen anhand zweier Mundarten. Erlangen. Singer, Hans-Rudolf. 1984. Grammatik der arabischen Mundart der Medina von Tunis. Berlin—New York: W. de Gruyter. Spitaler, Anton. 1962. “al-h amdu li-llāhi lladī und Verwandtes. Ein Beitrag zur mittel- und neuarabischen Syntax.” Oriens 15, 1962, 97–114 = Anton Spitaler, Philologica. Beiträge zur Arabistik und Semitistik, ed. by Hartmut Bobzin, 230–247, 248 (“Zusätze”). Wiesbaden: Harrassowitz, 1998. Vanhove, Martine. 1993. La langue maltaise. Etudes syntaxiques d’un dialecte arabe ‘périphérique.’ Wiesbaden: Harrassowitz. Woidich, Manfred. 1980. “illi als Konjunktion im Kairenischen.” Studien aus Arabistik und Semitistik. Anton Spitaler zum siebzigsten Geburtstag von seinen Schülern überreicht, ed. by Werner Diem and Stefan Wild, 224–238. Wiesbaden: Harrassowitz. ——. 1989. “illi ‘dass,’ illi ‘weil’ und zayy illi ‘als ob’: zur Reinterpretation von Relativsatzgefügen im Kairenischen.” Mediterranean Language Review 4–5, 1989, 109–128. Yoda, Sumikazu. 2005. The Arabic Dialect of the Jews of Tripoli (Libya). Grammar, Text and Glossary. Wiesbaden: Harrassowitz.


1. Introduction Le texte, bien connu, que nous traduisons en annexe, est extrait du Kitāb al-Īdāh fī ilal al-nahw d’az-Zajjājī. Az-Zajjājī « le Zajjâjien » est le surnom sous lequel est connu Abū l-Qāsim ‘Abd al-Rahmān b. Ishāq. Ce grammairien d’origine iranienne du IVe/Xe siècle doit son surnom au fait qu’il étudia à Bagdad auprès du grammairien Ibrāhīm b. as-Sarī az-Zajjāj (m. 311/923–4). Il s’installa ensuite en Syrie, à Alep, puis à Damas, avant de se rendre en Palestine, à Tibériade, où il mourut en 337/949. On connaît un peu moins d’une vingtaine d’ouvrages d’az-Zajjājī. Une dizaine environ a été publiée. Parmi ceux-ci, deux se distinguent : – le Kitāb al-Jumal, qui est un ouvrage de grammaire devant son nom au fait, non qu’il traite de phrases, mais qu’il est constitué de « notes de synthèse » ( jumal) sur les différents chapitres de la grammaire. Il a été publié en 1926 à Alger par Mohammed Ben Cheneb et republié à Paris en 1957. – le Kitāb al-Īdāh fī ilal al-nahw, qui n’est pas un ouvrage de grammaire, mais sur la grammaire. Il a été publié par Māzin Mubārak au Caire en 1959, puis republié à Beyrouth en 1973 et 1979. Vers la fin des années 60 et le début des années 70, il y eut, avec l’explosion de la linguistique, celle d’une sous-discipline : l’histoire de la linguistique. Le mouvement atteignit même les arabisants. De par sa nature même, le Īdāh d’al-Zajjājī attira l’attention, en particulier celle 1 Ce texte est la version écrite de la leçon ERASMUS faite au séminaire du Pr. Dr. Andreas Kaplony, à l’Orientalisches Seminar de l’Université de Zürich, le mardi 19 Avril 2005. Que les collègues et étudiants de l’Orientalisches Seminar soient remerciés pour leurs remarques et questions, dont a bénéficié la version finale.


pierre larcher

de Kees Versteegh, qui le traduisit dans le cadre de son MA (1971), et l’utilisa abondamment dans sa thèse Greek Elements in Arabic Linguistic Thinking (1977). Néanmoins, c’est seulement en 1995 que Versteegh publia cette traduction, sous le titre de The Explanation of Linguistic Causes. Az-Zajjājī’s Theory of Grammar. L’ouvrage est typique du IVe/Xe siècle, en ce que s’y révèle partout l’influence de la falsafa (ou philosophie hellénisante), à commencer sur la forme même de l’ouvrage : celle, dialectique, héritée de l’Antiquité grecque, par question et réponse. Notre texte constitue le chapitre XIV (89–90) de l’ouvrage. Il répond à la question de savoir pourquoi la grammaire a été nommée nahw en arabe. Notons que c’est le même mot de illa, pluriel ilal, qui apparaît dans le titre du chapitre et dans le titre de l’ouvrage. En revanche, dans la formulation de la question, au début du chapitre, apparaît celui de sabab. Cela peut amener à penser que les deux termes sont synonymes. Et c’est sûrement ce qui détermine Versteegh à traduire ilal par causes dans le titre de l’ouvrage et illa et sabab dans le texte du chapitre par le même mot de reason. Il existe cependant entre sabab et illa la même différence qu’en français entre cause et justification. La cause est objective et relève de l’ordre logique : A parce que B. La justification en revanche est intersubjective et relève de l’ordre dialectique : A car/puisque B. N’oublions jamais que le rapport des autres disciplines à la falsafa est dialectique : par les questions qu’elle pose, la falsafa les oblige à répondre, en se justifiant. Ce texte est capital, non seulement pour l’histoire de la grammaire arabe, mais encore celle de la langue arabe. Telle qu’elle a été comprise par la tradition arabe, la première tire en effet son origine de la « corruption » ( fasād) de la seconde.

2. Description Le texte nous dit où, quand et comment ce processus a lieu et il nous dit aussi en quoi il consiste. Où : à Basra, c’est-à-dire dans une des villes nouvelles créées à la suite de la conquête islamique, ce qui répond en même temps à la question du quand : Basra a été fondée en 16/637. Comment : par un double processus de sédentarisation des Bédouins et de mélange des populations arabes avec des populations non-arabes.

les origines de la grammaire arabe


Apparaissent dans le texte quelques termes fondamentaux. D’abord celui de Arab, dans l’expression abnā al-Arab (litt. Fils, enfants des Arabes), opposé tout à la fois à al-h ādira et abnā al-Ajam (litt. Fils, enfants des non-Arabes). Cela veut dire que Arab ne s’oppose pas seulement ici à Ajam comme Arabes à non-Arabes (et, plus particulièrement, dans le contexte local, à Persans), mais encore comme Bédouins à sédentaires. Au témoignage même du Lisān al-Arab (désormais LA) de Ibn Manzūr, m. 711/1311 (art. hdr) : « al-h adar, al-h adra et al-h ādira sont le contraire de al-bādiya et il s’agit des villes, des villages et de la campagne » (alh adar wa-l-h adra wa-l-h ādira xilāf al-bādiya wa-hiya al-mudun wal-qurā wa-r-rīf ). Al-h ādira désigne donc bien le pays sédentaire par opposition à al-bādiya ou pays bédouin. Voilà pour la sédentarisation. Venons en maintenant au mélange des populations. Celui-ci n’est pas évoqué explicitement au travers du terme habituel de muxālata, mais implicitement au travers de celui de muwalladūna-īna. On traduit ordinairement par « métis » (< lat. mixticius « mélangé »). On a voulu voir dans muwallad l’étymon de « mulâtre » (Kazimirski 1846–7, art. wld), via l’espagnol mulato, mais ce dernier terme se rapporte peut-être plus simplement au latin mulus/mula (« mule(t) », cf. en espagnol même, outre mulo/mula « mule(t) », muleto/ muleta « jeune mule(t) »), à tout le moins a subi une contamination de cette famille lexicale. Dans la même veine l’article muwallad de EI2 indique qu’il s’agit d’un « terme appartenant au vocabulaire des éleveurs et désignant le produit d’un croisement (tawlīd) entre deux races animales différentes, donc un hybridé, un sang mêlé » et que c’est par analogie que le terme a été étendu aux humains. Mais l’article ne donne aucune référence pour ce sens, que nous ne trouvons pas, par exemple, dans LA (art. wld). La question se pose donc au linguiste de l’articulation de la désignation historique du terme et de sa signification. Morphologiquement, muwallad est le participe passif du verbe wallada. Wallada est le factitif du verbe de base, mais il renvoie à l’actif ou au passif de celui-ci, donc à walada-yalidu ou à wulida-yūladu, selon qu’il est doublement ou simplement transitif. Le verbe de base walada-yalidu, simplement transitif, signifie « engendrer un enfant » (et plus particulièrement « accoucher d’un enfant », s’il se dit d’une femme, ou « mettre bas un petit », s’il se dit d’une femelle). Le verbe doublement transitif wallada-hā -hu signifie « faire en sorte qu’une femme accouche ou qu’une femelle mette bas », c’est-à-dire l’aider à accoucher ou à mettre bas. Muwallida est un des noms de la sage-femme (qābila). Le verbe simplement transitif wallada signifie « faire naître quelqu’un » ou « générer quelque chose ».


pierre larcher

Dans les deux cas, cependant, muwallad signifie, comme seul objet ou second des deux objets de wallada, « engendré ». C’est sûrement par l’idée de « mis au monde » que le terme muwallad a pris, tout à la fois par métaphore et généralisation, le sens de « tout ce qui est nouveau, moderne » (al-muh dat min kulli šay LA, art. wld). Il peut se dire, soit de quelqu’un, soit de quelque chose : il se dit en particulier des « poètes modernes » (al-muwalladūn min aš-šuarā) et des néologismes (summiya al-muwallad min al-kalām muwalladan idā istah datūhu wa-lam yakun min kalāmihim fīmā madā « ce qu’il y a de muwallad dans le parler a été ainsi appelé, quand on le produit, sans qu’il ait existé dans le parler auparavant »). Comme tout ce qui est nouveau, le terme peut s’entendre en mauvaise part comme quelque chose de fabriqué, controuvé, apocryphe. C’est sans doute par une extension de ce dernier sens que très tôt le terme a pris le sens de « non purement arabe », pouvant se dire, là encore, soit de quelqu’un soit de quelque chose, cf. LA, art. WLD arabiyya muwallada wa-rajul muwallad idā kāna arabiyyan ġayr mah d « de l’arabe ou un homme muwallad(a), s’il n’est pas purement arabe ». Historiquement, le terme s’est dit des enfants nés, à la suite des conquêtes islamiques, d’unions mixtes, généralement entre des pères arabes et des mères non arabes. Si donc l’on suit le mouvement sémantique suggéré par LA, le sens de « métis » n’est pas à mettre au départ, mais au contraire à l’arrivée d’un processus d’évolution sémantique . . . Notons que ce dernier sens pourrait aussi s’atteindre par un simple et banal processus de tadmīn, consistant à « faire entrer » dans un mot le sens de toute une collocation, muwallad étant mis pour muwallad min muxālatat al-Arab al-Ajam (« issu/produit du mélange des Arabes et des non-Arabes »).2 Maintenant, en quoi consiste précisément ce processus de « corruption de la langue » ? Celui-ci est décrit au travers d’une anecdote mettant en scène Abū l-Aswad ad-Du’alī et sa fille. La tradition a vu dans ce personnage du Ier/VIIe siècle le « père » de la grammaire arabe. Pourquoi lui plutôt qu’un autre ? D’abord, parce qu’il est d’origine arabe : sa généalogie complète, telle que donnée par les Tabaqāt (21) de Zubaydī, m. 379/989–90, le rattache aux Kināna, tribu

2 C’est un processus fondamental, tant dans le lexique de l’arabe classique (e.g. jihād « guerre sainte » mis pour jihād fī sabīl li-llāh « combat pour Allah », siyāsa « politique », mis pour siyāsa madaniyya (« gouvernement de la cité ») que dans celui de l’arabe moderne (tālib « étudiant » mis pour tālib al-ilm « celui qui cherche le savoir », amīn « secrétaire » mis pour amīn as-sirr « dépositaire du secret »).

les origines de la grammaire arabe


de la région de la Mecque. Il serait né vers 606 de notre ère et aurait séjourné chez les Qušayr d’Arabie centrale (nous verrons ultérieurement l’importance de ces notations). Ensuite, rallié à Alī, il est nommé par ce dernier cadi puis gouverneur de Basra, en 36/656, mais la disparition de Alī en 41/661 ne l’empêche pas, comme le montre notre texte (et d’autres sources) d’entretenir des relations avec le gouverneur Umayyade Ziyād b. Abīhi. Il serait mort vers 69/688. Enfin, il était également poète. Le personnage est donc le prototype de l’arabe sédentarisé, que son origine et son parcours désignent comme un maître et un gardien de la langue. Il est de Basra, dont la tradition ultérieure fera le siège de l’école dominante de la grammaire arabe, et voir dans un Basrien le père de la grammaire arabe n’est rien d’autre qu’une manière de signaler l’ancienneté du travail grammatical dans cette ville et surtout son antériorité par rapport à l’école rivale de Kūfa. Il est consensuel socialement et politiquement : nomades et sédentaires, partisans de Alī comme des Umayyades peuvent s’en réclamer. Venons en maintenant à l’anecdote elle-même. La fille dit à son père mā ašaddu l-h arri. Celui-ci interprète le propos comme une question sur la chaleur la plus intense (« Quelle est la chaleur la plus intense ? ») et répond, par suite, ar-ramdau fī l-hājira, c’est-à-dire « la canicule en plein midi ». Mais la fille rejette cette interprétation lam asalka an hādā « je ne t’ai pas demandé cela », ajoutant « je me suis étonné de l’intensité de la chaleur » (innamā taajjabtu min šiddati l-h arri). « Alors dis, rétorque son père, mā ašadda l-h arra (« Quelle chaleur intense !») ». Autrement dit le père reproche à sa fille d’avoir confondu les structures interrogative (istifhām) et exclamative (taajjub) et d’avoir employé l’une pour l’autre. C’est cela un lah n, autre terme fondamental apparaissant dans le texte : non pas une faute de langage en général, mais une faute contre la flexion désinentielle, casuelle et modale, en particulier. C’est cette flexion qu’on appelle en arabe même irāb. Une telle faute serait d’autant plus grave qu’elle créerait un quiproquo. De cette anecdote, il existe plusieurs versions (comme le reconnaît le texte lui-même) : Versteegh (1997b, 59) cite la version donnée par Sīrāfī (m. 368/979), dans son Axbār (éd. Krenkow, 19) avec les exemples de mā ah sanu as-samāi « What is the most beautiful thing in the sky? » / mā ah sana s-samāa « How beautiful is the sky! ». A travers cette anecdote, il s’agit en fait de mettre au centre de la arabiyya ou « langue des Arabes » le irāb (mot étymologiquement lié à Arab) et de suggérer que cet irāb est pertinent, en ce qu’il distingue des


pierre larcher

significations (muwaddih , mubayyin, mufarriq, munabbi . . . li-maānī al-luġa), pour citer ici quelques-uns des mots que l’on relève dans les sources. Et c’est l’aggravation de la situation qui amène à la constitution de la grammaire, son nom arabe de nahw étant justifié par le fait que le grammairien se fait l’indicateur de la « voie à suivre » (unh ū hādā n-nahw). Nahw est en effet le masdar du verbe nah ā-yanh ū, qu’on emploie toujours comme circonstanciatif, figé à l’accusatif (nahwa), de sens « vers ». Peu importe si cette étymologie est fantaisiste ou non. Ce qui nous intéresse ici, ce sont les différentes interprétations qu’un linguiste est susceptible de faire de ce texte pour l’histoire de la langue.

3. Interprétation 3.1

De l’histoire . . .

Au XIXe siècle, la linguistique, née au tournant du XVIIIe et du XIXe, sous forme de la grammaire comparée (des langues indo-européennes), devient historique. On ne cherche plus seulement à reconstruire en amont des protolangues (Ursprache). Plus modestement, on cherche à retracer en aval l’évolution des langues existantes. La linguistique historique est une spécialité essentiellement allemande. On ne s’étonnera donc pas que ce soient les arabisants allemands qui, les premiers, se sont intéressés à l’histoire de l’arabe. Ils réinterpréteront le fasād al-luġa de la tradition arabe, caractérisée par les lah n ou fautes de irāb, comme le signe d’une évolution d’un type ancien arabe (en allemand Altarabisch et en anglais Old Arabic) vers un type néo-arabe (en allemand Neuarabisch et en anglais New ou Neo-Arabic). Le type ancien arabe est évidemment caractérisé par l’existence d’une flexion désinentielle, casuelle et modale, le type néo-arabe par la disparition de cette flexion. Le type ancien arabe est donc plus synthétique et le type néoarabe plus analytique. Corollairement, dans le type ancien arabe, l’ordre des mots est plus libre, mais, dans le type néo-arabe, moins libre. Quand le type ancien arabe commence à se dégrader en type néo-arabe, nous entrons dans le moyen arabe (en allemand Mittelarabisch et en anglais Middle Arabic). C’est Heinrich Leberecht Fleischer (1801–1888) qui projettera, implicitement dans un article de 1847, explicitement dans un article de 1854, sur l’arabe cette tripartition célèbre en linguistique historique. Voici ce qu’il écrit en 1854 (2–3) :

les origines de la grammaire arabe


Als heilige Sprache des Islam, Organ der Gelehrsamkeit und höhern Wissenschaftlichkeit, Mittelpunkt oder vielmehr ausschliesslicher Gegenstand aller Schulphilologie, steht das Altarabische seinem Abkömmling, dem Neuarabischen, in der Anschauung des Morgenlandes selbst schroff gegenüber. Nur jenes heisst bei den Gelehrten al-luġah, die Sprache, al‘arabiyyah, das Arabische schlechthin, dieses al-lisān al-‘āmm oder alāmmī, die gemeine Mundart, la lingua volgare.

Il est clair, d’après la description même qu’en donne Fleischer, que l’ancien arabe est l’arabe classique et le néo-arabe l’arabe dialectal et non moins clair que si les deux variétés coexistent en synchronie, l’arabe dialectal est explicitement compris comme étant historiquement le « descendant » (Abkömmling) de l’arabe classique. Un peu plus loin (4), il mentionne le moyen arabe. Il l’avait déjà exactement décrit (du point de vue de la linguistique historique) en 1847, à propos de la langue d’un codex gréco-arabe (155), qu’il compare à celle des Mille et une nuits, auxquelles il avait consacré sa dissertatio en 1836 : « Wie in der Tausend und Einen Nachten sind auch hier einzelne jener ältern Formen mit der neuern gleichsam noch im Kampfe begriffen ; willkürlich tritt bald die eine, bald die andere ein ». L’état moyen d’une langue se caractérise en effet par l’alternance, en synchronie, d’éléments interprétables, en diachronie, comme relevant encore de l’état ancien (älter) ou déjà de l’état moderne (neuer). Un de ses élèves, Ignaz Goldziher (1850–1921), dans un écrit de jeunesse rédigé en hongrois et aujourd’hui traduit en anglais, compare explicitement la relation entre ancien arabe et néo-arabe à celle du latin et des langues romanes, appelées jadis néo-latines (Goldziher 1994, 20) : As French abandoned the case inflection of Latin and developed the Roman synthesis into analysis, making de l’homme from hominis, so did the living Arabic of today dissolve the old rajulin into metā r-rajul ; as latin scrip-si developed into French j’ai écrit (. . .), so was Old Arab[ic] aktubu turned into biddi aktub ou bi-aktub.

Cette conception « allemande » de l’histoire de l’arabe se retrouve au XXe siècle, en particulier chez Johann Fück (1894–1974) dans son grand ouvrage Arabīya (Fück, 1955[1950]), et, aujourd’hui encore, chez Joshua Blau, le grand maître du moyen arabe (e.g. Blau 2002, 16). Objectivement, cette conception rejoint celle que s’en faisait au VIIIe/ XIVe siècle Ibn Xaldūn (m. 808/1406) et qu’il expose à deux reprises dans la Muqaddima : une première fois dans la section 22 (675–7), intitulée fī luġāt ahl al-amsār (« Des parlers des habitants des villes ») du chapitre IV, consacré aux « villes et pays », et une seconde aux chapitres


pierre larcher

47 et 48 (1073–1080), consacrés respectivement aux parlers nomades et aux parlers sédentaires de son temps, du chapitre VI, lui-même consacré aux « sciences ». Pour Ibn Xaldūn, ce qu’il appelle « langue de Mudar » (luġat Mudar, al-lisān al-mudarī) est la « langue première » (allisān al-awwal) et « originelle » (al-lisān al-aslī) de l’Arabie et la langue du Coran et du h adīt. Cette langue est parlée de manière d’autant plus châtiée ( fasāh a) que ceux qui la parlent sont plus éloignés des autres nations, « la manière de parler des Qurayš étant pour cette raison la plus châtiée et la plus pure des manières de parler arabes » (wa-li-hādā kānat luġat Qurayš afsah al-luġāt 3 al-arabiyya wa-asrah uhā): si l’on retraduit en termes géographiques la généalogie, l’appellation « langue de Mudar » revient à désigner le centre et l’ouest de l’Arabie comme le domaine de l’arabe fasīh . Ibn Xaldūn fait ici la synthèse de deux thèses, sur lesquelles nous reviendrons ci-dessous : d’une part la thèse philologique (qui voit ce domaine comme constitué de deux sous-domaines dits Hedjaz et Tamīm),4 et d’autre part la thèse théologique (qui identifie, sur la base de Cor. 14:4, la langue du Coran à la luġat Qurayš et, sur une base dogmatique, la luġat Qurayš à la luġa al-fush ā), tout en les croisant avec une thèse philosophique, issue du Kitāb al-h urūf de Fārābī (m. 339/950),5 et qui est le corrollaire de la thèse liant « corruption » et « mélange ». A l’inverse, cette langue est déjà corrompue dès l’époque préislamique, là où les Arabes sont en contact avec d’autres nations, et se corrompt encore davantage après la conquête islamique et les nouvelles fondations urbaines, et donc en milieu sédentaire plus encore que nomade, jusqu’à donner naissance à de nouvelles langues. L’originalité d’Ibn Xaldūn est en effet de ne pas considérer les dialectes comme de simples formes dégradées de la « langue de Mudar », mais comme des variétés autonomes par rapport à celle-ci et distinctes d’elle, en ce qu’elles ont substitué à la syntaxe basée sur la flexion désinentielle une syntaxe de position (at-taqdīm wa-t-taxīr).6


C’est cette expression qui donne, par réécriture, celle de al-luġa al-fush ā. Sur cette subdivision, cf. Rabin (1951). On comprend pourquoi Abū l-Aswad d-Du’alī, natif du Hedjaz, est dit avoir fait un détour par l’Arabie centrale . . . 5 Du moins la version de ce texte connue par le Muzhir (I:211–212) ou, mieux, le Iqtirāh (20) de Suyūtī (m. 911/1505), non celle publiée par Mahdi en 1969. Sur les deux versions de ce texte, cf. Langhade (1994, 248–258) et Larcher (2006a). 6 Sur Ibn Xaldūn et l’histoire de l’arabe, cf. Versteegh (1997a, 153–165) et Larcher (2006b). 4

les origines de la grammaire arabe 3.2


à la sociolinguistique

En citant l’expression italienne de lingua volgare, certes attirée par celle de al-lisān al-āmm(ī),7 Fleischer montre qu’il n’a pas seulement en tête la linguistique historique, mais aussi le modèle italien. Or, dans ce modèle, la « langue vulgaire » n’est évidemment pas désignée en termes diachroniques, c’est-à-dire historiques, comme « descendant » du latin ; elle l’est au contraire, en termes synchroniques et sociolinguistiques, comme langue du vulgum pecus, par opposition au latin « langue des clercs ».8 Mais, là encore, cette conception, qui, pour l’arabe, sera baptisée ultérieurement par les arabisants diglossie,9 rejoint objectivement celle que se font les auteurs de langue arabe, à commencer d’ailleurs par az-Zajjājī lui-même. Au chapitre XVII, intitulé bāb dikr al-fāida fī taallum alnahw, il demande (95) : « à quoi sert d’apprendre la grammaire, la plupart des gens parlant naturellement10 sans flexion désinentielle, qu’ils ne connaissent pas, tout en comprenant les autres et en [se] faisant comprendre d’eux » ( fa-mā al-fāida fī taallum an-nahw wa-aktar an-nās yatakallamūn alā sajiyyatihim bi-ġayr irāb wa-lā marifa minhum bihi fa-yafhamūn wa-yufhimūn ġayrahum mitl dālika).

7 Fleischer ne donne aucune référence pour ces deux expressions. C’est dommage, car si, à l’époque où il écrit (milieu du XIXe siècle), l’expression al-arabiyya est, comme il le note, couramment utilisée, par une métonymie significative, pour désigner l’arabe classique, c’est l’expression de al-luġa ad-dārija qui est utilisée pour désigner l’arabe dialectal. L’expression d’al-luġa al-āmmiyya (vs al-luġa al-fushā) n’apparaîtra que vers la fin du XIXe siècle, du moins comme nom de cette variété, mais dès le Moyen Age, on la rencontre pour désigner un « vulgarisme » au sein de la langue. 8 La comparaison avec la situation italienne ne peut d’ailleurs être poussée trop loin sans aporie. Le domaine arabe n’a pas connu la révolution qu’a connue le domaine roman (et, mutatis mutandis, l’Europe entière), à savoir la promotion des langues « vulgaires » au rang de langues littéraires, ce qui fera du latin (et seulement pour un temps, plus ou moins long selon les pays) le véhicule de la seule culture savante. Ainsi, après Il cantico delle creature (1226) de Saint François d’Assise (1182–1226), Dante Alighieri (1265–1321), logiquement, écrit La Divine Comédie en langue vulgaire, mais traite de celle-ci en latin (De vulgari eloquentia). 9 Ce terme, venu de la linguistique néo-hellénique (1885), a été explicitement introduit en linguistique arabe par William Marçais (1874–1956), dans un article de 1935, avant que le concept ne soit théorisé, à partir de l’arabe et d’autres langues, par Charles A. Ferguson (1921–1998), dans un article de 1959. Pour le détail, cf. Larcher (2003). 10 Blanc (1979, 165, n. 20) traduit par « spontaneously » et Versteegh (1995) par « intuitively ».


pierre larcher

Il revient sur ce point à la fin du chapitre (96) : « Quant aux gens du commun qui parlent l’arabe sans flexion désinentielle, on les comprend. Mais cela est seulement possible pour ce qui est bien connu et d’usage courant, ce dont on a une connaissance familière et est usité. Mais si, d’aventure, l’un d’eux se risquait à éclaircir une ambiguïté, sans le faire comprendre au moyen de la flexion désinentielle, il ne le pourrait pas » ( fa-ammā man takallama min al-āmma bi-l-arabiyya bi-ġayr irāb fa-yufham anhu fa-innamā dālika fī al-mutaāraf al-mašhūr wa-lmustamal al-malūf bi-d-dirāya wa-law iltajaa ah aduhum ilā al-īdāh an manā multabis min ġayr fahmihi bi-l-irāb lam yumkinhu dālika).

Alors que le chapitre XIV concerne le Ier/VIIIe siècle, le chapitre XVII concerne l’époque d’al-Zajjājī lui-même, c’est-à-dire le IVe/Xe siècle. La situation décrite dans ce chapitre semble pouvoir être interprétée comme l’aboutissement du processus décrit au chapitre XIV. Au Ier/VIIIe siècle le type ancien arabe commence à se dégrader en type néo-arabe. Trois siècles plus tard, cette dégradation a abouti, non à une substitution d’un type à l’autre, mais, le type ancien arabe subsistant, à une coexistence des deux, chaque type étant caractérisé non seulement linguistiquement par la présence/absence du irāb, mais encore socialement et culturellement : dire que le type non fléchi est l’expression « naturelle » de « la plupart des gens » revient à dire que le type fléchi est celle non seulement d’une minorité, mais encore l’expression artificielle de cette minorité (artificielle, puisque scolairement apprise et non naturellement acquise et pratiquée !). Qualifier explicitement la majorité de āmma revient non seulement à qualifier implicitement la minorité de xāssa , mais encore semble préfigurer l’appellation de al-luġa al-āmmiyya, aujourd’hui utilisée pour désigner l’arabe dialectal. Enfin reconnaître par deux fois que l’absence de flexion désinentielle ne nuit en rien à la communication explique que, pour sauver cette flexion (son gagne pain !), le grammairien distingue entre communication quotidienne et communication savante et cherche, voire fabrique, des structures ambiguës, que seule la flexion, censément pertinente, est à même de désambiguïser (on voit tout de suite que l’exemple proposé au chapitre XIV relève exactement de ce cas de figure et, comme tout arabisant en a fait l’expérience, ce petit jeu perdure, jusqu’à aujourd’hui, dans certains milieux « puristes » !). La vérité oblige cependant à dire qu’il subsiste un doute sur cette interprétation. On doit se rappeler ce que dit al-Jāhiz (m. 255/869) dans le Bayān (I:137). Il n’oppose nullement āmma à xāssa comme la masse illettrée à l’élite lettrée, mais seulement, parmi les gens lettrés, les gens ordinaires aux happy few. C’est exactement le sens que les spécialistes

les origines de la grammaire arabe


donnent à āmma dans la littérature spécialisée des ouvrages de lah n alāmma.11 Il n’y a rien là que de logique : lah n présuppose irāb ; par suite, s’il n’y a pas irāb, il n’y a pas lah n ! Si l’on projette le sens « jāhizien » de āmma sur le texte d’az-Zajjājī, il ne s’agirait plus alors de deux variétés d’arabe, mais seulement de deux registres d’une même variété : l’un « soutenu » (avec réalisation du irāb) et l’autre « relâché » (sans réalisation de ce irāb). Jusqu’à aujourd’hui l’école enseigne le irāb qui « terrorise » (irhāb) les apprenants, parce que sa réalisation est source de lah n, d’où le sage conseil ijzim taslam « Supprime la voyelle brève finale, tu seras préservé de l’erreur ! ». Un seul élément du texte d’az-Zajjājī peut faire pencher la balance en faveur de la première interprétation, plutôt que de la seconde : c’est la notation que les locuteurs de l’arabe sans flexion désinentielle « n’ont pas connaissance de celle-ci ». La āmma, au sens d’al-Jāhiz, ne l’ignore pas ; elle en a seulement une connaissance imparfaite12 . . .

4. Discussion On le voit : tout tourne autour du irāb, tant dans le texte d’az-Zajjājī que dans l’interprétation, tout à la fois historique et sociolinguistique, qui en est faite. Pourtant, dès avant le XIXe siècle et l’essor de la linguistique historique, une tout autre tendance était apparue chez certains arabisants. Dès le XVIIIe siècle, des arabisants, également allemands d’ailleurs, comme Johann Davis Michaelis (1717–1791) et Johann Gottfried Hasse (1759–1806), avaient développé ce que Gruntfest (1991), qui l’a étudiée, appelle « A Early theory of Redundancy of Arabic Case Endings ». Notons qu’ils réinterprétaient des idées déjà exprimées, au XVIIe siècle, par l’Italien Antonius ab Aquila (Antonio dell’Aquila), Franciscain envoyé en mission auprès des Chrétiens d’Alep et auteur de la première

11 On lira avec profit l’article Lah n al-āmma, de EI2, dû à Charles Pellat (1914– 1992). 12 On notera que dans l’unique manuscrit, daté de 617H, qui sert de base à l’édition du Īdāh , il y a dans l’avant-dernier paragraphe de ce chapitre XVII, une magnifique faute de irāb : lan yumkin ah ad [corrigé par l’éditeur en ah adan] min al-muwalladīn iqāmat-hu illā bi-marifat an-nahw (« Nul, parmi les muwalladīn, ne pourrait l’établir (la poésie), sauf par la connaissance de la grammaire »). L’absence du alif révèle au minimum qu’il n’est pas prononcé, voire suggère que ah ad a été traité comme le sujet de yumkin et par suite iqāma comme l’objet, ce que peut également suggérer le masculin yumkin.


pierre larcher

grammaire de la « langue (arabe) vulgaire » (1650).13 Notons également que tous les arabisants du temps savaient le latin et le grec : quand on a souffert dans sa jeunesse sur une version grecque ou latine, on voit tout de suite la différence entre grec et latin classiques et arabe classique. En grec et en latin, c’est la déclinaison qui permet de construire la phrase ; en arabe, c’est la construction de la phrase qui permet de restituer la déclinaison . . . Les grammairiens arabes le reconnaissent eux-mêmes implicitement, en couplant, dans leur théorie, le irāb au amal : le irāb est l’effet de l’ « action » d’un élément dit āmil sur un autre dit mamūl, ainsi lan, dit nāsib, qui détermine le nasb (subjonctif ) du mudāri (inaccompli) dit mansūb. Par ailleurs, le fait même que les grammairiens arabes cherchent des cas où cette flexion serait pertinente suffit à prouver qu’elle ne l’est pas ! On voit tout de suite le caractère artificiel de l’exemple proposé par az-Zajjājī. Dans beaucoup de langues, structures interrogatives et exclamatives sont confondues, étant distinguées par l’intonation, ainsi en français ou en allemand Quel artiste/Welcher Künstler ? vs Quel artiste/Welch ein Künstler ! Les exemples donnés de l’ « aggravation de la situation », qui font défaut chez az-Zajjājī lui-même, mais qu’on trouve chez d’autres auteurs, ne sont guère plus convaincants. Ainsi le fameux tuwuffiya abānā wataraka banūna (« notre père est mort et a laissé des fils ») est si peu probant pour un fasād al-luġa interprété en termes historiques que Versteegh (1997b, 51), qui le cite d’après le Nuzha d’Ibn al-Anbārī (m. 577/1181), l’interprète en termes sociolinguistiques (alors même que l’anecdote concerne le temps de Ziyād, donc le Ier/VIIIe siècle). En termes historiques, on se serait en effet attendu à l’emploi des formes néo-arabes abū et banīn (respectivement nominatif de la flexion triptote abū/ā/ī et cas régime de la flexion diptote banū/ī(na) du type ancien arabe), de sorte qu’avec la phrase donnée en exemple, il n’y aurait pas eu de « faute » ! En termes sociolinguistiques, en revanche, la coexistence de deux types dont l’un a plus de prestige que l’autre peut amener le locuteur, dans une circonstance formelle, à substituer à l’unique forme du type néo-arabe, qu’il emploie ordinairement, mais pense fautive, l’autre forme du type 13 Cf. Fück 1955, 78. Fück considère que la Fabrica overo Dittioniario della lingua volgare arabica et italiana (1636) de Dominicus Germanus de Silesia (1588–1670) n’est pas, malgré son titre, un dictionnaire, mais une introduction, presque sans valeur, à l’arabe vulgaire. En tout cas, on voit que ce sont des clercs, Italiens ou liés à l’Italie, qui, sous l’appellation, valant signature, de « langue vulgaire », sont les « inventeurs » de l’arabe dialectal. Cette double qualité, jointe au fait qu’ils étaient des hommes de terrain et non de cabinet, les y prédisposait.

les origines de la grammaire arabe


ancien arabe, qu’il n’emploie pas et pense correcte : pour éviter une faute, il en commet une autre ! Certes, ces arabisants voyaient dans les désinences casuelles une invention des grammairiens arabes, position encore défendue au XIXe siècle par Aloïs Sprenger (1813–1893). Si personne aujourd’hui ne croit plus cela, en revanche la redondance de la flexion en arabe classique amène beaucoup d’arabisants à douter que l’histoire de l’arabe consiste en une évolution du type ancien arabe vers le type néo-arabe, caractérisés respectivement par la présence et l’absence de cette flexion. Deux positions se sont fait jour. L’une est très répandue chez les arabisants. L’akkadien possédant une flexion casuelle triptote et d’autres langues sémitiques exhibant des restes de flexion casuelle, les arabisants admettent que la flexion casuelle de l’arabe classique est un trait de haute antiquité, qui s’est maintenu dans le seul registre poétique de la langue. Sous l’influence de la représentation diglossique de l’arabe, rétoprojetée sur l’histoire de la langue, ce registre poétique est souvent vu comme une langue commune (koinè), véhiculaire, opposé aux vernaculaires que sont les anciens dialectes arabes. Cette koinè serait également la langue du Coran, à quelques « hedjazismes » près. Malgré les apparences, cette hypothèse arabisante n’est pas si différente de la thèse théologique ! Celle-ci, on l’a vu, se résume en une double identification : l’une, opérée sur une base scripturaire, de la langue du Coran avec la « langue de Qurayš » et l’autre, opérée sur une base purement dogmatique, de la langue de Qurayš avec al-luġa al-fush ā. Il y a un siècle, Vollers (1906) acceptait la première identification, mais refusait la seconde. Il supposait en effet que le Coran avait d’abord été énoncé et écrit dans le vernaculaire de la Mecque, parler ouest-arabique dépourvu de irāb, avant d’être réécrit dans la langue véhiculaire de la poésie, parler est-arabique pourvu de ce irāb. Kahle (1959[1947]) reprit cette hypothèse en l’atténuant : si le ductus coranique (rasm) reflète le vernaculaire de la Mecque, les qirāāt, elles, reflètent la langue véhiculaire de la poésie. La plupart des arabisants (cf. Blachère 1952, 66–82) refusent la première identification et, par suite, la seconde, mais acceptent la troisième, découlant de la suppression du moyen terme, i.e. langue du Coran = al-luġa al-fush ā. Or, si l’on relit le célèbre texte du Sāh ibī (52–53) d’Ibn Fāris (m. 395/1004), dont la source n’est autre que celui attribué à al-Farrā’ (m. 207/822) et jadis exhumé par Kahle (1959[1947]), on s’aperçoit que pour concilier vérité théologique et vérité philologique (i.e. le fait que la langue du Coran, identifiée à la luġat Qurayš, exhibe des traits, les fameux « hedjazismes », qui ne sont pas ceux de la luġa al-fush ā) ces auteurs imaginaient un


pierre larcher

scénario expliquant comment la luġat Qurayš était devenue la luġa al-fush ā : comme produit d’un processus de koinéïsation justifié par le fait que, la Mecque étant un centre de pèlerinage intertribal, les Qurayš avaient ainsi pu « sélectionner » (taxayyur) le meilleur de chaque parler arabe !14 Mais une autre position, originale et marginale, est en train de gagner du terrain. La poésie archaïque pratique la qāfiya mutlaqa (« rime absolue »), c’est-à-dire réalise les voyelles brèves u, a et i, avec ou sans tanwīn, uniformément comme des voyelles longues ū, ā, ī. C’est dire si le tanwīn ne sert à rien !15 La récitation psalmodiée du Coran (tajwīd) va plus loin : elle pratique généralement la qāfiya muqayyada (« rime liée »), c’est-àdire supprime les voyelles brèves u, a et i en finale. Elle supprime également -un et -in, et réalise -an comme -ā. Cette prononciation pausale de -an en -ā est la seule trace de flexion casuelle qui apparaisse dans le maigre matériel épigraphique conservé16 et c’est aussi sa seule manifestation dans les dialectes modernes (marh aban ou ahlan wa-sahlan prononcés marh abā et ahlā w-sahlā). Le cas du suffixe relateur -(V)n qu’on trouve dans maint dialecte ancien et moderne est à disjoindre, dans la mesure où sa fonction (marquer la relation mawsūf/sifa, celle-ci pouvant être un N, un SP ou une phrase) ne le relie en rien au tanwīn de l’arabe classique et amène à penser que c’est la conservation d’un trait archaïque (cf. Ferrando 2000). Qu’on doive supprimer les voyelles brèves en finale (y compris donc les voyelles flexionnelles) est évidemment un argument en faveur de la non pertinence de cette flexion sur le plan syntaxique.17 En revanche, qu’en dehors de la pause, elle soit réalisée, suggère que cette flexion n’est pas syntaxique, mais prosodique. L’idée était déjà émise au XIXe siècle par Johann Gottfried Wetzstein (1815–1905), dans un article paru en 1868. Les historiens de la grammaire arabe savent aujourd’hui, grâce à Versteegh (1981[1983]), qu’il s’est trouvé au moins un grammairien arabe, Qutrub (m. 206/821), pour anticiper cette idée. Nous connaissons sa théorie grâce à az-Zajjājī (ch. VII, 69–71, bāb al-qawl fī al-irāb


Sur ces deux textes, cf. Larcher 2004 et 2005b. Comme le suggère en outre la réalisation du ā long résultant de la qāfiya mutlaqa comme . . . -an (tanwīn at-tarannum) attribuée aux Tamīm, dans la récitation poétique (inšād). Dans les deux cas, on peut donc avoir l’article et voir et/ou entendre ā/an. 16 Un exemple dans l’inscription d’Umm al-Jimāl (Ve ou VIe siècle ap. JC ?), avec un mot lu successivement par Littmann en 1929 et 1949 comme ġiyāran et ġafran. Nous parlons bien sûr ici de la seule flexion marquée par des voyelles brèves. 17 Cf. l’analyse qui est proposée de Cor. 85 :21–22 dans Larcher (2005a). 15

les origines de la grammaire arabe


limā daxala fī-kalām « pourquoi la flexion s’est-elle introduite dans le discours ? »), qui explique que pour Qutrub les voyelles de flexion sont liées à l’enchaînement (wasl) des mots dans la phrase. Des idées, sinon identiques, du moins comparables sont défendues aujourd’hui par Owens (1998). Un bon argument en faveur de la conception prosodique et non syntaxique de la flexion peut être trouvé dans le traitement contradictoire de Tamūd dans le Coran. Contradictoire, dans la mesure où il est traité partout comme un diptote sans tanwīn Tamūd-u/a, sauf en 11 : 68 ; 25 :38 ; 29 :38 ; 53 :51, où apparaît un alif. Alors que H afs ‘an ‘Asim (Coran du Caire) « neutralise » ce alif, Warš ‘an Nāfi‘ (Coran du Maghreb) le traite bien, en ces endroits, comme un triptote avec tanwīn et lit Tamūdan. Pourquoi, donc, deux déclinaisons pour un même mot, alors qu’aucune des deux ne sert à rien sur le plan syntaxique (en 17 :59, nous avons un Tamūd complément d’objet sans alif) ? On a tôt fait d’observer que dans trois des quatre cas où apparaît ce alif, Tamūd est coordonné à Ād, également muni de ce alif (et partout traité lui comme un triptote Ād-u/ a/i-n). L’apparition du alif peut s’expliquer par un simple et banal ajustement forme/sens : c’est la coordination syntaxique de deux mots qui, sémantiquement, vont ensemble, comme noms de peuples « reprouvés », qui explique qu’on donne au second le traitement syntaxique du premier. L’explication ne vaut cependant pas pour le quatrième cas (11 :68), où Tamūd apparaît seul : (ka-al-lam yaġnaw fīhā) a-lā inna Tamūda (H afs)/an (Warš) kafarū rabbahum a-lā budal-li-Tamūd[a]. Sauf à lire ce alif comme il est écrit, c’est-à-dire non comme -an, mais bien comme -ā, auquel cas il forme aussitôt assonance avec les nombreux -ā de son environnement syntaxique immédiat, que nous avons mis en gras . . . On peut aller plus loin. Le début du verset, que nous mettons entre parenthèses, va avec ce qui précède. Le reste s’organise en un parallélisme marqué par l’anaphore de a-lā. On voit que si on lit, non budan, d’où, par assimilation (idġām) du nūn au lām budal-li-, mais budā, on n’a plus seulement alors une assonance en ā, mais encore une rime interne en -dā, soit : a-lā inna Tamūdā kafarū rabbahum a-lā budā li-Tamūd « Holà ! Oui, les Thamoud, ils ont été infidèles à leur seigneur ! Holà ! Arrière aux Thamoud ! ».18 18 Tamūd est loin d’être un cas unique. On peut également citer salāsilā de 76 :4 et qawārīrā de 76 :15–16. Par ailleurs on a compris que, pour ma part, je ne considère pas -ā comme la prononciation pausale de -an, mais -an comme une réinterprétation ultérieure d’un -ā « fondamental ». L’inscription du Jabal Usays (528–9 ap. JC) montre


pierre larcher

On ne perdra pas de vue que l’histoire de l’arabe est toujours dépendante des idées du moment. Ainsi Versteegh (1984), surfant sur la vague des études créoles, a proposé de réinterpréter le fasād al-luġa dans les termes d’un processus de pidginisation-créolisation. La conquête islamique a pour effet de mettre en contact des populations arabophones et des populations non arabophones. On peut donc imaginer, pour les besoins de la communication, l’émergence de langues de contact hétérogènes, ce qu’on appelait jadis des sabirs et qu’on appelle aujourd’hui des pidgins (Fück 1955 [1950], 8, en fait déjà l’hypothèse). Si les enfants qui naissent d’unions mixtes (les muwalladūn) la prennent pour langue maternelle, cette langue de contact devient un créole. Et, enfin, si au contact de l’arabe, le créole se réarabise (ou se décréolise), ce pidgincréole décréolisé peut être vu comme le point de départ des dialectes arabes, sur le modèle du Juba-Arabic au Sud Soudan.19 L’hypothèse de Versteegh suscita un vaste débat et fut généralement rejetée (cf. Holes 1995, 19–24). Il me semble néanmoins qu’on ne devrait pas jeter le bébé avec l’eau du bain !20 On ne peut exclure a priori que des processus de ce genre aient joué, localement, un rôle. Pour ma part, je me contenterai de signaler et souligner ici une phrase de la Muqaddima d’Ibn Xaldūn qui m’a fait parler, pour tenter une comparaison éloquente, de « case de l’Oncle Tom sur les bords du Tigre et de l’Euphrate ».21 Traitant de la genèse des parlers sédentaires dans les différentes régions du monde musulman (Maghreb, Machrek, Andalousie), il écrit à propos du second (1077–1078) : « De même, quand les Arabes eurent vaincu les nations de l’Orient, Persans et Turcs, et se furent mélangés à eux, les langues de ceux-ci se diffusèrent parmi eux, par l’intermédiaire des laboureurs, des paysans et des captifs qu’ils prenaient comme intendants, sages-femmes, pères nourriciers et nourrices : leur langue se corrompit du fait de la corruption de

qu’avant l’invention du tā marbūta les deux réalisations d’un même morphème se traduisent par deux graphies : -h à la pause, mais -t en liaison (il reste des traces de cet état de choses dans le ductus coranique, avec des exemples de rah mat ou nimat). On voit d’autant moins pourquoi les deux réalisations phoniques du tanwīnan de l’arabe classique auraient donné lieu dans le matériel préclassique à une graphie unique (correspondant à -ā) que, dans les textes du moyen arabe, le suffixe relateur -(V)n se traduit dans la graphie par un n. 19 Ma collègue Catherine Miller, consultée, m’indique qu’à l’heure actuelle le JubaArabic est encore loin d’être un dialecte arabe. En outre les choses vont un peu dans tous les sens, manifestant des tendances contradictoires. 20 Cf. d’ailleurs, Versteegh lui-même (2004). 21 Cf. Larcher (2006b).

les origines de la grammaire arabe


l’habitus, au point de muter en une autre langue » (wa-kadā al-Mašriq lammā al-Arab ġalaba alā umamihi min Fāris wa-t-Turk fa-xālatahum wa-tadāwalat baynahum luġātuhum fī l-akara wa-l-fallāh īn wa-s-subiyy alladīna ittaxadūhum xawalan wa-dāyātin wa-adh āran wa-marādi fa-fasudat luġatuhum bi-fasād al-malaka h attā inqalabat luġa uxrā).

Il serait évidemment intéressant de remonter ici aux sources mêmes d’Ibn Xaldūn. Bien sûr, la notation va dans le sens de l’argumentation d’Ibn Xaldūn (interruption du processus « naturel » d’apprentissage de la langue de Mudar, par transmission d’une génération à l’autre d’Arabes, qui seule crée l’ « habitus »). Mais même le plus « intégriste » des créolistes ne peut nier que la phrase soulignée décrit ce que ce genre de créolistes considère être la base objective, sur le plan social, d’un processus de créolisation : une population servile, parlant une ou d’autres langues que celle de ses maîtres. En outre deux, voire trois, des quatre catégories mentionnées renvoient à l’étymologie même de créole (< lat. criare « nourrir, élever »), terme qui peut désigner aussi bien les enfants blancs que les esclaves noirs et leurs langues (généralement le parler des Noirs, mais au moins en un cas, celui de Saint-Barthélemy dans les Antilles, celui des Blancs). La linguistique arabe est aujourd’hui sous l’influence de la sociolinguistique américaine « variationniste », issue des travaux de William Labov (né en 1927). Cela a amené au dépassement de la représentation diglossique de l’arabe. Mais cela a aussi amené, en matière d’histoire de la langue, à renouer avec la vision des plus anciens grammairiens arabes, notamment Sībawayhi (m. 177/793 ?) et al-Farrā’. Ceux-ci concevaient l’arabe tout à la fois comme une langue une et plurielle, luġa faite de luġāt, c’est-à-dire non pas de variétés autonomes, mais de variantes, tribales ou régionales, bonnes ou mauvaises (cette hiérarchie suffit bien sûr pour ne pas en faire des « variationnistes » avant l’heure !). Aujourd’hui, l’arabe dit classique est généralement compris, non comme un état de la langue, mais comme une langue standardisée, c’est-à-dire retenant certaines luġāt et en éliminant d’autres. Elle est un aboutissement, non un point de départ, et, par suite, les dialectes arabes modernes ne sauraient en être des « corruptions ». Dans la mesure où beaucoup des traits décrits par les grammairiens arabes (notamment les plus anciens d’entre eux : Sībawayhi et al-Farrā) se retrouvent, identiques ou analogues, dans les parlers arabes d’aujourd’hui, il n’y a pas de raison de penser qu’ils ne prolongent pas les anciens parlers arabes, selon des modalités complexes en fonction des lieux et des temps (c’est le travail de la dialectologie historique


pierre larcher

que de débrouiller l’écheveau). Simplement, même si les parlers arabes d’aujourd’hui relèvent uniformément de ce qui est pour la linguistique historique le type néo-arabe, on se gardera bien de croire qu’ils descendent uniformément d’un type ancien arabe. L’opposition des deux types, compte tenu de la non-fonctionnalité de la flexion désinentielle en arabe classique même, apparaît très largement outrée et, par suite, ils ont pu très bien coexister à date ancienne et, non pas en opposition, mais en continuité . . . Un argument en ce sens peut d’ailleurs être trouvé dans la « faute » relevée dans l’un des deux plus vieux papyrus datés (22 de l’Hégire), où l’on a Abū Qīr pour Abī Qīr (cf. Diem 1984, 271). Abū Qīr n’est pas un nom arabe, mais l’arabisation d’un nom grec, qui est Apa Kyros. Pour parvenir à Abū Qīr, il faut donc bien passer par Abā Qīr, réinterprété comme l’accusatif d’une flexion triptote Abū/ā/ī. Cela n’empêche pourtant pas le scripteur d’utiliser le « nominatif » Abū Qīr à une place réclamant le « génitif » ! Le papyrus étant contemporain de la conquête de l’Egypte (celle-ci, commencée en 18 de l’Hégire, s’achève en 25 avec l’occupation définitive d’Alexandrie) et, plus précisément de la fondation de Fustā t  (22H), il paraît difficile, pour ne pas dire impossible, de voir dans ce fait le signe d’une évolution d’un type à l’autre, liée au « métissage » des populations dans les centres urbains nouvellement créés à la suite de la conquête ! Et si on a donc bien ici, en même temps, référence implicite au type ancien arabe et utilisation explicite du type néo-arabe, il vaut mieux y voir la continuation d’une situation de langue « plurielle », non encore standardisée. Un autre argument dans le même sens peut être trouvé dans le Coran même, avec les incertitudes de la flexion visible, casuelle ou modale (pour des exemples et une analyse, cf. Larcher 2005a).

5. Conclusion Le récit traditionnel des origines de la grammaire arabe, née de la « corruption de la langue », a l’apparence d’un récit historique. Les orientalistes du XIXe siècle, siècle d’histoire et d’esprit critique, s’y sont laissé prendre, même si, en 1906, Karl Vollers jeta un fameux pavé dans la mare ! Cent ans après, les linguistes arabisants voient dans l’arabe dit classique, comme dans toute autre langue classique, une construction. Ils verront alors ce récit comme une reconstruction, ayant pour seul but la légitimation de cette construction.

les origines de la grammaire arabe


6. Références 6.1

Sources primaires

al-Fārābī, Abū Nasr. 1969. Kitāb al-h urūf (Alfarabi’s Book of Letters), éd. Muhsin Mahdi, coll. Recherches publiées sous la direction de l’Institut de Lettres Orientales de Beyrouth n°46, Première série : Pensée arabe et musulmane. Beyrouth : Dar el-Machreq. al-Jāhiz, Bayān = Abū Utmān Amr b. Bahr al-Jāhiz, Kitāb al-bayān wa-t-tabyīn, éd. Abd as-Salām Muhammad Hārūn, 4 parties en 2 vols, Le Caire, 1367/1948. Ibn Fāris, Sāh ibī = Abū l-H usayn Ahmad Ibn Fāris as-Sāh ibī fī fiqh al-luġa wa-sunan al-arab fī kalāmihā, éd. Moustafa El-Chouémi. Coll. Bibliotheca philologica arabica, publiée sous la direction de R. Blachère et J. Abdel-Nour, vol. 1. Beyrouth : A. Badran & Co. 1383/1964. Ibn Manzūr, Lisān al-Arab = Muhammad b. Mukarram b. Alī b. Ahmad al-Ansārī al-Ifrīqī al-Misrī Jamāl ad-Dīn Abū l-Fadl Ibn Manzūr. Lisān al-Arab al-muh īt. Ed. par Yūsuf Xayyāt, 4 vols. Beyrouth: Dār Lisān al-Arab. s.d. Ibn Xaldūn, Muqaddima = Walī d-Dīn Abd ar-Rahmān b. Muhammad Ibn Xaldūn. al-Muqaddima, t. I du Kitāb al-ibar. Beyrouth: Maktabat al-madrasa et Dār al-kitāb al-lubnānī, 1967. Suyūtī , Iqtirāh = Jalāl ad-Dīn Abd ar-Rahmān Abū Bakr as-Suyūtī, Kitāb al-Iqtirāh fī ilm usūl an-nahw, H aydarābād, 1359 H. [reimp. Alep : Dār al-Maārif, s.d.]. Suyūtī , Muzhir = Abd ar-Rahmān Jalāl ad-Dīn as-Suyūtī al-Muzhir fī ulūm al-luġa wa-anwāihā, éd. Muhammad Ahmad Jār al-Mawlā, Alī Muhammad al-Bajāwī et Muhammad Abū l-Fadl Ibrāhīm, 2 vols. Le Caire : Īsā l-Bābī l-H alabī. s.d. az-Zajjājī, Abū l-Qāsim. 1973. al-Īdāh fī ilal an-nahw, éd. Māzin Mubārak, 2ème edition, Beyrouth, Dār an-Nafāis. ——. 1957. Al-Jumal. Précis de grammaire arabe publié avec une introduction et un index par Mohamed ben Cheneb, coll. Etudes arabes et islamiques, Première série : Manuels. Paris : Klincksieck. Zubaydī, Tabaqāt = Abū Bakr Muhammad b. al-H asan az-Zubaydī al-Andalūsī, Tabaqāt an-nahwiyyīn wa-l-luġawiyyīn, éd. M. Abū l-Fadl Ibrāhīm, coll. D axāir al-Arab 50. Le Caire, Dār al-Maārif, 1984. 6.2 Sources secondaires Blachère, Régis. 1952–1964–1966. Histoire de la littérature arabe des origines à la fin du XV e siècle de J.C., I, II et III. Paris : Adrien-Maisonneuve. Blanc, Haïm. 1979. “Diachronic and Synchronic Ordering in Medieval Arab Grammatical Theory.” in Studia Orientalia Memoriae D.H. Baneth Dedicata. Jerusalem: The Magnes Press, The Hebrew University. 155–180. Blau, Joshua. 2002. A Handbook of Early Middle Arabic, The Max Schloessinger Memorial Studies Monographs 6, Institute of Asian and African Studies, Faculty of Humanities, The Max Schloessinger Memorial Foundation and The Hebrew University, Jerusalem. Diem, Werner. 1984. “Philologisches zu den arabischen Aphrodito-Papyri.” in: Der Islam 61, 251–275. EI2 = Encyclopédie de lIslam, nouvelle édition, Leiden, Brill, 1960–. Ferrando, Ignacio. 2000. « Le morphème de liaison /an/ en arabe andalou : Notes de dialectologie comparée », in : Oriente moderno, 80:1, 25–46. Ferguson, Charles A. 1959. “Diglossia.” Word, 15:2, 325–340. Fleischer, H. 1847. “Ueber einen griechisch-arabischen Codex rescriptus der Leipziger Universitäts-Bibliothek.” ZDMG, I, 148–160 [repris dans Kleinere Schriften, 1885– 1888 t. III, ch. XXII, 378–388].


pierre larcher

Fleischer, Heinrich. 1885–88. “Ueber arabische Lexicographie und Ta‘ālibī’s Fikh alluġah.” Berichte über die Verhandlungen der Königlich Sächs. Gesellschaft der Wissenschaften. Philol.-histor. Cl. 1–14 [repris dans Kleinere Schriften, 1885–1888, t. III, ch. IX, 152–166]. Fück, Johann. 1955 [1950]. Arabīya. Recherches sur l’histoire de la langue et du style arabe, Paris, Didier [tr. fr. de Arabīya. Untersuchungen zur arabischen Sprach- und Stilgeschichte, Abhandlungen der sächsischen Akademie der Wissenschaften zu Leipzig, Philologisch-historische Klasse. Band 45, Heft 1, Berlin, Akademie Verlag, 1950]. ——. 1955. Die arabischen Studien in Europa bis in den Anfang des 20. Jahrhunderts. Leipzig: Otto Harrassowitz. Goldziher, Ignaz. 1994. On the History of Grammar among the Arabs. An Essay in Literary History, translated and edited by K. Dévényi et T. Iványi, Studies in the History of the Language Sciences, 73, Amsterdam: Benjamins, 1994 [traduction anglaise de “A nyelvtudomány története az araboknál”, Nyelvtudományi Közlemények 14, 307–375, 1878]. Gruntfest, Y. 1991. “From the History of Semitic Linguistics in Europe: an Early Theory of Redundancy of Arabic Case-endings.” in: K. Dévényi et T. Iványi, eds. Proceedings of the Colloquium on Arabic Grammar, The Arabist. Budapest Studies in Arabic 3–4. Budapest. 195–200. Holes, Clive. 1995 [2004]. Modern Arabic. Structures, Functions, and Varieties. London: Longman [Revised Edition, Washington D.C.: Georgetown University Press]. Kahle, Paul. 1947 [1959]. The Cairo Geniza, First Edition: 1947, Second Edition: 1959. Oxford: Basil Blackwell. Kazimirski, A. de Biberstein. 1846–7. Dictionnaire arabe français, 2 vols. Paris : Théophile Barrois. Langhade, Jacques. 1994. Du Coran à la philosophie. La langue arabe et la formation du vocabulaire philosophique de Farabi. Préface de Jean Jolivet. Damas : IFEAD. Larcher, Pierre. 2003. « Diglossie arabisante et fush ā vs āmmiyya arabes : essai d’histoire parallèle ». Auroux, Sylvain et al. eds. History of Linguistics 1999. Selected Papers from the Eighth International Conference on the History of the Language Sciences (ICHoLS VIII), Fontenay-St.Cloud, France, coll. SIHoLS 99. Amsterdam: Benjamins. 47–61. ——. 2004. « Théologie et philologie dans l’islam médiéval : relecture d’un texte célèbre de Ibn Fâris (Xe siècle) », dans Le discours sur la langue sous les régimes autoritaires, Cahiers de l’ILSL, n° 17. Université de Lausanne. 101–114. ——. 2005a. « Arabe préislamique, arabe coranique, arabe classique : un continuum ? », dans Karl-Heinz Ohlig et Gerd-Rüdiger Puin (Hrsg) : Die dunklen Anfänge. Neue Forschungen zur Entstehung und frühen Geschichte des Islam. Berlin : Verlag Hans Schiler. 248–265. ——. 2005b. ‘D’Ibn Fāris à al-Farrā’ ou un retour aux sources sur la luġa al-fush ā’, Asiatische Studien/Etudes asiatiques, LIX, 3, 797–814. ——. 2006a. « Un texte d’al-Fārābī sur la ‘langue arabe’ réécrit ? » Lutz Edzard and Janet Watson (eds.). Grammar as a Window onto Arabic Humanism. A Collection of Articles in honour of Michael G. Carter. Wiesbaden: Harrassowitz. 108–129. ——. 2006b. « Sociolinguistique et histoire de l’arabe selon la Muqaddima d’Ibn Xaldūn (VIIIe/XIVe siècle) ». Pier Giorgio Borbone, Alessandro Mengozzi e Mauro Tosco (eds.), Loquentes Linguis Studi linguistici e orientali in onore di F.A. Pennacchietti / Linguistic and Oriental Studies to Honour F.A. Pennacchietti / Lingvistika kaj orientaj studoj honore al Fabrizio A. Pennacchietti. Wiesbaden : Harrassowitz. 425–435. Littmann, Enno. 1929. « Die vorislamisch-arabische Inschrift aus Umm ij-Jimāl », ZS, VII, 197–204. ——. 1949. Arabic Inscriptions (Syria : Publications of the Princeton University Archaeological Expedition to Syria in 1904–5 and 1909. Division IV, Section D). Leiden: E. J. Brill. Marçais, William. 1930. « La diglossie arabe », in L’Enseignement public—Revue pédagogique, tome 104 n°12 (1930), 401–409.

les origines de la grammaire arabe


Owens, Jonathan. 1998. “Case and proto-Arabic.” Bulletin of the School of Oriental and African Studies, I:61/1, 51–73 et II:61/2, 215–227. Rabin, Chaïm. 1951. Ancient West-Arabian. London: Taylor’s Foreign Press. Versteegh, C. H. M. 1977. Greek Elements in Arabic Linguistic Thinking. Leiden: E. J. Brill. Versteegh, Kees. 1981 [1983]. “A dissenting grammarian: Qutr ub on declension.” Historiographia Linguistica 8:2–3, 403–429 [repris dans Cornelis H.M. Versteegh, Konrad Koerner and Hans-J. Niederehe, eds. The History of Linguistics in the Near East, Studies in the History of Linguistics, 28. Amsterdam: Benjamins. 1983.167–193]. ——. 1984. Pidginization and Creolization: The Case of Arabic. Amsterdam Studies in the Theory and History of Linguistic Science, Series IV-Current Issues in Linguistic Theory. Amsterdam: Benjamins. ——. 1995. The Explanation of Linguistic Causes, Az-Zajjājī’s Theory of Grammar. Introduction, Translation, Commentary. Studies in the History of the Language Sciences 75. Amsterdam: John Benjamins. ——. 1997a. Landmarks in Linguistic Thought III. The Arabic Linguistic Tradition. Coll. History of Linguistic Thought. London: Routledge. ——. 1997b [2001]. The Arabic Language. Edinburgh: University Press [2ème éd. 2001]. ——. 2004. “Pidginization and Creolization revisited: The Case of Arabic.” dans Martine Haak, Rudolf de Jong and Kees Versteegh, eds. Approaches to Arabic Dialects. A Collection of Articles presented to Manfred Woidich on the Occasion of his Sixtieth Birthday. Leiden: E. J. Brill. 343–357. Wetzstein, Johann Gottfried. 1868. “Sprachliches aus den Zeltlagern der syrischen Wüste.” Zeitschrift der Deutschen Morgenländischen Gesellschaft 22. 69–194.

7. Traduction du texte d’al-Zajjājī Pourquoi cette sorte de science a été appelée nahw Si l’on demande pour quelle raison cette sorte de science a été appelée nahw et s’est vue attribuer ce nom, on répondra : la raison en est l’histoire rapportée au sujet de Abū l-Aswad ad-Dualī. Lorsqu’il entendit parler les métis d’Arabes à Basra, il réprouva les fautes de langage qu’ils commettaient, du fait de leur contact avec la vie sédentaire et les enfants des non-Arabes. Une fille à lui dit un jour : « Papa, mā ašaddu l-h arri ? [quelle est la chaleur la plus intense ?] »—« La canicule en plein midi, ma petite fille», lui répondit-il, ou quelque chose de ce genre, car il y a divergence dans le récit. « Je ne t’ai pas demandé cela, lui dit-elle, je me suis seulement étonnée de la chaleur intense ».—« Alors dis, reprit-il, mā ašadda l-h arra ! [quelle chaleur intense !] ». Et d’ajouter : « Nous appartenons à Dieu; la langue de nos enfants s’est corrompue ». Il songea à faire un ouvrage, où il rassemblerait les fondements de l’arabe, mais Ziyād l’en empêcha. « Nous ne croyons pas, dit-il, que les gens se fient à ce livre, ni qu’ils abandonnent la [bonne] langue et cessent de tirer la pureté [linguistique] de la bouche des Arabes ». Et, ce, jusqu’à ce que les fautes de langage se répandent, deviennent nombreuses et affreuses. Alors, il lui ordonna de faire ce qu’il lui avait interdit. Et Abū l-Aswad fit un livre


pierre larcher

contenant la syntaxe de l’arabe et dit aux gens : « suivez cette voie, c’està-dire allez dans ce sens » ; nahw signifie direction et c’est pourquoi la grammaire a été appelée nahw. On dit qu’il fut le premier à écrire dans un ouvrage que le discours est nom, verbe et particule, dotée d’une valeur sémantique. Interrogé à ce sujet, il déclara : « je l’ai emprunté au commandeur des croyants Alī b. Abī Tālib (Dieu étende ses bénédictions et son salut sur lui !) ». Il arrive qu’un nom, un qualificatif ou un surnom l’emporte pour une chose. Celle-ci est alors connue sous ce nom spécifiquement, à l’exclusion de tout autre objet entrant dans la compréhension de ce nom. Le fiqh, on le sait, est l’intelligence des choses. On dit « faqihtu le récit » aussi bien que « je [l’] ai compris » et un homme faqīh ou faqih, c’est-à-dire qui comprend. Puis le fiqh est devenue la science religieuse, spécialement. Et quand on dit un homme faqīh, on vise seulement l’homme savant en matière de Loi, même si toute personne qui comprend une science et y excelle est un faqīh en cette science. Et, de même, tibb est l’habileté. C’est de là que l’on dit un homme de tibb et tabīb, s’il est habile. Puis tabīb est devenu inspérable de ceux qui s’intéressent à la science des philosophes ayant pour effet la conservation de la santé et, plus spécialement, permettant de la recouvrer. Les exemples de ce genre de choses abondent.

SĪBAWAYHI’S VIEW OF THE ZARF AS AN ĀMIL Aryeh Levin The Hebrew University, Jerusalem

1. Introduction 1.1

The meaning of the term zarf

In Sībawayhi’s terminology the term zarf (plural zurūf ), designates an expression denoting place or time. The zarf is an accusative as in al-laylata ‘to-night’,1 or a combination of an accusative + genitive as in maahu ‘with him’,2 xalfa-ka ‘behind you’3 and yawma l-jumati ‘Friday’,4 or a combination of a h arf jarr + genitive, as in fī-hā ‘in it.’5 1.2

The syntactic status of the zarf

In given syntactic constructions the zarf is an indispensable part of the sentence, while in others it is a dispensable part. When the zarf is an indispensable part it occurs as a predicate in some types of the nominal sentence, as in the examples abdu llāhi fīhā (Sīb. I:222, 18) and fīhā abdu llāhi (Sīb. I:222, 17) “ Abdallah is in it”6 and abdu llahi fīhā qāiman “ Abdallah is standing in it” (Sīb. I:222, 15). A zarf also occurs as the indispensable predicate in sentences beginning with inna and the kind of kāna the later grammarians called kāna an-nāqisa, as in the examples inna fīhā zaydan and inna zaydan fīhā “verily Zayd is in it” (Sīb. I:222, 20); inna zaydan lafīhā qāiman “verily Zayd is standing in it.” (Sīb. I:242, 12); mā kāna fīhā ah adun xayrun minka “nobody better than you was in it” (Sīb. I:21, 7) and mā kāna 1

See Sīb. I:176, 17–20. The form maa is conceived of by Sībawayhi and the other grammarians as a noun taking the accusative (see Sīb. I:177, 14–15). When following the particle min ma takes the genitive, as in the example min maihi—‘from him’ (ibid.). 3 See Sīb. I:170, 17–20. 4 See Sīb. I:176, 17–20. 5 See Sīb. I:207, 20–21. 6 See Sīb. I:222, 14–20. Cf. Sīb. I:207, 20–21. 2


aryeh levin

ah adun mitluka fīhā “nobody like you was in it” (ibid.). When the zarf is the indispensable predicate of the sentence it is called mustaqarr,7 lit., ‘a place where someone is.’ As a grammatical term mustaqarr designates ‘an indispensable predicate of the nominal sentence, denoting the place where the subject is.’8 The term mustaqarr is sometimes restricted by an expression denoting its grammatical quality as an indispensable part of the sentence, as in mustaqarran taktafī bihi ‘a predicate denoting the place where the subject is, with which you content yourself [when intending to express a complete sentence]’ (Sīb I:21, 11), and mustaqarran lizaydin yastaġnī bihi s-sukūtu ‘a predicate denoting the place where Zayd is [occurring in the sentences inna fīhā zaydun and inna zaydan fīhā], with which a complete sentence can be satisfied as its complement.’9 (Sīb. I:222, 20–21). When the zarf is not an indispensable part of the sentence it is called ġayr mustaqarr ‘not a mustaqarr.’10 It is said that the zarf in this case is mulġan or laġw ‘a dispensable zarf which does not operate as an āmil’ (see below § 4). The form mustaqarr sometimes occurs in combinations referring to a dispensable part of the sentence. These combinations include restrictive expressions indicating that the whole combination refers to a dispensable part: in referring to the example fīhā abdu llāhi qāimun “Abdallah is standing in it” (Sīb. I:223, 2), where fīhā is a dispensable part of the sentence, Sībawayhi says that fīhā here is a mustaqarr lil-qiyām ‘an expression denoting the place where the act of standing [expressed in the predicate qāimun] takes place.’11 Similarly, in referring to the examples fīha abdu llahi qāiman and abdu llāhi fīha qāiman “Abdallah is standing in it” (Sīb. I:222,15), Sībawayhi says that qāiman is a h ālun mustaqarrun fīhā ‘[an expression denoting] a h āl (= a state) where [the subject abdu llāhi] is.’ Note that the above combination refers to a part of the sentence that is a h āl and not a zarf.


See Sīb, I:21, 7–17; 222, 14–22. This definition is inferred from Sīb. I:222, 14–22; See also Sīb. I:21, 4–11; as-Sīrāfī III, 11, 9–14; as-Sīrāfī according to Jahn, 1895, I/2, 73, note 16. 9 Lit. “with which silence can be satisfied [after expressing a complete sentence].” 10 See Sīb. I:21, 14–15. 11 See Sīb. I:223, 1–9. 8

sĪbawayhi’s view of the zarf as an āmil


2. The amal in sentences containing a zarf 2.1 A zarf cannot operate as the āmil producing the nominative in the subject Sībawayhi believes that in nominal sentences a zarf cannot operate as the āmil producing the nominative in the subject (= al-mubtada), irrespective of whether the zarf occurs as an indispensable predicate (= mustaqarr) or as a dispensable part of the sentence. Hence, in a sentence like fīhā abdu llāhi qāiman (Sīb. I:222,15), the āmil of the subject abdu llāhi is not the predicate fīhā, but the abstract āmil called al-ibtidā.12 In Sībawayhi’s view, the sense of al-ibtidā is ‘the act of putting the noun in a position where it is unaffected by any word operating as an āmil.’13 The view that the zarf cannot be the āmil of the subject derives from the notion that in a nominal sentence, a word operating as an āmil producing the nominative must be logically identical with the noun affected by it. For example: in the sentence abdu llāhi axūka “Abdallah is your brother” (Sīb. I:6, 11), the subject abdu llāhi is logically identical with the predicate axūka, since Abdallah is your brother and your brother is Abdallah.14 Hence the subject abdu llāhi is the āmil producing the nominative in the predicate axūka.15 In contrast, in fīhā abdu llahi qāiman, the predicate fīhā is not identical with the subject abdu llāhi, since it is an expression denoting the place where the subject is, and hence fīhā cannot be the āmil producing the nominative in the subject abdu llāhi.16 As a result, this sentence does not include any word that can operate as an āmil producing the nominative in the subject abdu llāhi. Hence, abdu llāhi takes the nominative because of the effect of al-ibtidā.17 Sībawayhi contends that examples beginning with inna, such as inna fīha zaydan (Sīb. I:222, 20), confirm that fīhā does not produce the


See Sīb. I:222, 14–223, 18. See Levin, (forthcoming), mubtada, § 4.1. 14 This notion is discussed in detail in Levin, 1979, 199–202; Levin, 2002, 359, 15–360, 11; Levin, 2006, 110–111, §5; Levin, forthcoming, Cahiers linguistiques, § 3.2. 15 For this notion see Sīb. I:239, 5–9. 16 See Sīb. I:222, 14–223, 18. 17 See Sīb. I:222, 14–19. 13


aryeh levin

nominative in the subject, since in the above example inna is the āmil of the subject zaydan, as is shown by its accusative ending.18 Although Sībawayhi believes that the zarf cannot be the āmil of the mubtada, it is inferred that he holds that the zarf can produce the accusative in nouns occurring as a h āl, as in nominal sentences of the type abdu llāhi fīhā qāiman and fīhā abdu llāhi qāiman “Abdallah is standing in it” (Sīb. I:222, 15).19 In his view, in these examples, the āmil producing the accusative in the h āl qāiman is the zarf fīhā.20 Sībawayhi also explicitly says that the zarf is the āmil producing the accusative in words denoting measures of distance occurring as a tamyīz 21 (see below § 3). 2.2 Arguments confirming that in Sībawayhi’s view the zarf is the āmil of the hāl Sībawayhi does not explicitly say that the zarf is the āmil producing the accusative in the h āl in certain constructions of the nominal sentence. However, his view in this respect is inferred from some places in the Kitāb text. 2.2.1 In his discussion of sentences beginning with mā kāna, Sībawayhi refers to two types: (i) sentences where the zarf fīhā occurs as a mustaqarr, i.e., as the indispensable predicate of the sentence, as in mā kāna fīhā ah adun xayrun minka “nobody better than you was in it” (Sīb. I:21, 7). (ii) sentences where the zarf fīhā occurs as a dispensable part of the sentence (mulġan or laġw), as in mā kāna ah adun xayran minka fīhā (Sīb. I:21, 10). In this discussion Sībawayhi explicitly says that a zarf which is a mustaqarr can operate as an āmil.22 He also adds here that since a mustaqarr


Sīb. I:222, 19–223, 1. Sībawayhi refers to the accusative in these examples both as a h āl (Sīb. I:223, 1) and a xabar (Sīb. I:222, 140). For this special use of xabar see Levin, 1979, 193–196, § 2.4. 20 See Sīb. I:222, 4–9. See also Sīb. I:167, 11–16; Sīb. I:218, 6–16, especially lines 12–13. See below § 2.2– 21 The later grammarians’ term tamyīz does not occur in the Kitāb. 22 See Sīb. I:21, 7–19. 19

sĪbawayhi’s view of the zarf as an āmil


can operate as an āmil, it is preferable to put it in a position where it precedes the noun affected by it. The more the speaker makes it precede this noun the better, he says. In contrast, he says that when the zarf is not a mustaqarr, it is preferable to put it at the end of the sentence, or at least in a position close to the end.23 2.2.2 In his above discussion, dealing with the possibility that a zarf can operate as an āmil, Sībawayhi does not give any example illustrating this point. However, it is clear that his statements in this respect refer to examples such as fīhā abdu llāhi qāiman and abdu llāhi fīhā qāiman, discussed elsewhere in the Kitāb.24 Since Sībawayhi explicitly says that in these examples the āmil of the mubtada abdu llāhi is not the zarf, but al-ibtidā (see above § 2.1), it is inferred that the āmil producing the accusative in qāiman, which is a h āl, is the zarf fīhā. It should be stressed that the zarf fīhā is the only part of the sentence which can be the āmil producing the accusative in qāiman, since the mubtada abdu llahi is logically identical with qāiman, and a noun which is logically identical with another noun can produce in it only the nominative, as in fīhā abdu llāhi qāimun (see above § 2.1). 2.2.3

The above inference is supported by the following considerations: In his discussion of the example laka š-šāu šātan bidirhamin šātan bidirhamin “[The sale of] the sheep became binding on you25 [in a situation where their price is] one dirham a sheep, one dirham a sheep” (Sīb. I:167, 13), Sībawayhi says that the accusative šātan occurring twice in this example is a h āl. He explains how the accusative is produced in šātan as follows: wasāra laka š-šāu idā nasabta bimanzilati wajaba š-šāu kamā kāna fīhā zaydun qāiman bimanzilati staqarra zaydun qāiman “when you put [šātan] in the accusative [the utterance] laka š-šāu is equivalent to wajaba š-šāu, just as [the utterance] fīhā zaydun qāiman is equivalent to istaqarra zaydun qāiman” (Sīb. I:167, 13).26 Since Sībawayhi believes that in all verbal sentences the verb is the


See Sīb. I:21, 9–19. See Sīb. I:222, 14–224,2. 25 The translation here is based on Lane’s rendering of the expression wajaba l-bayu (see Lane VIII, 2922A,15–17). According to Sībawayhi laka š-šāu is equivalent to wajaba š-šāu. 26 See Sīb. I:167, 11–16. Cf. Sīb. I:222, 14–223, 1. 24


aryeh levin

āmil of the h āl,27 it is clear that in his view, the āmil of šātan in wajaba š-šāu šatan bidirhamin šātan bidirhamin is the verb wajaba, while the āmil of qāiman in istaqarra zaydun qāiman is the verb istaqarra. Since Sībawayhi holds that laka and fīhā are equivalent to wajaba and istaqarra which operate as awāmil, it is inferred that in his view laka and fīhā also operate as awāmil in the corresponding examples laka š-šāu šatan bidirhamin šātan bidirhamin and fīhā zaydun qāiman respectively. For similar examples where a combination of li + genitive or li + a relative clause operates as an āmil see Sīb. I:223, 18–224, 2. In referring to the example fīhā zaydun qāiman fīhā “In it is Zayd, in it” (Sīb. I:238, 15) Sībawayhi says: fainnamā ntasaba qāimun bistignāi zaydin bifīhā “[The h āl qāiman] takes the accusative because of the fact that when [the first] fīha is added to zaydun, it makes the sentence complete.”28 The significance of this wording is that since the first fīhā in the above utterance is the indispensable predicate making the sentence complete, it is the āmil producing the accusative in the h āl qāiman in fīhā zaydun qāiman fīhā.

3. The zarf as an āmil of a tamyīz denoting a measure of distance Sībawayhi explicitly says that in the example dārī xalfa dārika farsaxan “the wandering territory of my tribe is situated in a distance of one parasang from that of yours” (Sīb. I:176, 6–7), the āmil producing the accusative in farsax is the zarf xalfa dārika, which is the indispensable predicate of the sentence.29 He adds that this zarf produces the accusative in farsaxan and in other measures of distance, since it is not logically identical with farsaxan, and since farsaxan is not the sifa of xalfa dārika. He compares the effect of xalfa dārika on farsaxan with that of išrūna on dirhaman, in the example išrūna dirhaman.30


See Sīb. I:15, 18–22. See Sīb. I:238, 14–18. 29 See Sīb. I:176, 6–12. For xalfa as a zarf see Sīb. I:177, 14. 30 See Sīb. I:176, 6–11. For the amal in išrūna dirhaman see Carter, 1972. See also Levin, (forthcoming), Cahiers linguistiques, § 3.2. 28

sĪbawayhi’s view of the zarf as an āmil


4. The ilġā of the zarf According to Sībawayhi, one of the syntactic qualities of the zarf is alilġā, lit. ‘the abolishment.’ It is inferred from the text of the Kitāb that this term, when referring to a zarf occurring in certain syntactic constructions, denotes the abolishment of the status of the zarf as an indispensable predicate of the sentence. In this case al-ilġā is opposed to al-istiqrār “the occurrence of the zarf as an indispensable predicate.”31 In other syntactic constructions al-ilġā denotes the abolishment of the status of the zarf both as an indispensable predicate and as an āmil. In these constructions al-ilġā is opposed both to al-istiqrār and al-imāl, i.e., “the appliance of the amal of the zarf to the case ending of a certain noun.”32 The following examples illustrate the contrast between al-ilġā on the one hand, and al-istiqrār and al-imāl on the other hand in certain syntactic constructions: (1) In the example mā kāna fīhā ah adun xayrun minka “nobody better than you was in it” (Sīb. I:21, 7), the zarf fīhā is the indispensable predicate of the sentence33 (= al-mustaqarr). In contrast, in mā kāna ah adun xayran minka fīhā (Sīb. I:21, 10), the ilġā of the zarf fīhā takes place, since the indispensable predicate is xayran minka, and hence the status of fīhā as the indispensable predicate is abolished.34 (2) In the examples abdu llahi fīha qāiman and fīhā abdu llāhi qāiman (Sīb. I:222, 15) the zārf fīhā is the indispensable predicate,35 and it is also the āmil of the h āl qāiman . In contrast, in fīhā abdu llāhi qāimun (Sīb. I:223,2) the ilġā of fīhā takes place, since qāimun is the predicate, and fīhā loses its status as an indispensable predicate and as an āmil.36


See Sīb. I:21, 14–15. The contrast between ilġā and imāl is inferred from Sīb. I:21, 9–14, where Sībawayhi uses the forms āmilan and yamalāni in contrast to al-ilġā and alġayta. For the occurrence of the term imāl in the Kitāb, also when referring to other awāmil, see Troupeau, 1976, 149, voc. imāl. 33 The expression ‘indispensable predicate’ here is based on Sībawayhi’s view of xabar kāna (see Levin, 1979, 203–205, § 2.6). 34 See Sīb. I:21, 7–14. 35 See Sīb. I:222, 14–223,1. 36 See Sīb. I:223, 1–2. 32


aryeh levin

Sībawayhi compares the ilġā of the amal of the zarf with that of the verbs later called afāl al-qulūb.37 He says that when the ilġā of the zarf takes place, it is preferable to pronounce it at the end or close to the end of the sentence. In contrast, when the zarf is a mustaqarr occurring as an āmil, it is preferable to put it at the beginning of the sentence, like verbs such as āzunnu and ah sibu, which are pronounced at the beginning of the sentence when they are awāmil.38 Sībawayhi illustrates two types of taqdīr construction of sentences containing a zarf mulġan: (i) In referring to the sentence inna bika zaydan maxūdun “Zayd is enchanted by you” (Sīb. I:242, 2), where the ilġā of the zarf bika takes place, Sībawayhi says that when the speaker expresses this sentence, it is as if he were saying inna zaydan maxūdun. Similarly, when saying inna fīka zaydan la rāġibun “Zayd covets you” (Sib. I:242, 5) it is as if the speaker were saying inna zaydan rāġibun.39 These taqdīr constructions illustrate the notion that when the above sentences are pronounced it is as if the zarf bika and fīka are not spoken, and hence they cannot operate as the āmil producing the accusative in maxūdun and rāġibun respectively. It appears that Sībawayhi holds this view in order to solve a theoretical difficulty arising from one of the main principles of the theory of amal: in his view, the effect of an āmil producing the nominative or the accusative in the noun is always applied, irrespective of whether this āmil is an indispensable part of the sentence or not.40 This principle seems to be violated if one assumes that when a zarf such as fīhā is an indispensable predicate, as in fīhā zaydun qāiman, it is the āmil producing the accusative in the h āl qāiman, but when fīhā is a dispensable part, as in fīhā zaydun qāimun, its amal is abolished. In order to solve this difficulty Sībawayhi says that when the amal of the zarf is abolished, the zarf does not occur in the taqdīr construction. Since according to the grammarians the relevant construction, as far as grammatical analysis is concerned, is that of the taqdīr, Sībawayhi assumes that fīhā

37 The ilġā of the amal of this category of verbs is discussed in chapter 31 of the Kitāb (= Sīb. I:49, 4–52, 15). 38 See Sīb. I:21, 10–13. 39 See Sīb. I:242, 2–8. Sībawayhi’s words expressing this notion are very clear: . . . kaannaka aradta inna zaydan rāġibun wainna zaydan maxūdun walam tadkur fīka walā bika faulġiyatā hāhunā kamā ulġiyatā fī l-ibtidāi (Sīb. I:242, 7–8). 40 See Sīb. I:223, 6–13.

sĪbawayhi’s view of the zarf as an āmil


does not occur in the taqdīr construction,41 and hence it is clear that its amal cannot be applied. (ii) Sībawayhi says that when the speaker expresses the sentence fīhā abdu llāhi qāimun he intends it is as if he were saying abdu llāhi qāimun fīha. This taqdīr construction illustrates the view that when the zarf is not an indispensable predicate and hence is not an āmil, the speaker imagines that it is as if he were pronouncing the zarf at the end of the sentence, since as regards grammatical theory it is preferable to pronounce a zarf which is not an āmil at the end of the sentence, or at least in a position close to the end (see above § 2.2.1). Sībawayhi says that when the ilġā of the zarf occurs in an example like fīhā abdu llāhi qāimun the speaker imagines that it is as if he were saying abdu llāhi qāimun fīhā, since in this taqdīr construction the zarf which is not an āmil occurs at the end of the sentence.42

5. An interpretation of a difficult passage from the Kitāb (Sīb. I:207, 17–21) Sībawayhi’s discussion in Sīb. I:207, 17–21 is one of the most difficult passages in his text. It contains some points relevant to the topic of this paper. In referring to the example marartu birajulin maahu saqrun sāidan bihi ġadan “I passed by a man with a hawk who was intending to hunt with it tomorrow” (Sīb., Hārūn II:52, 6–7),43 Sībawayhi says: walam annaka idā nasabta fī hādā l-babi faqulta marartu birajulin maahu saqrun sāidan bihi ġadan fal-nasbu alā h ālihi lianna hādā laysa bibtidāin walā yušbihu fīhā abdu llāhi qāimun ġadan lianna z-zurūfa tulġā h attā yakūna l-mutakallimu kaannahu lam yadkurhā fī hādā l-mawdii, faidā sāra l-ismu majrūran aw āmilan fīhi filun aw mubtadaun lam tulġihi liannahu laysa yarfauhu l-ibtidāu, wafī z-zurūfi idā qulta fīha axawāka qāimāni yarfauhu l-ibtidāu

41 For taqdīr constructions which are shorter than their corresponding literal constructions see Levin, 1997, 146–148, § 3.3. 42 See Sīb. I:222, 22–223,6. 43 Derenbourg’s edition has āidan instead of sāidan (see Sīb. I:207, 17–18). However, Hārūn’s version sāidan is supported by Sīb. I:206, 8 in Derenbourg. The version sāidan also occurs in all the later grammarians’ treatises (see, for example, al-Fārisī, I: 250, 7–10; as-Sīrāfī, VI: 131, 7 (in a quotation from Sībawayhi’s text).


aryeh levin know that if you put a noun in the accusative in [syntactic constructions of] this type,44 and you say marartu birajulin maahu saqrun sāidan bihi ġadan, the accusative ending [occurring in sāidan] remains unchanged, because this [utterance]45 is not [an independent sentence] occurring at the beginning of the utterance,46 and it does not resemble [the nominal sentence] fīhā abdu llāhi qāimun ġadan (= Abdallah is standing in it tomorrow) [where the form qāim can take either the nominative or the accusative],47 because [when expressing sentences such as fīhā abdu llāhi qāimun ġadan ] the [amal] of the expressions denoting place (= zurūf ) is abolished, [and the feeling of] the speaker is that [when expressing his literal utterance he intends] it is as if he were not pronouncing the zarf in this place at all.48 When the noun [rajulin] takes the genitive [in the example marartu birajulin maahu saqrun sāidan bihi ġadan], or when it is affected by a verb [as in the example raaytu rajulan maahu saqrun sāidan bihi ġadan]49 or [when it is affected by the mubtada [in an example such as hādā rajulun maahu saqrun sāidan bihi ġadan]50 you do not abolish the amal [of the zarf maa which produces the accusative in sāidan], because the ibtidā does not produce the nominative in the noun [rajulin in the example marartu birajulin maahu saqrun sāidan bihi ġadan], [and hence this noun is not a mubtada which can be the āmil producing the nominative in the word sāid]. [On the other hand, in sentences beginning with] a zarf,51 when you say fīhā axawāka qāimāni (= your two brothers are in it), the ibtidā produces the nominative in [the mubtada, which is axawāka]52 [so this mubtada is the āmil producing the nominative in the xabar, which is qāimāni (Sīb. I:207, 17–21).53

44 I.e., in syntactic constructions including a relative clause beginning with a zarf, such as marartu birajulin maahu kīsun maxtūmun alayhi “ I passed by a man having with him a sealed sack (Sīb. I:207, 15–16). Many examples of this type are discussed in Chapter 112 of the Kitāb (=Sīb. I:206,5–210,2). 45 I.e., the clause maahu saqrun sāidan bihi ġadan. 46 The term al-ibtidā here denotes “a position occurring at the beginning of the utterance.” For al-ibtidā in this sense, see, for example, Sīb. II:295, 16; 296, 11–15; 297, 3–6; 362, 16–23. 47 See Sīb. I:222, 14–223, 2. See above § 4. 48 I.e., when the speaker says fīhā abdu llāhi qāimun he intends it is as if he were saying abdu llāhi qāimun (for this notion see as-Sīrāfī, VI:135, 18–136, 5; see above § 4). 49 This example does not occur in the Kitāb. It has been introduced here according to al-Fārisī’s interpretation (See al-Farisī, I:251, 3–5) in order to explain Sībawayhi’s intention. 50 This example does not occur in the Kitāb. It is introduced according to al-Fārisī’s interpretation (see al-Farisī, I:251, 3–7), in order to explain Sībawayhi’s intention. 51 In Sībawayhi’s manner of expression a combination such as fī az-zurūf denotes the sense of “in sentences beginning with a zarf ”. Similarly, the expression fī l-fil denotes “in sentences beginning with a verb” (see Sīb. I:17, 17); fī daraba—“in a sentence beginning with daraba” (see Sīb. I:16, 18–20); fī kāna—“in a sentence beginning with kāna” (Sīb. I:17, 12). 52 For this notion see Sīb. I:222, 14–223, 18. 53 Ibid.

sĪbawayhi’s view of the zarf as an āmil


The following remarks and conclusions are inferred from the above passage. These conclusions are supported by other texts in the Kitāb, discussed in this paper. (1) A zarf can operate as an āmil only when it is a mustaqarr, i.e., only when it is the indispensable predicate of a certain sentence or of a certain clause. (2) The zarf cannot be an āmil producing the nominative in the subject or in the predicate. In a nominal sentence such as fīhā abdu llāhi qāimun the mubtada abdu llāhi takes the nominative because of the amal of al-ibtidā, and abdu llahi is the āmil producing the nominative in the predicate qāimun. (3) In a sentence such as marartu birajulin maahu saqrun sāidan bihi ġadan, the noun rajulin cannot be the āmil producing the nominative in sāid, since rajulin is not affected by the ibtidā, and hence it is not a mubtada. In Sībawayhi’s view, in a nominal sentence, only a mubtada can produce the nominative in a noun occurring as a predicate. The noun affected by the mubtada must be logically identical with it. Since the sentence marartu birajulin and the clause maahu saqrun sāidan bihi ġadan do not contain a mubtada logically identical with sāid, sāid cannot take the nominative. The only word, which can be the āmil of sāid in the above utterance is the zarf maa, which can produce the accusative in sāid. Hence it is impossible to abolish the amal of maa, since if its amal were to be abolished the word sāid would remain without an āmil. (4) But in fīhā axawāka qāimāni , the mubtada axawāka takes the nominative because of the amal of al-ibtidā, and hence it can be the āmil of the predicate qāimāni. It is also possible to say fīhā axawāka qāimayni “in it are your two brothers standing.” In this structure axawāka takes the nominative because of the amal of al-ibtidā, and fīhā, which is a mustaqarr, is the āmil producing the accusative in the h āl qāimayni. (5) The form ibtidā contained in the expression lianna hādā laysa bibtidāin denotes “an expression occurring at the beginning of the utterance.” It does not denote here any of the terms it designates in Sībawayhi’s terminology of the nominal sentence. The words lianna hāda laysa bibtidā express the notion that the clause maahu saqrun sāidan bihi is not an independent sentence occurring at the beginning of the utterance. (6) There is another argument which, according to as-Sīrāfī, prevents the ilġā of maahu in the clause maahu saqrun sāidan bihi ġadan:


aryeh levin

maahu, as-Sīrāfī says, includes the antecedent -hu , referring to rajulin in marartu birajulin maahu saqrun sāidan bihi ġadan. If the ilġā of maahu were applied, it would have been dropped from the taqdīr construction, which in this case would be marartu birajulin saqrun sāidun bihi ġadan. This taqdīr construction cannot exist, since the clause contained in it does not include an antecedent referring to rajulin. Since in the grammarians’ view the taqdīr construction is the relevant one as far as grammatical analysis is concerned, it is impossible to apply here the ilġā of maahu, since this would create an unacceptable taqdīr construction.

6. Conclusions 1

Syntactically, Sībawayhi distinguishes between two kinds of a zarf: 1.1 A zarf which is an indispensable predicate of a nominal sentence, as fīhā in the example fīhā abdu llāhi qāiman “In it is Abdallah standing.” This kind of zarf is called mustaqarr ‘[A zarf] denoting the place where the subject is.’ 1.2 A zarf which is a dispensable part of the sentence, as fīhā in the example abdu llāhi qāimun fīhā “Abdallah is standing in it.” This kind of zarf is sometimes called mulġan or laġw or ġayr mustaqarr. 2 A zarf which is an indispensable predicate (= mustaqarr) is liable to operate as the amil producing the accusative in a part of a sentence occurring as a h āl or a tamyīz denoting a measure of distance. For example: (i) In fīhā abdu llāhi qāiman, fīhā is the āmil of the h āl qāiman. (ii) In dārī xalfa dārika farsaxan “the wandering territory of my tribe is behind that of yours, at a distance of one parasang”, the zarf xalfa dārika is the āmil of the tamyīz farsaxan. 3 In contrast, a zarf which is not an indispensable part of the sentence cannot operate as an āmil. In Sībawayhi’s view, this zarf undergoes the process of al-ilġā, i.e., the process of the abolishment of its status as an indispensable predicate and as an āmil. Hence, this zarf is called mulġan or laġw “[a zarf whose] status as an indispensable predicate and as an āmil is abolished.”

sĪbawayhi’s view of the zarf as an āmil


4 Sībawayhi illustrates two types of taqdīr construction of sentences containing a zarf mulġan: 4.1 In referring to examples like inna bika zaydan maxūdun “Zayd is enchanted by you” (Sīb. I:242, 2), where the ilġā of the zarf bika occurs, Sībawayhi says that when the speaker expresses this sentence, he intends it is as if he were saying abdu llāhi maxūdun. This taqdīr construction illustrates the notion that when pronouncing the above sentence the speaker intends that it is as if the zarf bika is not pronounced, and hence it cannot operate as the āmil producing the accusative in maxūdun. 4.2 When the speaker expresses the sentence fīhā abdu llāhi qāimun he intends it is as if he were saying abdu llāhi qāimun fīha. This taqdīr construction illustrates the view that when the zarf is not an indispensable predicate and hence is not an āmil, the speaker imagines that it is as if he were pronouncing the zarf at the end of the sentence, since as regards grammatical theory it is preferable to pronounce a zarf which is not an āmil at the end of the sentence, or at least in a position close to the end.

7. References 7.1

Primary sources

al-Fārisī, Abū Alī al-H asan b. Ahmad b. Abd al-Ġaffār. (d. 377/987). at-Talīqa alā Kitāb Sībawayhi. Iwad b. H amad al-Qūzī, ed. 1410 A.H. = 1990. Cairo. Sībawayhi. (d. 177/793). Le livre de Sībawaihi. Traité de grammaire arabe. Hartwig Derenbourg, ed. 1881–1889. Paris. 2 vols. Abū Bišr Amr b. Utm  ān b. Qanbar. (d. 177/793). al-Kitāb. Kitāb Sībawayhi. Abd as-Salām Hārūn, ed. Cairo, 1977. 5 vols. as-Sīrāfī, Abū Saīd. (d. 368/979). Šarh Kitāb Sībawayhi. Ramadān Abd at-Tawwāb and others, eds. Cairo, 1988–2004. 6 vols. 7.2

Secondary sources

Carter, Michael G. 1972. “ ‘Twenty Dirhams’ in the Kitāb of Sībawayhi,” Bulletin of the School of Oriental and African Studies. London 35, 485–496. Jahn, G. 1895. Sībawaihi’s Buch über die Grammatik, übersetzt und erklärt von G. Jahn. Vol. I, second pagination. Berlin. Lane, E.W. 1863–1893. Arabic-English Lexicon. London (8 volumes). Levin, Aryeh. 1979. “Sībawayhi’s view of the Syntactical Structure of kāna wa-axawātuhā,” Jerusalem Studies in Arabic and Islam 1, 185–211. ——. 1997. “The Theory of al-Taqdīr and its Terminology. Jerusalem Studies in Arabic and Islam 21, 142–166.


aryeh levin

——. 2002. “An Interpretation of a Difficult Passage from the Kitāb.” Jerusalem Studies in Arabic and Islam 27, 356–362. ——. 2006. “An Interpretation of Two Difficult Passages from al-Kitāb Referring to the Āmil in Elliptical Sentences.” Jerusalem Studies in Arabic and Islam 31, 107–117. ——. (forthcoming). “Sībawayhi’s View of al-mubtada and al-ibtidā.” Moshe Bar-Asher’s Festschrift. ——. (forthcoming). “The āmil of the xabar in Old Arabic Grammar.” In: Cahiers linguistiques de Inalco. Troupeau, Gérard. 1976. Lexique-Index du Kitāb de Sībawayhi. Paris: Klincksieck.


1. Introduction In medieval Arab grammatical tradition, the two basic sentence types jumla filiyya and jumla ismiyya, are normally defined by the first occurring predicative constituent. A verb followed by its subject signals a jumla filiyya (‘verbal sentence’), e.g. daraba abdu-llāhi zaydan (“Abdullāh hit Zayd”). By contrast, a sentence introduced by a nominatival noun is a jumla ismiyya (‘nominal sentence’), e.g. zaydun rajulun (“Zayd is a man”). This binary division corresponds with the grammarians’ theory of amal (‘regimen’), or, to be more specific, with two basic types of amal. A jumla filiyya correlates with a verbal āmil (‘operator’), whereas a jumla ismiyya corresponds with ibtidā, which is considered an abstract āmil.1 The ibtidā is normally said to consist of a zero phonological āmil and the predicatival relationship between the mubtada and the xabar, respectively the subject and predicate in this type of sentence. The basic principle of amal, stipulating that the āmil should precede the mamūl, applies in both sentence types. In a jumla filiyya, the verbal āmil affects the complements following it; in a jumla ismiyya, the abstract āmil, ibtidā, occupies in principle a pre-mubtada position, from where it assigns the raf case to the mubtada; the latter, in turn, assigns raf  to the xabar (according to Sībawayhi’s [Kitāb I:239] version). As for cases such as daraba zaydan abdu-llāhi and rajulun zaydun, these were presented as cases of taqdīm wa-taxīr (‘preposing and postposing’), i.e. as the inverted versions of daraba abdu-llāhi zaydan and zaydun rajulun; in other words, as inverted jumla filiyya and jumla ismiyya respectively.


For a detailed discussion of this correspondence, see Levin 1985.


yishai peled

However, the linkage between the concept of two sentence types on the one hand, and the theory of amal on the other, turned out to constitute a major problem with regard to the binary division into jumla filiyya and jumla ismiyya. The grammarians realized that such types as daraba abdu-llāhi zaydan and zaydun rajulun, with their inverted versions, leave various constructions that do not easily fit into any of the two categories. The two apparently most problematic cases may be represented by the two model sentences qāimun zaydun (“Standing is Zayd”), where a participle is followed by a definite noun phrase, and zaydun fī d-dār, or fī d-dāri (or fīhā) zaydun (“Zayd is in the house/in it”), where the predicate position is occupied either by a definite prepositional phrase or by an adverbial phrase such as hunā, hunāka etc. In the latter type, a definite subject noun may either precede or follow the adverbial/prepositional predicate; an indefinite subject must obligatorily follow its predicate. As it were, sentences such as qāimun zaydun and fī d-dāri (or fīhā) zaydun may be considered as cases of an inverted jumla ismiyya, pragmatically motivated. And modern linguists, uncommitted to the theory of amal, would probably regard them as such. Yet for many of the medieval grammarians, they represent, rather, a sentence type in its own right. To be more precise, fī d-dāri zaydun is explicitly presented as such by some of the grammarians; qāimun zaydun, by contrast, is often dealt with in a way that leads one to believe that it was considered by certain grammarians as representing a third sentence type.2 I would argue that the controversies that arose over these (and other) constructions point to what may be viewed as gaps in the medieval theory of sentence types. In other words, an attempt will be made to show that the theory, based on a binary conception of jumla filiyya and jumla ismiyya, representing two types of amal, was highly vulnerable and far from stable. This paper concentrates on the theoretical problems presented by the above two constructions. We start with qāimun zaydun.

2 To be sure, there were grammarians who analyzed both constructions as an inverted jumla ismiyya with a fronted xabar; others accepted more than one type of analysis. For a discussion, cf. Ibn Abī r-Rabī, Basīt I:583ff.

the medieval arabic theory of sentence types


2. qāimun zaydun In his bāb al-ibtidā, Sībawayhi (Kitāb I:239) discusses the option of mubtada-xabar inversion. His starting point is that the standard preferred (al-h add) structure is for the mubtada to precede the xabar rather than the reverse, much as the standard word order in the verbal sentence is for the fāil to precede the maf ūl. When dealing with inversion, he singles out qāimun zaydun as a markedly complex case deserving special attention. He quotes his teacher al-Xalīl as saying that qāimun zaydun is an ill-formed (qabīh ) sentence unless analyzed as the inverted version of zaydun qāimun. As is well known, some of the later grammarians made the point that such an inversion is quite problematic, since it places the mamūl before the āmil, and the Kūfans saw a further problem in that it makes the pronoun implicit in the participial form qāimun precede its antecedent (al-idmār qabla d-dikr—see, e.g. Ibn al-Anbārī, Insāf I:65). But these problems were easily dismissed by the claim that qāimun zaydun represents a secondary ( far) or surface (lafz) structure, whereas in the basic structure (manā, niyya, taqdīr) zaydun, the zāhir and āmil precedes qāimun, the mamūl, with the implicit pronoun (the mudmar) referring back—as required—to zaydun (for a detailed discussion, see, e.g. Ibn al-Anbārī, Insāf I:65–66, 68). However, the real problem with qāimun zaydun was associated with a different—verbal—analysis of the participle, known to have been advocated by some of Sībawayhi’s contemporaries. Citing al-Xalīl, Sībawayhi points out that: fa-idā lam yurīdū hādā l-manā wa-arādū an yajalūhu filan ka-qawlihi yaqūmu zaydun wa-qāma zaydun qabuh a li-annahu smun “If, however, they do not accept this analysis [= inversion], and want to treat [qāimun] as a verb, in analogy to such sentences as yaqūmu zaydun and qāma zaydun, this should be rejected, because [qāimun] is a noun”— (Sībawayhi, Kitāb I:239).

Yet Sībawayhi immediately makes it clear that under certain conditions an active participle, while categorized as a noun, may implement a verbal function (yajrī majrā l-fil). This could be accepted (h asuna indahum), he maintains, if the participle functions as part of an asyndetic relative clause (sifa) linked to some antecedent (mawsūf), or, otherwise, governed by a preceding operator such as a mubtada. In other words, qāimun zaydun is disallowed with a verbal analysis, much as dāribun zaydan (“hitting Zayd”) is unacceptable as a complete sentence. However,


yishai peled

qāimun zaydun, as part of a larger sentence, is acceptable with a verbal analysis, in analogy to such cases as anā dāribun zaydan (“I am hitting Zayd”).3 As we shall see shortly, Sībawayhi’s approach to the verbal analysis of qāimun zaydun was later established as a firm principle in medieval Arab grammatical thinking. Ibn as-Sarrāj (Usūl I:59–60), whose position regarding mubtadaxabar inversion seems to be similar to that of Sībawayhi’s, readily accepts muntaliqun zaydun as an inverted version of zaydun muntaliqun. As for analyzing qāimun zaydun as analogous to yaqūmu zaydun where qāimun is not preceded by any ‘supporting’ element (see below), in other words, construing zaydun as fāil to muntaliqun—in principle, Ibn as-Sarrāj, much like Sībawayhi, regards such an analysis as misguided (qabuh a), yet he admits it as jāiz (‘acceptable’). What both grammarians seem to accept without reservation is that the noun following the participle may be analyzed as a kind of fāil provided that it is anchored (yatamidu alā, in Ibn as-Sarrāj’s words) to some preceding constituent. As an illustration of how this condition can be met, Ibn as-Sarrāj adduces such sentences as: marartu bi-rajulin qāimin abūhu (“I passed by a man whose father was standing”), zaydun qāimun abūhu (“Zayd— his father is standing”), a-qāimun abūka (“is your father standing?”). This rule would be developed by later grammarians into a general principle of itimād, designed to specify the conditions under which a non-verbal predicate may be analyzed as analogous to a verb preceding its subject.4 This principle stipulates that a non-verbal predicatival constituent, such as an active participle, or an adverbial/prepositional phrase (see section 3 below) may exercise amal upon the constituent following it (the subject) only if supported by (yatamidu alā) some element such as an interrogative particle, a relative pronoun, or, otherwise, when the clause as a whole functions as an asyndetic relative clause or as xabar to a preceding mubtada. In such cases, the first predicatival constituent is perceived as behaving analogously to a verb. And as a verblike constituent it acts as a āmil, assigning the raf  case to the following

3 Sībawayhi (Kitāb I:239) asserts that while the active participle and the verb may be similar in some respects, one must appreciate the difference between them. Other grammarians (e.g. Ibn al-Anbārī, Asrār, 70) pointed out that the active participle is weaker than the verb, and cannot, therefore, exercise verbal amal, unless supported by some preceding element (see below). 4 For the concept itimād as it is used in al-Xalīl’s Kitāb al-Ayn with reference to other grammatical structures, see Talmon 1997, 210.

the medieval arabic theory of sentence types


nominal constituent.5 For discussion, see, e.g. Ibn al-Anbārī, Asrār: 70; cf. Goldenberg 2002, 199–201. Ibn as-Sarrāj is aware, however, of the implications of a verbal analysis of qāimun zaydun for the theory of amal. He argues (Usūl I:60) that in qāimun abūka (“Your father is standing”), qāimun is assigned the raf  case by the ibtidā, and abūka is assigned the same case by the ‘verb’ preceding it. He indicates further that abūka fills a xabar position. In any event, both Sībawayhi and Ibn as-Sarrāj reject the use of dāribun bakran amrun (“ Amr hits Bakr”) as an independent sentence, on the ground that the active participle, while being analogous to the verb, is by definition a nominal, and as such cannot be made to function identically to a verb in terms of case assignment. The above examples, where the participle is linked to a preceding antecedent (mawsūf), a mubtada, or an interrogative particle, are viewed as analogous to the construction dāribun bakran when anchored, under the principle of itimād, to some external constituent (mah mūl alā ġayrihi), such as a mubtada, thus presenting a well-formed independent sentence (e.g. hādā dāribun bakran—“this [person] is hitting Bakr”) (Ibn as-Sarrāj, Usūl I:60; and cf. Sībawayhi’s position above; Levin 1985, 125–126). Like Sībawayhi and Ibn as-Sarrāj, as-Zajjājī (Jumal, 37–38) was aware of the theoretical problems raised by sentences consisting of an active participle followed by a noun phrase. In particular he demonstrated the implications for the grammatical agreement between the two constituents. To the extent that qāimun in qāimun zaydun is conceived of as xabar muqaddam (a fronted xabar), it must be replaced by qāimāni or qāimūna, once zaydun is substituted by a dual or a plural form respectively. But under the alternative analysis cited by as-Zajjājī, in which qāimun is assigned a verbal function, the active participle preceding its subject should invariably take the singular form. In other words, the proponents of this analysis would have qāimun az-zaydāni/az-zaydūna rather than qāimāni z-zaydāni and qāimūna z-zaydūna.6

5 Ibn Abī r-Rabī (Basīt I:585) remarks that some grammarians rejected the idea of an adverbial/prepositional phrase assigning case. They argued that such phrases were different in status (manzila) from the adjective. The latter, they argued, is capable of inflection, and as such is more powerful than the adverbial/prepositional phrase. Therefore, they concluded, the adverbial/prepositional phrase may not be analyzed as a case assigner even where the principle of itimād is met. I return to this issue later. 6 See, e.g. Ibn Abī r-Rabī (Basīt I:584), who also indicates that the proponents of akalūnī l-barāġīt must, by extension, say qāimāni z-zaydāni and qāimūna z-zaydūna,


yishai peled

This rule, as stated by as-Zajjājī, highlights the verbal function of the participle under this particular analysis, relating qāimun az-zaydāni/azzaydūna to qāma z-zaydāni/z-zaydūna. It should be noted that the medieval grammarians treated participles, as well as other types of adjective, as complex forms incorporating a personal pronoun (for discussion, see Goldenberg 2002, 195). Ibn Yaīš (Šarh I:87–88) states clearly that participles and other adjectives are derived from the verb, and that in virtue of having a “verbal meaning” (manā fil) they must have a fāil.7 However, construing muntaliqun zaydun as modeled on yantaliqu zaydun implies neutralizing the pronominal element in muntaliqun exactly as it is done in yantaliqu.8 The argument is that the participle, much like a regular fil, cannot assign the raf case twice (lā yarfau fāilayni).9 Zajjājī (Jumal, 38) indicates that in such cases the active participle introducing the sentence is assigned raf by the ibtidā, whereas the constituent following it is assigned the same case by “its verb” (bi-filihi— apparently referring to the active participle; cf. Ibn as-Sarrāj’s analysis above). Zaydun in qāimun zaydun, it is argued, replaces the xabar (yasuddu masadd al-xabar); and the participle preceding it, he points out, is invariably singular li-annahu qad jarā majrā l-fili l-muqaddami (“for it behaves analogously to a verb preceding [its subject]”).10 Observe that, unlike the vast majority of grammarians, Zajjājī did not make the point of linking the verbal analysis of qāimun zaydun to an obligatory application of the principle of itimād. It should, indeed, be noted that some of the later grammarians held a narrower version of this principle, restricting the use of a participle in sentence-initial position to cases where the participle is attached to a negative or interroga-

since in this version of the language the verb preceding the subject agrees with it in number and gender (for a detailed discussion of akalūnī l-barāġīt, see Levin 1989). 7 The grammarians, however, recognized that the personal pronoun incorporated in an active participle cannot qualify as fāil in the way an implicit personal pronoun in a verb can. Thus, while alladī daraba zaydun (“The one who hit is Zayd”) is a perfectly grammatical sentence, alladī dāribun zaydun is not, since, unlike alladī daraba, alladī dāribun cannot implement the function of a subject clause (see Jurjānī, Muqtasid I:463–464). 8 This is, perhaps, why the Kūfans, who rejected the analysis of qāimun zaydun as an inverted nominal sentence, could accept it as modeled on a verbal sentence: under the verbal analysis the pronoun in qāimun is disabled so there is no problem of cataphora (cf. above). 9 For Ibn Yaīš (Šarh I:87–88), then, a sentence such as zaydun qāimun abūhu (“Zayd, his father is standing”) consists of a mubtada (zayd) and a xabar, the latter analyzed as a complex construction consisting of a fil (qāimun) and a fāil (abūhu). 10 This type of analysis is attributed to Axfaš; see, e.g. Ibn Usfūr, Šarh I:341.

the medieval arabic theory of sentence types


tive particle: mā qāimun az-zaydāni (“the two Zayds are not standing”), a-qāimun az-zaydāni (“Are the two Zayds standing?”). According to Ibn Aqīl (d. 1367) (Šarh I:189), qāimun, in each of the last two sentences, functions as a mubtada’, whereas az-zaydāni is a fāil sadda masadd alxabar (a fāil substituting for the xabar). Indeed, this is the common formula employed by those later grammarians who adopted the verbal analysis of the construction in question (cf. Carter 1981, 189). At this point one might ask how frequent in classical Arabic are such sentences as qāimun az-zaydāni? I looked for this construction in the Qurānic text, but no example of it was attested. In all the recorded cases, a singular participle is followed by a singular noun phrase, or, otherwise, a singular feminine participle by a plural (non-human) noun phrase. It is interesting to note, however, that all cases display some kind of a ‘supporting element’. In the vast majority of examples, the construction in question functions as predicate to a preceding subject realized as either a referential nominal (typically, but not necessarily, in a sentence introduced by inna or one of its ‘sisters’), or, otherwise, as a nonreferential damīr aš-šan: wa-zannū annahum māniatuhum h usūnuhum (“They believed that their fortresses would protect them”—Q. 59:2), wa-huwa muh arramun alaykum ixrājuhum (“You are not allowed to expel them”—Q. 2:85). Huwa in the latter example functions as damīr aš-šan. One example was noted where the ‘supporting element’ is the interrogative particle a-: a-rāġibun anta an ālihatī (“do you loathe my gods?”—Q. 19:46). What is then the effect of the ‘supporting’ element that makes [zaydun/rajulun] qāimun abūhu or a-/mā qāimun abūhu an acceptable verbal construction, as opposed to qāimun abūhu? Al-Xalīl (see above) does not provide an elaborate answer. He argues, however, that a participle cannot easily replace a verb in pre-subject position, because it is an ism. A verb and a noun, he maintains, may in certain positions implement similar functions, but they must still be differentiated. Nor did later grammarians elaborate on the function of the ‘supporting’ element. But their discussion of the relevant cases might give us a clue. To phrase the question differently, how does the ‘supporting’ element impart further verbal force to the adjectival predicate that enables it to act analogously to a verb in such cases? If we compare the two constructions qāma zaydun and qāimun zaydun, we can see that the difference between the two is that the finite verb, while devoid of a pronominal element, is still inflected for person, whereas the participle is not. Lacking either a pronominal element or inflection for person, the participle is excluded as a


yishai peled

pre-subject verbal predicate in an independent sentence. The function of the ‘supporting’ element zaydun in zaydun qāimun abūhu, is to make up for the lack of person inflection in qāimun, and thus empower the latter to implement a verbal function.11 Further, as has been indicated, the ‘supportive’ function may be implemented not only by a noun but also by an interrogative or a negative particle. These are, in other words, further sources from which the adjective could derive a verbal force. Already Sībawayhi attributed a verbal effect to certain interrogative particles (Sībawayhi, Kitāb I:39f., 41f.—the latter dealing specifically with a-). Ibn Abī r-Rabī (Basīt II:712) points out that the interrogative particle a- requires a verb (alhamza tālibatun bi-l-fil), and that underlying (taqdīr) a-zaydun daraba amran is a-daraba zaydun amran (“Did Zayd hit Amr?”) (and cf. his similar attitude to hal—II:679). Regarding the construction at issue he maintains that underlying a-qāimun zaydun is a-yaqūmu qāimun zaydun, and that the former is derived from the latter by suppressing the redundant yaqūmu, for which qāimun serves as an exponent (tafsīr). This explains why the interrogative a- qualified in the grammarians’ view as a supporting element in sentences consisting of an adjectival predicate followed by a subject. However, the verbal analysis of such constructions as (a-)qāimun zaydun raises a difficult problem for the medieval theory of sentence types. For if a sentence consists of a mubtada followed by a fāil, how should it be categorized in terms of sentence types? The fact that the fāil is presented as replacing the xabar does not make the issue any simpler. For under the suggested analysis, a verb-acting constituent is considered, in terms of irāb, as a mubtada assigned the raf  case by the ibtidā. The conception of the participle in such cases as mubtada is quite understandable, given that it is a nominatival constituent in sentence-initial position. If one declines the inversion analysis of sentences such as qāimun zaydun, how else can one account for the raf case of qāimun? The main problem with this analysis lies in its stipulating that qāimun as a mubtada assigns raf to a fāil occupying a xabar position. 11 That, I believe, is what is intended by al-Xalīl (Sībawayhi, Kitāb I:239) when he refers to the participle in such cases as [kāna] sifatan jarā alā mawsūfin aw jarā alā smin qad amila fīhi “[the participle] is an adjective agreeing with a head or [otherwise] with a noun acting upon it”, and further when he says: lā yakūnu maf ūlan fī dāribin h attā yakūnu mah mūlan alā ġayrihi fa-taqūlu hādā dāribun zaydan . . . “dārib cannot take an object unless it is linked to some other constituent, as for example in hādā dāribun zaydan—this [person] is hitting Zayd”.

the medieval arabic theory of sentence types


Once we correlate qāimun zaydun with qāma zaydun, rather than with zaydun qāimun, with the consequence that qāimun is to remain in the singular irrespective of the number of the following noun (qāimun azzaydāni/az-zaydūna), one can hardly see how sentences such as qāimun zaydun under a verbal analysis, let alone qāimun az-zaydāni, may be viewed other than as cases of jumla filiyya. Indeed, when presenting the fāil, some of the grammarians, like Ibn Yaīš (Šarh I:74), indicate that the position preceding the fāil is available for a verb ( fil) or a nominal that is analogous to a verb (šabahuhu, mā huwa fī manā l-fil min al-asmā). In this latter category they normally include the active and passive participles, as well as such adjectives as h asan (sifa mušabbaha bi-smi lfāil—‘an adjective analogous to the active participle’, e.g. Ibn Yaīš, Šarh I:87). It is argued that in a sentence such as zaydun dāribun ġulāmuhu12 (“Zayd, his slave is hitting”), dāribun, much like yadribu, assigns raf  to ġulāmuhu. One may infer, then, that Ibn Yaīš would regard a sentence such as qāimun zaydun as a verbal sentence. Yet I have not recorded any explicit reference to this type of sentence as a jumla filiyya. In any case, the prevalent analysis of the construction under discussion was, as already indicated, mubtada+fāil sadda masadd al-xabar. It is not surprising, however, that the grammarians adhering to this analysis did not commit themselves to explicitly categorizing such sentences as either jumla filiyya or jumla ismiyya. A remarkable exception is Ibn Hišām al-Ansārī (d. 1360) who, in his famous book Muġnī l-Labīb, provides an elaborate discussion of Arabic sentence types. Ibn Hišām’s classification will be discussed in detail in section 5 below. As we shall see, he defined three sentence types (rather than two!) by the kind of constituent introducing the sentence. Thus, a sentence introduced by a nominal element is a jumla ismiyya. And among his examples of jumla ismiyya we find the sentence qāimun azzaydāni. Ibn Hišām was, indeed, aware of the controversy surrounding this sentence, indicating that it was accepted as a well-formed sentence by Axfaš and the Kūfans. As we saw above, qāimun az-zaydāni was adduced as an acceptable sentence in Arabic also by Zajjājī, but the latter did not classify it as jumla ismiyya.

12 Note, however, that in Zamaxšarī’s and Ibn Yaīš’s examples the construction at issue is itself a xabar following a mubtada. As we have seen, this is consistent with the principle of itimād.


yishai peled

In sum, then, the analysis of qāimun zaydun into a mubtada+fāil replacing a xabar appears to reflect a twofold attempt: (1) to make this kind of sentence conform to the principle of ibtidā; this is motivated by the fact that the first constituent is a nominal exhibiting a raf  case ending, and; (2) to apply the principle of verbal tadiya, given the verbal properties of the active participle. But, to the extent that this analysis holds, does not it follow (at least from the grammarians’ viewpoint) that qāimun zaydun represents a sentence type in its own right? To my knowledge, no such proposition has ever been advanced in medieval Arab grammatical literature.13

3. fīhā / fī d-dāri zaydun Only a small minority of the grammarians suggested that sentences such as fīhā zaydun (“In it there is Zayd”) should be regarded as representing a sentence type in its own right. They designated this type jumla zarfiyya, but differed on whether this term should or should not cover also sentences such as zaydun fīhā/fī d-dār (‘Zayd is in it / in the house”). I return to this later. At this point, let us examine the grammarians’ conception of this construction, starting with Sībawayhi. 3.1


In his bāb al-ibtidā (chapter 132), Sībawayhi does not develop any discussion of this sentence sub-type. But elsewhere in the Kitāb (I:170–171; cf. Levin 1987, 362, and Owens 1989, 224) he argues that in cases such as huwa xalfaka (“He is behind you”) it is the subject huwa that assigns the nasb case to xalfaka. Indeed, this is consistent with his argument (Kitāb

13 Badawi (2000, 8f.) claims that the grammarians recognized three types of sentences namely: filiyya, ismiyya and wasfiyya, introduced, respectively, by a verb, a noun and an adjective (a participle or otherwise). He emphasizes the use of different terms for the subject and predicate in each sentence type, indicating that in the jumlat wasf these are referred to as mubtada and fāil sadda masadd al-xabar. However, the term jumlat wasf has not been attested in the grammarians’ writings studied for the present work.

the medieval arabic theory of sentence types


I:239) that in zaydun muntaliqun, it is the subject that assigns raf  to the predicate.14 However, what about sentences such as fīhā zaydun, consisting of an adverbial/prepositional phrase followed by a nominatival noun phrase? Sībawayhi deals with these cases within the framework of his discussion of sentences such as fīhā abdu-llāhi qāiman and abdullāhi fīhā qāiman (“Abdullāh is in it, standing”—cf. Talmon 1993, 281). He starts his discussion analyzing qāiman as an accusatival xabar to abdu-llāhi. Then he goes on to indicate that abdu-llāhi in these cases: irtafaa bi-l-ibtidāi li-anna lladī dukira qablahu wa-badahu laysa bihi wa-innamā huwa mawdiun lahu wa-lākinnahu yajrī majrā l-ismi l-mabniyyi alā mā qablahu “is assigned the raf case by the ibtidā since the constituent occurring either before or after it [= the adverbial] is not it [= is not identical in reference with it], but rather signals its location. Yet [this adverbial] functions analogously to a noun built upon the [subject] preceding it”—(Sībawayhi, Kitāb I:222).

Sībawayhi, as can be seen, points out that fīhā is non-coreferential with abdu-llāhi, but rather refers to Abdullāh’s location (mawdi). But it is precisely this observation that underlies his endeavor to establish, first of all, the acceptability of fīhā abdu-llāhi/abdu-llāhi fīhā as a complete independent sentence. To this end he draws an analogy between fīhā abdu-llāhi and hādā abdu-llāhi (“This is Abdullāh”), claiming that in terms of completeness, the former, much like the latter, is a kalām mustaqīm (‘a correct sentence’) that h asuna [badahu] s-sukūtu (“that may appropriately be followed by silence”—Sībawayhi, Kitāb I:222 and 239–240). Similarly, abdu-llāhi fīhā is presented by him as analogous to abdu-llāhi axūka (“Abdullāh is your brother”): in both cases the second constituent is ‘built’ (mabnī) upon the first. As for abdu-llāhi, Sībawayhi states clearly that, whether preceding or following the prepositional phrase, it is assigned the raf  case by the ibtidā. But once a predicatival relationship is established between fīhā and abdu-llāhi, Sībawayhi reanalyzes the sentence assigning fīhā the function of xabar and abdu-llāhi the function of mubtada. The position of qāiman is then demoted to that of h āl (see Figure 1 below). He points out, however, that since fīhā represents the person’s location, fīhā abdullāhi is paraphrasable by istaqarra abdu-llāhi. In other words, fīhā

14 Note, however, that the amal in huwa xalfaka is presented by Sībawayhi as analogous to that in his model construction išrūna dirhaman (“Twenty dirhams”).


yishai peled mubtada








Figure 1

behaves analogously to the verb istaqarra. As we shall see in 3.2, the verb istaqarra, or otherwise the participle mustaqirrun, have since become the grammarians’ most common device for explaining the grammatical structure of sentences such as zaydun fīhā / fīhā zaydun. A further indication of Sībawayhi’s consideration of fīhā as a verb-like element is his statement (Kitāb I:223) that qāiman in the above sentence may, alternatively, be replaced by qāimun in the nominative. This, he explains, is the result of ‘abrogating’ (alġayta) fīhā. In the medieval grammarians’ writings, ilgā is normally used as a technical term denoting the annulling of amal. It is typically used with reference to potential awāmil, that is, elements that normally exercise amal upon other elements in the sentence (for discussion see, e.g. Peled 1992a, 150–152). One may infer, then, that Sībawayhi considered fīhā, in virtue of its acting analogously to istaqarra, as an āmil assigning nasb to qāiman in fīhā abdu-llāhi qāiman. As we shall see later, such adverbials as fīhā were considered by some early grammarians as an āmil assigning raf  to the following subject in such cases as fīhā zaydun. This view is typically attributed to the Kūfans. Yet, Sībawayhi (Kitāb I:223–224) then enters into an extensive discussion designed to exclude the possibility that fīhā in sentences such as fīhā abdu-llāhi qāimun is the āmil assigning raf  to abdu-llāhi. He draws an analogy between this sentence and bika abdu-llāhi maxūdun (“Abdullāh is fascinated by you”). He argues that an operator assigning case to an optional constituent (qāimun in the former sentence) has the same status (manzila) as an operator acting upon an obligatory constituent (maxūdun in the latter).15 Sībawayhi emphasizes that in both cases (as well as in similar ones adduced by him) the adjective is ‘built upon’ the noun, thus establishing a predicatival relationship between the two. The prepositional phrase, by contrast, is a laġw, i.e. a constituent that neither assigns nor receives amal. In fīhā


Indeed, in later grammatical writings, the model sentence bika zaydun maxūdun features regularly in the Basran arguments against the Kūfan claim that in fīhā zaydun it is fīhā that functions as the āmil assigning raf  to zaydun (cf. Ibn al-Anbārī, Insāf I:52–53).

the medieval arabic theory of sentence types


abdu-llāhi qāimun, fīhā is only designed to specify the location where the standing is taking place. Let us now return to the original construction fīhā abdu-llāhi to which Sībawayhi devotes a separate bāb, following his bāb al-ibtidā. In chapter 133 of the Kitāb, he deals with such cases as fīhā abdu-llāhi, tamma zaydun and ayna zaydun. Compared to his discussion above, Sībawayhi here seems to be less specific about the rāfi of zaydun: wa-lladī amila fīmā badahu h attā rafaahu huwa lladī amila fīhi h īna16 kāna qablahu “The operator assigning raf  to the following constituent is the same operator that assigned it the raf  case when that constituent was before it [i.e. before fīhā]”—(Sībawayhi, Kitāb I:239; cf. editor’s notes as well as the editions of Būlāq and Hārūn).17

Sībawayhi thus asserts that the operator assigning raf  to zaydun in fīhā zaydun is the same one that assigns the raf  case to zayd in zaydun fīhā, ‘where/when zayd occurs before it (i.e. before fīhā)’. Apparently, he leaves it to the reader to conclude that it is the ibtidā that functions as āmil in both cases. This could lead one to believe that for Sībawayhi, fīhā zaydun represents an inverted version of zaydun fīhā. Yet Sībawayhi does not make any explicit claim for taqdīm wa-taxīr in this particular case. He presents fīhā zaydun as a case in which fīhā: yaqau mawqia l-ismi l-mubtadai wa-yasuddu masaddahu “occupies the position of the mubtada and replaces it”—(Sībawayhi, Kitāb I:239; quoted also in Kouloughli 2002, 9).

As we saw in section 2, there were among later grammarians those who analyzed qāimun in qāimun zaydun as a mubtada followed by a fāil (replacing a xabar). It seems more than likely that they were influenced by the above Sībawayhian passage. Regarding ayna zaydun, Sībawayhi points out that ayna is paraphrasable by fī ayyi makān (“in which place?”), and emphasizes that ayna, as an interrogative, must obligatorily occur sentence-initially. In other words, fīhā zaydun is distinct from ayna zaydun only in that in the latter case the xabar occupies sentence-initial position obligatorily.


The words h aytu and h īna alternate in this position in two different versions of the text. 17 Talmon (1993, 283–284) confronts the long version cited here with a shorter one that to me looks rather obscure.


yishai peled

It is interesting to note that Sībawayhi does not include in his discussion such cases as fīhā rajulun (“In it there is a man”), where the subject, being indefinite, obligatorily follows the predicate (much like in ayna zaydun). Indeed, as we will see later, this construction received little attention from the grammarians, compared to fīhā zaydun. As we shall see, while the grammarians never failed to point out that the predicatesubject order in fīhā rajulun is obligatory, only a small minority of them regarded this construction as representing a sentence type in its own right. For the vast majority, fīhā rajulun, much like fīhā zaydun, represented an inverted jumla ismiyya. A detailed discussion of constructions with an obligatorily fronted xabar is provided in section 4 below. 3.2

The istaqarra/mustaqirrun hypothesis

Following Sībawayhi, the medieval grammarians continued to address themselves to the twofold problem posed by such sentences as zaydun fīhā and fīhā zaydun. This problem, as we have seen, consisted in establishing a predicatival relationship between the nominal and the adverbial/prepositional phrase, and accounting for the āmil assigning case to each. The suggested solution of positing an underlying linking element such as istaqarra/mustaqirrun gave rise to extensive discussions that developed into what may be referred to as the istaqarra/mustaqirrun hypothesis. As we saw in 3.1, the origins of this theory can be easily traced back to Sībawayhi’s Kitāb.18 Throughout the centuries, the istaqarra/mustaqirrun hypothesis has become the common strategy used by the medieval grammarians in order to account for such sentences as zaydun fī d-dār (or, for that matter, zaydun xalfaka, where xalfaka alternates with min xalfika—for discussion of the status of the zarf, see Levin 1987, 351–357).19 As for the grammatical status of the adverbial following the mubtada, here the grammarians differed. For Sībawayhi (Kitāb I:222), muntaliqun and fīhā were equally admissible as xabar for zaydun. In Ibn as-Sarrāj’s (Usūl I:62–63) view, the xabar in zaydun fī d-dār is the underlying mustaqir18 As we have seen, however, Sībawayhi used this device in dealing with sentences such as fīhā abdu-llāhi qāiman, to account for the nasb of qāiman; he did not employ it in cases such as fīhā zaydun or zaydun fīhā, pointing rather to the ibtidā, in both, as the āmil assigning raf  to zaydun. 19 For the Kūfan theory of xilāf (or muxālafa), see Astarābādī, Šarh I:214; Ibn alAnbārī, Insāf I:245–247; Mujāšiī, Šarh , 87, n. 216; and cf. Carter 1973).

the medieval arabic theory of sentence types


run which, when coreferring with the mubtada, is optionally omitted (h adf). The adverbial fī d-dār fills the xabar’s slot, acting as a complement to the omitted genuine xabar (for a similar view, see Ibn Usfūr, Šarh I:347). Zaydun xalfaka/fī d-dār is thus paraphrasable by zaydun mustaqirrun xalfaka/fī d-dār. Ibn as-Sarrāj points out that the omitted element normally conveys some meaning of existence, and is redundant because it is retrievable from the adverbial.20 He emphasizes that postulating an underlying constituent in such cases is obligatory, because, in itself, fī d-dār (or xalfaka) does not qualify as a predicate; that is, it does not predicate any quality of zayd (laysa bi-h adīt),21 but only specifies the location (mawdi) of Zayd. Jurjānī (Muqtasid I:274–275) argues that an adverbial, with either an explicit or implicit preposition, in principle presupposes a verb with which it is linked to form a syntactical unit (cf. Ibn al-Anbārī, Insāf I:246). Consequently, underlying fī d-dār, in zaydun fī d-dār, is the verbal clause istaqarra fī d-dār. This indeed represents the view of the majority of grammarians for whom positing an underlying (muqaddar) finite verb (istaqarra) is consistent with the clausal status of fī d-dār in sentences such as zaydun fī d-dār. Obviously, once established as the (head of the) xabar, istaqarra/yastaqirru is referred to as the āmil, assigning case to the following adverbial/prepositional phrase. However, the assumption of an underlying finite verb in cases such as zaydun fī d-dār was not universally accepted. The controversy here is linked to the fact that not all the grammarians regarded fī d-dār in zaydun fī d-dār as having a clausal status. Some grammarians posited a participle rather than a verb as the underlying element linking the adverbial/ prepositional phrase to the preceding mubtada. We have just seen that Ibn as-Sarrāj was one of the proponents of this hypothesis; indeed, he

20 Ibn Usfūr (Šarh I:347–348) emphasizes that using an adverbial/prepositional phrase as a xabar substitute is only admissible when the deleted element is recoverable from the surface construction—otherwise, the xabar should appear in full. Thus, for example, zaydun fī d-dār is only allowed if it is intended to convey the meaning mustaqirrun fī d-dār, because fī, signalling a receptacle (wiā), is compatible in meaning with istiqrār (‘staying’). If, however, zaydun fī d-dār is intended to convey the meaning of dāh ikun fī d-dār (“[Zayd] is laughing in the house”), then the word dāh ik must occur; for, unlike the meaning of ‘staying’, that of ‘laughing’ cannot be recovered from the preposition fī. Cf. Astarābādī, Šarh I:215, for linking elements like h āsil and kāin (‘be’); Levin 1987, 360. 21 Ibn as-Sarrāj’s use of the term h adīt in this case is significant, for it signalizes ‘predicate’ realized whether as fil or as xabar (cf. Goldenberg 1988, 46–49).


yishai peled

viewed sentences such as zaydun xalfaka as displaying a single-phrasedrather than a clausal xabar (Ibn as-Sarrāj, Usūl I:63). Astarābādī (Šarh I:215) cites Ibn as-Sarrāj and Ibn Jinnī as two grammarians who advocated the participle rather than the verb hypothesis, on the ground that the participle as a single phrase (mufrad) is compatible with the basic structure of the xabar. Another proponent of the participle hypothesis is Mujāšiī (Šarh , 87) who derives such sentences as zaydun amāmaka (“Zayd is in front of you”) and amrun min al-kirām (“ Amr is one of the honorable”) from the underlying (taqdīr) structures zaydun mustaqirrun amāmaka and amrun kāinun min al-kirām respectively.22 Mujāšiī makes it clear that for him an adverbial/prepositional phrase in xabar position has the status of, and is therefore a substitute for, a participle (not a clause). Postulating a personal pronoun implicit in the participle, Mujāšiī argues further that this pronoun moves to, and resides in, the adverbial/prepositional xabar occupying the position of the deleted participle (in Mujāšiī’s words: wa-afdā d-damīru lladī kāna fī smi l-fāil ilā n-nāib anhu fa-statara fīhi).23 For further discussion of the istaqarra/ mustaqirrun hypothesis, see Jurjānī, Muqtasid I:275ff; Ibn al-Anbārī, Insāf I:245–247; Ibn Usfūr, Šarh I:344, 349–351. We can see that the istaqarra/mustaqirrun hypothesis is used by the grammarians to fit the construction zaydun fīhā/fīhā zaydun into their theory of amal, and, by implication, to their binary system of sentence types. Once zaydun in both zaydun fīhā and fīhā zaydun was recognized as a mubtada, both constructions could be said to represent a jumla ismiyya, the latter being an inverted version of the former. The xabar, when following the mubtada, is presented as either clausal or phrasal, depending on whether one assumes yastaqirru or mustaqirrun to be the underlying linking element. In both cases, this element is made accountable for the case of fī d-dār (see Kouloughli 2002, 13–16, for further discussion). However, the istaqarra/mustaqirrun hypothesis was not universally accepted. And as we shall see, the alternative hypotheses had substantial implications for the theory of sentence types.

22 Astarābādī (Šarh I:215) claims that the underlying element is obligatorily deleted, rejecting such sentences as zaydun kāinun fī d-dār. He indicates that Ibn Jinnī did allow such constructions, but points out that there is no evidence to support this position. 23 This is evidently Mujāšii’s way of claiming a xabar status for the adverbial/prepositional phrase. Astarābādī (Šarh I:216–217) points to Fārisī and his followers as advocating the same hypothesis. But Sīrāfī is cited by him as claiming that the pronoun is deleted as part of the linking constituent.

the medieval arabic theory of sentence types 3.3


Abū Alī l-Fārisī

Abū ‘Alī l-Fārisī (d. 987) is considered to be one of the first grammarians who advanced in an explicit way and developed the idea of sentence types in Arabic (cf. Owens 1988, 36–37). He defined each type, and spelled out the problematic nature of the dichotomy verb+noun (jumla filiyya) versus noun+noun (jumla ismiyya).24 Indeed, he was the first to present a detailed argument with the conclusion that zaydun fī d-dār is neither a jumla filiyya nor a jumla ismiyya. Having defined the two basic sentence types in Arabic, Fārisī turns to concentrate upon the construction represented by zaydun fī d-dār. Indeed it looks as though Fārisī’s definition of the two basic sentence types is meant as an introduction to his discussion of this particular construction (Fārisī, Askariyya, 105–109). He starts by indicating that, although such sentences are composed of a nominal element (i.e. the two nouns) and a particle (the preposition), they do not have the same status as inna sentences, where the particle enters into a sentence made up of two nouns. This is because fī d-dār is non-coreferential with zaydun. And since zayd and fī d-dār are not identical in reference, they cannot be analyzed simply as subject and predicate. However, given that zaydun fī d-dār is definitely a well-formed sentence in Arabic, one must assume some underlying (muqaddar) linking element to account for the predicatival relationship between its two constituents. As we saw in 3.2, this linking element must inevitably be either a noun or a verb (a particle does not bear any reference). To the extent that either of these can be posited, a sentence such as zaydun fī d-dār must eventually belong either to the verb+noun or to the noun+noun type.25

24 Anxious to provide accurate and valid definitions, Fārisī (Askariyya, 104–105) points further to the option of a particle (h arf) entering into either of the two defined jumla’s, to form a kalām. What the reader is invited to infer is that the resulting construction is an independent grammatical sentence whose basic type (i.e. filiyya or ismiyya) is unaffected. He exemplifies this by sentences introduced by hal, inna, mā, qad and lam. (As a matter of fact, the same principle had already been stated by Ibn as-Sarrāj, Usūl I:43.) 25 Fārisī (Askariyya, 109) draws a comparison between the case in question and address (nidā) expressions. He argues that yā zaydu (‘O, Zayd!’), much like fī d-dāri zaydun, consists of nominal elements and a particle, and constitutes an independent sentence. The difference between the two, he maintains, is that in the case of yā zaydu a verbal element should be assumed, which renders the address expression a sub-type of a jumla filiyya, whereas in the case of zaydun fī d-dār/fī d-dāri zaydun no such element can be posited (see below).


yishai peled

However, from this point onward, Fārisī’s argument continues in a direction designed to prove that this is not the case. In other words, a sentence such as zaydun fī d-dār, while conforming to the general principle governing the production of well-formed sentences, constitutes an exception in that it does not fall into any of the two basic categories, jumla filiyya and jumla ismiyya. In Fārisī’s words: a-lā tarā anna l-kalāma wa-in kāna lā yaxlū mimmā dakarnā fī l-asli faqad sāra lahu l-āna h ukmun yaxruju bihi an dālika l-asli “Notice that although a sentence must obey the basic principles indicated by us, in this case there is an [overriding] rule leading the sentence away from the basic principles” (Fārisī, Askariyya, 105).

What Fārisī is now trying to prove is that neither a verb nor a noun can be posited as a linking element establishing a predicatival relationship between zaydun and fī d-dār, a relationship modeled on that obtaining between the subject and predicate of a regular verbal or nominal sentence. And if it can be proved that neither a verb nor a noun can be posited as an underlying linking element between zaydun and fī d-dār, the conclusion must be that sentences such as zaydun fī d-dār represent a sentence type in its own right. He starts (Fārisī, Askariyya, 105) by adducing the sentence inna fī d-dāri zaydan, were the particle inna enters into the sentence fī d-dāri zaydun (his choice of this construction rather than zaydun fī d-dār is significant, as will be seen below.) Then he makes the following two points: (1) An underlying verbal link cannot be assumed, because a verb would exclude the use of inna. In other words, while fī d-dāri zaydun may be preceded by yastaqirru, inna and yastaqirru are mutually exclusive: inna fī d-dāri zaydan is a perfectly grammatical sentence in Arabic, but *inna yastaqirru fī d-dāri zaydan is disallowed. (2) A linking noun cannot be posited either, because that would amount to assuming—falsely—that inna exercises its effect (amal) upon zaydan across the underlying linking noun (Fārisī, Askariyya, 108). Having disqualified both noun and verb as possible underlying linking elements in cases such as zaydun fī d-dār and fī d-dāri zaydun, Fārisī (Askariyya, 108) argues further that in such cases the adverbial constituent as such cannot be claimed to implement a verbal function. This,

the medieval arabic theory of sentence types


he maintains, is borne out by the fact that the adverbial may not be preceded by a circumstantial phrase; a sentence such as *qāiman fī d-dāri zaydun is inadmissible, but it would be allowed if fī d-dār had a verbal value (a sentence such as qāiman dah ika zaydun is considered as perfectly acceptable by the grammarians.) All the above boils down to a rejection of the istaqarra/mustaqirrun hypothesis, and that, in turn, leads Fārisī to the conclusion that sentences such as zaydun fī d-dār/fī d-dāri zaydun should be considered neither as jumla filiyya nor as jumla ismiyya; they must be thought of, rather, as representing a sentence type in its own right. Note, however, that Fārisī did not assign the type of sentence under discussion any special designation. The term jumla zarfiyya to which we will be introduced below was coined in a later period. But if one is supposed to assume no underlying element linking the two predicatival constituents in zaydun fī d-dār/fī d-dāri zaydun, what is then the assigner of raf  to zaydun in such cases? As we saw in 3.1, Sībawayhi, who was not committed to any theory of sentence types, had no problem presenting the ibtidā as raf  assigner to zayd in both zaydun fī d-dār and fīhā zaydun. But for Fārisī, making a similar claim would imply classifying fīhā zaydun as a jumla ismiyya. Regarding the raf  assigner in zaydun fī d-dār Fārisī does not develop any elaborate discussion, apparently because in such cases one would automatically refer to the ibtidā as the raf assigner. However, when it comes to fīhā zaydun, the construction on which he focuses his discussion, Fārisī presents a clear position as to the rāfi of zaydun. Having shown that neither a verb nor a noun can be posited as a linking element, and having proved, further, that the adverbial itself cannot be claimed to function as a verb, Fārisī (Askariyya, 108–109) refers the reader to Abū l-H asan [al-Axfaš] (d. 733), explaining that these are the reasons why Abū l-H asan regarded the adverbial per se as the rāfi when preceding a noun functioning as muh addat anhu (‘of whom the message is predicated’, ‘subject’). Notice that it is not the term fāil that is used with reference to that noun, but rather muh addat anhu, a term that cuts across all sentence types. As we shall see, however, later grammarians did not refrain from using the term fāil in this particular context. Obviously, attributing the assigning of raf  to an adverbial/prepositional phrase constitutes a serious problem for the theory of amal. Since the formulation of this theory, the grammarians always insisted that the function of case assignment is implemented by either a verb or a particle. Various elements, notably active participles and other adjectives


yishai peled

were claimed to have a ‘verbal force’. But in our case, as we have seen, Fārisī argued that fī d-dār was not verbal enough to allow a circumstantial phrase to precede it. So one might ask what it is that qualifies fīhā as raf assigner. To my knowledge, this point has never been clarified by the grammarians. And it is no wonder that the concept of jumla zarfiyya, where a predicatival prepositional phrase assigns raf  to the following subject, remained marginal and never became part of mainstream medieval Arab grammatical thinking. Anyhow, for Fārisī, Axfaš’s position regarding the raf assigner in fīhā zaydun constituted further support for his thesis that this construction represents a sentence type in its own right.

4. Obligatory fronting of the xabar 4.1

Formal aspects

A remarkable feature of the grammarians’ (including Fārisī’s) discussion of the adverbial/prepositional xabar is that they base their argument on such sentences as zaydun fī d-dār/fī d-dāri zaydun, where the definite subject may either precede or follow the predicate (cf. Kouloughli 2002, 10, n. 7). But for the third sentence type advocated by Fārisī, sentences such as indī mālun (“I have money”) would surely be a better example. For in this case the subject follows the predicate obligatorily; reversing the order is disallowed. Most of the grammarians adduce such sentences by way of illustrating an obligatorily fronted (taqdīm) xabar. However, within the framework of their grammatical discourse, obligatory fronting of the xabar presents a major conceptual problem. A xabar, by definition, must follow, not precede, the mubtada. The very concept of an obligatorily fronted xabar appears to conflict with two fundamental principles in medieval Arab grammatical theory : 1. The formal principle stipulating that the āmil precede the mamūl—see, e.g. Ibn Abī r-Rabī, Basīt I:587); 2. The functional principle placing the constituent representing the ‘given’ information before the one representing what is ‘new’ for the addressee. To the extent that the mubtada is identified with the ‘given’, and the xabar with the ‘new’, such sentences pose a serious problem. Obviously, in cases such as fī d-dāri rajulun, the grammarians could not present rajulun fī d-dār as the asl of fī d-dāri rajulun, since the former is disallowed as an independent sentence in Arabic. One may argue,

the medieval arabic theory of sentence types


therefore, that in fī d-dāri rajulun, as well as in other cases of obligatory fronting to be discussed below, the very concept of taqdīm is not normally intended in the sense that the above structure is the output of reversing the order of some basic structure in which the indefinite rajulun precedes fī d-dār. Rather, one must assume that taqdīm in such cases is used in the sense of ‘placing in initial position’, with no transformation involved. Ibn Yaīš (Šarh I:86) argues that rajulun fī d-dār is excluded (1) because it could be wrongly interpreted as a noun phrase (with fī ddār functioning as the attribute—sifa—of rajulun) rather than as a complete sentence (see also Ibn Usfūr, Šarh I:343), and (2) in order to avoid introducing a declarative (wājib) sentence by an indefinite noun.26 The idea of transformation in cases such as fī d-dāri rajulun was not, however, universally excluded. Not surprisingly, it was suggested, albeit in a rather idiosyncratic manner, by Ibn Jinnī, a grammarian noted for his originality and for frequently advancing dissenting arguments incompatible with mainstream medieval Arab grammatical thinking. In his Sirr sināat al-irāb, Ibn Jinnī (Sirr I:276) uses the concept asl marfūd (‘a rejected basic construction’) in dealing with cases which he regards as transformed constructions, but whose underlying structure (asl) is inadmissible. The principle of asl marfūd is not explicitly applied by him to rajulun fī d-dār. But in his Xasāis (I:300) he maintains that, while mubtada-xabar is the basic word order of a jumla ismiyya, a certain intervening factor (ārid) might impose the reversal of that order. The occurrence of an indefinite mubtada at the beginning of an affirmative sentence constitutes in Ibn Jinnī’s view such an ārid, a kind of contingency imposing the movement of the mubtada into the second position in the sentence (cf. Peled 1992b, 105–106).27 This is regarded 26 Astarābādī (Šarh I:232) maintains that the problem of ambiguity between xabar and sifa is acute, owing to the common occurrence of an adverbial in xabar position in Arabic. He cites, however, one case where an adverbial xabar follows an indefinite mubtada, pointing out that it is perfectly acceptable when the sentence is used as an exclamation (duā). Astarābādī also remarks that fronting a non-adverbial xabar to an indefinite mubtada does not eliminate the ambiguity. Thus, if you transform rajulun qāimun into qāimun rajulun, rajulun could be analyzed as xabar of qāimun or as an apposition (badal) to it, whereas a fronted adverbial in similar cases is bound to be interpreted as xabar, due to its nasb case, whether explicitly marked (lafzan), or understood by position (mah allan). 27 Note that an indefinite mubtada introducing a negative or interrogative sentence is readily accepted by Ibn Jinnī. Thus he admits (Xasāis I:300) sentences such as hal ġulāmun indaka (“Is there a boy with you?”) and mā bisātun tahtaka (“There is no carpet under you”), claiming that they are communicatively useful, as opposed to sentences such as rajulun indaka (“A man is with you”). The argument is that one can


yishai peled

by him as a corrective procedure (islāh al-lafz):28 the second position, he reminds us, is in principle the xabar’s position; and since the xabar is essentially indefinite, the indefinite mubtada now fills the appropriate slot as far as (in)definiteness is concerned. He emphasizes, however, that in the underlying theoretical (muqaddar) level the mubtada precedes the xabar (Ibn Jinnī, Xasāis I:318).29 There are other cases adduced by the grammarians as examples of obligatory fronting of the xabar. The first example presented by Ibn Abī r-Rabī (Basīt I:587) is the interrogative construction ayna zaydun (“Where is Zayd?”). This grammarian strived to demonstrate that such cases of obligatory xabar fronting do not violate the basic principle that the āmil must precede the mamūl. Regarding ayna zaydun he indicated that the basic underlying (asl) structure in this case is a-zaydun fī d-dāri am fī s-sūqi am fī l-h ānūti (“Is Zayd at home, in the market, or in the shop?”). The word ayna is an economy device designed to replace both the adverbials and the interrogative particle a-, as well as the particle am. It implements a xabar function in virtue of its being a replacement for the adverbials. At the same time, it is obligatorily fronted as a substitute for the interrogative particle a-. Another semantic component in ayna is that of specification (tayīn), formally represented in the basic underlying structure by the particle am. The transformation from the basic to the final surface structure thus proceeds in the following stages: First, the xabar ( fī d-dāri am fī s-sūqi etc.) is moved to a frontal position immediately following a- (a-fī d-dāri am fī s-sūqi . . . zaydun— the interrogative a- is always positioned sentence-initially). Then ayna is introduced to replace all the constituents preceding zaydun. It is thus the adverbial semantic component in ayna that warrants its occurrence in pre-mubtada’ xabar position, whereas the interrogative component accounts for the obligatoriness of the movement. Ibn Abī r-Rabī indicates that the same applies to other interrogatives such as matā, kayfa, negate the existence of, or pose a question with regard to, an unknown entity (mankūr lā yuraf ), while there is no communicative value in predicating of an unknown entity affirmatively. 28 Similarly, Astarābādī (Šarh I:232) regards the obligatory fronting of the xabar in cases such as fī d-dāri rajulun as a corrective (musah h ih ) procedure designed to handle the indefiniteness of the mubtada. 29 Ibn Jinnī (Xasāis I:319–320) then refers to sentences displaying an indefinite mubtada in sentence-initial position. He argues, however, that these are not predicatival sentences, in the sense that they are meant to express a wish or imprecation rather than convey information. Another case is explained by him as paraphrasable by a negative sentence (cf. n. 27 above).

the medieval arabic theory of sentence types


man and mā: kayfa axūka (“How is your brother?”), man axūka (“Who is your brother?”), and the like. Ibn Abī r-Rabī’s second case of obligatory xabar fronting is the construction fī d-dāri rajulun. This, however, is dealt by him in pragmatic rather than in purely formal terms, and will, therefore, be reviewed in the following sub-section. The third case is exemplified by Ibn Abī r-Rabī (Basīt I:588) by the sentence alā t-tamrati mitluhā zubdan (“on the date there is butter of an equal amount”). He indicates that the reverse order (mitluhā zubdan alā t-tamrati) is disallowed since -hā in mitluhā is an anticipatory pronoun in both lafz (surface) and martaba (underlying) structures, thus violating the rule of the anticipatory pronoun (al-idmār qabla d-dikr—cf. section 2 above). For further discussion of the relationship between anaphora and the position of the xabar, see Astarābādī, Šarh I:232–233. Introducing his fourth and final case of obligatory xabar fronting, Ibn Abī r-Rabī (Basīt I:588) cites the exceptive sentences mā fārisun illā zaydun (“No one is a horseman but Zayd”) and mā fī d-dāri illā amrun (“No one is in the house but Amr”). In these two sentences the subject nominal occurs sentence-finally and is dominated by the exceptive particle illā. Reversing the order in such cases, Ibn Abī r-Rabī points out, would not violate the rules of Arabic grammar, but result in a sentence different in meaning from the original one. The sentence mā fārisun illā zaydun assigns to Zayd, and only to him, the attribute of horsemanship. This sentence, however, is neutral as to whether or not Zayd possesses other qualities as well. But if the order of constituents is reversed so as to make the subject zaydun precede the predicate, the resulting sentence mā zaydun illā fārisun unmistakably excludes the possibility that Zayd possesses any quality beside horsemanship. Similarly, the sentence innamā fārisun zaydun is equivalent in meaning to mā fārisun illā zaydun, whereas innamā zaydun fārisun is synonymous with mā zaydun illā fārisun, which explains why a mubtada-xabar order is inadmissible in this related case as well. Ibn Usfūr (Šarh I:353) adds two more cases where the xabar is obligatorily placed sentence-initially: 1. When the mubtada is a nominalized clause introduced by anna: fī ilmī annaka qāimun (“It is known to me that you are standing”); 2. When the xabar is a kam al-xabariyya phrase: kam dirhamin māluka (“How many dirhams you have!”). The first of these two cases is dealt with also by Astarābādī (Šarh I:233–234) who cites Fārisī as claiming that the adverbial/prepositional phrase in such cases exercises amal (raf ) upon the following anna


yishai peled

clause with no supporting element (cf. 3.3 above; for itimād, see section 2 above). Astarābādī (Šarh I:233) explains that the reason for the obligatory fronting of the xabar (whether adverbial or not) in such cases is that if the anna clause were placed sentence-initially, the word nna could be misread as inna rather than anna. For, between the two particles, it is the former rather than the latter that is associated with the initial position in the sentence. Astarābādī points out further that if the xabar precedes the anna clause it is bound to be correctly analyzed as xabar to the following clause as a whole rather than as a fronted constituent governed by anna, because a constituent within the scope of inna/ anna cannot be preposed to either of these particles. Furthermore, once the adverbial/prepositional phrase is established as the xabar of the following clause, then the particle heading that clause will be easily read as anna, because a mubtada’ clause, being a noun clause, cannot be introduced by inna. For further discussion of this issue, see Ibn Yaīš, Šarh VIII:59–60. As we have just seen, fī d-dāri rajulun was only one item, and not necessarily the first, on the list of constructions presented by the grammarians as examples of obligatory fronting of the xabar. But it was apparently the most difficult to deal with in purely syntactic terms. For one thing, like the related fī d-dāri zaydun, it presented a challenge to the theory of amal. For another, the indefiniteness of the mubtada could not, in itself, constitute sufficient grounds for ruling out its occurrence in sentence-initial position (cf. Astarābādī, Šarh I:202–207 for a detailed discussion of cases of an indefinite mubtada in sentence-initial position). The grammarians’ main formal explanation for the obligatoriness of predicate-subject order in this case was that an adverbial/prepositional phrase following an indefinite nominal could be wrongly interpreted as an attribute rather than a predicate. But as we have just indicated, sentences with an indefinite mubtada do occur in Arabic. Indeed, as we will see shortly, the strongest argument against *rajulun fī d-dār was pragmatic rather than syntactic. 4.2

fī d-dāri rajulun—pragmatic aspects

When examining the construction fī d-dāri rajulun in terms of information structure, most of the grammarians appreciated that sentences of this kind represent a special case. Indeed they recognized that in these cases it is the definite adverbial/prepositional phrase, occupying sentence-initial position that represents the given information, whereas the

the medieval arabic theory of sentence types


following nominatival indefinite phrase signals the new information, and not the other way around (see, e.g. Ibn Yaīš, Šarh I:86–87).30 Ibn Yaīš (Šarh I:86) argues that, judging by the meaning of sentences such as laka mālun (“You have money”), it is the definite complement of the preposition (in this case the pronoun -ka) that represents the muh addat anhu (‘what the sentence is about’), even though, formally ( fī l-lafz), it is the nominatival noun that implements that function.31 In support of his claim he indicates (1) that laka mālun is paraphrasable by anta dū mālin, and (2) that an indefinite nominal is inadmissible in the position of the complement of the preposition: *li-rajulin mālun is disallowed (lam yakun kalāman—‘is not an [acceptable] sentence’). Another attempt to relate the construction in question to a mubtadaxabar order was made by Ibn Abī r-Rabī. This grammarian presents a number of cases where the xabar is obligatorily fronted (cf. 4.1 above), taking great pains to demonstrate that fronting the xabar in these cases is wholly justified, if not on purely formal, then on functional/semantic, grounds. Regarding fī d-dāri rajulun, Ibn Abī r-Rabī states the following: fa-hādā yulzimu t-taqdīma wa-lā yajūzu taxīruhu fa-taqūla rajulun fī d-dāri li-annahu lā yubtadau bi-n-nakirati wa-innamā jāza l-i-btidāu hunā bi-n-nakirati li-anna l-maqsūda l-ixbāru an-i d-dāri bi-annahā maskūnatun wa-laysati n-nakiratu l-maqsūdata bi-l-ixbāri wa-kāna l-aslu an taqūla ad-dāru mamūratun bi-rajulin tumma arādū l-i-xtisāra fa-qālū fī d-dāri rajulun wa-alzamū d-dāra t-taqdīma li-annahā l-muxbaru anhā bi-l-h aqīqati “in such cases [the xabar] is preposed obligatorily. It may not be postposed to yield rajulun fī d-dār, because an indefinite noun may not fill a mubtada position. In our case, the mubtada32 may be indefinite, because the intention is to predicate of the house that it is inhabited, rather than to predicate of the indefinite noun. Underlying [our sentence] is the sentence ad-dāru mamūratun bi-rajulin. But for the sake of brevity, they say

30 According to Talmon (1993, 285–287), the idea is already attested in ninth century writings where the locative is typically referred to as sifa and the nominatival noun following it as xabar as-sifa. 31 This obviously rests on the assumption that in a sentence containing only one nominatival noun, it is this noun that should be construed as the muh addat anhu. This term, while referring literally to a pragmatic function, signals in the grammarians’ usage, the subject, irrespective of sentence type; its counterpart h adīt signals the predicate (see Goldenberg 1988, 46–49, for discussion). 32 Notice that yubtadau and ibtidā are both construed in this case as ‘used as mubtada’ or ‘implementing a mubtada function.’


yishai peled fī d-dāri rajulun. They obligatorily prepose the [phrase fī] d-dār, because it is really the house that is predicated of [i.e. the topic]” (Ibn Abī r-Rabī, Basīt I:587–588).

In other words, the sentence fī d-dāri rajulun is paraphrasable by ad-dāru mamūratun bi-rajulin (“The house is inhabited by a man”), where ad-dāru obviously functions as mubtada. Ibn Abī r-Rabī explains that the meaning intended by fī d-dāri rajulun is that the house is inhabited; it is ad-dār of which something is predicated, not the indefinite rajulun. The version fī d-dāri rajulun, Ibn Abī r-Rabī maintains, is preferred to ad-dāru mamūratun bi-rajulin for reasons of economy. The constituent ad-dār is obligatorily preposed, for it is the one that implements the function of muxbar anhu, which obviously entails that rajulun functions as xabar. (For the terms muxbar/muxbar anhu, see Goldenberg 1988, 46–51). A different, indeed exceptional, view on this issue is held by Jurjānī (Muqtasid I:308–309). Sentences such as indī mālun (“I have money”) are dealt with by him within the framework of his discussion of sentences with an indefinite mubtada. For Jurjānī, the mubtada in this particular case, albeit indefinite, implements the same pragmatic function as a definite sentence-initial mubtada. Mālun in the above sentence implements the function of mubtada li-ajli h usūli l-ixtisāsi fī l-xabari id kullu wāh idin lā yalamu anna indaka mālan “because the xabar signals some specification [regarding the mubtada], for it is not common knowledge that you have money” (Jurjānī Muqtasid I:308).

For Jurjānī, then, indī in its sentence-initial position makes the same contribution to the communicative value (ifāda) of the sentence as would a xabar occupying a post-mubtada position. Much like other grammarians, however, Jurjānī argues that in such cases the xabar is obligatorily fronted because mālun indī would be wrongly interpreted as a noun phrase, with indī analyzed as a complement (sifa) to mālun (see 4.1 above). Most of the grammarians, however, did not follow this line of thought. Rather, they discerned a discrepancy between the syntactic analysis of sentences such as fī d-dāri rajulun into xabar and mubtada, and the pragmatic functions of muxbar anhu and xabar implemented by fī ddār and rajulun respectively. They did, however, emphasize that in terms

the medieval arabic theory of sentence types


of definiteness sentences like laka mālun conform to the principle that the definite constituent, representing the given information, should precede the indefinite noun signalling the new information. In any case, the grammarians could not accept an analysis of fī d-dāri rajulun into a mubtada followed by a xabar, on the ground that if a sentence of this kind is introduced by inna, the noun phrase rajul automatically takes the nasb case, which marks it unmistakably as subject. For discussion, see Kouloughli (2002, 16–17), and his references.

5. Ibn Hišām al-Ansārī’s tripartite division 5.1


Returning now to the question of the raf assignment in cases such as fīhā zaydun, most of the grammarians from Sībawayhi onwards regarded ibtidā as the operator assigning raf to zaydun. This, however, was by no means universal, as we have seen. Indeed, Ibn al-Anbārī (Insāf I:51–55) attributes this position to the Basrans while presenting the other (much less common) position as Kūfan. We learn that the Kūfans, as well as the Basran grammarians Mubarrad and Axfaš (cf. 3.3 above), regarded the adverbial/prepositional phrase in the above case as the assigner of raf to the following noun zaydun.33 Ibn al-Anbārī maintains that both the Basrans and the Kūfans resort to Sībawayhi for support for their respective claims. The Basrans obviously point to chapter 133 in the Kitāb, where, as we have seen, Sībawayhi refers to the ibtidā (though without using the actual term) as the āmil assigning raf  to zaydun. The Kūfans, for their part, draw upon a number of cases where, according to Sībawayhi, an adverbial assigns raf  to a following noun (Ibn al-Anbārī, Insāf I:52). It is upon such cases, Ibn al-Anbārī argues, that the Kūfans base their claim that in fīhā zaydun it is fīhā that should be regarded as the rāfi of zaydun. Astarābādī (Šarh I:218) indicates that the analysis of zaydun in fī d-dāri zaydun as fāil of the adverbial/prepositional phrase was advanced by the Kūfans, as well as by Axfaš in one out of two statements he made

33 This position is clearly evidenced in Farrā’s Maānī l-Qurān (e.g. I:195–196; III:133). See Talmon (1993, 279) for further details and references.


yishai peled

on the subject.34 The argument in both cases is that the adverbial/prepositional phrase has a ‘verbal force’ (manā l-fil), analogously to qāim in qāimun zaydun. In Astarābādī’s view, the Kūfans’ position emanates from their categorical objection to xabar fronting, irrespective of whether the xabar is a phrase or a clause. This objection, in its turn, is designed to forestall the occurrence of an anticipatory pronoun. To this, however, Astarābādī offers an outright rejection, claiming, as could be expected, that the anticipatory pronoun occurs only in the surface structure; in the basic structure the mubtada precedes the xabar with no anticipatory pronoun involved (cf. section 2 above). As for Axfaš, according to Astarābādī he did not object to xabar fronting, and (in his other statement) indeed regarded the ibtidā as the assigner of raf to zaydun in fī d-dāri zaydun. Axfaš’s position here is divergent from the one cited above. It rests upon two assumptions: 1. The verbal force of the adverbial/prepositional phrase is weaker than that of the adjective. 2. The acceptability of the construction fī dārihi zaydun (“in his house Zayd [is located]”). In this case, the option of analyzing zaydun as fāil is excluded on the ground that it would lead to an unacceptable anticipatory pronoun (since the pronoun, under this analysis, would be cataphoric both in the lafz and manā configurations—cf. Astarābādī, Šarh I:202). The debate, as could only be expected, concludes with the Basrans having the upper hand. There is evidence to suggest, however, that, as in many other cases, the position dismissed as ‘Kūfan’ represented a view that was much more widely accepted than the ‘mainstream’ medieval grammarians would have us believe. As we shall see in the next sub-section, it remained viable centuries later, in the writings of one of the most prominent medieval grammarians, Ibn Hišām al-Ansārī. The relevance of this issue to our discussion is clear: it is closely related to the question of whether or not sentences such as fīhā zaydun represent a sentence type in its own right.

34 Ibn al-Anbārī (Insāf I:51) adds Mubarrad to the proponents of this kind of analysis (and cf. Ibn Usfūr, Šarh I:158–159).

the medieval arabic theory of sentence types 5.2


Ibn Hišām’s categorization and definitions

Coming now to the 14th century grammarian Ibn Hišām al-Ansārī (d. 1360), we find both a critical discussion of the concept jumla as compared with kalām, and an unambiguous division into three sentence types. Recall that Fārisī’s starting point was that there were two basic sentence types in Arabic. This was followed by an elaborate argument designed to prove that zaydun fī d-dār represented a sentence type in its own right. For Ibn Hišām the tripartite division is an established linguistic fact, and he makes it the starting point of his discussion. He does not, however, ignore the problems raised by this division, as we shall see. Ibn Hišām (Muġnī, 492) starts by defining the concepts kalām and jumla. The latter is defined by him as a construction made up of either fil + fāil or, otherwise, mubtada + xabar. Then, however, he distinguishes three types of jumla, ismiyya, filiyya and zarfiyya, introduced respectively by a noun, a verb and an adverbial (zarf) or a prepositional phrase ([ jārr wa-] majrūr).35 Being aware of the problem arising from his defining each type by the initially occurring constituent, he remarks (Muġnī, 492) that the definitions refer only to predicative constituents (musnad and musnad ilayhi). Thus, the sentence a-zaydun axūka (“Is Zayd your brother?”) and the conditional clause in qāma zaydun (“If Zayd stands up”) are, respectively, ismiyya and filiyya, even though the first noun in the former and the verb in the latter are each preceded by a particle. Once the problem of the particles is settled, Ibn Hišām appears to be remarkably uncompromising in applying his principle, that Arabic sentence types must be defined by the initial predicative constituent. And this, indeed, leads to some conspicuous peculiarities. The jumla ismiyya, for instance, is exemplified by him by the following three sentences: zaydun qāimun (“Zayd is standing”), hayhāti l-aqīq (“How far is the ravine!”) and qāimun az-zaydāni (“Standing are the two Zayds”). While the first of the three sentences is straightforward, the other two are not. The word hayhāt in the second example is regarded by the medieval grammarians as ism fil representing, as the term suggests, a special word category whose members are considered as neither nouns nor verbs (see, e.g. Astarābādī, Šarh III:165ff. for details). Such asmā

35 A fourth type, jumla šartiyya, which, as he indicates, was proposed by Zamaxšarī, is rejected by Ibn Hišām on the ground that the conditional clause should be categorized as jumla filiyya.


yishai peled

al-af āl, in constructions like hayhāti l-aqīq, are normally described as occupying a verb position, with the implication that the following noun implements the function of fāil. However, since hayhāt as an ism fil is viewed as a special kind of noun (rather than as a kind of verb), a sentence introduced by it must, according to Ibn Hišām’s rigid principle of classification, be regarded as a jumla ismiyya rather than as jumla filiyya. The special problems relating to the third case (qāimun az-zaydāni) have already been discussed in section 2 above, and will not be repeated here. It is, however, noteworthy that for Ibn Hišām the fact that qāimun az-zaydāni is introduced by a participle (viewed by the grammarians as a nominal element) was sufficient for classifying this sentence as jumla ismiyya; the fact that qāimun is followed by a noun in the dual form did not require, in his eyes, any further argument or elaboration. The jumla filiyya is illustrated by Ibn Hišām by six sentences introduced by a verb. Out of these, five are straightforward: he uses qāma, yaqūmu, qum, duriba and zanna to demonstrate that a jumla filiyya may be introduced by any finite verb form. Specifically, by adducing the sentence zanantuhu qāiman (“I believed him to be standing”), Ibn Hišām makes the point that a cognitive verb may, like any other verb, introduce a jumla filiyya. As is well known, in medieval Arab grammatical theory, from Sībawayhi onwards, cognitive verbs such as zanna (zanna wa-axawātuhā—‘zanna and sisters’) are presented as analogous to inna and kāna (and their respective ‘sisters’), in that they enter into (yadxulna alā) sentences composed of a mubtada and xabar, nullifying in the process the abstract operator ibtidā, and assuming in its stead the function of a formal āmil assigning case to both nominal constituents in the sentence (see, e.g. Sībawayhi, Kitāb I:6; Ibn Yaīš, Šarh VII:77–78). In any case, the status of zanna as a verb was never disputed. By contrast, the verbal status of kāna was a matter of controversy among the grammarians. The vast majority of grammarians considered kāna as a semantically deficient (nāqis) verb,36 in the sense that it lacks the semantic component of action.37 As such, its only function is to signal the time of the nominal sentence into which it ‘enters’. Significantly, 36 Unless signalling existence, in which case it is labelled kāna at-tāmma ‘complete kāna’, and treated as an ordinary verb. 37 This had been recognized already by Sibawayhi, although in his account (Sibawayhi, Kitāb I:16) he depicted kāna as analogous to daraba in terms of transitivity (tadiya): like daraba it takes two nominal complements, one in the nominative, the other in the accusative.

the medieval arabic theory of sentence types


while acknowledging the predicative relationship between zaydan and qāiman in zanantu zaydan qāiman, the grammarians analyze the two nominal constituents as direct objects (maf ūl) of zanna. By contrast, in kāna zaydun qāiman, zaydun and qāiman are normally analyzed in terms of a jumla ismiyya, namely as mubtada and xabar, or otherwise, as ism kāna and xabar kāna—respectively. True to his rigid definitions of the three sentence types, Ibn Hišām includes in his examples of jumla filiyya the sentence kāna zaydun qāiman (“Zayd was standing”). And this already seems to run counter to the mainstream conception of jumla filiyya. Indeed, the grammarians do not normally bring kāna into their discussions of sentence types. But given their special treatment of this verb, it is highly unlikely that they would have classified kāna zaydun qāiman as jumla filiyya just by the verbal morphology of kāna. But the most interesting for our present discussion is Ibn Hišām’s jumla zarfiyya. This he illustrates by the two sentences a-indaka zaydun (“Is Zayd with you?”) and a-fī d-dāri zaydun (“Is Zayd in the house?”), where the first predicatival constituent is an adverbial and a prepositional phrase, respectively. Ibn Hišām points out that sentences such as these can qualify as jumla zarfiyya only idā qaddarta zaydan fāilan bi-z-zarfi wa-l-jārri wa-l-majrūri lā bi-listiqrāri l-mah dū  fi wa-lā mubtadaan muxbaran anhu bihimā “if you assume zayd to be a fāil [acted upon] by the adverbial/prepositional phrase, not by a deleted [verb/participle conveying the meaning of] istiqrār, and [only if] you do not analyze zayd as a mubtada for which the adverbial/prepositional phrase serves as xabar” (Ibn Hišām al-Ansārī, Muġnī, 492).

The significance of this passage lies in that it seems to suggest that Ibn Hišām’s sentence-type definitions were not as rigid as they appeared to be when we looked at his definitions and illustrations of the jumla ismiyya and the jumla filiyya. It now turns out that for him, the predicative constituent that comes first in the sentence is not in itself the only criterion for determining the type of sentence. Rather, for a predicatival constituent to qualify as sentence-type identifier it must act as āmil upon the second predicatival constituent. Another important point to note is that in both of Ibn Hišām’s examples the adverbial/prepositional phrase is preceded by the interrogative particle a-. This may be taken to suggest that by the time of Ibn Hišām the principle of itimād, which, as we have seen (section 2 above), can be


yishai peled

traced back to Sībawayhi, had already been firmly established in medieval Arab grammatical thought. When Ibn Hišām states that a sentence can only qualify as a jumla zarfiyya if the second constituent functions as fāil to the first, illustrating this with examples displaying the interrogative particle a- preceding the first predicative constituent, one is bound to conclude that for him a jumla zarfiyya is a sentence whose first predicatival constituent is an adverbial/prepositional phrase, where that phrase acts as a verb, thus assigning raf to the following constituent on the strength of the principle of itimād. Ibn Hišām then goes on to make a critical remark directed at Zamaxšarī. He indicates that Zamaxšarī exemplified jumla zarfiyya by the phrase fī d-dār in zaydun fī d-dār (cf. Ibn Yaīš, Šarh I:88). This position, he argues, is based: alā anna l-istiqrāra l-muqaddara filun lā ismun wa-alā annahu h udifa wah dahu wa-ntaqala d-damīru ilā z-zarfi bada an amila fīhi “on [the assumption] that [the underlying word conveying] istiqrār is a verb, not a noun, and that that verb was deleted alone while the pronoun implicit in it moved to the adverbial phrase, after [the verb] had exercised amal upon the adverbial” (Ibn Hišām al-Ansārī, Muġnī, 492).

Recall that zaydun fī d-dār was described by Fārisī as representing a sentence type in its own right on the ground that, in his view, neither a verb nor a noun could be posited as a linking element between zayd and fī d-dār. Here Zamaxšarī is quoted as elevating fī d-dār to the status of a clausal xabar ( jumla zarfiyya). This is done, so the argument goes, by positing an underlying verb that is deleted while the pronoun implicit in it is transferred to the adverbial, the verb having exercised amal upon that adverbial. In other words, the clausal status of fī d-dār stems from the pronoun it receives from the deleted verb istaqarra. What we see here is, indeed, another attempt to account for the predicatival relationship between zaydun and fī d-dār, two non-coreferential elements, as well as for the irāb of the xabar. However, this attempt is based on the istaqarra hypothesis (3.2 above), and that is precisely the reason why it is rejected by Ibn Hišām. Zamaxšarī’s analysis is incompatible with Ibn Hišām’s conception of jumla zarfiyya. For Ibn Hišām, once an underlying verb is assumed, the clause should be regarded as a jumla filiyya; exactly as, when one posits a xabar-mubtada relationship between fī d-dār and zayd, the sentence must be considered as jumla ismiyya. This will be further clarified below. In 3.3 we pointed to Fārisī’s reference to Axfaš, who had attributed to the adverbial/prepositional phrase the function of operator assigning

the medieval arabic theory of sentence types


raf to the nominal following it. He did not, however, refer to that nominal explicitly as fāil. The term muh addat anhu which he used, signals in medieval Arabic grammatical literature the subject of the sentence, whether a fāil or a mubtada. Indeed, the specific grammatical status of the nominatival constituent, as determined by the āmil assigning it the raf case, never ceased to be a topic of debate among the grammarians. Yet one thing emerges quite clearly. The analysis of zaydun as fāil in both fī d-dāri zaydun and qāimun zaydun, normally attributed to the Kūfans and Axfaš (see section 2 above), always comes up when these two constructions are discussed. It was never abandoned. However, of these two constructions, it is only fīhā/fī d-dāri zaydun that was considered, albeit by a small number of grammarians, as a sentence type in its own right. The reason for this should by now be clear. The opening predicative constituent in each of the three sentence types was regarded as an operator (āmil) assigning case to the following constituent(s): sentence types were unmistakably correlated with amal types. And since the participle (qāimun) could not be viewed as other than a verbal or a nominal element, it could not be regarded as introducing a sentence type in its own right. By the same token, a sentence such as zaydun fī ddār, introduced as it is by a nominatival noun, could only be defined as jumla ismiyya. The concept jumla zarfiyya was by and large associated with cases where a zarf could be claimed to be a āmil assigning raf  to the nominal following it. As we will see in the next sub-section, it was Ibn Hišām, an eminent proponent of the tripartite division, who also appreciated and spelled out the problems arising from the actual notion of sentence types in Arabic, whether two or three. 5.3


As we have seen throughout, the problems the grammarians encountered in categorizing Arabic sentences stemmed from the fact that their conception of sentence types was deeply embedded in the theory of amal. This is manifested also in the way these problems are illustrated by Ibn Hišām (Muġnī, 493–497). He offers an illuminating discussion of ten cases where a sentence can be construed as either a jumla filiyya or a jumla ismiyya, or, otherwise, raise a controversy among grammarians as to the right categorization. Significantly, no case is cited as an unambiguous jumla zarfiyya. Since the basic arguments recur throughout his discussion, I will review only four of his examples that, I believe, well illustrate the problematic aspects of the traditional categorization of sentence types.


yishai peled

Let us start with Ibn Hišām’s fourh example mādā sanata (“What have you done?”). He points out that this sentence may be paraphrased as either mā lladī sanatahu, or as ayya šayin sanata. Since lladī sanatahu is a nominalized constituent, the sentence, according to the first paraphrase, must be categorized as a jumla ismiyya. Ibn Hišām indicates that the first constituent mā is analyzed as a fronted xabar by Axfaš, and as mubtada by Sībawayhi. By contrast, the proponents of the second paraphrase, ayya šayin sanata, would categorize the same sentence as a jumla filiyya, analyzing ayya šayin as a fronted direct object. (And see, further, Ibn Hišām’s discussion of the sentence mādā sanatahu.) Ibn Hišām’s sixth example reads qāmā axawāka (“Your two brothers stood up”). This sentence is presented by him as acceptable, subject to specific types of analysis. (To what extent this construction was in actual use in medieval Arabic is immaterial for the present discussion). First, the sentence could be categorized as jumla filiyya if (1) the ending -ā in qāmā is interpreted as a dual-marking particle (h arf tatniya), much as the -t in qāmat Hindun is analyzed as a feminine marker (and not as a pronoun); or alternatively if (2) the ending -ā is interpreted nominally and the following axawāka is analyzed as apposition (badal) to it. Second, qāmā axawāka may be categorized as jumla ismiyya with a fronted xabar (with the ending -ā interpreted nominally and axawāka analyzed as a postposed mubtada). Note that Ibn Hišām does not mention the possibility of analyzing qāmā axawāka as a jumla filiyya with axawāka functioning as fāil to qāmā (luġat akalūnī l-barāġīt see above, section 2, n. 6). The seventh example presented by Ibn Hišām is nima r-rajulu zaydun (“What a nice man is Zayd”). This sentence, he explains, may be viewed as an inverted jumla ismiyya, with nima r-rajulu functioning as a preposed xabar to zayd. Under an alternative analysis, however, zaydun could function as xabar to a deleted mubtada. Ibn Hišām argues that under the latter analysis, nima r-rajulu zaydun consists of two asyndetically coordinated clauses, the first one (nima r-rajulu) verbal, and the second nominal. But perhaps the most interesting is Ibn Hišām’s second example, where he makes the following statement regarding a-fī d-dāri zaydun and a-indaka amrun: fa-innā in qaddarnā l-marfūa mubtadaan aw marfūan bi-mubtadain mah dū  fin taqdīruhu kāinun aw mustaqirrun fa-l-jumlatu ismiyyatun dātu xabarin fī l-ūlā wa-dātu fāilin muġnin an-i l-xabari fī t-tāniyati wa-in qaddarnāhu fāilan bi-staqarra fa-filiyyatun aw bi-z-zarfi fa-zarfiyyatun

the medieval arabic theory of sentence types


“if we analyze the nominatival constituent as mubtada, or otherwise, as a nominal assigned the raf  case by a deleted mubtada such as kāinun or mustaqirrun, then the sentence should be considered as nominal, with a xabar under the first analysis, or with a fāil replacing the xabar under the second. If, however, we analyze it [i.e. zaydun or amrun] as fāil of [an underlying] istaqarra, then the sentence is verbal; if [the operator assigning raf to the fāil] is the adverbial, then the sentence should be considered as a jumla zarfiyya” (Ibn Hišām al-Ansārī, Muġnī, 494).

Here Ibn Hišām offers four ways for analyzing sentences such as a-fī d-dāri zaydun, correlating each analysis with a different sentence type. These are the four options as presented in the above passage, in Ibn Hišām’s order: 1. zaydun could be analyzed as mubtada. This would imply that the sentence is a jumla ismiyya, with the adverbial/prepositional phrase implementing the function of (a preposed) xabar. 2. We could posit an underlying mubtada, such as kāinun or mustaqirrun, assigning the raf  case to zaydun. In this case zaydun would implement the function of fāil replacing (muġnin an) the xabar. The sentence under such an analysis would be regarded, according to Ibn Hišām, as jumla ismiyya. 3. zaydun could be analyzed as fāil assigned the raf  case by the underlying verb istaqarra. In this case the sentence would be considered as jumla filiyya. 4. If, however, we analyze zaydun as a fāil receiving its raf case from the preceding adverbial/prepositional phrase, then the sentence should be regarded as jumla zarfiyya. In Ibn Hišām’s view, then, a fāil is not necessarily preceded by a finite verb. But it is only when the operator assigning raf to the fāil is a finite verb (whether explicit or underlying) that the sentence may be conceived of as jumla filiyya. When the raf  assigner is a participle (whether explicit or underlying) or an adverbial/prepositional phrase, the sentence must be conceived of as a jumla ismiyya in the first case, and as a jumla zarfiyya in the second, even though the actual use of the term fāil suggests that the participle and the adverbial / prepositional phrase in such cases behave analogously to a verb. As can be seen, Ibn Hišām presents the four options without any attempt to ‘defend’ his categorization. His analyses are consistent with his definitions of the three sentence types (see 5.2 above), and manifestly reflect the controversies relating to the constructions in question.


yishai peled

The proponents of the first option would presumably regard a-fī d-dāri zaydun as the inverted version of a-zaydun fī d-dār. The occurrence of the interrogative a- in this case is irrelevant, as is the case also under the third analysis, where the sentence is presented as an unmistakable jumla filiyya. Indeed, positing an underlying verb like istaqarra in order to account for the raf  case of zaydun in sentences of this kind was common practice among the grammarians, as we saw in 3.2. What is really remarkable in Ibn Hišām’s third analysis is that it leads to the important conclusion that under a certain analysis a sentence such as a-fī d-dāri zaydun could be conceived of as jumla filiyya. Under the second analysis, zaydun implements the function of fāil following a deleted mubtada, thus occupying a xabar position. The adverbial/prepositional phrase, under this as well as under the third analysis (see above), would be analyzed as an adjunct. Obviously, the second analysis is reminiscent of the analysis of (a-)qāimun zaydun into a mubtada followed by a fāil replacing the xabar, as we saw in section 2. Note that, unlike Sībawayhi (Kitāb I:239; and cf. 3.1 above), Ibn Hišām does not view the prepositional phrase as occupying a mubtada position. Rather, the mubtada in this case is an underlying participle. Here, at any rate, a sentence whose subject is labeled fāil is categorized as jumla ismiyya. As we have already indicated, the fourth analysis is consistent with Ibn Hišām’s theory of three sentence types, each defined by the predicative constituent introducing the sentence (and acting as āmil upon the second constituent). But we have already seen (3.3 above) that it remains unclear how an adverbial/prepositional phrase can function as a verb assigning raf to a following nominal constituent. Following Sībawayhi, it was often argued that such a phrase may act analogously to a verb when preceded by a ‘supporting’ element such as the interrogative particle a- (itimād—cf. sections 2, 4.1 above). But does that in itself warrant categorizing the construction a-fī d-dāri zaydun as representing a sentence type in its own right? The concept of jumla zarfiyya does not seem to have been seriously discussed in the writings of the medieval Arab grammarians. Evidently, the vast majority found it difficult to fit the concept of jumla zarfiyya into their theory of amal. Indeed, this is manifested even in Ibn Hišām’s position, which does not present a-fī ddāri zaydun as a straightforward jumla zarfiyya. Rather, it makes it clear that the actual identification of a sentence as jumla zarfiyya is essentially dependent upon conceiving the adverbial/prepositional phrase as a āmil assigning raf to the following nominal. The other two types,

the medieval arabic theory of sentence types


in contrast, could be determined straightforwardly, since both the verb and the ibtidā were established awāmil in the medieval theory of amal. Hence the grammarians’ adherence to the binary system of jumla filiyya and jumla ismiyya.38

6. Summary Since the grammarians’ theory of sentence types grew out of, and has always been closely related to, their theory of amal, it is not surprising that in elaborate discussions of Arabic sentence types, particularly those of Fārisī and Ibn Hišām, problems relating to the categorization of certain constructions were couched in terms of case assignment (amal). The basic types of jumla filiyya and jumla ismiyya are shown throughout to represent two types of amal: verbal tadiya and ibtidā. Sentences such as qāimun zaydun and fī d-dāri zaydun/rajulun are shown to be problematical in terms of amal. With regard to qāimun zaydun, we have seen that many grammarians advocated the rather awkward analysis of mubtada+fāil sadda masadd al-xabar. This was designed to deal with the essentially nominal nature of the participle occurring sentence-initially, as well as with its verb-like behavior in this particular case. Apart from Ibn Hišām who regarded this construction as an example of jumla ismiyya, the proponents of the above analysis did not commit themselves to any clear-cut categorization of this particular structure. Regarding fī d-dāri zaydun/rajulun, the very fact that this construction displays an adverbial/prepositional predicative constituent in sentence-initial position, gave rise to the hypothesis that it represents a sentence type in its own right, a jumla zarfiyya. And it comes as no surprise that this was associated with the hypothesis that in such cases it is the adverbial/prepositional phrase that assigns raf  to the nominal constituent following it. Obviously, this hypothesis and the long established istaqarra/mustaqirrun hypothesis were mutually exclusive. In 3.3 we saw Fārisī’s attempt to refute the istaqarra/mustaqirrun hypothesis, arguing from the theory of amal. This was his line of defending a tripartite sentence-type system.

38 For a modern study advocating a three-type division, see Kouloughli (2002, 21–24) who argues that sentences such as fī d-dāri rajulun/zaydun (referred to by him as ‘locative sentences’) should be viewd as representing a sentence type in its on right, since they exhibit a number of syntactic and semantic properties not shared by ‘regular’ topiccomment sentences.


yishai peled

For him, indeed, claiming that fīhā zaydun represented a sentence type in its own right was tantamount to presenting fīhā as the assigner of raf  to zaydun. But such an argument could never be accepted by the majority of grammarians, because it was considered a major violation of a central principle of the theory of amal. The vast majority of grammarians did find the istaqarra/mustaqirrun hypothesis a convenient tool for fitting such constructions as zaydun fīhā and fīhā zaydun into their theory of amal. The question of sentence type was apparently secondary. Once they established the status of mubtada to zaydun in both cases, they could argue that both represent a jumla ismiyya, irrespective of whether the underlying element linking the adverbial/prepositional phrase to the mubtada is a verb or a participle. Fīhā zaydun was thus conceived of as the inverted version of zaydun fīhā. To Ibn Hišām, as we have seen, this was unacceptable. For, if underlying fīhā zaydun is the structure yastaqirru fīhā zaydun, it follows that zaydun is assigned the raf  case by the underlying verb occupying sentence-initial position. And Ibn Hišām, indeed, drew the conclusion following from that assumption, namely that under the above analysis fīhā zaydun must be categorized as jumla filiyya. But this position of Ibn Hišām’s is clearly exceptional in the medieval grammatical literature. The majority of grammarians, conforming to the theory of amal, did posit istaqarra as the underlying case assigner to zaydun, but they never went as far as categorizing fīhā zaydun as jumla filiyya. Rather, fīhā zaydun has always been regarded an ibtidā construction, clearly associated with jumla ismiyya.

7. References 7.1 Primary sources Astarābādī, Šarh = Radī ad-Dīn Muhammad b. al-H asan al-Astarābādī, Šarh Kāfiyat Ibn al-H ājib. Emil Badī Yaqūb, ed. Beirut: Dār al-Kutub al-Ilmiyya, 1998. Fārisī, Askariyya = Abū Alī l-Fārisī, al-Masāil al-askariyyah. Muhammad aš-Šātir Ahmad Muhammad Ahmad, ed. Cairo: Matb aat al-Madanī, 1982. Farrā, Maānī = Abū Zakariyyā Yahyā b. Ziyād al-Farrā, Maānī l-Qurān. Ahmad Yūsuf Najātī and Muhammad Alī an-Najjār, eds. Cairo: ad-Dār al-Misriyya li-t-Talīf wa-t-Tarjama, 1955–1972. Ibn Abī r-Rabī, Basīt = Ubaydallāh b. Ahmad b. Ubaydallāh Ibn Abī r-Rabī, al-Basīt fī šarh jumal az-Zajjājī. Ayyād b. ‘Īd at-Tabītī, ed. Beirut: Dār al-Ġarb al-Islāmī, 1986. Ibn al-Anbārī, Asrār = Abū l-Barakāt Abd ar-Rahmān b. Muhammd b. Abī Saīd alAnbārī, Kitāb Asrār al-Arabiyyah. Muhammad Bahjat al-Baytār, ed. Damascus: Matbūāt al-Majma al-Ilmī al-Arabī bi-Dimašq, 1957. Ibn al-Anbārī, Insāf = Abū l-Barakāt Abd ar-Rahmān b. Muhammd b. Abī Saīd alAnbārī, Kitāb al-Insāf fī masāil al-xilāf bayn an-nahwiyyīn al-basriyyīn wa-l-kūfiyyīn. Muhammad Muhī ad-Dīn Abd al-H amīd, ed. Beirut: Dār al-Fikr, n.d.

the medieval arabic theory of sentence types


Ibn Aqīl, Šarh = Bahā ad-Dīn Abdallāh Ibn Aqīl, Šarh Ibn Aqīl alā alfiyyat Ibn Mālik. Muhammad Muhyī ad-Dīn Abd al-H amīd, ed. n.p: Dār Sab, n.d. Ibn Hišām, Muġnī = Jamāl ad-Dīn Ibn Hišām al-Ansārī, Muġnī l-labīb an kutub alaārīb. Māzin al-Mubārak and Muhammad Alī H amdallāh, eds. Beirut: Dār al-Fikr, 1985. Ibn Jinnī, Sirr = Abū l-Fath Utmān Ibn Jinnī, Sirr sināat al-irāb. H asan Hindāwī, ed. Damascus: Dār al-Qalam, 1985. Ibn Jinnī, Xasāis = Abū l-Fath Utmān Ibn Jinnī, al-Xasāis. Muhammad Alī an-Najjār, ed. Cairo: al-Haya l-Misriyya l-Āmma li-l-Kitāb, 1986–1988. Ibn as-Sarrāj, Usūl = Abū Bakr Muhammad b. Sahl Ibn as-Sarrāj, Al-Usūl fī n-nahw. Abd al-H usayn al-Fatlī, ed. Beirut: Muassasat ar-Risāla, 1987. Ibn Usfūr, Šarh = Alī b. Mumin b. Muhammad b. Alī Ibn Usfūr, Šarh jumal azZajjājī. Sāhib Abū Janāh, ed. Mosul: Ihyā at-Turāt al-Islāmī, 1980–1982. Ibn Yaīš, Šarh = Muwaffaq ad-Dīn Yaīš b. Alī Ibn Yaīš, Šarh al-Mufassa l. Cairo: Maktabat al-Mutanabbī, n.d. Jurjānī, Muqtasid = Abd al-Qāhir al-Jurjānī, Kitāb al-muqtasid fī šarh al-Īdāh . Kāzim Bahr al-Marjān, ed. Baghdad: Dār ar-Rašīd li-n-Našr, 1982. Mujāšiī, Šarh = Abū l-H asan Alī b. Faddāl al-Mujāšiī, Šarh uyūn al-irāb. Abd alFattāh Salīm, ed. Cairo: Dār al-Maārif, 1988. Sībawayhi, Kitāb = Abū Bišr Amr b. ‘Utmān Sībawayhi, Al-Kitāb. Hartwig Derenbourg, ed. Hildesheim and New York: G. Olms, 1970. Zajjājī, Jumal = Abū l-Qāsim Abd ar-Rahmān b. Ishāq az-Zajjājī, Kitāb al-jumal fī nnahw. Alī Tawfīq al-H amad, ed. Beirut: Muassasat ar-Risāla, Dār al-Amal, 1988. 7.2

Secondary sources

Badawi, El-Said M. 2000. “Ray fī maānī l-irāb fī fush ā t-turāt: H ālat al-jumla lismiyya.” Diversity in Language: Contrastive Studies in Arabic and English Theoretical and Applied Linguistics, Z. M. Ibrahim, S. T. Aydelott and N. Kassabgy, eds., 1–20. Cairo: The American University in Cairo Press. Carter, Michael G. 1973. “Sarf et H ilāf: Contribution à l’histoire de la grammaire arabe.” Arabica 20–3, 292–304. ——. (ed.). 1981. Arab Linguistics: An introductory classical text with translation and notes (Širbīnī’s Nūr al-sajiyyah). Amsterdam: John Benjamins. Goldenberg, Gideon. 1988. “Subject and Predicate in Arab Grammatical Tradition.” Zeitschrift der Deutschen Morgenländischen Gesellschaft 138–1, 39–73. ——. 2002. “Two Types of Phrase Adjectivization.” “Sprich doch mit deinen Knechten aramäisch, wir verstehen es!”: 60 Beiträge zur Semitistik, Festschrift für Otto Jastrow zum 60. Geburtstag, W. Arnold and H. Bobzin, eds. 193–208. Wiesbaden: Harrassowitz. Kouloughli, D. E. 2002. “On locative sentences in Arabic.” Zeitschrift für arabische Linguistik 41, 7–26. Levin, Aryeh. 1985. “The distinction between nominal and verbal sentences according to the Arab grammarians.” Zeitschrift für arabische Linguistik 15, 118–127. ——. 1987. “The views of the Arab grammarians on the classification and syntactic function of prepositions.” Jerusalem Studies in Arabic and Islam 10, 342–367. ——. 1989. “What is meant by akalūnī l-barāġītu?” Jerusalem Studies in Arabic and Islam 12, 40–65. Owens, Jonathan. 1988. The Foundations of Grammar: An Introduction to Medieval Arabic Grammatical Theory. Amsterdam: John Benjamins. ——. 1989. “The Syntactic Basis of Arabic Word Classification.” Arabica 36, 211–234. Peled, Yishai. 1992a. “Amal and Ibtidā in Medieval Arabic Grammatical Tradition.” Abr-Nahrain 30, 146–171. ——. 1992b. “Cataphora and Taqdīr in medieval Arabic grammatical theory.” Jerusalem Studies in Arabic and Islam 15, 94–112.


yishai peled

Talmon, Rafael. 1993. “Two Early ‘non-Sībawaihian’ Views of amal in Kernel-Sentences.” Zeitschrift für arabische Linguistik 25, 278–288. ——. 1997. Arabic Grammar in its Formative Age: Kitāb al-Ayn and its attribution to Xalīl b. Ah mad. Leiden: E. J. Brill.


1. Introduction: the Qurān The history of the Arabic language is indelibly marked by the fact that the Qurān has made Arabic a prophetic language with its own holy book and a worldwide appeal. Arabic had been a mantic language in pre-Islamic times. When the poet recited verses or when the soothsayer uttered his sayings, reciter and listener were sure that behind his voice there was another voice. This voice ‘really’ speaking was that of a higher power. With the Qurān, the mantic voice behind the voice of the Prophet Muhammad became in the believer’s ear the voice of the one and only God. Muslim dogma and the consensus of the unlearned considered the Qurān to be the direct, undiluted Arabic word of God. The status of Classical Arabic, the standardization of Arabic including the development of Modern Standard Arabic, the diglossia Standard Arabic versus Arabic dialects, the nature of the Arabic vocabulary, Arabic orthography, Arabic style and vocabulary—all are unthinkable without the ‘Qurānic fact’. The history of the Arabic language down to our times cannot be written without constant reference to the Qurān. Conversely, the Qurān is deeply marked by its ‘Arabness.’ The Qurān is the first literary document in Arabic. It is also the first Arabic document to mention the Arabic language. In contradistinction to the attitude of the Jewish Bible and the Christian New Testament toward their own linguistic forms, the language of the Qurān is an important topic of Qurānic self-reflection. While neither the Bible nor the New Testament refer to their Hebrew (or Aramaic) and Greek literary forms, the Arabic character of the Qurān is in its self-view a cornerstone of its divine quality. The adjective arabī in the Qurān always refers to the language, never to a tribe or a social class. The term as an ethnic designation is also


stefan wild

very rare in Jāhiliyya poetry.1 In the Qurān, arabī refers always to the language of the holy text. It occurs eleven times in the Qurān, and only in Sura’s traditionally dated to the Middle and late Meccan period. In six passages, the adjective arabī is a qualification of the word Qurān, a word meaning primarily ‘recitation, reading aloud . . .’ (Sura 12:1–2; 20:113; 39: 27–28; 41:1–4; 42:7; 43:1–4). The conjunction lisān arabī ‘Arabic tongue’ occurs three times in the Qurān (Sura 16:103; 26:195, 46:12), and is used to describe the language of the Qurān. The conjunction h ukm arabī (Sura 13:37) ‘an Arabic judgment’ also refers to the holy text, and in 41:44, in which the possibility of a Qurān ajamī ‘a non-Arabic Qurān’ is dismissed, the adjective arabī again refers to the holy text. The Arabic quality of Qurānic revelation could scarcely be more solidly established. On the other hand, the Qurān seems indifferent to the linguistic shape of preceding revelations. In the Qurān, the only language used is Arabic. Arabic is also the only language mentioned by name. The Qurān does not specify in which languages Nūh, Ibrāhīm, Ismāīl, Mūsā, Īsā or other prophets and messengers spoke to their peoples or in which languages their holy books might have been. God speaks Arabic to Adam and his wife, Satan whispers in Arabic (Sura 20:120), the angels and the jinn speak Arabic (Sura 72:1–15), Moses addresses the Pharaoh, Joseph addresses the Egyptian minister’s wife in Arabic, Jesus speaks Arabic from his cradle, D ū l-Qarnayn and the People of the Cave—they all use Arabic. Every single soul is made to speak Arabic at the Day of Judgment, animals like the ant (Sura 27:18) or the hoopoe (Sura 27:22), even inanimate entities like Hell (Sura 50:30) speak Arabic. Everybody and everything that speaks in the Qurān must necessarily speak Arabic, because Arabic is the only language used throughout the Qurān. But the intention of the text is in no way to convey that all mankind throughout history shared and will share the same language. I do not know of any exegete who concluded from the Qurānic accounts that the language used between Mūsā and the Egyptian Pharaoh was Arabic or that the language used between the Egyptian notable’s wife and her lady-friends must have been Arabic, or that the Messiah spoke in Arabic—just because the Qurān reproduces their words in Arabic. It is a different matter for Adam and Ismāīl (see below).


Cf. Agha and Khalidi: Poetry and Identity 70.

divine, prophetic, and heroic arabic


On the other hand, the Qurān does mention the existence of languages other than Arabic and emphasizes that the plurality and variety of human languages is a sign of divine grace. The divine creation of different languages is as important as the creation of heaven and earth: wa-min āyātihī xalqu s-samawāti wa-l-ardi wa-xtilāfu alsinatikum wa-alwānikum īnna fī dālika la-āyātin li-l-ālimīn And of His signs is the creation of heavens and earth and the variety of your tongues and hues. Surely in this are signs for people who know (Sura 30:22).

Even as God created man and woman, heaven and earth, he created different colors (alwān, or: kinds of human skin) and different human languages (cf. the enumeration in Sura 30:20–25). The existence of different languages is one of God’s “signs for those who know.” Unless the word āya is here taken to mean a portent of warning, such a view seems difficult to reconcile with the myth of the Babylonian tower (Gen 11: 1–9), according to which the origin of a multitude of human languages is divine punishment. Sura 30:22 may even be an inter-textual stab at the narrative of the Babylonian Tower. There were, however, extra-Qurānic traditions that preserved the motif that the difference between human languages was due to an act of divine wrath (p. 196). A second important element in the Qurānic linguistic outlook flows forth from this esteem of other languages. In the course of history, God sent each messenger (rasūl) to a specific people (qawm), and this messenger brought the divine message to that people in its language. And We have sent no messenger save with the tongue of his people that he might make all clear to them wa-mā arsalnā min rasūlin illā bi-lisāni qawmihī li-yubayyina lahum (Sura 14:4).

The primary raison d’être of the Qurān is that the Prophet Muhammad’s message was in Arabic. Whereas the other prophets and messengers had been sent earlier with a message in the languages of their peoples, who did not speak Arabic, Muhammad is sent to the Arabs. The Arabic language vouchsafes the understanding of the Arabic-speaking audience. For a great part of the Qurān, the fact that this revelation was in Arabic was the most important difference between Muslim revelation and all previous revelation. These Qurānic elements influenced a theological-linguistic scenario that gave rise to a particular image of linguistic history. The claim that


stefan wild

Arabic as the language of the Qurān had become a divine language, led to the construction of a religious pre-history of Arabic. There were three competing strands of knowledge that were interwoven to form a comprehensive imaginaire of the history of Arabic and by implication of human languages in general: 1. the philosophical question of the origin of language in general; 2. exegesis of the Qurān, including the Qisas al-anbiyā literature, i.e., reports on earlier prophets, which depended heavily on Biblical and Aggadic material; and 3. historical traditions about pre-Islamic history, including the genealogies of the Arab tribes. One of the main issues was the origin of Arabic, an issue often framed as the question ‘Who was the first human being to speak Arabic?’ MuslimArab scholarship tried sometimes to disentangle, sometimes to combine and often to harmonize the heterogeneous strands of philosophical speculation, exegetical H adīt, and genealogical traditions. In many cases, however, contradictory reports were just left standing side by side. An extensive and useful overview of much of the Arabic material was given by as-Suyūtī in his Muzhir.2 The following remarks will outline the main lines of this colorful and often contradictory linguistic yarn.

2. Adam’s Arabic and Nūh’s Syriac The basic divide between Arab philosophers and theologians concerned with the origin of language was whether it rested on human convention (isti lāh , tawādu) or on a divine act of revelation (ilhām, wahy, tawqīf ).3 The Platonic controversy whether human language was what it was physei or thesei, by nature or by imposition, here took a new form. A further problem was whether, theologically speaking, Arabic had a special linguistic status that set it apart from all other languages.

2 Andrzej Czapkiewicz, The Views of the Medieval Arab Philologists on Language and its Origin in the Light of Al-Suyuti’s Al-Muzhir, Krakow 1989. Czapkievicz’s translations are sometimes hard to understand. 3 For the difference between tawqīf and ilhām, cf. J. van Ess, Theologie und Gesellschaft IV 325.

divine, prophetic, and heroic arabic


An important part of tradition seems to be in general agreement that Adam, who is in Islam a prophet, was the first human being to speak Arabic. The ‘revelationists’ based their view mainly on the early exegesis of Qurān 2:31 wa-allama Ādama l-asmāa kullahā “and He (God) taught Adam all names.” Weiss explains: ‘According to at -Tabarī, the majority of early exegetes, including the noted Companion and tafsīr-authority Ibn Abbās, interpreted this verse as meaning that God taught Adam the Arabic names of all existing things. This meant that Adam’s language was revealed by divine teaching. It also clearly implied that God taught Adam language in its entirety.4 According to Ibn Jinnī (d. 392/1002), most Mutazilites, on the other hand, taught that all languages including Arabic rested on convention, not on revelation.5 Even those who accepted the revelationist theory could argue that Sura 2:31 meant that Adam was taught the names of all things created in all languages, in Arabic, of course, but also in Persian, Syriac, Hebrew, Greek, and all other languages as Ibn Jinnī explains. Adam and his offspring spoke all these languages. But in the course of time, Adam’s offspring spread over the world and each of them stuck to one of these languages until this language became his main language and other languages eluded him because of his lack of contact with them (qad fussira hādā bi-an qīla inna llāha subh ānahū allama Ādama asmāa jamīi l-maxlūqāti bi-jamīi l-lug āti l-arabiyyati wa-l-fārisiyyati wa-lsuryāniyyati wa-l-ibrāniyyati wa-r-rūmiyyati wa-g ayri dālika min sāiri l-lug āti fa-kāna Ādamu wa-waladuhū yatakallamūna bihā tumma inna waladahū tafarraqū fī d-dunyā wa-aliqa kullun minhum bi-lug atin min tilka l-lug āti fa-g alabat alayhi wa-dmahalla anhu mā siwāhā li-budi ahdihim bihā, Ibn Jinnī, Xasāis I 41). This model safeguarded the priority of Arabic; Arabic as the language of the divine word must be prior to all other languages, whether their origin was revealed or due to convention. But according to some, all other languages were also revealed to Adam by God. Others with a different timetable thought that different linguistic communities emerged only after the Flood. The idea that Adam had known all human languages found a parallel in the belief of some Shiites in the 9th century that the Prophet Muhammad had known


Weiss, Muslim Discussions 37. aktaru ahli n-nazari alā anna asla l-lug āti innama huwa tawādu wa-sti lāh lā wahy wa-tawqīf (Al-Xasāis I 40–41). 5


stefan wild

all languages.6 The plurality of human languages is sometimes seen as due to a benevolent divine intervention: “He (God) shaped every nation conforming to a language in which He made them speak and which He made easy for them.”7 God had made His Arabic revelation ‘easy’ for the Arabs (e.g. Sura 54:17); in a similar way, earlier revelations had been made in other languages in order to make them ‘easy’ for their listeners. The most widely accepted historical model that explained the existence of languages other than Arabic was built on the supposition that Arabic had been at a certain time the universal language. The idea of a universal common language can also be found in Gen 11:1, where, however, this language does not have a name. For most Arabic scholars, Arabic was the earliest existing language.8 But this universality had at some point in history come to an end, and decadence, corruption, and confusion ( fasād, tah rīf, tabalbul) had set in. This statement was hard to reconcile with the Qurānic assertion that God himself had created the variety of human languages (Sura 30:22). The reason for the development of other languages was nevertheless frequently seen either in a divine act of punishing Adam or mankind, or in a general confusion (balbala), which ‘mixed up’ what had been one common human language in a gradual process of ‘corruption’ (tah rīf ). Tah rīf and tabalbul were sometimes expressly linked to a punitive act of God, sometimes they seem to be seen more as a general tendency of history. According to Ibn Asākir’s chronicle, Arabic was Adam’s language in Paradise until he disobeyed God. Then God deprived him of Arabic and he started speaking Syriac, evidently considered a lesser language or a corrupted form of Arabic. When Adam repented, however, God gave the Arabic language back to him.9 A second less ‘revelationist’ tradition was the following: The first language with which Adam came down from Paradise was Arabic. When the contact with Arabic became far and distant, Arabic was corrupted (h urrifa) and became Syriac (suryāniyya) which is a Nisba to ard sūrā or Suryāna. This is the land of al-Jazīra, where Nūh and his


Van Ess, Theologie und Gesellschaft IV 324. az-Zubaydī, Tabaqāt an-nahwiyyīn (ed. Muhammad Abū l-Fadl Ibrāhīm), Cairo 1954/1373, 1 jabala kulla ummatin mina l-umami alā lug atin antaqahum bihā wayassara lahum. 8 lug atu l-arabi asbaqu l-lug āti wujūdan, as-Suyūtī , Muzhir I 28, 4. 9 as-Suyūtī, Muzhir I 30, 10. 7

divine, prophetic, and heroic arabic


people lived before the flood. Syriac resembled Arabic, but it was corrupted (muh arraf ), and Syriac was the language of all people in Nūh’s ark except for one man, whose name was Jurhum, and his language was still the original Arabic. When they left the ark, Īram b. Sām married one of his (Jurhum’s) daughters. From them the Arabic language came down on his offspring Aws Abū Ād, Ubayl, Jāir Abū Tamūd, and Jādis. Ād was given the name of Jurhum, because he was their maternal forefather. And Syriac stayed with the offspring of Arfaxšad b. Sām until it reached Yašjub b. Qahtā n. He was in Yemen. There the Banū Ismāīl settled and the Banū Qahtā n learned from them the Arabic language.10

It is interesting to see that the term tah rīf, which is the usual word denoting the falsification of the revealed scriptures by Jews and Christians, is here used to explain a fact of imagined linguistic history : the ‘corruption’ and subsequent loss of Arabic. According to this account, Nūh’s language was not Arabic, but Syriac. Arabic had survived only with Jurhum and his tribe. Abdalmalik b. H abīb (d. 238/852) developed a stemma of prophetic languages. According to him, these prophetic languages are Arabic, Syriac, and Hebrew. All the sons of Israel spoke Hebrew, the first to speak it was Ishāq. Syriac was the language of five prophets: Idrīs, Nūh, Ibrāhīm, Lūt and Yūnus. And twelve of the Prophets spoke Arabic: Adam, Shīt, Hūd, Sālih, Ismāīl, Šuayb, al-Xidr, ‘the three in surat Yāsīn’ i.e. the three nameless messengers who were sent as the ‘Companions of the City’ (ash āb al-qarya, sura 36:13ff.), Xālid b. Sinān al-Absī, the legendary forerunner of the Prophet Muhammad, and the Prophet Muhammad himself.11 Others added more and more details. According to al-Azraqī, Nūh had eighty men with their families on his ark. When the ark came to a halt on Mount Jūdī, Nūh founded a village called Tamānūn (‘eighty’). The next morning, they found that their tongues/languages had been confused and that there were now eighty languages, one of which was Arabic. They did not understand each other any more.12 This account offers no explanation for the emergence of these eighty languages, but

as-Suyūtī, Muzhir I 30,-6 according to Abdalmalik b. H abīb. Claude Gilliot and Pierre Larcher “Language and Style of the Qurān.” EQ 3, 109– 135, especially “The mythical narratives on the superiority of Arabic” 118ff. 12 fa-btanā qaryatan wa-sammāhā Tamānin fa-asbah ū dāta yawmin wa-qad tabalbalat alsinatuhum alā tamānīn lughatan ih dāhā l-arabiyyatu (al-Azraqī, Axbār Makka, /ed. F. Wüstenfeld) vol. 1, 20. 10 11


stefan wild

is based on the idea that Nūh and his followers originally did not speak Arabic. In other accounts, Ibrāhīm is said to have spoken two languages, Syriac and Hebrew. When he fled from the persecution by Namrūd in Kūta in the land of Babylon, his language was Syriac (lisānuhū suryānī). Namrūd wanted to catch him and gave order that anybody speaking Syriac should be arrested and brought before him. But when Ibrāhīm crossed the Euphrates from H arrān, God ‘changed’ his language and Ibrāhīm miraculously started speaking Hebrew (ibrānī). He escaped because his persecutors did not know this language.13 A further version closer to the narrative of the Babylonian Tower is reported by at- Tabarī: Namrūd has a high building (sarh ) made, until it reaches the sky. God destroys the building and on that day the languages of people become confused from fright. They then speak in seventythree languages. This is why the place was called Bābil. The language of mankind before that was Syriac.14 A different kind of etiological explanation is linked to Yarub (see below section 4). The etymological connection between balbala and the name Bābil is often invoked and finds its way into Arabic lexicography.15 This connection between the Arabic word for ‘confusion’ (balbala) and the name of Bābil is prefigured in the Bible, in which the name of the city of Babel is linked to the Hebrew verb balal, also meaning ‘to confuse’ (Gen 11:7). The number of existing languages is given in different ways. According to al-Masūdī, there were seventy-two languages divided under Nūh’s sons: the descendants of Sām spoke nineteen languages, the descendants of H ām seventeen, and those of Yāfit  thirty six.16 As far as I can see, nobody in pre-modern times ever claimed that the prophet Ibrāhīm spoke Arabic—although his unequalled importance for the link between earlier monotheistic religions and Islam, especially through his role in building the Kaba, would have made such a claim attractive. Abraham is called a Muslim in the Qurān (h anīf muslim


Ibn Sad, Tabaqāt ed. Eugen Mittwoch, Leiden 1905, I:1 21, 14; a similar tradition in Ibn Mutarrif at -Tarafī, Qisas al-anbiyā no. 124. 14 fa-tabalbalat alsunu n-nāsi min yawmaidin mina l-fazai fa-takallamū bi-talātati wa-sabīna lisānan fa-li-dālika summiyat Bābil wa-innamā kāna lisānu n-nāsi qabla dālika s-suryāniyyata., At- Tabarī, Tārīx I 322. 15 Lisān al-Arab s.v. bll: summiyat ardu Bābil li-anna llāha h īna arāda an yuxālifa bayna alsinati banī Ādama baatha r-rīh an fa-h ašarahum min kulli ufuqin ilā Bābil fabalbala llāhu bihā alsinatahum tumma farraqahum tilka r-rīh u fi l-bilād. 16 al-Masūdī, Murūj ad-dahab, ed. Barbier de Meynard I 78.

divine, prophetic, and heroic arabic


sura 3:67)—but never an Arab. This claim was put forward only by zealous Arab nationalists. When in 1999 Pope John Paul II announced his plan to visit Ur in Iraq, members of the ruling Bat-party criticized that the Pope in his announcement had not spoken of Ibrāhīm as “an Arab.” They argued: “Ibrāhīm was born in the land of the Arabs and he lived in Iraq, in the Sumerian city of Ur. When he was chased out of his place of birth because of his monotheist creed, Ibrāhīm started his combative journey through the lands of the Arabs, H arrān, Palestine, Egypt, and Mecca . . .” ‘Combative’ was a favorite attribute of praise in the Bat-party. The Batīs do not expressly identify Arabic as Ibrāhīm’s language—even though the reader is forced to conclude that the language that the combative Arab Ibrāhīm spoke must have been Arabic.17 Such a linguistic myth could be elaborated. There is frequently a tendency in modern popular Arab discourse to call ‘Arabic’ what is elsewhere called ‘Semitic,’ in order to extol the importance of Arabic. The Canaanite tribes preceding the Israelite settlers are then called ‘Arab Canaanite’ tribes, and the variant of Akkadian to be found in the Ebla texts is called ‘Arabic.’18 For the claim that the Aramaic of the Nabatean inscriptions is ‘really’ Arabic, see below.

3. Ismāīl’s Arabic Next to Adam, the most important figure with whom the introduction of Arabic is firmly connected is Ibrāhīm’s son Ismāīl. In one of the most important foundational Muslim narratives, God orders Ibrāhīm to migrate to Mecca with young Ismāīl and the latter’s mother Hagar. Ibrāhīm builds the Kaba together with Ismāīl. There they meet members of the (Arabic-speaking) Jurhum tribe. Ismāīl grows up with their children, learns how to shoot (the bow) and to speak in their language, and he takes a Jurhum-wife.19 Another report says, without reference to where and how, Ismail learned Arabic: ‘Sarah gave Hajar to Ibrāhīm, he slept with her and she bore him Ismāīl, who was Ibrāhīm’s eldest son. His name used to be Išmōēl which was later arabicized (wa-kāna smuhū

17 Amatzia Baram, ‘Der moderne Irak, die Baath-Partei und der Antisemitismus’ in Jahrbuch für Antisemitismusforschung 12 (2003) 99–119 p. 114. 18 Welt des Islams 21 /1982, 240f. 19 Ibn Qutayba, al-Maārif, ed. Tarwat Ukāša, Cairo 1960, 34.


stefan wild

šmwyl fa-uriba).’20 This Arabic tradition closely follows the Biblical account (Gen 21:20)—even the bow is already there—and preserves in Išmōēl a dialectical Hebrew form of the classical Hebrew Yišmāēl. Abraham had given his son a Hebrew name, because he spoke Hebrew. The link between Ismāīl and the Arabic language became part of the awāil-literature. The Prophet had said: all Arabs descend from the offspring of Ismāīl b. Ibrāhīm. The question who was the first to speak Arabic could, therefore, be answered and the answer is contained in a short didactic dialogue in which an Abū Jafar Quray b. Uqba b. Bašīr asks and a Muhammad b. Alī answers: Q: Who was the first to speak Arabic? A: The first to speak Arabic was Ismāīl b. Ibrāhīm, when he was thirteen years old. Q: And what did people speak before that? A: Hebrew. Q: And what was God’s language that was sent down to his messengers and servants? A: Hebrew.21

Here, Adam is not even mentioned as the first recipient of Arabic. And God’s language of pre-Arabic revelation is simply Hebrew. In awāilcollections, the particular that Ishmael was the first to write in Arabic is not too frequent.22 See below p. 200. The report that Ismāīl picked up Arabic from Arabic-speaking tribes was widely disseminated, but there was a competing tradition saying that Ismāīl received the Arabic language by revelation on the day he was born. The three other sons of Ibrāhīm stuck to their father’s (Hebrew) language.23 According to a H adīt, the Prophet Muhammad recited (the verse) ‘an Arabic Qurān for people who know’ (Sura 41:3); then he said: ‘Ismāīl received this Arabic language by a great revelation’ (ulhima Ismāīlu hādā l-lisāna l-arabiyya ilhāman). This tradition contradicted the view that Ismāīl had learnt Arabic from the Jurhum tribe. In the latter case, the tribe must have logically preceded Ismāīl in speaking Arabic. Another H adīt mentions Ismāīl’s age: when he learned Arabic, he was 14 years old.24


Ibn Sad, Tabaqāt I/1 23, 9. Ibn Sad, Tabaqāt I/1 24, 16. Tottoli p. 82. Cf. Sibt  Ibn al-Jawzī, Mirāt 310; Ibn Asākir, Tārīkh II 331. 22 See the comment of Šiblī, Mah āsin al-wasāil fī marifat al-awāil 143. 23 Ibn Sad, Tabaqāt I/1 24, 21. 24 as-Suyūtī, Muzhir I 22, 13. 21

divine, prophetic, and heroic arabic


Ibn Sad transmits a report that contradicts the idea that Ismāīl was born with Arabic and traces Arabic only to Ismāīl’s offspring. According to this tradition, Ismāīl never spoke Arabic, because in his filial piety he did not deem it permissible to act differently from his father. According to this report, the first of his offspring to speak Arabic were the Banū Rala bint Yašjub b. Yarub b. Lūdan b. Jurhum b. Āmir b. Saba b. Yaqtān b. Ābir b. Šālikh b. Arafxašad b. Sām b. Nūh.25 Nevertheless, the fact that Ismāīl ‘forgot’ his father’s Hebrew is expressly stated in a tradition traced to Muhammad b. Salām:26 ‘The first one to speak Arabic and to forget his father’s language was Ismāīl.’ Muslim Arab tradition, therefore, agrees that Ibrāhīm’s language was not Arabic, whereas there is disagreement on whether it was Ismāīl who was the first to speak Arabic and, if so, when and how Ismāīl learned it. For the tribal aspects of such traditions, cf. below under 4. Yarub’s Arabic. The most frequently quoted account, successful as a canonical H adīt, was, however, that Ismāīl learned Arabic from the Jurhum tribe in his youth (wa-shabba l-ghulāmu wa-taallama l-arabiyyata minhum, i.e., Jurhum, s. Buxārī, Sah īh , Anbiyā 21). He marries twice, in both cases a wife from the Jurhum tribe. This again meant, of course, that Ismāīl was, strictly speaking, not the first one to use Arabic, because he had to learn it from somebody else. In some reports, the gift of the Arabic language is mentioned next to other privileges of Ismāīl: the traditionists report that Ismāīl was the first one to speak Arabic, the first one to build the h aram after his father Ibrāhīm, and the first one to install the rites of pilgrimage. He was also the first to ride full-grown horses, which before were wild and could not be ridden. Some say: ‘Ismāīl was the first whose tongue God opened to speak Arabic. And when he grew up, God gave him the Arabic bow.’ This report implies that God revealed Arabic to Ismāīl and that Ismāīl did not have to learn it from the Jurhum.27 This report became part of Adab literature. At- Taālibī mentions it in his Latāif al-Maārif. ‘The first person to speak Arabic was Ismāīl, peace be upon him; all the Arabs came subsequently from his progeny, except for three tribes, those of Auzā, H adramawt and Taqīf. He was the first to

25 26 27

Ibn Sad, Tabaqāt I/1 24, 22. as-Suyūtī, Muzhir I 32, -2. al-Yaqūbī, Tārīx I 22.


stefan wild

ride horses, these mounts being originally wild horses which had never been ridden before.’28 There was also some concern about the fact that the long time that had elapsed between the period in which Ismāīl had started speaking Arabic and the time of the Prophet Muhammad made it difficult to accept that the Arabic language should not have changed during this interval. The Prophet gave the answer: ‘Ismāīl’s language had been obfuscated (darusat). But Jibrīl came and made me retain the language and I retained it,’29 i.e. the Qurānic message saved Arabic. In other versions, it is not only the Arabic language whose origin is attributed to Adam or Ismāīl, but also the Arabic script, which is in this case imagined as written on clay tablets: ‘The first who installed the Arabic script, the Syriac script, and all other scripts was Adam, 300 years before his death. He wrote them in clay and baked them. When the earth was hit by the flood, each people received its script and used it to write. Ismāīl b. Ibrāhīm received the Arabic writing.’30 Some of these conflicting mythological reports are woven together and transformed into scholarship by modern Wahhabi scholars. Taking such accounts as factually historical, Muhammad Musta fā al-Azamī is led in a recently published book31 to a re-writing of linguistic history on a grand scale. He asks himself the question ‘What language did the Nabateans speak?’ And the answer is: Growing up in Makkah from his earliest childhood Ismāīl, eldest son of Ibrāhīm, was raised among the Jurhum tribe and married within them twice. This tribe spoke Arabic, and so, undoubtedly, must have Ismāīl. The Jurhum Arabic probably lacked the sophistication and polish of the Quraishi Arabic, preceding it as it did almost by two thousand years. Ibn Ushta records a statement from Ibn Abbās, that the first person to initiate set rules for the Arabic grammar and alphabet was none other than Ismāīl. Eventually, Allah commissioned Ismāīl as a messenger and prophet, to call his people for the worship of the one true God Allah, to

28 In: C.E. Bosworth (ed.), The Book of Curious and Entertaining Information. The Latif al-marif of al-Thaālibī, Edinburgh 1968, p. 40, cf. Lammens, La cité arabe de Taif à la veille de l’Hégire, Beirut 1922, 57–68. 29 as-Suyūtī, Muzhir I 35, 3. 30 as-Suyuti, Itqān IV 143 quoting Ab Bakr M. b. Abdallāh b. Muhammad b. Ušta al-Isbahānī (d. 360) and his Kitāb al-masāh if after Kab al-Ahbār: cf. as-Suyūtī, Bug ya 59; as-Safadī, Wāfī III 347; and Nöldeke GdQ II 53 and GdQ III 1 fn. 2. 31 Muhammad Musta f al-Azam, The History of the Quranic Text. From Revelation to Compilation. A Comparative Study with the Old and New Testaments, UK Islamic Academy, Leicester 2003.

divine, prophetic, and heroic arabic


establish prayers and pay alms to the poor. Since Allah sends every messenger in the language of his own people, Ismāīl must have preached in Arabic. Genesis credits Ismāīl with twelve sons, among them Nebajoth/ Nabat; born and nurtured in these Arabian surroundings they must have adopted Arabic as their mother tongue. These sons may have preserved their father’s message by using the prevailing Arabic script; certainly, they would not have resorted to whatever script was then current in Palestine (Ibrāhīm’s homeland), since two generations had already lived in Arabia. When Nabat subsequently migrated northwards he must have taken the Arabic language and alphabet with him. It was his descendants who established the Nabatean Kingdom (600 B.C.E.–105 C.E.).

Al-Azamī dates Ismāīl and his early Arabic at around 1400 B.C.E. and concludes: The Nabatean language and script were ‘. . . a form of Arabic’ (121). Such a sentence disregards that the Nabateans’ spoken language was indeed Arabic, while the texts they wrote were in Aramaic. The thesis that Nabatean is a form of Arabic comes close to a linguistic panorama, which sees in all languages usually called ‘Semitic’ a form of Arabic. ‘The Arabic language and script, in their primitive forms, gave birth to the Nabatean and most probably predated the Syriac’ (121) is but one example of such a view.

4. Yarub’s Arabic A further source in tracing the first Arabic speaker was tribal history. The Arabs, in their own self-view, were not only a linguistic community, but also marked by common ancestors. Descent was a primary symbol of a tribal community, and the self-esteem and prestige of individuals and communities were linked to the purity of descent. The efforts of the Arab genealogists to establish a link between living Arab tribal groups and past forefathers were, of course, highly tenuous. The skeptical observation of Ibn H azm (d. 456/1064) that ‘on the face of the earth there is no one whose descent from them is verifiable’ did not prevent the emergence of the most speculative lineages.32 By and large, the Arab tribes claimed descent from one of two ancestors, either a North Arabic origin

32 Ibn H azm, Jamharat ansāb al-arab, ed. Lévy-Provencal, Cairo 1948, 8, quoted by EI 2nd ed. I 546.


stefan wild

connected to the North Arab Adnān (also to Maadd and Nizār, later Qays) or a South Arabic origin connected to Qahtā n (also called “Yamanis,” later also Kalb). Some tribes such as Ād, Tamūd, Īram, Jurhum, Tasm and Jadīs were believed to have disappeared before Islam. Hostility between the offspring of Qahtā n and those of Adnān emerged after the advent of Islam in the form of the cleavage between the Ansār in Medina and the Qurayš. The fact that the Prophet Muhammad belonged to the Qurayš brought enormous prestige to the Adnān, i.e., the North Arabs. The narrative that linked Ismāīl to Arabic was counterbalanced by the argument that Ismāīl had learned Arabic from the Jurhum, a South Arabic tribe. While al-Masūdī accepted that Ismāīl had been given Arabic by God, he did not deny that Yarub b. Qahtā n, the ancestor of the Yamanis, was the first to speak Arabic. ‘Which of the three Arab group, the extinct Arabs, the Nizāris, or the Yamanis was the first to speak Arabic? The Nizāri and Yamani groups vigorously upheld their own claims to this honour. The Nizāris held that Ismāīl was given the language by God, while the Yamanis contended that Ismāīl had learned the language from a Yamani tribe living in Mecca.33 ‘This conflict of claims masked a deeper social and political conflict between the two groups. Al-Masūdī, while granting that Yarub Ibn Qahtā n, the ancestor of the Yamanis, was the first to speak Arabic (Murūj sec. 71), believes that Ishmael too was granted this honour by God, independently of his association with the Yamanis.’34 The tribal family trees set up by Arab genealogists always have an agenda. When names are inserted or omitted, insertion and omission usually serve a purpose. When in the list of Qahtā n’s ancestors there are two new names, those of Yašjub and Yarub, inserted between Saba and Qahtā n (Yoqtan) in a family tree that is otherwise based on Gen 10, 1–32, this happens because these two names fulfill two important functions. Yarub symbolizes and personalizes the change from Syriac to pure Arabic, while Yašjub/Yaman gives his name to the land Yaman. The appearance and etiological function of persons with such telling names is a common feature of Arab genealogy.35 A tradition quoted by Yāqūt and traced to ad-Dīnawarī’s Mujālasa links Bābil with Yarub and Arabic as the heavenly language: ‘When God 33 Tarif Khalidi, Islamic Historiography. The histories of Masudi, Albany 1975, 116; conflicting claims are set forth in Tanbīh 79–83 and Murūj 996–99. 34 Tarif Khalidi, Islamic Historiography ibid. 35 Manfred Kropp, Geschichte der ‘reinen’ Araber II 379f.

divine, prophetic, and heroic arabic


assembled mankind in Babylon, He sent them an Eastern, a Western, a Southern, and a Northern wind. And He assembled them in Babylon and they assembled to see why they had been called. Suddenly, a voice called out: “He who has the West on his right, the East to his left, and faces Mecca—his is the language of the people of heaven.” Then Yarub b. Qahtā n stood up, and the voice said: “Yarub b. Qahtā n b. Hūd, you are the one.” And he was the first to speak Arabic.”36 According to Ibn Durayd, “the pure Arabs were seven tribes Ād, Tamūd, Imlīq, Tasm, Jadīs, Umayyim und Jāsim. Most of them have vanished except for some remnants dispersed in the sub-tribes. And Yarub, whose name was Muhazzim b. Qahtā n, received his byname because he was the first whose tongue switched from Syriac to Arabic. This is the meaning of al-Jawharī’s statement in as-Sih āh : ‘the first to speak Arabic was Yarub b. Qahtā n.’ ’’37 Here, the eponymon Yarub is linked to the root rb, another example for the favorite method of the genealogists of connecting proper names with imagined history, etymology with linguistic history. In this mixture of traditions, Yarub b. Qahtā n is finally also located in Babylon and celebrates his virtues in the style of a poetic self-praise. ‘The majority of historians among them, the author of Tawārīx alumam (al-Isfahānī), and the author of al-Maārif (Ibn Qutayba) tell us that Yarub b. Qahtā n was the first to speak in clear Arabic, the first to bear the crown in Yaman, the first who was greeted by his sons with the formula ‘May you avoid the curse and your day be happy!’ Ibn Hishām, the author of the ‘Book of Crowns’, which he wrote on the kings of H imyar reports: “It was Yarub who went with the Arabs to Yaman and settled there. Therefore Yaman was called after him, because Yarub’s name was ‘Yaman.’ ” Al-Bayhaqī says: “Yarub was the first to speak in clear Arabic and left the confused way of talking used by the Arabs. He and his offspring were successful and God bestowed on them the land of Yaman as heritage. To them belong the kings of Tubba, who conquered the regions of the earth. On the strength of this privilege they are ranked before Adnān, while the Adnān can boast the prophethood of Muhammad. . . .” At the time, in which the languages had already been confused, the Banū H ām had come to the highlands of

36 37

Yāqūt, Mujam al-buldān ed. F. Wüstenfeld, I 447, 19. as-Suyūtī, Muzhir I 31, -3.


stefan wild

Yaman. . . . And Yarub was the most high-minded youth in Babylon and spoke clear Arabic, while the others spoke in confused language.38 And Yarub recited the following verse: I am the youth favored by the richest gift, the happy one, well known for his virtue. I am the son of Qahtā n, the influential and rich, I spoke in Arabic, while the people were in (linguistic) confusion. (I spoke) in the clearest unambiguous language and in the perfect language of the kingdoms after me. anā l-g ulāmu dū n-nasībi l-ajzali / al-aymanu l-marūfu bi-t-tajammuli / anā bnu Qahtā na l-hammāmi l-aqyali / arabtu wa-l-ummatu fī tabalbuli / bi-l-mantiqi l-abyani g ayri muškili / wa-mantiqi l-amlāki badi l-kamali.

This is pure poetic self-praise in the rajaz meter as we know it from the earliest Arabic poetry. According to this report, Yarub b. Qahtā n was the first to speak clear Arabic and turned away from the ‘confused’ version of Arabic (namatu l-arabiyyati l-mubalbalati). He let his sons greet him with the greeting abayta l-lana (‘may you avoid the curse!’) and with ‘good morning!’ This Arabic is not so much seen as a prophetic language but as a heroic language spoken by a tribal noble forefather. The hero of this poem, Yarub, praises himself for speaking pure Arabic, a royal language of noble kings, while lesser Arab mortals used mixed and confused languages. In a further verse, Yarub predicts the coming of the Prophet Muhammad: Muh ammadu l-hādī n-nabiyyu l-mursalu / li-llāhi darru l-mājidi l-mustaqbili Muhammad, the guide, the God-sent Prophet—/ how praiseworthy is the blessed one who is coming.

Such a tradition combined tribal history with revelation. And to round things out, the report claims that the same Yarub was the first to recite Arabic poetry and to put it into meters. He invented the poetic genres and composed praise-poetry, self-praise, and love poetry.39 Earlier traditions see the Banū Hāshim and the Banū Yarub in fierce competition. When the rival of the Prophet Muhammad, Musaylima ‘the Liar,’ who traced his descent to the Banū Yarub, heard of Muhammad’s death, he hoped to outstrip the Prophet and is said to have recited the verse:

38 39

Kropp, I Arab. text 9ff. Kropp 11, transl. 149.

divine, prophetic, and heroic arabic


‘Passed away the prophet of the Banū Hāshim / and rose up the prophet of the Banū Yarub!’40

5. The Prophet Muh ammad’s Arabic As we have seen, after the advent of Islam, the unique character of Arabic was extrapolated by theologians, philosophers, and genealogists back to times immemorial. In their view, after the revelation of the Arabic Qurān, Arabic became even more of a special language. Arabic itself, not only its Qurānic form, was something sacred, superhuman. Yet, in a famous description, aš-Šāfiī (d. 204/820) explained the special character of Arabic mainly by its most important hermeneutical challenge, i.e., its vast and ramified vocabulary. According to him, this was the primary reason for the privileged status of Arabic: We do not know that any man except for a prophet can claim to know the Arabic language completely. However, nothing Arabic escapes the Arabs collectively, so that there would be nobody among the Arabs who knows it. The knowledge of Arabic among the Arabs resembles the knowledge of the traditions (sunan) among the scholars of religious Law. We do not know anybody who could claim to know all traditions, so that not a single one would escape him. But when the knowledge of traditions of all scholars is collected, the whole of tradition is found. When the knowledge of each of these scholars is divided, something of the traditions will escape each of them. But what escapes him will be found with someone else.41

The Prophet, who as an individual can claim to know Arabic completely, is, of course, Muhammad. His knowledge of Arabic is superhuman. But the Arabs collectively also know the whole language. The Prophet’s own mastering of Arabic was related to the Qurānic revelation, as well. When a man admires the Prophet’s rhetorical talent and says, ‘What a good speaker you are: We have not seen anybody speaking better Arabic than you!’ the Prophet answers: ‘This is my right. For the Qurān was revealed to me in clear Arabic.’42 And the Prophet could boast: ‘I am the best Arabic speaker’ (anā afsah u l-arab43 or anā

40 Ibn Katī r, Bidāya vi, 341, quoted according to M. J. Kister, ‘Musaylima’ in EQ 3 (2003) 462. 41 Aš-Šāfiī, Risāla, ed. Ahmad Muhammad Šākir, Cairo 1940, 42, 8ff. 42 as-Suyūtī, Muzhir I 35, 6. 43 Muzhir I 209, 2.


stefan wild

arabukum).44 It is not always clear whether such statements refer to Qurānic revelation or to the Prophet’s non-Qurānic words. In the same vein, the Prophet was immune against linguistic mistakes (lah n) in his speech. He was reported to have said: ‘I am from the Qurayš and I grew up among the Banī Sad—how should I commit a linguistic error?’ (anā min Qurayš wa-našatu fī banī Sad wa-annā fiyya l-lah n).45 The Banī Sad b. Bakr b. Hawāzin were the clan of Muhammad’s wet-nurse. In this case, the Prophet’s claim to linguistic excellence was definitely rooted in his tribal background and not in divine grace. Among the further signs of the Prophet’s unprecedented eloquence was that he was said to have coined expressions that allegedly had never been used in Arabic before him and that later became part of an elevated Arabic style, such as māta h atfa anfihī, ‘he died a natural death.’46 In some traditions, the Prophet became a second Adam. In a H adīt, he said: ‘My community appeared to me in water and clay. And I was taught all names even as Adam was taught all names.’47 Arab grammarians connect the linguistic quality of the Arabic of the Qurayš with pre-Islamic times and with the role the Qurayš played at the Kaba in Mecca. A privileged position vis-à-vis the Arabic language was also conferred on the companions of the Prophet—they also spoke nothing but the purest Arabic. The proof was that they never coined a new word (mā alimnāhumu sta lah ū alā xtirāi lug atin aw ih dāthi lafzatin lam tataqaddamhum).48 This was most probably connected to the fact that the companions were the guardians of the Prophet’s word and deeds. If their Arabic had been doubtful, it could have influenced the correctness of tradition. In a comparable manner, the ethical, philosophical, and religious letters and pronouncements, ascribed under the title Nahj al-balāg a (‘The Method of Eloquence’) to Alī b. Abī Tālib, were considered by many Shiites in essence and style as ‘second only to the Qurān.’49


Ibn Sad, Tabaqāt I 113. Abū t-Tayyib al-Lugaw, Marātib an-nahwiyyīn, ed. Muhammad Abū l-Fadl Ibrāhīm, Cairo n.d. , Maktabat Nahdat Misr 6,1. 46 Muzhir I 209, 5. 47 Muzhir I 35, 4. 48 Muzhir I 10, 9. 49 Nahj al-Balāg a—wa-huwa majmū mā xtārahu š-Šarīf Abū l-H asan Muh ammad ar-Rādī b. al-H asan al-Mūsawī min kalām Amīr al-muminīn Abī l-H asan Alī b. Abī T ālib, ed. Subhī Sālih, Beirut 1982; Moojan Momen, An Introduction to Shii Islam, The History and Doctrine of Twelver Shiism, New Haven/London 1983, 25. 45

divine, prophetic, and heroic arabic


6. The fate of Arabic after the Islamic conquest The Arabic language celebrated its greatest triumph with the vast expansion of the Islamic conquest and the concomitant expansion of the Arabic language. But most Arab grammarians and lexicographers did not see it that way. In an almost incredible volte-face, they saw the fate of the Arabic language sealed and its purity doomed to corruption precisely at the moment when the Muslim faith and the Qurān won over the world: wa-lam tazal il-arabu tantiqu alā sajiyyatihā fī sadri islāmihā wa-mādī jāhiliyyatihā h attā azhara llāhu l-islāma alā sāiri l-adyāni fa-daxala n-nāsu fīhi afwājan wa-aqbalū ilayhi arsālan wa-jtamaat fīhi l-alsinatu l-mutafarriqatu wa-l-lug ātu l-muxtalifatu fa-fašā l-fasādu fī l-lug ati l-arabiyyati.50 The Arabs did not cease to speak Arabic according to its disposition in the beginning of the era of Islam and in the pre-Islamic past until God let Islam prevail over all other religions. Then people entered Islam in masses and turned to it in flocks. In Islam all separate languages and different tongues came together and therefore corruption ( fasād) spread in the Arabic language.

This rather anticlimactic statement marks the beginning of Arabic as the language of an Islamic world civilization. For many an Arab grammarian and lexicographer, this period was at the same time the starting point of its decadence and corruption.

7. References Agha, Saleh Said and Tarif Khalidi. 2002/03. “Poetry and Identity in the Umayyad Age.” Al-Abhāth 50–51, 55–119. al-Azamī, Muhammad Musta fā. 2003. The History of the Quranic Text. From Revelation to Compilation. A Comparative Study with the Old and New Testaments. Leicester: UK Islamic Academy. al-Azraqī. 1858. Die Chroniken der Stadt Mekka (F. Wüstenfeld, ed.), vol. 1 Axbār Makka. Leipzig. Czapkiewicz, Andrzej. 1988. The Views of the Medieval Arab Philologists on Language and its Origin in the Light of Al-Suyuti’s Al-Muzhir. Krakow: Universitas Jagellonica Acta Scientiarum Litterarumque CMIX.

50 az-Zubaydī, Tabaqāt an-nahwiyyīn wa-l-lug awiyyīn (ed. Muhammad Abū l-Fadl Ibrāhīm), Cairo 1954, 1.


stefan wild

EQ 2001–2006 = Jane McAuliffe, ed. Encyclopaedia of the Qurān, vol. 1–5. Leiden: E. J. Brill Fischer, A. and A.K. Irvine. 1978. “Kahtā n.” EI 2nd ed., 4: 447–449. Gilliot, Claude and Pierre Larcher. 2003. “Language and Style of the Qurān.” EQ 3, 109–135, especially ‘The mythical narratives on the superiority of Arabic’, 118ff. Goldziher, Ignaz. 1873. “Beiträge zur Geschichte der Sprachgelehrsamkeit bei den Arabern.” Sitzungsber. d. Kaiserl. Ak. d.W. Wien, Phil-Hist.Kl. 73, 511–52. Ibn H azm. 1948. Jamharat ansāb al-arab. Ed. É. Levy-Provencal, Cairo. Ibn Mutarrif at-Tarafī, Abū Abdallāh M. b. A. al-Kinānī. 2003. Qisas al-anbiyā (‘The Stories of the Prophets’). Roberto Tottoli, ed. Berlin: Islamkundliche Untersuchungen 253. al-Jumahī Muhammad b. Sallām 1916. Tabaqāt aš-šuarā. Joseph Hell, ed. Leiden. Khalidi, Tarif. 1975. Islamic Historiography. The histories of Masudi. Albano. Kropp, Manfred. 1975. Die Geschichte der ‘reinen Araber’ vom Stamme Qahtā n aus dem Kitāb Nashwat at-Tarab fī tārīkh jāhiliyyat al-arab des Ibn Saīd al-Maghribī, hrsg. u. übersetzt von Manfred Kropp, Bd. I: Einleitung und Text, Bd. II: Übersetzung und Anmerkungen, Diss. Heidelberg. Loucel, Henri. 1963. “L’origine du langage d’après les grammairiens arabes.” Arabica 10 I 188–208; II 253–281; 11 (1964) III 57–72; IV 151–187. Rubin, Uri. 1990. “H anīfiyya and Kaba. An inquiry into the Arabian pre-Islamic background of dīn Ibrāhīm.” JSAI 13, 85–112. as-Suyūtī, Abdarrahmān Jalāladdīn. n.d. al-Muzhir fī ulūm al-lug a. Muhammad Ahmad Jād al-Mawlā Bek, Muhammad Abū l-Fadl Ibrāhīm, Alī Muhammad al-Bījāwī, eds. Cairo n.d., 2nd ed., Dār Ihyā al-Kutub al-arabiyya, vol. 1–2. Tottoli, Roberto, ed. 2003. The Stories of the Prophets by Ibn Mutarrif al-Tarafi, edited with an introduction. Berlin, 253. ——. 2002. Biblical Prophets in the Qurān and Muslim Literature. Richmond. Weiss, Bernard G. 1974. “Medieval Muslim Discussions on the Origin of Language.” ZDMG 124, 33–41.


1. Introduction Grammars of Hebrew written in Europe in the Renaissance by Christians could benefit from the Hebrew grammatical tradition. Johannes Reuchlin (1455–1522) quotes in his De rudimentis hebraicis (1506) Priscian and ‘Rabbi David’ (i.e. Qimhi) as well (Law 2002, 247–248). In grammars of Sanskrit written in Europe we see also that the framework of grammatical description has been derived from the Indian grammatical tradition. The German Jesuit Heinrich Roth (1620–1668), as Hauschild (1988, 13–14) observes, ‘uses with perfect familiarity the technical terms of Indian grammar [. . .] Roth stands entirely within the Indian grammatical tradition, and probably he used the practical grammar of Anubhûti Svarûpâcârya, called the Sârasvata Vyâkarana, which was in general circulation in Hindustân, Bihâr and Benares. [. . .]. Another candidate, though a less likely one, would be the grammar called Mugdhabodha, which was composed in the second half of the 13th century by Vopadeva, but the usage of which was more common

1 This article is an elaborated version of paper delivered at the IIIrd International Conference on Missionary Linguistics, Hong Kong and Macau, 12th–15th March 2005. The organization of the conference and participation in Hong Kong has been made possible by financial support of the Norwegian Research Council (Norges Forskingsråd) and the Language Centre of the Hong Kong University of Science and Technology. I have to acknowledge Kees Versteegh who inspired me to do research in the field of the History of Linguistics and particularly Michael Carter for his valuable corrections and suggestions. Thanks to my colleague Maria Cândida Barros, I came across the reference to the grammar of Lucas Caballero. Research has been made possible by the Radboud University (TCMO) where my 2 research on Pedro de Alcalá started. I continued this research topic at the University of Oslo, supported by the NFR-project OsProMil (Oslo Project on Missionary Linguistics). I am grateful to the Rogge Library (Strängnäs) for the reproduction of the MS. I gladly acknowledge Pierre Winkler for his translations from Latin.


otto zwartjes

in Bengal . . . The work of Roth was a ‘masterpiece,’ which does not differ considerably from current grammars, which similarly depend on the linguistic achievements of India’s own grammatical tradition’ (Hauschild 1988, 13–14). In the grammars of Japanese written by the Portuguese Jesuit João Rodrigues (1561–1634), particularly in his description of particles and verbal endings, we can also find information concerning the study of ‘tenifa’ or ‘tenivofa’ (the study of particles and verbal endings) from contemporary Japanese scholars (Maruyama 2004, 155). As has been demonstrated by Gregory James (2007), some missionaries describing the Tamil language, such as Bartholomäus Ziegenbalg (1682– 1719), were familiar with some ancient Tamil works on grammar. These missionaries worked under favourable conditions compared to their colleagues who described languages, particularly Amerindian and Austronesian, without any written tradition or an adequate ‘indigenous’ grammatical framework they could rely on. The use of Arabic grammatical terms in the first grammar written in Europe of vernacular Arabic, the Arte para ligeramente saber la lengua arabiga (1505) of Pedro de Alcalá (Order of St. Jerome) has been the subject of an article written by William Cowan (1981). In de Alcalá’s grammar, some technical terms were incorporated in the descriptive framework, including terms such as damir, temiz, masdar, amr and xucla. In this article, Spanish grammars of Arabic—vernacular and classical—written by Franciscans in Damascus and completed, copied, or printed in Spain in the 18th century occupy our attention, particularly the grammars of Francisco Caballero and Juan de la Encarnación (18th century) and Francisco Cañes (1730–1795). Of the first we have an unpublished manuscript, which has escaped the attention of researchers until today,2 and of the second a printed work has been conserved,

2 Bibliographical information concerning Francisco Cañes can be found in Schnurrer (1811, 79, no. 113) BICRES III, and in Monroe (1970) we find some historical background. However Cañes is not mentioned by Fück (1955), Dannenfeldt (1955), and Killean (1984) and particularly the grammar of Caballero has been neglected by all. After a century-long period of silence and total neglect, an important monograph on Bernardino González appeared recently, together with a facsimile edition of the dictionary (Intérprete arábico) and his grammar (Epítome) (Lourido Díaz 2005), not long after this paper had been delivered in Hong Kong. When the proofs were almost ready for publication, I received a copy of this monograph, courtesy of Emilio Ridruejo. Lourido Díaz (2005, I, 21–22) had traced seven manuscripts of the dictionary, and six copies of Epítome de la gramática árabe made by Bernardino González’ pupils, probably for their own use. One copy was completed by Blas Francisco de Salamanca in 1704, the second by Lucas Caballero and Juan de la Encarnación between 1709 and 1710, the so-called Tingstadius copy. Two copies were compiled in 1719, one from the El Escorial

inflection and government in arabic


studied sporadically by dialectologists describing the urban colloquial Arabic of Damascus (Lentin 1997), but which also escaped the attention of researchers working in the field of the history of linguistics. In the Damascene Franciscan tradition, the authors used a great number of Arabic grammatical terms, adapted in a hispanicised form, which substituted, accompanied or sometimes superseded Latin terminology : examples are: ‘arabicación’ (irāb), ‘moziones’ (h arakāt) or the ‘partículas nasbantes, chazmantes and charrantes.’ Obviously, these Franciscan authors were informed by the Maronite Christians, but in the prologue of the grammar of Francisco Cañes, we read that Spanish Franciscans not only were familiar with the work of Pedro de Alcalá, but also that they were inspired by north-European grammars and dictionaries of Arabic, such as the famous grammar written in Latin by Thomas van Erpen (= Erpenius 1585–1624) and lexicographical work of Jacob Golius

and the other from the University of Valencia and two further anonymous and undated manuscripts from the Real Academia de la Historia. Lourido Díaz states that all these works were calques of that of Bernardino González (Lourido Díaz 2005, I, 13). All the Latin grammars analysed in this article are also listed in two footnotes by Lourido Díaz (2005, I, 130 and 135) but very little importance is given to the influence of these on the grammatical tradition of the Franciscans linked to Damascus. After having consulted all these Latin grammars, we have come to the conclusion that the Spanish grammars of Arabic completed by Franciscans in Damascus or in Spain were heavily inspired by the Latin grammars, and in some cases they are Spanish calques, or translations of the Latin examples. This is particularly evident in the grammatical examples and the use of literally almost the same ‘orientalising’ terminology, inherited from the Arabic tradition. Thanks to the evidence of Lourido Díaz’ study, a direct link between the Franciscans in Damascus and Spain and the Holy Congregation of the Propaganda Fide and the San Pietro di Montorio can now be confirmed. Juan de la Encarnación learned Arabic from his teacher Lucas Caballero, a pupil of Bernardino González. The latter, in his turn, was a pupil of the Italian Fray Bonaventura da Molazzana, who taught at the San Pietro di Montorio and who arrived in Seville in 1693. It is known that the grammars and dictionaries used at the San Pietro di Montorio were those of Dominicus Germanus and Philip Guadagnoli, among others, and it is thus probable that Bernardino González had direct access to the ‘Italian’ grammatical tradition. It is also important for the purposes of this article to know that the work of Bernardino González was also obligatory in the curriculum for Spanish and Portuguese missionaries (Lourido Díaz 2005, I, 34). Morevoer, the Portuguese Arabist de Sousa was born in Damascus, so all these grammars are thus linked and use common sources. Germanus of Silesia was educated in the Holy Land, and was later an Arabic instructor in the El Escorial Monastery in Spain. Although Lourido Díaz’ monograph is without any doubt extremely important for all those interested in the bio-bibliographical data related to Bernardino González and his successors, little importance is given in it to the influence of Latin sources and almost nothing is said about possible Arabic sources, the significance of these works from the perspective of the history of linguistics or the history of Arabic. Are these grammars to be considered as key creative productions on the part of Spanish missionaries, or are they nothing more than a chain in a long tradition? In the future we hope to give an answer to this question.


otto zwartjes

(1596–1667). Other sources mentioned by Cañes are the grammars of Felipe Guadañoli,3 Antonio de Aguila,4 Agapito de Valle Flammarum,5 and Francisco Gonzalez,6 and the Franciscan grammarian also informs us that he used a “considerable number of manuscripts,” without specifying which sources these were. In Renaissance grammars, we find usually two subdisciplines in syntax. In the Spanish tradition, syntax, often called ‘construcción’, can be subdivided in ‘construcción de régimen’ and ‘construcción de concordancia’ and Francisco Cañes does not form an exception. In this article we concentrate on the first and we particularly attempt to analyze how the concept of āmil (often translated as ‘governor’) has been incorporated into this model. The subject of concord (agreement) has been analysed in Zwartjes (2007). The use of Arabic grammatical terms in these grammars will be analyzed, and we will concentrate on the following questions related to morphosyntax: – Which Arabic grammatical terms are used in these Spanish grammars particularly in the sections dedicated to inflection and agreement? – What did they mean and why did they use them? – Which sources could they have used?

2. The grammar and dictionary of Arabic of Pedro de Alcalá As we all know, Arabic speaking communities are diglossic. Pedro de Alcalá’s dictionary and grammar of Granadan Arabic is obviously a description of colloquial Arabic. His purpose was to teach the ‘ordinary people’ (‘los populares’) and not the language of ‘the wise’ (‘alfaquíes’). The aims to compose his dictionary are also slightly different if we compare them to other missionary dictionaries in the New World, where missionary composed dictionaries for their own use and for the novices from the Old World. As we can read in the prologue to his dictionary, Pedro de Alcalá wrote his dictionary not only for the Old Christians who wanted to learn Arabic, but also for the New Christians.7


Schnurrer (1811, 47, no. 72). Antonius Ab Aquila (Schnurrer 1811, 50, no. 78). 5 Schnurrer (1811, 59, no. 85), or Agapito à Valle Flemmarum (da Val di Fiemme). 6 I have not identified this author yet, but this could possibly be Bernardino González, as Lourido Díaz suggests (2005). 7 “Ca assi como los aljamiados (o cristianos viejos) pueden por esta obra saber el arauia, viniendo del romance al arauia: assi los arauigos (o nueuos cristianos), sabiendo leer la 4

inflection and government in arabic


As has been demonstrated in recent research, mainly from Federico Corriente (1988), it is true that the language described is predominantly colloquial,8 but at the same time it is also obvious that there is some interference between the colloquial and the classical registers. The main objective of all missionary grammars is practical: the teaching of a certain language. Nevertheless, missionary works can be predominantly didactic, showing many paradigms with few explanations and linguistic theory, while in other works the didactic-pedagogical approach is much more theoretical. Pedro de Alcalá also states that his grammar is more a practical introduction, not a learned theoretical work. His approach reveals itself to be universalist, since he observes that there are definitions and explanations in one language in respect to the expression of its concepts in its own terms, these are the same in all other languages regarding the expression of their own concepts. Thus for the same reason that this name ‘Pedro’ is a proper noun in Latin, it is also such in ‘Arabic’ (Alcalá 1505[1883], 2). This observation is characteristic of Renaissance grammars in general and can be found in many other sources from this period. As Vivian Law observed: Questions as ‘what is a proper noun?’, ‘what is a verb?’ ‘how many word classes are there?’ ‘what are properties of the conjunction?’ are as close to universal as any you are likely to find in a medieval grammar. Such concerns apply equally well to any European language; indeed, they had already been transferred from Greek to Latin. There is no inherent reason why they should not also be asked about Old Irish or Old Icelandic: one can find proper nouns (for instance) just as easily there as in Latin. (Law 2002, 191)

Obviously, Alcalá did not find it necessary to give his pupils definitions of the parts of speech, since they are the same for all languages. So one would wonder why the author decided to include Arabic grammatical terminology, which seems to be in contradiction with his own ‘universalistic’ approach. A possible explanation is that he did this only for ‘scientific’ reasons. He might have introduced them with the purpose to have a more adequate or sophisticated framework to fit in phenomena

letra castellana: tomando primero el arauia, ligeramente pueden venir en conocimiento del aljamia).” (Pedro de Alcalá 1505, prologue, ii v.) See also Cowan (1981, 358). [As the ‘aljamiados’, or old Christians can learn Arabic through this work, coming from Romance to Arabic, so the Arabs (or new Christians), having mastery of the Castilian alphabeth, taking first the Arabic, easily can have knowledge of the ‘aljamia’]. 8 His purpose was “hazer vocabulista de la habla comun y usada de la gente deste.” (ibid.) “to compose a dictionary of the common speech and used by the people.”


otto zwartjes

he was not so familiar with. As has been demonstrated by Cowen, this is only in a few cases a satisfactory explanation. Sometimes Arabic terms are only used as equivalents or translations of Latin terms. Clear examples are the names for the cases in the nominal paradigm. Pedro de Alcalá recognizes when dealing with ‘declensions’ that the Arabic noun has only one declension, since all Arabic nouns are invariable.9 However, admitting that there is only one declension, he states that Arabic nouns have six cases. We reproduce here the table according to Cowan (1981, 359): TABLE 1




Classical term

mubtedĕ ‘subject’ mudăf ‘genitive’ maxrŏr ‘dative’ mafuŭl10 ‘object’ munĕde ‘vocative’ zarf ‘adverb’








maf ūl





As William Cowan observed:10 they are not true cases in either the Arabic or the Latin sense of being an inflection added to a noun, a fact that Alcalá was quite aware of, but are regarded by him in the same way that modern theoreticians of case grammar regard the syntactic positions in an utterance. Such relations are not expressed by inflections, since colloquial Arabic has none, but through abitudines or conocimientos. The markers of the equivalents of these Latin cases are in fact combinations of prepositions and particles and the definite article (lil, maal, lal, etc.) (Cowan 1981, 359).

9 “La declinacion de todos los nombres arauigos es vna solamente. Porque todos los nombres arauigos son inuariables” (Alcalá 1505, capitulo nono). 10 In the original text, a small hamza in superscript is placed on the first vowel ‘u’.

inflection and government in arabic


The information given by William Cowan, however, is not complete, since translations from his dictionary are not included here. Although we do not find all the terms in his dictionary, we do find some of them: mubtedĕ is not only translated by Pedro de Alcalá as ‘nominative’, but also as ‘principio de oracion’ (Corriente 1988, 12), which is not without importance (see below). Ízm mudáf is rendered as ‘posesivo nonbre’.11 The term ‘abitud’ seems to be inherited from other grammars than the Latin and Castilian grammars of Antonio de Nebrija (c. 1444–1522). The term ‘habitudo’ is used by Ferdinandus Nepos in his Materies (completed between 1469 and 1485) and Juan de Pastrana Compendium grammatica (1462).12 Item per hanc regulam primo iuvenes component per unum casum tantum sic dicendo: la tierra ‘terra’; de la tierra ‘terre’, etc.; uel per duos, sic: la tierra del rey; uel per tres et quatuor et amplius, sic: o leyente la lection a los scolares en el general de las escuelas componitur ‘o legens lectionem scolaribus in generali scolarum’, dando cuilibet casui propriam habitudinem, interrogando cuius casus, numeri et cuius declinationis hoc principio. (Codoñer 2000, 90).

However, it is not so clear what Pedro de Alcalá’s exactly means with the term ‘abitud’ (pl. ‘abitudines’), since he uses it as a synonym of the definite article (‘Nota que porque ay algunas abitudines en cada vno delos casos que en alguna manera parescen preposiciones, porque se preponen alos casos, avn que en verdad no lo sean, mas son articulos’) (Alcalá 1505[1971], 26). In another section, Alcalá uses the term as a synonym of ‘preposicion’, since according to his observations, the maxrór case (see below) has the four ‘abitudines’ la, lal, li, lil, whereas the term ‘abitud’ is not used for ba, bal, bi, bil, fa, etc. which are described in the paragraph on the zarf case (Alcalá 1505[1971], 27). Here they are just called ‘prepositions’ and not ‘abitudines’ and there is no explanation

11 Other terms not mentioned by Cowen, are ‘jezme’, translated as ‘consonante’ (letra mazjun [sic] (Corriente (1988, 34), and ‘médde’, translated as ‘acento’ (Corriente 1988, 189). 12 “With this rule, the novices first build constituents with only one ‘case’ [= ‘head’ of the NP], saying “la tierra” ‘terra’, “de la tierra”, ‘terre’, etc., or with two ‘cases’ [= ‘head’ + complement of the NP], as “la tierra del rey”, or with four and so on, as: “reading the lesson in general for the scholars of the schools,” assigning the appropriate “habitudo” [= grammatical form] to whatever case [= grammatical function]. The term ‘habitudo casualis’ is also used by Nepos in relation with government: “Haec enim regula maxima est in construendo [. . .] quia talem casum regit dictio qualis fuerit habitudo casualis.” (ibid.).


otto zwartjes

why he used these different terms. In other sections, Alcalá uses ‘abitud’ as synonym of ‘conocimientos’ (see also Zwartjes 1993, and 1994). In Arabic grammatical theory a nominal sentence can be divided into ‘topic’ for which mubtada (lit. ‘what is begun with’) is normally used, and ‘comment’ (xabar), or ‘predicate’ (Owens 1988, 32), or according to Sībawayhi’s terminology also called mabnīy alā-l-mubtada (‘what is built upon what is begun with’) (Owens 1990, 45). Mudāf is the word Pedro de Alcalá uses for the genitive, which in Arabic tradition means literally ‘what is added’, i.e. possessed (Owens 1988, 34; 1990, 104). Majrūr from the same root as jarr (see below), means ‘pulling’, or governing the -i inflected form. Pedro de Alcalá follows in his sections about the prepositions the Latin system and tries to apply Arabic terms to them, without realizing that in the dialect he describes, case-endings are not used, and without realizing that in classical Arabic nominal declension, there are three inflectional vowels, the -u, the -a, whereas for the verbal inflection the three vowels -u, -a and ø (zero ending) can be distinguished. Pedro de Alcalá did not take the Arabic inflectional endings as starting point, but the Latin prepositions in alphabetic order: prepositions + accusative, prepositions + ablative, etc. and at the same time he translated the names of these ‘Latin’ terminology into Arabic: Capitulo XXXII. De las preposiciones. Hallamos en el Arauia todas las preposiciones que en la gramatica [latina], y ayuntadas a essos mesmos casos, que son mafúul y darf (que son acusatiuo y ablatiuo), y son las del acusativo las siguientes: A A

ad al

apud aynd

circa carib

circa qued

ante acábal

longe baád

ante cudim

[. . .] [. . .]

Las preposiciones del darf (que es ablativo) son las siguientes, conuiene saber: Con con con Ba bal bi

con bil

en fa

en fal

en en fi fil . . . (Alcalá 1505[1883], 26).

“Chapter XXXII. About the prepositions. We find in the Arabic language all the same prepositions as in Latin grammar, and they are combined with the same cases, which are mafūl and zarf (which are accusative and ablative), and those which can be combined with the accusative are the following: A. A

ad al

apud aynd

circa carib

circa qued

ante acábal

longe baád

ante cudim

[. . .] [. . .]

inflection and government in arabic


The prepositions of the darf (which is ablative) are the following: con con ba bal

con bi

con bil

en fa

en fal

en fi

en fil . . .”

As we can see, Pedro de Alcalá did not separate the prepositions from the definite article (al). Mafūl is the term used for ‘object.’ Sībawayhi distinguishes five subtypes: mafūl bihi ‘direct object’, mafūl fīhi ‘locative object’, mafūl maahu ‘accompaniment object’, mafūl lahu ‘reason object’, and mafūl minhu ‘object from it’. (Owens 1990, 160). For the ‘locative object’, instead of mafūl fīhi the term zarf is also used (Owens 1990, 51,141–151), which is the term Alcalá uses here for the ‘ablative’. Munādā is the direct translation of ‘vocative’. According to Cowan (1981, 360), Alcalá ‘was apparently trying to make unfamiliar material intelligible to his audience, but at the same time to avoid a direct equation with the Latin categories.’ In fact, the first might be true, but we must be aware that in his paradigm of the case-system, we do find an equation with Latin cases, and we never find any traces of the four traditional Arabic inflectional endings, -zero, -u, -a, and –i and never the original Arabic names for these inflectional endings are introduced here. Other technical grammatical terms in Alcalá’s grammar are: amir (‘conocimiento’;13 cl. Ar. damīr ‘conjunct pronoun’), temiz (cl. Ar. tamyīz ‘accusative of specification’),14 amr (‘imperatiuo’; cl. Ar. amr ‘imperative’), xucla ‘señal’; cl. Ar. šakl(a) ‘orthographic sign’:15 Es otrosi de notar, que los Arauigos non tienen letras vocales como los Latinos, mas tienen en lugar dellas ciertas señales, que ellos dizen xúclas, con las quales y con todos los caracteres suso dichos leen y escriuen lo necessario (Alcalá 1505[1883], 4). It has to be observed that the Arabs do not have the letters for the vowels as the Latins, but instead of them, they have certain signs, which they call xuclas, and with all the above-mentioned characters they read and write anything which is necessary.


Cf. Zwartjes (1994). Temĭç is also translated as ‘conocimiento’ in his dictionary (Corriente 1988, 197). 15 In his dictionary Alcalá translates the word aâléma as ‘(signo por) señal’ (Corriente 1988, 140). The señal is also used in his grammar for the article (señal de demostración), so this term could be the technical grammatical term, but as happens often in these grammars, it is not always possible to distinguish between language and metalanguage. 14


otto zwartjes

The origins of the mnemonic vowel terms with and without tanwīn (the ‘nunated’ forms) which expresses indefiniteness in Arabic are unknown to me and this could be a local teaching method, not recorded elsewhere: TABLE 2 THE VOWELS AND THE ‘NUNATED FORMS’ ACCORDING TO PEDRO DE ALCALÁ (1505[1883], 4)16 minibĕ minibĭ minibŭ minenbĕn mininbĭn minunbŭn minĕb

fath a kasra damma fath atān kasratān dammatān sukūn16

a i u an in un ø

The remaining Arabic terms analyzed by Cowan are alif cequin (alif sākin ‘silent alif ’), and in his dictionary we find iarab (‘oración de gramática’; clas. Ar. irāb (the inflectional endings ø, a, i, u (see below)17 and harf (‘letra’; Cl. Ar. h arf), which are not analyzed at all in the grammatical treatise. Summarizing the preceding paragraphs, we can conclude that in most cases Pedro de Alcalá could easily use Latin terminology, such as the names for the cases or the imperative. The use of an ‘exo-grammatical’ term damīr for the conjunct pronoun does make sense since traditional grammar did not have precise equivalents from contemporary sources yet. Probably, Pedro de Alcalá understood very well that the conjunct pronoun in Arabic can be used differently from the Spanish pronouns; they can also be affixed to prepositions and nouns, for instance and that explains probably the reason why he used the Arabic term. In the remaining cases, Arabic terms are used for mnemonic or pedagogicaldidactical reasons. It is questionable if these terms made his teaching

16 The terms fath a, kasra and damma are not found in his grammar which gives us the impression that Pedro de Alcalá did not know them. 17 Again, we find more information in the dictionary, neglected by Cowan: iaráb is translated as ‘declinacion de palabras’ (Corriente 1988, 134), which is not unimportant because the author avoids the term ‘noun’ here, since irāb is used for nouns and verbs as well.

inflection and government in arabic


method more attractive or more easily accessible to his pupils.18 Seen through our modern eyes, the grammar of Pedro de Alcalá was a real novelty, a pioneer work. His transcription system of the Arabic alphabet is the first in Europe, although his knowledge of classical Arabic was evidently limited; there are quite a few inconsistencies while using his own transliteration system. It is also obvious that Pedro de Alcalá did not have any knowledge of Arabic grammatical theory and his Greco-Latin approach is particularly visible in the lacking of insights in the analysis of derivations. As had been observed by Fück [Pedro de Alcalá] “erkennt nicht die Bedeutung des Wurzelbegriffs, so dass dem Leser der arabische Formenbau in der Nominal- wie in der Verbalbildung undurchsichtig bleibt” (1955, 33). Nevertheless, as has been stressed by several scholars at the conferences on missionary linguistics,19 it is easy to point at the shortcomings of these grammars and dictionaries measured by the standards of our own time, but when studied and analyzed in its own historical and cultural context, there is no doubt that there is still a wealth of material to be studied. As observed by Smith Stark “certain preconceptions about pre-modern descriptive work have resulted in its neglect among those in the Western tradition” (2005, 4). Monographs and studies on the description of Arabic by Spanish missionaries are still non-existent. Positive evaluations are scarce when the grammars are concerned, whereas the majority agrees that the lexicographical work of Pedro de Alcalá and others have been of great value. Dannenfeldt’s observation, according to whom “Both of these [Alcalá’s] works are based on solid

18 Missionaries usually emphasised that the language they were learning was ‘easy’ to learn, although others label the language under description as ‘difficult.’ They tried to use the most transparent and less ‘obscure’ paradigms and explanations. The reason to re-write existing grammars was almost always because predecessors were too ‘obscure’. Probably, the grammar of Pedro de Alcalá could be perfectly understandable without the use of Arabic grammatical terminology. The same could be said of the use of Hebrew posodical-grammatical terms by Oyanguren de Santa Inés in his grammar of Tagalog (1742, 208–209), such as milehal (stress on the penultimate syllable, instead of the usual ultimate syllable, in connection with stress assignment), athnach (semicolon or pause) and metheg (one type of the several secondary accents, avoiding the loss of vowels in pronunciation, or a sign, pointing a vowel, which usually would be reduced to schwa but which is to be fully pronounced in this particular place). The terms ’atnach and meteg are both so-called ‘cantillation marks’ in the Hebrew Bible from Masoretic times. Did the pupils of Oyanguren de Santa Inés know Hebrew, or is this pure pedantry or snobbism? 19 The International Conference on Missionary Linguistics took place in Oslo (2003), São Paulo (2004), Hong Kong and Macau (2005), Valladolid (2006), Mérida (Yucatán, 2007) and the sixth will be organized in Évora, Portugal.


otto zwartjes

philological methods and are evidences of humanist learning in Spain” (1955, 33),—when compared to Fück’s (1955, 33)—is without any doubt an exception. Although Pedro de Alcalá’s grammar served as the model for the Gramática arábigo-castellana (still unpublished) composed by the Hiernoymite Patricio José de la Torre (1760–1819),20 and although we have some evidence the Maronite Joseph Simon Assemani of Syrian origin (1687–1768)21 quoted from his grammar, Alcalá’s dictionary was particularly widely known and used throughout Europe. Johannes Gabriel Sparvenfeldius possessed probably Alcalá’s catechism, which has been appended to his grammar and his dictionary already in 1706 but in that period it had become already a rare book (Schnurrer 1811, 16).22 If the original work was not longer available, scholars used the original editions or the re-edition from 1776 from Patricio José de la Torre with transcriptions into Arabic characters, which was the base for the Supplément of the Dutch Orientalist Reinhart Dozy (1820–1883) (Monroe 1970, 38). One of the main purposes was to publish the work of Pedro de Alcalá in Arabic script, something that was completely unnecessary according to Pedro de Alcalá. Although we find an entire page with the Arabic alphabet in his grammar, an observation below it tells us “all characters can be substituted by Latin or ‘Castilian’ letters”: Estos son los caracteres y nonbres de las letras arauigas, las quales todas se pueden suplir con nuestras letras Latinas o castellanas, de manera, que para la comun algarauia no ay necessidad de las saber ni conocer todas, mas solamente quarto conuiene saber kha, dil , te, ay, cuyos sones no tenemos en nuestro ABC latino. (Alcalá 1505[1883], 3–4) These are the characters and names of the Arabic letters, which all can be substituted by our Latin or Castilian letters, so that there is no need to learn or to know them all for the common speech, but only four [are necessary], namely the kha, dil , t e, ay, whose sounds we do not have in our Latin ABC.

We have seen that for didactic reasons, Pedro de Alcalá used these mnemonic words, which are not derived from canonical grammatical works of the Arabic tradition. Other grammarians found a different


Schnurrer (1811, 88, no. 128). BICRES III, no. 84. Schnurrer (1811, 16) observes that Assemani quotes from his grammar ‘ex grammatical recitat,’ but also adds that he actually used material from Alcalá’s dictionary (‘non Grammaticae sunt, sed Vocabularii,’ ibid.). 22 Ioan. Gabr. Sparvenfeld. 1706. Catalogus Centuriae Librorum Rarissimorum Manuscript.& partim Impressorum, Arabicorum, Persicorum, Turcicorum, Graecorum, Latinorum, &c. Upsala: John Henry Werner. 21

inflection and government in arabic


solution. One of the most remarkable solutions can be found in the Fabrica Arabica (1640) of Dominicus Germanus of Silesia (1588–1670). In this case, we have translations of the Latin names of the cases into Arabic. In the following table we see these literal: TABLE 3 THE CASES ACCORDING TO DOMINICUS GERMANUS’ FABRICA ARABIC (1640)23



al-mutasammī al-mutawallid al-mustatī al-muštakī al-munādī al-mustaqti

Nominativus Genitivus Dativus Accusativus Vocativus Ablativus23

3. The grammars of Lucas Caballero, Juan de la Encarnación, and Francisco Cañes: ‘Grammars at the crossroads of two systems?’ 3.1


After the foundation of a great number of missions in the East, the importance of the Arabic language for preaching the Christian faith continued to increase. Paul V in a papal bull dated 1610 had commanded the various religious orders to teach Oriental languages in their colleges. In the early 16th century, Arabic was taught in Seville at the Colegio trilingüe.24 The Franciscans decided to found colleges in Salamanca, Alcalá, Paris and Toulouse for the teaching of Arabic, Greek and Hebrew (Monroe 1970, 26). According to Monroe, Bernardino González (c. 1665–1735) composed an Arabic dictionary in Seville, which was completed by Franciscans in Jerusalem in 1709 (Monroe ibid.), an unpublished work. José de León began to compile a new dictionary of Arabic and Bernardino González was sent to Damascus in order to complete his work.25 As Monroe 23 From the verbs sammā ‘to denominate,’ walada ‘to give birth,’ atā ‘to give,’ šakā ‘to complain,’ nadā ‘to call’ and qatta ‘to cut off, to disjoin,’ the tenth derived form means ‘to deduct,’ which seems to be an approximate translation of ‘auferre.’ 24 In this short account, there will be no space to summarize the study of languages during the Middle Ages in Muslim Spain. 25 Franciscans had already arrived in 1233 in Damascus. Propagating the Christian faith was not permitted by Sultan Malik al-Ašraf, but they took care of Europeans who settled there. After several cases of martyrdom and imprisonment, the Cadi of the city


otto zwartjes

observed, ‘the work was the most complete ever to have been composed in Spanish after that of Pedro de Alcalá. [. . .] The work of González and his collaborators was lost until Asín Palacios (1901) came across a copy containing additions up to the year 1727 made by the friars of Damascus’ (Monroe 1970, 27). One of the successors of Bernardino González was Francisco Cañes who settled in Damascus at the Spanish Franciscan College in 1757 (Monroe 1970, 28). Cañes’ grammar of Arabic has been printed in two different editions, a first in 1775 (Madrid, Don Antonio Perez de Soto, and a second in 1776 (Schnurrer 1811, no. 113, BICRES 959 and 971), entitled Gramatica arabigo-española, vulgar y literal. Con un diccionario arábigo-español, en que se ponen las voces mas usuales para una conversación familiar, con el texto de la Doctrina Cristiana en el idioma arabigo. I came across another copy from the same Franciscan tradition, composed by Lucas Caballero and Juan de la Encarnación as we can read in the colophon, which escaped the attention of scholars who have worked in this field. The manuscript has been identified by Magnus and Aare Mörner in his Spanien i svenska arkiv. The title of this manuscript is Compendio de los rudimentos y gramática árabe en que se da notizia de la lengua vernácula y Vulgar y algunas reglas de la literal Iustamente, 1709, and in the colophon 1710 (another author, Juan de la Encarnación, finished the text San Diego, Seville). The work is based on Bernardino González as we can read in the title, and Lucas Caballero, “lector actual Arabo en el Colegio de Damasco” composed (‘recopilado’) this manuscript, which has been donated by Johan Adam Tingstadius (1748–1827),26 bishop of Strängnäs, Sweden, from 1803, to the Rogge library, which has belonged administratively to the Royal Library of Stockholm since 1968. As the titles of the grammars of Cañes and Caballero demonstrate, the language under description is not only classical Arabic, but the urban dialect of Damascus. Apart from Alcalá’s grammar of colloquial Arabic of Granada, European scholars usually did not pay much attention to lower registers, so the linguistic works of these Franciscans work-

granted the privilege of being able to open a public chapel and in 1668 the Franciscans established themselves in a Maronite church, which they left in 1719 when they acquired a new church in the Christian Quarter of Bāb Tūma. The foundation of the college where Arabic was taught dates from this period. 26 Tingstadius was a professor in Oriental languages at Uppsala. He published, for instance in 1770, a treatise entitled Dissertatio philologica de natura et indole linguarum orientalium communi (Uppsala: Johan Edman) and in 1794 his Dictiones arabicae ex carmine Tograi, hebraismum biblicum illustrantes. Uppsala: Johan Edman.

inflection and government in arabic


ing in Damascus are of great importance. However, they were not the only grammarians who described non-Classical registers. Antonio ab Aquila’s grammar published in 1650 is not only a grammar of classical Arabic (“ad grammaticae doctrinalis intelligentiam”) but also colloquial Arabic (“ad vulgaris dumtaxat idiomatic”), probably the reason why he called the grammar “Arabicae linguae novae et methodicae institutiones.” Dominicus Germanus (Germanus of Silesia; 1588–1670)27 composed a dictionary in 1636 with the title Fabrica overo Dittionario della lingua volgare arabica et italiana, copioso di voci e locutioni, con osservare la frase dell’una e dell’altra lingua (Roma. Nella stampa della Sac. Congr. De Propag. Fide) followed three years later by his Fabrica linguae Arabicae cum interpretatione latina et italica, accommodata ad usum linguae uulgaris et scripturalis (Roma. Typis Sac. Congreg. De Prop. Fid.). In 1800 a work has been completed by Franciscus de Dombay (1758– 1810) with the title Grammatica linguae Mauro-Arabicae juxta vernaculi idiomatis usum, accessit vocabularium Latino-Arabicum (Vindobonae: apud Camesina)28 but, according to Schnurrer, this title is misleading; although this grammar describes the common speech of the people in the Maghreb (“Arabicus sermo in Mauritania quo vulgus uti solet”), he observes that all words are ‘good Arabic’ (“non sunt vulgari idiomati propria, sed omnia bene Arabica”) (Schnurrer 1811, 95).29

27 Germanicus was a teacher of Arabic at the mission school St. Peter in Montorio, Rome. He assisted with the preparation of the Arab Bible, he published dictionaries, and commentaries on the Qurān. He was teacher and translator at the court of Philip IV of Spain. 28 I have not been able to consult this grammar yet. 29 Although this is not the aim of this paper, I wish to show just a few elements from these sources that are important records of colloquial Damascene Arabic from the beginning of 1700. Particularly the word lists are full of colloquialisms, but also the grammar of Lucas Caballero has many colloquial elements, to mention a few: omission of vowels: muqatla instead of muqātala, the use of the -u- vowel as a prefix for the imperfect tense (64), b- future suffix (p. 24), which is colloquial (in Egypt it is the present tense). However, Caballero is not always consistent, we find both faaltu as faaltum (2 person pl.m.), the use of -ī instead of -īna for the second person feminine singular in taf alī (= ‘Haztu fem.’) (72). It is remarkable that sometimes we find even hybrid forms, such as antum faaltu. However, we find also classical elements, such as the use of the feminine plural in the verbal paradigms, which is not used in colloquial urban speech today. It is also significant that the order of the persons singular in the verbal paradigms is not the traditional one 73v–72r. Cañes has 3 (masc. Sing, 3 fem. Sing, 3. plur. 2 masc. Sing. 2 fem. Sing. 2 plur. 1 sing 1 plur. Instead of 3,3,2,2,1 (sing.), 3, 3, 2, 2, 1. I am grateful to Manfred Woidich for his comments on this footnote.


otto zwartjes

Grammars of Classical Greek were available throughout Europe, but Spanish priests were engaged in the writing of vernacular Greek too, as we can read in the prologue of Francisco Cañes (1775, s.n.).30 In the following paragraphs we will treat the use of Arabic terminology in both grammars separately. 3.2

Semantic and inflectional vowels and the concept of amal

The questions to be answered now are: which Arabic grammatical terms are used, particularly in the sections dedicated to inflection and government, what did they mean and why did Caballero and Cañes use them? Let’s start with the vowel system, according to the paradigm of Caballero are: TABLE 4A bu ba bi


O que inclina a V A que inclina a E E que inclina a I

Unlike Pedro de Alcalá, Caballero gives also in an appendix of this Chapter the Arabic technical terms for these vowels when they are used as inflectional endings, accompanied with translations into Spanish: Los Gramaticos a estas mociones dan otros nonbres conbiene a saber que significan, eleuacion, ereccion, y contraccion: al-raf, al-nasb, wa-l-jarr y estas mociones duplicadas llaman tanuín (tanwīn) esto es nunacion que es

30 It must be emphasized that missionary sources, often written in Spanish, are in many cases the only existent sources which can give us information of vernaculars once spoken in early ages of languages of which we only have more detailed information concerning the literary or classical register. Priests understood very well that in China the teaching of Mandarin was not so useful in regions were other dialects were spoken. Grammars of modern Greek circulated since 1638, the Grammatica linguæ græcæ vulgaris was printed by Simon Portius. The first Spanish grammar of modern Greek was composed by Pedro Fuentes, as we can read in the following quotation from the same prologue: “Por lo tocante á la lengua griega ha impreso su gramatica vulgar Fr. Pedro Fuentes observante, que residió en el Seminario de Nicosía en Chipre, y ahora está imprimiendo la gramatica literal.” 31 In their tables of the vowel system, both authors also give the names of the vowels in Arabic script, not reproduced here.

inflection and government in arabic


lo mismo que decir addicion de un sonido de nun o n y biene a sonar on, an, en vg. Racholon, Racholan, Racholen, Homo, Hominem, Hominis [. . .] pero esto ueras mas claro en el tratado del nonbre (77 r.). The grammarians give other names to these ‘motions’ [h arakāt], namely the names which signify ‘elevation’, ‘raising’, and ‘contraction’, ar-raf , an-nasb, wa-l-jarr, and these ‘motions’, when duplicated, are called tanwīn, which is ‘nunation’ which is the same as adding the sound of a nūn, or -n, which approximately sounds as on, an, en vg. Racholon, Racholan, Racholen, Man (nom.), man, (acc.), man (gen.), but this you will see more clearly in the chapter about the noun.”

Cañes’ table resembles that of Caballero, although there are some differences. The order of the vowels is different, Spanish translations of the Arabic vowel terms are given, and instead of de verb ‘inclinar’ we find ‘declinar’: TABLE 4B ba bi bu


Fatha, Apercion Kesra, Fraccion Domma, Colleccion

A que declina á e, y a clara. Ba, ó Be. E que declina en e, ó i claro. O que declina en o, y u. como Bo, ó Bu.

Also here, we find almost the same observation as above, which demonstrates that the works are closely related to each other and that they probably derived from a common source (or sources): Los gramaticos à las tres mociones dichas les dan otros nombres, es á saber: al Fatha le llaman ‘Nasbo’, esto es ‘ereccion’, al ‘Kesra Charro’, ‘contraccion’, y al ‘Domma Rafo’, ‘elevacion’. Suelen tambien duplicar las dichas mociones de esta suerte ( ) ( ), y entonces les dan el nombre ‘tanuin’, esto es, ‘nunacion’, ò ‘nun’ vocal, que viene á sonar ‘an’, ‘en’, ‘on’. (Cañes 1775, ff. 8–9).


The grammarians give other names to these three above-mentioned ‘motions’, namely: they call the Fatha ‘Nasbo’ which means ‘erection’, they call Kesra ‘Charro’ which means ‘contraction’, and Domma Rafo ‘elevation’. They are used to dublicate the above-mentioned ‘motions’ in this way: ( ) ( ), and they give them the name tanwīn which is ‘nunation’, or vocal nūn which sounds rougly like: an, en, on.

In Chapter 4, both authors deal with the ‘cinco signos’: (1) Secun (‘quietud’), which has according to both the synonym chezm (‘caballero’), or chiasmo (Cañes), (jazm) which Caballero translates as ‘anputacion’ and Cañes as ‘corte’;

226 (2) (3) (4) (5)

otto zwartjes Texdid (‘duplicacion, ó corroboracion’); Maddo (‘extension ó produccion’); Vasalo (‘union, ò conjuncion’); and Hamza (‘punzamiento’).

A chapter dealing with how to read Arabic without vowel signs was appended by Caballero, not present in the grammar of Cañes. Although there are differences, both grammars are from the same tradition and both authors (re-)formulated probably the lost version of Bernardino González, or quoted directly from other sources, such as Agapito à Valle Flemmarum, who almost has the same definitions as Cañes’s, although the vocalizations of the Arabic terms and the order is slightly different: TABLE 4C bu ba bi


Dzhamma, collectio Fathha, aperitio Kesra, fractio

o declinans ad u, & u claru ” a declinans ad e, & a claru ” e declinans ad i, & i claru ”

The ‘Tratado III’ deals with the noun. Here we find sometimes some parallels with the grammar of Pedro de Alcalá, particularly since the term ‘señal’ has been used in both sources, or ‘notificación’ for the article33 and the ‘abitudines’ of Pedro de Alcalá resemble much the ‘señales del nombre’. Since colloquial Arabic nouns are not inflected according to cases, we find in the grammars of Alcalá, Caballero and Cañes equivalents of the Greco-Latin cases for didactic reasons: Los arabes aunque en la lengua vulgar reconocen tres numeros en el nombre, es saber: ‘singular’, ‘dual’, y ‘plural’ no conocen distinction de casos. Y asi el nombre en qualquier caso termina con una misma voz. (Cañes 1775, 59). The Arabs, although they recognize in the colloquial speech three numbers in the noun, being, singular, dual and plural, they do not know the distinction in cases. And that’s why the noun ends with the same sound in whichever case.

32 The vocalization of Martelottus is slightly different: Dzhammon, Phathhon and Kafron. 33 Pedro de Alcalá uses ‘señal de demostración’. See Zwartjes (1992).

inflection and government in arabic


Elsewhere in his grammar, i.e. after the ‘Numeros cardinales’ and ‘Numeros ordinales’ we find a comparable description: Los Arabos en la lengua Vernacular o Vulgar no tienen terminacion de casos en el nombre assi como los Castellanos (16v.). In the vernacular or vulgar speech, the Arabs do not have the ending of cases in the noun as the Castilians have.

The case system according to Caballero is as follows: TABLE 5


Nominativo Vocatiuo

raf  un


Vir, o vir



jarr un


Virum o Virum Viri Viro A viro

(this column in Arabic script)

Acusativo Vocatiuo Genitiuo Datiuo Ablatiuo

What did these terms mean in the Arabic grammatical tradition? Before Sībawayhi, no distinctions have been made between the vowels which are used in classical Arabic for the declensional endings, and the other vowels, for instance: there was no disctinction between both vowels ‘i’ in the ‘genitive’ al-kitâb-i. An important novelty of Sībawayhi is that he distinguished the first ‘i’ that is non-declensional, from the final ‘i’ which is ‘declensional’ (Versteegh 1997, 19). When we analyze Arabic terminology in our 18th century grammars, the sections about the particles are even more interesting. Cañes and Caballero use both a metalanguage inherited from the Arabic tradition that had already been developed by Sībawayhi: These endings follow eight courses: accusative (nasb), genitive ( jarr), nominative (raf ), apocopate ( jazm), a-vowel ( fath ), i-vowel (kasr), u-vowel (damm), zero-vowel (waqf ). (Translation by Versteegh 1997, 36).

The main distinction is whether a certain vowel is declensional or not. The Arabic terms used by Caballero have all to be related to the Arabic term amal that generally is translated as ‘governance’, or ‘dependence/ dependency’, which resembles the 20th century theory of ‘government and binding’. As explained by Owens:


otto zwartjes The governor is said to govern the governed in some case or mode form (irāb): For the nouns these forms are: u nominative (raf ); a accusative (nasb); i genitive ( jarr); For the verb only the imperfect verb shows mode inflection: u indicative (raf ); a subjunctive (nasb); Ø jussive ( jazm).(Owens 1988, 39).

Carter demonstrated in several publications (1991, 1993, 1994)—dealing with different grammatical theories—that the translation of technical terms are in many cases inexact, problematic, anachronistic and can lead to misconceptions.34 Carter argued that the basic meaning of the concept of amal is different, stressing that the interpretation of Weiß: “amal und regere: ‘Es ist nicht recht verständlich’, wie man hier schwanken kann. amala fī kann ja doch gar nicht heißen ‘Gouverneur sein über. [. . .]’. amala fī ist ‘tätig sein, arbeiten’.” The translation ‘governor’ is according to Weiß an example of wishful thinking (‘Wunsch der Vater des Gedankes’), because the concept of governance was already widely used in the Greco-Latin framework. A translation that brings us closer to the original meaning “an etwas arbeiten, auf etwas einwirken” (Weiß 1910, 384). Particles (h urūf )35 can also be defined as ‘governors’, since they can ‘govern’ cases. The subclasses of some particles can be defined in terms of ‘dependency’, i.e. which nominal or verbal ending they ‘govern’. If a particle (h arf ) ‘governs’ a genitive, such particles are called h urūf jarr (Owens 1988, 10), for instance bi (‘by means of ’), or min (‘from’) etc. In both the Compendio delos Rudimentos y Gramatica Araba of Lucas Cauallero, as the Gramatica arabigoespañola, vulgar, y literal of Francisco Cañes we find hispanicised forms of these subclasses of particles.36 At the beginning of the chapter dealing with the parts of speech, both Caballero and Cañes give us the tripartite division of the parts of speech, which is an ‘Arabic’ division, according to Cañes (‘Pero los arabes las

34 The same happened when Romans translated Greek grammar. An illustrative example is the term casus accusativus which is a wrong translation of the Greek term ptosis aitiatike. It is not the “anklagender Fall,” but “das von der Handlung Betroffene, dasjenige, dem etwas gechieht.” (L. Lersch: Die Sprachphilosophie der Alten, Bonn 1838– 1841. vol. 2, 186, quoted in Carter 1993, 131). 35 H arf does not only mean ‘particle,’ since it has in fact much more meanings, such as ‘edge, letter, sound, word.’ See for a detailed overview of the most important meanings the first Appendix of Owens (1990, 245–248). 36 We have not been able to consult a dictionary, which is particularly devoted to the particles, the Diccionario de partículas árabes [s.a.], composed by Mariano Rizzi y Franceschi (18th century; BICRES III, no. 71).

inflection and government in arabic


[partes de la oracion] reducen a tres, que son: verbo, nombre, y particula (h arf ).’ As a matter of fact, the authors follow this tripartite division, but in the second chapter of the 4th treatise, which deals with the particles, the author prefers to follow the Latin model: Capitulo II. De las particulas separadas. Muchos son los generos, que hay de estas particulas, y para evitar confusion, las dividiremos al modo de los latinos, en ‘adverbios’, ‘conjunciones’, ‘preposiciones’, y ‘interjeciones’. (71). “Chapter II. About the separate [= not suffixed] articles. The subcategories of these particles are many, and in order to avoid confusion, we divide them, according to the Latin manner, into adverbs, conjunctions, prepositions and interjections.”

In the pages dealing with classical Arabic, a separate chapter following the final chapter in syntax, Francisco Cañes again prefers the Arabic model. In the following quotation he deals with them after the chapter on syntax describing classical Arabic: Aqui es preciso advertir, que asi como el nombre se declina por la variacion de las terminaciones, que tiene en los casos, como adelante se verá; igualmente el verbo en el ‘modsareo’ se conjuga de tres maneras: por variarse su terminacion en las personas. Esta variacion proviene, de que los arabes anteponen al verbo en el ‘modsareo’ dos generos de particulas, que se llaman ‘nasbantes’ y ‘chazmantes’. Se llaman ‘nasbantes’ del verbo ‘nasaba’ puso, plantó, fixó, porque en fuerza de ellas el ‘domma’ de la tercera radical del verbo se pierde, y convertido en la mocion ‘fatha’, la fixan, y plantan sobre la dicha radical, sin que padezca mutacion. Las ‘chazmantes’ se llaman asi del verbo ‘jazama’ secó,37 cortó, porque estas particulas le cortan la mocion á la tercera radical; y poniendole el signo ‘secun’, la dexan quiescente, ó liquida. (Cañes 1775, 109–110). Here, it is necessary to observe, that as the noun declines through the variation of its endings which they have in the cases, as we shall see below, so on the same manner the verb in the mudāri38 conjugates in three ways: by varying the endings according to persons. This variation results from the fact that the Arabs put before the verb in the mudāri two classes of parti-


Probably ‘seccionó’, since ‘secar’ means ‘to dry.’ Usually translated as “imperfect,” literally ‘the resembling verb’ (Baalbaki 2004, XIII, 23), because they resemble the nouns, since both share the same declensional vowels (irāb) -u (‘subject’ and ‘indicative mood’; and -a which is ‘direct object’ and ‘subjunctive mood’). The nouns do not have zero endings (‘apocopate’ or jussive mood in the verbal system), whereas the nouns have the -i ending (‘genitive’), which is not present in the verbal paradigm. 38


otto zwartjes cles, which they call ‘nasbantes’ and ‘chazmantes’ [= which co-occur with the nasb (a ending) and those which co-occur with the jazm (Ø ending)]. They are called nasbantes from the verb nasaba, to put up, to plant, to fix, because through their force, the domma [damma] of the third radical of the verb is cut off and converted into the fath a [a ending/‘motion’]; they fix this radical and settles down on it, while it does not suffer any change. The chasmantes [those which co-occur with the zero-ending] are called so from the verb jazama, to cut off, to truncate, because these particles cut off the vowel [‘motion’] from the third radical, and put on this the sign sukūn [‘motionless’, ‘vowelless’] and leave them ‘silent’ or ‘liquid’.

In Chapter IV, dealing with the noun, Cañes gives us three classes of particles, ‘charrantes, chazmantes y nasbantes’: Particulas ‘charrantes’ son unas preposiciones, que antepuestas al nombre le colocan en el caso ‘charro’, ó ‘genitivo’ (139) [. . .] Aqui se debe advertir, como en arabe lo mismo es de decir particula ‘nasbante’, que en latin preposicion de acusativo, y asi antecediendo al nombre le colocan en dicho caso, [. . .] Estas particulas á manera de los verbos, admiten afixôs, y rigen los nombres, colocando el sujeto en ‘nasbo’, ó ‘acusativo’, y el predicado en ‘rafeo’, ó ‘nominativo’, de suerte, que se viene hacer una permutacion del nominativo con el acusativo . . . (143) the particles which are ‘charrantes’ are some prepositions, which placed before the noun, put it in the jarr [i] ending, or genitive. Here it must be observed, that as in Arabic it is the same to call them ‘particular nasbante’, as in Latin, prepositions which combine with the accusative, and as such they are placed before the noun, they put it in this so-called case. [. . .] These particles, when combined with verbs, allow affixes, and govern the nouns, placing the subject in nasb, or accusative and the predicate in raf , or nominative, so that there is a permutation from the nominative to the accusative.

Caballero uses the same classification and terminology, although his hispanicised forms are slightly different: ‘particulas charrantes, nassuantes’, and ‘chesmantes’. Although in Cañes’ description, only classical Arabic is commented on, Caballero also dedicates a paragraph on the ‘particulas Chesmantes, ojo, Vulgares’ (particles which correspond with the optative ‘Utinam’ and other Latin forms, such as quare, quia, quoniam, etc.). If we compare the use of Arabic terminology in the works of Caballero and Cañes with those used by Erpenius, we can conclude that the latter did not use Arabic terms so extensively. Worthy of mention are for instance: Fatha, damma, kesra in the first book ‘De elementis . . .’. We find a Latinized form of the term jazm: ‘post gjezman constanter manent’

inflection and government in arabic


(1620, 20), also used as the verb Gjezmare, gjezmant, or in the passive form gjezmatur (47), verbum hamzatum (70). The term motiones is not used as a translation of h arakāt but it for the change a noun undergoes if the feminine ending is added to the masculine form.39 Nunnatio is used as well (141), but the Arabic terminology for inflectional endings as used by the Spanish missionaries is not recorded. 3.3

Possible sources

The use of non-Western metalanguage in itself has its own tradition. Not only Pedro de Alcalá used non-Western terminology, but in Northern Europe we see also that Hebrew grammars used Hebrew terminology in an adapted form.40 Which sources could the Franciscans have used? In the prologue of the grammar, Cañes informs us that he completed an eclectic grammar in agreement with the ‘taste of everyone’ (“que sea del gusto de todos”), using the most useful aspects (“he procurado aprovecharme de lo bueno que en ellas he visto” ibid.) of earlier sources and adding material from his own 16 years long experience: He procurado con el mayor cuidado y desvelo leer, y releer para el ajuste de esta, las gramaticas de Fr. Pedro de Alcalá, de Fr. Felipe Guadañoli, de Tomás Erpenio, de Fr. Antonio de Aguila, de Fr. Agapito de Valle flammarum, de Fr. Francisco Gonzalez, & c. Asimismo me he valido de un considerable numero de manu-scritos, que me han franqueado gustosos algunos aficionados á la lengua arabe. Finalmente he aprovechado lo que me enseñó el estudio, y la experiencia por espacio de diez y seis años, que estube predicando, y confesando en arabe en las misiones del Asia (ibid.). I have taken the greatest care to do my best to read and read over again the grammars of Fr. Pedro de Alcalá, Fr. Philip Guadagnoli, Thomas Erpenius, Fr. Antonio de Aguila, Fr. Agapito de Valle Flammarum, and Fr. Francisco González, and others. Likewise I have used a considerable amount of manuscripts, which some ‘aficionados’ of the Arabic language have passed to me. Finally I have benefited from what study and experience


“Motio est nominis Masculini in Foemininum converse; sitque additione terminationis foemininae” (Chapter V). 40 Johannes Reuchlin, who published his Hebrew grammar one year after the publication of Alcalá’s, introduces the verb dagessare: “quando he uel aleph repellunt nun passiue significationis, dagessatur prima ut . . .” (1974 [1506] Liber III, 590). See also Geiger (1871, 129). The presence of Oriental elements in Western grammars is ofcourse not only present when loans are used. Translations and ‘mistranslations’ produced sometimes terms which are not longer recognized as from ‘oriental’ origin. An example is the history and development of the concept of radix (root).


otto zwartjes have taught me over the period of sixteen years that I spent predicating and confessing in Arabic in the missions of Asia.

To start with the first grammar mentioned, Pedro de Alcalá’s Arte, we can conclude immediately that this grammar has not been the source for the terms used for the inflectional endings. Thomas Erpenius has been apparently used, but we observed that Erpenius reduced the ‘exotic’ grammatical terms in his grammars,41 although he maintained them in his Latin translation of Arabic treatises written by Ibn Ājurrūm and al-Jurjānī, which could have been also the direct sources of Cañes. Ibn Ājurrūm was born in Morocco in 1273–1274 and died in Fez in 1323. He is the author of a grammatical compendium entitled Muqaddima al-Ājurrūmīya Mabādi ilm al-Arabī where he exposes the inflectional system of Arabic, called irāb. This treatise on syntax has been widely used until the present day and it is one of the later works ‘downstream’ the long tradition starting with Sībawayhi. The Muqaddima was known in Europe since the 16th century.42 This work has been printed for the first time in Europe in 1592 (Medici, Rome). A translation by Peter Kirsten (1577–1640), into Latin appeared in 1610 (Breslae, 1610), followed by a translation of Erpenius (Leidae, 1617). In 1631, another translation appeared by Thomas Obicini (1585–1632).43 The Kitāb al-awāmil al-mi’a n-nahwiyya (“Book of the Hundred ‘Regentia’ ”) of Abd al-Qāhir al-Jurjānī (died in ca. 1080) has been translated by Erpenius in 1617 and published together with the Muqaddima: Grammatica Arabica dicta gjarvmia & libellus centum regentium cum versione Latina & comentarijs (Leidae, 1617). As we shall demonstrate below, scholars in Rome, such as Philip Guadagnoli, knew this work. Erpenius used in his translations of these works ‘loans’ from Arabic in his specified grammatical terminology. To mention an example:

41 In this article, we quote from the Rudimenta (1620). For a more complete analysis, the Grammatica Arabica, quinque libris methodice explicata a Thoma Erpenio (Leidae, 1613) and the Grammatica Arabica (Leidae, 1636) have to be taken into account. 42 See his article ‘Ibn Ājurrūm’ (Encylopedia of Islam. New Edition, 3, 697), and Ben Cheneb (1927, 381–382). 43 Thomas Obicini (1585–1632) was abbot of the Franciscan convent at Aleppo between 1613 and 1619 and in 1621 he returned to Rome where he founded the college at the St Peter Convent of Montorio where arabic was taught for the missionaries who were being prepared to spread the faith in the East. He was responsible for the supervision of the type designs of Oriental types at the Propaganda Press.

inflection and government in arabic


partes autem eijus sunt Rafa, Nasba, Chafda and Gjezma, e quibus convenit Nominibus Rafa, nasba and Chafda, non autem Gjezma: verbis vero Rafa, nasba, and Gjezma: non autem Chafda (f. 11)

However, as has been demonstrated by Fück (1955, 68), Erpenius replaced in his own grammars, when possible, Arabic terminology by Latin equivalents, and his Rudimenta could not have been either the source of inspiration of Caballero and Cañes either. Erpenius maintained the Arabic terminology of the vowels: TABLE 6 Nomen


Figura Potestas







Nunc a purum & clarum ut in amabam, nunc cum e mixtum, id est η Graecum, ut multi id nunc pronuntiant45 Nunc u purum & clarum, nunc cum o mixtum, id est o obscurum i simplex

Erpenius does not maintain the Arabic terms of the subclasses of the particles, as he rendered them in a Latinized form in his translation of Ibn Ājurrūm, but obviously he attempted to fit them into the Latin model.46 If we compare the translation into Latin of Erpenius with Obicini’s, we can conclude that the latter also used the Arabic terms for inflectional endings.47 Obicini firstly gives the Arabic term, written in the Arabic script, then a translation is given followed by a description or paraphrase with the purpose to explain the Arabic terms: ar-raf as ‘elevatio’, the definition of an-nasb is ‘accusativus, quasi patiens positum sub agente’ (without translation), al-xaf is rendered as ‘depressio, & amplectitur

44 In the left column the terms are also written in Arabic script, and in the second column the Arabic letter b is given, together with its appropriate vocalisation. 45 “Sometimes a pure and clear a as in amaba, other times mixed with an e i.e. as in Greek η as many pronounce it now,” etc. 46 For instance, we find definitions such as: de syntaxi Particularum: Praepositiones omnes tum separatae regunt genitivum ‘fī baitin . . .,’ instead of the terms ‘nasbantes,’ etc. Nevertheless, in his Rudimenta we still can find verbs as “gjezmare” (1620, f. 47). 47 “Rafaa, & Nasba, & Chafda, & Gezma. At nominibus ex ijs (conueniunt). Rafaa, & Nasba, & Chafda, non autem Gezma, Verbis verò ex eisdem, Rafaa, & Nasba, & Gezma, non autem autem Chafdha” (Obicini 1631, f. 3). Agapito à Valle Flemmarum has almost the same definition (1687, 194).


otto zwartjes

genitivum, dativum, & ablativum, quasi obliquus’, and finally, al-jazm as ‘abscisio, idest casus privatio, quasi à dictione cadentium scindat’. This means that Obicini decided to maintain Arabic terminology, and when explained appropriately, the learner could take advantage of these Arabic terms which were developed for describing the Arabic language. And now let’s come back to the concept of āmil fī. In Troupeau (1962), the most important translations in this period is listed: In Kirsten’s and Obicini’s translation we have ‘agens’, whereas the first also uses ‘efficiens’ and the latter uses also ‘regens’. Erpenius uses both ‘regens’ as ‘operans’. As Michael Carter observed, we see until today that the basic meaning has been neglected by many scholars: Anderseits kann man fast an eine Verschwörung glauben, die Grundbedeutung des Terminus amal auch heute noch zu unterdrücken, um den unbegründeten Mythos zu verewigen, der arabische Begriff vom strukturellen Verhältnis zwischen Satzteilen sei identisch mit dem lateinischen, d.h. eine Art ‘Rektion’. Daß die Grundmetapher der lateinischen ‘Rektion’ eine durchaus hierarchische, senkrechte Einordnung der Satzteile voraussetzt, die arabische amal ‘Operation’, aber im Gegenteil eine waagerechte Beziehung schildert, scheint die moderne Sprachwissenschaft nicht anerkennen zu wollen. (Carter 1993, 133).

Carter observed that Weiß was an exception and demonstrated that the translation by Kirsten unfortunately did not have any impact on later translations or interpretations: “Wie eine Stimme in der Wüste verbleibt noch der Aufsatz von J. Weiss. Die wörtliche und systemtreue Übersetzung operans von Kirsten hat sich leider nicht durchgesetzt” (Carter 1993, 134). In the first place we have to add that Franciscus Martelottus and Philip Guadagnoli (1596–1656), both neglected by Troupeau, also deserve our attention in this context, since we find in their grammars of Arabic a very precise analysis and translation of the concept. To start with the latter: the concept of al-irāb is translated as ‘arabicatio’ and amal as ‘efficientia’ or ‘operatio’, and al-awāmil as ‘regentia’: Ratione autem, quo vnu quodque in suo vel Casu vel Modo & quacumque dispositione locatur, quam Constructionem latini dicunt, Arabes interdum ‘al-irāb’, ‘Arabicationem’, proprius autem ‘amalun efficientiam’, seu ‘operationem’ appellant. Nomen autem illud, seu Verbum, seu Particula, eiusmodi operationem exercens, scilicet cuius vi aliquid in tali vel tali dispositione locatur, dicitur, ‘āmilun’ ‘operans’, ‘regens’. Locatum verò ex eius dicitur ‘mamūlun operatum’, ‘rectum’. Exinde dicuntur, ‘al-awāmilu regentia’. Quibus omnibus notis, nihil superest Grammatico. Collegit autem quidam,

inflection and government in arabic


cognomine Giargianius, in libello Regentia omnia, quem propterea nominavit ‘De Centum Regentibus’. Regentia enim, vel sunt ‘lafziyyatun explicita’, vel ‘manawiyyatun implicita’.48 (Guadagnoli 1642, f. 248). However, the system according to which a constituent has to be placed in the appropriate case or mode and on which convenient position has to be placed, this system which the speakers of Latin call ‘Constructio’ [= syntax], is system is called al-irāb by the Arabs which means ‘arabicization’, but they call this more appropriately amalun ‘efficiens’ [an act which produces a certain effect], or ‘operation’ [an act caused by force of an ‘operator’]. When a Noun, a Verb or a Particle produces such effects in this manner, and when by force of this effect something has to be placed in a certain disposition, we call this āmilun [‘producing a certain effect upon something’, ‘to govern’]. The element which has been effected is called mamūlun [‘the governed’]. Consequently, al-awāmilu are called regentia [‘governers’]. Although all this is well-known, nothing has been transmitted by ‘The Grammarian’,49 but someone with the name Giargianius has collected in a booklet all the ‘regentia’, which can be subdivided in two subclasses, the lafziyya (‘expressed’) and the manāwiyya (‘abstract’).

As we see, Guadagnoli’s description is not only accurate and complete, since the original Arabic is given and the literarly meaning had been maintained, but he also mentions his source, which is the hundred ‘regentia’ written by al-Jurjānī (‘Giargianius’).50 Do we find traces of these translations of amal other grammars written in Latin of this period? The answer is positive. In chronological order we shall summarize what other grammarians from Rome taught us, starting with one of the earlier grammars that appeared in Rome after the Medici translation of the work of Ibn Ājurrūm, the Institutiones of Franciscus Martelottus. Martelottus not only mentions Erpenius in his prologue (1620, 38),51 but he also deals with methodology. Should the Arabic model with

48 ‘Operators’ can be ‘expressed’ (lafziyy), and ‘abstract’ (manawī). “The first class are the particles or verbs or nouns that are either actually uttered or elided but understood, while the latter are abstract causes that do not involve uttered or restored linguistic elelements.” (Baalbaki 2004, XV, 23–58). This means that elided elements can produce ‘effects.’ 49 We could not identify this ‘Grammarian.’ 50 Also in other paragraphs we see direct translations in the work of Guadagnoli, which remain close to the Arabic original, such as ‘ignoratum,’ for majhūl, usually erroneaously translated as ‘passive,’ opposed to ‘cognitum’ for the ‘active’ (marūf ) (Guadagnoli, 1642, f. 255). 51 “Scripsit autem eleganter admodum de dictionibus hisce, quemadmodum & de litteris eruditissimus Orientalium linguarum in Leidensi Academia professor Thomas Erpenius uniusquam leuasset quoq. nobis huiis secundi libri labores partem” (1620, 38).


otto zwartjes

its technical terms be maintained, or should they be abandoned and replaced by Latin terminology? Martelottus decided to follow the traditional method when dealing with the single word classes (“Priora vero hic loci propriè de singulis dictionibus singillatim ordine debito, ac recta methodo explicanda sunt” (1620, 37), but we can infer from his preface, he obviously follows the Arabic model, and he does this explicitly.52 However, Martelottus does not leave the Arabic terminology out of his volume: Ubi omisso Arabicorum grammaticorum ordine Latinis admodum dissono, in rebus pluribus nobiscum, quoad methodum conuenisse comperimus. Caeterum quamvis à praedicta Arabum methodo deflectentes, nostro nos ordine Latinis magis consono procedamus, omnino grammaticalibus vocibus ubique utemur, eorumque ordines, ac procedendi rationes, divissiones, ac series enucleabimus, ut faciliro cunctis ad eosdem perlegendos authores pateat aditus. (Martelottus 1620, 38). Whenever the difference between Arabic grammars and the Latin way of construction was left out of consideration, we discovered that, with respect to the system of the language, Arab had many similarities with ours. After all, to what extent we ourselves may differ from this language system of the Arabs and are differing more and more as well from the Latin way of construction, we should still use in our research the Arab terminology and developing our knowledge we will explain the systems, the methods of construction and concatenation, in order to make it easier for us all to read the same authors.

Martelottus’s methodology is in our eyes extremely modern. It tries to bridge the gap between ‘exo- and endo-grammaticalization.’ An eclectic approach, combining the best elements of both traditions and their corresponding technical terms is the best way to understand the Arabic language. Martelottus also dedicates an entire chapter to inflection (“De irāb, seu Inflexione”), translated as ‘Arabificatio’, or ‘Arabicatio’ (1620, 98). The concept of āmil is translated again as ‘operans’ and the author quotes directly from the Arabic grammatical tradition: ‘irāb’ apud Grãmaticos ‘taġyīru awākhir al-kalām ilā al-ikhtilāf alawāmil:’ variatio ultimorum, seu extremitatum dictionum, ob diversitatem operantium (1620, 98).

52 In his grammar we find Arabic terminology extensively, together with the Latinized form, for instance the traditional Arabic classification of the consonants: “Chalchiiton, lahuiiaton, sciagiariiaton, asliiaton, natiiaton, dhalchiiaton, sciaphahiiaton, liniiaton” (1620, 35).

inflection and government in arabic


Variation of the final parts or extremities of the words, caused by the diversity of the operators (‘by the different effects produced by the operators’).

At the end of the 17th century, Agapito à Valle Flemmarum treats the particles in detail in his chapter entitled “De syntaxi Particularum,” and particularly describes the ‘effect’ they have as ‘operators’ on the inflectional system, translating al-h urūf al-āmila as: 1. 2. 3. 4. 5.

“particlae ‘operantes’ in nomen simplex Giarrantes, et in nomen simplex nasbantes, et in summam, et in verbum nasbantes, et ‘operantes’ in verbum Giezmantes” (1687, 268).

Another Franciscan, Antonio ab Aquila, also used Arabic terminology for the particles, although we do not find the same definitions. In his classification of the subclasses, we find a Latin terminology, such as ‘particulae coniunctae’ and ‘particulae separatae,’ with subclasses, such as ‘particula interrogationis, iurandi, status in loco, ad determinationem, optativi, conformandi, negationis, conditionalis’ (1650, Caput XXV, 388 ff.). However, he also uses Arabic terms, such as ‘particula Giàzemeh’ (f. 403). And finally, the question rises wether we find traces of the concept of āmil in the two Spanish grammars. The answer is again positive, as we shall demonstrate. Lucas Caballero gives rules for classical Arabic (‘gramática literal’) where particles have a certain ‘effects’: De las particulas charrantes Despues de las particulas que hemos puesto en el Arabo corresponden a nuestras quatro ultimas partes de la oracion, de las quales particulas, la mayor parte son literales; siguiendo el intento de dar algunas reglas de la grammatica literal, quiero explicar algunos efectos que tienen las particulas puestas o otras que a ellas siguen . . . (Cabellero 1709, f. 7r.; emphasis is mine) After the particles we have put in Arabic, four final parts of speech correspond with ours, whose major part are literary utterances: following our intention to give some rules of literary grammar, I want to explain certain effects these particles have when placed, or other which follow after them.

In the grammar of Cañes, we find exactly the same term, hispanicised as ‘operacion:’


otto zwartjes . . . quando se les une la particula mā quedan absolutamente privadas de su operacion, y regimen, de suerte que pierden la fuerza que tenian de colocar el sujeto en el caso ‘nasbo’ y se queda en el ‘rafeo’, ó nominativo, como ciertamente que escrivir á Pedro, ‘innamā sayaktubu Butrus’ (Cañes 1775, 144; emphasis is mine). . . . when combined with the particle mā they remain absolutely deprived of their ‘operation’ and ‘government’ so that they loose their force which they had before to put the nasb (a inflection, or direct object) and it remains in the raf (u ending, or subject) or nominative, as in ‘certainly to write to Peter’ ‘innamā sayaktubu Batrus’.

To sum up, we have found the following authors as possible sources of the two Franciscan grammarians of Damascene Arabic: Ibn Ājurrūm and al-Jurjānī (through translations of the Medici edition, Erpenius, Obicini and maybe also Kirsten) as the main Eastern sources. Western sources mentioned by name are Erpenius, Golius (mentioned by Cañes), Agapito à Valle and Guadagnoli. It has to be observed that missionary grammarians in Rome were familiar with some of the most important Arabic authors. The name of Al-Xalīl b. Ahmad al Farāhidī (died in 780) who codified and established a system of 15 meters has been mentioned by Guadagnoli. Since the grammars of Caballero and Cañes do not have a final chapter on prosody, they did not deel the need to use this source. Another prominent pioneer, who established the ‘foundations’ of Arabic grammar is of course Sībawayhi, whose work became known in the West through translations of Jahn. It is surprising that we find already in 1620 his name in the grammar of Martelottus, which is evidence for the fact that missionary-grammarians in Rome knew who he was. Dealing with the ‘verba ternaria,’ Martelottus explains that the ‘forma masdari’ is irregular, and in that context he mentions Sībawayhi: In verbis ternarijs, ut iam diximus, forma masdari irregularis est, omnes autem quidem Author Arabicus, nomine Sibauai ad 32. reuocauit, videlicet (1620, 213). In the ternary verbs, as we said before, the masdar-form is irregular; indeed, the Arabic author with the name Sībawayhi reduced all these to 32.

The same author, Martelottus, also mentions another Arabic source in his chapter dealing with the ‘Constructio Particularum’ where he divides the ‘particulas operantes’ in five subclasses, in agreement with an Arabic work called Lucerna, or Al-Misbāh , the “particulae operantes in nomen simplex giarrãtes, particulae operantes in nomen simplex nesbantes, in verbum nasbantes, in verbum gezmantes,” and the original text in

inflection and government in arabic


Arabic script is given in the same table as well. Although Martelottus does not give the name of the author, we think this work is probably the treatise written by al-Muta rrizī, entitled Al-Misbāh fī ilm an-nahw. Al-Muta rrizī (1144–1213) compiled this treatise which became a textbook in the madrasas of the East. The Misbāh itself wa sbased on three small grammatical monographs of al-Jurjānī (Lichtenstädter 1936, 847 and Sellheim (EI [New edition] 7, 773).

4. Conclusion Summarizing, we can conclude that Pedro de Alcalá’s use of Arabic terminology seems to be unsystematic and the reasons why he used them remain unclear, particularly when he uses the Arabic names for the cases. They do not reflect the Arabic inflectional endings and his model was obviously Latin grammar. The use of the term damīr is an exception, since the suffixed pronouns can be used in a different wayattached to verbs, nouns and particles—compared to Latin. Pedro de Alcalá’s mnemonic terms of the vowels stand alone, and we do not find any use of them in other works and in his dictionary we do not find the terms fath a, kasra, and damma. Although we find in Alcalá’s dictionary the translation ‘obrar’ for the Arabic root amala there are no traces of Arabic theory concerning ‘operators’ or ‘government’, related to this term. The earliest translations of al-Jurjānī and Ibn Ājurrūm are without any doubt an important milestone for the development of the study of nonWestern grammatical theory in the West and probably for some of them a real new ‘discovery,’ which can serve as an enrichment of the Western system, as Martelottus postulates. Direct influence of these works can be found in the grammars of Martelottus, Ab Aquila, Agapito and Guadagnoli, and the Franciscans in Damascus continued this tradition. Although Caballero and Cañes were not the pioneers themselves, they were probably the first grammarians who introduced extensively Arabic terminology in the Spanish metalanguage, as we have demonstrated. In Northern Europe, however, we see a different approach. Erpenius did not aim at orientalising Western grammatical terminology, except for the terms for the vowels and some other terms. The concept of āmil has followed its own course in grammatical theory and in the 20th century it was absorbed by anachronistic terms such as ‘government’ and ‘dependency’, as Carter demonstrated. Nevertheless, we have


otto zwartjes

demonstrated that the Franciscans were totally aware of the right connotations of āmil although we have to admit that sometimes they use ‘gobierno’ or ‘regimen’ as a synonym for ‘efficiens’ or ‘operans.’ Guadagnoli and Martelottus gave us without any doubt the most detailed analysis, and probably Caballero and Cañes have been inspired by their works. It was surprising that not only works of Ibn Ājurrūm, al-Jurjānī and al-Muta rrizī are mentioned by name by some of the grammarians working in Rome, but even Sībawayhi is mentioned by name in this relatively early period (1620). Missionary linguistics in Rome, particularly the achievements of scholars and teachers who published grammars in the seventeenth century at the Polyglot Press of the Sacra Congregatio de Propaganda Fide deserve to be studied more in detail in future. Particularly those authors who tried to combine exo- and endo-grammatical terminology and approaches have been innovative. How the learners of Arabic appreciated this ‘bridging’ approach is another question. Many scholars preferred in their teaching curriculum the more Latin-based grammar of Erpenius and his work was without any doubt a great success during many centuries. However, I agree with Martelottus that there is no reason to postulate that the Arabs differ from our system. We can also say, “we ourselves may differ from this Arabic language system.” While using their own terminology, which has been developed for their own linguistic phenomena, we will make progress in the understanding of not only the language but of the linguistic model as well.

5. References 5.1

Primary sources

Alcalá, Pedro de. [Before 5 Feb. 1505a]. Arte para ligeramente saber la lengua arauiga. Salamanca: Juan Varela. ——. [Before 5 Feb. 1505b]. Vocabulista arauigo en letra castellana. Salamanca: Juan Varela. ——. [ca. 1506a]. Arte para ligera mente saber la lengua arauiga emendada y añadida y Segunda mente imprimida. Salamanca: Juan Varela. ——. [ca. 1506b]. Vocabulista arauigo en letra castellana. Salamanca: Juan Varela. [Biblioteca Nacional, R 31638.] Re-edition: Paul de Lagarde. 1883. Petri Hispani de lingua arabica libri duo. Gottingae: In aedibus Dieterichianis Arnoldi Hoyer. Reprint: Osnabrück: Otto Zeller, 1971. Aquila, Antonio Ab. 1650. Arabicae linguae novae, et methodicae institutiones. Non ad vulgaris dumtaxat Idiomatis; sed etiam ad Grammaticae doctrinalis intelligentiam, per Annotationes in Capitum Appendicibus suffixas, accomodatae. Authore F. Antonio ab Aquila Or. Min. Sancti Francisci strict. Obser. Theologo, atque in Collegio Sancti Petri

inflection and government in arabic


Martis Aurei à Sacra Congregatione de propaganda Fide Arabicae linguae deputato lectore. Opus tum omnibus Arabicae linguae studiosis, tum potissumum Apostolicis Viris, per Asiam & Africam Fidem propagaturis, utile & necessarium. Romae: Typis Sac. Cong. De Prop. Fide. Caballero, Lucas. [Cauallero] 1709. Compendio de los Rudimentos y Gramatica Araba en que se da suficiente notizia de la lengua Vernacula o Vulgar y algunas Reglas de la literal Iustamente P.M.R.F. Bernardino Gonzalez hijo de la Proâ de la Concepzion en España Lector Jubilado en Arabo y Misionero Apostolico del Oriente y recoplada por el P. Fr. —— Mo Apostolico hijo de la Proa de los Angeles Lector actual Arabo en el colegio de Damasco. Ms. Rogge Library, Handskriftssamlingen, Sweden. Cañes, Francisco. 1775. Gramatica arabigoespañola, vulgar, y literal. Con un diccionario arabigo-español, en que se ponen las voces mas usuales para una conversacion familiar, con el Texto de la Doctrina Cristiana en el idioma arabigo. Madrid: En la Imprenta de Don Antonio Perez de Soto. Dombay, Franciscus de. 1800. Grammatica linguae Mauro-Arabicae juxta vernaculi idiomatic usum, accessit vocabularium Latino-Arabicum. Vindobonae: apud Camesina. Erpenius, Thomas. 1617. Grammatica Arabica dicta Gjarumia, & Libellus centum regentium cum versione latina & commentariis Thomae Erpenii. Leidae: Ex typographia Erpeniana Linguarum Orientalium. ——. 1620. Rudimenta lingvæ arabicae. Accedunt ejusdem Praxis Grammatica; & Consilium de studio Arabico feliciter instituendo. Leidae: Ex Typographia Auctoris. Germanus [of Silesia], Dominicus. 1639. Fabrica linguae Arabicae cum interpretatione Latina & italica, accomodata ad usum linguae vulgaris, & scripturalis. Romae: Typis Sac. Congreg. De Prop. Fide. ——. 1640. Fabrica Arabica copiosioribus impensis atque vberiori structura. Romae. González, Bernardino 2005 [c. 1705]. Intérprete Arábico. Edición y estudio preliminar por Ramón Lourido Díaz. 2 vols. Madrid: Real Academia de la Historia/Ministerio de Asuntos Exteriores y de Cooperación. ——. 2005 [c. 1705]. Epítome de la gramática arábiga. Edición y estudio preliminar por Ramón Lourido Díaz. 2 vols. Madrid: Real Academia de la Historia / Ministerio de Asuntos Exteriores y de Cooperación. Guadagnoli, Philippi. 1642. Breves Arabicae linguae Institutiones. Romae: Ex Typographia Sac. Congregationis de Propaganda Fide. Martelottus (= Martelotti), Franciscus. 1620. Institutiones Linguae Arabicae Tribus Libris distributae. In quibus uberrime quaecumque ad litteras, Dictiones & Orationem attinent, explicantur. Authore P. Francisco Martelotto Martinensi, Sacerdote, Theologo, Clericorum Regularium Minorum. Romae: Excudebat Stepahsnus Paulinus. Nepos, Ferdinandus. 2000.[c. 1468–1485]. Materies. (Ms. See Codoñer). Obicini, Franciscus. 1631. Grammatica Arabica Agrumia appellate. Cum versione Latina ac dilucida expositione. Adm. R.P.F. Thomae Obicini Noniensis, Diae. Nouariau, ord. Minorum Theologi, ac linguarum Orientalium in Collegio S. Petri Montis Aurei, de mandato Sacrae Congregationis Fidei propaganda Magistri. Romae: Typis Sac. Congregationis de Propaganda Fide. Oyanguren de Santa Inés, Melchor. 1738. Arte de la lengua japona. Dividido en quatro libros segun el Arte de Nebrixa, con algunas voces proprias de la escritura, y otras de los lenguages de Ximo, y del Cami, y con algunas perifrasis, y figuras. México: Joseph Bernardo de Hogal. ——. 1742. Tagalysmo elucidado, y reducido (en lo possible) â la Latinidad de Nebrija, con su Syntaxis, Tropos, Prosodia, Passiones, &c. y con la allusion, que en su uso, y composicion tiene con el Dialecto Chinico Mandarin, con las Lenguas Hebrea, y Griega. México: D. Francisco Xavier Sánchez, en la Calle de San Francisco. Reuchlin, Johannes. 1974[1506]. De rudimentis hebraicis libri III. Pforzheim. Reprint: Hildesheim – New York: Georg Olms Verlag.


otto zwartjes

Rodrigues, Ioão. (Rodriguez) 1604–1608. Arte da Lingoa de Iapam composta pello Padre —— Portugues da Copanhia de Iesv diuidida em tres Livros. Com Licença do Ordinario, e Svperiores. Nagasaqui: no Collegio de Iapão da Companhia de Iesv Anno. 1604 (-1608 – colofon.) Rodrigves, Ioam. 1620. Arte Breve da Lingoa Iapoa tirada da Arte Grande da mesma lingoa, pera os que comecam a aprender os primeiros principios della. Pello Padre Ioam Rodrigvez da Companhia de Iesv Portugues do Bispado de Lamego. Diuidida em tres Livros. Com Licenca do Ordinario, & Superiores. Amacao: no Collegio da Madre de Deos da Companhia de Iesv. Roth, Heinrich. 1988[ca. 1662]. Grammatica Linguae Sanscretanae Brachmanum Indiae Orientalis. Facsimile edition: Arnulf Camps & Jean-Claude Muller: The Sanskrit Grammar and Manuscripts of Father Heinrich Roth S.J. (1620–1668). Facsimile edition of Biblioteca Nazionale, Rome, Mss.Or. 171 and 172. Leiden: E. J. Brill. Valle Flemmarum, Agapiti à. 1687. Flores Grammaticales arabici idiomatis collecti ex optimis quibusque Grammaticis nec non pluribus Arabum monumentis, ad quam maximam fieri potuit brevitatem atque ordinem revocati: Studio, & Labore Fr. Agapiti à Valle Flemmarum Ordinis Minorum S. Francisci reformatorum Provinciae Tridentinae, in Semionario Patavino lectoris. Opus omnibus Arabicae Linguae Studiosis perutile, & necessarium. Patavii: Ex Typographia Seminarii Opera Augustini Candiani. Ziegenbalg, Bartholomäus. 1985 [1716]. Grammatica Damulica. Quæ Per Varia Paradigmata, Regulas & Necessarium Vocabulorum Apparatum, Viam Brevissimam Monstrat, Qua Lingua Damulica Seu Malabarica, Quæ Inter Indos Orientales in Usu Est, & Hucusque In Europa Incognita Fuit, Facile Disci Possit: In Usum Eorum Qui Hoc Tempore Gentes Illas Ab Idololatria Ad Cultum Veri Dei, Salutemque Æternam Evangelio Christi Perducere Cupiunt . . . Halæ Saxonum: Litteris & Impensis Orphanotrophei. Ed. Burchard Brentjes & Karl Gallus, Halle: Martin-Luther-Universität Halle-Wittenberg, 1985. 5.2

Secondary sources

Baalbaki, Ramzi. 2004. Grammarians and Grammatical Theory in the Medieval Arabic Tradition. Aldershot: Ashgate –Variorum. Ben Cheneb, Mohammed 1927. “Ibn Ādjurrūm.” Enzyklopädie des Islām. 2.381–382. Leiden – Leipzig: Brill – Harrassowitz. BICRES (see Niederehe). Carter, Michael G. 1991. “Review Owens 1988.” Journal of the American Oriental Society, 111.2: 395–397. ——. 1993. “Probleme bei der Übersetzung von Fachsprache am Beispiel des Arabischen.” In Frank, Armin Paul, Kurt-Jürgen Maaß, Fritz Paul, und Horst Turk, eds. Űbersetzen, Verstehen, Brücken bauen. Geisteswissenschaftliches und literarisches Űbersetzen im internationalen Kulturaustausch. Berlin: Erich Schmidt., 1.130–141. ——. 1994. “Review Owens 1990”. Journal of the American Oriental Society, 114.3: 472– 475. Codoñer, Carmen. 2000. Gramáticas Latinas de transición. Juan de Pastrana, Fernando Nepote. Introducción y edición crítica. Salamanca: Ediciones Universidad de Salamanca. Corriente, Federico. 1988. El léxico árabe andalusí según P. de Alcalá. (Ordenado por raíces, corregido, anotado y fonémicamente interpretado. Madrid: Universidad Complutense de Madrid. Cowan, William. 1981. “Arabic grammatical terminology in Pedro de Alcalá.” Historiographia Linguistica, 8.2/3:357–363. Dannenfeldt, Karl H. 1955. “The Renaissance Humanists and the Knowledge of Arabic.” Studies in the Renaissance, 2, 96–117.

inflection and government in arabic


Escavy, R., M. Hernández Terrés, A. Roldán, eds. 1994. Actas del Congreso Internacional de Historiografía Lingüística. Nebrija V-Centenario 1492–1992. Murcia, 1–4 abril 1992. Murcia: El Taller, vol. III: Nebrija y otros temas de historiografía lingüística. Frank, Armin Paul, Kurt-Jürgen Maaß, Fritz Paul, und Horst Turk, eds. 1993. Űbersetzen, Verstehen, Brücken bauen. Geisteswissenschaftliches und literarisches Űbersetzen im internationalen Kulturaustausch. Berlin: Erich Schmidt. Fück, Johann. 1955. Die arabischen Studien in Europa bis in den Anfang des 20. Jahrhunderts. Leipzig: Otto Harrassowitz. Geiger, Ludwig. 1871. Johann Reuchlin. Sein Leben und seine Werke. Leipzig: Duncker and Humblot. Hauschild, Richard. 1988. ‘Notes on the Content of the Three Manuscripts of Heinrich Roth.’ Arnulf Camps and Jean-Claude Muller: The Sanskrit Grammar and Manuscripts of Father Heinrich Roth S.J. (1620–1668). Facsimile edition of Biblioteca Nazionale, Rome, Mss.Or. 171 and 172. Leiden: E.J. Brill. James, Gregory. 2007. “The terminology of declension in Early Missionary Grammars of Tamil.” In Zwartjes, Otto, Gregory James and Emilio Ridruejo, eds., Missionary Linguistics III/Lingüística misionera III. Morphology and Syntax. Selected papers from the IIIrd and IVth International Conferences on Missionary Linguistics, Hong Kong/ Macau/Valladolid, 167–190. Juynboll, W.M.C. 1931. Zeventiende-eeuwsche beoefenaars van het Arabisch in Nederland. Utrecht: Kemink en Zoon. Kerkhof, Maxim, Hugo de Schepper, Otto Zwartjes, eds. 1993 España: ¿Ruptura 1492?. Amsterdam: Rodopi. Killean, Carolyn G. 1984. “The development of Western Grammars of Arabic.” Journal of Near Eastern Studies, 43.3: 223–230. Law, Vivian. 2002. The History of Linguistics in Europe. From Plato to 1600. Cambridge: Cambridge University Press. Lentin, Jerome. 1997. Recherches sur l’histoire de la langue arabe au Proche-Orient à l’époque moderne. Thèse de Doctorat d’État, Université de Paris III. Lichtenstädter, Ilse. “Al-Muta rrizī.” Enzyklopädie des Islām, 3:847. Also New edition: R. Sellheim, 7:773. Lourido Díaz, Ramón. 2005. “Estudio preliminar”. See: González (primary sources). Maruyama, Toru. 2004. “Linguistic Studies by Portuguese Jesuits in Sixteenth and Seventeenth Century Japan.” In Zwartjes, Otto and Even Hovdhaugen, eds. Missionary Linguistics [I]/Lingüística misionera [I]. Selected papers from the First International Conference on Missionary Linguistics, Oslo, 13–16 March 2003. Amsterdam: John Benjamins. 141–160. Monroe, James. T. 1970. Islam and the Arabs in Spanish Scholarship (Sixteenth century to the present). Leiden: Brill. Mörner, A.R. and Magnus Mörner. 2001. Spanien i svenska arkiv. Skrifter utgivna av Riksarkivet, 16. Stockholm: Riksarkiv, 60–61. Niederehe, Hans-Josef. 2005. Bibliografía cronológica de la lingüística, la gramática y la lexicografía del español (BICRES III). Desde el año 1701 hasta el año 1800. Amsterdam: John Benjamins. Owens, Jonathan. 1988. The Foundations of Grammar. An introduction to Medieval Arabic Grammatical Theory. Amsterdam: John Benjamins. ——. 1990. Early Arabic Grammatical Theory. Heterogeneity and Standardization. Amsterdam: John Benjamins. Schnurrer, Christianus Fridericus de. 1811. Bibliotheca Arabica. Halae ad Salam: I.C. Hendelii. Smith-Stark, Thomas. 2005. “Phonological description in New Spain.” In Zwartjes, Otto and Cristina Altman, eds.. Missionary Linguistics II/Lingüística misionera II. Orthography and Phonology. Selected Papers from the Second International Conference on Missionary Linguistics, São Paulo, 10–13 March 2004. Amsterdam: John Benjamins, 1–64.


otto zwartjes

Troupeau, G. 1962. “Trois traductions latines de la ‘Muqaddima’ d’Ibn Āğurrūm.” Études d’Orientalisme dédiées á la mémoire de Lévi-Provençal. Paris: Maisonneuve et Larose, 1.359–365. Versteegh, Kees. 1997. Landmarks in Linguistic Thought III. The Arabic Linguistic Tradition. London: Routledge. Weiß, Josef. 1910. “Die Arabische Nationalgrammatik und die Lateiner.” Zeitschrift der deutschen Morgenländischen Gesellschaft, 64:349–390. Zwartjes, Otto. 1993. “El artículo en las gramáticas pioneras de Nebrija y Alcalá y las gramáticas grecolatinas.” In Kerkhof, Maxim, Hugo de Schepper, Otto Zwartjes, eds. España: ¿Ruptura 1492?. Amsterdam: Rodopi, 261–286. ——. 1994. “Tradición e innovación en las gramáticas pioneras de Antonio de Nebrija y Pedro de Alcalá: la categoría gramatical del pronombre.” In Escavy, R., M. Hernández Terrés, A. Roldán, eds. 1994. Actas del Congreso Internacional de Historiografía Lingüística. Nebrija V-Centenario 1492–1992. Murcia, 1–4 abril 1992. Murcia: El Taller, 3:651–665. ——. 2007. “Agreement asymmetry in Arabic according to Spanish missionary grammarians from Damascus (XVIIIth century).” In Zwartjes, James, Ridruejo, eds., forthcoming. Missionary Linguistics III/Lingüística misionera III. Morphology and Syntax. Selected papers from the IIIrd and IVth International Conferences on Missionary Linguistics, Hong Kong/ Macau/Valladolid, 273–303. ——, and Even Hovdhaugen, eds. 2004. Missionary Linguistics [I]/Lingüística misionera [I]. Selected papers from the First International Conference on Missionary Linguistics, Oslo, 13–16 March 2003. Amsterdam: John Benjamins. ——, and Cristina Altman, eds. 2005. Missionary Linguistics II/Lingüística misionera II. Orthography and Phonology. Selected Papers from the Second International Conference on Missionary Linguistics, São Paulo, 10–13 March 2004. Amsterdam: John Benjamins. ——, Gregory James and Emilio Ridruejo, eds. 2007. Missionary Linguistics III/Lingüística misionera III. Morphology and Syntax. Selected papers from the IIIrd and IVth International Conferences on Missionary Linguistics, Hong Kong/ Macau 12–15 March 2005. Valladolid, 8–11 March 2006. Amsterdam: J. Benjamins.



1. Introduction The Arabic term waqf ‘pause’ is a grammatical concept. It is a verbal noun derived from the verb waqafa that could mean ‘to stop,’ ‘to come to a standstill’ or ‘to pause.’ This verb is used both in transitive and intransitive form. The term waqf is from the transitive verb waqafa. Another verbal noun wuqūf ‘standing still’ is derived from the intransitive verb waqafa (Farrāj 2001, 13–14). The term waqf is more commonly used and has a significant point of reference in both linguistic and religious connotations that mean to hold something at a pause or stop. This term waqf and its plural form ’awqāf mean an endowment in Islamic law which signifies the dedication of property or land that cannot be sold. The chain of speech utterances may be divided into spoken group events. This is usually regulated by the phonological, syntactic and semantic rules of the language. Normally words in the chain of the utterances of speech in prepausal forms either end with vowels or consonants. In Arabic vowels at the end of a speech event in prejunctional state are a sign of continuity and consonants signal waqf (H assān 1973, 270–271). When speaking or reading aloud with the prepausal mode almost all short vowels at the end of words, phrases and sentences will be dropped. This linguistic phenomenon, in Arabic, is referred to as pause forms. Oral reading and recitation have been highly emphasized in the Arab-Islamic tradition. Poets in Arabia in the pre-Islamic period composed their odes to be recited and the members of the poet’s tribe in turn memorized and recited these poems whenever the opportunity availed itself for them to do so. When the Qurān was first revealed it emphasized recitation. The very first verse states ‘recite or read aloud, iqra. The tradition of oral reading, in general, is engrained in the hearts and minds of most Muslims and Arabs. The Arabic sources that dealt with this topic were focusing on the rules of waqf in Classical Arabic with special emphasis on the rules of


salman h. alani

recitation in the Qurān. This aspect of waqf will not be covered in this paper. The emphasis is on the phenomenon of waqf with the explanation of the basic linguistic features of waqf of Modern Standard Arabic. Previous research on waqf, in western languages, is rather limited. There are four studies that made a basic contribution to waqf or pause. The following is brief account of these studies. 1.1

F.T. Mitchell (1990)

Mitchell wrote that pause “. . . relate(s) first and foremost to words that occur in apocopated form before pause, that is predominantly, final in the phrase and sentence.” He added that: “In discourse, it should not be expected that pause will always neatly correlate with grammatical phrasal divisions nor that the most appropriate divisions are always observed by speakers or readers, but the principle should always be followed that, wherever a pause is made, the preceding word should be pronounced in its pausal form.” He postulated the following rules: (i) a final short vowel, for example those of case, tense and mood is omitted; (ii) the sign of the indefinite, i.e., in nunation (-n) is omitted together with the preceding vowel, in the nominative (-un) and genitive (-in) cases of nouns and adjectives; (iii) accusative (-an) may be replaced by a long fath a (ā); (iv) the feminine singular and unit ending tā marbūtah is replaced by /-h/ in all three cases (Mitchell 1990, 99–100). This brief account of waqf by Mitchell covers the basic rules of pause of al-‘arabiyyah. It is intended to aid the learner of Arabic to pronounce and read aloud correctly. He used a short transliterated Arabic passage read aloud by what he called “. . . a speaker trained in the high Classical tradition” to illustrate the pause form rules. The comments and the pause rules almost mirror the traditional statements of the Arab grammarians on waqf (Mitchell 1990, 100–101). 1.2

Clive Holes (2004)

Holes defined waqf as: “pause is defined (rather vaguely) as an audible break in delivery” (Holes 2004, 63). The focus of his analysis is on the spoken Modern Standard Arabic. The illustrative examples he used were taken from Syrian radio broadcast and television news and speeches of a former Egyptian leader, Gamal Abdul Nasser. He (Holes) said that

linguistic analysis and rules of pause in arabic


pauses depend on what is said in the spoken Modern Standard Arabic, the audience and the speaker’s intentions. The conditions where prepause occur are: (i) final short vowels that appear in prejunctural position are omitted; (ii) the feminine gender morpheme tā marbūtah in certain nouns and adjectives in prepause, together with the /-u, -i, and -a/ are deleted. He added that /-ah/ is pronounced as /-a/; (iii) The inflectional suffixes /-un/ nominative, /-an/ accusative and /-in/ genitive were discussed individually. The /-un/ is omitted in prepause. The nature of the accusative /-an/ depends on its grammatical status. When it is used as an adverbial marker as in kullīyan ‘completely’ the majority of speakers retain it, however few realize it as /-ā/. When /-an/ is used as marker of an object of nouns or an adjective that agrees with one, it is sometimes retained and sometimes omitted. Holes analysis of the inflectional suffix /-in/ is very interesting. He made a distinction of /-in/ when it is used as a marker of a noun in the genitive case and /-in/ as a marker of an attributive adjective in agreement with a genitive head noun. The suffix /-in/ usually is omitted, however, Holes gave some examples where it is retained (Holes 2004, 63–68). 1.3

Muh ammad Farrāj (2004)

The focus of Farrāj’s long Arabic article is on the traditional grammatical concept of waqf. The emphasis of this article was on the rules of waqf in Classical Arabic Grammar and its application for the recitation of the Qurān. He stated that there are two basic features of waqf. These are wajh ‘manner’ and mah all ‘place.’ The purpose of mah alli ’l-waqf ‘place of pause’ is to regulate the speech event and leads to improving its division where juncture or pause may occur. The wajh, plural awjuh, provides the speaker or the reader with the guide for waqf so the speech will appear harmonized in its utterances. He added that waqf has a wazīfa ‘task’ which serves for clarity of pronunciation and meaning that help in understanding of what is being read or spoken (Farrāj 2004, 12–23). 1.4

Tamām H assān (1973)

H assān stated that waqf through its various means by its nature is a mifsal, ‘separator’ of speech where it is possible that the chain of speech may broken into spoken groups. Every one of these, when its meaning is complete, is considered a speech event. However, if the speech is not


salman h. alani

complete then the speech event may consist of more than one event (H assān 1973, 270).

2. The Arabic linguistic analysis of and rules for waqf 2.1

The waqf status of the tā marbūtah

One of the most common modes of waqf is applied to words that end with a tā marbūtah. This tā marbūtah is considered to be a morpheme that primarily marks the feminine endings of nouns and adjectives. However there are some feminine nouns and adjectives that do not always end in tā marbūtah. In Arabic script some feminine nouns and adjectives may be written with tā mabsūtah. This is the regular tā that appears at the end of words. This is especially true in the script of the Qurān. In fact we find sometimes the same word written with tā marbūtah and in other contexts it is written with a tā mabsūtah. The word rah ma ‘mercy’ for example is written in either tā marbūtah or tā mabsūtah. There are nouns and adjectives that end in a tā marbūtah that have a masculine meaning. Proper names such as H amzah, Talh ah, and several others are written with tā marbūtah. The rules of waqf apply to them as they apply to the feminine nouns and adjectives. Therefore it is not the gender of the word but rather the form of the word that determines the application of the waqf rules. The tā marbūtah in nouns and adjectives that appears in prepause form is deleted and replaced by /-h/. This deletion, of course, includes the vowels: /-i/, -/u/ and /-a/ in the definite and /-in/, /-un/ and /-an/ in the indefinite. The deletion takes place in all three cases: nominative, accusative and genitive. With prejunctural, on the other hand, the tā marbūtah and the vowels both in the definite and indefinite are retained. This is illustrated with the word madrasah ‘school’ as in the prejunctural forms below:

Indefinite Definite




madrasa-tun al-madrasa-tu

madrasa-tin al-madrasa-ti

madrasa-tan al-madrasa-ta

linguistic analysis and rules of pause in arabic


The case endings in the indefinite /-tun, -tin, -tan/ and the definite /-tu, -ti, -ta/ all will be deleted in the prepausal forms. The word madrasah ‘school’ and the word al-madrasah ‘the school’ both in definite and indefinite forms will have the same endings in the prepausal forms. The reason is that the /-h/, placed between slashes, is sometimes weakened to the point that one really cannot even hear it. This is especially the case of Modern Standard Arabic, read aloud or spoken by radio and television broadcasters. I have examined and analyzed the speech segments of several announcers and observed both the dropping off of the /-h/ and the retaining of it. In careful delivered speeches especially of religious nature the /-h/ is almost always retained. 2.2

The deletion of final short vowels

The final short vowels, in waqf, are deleted. This deals with al-h arakāti l-irābiyyah, the vowel marks that signal case endings, tense and mood. The following rules apply to nouns and adjectives derived from strong verb roots: (a) The case endings in the definite nouns are indicated by /-u/ in the nominative case, /-i/ in the genitive case, and /-a/ in the accusative case. All of these short vowels that mark the case endings are omitted in waqf. The following sentences illustrate the prejunctural and prepausal forms: Prejunctural forms

Prepausal forms


jāa-l-walad+u mina-l-walad+i raaytu-l-walad+a

jāa-l-walad mina-l-walad raaytu-l-walad

the boy came from the boy I saw the boy

(b) When the indefinite nouns and adjectives that end in ‘nunation,’ the sound /-n/ pronounced but not written. The case endings in the indefinite nouns and adjectives are indicated by /-un/ in the nominative case, /-in/ in the genitive case and /-an/ in the accusative case. Both nominative and genitive endings are omitted in prepausal forms. However the accusative case maker is changed into an alif /-ā/ that requires a special treatment. The word walad ‘boy’ in the following sentences illustrates both prejunctural and prepausal forms.


salman h. alani Prejunctural forms

Prepausal forms


jāa walad+un min walad+in raaytu walad+an

jāa walad min walad raaytu walad+aa

a boy came from a boy I saw a boy

(c) The indefinite accusative case ending is /-an/. Also certain types of adverbs end in /-an/. This ending is replaced by /-ā/ in prepausal forms as illustrated in the table above. In the case of adverbs some speakers retain the ‘nunation’ while others replaced it with /-ā/. However, in prepausal and words in isolation may be retained but oftentimes it changes to /-ā/ following the normal rules of waqf. The word id an meaning ‘if ’ is written orthographically either with nūn /n/ or with the alif tanwīn. In prepausal form pronounced as /-ā/. (d) Tense and mood in verb-vowel endings are omitted.


Prejunctural forms

Prepausal forms


katab+a yaktub+u lan yaktub+a

katab yaktub lan yktub

he wrote he writes he will not write

Words that end in long vowels

There are two main grammatical categories of words that end in long vowels. These are called in traditional Arabic grammar as the al-manqūs ‘defective’ as in the word al-wādī ‘the valley’ and al-maqsūr ‘shortened’ as in the word al-fatā ‘the youth.’ 1. The rules of waqf on words that end in the long vowel /-ī/ such as al-muh āmī ‘the lawyer’ which is derived from a finally weak verb h amā and its imperfect yah mī ‘to defend’ ends in /yā/. Nouns that are derived from finally weak verbs like h amā of the pattern of al-muh āmī always end /-ī/. This category of words should not be confused with words ending in yā-n-nisbah as in words like lubnāniyyun ‘Lebanese’ or even with words like zabyun ‘deer’ both of these words are written in Arabic script with /yā/, however the source of this /yā/ is not a final radical as the nouns

linguistic analysis and rules of pause in arabic


derived from finally weak verbs like h amā and its imperfect yah mī ‘to defend’ which ends in /yā/ as a radical (Farrāj, 2001, 71–72). The vowel endings of al-manqūs ‘defective’ are determined by the case endings and whether they are definite or indefinite. When words of the manqūs ‘defective’ are in the accusative case and indefinite they end in /-ā/ in the waqf as qābaltu muh āmyā ‘I met a lawyer’ after the omitting of the ‘nunation.’ When they are definite they end in /-ī/ as in qābaltu l-muh āmī ‘I met the lawyer.’ When the manqūs words are in genitive or nominative case they end in the omitting of the final vowel /-ī/ as in filwād‘in the valley’ and hādā wād ‘this is a valley. However when they are definite the words of the manqūs end in long vowel /-ī/ as in fil-wādī ‘in the valley’ and hādā l- wādī ‘this is the valley. 2. The rules of waqf of words that end in the long vowel /-ā/ such as al-fatā ‘the youth’ and which are primarily derived from finally weak verbs are called in traditional Arabic grammar al-maqsūr ‘the shortened.’ The waqf on these words is always end in an alif /-ā/ in all three case endings and regardless how orthographically they are written with alif maqsūrah or regular alif. What matters here is the pronunciation and not the script.

3. Concluding Remark It is worthwhile to mention that the rules of waqf are not always adhered to by readers and speakers of Modern Standard Arabic. The Arabic language for sometime has been going through processes of change. Some readers and speakers are not using the al-h arakāta-l-i‘rābiyyah ‘case endings.’ The often heard statement that states sakkin taslam ‘use sukūn and you will be safe’ reflects the state of affairs of the on going change of the Arabic language. Therefore the rules of waqf outlined above when considered should be used as guidelines not to be applied in a rigorous and strict manner.


salman h. alani 4. References

4.1 Primary source Ibn Yaīš. n.d. Šarh al-mufassa l. Cairo: Maktabat al-Mutanabbī. 4.2 Secondary sources Farrāj, Muhammad. 2001. Al-waqf wa-wazāifuhu inda n-nahwiyyīn wa-l-qurrā. Annals of Arts and Social sciences, Monograph 159 Volume 21. Kuwait: Kuwait University. H assān, Tammām. 1973. Al-luġa al-arabiyya mabnāhā wa-manāhā. Cairo: al-Haya al-Āmma lil-Kitāb. Holes, Clive. 2004. Modern Arabic Structures, Functions and Varieties. Revised edition. Georgetown, Washington, D.C.: Georgetown University press. Mitchell, T.F. 1990. Pronouncing Arabic. Oxford: Clarendon Press.

THE EXPLANATION OF HOMONYMY1 IN THE LEXICON OF ARABIC Georges Bohas and Abderrahim Saguer ENS – Lettres et Sciences Humaines, Lyon

1. Introduction: the theoretical framework and earlier analyses This article is a follow-up to our previous paper entitled “Sur un point de vue heuristique concernant l’homonymie dans le lexique de l’arabe.”2 By adopting a heuristic viewpoint, we take into account the fact that we have not yet finished exploring all the matrices of Arabic, and thus proceed by means of successive evaluations and provisional hypotheses.3 Certain points remain to be clarified; nevertheless, the level of explanation we can achieve has become clearer, as well as the explanatory methods that we are able to offer within the framework of the Theory of Matrices and Etymons (TME). Although there are already numerous publications on this subject, it is worth underlining that, within the TME framework, the lexicon is organized on three levels:4 1 / The matrix: a non linearly-ordered combination of a pair of phonetic feature vectors linked to a ‘notional invariant’; for example, {[labial], [coronal]} ‘to strike a blow.’5

1 It is worth defining homonymy in opposition to polysemy. Polysemy is ‘a word which brings together several meanings between which users can recognize a link’ (Nyckees, 1998: 194); the meanings are different but related. ‘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 different meanings, for example: flies ‘certain insects’ and flies ‘the opening at the front of a pair of trousers’ or to sound ‘to make a noise’ and to sound ‘to measure the depth of water’—different non related meanings. 2 Bohas and Saguer (2006). 3 That is: in a not-rigorously demonstrated manner but justified by reasons of internal coherence (see the website Einstein, Albert); and accepting that you cannot explain everything. 4 See Bohas (1997, 2000), Dat (2002). 5 This is a property of the language that was proved both formally and semantically by Bohas and Darfouf (1993), developed in Bohas (1997), which consists in the fact that


georges bohas and abderrahim saguer

2 / The etymon: a non linearly-ordered bi-consonantal base made up of two phonemes taken from a given matrix exhibiting both the features of this matrix and its ‘notional invariant’; for example, {b,t} ‘to strike a blow with a sharp object.’ 3 / The radical: an etymon that has developed by diffusion of the final consonant or by incrementation, or that results from the blending of two etymons; the radical includes at least one vowel and vectors the notional invariant; for example: /bvtar/, ‘to cut, to cut the tail’ (Bohas 2000, 9). The radical is the domain in which diverse morphological and Ablaut processes take place (Guerssel and Lowenstamm 1993, Segeral 1995). So far, ten matrices have been accounted for; most of these have already been subjected to in-depth studies (see footnote 14). Matrix 1

{[labial],6 [coronal]} Notional invariant: ‘to strike a blow’

Matrix 2

{[labial], [-voiced] }7 [+continuant]8 Notional invariant: ‘movement of air’

Matrix 3

{[labial], [pharyngeal]}9 Notional invariant: ‘(a) tightening’

Matrix 4

{[coronal], [pharyngeal]} [-dorsal]10 [-voiced]

a binary combination {a, b} is realized in the order a+b and in the order b+a while keeping the same notional invariant. 6 [labial] characterizes sounds produced with a constriction of the lips. For matrices 1, 2, 3, 6 we integrate on-going research which shows that the feature [labial] should not be restricted by [-sonorant] (see Mansouri, 2006). 7 [±voiced] Sounds produced with vibration of the vocal cords are said to be voiced ([+voiced]), whereas other sounds are said to be non-voiced ([-voiced]), see Dell (1973: 56). 8 [±continuant] Sounds with the feature [+continuant] are produced without interrupting the flow of air through the oral cavity, those with the feature [-continuant] are produced with total interruption of the flow of air at the oral cavity, see Halle (1991: 208). 9 [pharyngeal] characterizes segments that the Arabic tradition calls gutturals, that is: , h, , h , x, ġ and q. For the problems posed by the characterization of this class, see Kenstowicz (1994: 456ff). 10 [dorsal] characterizes sounds produced with a constriction created with the back of the tongue between the soft palate and the uvula (velar and uvular consonants; rear vowels).

explanation of homonymy in the lexicon of arabic


Notional invariant: ‘low voice, muffled, hoarse noise’ Matrix 5

{[coronal], [dorsal]} Notional invariant: ‘to strike a blow’11

Matrix 6

{[labial], [dorsal]} Notional invariant: ‘curvature’

Matrix 7

{[dorsal], [pharyngeal]} Notional invariant: ‘the cries of animals’

Matrix 8

{[+sonorant12], [+continuant]} [+lateral13] Notional invariant: ‘the tongue’

Matrix 9

{[+nasal], [+continuant]} Notional invariant: ‘the nose’

Matrix 10

{[+nasal], [coronal]} Notional invariant: ‘traction’14

The data on which we have based our study are to be found in the Kazimirski, and have been checked in the Qāmūs and/or the Lisān. When they are based on another source, this is mentioned.

2. Explanatory methods In the paper quoted above, we demonstrated that the homonymy of a radical may be attributed to three causes: A. the fact that it is the result of blending: it manifests the meanings of both the etymons that are its source.

11 See Diab (2005) who brings a modification to the formulation of the notional invariant of this matrix. 12 [±sonorant] Sounds with the feature [+sonorant] are produced with a constriction which does not influence the capacity of the vocal cords to vibrate spontaneously. Those with the feature [-sonorant] have a constriction which reduces the global flow of air and makes voicing more difficult. ‘Thus the natural state for sonorants is [+voiced] and for non sonorants (termed obstruents) is [-voiced],’ see Kenstowicz (1994: 36). 13 [±lateral] A [+lateral] sound is produced by making a constriction with the central part of the tongue while lowering one or both lateral edges so that air escapes around the side(s) of the mouth, see Kenstowicz (1994: 35). 14 For matrices 1 to 6, see Bohas (2000), Dat (2002), for an in-depth study of matrix 6, see Serhane (2003), Bohas and Serhane (2003), for matrix 7, see Bohas and Dat (2005), for matrices 8 and 9, see Bohas (to be published) and for matrix 10, see Saguer (2003).


georges bohas and abderrahim saguer

B. the fact that its etymon is the realization of several matrices: it manifests the meanings of these matrices. C. the fact that two etymonial analyses are possible, such as [nX]Y and n[XY]. Below, we illustrate each case with an example from Bohas and Saguer (2006). A. Homonymy resulting from blending Let us consider the verb ġaraza, which attests two meanings (hereafter senses): S1 – ‘to prick something with a needle, to drive in, to plunge (a sharp instrument), to plunge a tail into the ground to lay eggs (of locusts)’; S2 – ‘to give but very little milk (of pregnant camel).’ The same semantic load is found in ġārizun; S1 – ‘that drives in, plunges a sharp instrument, a goad into something; that plunges a tail into the ground to lay eggs (of locusts)’; S2 – ‘that gives but little milk (camel)’. Since it is not possible to establish a plausible semantic relationship between the two senses, this is an obvious case of homonymy. And yet we observe the existence of the following words: ġarra15 F. III : to be found in small quantities (of milk of a female); ġirārun : a small quantity, generally, such as a small quantity of milk in a female muġārrun : that has little milk in the udders (camel) Phonetically, the etymonial analysis can only be {ġ,r}, since their radical has no other consonant,16 and it is obvious that they attest sense S2. Moreover, the precise meaning of the verb razza is: ‘to plunge a tail into the ground to lay eggs (of locusts); to stick, drive in and fix firmly one object into another or into the ground.’ Thus it clearly attests sense S1 and is analysed as the etymon {r,z}.17 Therefore, the explanation is that ġaraza includes senses S1 and S2 because it results from blending of the

15 16 17

We use boldface for the segments that make up the etymon. We call these non ambiguous radicals. Non ambiguous radical.

explanation of homonymy in the lexicon of arabic


two etymons {ġ,r} and {r,z}. The way this blending occurs is represented in model A:18 A





Ck Sj

Cj Si More explicitly : ġr ‘lack of milk’ Si

Ci +

Ck19 SJ


rz ‘to drive a sharp object into’ Sj



Henceforth, in similar cases, we shall talk of explanation through blending. B. Homonymy through the realization of several matrices 1) An easy case Let us take the verb mata"a, which means: S1 – ‘to strike somebody with a stick’ S2 – ‘to tighten, stretch out a rope’ The m is [labial], the t is [coronal]: the etymon {m,t} can thus be a realization of matrix 1: {[labial], [coronal]} Notional invariant: ‘to strike a blow’ And for this reason assumes the sense S1 – ‘to strike somebody with a stick.’ But the m is also [nasal] and the etymon {m,t} can also be a realization of matrix 10:


See Bohas (1997: 175f and 2000: 49). The obligatory contour principle (OCP) explains the fusion of the two Ci into a single segment. See McCarthy (1986) for the definition and various applied examples referring to Semitic languages of this principle which forbids adjacent identical elements at the same level. Since, OCP has given rise to a multitude of studies the list of which would be superfluous here. 20 We use x to indicate blending. 19


georges bohas and abderrahim saguer

{[+nasal], [coronal]} Notional invariant: ‘traction’ And for this reason assumes the sense S2 – ‘to tighten, stretch out a rope.’ 2) A more complex case The verb natara attests the following meanings: S1 – ‘to disperse’ F. I : to scatter, disperse, disseminate F. II : to scatter a lot, in large quantities: intensive of F. I F. V : to be scattered, dispersed, to disperse F. VI : to be scattered, dispersed, to disseminate, to spread S2 – ‘actions concerning the nose’ F. I natura : to blow one’s nose F. II : to blow one’s nose F. II : to draw up water through the nostrils F. VIII : to blow one’s nose F. VIII : to draw up water etc. through the nostrils and to expel it through the nostrils S3 – ‘to pull, tear off/out’ F. I : to remove, to take the clothes off the body of somebody, to strip S4 – ‘to strike a blow with a sharp object’21 F. IV : to pierce somebody with a sharp instrument and to make the blood flow The first hypothesis we can formulate is that natara develops the etymon: {n,t} and that for this reason, it is a realization of matrix 9:22 {[+nasal] [+continuant]} Notional invariant: ‘the nose’ The phonetic substance of this matrix comprises, on the one hand, the two nasals, m and n, and, on the other, the various fricatives.

21 We will propose no explanation for this sense; as we said in the introduction, we have not yet explored all the matrices of Arabic, the notional invariant ‘sharp’ is without doubt important but as yet we know nothing of it. 22 See Bohas (to be published) for a detailed study of this matrix.

explanation of homonymy in the lexicon of arabic


The ramifications of the notional invariant are as follows: 1. the nose 1.1. the organ itself and what affects it 1.2. the specification of parts (the top, the sides) 1.3. to be sharp, > protruding, > to precede 2.1. specifications of the organ (big, small . . .) 2.2. animal or human presenting these specifications 3. to lift the nose: movement of pride or contempt 4. the nose and air: to breathe in; to breathe out; to perceive odours, to smell 5. the influence of the nose on the voice: nasal sound; similar animal cries (buzzing-grunting) 6. various liquids (mucus, phlegm) which pass through the nose We may note that, in this organization, senses S2 – ‘to blow one’s nose,’ ‘to draw up water through the nostrils,’ fit into both 6. and 4. But, since n is [+nasal] and t [coronal], {n,t} can also be a realization of matrix 10: {[+nasal], [+coronal]} Notional invariant: ‘traction’ Moreover, other realizations of the {n,t} etymon can be found in: natala [nt]l : ‘to remove the garment or breastplate’ našnaša [nš]nš : ‘to remove (one’s clothes)’ Sense S3 – ‘to pull, tear off/out’ of F. I: ‘to remove, take the clothes off the body of somebody, to strip’ is thus incorporated into this matrix. We can therefore understand why the {n,t} etymon is homonymic: if we take into account the t the feature [coronal], then we see that it realizes matrix 10 ‘traction,’ while if we take into account the feature [+continuant], then we see it realizes matrix 9 ‘the nose.’ In this case we shall speak of ambiguity originating from the fact that an etymon is the realization of several matrices. C. Homonymy due to the possibility of two etymonial analyses Let us pursue the analysis of natara. We have seen that it also attests the sense ‘to disperse, to scatter.’ It is true that we can establish both a phonetic and a semantic relation with: tarra [tr]r : to disperse, disseminate the etymon of which can only be {t,r},23 and with:


No ambiguity.


georges bohas and abderrahim saguer

tāra t[w]r : to be lifted and scatter in the air tartaratun [tr]tr : dispersion, dissemination. farata f[rt] : to be dispersed, disseminated (of a tribe) This leads us to analyse natara by positing {t,r} as etymon and the n as an initial crement. Therefore, the homonymy stems from the fact that in A and B the form is analysed as [nt]r, which may have two matrix links, and in C as n[tr]. We shall say in the latter case that homonymy arises because several etymonial analyses are possible.

3. The two levels of explanation In Bohas and Saguer (2006) we studied cases concerning radicals containing an n. In this paper, we shall extend the study to include the analysis of radicals containing an l. According to previous studies,24 l may have the status of initial crement/prefix25 or of matrix segment. Earlier, Hurwitz (1913, 55–60) had already recognized this prefix status: The preformatives are thus seen to possess a fairly definite, 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 i are to be connected etymologically with the reflexive n, and the preformative y may be considered to be a denominative stem.

When it has a matrix function, l may form an etymon with the second radical: [lx]y, or the third radical: l[x]y.26 In other words, a radical [lvxy] may present all its possibilities, as can be seen in the table below: Etymon:










The most recent are those of Saguer (2000, 2002, 2002b). We use ‘prefix’ when l has a semantic-grammatical value and ‘initial crement’ when it does not. 26 This expresses the fact that in this form the etymon is [ly] and x is an inset crement. 25

explanation of homonymy in the lexicon of arabic 3.1


The explanation through attribution to identified matrices

The verb labaxa demonstrates eight senses: S1 – ‘to be fleshy (of the body)’ S2 – ‘to beat, strike somebody’ S3 – ‘to kill somebody’ S4 – ‘to take a thing from somebody, from the hand of somebody’ S5 – ‘to get something out of somebody using trickery’ S6 – F. III: ‘to slap somebody in the face’ S7 – F. V: ‘to perfume oneself with musk’ S8 – ‘to say something foolish, insulting to somebody’ It should be noted that several senses are polysemically related; in other words, a plausible semantic relation can be established between them, as follows: – S2, S6 and S3 are manifestations of the notional invariant ‘to strike a blow’; ‘to kill’ is related to this concept through the cause>consequence relationship and S6 specifies the mode of action. – S4 and S5 come under the notional invariant: ‘to pull, bring something to oneself.’ Thus five homonymic senses remain: A S1 ‘to be fleshy (of the body)’ B S2+S6+S3 ‘to strike a blow’ S2 – ‘to beat, strike somebody’ S6 – F. III: ‘to slap somebody in the face’ S3 – ‘to kill somebody’ C S4+S5 ‘to pull, bring something to oneself ’ S4 – ‘to take a thing from somebody, from the hand of somebody’ S5 – ‘to get something out of somebody using trickery’ D S7 – F.V: ‘to perfume oneself with musk’ E S8 – ‘to say something foolish, insulting to somebody’ We are going to attribute each of these homonymic senses to a source matrix. Take sense A S1–labaxa: ‘to be fleshy (of the body).’


georges bohas and abderrahim saguer

It is easily noted that this form is semantically linked to: rabīxun r[bx] : fat, thick and with a soft and loose body habayyaxun h[bx] : young chubby : fat, thick, hard anbaxun [bx] muxabxabat [xb] xb : handsome and plump (camels) This leads us to deduce that labaxa should be analysed as a form incorporating the etymon {b,x} with initial incrementation of an l which has the role of a prefix denoting the sense we have called ‘static,’ following Joüon (1923, 95). This is defined as the appropriation by the subject of the quality or the state X. The Hebrew examples cited by Joüon are: kābed ‘he is heavy,’ qāton: ‘he is small’ (1923, 95). This etymon {b,x} is a realization of the matrix 6 {[labial], [dorsal]} which has ‘curvature’ as its notional invariant. One of the rst manifestations of curvature in the convex form is precisely the concept of fatness:27 fat, plump, robust, that which realizes a swollen form: ˆ, as in: bajja -F. VII bājilun fajia hijaffun

: to be plump, to have rounded flanks (of animals fattened by pasture) : fat, replete : to have a fat belly : who has a fat belly, paunch

For sense A, labaxa ’to be fleshy (of the body)’ is analysed as l[bx] and is thus a realization of the matrix {[labial], [dorsal]} which has the notional invariant ‘curvature.’ Let us consider sense B – S2+S6+S3 ‘to strike a blow’; S2 – ‘to beat, strike somebody’; S6 – F. III: ‘to slap somebody in the face’; S3 – ‘to kill somebody.’ It is easy to establish a link between the form and other manifestations of the etymon {l,b} such as: wabala w[bl] : to strike somebody with a stick labana [lb]n : to strike somebody violently, to batter somebody to death with a stick lah aba l[h ]b : to strike somebody with a sabre and for the implied sense: habila h[bl] : to lose one’s son through death


See Bohas (2000: 108).

explanation of homonymy in the lexicon of arabic


This enables us to establish that labaxa includes the etymon {l,b}, itself a realization of the matrix {[labial], [coronal]} which has ‘to strike a blow’ as its notional invariant. However, note the existence of the following set: laxxa [lx]x laxaba [lx]b laxama [lx]m

: to slap somebody in the face : to slap somebody in the face : to strike somebody in the face

This enables us to extract the etymon {l,x} connected, once again, to the sense ‘to strike a blow.’ The segment l is analysed as [coronal} and x as [dorsal], so that this etymon is a realization of the matrix {[coronal], [dorsal]}, which leads us to the conclusion that labaxa in the B sense results from blending between two etymons {l,b} and {l,x}, a B type blending (Bohas 2000, 50). For sense C. = S4.+S5. – ‘to pull, bring something to oneself ’: S4. – ‘to take a thing from somebody, from the hand of somebody’; S5. – ‘to get something out of somebody using trickery’ we shall justify an etymonial analysis l[b]x, that is, an etymon {l,x} with an inset crement b. Indeed, a phono-semantic relation can be established with: In the order xl: xalaja xalā (hly) xalxala saxala xalasa

: to pull, attract to oneself : to pull out : to completely strip a bone of its flesh : to take, remove, ravish through trickery : to take away, remove ravish in an instant, unexpectedly, to pull out F. III : to pull something from somebody; to seize; to grab something F. V : to remove, take away F. VI : to reciprocally pull out, and to pull something each from his own side F.VIII : to pull something towards oneself in a hurry (the same meaning as the first item of the list with the addition of celerity). In the order lx : malaxa : to forcibly pull something towards oneself seizing it with one’s teeth or hands F. V. : to pull out, to burst e.g. the eye of a prey (birds of prey) F.VIII : to pull, extricate, pull out (a tooth, an eye); to pull out of the scabbard (a sabre, etc.)


georges bohas and abderrahim saguer

Therefore the etymon {l,x} carries two homonymic senses: ‘to strike a blow,’ as a realization of the matrix {[coronal], [dorsal]} and ‘to bring something to oneself ’ as a realization of another matrix, which remains to be studied and, we suspect, includes the features: {[+approximant28], [+continuant]} [coronal] with the notional invariant: ‘to bring something to oneself ’ Besides the words above, this matrix manifests itself in: Etymons with l salla salaba salata salaxa salaba sahala halaba halata halada lahā lahaba lahata laxasa F. II halaa halaba halata halama šalaha F. II


: to pull, extract gently one object from another : to snatch something with force from somebody : to extract, to pull : to skin, to remove the skin from a sheep; to remove one’s clothes and shirt : to pull, to extract the marrow from bones : to peel, to strip bark, skin : to pull out hair, bristles, horsehair : to peel something by removing the skin, bark : to pull, extract, pull out something from its place : to remove the inner bark of a tree, wood : to remove the bark from a piece of wood, to strip it of its bark : to remove the bark from a piece of wood, to peel it : to remove, to extract the purest part : to remove the flesh from the skin of a dead or bled white animal : to take the milk, to milk : to pull out, to remove in flakes : to remove tinea from the skin : to strip, to remove the clothes from somebody

This composition, [+approximant], seems too complex to characterize the class r, l. [coronal] This is due to the fact that, according to Yeou and Maeda (1994), the pharyngeals and uvulars of Arabic are also characterized by the feature [+approximant]. Indeed, the continuation of our studies will enable a discussion of this point. If the gutturals do figure in this matrix, and assuming that it may then be formulated simply as {[+approximant], [+continuant]}, this would constitute proof that the gutturals and r, l are indeed members of the same class: [+approximant].

explanation of homonymy in the lexicon of arabic Etymons with r haraba harasa harada saraqa sarā xaraša xarata xarafa 'araza 'ar'ara

'araqa 'arama 'ariyā qa'ara


: to strip somebody, to pillage (a caravan, a tribe) : to steal something; in fields, pastures : to take, by milking, all the milk from a camel : to steal : to remove, separate, distance something from somebody : to attract a camel to oneself with a hooked stick : to strip of bark and make even : to pick a piece of fruit from a tree : to pull with force : to move the cork of a bottle to uncork it; to remove, take out the cork; to pull, to pull out, to burst an eye : to strip a bone of the flesh that was on it by eating it : to eat the flesh that sticks to the bone : to be naked, stripped of one’s clothes (cause> consequence relation) : to pull out with the root, from top to bottom and to cause to fall

This abundance of data, the notional domain of which has still to be organized, is nevertheless sufficient to justify the existence of this matrix: [+approximant], [+continuant] [coronal] ‘to bring something to oneself ’ This data also justifies the analysis that posits that in labaxa sense C = S4+S5 ‘to pull, bring something to oneself ’: S4 – ‘to take a thing from somebody, from the hand of somebody’; S5 – ‘to get something out of somebody using trickery’ is a manifestation of this matrix. As to labaxa, D S7 F.V – ‘to perfume oneself with musk’ it is analysed as l[bx] in which the etymon {b,x} is found, as in: baxara [bx]r : to perfume somebody or something with incense baxira : to smell bad : who has unpleasant, fetid breath abxaru baxxa [bx]x : to snore during sleep This etymon is a realization of matrix 2. {[labial], [-voiced]} [+continuant]


georges bohas and abderrahim saguer

which combines the labials b and f with voiceless fricatives. The ramifications of the notional invariant of the conceptual field are: – movement of air: wind, breath – expulsion of air in man or animal 29 ¤ consequences (various odours) Below we provide some realizations of the matrix with f: { ft} nafata { fh} fahh a fahfah a fahā -F. II fāha/fawah a/ lafaha nafaha { fx} faxxa fāxa/fawaxa/ nafaxa

: to blow (on something) : to hiss (snake); to hiss during sleep : to be hoarse : to season food : to spread one’s perfume; to smell good or bad : to blow (of warm wind) : to spread one’s odour; to blow (of cold wind) : to snore and hiss talking of somebody who sleeps; to spread (of an aroma) : to spread (of an odour); to hiss (wind); to release wind (of a man) : to blow with the mouth; to break wind

Finally, sense: E S8 – ‘to say something foolish, insulting to somebody’. Note that labaxa commences with an l [+lateral] and nishes with a x which is a [+continuant] segment. The analysis is thus l[b]x, of the etymon {l,x}, which is also present in: : to be very talkative and say a lot of rubbish laxiya xatila : to talk a lot and to say only trivialities In this case, this etymon is a realization of matrix 8, exactly like: {l, ġ} laġā : to speak, in general to say futile things, to utter vain, flippant or careless words, to make mistakes in talking, to make an error


This sign indicates that a semantic relationship exists, here: cause>consequence.

explanation of homonymy in the lexicon of arabic laġiya F. X

Matrix 8


: to make an error talking, to be mistaken : to gather locutions, idioms, or to pay attention to words and locutions (luġāt), particularly of nomadic Arabs, i.e. to draw from them knowledge of Arabic words {[+sonorant], [+continuant]} [+lateral] Notional invariant: ‘tongue’

As we shall often be returning to the organization of the notional field of this matrix, it is worthwhile examining it here:30 The ramifications of the notional invariant include: 0. The tongue and its characteristics 1. The tongue and the physical actions that are its peculiarities 1.1. to make an action of the tongue 1.2. to seize, to pull with the tongue 1.3. to lick 1.3.1. consequence (1): to moisten and stick 1.3.2. consequence (2): to be smooth, polished 1.4. to savour, to taste 2. The tongue as an instrument of language: to speak, to speak in various ways, to be talkative, to malign, to wrestle in words with somebody, to hurt with malevolent words; to speak with authority > to order. 3. The tip of the tongue: pointed, to be sharp, to become a point and thus to prick. Here we have a relationship of the type part/whole: only the tip of the tongue is taken into consideration. Note a possible interference with 2. – he who has a pointed tongue is likely to utter hurtful remarks. We link to 3. the cases in which the pointed characteristic is explicitly given as cause. 4. Tongue of fire: to blaze, to burn, to scorch. Consequently, all the senses expressed by this radical are made explicit in the complex diagram below which constitutes its lexicogenetic tree, and traces its phono-semantic composition, thus accounting for the homonymy :


This matrix is studied in detail in Bohas (2006).


georges bohas and abderrahim saguer

Matrix level M6





{[labial], {[labial], {[coronal], {[+lateral], [+continuant]} [coronal]} [dorsal]} [+continuant]} ‘air> odour’ ‘to strike a blow’ ‘to strike a blow’ ‘the tongue’

{[labial], [dorsal]} ‘curvature’ Etymonial level {b,x}1








Radical level


Once this tree is exhaustively constructed, the analysis is finished. In the present case, we have managed to link each sense to a matrix. This is not always possible. 3.2

The explanation through attachment to etymons

In other cases, the analysis only enables the etymons to be identified without reaching the matrix level. Consider the verb laata which has the senses: S1 – ‘to hurt, to strike, to injure, to harm somebody (either by shooting an arrow at him, or by looking at him with the bad eye’ S2 – ‘to print on an animal, on the neck, a mark with hot iron, to mark it’ S3 – ‘to hurry, to hasten’ S4 – ‘to delay paying a debt, to put off its payment’ S5 – ‘to go to pasture (livestock)’ Each sense is homonymic in relation to the others, but, in each case, it is possible to establish a phono-semantic relation with other words, enabling us to identify etymons, as we shall demonstrate. For S1 – ‘to hurt, to strike, to injure, to harm somebody (either by shooting an arrow at him or by looking at him with the bad eye.’ Taking into account:

explanation of homonymy in the lexicon of arabic


laata l[]t lahat l[h]t lāta l[w]t

: to touch, attain somebody, e.g. with an arrow : to strike, to attain somebody with an arrow : to strike, to attain somebody with an arrow or by looking at him with the bad eye we may establish that the three verbs all demonstrate the phonetic constant lt and very similar meanings (indeed for laata and lāta, the meaning is identical), enabling the identification of an etymon {l,t}. For S2, compare laata ‘to print on an animal, on the neck, a mark with hot iron, to mark it’ with the elements of the following paradigm: 'alata : to mark (a camel) on the neck with a transversal mark 'alama : to mark, to distinguish by a mark, by some sign 'alaba : to mark something, either by making incisions or by pressure ra'ala F. II : to incise the ear of a beast to mark it This enables us to establish the presence of a phonetic constant l and a semantic constant ‘to mark’, thus identifying the etymon {l,' }. Let us add the form: laxafa : to print a large cautery, to make a large burn on somebody in which lx constitutes either a variant31 of l or a clue to identifying a matrix. For S3, it is worth comparing la'ata—‘to hurry, to hasten’ with: 'atā F. II [' t]w : to hurry somebody, to tell him to hurry 'abata ' [b]t : to throw a horse into the race so as to make it sweat : to go, to advance rapidly and by lengthening the hata'a h[t' ] neck (of camel) 'atawwad ['t]d : rapid, hurried and tiring (voyage) This enables us to identify the phonetic constant ,t and the semantic constant ‘to go rapidly,’ which leads us to deduce that laata should be analysed as an etymon {',t} with the incrementation prefix l bearing the reflexive meaning.32 Here again, note the variant: xāta : to pass rapidly

31 32

We explain this notion the following pages. Which corroborates Hurwitz earlier quoted in subsection 3.


georges bohas and abderrahim saguer

We may relate S4 – laata: ‘to delay paying a debt, to put off its payment’ to: tāla t[w]l F. II : to allow one’s debtor an extension talā [tl]w : to wait, to be waiting, to defer talla [tl]l : to allow an extension, to give respite to one’s debtor latat [lt]t : refusal to pay or to recognize what one owes to somebody The phono-semantic constant presented by the above data enables us to identify the etymon {l,t}. Note that S3 and S4 are two contradictory senses: ‘to go rapidly’ and ‘to delay.’ This enantiosemy33 can be explained by the fact that the word is analysed as a blending of two etymons with opposing senses: lt ‘to delay’ and x t ‘to go rapidly’ (D type blending: (Khatef, 2003 and 2004)). This enantiosemy is homonymic, since the two senses have nothing in common. The enantiosemy, which is usually presented as a ‘quirk of Arabic’ is, in fact, trivially predicted by TME (for more examples, see Bahri 2003). Finally, for laata S5 – ‘to go to pasture (livestock)’ a semantic link may be established with: : to graze in this or that place; to go to graze freely; ra'ā to put to graze, to take to graze : to leave to graze freely rata'a F. IV 'āra : to go away, to move away (of, amongst others, a horse that goes off grazing here and there) : to graze freely raba'a These forms are obviously related to the etymon {r,' } linked to the sense ‘to go to graze (freely).’ It seems plausible that the etymon l from laata S5 ‘to go to pasture (livestock)’ is an allophone of this etymon, r and l being from the same class [+approximant] [coronal]. The definition of the allophones of etymons is set out in Bohas and Dat (2007):

33 We use enantiosemy (i.e. reverse semantics) for words that mean something and its opposite, such as ‘big and small,’ 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.’

explanation of homonymy in the lexicon of arabic


Phonetic evolution may provoke a modification of the signifier and bring about the appearance of what we will call etymon allophones, which are phonetic variants of matrix etymons (from the set of a given matrix). Most frequently, the etymon allophones bring into play acoustic factors, which explains the confusion of segments in the communicational process. We posit that if: [b] /  {a , _} and [c] /  {a , _} where  {a,b} is a matrix etymon and [b] and [c] are phonemes with one or several shared phonetic features other than the vector of features demanded by the matrix combination, and that they correspond to two lexical items that are conceptually related, linkable (not necessarily identical), then [b] and [c] are the free variants of the phoneme (belonging to the paradigm defined by the vector of features) that enters into the composition of the matrix etymon. The allophonic forms, which enlarge the logical number of etymons belonging to a binary matrix of features, characterize the etymons, enlarged or not, the articulation of which is weakened34 or loosened in verbal communication; these allophones are, over time, recovered and incorporated into the lexicon of the language. These are free phonetic variants—historical and/or dialectal—of successful (widespread) innovations that co-exist with the source-forms, such as the elements in the following list (from the lexicon of Hebrew). s ¤ s / š /


: to demolish, to knock over, to knock down, to pull out nātas : to break, to destroy sādāh Niph. (hapax) : to be desolated, ravaged śādad Pi. : to break up lumps of soil, to harrow, to level a piece of land šādad : exercise violence, to desolate, to wreck, to destroy, to devastate In all these examples, the existence of the minimal pair engenders no major lexical opposition, the semic difference often resulting from the translation.

This argumentation enables us to consider laata S5 ‘to go to pasture (livestock)’ as a realization of the etymon {l,' } allophone of {r,' } given the (obvious) semantic relation and the phonetic relation we have established, but, in our analysis of this word, we have yet to trace it to a clearly identified matrix.

34 ‘The natural tendency of the speaker is to limit effort of his speech and to avoid sharp shifts in the use of speech organs.’ (Lipinski, 1997, 186).


georges bohas and abderrahim saguer

laata, which realizes the etymons: {',t}, {l,t} and {l,' }, is therefore typical of the second level of explanation, namely identification of the etymons. When we go on to study specific cases in the fourth part, we will move from one of the two levels of explanation to the other. In some instances we can link the etymon to a matrix; in others we may only link the radical of a word to an etymon, while in others, we can provide no analysis for the simple reason that, as things stand, identification of the matrices and etymons is not yet complete. The objective we pursue is to fully organize the lexicon into matrices. Is this a chimera, as some do not fear to say? In our eyes, the best answer is provided by Darwin: [. . .] it is those who know little, and not those who know much, who so positively assert that this or that problem will never be solved by science.35

4. Case study Within the framework defined above, we shall now analyse some words. 4.1

lah ana

S1 – ‘to become fond of somebody, to fall in love with somebody’ S2 – ‘to speak Arabic badly’ S3 – ‘to speak a particular slang with somebody so as not to be understood by others’ S4 – ‘to understand, to hear a word, an expression (link speak/understand)’ Let us start by studying sense S1 – ‘to become fond of somebody.’ A phonetic and semantic relation can be easily established with: h anna [h n]n : to groan, to make a groan of tenderness (of certain animals e.g. a female camel showing tenderness to her young) nāh a n[w]h : to coo, to groan (of pigeons) h anh ana [h n]h n : to have and to show tenderness, compassion, with emotion and worry

35 Charles Darwin: ‘The Descent of Man and Selection in Relation to Sex,’ 4th paragraph, quoted in Quiniou (2006); for the full text visit

explanation of homonymy in the lexicon of arabic


h anā [h n]w

: to have great tenderness for somebody (also used of a mother who through love of her children, does not want to re-marry) h aniba [h n]b and F. II : to feel a sentiment of pity, of compassion for somebody sah ana s[h n] F. III : to treat somebody with goodness The above leads us to deduce that, for this sense, lah ana should be analysed as a form incorporating the etymon [h ,n] through initial incrementation with an l which acts as a prefix marking the middle voice. In the semantic organization of matrices, the starting point is a concrete sense. In this we agree with 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 under study is particularly illuminating. The word h anna which has the abstract sense ‘to be moved,’ ‘to have compassion for somebody,’ ‘to feel great tenderness for somebody’ has the matrix sense ‘to groan, to make a groan of tenderness (of a female camel).’ Exactly like anna ‘to groan’ which comes from the same matrix: Matrix 4 {[coronal] , [pharyngeal]} [-dorsal] [-voiced] Notional invariant: ‘low voice, muffled, hoarse noise’ From the physical groan, as a noise, we pass to why we groan, to what the groan expresses. In h anna, the two senses are maintained, whereas, in lah ana, only the abstract sense appears. We have thus identified the matrix of which lah ana ‘to become fond of somebody’ is a realization. Let us now examine senses S2 and S3 which are clearly linked. S2 – ‘to speak Arabic badly,’ S3 – ‘to speak a particular slang with somebody so as not to be understood by others.’ For these senses, the analysis [lh ]n is used, in other words, the etymon is {l,h}, which is itself a realization of Matrix 8 {[+lateral], [+continuant]} linked to the notional invariant ‘tongue;’ furthermore, we have already seen that ‘to speak’ is one of the developments of the ‘tongue’ notional invariant:


georges bohas and abderrahim saguer

2. The tongue as an instrument of language (cf. 3.1 above): to speak, to speak in various ways, to be talkative, to malign, to wrestle in words with somebody, to hurt with malevolent words; to speak with authority > to order.36 The same sense is found in: laxxa [lx]x : to be unintelligible, to speak (especially Arabic) in an unintelligible way laxlaxāniyy [lx]lx : who has difficulty speaking or only speaks lecherous language laxiya [lx]y : to be very talkative and say a lot of rubbish laġā [lġ]w : to say futile things, to utter vain, flippant or careless words S4 – ‘to understand, to hear a word, an expression’ that can be found in lah ina with ‘to be intelligent,’ seems to accept the same analysis. The causal relation we assume, i.e. speak > understand > be intelligent is, in fact, explicit in Syriac; in the latter, mlīlā means both ‘who speaks’ and ‘who is intelligent’ or in Greek in which logikós covers two series of meanings: I. ‘which concerns speech’ and II. ‘which concerns reason.’37 Therefore the lexicogenetic tree of the word may be constructed: M8


{[+lateral], [+continuant} ‘tongue’

{[consonantal], [pharyngeal]} ‘muffled voice . . .’



[lh ]n

l[h n] lah an

36 37

See for the organization of the conceptual field above at the end of subsection 3.1. See Bailly (1950).

explanation of homonymy in the lexicon of arabic 4.2



Reorganizing the Kazimirski, the following senses can be said to be included: S1 – ‘to head into the interior of the land’ S2 – ‘to prick somebody, to bite (of a scorpion or snake)’ S3 – ‘to hurt somebody with malevolent words, with features of satire’ S4 – ‘to malign somebody’ S5 – F. IV ‘to sow enmity between men’ These five senses can be reduced to two: A ‘to make one’s way into the land’ B in relation with ‘tongue,’ as we shall show. As far as the first (A.) sense (‘to head into the interior of the land’) is concerned, this form is semantically linked to: nasaa n[s] : to head into the interior of the land, country tasaa [ts]38 x [s] : to travel, to cross a country, to head into the interior of the land šasaa š[s] : to be remote, to be situated at a great distance This fact leads us to deduce that lasaa should be analysed as a form incorporating the etymon {s,' } through initial incrementation of a prefix l marking the reflexive meaning, in the same way as the n at the beginning of the form nasaa. For the other (B) senses: S2 – ‘to prick somebody, to bite (of a scorpion or snake)’ S3 – ‘to hurt somebody with malevolent words, with features of satire’ S4 – ‘to malign somebody’ S5 – F. IV ‘to sow enmity between men’ The etymon {l,s} that the radical develops is a realization of Matrix 8 organized around the notional field ‘tongue’ as was earlier discussed at the end of subsection 3.1. Of these four senses, S2 – ‘to prick somebody, to bite (of a scorpion or snake)’ is linked to S3., like: lasaba : to prick somebody (said properly of a snake) 38 Tassa: F. II: ‘to head into the interior of the land, country.’ The radical thus results from the blending of the two etymons.


georges bohas and abderrahim saguer

S3 – ‘to hurt somebody with malevolent words, with features of satire’ is linked to S2 or S3, as is: ladaga : to hurt somebody with one’s tongue, i.e. with a biting remark S4 – ‘to malign somebody’ is linked to S2, like: lasana : to give blows with the tongue, i.e. to malign somebody, to tear him apart S5 – F. IV ‘to sow enmity between men’ is in a cause > consequence relation with S2 and S3. We can therefore show the lexicogenetic tree which, for {l,s}, goes back to the matrix, and which, for the etymon {s,' } remains at the etymonial level: M8


{[+lateral], [+continuant]} ‘tongue’ {l,s}

{s,' }



‘to head into the land’



lasafa and lasifa

This entry covers the following senses: S1 – ‘to fit by placing one next to the other and one over the other (e.g. stones in constructing a wall, in building)’ S2 lasifa – ‘to be dried and stuck to the bones (of skin on a very scrawny body)’ S3 – ‘to wind a strap of sinew around the base of an arrow’ S4 – ‘to shine, to gleam’ These four senses do seem to be in homonymic relation, as it is difficult to see how one could draw some sort of plausible semantic relation between them.

explanation of homonymy in the lexicon of arabic


For S1 – ‘to fit by placing one next to the other and one over the other (e.g. stones in constructing a wall, in building),’ comparison with the words below leads us to identify a common etymon {s,f } ‘to arrange, to organize:’ saffa [sf ]f : to arrange in order xasafa x[sf ] : to fit and join solidly sannafa s[n]f : to compose, to make (a work, a book) For S2, we recognize matrix 8 as in item: 1.3.1. consequence (1): to moisten and stick, as in: lassa [ls]s F. VIII : to attach oneself, and to stick strongly lasiqa [ls]q : to be stuck to the bones lasiqa [ls]q : to be stuck lasaġa [ls]ġ : to be dried and stuck to the bones (of skin or a very scrawny body) For S3 – ‘to wind a strap of sinew around the base of an arrow,’ consider: : to wind a solid strap or a flattened sinew around rasafa r[sf ] the tip of an arrow to make firm the iron that has been fitted 'afasa [ fs] : to wind a ifās around the mouth of a bottle 'asaba [sb] : to bandage, to wind a headband, bandage around; to put a dressing on (the head, a member) Examining the above, we may identify the etymon {s,f }, which is itself a realization of Matrix 6, f [labial] and s [dorsal],39 of which the notional invariant is ‘curvature;’ and of which ‘to surround’ and ‘to wind around’ are consequences.40 Finally, lasafa in sense S4 – ‘to shine, to gleam’ can be brought into relation with: walafa w[lf ] : to shine time after time with repeated flashes that come in uninterrupted succession (of lightening) jafala : to shine

39 40

Emphatics are characterized by the features [dorsal], [pharyngeal], [coronal]. See Bohas (2000: 115–117).


georges bohas and abderrahim saguer

The above enables us to identify the etymon lf ‘to shine.’ As to the matrix of which it is apparently a realization, it should contain the feature [+approximant], as in: balaqa F. VIII : to shine, to gleam baraqa : to shine, to gleam, to be shiny lamah a : to shine lamaa : to shine ramah a : to shine (of lightening) In any case, as in walafa and lasafa, the composition {[+approximant], [labial]}41 connected to the notional invariant ‘to shine’ can be observed. This is a complex matrix that has yet to be studied. Suffice it to say that, in the present state of our knowledge, the etymon has been positively identified, while the matrix has not. We can trace the lexicogenetic tree thus: M8



{[+lateral], {[+approximant] ?} [+continuant]} ‘tongue’ ‘shininess’

{[labial], [dorsal]} ‘curvature’


{l,f }

{s,f }1

{s,f }2

[ls] f

l[s] f

l[sf ]

l[sf ]

‘to organize’



lasaba and lasiba

S1 – ‘to prick somebody (of snake)’ S2 – ‘to give somebody a lash with a whip’ S3 – lasiba ‘to lick; to attach and to stick to something’

41 The same remark as in footnote 28 is appropriate here. If gutturals appear to be a part of this matrix, then they join the class of approximants, otherwise, [approximant] will have to be restricted by the addition of [coronal].

explanation of homonymy in the lexicon of arabic


Take S2 to start with. This form is related both semantically and phonetically to: saba a [sb] : to whip somebody with a whip until he bleeds This fact leads us to deduce that lasaba should be analysed as a form incorporating the etymon {s,b} through initial incrementation of the l, which has no semantic value. The etymon {s,b} comes from Matrix 1 {[labial], [coronal]}, the notional invariant of which is ‘to strike a blow,’ with the specification, in this case, of the means ‘with a whip,’ and other specifications in: rabasa : to strike with the hand safaa : to strike, to give a blow (especially of birds, when, in fighting, they give each other vigorous blows with their wings) sāfa : to strike somebody with a sabre nasama : to strike the ground with a foot The two other senses come from the matrix {[+lateral], [+continuant]} ‘tongue,’ the first enters the heading: 3. The tip of the tongue: pointed, to be sharp, to become a point, and thus to prick. As in: lasaa : to prick somebody, to bite (of scorpion or snake) lasana : to form a point, to give a pointed form, (e.g. to a shoe, etc.) : to prick (of scorpion) salama : to prick somebody, to make a bite (of a snake) F. IV passive : uslima: to be pricked by a snake And the second the heading: 1.3. to lick and 1.3.1 (see the end of subsection 3.1 above) as in: lassa : to lick (a vase, a frying pan, etc.) lasada : to lick (a vase) ladasa : to lick : to lick lahasa Drawing up the lexicogenetic tree is therefore an easy matter:


georges bohas and abderrahim saguer M8


{[+lateral], [+continuant]} ‘tongue’

{[labial], [coronal]}





‘to strike a blow’



lah afa

Whereas the analysis of lasaba presents no particular difficulty, that of lah afa is formidably complex, which is why we have left it till last. Indeed, this verb attests several senses, as follows: S1 – F. I: ‘to lick something’ F. IV: ‘to pressure somebody, to trouble, to ask with insistence’ S2 – F. I: ‘to envelop somebody in a sheet, a blanket’ F. III: ‘to help, to assist somebody’42 F. IV: ‘to clothe and envelop somebody in a garment’ F. V: ‘to envelop oneself in a sheet, in a cloth, in a piece of material’ F. VIII: ‘to envelop oneself in a piece of material’ S3 – F. I P.: ‘to suffer losses in one’s belongings, flocks, etc.’ 43 F. IV: ‘to do something bad to somebody, to cause him some harm’ S4 – F. IV: ‘to pull out (e.g. somebody’s nail)’ S5 – F. IV: ‘to burn, to have something consumed in fire’44 Let us start with S1. The radical includes l and the fricative h , and can be a realization of Matrix 8: {[+sonorant] , [+continuant]} [+lateral]

42 43 44

This sense is linked to this block, see below. ‘P.’ indicates the passive. Attested in Asās al-balāġa.

explanation of homonymy in the lexicon of arabic


Notional invariant: ‘tongue’ ‘to lick’ coming under the heading 1.3.45 The following forms develop the etymons {l,h} and {l,' }, that are analogous realizations of this matrix: lahika [lh ]k : to lick lahisa [lh ]s : to lick : to lick la'iqa [l]q la'ā [l]w F. V : to lick The meaning expressed by F. IV that we have placed under sense S1, ‘to pressure somebody, to trouble, to ask with insistence,’ seems to correlate well with the idea of licking, exactly as in French in which ‘coller,’ ‘to stick’ can mean46 ‘to impose one’s presence on somebody’ and ‘collant’ ‘sticky:’ ‘that which one cannot rid oneself of.’ Thus, we have the following chain: lick >stick >insist, the sense that we find in other manifestations of the etymon {l,h}: lahh a [lh ]h : to insist on something to somebody, to persist in asking him (for) something hafala h [ f ]l F. V : to insist halata [h l]t : to insist, to pressure, to scold somebody, and to swear mahala m[h l] F. V : to insist and to pressure somebody For S.2, the emergence of the sense ‘to envelop’ undoubtedly comes from the blending of the two etymons: lf x h f, which can be clearly seen in: laffa : to envelop, to twist, to surround with something h affa : to surround somebody with something, to envelop with something Is it possible to go further and identify the matrix from which these etymons stem? Without doubt, if we take into consideration the following data: laffa : (also means:) to gather, to collect from all sides daffa : to gather, to bring together47 A hypothesis defended by Cantineau (1951), amongst others, is that in protosemitic there was a lateralized emphatic that we shall write dl. In Bohas and Janah (2000) it was argued that this dl had split in Arabic into

45 46 47

See above subsection 3.1. See the Petit Robert See also: mudāf d[y] f : ‘surrounded, attacked, encircled on all sides’.


georges bohas and abderrahim saguer

two phonemes: d and l. A word bearing one meaning had thus given birth to two words with the same sense (modulo a few nuances). Thus we would have: dlaffa ‘to bring together’



‘to gather, to bring together’ ‘to gather, to collect from all sides’

This means that the l of la a is not a ‘true’ l, but lexically a dl and that, as such, it has the feature [pharyngeal] of the emphatics. The pair: laffa : to envelop, to twist, to surround with something h affa : to surround somebody with something, to envelop with something is in fact a dlaffa/h affa pair and the etymons dlf/h f are realizations of Matrix 3: Matrix 3 {[ labial] , [ pharyngeal]} Notional invariant: ‘(a) tightening’ The relation between ‘to envelop,’ ‘to surround’ and ‘to tighten’ is merely one of cause to consequence. The sense expressed by F. III: ‘to help, to assist somebody’ seems to be accounted for in terms of metaphor: to help is ‘to surround somebody with assistance, protection or affection.’ As an argument in favour of this relation, we have the verb h affa that explicitly shows this meaning, since, in the Kazimirski, after ‘to surround’ comes the sense ‘to be constantly around somebody, and to be attentive to serve or protect him.’ S3 constitutes a single semantic block: a : F. I P. : to suffer losses in one’s belongings, flocks, etc. b : F. IV : to do something bad to somebody, to cause him some harm Let us bring lah afa into relation with the forms which manifest the same properties: falla F. IV : to lose one’s flocks falaa : to lose, to reduce to nothing faliya : to be cut, separated from the rest of the body lafaa : to peel, to skin wafala : to peel something by removing the bark We introduce an etymon {l,f }; l is [coronal] and f [labial]. This etymon can thus be a realization of Matrix 1:

explanation of homonymy in the lexicon of arabic


{[labial], [coronal]} Notional invariant: ‘to strike a blow’ ‘Loss, harm’ comes under the heading B.3,48 global consequence, as in: h afata : to destroy, to lose talifa : to perish As for S4 – F. IV: ‘to pull out (e.g. somebody’s nail)’ it is a realization of a Matrix under study: {[+approximant], [+continuant]} [coronal] with the notional invariant: ‘to bring something to oneself.’ This manifests itself in the words quoted earlier in section 3.1. As f and h are both continuants, it seems reasonable to consider the radical as a blending of the two etymons that both realize this matrix: lh x lf, B type blending.49 There remains sense S5 – F. IV: ‘to burn, to have something consumed in fire.’ In this meaning, we can relate lahafa to: lafaha : to burn, to cause harm through its intensity (of fire, or a very warm wind) fayh : heat caused by a star sahafa : to burn, to have something consumed in fire The above relation reveals the etymon { f,h}; the semantic relation with ‘to blow’ remains to be established in order to link it to Matrix 2. As for the residual cases F. II : to let the bottom of one’s garment scrape the ground, to wear it very long so that it trails, by extension: to walk proudly F. IV : to come to the foot of a mountain and lih f: the foot of a mountain At this stage in our research, all that can be noted is that the latter is perhaps to be related to h āffatun: ‘edge, margin, extremity,’ although we cannot establish this with certainty. All these comparisons are of limited interest, but it is worth remembering, as we said at the start, that certain points are still unclear.

48 49

See Bohas (2000: 79). See Bohas (2000: 50).


georges bohas and abderrahim saguer

Consequently, we shall limit ourselves to those senses that are clear, in order to construct the lexicogenetic tree: M8


{[+lat], [+cont]} {[labial], [coronal]} ‘tongue’ ‘strike a blow’



{[phar], [lab] } ‘tightening’

{[+approx], [+cont]} ‘to bring to oneself ’


{l,f }

{h,f }

{l,h} x {l,f }

[lh ] f

l[h ] f

l[h f ]

[l,h ] x [l,f ]

lah af

5. Conclusion Our method is thus distinct from that of the partisans of the triconsonantal root. For them, it is su cient to identify the three consonants in order to consider the analysis complete, even if semantic incongruities and incompatibilities are evident, and even if this identi cation provides no explanation of phono-semantic links between words, such as homonymy and enantiosemy. What does reassure them, however, is that they can pride themselves on having reached a state of certainty. . . . that the root of maktab = —ktb is a certainty! Yet why should the root of istadaytu be —dw rather than —dy, and, if it is —dw, what phonetic motivation is there to be found in istafaltu to justify the passage from w to y in istadaytu? Indeed, this certainty is not as de nitive as they would have us believe. With our approach based on argumentative reasoning, we might make mistakes: one word might, perhaps, be matched up with another on the basis of such and such a property without us having noticed this relation. Given the explanatory results of our approach, which adopts a heuristic point of view, this is a risk we assume. The reader will have noticed that all our research has been carried out taking the lexicon of Arabic as a synchronic whole. As we have often repeated, in the study of the lexicon, it is vain to go back to a previous

explanation of homonymy in the lexicon of arabic


biliteral stage in order to diachronically derive a triliteral stage. In other words, the old debate—biliteral or triliteral—is immaterial. The binary composites (the matrices and the etymons) and the ternary composites are there before our eyes: you only have to open a dictionary to find them. Every radical is, according to the level of explanation, binary or ternary, as the explanation of homonymy demonstrates all too well. Admittedly, no one is obliged to overstep the conception, invented by the Arab grammarians, of the tri- or quadri-consonantal root. However, to restrain oneself to this is to forego the opportunity to provide explanations for the phono-semantic phenomena contained in the lexicon, phenomena that everyone may observe. To work with a theory that posits that the Sun rotates around the Earth enables one to explain a certain number of observable phenomena;50 but to work with a theory that posits that the Earth rotates around the Sun enables one to explain a greater range of phenomena, and fits in better with other knowledge we have of the movement of the stars.

6. References Asās al-balāġa = Abū l-Qāsim Mahmūd b. Umar az-Zamaxšarî, Asās al-balāġa, Abdarrahīm Mahmūd, ed. Bayrūt: Dār at -ti bāa wa-n-našr. Bahri, A. 2003. L’énantiosémie en arabe, Doctoral thesis. University Paris 8. Bailly, A. 1950. Dictionnaire grec français. Paris: Hachette. Bohas, Georges. 1997. Matrices, étymons, racines, éléments d’une théorie lexicologique du vocabulaire arabe. Paris: 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, 11–41. ——, 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 A. Janah. 2000. “Le statut du dād dans le lexique de l’arabe et ses implications.” Langues et Littératures du Monde Arabe, 1, 13–28.


We find in The Legend of Alexander: ‘For the Sun is the servant of The Lord, which interrupts its course neither day nor night.’ This idea that the stars are in the service of God seems to date back at least to Bardesane (born in 154): ‘Neither the Sun, nor the Moon, nor the other beings which are superior to us in any thing have received power over themselves, they are on the contrary subject to a law and, consequently, they do what they have been ordered and never anything else. The Sun never says ‘I shall not rise at the given hour,’ nor the Moon ‘I will no longer have phases, I will neither wax nor wane,’ . . . All these creatures are servants and remain subject to a law: they are instruments of the wisdom of The Lord, Who is infallible.’ (See Teixidor 1992).


georges bohas and abderrahim saguer

——, and M. Dat. 2005. “La matrice acoustique {[dorsal], [pharyngal]} en arabe classique et en hébreu biblique, première esquisse.” Mélanges de l’Université Saint-Joseph, LVIII, 125–143. ——, and M. Dat. 2007. Une théorie de l’organisation du lexique des langues sémitiques : matrices et étymons, Lyon: ENS éditions. ——, and A.R. Saguer. 2006. “Sur un point de vue heuristique concernant l’homonymie dans le lexique de l’arabe.” In Edzard, L. & J. Watson (eds), Grammar as a Window onto Arabic Humanism. A Collection of Articles in Honour of Michael G. Carter, Wiesbaden, Harrassowitz Verlag, 130–154. ——, and R. Serhane. 2003. “Conséquences de la décomposition du phonème en traits.” In: Angoujard, J.-P. and S. Wauguier-Gravelines (Eds.): Phonologie. Champs et perspectives. Lyon: ENS éditions, 131-155. Cantineau, Jean. 1951. “Le consonantisme du sémitique.” Semitica, IV, 79–94. Dat, M. 2002. Matrices et étymons. Mimophonie lexicale en hébreu biblique. Doctoral thesis, Lyon: Ecole Normale Supérieure Lettres et Sciences Humaines. Dell, F. 1973. Les règles et les sons: introduction à la phonologie générative. Paris: Hermann. Diab, S. 2005. La matrice {[coronal], [dorsal]}, Les étymons impliquant le jīm, Masters 2 Research Paper. Lyon: École normale supérieure lettres et sciences humaines. Guerssel, M. and J. Lowenstamm. 1993. The Derivational Morphology of the Classical Arabic Verbal System. ms. UQAM and University Paris VII. Halle, M. 1991. “Phonological Features.” In W. Bright (ed.): Oxford International Encyclopedia of Linguistics. Oxford: Oxford University Press, 207–212. Hurwitz, S. 1913 [1966]. Root-Determinatives in Semitic Speech, a Contribution to Semitic Philology. New York: Columbia University Press. Joüon, P. 1923. Grammaire de l’hébreu biblique. Rome: Institut biblique pontifical. Kazimirski, A. de Biberstein. 1860. Dictionnaire arabe français, Paris: Maisonneuve et Cie. Kenstowicz, M. 1994. Phonology in Generative Grammar. Oxford UK : Blackwell. Khatef, L. 2003. Statut de la troisième radicale en arabe: le croisement des étymons, Doctoral thesis. University Paris VIII. ——. 2004. “Le croisement des étymons: organisation formelle et sémantique.” Langues et Littératures du Monde Arabe, 119–138. Lisān = Jamāl ad-Dīn Abū l-Fadl Muhammad b. Mukarram b. Alī b. Ahmad b. Abī l-Qāsim b. H abqa Ibn Manzūr, Lisān al-Arab, s.d. Abd Allāh Alī al-Kabīr, Muhammad Ahmad H asab Allāh, Hāšim Muhammad aš-Šādilī, eds. Cairo: Dār al-Maārif. Lipinski, E. 1997. Semitic Languages. Outline of a Comparative Grammar. Leuven: Peeters. Mansouri, W. 2006. La place du trait [sonorant] dans les matrices de l’arabe. Masters 2 Research Paper, Lyon, École normale supérieure lettres et sciences humaines. McCarthy, J.J. 1986. “OCP Effects: Gemination and Antigemination.” Linguistic Inquiry, 17,2. 207–263. Nyckees, V. 1998. La sémantique, Paris: Belin. Qāmūs = Majd ad-Dīn Muhammad b. Yaqūb al-Fayrūzābādī, Al-Qāmūs al-Muh īt. Bayrūt: Muassasat ar-Risāla. Quiniou, Y. 2006. “La mort scientifique de Dieu.” Le nouvel observateur. Hors-série, 38–41. Le Petit Robert, dictionnaire alphabétique et analogique de la langue française. 1967 [1993]. Paris: Dictionnaires Le Robert. Saguer, A.R. 2000. “L’incrémentation des préfixes dans le lexique de l’arabe. Le cas du n.” Actes du colloque Journées de linguistique arabe et sémitique, Langues et littératures du monde arabe, 1, 57–82. ——. 2002a. “L’incrémentation des préfixes dans le lexique de l’arabe. Le cas du m.” Langues et littératures du monde arabe, 3, 29–57.

explanation of homonymy in the lexicon of arabic


——. 2002b. Z āhirat al-isbāq fī l-judūr al-arabiyya. Agadir: publications de l’Université Ibn Zuhr. ——. 2003. “La matrice {[+nasal), [coronal)}, «traction» en arabe. Première esquisse.” Langues et littératures du monde arabe, 4, 138–183. Segeral, Ph. 1995. Une théorie généralisée de l’apophonie, Doctoral thesis. University Paris 7. Serhane, R. 2003. Etude de la matrice {[labial], [dorsal]} en arabe, Doctoral thesis. University Paris 8. Teixidor, J. 1992. Bardesane d’Edesse, la première philosophie syriaque. Paris: Les éditions du Cerf. Yeou, M. and S. Maeda. 1994. “Pharyngales et uvulaires arabes sont des approximantes: caractérisation acoustique.” 20ème Journées d’Études sur la Parole, Trégastel, 409–414.


1. Introduction I would like to start this paper with some terminological notes. I use the dichotomy ‘community language’ versus ‘superimposed language’ to refer to the typical unequal social-economic status of the bilingual speaker’s languages. These terms refer to local as well as global power relations and their sociolinguistic consequences. For instance, whether in Portugal, Brazil or the United States, Portuguese speaking people learn English in order to gain access to valuable information and upward social mobility, i.e. education, media and employment. I will use the term ‘socially dominant’ for the language the individual speaker is most exposed to in her daily life. This could be either the community or the superimposed language, depending on the local situation. Thus English is more likely to be socially dominant for a particular Portuguese/English bilingual member of the Portuguese community in the US, while Portuguese will be socially dominant for most bilinguals living in Portugal. The terms ‘matrix language’ and ‘embedded language’ are grammatical notions referring solely to local syntactic units of analysis in bilingual speech. The higher order constituent is the matrix in which lower order constituents are embedded. In mixed sentences higher and lower order constituents are in different languages. In most instances the community language functions as the matrix language and superimposed language elements are embedded. However, the reverse occurs in a minority of cases so the terms should not be confused.1

1 Since the early 1990s, Carol Myers-Scotton has been the most influential promoter of the insertion approach to code-switching and the terms matrix and embedded language.


louis boumans

The present article deals with verbs from a superimposed language that function as embedded elements in community language discourse. The community language may or may not be socially dominant, and I argue that this makes a difference for the way in which foreign verbs are embedded.

2. The integration of foreign verbs There are three ways in which foreign verbs are integrated in the matrix language, two of which are common. One is the complete morphological integration. Some basic form of the foreign verb, typically the verb stem or the infinitive, is treated as the verb stem of the receiving matrix language, and verbal categories of the latter are expressed by matrix language morphology. Gloss (1) is a Moroccan Arabic/French example showing the French verb stem montr- (from montrer [m ~tre] ‘to show’) with an Arabic prefix and suffixes.2 (1) waš ġadi y-montr.i-w-l-ek . . . Q FUT 3-show-PL-to-2SG ‘Are they going to show you . . .?’ MA3/French (Wernitz 1993, 308)

The insertion of foreign verb stems without overt morphological integration in matrix languages lacking verbal morphology, such as various Austronesian languages (Van Staden 1999) can be considered as a subcategory of the morphological integration strategy, even if the integration is not overtly expressed by ML morphemes.

I concur with the fundamentals of her original Matrix Language Frame (MLF) model (Myers-Scotton 1993), except for the definition of the matrix language. In my view, all syntactic constituents function as a matrix for lower-order constituents, whereas in the MLF model only the Complementizer Phrase functions as a matrix. I refer to earlier work for more details on this approach to code-switching (Boumans 1998, Boumans and Caubet 2000, Boumans 2002). 2 The vowel i is not a proper suffix. Embedded French verbs are modelled on a class of Arabic verbs ending in a vowel. This vowel is subject to a/i ablaut. Cf. Caubet (1993), Boumans (1998), and Boumans and Caubet (2000). 3 The following abbreviations are being used: in the main text: MA Moroccan Arabic; in the glosses to numbered examples: 1,2,3 first, second, third person; ACC accusative; AGR agreement; ART article; AUX auxiliary; DEF definite article; FUT future tense; IMPF imperfective; INF infinitive; M masculine; NEG negation; PASTPART past participle; PL plural; PROGRPART progressive participle; PRT preterit; REL relative clause marker; SG singular.

the periphrastic bilingual verb construction


The other common strategy is the periphrastic construction. This strategy is equally common in the world’s languages and can also be illustrated with Moroccan Arabic, this time in contact with Dutch: (2) bġa y-dir li-h aanvall-en want 3-do to-3SG attack-INF ‘It [the bird] wanted to attack him.’ MADutch (Jamal, 16, Utrecht 2000)

In his book Bilingual Speech (2000), Muysken presents a typology of verb integration with special attention for the periphrastic type. In the periphrastic construction, he argues, the foreign verb can be a nominalization, an infinitive or an adjunction. In the first case, the nominalized verb is the complement of the ‘helping verb,’ in the case of an infinitive the helping verb must be analysed as an auxiliary. In the adjunction analysis, the foreign verb and the ‘helping verb’ form a kind of verbal compound. Which of these three analyses is most appropriate does not only depend on the language pair involved. As I discussed elsewhere (Boumans 1998; 2000), strategies and grammatical constructions differ among individuals belonging to the same bilingual community. Compare examples (2) and (3). In Jamal’s utterance in (2), the patient participant of the Dutch transitive verb aanvallen ‘to attack’ is expressed as the MA indirect object li-h ‘to-him’. The Patient role is expressed as an indirect object, because the Dutch verb occupies the Direct Object position of the MA transitive verb dar ‘to do’. This is in fact the common construction in the Dutch Moroccan community. Now in Samir’s example in (3), the patient of the transitive verb controleren ‘to supervise’ surfaces as the direct object suffix -hom ‘them’. The construction in (2) can be described as a helping verb plus a verbal noun object, while the one in example (3) is in fact very similar to the monolingual Dutch auxiliary construction. Compare (3) with its Dutch translation in (4). Note that the pronominal direct object of the main verb cliticizes to the finite auxiliary, even if an adjunct constituent (morgen ‘tomorrow’) comes between the two verbs. (3) škun ġadi y-dir-hom controler-en? who FUT 3–do-3PL supervise-INF ‘Who will supervise them?’ MA/Dutch (Samir 20, Nijmegen 1991) (4) wie gaa-t hun (morgen) controler-en who go-2SG 3PL tomorrow supervise-INF ‘Who will supervise them (tomorrow)?’ Dutch translation of (3)


louis boumans

The variation between speakers, and even between utterances of the same speaker, can be explained by the varying levels of competence in both languages, and by the effect of grammaticalization and conventionalization (Backus 1996). The third, less common strategy for the incorporation of foreign verbs deserves mentioning here for the sake of completeness. This strategy consists in inserting inflected verb forms rather than verb stems. The inflected foreign forms are mapped onto the ML paradigm and express ML verbal categories. A famous case in point is Mednyj Aleut, also known as Copper Island Aleut (Thomason and Kaufman 1988, 233–8). In this variety of Aleut, Russian inflectional patterns in finite verbs replaced Aleut ones while most other grammatical subsystems remained intact. Crucially, the Russian verbal inflections express Aleut tense and aspect categories. Similarly, Igla (1991) reports on a dialect of Romani nowadays spoken in a suburb of Athens in Greece. The speakers of this dialect of Romani moved in from Turkey in the 1920s but no longer speak any Turkish. However there are still approximately 30 verbs of Turkish origin that continue to be inflected with Turkish suffixes while following the Romani verbal paradigm, cf. Table 1. That is, the Turkish inflectional suffixes express Romani inflectional categories. TABLE 1 THE PRESENT TENSE OF TURKISH VERBS AND OF VERBS OF TURKISH AND NON-TURKISH ORIGIN IN AJIA VAVARA ROMANI (FROM IGLA 1991, 52)

present sg 1 sg 2 sg 3 pl 1 pl 2 pl 3


Ajia Vavara Romani

‘to write’ yaz-ar-m yaz-ar-sn yaz-ar yaz-ar-z yaz-ar-snz yaz-ar-(lar)

‘to write’ yaz-ar-um yaz-ar-sun yaz-ar yaz-ar-us yaz-ar-sunus yaz-ar-(lar)

‘to bring’ an-av an-es an-el an-as an-en an-en

(5) i thagarni kurta-du len e rom-en ART queen save-PRT REL ART gypsy-ACC.PL ‘(..) the queen saved (them) the gypsies.’ Romani/Turkish (Igla 1991, 53)

This latter strategy is actually very rare in language contact situations. However, it may be more common in the case of two closely related languages or varieties. For instance Malkiel (1986) draws attention to the Spanish conjugation pattern of a number of Portuguese verbs. Also

the periphrastic bilingual verb construction


non-tensed verb forms such as participles, infinitives and imperatives are quite readily interchangeable between languages, cf. (6). (6) Le estaba poniendo atención qué estaba 3.IO was.3SG put.PROGRPART attention what was.SG recorded record.PASTPAR ‘He was paying attention to what was recorded.’ Spanish/English (Pfaff 1979, 300)

3. Explanations for the distribution of the two common strategies Why do speakers opt for one strategy rather than the other? The two common strategies in particular, morphological integration and periphrastic constructions, allow us to compare the languages and linguistic situations involved. 3.1

Characteristics of the matrix language

When one makes a list of all language contact situations and the attested strategies for the integration of foreign verbs, it becomes clear that genetic or areal typological factors play an important role. In the IndoIranian and Turkic languages, for instance, the periphrastic construction appears in virtually all language contact situations. In connection with this, Muysken (2000, 194) speaks of ‘a large “linguistic area”, in this case, stretching from Sri Lanka to Greece.’ On the other hand, most western European languages seem to prefer the morphological integration strategy. A plausible explanation for the areal bias is that speakers who are accustomed to a certain strategy of incorporating foreign verbs will reuse this strategy in new contact situations. Some examples of this kind of ‘bilingual knowledge’ will be discussed below. Still, the two MA examples cited in (1) and (2) show that the matrix language is not the only factor deciding which incorporation strategy speakers will use. 3.2

Characteristics of the embedded language

One way to explain the difference between MA/French and MA/Dutch code-switching is to look for differences in the superimposed embedded languages. Heath (1989) argues that the phonological shape of the French verbs facilitates their incorporation in Arabic. Bilingual speakers associate the stressed final vowel /e/ of most French infinitives and other


louis boumans

verb forms with the final /i/ of the prefix conjugation (imperfective) of a class of MA verbs. Dutch infinitives typically end in an unstressed suffix -en that is pronounced as a schwa. It is not obvious whether this makes them phonologically less similar to the MA finite verb in /i/ than French infinitives. Firstly, MA phonology does not allow for the schwa in open syllables. This may lead Moroccan listeners to ignore word-final schwas in Dutch, and interpret the infinitives as consonant-final. Alternatively, however, Moroccans may interpret the Dutch final schwa as a full front or back vowel. Both tendencies can be observed in the speech of Moroccan learners of Dutch. Moreover, the vocalic ending in French verbs cannot be decisive, since Spanish, Italian and English verbs are morphologically integrated in North African varieties of Arabic in the same way as French verbs. Many of these Romance and English infinitives end in a consonant. Arabic/Spanish language contact is still common in the (formerly) Spanish occupied northern parts of Morocco (cf. Heath 1989, Herrero Muñoz-Cobo 1996). Cohen (1912) notes interesting observations on the Jewish dialect of Algiers, where the Spanish (or Lingua Franca) infinitive ending -ar is even extended to embedded French verbs. E.g. refuzarit from French refuser [rıfyze] in (8). Numerous verbs of Italian and English origin are found in Maltese, another variety of Maghribian Arabic (Aquilina 1965 [1987], Camilleri 1994, Mifsud, 1995).4 The stem extension -ja- in Maltese verbs of English origin, as in (10), is a reflection of the Sicilian ~ Italian infinitive suffix -are. (7)

frin-ar-t ~ frin-ar.i-t brake- INF-1SG ~ brake- INF.STEM EXTENSION -1SG (Sp. frenar) ‘I braked’ Tetouan Arabic/Spanish (Heath 1989, 184)


refuz-ar-it refuse-SPANISH INF-1SG (Fr. refuser) ‘I refused’ Jewish Arabic of Algiers/French (Cohen 1912, 432)


ti-ppartiċipa-w 3-participate-PL (It. partecipare) ‘you (pl) participate’ Maltese/Italian~ Sicilian (Camilleri 1994, 437)

4 As a matter of fact, the type of verb integration illustrated in (1) does not occur in the Middle Eastern varieties of Arabic. Instead the periphrastic construction is more common in that region.

the periphrastic bilingual verb construction


(10) ni-bbliċċ.ja-ha 1-bleach.STEM EXTENSION-3SG ‘I bleach it’ Maltese/English (Camilleri 1994, 443)

Thus, verb stems from various embedded languages and with diverse phonological characteristics can be morphologically integrated into Moroccan or Maghribian Arabic, and phonology does not seem to be an explanation for the periphrastic construction in the case of MA/Dutch. A second way in which the embedded language might influence the selection of the verb integration strategy is the use of periphrastic verb constructions in the embedded language itself. Following this line of thought, the bilingual MA/Dutch periphrastic construction could be inspired by periphrastic constructions with doen ‘to do’ in standard or non-standard Dutch.5 Again, the comparison with the other language pairs speaks against this hypothesis. Periphrastic constructions with a ‘do’ verb are much more common in English than they are in Dutch, while their abundance in French and Spanish may be similar to that in Dutch.6 Therefore, the occurrence of ‘do’ constructions in monolingual Dutch likewise does not explain why the periphrasis strategy is chosen to insert Dutch verbs in MA matrix clauses. 3.3

Characteristics of the sociolinguistic setting

In the case of the North African language contacts, neither the host language nor the embedded language is the sole factor determining the way in which the loan verbs are integrated in Arabic. For this reason I conjecture that the sociolinguistic setting in the Netherlands, where MA is a minority language, is responsible for the periphrastic construction in MA/Dutch. As a more general hypothesis, I suggest situations in which language contact is more intense, like migration to an urban industrialised society, favour the use of the periphrastic construction.7

5 Jacomine Nortier and Roeland van Hout made this suggestion when we discussed this paper at the SS14 workshop on borrowing in Gent, April 4, 2002, and on earlier occasions. 6 With the exception of certain infrequent constructions, do-periphrasis in Dutch is associated with non-standard regional varieties or child language. Cf. Nuijtens (1962, 154–57), Giesbers (1984), and Cornips (1994). 7 One could argue that the morphological integration of French verbs is found both in North Africa and in the North African diaspora in Francophone Europe and Canada. In this particular case, however, code-switching strategies that were already established in the homeland have simply been maintained in the diaspora.


louis boumans

In order to shed some light on this issue and to test the ‘intensity of contact’ hypothesis, the following two sections survey verb integration in two additional language pairs for which both integration strategies are attested: Greek/English and Portuguese/English. In both cases the influence of English has been described for immigrant communities in Anglophone countries as well as in the Greek and Portuguese speaking homelands.

4. Greek/English 4.1


The morphological integration of Italian verbs consists of adding Greek suffixes to the Italian infinitive in -are.8 (11)

It. posare ‘to pose’  

It. schizzare ‘to sketch’ Mainland Greek/Italian (Van Dijk-Wittop Koning 1963)


fundáro It. fondare ‘to anchor’ barkáro It. barcare ‘to board’ kompletáro It. completare ‘to fill up [cargo]’ Mainland Greek/Italian (Hartley 2001)

According to Swanson the ‘mildly productive verbal suffix -aro’ entered the Greek language in late Byzantine times (1958, 40). The Italian infinitive marker has been generalised to verbs of French and English origin as well: (13)

 Fr. débuter ‘to make one’s debut’  

Fr. lancer ‘to launch’ Mainland Greek/French (Contossopoulos 1978, 42)


flertaro to flirt stoparo to stop (a machine) sutaro to make a shot (soccer, basketball) Mainland Greek/English (Swanson 1958)

8 The authors on loanwords and codeswitching in Greek make use of different writing and spelling conventions. The spelling of the source publication is retained in the examples cited here.

the periphrastic bilingual verb construction (15)



/parkaro/ ‘to park’ 

/manadzaro/ ‘to manage’  

/triparo/ ‘to trip’ Mainland Greek/English (Apostolou Panara 1991, 50)

Interestingly, the recycling of a foreign verb marker like the Italian -are to mark verbs from a third language is yet another example of how ‘bilingual experience’ influences the way in which speakers treat words from foreign languages. Note the close parallel with -ar in the Jewish Arabic dialect of Algiers (8), and the -ja stem extension after English verb stems in Maltese (10).9 Most sources on loan verbs in Modern Greek mention only this -ar- extension with complete morphological integration of the foreign word. However, Apostolou Panara, while asserting that “verbs all adapt,” i.e. are morphologically integrated, mentions that ‘in some cases, along with the single lexeme we attest a periphrasis with the Greek verb /kano/ (‘to do, to make’) preceding a noun or a gerund’ (1991, 50), cf. the examples in (16). A few verbs occur exclusively as part of the periphrastic construction (17), whereas the periphrasis is judged ungrammatical in most cases (18). (16)

/flertaro/ besides /kano flert/ ‘to flirt’ /manadzaro/ besides /kano manadzing/ ‘to manage’ /stokaro/ besides /kano stok/ ‘to stock’ Mainland Greek/English (Apostolou Panara 1991, 50)


/kano kambing/ but not */kambaro/ ‘to camp’ /kano serfing/ but not */serfaro/ ‘to surf ’ Mainland Greek/English (Apostolou Panara 1991, 51)


/triparo/ but not */kano trip/ ‘to trip’ /parkaro/ but not */kano park/ ‘to park’ Mainland Greek/English (Apostolou Panara, 1991, 51)

It is possible that a closer study of the English vocabulary in Modern Greek would reveal regularities that govern the use of periphrasis. Periphrasis with káno could be associated with particular types of events, for instance habitual or durative activities. The periphrastic construction may be more common with recent English loans, or in the speech

9 In Dutch a similar situation obtains with the French infinitive marker -er [er] (Treffers-Daller 1994), which is sometimes used for English loanwords as well, e.g. formatter-en ‘to format’.


louis boumans

of speakers who are more exposed to English. kano kambing may be a calque after the French expression faire du camping (cf. káno ski in (24) below, possibly after French faire du ski). The available data do not allow for such conclusions at the moment. Still, I think it is safe to conclude from the literature that in Greece the periphrastic construction occurs as a minor pattern with some verbs only, while morphological integration is the rule. 4.2


Cyprus has been under British rule for more than eighty years, during which English was the language of administration. First, from 1878 on Britain administered the island in agreement with the Ottoman Empire. Then Cyprus was annexed by Britain when the Ottoman Empire enters into World War I on the side of Germany, and subsequently the island became a British Crown colony under the British rule. In 1960 Cyprus gained independence, and (Modern Standard) Greek became the language of administration. The English language remained influential through the tourism industry and the large international community on the island. There are a number of studies dealing specifically with the influence of English on Cypriot Greek. Papapavlou (1997) cites a list of English loan words found in written sources and tape-recorded speech. His list includes 23 verbs in the -aro conjugation. He makes no mention of the periphrastic construction. (19)

flεrdάro ‘to flirt’ rej˘istrάro ‘to register’ riskάro ‘to take a risk’ čarcάro ‘to change’ (money) ‘to check’ (inspect) čakhάro Cypriot Greek/English (Papapavlou 1997)

Similarly, Evripidou, also on the basis of written sources and recorded speech, lists 19 English-origin Cypriot Greek verbs, all of them integrated in the -aro class (2001). Goutsos discusses Greek/English codeswitching among members of a middle-class Cypriot family in Limassol (2001). This study differs from the other ones, as Goutsos does not list English loan words but rather focuses on the discourse functions of language choice in the conversations. He does not comment on integrated verbs of the -aro conjugation. He does cite three instances of the

the periphrastic bilingual verb construction


periphrastic construction. As in Apostolou Panara’s data, there are two types of construction, one in which the English verb is represented by the infinitive (or verb stem), see (20) and (21), the other with the gerund (22). (20)

éθθa kámno ‘shower’ ‘I will shower’ Cypriot Greek/English (Goutsos 2001, 203)


ói ðjóti ótan kámnis ‘wash’ ‘no because when you wash’ Cypriot Greek/English (Goutsos 2001, 203)


. . . lálun tus pu na érti i jajá su ená kámni ‘swimming’ ‘I was saying to them, when your grandma comes, she will swim.’ Cypriot Greek/English (Goutsos 2001, 203)

The periphrastic construction is not a pervasive phenomenon in the speech of Goutsos’ informants, nor is code-switching in general (Goutsos 2001, 216). The examples also show the absence of an established convention on how to insert the English verb into the periphrastic construction. Still it is noticeable that the do-construction is recorded on Cyprus and not in mainland Greece. Goutsos comments: ‘Compared to SMG [Standard Modern Greek], CG [Cypriot Greek] draws more on English resources for borrowing and creates “mixed” compound forms with de-lexicalized verbs’ (2001, 204). Although Apostolou Panara shows that the periphrastic construction does occur in mainland Greek, Goutsos’s comment confirms the impression that it is a marginal phenomenon. If it is indeed true that the periphrasis is more common in Cyprus, it is attractive in the light of the present discussion to explain this difference as resulting from the greater impact of English on the language of the former crown colony. Moreover, Goutsos’s examples stem from a really bilingual setting, whereas the loanwords collected by the other authors stem mostly from written sources that do not presuppose a bilingual readership. 4.3

Mainland Greek in the Diaspora

The oldest study on American Greek mentions just one English-origin verb: mouvaro ‘to move’ (Lontos 1925–26). However, this particular form probably goes back to an Italian form muevere, as Hartley (2001) suggests in his paper on loan words in Greek nautical terminology.


louis boumans

A much more elaborated study of American Greek is provided by Seaman. He mentions the same form muváro (1972, 165), but it becomes clear that the periphrastic construction with káno ‘to do, make’ is the productive way to incorporate English verbs in his data: ‘In verbal loans from English, /jíno/ seems to be the auxiliary for passives and /káno/ for actives’ (1972, 166). His examples of the passive construction with jíno ‘to be, become’ are ambiguous, as they might also be analysed as cases of a copula plus predicate, and there are only a handful of examples. On the other hand, Seaman cites 47 examples of the periphrastic construction with káno. In 33 cases the embedded English element is unambiguously a verb; 11 cases may involve either a verb or a noun, e.g. control, welcome. Some examples are reproduced below (see also Muysken 2000, 212).10 (23)

káno, ‘cover up’ ‘I cover up’ kánis, ‘brush’ ta δóndja su ‘You (sg) brush your teeth’ ékane, ‘punch’ ‘s/he punched [a meal ticket]’ American Greek (Seaman 1972, 167–68)

Tamis found the same construction in Australia (1986, cited in Muysken 2000, 212). The construction is also known from Canada (24) and Brussels (25). Compared with Seaman’s study, the information on the latter diaspora communities is rather scarce. It is not clear whether in these cases the periphrastic construction is really the productive strategy for incorporating English or French verbs. The Canadian examples cited in (24), involving computer terminology and sports, remind of the use of the periphrastic construction in Greece described by Apostolou Panara. (24)

káno ski káno save [computer] káno print [computer] káno jogging Montreal Greek/English (Muysken 2000, 213, based on Hatzidaki p.c.)


káno déménager ‘to move house’ Greek/French (Muysken 2000, 213, based on Hatzidaki p.c.)

10 In Table 7.7 on p. 212 of Muysken’s book the words jíno and káno have been reversed.

the periphrastic bilingual verb construction 4.4


Cypriot Greek in London

There are a number of publications (and many unpublished theses) on the language of the Greek Cypriot community in London. Gardner-Chloros (1992) mentions the occurrence of morphologically integrated verbs like tekaro ‘to check’, which is not surprising as these words are found in Cyprus (and Greece) as well. In addition, however, she points out the periphrasis with káno as an innovative construction, occurring with English ‘adjectives, nouns, participles . . .’ (Gardner-Chloros 1992, 127). (26)

kamno use ‘to use’ kamno respect ‘to respect’ kamno developed ‘to develop’ kamno spelling ‘to spell’ Cypriot Greek in London (Gardner-Chloros, 1992, 127)

Zarpetea compares the language use of three generations of London Cypriots (1995). She notices the use of both integrated loan verbs and the periphrastic construction in the first as well as the second-generation immigrants. The periphrastic construction still occurs in the speech of the third generation, young children who use only little Greek: (27)

έκαμα wash τα cups, do you want me to do the plates now mum? ‘I have washed the cups, . . .’ Cypriot Greek/English (Zarpetea 1995, 581)

As a final note on morphologically integrated English verbs in immigrant Greek, I would like to draw the attention to the fact that there is much overlap between the lemmas mentioned in the various sources. This is particularly striking in Zarpetea’s paper. Out of the seven verbs she cites, five also occur in other sources, cf. Table 2. TABLE 2


Zarpetea 1995, 578

other sources

Κανςελλάρω ‘to cancel’

Gardner-Chloros 1992, 126, Papapavlou 1997, 232, Evripidou 2001, 24 Papapavlou 1997, 240, Evripidou 2001, 24 Papapavlou 1997, 240 Apostolou Panara 1991, 50, Papapavlou 1997, 236 Lontos 1925–26, 309, Seaman 1972, 167, Hartley 2001

   ‘to check’ τσιαρτσιάρω ‘to charge’ παρκάρω ‘to park’ μουβάρω

‘to move’

μπουκκάρω ‘to book’ πακκεττάρω ‘to packet’


louis boumans

There is no such overlap in the examples cited for the periphrastic construction with kámno (mainland Greek káno).11 This indicates that the morphologically integrated verbs found in London may well have been coined in the more monolingual setting in Cyprus. As noted above, μουβάρω ‘to move’ is probably not even of English origin. This is a further indication that in the Cypriot diaspora, the periphrastic construction replaces morphological integration as the most productive strategy for the incorporation of new English verbs. The chances are that the same development has taken place in the American Greek community studied by Seaman (1972).

5. Portuguese/English I have been able to trace surprisingly few studies on English loan words in Portuguese, whether European or Brazilian. Compared with Greek, there are also few studies on Portuguese speaking communities in Anglophone countries. The picture that arises from the available data is approximately the same as for Greek. 5.1


The website of the Instituto Superior Politécnico de Viseu in Portugual hosts a short list of English loanwords in Portuguese, compiled by two students and their English teacher (Queiroz, Rodrigues and McKenny 1999, accessed June 2002). The authors are not explicit about their data sources, except that visitors of the web page are invited to contribute observations. As it concerns a Portuguese project, however, I assume that most of their sources are likewise of Portuguese, rather than Brazilian, origin. Their list of English loan words contains four verbs integrated in the Portuguese conjugation class ending in -ar (28), in addition to the

11 Apostolou-Panara, Goutsos, Gardner-Chloros and Zarpatea cite 14 different English verbs in the periphrastic construction. kámno wash is shared between Goutsos and Zarpatea. kámno use is cited in the papers by Gardner-Chloros and Zarpetea. But these two papers must be based on (partly) the same data. I infer this from the fact that the example     very busy ‘I know that you have (are) very busy’ is cited in both (Gardner-Chloros 1992, 128; Zarpatea 1995, 578).

the periphrastic bilingual verb construction


denominal form stressar (from ‘stress’) and the de-adjectival compactar (‘to compress’ from compact, computer terminology). (28)

draftar ‘to draft’ driblar ‘to dribble’ linchar ‘to lynch’ snifar ‘to sniff [drugs]’ European Portuguese/English (Queiroz et al. 1999)

In addition this list contains six deverbal English nouns denoting activities in the domains of sports and information and communication technology: jogging, kickboxing, surf; scan, upgrade and zap. The web site does not indicate how these words are used in Portuguese. Nicholas Hurst (p.c. 18 Apr 2002) cites examples of periphrastic constructions involving the Portuguese verb fazer ‘to do’ (29). The examples show that there is much variation: as in Greek contexts, the English verb is used in either the citation form (infinitive) or the gerund. Moreover, the indefinite article um may or may not be present. (29)


fazer zapping fazer um scanning fazer surf fazer um upgrade European Portuguese/English (p.c. Nicholas Hurst 18 Apr 2002)


Kennedy (1971) and Harmon (1994) quote a number of English verbs occurring in Brazilian dictionaries and other written sources. All are integrated into the so-called first conjugation class ending in -ar: (30)

chutar ‘to shoot [ball game]’ drenar ‘to drain’ ranquear ‘to rank’ treinar ‘to train’ Brazilian Portuguese/English (Kennedy, 1971)

Harmon does not mention the construction with fazer. Kennedy cites just one example of a somewhat enigmatic periphrastic construction: fazer o footing, translated in English as ‘take a stroll.’

306 5.3

louis boumans American Portuguese

There is a classical study of the language of Portuguese immigrants in the US by Pap (1949). He gives many examples of morphologically integrated English verbs. These are added to the most frequent en most productive Portuguese conjugation class in -ar, with the necessary phonological adaptations. In some cases the English verb stem is extended with an -e- such that the infinitive ends in -ear. (31)

raid-ear ‘to ride’ canec-ar ‘to connect’ damp-ar ‘to dump’ s(e)leir-ar ‘to slide’ chinj-ar ‘to change’ raiv-ar ‘to drive’ North American Portuguese/English (Pap 1949, 95–100)

While morphological integration of English verbs is the most common pattern, Pap also lists a large number of instances of a periphrastic construction with fazer. The nominalized English verb stem occupies the direct object position. In contrast with the contemporary European examples cited by Hurst, the foreign verb stem is preceded by the Portuguese masculine definite article o. (32)

fazer o boda ‘to bother’ fazer o spoil ‘to spoil’ fazer o save ‘to save’ fazer o find out ‘to find out’ fazer o give up ‘to give up’ North American Portuguese/English (Pap 1949, 105–106)

6. Discussion The data from North African varieties of Arabic in contact with European languages suggest that diaspora communities are more inclined to use the periphrastic alternative for incorporating foreign verbs. The studies on Greek and Portuguese in contact with English do not unambiguously confirm this. Morphological integration as well as periphrastic constructions are reported for all contact situations, whether in the Greek and Portuguese speaking countries or in the respective immigrant communities in Anglophone countries.

the periphrastic bilingual verb construction


Broadly speaking, however, the literature on Greek and Portuguese does point in the same direction. Integration in the productive conjugation classes, Greek -áro and Portuguese -ar, is predominant in the studies on English loanwords in Greece, Cyprus, Portugal and Brazil. ‘Do-constructions,’ on the other hand, feature prominently in the studies on migrant communities. A caveat is in order: studies on the influence of English as a foreign language are traditionally oriented toward loan words, i.e. words in isolation, in written sources. These studies tend not to identify foreign verbs in the periphrastic construction as verbs at all, which can easily lead to the suggestion that embedded verbs are either absent or consistently integrated into the morphology of the matrix language. For obvious reasons, studies on bilingualism in migrant communities focus on oral communication. Thanks to the recognition of code-switching as an interesting linguistic phenomenon, the latter studies also pay more attention to the higher degrees of linguistic organisation like clauses and utterances. Hence the emphasis on one verb strategy or the other may be to some extent an artefact of the different research traditions. Still, at least the Greek case supports my hypothesis, in particular the studies by Apostolou Panara on Greece and Seaman on the Greek community in the US. Both authors discuss the different types of verb integration, and designate one as the default strategy: morphological integration in Greece and periphrasis in the US. The studies on the London Cypriots suggest that periphrasis is the most productive strategy there as well. With respect to the sociolinguistic factors in verb integration, I conclude that, all other things being equal, morphological integration is the norm in homeland settings where the matrix language is socially dominant, while the periphrastic construction is found in migrant communities where the embedded language is dominant. The migrant communities concerned here are actually undergoing language shift. Stated in more general terms, the periphrastic construction characterises situations of intense contact with the language of the embedded foreign verb. As mentioned above, other factors may override the sociolinguistic ones, as in the case of matrix languages that opt for the same integration strategy in no matter what sociolinguistic setting.

308 6.1

louis boumans Automatization of superimposed language practices

The next question is why speakers more often revert to periphrasis in situations of more intense contact. I can think of the following explanations: Firstly, automatization of superimposed language practices: In general the degree of integration of foreign words is inversely proportional to the speaker’s knowledge of the embedded language. With increasing practise and competence, the linguistic habits characteristic of that foreign language become entrenched in the speaker’s brain, and can more easily be reproduced. This even results in the ‘disintegration’ of formerly integrated foreign words, a process known as denativization (Haugen 1950). In Morocco, for instance, ‘French-Moroccan Arabic bilinguals now prefer to reborrow the same French terms in shapes closer to the French prototype’ (Heath 1986, 114). Denativization has often been described with respect to phonology of loan words from English, a language whose influence is expanding (cf. Khubchandani 1968, Hasselmo 1969, De Reuse 1994, Van Ness 1994). I would like to stress this point, because it is a commonly held belief that loanwords become, on the contrary, more integrated over time (Nortier and Schatz 1988, Heath 1989). The replacement of morpho-phonological integration by periphrasis as the productive strategy for the integration of foreign verbs fits into the denativization process, since the periphrastic construction entails little or no phonological or morphological adaptation to the matrix language. This type of change could be taking place in for instance Greece or Portugal, where exposure to English is still increasing. Denativization refers to a diachronic change in situations where language contact increases over time. However, the underlying mechanism of automatization of superimposed language practices also explains the differences between immigrant and homeland bilingualism. 6.2

Loss of community language practices

Secondly, attrition or loss of community language practices: the morphological integration of foreign verb requires the application of matrix language morphological procedures. These procedures are typically the same as those used for deriving verbs from nouns or adjectives. The loss or erosion of community language morphology, a common effect of decreased use of the language and incipient shift (Andersen 1982,

the periphrastic bilingual verb construction


El Aissati 1997), cuts off morphological integration as a strategy for embedding foreign verbs. 6.3

Collocational complements

Thirdly, triggering by collocational complements of the embedded verbs: Many examples of periphrasis with foreign verbs involve a collocation with a foreign direct object noun or other complement, cf. wash and cups in Greek/English example (27) above and diep, gesprek and voeren in the Arabic/Dutch example below. (33)

ma ne-qder š eh n-dir NEG 1-can NEG er 1-do diep-e gesprekk-en voer-en deep-AGR conversation-PL conduct-INF ‘I can’t er carry out deep conversations.’ MA/Dutch, Jamal, 20, Eindhoven 1991 (Boumans 1998, 245)

The periphrasis leaves the original embedded language collocation intact. Therefore the collocation might trigger this construction. The usage of foreign collocations presupposes a higher level of competence in that language and hence more intense exposure to it.

7. Conclusions In nearly all cases, foreign verbs are combined with matrix language inflection in either of two ways: they are morphologically integrated and inflected with ML affixes (if any), or they are embedded in a periphrastic construction with an inflected ML verb. Some MLs, such as Magribian Arabic, Portuguese and Greek, allow for both solutions. A comparison of Magribian, Portuguese and Greek bilingual communities suggests that the periphrastic integration of foreign verbs is favored in situations of more intense contact with the superimposed language.

8. References Andersen, R. 1982. “Determining the Linguistic Attributes of Language Attrition.” In R. Lambert and B. Freed, eds. The Loss of Language Skills. Rowley, MA: Newbury House, 83–118. Apostolou Panara, A. 1991. “English Loanwords in Modern Greek: an Overview.” Terminologie et Traduction, 1.91, 45–60.


louis boumans

Aquilina, J. 1965 (1987). Teach Yourself Maltese. Valetta: Hodder and Stoughton. Backus, A. 1996. Two in One: Bilingual Speech of Turkish Immigrants in the Netherlands. Tilburg: Tilburg University Press. ——. 2001. Synchronic Register Variation in Monolingual and Bilingual Turkish (project description). Boumans, L. 1998. The Syntax of Code-switching. Analysing Moroccan Arabic/Dutch Conversations. Tilburg: Tilburg University Press. ——. 2000. “Periphrastic Verb Constructions in Moroccan Arabic/Dutch Code-switching.” In A. Fenyvesi and K. Sandor, eds. Language Contact and the Verbal Complex of Dutch and Hungarian. Working Papers of the 1st Bilingual Language Use Theme Meeting. Szeged: University of Szeged, Teacher Training College, 67–84. ——. 2002. “Repetition Phenomena in Insertional Codes-witching.” In A. Rouchdy, ed. Language Contact and Language Conflict Phenomena in Arabic. London: Curzon, 279–316. ——, and Caubet, D. 2000. “Modelling Intrasentential Code-switching: a Comparative study of Algerian Arabic/French in Algeria and Moroccan Arabic/Dutch in the Netherlands.” In J. Owens, ed. Arabic as a Minority Language. Berlin: Mouton de Gruyter, pp. 113–80. Camilleri, A. 1994. “Language contact between Maltese and English: code-switching and cross-linguistic influence.” In D. Caubet and M. Vanhove, eds. Actes des premières journées internationales de dialectologie arabe de Paris. Paris: INALCO, 431–49. Cohen, M. 1912. Le parler arabe des juifs d’Alger. Paris: H. Champion. Contossopoulos, N.G. 1978. L’influence du français sur le grec: emprunts lexicaux et calques phraséologiques. Athènes: s.n. Cornips, L. 1994. “De hardnekkige vooroordelen over de regionale doen+infinitiefconstructie.” Forum der Letteren, 35.4, 282–94. De Reuse, W. 1994. “English Loanwords in the Native Languages of the Chukotka Peninsula.” Anthropological Linguistics, 36.1, 56–68. El Aissati, A. 1997. Language Loss among Native Speakers of Moroccan Arabic in the Netherlands. Tilburg: Tilburg University Press. Evripidou, D. 2001. Lexical Borrowing: A Study of English Loanwords in the Greek Cypriot Dialect. Lancaster: Lancaster University. Gardner-Chloros, P. 1992. “The Sociolinguistics of the Greek Cypriot Community in London.” Plurilinguismes, 4, 112–36. Giesbers, H. 1984. “Doe jij even lief spelen? Notities over het perifrastisch doen.” Mededelingen NCDN, 19, 57–76. Goutsos, D. 2001. “A Discourse-analytic Approach to the Use of English in Cypriot Greek Conversations.” International Journal of Applied Linguistics, 11:2, 194–223. Harmon, R.M. 1994. “Aspectos lingüísticos dos empréstimos em português.” Hispania, 77:3, 463–469. Hartley, A.H. 2001. Loanwords in Modern Nautical Greek. (2002). Hasselmo, N. 1969. “On Diversity in American Swedish.” Svenska Landsmål och Svenskt Folkliv, 92, 53–72. Haugen, E. 1950. “The Analysis of Linguistic Borrowing.” Language, 26:2, 210–31. Heath, J. 1986. “Hasta la Mujerra! and other Instances of Playful Language Mixing in Morocco.” Mediterranean Language Review, 2, 113–6. ——. 1989. From Code-Switching to Borrowing. A Case Study of Moroccan Arabic. London: Kegan Paul International. Herrero Muñoz-Cobo, B. 1996. El árabe marroquí: aproximación sociolingística. Almería: Universidad de Almería. Igla, B. 1991. “On the Treatment of Foreign Verbs in Romani.” In P. Bakker and M. Cortiade, eds. In the Margin of Romani. Gypsy Languages in Contact. Amsterdam: Inst. ATW, Univ. Amsterdam, 50–5.

the periphrastic bilingual verb construction


Kennedy, J.H. 1971. “The Influence of English on the Vocabulary of Brazilian Portuguese.” Hispania 54:2, 327. Khubchandani, L.M. 1968. “The Gender of English Loanwords in Sindhi.” In B. Krishnamurti, ed. Studies in Indian Linguistics. (Professor M.B. Emeneau sastipurti volume). Poona: Deccan College, 180–88. Lontos, S.S. 1925–26. “American Greek.” American Speech, I, 307–10. Malkiel, Y. 1986. “A Spanish Conjugational Model Superimposed on Portuguese.” Mediterranean Language Review 2, 51–66. Mifsud, M. 1995. Loan Verbs in Maltese. A Descriptive and Comparative Grammar. Leiden: E.J. Brill. Muysken, P. 2000. Bilingual Speech. A typology of Code-Mixing. Cambridge: Cambridge University Press. Myers-Scotton, C. 1993. Duelling Languages. Grammatical Structure in Codeswitching. Oxford: Clarendon Press. Nortier, J. and Schatz, H. 1988. “Van éénwoordwisseling naar ontlening, een vergelijkend onderzoek.” Forum der Letteren, 29, 161–78. Nuijtens, E.T.G. 1962. De tweetalige mens—een taalsociologisch onderzoek naar het gebruik van dialect en cultuurtaal in Borne. Assen: Van Gorcum. Pap, L. 1949. Portuguese-American Speech. New York: King’s Crown Press. Papapavlou, A. 1997. “The Influence of English and its Dominance in Cyprus: Reality or unfounded fears?” Journal of Mediterranean Studies, 7:2, 218–49. Queiroz, A., Rodrigues, J. and McKenny, J. 1999. Empréstimos da língua inglesa para a portuguesa. English loan words and phrases in Portuguese. Escola Superior de Educação de Viseu. 2002. Seaman, P.D. 1972. Modern Greek and American English in Contact. The Hague: Mouton. Swanson, D. 1958. “English Loanwords in Modern Greek.” Word, 14, 26–46. Tamis, A. 1986. The State of the Modern Greek Language as Spoken in Victoria. Melbourne, University of Melbourne. Thomason, S.G. and Kaufman, T. 1988. Language Contact, Creolization and Genetic Linguistics. Berkeley: Univ. of California Press. Treffers-Daller, J. 1994. Mixing Two Languages. French-Dutch contact in a comparative perspective. Berlin: Mouton de Gruyter. Van Dijk-Wittop Koning, A.M. 1963. De continuïteit van het Grieks: onderzoek naar de herkomst van de woordvoorraad van het Nieuwgrieks. Zwolle: W.E.J. Tjeenk Willink. Van Ness, S. 1994. “Die Dimensionen lexicalischer Entlehnungen im Pennsylvaniendeutschen von Ohio (USA).” Zeitschrift für Dialektologie und Linguistik, 61.3, 279–97. Van Staden, M. 1999. “Where does Malay end and Tidore begin?” In R.A.C. Dam, ed. Perspectives on the Bird’s Head of Irian Jaya, Indonesia. Proceedings of the conference Leiden, 13–17 October 1997. Amsterdam: Rodopi, 691–716. Wernitz, C.J. 1993. Bedingungen und Voraussetzungen für Sprachwechsel. Eine Untersuchung zum Sprachwechsel bei bilingualen Marokkanern in Frankreich. Frankfurt: Peter Lang. Zarpetea, P. 1995. “Code-switching and Lexical Borrowing (loanwords) in the Speech of Three Generations of Greek Cypriots in London (Harringey).” Studies in Greek Linguistics. Proceedings of the 16th Annual Meeting of the Department of Linguistics. Thessaloniki: Aristotle University, 576–87.


Pour Kees Versteegh Aut cuspis sic vita fluit, dum stare videtur.

1. Introduction1 Les schèmes du verbe augmenté (mazīd) en arabe sont traditionnellement associés à des valeurs sémantiques de base, telles que l’itération, l’intensivité, la causativité et la factitivité, la réflexivité, etc. Si les listes d’exceptions qui accompagnent ces descriptions indiquent que le problème, malgré des avancées significatives, est loin d’être résolu de manière satisfaisante, la question des valeurs sémantiques associables aux trois schèmes simples (mujarrad) ou non-augmentés faula, faila et faala2 semble offrir encore plus de résistance à l’analyse. La description de ces schèmes dans les grammaires arabisantes est, lorsqu’elle existe, bien plus sommaire que pour les verbes augmentés.3 1 Je reprends ici, sous une forme largement revue, le chapitre 3 de mon cours de préparation à l’agrégation d’arabe du Centre national d’enseignement à distance (CNED), 2002 et 2003. Par souci d’explicitation, j’ai présenté quelque peu en détail le cadre conceptuel nécessaire au traitement de cette difficile question, et conservé des définitions ou des indications qui pourront paraître évidentes, mais ne le sont pas nécessairement. Ce travail doit beaucoup au cadre théorique et aux travaux de Jean-Pierre Desclés, ainsi qu’aux séminaires que nous avons animés, et continuons d’animer ensemble, depuis 2001, sur les valeurs associées en contexte aux formes verbales en arabe et en français. Je remercie également les membres de son équipe de recherche, notamment Brahim Djioua, pour de nombreuses et fécondes discussions. 2 La convention qui prévaut dans la tradition linguistique arabe désigne ce que nous appelons les consonnes radicales 1, 2 et 3 (de la racine sémitique) respectivement par f,  et l. 3 Ainsi pour les grammaires en langues occidentales : Neyreneuf et Al-Hakkak (1996, 35–42) signalent “des nuances de sens . . . pour chaque forme dérivée”, mais non pour les verbes simples (id., 28–29). De même, les descriptions consacrées par Blachère et


joseph dichy

Il suffit d’observer qu’ils demeurent désignés dans la tradition arabisante occidentale comme la ‘Forme I’. Cette désignation relève d’une idée générale qui répond, à défaut de décrire les données linguistiques, à une cohérence interne. La ‘Forme I’ est appelée ‘verbe primitif ’ par A. Sylvestre de Sacy, par référence au ‘verbe nu’ mujarrad des grammaires arabes (1831, I : 123). A cette forme primitive, Sylvestre de Sacy associe les grandes divisions sémantiques générales des verbes (id. 121–122) sur lesquelles il fondera ensuite son analyse des formes dérivées (id. 129–143). Mais il n’identifie pas de valeurs sémantiques associées en propre aux schèmes faula, faila et faala, dans lesquelles il ne voit qu’une ‘forme primitive’ ou ‘nue’ (mujarrad) à partir de laquelle les formes dérivées ou ‘augmentées’ (mazīda) sont construites. Si la tradition arabisante a pu ainsi rapporter ces trois schèmes à une seule ‘forme’, c’est sans doute aussi pour une autre raison : ces derniers apparaissent en effet comme instables, tant du point de vue formel, où l’on observe des modifications de la voyelle de la 2e radicale (section 2 ci-dessous), que du point de vue sémantique, où les notions très générales de transitivité, de ‘verbes d’action’ ou ‘d’état’ par lesquelles on a essayé de les caractériser rencontrent un nombre élevé de contre-exemples. C’est ainsi que les trois pages d’indications essentiellement formelles consacrées à cette question par la grammaire d’aš-Šartūnī sont précédées de cet avertissement : “On ne peut établir [la forme des verbes relevant de] ces schèmes (awzān) qu’en ayant recours à des ouvrages lexicographiques (1910, 11). Faut-il donc renoncer ? Il y a de solides raisons de penser que non. Plusieurs essais de mise en ordre et d’analyse ont été tentés. La synthèse la plus récente de ces tentatives est, à ma connaissance, celle de Pierre Larcher (2003, chap. II), qui comporte en outre des propositions originales. Signalons également les analyses incontournables d’André Roman (1983, II : 886–900). D’autres auteurs, au premier rang desquels Marcel Cohen, Paul Joüon, Jean Cantineau et Henri Fleisch ont été utilisés. En outre, un nombre considérable d’observations et d’analyses a été effectué dans les textes des sciences linguistiques arabes médiévales. Je ferai largement usage de ces travaux, sans toutefois les discuter en détail, pour Godefroy-Demombynes (1952, 49–70) aux “formes dérivées” contrastent avec l’absence d’analyse de la “Forme I”. Belot, (1922, 15), Badawi, Carter et Gully (2004, 60) ne signalent pour la forme “primitive” ou “de base” (al-fil al-mujarrad), de manière très prudente, que la transitivité ou son absence en fonction de la voyelle médiane de faala, faila et faula. On trouve en revanche des propositions de description du sens dans Caspari-Uricoecha (1881, 32–33), Wright (1896–98 I : 30), Brockelmann (1948, 35) ou Boormans (1967, 10), ainsi, naturellement, que dans la tradition linguistique arabe (voir la synthèse de Nūr ad-Dīn 2002, 177–186).

faula, faila, faala


éviter d’allonger indûment cette étude. A l’entêtant problème posé par la dispersion sémantique des verbes de schème simple en arabe j’ai cherché à apporter une solution fondée sur une définition aussi rigoureuse que possible des termes utilisés, ainsi que sur une conception, me semble-t-il, renouvelée de la relation entre sens et forme dans les schèmes (au sens que prend ce terme dans le domaine sémitique) en arabe (Dichy 2003). Les propositions qui suivent s’appuient en outre sur un dépouillement systématique des données lexicales dans le roman de Tawfīq al-H akīm, Yawmiyyāt nāib fī l-aryāf (abrégé en Yawm.) d’une part, et dans plusieurs dictionnaires arabes de l’autre (notamment Hans Wehr-Cowan 1979 et al-Mujam al-Wāsīt). J’ai fait largement usage de la base de données lexicale DIINAR.1 (“Dictionnaire informatisé de l’arabe, version 1” – Dichy, Braham, Ghazeli et Hassoun 2002 ; Dichy et Hassoun 2005 ; http ://, du concordancier informatisé construit par R. Abbès (2004), ainsi que de la liste indexée de 20.000 verbes réalisée pour les besoins du Bescherelle des Verbes arabes (Ammar et Dichy 1999). Les références à ces dernières sources ne sont pas données, à la différence des textes arabes médiévaux, des travaux arabisants de l’époque moderne ou, au besoin, des dictionnaires ; Yawm est indiqué avec le numéro de la page suivi de la ligne lorsque le sens doit être interprété en fonction du contexte.

2. Les schèmes du verbe non augmenté, aspects formels 2.1

L’alternance vocalique de la voyelle de la 2e radicale

Les trois schèmes de base du verbe simple ont été identifiés dans les sciences médiévales arabes du langage à partir de la voyelle de la deuxième radicale ou voyelle médiane du verbe au paradigme du mādī (ou ‘suffixé’),4 respectivement u, i et a. Les trois familles morphologiques correspondantes sont énumérées sous forme d’exemples (amtila)

4 Les termes d’accompli et inaccompli, d’achevé et d’inachevé ou de perfectif et imperfectif, pourtant reçus, comportent un problème, celui de désigner la forme morphologique par un terme dénotant l’une des valeurs aspectuo-temporelles que cette forme peut prendre en contexte. D’où le choix de désigner ces paradigmes verbaux par leur nom arabe ou par le trait morphologique qui les caractérise. Ainsi, le mādī (traditionnellement ‘accompli’), dans lequel le morphème de personne précède la base du verbe, correspond ici au ‘suffixé’ ; le mudāri (traditionnellement ‘inaccompli’), dans lequel ce morphème précède la base, est désigné du terme de ‘préfixé’ (cf. Moscati, éd. 1964, 131–2 : “préfix-conjugation” vs “suffix-conjugation”).


joseph dichy

par Sībawayhi (m. vers 180/796) dès l’introduction du Kitāb (I : 12), et reprises au cours de ce traité au moyen des conventions faala, faila et faula (désignées comme des ‘formes construites’, abniya) avec des indications morphologiques, syntaxiques et sémantiques (Kitāb IV : 5–67, notamment 38). Je laisse de côté ici la question, qui a été posée, de savoir si la forme à sujet non exprimé (dite ‘passive’) fuila ne constituerait pas un schème à part entière.5 Selon la reconstruction du proto-système de la langue arabe, telle qu’elle a été proposée en diachronie (et en tout cas dans l’organisation qui sous-tend le système de la langue), la forme simple du verbe comporte trois schèmes, représentés dans le tableau ci-dessous (voir notamment Cantineau 1950, 77 ; Fleisch 1957 et 1968, 115–119 ; Roman 1983, II : 894) : Mād. ī (suffixé)

Mudāri' (préfixé)

• Schème simple 1


yaf ulu

• Schème simple 2


yaf alu

• Schème simple 3


yaf ilu



Selon Moscati, éd. (1964 : 122), qui se situe dans la perspective des études sémitiques comparées, “l’antiquité de ce schéma vocalique à trois termes en arabe est confirmé par quelques-unes des manifestations les plus anciennes du groupe sémitique du Nord-Ouest, c’est-à-dire, par l’amorite, l’ougaritique, et les gloses de Tell Amarna . . .”6 Dans cette reconstruction, comme le souligne Roman (loc. cit.), les schèmes sont toujours identifia-

5 Voir sur ce point Cantineau (1950) ; Roman (1983 II : 897–900 ; 2005, 33), qui voit dans la première voyelle la marque de la diathèse (subjective ou objective) ; Fleisch (1957 ; 1968, 246 ; 1979) ; Larcher (1996 ; 2003, 26–28). Nūr ad-Dīn (2002, 185–186) signale des traités tardifs, dans lesquels la forme fuila est considérée comme un quatrième ‘principe’ (’asl), i.e., un quatrième schème du verbe simple, et non une forme de la conjugaison (comme dans , m. en 643/1245, Šarh al-Mulūkī : 30). 6 “The antiquity of this threefold vocalic scheme in Arabic is confirmed by some of the oldest manifestations in North-West Semitic, i.e. Amorite, Ugaritic, and the Tell Amarna glosses. In the prefix-conjugation the variation in the second vowell is at least partly paralleled : u or i corresponding to a, and a to i, while u generally remains” (Moscati, éd. 1964, 122).

faula, faila, faala


bles en fonction de la voyelle de la deuxième consonne radicale (en gras dans la transcription), et ce, dans chacun des deux grands paradigmes de la conjugaison. Dans l’état de langue arabe tel qu’il nous est parvenu, dès les époques classiques, les trois schèmes simples ci-dessus ne sont plus directement identifiables, comme le rappelle le tableau suivant :

• Schème simple 1 • Schème simple 2

Mādī (suffixé)

Mudāri' (préfixé)


yaf ulu


yaf ilu

Sous-catégorie formelle

yaf ilu

• Schème simple 3 Sous-catégorie formelle 1 Sous-catégorie formelle 2

yaf alu


yaf ulu yaf alu


Cette situation continue de permettre l’identification au mādī (suffixé) de trois schèmes distingués au moyen de la voyelle qui affecte la 2e radicale, mais non au mudāri (préfixé). Lorsque l’on traite d’aspects formels, comme dans la conjugaison ou dans un dictionnaire, il est donc nécessaire de caractériser le verbe par l’alternance vocalique de la voyelle médiane aux deux paradigmes du suffixé et du préfixé, exemple : daraba/yadribu ‘frapper’, d’alternance vocalique a/i. (J’écrirai ici, selon une convention reçue : daraba i.) 2.2 Les zones de stabilité ou d’instabilité formelles de l’alternance vocalique Comme le montre la figure 2, le schème simple 1 (alternance u/u) est formellement stable. Dans les schèmes faila et faala en revanche, on trouve des sous-catégories formelles dans lesquelles la voyelle de la 2e radicale varie. Ces variations sont souvent corrélées à la présence, au sein de la racine, d’une ‘cause de transformation’ (illa) :7 deuxième et


Sur le statut épistémologique de cette notion dans la tradition linguistique arabe


joseph dichy

troisième radicales identiques, consonne radicale /w/ ou /y/, ou encore radicale // en première, deuxième ou troisième position. Les relations entre ces ‘causes de transformation’ et les variations observées dans les schèmes du verbe simple ont été, sinon expliquées, du moins largement inventoriées dans les sciences linguistiques arabes médiévales (voir résumé dans aš-Šartūnī, 1910 : 10–14 ; Qabbāwa, 1998 : 85–94). 2.2.1

La sous-catégorie formelle faila/yaf ilu du schème simple 2

La sous-catégorie formelle du schème simple 2 ( faila/yaf ilu) est corrélée dans la plupart des cas à une racine de première radicale w ou y. Elle ne concerne qu’un petit nombre de verbes. Trois d’entre eux, repris de Sībawayhi (Kitāb IV : 38–39) et régulièrement cités dans les traités linguistiques arabes médiévaux ou par les grammaires, relèvent de racines sans ‘cause de transformation’ : h asiba i ou a ‘croire’, ‘estimer’ ; baisa i ou a ‘être misérable’ ; naima i ou a ‘être doux, fin’. Ils peuvent être considérés comme négligeables, car ils supportent également une réalisation a de la voyelle médiane du mudāri ‘préfixé’.8 2.2.2

La sous-catégorie formelle faala/yaf alu du schème simple 3

Une seule modification de la voyelle médiane est suffisamment régulière. Elle est largement signalée dans les ouvrages arabes médiévaux (cf. par exemple, al-Zajjājī, Jumal, 396–397). Il s’agit de la ‘sous-catégorie formelle 2’ du schème simple 3 ci-dessus ( faala/yaf alu) : elle se produit – pour simplifier – si la 2e ou la 3e radicale du verbe est l’une des six ‘consonnes d’arrière’ (h urūf al-h alq), i.e. appartient à l’ensemble {’, h, h,  , ġ, x}. Cette règle s’applique selon une gradation, de manière hautement fréquente pour les deux consonnes les plus antérieures (les consonnes glottales  et h), et nettement moins fréquente pour les deux

médiévale, et notamment chez Ibn Jinnī (m. en 393/1002), voir Guillaume 1984. La liste et les combinaisons des ‘causes de transformation’ (il peut y en voir plus d’une dans une racine donnée) ont été présentés sous une forme modélisée dans Dichy (1993) et Ammar et Dichy (1999, 31–34). 8 Voir par exemple Ibn Xālawayh, m. en 370/981, Laysa fī kalām al-arab “Point ne se trouve dans le langage des Arabes . . .” (sous-entendu ‘hormis’, ‘à l’exception de’), chap. 10 ; az-Zajjājī, m. en 337/949, Jumal : 398 (plus cohérent que le précédent). Sur le statut de la “classe h asiba/yah sibu”, notamment chez Ibn Jinnī, m. en 392/1002, cf. Guillaume (1984, 427–433).

faula, faila, faala


consonnes les moins antérieures (ġ et x) (cf. Ibn Yaīš, Šarh al-Mufassa l VII : 153–154, repris in Ammar et Dichy 1999 : 24). 2.2.3

La sous-catégorie formelle faala/yaf ulu du schème simple 3

La sous-catégorie formelle 1 du schème simple 3 ( faala/yaf ulu) ne paraît pas, quant à elle, pourvoir être l’objet de corrélations systématiques entre la voyelle de la 2e radicale et des traits phonétiques affectant les consonnes du verbe. On trouve ainsi un nombre non négligeable de verbes dont le préfixé (mudāri) admet les deux voyelles médianes i et u, ainsi : azafa/yazifu et yazufu (an) ‘se détourner (de)’ ; qafala/yaqfilu et yaqfulu ‘revenir (de voyage)’, ‘rentrer’, etc. La présence, dans de tels exemples, des deux réalisations met en défaut toute corrélation entre les traits phonétiques des consonnes radicales et la voyelle médiane dans la langue arabe telle qu’elle nous est parvenue, avec ses normes stabilisées par les travaux des lexicographes et l’autorité des grammairiens.

3. Aspects sémantiques, concepts de base Considérons maintenant les valeurs sémantiques associables aux schèmes simples. Les deux notions qui ont été traditionnellement utilisées constituent un cadre conceptuel dont on verra qu’il ne peut suffire. Il s’agit du trait syntaxique de transitivité (§ 3.1.1) et de la distinction entre verbes d’action et verbes d’état (§ 3.1.2). Ces notions devront être reprises dans un cadre conceptuel élargi (sections 3.2 à 3.4). Les descriptions proposées ci-dessous s’appuieront sur une série d’exemples, choisis pour illustrer la plupart des types de valeurs sémantiques qui caractérisent ces schèmes. 3.1 Les deux principales notions traditionnellement utilisées pour décrire le sens des schèmes simples 3.1.1 Le trait syntaxique de transitivité/intransitivité La première distinction que l’on rencontre dans les grammaires arabisantes comme dans les sciences médiévales arabes du langage est de nature syntaxique. En fonction de la voyelle a, u ou i affectant la deuxième consonne radicale, ces verbes sont dits :


joseph dichy

– pour le schème faala, transitifs, bien que comportant un nombre considérable d’intransitifs, – transitifs et intransitifs pour le schème faila, et – toujours intransitifs pour les verbes en faula.9 Les traits syntaxiques de transitivité ou d’intransitivité sont cruciaux dans la théorie de la rection qui prévaut dans la tradition linguistique arabe médiévale. Leur incidence sur le sens des verbes n’est toutefois pas directe : elle en est plutôt une conséquence qu’une cause. Une autre question que la théorie de la rection laisse dans l’ombre est celle de la transitivité ‘indirecte’, dans laquelle le complément du verbe est un syntagme prépositionnel. La transitivité, envisagée sous ses deux formes, directe et indirecte sera naturellement incluse dans les valeurs sémantiques associées aux verbes des schèmes simples, particulièrement en ce qui concerne le schème faala. 3.1.2

La partition traditionnelle entre ‘verbes d’action’ et ‘verbes d’état’

Les arabisants et les sémitisants occidentaux ont généralement fondé leurs analyses des verbes des schèmes simples sur une partition générale en ‘verbes d’action’, ou ‘actifs’, et verbes d’état ou ‘statifs’, les seconds étant nécessairement intransitifs, et les premiers, transitifs ou intransitifs selon les cas. Dans son principe, le schème faala aurait ainsi correspondu à des ‘verbes d’action’ (transitifs ou intransitifs), le schème faula à des ‘verbes statifs’ toujours intransitifs exprimant un “état permanent ou une qualité inhérente naturelle”, et le schème faila à des verbes “généralement intransitifs” indiquant “soit un état ou une condition temporaire, soit une qualité purement accidentelle d’une personne ou d’une chose” (Wright 1896–98 I : § 36–38). Dans cette description, le schème faila pose un problème, qui a été traité en recourant à la notion de verbe moyen d’E. Benveniste (section 3.4 ci-dessous). Si la notion de verbe d’état, une fois redéfinie, résiste à l’analyse, celle de ‘verbe d’action’ ne laisse pas, dans l’usage qui en a été fait, d’apparaître comme indifférenciée. Que faire, par exemple de verbes comme zanna u ‘croire’, saqata u ‘tomber’, halaka i ‘périr’, wajaba i ‘devoir’, šamala u ‘englober’ ou raā a ‘voir’, qui relèvent tous du schème faala, réputé cor9 Cf. Sībawayhi (Kitāb IV : 38), ainsi par exemple que az-Zamaxšarī (Mufassa l, 277– 279). L’intransitivité des ‘verbes d’état’ est rappelée notamment par Wright (1896–98, vol. I, § 36–38) ou Fleisch (1957 ; 1979). Pour l’hébreu, voir par ex. Gesenius-KautzschCowley (1910, § 41–43) ou Joüon (1923, § 40–41).

faula, faila, faala


respondre aux ‘verbes d’action’ ? Et que dire des paires akala u ‘manger’, de schème faala vs. šariba a ‘boire’, de schème faila, ou raā a ‘voir’, vs. samia a ‘entendre’ ? La réponse consistant à poser une catégorie intermédiaire ou ‘moyenne’ (Marcel Cohen 1929 ; Joüon 1930 ; Fleisch 1957), qui concerne au premier chef le schème faila tout en incluant des verbes en faala, ne résout que très partiellement le problème. Les processus de glissement sémantique entraînant le passage d’un ‘verbe d’action’ à un ‘verbe d’état’, s’ils ont donné lieu à quelques fines observations,10 n’apportent pas non plus de solution de portée générale. Mais le principal problème, comme la section 4.3 l’illustrera, est que l’apparente transparence de la notion de ‘verbe d’action’ aura masqué la complexité des données et la forte dispersion sémantique du schème faala. On verra dans la suite de ce travail qu’il n’est de fait pas nécessaire de recourir aux ‘verbes d’action’ pour décrire les verbes en faala, malgré l’intérêt que présente par ailleurs cette notion. Passons maintenant à la définition des propriétés fondamentales nécessaires à l’analyse des schèmes simples : l’agentivité (section 3.2), les notions de procès, d’événement, de processus et d’état (sect. 3.3), celle de verbe moyen ou de diathèse interne (sect. 3.4). 3.2

La propriété d’agentivité

L’agentivité se définit comme la participation de l’agent au procès dénoté par le verbe, le critère de cette ‘participation’ étant le contrôle (Lyons 1978/90, 3.4 ; Desclés 1994) de l’agent sur ce procès. Il y a en arabe trois degrés d’agentivité :11 l’agentivité pleine dans laquelle l’agent contrôle effectivement le procès ; partielle, dans laquelle il n’exerce qu’un contrôle imparfait, et la non agentivité, dans laquelle il ne jouit d’aucun contrôle (l’agent existe, mais il ne coïncide nullement avec le sujet grammatical). Ces trois degrés s’opposent eux-mêmes à l’agentivité neutralisée, qui correspond à des procès dans lesquels il n’y a aucun agent, ce qui se

10 Cf. par exemple pour l’hébreu, Joüon 1923, § 41, pour l’arabe, Larcher 2003, 27–28. 11 Cf. Roman (1990, 42–43 ; 1999/2005, B-2.3.2). La définition ci-dessus de l’agentivité est en partie différente. Cf. également l’analyse de Fleisch (1968, 116–117), qui pose une bipartition du verbe arabe : “1° le verbe à agent (le sujet étant considéré en tant qu’agent) ; le verbe de qualité (le sujet étant simplement le qualifié)”, et distingue, dans la première catégorie “l’agent pur et simple” de “l’agent intéressé”.


joseph dichy

présente typiquement dans les verbes à valeur d’état caractéristique (§ 4.1.1 ci-dessous). Exemples (verbes de schème simple) :12 (a) verbes d’agentivité neutralisée (valeur sémantique ‘être x’) ou nonagentifs (valeur : ‘devenir x’) : – schème faula : karuma u, ‘être, devenir karīm’ ‘noble-et-généreux’ ; šarufa u ‘devenir’, puis ‘être šarīf , i.e. ‘noble, de premier rang social’ ; saġura u ‘devenir, être saġīr’, ‘petit’ ou : ‘être très jeune’ ; latufa u ‘être, devenir latīf’, ‘subtil’, ‘fin, gracieux’ ; – schème faila : baliha a ‘devenir ou être ablah’, ‘faible d’esprit’, ‘stupide’ ; salia a ‘devenir ou être chauve (asla) ; kariša a ‘être ou devenir ridé, craquelé’ (notamment : peau) ; – schème faala : xadaba i ‘devenir, être vert (plante, herbage, lieu . . .)’ ; xaffa i ‘devenir, être xafīf’, ‘léger’ (antonyme de schème faula : taqula u ‘devenir, être taqīl’, ‘lourd’) ; (b) verbes non-agentifs (le sujet grammatical n’est pas l’agent du procès dont il est le siège) : – schème faila : dasiqa a ‘se remplir (à ras bords)’, ‘déborder’ (bassin . . .) ; marida a ‘tomber malade’ ; qarima a ‘avoir faim de viande’ (par effet de privation) ; – schème faala : sakana u, ‘se calmer’ (mer, vent, bruit . . .) ; waqaa a ‘tomber’ et également : ‘se trouver’ (ville, montagne, lieu . . .) ; halaka i ‘périr’ ; (c) verbes partiellement agentifs : – schème faila : ġadiba a ‘être, se mettre en colère’ ; h azina a ‘être’ ou ‘devenir triste’, ‘éprouver de la tristesse’ ; wamiqa i ‘chérir’ ; ašiqa a ‘être’ ou ‘tomber passionnément amoureux’ ; samia a ‘entendre’ (sens de l’ouïe) ; – schème faala : raā a ‘voir’ (vue) ; šamma u ‘sentir’ (odorat) ; (d) verbes pleinement agentifs : – schème faila : šariba a ‘boire’ ; rakiba a ‘chevaucher, monter (pour se déplacer)’ ; h amida a ‘louer’ (louange),


Certains des exemples donnés dans ce travail existent aussi avec des formes correspondant à d’autres schèmes du verbe simple, quelquefois avec le même sens. Ces autres formes ne sont pas reprises ici, malgré l’intérêt qu’eût représenté la discussion.

faula, faila, faala


– schème faala : daraba i ‘frapper’ ; jalasa i ‘s’asseoir’ ; āda u ‘revenir’ ; dahaba a ‘partir’, ‘aller’ ; tabaxa u (ou a) ‘faire cuire’ ou ‘bouillir’, ‘cuisiner’ ; talaa a ‘monter’ ; akala u ‘manger’ ; qāla u ‘dire’. La sous-catégorie (a) inclut des verbes des trois schèmes et les sous-catégories (b) à (d), des verbes des schèmes faila et faala. Ces exemples, qui anticipent par leur classement sur la suite de l’exposé, montrent que la propriété d’agentivité, malgré l’importance qui est la sienne dans la description sémantique des schèmes simples, ne peut, à elle seule, constituer le principe qui en permettrait le classement. 3.2.1

Agent, patient, sujet grammatical

Il est essentiel, pour bien comprendre la notion d’agentivité, de distinguer les termes d’agent (et corollairement, de patient) du sujet grammatical du verbe. Les formes verbales de l’arabe incluent en effet toujours leur pronom sujet, qui correspond aux morphèmes de personne suffixés au mādī et préfixés au mudāri, et coïncide avec le sujet grammatical du verbe. Ainsi : jāat l-amīra correspond mot-à-mot à : ‘elle est venue la princesse’ (Ammar et Dichy 1999, 12–14). Cette structure est occultée par l’emploi indifférencié, dans la tradition linguistique arabe, de fāil pour l’agent comme pour le sujet grammatical. L’agent ( fāil) est ce qui effectue le procès décrit par le verbe. Dans les verbes décrivant une action, l’agent est l’auteur de celle-ci. Dans les verbes de perception ou de sentiment, par exemple, il est le siège de celui-ci. Dans les verbes décrivant un état, le procès, comme on vient de le voir, ne réfère pas à un agent : le sujet grammatical inclus dans la forme verbale est le siège de cet état, qu’il s’agisse d’une personne comme dans raufa u ‘être, devenir ou se montrer doux, compatissant’ ou d’une chose, comme dans daula u, ‘devenir/être étroit ou petit (quantité)’. Dans les verbes d’agentivité partielle ou entière, l’agent coïncide avec le sujet grammatical. Ce dernier, à la voix dite ‘passive’, coïncide avec le patient (qui ‘subit’ le procès décrit par le verbe), exemple : suiltu ‘j’ai été questionné’. Dans les formes non-agentives, le sujet grammatical ‘subit’ également le procès ou en est le siège, exemples : – saqatat alā l-lardī ‘elle est tombée par terre’ (qu’il s’agisse d’une personne ou d’une chose) ; – darasa r-rasmu ‘la trace (du campement) s’est effacée’ : l’agent est ici le vent du désert. (Ce thème est classique dans la poésie arabe ancienne ;


joseph dichy

cf. la fin du 6e vers de la Muallaqa ‘grande ode’ du poète emblématique de l’anté-islam Imru al-Qays : Wa-hal alā rasmin dārisin min muawwalī, ‘Sur traces effacées, qui s’en irait pleurant ?’) 3.3 Les propriétés d’événement, de processus et d’état (à partir de Desclés 1994) Le terme général de procès (auquel correspond en anglais celui de situation)13 traduit le fait que le sens du verbe est inscrit dans une durée, c’est-à-dire, dans un intervalle de temps. Cette durée peut être perçue de trois manières, correspondant aux notions aspectuelles d’état, d’événement ou de processus. La définition de ces termes premiers est empruntée directement à J.-P. Desclés.14 À première vue, ces définitions pourraient apparaître comme détournées : les trois termes aspectuels d’événement, de processus ou d’état, qui décrivent des phénomènes affectant des verbes en en contexte sont réutilisés, dans ce travail comme dans ceux dont la publication suivra, pour désigner dans la sémantique des schèmes de l’arabe des primitives incluses dans le sens grammatical associé aux schèmes simples et à certains schèmes augmentés. Mais à y regarder de plus près, s’agissant de valeurs sémantiques primitives, il n’y a rien d’étonnant à ce qu’elles puissent également être observées dans les propriétés lexicales de ‘familles morphologiques’, telles que les schèmes simples et les schèmes augmentés. La mise en lumière de ces observations constitue une première contribution théorique du travail présenté ici à la définition de ces valeurs primitives. Leur confrontation aux données de l’arabe m’a en outre conduit à en proposer des extensions, particulièrement en ce qui concerne les sous-catégories de l’état (§ 3.3.3).


Le terme de procès ne se traduit pas directement en anglais. Comrie (1985, 5), Lyons (1978/1990, 3.4) ont recours au terme de ‘situation’ pour recouvrir les états d’une part, les événements et les processus de l’autre. 14 Mais voir aussi Comrie (1976, 1985), Lyons (1978/1990), ou, différemment, la synthèse de Maingueneau (1994). La théorie “des invariants cognitifs du langage et des primitives sémantico-cognitives” de J.-P. Desclés est sous-jacente aux analyses présentées ici, notamment : l’idée que “les primitives sont des opérateurs abstraits déterminés par des propriétés formelles”, l’hypothèse selon laquelle elles sont “ancrées sur la perception et l’action”, ainsi qu’un nombre important de points qui découlent de ces conceptions (Desclés 1990 ; 1994 ; Desclés et Guentchéva à paraître). Le lecteur familier de ces travaux notera enfin que le terme de ‘schème’ est employé ici à la manière des travaux arabisants et sémitisants. On distinguera donc cet usage de celui des “schèmes sémantico-cognitifs” (Desclés, 1990 ; Abraham, 1995 ; Desclés et al., 1998).

faula, faila, faala 3.3.1



“L’aspect ‘événement’ [. . .] est l’expression d’une occurrence qui apparaît sur un arrière-fond stable [. . . et] établit une coupure entre l’avant événementiel (‘pas encore’) et l’après événementiel (‘ne plus’)”. La durée de l’événement est donc bornée à gauche par un avant et à droite par un après. Cet après peut coïncider avec un état engendré par l’événement, ou état résultatif (Desclés 1994, 73–75).15 Les exemples ci-dessous (verbe concerné en gras) illustrent quelques propriétés fondamentales de l’aspect événement : – mā kidtu hādā s-sabāh a aršifu finjāna l-qahwa alā maktabī h attā waradat išāra tilifuniyya . . . (Yawm., 123), ‘À peine, ce matin-là, avais-je siroté le café [posé] sur mon bureau, qu’un télégramme téléphoné survint . . .’ (mot-à-mot, ‘lorsque apparut’). La première proposition de cette phrase (mā kidtu . . .aršifu ‘à peine . . . avais-je siroté’) constitue l’arrière-plan sur lequel se détache l’événement h attā waradat ‘lorsque apparut’, événement qui détermine au sein de la situation décrite, un avant et un après. – wasala dayfunā mundu sāa ‘notre invité est arrivé depuis une heure’ : de l’événement ‘est arrivé’ résulte un état, dont la durée englobe le moment de l’énonciation. Le locuteur rapporte l’événement de l’arrivée de l’invité, et exprime par les mêmes mots l’état dans lequel son interlocuteur et lui se trouvent, et qui correspond à l’après du procès : leur invité ‘est là’ depuis une heure. On dira, en simplifiant, que l’événement engendre l’état qui résulte de lui. – ġariqa r-rajulu ‘l’homme s’est noyé’ (ou ‘se noya’) : l’événement décrit par le procès engendre l’état résultatif ‘noyé’, exprimé en arabe par l’adjectif ġarīq, associé à ce verbe. 3.3.2 ‘Processus’ “L’aspect ‘processus’ exprime un changement saisi dans son évolution interne. Tout processus exprime nécessairement un changement initial (. . .) qui indique le début du processus”. Le processus peut être décrit 15 Je fais dans ce travail un usage générique (ou hypéronymique) du terme d’état résultatif. Ce dernier inclut ici “l’état résultant”, qui est par ailleurs très utile pour l’analyse des valeurs aspectuo-temporelles en contexte, par exemple pour la description de “l’état-présent, concomitant à l’acte d’énonciation” (Desclés, 1994, 73). Ce choix, limité au présent travail, est dû au fait que les valeurs sémantiques associées aux familles morphologiques constituées par les schèmes ne sont pas sensibles à cette distinction contextuelle.


joseph dichy

comme une succession d’états. Il peut, selon le cas, s’orienter ou non vers un terme explicite, qui correspond alors au dernier état du processus ou état résultatif (Desclés 1994, 75–80). Exemples : – Dans la phrase ci-dessus, le syntagme mā kidtu aršifu finjāna l-qahwa ‘à peine avais-je siroté le café’ exprime un processus dont le terme est spécifié : à un moment donné, qui correspond à l’état final du processus, le breuvage est entièrement bu. La phrase dit mot-à-mot : ‘. . . siroté la tasse de café’ ; la valeur sémantique associée à cette expression dans le lexique de l’arabe implique que le contenu de la tasse a été bu en entier.16 – Dans ġarabati š-šamsu ‘le soleil s’est couché’ ou hal turīdu an tujanninanī ? ‘tu veux me rendre fou (folle) ?’ c’est également dans la valeur lexicale du verbe que le terme du processus (le coucher du soleil, la ‘folie’) est inscrit. – Le sens de amiya a ‘perdre la vue’ comporte un terme, qui correspond dans le lexique à l’adjectif amā ‘aveugle’. Ce verbe décrit, selon le contexte, soit un événement, soit un processus. 3.3.3 ‘État’, types d’états et catégories dérivées ou corrélées “L’aspect ‘état’ exprime la stabilité de la situation référentielle représentée, c’est-à-dire qu’il réfère à une absence de mouvement et de changement”. L’état décrit un procès qui n’est borné ni à gauche (en son commencement) ni à droite (le terme n’est pas spécifié), et ne fait référence à aucun agent (Desclés 1994, 71–73).17 Il y a plusieurs types d’états. Deux d’entre eux, ‘l’état caractéristique’ (§ 3.3.3.a) et ‘l’état résultatif ’ ou ‘état acquis’ (§ 3.3.3.b) intéressent directement les schèmes du verbe simple. Le second, qui est produit par un changement d’état, correspond à un glissement sémantique étroitement corrélé à l’usage des verbes d’état en discours. L’analyse ci-dessous débouche sur l’adjonction, à la valeur sémantique de l’état caractéristique’, de celle de l’acquisition d’état’. Cette dernière 16

Ainsi, le dictionnaire glose l’expression rašifa l-ināa, mot-à-mot ‘siroter un récipient’ comme “en aspirer le contenu jusqu’au bout” (al-Mujam al-Wasīt : racine /r-š-f/) ; de même, šariba l-ināa ‘boire un récipient’ correspond à šariba kulla mā fīhi ‘en boire tout le contenu’ (Ilyās et Nāsīf 1995, 164). 17 Comparer avec la définition de Maingueneau (1994, 64) : “Les états . . . n’ont ni début, ni fin, ni milieu, ils ne supposent ni agent ni changement (ex. Luc est paresseux).” À la différence de Comrie (1976, 1985), Lyons (1978/1990), Desclés (1994) et du présent travail, cet auteur n’oppose pas état, événement et processus.

faula, faila, faala


notion relève de celles, plus générales, d’état résultatif et de changement d’état. L’acquisition d’état constitue en effet la principale valeur sémantique corrélée à l’état caractéristique (§ 3.3.4). 3.3.3.a La notion d’état caractéristique Les données de l’arabe m’ont conduit à proposer une sous-catégorie de la notion générale d’état, celle d’état caractéristique (Dichy 2002/2003). Ce dernier est exprimé par un verbe associé dans le lexique à un adjectif, traditionnellement appelé adjectif ‘assimilé’ ou ‘analogue’ [au participe] (sifa mušabbaha – ci-après § 4.1.2). L’état caractéristique – par opposition à l’état non caractéristique –, décrit une propriété ou un attribut de l’objet, posé, selon le point de vue, comme dominant, intrinsèque, inhérent ou inscrit dans la nature de celui-ci. Cet état peut, comme on le verra au paragraphe suivant, être acquis à l’issue d’un événement ou d’un processus. Le terme d’état caractéristique, dont le contenu sera confronté aux données du schème faula à la section 4.1, est préférable à l’emploi habituel des grammaires arabisantes qui oppose, dans les verbes du schème simple, ‘qualité permanente’ ou ‘durable’ à ‘qualité temporaire’ (par exemple, Wright 1896–98, I : § 36–38). La durée relative d’un état – tout comme celle d’un processus ou d’un événement – correspond en effet à l’une des valeurs aspectuo-temporelles attribuées à une forme verbale donnée en fonction du contexte, et ne doit pas être confondue avec une valeur sémantique morpho-lexicale, associée comme telle à un schème donné. – Ainsi, h azina a ‘être ou devenir triste’ (de schème faila, adjectif correspondant : h azīn) est traditionnellement cité comme exemple de ‘qualité temporaire’ ou ‘transitoire’.18 Il réfère toutefois à une ‘qualité durable’ dans les énoncés suivants : fa-h azina ilā āxiri umrihi ‘Et il fut triste jusqu’à la fin de ses jours’ ; inna lladīna āmanū . . . lā xawfa alayhim wa-lā hum yah zanūn ‘Ceux qui ont eu foi . . . aucune crainte


Ce verbe est cité par Caspari-Uricoecha pour illustrer “un état transitoire et passager ou bien une propriété ou une qualité qui n’affecte l’objet du verbe que pendant peu de temps” (1881, 32) ; par Wright, comme illustrant “a temporary state or condition, or a merely accidental quality in persons or things” (1896–98 I : 30) ; par Brockelmann (1948, 35) comme exemple de verbes exprimant “toujours une qualité ou une situation accidentelle ou temporaire” (“Die Form faila steht durchweg für zufällige, vorübergehende . . . Eigenschaften und Zustände”) ; ou Boormans (1967, 10), comme décrivant un “état passager”.


joseph dichy

[à avoir] pour eux, il ne connaîtront [éternellement] aucune tristesse’ – mot-à-mot : ‘et aucunement ne s’attristeront’ (Coran 2 : 62). H azīn sera donc mieux décrit comme un adjectif d’état non caractéristique. – Inversement, le verbe xalua u ‘mériter le nom de dépravé’, ‘être un dépravé’ (adjectif correspondant : xalī), qui relève du schème faula, réputé renvoyer à des verbes de ‘qualité permanente’ ou ‘durable’ est glosé dans le dictionnaire par l’expression taraka l-h ayā wa-rakiba hawāhu, ‘abandonner [toute] honte et se livrer à ses passions’ (al-Mujam al-Wasīt : racine /x-l-‘/). La qualité de xalī ‘dépravé (notoire)’ n’est donc pas ‘permanente’. Xalī  peut en revanche être d’écrit – indépendamment toute référence à une durée – comme un adjectif d’état caractéristique. Exemples de ces deux sous-catégories de la notion d’état : – verbes associés à un adjectif d’état caractéristique : tāla u ‘être tawīl’ ‘long’ ; hayifa a ‘être ahyaf ’, ‘avoir la taille élancée’ ; awira a ‘être awar’, ‘borgne’ ; baxula u ‘être baxīl’, ‘avaricieux’ (au sens de ‘être avare par nature’)19 – verbes associés à un adjectif d’état non caractéristique : farih a a ‘être farih ’, ‘joyeux’ (l’un des contraires de h azina ci-dessus) ; yaisa a ‘être yāis’ ‘désespéré’ ; atiša a ‘être atšān’ ‘ayant soif ’, ‘avoir soif ’ ; marida a ‘être marīd’ ‘malade’. . . 3.3.3.b Verbes d’état vs verbes de changement d’état produisant un état résultatif Pour H. Fleisch, cependant : Le verbe de qualité n’est pas un statif. Il signifie ‘acquérir une qualité’, autrement dit : ‘devenir tel’ (d’après la qualité en question) : karuma = ‘devenir

19 Les propriétés de l’agentivité ne relèvent pas d’invariants linguistiques, mais de la représentation du procès véhiculée par le vocabulaire de chaque culture. Ainsi, dans la culture arabe médiévale, on peut, soit (a) se montrer avare envers quelqu’un dans une situation donnée, ou (b) être avare par nature ( fī t-tab) : cf. par exemple la préface du Kitāb al-Buxalā ‘Livre des avares’ d’al-Jāhiz, m. en 255/868 ; semblablement l’avarice est, à l’époque de Molière considérée comme pouvant être inscrite dans le caractère de celui qui, comme Harpagon, est un avaricieux. Le premier cas (a) est exprimé au moyen du verbe pleinement agentif baxila a alā ‘se montrer avare envers qqun’, et le second (b), par le verbe d’agentivité neutralisée baxula u ‘être un avaricieux’ (l’avare est le siège du vice inscrit dans sa nature).

faula, faila, faala


karīm (généreux)’ ; ou bien, comme conséquence de l’acquisition : ‘avoir une qualité’, soit : ‘être tel’, c’est un résultatif, alors karuma = ‘être karīm’ (généreux) (1968, 117).

Fleisch suit vraisemblablement ici la définition que donne al-Mubarrid de la valeur sémantique des verbes en faula : innamā huwa li-l-h āl allatī yantaqil ilayhā l-fāil . . . nahwu karuma wazarufa wa-šarufa “elle (cette classe de verbes) désigne uniquement l’état dans lequel entre le sujet [du verbe] (mot-à-mot : ‘l’état vers lequel se déplace le sujet’), exemples : karuma, zarufa et šarufa” (Muqtadab I : 209 – traduction des exemples : ‘devenir’ et donc ‘être noble-et-généreux’, ‘devenir’, puis ‘être raffiné’ et ‘devenir’, puis ‘être prééminent, illustre, noble’).

Ces deux descriptions font de l’état caractéristique une valeur seconde par rapport à la valeur ‘devenir qqch’. Or Sībawayhi décrit les verbes de schème faula comme dénotant “les caractères qui peuvent être dans les choses” (al-xisāl allatī takūnu fī l-ašyā – Kitāb IV : 28). Ces termes ont été par la suite largement repris dans les sciences linguistiques arabes médiévales (cf. par ex. az-Zamaxšarī, Mufassa l : 279). Dans cette conception, la valeur fondamentale des ‘verbes de qualité’ est stative, alors que chez al-Mubarrid et Fleisch, elle correspond au changement d’état. Mais faut-il vraiment opposer ces conceptions ? Est-il possible de décider de la valeur sémantique qui serait ‘première’ par rapport à l’autre ? Considérons les exemples suivants : – H asuna u ‘être h asan’, ‘beau’, qaduma u ‘être qadīm’, ‘ancien’ ou hayifa a ‘être ahyaf ’, ‘qui a la taille élancée’ désignent à la base un état caractéristique. – Šamasa i ou u ‘devenir/être ensoleillé’ (jour) ; ġasaqa i ‘devenir/être noir, obscur’ (nuit) réfèrent fondamentalement à un changement d’état. Il s’agit de verbes signifiant ‘devenir qqch’, décrivant un processus qui aboutit à un état résultatif non caractéristique. – Ġadiba a ‘se mettre/être en colère’ ou jāa u ‘avoir faim/être affamé (= ayant faim)’ correspondent fondamentalement à des verbes de changement d’état, dans lesquels le ‘nouvel état’ (non caractéristique) qui affecte le sujet grammatical résulte du procès. Les adjectifs correspondants, ġadib, ġadbān ‘en colère’ et jāi, jawān ‘ayant faim’ sont des états résultatifs non caractéristiques. – Le verbe awira a ‘devenir/être borgne’ désigne – selon le contexte – soit un changement d’état (‘devenir qqch’) débouchant sur l’état caractéristique ‘être awar’, ‘borgne’, soit cet état caractéristique lui-même (qui peut être résultatif ou non).


joseph dichy

– Un verbe d’état caractéristique – correspondant fondamentalement à un état, comme h asuna u ‘être beau’ ou hayifa a ‘avoir la taille élancée’ ci-dessus – peut aussi être amené par le contexte à désigner le résultat d’un événement ou d’un processus, exemple : hayifa l-āna l-fatā ‘le jeune homme est maintenant ahyaf ’, ‘a maintenant une taille élancée’ (l’énoncé présuppose : ‘maintenant qu’il a grandi’). Ces exemples montrent qu’à la question posée (ce qui, de la valeur sémantique stative ou du changement d’état serait ‘premier’ par rapport à l’autre), il n’y a pas de réponse générale, qui demeurerait valable dans tous les cas. La réponse doit au contraire être proposée pour chaque verbe, en fonction de son sens lexical (et des connaissances encyclopédiques associées à celui-ci). C’est ainsi que : – Šamasa i ou u ‘devenir/être ensoleillé’ ou jāa u ‘avoir faim’ décrivent d’abord un changement d’état (‘devenir qqch’), et en second lieu, l’état non caractéristique correspondant. – Le sens des verbes d’état caractéristique qaduma u ‘être ancien’ ou jayida a ‘avoir le cou long’ (signe de beauté) est peu compatible avec un contexte dans lequel ces verbes prendraient la valeur ‘devenir tel’ (bien qu’un tel contexte demeure, comme on vient de l’indiquer, possible). – Awira a, qui décrit un état caractéristique, est compatible tant avec la valeur stative ‘être borgne’ qu’avec celle du changement d’état (‘devenir tel’), sans qu’il soit possible de décider de la valeur qui dériverait de l’autre. 3.3.4. L’événement de l’acquisition d’état, cas particulier du changement d’état et de l’état résultatif Mais il convient, pour comprendre la présence dans ces verbes des deux valeurs stative et d’acquisition d’état, d’interroger plus avant la notion d’état résultatif. Ce dernier est engendré par un procès (événement ou processus) correspondant à un changement d’état (Desclés 1994, 71 et suiv. ; Desclés et Guentcheva à paraître, chap. II). Dans les données suivantes, cet état met en jeu des notions dans lesquelles les propriétés de l’état, de l’événement et du processus se trouvent combinées : (a) État résultatif (caractéristique ou non) engendré par un événement : – Le verbe amura u – également réalisé amira a – est glosé dans les dictionnaires comme ‘devenir émir’, notamment par succession (al

faula, faila, faala


Mujam al-Wasīt, al-Munjid, Hans Wehr-Cowan 1979). Ce verbe décrit, dans cette acception, l’événement d’un changement d’état, et le nom-adjectif associé amīr, qui désigne l’état caractéristique ‘émir’, ‘prince’, ‘celui qui gouverne un émirat’, correspond à un état résultatif. Mais il peut également, comme nous le verrons au paragraphe suivant, désigner l’état caractéristique correspondant, indépendamment d’un événement ou d’un processus qui l’aurait précédé. – H azina a ‘être’ ou ‘devenir triste’, ‘s’attrister’ désigne l’état non caractéristique h azīn ‘triste’ ou l’entrée dans cet état. L’événement que constitue cette entrée peut être causé par un autre événement, par exemple l’arrivée d’une mauvaise nouvelle : lammā samia xabara axīhi, h azina šadīda l-h uzni ‘lorsqu’il entendit les nouvelles de son frère, il tomba dans une profonde tristesse’ (mot-à-mot : ‘il s’attrista d’une intense tristesse’). L’adjectif h azīn, décrit ci-dessus (§ 3.3.3.a) comme un adjectif d’état non caractéristique peut donc en outre correspondre à un état résultatif, h azina relevant, dans ce dernier cas, des verbes de changement d’état. (b) État résultatif (caractéristique ou non) correspondant au dernier état d’un processus : Comme rappelé ci-dessus, un processus correspond à une succession d’états. L’état résultatif peut donc coïncider avec le dernier état du processus qui le produit. Exemples : – azuma šanuhu šayan fa-šayan ‘sa situation sociale est devenue peu à peu (ou ‘devint peu à peu’) de grande importance (azīm)’ ; – sawida a ‘devenir ou être aswad’ ‘noir’ ; saġura u ‘devenir ou être (socialement ou quantitativement) saġīr’ ‘petit’ ; – xadira z-zaru ‘la récolte a verdi’ ou ‘verdit’ (Munjid, racine /x-d-r/ ; Ilyās et Nāsīf 1995, 113). Ces données montrent que la notion d’état résultatif est loin d’aller de soi. Elles appellent deux remarques générales : 3.3.4.a Valeur aspectuo-temporelle du suffixé (mādī) et événement engendrant un état exprimé ou non par un adjectif assimilé’ On connaît en français la valeur ‘d’accompli du présent’ que peut prendre le passé composé et qui indique – dans une énonciation qui inclut à la fois le passé composé et le présent de l’énonciateur – “le résultat présent d’un procès accompli antérieurement” (Maingueneau 1994,


joseph dichy

67–68). Desclés (1994, 73) décrit ce résultat comme “un état-présent concomitant à l’acte d’énonciation”. “Comparons par exemple : (1) Alors tu te dépêches ? – Je finis ! (2) Alors tu te dépêches ? – J’ai fini !” (Leeman-Bouix 1994, 61–62). En (1), l’interlocuteur est en train d’achever quelque chose, le verbe conjugué au présent a donc valeur d’inaccompli ; en (2), le passé composé exprime un processus à la fois accompli et achevé (voir, pour cette distinction, Desclés 1994, 76–80). En arabe, la conjugaison au mādī (suffixé) comporte également la propriété d’engendrer un état résultatif, exemples : – saimtu takālifa l-h ayāti . . . ‘je suis las des fardeaux [dont nous charge] la vie’ (Muallaqa ‘grande ode’ du poète préislamique Zuhayr b. Abī Sulmā, 46e vers) : la forme conjuguée au suffixé (mot-à-mot : ‘j’ai été’ ou ‘je fus las . . .’) engendre l’état résultatif ‘être las’. La traduction française : ‘je suis las’ est contrainte d’exprimer cet étatprésent (au sens de Desclés 1994, 73), mais elle ne peut ‘rendre’ par une forme verbale l’événement qui l’engendre. En arabe le mādī (suffixé) de saimtu englobe ces deux interprétations (cf. l’exemple de wasala au § 3.3.1) ; – atištu ‘j’ai soif ’ : la forme au suffixé (mot-a-mot : ‘j’eus’ ou ‘j’ai eu soif ’) engendre l’état ‘avoir soif ’ (en traduction : ‘j’ai soif ’ – même commentaire que ci-dessus), qui pourrait être glosé comme : ‘la soif s’est maintenant installée en moi’ (‘ou s’installa désormais en moi’). Il est essentiel, pour comprendre le sens des schèmes associés à la notion d’état, d’observer que ces exemples illustrent la présence en arabe de deux catégories différentes de verbes, selon que l’état résultatif est exprimé ou non par un adjectif associé au verbe par la dérivation morpho-lexicale. L’état résultatif engendré par saimtu n’est pas lexicalisé : le cas est donc comparable – mutatis mutandis – à l’usage du passé-simple français. Pour atištu, l’état qui résulte du procès correspond à une entrée lexicale (i.e. à un adjectif ‘assimilé’). Ainsi :

faula, faila, faala


En reprenant les exemples (a) du § 3.3.4, qui décrivent des événements : – L’énoncé amura š-šāiru (‘le poète est devenu/devint le prince [régnant]’ ou le ‘poète règne/régna’) peut être glosé comme : ‘la position de prince est maintenant (ou fut désormais) occupée par le poète’. Le sens du verbe amura u est donc bien ‘devenir’, puis ‘être émir’, et non uniquement ‘devenir émir’, comme dans les dictionnaires cités plus haut. – lammā samia xabara axīhi, h azina šadīda l-h uzni ‘lorsqu’il entendit les nouvelles de son frère, il tomba dans une profonde tristesse’ correspond à : ‘lorsqu’il entendit . . . l’état de tristesse s’installa en lui’. Les exemples donnés en (b), qui portent sur des processus, méritent un mot d’explication : l’achèvement du processus parvenu à son dernier état constitue lui-même un événement (Desclés, 1994, 72, 77–80) : – Dans xadira z-zaru ‘la récolte a verdi’ ou ‘verdit’ cet événement correspond au moment où la récolte parvient au degré de mûrissement attendu (l’état de verdeur). – Dans azuma šanuhu šayan fa-šayan ‘sa situation sociale est devenue/devint peu à peu de grande importance’, cet événement est celui de l’achèvement du processus ‘devenir peu à peu’, qui atteint alors son dernier état (celui de ‘grande importance’). Dans tous ces cas, (1°) il y a un avant et un après de l’événement du changement d’état et de l’installation du ‘nouvel état’, et (2°) cet état est désigné dans le lexique par un adjectif ‘assimilé’ (sifa mušabbaha). L’évènement pourrait de ce fait être glosé comme ‘atteindre’ ou ‘acquérir l’état x’, dans lequel x correspond à un adjectif ou à un nom-adjectif : atšān ‘qui a soif ’ ; amīr ‘émir’ ; h azīn ‘triste’ ; axdar ‘vert’ ; azīm ‘de grande importance’. 3.3.4.b Les relations asymétriques ‘être x’ o ‘devenir x’ et ‘devenir x’ Ÿ ‘être x’ La présence dans le lexique d’un adjectif ‘assimilé’ permet de mieux saisir le lien entre ‘être x’ (valeur stative) et ‘devenir x’ (valeur d’acquisition d’état) qui s’établit entre les verbes associés à un adjectif d’état caractéristique ou non caractéristique d’une part, et les verbes de changement d’état de l’autre. Ainsi :


joseph dichy

– qadumat al-yawma dawlatuhum ‘leur dynastie est devenue aujourd’hui ancienne’ peut être glosé comme : ‘leur dynastie est devenue étant ancienne’ (ou ‘devint étant ancienne’). Il s’agit d’un exemple de verbe d’état caractéristique (qaduma u ‘être ancien’) dans lequel on observe un glissement sémantique en contexte de la valeur stative (‘être qadīm’ ‘ancien’) à celle d’acquisition d’état (‘être devenu étant qadīm’). C’est la rencontre au sein de l’énoncé du sens lexical ‘être ancien et de la valeur aspectuo-temporelle d’achèvement associée au mādī (suffixé) ellemême ‘renforcée’ ici par le marqueur temporel al-yawma ‘aujourd’hui’, qui permet le glissement sémantique ‘être x’ o ‘devenir x’ observé. – ġadiba r-rajulu’ ‘l’homme est (ou fut) en colère (ġadib)’ correspond à : ‘l’état être en colère s’est accompli en lui’. Cet exemple est celui d’un verbe d’acquisition d’état non caractéristique, dans lequel on rencontre un passage de la valeur de changement et d’acquisition d’état (‘devenir en colère’) à la valeur d’état résultatif acquis (‘être en colère’), la seconde étant engendrée par la première. Il peut être glosé comme : ‘être devenu étant ġadib (en colère)’. C’est également la rencontre en contexte du sens lexical ‘être en colère’ et de la valeur d’achèvement que prend le mādī (suffixé) qui permet l’engendrement de l’état acquis ‘être x’ par la valeur de changement et d’acquisition d’état ‘devenir x’. Il y a donc deux cas généraux de modification sémantique corrélés aux verbes d’état. Les verbes dont la valeur de base est celle de l’état caractéristique connaissent un glissement sémantique ‘être x’ o ‘devenir x’ (glissement de l’état au changement et à l’acquisition d’état). Dans les verbes exprimant fondamentalement un changement correspondant à l’acquisition d’un état (caractéristique ou non), l’état acquis est engendré par l’événement qui coïncide avec l’achèvement du processus d’acquisition, ce qui peut être représenté schématiquement comme : ‘devenir x’ Ÿ‘être x’ (dans lequel le symbole ‘Ÿ’ se lit ‘engendre’). 3.3.5 Les deux valeurs sémantiques associées à la notion d’état : l’état caractéristique et l’acquisition d’état L’analyse ci-dessus conduit donc à distinguer, pour les verbes associés à un adjectif d’état caractéristique, d’une part l’état caractéristique (‘être x’) et de l’autre l’acquisition d’état (‘devenir x’), qui correspond à un changement d’état. Il leur est associé, respectivement, des adjectifs d’état caractéristique et des adjectifs d’état acquis.

faula, faila, faala


L’examen des données relatives aux verbes associés à un adjectif d’état non caractéristique permet en outre d’observer que ces verbes relèvent tous du changement d’état : ils appartiennent de ce fait à la catégorie des verbes d’acquisition d’état. A condition de tenir compte des glissements sémantiques décrits plus haut (‘être x’ o ‘devenir x’ et ‘devenir x’ Ÿ ‘être x’), l’analyse de Fleisch et la conception d’al-Mubarrid citées plus haut (§ 3.3.3.b), qui ne s’applique que partiellement aux verbes associés à un adjectif d’état caractéristique, s’applique en revanche pleinement aux verbes associés à un adjectif d’état non caractéristique, que la tradition arabisante appelait ‘état temporaire’. On en a examiné ci-dessus plusieurs exemples. 3.4 Benveniste revisité : la notion de verbe moyen (ou diathèse interne) ; le médio-passif L’autre grande distinction sur la définition de laquelle il convient de revenir est celle de ‘verbe moyen’. On appelle verbe moyen selon la définition d’E. Benveniste, un verbe qui décrit “un procès dont le sujet est le siège” :20 Ici, le sujet est le lieu du procès, même si ce procès, comme c’est le cas pour lat. fuor [‘jouir, avoir profit de’] ou sanskr. manyate [‘éprouver une agitation mentale’], demande un objet ; le sujet est centre en même temps qu’acteur du procès ; il accomplit quelque chose qui s’accomplit en lui, naître, dormir, gésir, imaginer, croître, etc. Il est bien intérieur au procès . . . (1950/1966, 172).


Ambiguïté de cette définition ; les verbes médio-passifs

La formulation ci-dessus pose un double problème : (a) Dans les verbes d’état aussi on pourrait dire que “le sujet est le lieu du procès”, même si l’on voit bien qu’il ne peut en être “l’acteur”. Les verbes d’état, dans lesquels l’agentivité est neutralisée, n’ont pas de diathèse, et ne constituent en aucun cas une sous-catégorie du verbe moyen. Cette

20 A propos de la notion de verbe moyen en arabe classique, outre M. Cohen (1929), Joüon (1930), Cantineau (1950) et Fleisch (1957 ; 1968, 116–117), comparer la description ci-dessus avec P. Larcher (1995, 295–6 ; 2003, 22–26) ou, pour l’arabe dialectal égyptien, de C. Audebert (2002).


joseph dichy

restriction, qui est implicite dans la définition ci-dessus, se déduit également de cette remarque : Il ne s’agit donc nullement de faire coïncider la différence de l’actif au moyen avec celle des verbes d’action et des verbes d’état (Benveniste 1950/1966, 172).

(b) Si le sujet est “acteur du procès” dont il est le siège, s’il “accomplit quelque chose”, cela suppose qu’il jouit d’un certain degré d’agentivité. La propriété d’agentivité est incluse de même dans la notion de diathèse à laquelle Benveniste a recours, notamment lorsqu’il propose en conclusion de l’article cité, de substituer “diathèse interne” (par opposition à la “diathèse externe”) au terme de “verbe moyen” (par opposition au “verbe actif ” ou “externe”). Or le passage cité plus haut donne comme exemples de ce type de verbes naître, gésir et croître, qui sont non agentifs. Pour traiter cette difficulté, il convient d’emprunter au même article (mais non sur ce point à l’analyse) d’E. Benveniste la notion de verbe médio-passif, pour décrire les verbes non agentifs – hors verbes d’état bien entendu – comme namā u ‘croître’, ‘grandir’, waqaa a fī ‘se trouver’, ‘être situé à’ ou ‘dans’, madā ī ‘passer’ (temps), māta u ‘mourir’, etc.21 3.4.2 Diathèse interne vs diathèse externe et propriétés de transitivité/ intransitivité Le ‘verbe moyen’ s’oppose au verbe ‘actif ’ ou ‘externe’ : De cette confrontation [entre deux types de verbes] se dégage assez clairement le principe d’une distinction proprement linguistique, portant sur la relation entre le sujet et le procès. Dans l’actif, les verbes dénotent un procès qui s’accomplit à partir du sujet et hors de lui.22 Dans le moyen, qui est la diathèse à définir par opposition, le verbe indique un procès dont le sujet est le siège ; le sujet est intérieur au procès” (Benveniste, 1950/1966, 172). L’actif est “une production d’acte, révélant plus clairement encore la position extérieure du sujet relativement au procès ; et le moyen servira à définir le sujet comme intérieur au procès (op. cit. 173).


Je fais mienne ici une suggestion de Jean-Pierre Desclés, communication orale. Il est intéressant de noter – à titre de comparaison – que Sībawayhi considérait que la transitivité fait entrer un verbe “dans le chapitre des actions visibles ou audibles” ( fī bāb al-amāl allatī turā wa-tusma – Kitāb IV : 6). Cette description est reprise dans le commentaire d’al-Sīrāfī (en note de la même page) et dans Ibn Ya‘īš, Šarh al-Mufassa l VII : 157. 22

faula, faila, faala


Le verbe “actif ” ou “externe” est par définition un verbe transitif : Ainsi se constituent à partir du moyen des actifs [. . .] qui se caractérisent toujours par ceci que le sujet, posé hors du procès, le commande désormais comme acteur, et que le procès, au lieu d’avoir le sujet pour siège, doit prendre un objet pour fin (ibid.).

Rappelons que la transitivité doit s’entendre, en ce qui concerne les données traitées ici, comme directe ou indirecte (i.e. à complément prépositionnel). Exemples de verbes de diathèse externe : h akama šayan ‘maîtriser’ qqch. ; amlā alā fulānin šayan ‘dicter qqch. à qqun’ ; dafaa a ‘pousser, repousser’ ou ‘payer’ ; kataba u ‘écrire’ ; qāla u ‘dire’ . . . Les verbes moyens ou de “diathèse interne”, dans la définition de Benveniste, peuvent être aussi bien intransitifs que transitifs directs ou indirects. Exemples de verbes du schème simple : – verbes moyens intransitifs : baqiya a, ‘rester, demeurer’ lorsque le sujet est humain ; farih a a ‘se réjouir’, ‘éprouver de la joie’ ; – verbes moyens transitifs directs : malla (d’alternance vocalique i/a) ‘être ennuyé par’, ‘éprouver de l’ennui [à cause] de’ ; xašā i ‘craindre’ ; – verbes moyens transitifs indirects : šaara u bi-, ‘sentir’, ‘ressentir’ ; raġiba a fī ‘désirer qqch’, et son contraire raġiba a an ‘désirer éviter ou s’épargner qqch’.

4. Valeurs sémantiques de base des schèmes simples Le cadre conceptuel étant posé, considérons maintenant les valeurs sémantiques associées aux trois schèmes faula, faala et faila. 4.1 4.1.1

Les deux valeurs sémantiques du schème faula Les faula à valeur d’état caractéristique

Les verbes de ce schème sont traditionnellement décrits comme exprimant un état ou une ‘qualité durable’ ou ‘constante’, par opposition aux


joseph dichy

verbes en faila, qui dénoteraient un ‘état temporaire’ ou ‘instable’ (cf. par exemple az-Zamaxšarī, Mufassa l : 230). A cette référence à la durée, j’ai proposé de substituer la notion d’état caractéristique, selon la définition ci-dessus (§ 3.3.3.a). Ces verbes décrivent effet une qualité ou un état posé comme ‘constitutif ’ ou ‘caractéristique’ du sujet grammatical. Wright désignait ce dernier comme correspondant à “une qualité inhérente naturelle” (“a naturally inherent quality” – 1896–98, I : § 36–38), et Sībawayhi, déjà cité, comme “les caractères susceptibles d’affecter les choses”, mot-à-mot “ qui peuvent être dans les choses” (al-xisāl allatī takūnu fī l-ašyā – Kitāb IV : 28). Dans les faula à valeur d’état caractéristique, le procès ‘état’ est, comme on l’a vu, d’agentivité neutralisée. Ces verbes sont, comme il est attendu, intransitifs.23 Exemples (hors glissement sémantique de type ‘être x’ o ‘devenir x’) : qabuh a u ‘être qabīh ’ ‘laid’ (valeur physique ou morale) ; h asuba u ‘être h asīb’ ‘noble, de haute naissance’ ; h aluma u ‘être h alīm’ ‘clément’ ; sabuh a u ‘être sabīh ’ ‘brillant, radieux’ (visage) ou ‘beau, joli’ (garçon, personne) ; katufa u ‘être katīf ’ ‘épais, dense’. Faula est le schème simple le plus stable d’un point de vue formel. Nous avons vu (figures 1 et 2) que la voyelle de sa deuxième consonne radicale demeurait inchangée au paradigme du préfixé (mudāri). Ce schème est aussi le plus stable sémantiquement. Sa valeur de base, l’état caractéristique, est toutefois l’objet d’un glissement de sens qui lui est étroitement corrélé, le changement et l’acquisition d’état (§ 3.3.3.b et 3.3.4 ci-dessus). 4.1.2

Les faula à valeur d’acquisition d’état caractéristique

Ce glissement sémantique, très fréquent, est illustré par l’exemple : ġayra anna hādihi l-jamāa . . . d. a'ufa h udūruhā l-alaniyy . . . mundu an . . . ‘Mais la présence publique de ce rassemblement . . . s’est affaiblie (mot-à-mot, ‘est faible’) . . . depuis que . . .’ (Majallat an-Nahār, 2 juillet 2006, 13, col. 1).

23 Cf. Sībawayhi : “Il n’existe pas dans le langage des Arabes de verbe transitif en faula” (laysa fī l-kalām faultuhu mutaaddiyan – Kitāb IV : 38).

faula, faila, faala


Des exemples similaires ont été vus aux § 3.3.4. a et b. Le contexte indique clairement un début du procès (‘depuis que . . .’) : le verbe daufa u (adjectif d’état caractéristique : daīf ‘faible’) ne réfère donc pas à un état. La forme conjuguée au mādī exprime ici un événement, qui coïncide avec l’achèvement du processus décrit par le verbe (l’affaiblissement) : ce processus a atteint son ‘dernier état’, ce qui détermine un avant et un après. Cet après correspond à un état résultatif que l’on peut décrire comme un état acquis : cette ‘présence publique’ est ‘devenue faible’ ou ‘est désormais faible’. On observe, à travers les usages illustrés par cet exemple et ceux qui ont été analysés plus haut, que la valeur sémantique ‘être x’ incluse dans celle de l’état caractéristique se trouve très souvent réalisée en contexte comme un changement d’état (‘devenir x’). Ce changement a, à son tour, la propriété d’engendrer un état résultatif. Un tel phénomène n’est nullement propre à l’arabe. On peut comparer, au moins partiellement, avec les exemples suivants, qui sont empruntés au français : – Ah ! Maintenant, tu es beau ! (dit une maman à son fils, qui revenait en traînant les pieds de chez le coiffeur). – Lucie est désormais notre ancêtre. Dans les deux cas, les prédicats ‘être beau’ et ‘être notre ancêtre’ paraissent incompatibles avec un contexte qui leur assignerait un commencement ou un terme. Or les marqueurs temporels ‘maintenant’ et ‘désormais’ présents dans ces exemples assignent aux procès un début, ce qui a pour effet d’opérer une translation des prédicats verbaux ci-dessus de la catégorie de l’état vers celle du changement et de l’acquisition d’état. C’est comme si l’on disait : ‘tu es devenu beau’ (maintenant que tu as les cheveux coupés) et ‘Lucie est devenue l’ancêtre de l’humanité’ (depuis sa découverte et les recherches qui ont suivi). Le dictionnaire de Hans Wehr-Cowan (1979) traite le plus souvent – mais pas systématiquement – le glissement sémantique de l’état caractéristique vers l’acquisition d’état comme relevant des propriétés lexicales des verbes concernés, exemples : (a) glissement sémantique traité comme inclus dans le sens du verbe : daufa u ‘to be or become weak’ ; katura u ‘to be . . . numerous’ – suivi de : ‘to increase, augment’ ; saġura u ‘to be or become small’ (1er sens de ce verbe) ;


joseph dichy

(b) glissement sémantique non pris en compte dans le sens du verbe : h asuna u ‘to be handsome’, ‘beautiful’ ; qabuh a u ‘to be ugly’ ; zarufu u ‘to be charming’ (. . .) ‘witty’ ; saġura u (2e ‘sens’ de ce verbe) : ‘to be young’ ; qaduma u ‘to be old, ancient’. Si ce dictionnaire semble hésiter, c’est que la question de savoir si ce glissement sémantique est assez fréquent pour être inclus dans le sens lexical du verbe n’est pas toujours facile à trancher. Les traductions proposées dans Wehr-Cowan proviennent vraisemblablement des usages attestés dans les fiches des auteurs. Ces données montrent en tout état de cause que, pour un verbe donné du schème faula, les deux cas présentés plus haut (§ 3.3.4.b) se rencontrent. Voyons de quelle manière : 4.1.2a Le glissement sémantique ‘être x’ o ‘devenir x’ Dans les verbes en faula dont le sens est fondamentalement celui de l’état caractéristique, comme dans h asuba u ‘être h asīb’ ‘noble, de haute naissance’, un contexte d’acquisition de cet état est presque toujours possible. Cette valeur doit alors être considérée comme sémantiquement seconde, qu’elle soit inscrite dans le lexique ou qu’elle soit limitée à un énoncé donné. Le glissement sémantique observé correspond alors au schéma : ‘être x’ o ‘devenir x’, exemples : – h asuba l-āna jāruna ‘notre voisin maintenant est devenu un noble’ (énoncé ironique) ; – karuma r-rajulu ‘l’homme s’est montré noble-et-généreux’. Dans ces deux énoncés, la présence d’un adjectif d’état caractéristique associé par les structures morpho-lexicales de la langue aux verbes en faula suggère une valeur attributive : ‘notre voisin a acquis le titre de h asīb’, i.e. de ‘noble de haute naissance’ (d’où l’ironie) ; ‘l’homme a mérité le nom de karīm’, ‘noble-et-généreux’. 4.1.2.b L’engendrement de l’état acquis par l’événement ou le processus du changement d’état (‘devenir x’ Ÿ ‘être x’) Dans les faula dont le sens de base est celui de l’acquisition d’un état, comme dans badua u ‘devenir incomparable, sans pareil’ (al-Mujam al-Wasīt et al-Munjid, racine /b-d-‘/), le contexte associe au verbe un état acquis, qui est engendré, soit au terme du processus de changement d’état (l’état acquis correspond alors au dernier état du processus), soit par un événement. Exemple :

faula, faila, faala


– badua ibn sīna bi-ilmihi xayran wa-šarran ‘Avicenne fut incomparable par sa science, en bien comme en mal’. Du vivant du philosophemédecin, l’état acquis eût pris une valeur d’état-présent dont la durée inclut le moment de l’énonciation (Desclés, 1994, 73), la traduction étant alors ‘Avicenne s’est révélé . . .’ ou ‘se révèle . . .’ Comme dans les exemples du paragraphe précédent, le verbe en faula peut être interprété comme prenant une valeur attributive dont le prédicat est l’adjectif d’état caractéristique correspondant : ‘Avicenne mérita le nom de badī, incomparable et admirable’. Ces exemples réalisent le schéma : ‘devenir x’ Ÿ ‘être x’ (dans lequel ‘Ÿ’ se lit ‘engendre’). 4.1.3

Schème faula et adjectifs ‘assimilés’ et/ou participes

Si l’on écarte le glissement sémantique ‘être x’ o ‘devenir x’, le sens des verbes à valeur d’état caractéristique exclut les participes actif (ism al-fāil) et passif (ism al-maf ūl). Les valeurs aspectuo-temporelles de base des participes sont – en fonction du contexte – celles de l’achèvement (qui peut engendrer un état résultatif ) ou du progressif (dont le terme peut, ou non, être spécifié).24 Ces valeurs ne sont pas compatibles avec celles de l’état caractéristique, dont la durée n’est pas bornée et qui ne comporte aucune progression. En outre, le participe passif qui ne peut, quant à lui, être dérivé que d’un verbe transitif, est donc en quelque sorte doublement exclu. Les adjectifs ‘assimilés’ ou ‘analogues’ [au participe] (sifa mušabbaha) qui sont associés aux verbes en faula ainsi qu’aux verbes d’état caractéristique ou non relevant d’autres schèmes ‘compensent’ en quelque sorte l’absence de participes. Comparer par exemple les phrases nominales : – hum kuramā ‘ils sont généreux’ ; dans cette phrase, karīm, plur. kuramā, est l’adjectif ‘assimilé’ associé au verbe d’état caractéristique karuma u ; – (a) hum tālibūn ‘ils sont en train de demander/ils ont demandé’ ou (b) hum matlūbūn ‘ils sont demandés’ ; tālib et matlūb sont les 24 Pour les deux valeurs de base des participes (d’achèvement et progressive), comparer avec Roman (1990, 39–40). Pour une analyse développée de la valeur sémantique des participes en contexte et notamment du progressif à terme spécifié ou non, voir Dichy (2002/2003, § 5.2.2). En ce qui concerne la relation morphologique entre certains schèmes et l’adjectif ‘assimilé’ ou le participe actif, comparer la discussion menée ici avec Larcher (2003, 25–26).


joseph dichy

participes actif et passif de talaba u ‘demander’, qui ne relève pas des verbes d’état. Cette division entre d’une part les verbes décrivant un état et dotés d’un ‘adjectif assimilé’ et de l’autre les verbes auxquels sont associés des participes, rencontre toutefois un problème. Au chapitre consacré à l’adjectif ‘assimilé’ (sifa mušabbaha), az-Zamaxšarī (Mufassa l : 230) associe à celui-ci la ‘valeur sémantique d’immutabilité’ manā t-tubūt (qui renvoie à ce qui est tâbit ‘immuable’, ‘établi’ en quelque sorte hors énonciation du temps). Il oppose cette dernière à la ‘valeur d’occurrence dans le temps’ manā l-h udūt, valeur qu’il associe aux participes. Or il n’illustre pas ces derniers par des verbes décrivant des événements ou des processus (comme je l’ai fait ci-dessus avec talaba u ‘demander’), mais par des verbes associés à un adjectif d’état caractéristique. Le premier exemple donné est celui de h āsin, qui correspond à un participe actif de h asuna u ‘être beau’ (adjectif d’état caractéristique h asan). Az-Zamaxšarī souligne la ‘valeur d’occurrence dans le temps’ manā h udūt de la forme h āsin en insérant celle-ci dans les phrases : – huwa h āsin al-āna ou ġadan, mot-à-mot : ‘il est en train de devenir/il est devenu beau maintenant’, ou ‘il le deviendra demain’. – Les exemples suivants sont : kārim, participe actif de karuma u ‘être noble et généreux’, tāil, participe actif de tāla u ‘être ou devenir long’ et dāiq, participe actif du verbe d’état caractéristique dāqa i ‘être ou devenir étroit’ (adjectif d’état caractéristique dayyiq), pour lequel az-Zamaxšarī cite le verset : wa-dāiqun bihi sadruka, ‘alors que tu es angoissé par cela’, mot-à-mot, ‘ta poitrine étant oppressée (rendue étroite) par cela’ (Coran 11 : 12). Ibn Ya‘īš commente, dans ce verset, le choix du participe actif dāiq au lieu de l’adjectif dayyiq en indiquant que le premier réfère à “une étroitesse (i.e. à une oppression) [correspondant à] un accident [se produisant] dans le présent et non à un [caractère] immuable” (dīq ārid fī l-h āl ġayr tābit – Šarh al-Mufassa l VI : 83). Si les verbes associés à des adjectifs d’état caractéristique peuvent être compatibles avec le sens des participes actifs, c’est en raison : (a) soit du glissement de sens qui permet à ces verbes, lorsque leur sens premier est celui de verbes d’état, de prendre celui de l’acquisition d’état (glissement ‘être x’ o ‘devenir x’),

faula, faila, faala


(b) soit du fait que ces verbes réfèrent, fondamentalement, à l’acquisition d’un état caractéristique, c’est-à-dire, à un processus ou un événement engendrant un état résultatif (schéma ‘devenir x’ Ÿ ‘être x’). Les quatre exemples cités par az-Zamaxšarī relèvent de ces deux cas. Si h asuna ‘être beau’ relève de la première catégorie, on a vu qu’al-Mubarrid, cité au § 3.3.3.b, considérait karuma comme ayant une valeur d’acquisition d’état, et que le dictionnaire de Hans Wehr-Cowan oscillait entre la valeur de base à attribuer à chaque verbe. La valeur attributive se distribue, comme on l’a vu, sur les verbes relevant tant de (a) que de (b). Il apparaît clairement, en conséquence, que l’on ne peut réduire les valeurs sémantiques du schème faula au seul sens de ‘verbe d’état’ : il s’y ajoute l’acquisition d’état. 4.1.4 Deux conséquences de l’agentivité neutralisée ou de la non agentivité des faula L’agentivité neutralisée des verbes à valeur d’état caractéristique, et la non agentivité des verbes dénotant l’acquisition d’un état caractéristique entraînent également l’impossibilité pour ceux-ci de l’impératif : ce dernier n’est possible que si un certain degré de contrôle exercé par l’agent sur le procès est envisageable (c’est pourquoi les verbes pouvoir et devoir n’ont pas d’impératif en français). En outre, l’agentivité neutralisée est incompatible avec la transitivité. Il s’ensuit que la conjugaison des verbes correspondants exclut toutes les formes du passif (ce dernier n’étant possible qu’avec des verbes transitifs, dans le cadre de contraintes sémantiques précises).25 Le schème faula est relativement peu fréquent dans les textes, notamment en arabe littéraire moderne. On lui préfère sans doute les constructions adjectivales correspondantes. Cela ne signifie nullement que ce schème soit sorti de l’usage ou en voie de disparition, comme le montre l’exemple de daufa ci-dessus, emprunté à la presse actuelle. On trouve également assez souvent dans la prose contemporaine des emplois de construction impersonnelle, à la faveur, pour ainsi dire, de l’agentivité neutralisée : lā yah sunu an . . ., ‘il ne convient pas que . . .’ [Yawm., 79, 18] ; yajduru, ‘il convient’, suivi de an ou d’une forme infinitive . . .


Les critères qui permettent, interdisent ou restreignent l’apparition de l’impératif ou du passif en arabe ont été présentés dans Ammar et Dichy (1999, 19–20).

344 4.2

joseph dichy Les valeurs sémantiques de base du schème faila

Le deuxième schème du verbe simple est d’alternance vocalique i/a. Comme on l’a vu au § 2.2.1, une variante en i/i porte sur un petit nombre de verbes : sur 15 ou 16 verbes inventoriés par les traités de morphologie arabe médiévale, huit sont attestés en arabe moderne par le dictionnaire de Hans Wehr (cf. liste dans Ammar et Dichy 1999, 24). Cette variante, qui affecte essentiellement des verbes de 1re consonne radicale w ou y est purement formelle, et n’entraîne pas de différence de valeur sémantique avec le schème faila dont elle relève. Exemples : warita i ‘hériter’ ; waliya i ‘suivre’ ou ‘succéder à’. On rencontre dans les verbes du schème faila tous les degrés d’agentivité (pleine ou partielle) ou l’absence de cette propriété (non agentivité, agentivité neutralisée), ainsi que les différentes formes de transitivité ou de son contraire. D’où la partition des sens associables à ce schème qui a été observée par les auteurs qui se sont attachés à le caractériser.26 La valeur la plus générale est celle de verbes moyens ou de diathèse interne (au sens ‘revisité’ présenté plus haut), ce schème comportant en outre un nombre limité de verbes associés à un adjectif d’état caractéristique. Il y a quatre grandes catégories sémantiques de verbes en faila, distinguées principalement au moyen des propriétés d’agentivité et de transitivité. 4.2.1 Les faila associés à un adjectif d’état caractéristique (verbes d’état ou d’acquisition d’état) Un certain nombre de verbes d’alternance vocalique i/a (schème faila) relève des verbes associés à un adjectif d’état caractéristique. Ces derniers incluent tous les traits présentés plus haut à propos du schème faula, et comportent comme eux les deux grandes valeurs sémantiques étroitement corrélées ‘être x’ et ‘devenir x’. Les verbes en faila relevant de cette sous-catégorie sont intransitifs, d’agentivité neutralisée (valeur ‘être x’) ou non agentifs (médio-passifs, de valeur ‘devenir x’), et ne peuvent avoir de participe actif que dans les conditions, rarement réalisées, présentées plus haut (§ 4.1.3). Ils décrivent pour la plupart un état caractéristique qui correspond : 26 Outre M. Cohen (1929) et Joüon (1930), voir notamment : Fleisch (1979, 220 suiv.) ; Ammar et Dichy (1999, 23–24).

faula, faila, faala


– soit à une particularité physique ou psychique (défaut ou qualité)27 affectant un sujet humain ou un animal, domestique ou appartenant à un troupeau, – soit une particularité, souvent – mais pas nécessairement – connotée négativement, affectant une entité concrète ou abstraite. L’adjectif ‘assimilé’ est le plus souvent du schème af al, fém. falā. Une série limitée de verbes de cette sous-catégorie relève de racines à 2e radicale w ou y. J’en ai dénombré 71, dont 18 de 2e consonne y dans Ilyās et Nāsīf (1995). Ces verbes ont pour particularité de ne pas appliquer les règles de transformation phonologiques habituelles.28 Exemples : (a) Verbes à adjectif d’état caractéristique décrivant une qualité ou un défaut le plus souvent physique (humain ou animal) : jayida yajyadu ‘avoir le cou long’ (signe de beauté) – adj. ajyad, fém. jaydā ; awiza yawazu ‘être’ ou ‘tomber dans le besoin’ (adj. awaz) ; layisa yalyasu ‘être’ ou ‘se montrer courageux’, ‘tenir sa position (dans un combat)’ – adj. alyas ; rawia yawau ‘être’ ou ‘se montrer admirable (humain)’ pour son intelligence, sa beauté, son courage, etc. (adj. arwa) ; awira yawaru ‘être’ ou ‘devenir borgne’ (adj. awar) ; qawida a ‘avoir le dos et le cou allongés’ (cheval . . .). (b) Verbes à adjectif d’état caractéristique décrivant une particularité affectant une entité non animée concrète ou abstraite : awisa yawasu ‘être’ ou ‘devenir difficile, incompréhensible, abscons’ (affaire, explication) – adj. awas et awīs ; awija yawaju ‘être’ ou ‘devenir tordu, incurvé, tortueux’ (adj. awaj, fém. awjā) ; ġayisa yaġyasu ‘être’ ou ‘devenir doux, souple’ (adj. aġyas) ; qawira yaqwaru ‘être vaste (maison)’ – adj. aqwar. Un nombre également limité, mais plus important, de verbes en faila associés à un adjectif d’état caractéristique relève de racines sans transformation, exemples : 27 Voir notamment : Sībawayhi, Kitāb IV : 17 ; az-Zamaxšarī, Mufassa l : 278. La synthèse ci-dessus provient de mon propre examen des données. 28 Rappelons que dans xāfa, yaxāfu ‘craindre’, d’alternance vocalique i/a, la 2e consonne radicale w est l’objet de transformations en fonction de la voyelle qui la suit et la précède immédiatement, à la différence des verbes de 2e radicale w ou y dont on trouve ici des exemples.


joseph dichy

(c) Verbes décrivant une qualité ou un défaut chez un humain ou un animal : baliha a ‘devenir faible d’esprit’ ou ‘être stupide, idiot’ (adj. ablah, fém. balhā) ; ramida a ‘devenir, puis ‘être chassieux (œil)’ ou (mot-à-mot) ‘devenir ayant’, puis ‘avoir l’œil chassieux (humain)’ – adj. armad ; darida a ‘perdre ses dents’, ‘devenir’ ou ‘être édenté’ (adj. adrad) ; sawifa a (mot-à-mot) ‘devenir ayant’, puis ‘avoir beaucoup de laine’, ‘être laineux’ (ovin ou caprin) – adj. aswaf. (d) Verbes décrivant une particularité affectant une entité concrète ou abstraite : xadira a ‘verdir’, puis ‘être vert’ (adj. axdar ‘vert’) ; sawida a ‘noircir’, puis ‘être noir’ (adj. aswad ‘noir’) ; xaliqa a ‘s’user’, puis ‘être usé’ (vêtement) – adj. axlaq ; sawisa a ‘se gâter’, ‘être rongé par des vers’ (viande, nourriture) – adj. aswas. (e) Verbes de cette sous-catégorie dont l’adjectif n’est pas de schème af al : yasira yaysaru ‘devenir’ puis ‘être facile’ (adj. yasir et yasīr). La valeur sémantique de base des verbes présentés en (c), (d) et (e) est celle de l’acquisition d’état, à la différence des verbes des sous-catégories (a) et (b). Dans leur sens de base, les premiers sont donc non agentifs, et correspondent à des verbes médio-passifs. 4.2.2

Autres verbes médio-passifs et non agentifs en faila

D’autres verbes du même schème sont non agentifs et intransitifs. Ils correspondent, non à des verbes à diathèse interne ou moyens, comme dans les deux paragraphes suivants, mais à des verbes médio-passifs. Exemples : – verbes de sujet grammatical non-humain : rasila a ‘être longue et pendre’ (chevelure) ; rawiya a ‘être irrigué, arrosé’ (terre) ; xadiba a ‘verdir’ (arbre, jardin, terre) ; – verbes de sujet grammatical humain : ġariqa a ‘se noyer’ (au sens propre), ‘mourir noyé’ – adj. ‘assimilé’ ġarīq ; marida a, ‘être ou tomber malade’, adjectif ‘assimilé’ : marīd.

faula, faila, faala


Tous les verbes de cette catégorie ne sont pas associés, dans les dictionnaires, à un adjectif ‘assimilé’. Dans le cas des deux derniers exemples ci-dessus, il est à noter que le participe actif est attesté, mais avec une différence de sens du plus grand intérêt pour la distinction entre les valeurs sémantiques du participe actif et celles de l’adjectif ‘assimilé’ (cidessus, § 4.1.3) : Le Munjid al-luġa l-arabiyya l-muāsira (2000) donne l’exemple de anqaztu ġāriqan ‘j’ai sauvé un noyé’ (racine /ġ-r-q/), et Roman : lā takul hādā t-taāma fa-innaka māridun in akaltahu ‘Ne mange pas de cette nourriture, car tu seras malade si tu en manges’ (1999, 71, d’après le Lisān al-arab, XIIIe s., racine /m-r-d/). Comparons ces emplois. Les deux adjectifs ‘assimilés’ ġarīq et marīd correspondent à des états résultatifs engendrés par l’événement ou coïncidant avec la dernière étape du processus décrit par le verbe (Desclés 1994) : le processus ‘se noyer’ a pour dernière étape ( !) ‘être mort noyé’, décrit par l’adjectif ġarīq, de même que ‘tomber malade’ entraîne l’état décrit par l’adj. marīd. Par contraste, les participes actifs correspondants ġāriq et mārid ont une valeur de progressif à terme spécifié (Dichy 2002/2003, 34, 36, 60) : – le procès de ‘noyade’ n’a pas atteint son terme dans l’exemple du Munjid (le sauvetage a pu avoir lieu) ; – le terme (ou l’achèvement) du procès ‘tomber malade’ est projeté dans le futur dans l’exemple du Lisān cité par A. Roman. Ce dictionnaire glose, significativement māridun à la suite de la phrase ci-dessus par ay tamradu ‘c’est-à-dire : tu seras malade’. Il est significatif – ce qu’atteste notamment l’exemple du Munjid al-luġa l-arabiyya l-muāsira – que ces dérivations restent vivantes en arabe contemporain, même si leur fréquence est faible. 4.2.3 Les verbes de schème faila à diathèse interne (ou moyens) et d’agentivité partielle Un nombre beaucoup plus important de verbes du schème faila correspond à des verbes à diathèse interne, ou verbes moyens : le procès affecte le sujet, que le verbe soit transitif ou non (ainsi : ‘rire de qqun’ ou ‘rire’). Ces verbes sont transitifs indirects ou intransitifs et d’agentivité partielle


joseph dichy

(l’agent, n’étant pas autonome par rapport au procès, ne conserve qu’un contrôle limité sur celui-ci). Ils peuvent soit admettre un participe actif, mais non un adjectif ‘assimilé’, soit être associés dans le lexique à un adjectif ‘assimilé’, et n’admettre de participe actif que dans les constructions vues au paragraphe précédent et en 4.1.3. Exemples : (a) Verbes admettant un participe actif, mais non un adjectif ‘assimilé’ : – verbes intransitifs : baqiya a ‘rester, demeurer’ qq part (agent humain) [Yawm., 12, 12] ; nāma a ‘dormir’ ; dah ika a ‘rire’ (en ‘emploi absolu’, différent de : ‘rire de . . .’, ‘se moquer de . . .’) ; – verbes transitifs directs ou indirects : dah ika a min ou alā ‘rire de’, ‘se moquer de’ qqun ; nasiya a ‘oublier qqch’ ; samia a ‘entendre qqch’ ; saima a ‘s’ennuyer de . . .’ (qqch) [Yawm., 50, 10] ; h asiba a anna . . . ‘penser, croire, estimer que . . .’ ; alima a anna . . . ‘savoir que’ ou ‘apprendre que . . .’ alima a bi-, ‘avoir connaissance de . . .’ . (b) Verbes associés dans le lexique à un adjectif ‘assimilé’, mais n’admettant de participe actif que dans les constructions liées à ‘l’occurrence d’un procès’ (h udūt) – § 4.1.3 : – verbes intransitifs : atiša a ‘avoir soif ’ (adj. atšān) ; taiba a ‘éprouver de la fatigue’, ‘être fatigué’ (adj. taib) ; xafira a ‘éprouver un fort sentiment de timidité’, ‘être très timide’, notamment pour une jeune fille (adj. xafir);29 salima a ‘être sain et sauf ’, ‘échapper’ (à un mal ou à un danger’) – adj. salīm, mais également sālim ;30 – verbes transitifs directs ou indirects : h afiya a bi- ‘accueillir très chaleureusement’ qqun (adj. h afiyy) ; ġadiba a alā ‘être, se mettre en colère contre qqun’ (adj. ġadib). Dans les exemples que l’on vient de voir, la catégorie morphologique de l’adjectif ‘assimilé’ ne correspond pas à un état caractéristique, mais

29 Il existe un second adjectif ‘assimilé’, mixfar ‘grande timide’ : associé à ce dernier, xafirat ‘être une grande timide’ relève des verbes à valeur d’état caractéristique, dans lesquels l’agentivité est neutralisée. 30 Dans des cas tels que sālim l’adjectif ‘assimilé’ est dit, dans la grammaire arabe traditionnelle, prendre la forme (sīġa) du participe actif. Autres exemples : yā’is ‘désespéré’, bā’is ‘misérable’, māhir ‘adroit’, ‘habile’ (artisan) . . .

faula, faila, faala


plutôt à l’état résultatif produit pas le procès (événement ou processus) : le verbe salima produit ainsi salīm ‘sain et sauf ’, mais également sālim min . . . ‘sauf de . . .’ (mal ou danger). Dans un verbe exprimant un processus, comme dans ġadiba ‘se mettre en colère’, l’adjectif ġadib décrit le dernier état de ce processus ; si ġadiba se trouve, dans un contexte donné, décrire un événement, l’adjectif décrit l’état résultatif engendré par cet événement. Ces verbes permettent de préciser la notion d’agentivité partielle. Dans les exemples ci-dessus, le contrôle limité du procès par l’agent peut surtout être glosé négativement :31 refuser de ‘rester’ quelque part, voire d’être ‘sain et sauf d’un danger’ (et agir pour sortir de ce lieu ou de cette situation) ; combattre ou surmonter un sentiment ou une émotion (‘rire, ‘ennui’, ‘timidité’, ‘chaleur humaine’ ou ‘colère’ envers quelqu’un), une sensation (‘sommeil’, fait ‘d’entendre’, ‘soif ’, ‘fatigue’), un ‘mouvement de la pensée’ ou ‘processus cognitif ’ (‘croire’, ‘oublier’, ‘savoir’), etc. On notera enfin que la relation entre le verbe et, soit un participe actif, soit un adjectif ‘assimilé’, est lexicalisée (et non compositionnelle, comme elle l’était sans doute aux plus anciennes époques de la langue arabe) : on doit recourir aux connaissances lexicales pour la construire, et non aux seules connaissances grammaticales ou à une déduction à partir du sens du verbe. 4.2.4 Les verbes du schème faila à diathèse interne, transitifs et d’agentivité entière L’autre grande partie des verbes du schème faila est constituée de verbes également moyens (ou à diathèse interne), d’agentivité entière et de régime syntaxique transitif direct ou indirect. Dans cette configuration, le procès comporte un agent autonome contrôlant effectivement (du fait de l’agentivité entière) l’événement ou le processus décrit par le verbe et un ‘patient’ (du fait de la transitivité). Il englobe en outre d’une manière ou d’une autre l’agent ou se déroule au moins partiellement en lui (diathèse

31 Cf. Roman (1990, 42–43), qui décrit l’agentivité partielle comme une agentivité réactive.


joseph dichy

interne). Ces verbes ont un participe actif, et n’admettent pas d’adjectif ‘assimilé’. Exemples : šariba a, ‘boire’ ; radiya a bi- ‘être satisfait de . . .’ ; lah iqa, ‘suivre, poursuivre qqun’, lah iqa bi-, ‘suivre qqun’ au sens de ‘le rejoindre’ [Yawm., 66, 21] ; rakiba a, ‘monter, chevaucher’ ; watiqa i bi-, ‘faire confiance à (qqun)’. 4.3

Les valeurs sémantiques de faala et leur dispersion

Le troisième schème de la forme simple du verbe est faala. Il comporte comme on l’a vu une alternance vocalique de base ( faala i) et deux variantes résultant de divers conditionnements phonétiques ( faala u et faala a). On pourrait poser que ce schème comprend majoritairement des verbes relevant de la catégorie des verbes ‘actifs’ ou ‘externes’ qui s’oppose structurellement à celle des verbes moyens, au sens donné à ces termes dans l’article de Benveniste cité plus haut (section 3.4). Ce trait serait alors ce qui distingue faala des schèmes faila et faula. Or il existe un nombre très élevé de verbes ne répondant pas à cette définition ‘de principe’. Nous avons vu que ce schème connaissait une importante dispersion sur le plan formel : l’ordonnance systématique du tableau des schèmes de base du verbe simple (fig. 1) est principalement rompue dans le tableau des schèmes du verbe simple et de leurs sous-catégories (fig. 2) par les trois alternances vocaliques possibles du schème faala. Les exemples ci-dessous montrent en outre que cette dispersion se vérifie également du point de vue sémantique. Considérons d’abord les catégories émanant des propriétés d’agentivité et de transitivité : – Verbes pleinement agentifs et transitifs (directs ou indirects) : amana u bi- ‘avoir foi en’ ou ‘dans’ ; qāla u ‘dire’ ; saala a ‘interroger’, ‘questionner’ ; sāh a i bi- ‘crier’ vers, en direction de, contre qqun, ‘le héler’ ; āda u ilā ‘revenir à’. – Verbes pleinement agentifs et transitifs indirects, auxquels est associé un adjectif exprimant un état résultatif, ou état acquis : latafa u bi- ‘se montrer gentil, prévenant envers qqun’ (adjectif ‘assimilé’, exprimant un état résultatif : latīf – cf. Coran 42 : 19) ; daxala u fī ‘entrer dans’ un clan ou une tribu, ‘se placer sous sa protection’ (adj. exprimant un état résultatif : daxīl) ;

faula, faila, faala


– Verbes pleinement agentifs et intransitifs : makata u ‘rester, demeurer (sujet humain)’ ; jalasa i ‘s’asseoir’ ; haraa a ‘se précipiter’ (tête baissée) ; sakata u ‘se taire’ (après avoir parlé) ; samata u ‘se taire’, ‘garder le silence’. – Verbes d’agentivité partielle transitifs (directs ou indirects) : raā a ‘voir’ ; fāza bi- ‘gagner, obtenir qqch’ ; šaara u bi- ‘sentir’, ‘ressentir’ qqch. – Verbes d’agentivité partielle intransitifs : zahara a ‘apparaître (intentionnellement)’ lorsque l’agent est humain32 [Yawm., 32, 22] ; warada i, ‘apparaître’ intentionnellement, ‘arriver’ (avec un agent humain). – Verbes non agentifs intransitifs : – madā i ‘passer, se passer’ (laps de temps, événements ou processus dont on prend en compte la durée) [Yawm., 32, 16] ; badā u ‘paraître’ (avec un sujet non humain) [Yawm., 32, 17] ; h āna i ‘se produire accidentellement’ [Yawm., 61, 15] ; ġaraba u ‘se coucher’ (soleil) ; māta u ‘mourir’. – Verbes non agentifs transitifs (directs ou indirects) : waqa a alā ‘tomber sur’ [Yawm., 32, 17] ; h awā i ‘contenir, inclure, comprendre’. – Verbes d’agentivité neutralisée ou non-agentifs et intransitifs, associés à un adjectif ‘assimilé’ décrivant un état caractéristique (verbes d’acquisition d’état) : – dāqa i ‘être ou devenir étroit’ (adjectif d’état caractéristique : dayyiq) ; jāda u ‘être ou devenir excellent (adj. jayyid) ; xaffa i ‘être ou devenir léger’ (adj. xafīf ) ; šadda u ‘être ou devenir intense, dur ou violent’ (adj. šadīd) ; samala u ‘devenir ou être rigide ou sec’ (adj. sāmil et samīl). Reprenons cette analyse en classant les verbes de schème faala selon les grandes catégories présentées plus haut. On a :

32 L’agentivité est susceptible de varier selon que le sujet grammatical réfère à un humain (āqil) agissant intentionnellement ou non, ou à un non humain (ġayr āqil). Avec un sujet grammatical non humain, ou humain, mais dépourvu, dans le contexte, d’intentionnalité, zahara ou warada sont des verbes non agentifs. Un autre exemple (du schème faila cette fois) est dans Yawm. (16, 15) : fa zafira n-nawmu bi-jufūnī, ‘le sommeil triompha de mes paupières’. Dans cette phrase, le verbe zafira perd le trait d’agentivité qu’il aurait avec un sujet grammatical humain, et devient non agentif (le sommeil ne disposant d’aucun contrôle sur le procès, dont il n’est d’ailleurs pas l’agent). En revanche, waqaa ‘tomber’ demeure non agentif, que le sujet soit humain [Yawm. 16, 12] ou non humain [Yawm. 16, 20 et 32, 17].


joseph dichy

(a) Verbes à valeur d’état caractéristique ou d’acquisition d’état caractéristique – repris en partie des exemples ci-dessus : dāqa i ‘être ou devenir étroit’ (adjectif d’état caractéristique : dayyiq) ; lāna i ‘être ou devenir souple’ (sens physique), ou ‘tendre’ (sens psychologique affectant un sujet humain) (adj. layyin) ; jāda u ‘être ou devenir excellent’ (adj. jayyid) ; sāda u ‘être ou devenir le chef ’ (adj. sayyid) ; xaffa i ‘être ou devenir léger’ (adj. xafīf ) ; qalla i ‘être ou devenir peu nombreux’ (adj. qalīl) ; šadda i et u ‘être ou devenir intense, dur ou violent’ (adj. šadīd) ; samala u ‘devenir ou être rigide ou sec’ (adj. sāmil et samīl) ; fasada u et i ‘se corrompre’, ‘pourrir’, i.e. pour une viande, de la nourriture ou les mœurs d’une personne, etc. : ‘devenir’, puis ‘être corrompu’ (adj. fāsid, fasīd). (b) Verbes à diathèse interne ou moyens – repris des exemples ci-dessus : – Verbes associés dans le lexique à un adjectif ‘assimilé’ exprimant un état résultatif : latafa u bi- ‘se montrer gentil, prévenant envers qqun’ (adj. latīf ) ; daxala u fī ‘entrer dans’ un clan ou une tribu, ‘se placer sous sa protection’ (adj. daxīl) ; – Verbes sans adjectif ‘assimilé’ : amana u bi- ‘avoir foi en’ ou ‘dans’ ; makata u ‘rester, demeurer’ (agent humain) ; jalasa i ‘s’asseoir’ ; haraa a ‘se précipiter’ ; raā a ‘voir’ ; āda u ilā ‘revenir à’ ; fāza bi-, ‘gagner, obtenir qqch’ ; zahara a ‘apparaître’, ‘se manifester’ intentionnellement (agent humain) [Yawm., 32, 22] ; warada i ‘apparaître’ intentionnellement, ‘arriver’ (agent humain). (c) Verbes médio-passifs – repris des exemples ci-dessus : madā i ‘passer, se passer” (laps de temps, événement ou processus dont on prend en compte la durée) [Yawm., 32, 16] ; badā u ‘paraître’ (avec un sujet non humain) [Yawm., 32, 17] ; h āna i ‘se produire accidentellement’ (événement) [Yawm., 61, 15] ; ġaraba u ‘se coucher’ (soleil) ; māta u ‘mourir’ ; saqata a ‘tomber’, ‘chuter’ ; h awā i ‘contenir, inclure, comprendre’. (d) Verbes à diathèse externe (transitifs) : qatala u ‘tuer’ ; daraba i ‘frapper’ ; h asaba i ‘compter’ (des objets) ; madda u ‘tendre’, ‘étendre’ ; sah ara a ‘charmer, ‘ensorceler’ ; sajana u ‘emprisonner’ ; xadaa a ‘tromper, duper, leurrer’ ; fasala i ‘séparer, diviser’ ; qataa a ‘couper’ ; fatah a a ‘ouvrir’ ; bāa i ‘vendre’ ; qāla u ‘dire’ ; saala a ‘interroger’, ‘questionner’ ; talaba u min an . . . ‘de-

faula, faila, faala


mander à qqun de . . .’ ; sāh a i bi- ‘crier’ vers, en direction de qqun, ‘le héler’ [Yawm., 61, 11].33 4.3.1 Les corrélats formels de la présence de verbes à adjectif d’état caractéristique dans le schème faala Les verbes à adjectif d’état caractéristique (état caractéristique ou acquis) du schème faala, – dont on aurait attendu la présence parmi les faula –, sont associés à des corrélats formels. Ils relèvent : – soit des verbes ‘concaves’ (ajwaf, i.e. de 2e consonne radicale w ou y), pour lesquelles la réalisation de l’alternance vocalique u/u ne semble attestée que dans le cas de tāla/yatūlu – adj. tawīl – (bien que l’on puisse également voir dans cette forme un verbe en a/u) ; – soit des verbes ‘redoublés’ (mudaaf, de 2e et 3e consonnes radicales identiques), pour lesquels il n’existe qu’un tout petit nombre de verbes en faula, qui sont signalés comme des exceptions par Sībawayhi.34 Deux verbes ‘redoublés’ en u/u sont attestés en arabe contemporain (Hans Wehr-Cowan 1979) : šarra (1e pers. šarurtu) u ‘se monter’ ou ‘être mauvais, malfaisant’ (adj. šarīr) et labba (1e pers. labubtu) u ‘se montrer’ ou ‘être intelligent’ (adj. labīb – autre alternance vocalique : i/a). Ces deux verbes sont en outre donnés par le dictionnaire d’al-Fayrūz Ābādī, dans la glose de h abba (1e pers. h abubtu) u ilā ‘devenir un proche [qui reçoit des marques d’amitié] de’ qqun (adj. h abīb), comme étant, avec ce verbe, les seuls membres de cette classe formelle.35

33 Les verbes de parole du schème simple sont en arabe des verbes à diathèse externe. Le passage à une diathèse interne se fait au moyen de schèmes augmentés incluant ce que Roman (1990 ; 1999/2005) appelle le morphème-écho t – voir Ammar et Dichy (1999, 27–28) ; Dichy (2002/2003, § 4.2). Exemples : sala a ‘interroger’ [schème simple, verbe à diathèse externe] vs sāala an ‘interroger, questionner (qqun) sur’, puis tasāala an ‘s’interroger sur’ [schème augmenté incluant le morphème-écho, verbe à diathèse interne] ; qāla u ‘dire’ [schème simple, verbe de diathèse externe] vs qawwala ‘attribuer (des paroles)’ à qqun, puis taqawwala ‘prétendre, alléguer’ (au profit, dans l’intérêt, en faveur de soi-même) [schème augmenté incluant le morphème-écho, verbe de diathèse interne]. 34 Sībawayhi indique qu’il “n’y a presque pas de verbes en faula” de 2e et troisième radicale identiques (ou ‘redoublés’, mina t-tadīf – Kitāb IV : 36–37). Ibn Xālawayh (Laysa . . . : 27), signale l’absence de verbes ‘redoublés’ en faula, à l’exception des deux cas de labba u (déjà cité par Sībawayhi, loc. cit.) et azza u ‘avoir peu de lait’ (dit d’une chèvre). 35 Malgré l’indication citée, on trouve dans le dictionnaire d’al-Fayrūz Ābādī trois autres verbes : azza (3e pers. fém. sing. azuzat) u ‘voir (pour une chèvre) son lait diminuer’, ‘avoir peu de lait’ (adj. azūz), ainsi que damma (1e pers. damumtu) u ‘devenir’ ou ‘être très laid, repoussant’ (adj. damīm), et fakka (1e pers. fakuktu) u, ‘se montrer’ ou ‘être


joseph dichy

Toutefois, il ne s’agit pas d’une règle opérant dans les deux sens : tous les verbes ‘concaves’ ou ‘redoublés’ de schème faala ne relèvent pas, il s’en faut de beaucoup des verbes à adjectif d’état caractéristique. Les exemples de samala u ‘devenir ou être rigide ou sec’ et de fasada i et u – voir § 4.3, exemples (a) – montrent qu’il existe également un certain nombre de faala à adjectif d’état caractéristique qui ne relèvent d’aucune des deux catégories formelles que l’on vient d’indiquer. Leur sens de base est toutefois celui de l’acquisition d’un état caractéristique plutôt que celui de verbes d’état. 4.3.2 Un glissement sémantique partiellement responsable de la présence de verbes à diathèse interne et de verbes médio-passifs dans le schème faala Les verbes à diathèse externe sont agentifs et transitifs : le procès verbal ‘affecte’ un objet externe à l’agent. Un glissement sémantique très répandu et observable dans plusieurs langues peut toutefois modifier la diathèse de ces verbes, qui devient alors interne. Deux cas se présentent, selon que la valeur sémantique résultant de ce glissement est agentive ou non. Dans le premier, le sens produit est celui d’un verbe ‘moyen’ ou a diathèse interne : le verbe est agentif et le procès affecte l’agent. Dans le second, le sens produit par le glissement est médio-passif : le procès, qui est non agentif, affecte le sujet grammatical du verbe (l’agent existe, mais il n’est pas désigné). Comparons les données de plusieurs langues : – Verbes agentifs à diathèse interne, à partir de verbes à diathèse externe : – En français, le sens de base du verbe transitif plonger est celui d’un procès à diathèse externe dans lequel un agent plonge un objet dans un liquide, ex. : Pierre plongea sa chemise dans l’eau de la lessive. Ce verbe a pour pendant une forme pronominale exprimant une diathèse interne, se plonger, dont le sens est métaphorique, ex. : se plonger dans un livre (mais non *dans la piscine). En arabe, les sens de plonger et se plonger sont associés au verbe ġatasa i. Joüon donne un exemple similaire en hébreu : taval ‘tremper’ et ‘se tremper’ (1923, 95). La différence est dans la présence en français d’une marque morphosyntaxique, celle des verbes pronominaux. stupide’ ou ‘faible’, ces deux derniers verbes pouvant également être d’alternance i/a. Cf. aussi Ibn Xālawayh, Laysa . . . : 73 (note de l’auteur).

faula, faila, faala


Observons toutefois qu’en français, de manière en partie similaire à taval en hébreu et ġatasa en arabe, il existe, outre les deux sens que l’on vient de voir, une forme intransitive ‘moyenne’ plonger, de sens physique, sans marque morphosyntaxique associée.36 Tous ces exemples ont en commun d’être lexicalisés (la relation forme-sens ne résulte par d’un phénomène de dérivation compositionnel, mais est inscrite dans le lexique). – Le verbe à diathèse externe daraba i ‘frapper’, ‘battre’ – en anglais, ‘to hit’, ‘to beat’ – est cité dans tous les ouvrages, en langue arabe ou en langues occidentales, pour illustrer sens du schème faala, l’exemple mille fois reproduit étant : daraba zaydun amran ‘Zayd a frappé Amr’. Il comporte toutefois un sens dans lequel le procès exprimé par le verbe ne rencontre pas d’objet externe à l’agent, mais affecte ce dernier (le verbe est donc à diathèse interne), comme dans l’expression figée daraba fī l-ardi ‘parcourir la terre’ (ou une contrée, une région . . .). Un glissement sémantique analogue se produit en français dans les expressions battre les chemins (sens propre), ou battre la campagne (sens figuré). – Verbes médio-passifs (et donc non agentifs), à partir de verbes à diathèse externe : – Dans axada qalbuhu yadribu fī sadrihi ‘son cœur se mit à battre dans sa poitrine’, le verbe daraba prend une valeur non agentive ; le procès affecte le sujet grammatical. Il est significatif que le même glissement sémantique est observable en français avec battre et en anglais avec to beat dans les expressions correspondantes. – Le verbe à diathèse externe šaqqa u ‘fendre’, ‘briser’ (en anglais ‘to break’) prend également le sens de ‘éclore’ (bouton de fleur . . .), c’est-à-dire, ‘fendre sa gaine’, ou, pour une dent ‘pousser’. Ces deux sens sont associés, respectivement, aux ‘formes infinitives’ (masdar) šaqq et šuqūq. Comme dans tous les exemples ci-dessus, le sens médio-passif ou à diathèse interne dérive du sens transitif et à diathèse externe : il s’agit d’un phénomène de polysémie, inscrit dans le lexique. 36 Toutefois, à la différence des exemples empruntés à l’arabe et à l’hébreu, le sens physique de plonger – intransitif et ‘moyen’ – cité ici comporte en français un ajout sémantique (idée de ‘plongeon’), les données attachées à ce verbe étant relativement complexes. On a, par exemple, soit : plonger dans la piscine de son jardin, soit plonger les mains dans l’eau froide du bain, mais on ne peut plonger (= ‘faire un plongeon’) dans son bain . . .


joseph dichy

4.3.3 Les faala à diathèse interne ou médio-passifs ne correspondant ni à un corrélat formel ni à un glissement sémantique Il existe par ailleurs au sein du schème faala des verbes dont le sens est – ou en tout cas paraît être – originellement celui de procès à diathèse interne ou médio-passifs, et qui ne relèvent pas formellement des verbes ‘concaves’ ou ‘redoublés’. Exemples : – faala à diathèse interne : dahaba a ‘partir’, ‘aller’, ‘s’en aller’ (agent humain) ; mašā i ‘marcher’ ; daxala u ‘entrer’ (agent humain) ; akala u ‘manger’ (alors que šariba a ‘boire’ est, comme attendu, de schème faila) ; lamasa a ‘toucher’ et raā a ‘voir’ (mais samia a ‘entendre’ relève de faila) . . . – faala médio-passifs : zahaqa a ‘périr’, ‘s’évanouir, disparaître’ ; waqaa a ‘tomber’ ou ‘se trouver’, ‘être situé’ (pour un lieu, une ville, etc.) ; dahaba a ‘s’en aller’ (sujet grammatical non humain) ; fasada u et i ‘se corrompre’, ‘pourrir’ (adj. fāsid, fasīd). 4.3.4 Dispersion et régularité dans le schème faala Les exemples ci-dessus montrent qu’on trouve parmi les verbes en faala toutes les catégories et propriétés sémantiques associées aux deux autres schèmes. À la dispersion formelle présentée au début de ce travail semble donc répondre une dispersion sémantique équivalente, dont on constate à l’examen qu’elle ne semble pas pouvoir être réduite à des corrélats formels ou à des glissements sémantiques. Une grande régularité sémantique peut toutefois être observée : les verbes à diathèse externe apparaissent exclusivement dans le schème faala. Cette zone de stabilité est, selon toute vraisemblance, rendue possible par l’orientation particulière du glissement sémantique qui permet à un verbe à diathèse externe de devenir un verbe à diathèse interne ou un verbe médio-passif lorsque le procès verbal ‘revient’ sur l’agent ou sur le sujet grammatical, ou, selon la formule de Benveniste, quand le sujet devient “intérieur au procès” (1950/1966), exemples : – wadaa a ‘poser’, ‘déposer’ est un verbe à diathèse externe : l’agent effectue l’action d’imprimer à un objet le mouvement correspondant. Le même verbe prend le sens à diathèse interne de ‘accoucher’, qui peut être soit intransitif, soit transitif (Coran 3 : 36 et 46 : 15). L’agent devient, dans ce cas, le lieu du procès. De même, dans h amala i, le

faula, faila, faala


sens à diathèse externe ‘porter’ (un objet . . .) devient, avec une diathèse interne : ‘porter [un enfant]’, ‘devenir’ ou ‘être enceinte’. – Le verbe wasala i ‘joindre’, ‘unir’ une chose à une autre (avec la prép. bi-) est, dans son sens premier, à diathèse externe ; il est transitif à un double objet (de manière, respectivement, directe et indirecte). Au sens de ‘être uni’ à quelqu’un par un lien de parenté, d’amitié . . . et de ‘fréquenter assidûment’ quelqu’un, l’agent est en même temps l’un des deux objets du procès : le verbe est alors à diathèse interne et transitif (à un seul objet). Ces exemples, qui s’ajoutent à ceux qui ont été présentés au § 4.3.2, illustrent un processus qui opère dans un sens, le sujet grammatical ou l’agent devenant le lieu du procès (comme dans wadaa et h amala) ou l’objet de celui-ci (comme dans wasala). On peut en revanche formuler l’hypothèse que le glissement sémantique de la diathèse interne vers la diathèse externe n’est pas possible, compte-tenu des données de l’arabe : une fois le sujet ‘impliqué’ dans le procès, le verbe conserve cette propriété sémantique. Le passage de la diathèse interne (exemple : jalasa i ‘s’asseoir’) à une diathèse externe s’opère en construisant un autre verbe, de même racine, mais de schème différent, ainsi : ajlasa ‘faire asseoir’ (schème IV, af ala, de valeur causative-factitive). Le passage d’un verbe d’état caractéristique à un verbe à diathèse interne se produit également en changeant de schème, ainsi : latufa u ‘être ou devenir gentil’ ou ‘subtil’ est de schème faula ; latafa u ‘traiter quelqu’un avec gentillesse’, de schème faala, est à diathèse interne, la gentillesse affectant l’agent. Ces données d’observation permettent donc de formuler une hypothèse de portée descriptive.

5. Structure de la relation entre sens et forme dans les schèmes simples La dispersion de la relation entre sens et forme dans les schèmes du verbe simple n’est donc pas irréductible. Les raisons de cette dispersion ne sont pas, me semble-t-il, à chercher dans la ‘détérioration’ d’un ‘schéma originel’ ou d’un ‘proto-schéma’ qui entrerait en cohérence avec celui des relations formelles entre les deux paradigmes de la conjugaison objet de la figure 1. S’il paraît raisonnable de partir d’un proto-schéma


joseph dichy

présentant sur le mode hypothétique la relation entre sens et forme des schèmes du verbe simple dans son principe, il est essentiel de ne pas s’arrêter à cette hypothèse. Celle-ci est fondée sur l’idée que la relation sens l forme dans les schèmes serait essentiellement, voire originellement, de nature bijective. Or je montrerai plus loin que ce n’est guère le cas. 5.1 Le ‘proto-schéma’, à reconsidérer, des relation entre sens et forme dans les schèmes simples Le principe de ce proto-schéma associerait, de manière exclusive, la valeur sémantique des verbes à adjectif d’état caractéristique au schème faula, celle des verbes à diathèse interne ou médio-passifs à faila, et celle des verbes à diathèse externe à faala : Schème

Sens grammatical


Verbes à adjectif d’état caractéristique (état ou acquisition d’état)


Verbes à diathèse interne ou médio-passifs


Verbes à diathèse externe


Toutes les relations ci-dessus sont bijectives (ce qui est représenté par une flèche à deux têtes l). Il est peu vraisemblable qu’un tel système ait jamais existé en l’état, car, en sus d’inéluctables transformations morphophonologiques, des glissements sémantiques d’une catégorie à l’autre et divers effets de figement se sont nécessairement fait jour dès les toutes premières époques de la langue. Un tel glissement est déjà perceptible, dans la figure 3, dans le fait que les verbes à adjectif d’état caractéristique peuvent se réaliser comme des verbes d’acquisition d’état, qui sont médio-passifs. Or les verbes en faila peuvent également correspondre à des médio-passifs. C’est donc plus à une reconstruction hypothétique du principe qui gouverne ces relations qu’à une reconstitution diachronique que ce schéma nous convie. Ce principe est à reconsidérer de deux points de vue : il masque l’arbre conceptuel des notions qui le sous-tendent (§ 5.2) ; il est mis en défaut par les grandes lignes de glissements sémantiques ou de modifications formelles observables dans les schèmes simples (§ 5.3).

faula, faila, faala 5.2


L’arbre conceptuel des notions incluses dans cette structure

Conceptuellement, les grandes notions qui nous ont permis d’étudier les sens grammaticaux des schèmes simples ne sont pas à mettre sur le même plan : comme on l’a vu plus haut, les verbes à diathèse interne (ou moyens) et médio-passifs d’un côté, et les verbes à diathèse externe de l’autre sont dans une relation d’opposition. Or, les uns et les autres peuvent correspondre à un événement ou à un processus, mais non à un état – opposition fondamentale que masquait la distinction couramment répandue entre ‘verbes d’action’ et ‘verbes d’état’. L’ensemble constitué par les verbes à valeur d’état caractéristique s’oppose d’une part aux verbes médio-passifs – qui incluent les verbes d’acquisition d’état caractéristique ou non –, et aux verbes et à diathèse interne ou externe de l’autre. L’agentivité enfin opère elle même une partition entre les verbes à valeur d’état caractéristique, dans lesquels cette propriété est neutralisée, et les verbes médio-passifs, qui sont non-agentifs. Par contraste, les verbes à diathèse interne ou externe sont, selon le cas, entièrement ou partiellement agentifs. Ces oppositions sont résumées dans l’arbre conceptuel suivant : notion aspectuelle d’état (agentivité neutralisée)

Verbes à valeur d’état caractéristique


notions aspectuelles d’événement ou de processus non agentivité

Verbes médio-passifs ( y compris d’acquisition d’état)

agentivité entière ou partielle diathèse interne

diathèse externe

Verbes à diathèse interne

Verbes à diathèse externe


Conventions : Les propriétés sémantiques sont en caractères gras ; les catégories sémantiques des verbes du schème simple sont en italiques. On notera qu’en raison de phénomènes de polysémie, un même verbe peut appartenir à des catégories sémantiques différentes : les propriétés ci-dessus peuvent donc correspondre soit à des catégories lexicales, soit à des occurrences en contexte.


joseph dichy

5.3 Glissements sémantiques à l’œuvre dans la relation sens-forme dans les schèmes simples Si l’on part de l’aspect formel des schèmes, on observe que : (a) tous les verbes en faula correspondent à un verbe à adjectif d’état caractéristique (valeurs d’état caractéristique ou d’acquisition de celui-ci) ; (b) presque tous les verbes en faila relèvent de l’une ou l’autre des valeurs du verbe à diathèse interne ou du médio-passif. Ce schème inclut par ailleurs un sous-ensemble, réduit en nombre, de verbes à adjectif d’état caractéristique ; (c) les verbes en faala, cependant, n’appartiennent que partiellement aux verbes à diathèse externe. Si l’on part au contraire du sens grammatical, on observe que : (d) tous les verbes à diathèse externe sont inclus dans le schème faala ; (e) les verbes à adjectif d’état caractéristique sont en contrepartie représentés dans les trois schèmes ; (f) les verbes moyens et médio-passifs sont répartis entre les schèmes faila et faala (dans le schème faula, les verbes d’acquisition d’état sont médio-passifs). Les deux relations (a) et (d) sont de la forme ‘tous les . . . sont . . .’ ; la relation (b) est du type ‘presque tous les . . . sont . . .’. Ce qui précède peut être représenté par la figure suivante : Schème faula faila faala FIGURE 5

Sens grammatical • Verbes à adjectif d’état caractéristique (état ou acquisition d’état) • Verbes à diathèse interne ou médio-passifs • Verbes à diathèse externe LES RELATIONS ENTRE SENS ET FORME DANS LE VERBE SIMPLE

faula, faila, faala


Conventions : – Les flèches dont le trait est doublé ( ) indiquent une relation mettant en jeu la totalité des éléments de l’ensemble considéré. La convention se lit : pour tout schème ou sens grammatical situé au début de la flèche, le sens ou le schème est celui qui est indiqué par la pointe de la flèche (cf. section 4.1). – La flèche interrompue de trait doublé ( ) indique une relation mettant en jeu la quasi totalité des éléments de l’ensemble considéré, avec toutefois un ensemble d’exceptions clairement identifiable. La flèche – il n’y en a qu’une – se lit : pour presque tout schème situé au début de la flèche, le sens grammatical est celui qui est indiqué par la pointe de la flèche (voir section 4.2). – Les flèches en pointillés ( ) indiquent une relation qui ne concerne qu’un nombre réduit et fermé de verbes, dans l’inventaire duquel apparaissent des contraintes morphologiques, comme par exemple dans les verbes d’état caractéristique du schème faala (§ 4.2.1 et 4.3.1). – La flèche en trait plein ( ) signale une relation entre sens grammatical et schème qui concerne un nombre important et en tout état de cause ouvert de verbes. Ces verbes sont produits par un processus identifiable de glissement sémantique (§ 4.3.2 et 4.3.3).

6. Le décrochage partiel entre sens et forme : une relation non bijective Le décrochage partiel entre sens et forme illustré ci-dessus est dû au jeu de deux types de contraintes : formelles (incluses dans les flèches transversales en pointillé) et sémantiques (flèche transversale en trait plein). Pour comprendre la structure générale à laquelle on est confronté, il faut tenir compte du fait que la relation entre les schèmes et leur sens grammatical n’obéit par à une relation bijective. On rencontre en effet, soit des relations schème o sens grammatical, comme dans faula et faila, soit une relation schème m sens grammatical, comme dans faala, mais non la relation bijective (à deux sens) schème l sens grammatical, comme le laisserait croire une conception naïve du langage. Un tel décrochage est fréquent dans les langues : on le rencontre, typiquement, dans la relation entre les actes de langage (Searle) et leurs


joseph dichy

réalisations dans une langue donnée. En reconnaissance, il est également très fréquent dans les valeurs aspectuo-temporelles associées en contexte à une forme verbale : une même forme supportant plusieurs interprétations, c’est la mise en rapport de celle-ci avec divers indices contextuels qui permet au processus de compréhension d’opérer (Desclés et al. 1998 ; Desclés et Guentchéva, à paraître). Le repérage d’indices est particulièrement important dans le processus de lecture en arabe, l’écriture courante étant dépourvue des signes diacritiques secondaires dits de “vocalisation”. Dans la constitution du lexique arabe, la relation entre les sens grammaticaux associés à un schème donné et les sens lexicaux qui s’adjoignent à un nom ou à un verbe est soumise à l’effet du “principe de figement lexical” (PFL), qui soumet le sens grammatical des schèmes, dès le niveau du mot, à des écarts sémantiques (Dichy 2003, 204–208). Dans le verbe simple, des glissements sémantiques tels que ceux qui s’instaurent entre les valeurs de verbe d’état (‘être x’) et d’acquisition d’état (‘devenir x’) ou celui qui permet à certains verbes à diathèse externe de prendre une valeur de diathèse interne, soumettent les relations entre le sens grammatical des schèmes et leur forme à de fortes pressions. Le ‘protoschéma’ de ces relations (figure 3) persiste toutefois dans le schéma des relations observables (figure 5), mais avec des connexions entre sens et forme qui, n’étant pas bijectives, s’inscrivent dans la structure générale des rapports entre le sens et la forme que l’on observe dans le processus de compréhension du langage humain.

7. Références bibliographiques 7.1 Sources primaires Ibn Xālawayh, al-H usayn b. Ahmad (m. en 370/981). Laysa fī kalām al-arab, éd. Ahmad A. At tā r. Beyrouth : Dār al-Ilm li-l-malāyīn, 1979. Ibn Yaīš (m. en 643/1245). Šarh al-Mulūkī fī t-tasrīf, éd. Faxr ad-Dīn Qabbāwa. Alep : Al-Maktaba l-arabiyya, 1973. ——. Šarh al-Mufassa l. Le Caire : Maktabat al-Mutanabbī / Beyrouth : Ālam al-kutub, 10 tomes en 2 vols. al-Mubarrid (ou al-Mubarrad), Abū l-Abbās (m. en 285/898). Al-Muqtadab, éd. Muhammad A. Udayma. Le Caire : Al-Ahrām, 1399/1978, 4 vols. Sībawayhi (m. v. 180/796). Al-Kitāb, éd. Abd as-Salām M. Hārūn. Le Caire : Al-Haya l-misriyya l-āmma li-l-kitāb, 1977, 5 vols. Yawm. = Tawfīq al-H akīm. 1937. Yawmiyyāt nāib fī l-aryāf. Le Caire : Dār Misr li-t tibāa. az-Zajjājī, Abū l-Qāsim (m. v. 340/952). Al-Jumal fī nahw, éd. Alī T. al-H amad. Beyrouth : Muassasat ar-risāla, 1988 (4e éd.). az-Zamaxšarī, Abū l-Qāsim. Al-Mufassa l fī ilm al-arabiyya. Beyrouth : Dār al-Jīl, s.d.

faula, faila, faala 7.2


Sources secondaires

Abbès, Ramzi. 2004. La conception et la réalisation d’un concordancier électronique pour l’arabe. Thèse de doctorat en sciences de l’information, Lyon, ENSSIB/INSA. Abraham, Maryvonne. 1995. Représentations sémantico-cognitives des verbes (échantillon des verbes de mouvement) : problèmes et méthode. Thèse d’informatique, Paris : EHESS. Ammar, Sam et Joseph Dichy. 1999. Les verbes arabes. Paris : Hatier (coll. Bescherelle). Audebert, Claude. 2002. “Verbes actifs et moyens dans le parler du Caire : une suite.” Annales islamologiques 36, Le Caire, I.F.A.O. Badawi, Elsaid, Michael G. Carter, and Adrian Gully. 2004. Modern Written Arabic. A comprehensive Grammar. London and New York : Routledge. Belot, J. B. 1922. Cours pratique de langue arabe. Beyrouth : Imprimerie catholique. Benveniste, Émile. 1950/1966. “Actif et moyen dans le verbe,” in : Problèmes de linguistique générale. Paris, Gallimard, t. 1 (1966), 1950, repris in 1966, ch. XIV, 168–175. Blachère, Régis et Maurice Godefroy-Demombynes.1952. Grammaire de l’arabe classique. Paris : Maisonneuve (3e éd.). Boormans, Maurice. 1967. Grammaire d’arabe littéral. Feuilles de travail. Rome : Institut pontifical d’études arabes (document miméographié). Brockelmann, Carl. 1948. Arabische Grammatik. Leipzig : O. Harrasowitz. Cantineau, Jean. 1950. “La notion de ‘schème’ et son altération dans diverses langues sémitiques.” Semitica III. 73–83. Caspari, C.P. 1881. Grammaire arabe. Trad. E. Uricoechéa. Paris : Maisonneuve. Cohen, Marcel. 1929. “Verbes déponents internes (ou verbes adhérents) en sémitique.” Mémoires de la Société de Linguistique de Paris XXIII : 4, 225–248. Comrie, Bernard. 1976. Aspect. An Introduction to the study of verbal aspect and related problems. Cambridge University Press. ——. 1985. Tense. Cambridge University Press. Desclés, Jean-Pierre. 1990. Langages applicatifs, langues naturelles et cognition. Paris : Hermès. ——. 1994. “Quelques concepts relatifs au temps et à l’aspect pour l’analyse des textes.” Studia kognitywne 1, SOW, Varsovie. 57–88. ——, Valérie Flageul, Christiane Kekenbosch, Jean-Marc Meunier et Jean-François Richard. 1998. “Sémantique cognitive de l’action : 1. Contexte théorique. 2. Étude expérimentale de la catégorisation des verbes d’action.” Langages 132 (Cognition, catégorisation, langage). 28–47 ; 48–68. —— et Zlatka Guentchéva. À paraître. Aspectualité, temporalité : une approche cognitive et formelle à partir des langues. Document miméographié. Dichy, Joseph. 1993. “Knowledge-system simulation and the computer-aided learning of Arabic verb-form synthesis and analysis.” Processing Arabic Report, 6/7. T.C.M.O., Université de Nimègue. 67–84 ; 93–95. ——. 2002/2003. Structure de la dérivation lexicale en arabe : sens et forme des verbes et des dérivés nominaux les plus immédiats, Cours de préparation au CAPES d’arabe, session 2003, question de linguistique. Paris : C.N.E.D. ——. 2003. “Sens des schèmes et sens des racines en arabe : le principe de figement lexical (PFL) et ses effets sur le lexique d’une langue sémitique.” In Sylvianne Rémi-Giraud et Louis Panier, dirs. La polysémie ou l’empire des sens. Lexique, discours, représentations. Presses universitaires de Lyon. 189–211 (www.concours-arabe.paris4.sorbonne. fr/cours/dichy.doc) —— et Mohamed Hassoun. 2005. “The DIINAR.1 – « ! » Arabic Lexical Resource, an outline of contents and methodology.” The ELRA Newsletter, vol. 10, 2, April-June 2005. 5–10. ——, Abdelfattah Braham, Salem Ghazali et Mohamed Hassoun. 2002. “La Base de connaissances linguistique DIINAR.1, (DIctionnaire INformatisé de l’ARabe – version 1).” In A. Braham, éd. Actes du colloque international sur Le Traitement automatique de l’arabe. Tunis : Université de La Manouba. 45–56.


joseph dichy

Fleisch, Henri. 1957. “Études sur le verbe arabe.” Mélanges Louis Massignon. Institut Français de Damas. 153–181. ——. 1968. L’arabe classique. Esquisse d’une structure linguistique. Beyrouth : Dār alMachreq. ——.1979. Traité de philologie arabe vol. 2. Beyrouth : Dār al-Machreq. Ġalāyīnī, Musta fā. 1912. Jāmi ad-durūs al-arabiyya. 19e éd. revue par Muhammad A. an-Nādirī. Beyrouth, al-Maktaba l-Asriyya, 1994, 3 vols. en un. Guillaume, Jean-Patrick (1984), “Quelques aspects de la théorie morpho-phonologique d’Ibn Jinnī. À propos des verbes à glide médian.” In G. Bohas et J.-P. Guillaume, éds. Étude des théories des grammairiens arabes. I. Morphologie et phonologie. Institut français de Damas. 338–490. Ilyās, Jūzīf et Jirjis Nāsīf. Mujam ayn al-fil. Beyrouth : Dār al-Ilm li-l-malāyīn, 1995. Joüon, Paul. 1923. Grammaire de l’hébreu biblique. Rome : Institut pontifical, rééd. 1965. ——. 1930. “Sémantique des verbes statifs de la forme qatila (qatel) en arabe, hébreu et araméen”. Mélanges de l’Université Saint Joseph, XV : 1, 3–32. Larcher Pierre. 1995. “Où il est montré qu’en arabe classique la racine n’a pas de sens et qu’il n’y a pas de sens à dériver d’elle.” Arabica XLII. 291–314. ——. 1996. “Dérivation lexicale et relation au passif en arabe classique.” Journal asiatique 284, 2. 265–290. ——. 2003. Le système verbal de l’arabe classique. Aix-en-Provence : Publications de l’Université de Provence. Leeman-Bouix, Danielle. 1994. Grammaire du verbe français, des formes au sens. Paris : Nathan. Lyons, John. 1978/1990. Sémantique linguistique, trad. franç. de la 3e éd., revue et corrigée de Semantics II, Cambridge University Press (1978). Paris : Larousse, 1990. Maingueneau, Dominique. 1994. L’énonciation en linguistique française. Paris : Hachette. Moscati, Sabatino, éd. 1964. An Introduction to the Comparative Grammar of the Semitic Languages. Phonology and Morphology. Wiesbaden : O. Harrassowitz (2e éd., 1969). Al-Mujam al-Wasīt. 1973. Ibrāhīm Anīs, Abdalhalīm Muntasir, Atiyya as-Sawālihī, Muhammad X. Ahmad, eds. Le Caire : Dār al-Maārif, 2e éd. (1e éd. 1960). Al-Munjid fī l-luġa l-arabiyya l-muāsira. 2000. Subhi H amwī, éd. Beyrouth : Dār alMašriq. Neyreneuf Michel et Ghalib Al-Hakkak. 1996. Grammaire active de l’arabe. Paris : Livre de Poche. Nūr ad-Dīn, Isām. 2002. Abniyat al-fil fī Šāfiyat Ibn al-H ājib. Beyrouth : Dār al-Fikr al-lubnānī (1e éd. 1982). Qabbāwa, Faxr ad-Dīn. 1998. Tasrīf al-asmā wa-l-af āl. Beyrouth : Maktabat al-Macārif, 3e éd. Roman, André. 1983. Étude de la phonologie et de la morphologie de la koinè arabe. Marseille : Jeanne Laffitte, 2 vols. ——. 1990. Grammaire de larabe. Paris : P.U.F. (coll. “Que sais-je ?”). ——. 1999/2005. La création lexicale en arabe, ressources et limites de la nomination dans une langue humaine naturelle. Presses Universitaires de Lyon, 1999. 2e éd. revue et augmentée, Presses Universitaires de Lyon et Université de Kaslik, 2005. aš-Šartūnī, Rašīd. 1912. Mabādi al-arabiyya. Beyrouth : Imprimerie catholique (1e éd.). Sylvestre de Sacy, Antoine. 1831. Grammaire arabe à l’usage des élèves de l’École spéciale des langues orientales vivantes. Paris : Imprimerie royale (2e éd.). Réimpr. photomécanique. Paris : Institut du Monde arabe, 2 vols. s.d. Versteegh, Kees. 1997/2003. The Arabic Language. New York : Columbia University Press, 1997. Trad. arabe : Al-luġa al-arabiyya, tārīxuhā wa-mustawayātuhā wa-tatīruhā, trad. Muhammad aš-Šarqāwī. Le Caire : Al-Majlis al-alā li-t- taqāfa, 2003. ——. 2004. “Meanings of speech. The category of sentential mood in Arabic grammar.” In Joseph Dichy et Hassan Hamzé, éds. Le voyage et la langue, Mélanges en l’hon-

faula, faila, faala


neur d’Anouar Louca et André Roman. Damas : Institut Français du Proche-Orient, 269–287. Wehr, Hans, John M. Cowan (ed). 1979. A Dictionary of Modern Written Arabic (ArabicEnglish), edited by J. M. Cowan, 4th edition. Wiesbaden : O. Harrassowtiz. Wright, W. 1896–98. A Grammar of the Arabic Language. Réimpr. Beyrouth : Librairie du Liban, 1974.


1. Introduction Featuring is the affixation of, general as well as language dependent, ‘second-level’ labels (such as gender1 or ±animate) to ‘first-level’ nonterminals (such as noun, verb or particle) in a formal description of the syntax and semantics of a specific natural language. Such a formal grammar may serve as input for the automated processing of the natural language (NLP) concerned: Modern Standard Arabic. About 98% of Modern Standard Arabic texts, whether in printed or electronic form, are represented in a non-vocalized shorthand form. The analysis of this kind of data, automated or otherwise, has to cope with an exponential combinatorial ambiguity unless: – one considers the units of linguistic description beyond word level such as constituent, sentence, paragraph or even text-level; – one combines this description with an adequate, coherent, consistent, and as exhaustive as possible, featuring system for analysis purposes. Therefore, our objective is: the design of a featuring system as a tool for the disambiguation2 of undesired analysis results in the automated processing of Arabic text data.3

1 Feature names and feature values are represented in the text in Italics. We use two types of features: inherent and inherited features. The former represents the intrinsic semantic value of the entry concerned. The latter follows as modifier the entry it modifies according to language dependent concord and agreement rules. 2 We use the term ‘disambiguation’ for attempts to obtain a single and most probable syntactically and semantically correct analysis of input data in an automated processing environment. 3 Language phenomena such as homonymy, polysemy and even antinomy should also be controlled by means of adequate featuring.


everhard ditters

A comprehensive inventory of general or universal linguistic features is not (yet) available for Modern Standard Arabic. This is mainly due to the complexity of the description of semantic features. Usually, linguistic theories and descriptive language models tend to become language independent. However, we prefer—while exploiting current, language independent linguistic theories and descriptive models—to describe the features that are specific for the language under consideration. If the ‘linguistic model chosen for the description’ may have some impact on the organization of the featuring system, it remains possible to use heuristics and to refine the formal description any time we consider it worthwhile in order to avoid undesired ambiguities. So we are free to insert or to combine features such as agreement and concord,4 in order to account for regularities, relationships and dependencies occurring between different constituents, or between different elements within a constituent. Considering Modern Standard Arabic language description, we distinguish three layers of featuring: a morphological, a syntactic and a semantic layer. They describe the form, the function and the meaning of elements used in a, agreed upon by convention and expandable by language evolution, system of information interchange, tailored to the needs of users of any natural language system. This tri-partition roughly coincides with traditional classifications into orthography, phonology, morphology, syntax, stylistics, discourse, and even semantics.5 A description of the Arabic phonemes for analysis purposes does not require a differentiation in terms of their production characteristics. A listing of phonemes (26 consonants, three glides [semi-consonants or semi-vowels],6 and three vowels [in a short and long variety]), together with a number of graphemic alternatives, constitutes the basis for our phonemic description of Modern Standard Arabic.

4 We distinguish between concord and agreement and reserve the latter to describe regular patterns in relationship between an explicit agent and its predicate in a verbal sentence (Sv) involving gender (Ditters 1992, 169, n. 13; Kihm 2006, 14–15; cf. also: Bahloul 2006, 43–48). We use the term concord for feature-value sharing within the noun phrase (NP) between the head and its modifiers, as well as in the nominal sentence (Sn) for the marked relation between the topic and comment, involving, if applicable, definiteness, gender, number, and case. 5 We are definitely not trying to introduce a formalized dynamic description of the Arabic concept of the world. We rather follow a static semantic approach using finite enumeration of pertinent (static) semantic features. See also subsection 4 below. 6 For coherency within our description, we adopt Sībawayhi’s inventory of 29 Arabic consonants (Hārūn 1982, 4, 431).

featuring as a disambiguation tool in arabic nlp


The introduction of a feature such as verbtype may well account for alternative realizations of the finite forms of so-called ‘weak’ verbs. A feature like complement structure7 controls the number (argument-1 [arg-1], [arg-2], [arg-3] etc.) as well as the format (noun phrase (NP), prepositional phrase (PP), complement clause (CCL) etc.) of possible complements of a lexical verbal entry. By means of predicate-subject and predicate-object matching features we are in a position to considerably refine our formal description and to distinguish, for example, between ordinary and metaphorical language use. However, for the disambiguation of analysis results our main task at this moment remains: to draft a general pattern for nominal and verbal featuring; and to compose a basic set of (semantic) features and feature-clusters embedded in an adequate description of verbal complement structure. A remark about ‘the linguistic model chosen for the description’ may be useful. We still opt for a Phrase Structure Grammar (psg) model based on immediate constituency (ic), but enlarged with a second level of description to account for prevailing relationships and dependencies within a Modern Standard Arabic sentence.8 We even foresee the introduction of a third level of description, grouping together pertinent semantic features with information about their characteristic form of realization at syntax level. Inspired by Saad’s (1982) syntactic-semantic study of the Classical Arabic verb in Case Grammar (cg) terms, we will use a number of Fillmore’s (1968) cg ideas as well as later developments for the sub-categorization and patterning of verbs in Modern Standard Arabic.9 Finally some words about the processing environment. We use the agfl-system,10 a parser generator for research and applications in 7 To be elaborated upon later at syntax level when speaking about: tri-, bi-, mono-, and intransitive verbs or the reduction to a one-argument realization used to uniquely emphasize the semantic load of the verbal entry ; and at the semantic level while speaking about: semantic feature hierarchy; the minimal and the maximal realization of complement structure; compulsory and optional complement structure realization; case roles; and verb sub-categorization. 8 For other applications such as information retrieval (IR) or text-summarizing (TS), we use a Dependency Grammar (DG) approach since it appears to express more adequately semantic properties in terms of nodes and relations between nodes (Ditters and Koster 2004). 9 Only recently we decided to include al-Saffār (1979) in our research for his semantico-syntactic features in Case-Grammar terms. 10 For more information about the AGFL formalism, now in its 2.4th Windows version, see Koster (1991) and For more information about the formal description of Arabic, see Ditters 1992; 2003a; 2003b; and forthcoming.


everhard ditters

nlp. The acronym refers to Affix Grammars over Finite Lattices.11 The parsers generated by the system are top-down recursive backup parsers (Koster 1974, Meijer 1986), based on non-deterministic concepts and the unification-principle (Nederhoff 1993). The agfl-formalism is part of the family of two-level grammars: a context-free grammar is augmented with set-valued features for expressing agreement between syntactic categories. The formalism is, in principle, suited for describing morphological and syntactic structure, as well as finite semantics in terms of ±animate, ±concrete, ±human, ±volition and many others. We discuss feature-sets at morphologic (§ 2), syntactic (§ 3), and semantic level (§ 4). We end with a conclusion (§ 5) and a list of references used (§ 6).12

2. Morphology Strictly speaking, morphology relates to the formal description of the individual Arabic parts of speech: verb, noun and particle, and, if needed or useful, of their differentiation into a further sub-categorization.13 In a broader sense, morphology also comprises feature names and values such as: aspect, case, definiteness, derivation, gender, number, person, tense, voice, and many other features used at syntax level. Morphology, finally, has to account for punctuation marks and other textual ‘noise’, normally more conveniently stored in the lexicon module. Literary Arabic has been described as a predominantly root and pattern language type.14 Moreover, vowel-pattern variation combined

11 Here, the term affix, a variable with a finite set of values, has to be taken in its formal and not in its linguistic sense. 12 Within the morphological, syntactic and semantic sections, we use the traditional Arabic language parts-of-speech (POS) differentiation into noun, verb and particle as headings for the subsections. In these subsections we only discuss the, for us, relevant features. 13 Cf. Sībawayhi’s tri-partition (Hārūn 1982, 1:12) and the subdivision of POS into 7 classes by alSāqī (1977, 214). 14 Cohen (1970, 49 ff.) has been one of the first to describe, for automated Arabic language processing, a frame of mostly three, sometimes four and rarely five, consonants or semi-consonants filled with a combination of vowels (including the absence of a vowel at a certain slot) expressing semantic differentiations to the global meaning of the consonantal root combination. Elements of a small subset of the phoneme inventory are used to produce other derivates of the base frame, whether of the category ‘verb,’ ‘noun’ or ‘adjective,’ with their own specific variation on the global meaning of the consonantal root combination.

featuring as a disambiguation tool in arabic nlp


with the insertion of a finite set of auxiliary phonemes triggers a more or less ‘regular’, more or less ‘predictable’ change in the semantic load of the base root meaning. The Arabic phoneme inventory has extensively been described elsewhere.15 As far as the consonants are concerned, the description comprises a cluster of features referring to their way and place of production, as well as to their marking for velarization and degree of voicing. As far as the vowels are concerned, they have been described by their place and the openness of that place, set out against the position of the tongue in their production process. For analysis purposes, we have the Arabic grapheme inventory with variant scriptures for alternatives occurring in initial, median, end or independent position. In order to account for fully, partially or nonvocalized Arabic input data, the listing of the Arabic vowel graphemes will include an empty vowel realization. 2.1


Assigning morphological features and feature values to verbs, we account for: the realization of finite verb occurrences on the basis of number and place of the root consonants (radical, r1,16 r2, r3, and less frequent: r4 and r5); the realization of the vowel (a, i, u) of the second consonant (C2) of the base (Stem I) verb realization in perfect and mood aspect (vowperf, vowimperf ); and the realization of prefixed or infixed vowels (vow or sukūn). Morphological features, attached to elements of a finite verb realization at syntactic level (see also § 3.1 below), affect values concerning: aspect, tense,17 voice, person, gender and number. For the description of a finite 15

Cf. Fleisch mainly used three, sometimes four, features to define the elements of the consonantal system, which ends up in a 28 u 4 matrix (1961, 1:56–65) or a 16 u 5 matrix (1968, 19); Versteegh (2001, 20) used a 9 u 6 matrix; Saad (1982, 6) used a 7 u 9 matrix for the consonants and a 4 u 2 matrix for the vowels. See also the phoneme featuring of Bohas and Saguer (2007, 255 ff.) in this volume. 16 A listing of the Arabic phonemic (consonant and vowel) system is given in the lexical module(s). For analysis purposes, a single description of the r1 suffices. For generation purposes, features describing occurrence incompatibilities between an r1, r2 and r3 (r4, r5) should be provided for. 17 As Badawi et al. (2004, 362 ff.) do, we distinguish between a perfect and a nonperfect (mood) value of the feature aspect of the verb. In combination with modal and temporal verbs and/or adverbials a complete range of temporal and aspectual differentiations (±tenses) can be described (see on aspect and tense also Eisele 2006, 195–201; Bahloul 2006, 506; and Reese 2006, 50–53).


everhard ditters

verb form of the so-called ‘weak’ verbs with alternative realizations in its conjugation, a variable verbtype accounts for alternative verb realizations. To account for the differentiation in base and derived stems we use the feature derivation. With that we can list, in the form of a context-free metagrammar, the following feature names or ‘non-terminal affixes’ (in capital letters) with their finite-set of values or ‘terminal affixes’ (in lower case), connected to a finite verb form:18 ASPECT MOOD TENSE VOICE PERSON GENDER NUMBER VERBTYPE

:: :: :: :: :: :: :: ::


:: :: :: :: :: :: :: :: :: :: :: ::

perfect; MOOD. indicative; subjunctive; jussive; imperative.19 temporal; modal. active; passive. first; second; third. feminine; masculine. singular; dual; plural. r1r2r3; r2=r3; r1=w; r1=y; r2=a; r2=w; r2=y; r3=w; r3=y; r1=w, r3=y; r2=w, r3=y; r1=hamza; r2=hamza; r3=hamza. base; DERIVED. ii ; iii ; iv ; v; vi; vii; viii; ix; x. three; four; five. alphabet. R1. R1. R1. R1. vowel; sukūn. VOW. VOW. VOW.

18 We will use the following conventions for the formal description of features and the finite-set of values at the second level of description which closely follow the AGFL convention: – feature names are written in upper case; – feature values are written in lower case; – the rewrite symbol is a double colon ‘::’; – a single left-hand entry is rewritten in one ore more feature names and/or feature values at the right-hand side; – alternative realizations at right-hand side are separated by a ‘semicolon’; – options in the right-hand side are separated by a ‘vertical bar’; – a rule will be closed by a period ‘.’. 19 The archaic energetic-1 and energetic-2 finite verb forms are not accounted for in our formal description of Modern Standard Arabic.

featuring as a disambiguation tool in arabic nlp 2.2



Features, at the morphological level assigned to elements of the category noun, concern variables such as: person, gender and number.20 For the description of nominal plural building, it is worthwhile to differentiate between an external and internal formation. Besides person, gender and number, the feature case is a pertinent part of the declension system, in Modern Standard Arabic closely connected with the ‘type’ of declension. For the description of a differentiation in base and derived stems for deverbal nouns (active and passive participles, infinitives, nouns of time, place, doing an action once, or referring to instruments used in an occurrence), we use the feature derivation. By means of morphological base rules, we describe the subset denominatives of the category noun, such as: individuality nouns, multitude nouns, vessel nouns, relation nouns, abstract quality nouns and diminutives.21 The feature variable determination with values as definite, indefinite or morphologically ‘neutral’ is another inherent or inherited (at syntax level) characteristic of elements of the category noun. Elements of the sub-classes demonstratives, personal pronouns, proper nouns, and relative pronouns are inherently definite. Elements of the sub-classes indefinite and interrogative pronouns definitely possess an inherent indefinite value. As far as the ‘definite article’ is concerned, we follow Wright’s interpretation of the Arabic grammatical tradition in describing this article on the basis of its deictic and not of its generic value as a sub-class of the demonstratives.22 Furthermore, elements of the sub-classes adjectives, common nouns, elatives, numerals, and quantifiers may, or may not, receive a definite or an indefinite value depending on their function and occurrence in phrases at syntax level. So we list the following morphological feature rules for a noun: CASE DECLENSION

:: nominative; NOMINATIVE.23 :: invariable; diptote; triptote.

20 Since identical looking feature values are attached to distinct non-terminal names (different category labels), we can reuse these feature names without any risk for undesired ambiguities. 21 Cf. Wright 1974, 1:109–177. 22 Cf. Wright 1974, 1:264–270. See also Fleisch 1961, 1:339–347. 23 The AGFL formalism allows for the use of logical markers such as ‘+’, ‘-’, and others in combination with feature names and values.




:: individuality; multitude; vessel; relation; quality; diminutive. :: base; DERIVED. :: ii; iii; iv; v; vi; vii; viii; ix; x. :: definite; indefinite. :: actpart; paspart; infinitive; time; place; once; instrument. :: feminine; masculine. :: genitive; accusative. :: singular; dual; PLURAL. :: first; second; third. :: external; internal.


According to the traditional description24 of particles we may distinguish between: adverbs, conjunctions, interjections, and prepositions.25 A formal description should include: PARTICLE TYPE




A for us still useful sub-categorization of adverbs, at morphological level, is that in invariable, as far as case marking is concerned, and bound and free forms, as far as the orthographic representation is concerned. ADVERB


:: bound; free; invariable.


As far as the coordinating particles are concerned, a distinction should be made in cumulative and selective particles because of the importance of number value in concord and agreement phenomena at syntax level. We then divide the selective particles into: alternative, consecutive, explicative, exclusive, inclusive, restrictive, and successive elements.26 This sub-categorization will be extensively used at the next higher level of description. 24

Cf. Wright, 1974, 1:282–296. CF. Badawi et al. 2004, 174–219; Cantarino, 1974–5, 2:253 ff.; El-Ayoubi et al. 2003, 1:2, 275–460. 26 Cf. Ditters 1992, 222–228. 25

featuring as a disambiguation tool in arabic nlp


For the conditional particles we prepare their application at syntax level with a differentiation as to the nature of the condition in: possible, real and unreal. CONJUNCTION COORDINATING SELECTIVE CONDITIONAL


:: COORDINATING; CONDITIONAL. :: cumulative; SELECTIVE. :: alternative; consecutive; exclusive; explicative; inclusive; restrictive; successive. :: possible; real; unreal.


Elements of this sub-section of particles, rather close to discourse and formalized language use, can, as far the morphological level is concerned, be divided into vocatives and exclamations.27 However, we also have to decide whether or not ‘proverbs’ and ‘frozen expressions’ should be lodged among the interjections and labeled as adverbs or adverbials at syntax level. In any case we also like to mention here the category of formulaic greetings.28 INTERJECTION


:: vocative; exclamation; f-expressions; proverbs; greetings.


The standard differentiation of prepositions at this level is into a primary and a secondary group. The secondary group consists of an open set of noun-derived entries, marked for their function by means of a definite accusative case value and, at syntax level, engaged in a ‘construct state’-like link with the prepositional complement, itself marked with genitive case value (NPgen). The primary group comprises a finite-set of non-derived entries, some of which are, in their orthographic representation, directly bound to the prepositional complement while others are unbound.29 PREPOSITION PRIMARY

:: PRIMARY; secondary. :: bound; unbound.

27 A still poorly described domain of ‘frozen’ or ‘set’ expressions like: greetings, insults, proverbs and similar insertions should be included here. Cf. Bergman 2007, 136–137. 28 Cf. Elzeiny 2007, 202–207. 29 Cf. El-Ayoubi et al., 2003, I:2:574–592.


everhard ditters 3. Syntax

At syntax level, elements of the earlier mentioned word categories and sub-classes are assembled into constituents, labeled: verb phrase (vp), noun phrase (np), and particle phrase, sub-categorized according to the element realizing the compulsory head function.30 These phrases realize specific compulsory and optional functions within the two main units of linguistic description at sentence level: the nominal (sn) and the verbal (sv) sentence.31 It may be clear, that the combinatorial behavior of elements within a constituent, and that of constituents at sentence level, comprising mutual relationships and dependencies, is controlled by features and feature-condition rules. In Modern Standard Arabic text data one distinguishes within the nominal sentence (sn) the compulsory topic and comment functions (Figure 1a). In a verbal sentence (sv) we are dealing with an compulsory predicate function and several optional complements (Figure 1b). Optional sentence adverbials can occur in both sentence types. A np with a nominative case value for the head (npnom) always32 realizes the topic function.33 The comment function is realized by: a np34 with nominative case value for the head (npnom); an adjp with nominative case value for the head (adjpnom); a vp; a pp; or an adverb phrase (advp). A vp always fills in a sv the predicate slot. Sentence adverbials (sadv) are realized by: a np with accusative case value of the head (npacc); a pp; an advp; a cp; or a ccl.35

30 As mentioned in § 1, a particle phrase could be subdivided into a prepositional phrase (PP), an adverbial phrase (ADVP), an interjectional phrase (IP), and (to remain consistent) a conjunctive phrase (CP), or a complement clause (CCL). 31 Distinct concord and agreement phenomena mark a productive syntactic (and semantic) distinction between a nominal and a verbal sentence type in MSA. Discourse sensitive emphasis on the topic agent should be maintained, side by side with the possibility of emphasizing the action performance of, sometimes, the same agent in typical VSO-oriented approaches. Cf. Ditters 2001, 31–37. 32 The optional occurrence of a particle, like inna with an emphasizing semantic load and governing its complement by an accusative case value, only represents an alternative realization within the base structure. 33 An alternative topic realization is a CCL (complement clause), introduced by the emphatic particle inna, governing the head of the following NP in the accusative case. 34 A sub-class of the nouns is constituted by different subsets of adjectives. We should add the adjective phrase (ADJP) with a nominalized adjective in head position as possible alternative for a head or modifier function in the sentence. 35 For detailed structural descriptions of phrases in MSA, see Ditters 1992, chapters III and IV, and for a formal description of sentence structures see Ditters forthcoming.

featuring as a disambiguation tool in arabic nlp


In Figure 1a are listed the function slots within the sn, as well as the categories able to fill these slots. One may easily see how many different realizations of a nominal sentence are accounted for in this diagram where alternatives are represented within square brackets and optional elements within parentheses. FIGURE 1A







[NPnom]37 [NPacc]38

[ADJPnom] [ADVP] [CCL] [NPnom] [PCCL] [PP] [VP]

[ADVP] [CCL]39 [NPacc] [PP]

Figure 2

In Figure 1b are listed the function slots within the sv, as well as the categories able to fill these slots. One may easily see how many different realizations of a verbal sentence are accounted for in this diagram By means of variables as predicate-subject match (psmatch) and topiccomment match (tcmatch) we describe regular patterns in concord and agreement between the elements involved in a sv or sn, checking at the same time the matching of a verb-argument-1 relation as well as the compatibility of the elements involved in a topic-comment occurrence. In the same way we control the occurrence of other argument realizations of a verbal entry by means of a predicate-object match (pomatch).

36 The AGFL-formalism allows for free sequence variation at any level of description (this means for us: at function and category level) of entries within the formal description and processing. Therefore we do not need to list all mathematically possible realizations. 37 An alternative realization in the form of a CCL filler in a topic slot, as recorded from Classical Arabic data (Qurān 2:184): an tasūmū (topic) xayrun la-kum (comment), is, for its poor frequency in modern text data, not accounted for in our formal description of the Sn. 38 We include this alternative realization to account for an absolute negator-head combination in topic position (see below § 3.2). 39 In our formal description of Modern Standard Arabic we also include under the variable name CCL the protasis (the condition posed) in conditional and hypothetical sentences.


everhard ditters FIGURE 1B







[ADVP] [CCL] [NPacc] [PP]

Figure 3


Verb Phrase

At an early stage in the Arabic grammatical tradition, regular patterns in syntactic behavior and the meaning of the elements involved in verbal constructions brought grammarians to a general description of the verbal complement structure40 and the grouping together of verbs.41 Briefly summarizing this tradition we can say, that a finite-verb realization (marked for perfect aspect, indicative, imperative mood or their alternatives)42 functions as the compulsory head of a vp. In a minimum vp configuration, an implicit agent provides values for gender, number and person of the verbal head. Particles may precede the head realizing optional pre-modifying (prem) functions such as a modal, aspectual or temporal modifier, or carrying a negation value (neg). In a maximum vp configuration, an explicit agent, objects (compl) as well as other optional adverbial (adv pom) or peripheral (phr pom) modifiers may follow the head. The general structure of a vp is presented in Figure 2, where alternatives are included in square brackets and optional elements in parentheses:


Within the grammatical tradition grammarians spoke about: mutaaddin (transitive) and ġayr mutaaddin (intransitive), further differentiated into and complemented with: maf ūl bi-hi (direct object), maf ūl mutlaq (absolute object), maf ūl fī-hi (object of time or place), maf ūl la-hu or li-ajli-hi (object of cause or reason), maf ūl maa-hu (concomitative object), hāl (circumstantial object), tamyīz (object of specification). Cf. Sībawayhi’s al-Kitāb. Hārūn, 1:34–54, 297–310, 367–384; see also Fleisch 1968, 177–185. 41 We mention briefly: incomplete verbs (al-af āl an-nāqisa), the verbs of hope (af āl ar-rağā), the verbs of beginning (af āl aš-šurū), the verbs of the heart (af āl al-qulūb), the verbs of praise and blame (af ‘āl al-madh wa-d -d amm), the verbs of approximation (af ‘āl al-muqāraba), and the verbs of esteem (af ‘āl at-tafdīl). Cf. a.o.: Ayoub 1980; Cuvalay 1994, 1996; see also Wright, 2:47–52. 42 The negation of the perfect aspect by means of the particle lam followed by a finite verb form in the jussive or the negation particle lan governing a finite verb form in the subjunctive.

featuring as a disambiguation tool in arabic nlp FIGURE 2











[ADJPacc] [CCL] [NPnom] [NPacc] [PP] [VP]


[ADJPacc] [NPacc] [PCCL] [PP] [VP]

An active or passive voice value of the verbal head, in combination with the variable transitivity44 (with values as intransitive or transitive), affects the realization of a verbal complement structure, not only in the number of compulsory or optional arguments, but also in the form of these arguments as fixed in the lexically given DNA-structure of the verbal entry concerned. With the term ‘form’ we here refer to the argument realization by means of a np with nominative case (npnom) for an explicit agent, a np with accusative case (npacc) for direct object(s) or its equivalent,45 and any subject or object attribute,46 a complement clause (ccl), a co-referential vp for verbal extension,47 a pp as prepositional object or its equivalent.48 Each finite verb occurrence in MSA, intransitive and transitive alike, has in its complement structure a slot for an implicit agent or argument1 realization in the form of a suffixed pronoun (in past tense) or a discontinuous pre- and suffixed pronoun combination (in present tense).


Ditters 1992, 286 ff. We prefer the term ‘transitivity,’ with the values: intransitive, transitive, bi-transitive and tri-transitive to the term ‘valency’ or valence since the latter is less apt to determine non-optional core meanings. 45 A complement clause (CCL) introduced by the particle an or anna. 46 In de form of an NPacc or an indefinite adjective phrase marked for accusative case value (ADJPindef,acc). 47 With the variable name verbal-extension we describe, in our formal grammar, the occurrence of co-referential verb clusters, sometimes called ‘auxiliaries,’ such as: incomplete verbs (kāna or laysa, ‘to be(come)’ or ‘not to be’); inceptive verbs (badaa, ‘to begin’ and the like); continuity verbs (istamarra, ‘to continue’ and the like); anticipation verbs (kāda, ‘to be on the point’ and the like). Cf. Haak 2006, Badawi et al. 2004, 422 ff. 48 A prepositional complement clause (PCCL) introduced by a preposition followed by the particle an or anna. 44


everhard ditters

Combined with an explicit agent in the form of a npnom, the finite verb realization immobilizes into a 3rd person singular realization liable to vary in gender value. For monotransitive verbs with a direct object in the form of a npacc, in arg-2 position, this argument will assume the arg-1 position in case of a passive voice value of the verbal head. The same holds for bitransitive or tritransitive verbs: the first npacc direct object will assume the arg-1 position in case of passive voice realization. We may summarize the verbal complement structure in Modern Standard Arabic in two tables. The first matches the features form, function, and transitivity against the number of possible completive arguments. Together with the semantic load of the verbal entry concerned, we have the actors for the predicate-subject (psmatch) and the predicate object matching (pomatch). TABLE 1



function agent v-extension transitivity intrans/trans intrans/trans form [NPnom] [VPi]




object-1 object- 2 monotrans bitrans [NPacc] [ADJPindef,acc] [PPprep] [NPacc] [CCLan/anna/inna] [PPprep] [PCCLprep, an/anna]


object-3 indirect object tritrans intrans/trans [NPacc] [PPprep] [ADJPindef,acc]

The following table (Table 2) presents, in a general way, slots and possible filler for (mainly circumstantial) elements occurring within the complement structure of a lexically given verbal entry. As far as the featuring is concerned, it is gradually being introduced in the form of subscripts to the phrasal heads at syntax level. TABLE 2






function situation absolute source goal time form [NPacc] [NPacc] [PPmin/an] [PPprep] [NPacc,time] [CCLwa] [PPprep,time]




place instrument result [NPacc,place] [PPprep] [NPacc] [PPprep,place] [PPprep]

For the description of predicate-subject and predicate-object relations in MSA (psmatch and pomatch), we have so far exploited the Arabic grammatical tradition. Coming to Modern Standard Arabic, we like to

featuring as a disambiguation tool in arabic nlp


adopt, in the next section (§ 3.1), Saad’s work (1982, 20–26) in a ‘Case Grammar’ perspective for Classical Arabic. Some of the features linked with a lexical verbal entry are listed below: AGREEMENT [PSMATCH [POMATCH TRANSITIVITY MONOTRANS BITRANS TRITRANS


:: :: :: ::

+human; –human. ]49 ] intransitive; MONOTRANS; BITRANS; TRITRANS. :: acc; prep. :: acc2; accprep; prep2. :: acc3; acc2prep; accprep2; prep3.

Noun Phrase

We distinguish within the np one compulsory and several optional functions. Only a noun (or a nominalized adjective) can realize the compulsory head function. Among the optional functions we mention: negator (neg), a mutually exclusive predeterminer (predet) and postdeterminer (postdet), postmodifier (pom), and, for some head realizations, a complement (compl).50 In the following general structure of the np (Figure 3), optional functions are included in parentheses and mutually exclusive functions in square brackets.51 Relationships and dependencies within the np are specified and controlled by features and feature-condition rules. In a neg-head combination it is the negative particle lā that imposes the unique occurrence of an indefinite common noun in head-position provided with a definite version of the accusative case value. Personal pronouns and proper nouns, interrogative and indefinite pronouns (of the subset-type: man and mā) in head-position, are inherently marked as definite and indefinite respectively, thus they may only occur with an optional pom.


These rules will be discussed later. For example: a masdar in head function with a subjective complement in the slot for postdeterminer (NPgen) and an objective complement in the slot for COMPL (NPacc), both realizations in the form of a suffixed pronoun thus realizing a postdet and a compl function: hubb-ī-hā = love-mine-her ‘my love for her.’ 51 In Figure 3 we have included neg among the reciprocally exclusive function realizations. However, here we underline the incompatibility of a predet and a postdet in a np. Cfr. the examples in Cantarino 1974, I:114; 1975, II:220–222), Badawi et al., 2004, 464–466. El-Ayoubi et al. (2001–2003, I:1–2) do not mention a neg; neither in the ‘Vorfeld der Nominalgruppe,’ nor in the ‘Nachfeld der Nominalgruppe.’ 50


everhard ditters FIGURE 3




Fillers [NEG PARTlā]



([POSTDET]) [NPgen]


[PP] [NPacc]

Only (de)verbal nouns (those marked with acc, or acc2 for transitivity) as well as some nominalized adjectives in head-position may occur with a compl.53 Other relationships and dependencies concerning definiteness, person, gender, number, and case, will be discussed below. 3.2.1


Elements of the noun subsets proper nouns and personal pronouns have the value definite for definiteness with consequences for post-modifying elements at constituent and sentence level. Elements of the noun-subsets indefinite pronouns and interrogative pronouns have the value indefinite. Common nouns, (de)verbal nouns, and nominalized adjectives receive their value for definiteness by the occurrence of a predet or a postdet modifier. 3.2.2


Only elements of the noun subset personal pronouns vary in their value for person ranging from first, second to third, while varying in number and case feature values. All elements of other noun subsets bear the fixed value third for the feature person. 3.2.3


Occurrences of the subsets proper nouns, personal pronouns, and common nouns possess a lexically given value for the feature gender, whether this


Ditters 1992, 163 ff. For a more detailed discussion about adjectives in the context of the ellipsis of the head of an NP and of nominalized adjectives see Ditters (forthcoming). 53

featuring as a disambiguation tool in arabic nlp


is marked morphologically by a default null-value, by one of the feminine markers or not. Elements of the subsets indefinite and interrogative pronouns bear the default gender value masculine, but may, in retrospective and supported by the unification-principle of the agflformalism, receive an inherited feminine value provided by the context in which they occur beyond sentence level. At syntax level we may have to deal with the interference of semantic feature values such as ±animate, ±concrete, ±human, ±volition etc., here mainly related to as: concord and agreement phenomena. 3.2.4 Number The singular and plural realization of elements of the subsets ‘proper nouns’, ‘personal pronouns’, and ‘common nouns’ are given lexically. A dual number value realization is normally the result of the application of a grammar-rule except for the personal pronouns. Only infinitives or masdars always have a singular number value. Apparent evidence to the contrary is caused by the exploitation of masdars for the creation of an abstract lexicon. 3.2.5 Case Case is a syntax dependent application feature. Its values (nom, gen, and acc) are related to the function the element concerned has at phrasal or higher levels of linguistic description. A null-value for case, any other ambiguity concerning diptotic np-head occurrences, and gender issues for a first person realization have to be disambiguated beyond phrasal and sentence levels of description. 3.2.6


Earlier we mentioned feature-clusters. In some of them np’s are involved (concord, agreement, ps-match, po-match, and tc-match). In this section we discuss the occurrence of a noun in the head-position of a phrase at syntax level. Feature values of such a noun should be capable to match with elements occurring within the same constituent or with feature values of heads of other constituents within a given context. Summarizing some conclusions at this point, we list the following formal rules connected with a lexical noun entry at syntax level:




:: :: :: :: :: :: :: ::

+human; –human].54 full; partial.] ] ] ] accusative; genitive; nominative; null. definite; indefinite. demonstrative; personal; proper; relative; indefinite; interrogative; adjective; common; elative; numeral; quantifier; verbal.

Particle Phrase

As far as the occurrence of particles at syntax level is concerned: earlier we distinguished between an adverb phrase (advp), a prepositional phrase (pp), an interjectional phrase (ip), and a conjunctive phrase (cp). 3.3.1

Adverb Phrase

At the morphological level, we sub-categorized the adverbs into: invariable, as far as case marking is concerned, and bound or free forms, as far as the orthographic representation is concerned. One feature of the bound forms involves the temporal aspect of a verb realization itself marked by a non-perfect or mood tense (sa, directly linked to a finite verb form, then suggests a future temporal aspect).55 The interrogative a affects the whole of a succeeding verbal or nominal sentence. The affirmative la mainly occurs, in Modern Standard Arabic, as a premodifier to the particle qad in an advp. At syntax level, the bound and free forms are further differentiated into, among others, time, degree and manner adverbials, attached to verbs and adjectives as well. Finally, we have to distinguish between adverbs as head of an advp, on the one hand, and other realizations of an adverbial function at phrasal or sentence level such as pp, a npacc or a pccl at the other. Formally, an adverb phrase (advp) can be described as a, at syntax level occurring, constituent that, in its minimal configuration, has an adverb in the compulsory head function. Optional functions in an advp can be labeled as a pre- or postmodifier of the head. A premodifier may, for example, figure as a discontinuous negative particle to adverbials as


These rules will be discussed later. The so-called TMA (tense, mood and aspect) auxiliary particles have been discussed by Kinberg (2001). See also Eisele 2006. 55

featuring as a disambiguation tool in arabic nlp


qattu ‘never’ or faqat ‘no more’,56 or in the form of a preposition with an head as in min qablu ‘previously.’ An optional postmodifier can be realized by a PP.57 FIGURE 4








[ADV] [NPacc]


For the adverb we suggest the following combination of feature names with corresponding values:59 ADVERB



:: ARGUMENTATION; CONFIRMATION; DISCOURSE ; FOCUS; FREQUENCE; interrogative; LOCAL; MEASURE; MODALITY; SEQUENCE; TEMPORAL. :: evaluative; comparative; CONCLUSIVE. :: positive; negative. :: certain; affirmative; limitative. :: greetings; wishes; emphasis. :: specifying; inclusive; exclusive; cumulative. :: number; repetition; iteration; coinciding. :: position; dimension; direction. :: degree; quantity. :: expression; noun; adjective; nisbe. :: past; neutral; future; SUFFIXED; relative. :: coinciding; preceding; succeeding. :: idin; dāka; pronoun.

Prepositional Phrase

Within a prepositional phrase we distinguish two compulsory functions: a prepositional complement (pcompl) and a linker (plink) that connects that complement to the next higher level of linguistic description. The complement function can be realized by: an adverb phrase 56

Cf. El-Ayoubi et al. 2003, I:2, 275–460, and 406. Badawi et al. 2004, 161–174. In the discussion about the NP we already mentioned the use of an accusative NP (NPacc) for adverbial purposes. 58 Ditters 1992, 210 ff. 59 Cf. El-Ayoubi et al. 2004, I:2, 284 ff., and Wright 1974, I:282 ff. 57


everhard ditters

(advp); a complement clause (ccl) introduced by the particle an or anna (cclan|anna); a noun phrase with genitive case value (npgen); or a verb phrase with a head marked for subjunctive mood (vpsubj). The prepositional linker function is usually realized by an element of the particle-subset of prepositions. However, there are also expanded constructions in which a preposition, followed by an infinitive or abstract noun, realizes the header function,60 and prepositional clusters in which an infinitive, marked with an indefinite accusative case value (npacc) and followed by a preposition realizes the linker function.61 FIGURE 5






COMPL [ADVP] [CCLan|anna] [NPgen] [VPsubj]63

A good basis for a discussion about semantic features in the next section certainly is the sub-categorization of prepositions into: time, place, and ideal (or manner) features (Wright, 1974: II:129). With corpus based evidence, forwarded by Cantarino (1974, II: 262 ff.) and El-Ayoubi et al. (2004, I, 2:466 ff.) we are able to further differentiate this subcategorization of prepositions into features values:64 PREP IDEAL


:: IDEAL; PLACE; TIME. :: adversative; benefactive; causal; COINCIDING; comparative; content; discourse; explicative; final; HYPOTHETIC; instrumental; measure; modal; partitive; possessive; specification; substitution. :: comitative; simultaneous. :: concessive; conditional. :: destination; direction; local; partitive; position. :: direction; partitive; temporal.

Cf. El-Ayoubi et al. 2004, I:2,574. Cf. El-Ayoubi et al. 2004, I:2,584 ff. 62 Ditters 1992, 213 ff. 63 Not to confuse with the inflectional phrase as in the Government and Binding approach (cf. Fassi Fehri 1993, 16 ff.). 64 See also Badawi et al. 2004, 167 ff. 61

featuring as a disambiguation tool in arabic nlp


3.3.3 Interjectional phrase or clauses Earlier (§ 2.3.3) we differentiated the interjectional phrase (ip) into: vocatives; exclamations;65 frozen expressions, proverbs and greetings. Two of them, vocatives and exclamations, lend themselves easily for a general structure description (see Figure 6). Other optional ip’s ( f-expressions, proverbs, greetings, and insults) may provisionally receive, in our formal description of Modern Standard Arabic, the label of a phrasal or sentence adverbial since the syntactic structure of such an insertion can easily be described by means of the basic general syntactic patterns we already suggested. However, the semantic value of their occurrence in a given context has still to be evaluated and described.66 FIGURE 6









[INTERJ] [NPacc] [Sv]

[NPnom] [NPacc] [PPilā|bi|alā] [PPli] [Sadv]


A few comments may make this scheme more transparent. The line of slots tells us about the compulsory function of an ip head (i-head) with an optional tail in case of a discontinuous realization like wā-āh, be it for Modern Standard Arabic a bit archaic, as in wāzaydāh ‘Woe upon Zayd!’ (Badawi et al., 2004:37). It further mentions the optional occurrence of a premodifier (prem) or a complement (i-compl) of the i-head. In the filler-section alternative realizations are listed, also comprising a possible sv-sadv combination as man yaqtul yuqtal ‘Who kills will die killed.’ For this reason we included the term clause in the heading of this subsection.

65 66

Cf. Firanescu 2007, 79–81. Cf. Bergman 2007, II:136–137.

388 3.3.4

everhard ditters Conjunctive phrase (cp) or Clause (ccl)

Following the Wright tradition, at the morphological level, we made a distinction between a separable and an inseparable occurrence of conjunctions, as well as between connective and conditional conjunctive particles (Wright, 1974, I:290–294). At syntax level, however, we also prefer to distinguish between coordinating and subordinating conjunctions. Coordination The compulsory head-function in the cp is realized by an element of the subset coordinators of the conjunctions. An example of an optional premodifier would be the ‘narrative connector’ wa preceeding the adversative coordinator lākin ‘but’ (Badawi et al., 2004:555). The compulsory complement function may be realized by a variety of alternatives (Figure 7). FIGURE 7









[NP] [VP] [ADJP] [ADVP] [Sn] [Sv] Subordination The compulsory head function in a complement clause is realized by an element of the subset subordinators, which may contain a null value, for example, in caseof the occurrence of a purposive li followed by a verb with a subjunctive mood value. In our formal grammar of Modern Standard Arabic, the compulsory complement function may vary between a sn and a sv.

67 The conjunctive head may be empty is case of asyndetic coordination (Badawi et al. 2004, 539 ff.

featuring as a disambiguation tool in arabic nlp FIGURE 8










[Sn] [Sv]

The variables and features w